Archives of Insect Biochemistry and Physiology 26:83-95 (1 994)

Isomeric and Quaternary Properties of Homogenous 33 kDa Protein From the Venom of Chelonus Near curvimaculatus Davy Jones, Anuradha Krishnan, Neville Sarkari, and Mietek Wozniak Graduate Center for Toxicology, University of Kentucky, Lexington

The 33,000 Dalton venom protein of Chelonus near curvimaculatus was charac- terized for structural properties of charge, quaternary associations, and relationship to polydnavirus encoded proteins. Homogenous isoforms of the protein were isolated from the venom by sequential steps of 1 ) microdissection, 2) separation based on charge (Mono-Q column HPLC or narrow-range electrofocusing), and 3) centrifugal filtration based on molecular weight using Centricon microconcentra- tors. The purified protein dimerized under native conditions, and this quaternary association became denaturation resistant under certain conditions. Chemical modification of lysine epsilon amino groups did not disrupt such dimerization. The cDNA for the protein did not possess high similarity to any sequence encoded in the polydnavirus, as indicated by results of Southern blotting, but does possess similarity in its repeats to the repeats of the immunologically protec?ive surface glycoprotein of Leishmania amazonensis. 0 1994 WiIey-Liss, Inc.

Key words: parasitoid, polydnavirus, dimerization, toxin, GP46/M-2

INTRODUCTION

Venoms and toxins have been useful molecular probes in both fundamental and practical disciplines. Snake venom proteins have been identified that bind to acetylcholine receptors or axonal ion channels, and these venom proteins have proven useful as affinity ligands for purifying or identifying components of their target sites 111. Tetradotoxin from puffer fish is another example of a toxin that has been widely used as an experimental probe on the nature of ion channels [21.

The venoms of insects and other arthropods have also yielded powerful probes of components of physiological and biochemical pathways. Honey bee

Acknowledgments: Neville Sarkari is a participant in NSF Research Experience for (REU) Program.

This research was supported in part by NIH grant GM33995 and fundsfrom the Universityof Kentucky Research and Graduate Studies Program.

Received February 16, 1993; accepted April 23, 1993.

Address reprint requests to Davy Jones, Graduate Center for Toxicology, University of Kentucky, Lexington, KY 40506.

0 1994 Wiley-Liss, Inc.

84 Jones et al.

melittin has assisted in probing the structure of cell membranes 131 and has even been used as an agent for the isolation of nuclei [4]. Melittin was the first protein whose mode of action was discerned from its primary structure [51. Scorpion venom proteins have been used as affinity ligands to isolate components of insect sodium channels [61, further demonstrating the usefulness of venom components as probes for previously unknown or unreachable components of biological processes.

Parasitic wasp venoms are becoming appreciated as a vast but relatively untapped source of experimental and practical tools. Paralyzing components of wasp venoms have been used in basic studies as probes of the insect neuronal synapse 171. Such venoms have been proposed for practical use in insect control [8-11]. A wide variety of experimental uses on nonparalyzing wasp venoms is suggested by their wide variety of effects [ill. Nonparalyzing venoms have been implicated in the survival of larval endoparasites [12-141, in endocrine disruptions in the host 1151, and in biochemical interactions with polydna- viruses [16].

Egg-larval endoparasites in the genus Chelonus have been shown to require the participation of their venom for survival of the larval endoparasite [171. Without the venom, endoparasites appear to become encapsulated [171. An active venom component in this effect is a 33 kDaX protein 1121, which has been cloned, sequenced, and biochemically characterized [18]. This protein has a number of unusual features, such as a repeating 14-residue motif, the functional requirement of which is unknown.

The purpose of the present study was to further characterize the protein and its functional properties. In this paper we identify several isoforms of the protein and examine the quaternary interactions among protein molecules.

MATERIALS AND METHODS Insects and Venom

Host Trichoplusia ni and parasites Chelonus curvimaculatus were reared as described previously 1191. Venom proteins were obtained by homogenization of adult female venom glands in 10 mM phosphate buffer, pH 7.5.

Analytical and Preparative Methods

Venom proteins were fractionated by HPLC on a Mono-Q column as described by Taylor and Jones [12]. Isoelectric focusing was performed as described by Jones et al. [20] and the proteins eluted from the gel fractions in 10 mM phosphate buffer, pH 7.5. Fractions containing the 33 kDa protein (as deter- mined by analytical SDS-10% PAGE [211) were concentrated with a Centricon centrifugal microconcentrator (Amicon, Beverly, MA). Immunoblotting was performed as described previously, using either antiserum against total venom proteins or against the N-terminal octapeptide of the 33 kDa protein [12,221.

*Abbreviations used: Da = Dalton; IEF = isoelectric focusing; SDS-PACE = sodium dodecyl sulfate polyacrylamide gel electrophoresis.

Structure of 33 kDa Venom Protein 85

Native Quaternary Conformation As a test of any quaternary associations, the native 33 kDa protein was

analyzed for native size by a kinetic analysis (Ferguson plots) of its electro- phoretic migration on native gels of differing acrylamide concentration [23,24]. Unpurified venom proteins extracted from venom glands were loaded directly onto a preparative native polyacrylamide gel, which after running was sliced into forty 1 mm fractions that were loaded into individual wells of an SDS polyacrylamide gel. Silver staining then identified the fraction containing the 33 kDa protein, thus providing its location on native gels. The kinetics of its change in migration rate as a function of acrylamide concentration on native PAGE then provided the estimate of its native size [23,24]. Similar kinetic analyses were performed with purified 33 kDa protein.

Polydnavirus DNA Extraction and Analysis Polydnavirus DNA was extracted from female ovaries as described pre-

viously [25]. The native viral DNA was fractionated in 1% agarose gels, was blotted to nitrocellulose and subjected to Southern analysis at 65°C using as a probe 32P-labeled cDNA (by random-primer labelling) for the33 kDa protein.

Chemical Modification of 33 kDa Protein Iodination of the 33 kDa protein with Bolton-Hunter reagent was performed

as described in the supplier’s instructions (ICN, Irvine, CAI. Briefly, samples of purified 33 kDa protein were incubated with [12511 Bolton-Hunter reagent, and after the reaction the samples were dialyzed and concentrated.

RESULTS Isoforms in Crude Venom

Narrow-range isoelectric focusing fractionated the venom components into a number of discrete bands, as visualized by silver staining (Fig. 1, lane A). These venom proteins were analyzed with antiserum generated against total venom proteins (Fig. 1, lane B) and antiserum against the N-terminal octapeptide (Fig. I, lane C). Probing of these components with antiserum specific for the N-ter- minus of the 33 kDa protein specifically yielded only two signals (Fig. lC>, indicating the presence of at least two forms of the protein with the same N-terminus.

HPLC Separation of Isofoms Since immunoblotting following SDS-PAGE yielded a single band [ 12,221,

whereas following IEF two bands were observed, it was likely that the 33 kDa protein existed as two isoforms, differing in charge. As an approach to prepa- rative isolation of these two forms, HPLC separation on a Mono-Q anion exchange resin was performed. Figure 2 shows SDS-PAGE analysis of protein eluting in the fractions containing the 33 kDa protein (verified by immunoblot; not shown). It can be seen that the 33 kDa protein eluted as two incompletely separated peaks. No other venom proteins were observed in the HPLC fractions containing either form of the 33 kDa protein.

86 Jones et al.

A B C

Fig. 1. immunological detection of two isoforms of the 33 kDa protein in the venom. lane A: Venom gland proteins detected by silver staining after separation by pH 4-6.5 electrofocusing. lane 8: Detection of separated venom proteins with polyclonal antibodies generated against total venom proteins. Lane C: Detection of two isoforms of 33 kDa protein with antibodies generated against N-terminal octapeptide of the protein. Arrows indicate positions of detected isoforms. The top of the figure is the acidic side of the IEF gel.

Preparative IEF

Complete preparative separation of the two isoforms observed on HPLC was achieved by either stretching of a narrow-range pH gradient (pH 4-6.5) across a 23 cm distance and slicing 2.5 mm fractions or by running a 9 cm gel and slicing 1 mm fractions. As shown in Figure 3, the two isoforms were completely isolated, with at least several empty fractions separating them. Further, no other

Structure of 33 kDa Venom Protein 87

r

I 0

r

0.28

- 5

0.26 2 0 0

0 0.24 2

-

-

- I I I

0.22

FRACTION NO.

Fig. 2. Elution of two isoforms of 33 kDa protein by HPLC with Mono-Q column. lower panel: Profile of proteins eluting as detected by absorbance at 280 nrn. Upper panel: Silver staining of analytical SDS-PAGE of eluted fractions. Two peaks of the 33 kDa protein are observed. No other venom proteins were detected in these fractions.

venom proteins were present in the fractions containing either isoform. A low abundance, and more basic, third isoform of the same Mr of 33,000 was also detected (not shown). These results showed that preparative IEF under these conditions was an effective approach in securing preparative amounts of indi- vidual isoforms for further analysis.

Relationship to Polydnavirus The possibility that the transcript coding for the 33 kDa polypeptide was

related to sequences encoded on the polydnavirus genome [251 was also inves- tigated. Probing of the native virus genome (Fig. 4, lane 1) with cDNA for the 33 kDa protein yielded no signal (Fig. 4, lane 3), even after long exposures of the autoradiogram, whereas the probe hybridized very strongly with itself, as a positive control (Fig. 4, lane 2).

Quaternary Interactions of 33 kDa Protein On native gels, the 33 kDA protein comigrated with the 52 kDa protein and

was visualized as the most intense band after silver staining (Fig. 5, lanes A,B).

88 Jones et al.

I I I I I I I M.W

97 66

31

Fig. 3. Separation of isoforms of 33 kDa protein by electrofocusing venom proteins across a 9 cm pH gradient of 4-6.5, and elution of the isoforms from 1 mm gel fractions. Proteins in each fraction were visualized by silver staining. Arrows indicate the positions of the major isoforms and one of the two minor isoform. A third, more basic isoform (minor isoform-2) focused several fractions in the more basic direction (not shown). Overdevelopment of the silver staining did not reveal any other proteins in those fractions containing each isoform. Molecular weight markers (M.W.) are shown, with their size in kDa along the right side.

Kinetic analysis with Ferguson plots of the migratory behavior of this 33 kDa protein band in native polyacrylamide gels of differing acrylamide concentra- tion resulted in an estimated size for the 33 kDa protein of approximately 35 kDa. This result demonstrates that the 33 kDa protein identified in the major native band in Figure 5 (lanes A,B) corresponds to the monomeric form of that protein (average of three independent experiments).

Purified 33 kDa protein migrated as at least two distinct bands on native PAGE (Fig. 6, lane A). The fastest migrating form comigrated with the monomeric form identified in the analysis of crude venom (Fig. 6, lane B). The migration kinetics of the second, slower migrating form revealed it to possess a molecular size of approximately 70 kDa, demonstrating it to be the dimeric form.

Interestingly, it was observed that when purified 33 kDa protein was highly concentrated, it dimerized into a form that did not dissociate under denaturing SDS-PAGE conditions (Fig. 7, lane A). When purified 33 kDa protein dimerized in this manner was iodinated with [12511 Bolton-Hunter reagent, such modifica- tions (of epsilon amino group of lysine residues) did not disrupt this denatura- tion-resistant dimerization (Fig. 7, lane B).

Structure of 33 kDa Venom Protein 89

1 2 3

Fig. 4. Southern analysis of polydnavirus genome using 33 kDa protein cDNA as a probe. lane 1: Ethidium bromide staining of DNA of segmented polydnavirus genome, which contains at least 20 different segments ranging in size from 8-27 kb [31]. lane 2: Autoradiograph of Southern blot of cDNA clone of 33 kDa protein (positive control). lane 3: Autoradiograph of Southern blot of polydnavirus genomic DNA (no signal was observed).

DISCUSSION

The identification of minor isoforms of the 33 kDa protein is the first report of the existence of isoforms of a parasitic wasp venom protein. The existence of these isoforms could be the basis for the observation that CNBr cleavage of this protein yields both 27.5 and 27 kDa fragments but never just one of the fragments 112,221. If a methionine residue occurred near the very C-terminus of the protein in one isoform but not in the other two, CNBr fragmentation would yield the observed result.

An immediate question is whether all three isoforms have functional activity, such as the dimerization activity described here or the activity promoting endoparasite survival described by Taylor and Jones [121. If each isoform has such an activity, then it may be possible to identify residues or regions that are not conserved, and which therefore are less likely to be necessary for activity. Such a strategy has been very useful in identifying those

90 Jones et al.

A B C D

97

66

31

Fig. 5. Determination of fractional location of 33 kDa protein following native PAGE. lane A: Silver stain of analytical preparation of purified 33 kDa protein run in parallel to preparative lane. Lane B: Silver stain of proteins (following SDS-PAGE) that were in the native PAGE gel fraction that contained the 33 kDa protein (arrow), which comigrated on native PAGE with the 52 kDa protein. lane C: Silver stain of total venom gland proteins following SDS-PAGE. lane D: SDS-PACE molecular size markers, with size in kDa shown along the right side (arrow denotes 33 kDa protein).

residues in snake venoms that are evolutionarily conserved due to their partici- pation in protein function [26].

The data presented here show that the sequence in the transcript for the 33 kDa protein is not closely similar to sequences encoded in that CcV polydna- virus genome. This result suggests that a different situation exists for this protein as compared with that described by Webb and Summers 1271. Those researchers found that some transcript sequences for venom gland proteins from Campoletis sonorensis have similarity to sequences encoded in the CsV polydna- virus. Although the first nine aminoacids of the 33 kDa protein (I-F-SF-D-D-L-V-C)

Structure of 33 kDa Venom Protein 91

M G

NATIVE PAGE

Fig. 6. Dimerization of purified 33 kDa protein under native PAGE conditions. Lane A: Monomeric and dimeric (arrow) forms of homogeneous 33 kDa protein. lane B: Total venom proteins (arrow denotes band containing 33 kDa protein, as deduced from experiment shown in Figure 5).

have a weak resemblance to the N-terminus of honey bee phospholipase (I-I-Y-P-G-T-L-W-C), the remainder of the protein possesses no detectable similarity [3] .

There is, however, immunological similarity between the 33 kDa venom protein and several 4 2 4 6 kDa proteins in the venom of Ascogaster quadridentutu [28]. The antiserum used to detect similarity possesses its epitopes on the series of 12 repeats that form the major part of the primary structure of the 33 kDa protein 1181. These data suggest that at least part of the repeating structure is evolutionarily conserved. At this time it is unclear whether the sequence of the repeats is conserved because of the functional necessity of specific residues per se or because of the need to preserve the ability of the protein to mutate to new structure by duplication or deletion of repeats (i.e., mutation by homologous recombination). The conservation of the repeats suggests several testable pos- sibilities on functional mechanisms of action of the protein. In an ice nucleation protein, it has been shown that the repeats bind water molecules into an ordered array, promoting ice crystallization [29]. In a variation of this theme, the repeats may function as either multivalent chelators of a single target molecule or perhaps by promoting aggregation through contacts with multiple target mole- cules (e.g., bivalent antibody molecules).

The fact that the protein dimerizes suggests that there may be evolutionary constraints on evolution of the protein with respect to both the residues active biologically and the residues functioning in dimerization. Advantage can be

92 Jones et al.

A B C

97

66

45

Fig. 7. Denaturation/modification resistance of the dimeric form. Proteins are visualized by silver staining. Lane A: Denaturation-resistant form of 33 kDa protein occurring following concentration of homogeneous preparation by Centricon centrifugal concentration. Lane B: Dimeric form of protein (shown in lane A) not disturbed by chemical modification with ['251]Bolton-Hunter reagent. (Protein is visualized by silver staining in lane A, autoradiography in lane B.) Lane C: Molecular weight markers. The size in kDa is shown along the right side.

taken of dimerization toward identi.fication of residues participating in interac- tion with target molecules. That is, residues participating in dimerization are unlikely to also participate in interaction with the target (e.g., dimerization domain of the steroid receptor is different than hormone- binding or DNA-bind- ing domains [301).

In purification of the 33 kDa protein, it was necessary for some experiments to use highly concentrated material. We repeatedly observed that under such conditions the protein dimerized to an approximately 70 kDa form that was stable even under SDS-PAGE denaturing conditions (Fig. 7). No aggregate higher than 70 kDa was observed.

As a test of the participation of lysine residues in this dimerization, purified protein which was dimerized was radiolabeled with Bolton-Hunter reagent. As shown in Figure 7, amidation of epsilon lysine residues with the iodinated reagent failed to disrupt dimerization, suggesting that the amidated lysine residues are not essential for dimerization. If this hypothesis is correct, then reaction with dimethylsuberimidate (which cross-links proteins through epsi-

Structure of 33 kDa Venom Protein 93

lon amino groups of lysine residues) should not result in covalent crosslinking of the dimer subunits. In preliminary experiments we have not found dimethyl- suberimidate to cross-link the subunits (not shown).

Intriguing is the denaturation-resistant condition of dimerization conditions observed to occur spontaneously when the protein is highly concentrated in pure form. The protein occurs in the venom gland reservoir at much higher concentrations than that which resulted in denaturation-resistant dimeriza- tion of the pure protein [31]. It may be that the venom gland contains additional components that prevent the form of dimerization observed for purified protein.

Our previous studies on the tertiary structure of the protein suggested that the 12 tandem repeats form the exterior of the protein 1181. It is therefore likely that the dimerization interface involves one or more of the 12 repeats. Since the sequences of the repeats are highly similar, it is unclear why only dimers were detected and not higher-order aggregates. No repeating structure involving more than one repeat (e.g., leucine-zipper [32]) is easily discerned from the primary structure.

Further analysis of the biochemical characteristics of the 33 kDa protein is likely to provide new information on structural and functional properties of wasp venom proteins that are not presently suspected or understood. It now appears that venom proteins organized as the 33 kDa protein have discrete structural components necessary for tertiary structure, quaternary associations, and interaction with target molecules, all of which place constraints on the molecule as it evolves through insertion and deletion of its tandem repeats. It is of considerable interest that the conserved ser-gly-ser motifs in each repeat [191 is similar to the ser-gly-ser/thr motif found in the tandem repeats of the protective surface glycoprotein of Leishmania amazonensis [33].

LITERATURE CITED

1. Beneski DA, Catterall WA (1980): Covalent labeling of protein components of the sodium channel with a photoactivable derivative of scorpion toxin. Proc Natl Acad Sci USA, 77539.

2. Mosher HS (1986): The chemistry of tetrodotoxin. Ann N Y Acad Sci 479:32.

3. Banks BEC, Shipolini RA (1986): Chemistry and pharmacology of honey-bee venom. In Piek T (ed): Venoms of the Hymenoptera. London: Academic Press, Inc., pp 330416.

4. Smith PJ, Friede MH, Scott BJ, von Holt C (1988): Isolation of nuclei from melittin-destabilized cells. Anal Biochem 169:390.

5. Windholz M (ed) (1983): Merck Index. Rathway, NJ: Merck and Co., p 913.

6. de Lima ME, Couraud F, Lapied €3, Pelhate M, Ribeiro-Diniz C, Rochat H (1988): Photoaffinity labelling of scorpion toxin receptors associated with insect synaptosomal Na’ channels. Biochem Biophys Res Commun 151:187.

7. Walther C, Zlotkin E, Rathmayer W (1976): Action of different toxins from the scorpion Androctonus australis on a locust nerve-muscle preparation. J Insect I’hysiol22:1187.

94 Jones et al.

8. Beard IU (1971): Arthropod venoms as insecticides. In Jacobson M, Crosby DG (eds): Naturally Occurring Insecticides. New York Marcel Dekker, pp 242-270.

9. Jones D (1984): Use of parasite regulation of host endocrinology to enhance the potential of biological control. Entomophaga 31:153.

10. Coudron TA (1991): Host regulating factors associated with parasitic Hymenoptera. In Hedin PA (ed): Naturally Occurring Pest Bioregulators. ACS Symposium Series No. 449. Washing- ton, DC: American Chem SOC, pp 4145.

11. Jones D, Coudron T (1993): Venoms of Parasitic Hymenoptera as Experimental Tools. In Beckage NE, Thompson SN, Federici B (eds): Parasites and Pathogens of Insects. San Diego, CA: Academic Press, pp 227-244.

12. Taylor T, Jones D (1990): Isolation and characterization of the 32.5 kDa protein from the venom of an endoparasitic wasp. Biochim Biophys Acta 103537.

13. Kitano M (1986): The role of Apanteles glomeratus venom in the defensive response of its host, Pieris rapae cruciuora. J Insect Physiol32:369.

14. Tanaka T, Vinson SB (1991): Interaction of venom with the calyx fluid of three parasitoids, Cardiochiles nigriceps, Microplitis croceipes (Hymenoptera: Braconidae), Campaletis sonorensis (Hymenoptera: Ichneumonidae) in effecting a delay in the pupation of Heliotkis virescens (Lepidoptera: Noctuidae). Ann Entomol SOC Am 84:97.

15. Coudron TA, Kelly TJ, Puttler B (1990): Developmental responses of Trickoplusiu ni (Lepidop- tera: Noctuidae) to parasitism by the ectoparasite Euplectrus plutkypenae (Hymenoptera: Eulo- phidae) Arch Insect Biochem Biophys 13:83.

16. Stoltz DB, Belland EP, Lucarotti CJ, and MacKinnon EA (1988): Venom promotes uncoating in vitro and persistence in vivo of DhTA from a braconid polydnavirus. J Gen Virol69903.

17. Leluk J, Schmidt J, Jones D (1988): Characterization and functions of the proteins of hymenopteran venoms. In Sehnal F, Zabza A, Denlinger DL (eds): Endocrinological Fron- tiers in Physiological Insect Ecology, vol. 10. Wroclaw: Wroclaw Tech University Press, pp 457-460.

18. Jones D, Sawicki G, Wozniak M (1992): Sequence, structure, and expression of a wasp venom protein with a negatively charged signal peptide and a novel repeating internal structure. J Biol Chem 26714871.

19. Jones D (1986): Chelonus sp.: Suppression of host ecdysteroid and developmentally stationary prepupae. Exp Parasitol61:lO.

20. Jones D, Jones G, Rudnicka M, Click AJ, Sreekrishna SC (1986): High resolution isoelec- tricfocusing of juvenile hormone esterase activity from the hemolymph of Trichoplusia ni (Hubner). Experientia 41:45.

21. Laemmli UK, Favre M (1973): Maturation of the head of bacteriophage T4. DNA packaging events. J Mol Biol80:575.

22. Jones D, Wozniak M (1991): Regulatory mediators in the venom of Ckelonus sp.: Their biosynthesis and subsequent processing in homologous and heterologous systems. Biochem Biophys Res Commun 178:213.

23. Ferguson KA (1964): Starch gel electrophoresis-application to the classification of pituitary proteins and polypeptides. Metabolism 13985.

Structure of 33 kDa Venom Protein 95

24. Hendrick JL, Smith AJ (1968): Size and charge isomer separation and estimation of molecular weight of proteins by disc gel electrophoresis. Arch Biochem Biophys 126:155.

25. Jones D, Sreekrishna S, Iwaya M, Yang J, Eberely M (1986): Comparison of viral ultrastructure and DNA banding patterns from the reproductive tracts of eastern and western hemisphere Chelonus sp. (Braconidae: Hymenoptera) J Invertebr Pathol47105-115.

26. Menez A, Boulain JC, Faure G, Couderc J, Liacopulos P (1982): Comparison of the "toxic" and antigenic regions in toxin alpha isolated from Naja nicrigollis. Toxicon 20:95.

27. Webb BA, Summers MD (1990): Venom and viral expression products of the endoparasitic wasp, Campoletis sonorensis share epitopes and related sequence. Proc Natl Acad Sci USA 874961.

28. Jones D, Taylor T, Farkas R, Chelliah J, Haene B, Brown J, Reed-Larsen (1990): Intercession of parasitic wasps (Cheloninae) in host developmental and biochemical pathways. In Hoshi M, Yamashita 0 (eds): Advances in Invertebrate Reproduction, vol. 5. Amsterdam: Elsevier Science Publishers, pp 157-162.

29. Green RL, Warren GJ (1985): Physical and functional repetition in a bacterial ice nucleation gene. Nature 317645.

30. Evans RM (1988): Steroid and thyroid hormone receptors as transcriptional regulators of development and physiology. Science 240889.

31. Jones D, Leluk J (1990): Venom proteins of the endoparasitic wasp Chelonus near curvima- culahcs: Characterization of the major components. Arch Insect Biochem Physiol13:95.

32. Turner R, Tjian R (1989): Leucine repeats and an adjacent DNA binding domain mediate the formation of functional cFos-cJun heterodimers. Science 243:1689.

33. Lohman KL, Langer PJ, McMahon-Pratt D (1990): Molecular cloning and characterization of the immunologically protective surface glycoprotein of GP46/M-2 of Leishmania amazonensis. Proc Natl Acad Sci USA 87:8393.