Itrprinlrti friim /'roc. \nl. Am,I. Sri. /',S'/1 Vol. 70, No. 2, pp. 480-484, February ]<)7.'{ Kvidem-.e Cor Boundary Lipid in Membranes (iiiriiihrauoiis cytochrome oxidasc/spin label/electron spin resonance) I'ATIilCIA ('. Mm*', o. HAYEK ORIKKITII*, RODERICK A. CAl'ALDIf, AND OARRET VANDERKOOlf •liiMiiiite of Molecular I4J..I..«>- imrl Department of Chemistry, University ofOregon, Eugene, Ore. 97403; and tlnstituto for ICruyme Kcscaicli, I invcisily of Wisconsin, Madison, Wis. 53700 i'liiiiiiiiiiiindi'l hi/ V. liorhlhridr, December /,, l!)7g ABSTH\OT (Mochromc oxidase (EC 1.9.3.1) isolated from hccf-hcarl mitochondria with an appropriate phospholipid content forms vesicular structures. Lipid-protein inter actions in this model membrane system were studied with the,lipid spin label, 16-doxylstearic acid. As the phospho- lipid/prolein ratio is varied, two spectral components are ohserved. Al low phospholipid/prolcin ratios (<0.19 mg of phospholipid per mg of protein) the lipid spin label is liifCll!) immobilized. At higher phospholipid content an additional component characteristic of fluid lipid hilayers is evideiil. By summation of digilalized spectra and sub- sequent integration it was shown that all composite spec tra could he approximated by assuming only two com ponents arc present, and that the amount of phospholipid hound In Ihr protein is independent of the extent of the fluid hilaw-r region. The experimentally determined amount of phospholipid for maximum occupancy of pro- Iciii-hound niies is ahouI 0.2 mg of phospholipid per 1.0 nig itl protein. Iialiulalions show that this ralio is con sistent with a single layerof phospholipid surrounding the protein complex. The data are interpreted as evidence for a boundary of immobilized lipid between the hydrophobic protein and adjacent fluid hilayer regions in this mem brane model system. There is compelling evidence for theexistence ofphospholipid hilayers in biological membranes. The results ofx-ray diffrac tion studies (]), differential thermal analysis (2), and spin labeling (3) have all indicated a similarity in lipid behavior between phospholipid hilayers and membranes. Evidence is also accumulating for the existence of globular amphipathic membrane proteins extending into orthrough thelipid regions of the membrane (4). Assuming this general model to be the case, there must exist a boundary between, the fluid bilayer region and the membrane proteins. An interesting question arisesas to the properties of the lipid-protein interface. Initi ally, a well-defined membranous preparation with relatively high protein content and easily characterized functional properties isa good system in which to examine lipid-protein interactions. For this study, we chose cytochrome oxidase (K<* 1.9.3.1), which forms a model membrane system with partially-characterized components and functional properties (5, 6). At phospholipid/protein ratios of 0.3-0.7 (w/w), cyto chrome oxidase, prepared by the general method ofSunetal. (7), spontaneously forms membranous vesicles (Fig. 1). These vesicular structures are essentially an artificial membrane system, but it is reasonable to suppose that interactions be tween the cytochrome oxidase protein complex and the phos pholipids are meaningful in relation to similar associations Abbreviations: KSU, electron spin resonance; 16-doxylstearic 'id, the 4',4'-dimethyloxazoladine-A-oxyl derivative of 16- ketostearic acid. in the intact inner mitochondrial membrane. In this paper we report a study of membranous cytochrome oxidase of various phospholipid contents that contains a small concen tration of the spin label, 16-doxylstearic acid (8). The properties of the protein-lipid boundary and the bilayer region are examined by electron spin resonance (EXR) spec troscopy. MATERIALS AND Millions Preparation of Cytochrome Oxidase. Beef-heart mitochondria were prepared by the method of Crane et al. (!)), except that 10 mM Tris-HOl (pH 7.8) replaced the phosphate buffer. These mitochondria were used to prepare membranous cyto chromeoxidaseby the general method of Hun etal. (7). Briefly, mitochondrial paste was suspended at 30 mg/ml in 0.25 M sucrose-10 mM Tris-IICI (pJI 7.1), containing I m,M EDTA, Triton X-114 added to 0.6 mg of Triton per mg of pro tein, and solid KOI added to a final concentration of 0.2 M KC1. After it wasstirred for 30 miu on ice, the suspension was centrifuged three times (78,000 X g for 60 min); each time the supernatant and pellet were discarded and the fluffy middle layer was retained. After the final centrifugation, protein concentration was adjusted to 20 mg/ml. The mixture was treated with Triton X-100 (1 mg/mg of protein) and solid KC1 to 1 M. It was stirred on ice for 30 min and centri fuged at 105,000 X g for 30 min. The pellet was washed twice by centrifugation or by overnight dialysis against Tris- IICI (pH 7.1). The cytochrome oxidase preparation was stored at a concentration of 20 mg/ml at —20°. These membranous cytochrome oxidase preparations varied in phospholipid content from 0.33-0.49 mg of phospholipid per mg of protein, assuming an average phospholipid molec ular weight of 775. Additional lipid was incorporated by the method of Fleischer and Fleischer (10); unincorporated lipid wasremoved by sucrose density centrifugation. The maximum phospholipid/protein ratio attained was 0.73, which corre sponds closely to the value of 0.70 calculated from the data of Chuang et al. (11) as the maximum phospholipid incor poration they could obtain. Phospholipid content was reduced by successive 30-min extractions with cold 95% aqueous acetone (10) followed by 480 31 I'roe. Nut. Acad. Sri. USA 70 (H)7:1) centrifugation, rotary evaporation, and suspension in buffer. Cytochrome oxidase lipids consisted of the pooled super- natants fnmi the successive acetone extractions. Lipids ex tracted in tins fashion contain an estimated 5-7% protein. in the manner described, a series of cytochrome oxidase samples were prepared that ranged from 0.10 mg of phospho lipid per mg of protein to 0.73 mg of phospholipid per mg of protein. All samples had a heme concentration of 7.8-8.5 nmol of heme a per mg of protein and had the character istic optical spectrum of native cytochrome oxidase. All were active,as estimatedby the method'of Smith (12). Phosphorous was measured by the method of ('hen et al. (13), protein was determined by the method of Lowry et al. (14), heme a was estimated by the methoil of Williams (15), and cytochrome oxidase activity was determined by the method of Smith (12). Electrophoresis on 5% acrylamide gels containing 1%, sodium dodecyl sulfate and 5 mM 2-mer- captoethanoi (Ml) was used to characterize the cytochrome oxidase preparations, with coomassie blue as the protein stain. All major protein bands appeared to be similar among the preparations containing various amounts of phospho lipid. The morphology of the various cytochrome oxidase preparations was checked by electron microscopy, with either 1% uranyi acetate (pH 4) or 1% phosphotungstic acid (pH 5) as negative stains. At ratios of 0.24 mg of phospholipid per nig of protein or less, cytochrome oxidase appeared amor phous or sheet-like by electron microscopy; at ratios of 0.33 or higher, the cytochrome oxidase was vesicular, similar to earlier observations (11). Preparations of Spin-Labeled Cytochrome Oxidase. Cyto chrome oxidase suspensions were washed twice by centri fugation and suspended in 10 mM phosphate buffer (pH 7.0). The spin label, Hi-doxy Istearic acid (Syva Associates), in chloroform solution was evaporated on the bottom of a vial, FlO. 1. Klectron micrograph of membranous cytochrome oxidase (mg of phospholipid per mg of protein ratio is 0.49), negatively stained with 1% phosphotungstic acid (pH 5) showing the vesicular structure. Micrograph taken by Mr. William Oolquhoun. (liar - 0.1 /im). Boundary Lipid in Membrane: NORMALIZED TO CENTER PEAK HEIGHT mq Phospholipid mo, Pfotefo (0.10) 1/ 1 b ' './(0.241 (0.33) i (0.49) (no protein) A, f NORMALIZED TO CONCENTRATION V f 481 Fig. 2. ESR spectraof 16-doxylstearic acid spin label in buf fered aqueous dispersions of membranous cytochrome oxidase with various lipid contents. The lipid to protein ratio expressed as mg of lipid per mgof protein is indicated at the far left. Left, spectra normalized to the center-line height; ru/hl, the same spectra normalized to give equivalent values after two integra tions, i.e., the right column representsconstant concentration. which was then aspirated to remove traces of chloroform, the buffered cytochrome oxidase suspension was added to the vial containing the thin film of spin label, and the sus pension was thoroughly mixed by brief low-power sonicntion (Heat Systems Ultrasonics, Inc. bath sonifier, 40 W, 2 mm) and then allowed to stand on ice for 30 min before a sample was withdrawn for ESR. spectroscopy- Labeling was kept constant at 25 nmol of spin label per mg of protein. No spin- spin interactions were seen at double this concentration, evenin the samples containing the lowest lipid concentration. ESR Measurements. All ESR measurements were made on a Varian E-3 spectrometer interfaced with an 8K Varian 020/i digital computer, with the scan speed controlled by an external oscillator.The first-derivativespectra were digital- izedand stored on paper tape for later replotting, integration, subtraction, or addition (17). The usual spectrometer settings were microwave power 5 mW, modulation amplitude 1 O, scan range 100 G. The filter timeconstant varied from 0.3 to 1 sec for scan speeds of 10-30 min, depending on the line width associated with the sample. All spectra were recorded at room temperature unlessotherwise noted. RESULTS AND DISCUSSION Spin-labeled cytochrome oxidase ESR spectra were recorded for preparations of spin-labeled cytochrome oxidase containing increasing concentrations oflipid. The amount of lipid in each ofthesamples isexpressed as the ratio of mg of phospholipid per mg of protein; the samples contained lipid/protein ratios of 0.10, 0.15, 0.19, 0.24, 0.25, 0.33, 0.39, 0.43, 0.49, and 0.73. In addition, a sam ple of lipids extracted from membranous cytochrome oxidase served as a reference sample. The spectra of representative samples are shown in Fig. 2, arranged in order of increasing phospholipid content. The J 482 Chemistry: .lost et al. SUMMED SPECTRA 100 0 e 0 100 #Fl»i. '•>. Synthesized spectra obtained by summation of vari ous amounts of spectrum a (lipid/protein ratio of 0.10) and spec- trttm e. (lipids extracted from membranous cytochrome oxidase) of Fig. 2. The ratios on the leftare the fraction of the total absorp- I ion contributed by each spectrum (a:e). The summed spectra have then been normalized to the same center-line height. Note the general similarity of the synthesized spectra b, c, and d to the corresponding experimental spectra (Fig. 2, b, c, and d). spectral data are scaled in two different ways to illustrate different features. In the left column the spectra are scaled to an arbitrary center-line height, while in the right column all spectra arc scaled to reflect the same concentration of Ilie spiii label as in spectrum a. This procedure allows direct comparison of the spectral line shapes. At the lowest, lipid content (Fig. 2a), the ESR spectrum is characteristic of strong Immobilization of the spin labels. As the lipid content of the membrane increases, a second, more mobile spectral component is easily distinguished (see dashed arrmi) in spectrum b). With further increase in lipid content, the relative intensity of the more mobile feature increases until the bound component is completely obscured (spectrum d). This is particularly evident when these spectra are scaled to the same concentration (Fig. 2, right). Because of the narrow line width, the line heights of the mobile com ponent art' large and tire off scale in spectra b-e, Fig. 2, right. The ESU data of Fig. 2 point out three interesting facts: (i) the presence of strongly immobilized spin label at low lipid levels, (ii) the appearance of a very much more mobile component in membranes with higher amounts of lipid, and (Hi) the composite nature of the spectra associated with inter mediate lipid content. We will now consider each of these facts separately. In spectrum a of Fig. 2, the arrows indicate the position of the outer hyperhne extrema. In the absence of molecular motion (e.g., at low temperature) the distance between these two extrema is 2 A„, where AM is the principal value of the electron-nuclear hyperfine interaction measured along the nitroxide z axis. As molecular motion increases, these outer most lines move in, so that a measurable spectral feature, 2 A,„, serves as tin indicator of relative molecular motion. The value ol 2 A,„ for spectrum a (measured between the two solid arrows) is M O; the same sample at 77° K has 2 Am = 2A„ = 09 O. Since 2 A„, is very nearly as large as 2 A2Z, it is apparent that the spin label in cytochrome oxidase with low lipidcontent is almost completely immobilized. Proc. Nat. Acad. Set. USA 70 (l!)7S) In contrast to the strong immobilization seen in spectrum a, of Fig. 2, the spectrum of the spin label in lipids alone is characteristic of a fluid environment (spectrum e). In previous studies, similar line shapes were observed when the same spin label was present in liposomes of egg lecithin. Thus, the degree of fluidity of the lipid bilayers composed of cytochrome oxidase lipids is comparable to that of egg lecithin bilayers. This spectrum serves as a useful referent* with which the composite spectra may be compared. The "N splitting constant measured from the mobile components of spectra b a" are 14.2 ± 0.2 O, a value typical of hydrocarbon environ ments. Tin- same spin label in phosphate buffer has a sub stantially larger splitting constant (i.e., 15.5 t 0.2 0). This solvent effect and the line shape rule out the possibility that the mobile component arises from spin label in the aqueous phase. Clearly the mobile component present in spectra b-d is similar to spectrum e. Based on this evidence, we can con clude that at higher lipid/protein ratios, cytochrome oxidase membranes containfluid lipid bilayerregions. Furthermore, it is clear from Fig. 2 that there are two spectral components present, rather than a continuum. Thus, tv)0 distinctlipid environments are present in cytochrome oxidase ' membranes. Quantitative spectral analysis The spectra of Fig. 2 suggest that two distinct phospholipid environments are present in this model membrane system. Three interesting questions arise at this point, (i) Do all spectra of Fig. 2 represent combinations of the mute two spectral components? (ii) If so, what are the ratios of the spin-label concentrations giving rise to the two components? (Hi) Are the number of immobilized lipid sites constant over a wide phospholipid range? The questions can he approached by summation of various proportions of the putative com ponents. For this purpose, the individual components were assumed to be the spectra from the cytochrome oxidasesample of lowest lipid content (0.10 mg of phospholipid per mg of protein) and from the sample of cytochrome oxidase lipids dispersed in buffer (Fig. 3, spectra a and c, respectively). These two spectra were digitalized, aligned horizontally with external #-value references, and then summed in different proportions to give the synthesized spectra 6, c, and d, Fig. 3. These summed spectra are a good approximation of the cor responding experimental spectra (Fig. 2, b, c, and d). The composite spectra do indeed appear to be approximated by various combinations of the same two spectral components. A digitalized spectrum can easily be integrated twice (17) to determine the relative concentration of spin label. In the summed spectra the relative concentrations were determined by integration of the individual components. By this pro cedure, i.e., summing to simulate experimental composite spectra and then integrating to determine relative contribu tions of each component, the proportions shown at the left of Fig. 3 were calculated. Thus, it is possible to estimate the relative spin-label concentrations contributing to the com posite spectrum. The third question concerns the amount of immobilized phospholipid relative to the amount of protein present. As suming the distribution of the lipid spin label faithfully re flects the distribution of phospholipids in the two environ ments, the amount of bound lipid, Cb, in units of mg of phos pholipid per mg of protein, is simply C„ = C,x, where 0, I'nr. Sal. Acad. Sri. f'.S'.l 70 (1978) is the experimental value for total phospholipid of flu; mem brane preparation in units of mg of phospholipid per mg of protein and x is the fraction of the total absorption contri buted by the bound component. From Fig. 3. the values of x lor tlic three iiicnibrane preparations (Fig. 3, '*, C, and d) are (1.05, D.09, and (1.34, respectively. Therefore, the corresponding <:ilcul.it<-'l values of (',, are 0.23, 0.23, and 0.17 mg of phos pholipid per iiu; of protein. These values of (',, are remarkably similar, considering the approximations involved, and wo conclude that the amount of phospholipid bound to the protein in independent of the extent of the fluid bilayer region. Mem branes formed of cytochrome oxitlase thus appear to require about 0.2 mg of phospholipid for maximum occupancy of protein-bound sites. We have also determined ('„ by spectral subtraction of the bound component from each of the spectra., as well as by the summation technique, and also by solving simultaneous equations derived by subtraction of one composite spectrum from another 'the range of values lor V.„ arrived at with these three techniques is 0.17 0.23 mg of phospholipid per mg of protein, confirming that (',, ~ 0.2 mg of phospholipid per mg of protein. It is possible, of course, to have less than full occupancy of the bound sites. For example spectrum a of Figs. 2 and 3 contains 0.10 mg of phospholipid per mg of protein, and only 50% of the sites are occupied (C„ = 0.1). I'reparations where C, < 0.10 show only the bound compo nent, whereas, when C, > 0.24, the spectra alsoshowa mobile component (Fig. 2). Evidently, as the lipid content increases, the bound sites arc fully occupied, and phospholipid in excess of this amount forms fluid bilayers. estimation of boundary lipid Given the observation that about 0.20 mg of phospholipid per mg of protein is immobile in the cytochrome oxidase membranes, the obvious question arises as to how this can be accounted for in molecular terms. The immobilization is evidently an effect of the protein on the lipid, since no counter part is found in the spectrum of the lipids extracted from the membranes. As observed by electron microscopy, the protein complexes have an irregular shape in the plane of the membrane (18). This shape can, however, be approximately represented by a rectangle of 52 X 60 A, based on measurements made on the micrographs. A plausible hypothesis is that the first lipid layer around the protein complex corresponds to the im mobilized component. We can calculate the number of lipid molecules that can be accommodated in a layer that is one aliphatic chain thick and then compare the value obtained with the measured amount found. The diameter of an extended aliphatic chain is about 4.8 A, as deduced from x-ray diffraction studies of hexagonal close-packed arrays of phospholipids (19). The perimeter of a rectangle 52 X 00 A equals 224A. Division of the perim eter by 4.8 A yields 47 aliphatic chains. This number must be divitled by 2 to get the number of equivalent phospholipid molecules, but must also be multiplied by 2 since the bilayer arrangement is assumed, giving 47 first-layer phospholipids per protein complex. (An equivalent phospholipid molecule is defined to contain one phosphorous atom and 2 aliphatic chains. On this basis, one real molecule of cardiolipin cor responds to 2 equivalent phospholipid molecules.) From the molecular weights used for the protein complex (i.e., 210,000) Boundary Lipid in Membranes 4S3 Fin. 4. Diagrammatic representation of a single protein complex and associated phospholipid in membranous cyto chrome oxidase. and the phospholipid molecules (i.e., 775) (IS), the result is obtained that 0.17 mg of phospholipid per mg of protein can be accommodated in the first layer around one protein complex. This is, of course, an approximation; the irregular real perimeter ofthe protein would tend to increase thisvalue, while protein-protein contacts would tend to decrease it. Also, the first layer was defined to consist of one layer of aliphatic chains; the immobilization effect might extend some what beyond that. With these reservations in mind, we note that the amount of lipid that can be accommoilatcd. in the first (boundary) layer ascalculated is very close to the observed, amount immobilized. This hypothesis is summarized diagrammatically in Fig. 4, where theprotein complex of membranous cytochrome oxidase is shown extending through the phospholipid region. The lirst layer of phospholipid is indicated as boundary lipid sur rounding the hydrophobic regions of the protein complex. Beyond the boundary lipid is the more-fluid phospholipid bilayer. This model emphasizes the two lipid regions, bound ary lipid and fluid bilayer, but oversimplifies the structure in that any protein-protein contacts are not shown and the demarcation between boundary lipid and fluid bilayer regions is exaggerated. The details of the lipid-protein interaction are unknown. However, the binding surfaces of the protein complex must be irregular, since the hydrophobic side groups of the polypeptide chains form thepotential wells responsible for immobilization of the lipid spin labels. The hypothesis of boundary lipid is adequate to explain the immobilization data. Another conceivable explanation is that the bound lipids are distributed more or less uniformly throughout the protein complex. Although there may be a few phospholipid molecules surrounded by protein, two ob servations argue against significant amounts of bound phos pholipid within the protein complex. First, Vanderkooi et al. (18) have shown that measurements from electron micro graphs (of crystalline regions that occur in some preparations of membranous cytochrome oxidase) are consistent with the interpretation that all phospholipid is distributed between the protein complexes. Second, the spin labels areobserved to exchange between the boundary layer and fluid bilayer re gions. This is most evident in the second column in Fig. 2. In this figure, the ratioofspin label to protein is held constant, while the total phospholipid increases from 0.10 to 0.49 mg ofphospholipid per mg ofprotein. The height of the immobil ized spectral component decreases monotonically as a func- I IS I I 'lieiiii 11\ . Jo,11, rl al. iion ol mcie:i mi. phospholipid, indicating the ability to ex change U-tneeii ret'.liMt win-re flic labels arc immobilized ami "g 'I Hind bilayer. Therefore, although we rumml eom- plelelv rule out llic po -ibilify of phospholipid buried in the protein, Hie I'.SI: ,|..|i;i ;il(. most easily aeeoimled for in terms ol (he I Hilary lipid model (Fig. I). In ' ii.'ii'. we have shown that the model membrane burned by cylochroine oxidase and phospholipid contains two distinct, phospholipid regions, differing markedly in fluid ify. We interpret ourdata to support the concept that protein complexes extend into or through the bilayer region and that the hydrophobic protein surface tightly binds a layer of lipid, effectively reducing the phospholipid participating in fluid bilayer formation. Given the amphipathic properties of many membrane proteins, the existence of boundary lipid may be a general phenomenon and would tend to reduce the amount of phospholipid available to form fluid bilayer regions in bio logical membranes. We acknowledge Ilie skillful technical assistance of Miss Dee Bright limn and Mrs. Annette Williamson. U.A.C. and (i.V. are grateful lo I'rof. 11. 10. IJreen for encouragement and support. This work wii supported by Crunt CA10:'..'i7 from the National Cancer Institute-mid Grant GM-I2H47 from the Institute of General Medical Science-. O.II.G. acknowledges Career Develop ment (Irani no. I-Kl-CA-2:s,:s.-,9 for support. R.A.C. thanks the Wellcome Trustee* for the award of a Wellcome Research Travel < bant. 1. Wilkins. \l ||. |\, Blaiirock, A. 10. & lOngehnan, D. M. (I *»V I I Xuhirr ,\ew Hail. 230, 72-70. 2. Slcuii, .1. M., Tourtcllotle, .M. 10., lteinert, ,). C, MdOl- haney, Ii. .M. & Rader, l!. L. (1909) Proc. Nat. Acad. Set UNA 63, 101-109. o. McConnclI, II. M. & McFarland, B. C. (1970) Quart. Rev. Iitophus. 3, 91 136; Jost, P.,Waggoner, A. S. &Oriffith, O. H. (1971) in Structure and Function of Biological Membranes, ml; Rofhfield, L. (Academic Press, New York), pp. 84-144. 7. S. 9. 10. II. 12. 13. 14. Ifi. Hi. 17. IS. 19. Proc. Nat. Ariul. Sri. I SA in U/',":'{) Smgcr, S, .1 A Nicolson, (i. I, f1972 I Srimn 17'., ,,SI 7;i|; Marelieii, V T, IMI.i.-U. |*. VV\, .I..,, It-..n. |j |, Scgrc-.l, .(. p. A Scott, II. K. (19721 I'rtm: \„l I,,,,/ *,-,' USA (>i), ill.", I 11<>: Sleek, T. I,. <|972) in Membmiu ft, march, ed. Fox, <'..!•'. (Ai-ademic Press. New Voiki, pp 71 !KS; \.-mdcrkoui, (J. IP.172) Aim. A.). ,\rnd, Sri |«S, 0 15; ('apaldi, Ii. A. .V (been, I). 10. I|!(~2l /•'/•.7,'.s' /,,,/ 2f>, 2tl.i 209. C.'ipalih, If. A. .v. Ilaviislii, || (1972) ILLS l.rlt 20 201 203. .las.-iilis. A. A., Nemecek, I. II., Severiiui, I. I., Skuln.cli.-v, V. P. &Smirnova, S. M. (1972) lliochim. Hiophys. Aria 27% 4K5-480; Ilinkle, P. C, Kim, .1. J. & Raeker !•; (1972) ,/. liiol. Chem. 247, 1.'538-1339. Sun, F. F., Prezbindowski, K. S., Crane, F. L. & Jacobs, E. 10. (1968) Hiochim. Hiophys. Acta 153, 804-818. Jost, P., Libertini, L. J., Hebert, V. C. & Griffith () II (1971) J. Mot. liiol. 59, 77-98. Crane, F. L., Olenn, J. L. &Green, I). 10. (1850) Hiochim Hiophys. Acta 22, 475-487. Fleischer, S. & Fleischer B. (1907) in Methods in Enzy- motogy, eds. Estabrook, R. W. & Pullman, M. 10. (Academic Press, New York), Vol. X, pp. 400-433. Ohuang, T. F., Awashti, Y. C. & Crane, F. L (197(1) five. Indiana Acad. Nri. 19(11) 79, 110-120. Smith, I,. (I9.r).r,) in Methods of Biochemical Aiiali/sis, ed. Click, D. (Intcrscience, New York), Vol. II, pp. 427-434. Chen, P. S., Toribara, T. Y. & Warner, II. (I9.-.0) Amil Chem. 28, 1750- I75S. bowry, O. II., Rosebrough, N. J., Fair, A. b.