Fresenius' Joumal of
Fresenius J Anal Chem (1994) 350:538-543
© Springer-Verlag 1994
Rapid field screening test for determination of 2,4,6-trinitrotoluene in water and soil with immunofiltration Claudia Keuchel, Reinhard Niessner Institute of Hydrochemistry,Technical Universityof Munich, Marchioninistrasse 17, D-81377Mtinchen, Germany Received:25 March 1994/Revised:20 May 1994/Accepted: 1 June 1994
Abstract. A rapid field screening test for 2,4,6-trinitrotoluene (TNT) in water and soil is described. The immunofiltration assay is based on a simplified enzymelinked immunosorbent assay (ELISA) performed in a pre-packed portable device. Performance of the test was assayed in spiked water, methanolic soil extracts (dilution 1: 10) and natural water samples. A quantitative colour response to concentrations of TNT in the range of 1 to 30 gg/1 in water and 50 to 1000 gg/kg in soil is demonstrated. The relative average standard deviations are 11.9% and 14.1%, respectively. High correlations between immunofiltration, ELISA and gas chromatography have been demonstrated. The cross reactivities of the antibody are comparable in both techniques.
Introduction The surroundings of former sites of ammunition production and filling stations are often contaminated with explosives. This is a result of unsuitable disposal practices, occasional accidents during the manufacture and destruction after World War II. In Germany there are more than 2000 sites where contamination is suspected [1]. However contamination of explosives did not take place only in the area of former sites. In the USA, one of the Army's most serious pollution problems is the disposal of water used to clean equipment and interior surfaces at ammunition manufacturing facilities [2]. Only few mechanisms potentially affect the environmental fate of TNT, for example photodecomposition. The dominant degradation pathway of TNT is microbial transformation [3, 4]. The principal reaction is the reduction of nitro groups to amino groups. Only with few microorganisms has cleavage of the TNT ring structure been reported [5]. TNT and especially its transformation products are known to be toxic and, in some cases, mutagenic to microorganisms in the Ames test I-6, 7].
Dedicated to Professor Dr. Dieter Klockow on the occasion of his 60th birthday Correspondence to: R. Niessner
Various instrumental techniques have been employed for the determination of nitroaromatics, especially gas chromatography [8-10] and high performance liquid chromatography [11-13]. For on-site testing, several screening methods exist but with low sensitivity [-14, 15]. In recent years immunological assays have been developed for the identification and quantification of compounds of environmental concern [-16]. They benefit from the extraordinary selective antibody/antigen interaction. Usually immunoassays are performed in microtiter plates and need certain equipment. Especially for on-site testing, research has focused on new designs, such as immunological test strips and multilayer devices, which offer easier performance. These methods are extensively used in medicine but only a few applications exist in environmental monitoring [-17-20]. This article reports the development of a immunofiltration assay for the determination of TNT in water and soil. The test is membrane based and uses the wick-like properties of a cotton pad to suck the reagents through the membrane.
Experimental Materials
Polyclonal goat anti-mouse serum was obtained from Sigma (Deisenhofen, FRG). The monoclonal anti-TNT antibody was made available by Strategic Diagnostics Inc. (Newark, USA), free of charge. All nitroaromatic compounds were from Promochem (Wesel, FRG). The enhancer solution was obtained from Bibby Dunn (Asbach, FRG). The sodium salt of a humic acid was purchased from Aldrich (Steinheim, FRG). Other chemicals used were of high analytical grade. Reinforced cellulose nitrate (NC) membrane (pore size 0.2 ~tm), filter paper (blue ribbon) and glass fiber filter (No. 8) was purchased from Schleicher + Schiill (Dassel, FRG). The devices for the immunofiltration and the cotton pads were obtained from Pall (Dreieich, FRG). Three standard soils were purchased from the agricultural experimental station (Speyer, FRG). They have different size distribution of particles and differ in their
539 content of organic carbon. They are characterized as sandy (S), loamy (L) and rich in humic substances (H).
Lid
Membranewith adsorbedantibodies Filterpaper
Apparatus Reflectance measurements were made with a portable reflectometer RQflex (Merck, Darmstadt) with a 6 V power supply. The instrument is equipped with four light-emitting diodes (LED) with emission maxima at 566 nm (green) and 659 nm (red). The results were expressed as percentage reflectance R [%], referenced to an internal standard. For qualitative or semi-quantitative determinations, it is also possible to compare visually the colour intensity of the membrane with a negative control or with test cards.
~
Cottonpad
Vessel
Fig. 1. Constructionof immunofiltrationdevice
Sample preparation Groundwater from wells near former ammunition plants in Hessia were analysed without sample preparation. Drinking water from the city of Munich was spiked with different amounts of TNT. To some of the water samples, humic acids were added to give a content of organic carbon of about 2 mg/1. Soils were spiked with TNT solution and stored at 4°C. After two months they were extracted with methanol for 2 rain. For a field test, shaking is superior to extraction in an ultrasonic bath. The extract was centrifuged for 10 rain at 17000 rpm. This procedure is convenient for the laboratory because a large number of samples can be analysed simultaneously. Filtration of the extract with a glass fiber filter yielded similar results. The advantage of the filtration procedure is that it can be carried out on-site. A 1:10 (v/v) dilution with water of the supernatants were analysed with immunofiltration or with ELISA.
Tracer synthesis The tracer was prepared by conjugating the aminohexanoic acid derivative of 2,4,6-trinitrophenyl to horseradish peroxidase by a N-hydroxysuccinimide/carbodiimide method as described in Ref. [21].
Assay procedure All solutions were prepared just before use and diluted if necessary. Mouse antiserum and the TNT antibody solution were diluted 1:100 in phosphate buffered saline (PBS, 0.08 tool/1 phosphate, 0.145 tool/1 NaC1, pH 7.6). Trinitrophenyl-peroxidase conjugate (TNP-POD, tracer) was diluted 1 : 5000 in the same buffer. The assay of the peroxidase was performed with 3,3',5,5'-tetramethylbenzidine/H202 in citrate buffer (0.2 tool/1 citrate, 0.01% sorbic acid potassium salt, pH 3.8). Since a dye is needed which is insoluble in aqueous buffers and precipitates on the membrane, an enhancer solution was added at a ratio of 1:10 (v/v). The following solutions were added in portions of 100 gl to the upper membrane and allowed to drain through the membrane: 100 ~tl mouse antibody, 200 gl TNT antibody, 100 ~tl blocking solution, 100 gl PBS buffer, 200 ~tl sample or standard, 100 gl tracer, 100 gl PBS solution, 200 gl substrate solution. After the colour development the reaction was stopped by the addition of 100 gl citrate buffer. After draining, the coloured membrane was removed from the device, placed in the reflectometer and the reflectance was measured at 573 nm. Different solutions were tested for blocking the remaining binding sites on the membrane after addition of antibodies: casein, bovine serum albumin, 7-globulin, ovalbumin, gelatin and carboxymethylcellulose.
Preparation of immunofiltration devices The immunofiltration assay was performed in a white plastic device equipped with three different layers (Fig. 1). A cotton pad with a thickness of approximately 10 mm was placed in the lower part of the device. The second layer was a filter paper (app. 20 x 20 mm) to achieve regular flow rates. A porous membrane capable of adsorbing the antibodies was cut into 20 x 20 mm pieces and placed on top. The membranes were fixed with a plastic lid with a circular hole of 10 mm diameter in the centre.
Verification For comparison studies, TNT soil extracts were measured by a TNT immunoassay performed in microtiter plates, with polyclonal antibodies as described [21]. The environmental water samples were analysed with immunofiltration, ELISA and gas chromatography. Solid phase extraction (C18 phase) of water samples and determination of TNT with GC-NPD was performed by the Stadtwerke Frankfurt (HP 5890 equipped with an N P D
540 (275 °C), Permabond SE-54-DF-1.0, 25 m, 0.32 mm i.d., helium as carrier gas, on-column, temperature program: 38 °C (2 min), 16°C/min to 70 °C, 10°C/rain to 200°C, 8 °C/min to 260°C (15 rain)).
Results and discussion
The ELISA in microtiter plates is useful in the laboratory when large numbers of samples have to be analysed. There is, however, a need for a simple test without complicated instruments that would have a visual endpoint. A membrane-based direct competitive ELISA was developed in which the antibody is immobilized on a porous solid support and all solutions are sucked through the membrane. The wick-like properties of the cotton pad and the filter paper control the flow of sample and reagents. Figure 1 shows the construction of the immunofiltration device.
Membranes In initial experiments different membrane types were tested; these varied in the type of immobilization (adsorptive or chemically bound), pore size and capacity of antibody binding. The best results are obtained with cellulose nitrate. The adsorption of antibodies is permanent and no bleeding is observed. Background colour resulting from unspecifically adsorbed tracer is low and the developed colour is intense. Supported cellulose nitrate was used to achieve mechanical stability. Decreasing pore size results in a decreasing flow rate and longer contact of the liquidphase reagents with the solid phase. A membrane with a pore size of 0.2 ~tm has shown to be adequate.
Blocking The influence of different blocking solutions is illustrated in Fig. 2. The lower the reflectance, the more intense is the
colour on the membrane. The assay was performed as described above with the exception that no TNT antibody was immobilized. Thus the colour is the result of unspecifically adsorbed tracer. The values are compared with a membrane without tracer treatment (89% reflectance). The reflectance from blocking with casein, bovine serum albumin and ovalbumin is comparable to that from the untreated membrane, y-Globulin, gelatin and carboxymethylcellulose do not have the desired blocking efficiency.
Reflectance measurements The reaction of peroxidase with the substrate leads to a blue coloured, charge transfer complex between reduced and oxidized TMB. In the presence of the enhancer, a water-insoluble product is produced with a broad absorption range and a maximum at 580 nm (data not shown). Routinely the reflectance of the membrane is measured with the green light emitting diode ()~. . . . . . = 566 nm).
Assay optimization The monoclonal TNT antibody used in all experiments does not adsorb on the cellulose nitrate membrane. Therefore anti-mouse serum must be used for precoating. Different concentrations and volumes of antibodies, sample, tracer and substrate solution were examined. The test is optimized with regard to sensitivity, colour intensity and assay time. Figure 3 shows the membrane-based calibration graph with concentrations of TNT ranging from 0.01 gg/1 to 1000 gg/1. The test is performed as described in the experimental section. The lower the concentration of TNT, the more colour is developed and the lower is the reflectance. For example, a reflectance of 40% corresponds to a deep blue colour. The curve can be fitted with a 4-parametric equation as described in Ref. [-22].
90-
I00 90. 80,
80-
70, 80,
!
50.
7060-
40 $, ao, 20, I0. 0
40,
Cas
,
BSA Glob OVA Gel CMC none Blocking Substances
Fig. 2. Reflectance of membranes treated with different blocking solutions. Straight line: Without tracer addition (blank). Abbreviations: Casein (cas), bovine serum albumin (BSA), y-globulin (glob), ovalbumin (OVA), gelatin (gel), carboxymethylcellulose (CMC)
30
I
0.01
i iiiillq
i iiiiiii
0.1
I
1
i
iiiiiii
I
10
i i IItlll I
I 11111[I I
100
1000
C o n c e n t r a t i o n TNT [p,g/l] Fig. 3. Calibration graph for TNT assay (average ± standard deviation of six replicates). Conditions as described in the experimental section. Min. reflectance 37%, max. reflectance 87%, 50% inhibition 4.5 gg/1
541 F o r 50% inhibition in binding, T N T concentrations of 4.5 pg/1 are required.
Stop reaction A very i m p o r t a n t step in carrying out the assay is the stopping of the colour generation; this is done by simply adding buffer solution. Figure 4 illustrates the colour formation of the membranes, as a function of time, after withdrawing the m e m b r a n e from the device. W i t h o u t the washing step, the colour continues to intensity for a period of a b o u t 6 min due to further reaction of peroxidase with substrate solution that remained in the pores of the membrane. Thus it is i m p o r t a n t to include a washing step; this ensures that the colour intensity remains constant for a b o u t 6 min and slightly decreases after that period (compare Fig. 4).
Table 1. Cross-reactivity [%] of nitroaromatic compounds ~ Compound 2,4,6-Trinitrotoluene 1,3,5-Trinitrobenzene Tetrylb 2,4-Dinitroaniline 2-Amino-4,6-dinitrotoluene 2,4-Dinitrotoluene 4-Amino-2,6-dinitrotoluene 1,3-Dinitrobenzene 2,6-Dinitrotoluene 2,4-Dinitrophenol 2-Methyl-5-nitroaniline 2-Methyl-3-nitroaniline 4-Amino-2-nitrotoluene Hexogen° Octogen d Nitrobenzene 2-Nitrototuene 4-Nitrotoluene Toluene
Immunofiltration
Microtiter plate
100 29.8 25.5 12.2 10.8 6.6 0.5 0.2 < 0.1 < 0.1 < 0.1 < 0.1 < 0.1 < 0.1 < 0.1 < 0.1 < 0.1 < 0.1 < 0.1
100 19.8 36.5 6.2 6.3 3 0.7 0.3 0.1 < 0.1 < 0.1 < 0.1 < 0.1 < 0.1 < 0.1 < 0.1 < 0.1 < 0.1 < 0.1
a Concentration yielding 50% inhibition b N-Methyl-N,2,4,6-tetranitr oaniline RDX, Hexahydro- 1,3,5-trinitro-1,3,5-triazine a HMX, 1,3,5,7-Tetranitro-l,3,5,7-tetrazocan
Cross-reactivities The antibodies were analysed for reaction with other nitro c o m p o u n d s and with m a j o r metabolites (Table 1). The cross-reactivity d a t a are c o m p a r e d to data obtained with an E L I S A in microtiter plates using the same antib o d y and the same tracer. The magnitude of cross-reactivities of the c o m p o u n d s under investigation is c o m p a r a ble in b o t h analytical procedures. Structurally related c o m p o u n d s with three nitro groups, 1,3,5-trinitrobenzene and tetryl (N-methyl-N,2,4,6-tetranitroaniline), show high cross-reactivities. Aromatics with two nitro groups have an intermediate or no reaction. In general, other nitroaromatic c o m p o u n d s or explosives such as hexogen (RDX, hexahydro-l,3,5-trinitro-l,3,5-triazine) and octogen (HMX, 1,3,5,7-tetranitro-l,3,5,7-tetrazocan) are inactive. It can be concluded that the m o n o c l o n a l a n t i b o d y can recognize nitro groups in meta positions especially if they are not separated by another substituent. The difference in the cross-reactivities between immunofiltration and microtiter plate is not astonishing because Weller
Table 2. Recoveries of TNT from spiked water samples Concentration spiked [gg/1]
Concentration found [gg/1] a
1.0 2.o 3.0 4.0 5.o 7.5 10.0 2o.o 30.0
1.0 4- 0.1 1.7 4- 0.1 2.3 4- 0.3 3.6 4- 0.4 5.1 4- 0.5 8.3 4- 1.6 10.9 4- 1.0 31.7 4- 1.8 51.4 4- 6.1
1.0b 3.0b lo.ou
1.2 4- 0.2 3.6 4- 0.4 11.3 4- 1.4
a Average 4- standard deviation of three replicates b Contain 5 mg/1 humic acid
found that cross-reactivities are dependent on the incubation time of the tracer [22]. In the immunofiltration test the incubation time is substantially lower.
5046-
,•
Matrix effects 353025-
2O o
i
g
g
tb
m
Time [n-d~
Fig. 4. Influence of post-washing after substrate development. C) With stopping, • without stopping
Different concentrations of T N T were spiked into tap water from the city of Munich. Additional samples were prepared using the same water spiked with T N T and humic acid at a concentration of 5 mg/1 (organic carbon about 2 rag/l). This should represent natural water with a content of organic carbon of a typical order of magnitude. The samples were allowed to stand for one day at 4 °C and were analysed with the immunofiltration assay. The results are presented in Table 2. Concentrations as low as 1 gg/1 T N T can be detected. An average relative standard deviation of 11.9% was calculated for all spiked water samples.
542
Water samples Eight water samples from wells were analysed with G C and immunofiltration. F o u r samples fall below the detection limit of both G C ( < 0.2 ~tg/1) and immunofiltration ( < 1 gg/1). N o false negatives are observed. The results are compared in Table 3. Although immunofiltration is a fast and convenient screening method the results are very close to the G C values.
10 5:
5
Soil samples To test the effect of soil extract on performance characteristics, the assay was run in the presence of untreated soils. N o effects were seen up to 10% methanolic soil extract; however data were more subject to variations. Soil samples spiked at 50, 100, 500 and 1000 rtg/kg were extracted with methanol. Though it is obvious that shaking for 2 min is not an adequate method to take out all extractable analyte, this method was used because the aim was to develop a rapid screening method for on-site use. Manually shaking for less than 5 min is also used in other screening tests for T N T in soil 1-14, 24]. Shaking is convenient and needs no instrumental equipment. The recoveries were between 60% and 25% with an average relative standard deviation of 14.1% (compare Fig. 5).
Table 3. Comparison of TNT concentrations in wells using GC and immunofiltration procedures Sample no.
GC [ g g / 1 ]
1
< 1.0b < 1.0 < 1.0 < 1.0 4.4 -t- 0.4 5.1 22.0 21.6 _+1.7
< 0.1 <0.1 <0.1 3.6 4.3 16.5 18.4
a Detection limit GC b Detection limit immunofiltration
,00, 9ot 80
[] Humlc Loam
70
Sand
8
ao1
20
tO 50
100
50
100
500
Immunofiltration [Ixg/kgl Fig. 6. Correlation of TNT soil extracts determined by immunofiltration or ELISA (e Humic soil, * loamy soil, • sandy soil). y = 1.12 x x - 3.68, r = 0.980 (n = 12)
The recoveries depend on spike level and type of soil The higher the content of TNT, the lower the recovery. The sandy soil has lower recoveries than the loamy soil and the soil rich of humic substances. Though the recoveries are not quantitative this method is a useful tool to detect contaminated sites. In order to verify the results of the soil samples with another method, the samples were analysed with an ELISA known to have a good correlation with gas chromatography [23]. As shown in Fig. 6 a linear relationship is obtained with a slope of 1.12, an intercept of - 3.68 and a correlation coefficient of 0.980 (n = 12).
Immunofiltration [gg/1]
< 0.1 a
2 3 4 5 6 7 8
10
500
Conclusions
An immunofiltration assay has been developed for the determination of TNT. Compared to enzyme linked immunosorbent assay (ELISA) in microtiter plates, this test design allows a very rapid analysis, because the diffusion distances between the reagents are very short. The presented method approved to be very robust in terms of matrix effects and methanol content of the soil extract. A dilution of 1 : 10 (v/v) of the soil extract is sufficent to avoid interferences. No additional clean-up is necessary. For natural water samples, the correlation between immunofiltration and G C is high. The assay time of about 6 min per sample could be reduced to 4 rain for on-site testing if antibody coating and blocking was set up in the laboratory. These results suggest that the method is well suited for rapid on-site analysis of T N T in water and soil. The variability for the immunofiltration assay is somewhat larger than for ELISA. One source of error is probably due to the efficiency and uniformity of antibody coating. Another variable is the flow rate of the reagents through the test device. These problems need to be overcome during industrial production.
i000
Spike level [txg/kg] Fig. 5. Recoveriesfrom three different soil samples with several spike levels (average + standard deviation of three replicates)
Acknowledgements. The authors wish to thank E. Merck (Darmstadt) for financial support and for supplyingantibodies and the reflectometer. We are grateful to H. Korpien and H. Rohner (Stadtwerke Frankfurt a. M.) for the water samples and the GC measurements.
543
References 1. Haas R, Stork G (1989) Fresenius Z Anal Chem 335 : 83%846 2. Spalding RF, Fulton JW, (1988) J Contam Hydrol 2 : 139-153 3. McCormick NG, Feeherry FE, Levinson HS (1976) Appl Environm Microbiol 31 : 949-959 4. Won WD, Heckly RJ, Glover D J, Hoffsommer JC (1973) Appl Microbiol 27:513-516 5. Fernando T, Bumpus JA, Aust SD (1990) Appl Environm Microbiol 56 : 1666-1671 6. Levine BS, Furedi EM, Gordon DE, Barkley JJ, Lish PM (1990) Fund Appl Toxicol 15:373-380 7. Spanggord RJ, Mortelmans KE, Griffin AF, Simon VF (1982) Environm Mutagenesis 4:163-179 8. Levsen K, Mugmann P, Berger-Preiss E, Volmer D, Wiinsch G (1993) Acta Hydrochim Hydrobiol 21: 153-166 9. Hable M, Stern C, Asowata C, Williams K (1991) J Chromatogr Sci 29 : 131-135 10. Jenkins TF, Walsh ME, Schumacher PW, Miyares PH (1989) J Assoc Off Anal Chem 72 : 890-899 11. Lafleur AL, Morriseau BD (1980) Anal Chem 52: 1313-1318
12. Mascarinec MP, Manning DL, Harvey RW, Griest WH, Tomkins BA (1984) J Chromatogr 302:51-63 13. Jenkins TF, Legett DC, Grant CL, Bauer CF (1986) Anal Chem 58 : 170-175 14. Jenkins TF, Walsh ME (1992) Talanta 39:419-428 15. Heller CA, Greini SR, Erickson ED (1982) Anal Chem 54 : 286-289 16. Van Emon JM, Mumma RO (1989) Immunochemical Methods for Environmental Analysis. ACS Symposium Series 442, Washington 17. Schneider E, Dietrich R, M~irtlbauer E, Usleber E, Terplan G (1991) Food Agric Immunol 3 : 185-193 18. Morber C, Weil L, Niessner R (1993) Fresenius Environm Bull 2:151-156 19. Siebert STA, Reeves SG, Durst RA (1993) Anal Chim Acta 282:297 305 20. Rittenburg JH, Fitzpatrick DA, Stocker DR, Grothaus GD (1991) Brighton Crop Prot C o n f - Weeds 1:281-288 21. Keuchel C, Weil L, Niessner R (1992) Anal Sci 8 : 9-12 22. Weller MG, Weil L, Niessner R (1992) Mikrochim Acta 108 : 29-40 23. Keuchel C, Weil L, Niessner R (1993) Vom Wasser 81:7-15 24. Medary RT (1992) Anal Chim Acta 258:341-346