Pyrrolizidine alkaloids in tea, herbal tea and iced tea beverages– survey and transfer rates

ABSTRACT

The transfer rate of 37 pyrrolizidine alkaloids (PA) found in ten naturally contaminated teas and herbal teas to their brews was studied in detail. Mixed herbal, peppermint, red bush, senna, black tea and green tea infusions were prepared according to the ISO guide and vendor’s instructions, respectively, and parameters like herb-to-water ratio, steeping time and multiple extractions studied. In general, a transfer rate of 38–100% (median 95%) for brews following vendor’s instructions was determined. The total concentration range of PA in these ten samples was 154–2412 ng/g (median 422 ng/g) in the herb and for single analytes 0.1–170 ng/g. Seven of the 37 PA occurred unexpectedly; these were tentatively identified and quantified by liquid chromatography-high resolution mass spectrometry (LC-HR-MS), since their contributions to total PA-content matter. Additionally, 46 iced tea beverages were analysed for their PA-load, determined to be in the range 0–631 ng/L (median 40 ng/L). The applied solid-phase extraction (SPE) clean-up turned out to be capable of separating PA in the free base pyrrolizidine alkaloids (PAFB) and their N-oxides (PANO) in a two-step elution, which was a valuable tool to support identification of unexpected PA. Further, atropine was found in 50% of the ten tea herb samples (range: 1–4 ng/g) and in 13% of the iced tea beverage samples (range: 2–65 ng/L).

KEYWORDS:

• Pyrrolizidine alkaloids
• tea
• herbal tea
• iced tea
• transfer rate
• selective elution
• LC-HR-MS

Introduction

After (packaged) water, tea is the most consumed beverage worldwide. Annual tea consumption for 2021 is estimated to be 297 billion litres, that is, a per capita consumption of 37 litres per year (Bolton 2018). Tea is an important economic factor and its quality depends, among others, on harvest conditions and good agricultural practice (Codex 2014). It is in this context that PA were found unexpectedly in tea and herbal tea (BfR 2013), due to contamination with PA-containing weeds. A very wide variety of weeds are able to contaminate tea and herbal tea with PA. Several hundred 1,2-unsaturated PA are known and are produced mainly by plant species belonging to the families Asteraceae (tribes Senecioneae and Eupatorieae), Fabaceae (genus Crotalaria) and Boraginaceae (most genera) (Hartmann and Witte 1995; EFSA 2011). Since several PA are regarded as both hepatotoxic and carcinogenic, but the tea plants are not producing the contaminants, this health-related issue was addressed with numerous analytical and regulatory efforts.

The European Food Safety Authority (EFSA) concluded in its 2011 evaluation (EFSA 2011) that 1,2-unsaturated PA may act as genotoxic carcinogens in humans, based on the knowledge of metabolism, activation, DNA formation, genotoxicity, and carcinogenicity. In the same document, EFSA assessed the risk for genotoxic carcinogenicity and adjusted the reference point for the margin of exposure (MOE) approach in 2017 (EFSA 2017). The EFSA selected the benchmark dose lower confidence limit of 237 µg/kg body weight per day, calculated for a 10% increase in the incidence of liver haemangiosarcoma in female rats exposed to riddelliine as the new reference point for the chronic risk assessment of PA. Applying an MOE of 10,000 leads to an intake that would be of low concern from a public health point of view, that is, an adult of 70 kg should not consume more than 1.6 µg PA per day. The EFSA also concluded in its dietary exposure assessment that tea and herbal infusions were by far the main average contributors to the total exposure to PA (EFSA 2016). In response to this issue, the EU Commission published maximum levels for tea and herbal infusions in December 2020 (EU 2020). The levels range from 1.0 µg/kg for certain products in liquid form for infants and young children to 400 µg/kg for certain dried herbs for herbal infusions. The levels refer to the sum of a defined list of 35 PA (PAFB and PANO). The regulation also sets maximum levels for all other dried herbs and for some specific fresh or frozen herbs. Additionally, maximum levels for certain food supplements were set. However, the regulation will not come into force until 1 July 2022 and products lawfully placed on the market before this date may remain on the market until 31 December 2023.

To gain toxicological relevant exposure data, an important parameter is the transfer rate of PA from tea and herbal tea to the brew. Sparse and even contradictory data are available at present only. Lüthy et al. (1980) published a transfer rate of 80% for senkirkine from coltsfoot, Engeli (2014) 84–103% for nine PA from peppermint herbs, Mulder et al. (2018) about 85% for 25 PA from 38 samples and on the other hand 16–28% for 30 PA from 12 green, 16 black and 29 herbal or red bush samples (Picron et al. 2018).

In this study, we aimed to clarify the transfer rates of PA under various conditions, estimate the exposure due to iced tea beverage consumption and trace the PA-content of selected commercial herbal tea and iced tea beverage samples from 2013 to 2020.

Material and methods

Chemicals and reagents

Methanol was from Honeywell (LC-MS grade, Grogg AG, Stettlen, Switzerland). Sulphuric acid, formic acid and aqueous ammonia solution 25%, all of analytical grade, were from Merck (Grogg AG, Stettlen, Switzerland). Ultrapure water was used (Elga Purelab ultra, Labtec, Villmergen, Switzerland). Echimidine (EM), erucifoline N-oxide (EFO), europine (EU), europine N-oxide (EUO), intermedine (IM), intermedine N-oxide (IMO), jacobine (JB), jacobine N-oxide (JBO), lasiocarpine (LC), lycopsamine (LY), lycopsamine N-oxide (LYO), monocrotaline (MC), senecionine (SC), senecionine N-oxide (SCO), senkirkine (SK), senecivernine (SV), senecivernine N-oxide (SVO) and trichodesmine (TD) were from Phytolab (Vestenbergsgreuth, Germany). Jacoline (JL), retrorsine (RT) and seneciphylline (SP) were from Cfm O.Tropitzsch (Marktredwitz, Germany). Heliosupine (HS) and rinderine (RN) were from Phytoplan (Heidelberg, Germany). Echivulgarine (EV) was isolated in-house (Lucchetti et al. 2018) and echimidine N-oxide (EMO), heliotrine N-oxide (HTO), monocrotaline N-oxide (MCO), lasiocarpine N-oxide (LCO), retrorsine N-oxide (RTO) and seneciphylline N-oxide (SPO) were N-oxidised in-house starting from the free bases according to Cymerman Craig and Purushothaman (1970), except that dichloromethane instead of trichloromethane was used. Atropine (ATR) and the internal standard homatropine (HTR) were from TCI (Eschborn, Germany) and D3-atropine (D3-ATR) was from CDN isotopes (Quebec, Canada).

For each analyte, except for ATR, EFO, EMO, HT, HTO, JL, LCO and RN which were available for qualitative purposes only, a stock solution in methanol (c = 100 µg/mL) was prepared and stored at −20°C. For pre-extraction spiking purposes, a multi-analyte standard was freshly prepared in methanol (c = 1 µg/mL), which was further diluted to 0.1 and 0.01 µg/mL with methanol for post-extraction fortification experiments, that is, matrix calibration.

Tea, herbal tea and iced tea beverage samples

All samples for infusion preparation and the iced tea beverages were from the Swiss market, procured during 2013 to 2020 and comprised eight teas and herbal teas for infusion preparation and 16 herbal teas, to study PA-content over the years, all in tea bags; 46 iced tea beverages, six blank teas and herbal teas, all from loose tea; and three blank iced tea beverages. Two validated reference materials from ring trials, red bush (P1706-RMRo 2017) and peppermint herb (Uhlig et al. 2019), were additionally treated as samples to verify accuracy during method procedures.

Herbs were combined to an aggregate of 30 to 50 g, each, and stored at room temperature.

Infusion preparation and extraction of PA

Samples were analysed in duplicate and herbs ground to a fine powder (particle size 250–500 µm) by an IKA A11 mill (IKA, Staufen, Germany).

Three extraction procedures were applied for infusion preparation. To determine the total PA-content, samples were extracted with 0.05 M aqueous sulphuric acid at room temperature. This procedure consisted of three consecutive extractions to make sure of complete removal of PA. To study their transfer to the brew, three consecutive extractions according to the ISO 3103 protocol (ISO 1980) were applied, then in a fourth extraction possible remaining compounds removed with 0.05 M aqueous sulphuric acid to ensure full PA-extraction. Further, infusion preparation according to vendor’s instructions was conducted, terminated again by a final extraction with 0.05 M aqueous sulphuric acid. Table 1 gives an overview on all applied extraction procedures and conditions.

Table 1. Overview on all applied extraction techniques and conditions for tea herbs. Additionally, capped vessels were shaken horizontally at 250 rpm

procedure	Sulphuric acid	ISO	Vendor’s
Herb type(s)	All (n = 2x10)	All (n = 2x10)	Black (n = 2x2)	Green (n = 2)	Mixed herbal (n = 2x2)	Peppermint (n = 2x2)	Red bush (n = 2x2)	Senna (n = 2)
Herb-to-water ratio	1:50	1:50	1:100	1:125	1:100	1:143	1:100	1:100
Extraction 1	0.05 M aqc. H2SO4	Boiling water	Boiling water	Water, 80°C	Boiling water	Boiling water	Boiling water	Water, 80°C
Steeping time [min]	15	6	4	2	10	6	6	10
Capped/uncapped	Capped	Capped	Uncapped	Uncapped	Capped	Uncapped	Uncapped	Uncapped
Extraction 2	0.05 M aqc. H2SO4	Boiling water	0.05 M aqc. H2SO4	Boiling water	0.05 M aqc. H2SO4	0.05 M aqc. H2SO4	0.05 M aqc. H2SO4	0.05 M aqc. H2SO4
Steeping time [min]	15	6	15	1	15	15	15	15
Capped/uncapped	Capped	Capped	Capped	Uncapped	Capped	Capped	Capped	Capped
Extraction 3	0.05 M aqc. H2SO4	Boiling water		0.05 M aqc. H2SO4
Steeping time [min]	15	6		15
Capped/uncapped	Capped	Capped		Capped
Extraction 4		0.05 M aqc. H2SO4
Steeping time [min]		15
Capped/uncapped		Capped

The ISO-procedure was adapted to the use of 25 mL centrifuge filter tubes (Nylon, pore size 0.22 µm, Thermo Scientific, Reinach, Switzerland) to allow for multiple extractions. An aliquot of 0.4 g homogenised sample was transferred to the membrane-containing filter unit and D3-ATR added as quality standard (c = 200 ng/g). Then, 20 mL of boiling water were poured into the graduated filtration inlay. After capping and 6 min of steeping time under horizontal shaking at 250 rpm, the sample was centrifuged at 2200 g for 15 min and the filtrate collected. The extraction procedure was repeated two more times, each with 20 mL of boiling water. Finally, the filtration residue was extracted with 20 mL 0.05 M aqueous sulphuric acid for 15 min at room temperature under shaking, then centrifuged and the resulting four filtrates cleaned up by SPE separately.

For tea and herbal tea preparation according to vendor’s instructions, the same procedure applied, but with adapted herb-to-water ratios (HWR = 1:100 to 1:143) and variations in the number of extraction steps (2–3), water temperatures (80°C or boiling) and whether the centrifuge tube was capped during steeping time or not.

Prior to SPE, analytes containing solutions were adjusted to pH≈2 under dropwise addition of concentrated sulphuric acid where necessary and syringe-filtered (≤1 µm) in case of precipitate formation. The cleanup was achieved by automated extraction on an Aspec GX-274 robot (Gilson, Mettmenstetten, Switzerland). Aliquots of 17 mL of sample solution from herbs and brews or 300 mL of iced tea beverages were passed through a SPE PCX cartridge (Bond Elut Plexa, Agilent, MSP Kofel, Zollikofen, Switzerland), previously conditioned with 5 mL methanol and 5 mL water. The loaded column was washed consecutively with 5 mL water and 5 mL methanol. The analytes were eluted by 8.5 mL ammonia solution in methanol (1.5%, v/v), the solvent evaporated to dryness under nitrogen at 40°C and the residue dissolved in methanol/water/formic acid (5:95:0.1%, v/v/v) to result in a matrix strength of 0.5 g/mL for herbs and their brews, and 500 mL/mL for iced tea beverages.

For selective PANO vs PAFB elution, the acidic solution containing the analytes was adjusted to 5<pH<7 through dropwise addition of aqueous ammonia solution (25%) prior to SPE. The clean-up was the same as described above, but 8.5 mL pure methanol as first eluent were used, leaching out the PANO solely. Subsequently, the SPE-cartridge was eluted a second time, by 8.5 mL ammonia solution in methanol (1.5%, v/v), releasing the residual PANO and the PAFB (and SK).

Identification and quantification by LC-HR-MS

The chromatographic system consisted of a Shimadzu Prominence binary gradient system (Shimadzu, Reinach, Switzerland), degasser, autosampler, static mixer and column oven. Chromatographic separations were performed on a Kromasil C18 (125 × 2 mm, 3.5 µm particle size) analytical column (Macherey-Nagel, Buchs, Switzerland). The flow rate was 0.2 mL/min, column temperature was maintained at 30°C and injection volume was 5 µL. A gradient programme was used. The solvents were water and methanol, both containing 0.1% formic acid (v/v). The programme was 5% methanol at 0 min, then ramped linearly from 0 to 10 min to 23% methanol, then from 10 to 22 min to 37% methanol and 22–23 min to 95% methanol, held for 3 min at this condition, then the column re-equilibrated for 9 min at 5% methanol.

Mass spectrometric detection was achieved by a Bruker maXis 4G+ Qq-TOF mass spectrometer (Bruker, Bremen, Germany), equipped with an electrospray ionisation interface operated in positive ion mode. Source parameters were plate offset 500 V, capillary voltage 4.5 kV, dry temperature 200°C, nebuliser gas pressure 150 kPa and nitrogen dry gas flow rate 8 L/min. Typical resolution at m/z = 350 was 35,000, with an allowed deviation for determinations of ≤2 mDa. Collision induced dissociation (CID) fragment spectra of mass selected ions were generated by nitrogen as collision gas at 35 eV ion energy. Instrument calibration was achieved by incorporating sodium formate solution at the beginning of every run by a loop injection. For instrument control, data acquisition and processing, Compass 1.7 and TargetAnalysis 1.3 were used.

Identification of compounds was accomplished by comparing retention time (< ±2.5%), high mass resolved molecular ion (≤ ±2mDa) and three fragment ions (≤ ±2mDa) to standard substances or literature. The selective PANO vs PAFB elution served as additional identification verification.

For quantification, a matrix-matched methodology applied. Blank samples of teas and herbal teas (n = 6), infusions from these, prepared according to the ISO guideline and blank iced tea beverages (n = 3) were extracted as described above, but the dried residue from SPE-clean-up was dissolved in methanol and portioned in vials for the seven calibrate levels applied, each. Standard mix containing 24 PA and D3-ATR was added for a final concentration range of 0, 5, 20, 250, 500, 750 and 1000 ng/g for tea, herbal tea and their brew, and 0, 0.0125, 0.050, 0.625, 1.25, 1.875 and 2.5 ng/mL for iced tea beverages. Additionally, HTR was added as internal quality standard (c = 200 ng/g and 0.5 ng/mL, respectively). Subsequently, the vial contents were evaporated to dryness and reconstituted in LC eluent, in order to achieve final matrix strength of 0.5 g/mL for tea, herbal tea and their brew or 500 mL/mL for iced tea beverages. The protonated molecular ion MH⁺ was used as quantifier, while three specific fragment ions from CID were surveyed as qualifiers. For PA without identical analyte calibration curve, the most similar one was taken (Table 2).

Table 2. List of analysed substances. A capitalised O in the name of PA depicts its N-oxide and putative trivial names are given in brackets. The retention time (rt) of the analyte is given, identification is based on reference substance (ref) or literature (1 = (Boppré et al. 2005), 2 = (Kaltner et al. 2019), 3 = (van de Schans et al. 2017), 4 = (Cheng et al. 2017), 5 = (Colegate et al. 2012)) and analytes were found in iced tea beverages (b) and/or tea herbs (h)

Acronym	(Tentative) name	Sum formula	rt [min]	MH+	Qualifying MS/MS fragments @ 35 eV			Identification	Quantified with	Found in
ATR	Atropine	C17H23NO3	15.5	290.1751	124.1121	93.0700	91.0540	ref	D3-ATR	b, h
D3-ATR	D3-atropine	C17D3H20NO3	15.5	293.1939	127.1309			ref	D3-ATR
EFO	ErucifolineO	C18H23NO7	10.1	366.1547	278.1383	218.1171	136.0757	ref	JBO	b
EM	Echimidine	C20H31NO7	19.7	398.2173	238.1457	220.1331	138.0913	ref	EM	b, h
EMO	EchimidineO	C20H31NO8	19.2	414.2122	396.2017	352.1755	254.1387	ref	EM	b, h
EU	Europine	C16H27NO6	9.9	330.1911	254.1387	156.1020	138.0913	ref	EU	h
EUO	EuropineO	C16H27NO7	10.8	346.1860	328.1746	256.1173	172.0968	ref	EUO	h
EV	Echivulgarine	C25H37NO8	26.5	480.2592	380.2068	322.1649	220.1331	ref	EV	b, h
EVO	(EchivulgarineO)	C25H37NO9	26.5	496.2541	396.2017	338.1598	220.1331	(1)	EV	b, h
HS	Heliosupine	C20H31NO7	20.3	398.2173	238.1457	220.1331	120.0808	ref	HS
HT	Heliotrine	C16H27NO5	13.8	314.1962	156.1020	138.0913	120.0808	ref	IM	h
HTO	HeliotrineO	C16H27NO6	15.0	330.1912	172.0968	155.0941	138.0913	ref	IMO	b, h
HTR	Homatropine	C16H21NO3	11.5	276.1594	124.1121			ref
IM	Intermedine	C15H25NO5	9.5	300.1805	156.1020	138.0913	120.0808	ref	IM	b, h
IMO	IntermedineO	C15H25NO6	11.3	316.1756	172.0968	155.0941	136.0757	ref	IMO	b, h
JB	Jacobine	C18H25NO6	9.7	352.1755	280.1538	234.1485	155.0941	ref	JB	b, h
JBO	JacobineO	C18H25NO7	10.5	368.1704	278.1383	190.1225	120.0808	ref	JBO	b, h
JL	Jacoline	C18H27NO7	6.3	370.1860	342.1911	326.1598	138.0913	ref	JB	h
JLO	(JacolineO)	C18H27NO8	7.4	386.1809	342.1547	136.0757	120.0808	(2)	JBO	h
LC	Lasiocarpine	C21H33NO7	24.6	412.2331	336.1807	238.1457	220.1331	ref	LC	b
LCO	LasiocarpineO	C21H33NO8	26.2	428.2286	410.2176	338.1598	254.1387	ref	LC	b
LY	Lycopsamine	C15H25NO5	10.1	300.1805	156.1020	138.0913	120.0808	ref	LY	b, h
LYO	LycopsamineO	C15H25NO6	11.9	316.1756	172.0968	155.0941	136.0757	ref	LYO	b, h
MC	Monocrotaline	C16H23NO6	6.8	326.1598	237.1361	194.1170	138.0913	ref	MC	b, h
MCO	MonocrotalineO	C16H23NO7	9.5	342.1550	296.1492	236.1282	137.0835	ref	MCO	h
RN	Rinderine	C15H25NO5	9.7	300.1805	156.1020	138.0913	120.0808	ref	IM	b, h
RNO	(RinderineO)	C15H25NO6	10.7	316.1756	172.0968	155.0941	136.0757	(3)	IMO	b, h
RT	Retrorsine	C18H25NO6	12.9	352.1755	324.1805	220.1331	138.0913	ref	RT	b, h
RTO	RetrorsineO	C18H25NO7	13.2	368.1704	340.1755	220.1331	136.0757	ref	RTO	b, h
RTOiso	(UsaramineO)	C18H25NO7	13.4	368.1704	352.1750	220.1331	154.0858	(4)	RTO	h
SC	Senecionine	C18H25NO5	16.4	336.1807	308.1856	220.1331	138.0913	ref	SC	b, h
SCO	SenecionineO	C18H25NO6	17.0	352.1758	324.1805	220.1331	136.0757	ref	SCO	b, h
SK	Senkirkine	C19H27NO6	21.2	366.1911	168.1034	150.0925	140.0706	ref	SK	b, h
SP	Seneciphylline	C18H23NO5	13.6	334.1653	306.1703	151.0753	138.0913	ref	SP	b, h
SPO	SeneciphyllineO	C18H23NO6	14.4	350.1596	322.1649	154.0864	138.0913	ref	SPO	b, h
SV	Senecivernine	C18H25NO5	15.9	336.1807	308.1856	220.1331	138.0913	ref	SV	h
SVO	SenecivernineO	C18H25NO6	16.2	352.1758	324.1805	220.1331	136.0757	ref	SVO	h
TD	Trichodesmine	C18H27NO6	12.5	354.1911	308.1856	222.1489	164.1068	ref	TD	h
TDiso1	(18-hydroxyincanine)	C18H27NO6	10.3	354.1911	308.1856	222.1489	120.0808	(see text)	TD	h
TDiso2	(18-hydroxyincanine)	C18H27NO6	11.0	354.1911	308.1856	222.1489	120.0808	(see text)	TD	h
TDO	(TrichodesmineO)	C18H27NO7	14.4	370.1860	324.1805	238.1438	137.0835	(5)	TD	h
TDOiso1	(18-hydroxyincanineO)	C18H27NO7	12.1	370.1860	324.1805	238.1438	137.0835	(see text)	TD	h
TDOiso2	(18-hydroxyincanineO)	C18H27NO7	12.4	370.1860	324.1805	238.1438	137.0835	(see text)	TD	h
unkn	Unknown	C20H31NO7	19.8	398.2173	238.1457	138.0913	120.0808	(see text)	EM	h
unknO	UnknownO	C20H31NO8	22.5	414.2122	254.1387	138.0913	136.0757	(see text)	EM	h

Statistical evaluations were performed using Systat V9 (SPSS Inc., Chicago, USA).

Method validation

Brewed sample preparations for validation purposes were made according to the ISO guide requirements (ISO 1980).

Method linearity was evaluated using matrix matched calibration curves for each sample type and analyte. Six blank tea and herbal tea samples and their brew were fortified at concentration levels of 0, 5, 20, 250, 500, 750 and 1000 ng/g with the multi-analyte standard. Three blank iced tea beverage samples were spiked at 0, 0.0125, 0.050, 0.625, 1.25, 1.875 and 2.5 ng/mL. For linear regression, the corresponding function was forced through the origin and the correlation coefficient used as quality indicator.

Recovery was calculated by fortifying blank samples at a concentration level of 200 ng/g (pre-extraction spike), applied to six teas, herbal teas and their brews and spiking 0.5 ng/mL to three blank iced tea beverages. For precision determination, intra- and inter-day variations were monitored at these concentration levels for blank samples of a green tea herb, its brew and an iced green tea beverage. A three- to five-fold replicate analysis was performed in-day, while for inter-day variability analyses were performed in duplicate over three consecutive days.

Matrix effect, that is, the ionisation effect on analytes, was evaluated by spiking blank sample extracts for above-mentioned samples (post-extraction spike). The matrix effect was then calculated by subtracting the ratio of the slope values of the matrix-matched calibration curve and the blank calibration curve to 1 and expressing it in relative percentage. Limits of detection (LOD) were estimated from the lowest spiking levels applied to blank matrix samples based on a signal-to-noise ratio of three. Similarly, limits of quantification were estimated at 1.5–3x LOD, depending on analyte and matrix.

The extraction efficiency was calculated from the ratio area of spiked sample (pre-extraction spike) vs area of spiked extract (post-extraction spike).

Accuracy and trueness of the quantitation method were investigated by processing two matrix reference materials from ring trials, that is, red bush (P1706-RMRo 2017) and peppermint herb (Uhlig et al. 2019) and comparing data by calculating the corresponding z-score of the analyte (see discussion below). It is calculated as difference between determined and referenced value, divided by the Horwitz target standard deviation, considering the modifications by Thomson (ISO 2015). An absolute z-score value of equal or less than two is regarded as acceptable, that is, determined and reference value under this condition are comparable.

Results and discussion

Method validation

Data concerning linearity, recovery, matrix effect and extraction efficiency are presented in Table S1 of the supplemental data collection. The chosen regression model fits our data well (Figure S1). The internal standard HTR was added to every matrix calibration series and vial (n = 18×7) and the determined calibration-series repeatability, expressed in terms of relative standard deviation (RSD), varied in a range of 7–10%.

Regarding linearity, 85–88% of the analytes had a correlation coefficient above 0.99, independent from extraction solvent used. For these, a linear relationship between analyte concentration and peak area in the range 0–1000 ng/g for tea, herbal tea and their brew and of 0–2.5 ng/mL for iced tea beverages is ensured. The coefficients of the remaining analytes were all above 0.96. The analytes EV, JB, JBO, MC and MCO were especially affected by these lower correlations, and all had elevated LOD.

As for the recovery results from blank matrix spiked samples, 48% of the analytes in tea and herbal tea (range 36–94%), 17% in brews (range 30–95%) and 68% in iced tea beverage samples (range 42–98%) were in the range 70–120%. The outlying analytes were hampered by low extraction efficiency and signal suppression effects.

Regarding the extraction efficiency, 82% of the analytes in tea and herbal tea, 54% in brews and 96% in iced tea beverage samples were in the range 70–120%. Low analyte transfer was observed for ISO-brews in general, compared to acidic extracted herbs and iced tea beverages. This is due to the different extraction solvents and acting times applied, that is, 15 min 0.05 M aqueous sulphuric acid for tea and herbal tea herbs, 6 min boiling water for brews or already solubilised analytes in acidified beverages. Especially SC and SPO in tea, herbal tea plus EM, EU, HS, IM, LY, RT, SC, SP and SV in brews showed low extraction efficiency. For brews, the less pH-susceptible PANO are obviously extracted more efficiently with 58–119% (median 84%) compared to the PAFB (and SK) with 43–93% (median 65%). On average, hot water extracted PA 20% less efficient than aqueous sulphuric acid at room temperature, deduced from the recoveries of 82 and 66% for acidic and hot water extraction of spiked blanks, respectively (Table S1).

As already stated, there is a significant negative matrix effect, i.e. ion signal suppression due to ionisation process and chemical property of the analytes. Notable matrix suppression was detected, independent from extraction solvent, for 37–43% of the analytes with values of −20% and less, that is, increasing matrix suppression; for 1–4% with even less than −50% matrix effect. Especially affected were EV, JB, JBO, MC and MCO.

To overcome and compensate the negative effects of extraction efficiency and matrix, matrix-matched calibration was applied for quantification.

The LOD values were in the range 0.1–2 (median 0.25) ng/g for tea herbs, 0.1–1.2 (median 0.3) ng/g in brews (based on the herb) and 0.3–5 (median 1.25) ng/L for iced tea beverages. In tea herbs, especially IM and JBO in black tea, EU in mixed herbal tea, EUO in peppermint and RTO, SCO, SPO, SVO in senna had elevated LOD (1–2 ng/g). In brews, the LOD of JBO in black tea, EU, EUO, JB and TD in mixed herbal tea, EUO in peppermint and SCO, SPO, SVO in senna were increased (1–1.2 ng/g) as were iced green tea beverage LOD for IM and LY (5 ng/L). Data on limit of detection and repeatability are presented in Table S2.

The intra- and inter-day precision expressed in terms of RSD was 3–14% and 5–17% for tea, herbal tea, 4–11% and 5–14% for brews and 4–11% and 4–14% for iced tea beverages, respectively, providing acceptable repeatability for all analytes.

The accuracy of this method was successfully demonstrated in a comparison with certified values of two validated reference materials. The derived z-scores for all comparable analytes (n = 11 in red bush, n = 12 in peppermint herb) were below two and were −1.44 to 0.59 for red bush and −1.40 to 1.30 for peppermint herb. For further details, see Table S3.

Analytical methodology

Within the limitations of the LC system, allowing for pressures only below 40 MPa, a long chromatographic run was chosen in order to separate as many isomers as possible. The retention time range for PA was from 6.3 (JL) to 26.5 min (EV), see Table 2.

For SPE, the sorbent types reversed-phase C18, SCX and PCX were tested. In accordance with Chen et al. (2019), PCX was chosen for its simplicity in use and good retention and elution properties. Additionally, under normal operation no pH adjustments or vacuum drying are required. Eluates result in clear, yellow solutions and consequently ion source and instrument contaminations are moderate.

During PCX-SPE optimisation, it turned out that a selective elution of PAFB vs PANO was possible through a two-step procedure. Prior to SPE, the acidic extraction solution containing the analytes had to be adjusted to 5<pH<7 with aqueous ammonia solution (25%). In the first SPE-clean-up step following the procedure described in the infusion preparation section, pure methanol was applied as eluent. Under these conditions, only PANO were eluted, depending on their concentration and matrix between 27% and 100%. The SPE cartridge was subsequently eluted with ammonia solution in methanol (1.5%, v/v), leaching out the residual PANO (73 to 0%) and the PAFB (and SK) (100%). The pH-susceptible PAFB, including SK, eluted only under basic conditions. Table S4 shows the elution rate-dependence on the pH prior to SPE for a peppermint sample containing the five PA-pairs EV/EVO, RT/RTO, SC/SCO, SP/SPO, SV/SVO and SK. At pH<6, PA elute mainly in the second, basic, elution step, while at pH = 6 only PANO elute and at pH = 8 mainly all PA are eluted already in the first step. Table S5 demonstrates the elution selectivity for samples of black tea, green tea, mixed herbal tea, peppermint, red bush and senna. Note that depending on tea or herbal tea type, some of the PA are missing, due to low concentration, elevated LOD or chemical noise. The affected PA were EU, HTO, JBO, JL, JLO, RN and an unknown PANO (unknO, see below). For the analysed samples, 27–100% (median 82%) of PANO were eluted and separated in the first elution step, while 100% of PAFB and SK eluted in the second step.

Generally, chromatograms of cleaned-up extracts were dominated by phenylalanine, tryptophan, tyrosine (Zhang et al. 2019), adenosine, 3-carboxy-1,2,3,4-tetrahydro-beta-carboline (Strehmel et al. 2014) and deoxyguanosine (Lu et al. 2006), see Table S6. For these, ion intensities were a factor of 10 to 100 above those of PA.

Infusion preparation and transfer rate

Heat dissipation of vessel

To check the compatibility of the vessel to the requirements of the ISO-guide, its heat dissipation characteristic was studied. It turned out that it had similar isolation capacity as a ceramic mug or pot, which was investigated in parallel, having an outer and inner diameter of 9 and 7.5 cm, respectively, and 11 cm in height. The heat dissipation within 6 min for the centrifuge tube and the ceramic pot when capped were 22 and 23°C, respectively, with 26 and 31°C, respectively, when left open. Therefore, we consider the temperature profile requirement of the ISO-guide as fulfilled, even though we used a double-walled spin-filter polypropylene tube as brewing vessel.

Identification and quantification of PA

In the ten teas and herbal teas under investigation, 37 different PA were found (Table 2). Among these, seven unknowns were detected and characterised. All seven unknowns shared the three retronecine-characteristic low-mass CID product ion triads m/z = 93 to 95, 119 to 121 and 136 to 138, while a prominent m/z = 136 and the presence of a protonated dimer 2MH⁺ were additionally indicative for N-oxides. For comparison reasons, in Table S7 the unknown PA are listed together with their isomeric analogues EM, EMO, RTO, TD and TDO. Details on identification of unknowns are given in Table S8.

RTOiso was tentatively identified as stereoisomer of RTO, possibly usaramine N-oxide (Cheng et al. 2017). Both TD-isomers, TDiso1 & 2, did not match their structure isomers axillaridine (Wiedenfeld et al. 1985; Asres et al. 2004) or 12-desoxyaxillarine (Wiedenfeld et al. 1985), which according to our structure S1 would be 13-deoxyaxillarine. Owing to the clues from MS/MS-experiments and the close structural relationship to incanine (Tashkhodzhaev et al. 1979), we putatively named TDiso1 & 2 as 18-hydroxyincanine (structure S1). In Crotalaria species, a similar PAFB was found, pumiline A, which carries an epoxy functional group in position C-13, C-18 (Fletcher et al. 2011). In analogy, and since typical N-oxidised TDiso1 & 2-fragments were detected, TDOiso1 & 2 were named 18-hydroxyincanine N-oxide. The unknown (Unkn) and unknO due to their echimidine group-typical fragment ions and neutral losses were ascribed to an open-chain diester PAFB and PANO, respectively (Yoon et al. 2015). Additional neutral losses of m/z = 32 (CH₄O, methanol), 46 (C₂H₆O) and 62 (C₂H₆O₂ = C₂H₄O+H₂O?) indicate the presence of a side-chain methoxy-group similar to LC and possibly an additional ethoxy or secondary alcohol group with a methyl substituent.

Regarding quantification, it is important to screen samples for unexpected PA, since they might make considerable contributions to total PA-content. Here, unexpected PA were characterised and quantified according to their structure type by the most similar PA for which matrix-matched calibration curves were available (Table 2). Of course, the derived values remain estimations. Still, equal approaches to quantify unexpected or unknown PA have been proposed by others (These et al. 2013; Rösemann et al. 2014; Mägde et al. 2020). Since samples have been finely ground for homogeneity purposes, their physical properties have been changed. This might lead to higher transfer rates over time compared to the unground materials. We judge this an acceptable compromise in view of the required reliability of acquired data.

Comparability of data

The summed-up value over all extraction steps for every analyte was compared among the three extraction techniques (aqueous sulphuric acid, brew according to ISO and vendor’s instructions) by analysis of variance. The Thomson-corrected Horwitz standard deviation (ISO 2015) of the PA-values was determined and test-statistic value (F) plus significance level (p) calculated for every data triple. In general and as shown in Table S9, values from the three extraction techniques do not significantly differ from each other (F(2,6)<5.14, p > 0.05). For those not complying (0.1%), a Bonferroni-corrected post-hoc test revealed the outlier to eliminate and significance parameters were adapted (F(1,4)<7.71, p > 0.05). These applied also to data pairs where, due to values close to LOD, analytes were found only in two out of the three extraction techniques.

The individual PA-values for all samples, analytes and extraction steps including their standard deviations are listed in Table S10. Analysis of variance for the relative standard deviations calculated from Table S10 reveals that RSD do not vary depending on weighed portion, herb-to-water ratio or concentration of PA, but on sample type, extraction technique, PA (=analyte) and extraction time (Table S11). Figure S2 shows the least squares means plot of RSD vs the applied extraction techniques. While ISO and vendor’s brew have similar RSD, they are lower for acidic extraction at room temperature. Since for all three extraction techniques the sample homogeneity is the same, we assume the extended extraction time applied for acidic extraction (15 vs 6 and 1–10 min) to provide lower RSD. From a plot RSD vs (total) analyte content (Figure S3), median values for RSD and PA concentration of 12% (range 1–71%) and 19 (range 0.1–1100) ng/g, respectively, were determined. The median analyte content of the ten measured samples compares well with a median content of 22 (mean 87) ng/g determined for teas and herbal teas procured from 2013 to 2020 (range 0 (19%) – 1163 ng/g, n = 138, not shown).

Ratio PANO vs PAFB

For the analysed ten teas and herbal teas, the ratio PANO vs PAFB is in the range 0.4–4.5 (median 2.0, n = 10). The ratios are below one only for green tea and mixed herbal tea 1, independent from the extraction technique applied, so possible temperature degradation during extraction should not be the cause for ratios below one (Hösch et al. 1996). The ratio is reported to vary depending on the season, but for fresh weeds, a ratio higher than one is expected (Flade et al. 2019). In our green tea sample, for all PA-pairs IMO/IM, JBO/JB, LYO/LY and unknO/unkn the free bases dominate, resulting in a PANO vs PAFB ratio of less than one (Figure S4). This trend complies with nine other green tea samples measured by us (data not shown) and is partly supported by Mulder et al. (2015), reporting green tea ratio values below one for IMO/IM and JBO/JB, but considerably above one for LYO/LY, though. This ratio-mismatch might be caused by additional herb treatment and is also noted for iced tea beverages, where 60% of PANO vs PAFB ratios are below one. On the other hand, it is common use to heat or dry green tea to prevent oxidation. That additional treatment, e.g. drying, is able to change the PANO vs PAFB ratio, was shown by Mulder et al. (2009), in whose study dehydration of Senecio-contaminated lucerne resulted in a ratio-decrease. The ratios LYO/LY (except in peppermint and red bush) and EVO/EV were also always below one. In the case of EVO/EV, inappropriate quantitation might be the reason, since EVO for lack of reference substance was quantified with EV. Additionally, RTO/RT ratios in red bush 1 and SVO/SV in peppermint herb 1 are below one.

Tea and herbal tea types investigated

The transfer rates for every extraction step and analyte for brews are listed in Table 3 and for acidic extraction in Table S12. Since the values are calculated from Table S10, the same constraints regarding (relative) standard deviations apply (see above). In order to keep an overview, in Table 3 and S12 the indications for standard deviations have been omitted.

Table 3. Transfer rate of every extraction step (extr) for ISO (top) and vendor’s (bottom) infusion preparation. Values are given in (%) and for overview purposes, their corresponding standard deviations have been omitted. These are derivable from the table S10 (median RSD 12%), from which the transfer rates have been calculated

ISO	extr	EM	EMO	EU	EUO	EV	EVO	HT	HTO	IM	IMO	JB	JBO	JL	JLO	LY	LYO	MC	MCO	RN	RNO	RT	RTO	RTOiso	SC	SCO	SK	SP	SPO	SV	SVO	TD	TDiso1	TDiso2	TDO	TDOiso1	TDOiso2	unkn	unknO
black1	1										92	64	89		87	74	89					64	86	88	61	78				100
	2										8	13	11		11	14	11					14	12	10	16	17
	3											13			1	7						9	2	1	6	4
	4											10				5						14			17
black2	1										90	51	100			71	89					52	87		35	81	100
	2										9	16				15	11					17	12		11	14
	3										1	8				7						9	2		9	6
	4											25				8						23			46
green	1									81	94	64	100	81	100	79	94										46											75
	2									12	6	27		16		14	6										26											15
	3									4		9		3		5											15											6
	4									2						2											13											4
mixed1	1					65	89			79	90					80	90			100	90		100
	2					17	9			13	9					13	9				9
	3					6	3			6	1					5	1				1
	4					13				3						2
mixed2	1					66	86				93					89	90					88	93		74	92		73	91	77	86							90	98
	2					14	13				7					9	10					10	7		14	8		15	9	12	14							10	2
	3					8	1									1						2			8			7		5
	4					12																			4			6		6
pepperm1	1					68	75															67	95		71	94	60	64	89	65	100
	2					17	21															16	5		16	5	27	16	11	14
	3					6	4															8			5		8	8		9
	4					9																8			8		5	11		12
pepperm2	1									91	99					77	99					77	94		69	94		73	93	74	89
	2									5	1					13	1					14	6		20	6		18	7	14	11
	3									4						4						6			5			5		9
	4															6						4			6			4		4
red bush1	1								95	84						83	94					68	88	89	62	87	71	69	90	73	76
	2								5	11						16	6					15	11	9	14	12	14	13	9	19	23
	3									4						1						7	1	2	8	2	8	6	1	4	1
	4									1						1						10			16		7	12		4
red bush2	1	64	96																			66	92	87	62	86	66	61	86	61	85
	2	26	4																			14	7	11	12	12	13	15	13	12	13
	3	7																				8	1	2	9	2	10	9	1	11	2
	4	6																				12			18		11	15		16
senna	1			89	98				98									83	100								100					89	86	89	100	100	100
	2			11	2				2									15														11	14	11
	3																	2
	4
vendor’s	extr	EM	EMO	EU	EUO	EV	EVO	HT	HTO	IM	IMO	JB	JBO	JL	JLO	LY	LYO	MC	MCO	RN	RNO	RT	RTO	RTOiso	SC	SCO	SK	SP	SPO	SV	SVO	TD	TDiso1	TDiso2	TDO	TDOiso1	TDOiso2	unkn	unknO
black1	1										96	71	94		100	81	97					59	95	95	44	92				100
black2	1										97	61	100			80	100					63	98		38	96	100
green	1									91	98	100	100	100	100	87	97										100											94
	2									6	2					8	3																					6
mixed1	1					85	93				98					88	98			100	98		100
mixed2	1					86	93			100	95					95	100					84	100		86	100		83	96	89	100							93	100
pepperm1	1					86	95															81	100		80	99	93	81	100	84	100
pepperm2	1									100	100					90	100					86	100		85	100		83	100	83	100
red bush1	1								100	89						84	98					70	96	98	69	97	80	77	97	80	100
red bush2	1	80	100																			67	97	99	70	97	74	68	97	61	97
senna	1			98	100			100	100									95	99													95	94	95	100	100	100

The ten samples contained 4–8 (median 6) PAFB (including SK) and 5–8 (median 6) PANO in the range 0.1–1100 (median 8.1) ng/g and 0.1–661 (median 23.5) ng/g, respectively. Unexpected PA were found in 60% of the samples. RTOiso was found in black tea 1 and both red bush samples (3–8% of total PA-content); unkn & unknO in green tea and mixed herbal tea 2 (3–4% of total PA-content); and senna contained TDiso1 & 2 and TDOiso1 & 2 (60% of total PA-content). Generally, transfer rates for PAFB and PANO of infusions prepared according to ISO instructions were 52–89% (median 73%) and 87–100% (median 91%), respectively. For samples prepared according to vendor’s instructions, PAFB and PANO transfer rates were 63–95% (median 84%) and 95–100% (median 99%), respectively (Table 3). For both extractions, PAFB were extracted less efficiently than PANO. PA had slightly lower transfer rates in ISO than in vendor’s extraction.

The green tea sample, following vendor’s instructions, was extracted twice: once with water at 80°C and 2 min steeping time, then with boiling water and 1 min steeping time. Regarding PA, the first extraction was already very efficient, since the median transfer rates for PAFB and PANO were 94 and 99%, respectively, and the second extraction quantitatively removed the remaining compounds (Table 4).

Table 4. Sample properties and transfer rates for the ten teas and herbal teas investigated. For transfer rate range (TR) and number of found PA (n) considerations, unkn, SK and TDiso1 & 2 have been assigned to PAFB and RTOiso, unknO and TDOiso1 & 2 to PANO. The composition information refers to vendor’s indication and have not been verified. 1) thyme (30%), elderflower (20%), linden blossoms (20%), anise (5%), blackberry leaves (5%), bitter fennel (5%), rose hip peels (5%), wild thyme (5%), liquorice root (5%). 2) sage, peppermint, blackberry leaves, curly mint, liquorice root, apple, honey granules. 3) Senna angustifolia (Vahl) Batka leaves (current botanical name: Senna alexandrina Mill.)

	black tea 1	black tea 2	green tea	mixed herbal 1	mixed herbal 2	peppermint 1	peppermint 2	red bush 1	red bush 2	senna
	low budget tea	bergamot flavouring	semi-fermented, mid-price segment	1)	2)	low or mid-price segment	reference material (Uhlig et al. 2019)	reference material (P1706-RMRo 2017)	without flavouring	3)
PAFB		incl. SK	incl. SK, unkn		incl. unkn	incl. SK		incl. SK	incl. SK	incl. SK, TDiso1 &2
n	6	4	6	4	8	6	6	7	7	7
range [ng/g]	0.1–36	4–68	0.9–82	0.5–1100	0.7–32	1.2–106	2.6–29	3.8–79	0.9–49	0.4–49
TR1 ISO [%] (median)	61–74 (64)	35–100 (52)	46–81 (77)	65–100 (79)	66–90 (77)	60–71 (66)	69–91 (75)	62–84 (71)	61–66 (61)	83–100 (89)
TR1 vendor’s [%] (median)	44–100 (71)	38–100 (63)	82–100 (94)	85–100 (88)	83–100 (88)	80–93 (83)	83–100 (85)	69–89 (80)	61–80 (69)	94–100 (95)
TR2 vendor’s [%] (median)			0–8 (6)
PANO	incl. RTOiso		incl. unknO		incl. unknO			incl. RTOiso	incl. RTOiso	incl. TDOiso1 &2
n	7	5	5	5	8	5	6	7	7	6
range [ng/g]	4–64	0.1–45	0.6–22	6–661	2–39	2.4–257	7–170	12–175	2–170	6–115
TR1 ISO [%] (median)	78–92 (88)	81–100 (89)	94–100 (97)	89–100 (90)	86–98 (92)	75–100 (94)	89–99 (94)	76–95 (89)	85–96 (87)	98–100 (100)
TR1 vendor’s [%] (median)	92–100 (95)	96–100 (98)	97–100 (99)	93–100 (98)	93–100 (100)	95–100 (100)	100	96–100 (98)	97–100 (97)	99–100 (100)
TR2 vendor’s [%] (median)			0–3 (1)
unknown (% of total content)	RTOiso (6)		Unkn & unknO (3)		Unkn & unknO (4)			RTOiso (3)	RTOiso (8)	TDiso1 & 2 (20)
										TDOiso1 & 2 (40)

PA-content of the infusions prepared from the mixed herbal tea 1 sample varied considerably depending on the extraction technique applied (acidic vs ISO vs vendor’s). While the contents of EV, EVO, RTO and the internal standard D3-ATR were independent from the extraction procedure, the values for IMO, LY, LYO, RN and RNO increased remarkably from acidic extraction over ISO-brew (6 min, capped) to vendor’s brew (10 min, capped). IMO increased from 82 to 142 to 648 ng/g, LY from 141 to 174 to 962 ng/g, LYO from 18 to 34 to 108 ng/g, RN from not detected to 0.5 to 2.5 ng/g and RNO from 59 to 99 to 439 ng/g (Table 3). This corresponds to an increase of approximately x1, x2 and x7 for the three extraction methods. The PANO vs PAFB ratio remained constant despite parameter changes, but the elevated temperature and time apparently favoured the release of additional LY/LYO-group PA. Mulder et al. (2018) stated an up to 2.4-fold content increase for infusions compared to aqueous acidic extraction at room temperature. Further, mixed herbal tea 1 due to its composition is close to a medicinal herbal tea sample, where ‘acidified mixtures of organic solvents at temperatures up to 100°C for short periods of time apparently leads to optimum yields of PA’ (Kopp et al. 2020).

Senna is from Senna angustifolia (Vahl) Batka leaves (current botanical name: Senna alexandrina Mill.), commonly used for medical purposes, i.e. to treat constipation.

Further details on the ten teas and herbal teas used for infusion preparation are presented in Table 4.

Transfer rate influencing parameters

Concluded from the above, the infusion preparations according to vendor’s instructions resulted in higher transfer rates compared to ISO-brews. Here and hereafter, the term ‘transfer rate’ refers to the first extraction step (TR1), if not otherwise noted. For both extraction techniques, transfer rates for PANO were above those of PAFB. While the difference between ISO and vendor’s brews could be caused mainly by the different herb-to-water ratios and/or steeping times and temperatures, PAFB- and SK-transfer to water was obviously hindered by chemical properties, i.e., polarity issues.

Since data may have been influenced by additional parameters like tea or herbal tea type, concentration of analytes, chemical properties of PA and whether the extraction vessel had been capped or not, a statistical evaluation was carried out. To get an overview on influencing parameters, a factor and correlation analysis was performed including 12 variables (Table S13). These accounted for transfer rates (TR1, TR2), (total) PA-content (CONC_TOT), extraction technique (EXTR_TECH), herb-to-water ratio (HWR), solvent temperature of 1^st extraction (TEMP1), steeping time of 1^st extraction (TIME1), capped/uncapped vessel (CAP_UNCAP), analyte type (PAFB_PANO), individual analytes (PA), chemo-structure group of PA (PA_GROUP) and tea or herbal tea type (MATRIX) for all tea and herbal tea samples and brews investigated (n = 351). For the PA_GROUP variable, PA were classified into sub-groups EMG, EMOG, LYG, LYOG, MCG, MCOG, SCG and SCOG (Table S13). Not grouped were JB, JBO, JL, JLO due to their enhanced polarity, i.e. low retention time compared to the other 12-membered macrocyclic diester PA, and SK with its otonecine structure, having an N-methylated tertiary amine group. All these five PA were still considered during statistical evaluation, but treated as single analytes.

In view of data reduction, factor analysis was run limited to a three-factor output, which is presented in Table S14. On factor one, five variables have high loadings, that is, EXTR_TECH, HWR, TEMP1, TIME1 and CAP_UNCAP. Factor one therefore considers the ‘experimental parameters’. On factor two, three variables have high loadings (TR1, TR2, PAFB_PANO = ‘transfer rates & PA-category’) and on factor three two variables (PA, PA_GROUP = ‘detailed properties of PA’). Factors one, two, and three account for 33, 17 and 15% of total variance, respectively. While all three factors together explain 65% of total variance, the factor ‘experimental parameters’ has most importance. The subsequent Spearman correlation analysis (Table S15) reveals the variables correlated with the transfer rates, i.e. mainly PA-category (PAFB_PANO) and with low correlation, steeping temperature and time (TEMP1, TIME1). Little correlation is assigned to concentration (CONC_TOT), extraction technique (EXTR_TECH), herb-to-water ratio (HWR), ‘detailed properties of PA’ (PA, PA_GROUP), tea or herbal tea type (MATRIX) and whether the extraction vessel was capped or not (CAP_UNCAP).

While part of our observations made above are confirmed statistically, others are not clearly identified, but will be investigated subsequently.

The data in Figure S5 demonstrate the dependence of the first extraction transfer rate from the extraction technique applied. While TR1 for PANO is always higher than for PAFB, SK under acidic extraction conditions is close to PANO, but under infusion conditions more resembles PAFB, i.e. the same polarity issue applies both for PAFB and SK. Acidic and ISO extraction share the same HWR (1:50), but differ in TEMP1, TIME1 and extraction solvent. Since mainly PAFB- and SK-transfer rates are affected from different extraction conditions, i.e. ISO-brews having lower TR1, poor analyte protonation seems to be the main cause for their lower transfer rates. On the other hand, the major difference between ISO and vendor’s infusions are the HWR. Lower HWR facilitates transfer of analytes, since vendor’s transfer rates are around 8–34% above ISO. The TR1 for vendor’s PANO is even slightly higher than for acidic PANO (median 99 vs 95%). The outlaying PAFB-values in Figure S5 for ISO (SC in black tea 2) and vendor’s (SC in black tea 1 & 2, RT in black tea 1) might indicate sample-matrix effects for these analytes. Generally, the first transfer rate decreases in the order PANO (vendor’s) > PANO (acidic) ≈ SK (acidic) > SK (vendor’s) > PANO (ISO) > PAFB (acidic) > PAFB (vendor’s) > PAFB (ISO) > SK (ISO). Effects already noted from validation data support these findings. There the efficiency difference of acidic extraction vs ISO-brew was observed as slope-deviation in matrix-matched calibration curves.

The transfer rate 1 vs individual PA, for the three extraction techniques separately, is plotted in Figure S6. PA have been sorted according to PA_GROUP-criteria and except for EMG, whose members have quite heterogeneous transfer rates, the group assignment for the other PA appears reasonable. Additionally, it turns out that the first transfer rates for JB/JBO are similar to SCG/SCOG, JL/JLO to LYG/LYOG and SK to LYG. For a simplified overview, TR1 vs PA_GROUP is shown in Figure 1. For this plot and subsequently, JB/JBO was integrated in the SC/SCO-group and JL/JLO omitted due to its low occurrence (n = 2x2). Generally, transfer rate 1 in brews decreases in the order MCOG > LYOG > EMOG ≈ SCOG > MCG > LYG > SK > EMG ≈ SCG, representing the polarity and proton activity of the analytes under investigation.

Figure 1. Box plot of the 1st transfer rate (TR1) vs PA groups, sorted by PAFB, SK, PANO and grouped by acidic (1st box), ISO (2nd box) and vendor’s extraction (3rd box)

From Figure S7 an increase of transfer rate with decreasing HWR is observed. From HWR 1:50 (n = 117) to 1:100 (n = 84) there is a noticeable increase in transfer rate for all PA-groups, while subsequent transfer rate-increase with decreasing HWR to 1:125 (n = 10) and 1:143 (n = 23) due to the low number of experiments is uncertain. Still, the transfer rates of the latter are always higher than those of HWR 1:50.

For ISO-brews and additionally as a control also for acidic extraction, multiple extractions were applied to study the step-by-step decrease of transfer rates. While for infusion preparation according to ISO, PANO are hardly found in the 3^rd brew (Figure S8), PAFB and SK are found even in the 4^th acidic extraction. Especially EM, EV, JB, RT, SC, SP, SV and as mentioned SK require an acidic ambient to be fully extracted under our conditions applied (Table 3). For acidic extraction of the ten tea and herbal tea samples, the first extraction median transfer rate for PAFB, PANO and SK was 89, 95 and 96%, respectively. The second extraction median transfer rates were for PAFB, PANO and SK 10, 5 and 4%, respectively. Little PA were found in the third extraction with median transfer rates for PAFB, PANO and SK of 1, 0 and 0% (Figure S9, Table S12).

With regard to toxicological aspects and since transfer rates and thus intake estimates depend on them, it is important to note that there is a difference between the ISO- and the vendor’s-transfer rate. Figure 2 displays the ratio of the first transfer rate of vendor’s vs ISO brews. The transfer rate differences for all PA (n = 115) are in the range −17 to 54% (median 8%). The offset line in Figure 2 represents this median difference for all PA-ratios. While the PANO due to their high transfer rates group in the upper right corner of the diagram, the PAFB (and SK) allow for detailed investigation. For simplicity and due to its close hot water extraction property compared to PAFB, SK was assigned to PAFB hereafter. Low abundant PAFB (Number 1: JB, JL, SK, unkn in green tea, SK in both peppermint herbs) tend to have high vendor’s vs ISO transfer rate ratios. Circle 2 is a cluster of EM, EV and RT, SC, SP, SV mainly from peppermint herbs. Numbers 3 refer to red bush samples with data points for RT, SV and LY. Data point number 4 belongs to RT in mixed herbal tea 2 and data points number 5 are from SC and RT in black tea 1 and SC in black tea 2. From this, we conclude that apart from HWR, there is a transfer rate dependence on the chemical properties of the PA, tea or herbal tea type and analyte concentration. To further study the influence of tea or herbal tea type, Figure S10 shows the dependence of the first transfer rate from sample type and brewing technique. Again, a notable difference in transfer rates, TR1(PAFB) < TR1(PANO) and TR1(ISO) < TR1(vendor’s), is observed. Additionally, the data show an enhanced matrix effect for both black tea samples by corresponding low transfer rates for PAFB. In Figure S11, the transfer rates for black teas are plotted against the averaged values of analytes. As previously discussed, the transfer rates especially of RT and SC, but also for JB, are low in black tea herbs. SK and SV do not share this behaviour, having transfer rates of 100%, due to their low abundance, around LOD (Table S10).

Figure 2. Plot of vendor’s vs ISO transfer rates. The dotted line denotes equal transfer rates, while most ratios lie on the straight offset line, representing the calculated transfer rate difference of 8%. Dots denote PAFB (including SK), crosses PANO. Numbers indicate low abundant PAFB (1), peppermint herb cluster (2), red bush (3), mixed herbal (4) and black teas (5) samples. For details, see text

To decide on the influence of steeping time, our data are rather weak. Since ISO-brews were always extracted at a steeping time of 6 min, the only interpretable data are from the vendor’s brews, HWR 1:100, with steeping times of 4 (n = 22), 6 (n = 26) and 10 (n = 36) minutes. Figure S12 depicts an increase in transfer rate for EMG (1%/min), LYG (3%/min), LYOG (1%/min), SCG (4%/min) and SCOG (1%/min) with increasing steeping time, while the information from the other analytes is not interpretable, since their values are around LOD (SK in black tea, 4 min) or have already transfer rates close to 100% at low steeping times (EMOG). Additionally, in case of SCG, the previously discussed matrix effect of black teas probably influences corresponding transfer rates, shown in Figure S12. Another example for the influence of steeping time has already been discussed for the mixed herbal tea 1 sample, where the total PA-content increases from 638 to 2412 ng/g (factor ~4) for a steeping time increase of 4 (from 6 to 10) minutes, while transfer rates are not distinct, though.

Due to the scarce data on possible influence of steeping temperature, 91% (n = 212) of brews were prepared with boiling water and only 9% (n = 22) with water at 80°C, applied even at differing tea and herbal tea-type samples, no inferences can be drawn from our data. For the CAP_UNCAP variable we would predict an increased transfer rate for capped vessels, but since the temperature difference of uncapped vs capped according to our heat dissipation measurements was only 0.7°C/min steeping time, and data origin from dissimilar tea or herbal tea-type samples, no conclusion can be made based on our experimental conditions (Table 1).

Survey

Iced tea beverages

A total of 25 different PA were found in the 46 iced tea beverage samples (Table 2), in the range 1 to 248 ng/L, while total contents ranged from 0 (17%) to 631 (median 40) ng/L. Additionally to the PA found in herbs, EFO, LC and LCO were detected. Since chromatographic peaks for some closely eluting isomers in iced tea beverages sometimes tended to be distorted and not well resolved, the determined values were ascribed to the main peak. This was especially the case for LY/LYO group members (IM, IMO, LY, LYO, RN, RNO) and for SC/SV. Figure 3 shows the distribution of the total PA-content for the 46 samples, lowest to highest. Two samples lie significantly above the others, a beverage declared to contain herbs from the alps (518 ng/L, 4 PA) and a beverage made from black tea extract (613 ng/L, 8 PA). While details on found PA for all samples are given in Table S16, the insert of Figure 3 displays the total PA concentration vs declared extract origin. Black tea extracts and alpine herbs contents lie above green tea extract, mixed herbal, unknown origin of extract, peppermint and fennel extract (Table S17). From the PA found in the beverages with undeclared origin of extract (HTO, IMO, LY, LYO), we conclude that these iced teas were prepared with black tea extracts. The predominant PA found in all positive samples are from the LY/LYO-group, for four beverages made of black tea extract additionally from the SC/SCO-group, being consistent with the PA composition of black tea herbs. Three out of five beverages from alpine herb extracts have high loads on IM/IMO, LY/LYO and RNO, especially on the N-oxides. This pattern is possibly due to a contamination with Eupatorium cannabinum (Roeder 1995; Kast et al. 2018). Literature quotes the PA-content of iced tea beverages as 0 (33%) −250 (median 40) ng/L, n = 9 (Engeli 2014) and 0 (25%) −7590 (median 30) ng/L, n = 28 (Picron et al. 2018). Both median values are in good agreement with our findings.

Figure 3. Total PA-content of iced tea beverages, sorted lowest to highest (n = 46). Letters in sample names denote unknown extract origin (?), alpine herbs (a), black tea (b), fennel (f), green tea (g), herbal tea (h), mixed herbal (m), peppermint (p) and numbers indicate the quantity of PA found. The insert of the Figure displays the PA-content distribution depending on declared origin of extract

The ratio PANO vs PAFB for iced tea beverages is in the range 0.05–7.4 (median 0.6, n = 27). A ratio below one might indicate additional herb or tea extract treatment, as mentioned above. For tea and herbal tea, this was especially the case for green tea herb, while for iced tea beverages ratios of black and green tea were below one and those containing mixed herbal, peppermint and alpine herb extracts were above one. Possibly, for those with ratio below one, extracts have been exposed to additional treatments like drying.

Figure S4 compares the median values for PANO vs PAFB-ratios derived from tea and herbal tea samples and iced tea beverages.

Products over time

Three herbal tea products with seven different lots and five iced tea beverages were monitored from 2013 to 2020 (Table S18). While the package for these samples remained the same over time, the contents obviously changed considerably, as shown by PA amounts found. In herbal teas, the observed deviations within lots are always smaller than the annual difference, therefore allowing for a trend interpretation over time. While a significant decrease for total PA in the mixed herbal tea (>80%) and the two peppermint herbs (>95%) was noted, content of PA tend to rise in four out of five ice tea beverage products (>80%).

Of course, the modest number of samples measured does not allow for a general statement, but there are signs of improvement concerning producer’s awareness, at least for the herbal tea products, but further surveys are necessary to judge future development. That PA-levels in different teas and herbal infusions are lower than in previous years was also observed by EFSA (2016).

Additionally, Table S18 demonstrates that homogeneity of sample is an issue in PA quantification. Comparing the relative standard deviations from herbs and beverages, median values of 14% (range 4–38%) and 4% (range 1–33%) result, respectively. While beverages represent homogeneous solutions, herbs require extended homogenisation preceding analysis.

Atropine as additional contaminant

Atropine was found in 50% of the teas and herbal teas, and in 13% of the iced tea beverages, quantified by the internal standard D3-atropine under matrix-matched conditions. The green tea herb sample, red bush 1 and both black teas contained 1 ng/g atropine; senna 2 ng/g; and both mixed herbal teas 4 ng/g. Six of the 46 iced tea beverages contained atropine in the range 2–65 (median 13) ng/L (Table S16). Scopolamine, whose concentration according to our experience is around 14% of that of atropine in tea and herbal tea samples (n = 48, 18% positive for scopolamine, not shown), was not detected.

Intake estimation and evaluation

From our own and most of the literature findings (Lüthy et al. 1980; Engeli 2014; Mulder et al. 2018) a general transfer rate of 80–100% for PA must be assumed for infusions, especially when prepared according to vendors’ and not ISO instructions (this study).

Assuming a consumption of 0.5 L of infusion per day for a regular consumer and applying the median values that we determined, that is, a PA-content of 22 ng/g herb, (vendor’s) transfer rate of 95% and a herb-to-water-ratio of 1:120, would lead to a daily intake of 87 ng of PA. For an individual of 70 kg body weight this leads to an MOE of about 190,000 according to the assessment of EFSA (2017). This means that for an adult regular consumer of herbal brews the intake from this source alone would be of low concern. However, if the mean value of 87 ng/g is used instead of the median for calculations, the MOE drops to 48,000. Assuming additionally a child of 15 kg body weight consuming the same beverage, we are already at the threshold of concern. Moreover, there are products far above the mean and other PA-sources are not yet taken into account. Therefore, efforts to reduce impurities in tea and herbs for herbal infusions must continue. The PA-intake from iced tea beverages is considerably lower than from infusions, we estimate by a factor of 6, and together with a worldwide consumption of 37 billion litres (2016; Conway 2020), i.e. a factor of 8 less than infusions, appears negligible, but might be important for high-level consumers or children.

Conclusion

Regarding quantification, it is important to screen samples for unexpected PA. Although quantitation of unknowns results in estimations only, this aspect should not be overlooked. In some of our samples, TD/TDO-isomers and the often ignored EV/EVO accounted for 60 and 17% of total content, respectively. While iced tea beverages represent uniform samples, tea and herbal tea require extended homogenisation prior to analysis. On the other hand, this is quite different from the consumer’s way of preparing infusions.

The undocumented feature of the applied SPE material allowed for selective PANO vs PAFB separation in a two-step elution. Together with fingerprint peaks from MS and MS/MS, this is a valuable qualitative identification tool, which could also be of interest and exploited for preparative purposes.

The median transfer rates for PA from brews according to ISO and vendor’s instructions were 87 and 95% (n = 2×117), respectively. In our data, decreasing herb-to-water ratio and increasing steeping time raised the transfer rate of PA, while tea or herbal tea type, proton affinity of PA and concentration had additional influence. Since there is a pH-dependence on the transfer rate especially for PAFB and SK, acidic tea or herbal tea types would suggest enhanced transfer rates, but it is unlikely that the pH of brewed herbs would be low enough to have an effect on PA-extraction. Additionally, more acid is expected to be released from tea or herbal tea with increasing brewing time, an issue that was not investigated by us.

The findings on transfer rate influencing parameters are in agreement with the results of the performed correlation analysis involving 12 parameters.

The herein commonly applied acidic extraction technique used to determine total PA-content, 0.05 M sulphuric acid at room temperature, turned out to be suboptimal in one out of ten cases, where boiling water gave rise to a factor of about seven higher PA-content instead.

In view of toxicological relevance, it is important to analyse tea or herbal tea samples as close to consumer habits as possible. It is a challenge to predict PA-content of infusions from the herbs used, since transfer rates of PA depend on many factors and may vary considerably. Therefore, PA-content of infusions rather than their herbs should be determined and the brew preferably be prepared according to vendor’s and not ISO instructions, in order to assess realistic consumer exposure.

Although there is growing awareness of tea and herbal tea contamination with PA-containing weeds, further efforts on solving and surveying this issue are required. Still, PA-free teas and herbal teas are most easily found among loose herbs, while bagged products are typically blended to attain a constant flavour profile over time.

Concerning risk assessment, intake from iced tea is low, but could be of concern for high-level consumers and children.

Acknowledgments

We are grateful to Simone Staiger, who drew our attention to senna.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed on the publisher’s website.

Correction Statement

This article has been republished with minor changes. These changes do not impact the academic content of the article.