COMPARATIVE EVALUATION OF TWO COMMERCIALLY AVAILABLE ELISA KITS FOR THE SCREENING OF MICROCYSTINS IN WATER SAMPLES Triantafyllos Kaloudis, Nicholas C. Thanasoulias, Katerina Dimitrou, Leonidas Kousouris and Philippos Tzoumerkas Athens Water Supply and Sewerage Company (EYDAP SA) Water Quality Control and Protection Division, Organic Pollutants Laboratory, 19014 Polydendri Attikis, Greece. Tel.-Fax: 2102144593, E-mail:
[email protected] Microcystins are cyclic peptides with hepatotoxic activity that are produced by several bloom-forming cyanobacterial genera in surface waters. Immunoassays are currently the most promising screening methods for the detection of microcystins in raw and treated water samples within the WHO provisional levels. Several ELISA kits are now commercially available for this purpose. However, laboratories should validate the applicability and performance of these kits according to their individual applications. This paper evaluates the performance of two different commercially available ELISA kits for the quantitative screening of microcystins in raw and drinking water samples. These kits were selected because they use different types of antibodies and different assay protocols. One kit is based on polyclonal antibodies raised against microcystin LR. Therefore cross-reactivity with other microcystin congeners is an issue of concern. The other kit uses polyclonal antibodies against ADDA, which is an unusual β-amino acid present in most of the known microcystin congeners. Thus it is advocated as a congener-independent assay. The performance of the kits was evaluated against quantitative action levels set in order to monitor the presence of microcystins in surface and drinking waters of EYDAP. Several issues concerning the validation requirements of such kits are discussed and results regarding their sensitivity, precision, accuracy, cross-reactivities, false positive/false negative percentages and matrix-specific performances are presented. Results showed that both methods are sensitive and can be used as screening tools for microcystins and nodularin, when the decision levels are set according to their specific performances. Matrix effects should be taken into account, since as it was found they can significantly alter or even totally suppress their responses. Validation and quality control schemes should include the use of certified standards in order to balance any deviations of the internal calibrators. Positive results should be confirmed with established chromatographic methods. Keywords: Cyanobacteria, microcystins, ELISA
INTRODUCTION Cyanobacteria, also known as “blue-green algae” or “cyanophyceae”, are a group of procariotic, photosynthetic, highly adaptable microorganisms that can grow and form blooms in limnic and marine environments (Mur et al., 1999). These microorganisms produce a wide range of compounds named “cyanotoxins” that are toxic to
invertebrates, vertebrates and mammals (Codd et al., 1999). Cyanotoxins fall into three groups of chemical structure: cyclic peptides, alkaloids and lipopolysaccharides. The most frequently reported cyanotoxins in fresh and brackish waters are microcystins and nodularins, which have a cyclic peptide structure and hepatotoxic activity. Microcystins (MC) are cyclic heptapeptides that are produced by the cyanobacterial genera Microcystis, Anabaena, Oscillatoria, Nostoc, Hepalosiphon and Anabaenopsis. Their common ring structure contains the unusual b-linked amino acid 3-amino-9methoxy-10-phenyl-2,6,8-trimethyldeca-4,6-dienoic acid (ADDA), forming a heptapeptide ring with three D-amino acids (alanine, β-linked erythro-methylaspartic acid, α-linked glutamic acid), N-methyldehydroalanine and two variable L-amino acids, R1 and R2. The large number of microcystin congeners results from the different combinations of R1 and R2 in their molecular structures. About 60 structural variants have been characterized so far. The toxicity of hydrophobic congeners such as MC-LR (L=leucine, R=arginine) is higher, while MC- RR (R1 and R2 are arginine) is among the least toxic (Sivonen & Jones, 1999). Nodularins are cyclic pentapeptides that are produced by the genus Nodularia that grows in marine and brackish waters. Nodularin congeners contain ADDA, Dglutamate, 2-(methylamino)-2-dehydrobutyric acid, D-methylaspartic acid and a variable L- amino acid. The most frequently reported nodularin contains L-arginine as the variant amino acid. Several incidents of animal and human poisoning from microcystins have been reported worldwide, with two major episodes in Australia and Brazil (Falconer, 2005). Following the increased concern about the public health risks associated with microcystins intake, the WHO recommended a provisional level of 1µg/L for total microcystin-LR concentration in drinking water (Falkoner et al., 1999). Accordingly, through the last years there has been an increased research interest in developing sensitive, accurate and cost-effective methods for the quantitative determination of microcystins and nodularins. State of the art techniques for the detection and quantification of cyanobacterial hepatotoxins (McElhiney & Lawton, 2005, Hawkins et al., 2005, Harada et al., 1999) include HPLC with UV or MS detection (Meriluoto, 1997, Spoof et al., 2003), immunoassay methods (ELISA) (Metcalf & Codd, 2003), protein phosphatase inhibition assay (Carmichael & An, 1999) and bioassays (Harada et al., 1999). An HPLC method with UV detection after solid phase extraction was recently published as an ISO 20179 international standard. Immunoassays are finding increasing applications for the analysis of microcystins and nodularins, since they combine high sensitivity with minimal sample preparation. They are considered as probably the most promising tools for the quantitative screening of cyanobacterial toxins. To date, there are four commercially available ELISA kits for this purpose. However, several performance criteria are required to be examined and rigorous performance testing is essential in order to obtain accurate, reliable and meaningful data from ELISA methods (Metcalf & Codd, 2003). The aim of this study was to evaluate the performance of two conceptually different commercially available ELISA kits, when used for the screening of surface and treated waters of EYDAP SA. The kits are based on antibodies grown against MC-LR and ADDA and their performance is assessed with respect to the specific application and laboratory conditions. No claim is made that the results obtained in our study represent the general performance or inherent characteristics of the two kits under study.
MATERIALS AND METHODS The two competitive-type ELISA kits used were: a. The Envirologix (Portland, USA) Quantiplate kit based on the polyclonal antibodies of Chu et al. (1989) that are raised against MC-LR. The standard assay protocol was used for the determinations, which includes the following steps: Addition of assay diluent to microwells coated with antibodies, addition of samples, incubation, addition of microcystin-enzyme (Horseradish Peroxidase-HRP) conjugate, incubation, washing, addition of tetramethylbenzidine (TMB) substrate, incubation, addition of stop solution (HCl) and reading the absorbance at 450nm. b. The Abraxis (Pennsylvania, USA) Microcystins kit based on the ADDA antibodies of Fischer et al. (2001). Since ADDA is a common part of the molecular structure of most microcystins and nodularins, this method is considered a congener-independent assay. The assay protocol steps are: addition of samples to microwells coated with ADDAprotein conjugate, addition of the antibody solution, incubation, washing, addition of anti-sheep IgG enzyme (HRP) conjugate, incubation, washing, addition of TMB substrate, incubation, addition of stop solution and reading the absorbance at 450nm. Both methods were used according to the exact instructions given by the manufacturers. Calibrations were done with the calibrators provided with the kits, by plotting %B/Bo against logC. A Tecan Infinite microplate reader (Tecan Group Ltd., Switzerland) equipped with a 450nm filter and Magellan software was used for the absorbance measurements. Certified standard solutions (10µg/ml in methanol) of MC-LR, MC-RR and Nodularin (NOD) were used for the spiking of samples (DHI Water & Environment, Hörsholm, Denmark). The exact concentration of each standard solution was given in the certificate of analysis, after spectrophotometric determination. Five different sample matrices were studied in microcystin – nodularin fortified sample experiments: a. Ultra pure (UP) grade (18.2 MΩ) laboratory water (USF Elga). b. Laboratory tap water, with the addition of 1ml per litre of 10g/L sodium thiosulfate solution (according to ISO 20179). c. Lake water (sampled from Mornos lake), filtered through a 47 mm glass microfiber filter paper with 1 µm retention size (Pall Corp., Michigan, USA), with the addition of sodium thiosulfate as above. d. Sea water (sampled from Saronikos gulf) and 50% sea water (diluted with HPLC grade water), filtered through a 47 mm glass microfiber filter paper, with the addition of sodium thiosulfate as above. This matrix was used for the evaluation of the response of the two methods to nodularin in sea and brackish waters. e. An extract that is an intermediate product of sample preparation according to ISO 20179. This matrix was used in order to check how both methods perform in endotoxin determinations, i.e. determinations of cyanotoxins that are released after extraction and lysis of cyanobacterial cells. The matrix was prepared according to ISO 20179 by filtration of 500 ml of lake water (sampled from Marathon Lake), extraction of the filters with 75/25 v/v methanol/water in an ultrasonic bath (Branson 1200 sonicator) and centrifugation of the extract at 4000 rpm for 10 min. The sensitivity and linearity of the response were assessed through calibration curves obtained with the calibrators contained in the kits as well as with a series of external
MC-LR standards in laboratory water, in the range of 0.1 to 2.0 µg l-1. Accuracy and precision studies were carried out by spiking matrix blanks (laboratory water, tap water and lake water) with MC-LR at 0.25 and 1.0 µg l-1 levels and determining MC-LR content at three different runs (different days, analysts, microplates), with a total of 15 samples for each matrix-level combination. The cross reactivity of the kits to MC-RR and NOD was assessed by determinations of spiked solutions (1 µg l-1) in the three matrices. Spiked samples of NOD (1 µg l-1) in sea water and 50% sea water were measured in order to evaluate the performance of the two methods for NOD determinations in marine and brackish waters. Finally, determinations of MC-LR were carried out in the intermediate extract of the ISO 20179 method with both kits. RESULTS AND DISCUSSION The linearity and sensitivity of the two methods were assessed by a series of five runs of calibrations with the calibrators included in the kits. Results of the linear fits obtained ( % B B0 = b0 + b1 log C ) are presented in Table 1. Limits of detection (LODs) were calculated using the curve fit data (LOD1) and by multiple (n) determinations of blank samples (LOD2). These LODs were in agreement or even lower than those reported by the manufacturers. Table 1: Calibrations and estimated LODs for the two kits. Kit
No of calibrators (range,µg l-1)
Envirologix 3 (0.16-2.5) Abraxis 5 (0.15-5.0)
No of runs 5 5
R 0.994 0.98
LOD1 µg l-1 0.141 0.063
LOD2 (n) µg l-1 0.147 (5) 0.087 (10)
LOD3 µg l-1 0.147 0.100
Table 2 presents the results for accuracy (%recovery) at the 0.25 and 1 µg l-1 levels of MC-LR for the three different matrices. Three runs of experiments, with 5 determinations in each run were carried out, giving a total of n=15 for each estimation. Mean recoveries for the Envirologix kit were in the range of 73-93%, The Abraxis kit gave higher results, especially for the tap and lake matrices, where mean recoveries were 133-189%. An ANOVA confirmed the significance of the above matrix effects (p=0.001). Deviations from 100% recovery in zero-matrix samples and kit variations could be due to differences in the calibrators provided with the kits, a problem often discussed in literature (Metcalf et al. 2000, Pettit & Simpson 2002). Table 2: % recoveries of MC-LR and corresponding standard deviations (n=15 in all calculations). Envirologix Abraxis %Recovery (s) %Recovery (s) Level/µg l-1 UP Water Tap Lake UP Water Tap Lake 1.0 93(14) 78(17) 78(20) 110(19) 133(28) 147(35) 0.25 78(21) 73(17) 77(17) 109(32) 155(28) 189(56) Table 3 presents results of intra and inter-assay precision. Intra assay is reported as the “within run” standard deviation of multiple determinations (n=5) across three runs. Inter-assay is reported as the overall standard deviation (n=15). For the Envirologix kit,
variation did not exceed 20% CV at any matrix. The Abraxis kit demonstrated increased variation especially for lake water matrices. Table 3: Precision data for the two kits Envirologix Sintra - Sinter Level/µg l-1 UP Water Tap Lake 1.0 0.12-0.14 0.15-0.17 0.11-0.20 0.25 0.03–0.05 0.03-0.04 0.03-0.04
Abraxis Sintra - Sinter UP Water Tap Lake 0.16-0.19 0.23-0.27 0.25-0.35 0.08-0.08 0.06-0.07 0.08-0.14
False positive results at the lowest assay levels were assessed by multiple determinations of matrix blanks (ultra pure, tap and lake water). The Envirologix kit had a 17% (n=18) of false positives at the 0.16 µg l-1 (lowest calibrator) level, while the Abraxis kit had a 6% at the 0.100 µg l-1 (LOD) level. False positive percentages were 0% for both kits, when the decision level was set at 0.250 µg l-1. On the same basis, false negative results at the 0.250 µg l-1 level were 15% for the Envirologix kit and 0% for the Abraxis kit. These results show that with a proper setting of the decision level a high screening efficiency can be achieved with both kits. Cross reactivity of the two methods to microcystin-RR (MC-RR) and Nodularin (NOD) is reported as the % MC-LR equivalent concentration. Results of the Envirologix kit showed an 78% cross reactivity to MC-RR and an 83% cross reactivity to NOD, compared to 54% and 68% as derived from kit documentation data. For the Abraxis kit the estimated cross reactivities were 73% for MC-RR and 153% for NOD, compared to 50% and 100% as reported by Fischer et al., 2001. Table 4 presents the results of the determinations (n=4) of 1 µg l-1 fortified samples of NOD in sea and 50% sea water and the results of 1 µg l-1 fortified samples (n=8) of MC-LR in the ISO method intermediate extract. Results demonstrated that the Envirologix kit performed well in nodularin determinations in sea and brackish waters. Slightly increased results have also been observed also by Metcalf et al. (2000) in MCLR determinations in saline water samples with this kit. The Abraxis kit gives extremely high values in these determinations. Another feature of this matrix-specific performance is that although the Envirologix kit gives a negative value (
