Classification of phospholipases A, according to sequence

Transcription

1 Eur. J. Biochem. 13, (1983) FEBS 1983 Classification of phospholipases A, according to sequence Evolutionary and pharmacological implications Mark J. DUFTON and Robert C. HIDER Department of Chemistry, University of Essex, Colchester (Received July 8/ctober 3, 1983) - EJB The sequences of 32 phospholipases A, (EC ) were systematically compared on the basis of polypeptide chain length and similarity at selected amino acid positions around the active site. Two difference matrices were constructed and the various groupings present in the data were expressed in dendrogram form. The two methods of comparison yielded different results, and this is seen as a consequence of separate aspects of phospholipase evolution being highlighted in each case. It appears that, although Elapid snake venom phospholipases are very similar in terms of overall conformation, the area around their active sites distinguishes them into two major groups, namely the Asian Elapids and the marine/australasian Elapids. Further, the Asian Elapids seem to have active-site vicinities which are closer to those in the mammalian pancreatic phospholipases. The relevance of the classifications to structure/activity relationships (especially j-neurotoxicity) and phospholipase evolution is discussed. When many homologous sequences are known for a particular type or class of protein, theoretical analyses of the data can prove to be most informative. The application of secondary structure prediction methods to the extensive phospholipase A2 (EC ) sequence data has already been described in the preceding paper [I], but there is much additional information to be obtained by systematically comparing the sequences amongst themselves. If difference matrices for all the phospholipase A2 homologues are constructed, then it is possible to derive dendrograms which will highlight the various subgroupings present in the sequence set. This form of classification can suggest evolutionary pathways in the phospholipase A, family and provide a means of correlating pharmacological, immunological and physical properties. An intercomparison of 29 phospholipase A, homologues has been made by Verheij et al. [2] but although a difference matrix was constructed, no dendrogram was derived. In their analysis, the difference between any two sequences was recorded as the total number of non-identical amino acid positions, with homology gaps treated as a 21st amino acid. The present study differs in that only certain features of the sequences were chosen for comparison. As described more fully in Materials and Methods, this was in an attempt to avoid some of the problems associated with the use of homology gaps. MATERIALS AND METHODS In order that the true homology of the phospholipase A, enzymes be revealed, it is necessary to insert numerous gaps in the sequences of some examples [l]. This implies that during the evolution of these proteins, some residues or peptide segments have been either deleted or inserted in certain lines of descent. This can cause difficulty because it is not always obvious where to insert gaps, even when the properties and genetic coding of neighbouring residues is taken into account. Such decisions are Enzyme. Phospholipase A, (EC ). crucial if aligned sequences are to be quantitatively compared, but compounding this is the problem of how to score the difference between a gap and an actual amino acid during the comparison process. In an effort to lessen these problems, two approaches were used. Comparison of chain length The first was to consider only chain length. These proteins contain several half-cystine residues that form the disulphide bridges, and some of these can be considered as immutable reference points in the group as a whole. Therefore, each sequence is expressable in terms of the numbers of residues between successive half-cystine residues, and given two sequences, the total number of differences on this basis is readily summed, as shown in the following example. Enhydrina schistosa phospholipase A, Bitis gabonica phospholipase A, differences : A difference score of 13 is thus recorded for this comparison. Note that the E. schistosa phospholipase A2 contains half-cystine residues (and therefore disulphide bridges) that are absent from the B. gabonica example. In the above analysis, unique half-cystines were treated as ordinary amino acids, so in. the data set as a whole, only those half-cystines common to all were used as reference points. Once all the homologues had been compared in this way, a difference matrix was constructed and systematic clustering was undertaken. Comparison of sequence in gap-free segments The second approach was to compare quantitatively the actual sequences but only at those positions in the aligned data set where there is little doubt as to the homology arrangement (i.e. where there is no apparent variation in local chain length).

2 546 Again, the half-cystine residues were used as reference points. The total number of positions chosen that fitted the criteria and were variable within the data set was 25. Each sequence was then compared with all the others on the basis of these positions and the differences expressed in terms of minimum mutation distance [3]. Thus, a second difference matrix was produced. The information present in the two matrices was analysed by first recordering the matrix elements to highlight the obvious subgroups. Then these groups and any remaining members were clustered according to the unweighted pair-group mean average method [4]. RESULTS The 32 phospholipase A, sequences are listed [I] in the preceding paper where it may be noted that three originate from mammalian pancreas (bovine, porcine and equine), 25 from Elapid snake venom (i.e. cobras, kraits, tiger snakes and sea-snakes) and five from Viperid venom (i.e. rattlesnakes and vipers) C Results of chain-length comparison The difference matrix obtained from the chain length comparison is shown in Fig. 1 and the relevant dendrogram in Fig. 2. Three major groups are identifiable in the data and these are as follows: (a) the enzymes from mammalian pancreas and taipoxin y from the venom of the Australian Taipan (Oxyuranus scutellatus); (b) the snake venom enzymes from cobra (Naja), ringhals (Hemachatus), krait (Bungarus), seasnake (Laticauda, Enhydrina), tiger snake (Notechis) and taipan (Oxyuranus); (c) the snake venom enzymes from rattlesnake (Crotalus), viper (Bitis) and pit viper (Trimeresurus). Thus, with the exception of taipoxin y, all the phospholipases are classified according to the three major animal families represented, namely mammals, Elapid snakes and Viperid snakes respectively. Within each group, most of the examples are very similar to each other, so there are no homologues whose affinity is not clearly defined according to the criteria used. Surprisingly, the Elapid enzymes have a greater similarity to the pancreatic enzymes than they do to the Viperid enzymes. Results of sequence comparison The 25 positions chosen for the comparison are indicated in the legend to Fig. 3. The difference matrix built from analysing these positions is shown in Fig.3 and an accompanying dendrogram in Fig. 4. The same general trend as before is seen in Fig. 4 because the Elapid phospholipases again appear to be more similar to the pancreatic types than the Viperid types. An important difference, however, is that the Elapid enzymes are divided into two subgroups by virtue of the closer similarity of the pancreatic enzymes to the examples from Naja, Hemachatus and Bungarus. Since the other Elapid species are Notechis, Laticauda, Enhydrina and Oxyuranus, the division is essentially between the Asian species and the marine/australian species. The three fl-bungarotoxins (Al, A2 and A3 from Bungarus multicinctus) are classified outside the overall Elapid/mammal grouping and hence form a separate class. The Viperidae enzymes are again clustered together and their relative remoteness from the other phospholipases is re-emphasised. Although not included in the above analyses, the sequences of the phospholipases Laticauda semifasciata I11 and IV have also been determined [2]. They are very similar to L. semifasciata I and, in terms of the sequence comparison method used herein, differ by only two and one DNA base changes respectively. They are also both shorter by one amino acid. In consequence, the enzyme denoted as Laticauda semifasciata in the matrices and dendrograms can be taken as equally representing the subgroup of variants I, I11 and IV. DISCUSSION The relationships deduced from comparing chain length and sequence are different, so it is necessary to decide on their individual significance. Chain-length comparison In a family of homologous proteins such as the phospholipases, deletions or insertions of amino acids within the polypeptide chain can be regarded as more significant evolutionary milestones than residue substitutions. This is because an insertion or deletion (called a gap event ) requires at the genetic level the simultaneous addition or removal of a triplet codon, whereas a change of amino acid can be brought about by a minimum of one base substitution. Furthermore, changes in chain length can exert considerable influence on folding pathways and structure nucleation, so the acceptance of gap events will be severely limited by conformational criteria.

3 54 22 Bovine pancreas 23 Porcine pancreas 24 Equine pancreas 6 Oxyuranus: Taipoxin Y 15 3jA niqricollis 9 Laja mossambica: CM I 1 Najamossambica: CM I1 11 Qj_a mossambica: CM I11 12 Naja melanoleuca: DE I 13 Najg melanoleuca: DE I1 14 Kajs melanoleuca: DE I11 16 N_aj_a najs a - 19 Bji iija oxiana 2 Hemachatus hemachates 1 Najn naja kaouthia: CM I1 18 Eaj_a naja kaouthia: CM I11 1 Notechis: Notexin 2 Notechis: I1 5 5 Oxyuranus: Taipoxin 6 4 Oxvuranus: Taipoxin a 3 Notechis: I1 1 Enhydrina: Myotoxin 25 Bunqarus: 6 Bungarotoxin A1 26 Bungarus: 6 Bungarotoxin A2 2 Bungarus: 6 Bungarotoxin A3 8 Laticauda semifasciata 21 Bungarus multicinctus 29 Crotalus adamanteus 3 Crotalus atrox 31 Bitis caudalis I 28 Trimeresurus okinavensis 32 Bitis gabonica Differences Fig. 2. Dendrogram constructed from the difference matrix in Fig. I I e c e Fig. 3. Difference matrix for 32phospholipases A, on the basis of minimum mutation distance between 25 selected residue positions. The positions considered were 1,2,3,4,6,,8,1,31,32,35,3,38,39,41,42,44,4,48,51,11,14,15,1 and 18 according to Table 1 of the preceding paper [l]. For sequence identification, seelegend to Fig. 1. Where two or more sequences are grouped together (e.g. 16,1,18) they are identical in terms of the comparison Reference to Fig. 5 shows that the areas where gap events have occurred (i.e. the areas used to derive the dendrogram in Fig. 2) are largely on the periphery of the molecule, which is where the greatest tolerance of this type of change could be expected. Therefore Fig. 2 can be interpreted as showing the fundamental conformational resemblance of the various phospholipase groups in terms of the major evolutionary changes that have occurred within the family.

4 548 I Bungarus: 6 Bunqarotoxin A3 i, 29 Crotalus adamanteus 3 Crotalus atrox 28 Trimeresurus okinavensis 31 Bitis caudalis Bitis qabonica Minimum Mutation Distance Fig. 4. Dendrogram constructed from the difference matrix in Fig N_aj_a niqricollis Najs mossambica: CM I - Najs mossambica: CM I1 N_aj_a mossambica: CM 111 N_aj_a melanoleuca: DE I1 Najs melanoleuca: DE I11 - N_aj_a melanoleuca: DE I Laj_a m s N_aj_a Laja kaouthia: CM I1 N_aj_a ndja kaouthia: CM 111 N_aj_a naja oxiana Hemachatus hemachates Bungarus multicinctus Bovine pancreas Porcine pancreas Equine pancreas Notechis: Notexin Notechis: I Enhydrina: Myotoxin Laticauda semifasciata Notechis: I1 1 Oxyuranus: Taipoxin a Oxvuranus: Taipoxin B Oxyuranus: Taipoxin y Bunqarus: B Bunqarotoxin AI Bunqarus: 6 Bungarotoxin A Fig. 5. Schematic representation of the tertiary structure of bovine pancreatic phospholipase A, [2]. Key: solid line, inter-cystine chain segments of constant length used for sequence comparison; broken line, inter-cystine chain segments of variable length used for deletion/insertion comparison. -S-S-, disulphide bridge With the exception of taipoxin y from Oxyuranus, the groupings in Fig. 2 correspond to the three different families involved. These are the mammals, the Elapid snakes (i.e. Naja, Hemachatus, Oxyuranus, Bungarus, Laticauda and Notechis) and the Viperid snakes (i.e. Bitis, Crotalus and Trirneresurus). If the dendrogram is regarded as an evolutionary tree, then the depicted relationship between these three groups seems highly unlikely. Clearly, Elapid snakes would be expected to be more closely related to the Viperid snakes than the mammals. The same anomaly has also been pointed out by Verheij et al. [2] in the discussion of their phospholipase A, difference matrix. These workers suggested that the rate of change acceptance may have been higher in the Viperid descent, possibly as a result of the different disulphide bridge arrangement compared to the other types. Clearly, the problem of rate variation is fundamental to the interpretation of dendrograms as evolutionary trees, but similarly important is the problem of gene duplication [5, 61. According to the sequence data, Notechis scutatus has at least three phospholipase genes, as have Oxyuranus scutellatus, Naja mossarnbica and Naja melanoleuca. Bungarus multicinctus seems to possess at least four. Above all, the classification of taipoxin y with the pancreatic phospholipases shows that the mammalian pancreatic and Elapid type enzymes can occur together in the same organism. There is confirmatory evidence that taipoxin y is correctly classified because it has additional residues on its N terminus that are homologous to those cleaved from the pancreatic enzymes during their activation []. Therefore, many of the sequences compared herein are paralogous (i.e. from different genes) rather than orthologous (i.e. from the same descendant gene) [8]. The degree of similarity recorded between paralogous sequences is thus an expression of the time since a gene duplication and not the time since divergence from a common ancestor of the two species represented. Indeed, some of the original duplications may have occurred in remote ancestors of both snakes and mammals. When apparently recent duplications are considered, such as those giving rise to the variants in Naja rnelanoleuca, for example, there could also be a problem arising from the pooling of venom from genetically distinct individuals. In other words, each individual may only possess one version of the gene, but it may be sufficiently changeable for several versions to be extant in the population. Hence, the anomalous relationships of Viperids, Elapids, and mammals implied in Fig. 2 may be due in whole or in part to the dendrogram essentially reflecting the history of gene duplication events in the phospholipase family. It is to be hoped that this question can be resolved by determining the complete

5 549 phospholipase gene complement of the animals concerned, possibly by characterising the enzyme from different tissues within the same animal. Evolutionary considerations aside, Fig. 2 suggests that in terms of overall conformation, there are basically three types of phospholipase A, and that the pancreatic type has the closest resemblance to the Elapid type. Since crystal structures are already available for a pancreatic enzyme and a viperid enzyme [9, lo], the above finding could be tested should a crystal structure become available for an Elapid enzyme. Sequence comparisons The significance of the relationships shown in Fig. 4 will depend to a large extent on the positions of the residues chosen. Fig. 5 shows the inter-cystine chain segments susceptible to gap events. These segments are, in consequence, the very areas where homology alignment is difficult, so the inter-cystine chain segments of constant length in Fig.5 are those from which the dendrogram in Fig.4 is derived. Therefore, this analysis differs from the previous approach by comparing the cores of the enzymes rather than their peripheries. Since much of the core region defines the active site and its immediate vicinity [I I], it is appropriate to regard Fig. 4 as showing the resemblance of most of this crucial area throughout the 32 homo1 ogues. In Fig. 4, the Elapid enzymes (excepting 8-bungarotoxins Al, A2 and A3) are divided into two subgroups by the closer similarity of the Naja, Hemachatus and Bungarus enzymes to the mammalian pancreatic enzymes. This contrast to the findings of the chain-length comparison (Fig. 2) suggests that the evolution of the sequence in the vicinity of the active site is not directly linked to the changes that have taken place in the peripheral areas. The same phenomenon can be seen in another family of proteins from Elapid snake venom, namely the related short neurotoxins, long neurotoxins and cytotoxins. In this instance, short neurotoxins and long neurotoxins have very similar functional sites even though there are pronounced main-chain differences. In contrast, although short neurotoxins and cytotoxins are very similar in gross structure, the amino acid compositions of their functional sites are completely different [12]. If gap events are to be considered rare in comparison to residue substitutions and therefore slower to accumulate, then Fig. 4 implies that the separation of the two Elapid phospholipase sub-groups has come about after the development of the Elapid enzyme type as defined in Fig. 2, i.e. probably within the history of the Elapid family. Certainly, the division of the species between the two Elapid subgroups, essentially Asian cobraslkraits in one subgroup and sea-snakesfaustralian types in the other, suggests this. It also appears that a faster rate of change has accompanied the Oxyuranus, Notechis, Laticauda and Enhydrina descent after the cobraslkraits (Naja, Hemachatus, Bungarus) had diverged as a separate family. Moreover, since Laticauda and Enhydrina are sea-snakes and Oxyuranus and Notechis are Australian Elapids seemingly derived from marine ancestors [13, 141, an acceleration in the rate of change could be associated with the ancestral adoption of a marine lifestyle and the taking of marine prey. An alternative explanation of the sub-division of the Elapid enzymes is evolutionary convergence between the core areas of the mammalian pancreatic enzymes and those of the Naja, Hemachatus and Bungarus enzymes. If this were the case, the impression of rate variation could be erroneous. However, whichever explanation is considered the more likely, the presence of paralogous sequences within the subgroups means that judgements about species evolution have to be made cautiously. In comparing Fig. 2 and 4, particular mention should also be made of the P-bungarotoxins and taipoxin y. The Al, A2 and A3 P-bungarotoxins show little resemblance to the other Elapid sequences in Fig. 4 and are only surpassed in terms of difference by the Viperid enzymes. Since there is another homologue in the same venom (Bungarus multicinctus) which is more similar to the NajalHemachatus types, the B-bungarotoxins do warrant a separate classification. Taipoxin y was highlighted previously because it is structurally very similar to the pancreatic enzymes, but in Fig. 4, this resemblance is shown not to extend to the active-site vicinity. Instead, it is classified along with the other taipoxins, though the allegiance is not strongly defined. An interesting property in common between the /3-bungarotoxins and taipoxin y is that both function in liaison with chaperone proteins in order to achieve pre-synaptic neurotoxicity [ The bungarotoxins require a covalently attached Kunitz protease inhibitor homologue while taipoxin y (which has only weak enzymatic activity and no toxicity itself) functions in association with taipoxins a and /3 [18]. In comparison to the other Elapid enzymes therefore, these exceptional homologues appear to be highly individual and may represent developments unique to the species concerned. This individuality is also seen in their half-cystine positions which are not precisely matched in any of the other types. The question of neurotoxicity Although it is difficult to obtain firm evolutionary evidence from the dendrograms, they do reveal at their simplest level information on the Elapid phospholipases which is of potential significance to their modes of action. This is because they distinguish two subgroups of.elapid enzymes (i.e. the Asian Elapid and the marine/australasian Elapid types) on the basis of core area sequence, but only one group on the basis of intercystine chain length (i.e. overall conformation). The latter finding is also implied by the consistent half-cystine placings and by secondary structural predictions with circular dichroic studies. Two observations which complement these results are as follows. Firstly, all the neurotoxic phospholipases (excepting the P-bungarotoxin group) are to be found in the sea-snakes and Australian Elapids [15, 16, 191. Secondly, when the relative hydrophilicity of each enzyme s interface recognition site is considered (as extrapolated from the interface proposed for the bovine pancreatic enzyme [l 11) there is again a marked division between Asian Elapids (predominantly hydrophobic interfaces) and the sea-snakes/australian Elapids (predominantly hydrophilic interfaces) [I]. This tends to confirm the validity of the species division shown for the Elapids in Fig. 4 because of the 2 postulated interface recognition site residues, only two homologous positions were considered in the sequence comparison herein. There is one homologue from an Australian snake (taipoxin a) which does have a predominantly hydrophobic interface, but this normally functions in conjunction with taipoxins /3 and y, both of which have a predominantly hydrophilic interface. There is thus a marked tendency for neurotoxicity to be associated with Australian Elapidfsea-snake phospholipases having a relatively hydrophilic interface, suggesting that development of the latter could be a prerequisite for this mode of action. Even so, there still remains the question of what finally determines the presence or absence of neurotoxicity.

6 55 Table 1. Selected phospholipase sequences The sequences of the neurotoxic Notechis scutatus Notexin, Notechis scutatus I1 5 and Enhydrina schistosa myotoxin (sequences 1, 2 and respectively) and the non-neurotoxic Laticauda semijhsciata I and Notechis scutatus I1 1 (8 and 3 respectively). The sequences of L. semifasciata I11 and IV, designated S(a) and 8(b) respectively, are also included. Those positions which have completely different compositions in the neurotoxic examples are indicated (a), together with the postulated interface recognition site residues (+). The homology alignment and numbering system is that used in the preceding paper [l] Q NLVPFSYLIPCANHGKRP-TWHYMDYGCYCGAGGSGTPVDELDRCCKIHDDCYDEAGK-KGC- 2. NLVOFSYLIQCANHGRRP-TRHYMDYGCYCGWGGSGTPVDELDKCCKVHDDCYDSAEK-KGC-. NLVPFSYVITCANHNRRS-SLDYADYGCYCGAGGSGTPVDELDRCCKIHDDCYGEAEK-PGC- 8. NLVQFSNLIQCNVKGSRA-SYHYADYGCYCGAGGSGTPVDELDRCC~IHDNCYGEAEK-~GC- 8(a) 8(b) NLVPFTYLIQCANSGKRA-SYHYADYGCYCGAGGSGTPVDELDRCCKIHDNCYGEAEK-MGC- NLVOFSYLIQCANTGKRA-SYHYADYGCYCGAGGSGTPVDELDRCCKIHDNCYGEAEK-MGC- 3. NLVOFSNMIQCANHGSRP-SLAYAD~GCYCSAGGSGTPVDELDRCCKTHDDCYARATKSY~Cwe* + * # + + m m l lld rn 13 FPKMSAYDYY-CGENGP-Y-CRNIKKKCLRFVCDCDVEAAFCFAKAPYNNAN~NID-TKKR-CO -_-_ SPKMSAYDYY-CGENGP-Y-CRNIKKKCLRFVCDCDVEA~FCFAKAPYNN~NWNID-TKKR-CP YPKMLMYDYY-CGSNGP-Y-CRNVKKKCNRKVCDCDVAAAECFARNAYNN~NYNID-TKKR-CK ---_ YPKWTLYTYESCTDTSP---C-DEKTGCQGFVCACDLEAAKCFARSPYNNKNYNID-TSKR-CK YPKLTMYNYYSCGTPSP---C-DDKTGCPRYYCACDLEAAKCFARSPYNNKNYNID-TSKR-CK ---_ YPKLTMYNYYSCGTQSP---C-NDKTGCQRYVCACDLEAAKCFARSPYNNKNYNID-TSKR-CK ---- TPYWTLYSWO-CIEKTP-T-C-DSKTGCQRFVCDCDATAAKCFAKAPYNKENYNID-PKKR-CO ** m*m.m m e **+ * Fig. 6. Tertiary structure of bovine pancreatic phospholipase A, [ZO], showing location of residues (@) that apparently dictate the presence or absence of p-neurotoxic properties (see Table I ). Residues thought to compose the interface recognition site of this enzyme are also indicated (+) [ll]. Broken lines indicate main points of variance between pancreatic and Elapid phospholipases A, If Fig. 4 is examined, there is a notable cluster of similar homologues [Notechis I1 1, Laticauda phospholipase I (and thereby Laticauda variants I11 and IV), Notexin, Enhydrina myotoxin and Notechis 1151 of which the first four are nonneurotoxic and the remainder are neurotoxic in their monomeric form. If their complete sequences are aligned and compared (Table l), 15 positions can be identified where the neurotoxic homologues have residues not seen in either of the four non-neurotoxic homologues. Further, ten of these positions are invariant in the three neurotoxins. When these 15 positions are seen in the context of the crystal structure of the bovine pancreatic enzyme (whose structure is predicted to be quite similar to the Elapid enzymes), it transpires that only one hypothesized interface recognition site residue is included amongst them (Fig. 6). The majority of the differences are in fact localised in one area of the molecule, namely in the /?-sheet section and its immediate vicinity. This suggests that the final determinants of neurotoxicity are mostly disposed in this region and away from the catalytic site and the greater part of the interface recognition site. It should be mentioned, however, that the interface residue highlighted (position 1) may have a high degree of importance. In the neurotoxic enzymes in Table 1, this residue is methionine and in the four non-neurotoxic examples it is either tryptophan or leucine. Two of these residues are the rarest of the amino acids and so the replacement of one rarity by another does seem significant. Drawing a further analogy with the other distinct family of snake venom proteins, the a-neurotoxins and cytotoxins, it is notable that a change from tryptophan to methionine in a position undoubtedly central to their mode of action is also associated with a major change in pharmacological properties [12]. Concerning the other positions highlighted in Table 1 and Fig.6, some general observations can be made about the differences in their overall composition in the neurotoxic and non-neurotoxic types. For instance, the neurotoxic types have a

7 551 higher content of positively charged and hydrophobic residues while the non-neurotoxic types have more seryl and threonyl residues. Possibly, the area in which most of these residues lie is another recognition site, or alternatively, a section of the structure which can influence the allosteric behaviour of the molecule. If it is a recognition site, its remoteness from the catalytic site and interface implies interaction with a moiety other than the substrate. A suitable candidate here could be an endogenous protein at the site of action, perhaps a component of the presynaptic membrane itself. REFERENCES 1. Dufton, M. J., Eaker, D. & Hider, R. C. (1983) Eur. J. Biochem. 13, Verheij, H. M., Slotboom, A. J. & de Haas, G. H. (1981) Rev. Physiol. Biochem. Pharmacol. 91, Fitch, W. M. & Margoliash, E. (196) Science (Wash. DC) 155, Sneath, P. H. & Sokal, R. R. (193) Numerical Taxonomy, Chap. 5, W. H. Freeman, San Francisco. 5. Wilson, A. C., Carlson, S. S. & White, T. J. (19) Annu. Rev. Biochem. 46, Peacock, G. & Boulter, D. (195) J. Mol. Biol. 95, Fohlman, J., Lind,P. &Eaker, D. (19) FEBSLett. 84, Fitch, W M. & Margoliash, E. (19) Evol. Biol. 4, Dijkstra, B. W., Kalk, K. H., Hol., W. G. J. & Drenth, J. (1981) J. Mol. Biol. 14, Keith, C., Feldman, D., Deganello, S., Click, J., Ward, K. B., Jones, E.. & Sigler, P. B. (1981) J. Biol. Chem. 256, Dijkstra, B. W., Drenth, J. & Kalk, K. H. (1981) Nature (Lond.) 289, Dufton, M. J. &Hider, R. C. (1983) CRC Crit. Rev. Biochem. 14, Underwood, G. (196) A Contribution to the Classification of Snakes, Trustees British Natural History Museum, London. 14. Mao, S.-H., Chen, B.-Y., Yin, F.-Y. & Buo, Y.-W. (1983) Comp. Biochem. Physiol. 4 A, Kondo, K., Toda, H., Narita, K. & Lee, C. Y. (1982) J. Biochem. (Tokyo) 91, Kondo, K., Toda, H., Narita, K. & Lee, C. Y. (1982) J. Biochem. (Tokyo) 91, Fohlman, J., Eaker, D., Karlsson, E. & Thesleff, S. (196) Eur. J. Biochem. 68, Lind, P. & Eaker, D. (1982) Eur. J. Biochem. 124, Lee, C. Y. (199) in Neurotoxins: Tools in Biology, (Ceccarelli, B. & Clementi, F., eds) vol. 3, pp 1-16, Raven Press, New York. 2. Dijkstra, B. W., Drenth, J., Kalk, K. H. & Vandermaelen, P. J. (198) J. Mol. Biol. 124, M. J. Dufton and R. C. Hider, Department of Chemistry, University of Essex, Wivenhoe Park, Colchester, Essex, Great Britain C4 3SQ