Proc. Nat. Acad. Sci. USAVol. 68, No. 12, pp. 3223-3227, December 1971

Mechanism of Action of Bacitracin: Complexation with Metal Ionand C55-Isoprenyl Pyrophosphate

(molecular model/chelating agents/ternary complex/cell wall synthesis/peptidoglycan)

K. JOHN STONE* AND JACK L. STROMINGERtBiological Laboratories, Harvard University, Cambridge, Massachusetts 02138

Contributed by Jack L. Strominger, August 26, 1971

ABSTRACT The inhibition by bacitracin of the en-zymatic dephosphorylation of C55-isoprenyl pyrophosphateis abolished by the addition of chelating agents. If, how-ever, the chelating agent is added after a preincubation ofbacitracin with a divalent cation and the lipid sub-strate, then its addition has little effect, indicating thatbacitracin, metal ion, and C55-isoprenyl pyrophosphateform a complex. Various divalent cations can participatein complex formation, but monovalent cations are in-effective. A direct demonstration of the formation of acomplex between the C,%-isoprenyl pyrophosphate andbacitracin in the presence of metal ions was obtained.Molecular models that show one possible conformationfor a complex between bacitracin and the C55-isoprenylpyrophosphate, in which the metal ion acts as a bridgebetween the two compounds, are presented.

Bacitracin is an antibiotic produced by strains of Bacilluslicheniformis. It is a mixture of closely related compounds, themain component of which (bacitracin A) is a cyclic polypep-tide with a peptide side chain (Fig. 1) (1-3). An unusual fea-ture of the structure is the occurrence of a thiazoline ringformed between the L-cysteine and L-isoleucine residues at theN-terminal end of the acylic peptide side-chain. The freeamino group of iisoleucine at the N-terminal end is adjacentto the thiazoline ring and is essential for antimicrobial activ-ity. The transformation of bacitracin A to bacitracin F (inwhich the amino group is replaced by a carbonyl group) re-sults in total loss of antimicrobial activity.A metal ion may be essential for the antimicrobial activity

of bacitracin (4-1 1). The antimicrobial activity is stimulatedby various metal ions and inhibited by metal-chelating agents,such as EDTA. Spectroscopic evidence for the interaction ofbacitracin with metal ions has been obtained. The chelationmay involve both the nitrogen of the thiazoline ring and theadjacent free amino-group in isoleucine; this chelation couldassist in stabilizing the thiazoline ring, thus hindering the de-amination to bacitracin F (10). It has been further postulatedthat in the zinc-bacitracin A complex, the zinc also coordi-nates through the imidazole of the histidine residue and thepeptide nitrogen of the histidine residue, thus forming a one-to-one complex with bacitracin (10).

Various biochemical lesions induced by bacitracin havebeen reported (9). Several effects, such as inhibition of inducedenzyme synthesis and stimulation of efflux of K+ ions, couldbe ascribed to alterations in cell-membrane function. Theantibiotic also inhibits incorporation of ['4C]aminoacids intocell walls and induces the accumulation of uridine-nucleotide

precursors of the wall under conditions in which incorporationof amino acids into protein is unaffected (12-15). However,since bacitracin affects protoplasts of bacteria, its action is notlimited to effects on cell walls. The precise site of action incell wall synthesis has been defined as inhibition of the de-phosphorylation of C55-isoprenyl pyrophosphate. This re-action is essential for regeneration of the lipid carrier requiredfor the cyclic synthesis of peptidoglycan (16, 17). In the pres-ent paper, data are presented which suggest that bacitracinacts by forming a complex with C55-isoprenyl pyrophosphate(a component of the cell membrane) and divalent cations. Themetal ion may serve as a bridge between the antibiotic andthe substrate.

MATERIALS AND METHODS

The C55-isoprenyl pyrophosphatase used was that pres-ent in membranes of Streptococcus fecalis. The membraneswere prepared and washed with EDTA and 2 M LiCl asdescribed previously (through step 2 of the procedure) (Staud-enbauer, W. L., and J. L., Strominger, J. Biol. Chem., sub-mitted). The residual LiCl was removed by washing withwater and then dialysis against water. Finally, the membraneswere lyophilized. This freeze-dried preparation could bestored for at least a year. Before use, it was suspended in waterto a protein concentration of 1.1 mg/ml.The substrate, [32P]C55-isoprenyl pyrophosphate, was pre-

pared as described.1 1 liter of cells of Micrococcus lutes wasgrown at 37°C to half-maximum growth (8 hr after a 1% in-oculum) in a medium containing 1% Difco Bacto-peptone,0.1% Difco yeast extract, and 0.5% NaCl, adjusted to pH7.5 with NaOH. Cells were harvested and transferred tomedium that contained 10 mCi of [32P]inorganic phosphateand only 10% of the yeast extract used in the normal growthmedium. 4 hr later, 160 mg of bacitracin was added and, aftera further 3 hr, the cells were harvested by centrifugation.Lipids were extracted from the cells by acetone, chloroform-methanol, and pyridinium acetate-butanol.4 The pyridiniumacetate-butanol fraction was washed four times with 0.5volume of water. The lower aqueous phases were backwashedonce with butanol, which was then combined with the organicphase. Pyridine (0.3 volume) was added and the mixture wastaken to dryness on a rotary evaporator. Mild saponificationthydrolyzed phosphatides to water-soluble products. The mix-ture was again taken to dryness, dissolved in chloroform-methanol 2: 1 and washed five times with an equal volume of50% methanol. The lower chloroform layer was evaporatedto dryness (lipid A). The aqueous phase was backwashed

$ J. Biol. Chen., submitted for publication.

* Present address: Roche Research Laboratory, Department ofBiochemistry, Liverpool University, Liverpool, England.t To whom reprint requests should be sent.

3223

3224 Biochemistry: Stone and Strominger

(L- Cys) (L-Ile')a 1

CH-CH-C2Hll | ~~CH3CH2 N

/ ~~NH2C-H

C0O D-Phe 1o

3 L-Leu 11 L-Ile L-Hiss

I I4 D-Glu 12 D-Orn L-Asp a

L- Ile - L- Lys D-AsnLO a 7

FIG. 1. Proposed structure of bacitracin A. -0 signifies a C-r.N bond. The structure of the L-cysteine and L-isoleucine residuesat the end of the acyclic side chain are detailed in order to illus-trate the thiazoline ring.

twice with chloroform. The combined chloroform layers weretaken to dryness (lipid B). Lipid A and lipid B were separatelysuspended by sonication in acetone and passed through a 1-ml column of basic silica-gel G (Woelm), which had beenwashed with acetone. Elution was done batchwise. Acetone(3 ml) eluted nonradioactive, nonpolar lipids; chloroform-methanol 1:1 (3 ml) and methanol (3 ml) eluted radioactivelipid-P-P. The chloroform-methanol and methanol fractionswere combined and examined by thin-layer chromatographyon silica gel G, with isobutyl ketone-acetic acid-water 8:5: 1as the developing solvent. Radioautography of the chromato-gram showed that both lipid A (5.0 X 107 cpm) and lipid B(2.1 X 107 cpm) were predominantly C55-isoprenyl-P-P (RF =0.33) although lipid A contained traces (about 5%) of mate-rial that chromatographed with [32P]C55-isoprenyl-P (RF =0.41). An important observation was that when highly radio-active lipid-P-P was filtered through glass wool or chromato-graphed on a column of silica gel, the lipid-P-P was irreversiblybound to the glass unless it had previously been silanized witha5% solution of dimethyldichlorosilane in toluene.

Bacitracin was generously given by the Upjohn Co., Kala-mazoo, Mich. The general methods and materials used havebeen described (17, 18).t

RESULTS

Effect of Metal-Chelating Agents on Bacitracin Inhibition ofDephosphorylation of Lipid Pyrophosphate. Bacitracin at lowconcentrations inhibits the dephosphorylation of Cs-iso-prenoid alcohol pyrophosphate by a specific pyrophosphatase(17). This dephosphorylation does not require metal ions, asevidenced by the fact that concentrations of EDTA and othermetal chelating agents up to 0.05 M have no effect on itsactivity. However, the addition of EDTA or other chelatingagents to a system that had been inhibited to the extent ofmore than 95% by the addition of bacitracin resulted in res-toration of activity to the amount found in the absence ofbacitracin (Fig. 2). This effect of EDTA is not due to somechemical alteration of the bacitracin molecule produced byEDTA, since after prior incubation of bacitracin with EDTA,

the addition of magnesium ions in slight excess of the amountof EDTA added resulted in restoration of the inhibition. Theeffect was not specific for Mg++, but could also be obtainedwith Cu++.

Effect of Prior Incubation of Bacitracin with Various Com-ponents of the Reaction Mixture in the Presence of Metal Ions.The sequence of addition of components of the reaction mix-ture was important in obtaining reproducible results. All ofthe above experiments had, therefore, been done by additionof the lipid substrate, metal, chelating agent, bacitracin, andenzyme in the sequence indicated. However, if bacitracin,metal ions, and substrate were incubated together before theaddition of EDTA, EDTA reversed the inhibition due tobacitracin by only about 30%. On the other hand, prior in-cubation of enzyme with bacitracin in the presence of a metalion resulted in normal enzymatic activity when EDTA wasadded before substrate. These data suggest that a complex isformed between bacitracin, metal ion, and substrate that canbe reversed to only a small extent by the addition of EDTA.Similar results were obtained with hydroxyethylenediamine-triacetate and ethylene-bis-(oxyethylenenitrilo)tetraacetate.

Effects of Various Metals on Formation of the Inactive Com-plex. When bacitracin was incubated with substrate alone,even in the absence of added metal, considerable inhibition of

60

40-

20 -

10-6 I0-o io-4 10-3 10-2Conc. of Chelating Agent (M)

FIG. 2. Restoration of activity of bacitracin-inhibitedVCw.iosprenyl pyrophosphatase by chelating agents. Incubationmixtures contained 10,000 cpm of [82P]Cm-isoprenyl pyrophos-phate (dried at the bottom of the tube); 10 Al of a mixture con-taining 0.4 M Tris-HCl (pH 8.2), 0.1% 2-mercaptoethanol, and1.2% deoxycholate; 5 j1 of 6 mM bacitracin; 10 ul of membranepreparation from S. fecalis (1.1 mg of protein per ml); and 10Al of chelating agent, to give the final concentration indicated.After incubation for 30 min at 370C, the reactions were stoppedby heating in a boiling water bath for 2 min. Assays were per-formed by spotting the incubation mixtures at the origin of apaper chromatogram, then chromatographing overnight inisobutyric acid-1 N ammonium hydroxide 5:3. The compoundswere located by radioautography; the Rfs were lipid-32P, 1.0,lipid-32P-32P, 0.95, and inorganic phosphate-32P, 0.3. The areasof radioactivity were cut out and counted in a liquid scintilla-tion spectrometer. EDTA, ethylenediaminetetraacetate; HEDTA,hydroxyethylenediaminetriacetate; EBONTA, ethylene-bis-(oxy-ethylene-nitrilo)-tetraacetate. Two other chelating agents, aa'-dipryidyl and (2-hydroxyethylimino)-diacetate had no effect.Solid lines are the experimental tubes. The dotted line is an in-cubation from which bacitracin was omitted, but to which chelat-ing agent was added; all of the chelating agents gave identical

Proc. Nat. Acad. Sci. USA 68 (1971)

control lines.

Mechanism of Action of Bacitracin 3225

TABLE 1. Effects of different metal ions on inhibition ofC5-isoprenyl pyrophosphatase by bacitracin

Cpm of 32Pi releasedMetal ion added Control Experimental

None 2046 685Mg++ 2440 269Ca++ 2085 138Cu++ 2163 99Zn++ 2129 104Mn++ 812 281Fe++ 660 228Co++ 462 287Cd++ 104 48Li + 2021 1899Na+ 1805 1965K+ 1812 1702

Experiments were performed essentially as described in thelegend to Fig. 2.The control was obtained by mixing substrate and metal ion to

a final concentration of 1 mM, then adding EDTA before baci-tracin and enzyme. In the experimental tube, substrate and metalion were mixed and then bacitracin was added before EDTA andenzyme. The concentration of EDTA was 1.4 mM.

activity on subsequent addition of enzyme was observed(Table 1). Treatment of enzyme, substrate, and bacitracinseparately with EDTA failed to alter the result of this experi-ment.§ It is presumed that this effect of bacitracin in theabsence of added metal is due to the presence of small amountsof metal ions that could not be removed under the conditionsused. As bacitracin is known to have a very high affinity formetals (5, 7), it would presumably also scavenge traces ofmetal ions in the buffers, glassware, etc.

Addition of several divalent cations, Mg++, Ca++, Cu++,and Zn++, further enhanced the inhibition. In each case, theinhibition was relieved if EDTA was added before addition ofbacitracin, but not if it was added after bacitracin. Severalother metals, Mn++, Fe++, Co++, and Cd++, also enhancedthe inhibition by bacitracin, although under our conditionsthese metals, especially Cd++, also caused some inhibition ofthe enzyme reaction by themselves. The monovalent cations,Li+, Na+, and K+, failed to enhance the inhibition by baci-tracin. More striking, however, is the fact that they relievedthe inhibition observed in the absence of added metal ion.This effect was presumably due to displacement from bacitra-cin of trace amounts of divalent cations by an excess of thesemonovalent cations.

The Effects of Bacitracin on Enzyme Systems Involving OtherSubstrates. A model to be presented below suggests that thedivalent cation forms a bridge between the pyrophosphateresidue in the C55-isoprenyl pyrophosphate and bacitracin.If this is so, then other compounds containing pyrophosphateresidues might also be expected to complex with bacitracin.The effects of bacitracin on enzymatic reactions involvinginorganic pyrophosphate, ATP, and ADP were thereforeexamined. Adenylate kinase, ATPase, pyruvate phospho-kinase, and inorganic pyrophosphatase were not inhibited by

bacitracin at concentrations up to 10 mg/ml. All of theseenzymes require Mg++ or another divalent cation for activityand were tested in its presence. In addition, it was alreadyknown from the point of inhibition of bacitracin in the cycleof cell wall synthesis that bacitracin does not affect any of theother reactions in the cycle that involve either Cs-isoprenylphosphate or C55-isoprenyl pyrophosphoryl-sugar compounds,e.g., the initial reaction in which UDP-MurNAc-pentapeptidereacts with C55-isoprenyl phosphate to form lipid-P-P-Mur-NAc-pentapeptide, or any of the subsequent reactions in which

K)

I.

50

30

10

Fraction Number

FIG. 3. Complexation of bacitracin with ['2P]Ca5-isoprenylpyrophosphate. (A) The radioactive lipid substrate (5.8 X106 cpm) in 6.5 ml of 0.1 M Tris buffer (pH 8.2), containing 0.35%deoxycholate, 0.02 mM MgC12, and 0.5 mM 2-mercaptoethanolwas applied to a column of Sephadex G-25 (1.4 ml in a 1-mlgraduated pipet). Fractions of 40,u were collected in a micro-pipet every 1.5 min: the entire fraction was counted. The first1 ml of the elution fluid was discarded. The elution of radio-activity was constant after collection of an additional 40 frac-tions. Bacitracin, 7.2 mg, dissolved in 40,ul of the radioactiveelution solution was added to the top of the column, and elutionwas then continued. (B) The same experiment was performedexcept that MgCl2 was omitted from the radioactive buffer-substrate mixture, and EDTA, at a final concentration of 40mMwas added. (C) The experiment was done as described in A,except that [32P]C6u-ficaprenol phosphate (19) (4.5 X 106 cpm)was substituted for the [32P]Cas-isoprenyl pyrophosphate.

50 Bacitrocin Added

30-

I0

50

s~~~~50~~~~~~~~~~~~~~~~~~.

30

10 _

.I

C

I

1..

0 40 80 120

§ EDTA was removed from enzyme and bacitracin by filtrationon a Biogel P-2 column, and from the lipid substrate by partitionbetween CHCls and methanol-water followed by chromatographyof the organic phase on silica-gel G.

i

L

Proc. Nat. Acad. Sci. USA 68 (1971)

3226 Biochemistry: Stone and Strominger

M___ _ _ ..... _ _~~~~~~~~~~~4 WKI

FIG. 4. Molecular models. (A) Bacitracin. The cyclic peptide is underneath with the branched D-asparagine residue at the right andthe ilysine residue, which is the branch point for the acyclic peptide, at the left. The acyclic peptide lies on top of the cyclic ring withthe thiazoline ring at the top center. The nitrogen in the thiazoline ring and the nitrogen in the terminal isoleucine residue are shown withtwo coordination points for the metal ion. The a-amino group of ilysine (hook at left) and the a-amino group of D-asparagine (hook atright) are shown as the positions for hydrogen bonding of pyrophosphate. The drawing is a tracing of the model in which the positions ofthe amino nitrogen atoms of the amino acid are numbered as in Fig. 1; 6 and 6' are the a- and eamino groups of ilysine, respec-tively. The shaded area in the drawing is the acyclic portion of the molecule. (B) Bacitracin complexed with metal ion and Cu-iso-prenyl pyrophosphate. The metal ion is the large atom at the center.

acetylglucosamine is added to the lipid intermediate (18).The effects of bacitracin on C55-isoprenyl phosphate phos-phatase (Willoughby, E.. and J. L. Strominger J. Biol. Chem.,submitted) and on the ATP-dependent phosphokinase thatcatalyzes the phosphorylation of free C55-isoprenyl alcohol(19) were also examined; neither enzyme was inhibited by upto 1 mg/ml of bacitracin. On the other hand, marked inhibi-tion of the biosynthesis of squalene and sterols from mevalonicacid, involving short-chain prenyl pyrophosphates as inter-

mediates, has been observed and will be reported subse-quently.

Complex Formation between Bacitracin and C55-isoprenylPyrophosphate. When a sample of [82P]Cm-isoprenyl pyro-phosphate was filtered through a column of Sephadex G-25,in a buffer containing deoxycholate, until it was eluted at aconstant rate, the addition of bacitracin resulted in a troughof radioactivity followed by a peak, indicative of the forma-

Proc. Nat. Acad. Sci. USA 68 (1971)

Mechanism of Action of Bacitracin 3227

tion of a complex (Fig. 3A) (20). This result is the oppositeof the effect observed in the earlier use of this method, in whicha peak was followed by a trough (20), and must indicate thatthe complex is included in the gel to a greater extent than isthe uncomplexed C55-isoprenyl pyrophosphate. Presumably,the free C55-isoprenyl pyrophosphate is included in a relativelylarge detergent micelle, which cannot penetrate the gel. Inthe presence of EDTA, complexation was reduced (Fig. 3B).No complex formation was observed when C55-isoprenyl phos-phate was used (Fig. 3C).

Molecular Models of Bacitracin and of C-isoprenyl Pyro-phosphate. Acid hydrolysis of bacitracin A yields smallamounts of two peptides containing the sequence phenyl-alanyl-isoleucine (1); it is believed that this observation indi-cates that these two amino acids must lie very close to eachother in the structure. A model of bacitracin was built so thatthe N-terminal amino acid of the acyclic side chain, i-iso-leucine, was close enough to interact with the carbonyl carbonof the D-phenylalanine residue in the cyclic portion of themolecule. The thiazoline ring holds the peptide nitrogen ofthe cysteine residue in the ring and the free amino group ofthe N-terminal isoleucine residue in a position suitable for theformation of a metal chelate (Fig. 4A). Moreover, this sitelies, facing inwards at the head of a groove into which pyro-phosphate can be fit. The metal ion will then coordinatethrough four points of attachment so that it forms a bridgebetween the bacitracin and the pyrophosphate. Attachmentis made through oxygens of adjacent phosphates in the sub-strate and through peptide nitrogens of the D-asparagine resi-due attached to the cyclic ring, and of the a-amino group ofi-lysine in the ring at the point where the acyclic peptide sidechain is attached to this nitrogen. This coordination is shownin Fig. 4B, with the hydrophobic C55-isoprenoid chain in anextended conformation. The requirement of isoprenyl pyro-phosphates for binding, however, indicate that the isoprenoidchain must also interact with the bacitracin. Subsequentstudies (manuscript in preparation) have shown that neitherinorganic pyrophosphate nor isopentenyl pyrophosphate iscomplexed in the presence of metal to bacitracin, but C15-farnesyl pyrophosphate forms a strong complex.1 It should beemphasized that no data are available regarding the conforma-tion of bacitracin. Other conformations are possible in whichthe isoprenyl pyrophosphate can be complexed through themetal ion to bacitracin, including one in which the acyclicpeptide side chain is folded on the opposite side of the peptidering to interact with the phenylalanine residue and anotherin which the histidine residue (just above the D-asparagineresidue as photographed) is brought forward to provide co-ordination sites for the metal ion. The latter conformation

One possible interpretation of the partial reversal of complexformation by EDTA is that EDTA can compete with a portionof the bacitracin-metal-substrate complex in which the hydro-phobic chain has not interacted with the bacitracin molecule(Fig. 4B), but it cannot dissociate the complex in which the iso-prenoid chain is fully associated with bacitracin. Another possi-bility is that the complexes formed with the various componentsof the bacitracin mixture used in this study differ, and thatEDTA can compete for the metal in some of the complexes butnot in others.

would account for the known interaction of the histidine resi-due in bacitracin with the metal ion (10, 11), but the possi-bility also needs to be considered that the metal-histidine co-ordination site would be replaced by metal-pyrophosphatecoordination sites in the complex. The present model, pre-sented as a basis for further study, accounts for the failure ofbacitracin-metal to complex with the C55-isoprenyl pyro-phosphoryl-sugar derivative (since there is no space for thesugar derivatives in the groove into which pyrophosphatefits) or with C55-isoprenyl phosphate (which cannot complextightly with the metal ion).

Finally, the general effect of bacitracin on various mem-brane functions (9) could be explained by the distortion ofmembrane structure that might be expected from the com-plexation of Cm-isoprenyl pyrophosphate with bacitracin,either due to the removal of the lipid from the membrane ordue to the penetration of the membrane by bacitracin. Simi-larly, the toxicity of bacitracin for animal cells could be dueto the complexation of bacitracin with short-chain prenylpyrophosphates found in animal tissues.

Supported by research grants from the U.S. Public Service(AI-09152 and AM-13230) and the National Science Foundation(GB-29747X).

1. Abraham, E. P., Biochemistry of Some Peptide and SteroidAntibiotics (John Wiley and Sons, New York, 1957).

2. Craig, L. C., W. Konigsberg, and R. J. Hill, in Ciba Founda-tion Symposium on Amino Acids and Peptides with Anti-metabolic Activity (J. A. Churchill, Ltd., London, 1958),p. 226.

3. Regna, P. P., in Antibiotics; Their Chemistry and Non-medical Uses (D. Van Nostrand Co., New York, 1959),p. 58.

4. Weinberg, E. D., in Antibiotics Annual 1958/59 (MedicalEncyclopedia, Inc., New York, 1959), p. 924.

5. Garbutt, J. T., A. L. Morehouse, and A. M. Hanson, J.Agr. Food Chem., 9, 285 (1961).

6. Adler, R. H., and J. E. Snoke, J. Bacteriol., 83, 1315 (1962).7. Weinberg, E. D., in Antimicrobial Agents and Chemotherapy

1964 (Ann Arbor, Mich. Amer. Soc. Microbiol., 1965),p. 120.

8. Snoke, J. E., and N. Cornell, J. Bacteriol., 89, 415 (1965).9. Weinberg, E. D., in Antibiotics: Mechanism of Action, ed.

I. D. Gottlieb and P. Shaw (Springer-Verlag, New York,1967), p. 90.

10. Craig, L. C., W. F. Phillips, and M. Burachik, Biochemistry,6, 2348 (1969).

11. Cornell, D. W., and D. G. Guiney, Jr., Biochem., Biophys.Res. Commun., 40, 530 (1970).

12. Abraham, E. P., and G. G. F. Newton, in Ciba FoundationSymposium on Amino Acids and Peptides with AntimetabolicActivity (J. A. Churchill, Ltd., London, 1958), p. 205.

13. Park, J. T., Biochem. J., 70, 2P (1958).14. Park, J. T., in Antimicrobial Agents Ann. 1960 (Plenum

Press, New York, 1961), p. 338.15. Mandelstam, J., and H. J. Rogers, Biochem. J., 72, 654

(1959).16. Anderson, J. S., P. M. Meadow, M. A. Haskin, and J. L.

Strominger, Arch. Biochem. Biophys., 116, 487 (1966).17. Siewert, G., and J. L. Strominger, Proc. Nat. Acad. Sci.

USA, 57, 767 (1967).18. Anderson, J. S., M. Matsuhashi, M. A. Haskin, and J. L.

Strominger, J. Biol. Chem., 242, 3180 (1967).19. Higashi, Y., G. Siewert, and J. L. Strominger, J. Biol.

Chem., 245, 3683 (1970).20. Hummel, J. P., and W. J. Dreyer, Biochim. Biophys. Acta,

63, 532 (1962).

Proc. Nat. Acad. Sci. USA 68 (1971)