STUDIES ON THE GROSS STRUCTURE, CROSS-LINKAGES, A?JD TERMINAL SEQCENCES Ih; RIBONUCLEASE BY CHRISTIAN 13. ANFINSEN, ROBERT R. REDFIELD, WARREN L. CHOATE, JUANITA PAGE, AND WILLIAM R. CARROLL (FOIL the Laboratory of Cellular Physiology, National Heart Institute, and the National Institute of Arthritis and Metabolic Diseases, National Institutes of Health, United States Department of Health, Education, and Welfare, Bethesda, Maryland) (Received for publication, October 2, 1953) Previous studies (1,2) have shown an asymmet'ric labeling of amino acids derived from difierent points along the chains of radioactive proteins. These results have suggested, a,s one major possibility, t'hat proteins are . synt,hesized by an assembly mechanism involving the condensation of preformed peptide fragments not in ready equilibrium with the pools of free amino acid. An attractive corollary of this general hypothesis is that such fragments may serve as common building blocks for a number of proteins. The studies of Sanger and his collaborators (3, 4) have elucidated the amino acid sequences in the insulin molecule and have made available methodology (3, 5) for similar st,udies on other proteins. The present report deals with the progress to dat,e of our investigations on the fine and gross structures of ribonuclease, particularly in regard to N and C terminal sequences and t,he nature of the cross-linkages in this single chain protein. It is hoped that,, as comparison of the fine structure of these two pancreatic proteins becomes possible, a more rational basis for experiments on pro- tein synt,hesis in this tissue can be devised. The homogeneity of each lot of commercial crystalline ribonuclcase (Armour and Company or Worthington Biochemical I,aborat~ory) used in these studies was checked ultracentrifugally, and, in most, cases, by clectrophoresis. These preparations showed a single N terminal amino acid (lysine) by the dinitrofluorobenzene (DNFB) method. Their specific enzyme activit'y corresponded to the accept,ed values in the literature (6). Analytica. values obtained for phenylalanine (`7) (3 residues per mole) and for cysteine estimated as cysteic acid (8 residues per mole) agreed wit,hin 5 to 10 per cent with t,he values obt,ained on ion exchange columns by Hirs, Moore, and Stein.' The preparations of pepsin, chymotrypsin, and trypsin used for degrada- 1 Hirs, W., Moore, S., and Stein, W. H., personal communication. 201 202 STRUCTURE OF RIBONUCLEASB tion were crystalline commercial products. Worthington carbosypep- tidase was recrystallized eight times and treated (8) with diisopropyl fluorophosphate (DFP) before use. In view of the laborious and time-consuming nature of the commonly used methods for ribonuclease assay, a rapid and fairly accurate procedure was developed for use in these studies. Ribonuclease (0 to 14 y) in 1.50 cc. of 0.1 M acetate buffer, pH 5.0, is added to 1.0 cc. of yeast nucleic acid (Schwarz) dialyzed 48 hours against water, final concentration 0.8 per cent. After incubation for 25 minutes at 25" the reaction is stopped lvith 0.5 cc. of 0.75 per cent uranium acetate in 25 per cent perchloric acid. Following removal of precipitated protein and suhstrat,e by centrifugation, 0.10 cc. of the supernatant fluid is diluted to 3.1 cc. with water and read 500 .400 s .300 w" .200 .I00 , 0 I I I I I I o FEB.1953 - o MAR. " 6 APR. " *MAY " ' JO .20 .30 .40 .50 .60 .70 KUNITZ UNITS FIG. 1. Calibration curve for the determination of ribonuclease activity at 260 rnp in the Beckman spectrophotometer. A st,andard curve with known levels of pure ribonuclease is run with each set, of determinations, although this is probably unnecessary in view of the excellent reproduci- bility of t,he procedure (see Fig. 1). Correction is made for the reagent blank determined by incubation wibhout enzyme. This blank increases slowly during storage of the substrate in the cold, but. appears to have no effect on reproducibility. The method gives linear results up to 0.60 Iiunitz unit (45 units per mg. of ribonuclease). The data of Fig. 1 indicate t'he stability of the reagents employed over a 6 month period. Results Gross Strwiurc-The ribonuclease molecule is charart,erized by a high degree of geometrical symmetry. Its low f/f0 ratio ((9) and Table I) and unit cell dimensions (30 X 18 X 48 A (10)) indicate the probable presence of an extensive system of cross-linkage. Earlier chemical studies by Brand .4XFINSl?K, REDFIELD, CHOATE, P.AGE, AND C.ZRROCT, 203 and his colleagues (11) and other experiments from this laboratory (12) made possible calculat,ions suggesting the presence of t,hree N terminal groups, and consequent,ly three peptide chains per mole (mol. n-t. 13,500). The N terminal and C terminal end-group analyses reported below, how- ever, point almost, certainly to a single chain structure, and this con- clusion is confirmed by ultracentrifugal and diffusion experiments. Kative ribonuclease sediments, under the conditions described in Table I, with an s~O,~, of 1.9 to 2.2 (see also (9, 12)). Following treatment with performic TABLE I Chemical data Physical data -__ ~.______ ____- Ribonuclense I -SH groups --SO*- groups Calculated per mole per mole ;:;gyi;: "o& x "`O&X Calculated mol. wt. f per cc. fa w. gm. per mole Native. Ositlizetl. 1 "0 ~ i.8 f 0.1 `2 / "7:: ::i: ;::ii ::i; Native ribonuclense and oxidized enzyme prepared as described in the text were dialyacd overnight against phosphate buffet, r/2 = 0.1, pH 7.2. SOD., was deter- minrtl in t,hc Spinco ultracent,rifuge and D. 00.~ in the Aminco-Stern electrophoresis apparatus with boundary sharpening by the method of Kahn and Poison (33). The analytic~al data of Hirs, Moore, and Stein indicate a molecular weight of about 13,500 gm. per mole. Upon performic acid oxidation, this value should increase 500 to 1000 gm. per mole owing to introduction of oxygen into cysteine, methionine, and tyrosirre residues. The dat.a above have been chosen from an experiment in which all figures reported were derived from studies on a single batch of enzyme. Calculn- tions of moles are based on finite concentrations of protein. These data, when com- pared with those of Itothen (9)) indicate the desirability of more extensive studies on the physical properties of ribonuclease, particularly in regard to diffusion measure- ment,s, and the estimation of frirtional ratios; such experiments are being rarrietl out t)J- one of us (W. 11. C.). acid, this constant falls to 1.35. Such a change in sediment,ation might be accounted for either by oxidative division of the molecule into two essen- tially equal fragments or by rupture of cross-linking disulfide bonds, re- sulting in the production of a derivative so coiled as to impart greater frictional characteristics. Although the first alternative is almost ruled out by the fact that dialysis of oxidized ribonuclease results in no loss of nitrogen from the dialysis sac, this point was more thoroughly established by diffusion measurements. Table I summarizes t,he data from an experi- ment in which the molecular weights of native and performic acid-treated ribonurlease are compared. Ribonuclease contains 8 cyst,eine residues," * Tbr presence of 8 eysteine residues permits t,he theoretical presence of sixteen 204 STRUCTURB OF RII3ONUCLE.G2? all bound in disulfide linkage as evidenced by the absence of -%I groups when tested by the met'hod of Boycr (14). Solution in 85 per cent formic acid for 30 minutes at room temperature causes minimal, if any, change in the protein, since, upon removal of t,he solvent in uacuo, t.hc enzyme activit#y is essentially completely recovered. The presence of II& during this 30 minute period (1 part of 30 per cent Hz02 to 9 parts of HCOOH), however, results in the complete oxidat'ion of cysteine sulfur to the cysteic acid form. Thus, following hydrolysis in a sealed tube with C, N acid, 8 moles of cysteic ac+d per mole of protein (Table II) can be separated chro- matogrnphicsally on Dowex 50 columns (H+ form, pH 7) (13) and detcr- mined calorimetrically (15). The above result,s strongly suggest t,hat. this prot)ein is cross-linked through four disulfide bridges. The physical studies described above and TABLE 11 Oxidized ribonuclease hydrolyzed Cysteic acid determined Cysteic acid per 811 ribonuclease fix PM *u 1 0.0722 0.563 7.8 2 I 0.50-l 3.90 7.7 3 I 0.119 0.940 7.9 In Il;xperiment 1, ribonuclease calculated from the Kjeldahl nitrogen value, as- suming 13,400 gm. per mole of ribonuclease and 16.5 per cent nitrogen. In Exper- iments 2 and 3, ribonuclease calculated from the dry weight of sample. All values orrected for 10 per cent loss of cysteic acid during oxidation of protein (34). 0 the end-group analysis below lead one to the tentative postulat8ion of a gross structure, such as is depicted in Fig. 2. Some support for t'his gen- eral pi&we is derived from the x-ray diffraction experiments of Carlisle and Swuloudi (10) whic>h indicated five crystallographic chains." Com- plete amino ataid analyses by Hirs, Moore, and Stein' lead to an estimate of 121 amino acid residues per mole of ribonuclease. Thus in this prelimi- nary suggestion, each of t,he five peptide folds depicted should contain, on the average, 24 amino acid residues with disulfide cross-links as in- dipeptide sequences of this amino acid in ribonuclease. Using the general method described by Flavin (13), we have, at. present, direct, degradat,ive evidence for seven different, cysteic acid sequences from a considerably larger family of chromatograph- ically separable di- and tripeptides of this amino acid. 3 In a more recent paper (16), Carlisle, Scouloudi, and Spier st,ate that further examination of the x-ray data suggesk the presence of six crystallographic chains rnthrr than five. We have, nevertheless, schematized the molecule as shown in Fig. 2, with five folds, since t,he present chemical dat,a are compatible with such a struc- t,ure. dicat,ed. It is clear that such a general st~ructure, when further spatially compressed by arrangement of the peptide chain in the a-helix coils sug- gested by Pauling and Corey (l'i), would result in a highly compact, sym- metrical molecule. The presence of 4 proline residues' in a five fold struc- ture is also compatible with t,he postulated (18) Ale of this amino acid as a center of direction reversal in peptide chains. N Ter+mid Residue of Ribonuclease~Dinitrophenyl ribonuclease (DNP ribonuclease) was prepared according to the usual methods for DNI' pro- tein (5). Acid hydrolysis was performed eit,her in concentrat'ed HCl or constant boiling HCl in sealed tubes at 105" for varying intervals of 2 to 18 hours. Ident'ification of the DKP amino acids was made by paper chromatog- raphy, by the systems of Blackburn and Loather (19), Biserte and Os- [N "d LYS-A"-THR-ALA I S 3 / [PRO] s 3 9 b=`ROl / n [PRO] 9 $ [MET,TYR.ALA,LEU,PHE]-VAL FIG. 2. Generalized gross structure of ribonuclease teaux (20), Monier and PQnasse (21), and finally the two-dimensional technique of Levy.4 For quantitat,ive determination, the DP\`P spots were eluted with 1 per cent sodium bicarbonate and their absorpt'ion measured at 350 rnp in the Beckman spectrophot,omet'er (20, 22). As previously reported (23), bis-DNP lysine was the only DKP amino acid detected in the ether ext,racts of t,he hydrolysat,es. No cu-DNP argi- nine or Ins;-DSI' histidine could be detect,ed in the aqueous phase. Determination oj Moles Bis-DhTP %ysine per Alole DNP Ribonzrclcase- Weighed samples (approximately 0.15 to 0.2 pnf of DXP ribonuclease dried over PJOs) jvere suhmitt,ed to acid hydrolysis at 105" in sealed tubes lvith 0.5 cc. of constant boiling HCl. After 2, 5, and 8 hours, duplicate samples were diluted with wat#er, ext,racted with ether, and chromato- graphed by the two-dimensional t,echniquc of Levy.4 Duplicate 0.20 C(M aliquots of a st,andard bis-DNP lysine solution were subject'ed to t)he same protnedures, and the unknowns compared to the standards after identical hydrolysis t.imes. Vsing Him, Moore, and Stein's figures for the amino a,cid composition 4 Levy, A. L., personal communication. 206 STRUCTURE OF RIBONUCLEASE of bovine pancreatic ribonuclease,* and adding 167 for each e-lysine (ten), N terminal residue (one), 0-tyrosine (six), and imidazolyl group of histidine (four), as used by Porter (5) and Sanger (22), to the molecular weight of ribonuclease based on Hirs, Moore, and Stein's figures, one arrives at a molecular weight of 13,341 + 21(167) = 16,848 for DNP ribonuclease. An average of 0.90 mole of his-DNP lysine per mole DNP ribonuclease was detected. The 2 hour sample was disregarded because solution of the sample was incomplete (Table III). No spots other than bis-DNP lysine and bis-DNP lysylglut~amic acid (see below) were detected. As noted below, since the next residue was found to be t,hreonine, it is a safe assumption that the N terminal group was completely hydrolyzed. In addition to the 5 to 10 per cent error in TABLE III &mlitative Determination of N Terminul Av~ino Acid Residue Sample ~ Amount detected after hydrolysis Time of hydrolysis Weight Lysine Lysyl- Total IN terminal Ip- glutamic sine per mole hrs. w. A-4 f--v PM GM lliOl0 5 2.35 0.140 0.100 0.036 0.136 0.97 r F3 3.35 0.199 0.150 0.03i 0.94 3.22 0.191 0.141 0.016 , 0.187 0.157 0.82 8 1 3.04 0.181 / 0.141 0.017 0.158 i 0.87 bverage..................................,.........................~ 0.90 t'he method (22), it is conceivable that the figure of 0.9 rather than 1.0 end-group per mole might be due to a higher rat,e of destruction when the DNP derivative is present in the protein than that which prevails when the amino acid derivat,ive (24) is free in solution. Alternatively, it may l)e that the presence of other products of hydrolysis catalyzes the dcstruc- tion of the derivative. &termination of N Terminal Sequence-40 mg. of DNP rihonucleasc \vere subjected to partial acid hydrolysis in 11 pi HCl for 72 hours at 37". After dilution, the hydrolysate was extracted with ether, ethyl acetate, and n-butanol (25). The exkacts were concent,rat,ed in 2~~210 and then sul)ject,ed to electrophoresis on XIunktcll So. 20 paper \vith 0.033 RI phos- phate buffer at pH 7.0 (26) as electrolyte. The main zone present in the ether extract, migrnt,ing at the rate of his-DSP lysinc, \vns extracted with bicarbonatr and identified as such by paper chromatography. A fainter zone present in the ether and ethyl acetate extracts proved, upon complete hydrolysis and chromatography, to be his-DNP lysylglutamic acid. No zones containing his-DNP lysine mere found on electrophoresis of t'he wbutanol extracts. A second hydrolysis for 20 hours, rather than 72 hours, shelved again only the two components, his-DNP lysine and his-DNP lysylglutamic acid, the latter compound being present in greater amount. Pepsin Hydrolysis of DNP Ribonuclease-40 mg. of DNP-ribonuclease were suspended in 0.01 M HCl, the pH was adjusted to 1.8 with 0.1 11 HCl, and the suspension diluted to 4.0 cc. &h 0.01 M HCl. Pepsin (Worthington, crystallized four times), 0.28 mg., was dissolved in a, drop of water and added to the suspension, which was then incubated with shaking, one-fourth of t,he initial volume being withdrawn aft,er -I and after 22 hours. These nliquots were ex;tracted with et'hyl acet'ate until no further color was removed. The extracts were subjected to paper electrophoresis, the zones eluted and hydrolyzed, and the result'ing ether-extractable DNP derivatives and the water-soluble residues (27) were chromatographed. Free bis-DKP lysine, dinitrophenol, and an electrophoretic band con- taining bis-DNP lysine plus glutamic acid, threonine, and alxnine in ap- proximately equimolar concentrations were identified. In order to determine the sequence of the amino acids in the t)etrnpeptide component, t,he remaining material from the above incubation was ex- tracted with et,hyl acetate. After removal of the ethyl acct)at,e, the ex- tracted material \vas hydrolyzed for 18 hours at 37" with 11 K HCl. The hydrolysate was diluted with water and extracted with ethyl acetate. This extract contained bis-DNP lysine (2+) and bis-DXP lysylglutamic acid (4+) determined by the methods used in the previous section. The aque- ous portion, after removal of HCl in vacua, was treated with fluorodinitro- benzene in 1 per cent trimethylamine acetate buffer, pH 9.5. The excess reagent was extracted from the alkaline solution, and the solution was dried under high vacuum. A few drops of G N HCl were added, and the sample was hydrolyzed in a sealed tube at 10.5" for 7 hours, diluted with water, and extracted with ethyl acetate. After hydrolysis with G N HCI in sealed tubes, the ethyl acetate-soluble and the aqueous portions of the hydrolysate were chromatographed4 (27). The former contained DNP threonine (4+) and a trace of DNP glutamic acid, and the latter, free alanine (4+). The over-all findings lead to an N terminal sequence in ribonuclease of Lys-Glu (or Glu-NHz)-Thr-Ala-. Whether the glutamic acid is in the form of the amide has not been ascertained. The evidence from the mi- gration of the compound during paper electrophoresis is inconclusive, but suggest,s that the glutamic arid is in the form of its amide, since the peptide moves more slowly than the neutral amino acid derivat,ive, DNP alanine. C Terminal Residue and Sequence-We have previously reported (23) 208 STRUCTURE OF RIBONUCLEhSE the presence of a C terminal valine residue in ribonuclease as determined I)y carboxypeptidase degradation, These experiments have now been repeated with low levels of DFP-treat,ed carboxypeptidase to permit an estimat'e of the order of appearance of amino acids during digestion. Ali- quots of the incubation mixture were taken initially and aft'er successive intervals and either chromatographed directly after removal of protein (28, 29) or treated with dinitrofluorobenzene (5). After removal of excess reagent,, DYP amino acids were extracted from hhe acidified solution with ether. The extract was freed of dinitrophenol by passage through a silica column wit&h water-snt,urated chloroform as the moving phase (30), and, TABLE IV A mine Acids Relend from Native Ribonuclease by DFP-Treated Carborypeptirlanr (Estimated As Thei? DNP Derivatives) DPN amino acid Valinc Pheuylalanine. I,eucine. Alanine. Tyrosine Methionine 15 min. -____ 0.031 0.019 0.024 0.009 Trace Absent Incubation time 1 hr. 3 hrs. 0.064 0.118 0.039 0.098 0.043 0.090 0.021 , 0.086 0.020 0.062 Trace 1 0.016 The values are expressed as ~350 readings obtained in 1 cm. cells in the Beckman spectrophotometer on solutions of each DNP derivative extracted from paper chro- mntograms (Levy, personal communication) with 1 per cent KaHC03 in a total volume of 2.5 cc. 2 cc. aliquots taken from G cc. of phosphate buffer, pH i.8,0.1 M, cont,aining 60 mg. of ribonuclease + 0.03 mg. of DFP-treated carboxypeptidase recrystallized eight times. thereafter, t,he DNP amino acids mere eluted with et,her. Two-dimen- sional chromatography by the method of Levy4 yielded spots which were cut out, and cluted with 1 per rent NaHCO, , the color being read at, 350 rn/* in the Beckman spectrophotometer. The data in Table IV confirm the presence of valine as t)he C terminal residue and suggest the subsequent position in the chain of phenylalanine (or leucine and isoleucine), alanine, tyrosine, and methionine, as indicated t'entatively in Fig. 2. The presence of t,hese amino acids, and no others, was confirmed in the paper chromatograms (27) run on aliquots not treated with DXFR. Si- multaneous 3 hour comrol incubations of ribonuclease and carhoxypepti- dase alone comained no detectable free amino acids, whether examined by visual amino acid chromatography (27) or by the quant,itative DKP pro- cedure above. ANFINSEN, REDFIELD, CHOATE, PAGE:, AND CARROLL 209 DISCUSSION A considerable simplification would be int,roduced int,o the study of the general problem of protein synthesis if it could be established that common peptide building blocks were used in the biosynthesis of more than one protein species. The present studies, alt,hough obviously rather long t,erm in nature, appear to us to be one approach to the examination of t,his point, since t,hey will ultimately allow the direct comparison of fine struc- ture in two prot,eins, insulin and pancreatic ribonuclease, synthesized in the same &sue. Although the ribonuclease molecule appears to be a single cross-linked chain, its general structural features, suggested by the present experiments and tent,atively summarized in Fig. 2, make the suc- cessful elucidation of its structure a hopeful possibility. Preliminary studies have been made on a variety of fragments separated from pepsin, chymotrypsin, and trypsin digests of native and DNP ribo- nuclease, before and after performic acid oxidation. These studies have shown a comforting reproducibility of digestion products and, toget,her with dcterminat,ion of peptide sequence, will be reported in a subsequent' communication. In earlier experiments (12) it was found that digest,ion of ribonuclease with pepsin yielded a product of which only a relatively small part was dialyzable and which still sedimented in the ultracent,rifuge with an ~20 ,w of about 1.6. Treatment with thioglycolate did not change this figure. In the light of similar reducing studies on insulin (31, 32) and in view of our more recent esperiment*s in which this pepsin-produced product has been made completely non-sedimentahle by performic acid osidation, these earlier result-s can probably now be explained by the incomplete cleavage of disulfide bonds in the pepsin-modified molecule by thioglycolate. Physical and chemical studies on ribonuclea,se indicate that this protein consists of a single chain arranged in a compact, folded structure, cross- linked through four disulfide bridges. The molecule, after performic acid oxidation, still contains only a single chain. Its N terminal sequence has been shown to be lysylglutamyl (or gluta- minyl) threonylalanine. The C terminal position is occupied by a EL- line residue, followed, back along the peptide chain, by phenylalanine. leucine or isoleucine, alanine, tyrosine, and methionine in an undetermined order. BIBLIOGRAPHY 1. Steinlxrg, II., and Anfinsen, C. B., J. Biol. (~kem., 199, 25 (1952) 2. Anfinsen, C. B., and Flavin, M., Federation Proc., 12, 170 (1953). 210 STRUCTURE OF RIBOPWCLEASE 3. Sanger, F., and Tuppy, H., Biochem. J., 49, 463 (1951). 4. Sanger, F., and Thompson, E. 0. P., Biochem. J., 63,353 (1953). 5. Porter, R. R., in Gerard, R. W., Methods in medical research, Chicago, 3, 256 (1950). 6. Kunitz, M., J. Biol. Cheln., 164, 563 (1946). 7. Udenfriend, S., and Cooper, J. R., J. Biol. Cherrl., 203, 953 (1953). 5. Jansen, E. F., Nutting, M.-D. F., Jang, R., and Balls, A. K., b. Biol. Chern., 179, 189 (1949). 9. Rothen, A., J. Gen. Physiol., 24, 203 (1940). 10. Carlisle, C. H., and Scouloudi, H., Proc. Roy. Sot. Londotb, Series A, 207, 496 (1951). 11. Brand, E., dnn. New York Acad. SC., 47, 187 (1946). 12. Anfinsen, C. B., J. Biol. Chem., 196, 201 (1952). 13. Flavin, M., Xalure, in press. 14. Boycr, I'., Abstracts, American Chemical Society, 123rd meeting, Los Angeles, 31c (1953). 15. Moore, S., and Stein, W. H., J. Biol. Chem., 176, 367 (1948). 16. Carlisle, C. H., Scouloudi, H., and Spier, M., Proc. Roy. Sot. London, Series B, 141, 1 (1953). 17. Pauling, L., and Corey, R. B., Proc. Roy. Sot. London, Series B, 141, 21 (1953). 18. Arndt, U. W., and Riley, D. P., Nalr,re, 172, 245 (1953). 19. Blackburn, S., and Lowther, A. G., Biochem. J., 48, 126 (1951). 20. Biserte, G., and Osteaux, R., Bull. Sot. chim. biol., 33, 50 (1951). 21. Monier, R., and Penasse, L., Compt. rend. Acad., 230, 1176 (1950). 22. Sanger, F., Biochem. J., 46,563 (1940). 23. rlnfinsen, C. B., Flavin, M., and Farnsworth, J., Biochim. et biophys. acta, 9, 468 (1952). 24. Desnuelle, P., Rovery, M., and Bonjour, G., Biochim. el biophys. a&, 6, 116 (1950). 25. Woolley, D. W., d. Biol. Chem., 179, 593 (1949). 26. Biserte, G., Biochim. et biophys. acla, 4, 416 (1950). 27. Redfield, R. R., Biochim. et biophys. acta, 10,344 (1953). 28. Partridge, S. M., Nature, 169, 496 (1952). 29. Thompson, A. R., Nature, 169,495 (1952). 30. Sanger, F., Biochem. J., 39, 507 (1945). 31. Stern, K. G., and White, A., J. Biol. Chem., 117,95 (1937). 32. Miller, G. L., and Andersson, K. J. I., J. Biol. Chem., 144, 465 (1942). 33. Kahn, D. S., and Polson, -4., J. Phys. and Colloid Chew., 61,816 (1917). 31. Schrnm, E., Moore, S., and Bigwood, E. J., Biochcm. J., in press.