M. V. LAYCOCK, A. W. MCCULLOCH, A. G. MCINNES, P. ODENSE, V. P. ...... a pressure of 0.2Torr (26 Pa), the spectra obtained were those of ..... ACS, Wash-.
Identification of domoic acid, a neuroexcitatory amino acid, in toxic mussels from eastern Prince Edward 1sland1 J. L. C. W R I G H TR. , ~ K. BOYD,A. S. W. DE FREITAS, M. FALK,R. A. FOXALL, W. D. JAMIESON, A. W. MCCULLOCH, A. G. MCINNES,P. ODENSE,V. P. PATHAK, M. A. QUILLIAM, M. V. LAYCOCK, M. A. RAGAN,P. G. SIM,P. THIBAULT, A N D J. A. WALTER Atlantic Research Laboratory, National Research Council of Canada, I41 1 0,uford Street, Halifux, N.S., Canada B3H 321
M. GILGAN Department of Fisheries and Oceans, Inspection Services Branch, P. 0 . Box 550, Halifax, N . S . , Canada B3J 2S7
D. J. A. RICHARD Department of Fisheries and Oceans, Inspection Services Branch, P.O. Box 270, Black's Harbour, N.B., Canada EOG IHO AND
D. DEWAR Canada Institute for Scientific and Technical Information, Natiot~alResearch Council of Canada, Ottawa, Ont., Canada KIA 0S2 Received August 15, 1988
J. L. C. WRIGHT,R. K. BOYD,A . S. W. DE FREITAS, M. FALK,R. A . FOXALL, W. D. JAMIESON, M. V. LAYCOCK, A. W. MCCULLOCH, A . G. MCINNES, P. ODENSE, V. P. PATHAK, M. A. QUILLIAM, M. A. RAGAN, P. G. SIM,P. THIBAULT, J. A. WALTER,M. GILGAN, D. J. A. RICHARD, and D. DEWAR. Can. J. Chem. 67,481 (1989). The causative agent of toxicity in cultured mussels from a localized area of eastern Prince Edward Island has been identified as domoic acid, a neuroexcitatory amino acid. The toxin was isolated by a number of different bioassay-directed separation techniques including high-performance liquid chromatography, high-voltage paper electrophoresis, and ion-exchange chromatography, and characterized by a number of spectroscopic techniques including ultraviolet, infrared, mass spectrometry, and nuclear magnetic resonance. The isolation and purification methods are described in detail and some new analytical data for domoic acid are reported. Key words: shellfish toxin, domoic acid, neurotoxin, bioassay-directed analysis.
J. L. C. WRIGHT, R. K. BOYD,A . S. W. DE FREITAS, M. FALK,R. A . FOXALL, W. D. JAMIESON, M. V. LAYCOCK, A . W. MCCULLOCH, A. G. MCINNES, P. ODENSE, V. P. PATHAK, M. A. QUILLIAM, M. A . RAGAN, P. G. SIM,P. THIBAULT, J. A. WALTER,M. GILGAN, D. J. A . RICHARD et D. DEWAR. Can. J. Chem. 67,481 (1989). On a identifie l'agent qui est la cause de la toxicit6 des moules d7Clevageprovenant d'une zone localisCe de la portion est de l'ile du prince Edouard; il s'agit de l'acide domoi'que, un acide amink neuro-excitant. On a is016 la toxine en faisant appel L un certain nombre de techniques de skparation basCes sur des essais biologiques parmi lesquelles on peut citer la chromatographie liquide B haute performance, 1'Clectrophorkse sur papier B voltage ClevC ainsi que la chromatographie d'Cchange ionique; on l'a caractCrisCe par un certain nombre de techniques spectroscopiques dont I'ultraviolette, l'infrarouge, la spectromCtrie de masse et la rtsonance magnCtique nuclkaire. On dCcrit en detail les mCthodes d'isolement et de purification et on prksente des donnCes analytiques nouvelles concernant I'acide domoi'que. Mots cle's : toxine des coquillages, acide domoi'que, neurotoxine, analyse baste sur des essais biologiques. [Traduit par la revue]
Introduction Shellfish poisons have been known since antiquity (1) and a recent review (2) lists the different families of marine toxins. From a Canadian perspective, the paralytic shellfish poisons (PSPs) have been of most significance to the shellfish industry (3). PSPs are products of the cold-water dinoflagellates, Gonyaulax spp., which are ingested by filter-feeding shellfish. Another group of marine toxins of potential significance to the Canadian shellfish industry are the diarrhetic shellfish poisons (DSPs). These toxins are produced by other cold-water dinoflagellates belonging to the genus Dinophysis, and although there is no unambiguous evidence that a DSP outbreak has occurred in Canada, there have been serious outbreaks in Japan and Europe (4). Towards the end of 1987, an outbreak of uoisoning in Canada was traced to cultured'blue mussels ( ~ ~ t i l e>ulis us L.) from a localized area of eastern Prince Edward Island (P.E.I.). During this period (November 11 to December 12) 153 cases 'NRCC No. 29840. 2 ~ u t h oto r whom correspondence may be addressed.
of acute intoxication related to ingestion of toxic mussels were documented (5), corresponding to about three cases per thou.~ included vornitsand portions of mussels c o n ~ u m e dSymptoms ing and diarrhea, which in some cases were followed by confusion, memory loss, disorientation, and even coma. Three elderly patients died. In the other most severely affected cases neurological symptoms still persist (5). The term Amnesic Shellfish Poison (ASP) has been proposed for this new shellfish toxin.4 Chemical analysis at ARL and other laboratories soon established that the toxicity was not due to any of the marine toxins , ~ to any of the known anthropothat are usually m ~ n i t o r e d nor genic xenobiotics (e.g., heavy metal^,^ pesticides, PCBs, and PAHs~).Significantly, extracts of contaminated mussels, when injected intraperitoneally into laboratory mice, produced a very characteristic and reproducible scratching syndrome and even3 ~Iverson, . J. Truelove, E. Nera, L. Tryphonas, J. Campbell, and E. Lok. Submitted for publication. communication. 4 ~ Hockin . (HWC. Ottawa). 5 ~ ~embel'la . ( ~ l %~, o n t - ~ o lPQ), i , personal communication. 6 ~W. . McLaren (NRC, Ottawa), personal communication. 7 ~Zitko . (DFO, St. Andrews, NB), personal communication.
482
CAN. J. CHEM. VOL. 67, 1989
tual death. In the interest of public health the entire East Coast shellfish industry was closed down. A task force was organized jointly by the Canadian Department of Fisheries and Oceans (DFO) and Health and Welfare Canada (HWC) to establish the extent of the contamination, the chemical nature of the toxin, and its origins. The Atlantic Research Laboratory (ARL) of the National Research Council of Canada, as part of the task force, was one of the laboratories that undertook the task of isolating and determining the chemical nature of the toxin. Within 5 days it was established at ARL that the mollusc toxin was domoic acid, a glutamate agonist (6) originally isolated some 30 years ago from the red alga Chondria armata Okamura (7, 8). This paper describes the methods used to isolate and identify domoic acid as the toxin in the contaminated mussels, and reports some new analytical and chemical data on this compound. The possible origin(s) of domoic acid in the 1987 outbreak are also discussed.
Domoic Acid
Results and discussion Toxic mussels and control (nontoxic) mussels were supplied by DFO (Charlottetown, P.E.I.). In all cases extracts of both toxic and control mussels were taken through identical procedures. This was done to distinguish the effects of the unknown toxic factor in the bioassay from inherent salt toxicity in some fractions, and from possible interference of other factors such as chromatographic mobile phases. Furthermore, this parallel approach would highlight additional or unusual signals present only in chromatograms and spectra of toxic mussel fractions. Strict control over the dosage used in the mouse bioassay was an important element of the procedure. A preliminary doseresponse curve indicated a narrow dynamic range, less than one order of magnitude, for the bioassay of the toxin(s). Extracts representing 0.3-2.5 g-equivalent of original wet tissue resulted in times of death (TOD) of 100-10min after intraperitoneal injection respectively. Volumetric control was used to afford an intraperitoneal dose in the range of 1.2-2.5 g-equivalents original tissue, independent of the actual amount of nontoxic material extracted. In addition it was important to verify that the symptoms induced in the bioassay by toxic fractions were exactly the same as those observed with extracts of toxic mussels. In initial experiments the digestive glands were separated from the remaining mussel tissue. Each tissue type was extracted with methanol :water (7 :3 v/v), concentrated, and then partitioned against dichloromethane. In no case was organic-soluble toxicity found. Aqueous extracts of both tissue types from control mussels displayed no toxicity but the digestive-gland extracts of toxic mussels were found to be toxic. The remaining tissue of toxic mussels displayed some low-level toxicity but this was attributed to incomplete separation of the two tissue types. However, in the interest of time saving during very-large-scale extractions, the entire mussel was processed. Later, when the toxin was firmly identified as domoic acid (I), other extraction procedures were studied. Quantitative extraction of 1 was achieved by boiling for 5 min, or by shaking for 24 h at room temperature, with either 50% aqueous methanol or water.8 Interestingly, the standard Association of Official Analytical Chemists (AOAC) method (5 min boiling, 0.1 M HC1) for extracting PSPs for mouse bioassay (9) was less efficient (75-85% recovery). Moreover, the concentration of 1 in the extract was found to decrease upon storage at room tempera-
ture (up to 50% loss in 1 week), although this loss could be reduced by storage at -15°C. The feasibility of using the AOAC method for simultaneous monitoring for PSP and domoic acid has been e ~ a m i n e d . ~ In the initial search for the toxin several methods of separation (high-performance liquid chromatography, hplc; high-voltage paper electrophoresis, hvpe; ion-exchange chromatography) were selected to avoid the possibility that in any one process the toxin might be obscured by other nontoxic compounds present in the extract. Furthermore, toxic fractions isolated by one scheme were cross-analyzed by another to verify that they were the same material, thus providing an internal check on the validity of the separation schemes. Following cleanup of the toxic aqueous extract on an XAD-2 column, the resulting toxic fractions were subjected to hvpe. This approach has the advantage of separating the fraction into acidic, basic, or neutral components. By slicing the chromatogram into sections, followed by extraction with distilled water, it was found that all the toxicity moved towards the anode at pH 6.5, indicating that the toxin was acidic. By repeating this experiment and by taking narrower slices it was eventually apparent that the toxic activity was associated with a band that stained yellow with ninhydrin and migrated on the trailing edge of the red-staining band for glutamic acid. Based on the premise that the toxic material was acidic and possibly a peptide, a reversed-phase hplc method was employed using a C18 stationary phase and gradient elution with wateracetonitrile mixtures containing 0.1 % trifluoracetic acid (TFA) to suppress ionization of the carboxyl groups. Detection was provided by a diode-array detector (dad), which also allowed continuous acquisition of uv-visible spectra. The hplc profile of the toxic hvpe fraction obtained from the XAD-2 column, compared with the equivalent fraction from control mussels, is shown in Fig. 1. Initial runs were monitored by uv absorption at 210 nm since most peptides and organic acids absorb at this wavelength. The chromatogram for the toxic fraction showed a unique peak at retention time 12.1 min (peak D). The equivalent region in the control chromatogram showed no peak. Significantly, this same peak was consistently observed in other extracts of toxic mussels but could not be seen in any extracts
'M. A . Quilliam, l?. G. Sim, A. W. McCulloch, and A. G. McInnes. Submitted for publication.
'J. F. Lawrence, C. F. Charbonneau, C. Menard, M. A. Quilliam, and P. G. Sim Submitted for publication.
I
WRIGHT ET AL.
0 600-
30
20
10
40
:b
4 002000 0
7
.
.
.
,
.
.
.
10
.
,
.
.
20
.
-
r
.
30
-
.
.
1
40
Time (min) FIG. 1. Reversed-phase hplc chromatograms of XAD-2 cleaned extracts of (a) toxic and (b) control mussels. The inset (c) is the uv-visible spectrum acquired at the apex of peak D (domoic acid). Peak T is due to the amino acid tryptophan. Conditions: 25 cm X 2.1 mm i.d. 5-p,m Vydac 201TP52 column at 40°C; 0.2 mLmin-' flow rate; gradient elution from TFA/H20 (0.1:99.9) to CH3CN/TFA/H20 (50:0.1:49.9) in 25 min and then to CH3CN/TFA (99.9:O. 1) at 35 min and hold to 50 min; detection provided by absorption at 210nm (20-nm bandwidth).
of control mussels. The uv-visible spectrum acquired at the apex of this peak revealed a single absorption maximum at 242 nm, suggesting the molecule contained two conjugated double bonds. By employing the same chromatographic conditions on a preparative hplc scale it was possible to collect fractions corresponding to peak D as well as other regions of the chromatogram for bioassay and chemical analysis. However, all the fractions obtained in the first series of preparative runs displayed toxic effects both in the control and toxic samples, although the symptoms were different from those induced by toxic mussel extracts. This problem of general toxicity was resolved by replacing the rodenticide TFA used in the mobile phase with 1% acetic acid. The peak D fraction was then found to be toxic to mice, with the correct symptoms. All the other fractions from aqueous methanol extracts were nontoxic, even at a fivefold higher dose in terms of weight equivalents of original tissue. When the toxic hplc fraction was subjected to hvpe, a yellow band was observed at a position corresponding to where toxicity was previously observed. Bioassays of aqueous extracts of slices of this electrophoretogram revealed that only the yellow band was toxic. Conversely, this band obtained by preparative hvpe of crude extracts gave an hplc peak at the same retention time as peak D, and the same uv spectrum. The fact that such different fractionation schemes yielded the same toxic factor was very important to the certainty of the result. Early on it was found the acidic toxin could also be eluted from a strongly basic ion-exchange resin (Dowex 1) using ammonia, and analysis of mussel extracts with a conventional
amino acid analyser revealed an unusual peak present only in toxic samples. The compound eluted between methionine and isoleucine; the ninhydrin-reaction product absorbed strongly at 440nm, consistent with a proline derivative. The acidic toxic factor could also be conveniently separated from a crude fraction using a weak anion-exchange resin. This proved to be a useful and facile method for purifying the toxin, and was also used to remove the toxin from the initial extract to assess the toxicity of the remaining residue. Concentrated residues from both toxic and control extracts obtained in this manner demonstrated some additional toxicity on mouse bioassay. However, the observed symptoms of general weakening and occasional cyanosis prior to death were quite different from those neurological symptoms induced by toxic mussel extracts. This additional toxic factor(s) was found to pass through a 500-Da ultrafilter membrane (Amicon YC-05) but once again the toxicity was the same for both control and toxic extracts. These symptoms of weakening, occasional cyanosis, and death have been observed with high salt concentrations; interestingly, similar symptoms have been observed in mice given intraperitoneal injections containing high ( 2 5 0 0 ppm) concentrations of zinc (10). The ir spectra of the early toxic fractions from the preparative hplc runs were highly variable, due to dependence on the pH at which the films were deposited. Also, the spectrum of the toxin was partly masked by absorption bands due to acetate or trifluoracetate salts, contaminants originating from either acetic or trifluoroacetic acid used in the chromatographic separations. Both of these complications were removed by deposit-
CAN. J. CHEM. VOL. 67, 1989
FIG. 2.
Spectra of amorphous domoic acid films on a CaF2 plate at increasing degrees of protonation from A to D.
ing the films from a solution of 0.1 M HCl. Acidification ensured a uniform and essentially complete protonation of the toxin molecules (shown by the absence of further spectral change upon evaporation from additional quantities of HC1). This procedure also replaced the acetate or trifluoracetate salts by the corresponding chlorides, which did not interfere with the spectrum, plus free acetic or trifluoroacetic acid, which rapidly evaporated during the preparation of the films. The spectrum of the toxin, when freed of such contaminants, contained absorption bands characteristic of non-ionized COOH (1215 and 1715 cm-', and a broad band with maxima at 3000, 2750, and 2600 cm-'), ionized COO- (1400 and 1612 cm-'), NH2+ (1565-1585 and 1950 cm-'), trans -CH=CH(966 cm-'), olefinic CH groups (3030 cm-I), and aliphatic CH groups (2980, 2935, and 2880cm-'). Figure 2 shows a typical series of spectra obtained at pH varying from about 5.0 in Fig. 2A to about 1.0 in Fig. 2D, the molecule being fully protonated at the latter pH. Further details of the ir spectra and of vibrational assignments are discussed elsewhere ( 1 I). Toxic and control fractions were continually screened by
mass spectrometry (ms). In view of the water solubility of the toxin, ionization by fast atom bombardment (FAB) using a glycerol: water matrix (2: 1 v/v) was selected. Since some early samples were heavily contaminated by chloride ions, the most useful spectra were initially obtained in the positive-ion mode where sodium adduct ions were the principal but less complicating problem. As the fractionation procedures succeeded in progressively concentrating the toxicity, the corresponding positive-ion mass spectra showed the increasing importance of an ion at m/z 312, which was absent from spectra of control samples. This observation, together with an ion at m/z 310 in the negative-ion spectra, suggested a molecular weight of 3 11Da for the toxin. Daughter-ion spectra of the positive FAB ions at m/z 312 displayed fragment ions at mlz294, 266, 248, 220, and 74 (Fig. 3). An important feature of the daughter-ion spectra was the series of losses of 46 Da (presumably HCOOH) characteristic of protonated carboxylic acids including amino acids (12, 13). Because the even-electron ion at m/z 312 must contain an odd number of nitrogens, it appeared likely that this ion
WRIGHT ET AL.
FIG.3. Fragment-ion mass spectrum, for collision-induced dissociation in rf-only quadrupole, of MH' ions (m/z 312) formed by FAB ionization of the toxin in a glycerol/water matrix. corresponded to a protonated molecule of a nitrogen-containing compound, perhaps an amino acid containing at least two carboxyl groups. This conclusion was substantiated by the successive losses of 44 Da (C02) from the [M - HI- anion. I' Finally, in the positive-ion spectra, the intense signal at mlz74 is characteristic of protonated a-amino acids and is ascribed (13) to the ion H2N=CH-COOH+. An alternative interpretation in terms of the odd-electron ion C3H602+' characteristic (14) is unlikely to be valid of the substructure HOOC-CH(CH3)for fragmentations of the even-electron species [M HIf studied here. Accurate mass measurement of the signal at mlz 3 12 gave a value of 3 12.146 +- 0.003 Da. Eleven molecular formulae are consistent with this value. However, by applying the additional restriction of an odd number of N atoms plus at least four oxygen atoms (at least two carboxyl groups, vide infra) the possibilities are reduced to just two, Cl5HZ2NO6and CI2Hz6N06S, for the protonated molecule. In principle it should be possible to distinguish between these possibilities based on the relative intensities of the m/z 312, 313, and 314 signals. The intensity of the signal at mlz 313 invariably fell in the range 16- 19% of that of the m/z 3 12 peak, favouring the Cl5HZ2No6 formula, but the intensity of the m/z314 signal was highly variable due to beam-induced interaction of the analyte with the matrix." Further support came from analysis by electron ionization (EI) of the tert-butyldimethylsilyl (TBDMS) derivative. Rapid ( 5 5 min) esterification of the three carboxyl groups gave a product whose spectrum displayed ions at mlz653.395 +- 0.002 ([MI+'), 596.327 + 0.002 ([M - tBu]+), and 494.314 + 0.002 ([M - tBuMe2SiO2CIt). After several hours reaction with the silylation reagent, intense signals were observed corresponding to a product in which the amine group was also derivatized: mlz 767.483 +- 0.002 ([MI"), 7 10.414 + 0.002 ([M - tBu]+), and 608.401 +- 0.002 ([M - tBuMe2Si02C]+).
+
'P. Thibault, M. A. Quilliam, W. D. Jamieson, and R. K. Boyd. Submitted for publication.
These data are consistent with a tricarboxylic amino acid of molecular formula C15H2]No6. The first useful 'H nmr spectra of toxic fractions from preparative hplc were obtained with solutions in dry DMSO-d6, this solvent having been chosen to avoid hydrogen exchange. Earlier 'H nmr experiments using THF-d8 had yielded broad spectra although, in retrospect, the domoic acid resonances are identifiable. The spectra in DMSO-d6 showed the compound to be largely free of other organic impurities and enabled resonances characteristic of a vinylic methyl group (1.81 ppm, broad singlet) and an allylic methyl group (1.27 ppm, doublet) to be identified. The connectivity of the olefinic protons, H-5', and the allylic methyl group was determined from their coupling patterns and a COSY experiment, establishing the partial structure CH3-CH-CH=CH-CH=C-CH3. Throughout the structure analysis, the uv, ir, ms, and nmr data were continually updated and used to search the Chemical Abstracts and its accompanying Registry File on the STN International on-line database (16). The search, in which the partial structural data, toxicological data, and the marine origin of the toxin were used together, led at first to a long list of compounds. However, as the structural information became progressively more detailed this list shrank rapidly until the neuroexcitatory amino acid domoic acid (1) and its isomers emerged as likely candidates. Meanwhile, 'H nmr spectra of the toxin were found to sharpen when D 2 0 was added to the DMSO-d6 solution, and in neat D 2 0 at pH 3.1 a 'H spectrum was obtained that agreed in every respect (chemical shifts to 0.02ppm, spin-spin couplings to 0.2Hz) with detailed published data for domoic acid (15, 17), taking into account that the published data were obtained at an unspecified pH. It also became apparent that the ir and ms data for the toxin were consistent with the domoic acid structure, and further spectral data were collected that completed the final identification. Further 'H and 13cnmr work (Table 1) was undertaken to confirm assignments and structural features of the toxin in D 2 0
TABLE1. Nuclear magnetic resonance data for domoic acid 2
3
4
6('H), p p m U v Multiplicityc
3.98 d
3.05 dddd
3.84 ddd
% n O e produced when H irradiatede
None
H-4, 4 . 7
6(I3c), ppma.f Multiplicity
67.1 bd
44.6 bd
Position
5P
5a
6a
6b
7
2-C02H
k
Position
1'
2'
3'
4'
5'
6'
% n O e produced when H irradiatede 6(13C), ppm",f Multiplicity
'Reference to TSP (sodium 3-trimethylsilylpropionate-2,2,3,3-D,) solution in D 2 0 contained in a concentric tube. Domoic acid solution at pH 3.40 t- 0.05, temperature, 20.0°C. bError ca. k0.02ppm. 's = singlet, d = doublet, t = triplet, q = quartet, m = multiplet, b = broad. dError ca. k0.2Hz. Spin-spin coupling connectivity confirmed by an 'H COSY spectrum. Some couplings determined by simulation. eError ca. t-0.5% for single protons, ca. *0.15% for methyl groups. Samples were not degassed or otherwise specially prepared for nOe measurements. fI3C assignments determined from I3C/'H heterocorrelation experiment and 'H-coupled I3Cspectrum. *The nOe (if any) between H-4 and H-5P was not measurable in these experiments due to the closeness of the resonances.
1 '-CH3
S1-Co2H
2
487
WRIGHT ET AL.
Domoic Acid Dose (pg)
FIG.4. Toxicity (expressed as reciprocal of time of death (TOD) of mouse) as a function of the amount of domoic acid injected into the mouse (as determined by hplc analysis). The toxicities of room temperature aqueous methanol (A) and boiling 0.1 M H C l ( 0 ) extracts of mussel tissue are similar to that of aqueous solutions of pure domoic acid (a).Each point is the mean of 2-4 replicate measurements of TOD for one solution, and the error bars represent an estimate of the standard deviation. The curve is a second-order polynomial leastsquares regression on all data points. solutions at pH 3.4. 'H COSY spectra and spectral simulation showed that the published 'H assignments (15, 17) were correct, as were 'H-'H spin-spin couplings apart from two minor differences probably resulting from our use of simulation to extract coupling constants. Homonuclear 'H nuclear Overhauser enhancement (nOe) experiments with our sample of domoic acid (Table 1) confirmed the Z, E configuration of the conjugated double bond system in the side chain, as well as the cis relationship of H-3 and H-4 in the prolyl ring. The only nOe's between this side chain and the ring were those between the C-1' methyl group and H-5a and H-6b, and between H-3' and H-4. These are consistent with one orientation of the side chain, but rotation about the C-4-C-1' bond is not necessarily precluded by the indetectability of nOe's that might be expected if rotation occurred. However, nOe between H-5' and both H-3' and H-4' shows that rotation is possible about the C-4'-C-5' bond. Only sketchy details on the assignment of the 13C nmr data of 1 have been published (7, 18) and a more complete account of the data and some reassignment is necessary. The 13C nrnr spectrum displayed 15 resonances composed of three carbonyl, four olefinic, and eight saturated carbons. All the carbons bearing hydrogen could be unequivocally assigned by a 2-D heterocorrelation experiment using the 'H nrnr shifts and coupling values previously assigned (15, 17) and reconfirmed in this work. Of the four remaining quaternary carbons the resonance at 133.8 ppm was assigned to C- 1'. The high-resolution 13c nmr data were used to distinguish the three remaining carboxyl resonances. The signal at 177.5 ppm ascribed to C-7 appeared as a doublet of triplets due to coupling with H-3 ( J = 3.0 Hz) and with H-6a and H-6b ( J = 7.2 Hz). The resonance for the C-2 carboxyl appeared as a doublet of doublets at 174.9 ppm (coupling to H-2 and H-3; J = 4.1 Hz) and the resonance for the C-5' carboxyl appeared as a complex multiplet centred at 181.9 ppm (coupling to H-6', H-5', H-4'). Finally, to determine if domoic acid accounted for all the observed toxicity, quantitative mouse and hplc assays were performed on a variety of extracts and on pure domoic acid
solutions. The data, presented in Fig. 4 in the form of a doseresponse curve, showed that no significant toxicity other than that due to domoic acid occurred in either the aqueous methanol or boiling acid (0.1 M HCl) extracts. The intraperitoneal dose required for a 30-min death time in the mouse bioassay was about 10 mg kg-' body weight (200 pg per average mouse). Pure domoic acid in aqueous solution produced the same symptoms in the mouse bioassay as did toxic-mussel extracts. The hplc assays have identified concentrations of domoic acid in toxic shellfish samples as great as 900 pg g-' fresh tissue, although the levels varied greatly between batches and averaged about 300 pg g-'. This is the first report of domoic acid (1) in shellfish and the first known occurrence of human intoxication from ingesting the neurotoxin. Domoic acid was originally isolated from the warm-water red macro alga Chondria armata over 30 years ago (7, 8), and subsequently several isomers and derivatives of 1 have been isolated from the same algal source (17, 19). There are early reports (20) of the use of domoic acid as an anthelmintic for young children in Japan; however, the doses used were at the 0.6 mg kg- level (20), an order of magnitude lower than those encountered by adult Canadians experiencing toxic effects from eating contaminated mussel^.^ Over the years there has been considerable interest in this group of metabolites because of their insecticidal properties (21) and because domoic acid (1) and the structurally related compound kainic acid (2) are potent members of a group of neurotoxic amino acids (22, 23). Biochemical, pharmacological, and physiological studies have shown that these excitatory amino acids act on three types of receptors in the brain, the hippocampus being the most sensitive structure in the central nervous system (CNS). These studies have been extensively reviewed (24-27). Very recently, the first electrophysiological evidence has been obtained that domoic acid activates kainate receptors and is more potent than kainic acid itself. " Furthermore, these electrophysiological results are in keeping with the clinical reports of patients hospitalized after ingesting the mussel toxin, and two post-mortem anatomopathological examinations are reported to reveal extensive damage of the hippocampus as well as spotty damage of thalamic and forebrain regions. Since domoic acid was not detected in mussels from elsewhere in P.E.I., it is unlikely that it is a metabolic product of the mussel itself. Moreover, toxic shellfish held in tanks of clean water for several weeks steadily purged themselves of domoic acid,12 a phenomenon often associated with dietary products of shellfish (2). As efficient filter-feeding bivalves, mussels process huge volumes of water relative to their body size (28) and at least some of the toxin in their tissues could be derived from low (undetectable) levels of domoic acid in sea water. However, the food vector is probably much more important than the water vector to account for the substantial quantities of domoic acid found in contaminated mussels. It is estimated that the total amount of domoic acid in mussels from the affected area was probably in excess of 6 kg, with essentially all of the toxin body burden associated with digestivegland tissue. The major food items for cultured mussels are phytoplankton and other particulate detrital organic material such as fragmented seaweed tissue from benthic macrophytes and particulate organic material of terrestrial origin camed into the marine
-
'
''
" G . Dobennel, L. Beauchesne, and C. de Montigny. Submitted for publication.
488
CAN. J. CHEM. \'OL. 67, 1989
environment bv surface-water runoff. The available evidence on distribution-and abundance of benthic macrophytes known to produce domoic acid, such as Alsidium corallinum C. Ag. (29), Chondria armata (6, 7), and its congener C . baileyana (Mont.) Harv. (unpublished result), effectively rule out these species as sources of the toxin in eastern P.E.I. Based on other shellfish toxin results, a more likely source of mussel contamination is the microalgal component of the plankton ingested by filter feeders. Plankton tows made in early December 1987 by L. A. Hanic (University of Prince Edward Island) revealed an intense diatom bloom in the affected area. This diatom was subsequently identified as Nitzschia pungens Grun. f. multiseries Hasle. l 3 When injected in mice, extracts of the plankton-tow samples produced symptoms identical to those produced by extracts from toxic mussels. As well, all toxic mussels harvested in December 1987 showed engorged digestive glands containing identifiable shell remnants of N . pungens; this was also true of toxic mussels harvested in late October and November, although no information is available on the plankton composition at that time. We observed domoic acid in plankton from the affected areas at levels sometimes in excess of 1% dry weight (30), sufficient to account for the levels observed in toxic mussels. This strong evidence that N . pungens was the major source of domoic acid in the toxic mussels is further supported by the demonstration of de novo production of domoic acid by laboratory cultures of N . pungens isolated from the affected area (34). Additional results from the St. Andrews Biological Station14 and from the University of Rhode Island1' document the very high probability that other phytoplankton such as Amphora coffaeformis (Ag.) Kiitz. are also primary producers of domoic acid. In view of the importance of this new shellfish toxin to the Atlantic regional fisheries and the possibility that it may be more widespread, it is essential to determine what other marine organisms can synthesize domoic acid.
Experimental All solvents were hplc grade from Anachemia. Distilled water was further purified by passage through a Millipore water-purification system (Milli-QR) equipped with ion-exchange and carbon filters. Trifluoroacetic acid (TFA) was purchased from BDH. Reference samples of domoic acid were obtained from Cambridge Research Biochemicals, England (Lot 12026), and Regis Chemical Company, Illinois, U.S.A. (Lot 2555). Melting points are corrected and were observed with a Kofler-type hot-stage apparatus. Optical rotations were measured using a Perkin Elmer model 141 polarimeter. Infrared spectroscopy Infrared spectra were recorded by microspectroscopic techniques using a BOMEM FT-IR spectrometer (model DA3.02) equipped with a "Spectroscope" microscope attachment made by Spectra-Tech Inc. The spectrometer was used with a KBr beamsplitter and a mercury cadmium telluride detector. Spectra in the region 4000-700 cm-' were recorded at a resolution of 2 cm-' with 400 interferograms accumulated for each spectrum. A microgram quantity of domoic acid, dissolved in water and with the pH adjusted if necessary with dilute hydrochloric acid, was evaporated on a small AgCl or CaF2 plate to yield a thin solid film approximately 3 mm in diameter and a few microns in thickness. Small circular areas of 0.1 mm diameter were 1 2 ~Scarratt . (DFO, Halifax), personal communication. 13C. J. Bird (ARL) and M. Poulin (National Museum of Canada, Ottawa), personal communication. 14J. Martin (St. Andrews Biological Station, DFO), personal communication. 15Y. Shirnizu (University of Rhode Island), personal communication.
visually selected, optically isolated by the microscope aperture, and their ir spectra recorded. The spectrum of an equivalent area of the empty plate was recorded as reference: the ratio of such spectral pairs yielded the transmittance spectrum, from which the absorbance spectrum was derived. Since the sample space was normally evacuated to a pressure of 0.2Torr (26 Pa), the spectra obtained were those of anhydrous material. Nuclear magnetic resonance spectroscopy All nrnr spectra were obtained with a Bruker MSL-300 spectrometer at room temperature (20°C), in 5-mm tubes, under the following conditions. 'H spectra: 300.13 MHz, one-pulse sequence, spectral width (SW) 4505 Hz or 2403 Hz, acquisition time (AQ) 1.82s or 3.41 s, 16 384 data points, delay between 45-deg rf pulses 8.6 s, data processed where necessary with exponential multiplication and zerofilling. Integrated intensities of signals were compared with those of a standard sucrose solution in D 2 0 using identical experimental conditions, allowing complete s in lattice relaxation for quantitative estimation of concentrations. H -COSY spectra (31): SW 2401 Hz and 90-deg pulses. I3C spectra: SW 16 667 Hz, AQ 0.983 s or 0.492 s, 30-deg pulses, and broadband 'H-decoupling. I3C coupled spectra used SW 135 14 Hz and AQ 2.43 s. Nuclear Overhauser enhancements were obtained by irradiating selected resonances for 10 s between acquisitions at a power level just sufficient to cause saturation (ca. 180 mV peak-peak into a 50-ohm load via a directional coupler), and subtracting these spectra from one obtained under the same conditions but with irradiation well removed from any resonances. D 2 0 was used as a solvent and spectra were referred to TSP in a concentric tube. The concentration of domoic acid in the original preparation was estimated from the 'H resonance areas to be 2.0(?0.1) mg in 0.7 mL D20, which was close to saturation at room temperature. A supersaturated solution of a later preparation of anhydrous domoic acid from mussels (9.4 mg in 0.6 mL D20) was used for I 3 C - ' ~correlation spectra. The 'H and I3cspectra reported in Table 1 were obtained from a saturated solution of the latter preparation at pH 3.4.
P
Mass spectrometry All mass spectrometric measurements reported herein were conducted using a VG Analytical ZAB-EQ tandem hybrid mass spectrometer, or a VG Masslab 20-250 quadrupole mass spectrometer, each equipped with a VG11-2505 datasystem. The FAB mass spectra were obtained using the ZAB-EQ instrument. Sample sizes in the range 10-100 nmol, in 1-10 yL of aqueous solution, were deposited on a gold probe tip, evaporated down to less than 1 yL, then mixed with about twice as much glycerol. Although this simple glycerol/ water matrix kept the background to a minimum, the matrix was acidified with HCl when working in positive-ion mode. An 8-keV x e O beam (Ion Tech FAB gun, 1-mA discharge current) was used. Conventional FAB mass spectra were obtained at a resolving power (10% valley) of 1500, scanning the range m/z 1300-50 Da with a cycle time of approximately 10 s. Mass assignments were reliable to within 0.1 Da. Masses of ions formed by FAB ionization were measured to better than 10 ppm by peak matching against either the matrix cluster ion at m/z 277 (3 glycerol molecules H) or, in a separate determination for a sample doped with sodium iodide, the Na312+cluster at m/z 323; resolving power was 8000 (10% valley definition). Electron ionization (EI) spectra of the tert-butyldimethylsilyl derivatives formed by reaction of domoic acid in dimethylformamide with N-methyl-N-(tertbutyldimethylsi1yl)trifluoracetamide (MTBSTFA, Regis Chemicals) were obtained using 70-eV nominal energy electrons. Accurate mass measurements of these EI ions were made by peak matching against perfluorokerosene at a resolving power of 8000. Tandem mass spectrometry experiments were done to obtain fragment-ion spectra of the ions at m/z 312 in the positive-ion FAB spectra. For these, the precursor ions were selected by the double-focusing BE stages of the instrument, decelerated, then focused into the rf-only quadrupole collision cell (q) containing argon at a pressure such that the intensity of the precursor ion signal was attenuated by 50-60% while the tandem quadrupole mass analyser (Q) was scanned. A collision energy (laboratory frame) of 30 eV was used. All tandem mass spectrometry data
+
489
WRIGHT ET AL.
were acquired in continuum mode, each spectrum representing the sum of at least 10 scans.
High voltage paper electrophoresis (hvpe) Concentrated aqueous-methanol extract (approximately 1 mL) was streaked across the middle of a sheet of Whatman 3MM chromatography paper and air-dried. The paper on either side of the sample zone was wetted with pyridine-acetate buffer at pH 6.5 (pyridine:acetic acid:water, 25:1:225 by volume) and electrophoresis was carried out at 3 kV for 40min in a Michl tank (32) with toluene as the heat exchanger. The paper was dried and vertical strips (1 cm wide) cut off and stained with cadmium acetate - ninhydrin (33). The remainder of the electrophoretogram was cut into horizontal zones, which were eluted with water and taken to dryness in vacuo. Amino acid analysis Dilute aqueous extracts (0.1 mL) were loaded on a Beckman 119CL automatic amino acid analyzer. The cation-exchange column (sulfonated polystyrene, 30 pm diameter) was equilibrated with 0.3 M sodium citrate buffer, pH 3.25, and eluted with 0.3 M sodium citrate buffers of pH 3.95 and pH 6.40. The dual-wavelength detector recorded absorbance~of ninhydrin reaction products at 570 and 440 nm. High performance liquid chromatography (hplc)and ultraviolet (uv) spectroscopy Domoic acid determinations were performed on a Hewlett-Packard model 1090M hplc equipped with a ternary DR5 solvent delivery system, variable volume (1-25 pL) injector and autosampler, built-in HP1040 diode array detector (DAD), and HP79994 data system. A 25 cm x 2.1 mm i.d. 5-pm Vydac 201TP52 column (Separations Group, Hesperia, CA) was operated at 40°C with 0.5 mL min-I 10% (v/v) acetonitrile in water with 0.1% (v/v) trifluoroacetic acid and 5-pL injection volumes. Detection was accomplished by monitoring absorption at both 210 and 242 nm with 10-nm bandwidths. Ultraviolet acquisition was either in the peak-triggered mode or in the continuous mode with a 640-ms scan interval. Prior to analysis, aqueous extracts (1-2 mL; pH 6-7) were applied to a prerinsed C I Bsolid-phase extraction cartridge (Supelco, Bellfonte, PA), eluted with 10% aqueous acetonitrile (3 mL) into a 5-mL volumetric, made up to volume, and filtered through a dry 0.22-pm filter. Calibration was accomplished using a solution of pure domoic acid isolated by heart-cutting the domoic acid peak from preparative hplc, and calculating domoic acid concentration from the literature molar absorptivity (7) at 242 nm (2.63 X lo4) and the observed optical density at 242 nm. The molar absorptivity of purified domoic acid was later redetermined in our laboratory and found to be pH-dependent. The value is (2.43 2 0.04) x lo4 at pH 2.0 and (2.61 -' 0.04) x lo4 at pH7.0. Isolation of domoic acid (i) By ion-exchange chromatography Digestive glands (572 g) were excised from frozen, cooked, mussel tissue (1826g) (processed by North Ocean Enterprises Ltd., Souris, P.E.I.) and homogenized with a mixture of distilled water (450 mL) and methanol (650 mL). The homogenate was stirred and heated at 65°C for 10 min. The cooled mixture was centrifuged (3500 rpm for 10 min) and the supernatant decanted. The residue was washed twice with 50% aqueous methanol (1 X 600 mL; 1 x 300 mL). The combined supernatants were concentrated in vacuo to give a viscous goldenbrown oily residue. This was taken up in water (200 mL); an aliquot (10 mL) was reevaporated to give a residue (1.62 g), shown by hplc to contain 35.8 mg of domoic acid. The residue (30.8 g), obtained by evaporating the remaining solvent (190 d ) , was triturated with methanol. This afforded an insoluble cream-coloured solid (1 1.13 g) and a methanol-soluble fraction (17.91 g). The solid was dissolved in water (325 mL), the pH adjusted to 5.0, and the solution centrifuged. The supernatant was placed on a BioRad AG 3-X4 column (2.5 X 25 cm; 107 mL resin; OH- form). After washing with water (700 mL) and 2 M acetic acid (200 mL), fractions (400 mL) containing domoic acid were eluted from the column with 4 M acetic acid. Evaporation gave residue A (69 1 mg).
The methanol-soluble fraction above was dissolved in water (200 mL), extracted with dichloromethane (3 X 200 mL), placed on a Bio-Rad Ag3-X4 column, and treated as described above. The domoic acid fraction (350 rnL) was evaporated in vacuo to give residue B (889 mg). Residues A and B were combined and crystallized from water to give domoic acid (208 mg), mp 200-205°C (subsequent crops gave an additional 273 mg of less pure material). Successive recrystallizations from water gave pure domoic acid dihydrate (180 mg), mp 215-216°C (lit. (8) mp 217°C). This melting point refers to loss of polarizing properties, and deformation, of the dihydrate crystal examined as soon as possible after preparation. Such crystals do not exhibit this well-defined behaviour on standing. -120.5" for anhydrous material; reported value (29) for dihydrate, -10so, equivalent to - 120.5" for the anhydrous compound. Anal. calcd. for CI5H,,NO6: C 57.87, H 6.80, N 4.50%; found (after drying at 50°C for 18 h): C 57.54, H 6.98, N 4.16%. (ii) By preparative hplc Shucked frozen mussels (= 800 g wet wt) were homogenized with methanol :water (7 :3 v/v ; 3 L) and allowed to stand at room temperature for 24 h with occasional stirring. The sluny was filtered through glass-fibre paper and the filtrate concentrated in vacuo. The concentrate was partitioned twice against dichloromethane to yield an organicsoluble extract (12.4 g) and a water-soluble extract (=65 g). Biological testing showed the toxicity to be present in the aqueous fraction. A portion of this aqueous extract (4 g) was applied to a column (40 x 3 cm) containing XAD-2 (Amberlite) resin. Fractions (200 mL) were collected during elution with a stepwise gradient from water through methanol. The toxicity was contained in the first two fractions. These were combined, concentrated in vacuo, and applied to a preparative reversed-phase column (Supelcosil LC-18 5 pm; 25 cm X 21.2 mm) and eluted with a linear gradient of 10% acetonitrile-water through 35% acetonitrile-water (1% acetic acid; flow rate 12 mLmin-I). The eluant was monitored at 242nm and domoic acid (retention time 10.5 min) was collected following repeated injections. After concentration, the resulting yellowish oil was dissolved in the minimum volume of hot water and, upon cooling, straw-coloured crystals of domoic acid formed. ~epeatedcrystallization yielded colourless needles, mp 212-214"C, identical (uv, ir, nmr) with authentic material.
Bioassays The biological assays were performed using female mice (CD-I, Charles River Canada Inc.) weighing 16-23 g. Each was injected intraperitoneally with an extract (1 mL) and observed for up to 3 h. At least two mice were used for each test. The symptoms varied depending on the dose of domoic acid. Mice with death times of less than 13 min exhibited circling movements, occasional barrel rotation, and convulsions. Mice with longer death times exhibited loss of coordination, scratching of shoulder area with back legs alternating from one leg to the other at 3 to 20-s intervals, occasional aggressiveness, and gnawing or biting.
Acknowledgements We acknowledge the efforts and cooperation of our D F O colleagues, particularly R . Bourque, J. Worms, and L . Lea (Gulf Region) and R . F. Addison and J. E. Stewart (ScotiaFundy Region), and those of our H W C colleagues including H . B. S. Conacher, S. W. Gunner, J. F. Lawrence, and E. Todd. We thank L . A. Hanic, University of P.E.I., for phytoplankton samples, often collected under extreme weather conditions. W e would also like to thank our N R C colleagues, J. W. McLaren (Ottawa) for heavy metal analysis, C. J. Bird and J. L. McLachlan (Halifax) for biological examination of algal and mussel samples, A. M . Backrnan and A. R. Taylor, CISTIARL, for library services, and F. E . Isaacs, Public Relations and Information Services, for a magnificent job under enormous pressure. It is a particular pleasure to acknowledge the extraordinary efforts of the A R L staff throughout this work.
490
CAN. 1. CHEM. VOL. 67, 1989
They include the following technical staff: C . A . Craft, E. W. Dyer, D. J. Embree, M . G. Flack, C . Gillis, M . Greenwell, W. R. Hardstaff, P. LeBlanc, N . I. Lewis, G . K. McCully, M . McInerney-Northcott, D. O'Neil, P. F. Seto, and D. F. Tappen, and the entire maintenance and administrative support staff, in particular K. V. Gray, W . J. Gray, H . 0 . Henderson, K. D . MacLeod, D. R . Robson, G. Saab, C . T. Vaughan, H. R. Watts, and B. Willis. Finally, w e thank Dr. Y. Ohfune, Suntory Institute for Bioorganic Research, Osaka, for the generous gift of a quantity of (-)-domoic acid obtained by total synthesis (ref. 15). 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.
11. 12.
J. Am. Chem. Soc. 98, 1301 13. C. W. TSANGand A. G. HARRISON. (1976). and R. VENKATURAGHAVAN. Mass spectral 14. F. W. MCLAFFERTY correlations. Advances in Chemistry. Ser. No. 40. ACS, Washington, D.C. 1982. p. 40. 15. Y. OHFUNEand M. TOMITA.J. Am. Chem. Soc. 104, 3511 (1982). 16. Anonymous. CISTI News, 5(4), 1 (1988). 17. M. MAEDA,T. KODAMA, T. TANAKA, H. YOSHIZUMI, T. TAKEMOTO,K. NOMOTO,and T. FUJITA.Chem. Pharm. Bull. 34, 4892 ( 1986). 18. M. MAEDA, T. KODAMA, T. TANAKA, H. YOSHIZUMI, K. NOMOTO, T. TAKEMOTO, and T. FUJITA.Tennen Yuki Kagobutsu Toronkai Koen Yoshishu, 27th, 616 (1985). 19. M. MAEDA,T. KODAMA, T. TANAKA, H. YOSHIZUMI, T. TAKEB. W. HALSTEAD. Poisonous and venomous marine animals of MOTO,K. NOMOTO,and T. FUJITA.Tetrahedron Lett. 28, 633 the world. Vol. 1. U.S. Government Printing Office, Washing(1987). ton, D.C. 1965. E. P. RAGELIS.Seafood toxins. ACS Symposium Series 262. 20. K. DAIGO.J. Jpn. Pharm. Assoc. 79, 353 (1959). T. TANAKA, Y. OHFUNE,K. NOMOTO, 21. M. MAEDA,T. KODAMA, Washington, D.C. 1984. K. NISHIMURA, and T. FUJITA.J. Pestic. Sci. 9, 27 (1984). A. W. WHITE.In Toxic red tides and shellfish toxicity in Southeast Asia. Edited by A. W. White, M. Anraku, and K. K. Hooi. . 21, 15 (1982). 22. R. ZACZEKand J. T. C O P ~ ENeurophamacology, 23. J. S. KIZER,C. B. NEMEROFF, and W. W. YOUNGBLOOD. PharIDRC, Ottawa, Canada. 1984. p. 110. T. YASUMOTO, M. MURATA, Y. OSHIMA, M. SANO,G. K. MATmacol. Rev. 29, 301 (1978). 24. J. T. COYLE.J. Neurochem. 41, 1 (1983). SUMOTO, and J. CLARDY. Tetrahedron, 41, 1019 (1985). T. M. PERL,L. BEDARD,T. KOSATSKY, J. HOCKIN,and R. S. 25. R. A. SCHWARTZ and Y. BEN-ARI.Adv. Exp. Med. Biol. 203 REMIS.Epidemic Intelligence Service Conference, Center for (1986). and H. J. OLVERMAN. Trends Neurosci. 10, 265 26. J. C. WATKINS Disease Control, Atlanta, GA. 1988. T. J. BISCOE,R. H. EVANS,P. M. HEADLEY, M. MARTIN, and J. (1987). 27. C. ANGSTand M. WILLIAMS. Transmissions, 3 , 1 (1987). C. WATKINS. Nature, 255, 100 (1975). T. TAKEMOTO and K. DAIGO.Chem. Pharm. Bull. 6,578 (1958). 28. M. E. WARDand E. AIELLO.Physiol. Zool. 46, 157 (1973). 29. G. IMPELLIZZERI, S. MANGIAFICO, G. ORIENTE,M. PIATELLI, and K. DAIGO.Arch. Pharm. (Weinheim), 293, T. TAKEMOTO S. SCUITO,E. FATTORUSSO, S. MAGNO,C. SANTACROCE, and D. 627 (1960). SICA.Phytochemistry, 14, 1549 (1975). Official method of analysis. Association of Official Analytical Chemists, Arlington, VA. 18.086- 18.092. 1984; W. N. ADAMS 30. S. S. BATES,C. J. BIRD,R. K. BOYD,A. S. W. DE FREITAS, M. FALK,R. A. FOXALL,L. HANIC,W. D. JAMIESON, and J. J. MIESCIER. J. ASSOC. Off. Anal. Chem. 63, 1336 (1980). A. W, MCCULLOCH, P. ODENSE,M. A. QUILLIAM, P. G. SIM,P. THIR. K. BOYD,A. S. W. DE FREITAS,R. A. FOXALL,W. D. JAMIESON, M. V. LAYCOCK, A. W. MCCULLOCH, M. A. QUILBAULT, J. A. WALTER,and J. L. C. WRIGHT. Atlantic Research Laboratory Tech. Report No. 57 (1988). LIAM,and J. L. C. WRIGHT.Atlantic Research Laboratory Tech. 31. A. BAXand R. FREEMAN. J. Magn. Reson. 44,542 (1981). Report No. 54 (1988). 32. H. MICHL.Monatsh. Chem. 82, 489 (1951). M. FALK.Can. J. Spectrosc. 33, 117 (1988). 33. J. HEILMANN, J. BAROLLIER, and E. WATZKE.Hoppe-Seyler's G. W. A. MILNE,T. AXENROD, and H. M. FALES.J. Am. Chem. Z. Physiol. Chem. 309, 219 (1957). Soc. 92, 5 170 (1970); P. A. LECLERQand D. M. DESIDERIO. 34. D. V. SUBBARAO, M. A. QUILLIAM, and R. POCKLINGTON. Org. Mass Spectrom. 7, 5 15 (1973); M. MEOT-NERand F. H. Can. J. Fish. Aquat. Sci. 45, 2076 (1988). FIELD.J. Am. Chem. Soc. 95, 7207 (1973).
