Quinolizidine alkaloids and phomopsins in lupin seeds and lupin containing food

Abstract

In recent years there has been growing interest in replacing (genetically modified) soya by lupin. Lupin seeds, flours and lupin containing food have been analyzed in order to assess the relevance of a potential health hazard given by mycotoxins and/or naturally occurring alkaloids. Since not all important alkaloids used for quantitation were commercially available, isolation of lupanine, 13alpha-hydroxylupanine and angustifoline from lupin flours of high alkaloid contents was performed. Alkaloids were analyzed by GC-MS/GC-FID in parallel, while the phomopsin mycotoxins were analyzed by ELISA, since chromatographic methods were not sensitive enough and required time-consuming sample cleanup. The analyzed lupin containing foods were free of phomopsins. In foods where lupin was only a minor constituent the alkaloid content was of no concern. However, roasted lupin beans intended as coffee surrogate had alkaloid contents close to the Australian intervention limit of 200 microg/g.

Full text

Journal of Chromatography A, 1112 (2006) 353–360
Quinolizidine alkaloids and phomopsins in lupin seeds
and lupin containing food
Hans Reinhard ∗, Heinz Rupp, Fritz Sager, Michael Streule, Otmar Zoller
Swiss Federal Office of Public Health, Division of Food Science, CH-3003 Bern, Switzerland
Available online 15 December 2005
Abstract
In recent years there has been growing interest in replacing (genetically modified) soya by lupin. Lupin seeds, flours and lupin containing
food have been analyzed in order to assess the relevance of a potential health hazard given by mycotoxins and/or naturally occurring alkaloids.
Since not all important alkaloids used for quantitation were commercially available, isolation of lupanine, 13␣-hydroxylupanine and angustifoline
from lupin flours of high alkaloid contents was performed. Alkaloids were analyzed by GC–MS/GC–FID in parallel, while the phomopsin
mycotoxins were analyzed by ELISA, since chromatographic methods were not sensitive enough and required time-consuming sample cleanup.
The analyzed lupin containing foods were free of phomopsins. In foods where lupin was only a minor constituent the alkaloid content was
of no concern. However, roasted lupin beans intended as coffee surrogate had alkaloid contents close to the Australian intervention limit of
200 ␮g/g.
© 2005 Elsevier B.V. All rights reserved.
Keywords: Preparative chromatography; Lupins; Quinolizidine alkaloids; Phomopsins; Dual column GC–MS/GC–FID; LC–DAD–ECD; LC–MS/MS; ELISA
1. Introduction
Lupins are leguminous plants like soybeans. Since lupin and
soya have comparable nutritive characteristics [1], lupins have
been discussed as possible substitute of genetically modified
soya in human foodstuffs. From the genus Lupinus more than
400 species are known, from which only four are of agronomic
interest: Lupinus albus (white lupin), Lupinus angustifolius (nar-
row leaf or blue lupin), Lupinus luteus (yellow lupin) and Lupi-
nus mutabilis (Andean lupin). Main producing countries are
Australia (L. angustifolius), Russia (L. luteus) and Poland (L.
luteus). In Europe especially L. albus and L. luteus are used
as green forage, as manure or are intended for human nutri-
tion. Apart from their benefits [1], lupins contain antinutritive
components such as saponins, tannins and flavonoids in varying
concentrations [2], but also naturally occurring bitter principles.
More than 150 alkaloids of the quinolizidine, piperidine and
indole group (Fig. 1) are known at concentrations up to 6%, con-
ferring the plant resistance to pathogens and herbivores. Though,
plant breeders have developed so called sweet lupins with alka-
∗Corresponding author. Tel.: +41 31 324 9386; fax: +41 31 322 9574.
E-mail address: hans.reinhard@bag.admin.ch (H. Reinhard).
loid contents below 0.05%, adequate for animal and human
consumption. The alkaloids vary in toxicity, i.e. teratogenicity
of anagyrine [3] and ammodendrine [4] as well as neurotoxicity
of other alkaloids has been documented [5]. Moreover, seeds
of L. angustifolius may additionally be infected by the fungus
Phomopsis leptostromiformis, which could result in contamina-
tion with phomopsin mycotoxins [6] and give rise to lupinosis, a
severe hepatic disease [7–9]. From the five phomopsins reported
in literature [10], only the structures of phomopsin A and B have
yet been elucidated [11] (Fig. 2). Regulations in Australia and
Great Britain demand to comply with a maximum contamination
of 200 ␮g/g alkaloids and of 5ng/g phomopsins in lupin-based
food [12]. Apart from toxicity, the possible allergenic potential
of lupins is under discussion; cross-allergenicity of peanut and
lupin has been shown [13]. Recently, several cases of allergenic
reactions in humans have been reported in Switzerland [14] but
lupin has not been included in the new EU and Swiss legislation
on the mandatory labeling of allergens [15].
The aim of the present work was to unambiguously detect
and quantify lupin alkaloids and mycotoxins in order to assess
a possible health hazard for food products intended for human
consumption on the Swiss market. To reach this objective, it
was necessary to isolate non-commercial alkaloids and develop
analytical methods for a variety of analytes.
0021-9673/$ – see front matter © 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.chroma.2005.11.079

354 H. Reinhard et al. / J. Chromatogr. A 1112 (2006) 353–360
Fig. 1. Exemplary structures of important representatives from quinolizidine,
piperidine and indole alkaloids found in lupins.
Fig. 2. Structure of phomopsin A (PhA), one of the five documented pho-
mopsins. In phomopsin B (PhB), the chlorine atom is replaced by hydrogen.
2. Experimental
2.1. Chemicals and immunoreagents
Gramine (Ord.# G1,080-6) was from Aldrich, lupinine
(Ord.# 01659L) and ␣-isolupanine perchlorate (Ord.# 14343L)
were form Apin (Abingdon, GB), citric acid monohydrate
(Ord.# 27490), N,O-bis(trimethylsilyl)trifluoroacetamide
(BSTFA)/1% trimethylchlorosilane (TMCS) (Ord.# 15238),
N,O-bis(trimethylsilyl)acetamide (BSA, Ord.# 15242),
orthophosphoric acid (85%, ord.# 79620) and pyridine (Ord.#
82702) were from Fluka (Buchs, Switzerland), bovine serum
albumin (BSA, Ord.# A9085), cytisine (Ord.# C2899),
sparteine (Ord.# S2126), phomopsin A (PhA, Ord.# P4829),
tetramethylbenzidine (TMB, Ord.# T2885), Tween 20 (Ord.#
P1379) and phosphate buffered saline tablets (PBS; Ord.#
P4417) were from Sigma, hydrogen peroxide (30%, ord.#.
8597), sodium azide (Ord.# 822335), sodium acetate trihydrate
(Ord.# 6267), sodium hydrogencarbonate (Ord.# 6329), sodium
carbonate (Ord.# 6392), sodium chloride (Ord.# 6404), sodium
hydroxide (Ord.# 6498) and sulfuric acid (Ord.# 731) were
from VWR Int. All these chemicals and the solvents used were
of analytical grade and applied without prior purification. High
purity water was produced by an EasyPure UV system (Mod.
D7402, Barnstead, Dubuque, US). Anti-phomopsin IgG and
phomopsin-horseradish peroxidase enzyme conjugate were
donations from CSIRO, Australian Animal Health Laboratory,
Geelong, Australia.
2.2. Instrumentation
2.2.1. GC
Gas chromatographic separations were performed on a
HP6890 (Agilent), coupled to a mass spectrometer (HP5973,
Agilent) and a flame ionization detector (FID, Agilent). Two
identical capillary columns (HP-1MS, length 30 m, i.d. 0.25 mm,
film thickness 0.25 ␮m, Agilent) were fitted by a two-hole fer-
rule to the injector and to the mass spectrometer and the FID,
respectively. The combined data acquisition of both detectors
resulted in different retention times of the corresponding ana-
lytes. This was overcome through a tool in the acquisition
software (MS ChemStation G1701BA rev. B.02.00, operated
in enhanced mode, Agilent) which requires a two point linear fit
to align the signals of two detectors. As alignment peaks hex-
adecane (C16) and hexacosane (C26) were chosen, also used
as reference peaks for retention index calculations (see Section
2.3.4). Derivatized samples were analyzed on a HP5890 gas
chromatograph coupled to a HP5972 mass spectrometer (both
from Agilent) and separated on a HP-1MS capillary column as
cited above.
2.2.2. Preparative LC
The system consisted of a Merck-Hitachi L6200 gradient
pump, an L3000 DAD (both from VWR Int.), a fraction collec-
tor (FC 203, Gilson) and a degasser (Gastorr GT-103, Omnilab,
Mettmenstetten, Switzerland).
2.2.3. Analytical LC
The chromatographic system consisted of two LC-10AD
high pressure gradient LC pumps (Shimadzu, Kyoto, Japan),
a static mixer (Lee Micro Mixer 10 ␮l, Westbrok, US) a SIL-
10A auto injector (Shimadzu) and a SPD-M10AVp UV diode
array detector (DAD, Shimadzu) with an electrochemical detec-
tor (ECD, Coulochem II, Esa, Chelmsford, US) in series or of a
triple quadrupole mass spectrometer system (MS/MS, API 3000,
ABI-MSD Sciex, Concord, Canada) as detector, respectively. A
Gastorr GT-104 degasser (Omnilab) was used and the temper-
ature of the column was kept at 20 ◦C by a Pelcooler column
oven (Portmann, Biel-Benken, Switzerland). Reversed phase
materials used were custom fills from Macherey-Nagel (Oensin-

H. Reinhard et al. / J. Chromatogr. A 1112 (2006) 353–360 355
gen, Switzerland) and included the stationary phases Inertsil
ODS3 and Spherisorb ODS1, both with 5 ␮m particle size,
which were all in columns of 250 mm ×4 mm and additionally
equipped with a pre-column Spherisorb ODS2 (8 mm ×4 mm,
5␮m particles) for LC–DAD–ECD. For LC–MS/MS Inertsil
ODS2 columns, 125 mm ×2 mm, 3 ␮m particles, were used.
2.3. Alkaloids
2.3.1. Isolation of reference alkaloids
For extraction 50 g of freshly ground lupins (particle size
<0.5 mm, milled on a Mod. ZM1, Retsch, Haan, Germany) of
a bitter variety of L. angustifolius were homogenized at high
speed in 500 ml 0.5 M hydrochloric acid for 5 min with a poly-
tron mixer (Kinematica, Littau, Switzerland). The slurry was
centrifuged at 4600 rpm for 25 min at 15 ◦C (Labofuge 400R,
Heraeus, Z¨
urich, Switzerland). A slightly turbid extract resulted,
which was adjusted to pH 9.1 with concentrated ammonia solu-
tion. The extract, ca. 250 ml, was then partitioned over 5 ×50 ml
ChemElut cartridges (Ord.# 1219-8009, Varian) and after 20 min
eluted with 2 ×100 ml of dichloromethane. The combined elu-
ates were evaporated to dryness. Depending on the lupin species
chosen the extraction resulted 300–600 mg crude isolate.
For pre-cleaning, crude isolate was dissolved in portions of
ca. 100 mg in diethyl ether saturated with 5% ammonia (v/v) and
charged on a 5 g silica cartridge (Ord.# 36950, Waters) which
had been conditioned with diethyl ether. The coated isolate was
then washed with 10 ml of solvent. The main component of
this elution fraction was lupanine. Subsequently, methanol was
added to the solvent in four steps (10, 25, 50 and 75%, v/v,
respectively) and elution continued with increasing methanol
content. The 10% methanol fraction contained lupanine, 13␣-
hydroxylupanine and angustifoline, the 25% methanol frac-
tion contained mainly 13␣-hydroxylupanine with portions of
lupanine and angustifoline, while the fractions with higher
methanol content were discarded. This pre-cleaning procedure
was applied several times until about 300–400 mg substance was
gained per fraction.
In a next step, isocratic column chromatography was run
on a 250 mm ×10 mm Nucleosil C18 column or on the better
performing 250 mm ×4.6 mm C18ec column (both with parti-
cle size 5 ␮m, Macherey-Nagel) with 5mM KH2PO4-methanol
(94:6) as mobile phase. Injection volumes were 20–100 mg iso-
lates. Fractions of about 20 ml were collected and analyzed.
From several separations, fractions of identical content were
combined and concentrated. This procedure led to an alkaloid
purity of about 90%. The methanol content of the mobile phase
was then changed to 12% (v/v, pH 2.7), resulting in shorter
retention times. In the subsequently resulting fractions only the
main zones were collected, while the front, end and mixed zones
were fed as collected fraction to repeated chromatographic sep-
aration, from which again the main zones were gained. The
collected main fractions of all three alkaloids had colors from
yellow to light brown. A further cleanup on silica cartridges
as described above largely removed this inconvenience. At this
point alkaloid purity had reached about 95%. Next, gradient
chromatographic separation on a 300 mm ×7.8 mm silica col-
umn (filled with C-Gel C640, particle size 15–35 ␮m, Zeochem,
Uetikon, Switzerland) with diethyl ether saturated with ammo-
nia (5%, v/v)–methanol as mobile phase was applied, varying the
eluent from (90:10) to (70:30). The collected main zones were
concentrated under nitrogen flow at <40 ◦C and remaining water
and ammonia removed through azeotropic evaporation with
ethanol. The residues were dissolved in methanol and passed
through a syringe filter (0.45 ␮m, PTFE, Titan, Infochroma,
Zug, Switzerland) in order to remove insoluble particles. Stor-
age of methanolic alkaloid solutions at 5 ◦C overnight resulted in
white, voluminous precipitates for all three alkaloids, identified
as fatty acids by GC–MS. To remove the precipitates, a cleanup
over magnesium silicate cartridges (0.5 g, Florisil Sep-Pak, ord.#
43405, Waters) was carried out with diethyl ether saturated with
ammonia (5%, v/v)–methanol. Methanol content was varied
from 0 to 5% (v/v). This procedure removed also the residual
slight coloration. The resulting isolates were of sticky, syrup-like
consistency. After removing insoluble components (<1%), crys-
tallization from hexane resulted in pseudo-crystalline angusti-
foline and 13␣-hydroxylupanine, while lupanine remained oily.
2.3.2. Sample preparation
Seeds were frozen in liquid nitrogen and milled to a particle
size of <0.5 mm (Mod. ZM1, Retsch). One gram of dry lupin
flour or up to 2.5 g of fresh lupin food product was weighed
in a centrifuge tube. Fifteen milliliters 0.5 M hydrochloric acid
were added and samples were kept for 10 min at room temper-
ature. The suspension was then homogenized for 3 min with
a polytron mixer, chilled to 5◦C and subsequently centrifuged
at 6000 rpm for 15 min at 5 ◦C (Labofuge 400R). The super-
natant was decanted in a tared flask and the procedure described
above was repeated with addition of hydrochloric acid. Super-
natants were combined, adjusted to pH 10 by adding ammonia
(aqueous solution 25%, v/v) and weighed. Twenty milliliters of
the alkalized solution were applied to a 20 ml ChemElut car-
tridge (Varian) and after 15 min the alkaloids were eluted by
two-fold 40 ml dichloromethane application. The solvents were
evaporated to dryness under reduced pressure. The residue was
dissolved in 5 ml methanol, transferred into a vial, evaporated to
dryness under nitrogen flow and stored at −20 ◦C. For GC anal-
ysis, 1 ml methanol containing 50 ␮g/ml caffeine for response
factor calculation, 5 ␮g/ml C16 and 5 ␮g/ml C26 for retention
index calculation was added. Subsequently, 1 ␮l of this solution,
bitter lupin extracts only after further dilution, was injected.
2.3.3. Derivatization
In addition to the sample preparation described above (Sec-
tion 2.3.2.), derivatization with BSA/TMCS/pyridine 8:1:1
(v/v/v), 10 min at 60 ◦C was used in cases were the analysis
was hampered by matrix peaks or lack of analytes detectability
without derivatization (see Section 3.1.2).
2.3.4. GC conditions
The GC equipped with two identical capillary columns in
parallel (HP-1MS, length 30 m, i.d. 0.25 mm, film thickness
0.25 ␮m) was operated with a temperature program starting at
80 ◦C held for 1 min, increasing to 300 ◦Cat15◦C/min and held

356 H. Reinhard et al. / J. Chromatogr. A 1112 (2006) 353–360
for 7 min at the final temperature. As carrier gas helium at a con-
stant flow of 1 ml/min was used. The injector was operated in
split mode (20:1) at a temperature of 250 ◦C. Injected volumes
were 1 ␮l. The same conditions applied for the GC HP5890, on
which derivatized samples were separated. The FID was oper-
ated with hydrogen at 40 ml/min, air at 450 ml/min, nitrogen as
makeup gas at 45 ml/min at a temperature of 250 ◦C.
Standard solutions were prepared by dissolving 2–5 mg alka-
loid as free base in 10 ml methanol. While gramine solution had
to be prepared daily, quinolizidine standards proved to be stable
for several days. Quantification was carried out by evaluating
the signals from the FID, unless derivatization was necessary.
In matrix spoilt samples 13␣-hydroxylupanine and angustifo-
line had to be quantified through peak evaluation from mass
spectra at m/z= 246 and 265, respectively (see Section 3.1.2).
For FID quantification, prior determination of response fac-
tors (RF) was required: peak areas of standard alkaloids were
related to caffeine (RF≡1), resulting in RF-values for lupa-
nine (0.45), 13␣-hydroxylupanine (0.58), angustifoline (0.59),
gramine (0.69) and sparteine (0.37). For ␣-isolupanine the RF
of lupanine was used.
Retention indices (RI) were calculated according to [16], with
hexadecane (C16, RI≡1600) and hexacosane (C26, RI≡2600)
as reference peaks. Unexpected trace alkaloids were identified
by RI and MS [24].
2.4. Phomopsins
2.4.1. ELISA conditions
The competitive ELISA was carried out as described by Than
et al. [17]. Briefly, microtiter plates (Ord.# 439454, 96 F bottom
wells from Nunc, Denmark) were coated with anti-phomopsin
antiserum in aqueous carbonate/NaN3coating buffer. The plates
were then incubated overnight at 4 ◦C. After 17–20 h of adsorp-
tion, the plates were washed with NaCl/Tween 20 aqueous
solution. Standards (range 5–1280 pg PhA/100 ␮l) or samples
(50 ␮l/well) and diluted phomopsin-enzyme conjugate (dilution
1:500, 0.5% BSA in PBS) were added and incubated after short
agitation for 3 h at room temperature. Following washing, TMB
substrate solution was added and after 15 min of incubation the
reaction was stopped by adding 0.5 M sulfuric acid. The opti-
cal density was measured at 450 nm (EIA Reader Mod. 2550,
BioRad) and a logistic dose response curve fitted through the
phomopsin standards data from which the PhA + B concentra-
tion of unknown samples was calculated. All determinations
were performed in four-fold.
According to Than [18], samples were prepared as fol-
lows: a representative sample of seed, e.g. 50g, was soaked
in 250 ml methanol–water (4:1, v/v) overnight at room temper-
ature and homogenized the next day for 3 min at high speed.
The resulting paste was stirred for 2 h and then centrifuged
at 2000 rpm for 10 min (Labofuge 400R). The supernatants
were stored at −18 ◦C until analysis. Samples were diluted
1:10 with assay buffer before analysis. Flour samples were
treated correspondingly to the above, only omitting the unnec-
essary homogenization, but were stirred during at least 2 h of
soaking.
2.4.2. LC–DAD–ECD
The mobile phase consisted of methanol-133 mM orthophos-
phoric acid pumped at a flow rate of 1 ml/min. Gradient elution
was achieved as follows: eluent composition changed from
(80:20) to (40:60) in 30 min, raised to (80:20) within 1 min
and the system conditioned for 5 min at (80:20). The DAD
was set to detection wavelengths 210 and 290nm, respectively.
PhA exhibits UV absorption of λMeOH
max nm (ε): 210(52297) [this
work], 210(50023) [19] and 290(16154) [this work], 290(16030)
[19], respectively. The ECD was equipped with an amperomet-
ric analytical cell operated in DC mode (Mod. 5040 from Esa,
cell potential 600 mV, gain range 100 nA).
2.4.3. LC–MS/MS
The mobile phase consisted of methanol-formic acid (0.1%,
v/v) at a flow rate of 0.2 ml/min. During linear gradient elution,
the eluent composition changed from (70:30) to (10:90) within
15 min, was then raised to (70:30) and kept constant for 5 min.
Under these conditions PhA eluted after 6 min. The mass spec-
trometer was equipped with an electrospray ion source operated
in positive ion mode at an ionization voltage of 5kV, heated at
350 ◦C. The nebulizer gas flow was set to 8l N2/min. The mass
transitions used for quantification were (precursor →product):
789 →323 m/zand 789 →226 m/z, acquired during 150 ms
(dwell time) each. Collision energies to generate the product
ion 323 and 226 m/zwere 39 and 57 eV, respectively. Nitrogen
was used as collision gas at a pressure of ∼1 Pa within the colli-
sion quadrupole (readout correct according to [20]). Both mass
selective quadrupoles were operated at unit resolution condi-
tions.
3. Results and discussion
3.1. Alkaloids
3.1.1. Physicochemical properties of isolates
To confirm identity and purity of isolates, several
physicochemical investigations were applied (Table 1).
Chromatographic purity was sufficient for lupanine and
13␣-hydroxylupanine, while angustifoline contained minor
impurities. Elemental analysis revealed a somewhat increased
oxygen content for 13␣-hydroxylupanine and angustifoline.
Retention index, response factor, melting point and mass
spectral fragments compare well with literature, angustifoline’s
melting point being somewhat low, though. The RF of 13␣-
hydroxylupanine was sensitive to the degree of GC injector
liner contamination and could subsequently rise to an RF value
of 0.75. The susceptibility of 13␣-hydroxylupanine had been
reported before [25], where a vast relative standard deviation
(RSD) for RF determination was described.
3.1.2. GC–FID and GC–MS analysis
The developed method for alkaloid determination proved to
be reliable and stable. Unfortunately, it was not possible to quan-
tify alkaloids by FID and confirm identity by MS. Lupin flours
from Australian provenience, i.e. from L. angustifolius, required
sample derivatization, since matrix peaks in the chromatograms

H. Reinhard et al. / J. Chromatogr. A 1112 (2006) 353–360 357
Table 1
Physicochemical properties of isolated alkaloids
Lupanine 13␣-Hydroxylupanine Angustifoline
GC–MS purity, cold on column (%) >99 >99 97
GC–MS purity, BSTFA derivatization (%) >99 >99 94
LC-DAD purity at 200 and 220 nm (%) >99 >99 97
Elemental analysis (calc. value, conformity) C 72.45 (72.54, 99.87%), H
9.66 (9.74, 99.18%), N 11.19
(11.28, 99.20%), O 6.34 (6.44,
98.44%)
C 67.77 (68.15, 99.44%), H 9.17
(9.15, 100.21%), N 10.17 (10.06,
95.94%), O 12.75 (12.10, 105.37%)
C 71.23 (71.76, 99.26%), H 9.31
(9.46, 98.41%), N 11.49 (11.95,
96.15%), O 6.17 (6.83, 119.7%)
Melting point (ref. value) (◦C) 38 (41 [21]) 171–172 (169.5 [21]) 74 (79–80 [22])
IR νKBr
max (cm−1) 2930, 1640, 1440 ([23]) 3420, 2925, 2360, 1630, 1440 3074, 2910, 1640, 1440
Retention index (ref. value)a2204 (2165 [24]) 2447 (2410 [24]) 2111 (2083 [24])
Response factor (ref. value)b0.45 (0.51 [25]) 0.58 (0.63 [25]) 0.55 (0.50 [25])
MS [m/z] from EI (relative intensity, %) M+= 248 (59); 136 (100), 149
(60), 248 (59), 150 (44), 247
(38). [24,26-28]
M+= 264 (41); 152 (100), 246 (55),
165 (44), 264 (41), 134 (38).
[24,26–28]
M+= 234 (0.2); 193 (100), 112
(54), 150 (17), 194 (13), 55 (10).
[24,27, 28]
aAccording to Kovats [16], with C16 and C26 as reference peaks.
bCompared to caffeine [25].
masked the alkaloid content (Fig. 3). Derivatization led to peak
shifts for 13␣-hydroxylupanine and angustifoline, allowing their
unambiguous detection. Lupinine could only be detected after
derivatization, being one of the minor constituents (0.6–2% of
Fig. 3. GC–MS chromatograms for underivatized seeds of L. luteus,L. albus and
L. angustifolius. Note that the ordinate is in log scale in order to visualize small
peaks: (1) C16; (2) gramine; (3) caffeine (IS); (4) sparteine; (5) ammodendrine;
(6) angustifoline; (7) ␣-isolupanine; (8) lupanine; (9) 13␣-hydroxylupanine and
(10) C26.
total alkaloid content), though. Derivatized alkaloids were quan-
tified by MS although FID would have been more favourable due
to better reproducibility. The limit of detection for trimethylsi-
lyl derivatives was a factor of 6–12 lower than for underivatized
alkaloids. We first used derivatization with BSTFA/TMCS, but
since for angustifoline two products formed and conversion was
incomplete, this procedure was dropped. Instead, sample deriva-
tization with BSA/TMCS/pyridine was used (see Section 2.3.3),
giving still rise to two angustifoline products but at complete
conversion, though. The first product mass spectrum showed
the expected loss of the allyl side chain as base peak and high-
est mass ion (m/z= 265, M-41+), while the second product
spectrum contained the ions m/z= 335 (14%), 309 (60%) and
265 (100%). Silylation artifacts are well known from literature
and in our case it seems that a CO2-incorporation, documented
especially for amino acids [29], i.e. carboxylate formation had
happened. The fragments of the second product are tentatively
interpreted as M-15+(loss of methyl), M-41+(loss of allyl
side chain) and M-85+(loss of CO2and allyl side chain after
rearrangement), respectively.
From the teratogenic alkaloids reported, only ammodendrine
could be detected in seeds of L. albus in trace amounts, but was
not detectable in ready-to-consume food.
Recoveries from a flour sample spiked at a 10 ng level
were 115% for lupanine, 43% for 13␣-hydroxylupanine, 49%
for angustifoline, 93% for cytisine and 139% for sparteine.
For trimethylsilylated 13␣-hydroxylupanine a recovery of 90%
resulted. From a standard solution recoveries (±RSD, n=3)
were: 99% (±6%) for lupanine, 94% (±11%) for 13␣-
hydroxylupanine and 50% (±28%) for angustifoline. The lim-
its of detection (LOD, S/N= 3) related to flour were 1 ␮g/g
for lupanine, 2 ␮g/g for angustifoline and 6 ␮g/g for 13␣-
hydroxylupanine. For trimethylsilylated angustifoline and 13␣-
hydroxylupanine the LOD was 0.3 and 0.5 ␮g/g, respectively.
To check reproducibility, a lupin flour was processed five
times and the RSD found was 3% for lupanine and 7% for
13␣-hydroxylupanine, while total alkaloid content was 97 and
77 ␮g/g, respectively. For a trimethylsilylated lupin flour pro-

358 H. Reinhard et al. / J. Chromatogr. A 1112 (2006) 353–360
Table 2
Content of alkaloids and phomopsin A and B in investigated seeds and flours
Seeds and flours AlkaloidsaPhomopsin A + B
“Bitter” (␮g/g) n“Sweet” (␮g/g) n(ng/g) n
L. albus 202 1 143–226 5 <0.1 4
L. luteus 0 500–895 4 0
L. angustifolius 10800–19800 3 44–2120 median: 138 18 <0.1–20b8
aSum of standard alkaloids (see Sections 2.3.4 and 3.1.2).
bOne unique positive sample, see text.
cessed five times, 13␣-hydroxylupanine was found to have an
RSD of 14% at a total alkaloid content of 70 ␮g/g.
The found alkaloid contents of seeds and flours are given in
Table 2, while Fig. 3 depicts typical chromatograms of the three
species studied. The samples have been grouped according to
their origin in L. albus,L. luteus and L. angustifolius and into
bitter or sweet varieties, according to supplier’s indications. For
quantitation, the found standard alkaloids (see Section 2.3.4)
were summed. For samples from the L. albus group, included
alkaloids were lupanine, 13␣-hydroxylupanine and angustifo-
line. The first two were directly quantifiable by FID, the lat-
ter was quantified by MS after trimethylsilylation. Traces of
␣-isolupanine, dihydroxylupanine, ammodendrine, albine and
13␣-angeloyloxylupanine or 13␣-tigloyloxylupanine were also
found, identified by RI and MS [24].IntheL. luteus group
gramine and sparteine were the predominant alkaloids directly
quantified from FID, while lupinine was present in minor con-
centrations of 3–10 ␮g/g, quantified by MS after trimethylsily-
lation. The samples of L. angustifolius showed an origin related
behavior: Samples from Australia had high matrix load and
were quantified similar to samples from L. albus. The remain-
ing samples contained noticeable amounts of ␣-isolupanine
(range 40–90 ␮g/g for the sweet, 237 ␮g/g for the bitter vari-
ety), which could be directly quantified together with lupanine,
13␣-hydroxylupanine and angustifoline by FID.
For L. albus, the alkaloid content of bitter and sweet varieties
was comparable, while for L. angustifolius there was a 10–20-
fold increase in alkaloid content for the bitter varieties (Table 2).
In samples of the L. albus group lupanine varied from 69 to
111 ␮g/g, 13␣-hydroxylupanine from 48 to 117 ␮g/g and angus-
tifoline from 9 to 12 ␮g/g. In L. luteus gramine varied in the range
450–893 ␮g/g, sparteine 2–50 ␮g/g and lupinine 3–12 ␮g/g.
The bitter varieties of L. angustifolius contained lupanine in
the range 4.0–12.9 mg/g, 13␣-hydroxylupanine 5.0–6.6 mg/g
and ␣-isolupanine 213–237 ␮g/g, while sweet lupin seeds
and flours had 18–862 ␮g/g of lupanine, 18–943 ␮g/g of
13␣-hydroxylupanine, 7–226 ␮g/g of angustifoline and those
not from Australian provenience additionally 37–90 ␮g/g ␣-
isolupanine.
Table 3 shows the alkaloid contents found in foods on
the Swiss market. According to the alkaloid pattern, mainly
lupins from L. albus and exceptionally from L. angustifolius
but not from L. luteus have been used, i.e. lupanine, 13␣-
hydroxylupanine and angustifoline were the principal alkaloids
found. ␣-Isolupanine, 13␣-angeloyloxylupanine or 13␣-
tigloyloxylupanine were present in trace amounts in two of the
coffee products, in the curryburger, the lupin tofu products and in
one of the lupin flour containing breads. For quantification, in all
food samples except coffee, lupanine and 13␣-hydroxylupanine
could be directly quantified from FID. The matrix in coffee
samples implied trimethylsilylation of 13␣-hydroxylupanine
and angustifoline. Unfortunately, even after trimethylsilylation
of angustifoline the matrix in all food samples except the lupin
snack product, where all three alkaloids were directly quan-
tifiable from FID, prevented data interpretation. Since alkaloid
ratios in our investigated seeds and flour measurements were
relatively constant within one species and have also been used
as taxonomical markers [30], it can be assumed, that the lost
angustifoline content is in the order of 5–10% of the summed
up value for lupanine and 13␣-hydroxylupanine. The values in
Table 3 have not been correspondingly corrected, though.
As for the found alkaloid contents, the coffee samples show
outstanding results and almost reach the Australian maximum
level for human exposure set at 200␮g/g of derived lupin prod-
ucts. While the other commercial products contain lupin flour or
isolates in limited content, the coffee powder is made of roasted
lupin grains alone. Alkaloid transfer from roasted lupin beans to
the ready-to-consume drink was about 50–70% at a water extrac-
tion temperature of 90 ◦C. Therefore, the possible advantage of
the absence of caffeine has to be weighed against an important
intake of toxic alkaloids.
3.2. Phomopsins
3.2.1. ELISA
The applied ELISA-kit proved to be sensitive and reliable,
but is not commercially available, yet. The kit detects both PhA
Table 3
Lupin containing foods investigated. Values of found alkaloids and phomopsin
A and B are given
Food products AlkaloidsaPhomopsin A + B
(␮g/g) n(ng/g) n
Lupin coffee 136–182 4 <0.1 3
Curryburger 4 1 <0.1 1
Lupin tofu untreated 12–15 2 <0.1 2
Lupin tofu with tomato 3 1 <0.1 1
Lupin bread (lupin flour
content 7-10%)
8–12 2 <0.1 1
Falafel with lupin tofu 3 1 0
Lupin snack 6 1 0
aSum of standard alkaloids (see Sections 2.3.4 and 3.1.2).

H. Reinhard et al. / J. Chromatogr. A 1112 (2006) 353–360 359
and PhB [17]. PhA is two to five times more toxic than PhB
[31], therefore it would be desirable to have a chromatographic
confirmation method being able to distinguish between the two
mycotoxins in the low ng/g range.
Typical RSD for samples was ±20%. The best sensitivity for
this assay was in the range 20–320 pg PhA/well and the limit
of detection was 0.1 ng/g related to the samples. The recovery
from lupin flour spiked with PhA was in the range 89–111% at
a concentration of 7 ng/g PhA.
The found PhA + B contents of seeds and flours are given
in Table 2. All samples except one were free of phomopsins,
expected especially in seeds and flours from L. angustifolius. The
only positive probe was optically striking, since attack of pho-
mopsis results in darkened grain hulls [6]. Thus, it was already
foreseen to dispose the positive sample and its corresponding
fodder batch.
Table 3 shows the PhA + B contents found in foods on
the Swiss market. Phomopsins were absent in all of the
analyzed food samples, i.e. no value was in the order of
magnitude of the Australian limit value of 5 ng/g for fin-
ished food products. For both L. albus and L. angus-
tifolius resistant varieties against phomopsis are known,
though. In the future, genetically modifications could enhance
resistance to lupin diseases, such as phomopsis infection
[32].
3.2.2. LC
The attempted chromatographic methods for phomopsins
detection failed, since the attainable limit of detection was too
high compared to those of ELISA. The detection limits for
PhA for DAD (at 290 nm) and ECD were 450 and 50 ng/g,
respectively, for a signal to noise ratio (S/N) of three from a
standard solution injected on a Spherisorb ODS1 column. For
LC–MS/MS, the achievable limit of detection was 1ng/g PhA
(S/N= 3).
Additionally, sample cleanup turned out to be crucial, while
for the ELISA procedure no cleanup was required. Although
several attempts were made to establish an efficient cleanup
method for PhA, i.e. cleanup of lupin flour samples over
C18, SAX, SiOH and Oasis according to the manufacturer’s
instructions, all efforts were unsuccessful. An LC-UV method
to detect PhA at a 500 ng/g level from lupin stubble had
been published before, using an Amberlite XAD-2 and a
cation exchange cleanup, requiring an analysis time of 2 days
[33].
In a lupin flour sample spiked with PhA chromatographic
interference between PhA and genisteine occurred, i.e. genis-
teine coeluted with PhA. The estimated genisteine content was
3.5 ␮g/g related to the lupin flour sample from L. angustifolius,
completely masking any potential PhA presence.
Precision of LC–MS/MS was evaluated by injecting 10␮l
aliquots of standard solution containing 3.2 ng PhA in a gradient
run. The ion transition 789 →226 m/zwas monitored after 4 h of
instrument’s warm-up, all other parameters were set as described
in Section 2.4.3. Reproducibility of retention time and peak area
over a 10 h run resulted in an RSD of 0.5% (n= 10) and 8.9%
(n= 10), respectively.
4. Conclusions
There’s no urgent call for action concerning the alkaloid
and phomopsin A + B contents in foods on the Swiss market.
Though, products made of seeds alone, such as roasted lupin
beans promoted as caffeine free coffee surrogate, have consid-
erable alkaloid contents and almost reach the Australian limit
for finished food products. Toxicological interpretation might
be indicated for these products. The Australian New Zealand
Food Authority has set a tolerable level of exposure to lupin
alkaloids for humans at 35 ␮g/kg/day [34].
From the alkaloids found in the Swiss market products it can
be concluded that mainly L. albus, rarely L. angustifolius and
never L. luteus had been used in production.
The allergenic potential of lupins seems to be of major con-
cern [14]. An appropriate declaration at inclusions of lupin
in wheat flour was suggested due to the crossed peanut–lupin
allergy potential [13].
In our hands, trimethylsilylation of alkaloids was not ade-
quate, since silylation artifacts were noted. Here, a CO2-
incorporation occurred, which according to [29] could be
avoided by adding diluted hydrochloric acid during the final
stage of drying the samples prior to derivatization. Further,
acylation instead of silylation could be more adequate for the
analysis of amino compounds [35].
Additional efforts should be taken to develop a confirma-
tion method for phomopsins in the low ng/g range, especially to
identify which one of the five phomopsins is present since the
differences in their toxicity is of relevance [31].
Acknowledgments
We gratefully acknowledge the help and support from John
A. Edgar and Khin A. Than, both from CSIRO, Australian Ani-
mal Health Laboratory, Australia, who obligingly offered us the
phomopsin ELISA-kit they had developed. We are also indebted
to Bernhard Zimmerli for his farsighted planning and accompa-
nying of the project and for sharing his versatile knowledge and
to Michael Beer for valuable contributions to this article’s style.
References
[1] R.J. van Barneveld, Nutr. Res. Rev. 12 (1999) 203.
[2] C.R. Sirtori, M.R. Lovati, C.M. Manzoni, S. Castiglioni, M. Duranti, C.
Magni, S. Morandi, A. D’Agostina, A. Arnoldi, J. Nutr. 134 (2004) 18.
[3] R.F. Keeler, E.H. Cronin, J.L. Shupe, J. Toxicol. Environm. Health 1
(1976) 899.
[4] R.F. Keeler, K.E. Panter, Teratology 40 (1989) 423.
[5] J. Pothier, S.L. Cheav, N. Galand, D. Dormeau, C. Viel, J. Pharm.
Pharmacol. 50 (1998) 949.
[6] J.A. Edgar, in: M., Dracup, J., Palta, (Eds.), Proceeding of the First Aus-
tralian Lupin Technical Symposium, Department of Agriculture, South
Perth, AU, 1994, p. 85.
[7] M.R. Gardiner, D.S. Petterson, J. Comp. Pathol. 82 (1972) 5.
[8] M.V. Jago, J.E. Peterson, A.L. Payne, D.G. Campbell, Aust. J. Exp.
Biol. Med. Sci. 60 (1982) 239.
[9] E.M. T¨
onsing, P.S. Steyn, M. Osborn, K. Weber, Eur. J. Cell Biol. 35
(1984) 156.
[10] P.H. Mortimer, J.W. Ronaldson, in: R.F. Keeler, A.T. Tu (Eds.), Hand-
book of Natural Toxins, vol. 1, Dekker, New York, US, 1983, p. 390.

360 H. Reinhard et al. / J. Chromatogr. A 1112 (2006) 353–360
[11] M.F. Mackay, A. van Donkelaar, C.C.J. Culvenor, J. Chem. Soc. Chem.
Commun. (1986) 1219.
[12] Anonymous, ACNFP Report on Seeds from the Narrow Leafed Lupin,
Appendix IX, MAFF Publications, London, GB, 1996, p. 107.
[13] D.A. Moneret-Vautrin, L. Guerin, G. Kanny, J. Flabbee, S. Fremont, M.
Morisset, J. Allergy Clin. Immunol. 104 (1999) 883.
[14] B. W ¨
uthrich, D. Mittag, B.K. Ballmer-Weber, Allergologie 27 (2004)
495, article in German.
[15] Anonymous, Directive 2003/89/EC and 2005/26/EC of the European
Parliament and Council, Brussels, BE, 2003 and 2005, respectively.
[16] E. Kovats, Helv. Chim. Acta 206 (1958) 1915.
[17] K.A. Than, A.L. Payne, J.A. Edgar, in: L.F. James, E.M. Bailey, P.R.
Cheeke, M.P. Hegarty (Eds.), Poisonous Plants, Iowa State University
Press, Iowa, US, 1992, p. 259.
[18] K.A. Than, CSIRO, Australian Animal Health Laboratory, Australia,
private communication.
[19] C.C. Culvenor, A.B. Beck, M. Clarke, P.A. Cockrum, J.A. Edgar, J.L.
Frahn, M.V. Jago, G.W. Lanigan, A.L. Payne, J.E. Peterson, D.S. Pet-
terson, L.W. Smith, R.R. White, Aust. J. Biol. Sci. 30 (1977) 269.
[20] B.A. Thomson, D.J. Douglass, J.J. Corr, J.W. Hager, C.L. Jolliffe, Anal.
Chem. 67 (1995) 1696.
[21] D. Lide. (Ed.), Properties of Organic Compounds on CD-ROM V6,
Chapman&Hall/CRC Press, Boca Raton, US, 2000.
[22] M. Wiewi´
orowski, J. Skolik, Roczniki chemii 33 (1959) 461, article in
German.
[23] F. Tosun, M. Tanker, T. ¨
Ozden, A. Tosun, Planta Med. 52 (1986) 242.
[24] M. Wink, in: P. Waterman (Ed.), Methods in Plant Biochemistry, vol. 8,
Academic Press, London, GB, 1993, p. 197.
[25] C.R. Priddis, J. Chromatogr 261 (1983) 95.
[26] M. Muzquiz, C. Cuadrado, G. Ayet, C. de la Cuadra, C. Burbano, A.
Osagie, J. Agric. Food Chem. 42 (1994) 1447.
[27] T. Aniszewski, J. Sci. Food Agric. 61 (1993) 409.
[28] F.W. McLafferty (Ed.), Wiley Registry of Mass Spectral Data, seventh
ed., John Wiley and Sons Inc., Hoboken, US, 2000.
[29] J.L. Little, J. Chromatogr. A 844 (1999) 1.
[30] T. Aniszewski, Sci. Legumes 1 (1994) 111.
[31] J.G. Allen, G.R. Hancock, J. Appl. Toxicol. 9 (1989) 83.
[32] M. Shankar, M.W. Sweetingham, W.A. Cowling, Euphytica 125 (2002)
35.
[33] G.R. Hancock, P. Vogel, D.S. Petterson, Aust. J. Exp. Agric. 27 (1987)
73.
[34] Anonymous, Australian New Zealand Food Authority, Technical
Report Series No. 3, 2001, p. 1. (http://www.foodstandards.gov.au/
srcfiles/TR3.pdf).
[35] R.P. Evershed, in: K. Blau, J.M. Halket (Eds.), Handbook of Derivatives
for Chromatography, second ed., Wiley, Chichester, GB, 1993, p. 51.