Anal. Chem. 1990, 62,2038-2043
2038
Identification of Archaeological and Recent Wood Tar Pitches Using Gas Chromatography/Mass Spectrometry and Pattern Recognition Erich W. H. Hayek, Peter Krenmayr,* and Hans Lohninger Institute of General Chemistry, Technical University of Vienna, Getreidemarkt 9, A-1060 Vienna, Austria
Ulrich Jordis, Wolfgang Moche, and Fritz Sauter* Institute of Organic Chemistry, Technical University of Vienna, Getreidemarkt 9, A-1060 Vienna, Austria
An analytical method has been developed for the asslgnment of recent and archaeological wood tar pitches to the species of trees used for their preparation. I t incorporates the prepurification by Kugelrohr dlstUlation and solklphase extraction followed by gas chrmatography/mass spectrometry analysis. Distribution patterns of vdatk, thermostable trlterpenokls and sterolds characterizing the bidogical origin of the pitches are evaluated by principal component analysis (PCA) and discriminant component analysis ( E A ) of the data. QuantHies of 10-100 mg of archaeological material can thus be characterlzed. Fourteen archaeological samples have been identified as birch bark pitches by comparison with recent pitches of known origin. I n addition 28 substances present In recent barks and recent pitches could be ldentlfied showing that the method developed can also be applled to the chemotaxonomy of broad-leaved trees.
INTRODUCTION Dark, ductile lumps ( I , 2 ) , black partially shiny coatings on pottery (I*), and residual glues on flint-stone arrowheads constitute a class of organic matter excavated repeatedly in swamps and tombs and found in urns ( I , 2). These finds have been termed, partly misleadingly, “tomb-resins”, “fumigating resins”, “fumigating pitches”, “urn resins”, “fumigating cakes”, or “resin cakes” (1-3,9). Yet, according to the manufacturing process, which constitutes in essence a “slow” pyrolysis (10-141, these mixtures of substances should be termed wood tar pitches (distillation residues of wood tars). Finds of this kind have been verified from the mesolithic period up to the Middle Ages, the origins of discovery stretching from southern Scandinavia to the Balkans (1-8, 15). The uncommonly broad range of applications of wood tar pitches in prehistoric times comprised, among others, use as medicines, incense substitutes, inside or outside coatings of pottery, agglutinants (e.g. arrowhead to shaft), and sealing compounds (1-5, 8, 9, 16). Wood tar pitches (mainly coatings of pottery) repeatedly have been analyzed since 1880, mainly via infrared or NMR spectra or by thin-layer chromatography (TLC) (2-4,8,17-20). Motivated by analogy to ethnological and folkloric traditions still existing in northern Europe, analytical work was focused on attempts to identify the material in question as tar from low-temperature carbonization of birch bark. Yet, recent papers demonstrated, that infrared and NMR spectroscopy and TLC are not sufficient to justify unequivocal characterizations ( 3 , 4, 17). In most cases samples were prepurified
* Corresponding authors.
Table I. Samples Used for Comparison number of samples (entries in mass spectra librarv) bark pitch Acer plantanoides Alnus incana Alnus glutinosa Betula alba Carpinus betulus Corylus avellana Fagus silvatica Fraxinus excelsior Juglans regia Prunus auium Quercus cerris Quereus petrea Ulmus minor
by-unspecific-solvent extraction. Although in two exceptional cases (where sufficient archaeological material was available) betulin-a triterpenoid-enediol occurring in birch bark in large quantities-was isolated and identified ( 4 2 11, a general methodology allowing the correlation of unknown wood tar pitches to particular species of trees was lacking. Tars from coniferous trees were characterized previously by gas chromatography/mass spectrometry (GC/MS) (22),focusing on tricyclic diterpenoids, after solvent extraction and
LC. Due to the isolation and identification of betulin in one particular archaeological sample @ I ) , our scrutiny focused on native deciduous trees, as triterpenoids, products of secondary metabolism of Dicotyledonae, occur mainly in deciduous trees, but not in coniferous trees (23).
EXPERIMENTAL SECTION Tetrahydrofuran (THF) and methanol (Riedel de Haen, p.a.) were distilled twice by use of a 1-m fractionating column. Referythrodiol erence substances betulin (lup-20(29)-ene-3@,28-diol), (olean-l2-ene-3/3,28-diol), friedelin (D:A-Friedoolean-3-on),lupeol (lup-20(29)-en-3/3-01),uvaol (urs-12-ene-3/3,28-diol), amyrine isomeric mixture, lupenone (lup-20(29)-en-3-on),and a-lupeol (lup-20(29)-en-3a-ol)were obtained from Carl Roth GmbH. Air-dried barks (24 h at 125 “C) of 13 species of decidious trees were used as raw materials (Table I). Preparation of authentic pitches (Table I) was carried out via descending destructive distillation: Samples of the barks were subjected to a low-temperature carbonizing process. The electric powered tubular stove (Figure 1) was heated to 330-370 “C for 1 h and the temperature raised to 620 “C for another hour. Pitches were obtained by evaporation of the tars to remove the volatile
0003-2700/90/0362-2038$02.50/00 1990 American Chemical Society
ANALYTICAL CHEMISTRY, VOL. 62, NO. 18, SEPTEMBER 15, 1990
2030
apectra of bark substances
spectra of substances of wood tar pitches
1
library
1
search
Figure 1. Tubular stove for the preparation of recent wood tars: (a,
e) glass wool: (b) glass tube filled with pieces of bark: (c)thermostat; (d) isolation and filament; (f) funnel; (9) sample collector.
mass apectra of archaeological samples
C
4
4
library search
retention time
4 Figure 2. Kugelrohr distillation: (a) glass wool; (b) sample; (c, d, e)
electrical heating and thermostat. compounds at 20 mbar and 80 "C. Yields of the pitches obtained ranged between 3.1% (Quercus cerris) and 14.3% (Betula alba). Kugelrohr distillation (Figure 2) of the pitches at O.OO6-0.06 mbar and 180-220 "C gave 46-86% oily or partially crystalline, yellow to brown products. Kugelrohr distillation of the archaeological samples or of dried barks without application of the carbonizing process yielded, under the same conditions, 1-21% of volatile products. One milliliter of a T H F solution of the distillates (10 mg/mL) was injected onto a SPE (silica phase column) (Analytichem) and eluted with 2 mL of THF. The solvent was evaporated under N2 and the residue redissolved in 1 mL of Me0H:THF = 8:2 and subjected to the SPE reversed-phase column (Analytichem). After elution with 2.5 mL of the same solvent mixture and evaporation of the solvents under N,, the samples were dissolved in 6 mL of THF, and 120 pg of PCB-209 was added as internal standard. This T H F solution was used for GC/MS, as the presence of MeOH interfered with the detection of the triterpenoids by activating the separation column. Separation was optimized using a capillary gas chromatograph, Carlo Erba Fractovap 4160, a 0.33 mm X 30 m DB5 capillary column (J&W, 0.1 pm) at a column pressure of 0.9 bar with helium as carrier gas and the following temperature program: on column injection at 45 "C, 50 "C/min to 260 "C, hold 3 min, 1.5 OC/min to 290 "C as final temperature. The GC was connected to the Finnigan MAT 8230 mass spectrometer via an open capillary interface. Separator temperature was 290 OC and ion source temperature 250 "C. Mass spectra were recorded at 70 eV electron energy, the mean MS cycle time was 1.25 s. Samples were characterized by means of patterns of corresponding GC peaks. For the selection of characteristic peaks mass chromatograms of the molecular ions and significant fragments of the triterpenoids and steroids with different numbers of hydroxy and acetate functional groups were recorded. A data base containing both the mass spectra of reference substances as well as mass spectra corresponding to characteristic GC peaks of recent pitches was created (Figure 3A). Mass spectra of unknown substances were characterized via a computerized library search using this data base. An identification was regarded conclusive if the result of the search withstood visual comparison with the spectrum in the data base and matched relative retention time (Figure 3B). The characterized peaks were quantified in relation to the internal standard. Data were transferred from the mass spectrometer data system to an external personal computer. Two data sets were generated: set 1 contained 86 features (intensities of characterized substances) representing 19 recent pitches, with emphasis on those occurring
1
visual check
1 catalog of
Figure 3. Flow chart for the selection of characteristic compounds (A) and the characterization of unknown samples (B).
only in trees of one family; set 2 contained all data of 38 recent samples (174 features). The data were processed by using methods of multivariate statistics, especially linear mapping procedures within EDAS (exploratory data analysis of spectra) (24,25). The number of features was reduced discarding those with the smallest variances of their intensities. Normalized data were obtained by application of eq 1 and used as input for the principal component analysis (PCA)
where u is a normalized feature value, u is a raw feature value, n is the number of features, i is the index of samples, and j and k are the index of features. Unsupervised and class-independent PCA was used to uncover grouping of the data. Elimination of principal components with minor variances or minor Fisher ratios reduced the number of descriptive variables. Autoscaling normalized the variance of each feature and transferred the mean to 0. Then DCA (discriminant component analysis) was performed, optimizing the separation of two given classes of data.
RESULTS AND DISCUSSION Archaeological and recent pitches can be related to the trees they were derived from by identification of patterns of triterpenoids and steroids. T h e method described utilizes both the thermal stability and volatility of triterpenoids and steroids (26,27)using Kugelrohr distillation for prepurification. Thirty-eight samples from 13 species of native trees were analyzed by GC/MS resulting in a mass spectral library containing 174 substances (Table I). Twenty-eight substances present in wood tar pitches and distillates from barks (mostly pentacyclic triterpenoids) were identified by comparison with standards and/or by matching the mass spectra with the NBS mass spectral library (Table 11). One compound, namely
ANALYTICAL CHEMISTRY, VOL. 62, NO. 18, SEPTEMBER 15, 1990
2040
Table 11. Identified Substancesn Alnus
maple glutinosa
Alnus incana
birch
hornbeam
hazel beech ash nut
Prunus avium
Quercus cerris
oak elm
B
1-(2,4-dihydroxy-6-meth-
oxyphenyl)-3-(2-hydroxyphenyl)-2-propen-l-one 30-norneohop- 13(18)-en-22one A-neoolean-3(5),12-diene allobetulone
BP
BP BP BP
(18a,l9u’-epoxy-oleanan-
3-one) B taraxerone (D-friedoolean-14-en-3one) BP taraxerol (D-friedoolean-14-en-38-
BP
BP
BP
B
BP
B
01)
taraxerolacetate friedelin (D:A-friedooleanan-3one) D:A-friedooleanan-2-one friedelanol (D:A-friedooleanan-38-
BP B B
B
01)
B
isomultifluorenone (D: C-friedooleanan-%en%one) fernene (D:C-friedo-B’:A ‘neogammacer-9(1lbene) heneicosylcyclopentane lanost-8,24-dien-3B-ol lup-20(29)-en-3-one lupeol (lup-20(29)-en-3/3-01) betulin (lup-20(29)-en-3P,28diol) a-amyrine (olean-12-en-3$-01) erythrodiol
BP
BP B BP
BP B BP
B BP BP BP
B
B
B B
B B B
B
B
B P
BP B BP
B B
B
BP
B
(olean-lZ-en-3/i’,28-diol)
olean-13(18)-en-3-one stigmastan-3,6-dione stigmast-4-en-3-one stigmast-5-en-3a-ol stigmasta-3,5-dien-7-one
B B
B B B B
B P B
BP BP
B B
B B B B
B BP B
B
P
B
B B B
B BP BP BP
B
stigmasta-5,24(28)-dien-301
a-amyrine (urs-12-en-3B-01) &tocopherol
P BP
B
Key: P, identified in pitch; B, identified in bark; *, high loading in data set 2 pointing to birch; f , high loading in data set 2 pointing to other trees.
lupan, could be detected in some archaeological samples, but not in any recent pitch or bark distillate. Examples of the GCIMS analyses are shown in Figures 4 and 5 to demonstrate (a) the reproducibility of the patterns of substances for different samples of the same species of trees (Figure 4A,B), (b) the differences between the patterns of substances for different species of trees (Figure 4C), and (c) the similarity of archaeological samples to recent birch pitches (Figure 5). The reliability of the classification of unknown samples via the presence or absence of characteristic compounds was tested in a blind study where one of the authors (E.H.) correctly classified five recent wood tar pitches from different species of trees. T o classify the archaeological samples, pattern recognition methods were applied. Concentrations of characterized substances were used as coordinates in n-dimensional space, in the expectation that those points representing samples
originating from the same species of wood would form clusters. To project the pattern points onto a two-dimensional plane, PCA and DCA were applied (24, 25, 28). As feature selection and normalization of data are critical points of the chemometric analysis of GC data (29),various methods (i.e. feature selection for a maximum of Fisher ratio, logarithmic transformation, or other normalization methods) were tested giving similar but less obvious results than the methods described (i.e. eq 1). PCA of the recent pitches shows good separation with regard to different types of trees (Figure 6). The separation of birch from other recent trees could be accomplished with a varying number of features, 10 to 30 being satisfactory for data set 1 (recent pitches) and 20 to 45 for data set 2 (all recent samples). Projecting the data of 14 archaeological samples (Table 111) onto the PCA plot of data set 1 allowed the unequivocal identification of 11 samples as birch bark pitches (Figure 6).
ANALYTICAL CHEMISTRY, VOL. 62, NO. 18, SEPTEMBER 15, 1990
I
,LA
B
m
z w
d
n
n
t
7
424 A
416
1,
I
426
h
h.
I
2041
A
414
I\
412 I
.
h
-_
410
I
A
408
R
I
A
396 ..I
A
394
C 2 3
C -
15
20
25
min
Flgure 4. Integrated ion intensity of birch bark distillates (A, B) compared to oak bark distillates (C). Numbers refer to unidentified entries of the library. In part C the mass chromatograms of the molecular ions and significant fragments of triterpenoids and steroids are displayed.
hol
C
1 2-DIEN
Fc
$~[~?!!~
25 -
1548
min
1773
A c-
lntegr Intensity
-
BETULIN
B
Figure 5. Integrated ion intensity of archaeological sample 10 (A) compared with a recent birch bark pitch (B). Numbers refer to entries of the library, identical compounds are marked by double arrows.
In order t o answer the question of the relationship of the remaining two samples, a discrimination analysis of data set 2 with birch bark as one class and all other barks as the other was made. PCA was carried out to reduce the dimensionality of the data by setting a minimum of variance or Fisher ratio of those principal components used as input for DCA. In general, autoscaling before DCA improves the class separation. Alternatively features with highest loadings in the selected principal components were chosen to generate a data set containing only 15 features of all recent samples. Eight of these features correspond to identified substances (marked in Table 11),another is assumed to be 19,2&epoxyolean-2-ene (apoallobetulin). Again PCA was carried out and five principal components with major Fisher ratios were selected for DCA. The classification ability was tested by the “leave a quarter
ai PCA: Principal Component I
A
E
A
H D K
ED I
variance 31.2 Z
Flgure 6. Principal component analysis of recent pitches using 30 selective features showing the separation of familles of trees and the
close relationship of the archaeological samples as birch tar pitch: (A) birch (Betula alba); (A) elder (Alnus glutinosa and Alnus incana); (B) hazel (Corylus avellana); (C) beech (Fagus silvatica): (D) maple (Acer plantanoides); (E) hornbeam (Carpinus betulus); (F) ash (Fraxinus excelsior); (H) nut (Juglans regia); (I) oak (Quercus petrea); (K) elm (Ulmus minor); (+) archaeological samples, which can be related to birch; (a)archaeological samples 7 and 8 (Table III), which cannot be related to either birch or hornbeam. out” method showing predictabilities better than 95% for each class (Table IV). Thus, the use of chemometric methods resulted in a striking reduction of the number of features. By use of any of the DCA methods described, 14 archaeological samples (glues on arrowheads and on pottery, coatings on pottery, and a large early Iron Age piece of raw wood tar pitch (Table 111)) could be identified as birch bark pitches (Figure 7), even in those cases where betulin could not be
2042
ANALYTICAL CHEMISTRY, VOL. 62, NO. 18, SEPTEMBER 15, 1990
Table 111. Archaeological Samplesa sample ID 1 2 3 4
5
6 rn
8
9 10 11 12 13 14 15
finding site
characterization of the object
dating (after Reinecke)
Pitten, Lower Austria Buchsberg, Lower Austria Kirchspiel, Denmark Spjald, Western Jutland, Denmark
bulgy bowl (omphalos type) with ancient glued joint surface find: coating on pottery shard (from bowl) lump of pitch found in the Herstedvester Moor coating of wooden box found in a depot of an ancient house Spjald, Western Jutland, Denmark lump of pitch found in a depot of an ancient house ancient glued joint on a pottery bowl (grave gift) Hochberg, Lower Austria Mondsee, Upper Austria coating on triangular flintstone arrowhead Mondsee, Upper Austria coating on triangular flintstone arrowhead Stillfried, Lower Austria coating on pottery shard (found in a sacrificial pit) Stillfried, Lower Austria contents of pottery bowl Stillfried, Lower Austria coating on pottery shard Stillfried, Lower Austria coating on pottery shard Stillfried, Lower Austria contents of pottery bowl Stillfried, Lower Austria loaf-shaped lump of pitch (found in a maybe "sacrificial" pit) Stillfried, Lower Austria charcoal and organic material (found in a pit containing a human skeleton)
Bronze Age (R. Br. C/D) Early Bronze Age (R. Br. A2/B1) Bronze Age Bronze Age (R. Br. C/D) Bronze Age (R. Br. C/D) Early Iron Age (R. H. C/D) Copper Age Copper Age Late Bronze Age (R. H. B) Early Iron Age (R. H. C/D) Late Bronze Age (R. H. B/C) Late Bronze Age (R. H. B/C) Late Bronze Age (R. H. B/C) Late Bronze Age (R. H. B) Late Bronze Age (R.H. B)
aSamples 1-14 are classified as birch bark derivatives, 15 is assigned as an animal fat.
Table IV. Predictive Abilities of Chemometric Classifications number of features used for PCA for DCA 40 40 15a
I
I :
?& predictability
for birch
for other trees
overall 97 100 100
5
100
96
10 .i
100 100
100 100
I:
Features with highest loadings in selected principal components II
detected. Sample 15 contained only three steroids (cholestan-type dienones) and cholest-5-en-3P-01 acetate as well as fatty acids but no triterpenoids. This suggests that the sample represents an archaeological animal fat.
ACKNOWLEDGMENT The authors wish to thank R. Strba (Agricultural University Vienna) and the staff of the "Bauhof Hutteldorf" of the Austrian federal forest administration for recent bark samples and their botanical identification; the Institute of Prehistory and Early History of the University Vienna (especially F. Felgenhauer), the Department of Prehistory of the Museum of Natural History, Vienna, and C. J. Becker and the first department of the National Museum, Copenhagen, Denmark, for providing archaeological samples; and K. Varmuza and W. Werther, Institute of General Chemistry, for critical discussions and the use of software packages PEDAS and EDAS. Registry No. 1-(2,4-Dihydroxy-6-methoxypheny1)-3-(2hydroxyphenyl)-2-propen-l-one,69707-17-1; 30-norneohop-13(18)-en-22-one, 128114-74-9;A-neoolean-3(5),12-diene,22586-84-1; allobetulone, 28282-22-6; taraxerone, 514-07-8;taraxerol, 127-22-0; taraxerolacetate, 2189-80-2; friedelin, 559-74-0; D:A-friedooleanan-2-one, 17947-04-5; friedelanol, 105370-95-4;isomultiflorenone, 22611-26-3; fernene, 1615-99-2; heneicosylcyclopentane, 6703-82-8; lanost-8,24-dien-3fi-ol, 79-63-0; lup-20(29)-en-3-one, 1617-70-5;lupeol, 545-47-1; betulin, 473-983; 6-amyrine, 559-70-6; erythrodiol, 545-48-2; olean-13(18)-en-3-one, 20248-08-2; stigmastan-3,6-dione, 77551-74-7; stigmast-4-en-3-one, 1058-61-3; stigmast-5-en-3n-01, 31793-83-6; stigmasta-3,5-dien-7-one, 2034a-amyrine, 638-95-9; 72-2; stigmasta-5,24(28)-dien-3-01,18472-36-1; d-tocopherol, 148-03-8.
3
b
I I I I I I I I I I I I
B rs+
*b + A.
+4
+
I
DCA: discriminant component
Fwhm ralio 100.7
Figure 7. Discriminant component analysis of birch bark versus all
other trees after previous PCA (40 features) and data reduction (10 principal components with major Fisher ratios) showing the classification of the archaeological samples as birch: (A)birch (Betub alba); ) . ( other trees; (0)archaeological samples 7 and 8; (+) other archaeological samples: (1) classification border.
LITERATURE CITED (1)
(2)
Thomsen. P. in Reailexikon der er: Berlin, 1926; pp 128-130. Sandermann, W. Technische
Vorgeschichte; Ebert, M., Ed.; SpringBeitrage zur ArcGoiogie 1965, 2 ,
58-73.
(3) Sauter, F. Archaeoi. AustrLaca 1967, 4 1 , 25-36. (4) Sauter, F.; Jordis, U. Forschungen in Stillfried 1980, 4 , 147-161. (5) Schwarze. E. Ausgrabungen und Fun& 1963, 8 , 17-18. (6) Behrens. H. Ausgrabungen undFunde 1960, 5 , 12-19. (7) Protz, H. Berliner Blatter fir Vor- und Frlihgeschichte 1985l66, 1 1 , 153-170. ( 6 ) Funke, H. Doctoral Thesis, Hamburg, 1969. (9) Hager, H.. List, P. H., Horhammer, L., Eds. Mndbuch der Pharmazeutischen Praxis, 4th ed.; Springer: Berlin/Heidelberg/New York, 19671980; Voi. 3, pp 434-438, and Vol. 6, pp 737-746. (10) Schweers, W In Chemische Technologic, 4th ed.; Winnacker, K., Kuchler. E.. Eds.; Hanser: Munchen, 1981; Vol. 5, pp 644-647. (11) Nikitin. N. 1. Chemie des Holzes: Akademie: Berlin, 1955; pp 416-453. (12) Bunbury. H. M. The Destructive Distillation of Wood; Benn Bros.: London, 1923; pp 259-269. (13) Bugge. G. Die Hoizverkohlung; Sammlung Gijschen: Berlin, 1925; pp 7-9, 36-41, 64-71, 86-91, 98-101. (14) Soltes, E. J.; Wiley, A . T.; Lin. S. C. K. Biotechnol. Bioeng. Symp 1981. 7 1 . 125-136
AM. Chem. 1 ~ 9 0 62, , 2043-2048 (15) (16) (17) (18) (19) (20) (21) (22) (23) 124) ~I
Schoknecht, U.: Schwarze. E.
Ausgrabungen und Funde 1967, 72, 205-209. Thoms, H. Handbuch der praktischen und wissenschsftlichenpharma l i e : Springer: Berlin, 1924-1931; Vol. 4, pari 2/1, pp 1608-1609, p 1766. Hadzi, D.; Cvek, F. Archeoloski vestnik 1976, 2 7 , 128-134. RottYnder, R. C. A. Archiiobg. Kwesmn&nzb&tt 1974. 4 . 95-98. RottlBnder, R . C. A. I n Die Schussenriedersiedlung im “SchlGsslesfeM”: Luning, J.. Zurn. H., Eds.. 1977. Heintzei, C. 2.Ethnologic 1880, 72, 375-378; 1881, 73, 241-242. Sauter, F.; Hayek, E. W. H.; Moche, W.; Jordis, U. 2. Natudorsch. 1867, 42‘2, 1151-1152. Evershed, R. P.; Jerman, K.; Eglinton. G. Nature 1885, 374, 528-530. Chaffee,A. L.; Hoover, D. S.;Johns, R. E.; Schweighardt. F. K. In 6io/Ogicei Markers h the Sedimentary Records: Johns, R. B.,Ed.; Elsevier: Amsterdam-Oxford-New York-Tokyo, 1986: pp 31 1-345. Varmuza. K.: Werther. W.: Lohninaer. H. In Software-Entwic&/unoin der Chemie 3;Gauglk, G.. Ed.; Sphger: Berlin-Heidelberg, 1989,pp 285-287
2043
(25) Werther, W.; Varmuza, K. In Software-Development in Chemistry 4 ; Gasteiger, G., Ed.; Springer: Berlin-Heidelberg, in press. (26) Windholz. M., Ed. The Merck Index, 10th ed.: Merck 8 Co.: Rahway. NJ, 1983; pp 1204, 4153, 5421. (27) Steiner, M. In Moderne Methoden der H/anzenana/yse; Paech, K., Tracey, M. V., Eds.; Springer: Berlin, 1955; Vol. 3, pp 128-135. (28) Varmuza, K. fanern Recognitlon in Chemistry; Springer: Berlin-Heideiberg, 1980. (29) Kvalheim, 0. A. Anel. Chim. Act8 1965, 777, 71-79.
RECEIVED for review January 26,1990.
Accepted May 31,1990. The authors wish to thank the Austrian “Fonds zur Forderung der wissenschaftlichen Forschung” for financial support (Project P 6811 C).
Enzyme-Linked Immunosorbent Assay Compared with Gas Chromatography/Mass Spectrometry for the Determination of Triazine Herbicides in Water E. M. Thurman,* Michael Meyer, Michael Pomes, Charles A. Perry, and A. Paul Schwab’ US.Geological Survey, Water Resources Division, 4821 Quail Crest Place, Lawrence, Kansas 66049
An enzyme-llnked lmmunosorbent assay (ELISA) was compared to a gas chromatography/mass spectrometry (GWMS) procedure for the analysis of triazine herbicldes and their metabolites in surface water and groundwater. Apparent recoveries from natural water and splked water by both methods were comparable at 0.2-2 pg/L. Soild-phase extraction (SPE) was examined also, and recoveries were determined for a suite of trlazine herbicides. A signfflcant correlatlon was obtained between the ELISA and GC/MS method for natural water samples that were extracted by SPE. Because ELISA was developed with an atrazine-like compound as the hapten with conJugationat the 2-position, It was selective for trlazines that contalned both ethyl and Isopropyl side chains. Concentrations for 50 % inhibition (IC,) were as follows: atrazine, 0.4 pg/L; ametryne, 0.45 pg/L; prometryn and propazlne, 0.5 pg/L; prometon, 0.7 pg/L; slmazlne and terbutryn, 2.5 pg/L; hydroxyatrazine,28 pg/L; deethylatrazlne and delsopropylatrazlne, 30 pg/L; cyanarlne, 40 pg/L; dkbalkylatrarine had no response. The comMnath of screening analysis by ELISA, which requires no sample preparatkn and works on 160 pL of sample, and conlltmation by GC/MS was designed for rapld, inexpensive analysis of triazlne herbicides In water.
INTRODUCTION The enzyme-linked immunosorbent assay (ELISA) has been shown to be a useful residue analysis method for herbicides (1-5). ELISA has been used extensively in clinical chemistry but has been introduced only recently into environmental chemistry on a commercial basis (2,3,5). The conventional analysis method for triazine herbicides (by gas-liquid chromatography with nitrogen-phosphorus detection, GC-NPD) Department of Agronomy, Kansas State University, Manhattan, KS 66506.
is sensitive and well characterized (6-9). Furthermore, both liquid chromatography (6,1&12) and gas chromatography/ mass spectrometry (13-18) have been applied successfully to the analysis of triazine herbicides. A major reason for the numerous methods of analysis of triazines, which are used extensively as preemergent herbicides for corn (Zea mays L.) and sorghum (Sorghum bicolor L.) is the fact that these herbicides are common contaminants in surface and groundwater (19). Thus, the triazine herbicides provide a challenging target for ELISA for rapid inexpensive analysis of surface and groundwater. This study examined the ELISA analysis method for triazine herbicides and their metabolites to determine cross reactivity, and to verify accuracy and precision against a reference GC/MS method. The study also checked for interference from naturally occurring humic substances and acetanilide herbicides that commonly occur in surface water. Gas chromatography/mass spectrometry was chosen as a reference method t o verify that the ELISA was responding to the triazines rather than other herbicides t h a t may be present. Because the ELISA method was designed to screen large numbers of water samples without sample preparation, an automated solid-phase extraction (SPE) method was designed also for rapid GC/MS confirmation. Objectives of the study were t o (1) compare ELISA t o GC/MS analysis for accuracy, precision, and cross reactivity for herbicides in natural water, (2) streamline the S P E extraction and GC/MS analysis by selected ion monitoring (SIM) for rapid confirmation, and (3) determine if the combination of ELISA screening and GC/MS confirmation for triazine herbicides would be a viable and inexpensive method for large water-quality surveys.
EXPERIMENTAL SECTION Reagents. Methanol (Burdick and Jackson, Muskegon, MI) and ethyl acetate (Fisher, Springfield, NJ) were pesticide-grade solvents. Ametryn, atrazine, prometon, prometryn, propazine, simazine, and terbutryn were obtained from Supelco (Bellefonte,
This article not subject to U S . Copyright. Published 1990 by the American Chemical Society
