Analytica Chimica Acta 399 (1999) 75–87

Potential of immunoextraction coupled to analytical and bioanalytical

methods (liquid chromatography, ELISA kit and phosphatase inhibition
夽
test) for an improved environmental monitoring of cyanobacterial toxins
Corinne Rivasseau, Marie-Claire Hennion ∗
Department of Environment and Analytical Chemistry (CNRS 657), Ecole Supérieure de Physique et de Chimie de Paris, 10, rue Vauquelin,
75231 Paris Cedex 05, France
Received 21 January 1999; received in revised form 5 May 1999; accepted 25 May 1999

Abstract
A new immunosorbent involving antigen–antibody interactions has been developed for the selective solid-phase extraction
of microcystins (heptapeptides synthesised by cyanobacterial algae) in environmental samples. Its evaluation is first described
as off-line solid-phase extraction followed by liquid chromatography or microchromatography. Due to the high affinity and
selectivity of the antigen–antibody interactions, extraction, concentration and isolation of microcystins in complex samples
occur in a single step. Selectivity is shown by the analysis of real samples and by comparison with the use of non-specific
octadecylsilica extraction sorbent. Especially, it is demonstrated that potential interferences from herbicides are not trapped by
the immunosorbent. Due to the cross-reactivity of the antibodies and the very similar molecular structures of the microcystin
variants, the immunosorbent was shown to trap the three commercially available standards of microcystins, as well as other
variants which occurred in algae cultures or real water samples and which have been identified using LC-mass spectrometry.
In case of blooms, rapid on-site determinations are achieved by coupling immunoextraction to bioanalytical methods such
as either a commercial ELISA kit or a phosphatase inhibition assay. The clean extracts free from any matrix interferences
and the easy-to-obtain enrichment factor of 10 greatly improve the determination of microcystins at the 0.1 ␮g l−1 in surface
water using these bioanalytical assays. The best available technique for a rapid monitoring of toxic blooms is the combination
of a simple immunoextraction with the phosphatase inhibition test because it combines a structure recognition tool with a
bioassay based on the toxicity mode. ©1999 Elsevier Science B.V. All rights reserved.
Keywords: Microcystins; Blue–green algae toxins; Solid-phase immunoextraction; ELISA; Phosphatase inhibition tests

1. Introduction

夽 Manuscript corresponding to the lecture given at ‘Immuno-
chemistry Summit VII & Third Workshop on Biosensors and
Biological Techniques in Environmental Analysis’, Las Vegas, NV,
USA, 1–3 December.
∗ Corresponding author. Tel.: +33-1-40-79-46-51; fax: +33-1-40-
49-47-76
E-mail address: marie-claire.hennion@espci.fr (M.-C. Hennion)
Water blooms of toxic cyanobacteria (blue–green
algae) represent an increasing environmental hazard
because many strains of cyanobacteria such as Mi-
crocystis, Oscillatoria, Anabaena and Nostoc produce
toxins, one major group being cyclic heptapeptides
named microcystins [1–4]. Reliable analytical meth-

0003-2670/99/$ – see front matter ©1999 Elsevier Science B.V. All rights reserved.
PII: S 0 0 0 3 - 2 6 7 0 ( 9 9 ) 0 0 5 7 8 - 4
76 C. Rivasseau, M.-C. Hennion / Analytica Chimica Acta 399 (1999) 75–87
ods as well as rapid on-site monitoring methods are
required in case of cyanobacterial blooms, since ev-
idence exists of the adverse effects of cyanobacte-
rial toxins for animal and human health [5–11]. Cur-
rent analytical methods require an extraction and a
clean-up procedure followed by separation using liq-
uid chromatography (LC) with UV diode array de-
tection (UV DAD) and mass spectrometry (MS) de-
tection [12–18] or capillary electrophoresis [19]. For
field analysis, enzyme-linked immunosorbents assays
(ELISAs) have been developed in laboratories. One
has been recently commercially available and its eval-
uation has been reported [20]. A matrix effect has
been observed which prevents obtaining of reliable de-
termination in surface water at the low ␮g l−1 level.
Another type of bioassay has been developed based
on the inhibition of phosphatase, since microcystins
have been found to be strong inhibitors of certain ser-
ine/threonine phosphatases, especially the Type 2A.
This assay was optimised using commercially avail-
able components, but an enrichment step was required
for analysis of surface water at a low level and some
non-negligible matrix effect was also noticed [21].
Therefore, sample pretreatment methods capable of
providing trace-enrichment and very clean extracts
in order to remove the matrix interferences are re-
quired for improved LC analysis and more reliabil-
ity of these bioassays with real samples especially at
low detection levels. Common procedures used for the
sample pretreatment utilise solid-phase extraction sor-
bents which are non-selective such as octadecylsilicas
or polymeric sorbents. Consequently, co-extraction of
analytes and interferences generally occurs and this
is a major problem when analytes of interest are at
trace-level and interferences at higher concentrations.
Selectivity can be greatly enhanced by using materials
involving antigen–antibody interactions, thus provid-
ing selective extraction methods based on molecular
recognition. Antibodies are covalently bonded onto an
appropriate sorbent to form a so-called immunosor-
bent (IS), to be packed into a solid-phase extraction
cartridge or precolumn. Since antibodies are highly
selective towards the analyte used to initiate the im-
mune response with a high affinity, the correspond-
ing immunosorbent may extract and isolate this ana-
lyte from complex matrices in a single step, and the
problem of the co-extraction of matrix interferences is
therefore circumvented [22]. The first commercial ISs
have been introduced for the clean-up of samples for
the analysis of aflatoxins using antibodies bonded onto
the sepharose [23]. Other sepharose- or silica-based
ISs have been described in the literature for the anal-
ysis of single analytes such as carbendazim, chlor-
toluron, atrazine or terbutylazine [24–27]. But since
an antibody can also bind to one or more analytes
with a structure similar to the one used for its prepara-
tion, ISs were also developed for trapping single ana-
lytes and their metabolites [28–30] or a whole class of
structurally related compounds. ISs have been tailored
for the extraction of groups of organic compounds
including triazine and phenylurea pesticides, BTEXx
(benzene, toluene, ethylbenzene and xylene isomers),
polyaromatic hydrocarbons (PAHs), benzidine and re-
lated azo dyes [31–40]. Validation studies using cer-
tified reference materials have demonstrated the relia-
bility of immunosorbents for the group of phenylureas,
triazines and PAHs [41–43].
The objective of the work reported here was to
prepare and to evaluate a new immunosorbent able
to extract microcystins and to facilitate their identifi-
cation. One problem for setting up new analytical or
bioanalytical methods for their monitoring is the lack
of commercially available standards, which have only
three variants nowadays whereas more than 50 vari-
ants have been isolated from algae cultures up to now.
There is real evidence for the environmental occur-
rence of other variants which have been identified us-
ing LC-MS in case of blooms [44,45]. Antibodies have
been synthesised against the commercially available
microcystin-LR, one most toxic variant and encoun-
tered worldwide. Since all microcystins have a com-
mon cyclic heptapeptide structure, the immunosorbent
is expected to trap other variants. A second objective
was to profit from the selectivity of the immuno-
extraction step to remove the matrix effect in complex
samples for setting up new rapid monitoring meth-
ods combining immunoextraction to either ELISA or
phosphatase inhibition bioassays.
2. Experimental
2.1. Chemicals
Microcystin-LR and -RR standards were pur-
chased from Sigma (Saint Quentin Fallavier, France),
 C. Rivasseau, M.-C. Hennion / Analytica Chimica Acta 399 (1999) 75–87
microcystin-YR standard from Calbiochem Nov-
abiochem (La Jolla, CA, USA). Concentrated so-
lutions of toxins were prepared in methanol. All
reagents were of LC grade or analytical reagent
grade. Methanol was purchased from Prolabo
(Fontenay-sous-Bois, France) and acetonitrile from
J.T. Baker (Deventer, The Netherlands). LC quality
water was obtained by purifying demineralised wa-
ter by a Milli-Q filtration system (Millipore, Saint
Questin en Yvelines, France). Phosphate buffer saline
(PBS) solution was a mixture of a 0.05 M phosphate
sodium buffer and 0.10 g l−1 sodium chloride solution
at pH 7.5. The phosphatase PP-2A was obtained from
Upstate Biotechnology (New York, USA), provided
in aliquots of about 5 ␮g protein, with 10 units activ-
ity, a unit being equivalent to 1 nmol pNPP hydrol-
ysed per min. The p-nitrophenyl phosphatase (pNPP),
dithiothreitol (DTT), EDTA, magnesium chloride,
calcium chloride Tris buffer, bovine serum albumin
were purchased from Sigma. ELISA EnviroGardTM
Microcystins Plate kits were obtained from Rhône
Diagnostics Technologies (Lyon, France).
2.2. Material
The instrumentation for microLC consisted of two
LC-10AS pumps (Shimadzu, Kyoto, Japan) con-
nected to an Accurate 1/16 microflow splitter (LC
Packings, Amsterdam, The Netherlands). The out-
let was linked to a four-port Valco CIAW valve
(VICI, Valco Europe, Switzerland) with a 2.5 ␮l in-
jection loop. UV detection was performed either
with a SPD-10A detector (Shimadzu) equipped with
a U-shaped microcell with an internal volume of
160 nl (LC Packings), or with a SPD-M10A pho-
todiode array detector (Shimadzu) equipped with a
microcell (prototype LC Packings). MicroLC-ES-MS
experiments were carried out on a benchtop HP1100
series LC/MSD set-up from Hewlett Packard (Wald-
bronn, Germany) which was equipped with a dual
air-cooled turbomolecular pump vacuum system
and incorporates a hinged swing-out spray cham-
ber enabling atmospheric pressure electrospray
(ES).
For the ELISA and phosphatase assays, the opti-
cal density was measured using a Ceres 900C BioTek
(Osi, Maurepas, France) microplate reader.
77
2.3. Stationary phases, solid-phase extraction
sorbents and analytical columns
Analytical microLC separations were performed
using a 25 cm × 1 mm i.d packed with 5 ␮m reversed
phase Hypersil BDS C18 (Hypersil SA, Les Ulis,
France). Off-line solid-phase extraction was per-
formed on 3 ml Bakerbond SPE C18 cartridges packed
with 500 mg of octadecyl silica. The immunosorbent
consisted in anti-microcystin-LR antibodies bonded
onto glutardialdehyde-activated silica particles of
27.5 nm pore size (Mallinckrodt Baker). Polyclonal
antibodies were prepared in the Laboratory of J.F.
Lawrence, as part of a collaborative program be-
tween ESPCI and Banting Research Centre, (Health
Protection Branch, Food and Research Division, Ot-
tawa, Canada). In order to induce an immunogen
response, microcystin-LR has been linked to the car-
rier protein bovine serum albumin in the presence of
a water-soluble carbodiimide according to the pro-
cedure described by Chu et al. for the production of
antibodies against microcystins [46]. The immunis-
ing reagent thus obtained have been injected to three
rabbits. The serum has been collected for the first
time in 3 months after immunisation. The IgG frac-
tion containing the antibodies of interest was isolated
from the pooled serum by affinity chromatography
and then covalently bound onto the silica according
to the procedure previously described in Refs. [31]
and [35]. The immunosorbent was then packed in
SPE cartridges, each one being filled with 250 mg of
the sorbent and stored at 4◦ C in a solution of PBS
containing 0.2% of azide. Each cartridge contained
20 mg of IgG fraction.
2.4. LC conditions
MicroLC was used for the analysis of extracts
from surface water and culture media and cellular
extracts. The mobile phase consisted of acetonitrile
5 × 10−3 M phosphate buffer acidified to pH 2 with
trifluoroacetic acid. The flow rate was 50 ␮l min−1 .
The elution gradient employed Solvent A made of
acetonitrile–phosphate buffer (25 : 75, v/v) and Sol-
vent B made of acetonitrile. For the surface water
analysis the gradient was 0% B to 20% B from 0 to
35 min, 30% B at 45 min, and 100% B at 50 min. For
the water algae sample, it was 10–20% B from 0 to
78 C. Rivasseau, M.-C. Hennion / Analytica Chimica Acta 399 (1999) 75–87

20 min, 40% B at 40 min, 60% B at 50 min and 100% 20 ml of a 80 : 20 methanol/water mixture and 10 ml
B at 55 min. of LC-grade water.
MicroLC-ES-MS experiments were performed us-
ing a 25 cm × 0.1 cm i.d. C18 column. The mobile 2.6. ELISA procedure
phase was composed of a linear gradient from 30
to 60% of acetonitrile in 40 min. The effluent at a
The EnvirogardTM test is a direct competitive en-
50 ␮l/min flow rate was introduced without any split
zyme immunoassay. Assays of standards or samples
in the ES set-up. The ES-MS was used in the positive
were performed following kit instructions. Briefly,
polarity mode.
a 100 ␮l volume of the unknown sample, standard,
calibrator or negative control, is introduced into the
2.5. Solid-phase extraction using C18 silica and
immunoextraction procedure well and incubated for 30 min at ambient temper-
ature. 100 ␮l of a microcystin–enzyme conjugate
solution is then added and incubated for another
Off-line preconcentration was carried out with
30 min at ambient temperature. The wells are emp-
100 ml samples using a C18 solid-phase extraction
tied and washed four times with 300 ␮l of ultrapure
cartridge. The adsorbent was activated with 5 ml of
water. 100 ␮l of substrate is added to each well and
methanol, then washed with 10 ml of de-ionized wa-
incubated for 30 min at ambient temperature. The
ter before sample percolation. The cartridge was then
substrate is transformed by the enzyme conjugate into
cleaned with 3 ml of an aqueous solution contain-
a blue compound. 100 ␮l of a 1 M hydrochloric so-
ing 20% methanol v/v and the toxins were desorbed
lution is added to stop the reaction and the solutions
with 3 ml of methanol acidified with 1% (v/v) trifluo-
turn yellow. The optical density or Absorbance A is
roacetic acid. The residue was evaporated to dryness
immediately recorded at 450 nm using the microplate
at 45◦ C under reduced pressure and dissolved in 50
reader. Standard curves were established by plotting
or 100 ␮l of acetonitrile–phosphate buffer (20 : 80,
the % of maximum absorbance versus the concentra-
v/v) acidified to pH 2. For ELISA or phosphatase
tion of the non-toxic calibrator provided in the kit or
assays, after evaporation of the solvent, the residue
the concentration of standard microcystin-LR, in log
was dissolved in 1 ml of de–ionized water containing
scale. For each run, the negative control and the three
0.5% of methanol.
calibrators (0.1 ␮g/l, 0.4 and 1.6 ␮g l−1 equivalent
Of-line preconcentration using the immunosorbent
microcystin-LR) were assayed at least in duplicate.
were as follows: the immunosorbent was conditioned
Standard calibration curves were drawn using com-
with 4 ml of PBS 0.1 M, then the sample was perco-
mercial microcystin-LR.
lated with the previous addition of 0.5% methanol. A
For an unknown solution, the concentration was di-
washing step was applied with 3 ml of pure water fol-
rectly estimated from standard curves drawn either
lowed by 3 ml of water containing 20% methanol. Des-
with non-toxic calibrators, or with standard solutions
orption was performed with 15 ml of a solution con-
in de-ionized water, or in the matrix sample. The
taining 80% methanol. A pure organic solvent was not
matrix effect was studied by constructing standards
selected because the IS is still at the laboratory-stage
curves for surface water samples and comparing with
study and the mixture of methanol and LC-grade water
the standard calibration curves prepared for de-ionized
(80 : 20) is more appropriate for its regeneration after
water or for the calibrator solutions given with the kit
use. The desorption solution was evaporated to dryness
to determine their parallelism.
at 45◦ C under reduced pressure and dissolved in 50 or
100 ␮l of acetonitrile–phosphate buffer (20 : 80, v/v)
acidified to pH 2. Regeneration of the immunosorbent 2.7. Phosphatase assay procedure
was made using 4 ml of PBS after being washed with
20 ml of a 80 : 20 methanol/water mixture and 10 ml The phosphatase PP-2A was isolated from hu-
of LC-grade water. When immunosorbents were not man reed blood cells as the heterodimer consisting
in use, they were stored at 4◦ C in a solution of PBS of 60 and 36 kDa subunits. PP-2A activity against
containing 0.2% of azide, after being washed with p-nitrophenyl phosphatase (pNPP) was determined
 C. Rivasseau, M.-C. Hennion / Analytica Chimica Acta 399 (1999) 75–87
by quantitation of p-nitrophenol (pNP) hydrolysed
from pNPP and pNP was estimated by an absorbance
measurement. The assay was performed at 37◦ C in
a 96-well microtiter plate using a total volume of
250 ␮l per assay. The optimised procedure for the best
sensitivity was as follows: 125 ␮l of sample or toxin
standards were introduced into the well together with
25 ␮l of 250 mM Tris at pH 8.4 containing 170 mM
MgCl2 , 20 mM DTT and 20 mM EDTA (initial con-
centrations). 50 ␮l of enzyme diluted in 20 Tris mM
pH 7.5 containing 1 mg/ml BSA and 0.1 mM MgCl2
were added. A 50 mU/ml PP-2A concentration was
chosen for routine experiments. After incubation for
8 min at 37◦ C, 50 ␮m of 100 mM pNPP in 50 mM Tris
pH 8.4 were added. The absorbance was recorded at
405 nm. For each unknown sample, the pNPP sam-
ple activity in the absence of exogenous PP-2A was
checked by carrying out a control assay with the
non-spiked sample as described above except that the
enzyme buffer did not contain any PP-2A.
2.8. Water and algae samples
Surface water samples were collected in rivers and
reservoirs from various areas of France, in order to be
representative of the various compositions of the ma-
trix. Their pH stand in the range of 7.3–7.9. Surface
water originated from Seine river (samples taken in
Paris) and from Clain river (samples taken in Poitiers,
France). Other surface water samples came from reser-
voirs located in the west of France All samples were
filtered on a Whatman GF/C filter (pore size = 1.2 ␮m)
and stored at 0◦ C before analysis during a period of
1–3 days before being analysed.
Cyanobacterial cultures were grown as described
in Ref. [14] and prepared as follows. The cells were
separated from the liquid medium by filtration through
a GF/C filter. Cellular components were extracted with
methanol. Toxins were isolated and concentrated from
5 ml using the immunosorbent.
3. Results and discussion
3.1. Structure of microcystins and expected
cross-reactivity of the antibodies
Class-extraction is possible if antibodies bonded to
the sorbent can cross-react with structurally related
79
compounds within the group of pollutants. More
than 50 microcystin variants with various toxicities
have been isolated up to now from various cyano-
bacyterial species. They consist of a common moiety
composed of seven amino acids, namely (– d-Ala1 –
l-X2 – d-MeAsp3 – l-Z4 –Adda5 – d-Glu6 –Mdha7 ),
where MeAsp stands for erythro-␤-methylaspartic
acid, Mdha for N-methyldehydroalanine, Adda for
3-amino-9-methoxy-2,6,8-trimethyl-10-phenyldeca-4,
6-dienoic acid, and X and Z for the variable amino
acids that give their name to the molecule, as shown in
Fig. 1 [6]. The 50 variants differ primarily in the two
l-amino acids plus methylation and demethylation on
the two unusual amino acids. Only microcystin-LR
(containing leucine and arginine), microcystin-YR,
(tyrosine and arginine) and microcystin-RR, (two
arginine) are commercially available.
The structural similarity represented in Fig. 1
allows expectation of some cross-reactivity what-
ever the variant selected. To date, few studies have
reported about the production of antibodies against
mycrocystins and their application in laboratory-made
ELISAs [16,46–51]. The first sensitive ELISA kit
made by Chu et al. [48] using anti-microcystin-LR
polyclonal antibodies has shown high cross-reactivity
with microcystin-RR and microcystin-YR variants,
but less reactivity with variants microcystin-LY
and microcystin-LA. Ann and Carmichael tested
the cross-reactivity of a laboratory-made kit using
similar anti-microcystin-LR polyclonal antibodies
with 18 microcystins and nodularin (monocyclic
pentapeptides with similar structure than those of
the heptapeptide microcystines) variants [16]. They
have found that the hydrophobic amino acid Adda
(3-amino-9-methoxy-2,6,8-trimethyl-10-phenyldeca-4,
6-dienoic acid) which has the (E) form at the C–6
double bond was essential for these toxins to ex-
press antibody specificity. Strong cross-reactivity
was obtained with a non-toxic (monoester of glu-
talic acid microcystin-LR) and methylated variants
of microcystin-LR. The commercial EnviroGardTM
ELISA kit used similar polyclonal antibodies and
profited from the high cross-reactivity with a non-toxic
microcystin variant and used it as a standard pro-
vided to calibrate the kit. Two other laboratory-made
kits using monoclonal anti-microcystin-LR antibodies
were shown to cross-react with microcystin-RR and
to a lesser extent to microcystin-YR [46,51].
80 C. Rivasseau, M.-C. Hennion / Analytica Chimica Acta 399 (1999) 75–87
Fig. 1. Common structure of the microcystin variants.
Therefore, microcystin-LR was selected as the tar-
get representative of the group, and also because of
its commercial availability, its toxicity and the fact
that it seems to be produced by many environmental
cyanobacterial strains [44].
3.2. Evaluation of the immunosorbent: Off-line
immunoextraction followed by liquid chromatography
3.2.1. Retention of microcystin-LR by the IS and
conditions for desorption
A typical SPE sequence using an immunosorbent
packed in a disposable cartridge is very similar to that
using a conventional C18 cartridge. The conditions for
desorption of analytes were first measured with the
analyte–antigen because it usually corresponds to the
highest affinity of the antigen–antibody interaction. A
spiked solution with 2 ␮g l−1 of microcystin-LR and
containing 0.5% of methanol (in order to avoid any
adsorption of the container due to the hydrophobicity
of microcystin-LR) was percolated through the IS and
desorption was performed by several successive 3 ml
solutions containing increasing amounts of methanol.
Results are reported in Table 1. The interaction is very
strong since it is necessary to add 80% methanol to
water and to increase the volume up to 9 ml to des-
orb the microcystin-LR. The volume could certainly
be decreased using 90% of methanol and strong acidic
conditions, but this step was not optimized for this
Table 1
Desorption of microcystin-LR from the immunoextraction sorbent.
Percentage extraction recovery determined in fractions of 3 ml
containing increasing amount of methanol. Mean value of three
experiments and RSD
Methanol content (%) 20 50 70 80 80 80
% Recovery <1 2 5 ± 4 21 ± 8 28 ± 7 32 ± 7
first evaluation to ensure a better regeneration of the
sorbent. Therefore, a volume of 15 ml containing 80%
methanol was selected for evaluating the properties of
the IS and this volume was further evaporated to dry-
ness. With such a sequence, the mean extraction per-
cent recoveries obtained when performing the three ex-
periments with 20 ml of water spiked with 2 or 5 ␮g l−1
were, respectively, 75 ± 8 and 77 ± 9%. Losses oc-
curred mainly during the evaporation step of the 15 ml.
It has been previously verified that these losses were
not due to evaporation to dryness [14,18].
This table also indicates that a washing step using
3 ml of water containing 20% methanol can be applied
to the immunoextraction cartridge without eluting the
microcystin-LR. This washing was shown to remove
some hydrophobic interferences likely to occur in real
samples and which are adsorbed onto the protein of
the antibodies by non-specific interactions [22,37]. Of
course, when the trapping of other microcystins will
be studied, the affinity can be lower and the effect of
this washing should be reconsidered.
 C. Rivasseau, M.-C. Hennion / Analytica Chimica Acta 399 (1999) 75–87
Table 2
Cross-reactivity as measured by the amount (% recovery) in the
effluent, washing (3 ml, 20% methanol) and desorption (15 ml,
80% methanol) solutions. (mean value of two experiments)
Effluent Washing Desorption
solution solution
Microcystin-LR 0 0 80 ± 7
Microcystin-RR 0 0 80 ± 6
Microcystin-YR 50 ± 5 28 ± 3 4±5
3.2.2. Capacity
It is defined as the maximum amount that can be
preconcentrated by IS. Owing to the small amount of
toxin available, it was not measured accurately, but
was only evaluated by percolating 20 ml of water con-
taining 1 and 3 ␮g of microcystin-LR (concentration
of 50 and 150 ␮g l−1 ) and by measuring the respec-
tive amounts trapped and non-trapped by the sorbent.
The capacity was evaluated to 0.45 ± 0.08 ␮g (n = 3),
which is a sufficient value for environmental analysis.
3.2.3. Cross-reactivity
It was measured using samples of 100 ml made of
pure water spiked with 1 ␮g l−1 of each commercially
available variant. A washing step with 20% methanol
was applied before desorption. The amounts were
measured in each desorption fraction, as well as the
part which has not been trapped by the IS and which is
in the effluent from percolation. Results are reported
in Table 2. The antibodies have the same affinity for
both microcystin-LR and microcystin-RR, whereas
microcystin-YR is just slightly retained, one-half be-
ing in the effluent and the other part eluted with 20%
methanol, thus showing a weak affinity. The structural
modification linked to the replacement of leucin or
arginin amino acid by tyrosin which contains a rigid
phenolic group is sufficient for decreasing the inter-
action. Another explanation can be that the different
ionisation states which depends on the two amino
acids, but according to our study for determining the
ionisation constants using microLC [18], the YR and
LR variants are anionic at pH 6 or 7 whereas the RR
variant is neutral. Therefore, the structural modifi-
cation is certainly responsible for lower recognition.
However, it should pointed out that the antibodies
were evaluated after a short period of immunisation
(3 months). Previous studies from our group and from
the group of Stevenson, have reported that the affinity
81
between antibodies and structurally related analytes
increased with the immunisation times in rabbits
[31,35,38]. Immunosorbents made with antibodies
having longer immunisation times are now under
evaluation. Cross-reactivity with other microcystins
can only be studied with algae extracts in which other
microcystins have been identified using LC-MS.
3.2.4. Repeatability
The repeatability was studied by performing three
experiments at three different dates using the same IS
with 20 ml of water containing 100 ng of each com-
mercial microcystin. An average RDS of 11% was
measured, which is a value similar to that obtained
using a C18 SPE cartridge.
3.2.5. Breakthrough volumes
The maximum sample volume that can be perco-
lated with a 100% recovery also depends on the affin-
ity of analytes with the antibodies. The breakthrough
volumes were evaluated similarly to experiments con-
ducted with other non-selective sorbents [52]. Increas-
ing volumes of water spiked with a constant amount
of toxins (100 ng) were percolated through the IS, tak-
ing care of not overloading the capacity of the IS.
No washing step was applied between percolation and
desorption. Recoveries in the range of 80% were ob-
served for sample volumes up to 300 ml for the RR
and LR variants, indicating good affinity for the an-
tibodies. The recovery for the YR variant which was
35% for a volume of 100 ml, according to results of
Table 2, but was lower than 10% for 300 ml.
3.3. Selectivity of the immunoextraction and
application to real samples
3.3.1. Analysis of samples contaminated with
herbicides
Toxin blooms occurs mainly in reservoirs in agricul-
tural areas and interferences with herbicides are likely.
A water sample of 20 ml was spiked with 100 ng of a
mixture containing six common herbicides (triazines
and phenylureas) and 100 ng of microcystin-LR. After
percolation the washing step using 3 ml of water con-
taining 20% methanol was applied. The effluent, the
washing and the desorption solutions were analysed
for each analyte separately and the results have been
82 C. Rivasseau, M.-C. Hennion / Analytica Chimica Acta 399 (1999) 75–87
Table 3
Selectivity of the anti-microcystin immunosorbent recovery (%) of herbicides and microcystin-LR in the effluent, washing and desorption
solution after immunoextraction of 20 ml of water spiked with 100 ng of each analyte
Recovery (%) Microcystin-LR Simazine Atrazine
Effluent 0 84 80
Washing solution 0 0 22
Desorption solution 81 0 0
reported in Table 3. They indicate that the herbicides
are mainly in the effluent. The part which is in the
washing solution increases with the hydrophobicity of
the herbicides, indicating that these analytes may have
been retained by non-specific interaction onto the pro-
teins of the antibodies. For the two less polar terbuty-
lazine and linuron (log Kow around 3), this washing
step is not sufficient as shown by the small amount in
the desorption solution. Increasing the volume of the
washing solution to 4 ml or increasing the methanol
content to 25% should have removed the totality of
these herbicides. Therefore the selectivity of the im-
munosorbent is good since only microcystins are in
the desorption solution.
3.3.2. Analysis of surface water: Comparison with
extracts obtained using C18 silica
Selectivity has also to be evaluated in real sam-
ples such as contaminated surface water. Fig. 2 rep-
resents the chromatograms corresponding to extracts
from 100 ml of water coming from a French reservoir
and spiked with 1 ␮g l−1 of microcystin-LR and -RR.
In Fig. 2a, the water has been percolated through the
immunosorbent followed by washing with 3 ml con-
taining 20% methanol and desorption. In Fig. 2b, the
same volume of the same sample, but non-spiked with
the microcystins, was percolated through a cartridge
packed with octadecylsilica and followed by the same
washing step before desorption with pure methanol.
The comparison between the two chromatograms il-
lustrates the selectivity provided by the IS, since the
large peak of co-extracted interfering analytes during
the first 15 min in Fig. 2b does not occur with the IS.
If the chromatogram of Fig. 2b was represented us-
ing the same sensitivity scale of 1 mAU per cm, the
large interfering peak should be five times larger and
should hinder the identification of the analytes of inter-
est from this water sample using a volume of 100 ml.
Isoproturon Diuron Terbutylazine Linuron
77 63 74 69
4 19 18 30
0 0 4 5
With the immunoextraction, in addition to the two mi-
crocystins, just a few unidentified analytes show up
in the chromatogram, but the overall base-line is clear
allowing easy identification of microcystins by their
UV spectrum given by the diode array detector at the
␮g l−1 level from a sample volume as low as 100 ml.
Detection limits are evaluated at 0.2 ␮g l−1 .
3.3.3. Analysis of algae culture samples
Various strains were laboratory-cultured in order
to identify the microcystins that they may produce.
Since concentrations of toxins are usually high in cul-
ture media and microLC is highly sensitive, then pre-
cocentration from 5 ml of water samples was suffi-
cient as shown in Fig. 3 where represents the chro-
matogram corresponding to the analysis of the culture
water sample of a strain of Microcystis aeruginosa, af-
ter 15 days of growth in the laboratory. Four variants
of microcystin-LR have been identified by microLC
coupled to MS in this sample. Peak 1 corresponds
to microcystin-LR and the demethyl variant, Peak 3
to (6(z)Adda) microcystin-LR and Peak 4 to methyl
microcystin-LR. We are now preparing a new IS with
antibodies with higher immunisation time in order to
identify the production of microcystin variants directly
from the algae. In contrast to water extracts which are
rather clean, the algae extracts contain many other an-
alytes with structures different from those of toxins.
One objective is to also use the immunosorbents for
the preparation of pure standards from mass culture in
one step without subsequent need of purification.
3.4. Immunoextraction coupled to ELISA tests
Due to the toxicity of some microcystin variants,
rapid monitoring methods are required in case of the
blooms. A commercial ELISA test has been evaluated
and since the antibodies involved in the assay are
 C. Rivasseau, M.-C. Hennion / Analytica Chimica Acta 399 (1999) 75–87 83

Fig. 2. Chomatograms corresponding to the microLC analysis of an extract from 100 ml of surface water sample after solid-phase extraction
using (a) the immunosorbent and (b) octadecylsilica. (a) Sample spiked with 1 ␮g l−1 of microcystin-RR (Peak 1) and microcystin-LR
(Peak 2) and (b) same water but non-spiked with microcystins. See Section 2 for microLC conditions and extraction procedures.

polyclonal antibodies raised against microcystin-LR, a
similar cross-reactivity can be expected. We have eval-
uated the ELISA using several types of surface water
which initially did not contain any microcystin-LR
[20]. Fig. 4 represents the dose-response curve drawn
with the non-toxic variant of microcystin provided
as a standard within the ELISA kit and the anal-
ysis of several surface water samples spiked with
microcystin-LR. Although the value obtained with
pure water spiked with 0.05 ␮g l−1 of microcystin-LR
was really significantly different from the blank
values, the kit is given with a limit of quantifica-
tion of 0.1 ␮g l−1 . Several drinking water samples
spiked with 0.1 ␮g l−1 were analysed and only a
slight matrix effect was noticed. The analysis of var-
ious surface water samples provided evidence for a
stronger matrix effect, as reported in Fig. 4, so that
a false-positive effect is obtained for surface water.
No false-negative effects were obtained. Especially,
in water samples from Reservoir B which has not
been spiked or spiked with 0.05 ␮g l−1 , a concentra-
tion of 0.12 ± 0.02 ␮g l−1 of microcystin-LR will be
measured. For surface water samples, we experienced
that the probability of false-positive was high when
the content was lower than 0.15–0.2 ␮g l−1 , and was
determined by ELISA, and therefore quantification
limits in these types of water were set at 0.2 ␮g l−1 . It
was shown that this problem of false-positive could
be partly overcome by applying an enrichment step
of a factor 10 using a C18 cartridge in order that the
sample concentration should meet the concentration
providing more reliable values around 50% B/Bo
values (0.30–0.40 ␮g l−1 ). However, this sample pre-
treatment did not always remove the matrix effect,
which is due to the non-selectivity of the C18 silica.
Since all the positive samples need to be confirmed
by analytical LC techniques, there is an interest in
decreasing the number of false-positives. That was
the main result achieved using the immunosorbent
in order to obtain an enrichment factor of 10 to dif-
ferent surface water samples non-spiked and spiked
with 0.04 ␮g l−1 of microcystin-LR. All the concen-
trations in spiked samples were found in the range
of 0.45 ± 0.10 ␮g l−1 which is a good result taking
into account extraction recoveries in the range of
80–100%. In addition to cleaner extracts, the enrich-
84 C. Rivasseau, M.-C. Hennion / Analytica Chimica Acta 399 (1999) 75–87
Fig. 3. Chomatogram corresponding to the microLC analysis of an
extract from 5 ml of the water from algae culture of Microcystis
aeruginosa PCC 7806 sampled after 15 days of growth. Extraction
using the immunosorbent. See Section 2 for microLC conditions
and immunoextraction procedure.
ment being based on the structural recognition, the
identity of the analytes is reinforced and the results
of the ELISA better guaranteed.
3.5. Immunoextraction coupled to phosphatase
inhibition tests
Bioanalytical methods based on the inhibition of
protein phosphatases are of particular interest since
the evaluated parameter was shown to be related to
the toxic effect of some variants of microcystins. Mi-
crocystins are some of the most potent and specific
in vitro inhibitors of phosphatases of Type 1 (PP1)
or Type 2A (PP-2A) and are also extremely potent
tumour promoters acting by the okadaic acid path-
way [8]. Several assays were described for the deter-
mination of okadaic acid with the drawback of us-
ing radioactive substrates or complicated phosphatase
preparations [53–55]. Tubaro et al. were the first to
develop a practical colorimetric bioassay using com-
mercially available phosphatase Type 2A for a fast and
sensitive assessment of okadaic acid contamination in
mussels [53]. Although microcystins were found in
marine environments and have been recently shown
to covalently bound in mussels tissues [56], their oc-
currence is mainly reported in surface water such as
reservoirs, lakes, ponds or low-flow rivers and there is
a need to detect the toxicity in water rapidly in case of
blooms. Therefore, the same commercially available
phosphatase PP-2A was used to optimise the phos-
phatase assay for a rapid and sensitive monitoring of
water samples. Two procedures [21] have been opti-
mised using the same phosphatase with different ex-
perimental conditions, one allowing a rapid estimation
of the concentration of microcystin-LR in the range
of 0.4–10 ␮g l−1 , which should be used to calculate
the dilution required for using the second procedure,
which has been optimised to be more accurate but with
a narrow working range between 0.2 and 0.8 ␮g l−1 in
the sample, as shown in Fig. 5. Since several solutions
will have to be added in the well for the assay, the
sensitivity of the test depends on whether the concen-
tration is calculated in the well or corresponds to that
in the sample, explaining the two reported scales. The
matrix effect from surface water samples was not re-
ally important, as shown by some points reported in
Fig. 5. But owing to the limit of quantification close
to 0.25 ␮g l−1 in real samples, an enrichment step is
very often applied.
This enrichment step performed using immunosor-
bents was applied and similar results as those ob-
served for the coupling of IS to ELISAs kits were
obtained, the most important one being the decrease
in the number of false-positives. Work is now under
study in order to isolate and identify microcystin vari-
ants from strains occurring in reservoir blooms by car-
rying out mass laboratory cultures. The knowledge of
both their individual retention by the immunosorbent
and their response to the phosphatase inhibition test
is required for the further validation of the combined
immunoextraction–bioanalytical phosphatase assay.
4. Conclusion
The advantages of a selective extraction and
clean-up in one step have been illustrated by sev-
eral examples combining off-line immunoextraction
 C. Rivasseau, M.-C. Hennion / Analytica Chimica Acta 399 (1999) 75–87 85

Fig. 4. ELISA calibration curves using the (䊏) non-toxic calibration standards provided with the kit, (䊉) de-ionized water spiked with
microcystin-LR and various surface water samples spiked with microcystin-LR: (䊐) river Seine water, (䊊) River Clain, (H) Reservoir A
and (N) Reservoir B. The error bar represents ±1 standard deviation from the mean and has only been reported on the calibration curve
drawn with the non-toxic calibrants (mean value from 5 replicates).

Fig. 5. Dose–response curve of PP-2A inhibition by microcystin-LR (n = 4) drawn with (䊉) pure water and in spiked surface water samples;
(䊐) river Seine water, (䊊) River Clain, (H) Reservoir A and (N) Reservoir B. (NB: The concentration in the x-axis corresponds to that
of the 125 ␮l sample introduced in the well and not to the concentration within the well).
86 C. Rivasseau, M.-C. Hennion / Analytica Chimica Acta 399 (1999) 75–87
with classical analytical techniques such as liquid
chromatography and bioanalytical techniques such
as ELISA or phosphatase inhibition assays. A clear
baseline enhances the overall sensitivity of the analy-
sis and allows the handling of lower sample volumes
in comparison to the non-selective extraction proce-
dures. Identification of analytes of interest is easier
and their confirmation is reinforced since trapping oc-
curs on the basis of molecular recognition. This point
should be considered especially when the IS are used
as a simple immunoextraction and clean-up tool be-
cause the removal of the matrix effect is accompanied
with trace enrichment. Other clean-up techniques are
based on a priority fractionation between analytes
and interferences. Here, it occurs because a highly
selective extraction is performed.
The environmental monitoring of microcystins and
the knowledge of toxins which are really released by
algae is difficult due to the lack of available standards.
It is well known that only analytes which are looked
for are analysed, which can explain why in many cases
only standards are detected. Therefore, the various
methods described here are complementary and their
combined use should contribute to an easier identifi-
cation of the released toxins during blooms and pro-
vide conditions for an efficient on-site monitoring of
potentially toxic blooms. Immunoextraction followed
by the phosphatase inhibition test is certainly a rele-
vant combination for a rapid and efficient on-site mon-
itoring of the toxicity of a bloom of microcystin al-
gae. Moreover, if as recently suggested, microcystins
are responsible for the new type of shellfish intoxi-
cation, namely hepatotoxic shellfish poisoning (HSP)
[57], then more complex matrixes will have to be anal-
ysed. The interest and efficiency of immunoextraction
have recently been demonstrated for the detection of
okadaic acid and related DSP toxins in shellfish ma-
trixes [58].
Acknowledgements
The SAUR Water Services (Maurepas, France) are
thanked for having supported the thesis of C. Ri-
vasseau. Dr. J.F. Lawrence and his collaborators from
Health Canada (Ottawa, Canada) are thanked for the
preparation of the antibodies. S. Martins (ESPCI,
Paris) is also acknowledged for her contribution.
References
[1] W.W. Carmichael, in: C.L. Owny, G.V. Odell (Eds.), Natural
Toxins: Characterization, Pharmacology and Therapeutics.
Proceedings of the 9th World Congress on Animal, Plant and
Microbial Toxins, Stillwater OK, Pergamon Press, Oxford,
1989, pp. 3–16.
[2] O.M. Skulberg, G.A. Codd, W.W. Carmichael, Ambio 13
(1984) 244.
[3] L.A. Lawton, G.A. Codd, J. Institution Water Environ.
Manage. 5 (1991) 460.
[4] P.C. Turner, A.J. Gamie, K. Hallinrak, G.A. Codd, Br. Med.
J. 300 (1990) 1440.
[5] P.R. Hawkins, M.T.C. Runnegar, A.R.B. Jackson, I.R.
Falconer, Appl. Environ. Microbiol. 50 (1985) 1292.
[6] W.W. Carmichael, J. Appl. Bacteriol. 72 (1992) 445.
[7] K.L. Rinehart, J. Appl. Phycol. 6 (1994) 159.
[8] R. Nishiwaki-Matsushima, T. Otha, S. Nishiwaki, M.
Suganuma, K. Kohyama, T. Ishikawa, W.W. Carmichael, H.
Fujiki, J. Cancer Res. Clin. Oncol. 118 (1992) 420.
[9] Y. Ueno, S. Nagata, T. Tsutsumi, A. Hasegawa, M.F.
Watanabe, H.D. Park, G.C. Chen, G. Chen, S.Z. Yu,
Carcinogenesis 17 (1996) 1317.
[10] I.R. Falconer, M.D. Burch, A. Steffensen, M. Choice, O.R.
Coverdale, Environ. Toxicol. Water Qual. Int. J. 9 (1994) 131.
[11] I.R. Falconer, in: Proceedings of the Symposium
Eutrophication, Causes, Consequences and Remediation,
Porto, 21–23 May 1995, pp.8–10.
[12] K.I. Harada, K. Matsuura, M. Suzuki, M.F. Watanabe, S.
Oishi, A.M. Dahlem, V.R. Beasley, W.W. Carmichael, J.
Chromatogr. 448 (1988) 275.
[13] L.A. Lawton, C. Edwards, G.A. Codd, Analyst 119 (1994)
1525.
[14] C. Rivasseau, M.C. Hennion, M.C.P. Sandra, J. Micro. Sep.
8 (1996) 541.
[15] T. Sano, K. Nohara, F. Shiraishi, K. Kaya, Int. J. Environ.
Anal. Chem. 49 (1992) 163–170.
[16] J. An, W.W. Carmichael, Toxicon 32 (1994) 1495.
[17] S. Nagata, T. Tsutsumi, A. Hasegawa, M.F. Watanabe, Y.
Ueno, Jpn. J. Toxicol. Environ. Health 41 (1995) 10.
[18] C. Rivasseau, S. Martins, M.C. Hennion, J. Chromatogr. A
799 (1998) 155.
[19] N. Bouaı̈cha, C. Rivasseau, M.C. Hennion, P. Sandra, J.
Chromatogr. B. 685 (1996) 53.
[20] C. Rivasseau, P. Racaud, A. Deguin, M.C. Hennion, Environ.
Sci. Technol. 33 (1999) 1520.
[21] C. Rivasseau, P. Racaud, A. Deguin, M.C. Hennion, Anal.
Chim. Acta. 394 (1999) 243.
[22] V. Pichon, M. Bouzige, M.-C. Hennion, Anal. Chim. Acta
376 (1998) 21.
[23] J.F. Lawrence, P.M. Scott, in: D. Barcelo (Ed.), Environmental
Analysis: Techniques, Applications and Quality Assurance,
Elsevier, Amsterdam, 1993, p.273.
[24] A. Marx, T. Giersch, B. Hock, Anal. Lett. 28 (1995) 267.
[25] G.S. Rule, A.V. Mordehal, J. Henion, Anal. Chem. 66 (1994)
230.
[26] D.H. Thomas, M. Beck-Westermeyer, D.S. Hage, Anal. Chem.
66 (1994) 3823.
 C. Rivasseau, M.-C. Hennion / Analytica Chimica Acta 399 (1999) 75–87
[27] S.J. Shahtaheri, M.F. Katmeh, P. Kwasowski, D. Stevenson,
J. Chromatogr. A 697 (1995) 131.
[28] J. Cai, J.D. Henion, Anal. Chem. 68 (1996) 72.
[29] R.B. Wong, J.L. Pont, D.H. Johnson, J. Zulalian, T. Chin,
A.E. Karu, in: J.O. Nelson, A.E. Karu, R.B. Wong (Eds.),
Immunoanalysis of Agrochemicals, Emerging Technologies,
American Chemical Society, Washington DC, vol. 586, 1995,
p.235.
[30] K.A. Bean, J.D. Henion, J. Chromatogr. A 791 (1997) 119.
[31] V. Pichon, L. Chen, M.-C. Hennion, R. Daniel, A. Martel, F.
Le Goffic, J. Abian, D. Barcelo, Anal. Chem. 67 (1995) 2451.
[32] V. Pichon, L. Chen, M.-C. Hennion, Anal. Chim. Acta 311
(1995) 429.
[33] V. Pichon, L. Chen, N. Durand, F. Le Goffic, M.-C. Hennion,
J. Chromatogr. A 725 (1996) 107.
[34] J.F. Lawrence, C. Menard, M.-C. Hennion, V. Pichon, F. Le
Goffic, N. Durand, J. Chromatogr. A 732 (1996) 277.
[35] V. Pichon, H. Rogniaux, N. Fischer-Durand, S. Ben Rejeb, F.
Le Goffic, M.-C Hennion, Chromatographia 45 (1997) 289.
[36] S. Ouyang, Y. Xu, Y.H. Chen, Anal. Chem. 70 (1998) 931.
[37] M. Bouzige, V. Pichon, M.-C. Hennion, J. Chromatogr. A
823 (1998) 197.
[38] A. Martin-Esteban, P. Kwasowski, D. Stevenson,
Chromatographia 45 (1997) 364.
[39] M. Cichna, D. Knopp, R. Niessner, Anal. Chim. Acta 339
(1997) 241.
[40] M. Bouzige, P. Legeay, V. Pichon, M.-C. Hennion, J.
Chromatogr. A, 846 (1999) 317.
[41] I. Ferrer, V. Pichon, M.-C. Hennion, D. Barcelo, J.
Chromatogr. A 777 (1997) 91.
87
[42] I. Ferrer, M.-C. Hennion, D. Barcelo, Anal. Chem. 69 (1997)
4508.
[43] I. Ferrer, M.-C. Hennion, D. Barcelo, Anal. Chem. 70 (1998)
4996.
[44] C. Rivasseau, Ph.D thesis, Speciality Analytical Chemistry,
University of PARIS VI, 7 May 1998.
[45] L.A. Lawton, C. Edwards, A.K.A. Beattie, S. Pleasance, G.J.
Dear, G.A. Codd, Natural Toxins 3 (1995) 50.
[46] F.S. Chu, X. Huang, R.D. Wei, W.W. Carmichael, Appl.
Environ. Microbiol. 55 (1989) 1928.
[47] R. Kfir, J. Johannsen, D.P. Botes, Toxicon 24 (1986) 543.
[48] F.S. Chu, X. Huang, R.D. Wei, J. Assoc. Off. Anal. Chem.
73 (1990) 451.
[49] C.M. McDermott, R. Feola, J. Plude, Toxicon 33 (1995) 1433.
[50] S. Nagata, T. Tsutsumi, A. Hasegawa, F. Yoshida, Y. Ueno,
F.W. Watanabe, J. AOAC Int. 80 (1997) 408.
[51] S. Nagata, H.T. Soutome, T. Tsutsumi, A. Hasegawa, M.
Sekijima, M. Sugamata, Y. Ueno, Natural Toxins 3 (1995) 78.
[52] M.C. Hennion, C. Cau-Dit-Coumes, J. Chromatogr. A 823
(1998) 147.
[53] A. Tubaro, C. Florio, E. Luxich, S. Sosa, R. Delle Loggia,
T. Yasumoto, Toxicon 34 (1996) 743.
[54] C.F.B. Holmes, Toxicon 29 (1991) 469.
[55] A. Takai, G. Mieskes, Biochem. 275 (1991) 233.
[56] J.F. Simon, J.P. Vernoux, Natural Toxins 2 (1994) 293.
[57] D.E. Williams, S.C. Dawe, M.L. Kent, R.J. Andersen, M.
Craig, C.B. Holmes, Toxicon 35 (1997) 1617.
[58] N. Delaunay, V. Pichon, J.P. Le Caer, M.C. Hennion, Anal.
Chim. Acta., submitted.