Anal. Chem. 1994,156, 3834-3839

Flow Injection Assay for the Neurotoxin ,8-ODAP Using an Immobilized Glutamate Oxidase Reactor with Prereactors To Eliminate Glutamate Interferences Ghirma Mogest and Glllls Johansson' Department of Analytical Chemistry, University of Lund, Box 124, S-221 00 Lund, Sweden

The neurotoxic amino acid, 8-Noxalyl-L-cr,B-diaminopropionic acid (&ODAP,ODAP) was oxidized by immobilized glutamate oxidase (GlOD) to produce hydrogen peroxide. The peroxide reacts with Trinder reagent in a reactor with immobilized horseradish peroxidase to form a red-colored quinone imine dye, which was detected spectrophotometrically at 512 nm. Determinations were made in a flow injection (FI) setup consisting of four packed-bed enzyme reactors containing GlOD (20 pL), catalase (20 pL), GlOD (250 pL), and peroxidase (50 pL) in series. Glutamate is oxidized quantitatively in the first reactor, but the hydrogen peroxide is destroyed in the second so that interferences from this substrate are removed. This step destroys only a few percent of the ODAP in the sample. Most of the remaining ODAP is oxidized in the third reactor. Injections of 20-pL ODAP standards gave a response curve which was linear within the range 10-650 pM. Phosphate buffer extracts of grass peas (latlryrus ssfivus) were purified by centrifugation and membrane filtration. Samples were injected into the FI setup to assay the toxin at a rate of 20 samples per hour. The 8-ODAP content of a batch of dry seed corresponded to 0.74% (w/w) with a relative standard deviation of 2.8%. Thermal treatment of ODAP standards at 80-90 OC reduced the response to 62% of that before heating. The decrease is due to B F= cr isomerization, and the experiment thus confirms that the method is selective for the toxic &isomer.

The grass pea (lathyrussativus, LS), also known as Guaya in Ethiopia and Khesari in India, is a leguminous plant grown in the tropical belt but to some extent also in temperate zones. LS is a very important crop in Ethiopia and the Indian subcontinent. It is a flood- and drought-resistant plant, harvested at a time of the year when other crops fail to grow. The edible seeds are very important nutritionally in the mentioned regions since they contain 26-30% protein. The lathyrus crop, however, contains the neurotoxin ODAP, which causes irreversible, disabling paralysis of the legs in humans when it is present in the staplediet for some months.' Epidemic occurrences of the disease have been reported in areas where famine is widespread, particularly during years of drought.*

There has been a considerable effort to minimize or eliminate the content of ODAP in foods made from LS peas. The most obvious option is to produce plants with no or low

Permanent address: Department of Chemistry, Addis Ababa University, Box

(1) Rao, S. L. N.; Malathi, K.; Sarma, P. S . In World Review ofNutrition and Dietetics; Bourne, G. H., Ed.; Karger AG: Basel, Switzerland, 1969; Vol. 10, pp 214-238.

(2) Haimanot, R. T.; Kidane, Y.; Wuhib, E.; Kalissa, A,; Alemu, T.; Zein, Z. A,; Spencer, P. S. Int . J . Epidemiol. 1990, 19, 664.

1176, Addis Ababa, Ethiopia.

3834 Analytical Chemistty, Vol. 66, No. 21, November 1, 1994

amounts of ODAP through breeding and genetic program^.^ This entails processing a large number of samples, requiring a fast and selective method for monitoring the neurotoxin. Furthermore, assaying the toxin is important in studying its metabolic fate in plant and animal tissues as well as its neurotoxic action in humans and other animals.

The most widely used method for determining the neu- rotoxin in foods and seed samples involves its alkaline hydrolysis to L-a,@-diaminopropionic acid (DAP), followed by reaction with o-phthalaldehyde (OPT) to give a colored product which can be detected at 420 nm.4 This is a batch method which requires an extraction time of 6 h and a 1-h sample pretreatment prior to the final determination. The method is not selective, since the nontoxic or much less toxic isomer, a-ODAP (a @-amino acid), also hydrolyzes to DAP. The natural occurrence of the a-isomer in the pulse is only about 5% of the total, but the toxin isomerizes during heating, giving an equilibrium with about 40% in the a-form. '~~ The need for a @-isomer selective method is therefore quite evident, since reporting of faulty and high levels of toxin, due to isomerization, is inevitable in assays of foods after cooking and ingestion.

A few HPLC methods for ODAP have been developed in recent years, but all of them are based on inconvenient and time-consuming off-line pretreatments. These include dansyl chloride and 9-fluorenyl methylchloroformate derivatiza- ti on^,^,* methods which do not differentiate between the two isomers. Work on HPLC separation of both the a- and @-isomer with phenyl isothiocyanate (PITC) derivatization has been reported r e~en t ly .~ A method for HPLC separation of neuroexcitatory amino acids was used to test the optical purity. It was based on the same off-line derivatization using o-phthaldialdehyde and chiral thiols. @-ODAP was eluted within about 25-40 mineg

Analytical schemes for routine quantitative enzymatic detection of this compound do not exist to the best of our knowledge. There is, however, a report about the use of DAP- ammonia lyase, from Pseudomonad sp., for the determination of ODAP after its hydrolysis to DAP.'O The method does not, evidently, discriminate between the toxin and the nontoxic

(3) Kou, Y.-H.; Lambein, F., Eds. Abstracts, Lrrthyrus and Lrrthyrism, Progress and Prospects, Second International Colloquium; Bangladesh, Dhaka, Dec. 1&12, 1993.

(4) Rao, S. L. N. Anal. Biochem. 1978, 14, 387. ( 5 ) Khan, J . K.; Kebede, N.; Lambein, F.; De Bruyn, A. Anal. Biochem. 1993,

(6) Abegaz, 8. M.; Nunn, P. B.; De Bruyn, A,; Lambein, F. Phytochemistry 1993,

(7) Geda, A.; Briggs, C. J.; Venkataram, S. J . Chromatogr. 1993, 635, 338. (8) Kisby, G. E.; Roy, D. N.; Spencer, P. S. J . Neurosci. Methods 1993,26,45. (9) Euerby, M. R.; Nunn, P. B.; Partridge, L. Z . J . Chromatogr. 1989,466,407.

208, 237.

33, 1121.

0003-2700/94/0366-3834$04.50/0 0 1994 American Chemical Society

isomer, and no reports on its use have appeared since then. Mehta et al." reported on the application of tryptophanase, an L-amino acid group-selective enzyme, in degrading ODAP to pyruvate, which was identified by derivatization in thin- layer chromatography. Attempts to use this reaction were not successful in this laboratory.12

We recently reported the result of a screening of possible enzymatic reactions and the finding that L-glutamate oxidase (GlOD, EC 1.4.3.1 1, from Streptomycessp.) catalyzed a direct oxidation of ODAP.3v12 Hydrogen peroxide and ammonia, typical products of amino acid oxidase catalyzed reactions, were identified as products of the neurotoxin. The oxidation product was assumed to be an a-keto acid, p-N-oxalyl-a- keto-p-aminopropionic acid (or 0-N-oxalyl-0-aminopyruvic acid).

HOOCCONHCH,CHNH,COOH + 0, + H 2 0 - HOOCCONHCH,COCOOH + H,02 + NH, (1)

The rate of the homogeneous reaction was slow, and the activity of the enzyme to this new substrate was only about 0.8% relative to that of L-glutamate. The KM was 0.24 mM as compared to 0.21 mM for glutamate. Development of a kinetic or equilibrium enzymatic assay for the toxin, using soluble enzyme, was considered to be impractical because of the slow reaction and the very high cost of enzyme per assay.

Earlier papers have reported on flow injection methods for glutamate using immobilized G10D.'3J4 In this paper, we present a flow injection method for determination of ODAP using immobilized GlOD and peroxidase for detection with the Trinder15 chromogenic reagent. Glutamate interferences were eliminated on-line by complete oxidation in a 20-pL GlOD prereactor for subsequent destruction of the product, hydrogen peroxide, in a catalase reactor.

EXPERIMENTAL SECTION Immobilization of Enzymes. On 150 mg of controlled pore

glass (CPG-10 with 0.12-0.20 mm particle size and 51 nm pore size, Serva) was immobilized 4 mg of GlOD from Streptomyces sp. (EC 1.4.3.1 1, 6.9 units/mg of solid at pH 7.4 and 30 OC, Yamasa Corp., Japan) after toluene phase silanization with (aminopropy1)triethoxysilane and glutar- aldehyde activation. l6 Based on the absorbance of the enzyme solution before and after immobilization, the coupling yield was 86%. The activity of the enzyme was 2.1 units/mg at pH 7 and 22 OC as determined from theoxidation rate of glutamate (5.0 mM), followed by the chromogenic reaction (see below). The remaining activity in the filtrate and the washings showed that 97% of the activity had been immobilized. Two batches of enzyme glass were prepared, the first (batch 1) with a fairly low protein/glass weight ratio (4/150) in order to optimize the coupling yield of the expensive enzyme. The kinetic data and the high immobilization yield prompted us

(10) Rao, D. R.; Hariharan, K.; Vijayalakshmi, K. R. J. Agric. Food Chem. 1974,

(11) Mehta, T.; Hsu, A.-F.; Haskell, B. E. Biochemistry 1972, 11, 4053. (12) Moges, G.; Solomon, T.; Johansson, G. Anal. Letr., in press. (13) Yao, T.; Kobayashi, M. I.; Wasa, T. Anal. Chim. Acta 1993, 231, 121. (14) Stalikas, C. D.; Karayanis, M. I.; Tzouwara-Karyanni, S. M. Analyst 1993,

(15) Barham, D.; Trinder, P. Analysr 1972, 97, 142. (16) Wetal1,H.H. InMerhodsinEnzymology;Mosbach,K.,Ed.;AcademicPress:

New York, 1976; Vol. XLIV, pp 14C-142.

25, 1146.

118, 723.

to use a ratio of 6/150 (batch 2) for the analytical determina- tions. The data above relate to batch 1.

Catalase (EC 1.1 1.1.6,19 900 units/mg solid, Sigma C-40) was immobilized on CPG (0.5 mg/50 mg of glass) using the same method as for GlOD. Horseradish peroxidase (HRP, EC 1.1 1.1.7, 268 purpurogallin units/mg of solid, Sigma P-8375) was immobilized by the diazotization method." Silanized glass was refluxed for 4 h with 3% p-nitrobenzoyl chloride in chloroform, also containing 5% triethylamine. After the mixture was washed with chloroform, the nitro group was reduced by boiling in 5% aqueous sodium dithionite. Dia- zotization of the reduced nitro group on 100 mg of glass was then performed at 0 OC with 30 mg of sodium nitrite in 2 M hydrochloric acid. In 0.5 mL of phosphate buffer (0.1 M, pH 8.5), 4 mg of HRP was allowed to react with the diazotized product at reduced pressure (30 min) and kept at 4 OC overnight for further reaction. The absorbance of 0.2 mg/ mL HRP solution was 0.36 at 403 nm (enzyme) and 0.13 at 280 nm (total protein). The estimated coupling yield was 83% for the enzyme and 85% for total protein.

The immobilized redox enzyme was packed in 20-pL (i.d. 1.0 mm) and 250-pL (i.d. 2.0 mm) plexiglass tubes. Ap- proximately 110 mg of enzyme glass was sufficient for a 250- pL reactor. The volumes of the catalase and HRP reactors were 20 (i.d. 1.0 mm) and 50 pL (i.d. 2.0 mm), respectively.

Reagents. L-Glutamic acid, L-aspartic acid (from Sigma), 4-aminophenazone, and 4-AP (BDH) were used as received. ODAP was synthesized according to the methods of Rao with L-aspartic acid and sodium azide as starting materials.18 1,2- Dichlorophenol-6-sulphonate (DCPS) was synthesized from 2,4-dichlorophenol (DCP, Merck 3774) and concentrated sulfuric acid according to the method of Barham and Trinder,l5 and it was stored at room temperature at pH 1. Enzymes and reagents used for assaying aspartate were all from Sigma.

The reagent for the FI system was prepared daily in degassed buffer and consisted of 2.5 mM DCPS, 0.5 mM 4-AP, and 0.5 mM EDTA with DCP (0.25 mM) as recom- mended by 0lss0n.l~ It was purified on-line with a 3 mL Isolute SPE column (International Sorbent Technology, Mid Glamorgan, U.K.). The reagent, together with 5 units/mL HRP, was also used in a batch mode to measure the activity of GlOD in solution and to determine the percentage conversion of ODAP, aspartate and glutamate in the steady-state experiments. The carrier in the flow systems was a 0.1 M phosphate buffer (pH 7) unless otherwise specified.

Steady-State Experiments. Kinetic information was ob- tained from continuous flow experiments with 100 pM ODAP, aspartate, or glutamate in phosphate buffer at pH 7 (also at pH 6 and 7.6 for ODAP), using the 20-pL GlOD reactor with enzyme glass from batch 1. The effluent at each flow rate was collected, and an aliquot was mixed with the reagent in a 1.0-mL cuvette. The absorbance of the formed dye was determined spectrophotometrically at 5 12 nm against a reagent blank. The percentage of oxidized substrate was evaluated from the molar absorptivity of hydrogen peroxide (standard- ized by permanganate titration) or glutamate (22 000 mol-' L cm-1).

(17) Olsson, B. Mikrochim. Acra 1985, 2, 211. (18) Rao, S. L. N. Biochemistry 1975, 14, 5218.

Analytical Chemistry, Vol. 66, No. 2 1, November 1, 1994 3835

25

512 nm W

+&-$ Carrier Record

W

Flgure 1. Complete manifold for flow injection determlnatlon of ODAP; reactors 1 (20 pL of GIOD), 2 (20 pL of catalase), 3 (250 pL of GIOD), and 4 (50 WL of HRP); G is guard column; the flow rates are 0.3 (carrier) and 0.12 mL min-' (reagent). Reactor 3 is packed with enzyme glass from batch 2.

Flow Injection Setup. Figure 1 represents the assembly employed in studying the flow injection system for determina- tion of ODAP. It consisted of two HPLC pumps (Gilson 307 and LKB 2150) which delivered the carrier buffer and the reagents. Samples were injected from a 20-pL loop with a Rheodyne 7125 injection valve. The sample stream (0.3 mL min-l) merged with the reagent stream (0.12 mL min-I) at a confluence point for mixing in a knitted Teflon tubing (i.d. 0.5 mm, length 0.5 m). The red quinone imine dye which was formed in the HRP reactor was then detected at 512 nm in a flow-through cell with an LKB 2151 UV-vis spectropho- tometer.

Procedures. ODAP in LS seed powder (seeds purchased from a market in Addis Ababa) and in glutamate-spiked samples was determined using the flow system. Extraction of the toxin from 30-100 mg of seed powder was made by 10 mL of 0.1 M phosphate buffer (pH 7) at room temperature (<25 "C) by agitation with a magnetic stirrer for 1-2 h. Particulate matter was separated by centrifugation and filtration through a 1.2-pm membrane filter. Proteins and other macromolecules were removed by an ultrafiltration membrane (Amicon, molecular weight cutoff 10 000) using a centrifuge at 4000 rpm.

The aspartate content was evaluated by incubating 100 pL of the extract with 1 unit of glutamic oxaloacetate transami- nase, 50 units of malic dehydrogenase, 2 mM a-ketoglutarate, 100 pM pyridoxal 5'-phosphate, and 100 pM NADH at pH 7.4. The coilsumption of NADH was monitored at 340 nm in a 1-mL cuvette.

RESULTS AND DISCUSSION Reaction Kinetics. In the previous work, it was shown that

the two-substrate, soluble, GlOD-catalyzed reaction followed pseudo-first-order kinetics when the initial ODAP concentra- tion, S , was well below its K,.I2 A pseudo-first-order substrate reaction in a packed-bed enzyme reactor can be described by the following equation:19

-dS/dt = KpsaPPS ( 2 ) KpsaPP represents the apparent first-order rate constant of the reactor. Integration of this equation gives

-In[ 1 - XJ = KpsaPP7 (3)

where T is the mean residence time and X is the fractional conversion in the reactor. The mean residence time is given

(19) Johansson, G.; Ogren, L.; Olsson, B. Anal. Chim. Acto 1983, 145, 71

0 3 0 1 0 2 3 3 0 4 3 5 06 07 0 8

Flow rate, ml/mir

Flgure 2. Conversion efficiencyof the 20 pL GIOD reactor in continuous flows of 100 pM ODAP or aspartate (pH 7). ~

Table 1. Percentage Converdono Calculated from Klnetlc Data for a Reactor Filled wlth Enzyme Glass from Batch 1

conversion, % mean residence time, s glutamate ODAP aspartate

0.5 88 1.4 0.30 1 .o 98 2.2 0.37 2.0 99.6 3.7 0.50 3.0 99.99 5.2 0.63

10 100 15 1.5 20 100 21 2.8 40 100 46 5.3

by

T = *R2Le,/Q = L/F (4)

where R is the radius, L the length, and et the total void fraction of the reactor. Q is the volumetric and F the linear flow rate. The total void fraction for the glass (0.8) was determined in an earlier work.19

The percentage conversions of 100 p M ODAP, aspartate, and glutamate in a 20-pL GlOD reactor with enzyme glass from batch 1 were evaluated (see Figure 2 for the former two). The oxidative conversion of ODAP decreased from 20.5% at 0.05 mL min-* to 2.5% at 0.75 mL min-l. The same reactor converted glutamate with 100% efficiency up to 1.0 mL min-I.

Plots of -In( 1 - X ) vs T for 100 p M ODAP at pH 6,7, and 7.6 in the flow rate range of 0.1-0.5 mL min-l were linear. The highest slope was obtained at pH 7, where the plot could be described by a line with the equation y = 0.0065 + 0.01 5 3 ~ . The corresponding plots for aspartate (y = 0.0024 + 0.001 3x), and glutamate 0, = 0.182 + 3 . 8 3 ~ ) were also linear. The latter line was based on data in the range 2-4.5 mL min-l. These equations can be used to predict conversion factors at various residence times, see Table 1. It can be seen that the glutamate is oxidized quantitatively within a residence time of less than 2 s, while ODAP and aspartate are oxidized only to a negligible degree. A reactor with a volume of 20 pL should give a mean residence time of 3.0 s for a flow rate of 0.32 mL m i d .

A reactor size of 250 pL will give a mean residence time of 40 s at a flow rate of 0.3 mL min-l, and the predicted ODAP conversion will then be 46%. The mean residence time can be doubled either by decreasing the flow rate to half or by doubling the reactor volume. The conversion can also

3836 Analyticel Chemlstty, Vol. 66, No. 21, November 1, 1994

m E: 024 t J -- 0.24 II m e, .- 5 0.16

a 0.08

0

0.00 0

0.00 p n I

0 200 400 600 800 1000 1200 Concentration, uM

Flgure 3. Responses toward (a) hydrogen peroxide, (b) glutamate, (c) aspartate, and (d) ODAP using the complete flow Injection setup shown in Figure 1.

be improved by using a more active enzyme preparation, and the analytical reactor used in this work was therefore made with glass from batch 2.

Design of a FI System for ODAP. Glutamate oxidase is known to be a very selective enzyme; the only known reacting amino acids are glutamate and aspartate. Our screening experiments showed that 8-ODAP should be added to the list.3J3 Its reaction rate falls between those of the other two substrates. The data in Table 1 show that glutamate in a sample should be oxidized completely in the reactor with a residence time of 2-3 s. Glutamate interferences were therefore removed by insertion of a 20-pL reactor, followed by a catalase reactor to destroy the formed hydrogen peroxide. About 5% of the ODAP will also be oxidized in the first reactor and thus escape detection, because the hydrogen peroxide is destroyed. Glutamate is predicted to be decomposed to 99.99% at a flow rate of 0.3 mL min-l.

A 250-pL enzyme reactor is inserted after the catalase reactor, see Figure 1, and according to Table 1 it should oxidize 46% of the remaining ODAP at a flow rate of 0.3 mL min-l (the mean residence time is 40 s) if filled with glass from batch 1. The trick for removal of the interference is to exploit the different kinetic propeties of the two substrates.

The curves in Figure 3 are plots of FI responses to hydrogen peroxide, glutamate, and aspartate in the complete FI system. The upper curve shows that the response to ODAP is linear up to 650 pM. The aspartate response is not noticeably affected by the first reactor, and it is still a theoretical interference, but this is of little practical concern since the aspartate concentrations in the pea extracts were negligible, as will be discussed below. Glutamate is destroyed up to 500 pM but will give an increasing interference if present at higher concentrations. The catalase reactor destroys hydrogen peroxide completely up to the studied limit, 2000 pM. Elimination of interferences with reactors containing an oxidase and catalase is not uncommon in FI, e.g., in elimination of glucose in sucrose determinations,*O but the use of the kinetic differences between the substrates is new to the best of our knowledge.

The analytical properties of the system are summarized in Table 2 and compared with those of the other substrates. Glutamate and hydrogen peroxide give the same response up

(20) Olsson, B.; Sthlbom, B.; Johansson, G. Anal. Chim. Acta 1986, 179, 209.

Table 2. Properties of the Analytlcal F I System wlth a 250 pI GIOD Reactor FIIIed wlth Enzyme Glass from Batch 2 but wlth No Prereactors (Sample Loop 20 pL)

sensitivity, linear relative RSD,' substrate au/mM range, pM response, % % (n = 5)

ODAP 0.54 10-650 80 1.5 aspartate 0.1 25-500 15 2 glutamate 0.68 5-2000 100 1.3 hydrogen peroxide 0.68 5-2000 100 1.5

Relative standard deviation for 100 pM substrate.

Table 3. Mean Recovery of ODAP Addttlonr ( n = 2) In the Presence of 500 pM Glutamate wtth the Complete F I Injectlon System Shown In Flgure 1

ODAP concentration, pM added found recovery, %

75 I8 104.0 150 151.2 101.3 250 258.3 103.3 500 5 1 1 102.2 600 594.6 99.0

0 0

0.1 80

3 Q 0.1 35

r cn c)

.- ," 0.090 Y

a 0.045

0.0001 " " " " " " 0.04 0.20 0.36 0.52 0.68 0.84 1.00

Flow rate, ml/min Flgure 4. Effect of total flow rate on the flow injection peak helghts for 150 pM ODAP. Injection volume was 50 pL, and carrlerlreagent flow rate ratio was 1:0.4. The enzyme glass was from batch 1; no prereactors were used.

to 1.00 mL min-l if the prereactors areremoved. This confirms that the glutamate reaction produces hydrogen peroxide quantitatively, as expected. The ODAP conversion is 80%, i.e., much higher than that given in Table 1. The reason is that the data in Table 1 were collected on glass from batch 1, whereas those in Table 2 are based on data obtained with glass from batch 2. The validity of the arguments for removal of glutamate interferences was tested with the prereactors in the system by determining the recovery of ODAP in the presence of 500 pM glutamate, see Table 3. The recovery has a mean relative error of 2.8% and an relative standard deviation of 1.24% (n = 5).

Optimization. The effects of flow rate, pH, and reagent concentrations were studied in a setup without prereactors and with a 50-pL sample loop. Data on flow rate and pH dependence were obtained with a reagent consisting of 8 mM DCPS, 1 .O mM DCP, and 1 .O mM 4-AP. Figure 4 shows the effect of the overall flow rate on the peaks of ODAP. Maximum output was obtained for a flow rate of 0.28 (0.20 + 0.08) mL min-I, but the throughput was rather low (about 14 samples per hour). Studies on other variables and calibrations of the system were therefore made at a total flow rate of 0.42 mL min-l.

AnalyricalChemishy, Vol. 66, No. 21, November 1, 1994 3837

0.30

a 0.24 1 I

? ! c 0.1 a

0.00 I I 5.0 5.4 5.8 7.2 7.6 8.0

PH Flgure 5. Effect of the carrier pH on the flow injection peaks of 150 pM glutamate (a) and ODAP (b). Experimental conditionsare the same as in Flgure 4.

A plot of the flow injection peak heights for ODAP and glutamate against the carrier pH is shown in Figure 5. The optimum carrier pH is 6.5-7.5 for both substrates. A recently reported two-channel FI GlOD reactor system for glutamate with soluble HRP in the Trinder reaction had an optimum pH of 7.8.14 In the present case, therefore, a significant pH shift, with a broad optimum, occurred due to the immobilization of HRP. The system was run at pH 7 during routine calibrations and other studies.. The concentration of the Trinder reagent was also varied, and the optimum, which is given in the Experimental Section, was selected for the applications described below.

Calibrations of the FI System. Injections of 20- and 50-pL standards produced calibration curves that were linear in the range 10-650 and 5-300 pM ODAP, respectively. The range with the 20-pL injection is sufficiently wide to cover typical ODAP concentrations. An extract of 100 mg of LS seed powder in 10 mL of buffer should typically contain 60-500 pM (corresponding to a content of 0 . 1 4 8 % in dry seed). Hence, the lathyrogenic compound can easily be determined in such samples. The sampling frequency of the system is 20 samples per hour (peak width 3 min), which is very high by standards of the nonenzymatic OPT or chromatographic methods. If desired, the flow rate can be more than doubled to increase the sampling rate to more than 40 samples per hour, with a sensitivity decrease to about 60% at 0.85 mL min-l. The detection limits were 6 and 3 p M for the smaller and larger sample volumes, respectively.

The sensitivities indicated in Table 2 remained essentially constant for more than 3 months. The activity of the amino acid oxidase with respect to ODAP had not decreased by more than 10% at the end of the period. When the reactors were not in use, they were stored in the carrier buffer at 4 "C. The long lifetime of the immobilized enzyme will offset the high initial cost and make the method economical. The good thermal stability of the enzyme2' indicates that the lifetime will be good even at tropical room temperatures.

Determinations of ODAP. Ground LS seeds were extracted with buffer and purified as described in the Experimental Section. The extraction in the phosphate buffer seems to be efficient since it is complete within 1 h (Table 4). No ODAP

~~ ~ ~ ~~

(21) Kusakabe, H.; Midorikawa, Y.; Kuninaka, A.; Yoshino, H. Agric. B i d . Chem. 1983, 47, 179.

Table 4. Extraction of ODAP from LS Seed Powder wlth 10 mL of Phosphate Buffer for Various Lengths of Time'

weight, mg time, min % (w/w) remarks sample extraction ODAP found,

30 30 0.58 50 60 0.74 50 60 0.75 without catalase reactor 50 90 0.70 50 120 0.74 50 120 0.76 spiked with glutamate

100 90 0.72

a ODAP determinations in the membrane-filtered extracts were made with the FI system shown in Figure 1 with GlOD enzyme glass from batch 2; sample loop 20 g k total flow rate 0.42 mL min-l. The data in the table are means of duplicate injections.

was found if the solids removed during centrifugation were washed with buffer and extracted a second time. Direct injection of the extract resulted in erratic and broad peaks after a number of samples, probably because proteins or cell fragments were adsorbed in the reactors. The reactors could be restored by washing with acid buffer (pH 5.7) until recovery. To minimize adsorption of impurities, the samples were cleaned further by ultrafiltration prior to injection. Data in Table 4 also show that removal of the catalase reactor results in essentially the same response, indicating that the glutamate content of this particular batch of seeds was low. Spiking with 300 p M glutamate had no effect. The mean value for the values with 1-2-h extraction time is 0.74% ODAP in the dry seed with a between-run relative standard deviation of 2.8% (n = 6 pairs). Two FI determinations were made on each extraction.

Previous workers have show that over 90% of the free amino acids in mature dry LS seeds are ODAP and homoarginine.22 GlOD is a very selective enzyme,21 and most of the common amino acids will be inert even in a reactor.14 No aspartate was found in the seed extract using the enzymatic assay described in the Experimental Section, and the FI system seems therefore to be selective for ODAP in such samples. The present method should therefore be quite attractive for ODAP determinations, not only in LS seeds but also in foods.

In our previous work we tried to determine ODAP at 37 "C using soluble enzyme.12 Low values were obtained due to a competing isomerization of the substrate to the CY form. This indicates the importance of making the extraction at room temperature, <25 OC. Abegaz et ale6 observed a 29% loss of P-ODAP due to isomerization when a solution was kept at around 22 "C for 1 week.

The effect of isomerization was demonstrated by keeping the ODAP standards at 80 O C for various times. Equilibrium was attained within 10 min for the 100 pM and 40 min for the 400 pM standard. Raising the temperature to 90 OC nearly halved the equilibrium time. A study of the complex kinetics of the isomerization process is in progress. The percentage of the toxin in the equilibrium mixture after incubation was 62% (relative standard deviation of 3.2%, n = 5). The a:@ equilibrium ratio 38:62 is in good agreement with the 40:60 ratio reported by previous worker^.^^^ This result confirms that the enzyme is selective for the toxic form. As pointed out in the introduction, such specificity has great

(22) Lambein, F.; Khan, J. K.; Kuo, Y.-H. Plonla Med. 1992, 58, 380.

3838 Analytical Chemistry, Vol. 66, No. 21, November 1, 1994

implications for assays ofthe neurotoxin incooked preparations of the lathyrus seeds.

ACKNOWLEDGMENT The authors wish to thank Professor B. M. Abegaz, Addis

Ababa, for gifts of P-ODAP. This work was supported by

grants from the Swedish Agency for Research Cooperation with Developing Countries (SAREC), and the Swedish Natural Research Council (NFR).

Received for W V ~ W 17, 1994. Accepted June 299 1994.'

Abstract published in Aduonce ACS Abstrucrs, August 1, 1994.

Analytlcel Chemism, Vol. 66, No. 21, November I , 1994 3839