SYMIPOSIUAM ON RELATIONSHIP OF STRUCTURE OF ]MICROORGANISMSTO THEIR IMMUNOLOGICAL PROPERTIES1

III. STRUCTURE AND BIOLOGICAL PROPERTIES OF SURFACE ANTIGENS FROM GRAM-NEGATIVEBACTERIA

KELSEY C. MILNER, ROBERT L. ANACKER, KAZUE FUKUSHI, WILLARD T. HASKINS,MAURICE LANDY, BERNDT MALMGREN,2 AND EDGAR RIBI

U.S. Public Health Service, National Institute of Allergy and Infectious Diseases, Rocky MountainLaboratory, Hamilton, Montana, and Laboratory of Immunology, Bethesda, Maryland

INTRODUCTION.................... 352NATIVE HAPTEN............................................................................... 354MICELLAR STRUCTURE IN CELL WALLS ............................... 359CONCLUSIONS............................... 366LITERATURE CITED............................... 367

INTRODUCTION

The so-called "endotoxins," which are foundin greatest abundance in the cell walls of certaingram-negative bacteria, continue to piquescientific curiosity because neither their essentialchemical structure nor their mode of action iswell understood, despite the great amount ofattention they have received. Endotoxic activityappears to be inextricably bound up with majorsomatic antigens believed to be required forimmune prophylaxis against infection. Therehas been a continuing interest, therefore, inattempts to effect a separation of toxic andantigenic properties. The substances in questionhave a stability to physical and chemical agentswhich is unusual among biologically activematerials; they are, however, labile to acids andbases, both of which destroy toxicity. Theoriesof the nature of the changes resulting in detoxi-fication by these agents have been reviewedelsewhere (9, 10, 24). When the treatment issufficiently mild, the chemical determinants ofantigenic specificity remain intact and can be

'This symposium was held at the AnnualMeeting of the American Society for Microbiology,Cleveland, Ohio, 7 May 1963, under the sponsor-ship of the Division of Medical Microbiology andImmunology, with Edgar Ribi as convener.

2 Present address: Department of Bacteriology,Karolinska Institutet, Stockholm, Sweden.

isolated in the form of polysaccharide haptens.It is the purpose of this paper to consider therelationship between hapten and completeendotoxic 0 antigen, to support with data oneview of this relationship, and then to enlargethat view-speculatively and with free use ofanalogy with other known systems of chainmolecules-into a hypothesis which appears tobe consistent with present knowledge.

In a further study of the effects of acid hydroly-sis on endotoxin, it was reported that endotoxindepolymerized to hapten with no detectableintermediate products (10). If, as mild acidhydrolysis of an endotoxin was allowed to pro-ceed, samples were removed at intervals, quicklycooled, and neutralized, it then could be shownthat, as the reaction continued, there was a con-tinuous increase in the amount of a small particleof apparent molecular weight in the range of10,000-40,000. The endotoxin component dis-appeared concomitantly. Furthermore, if ahighly refined endotoxin from Salmonellaenteritidis was employed as starting material,the conversion was essentially complete; thatis, the endotoxin was converted directly andentirely into the particle of small molecularweight. Chemically and immunologically, thesmall-sized component was the polysaccharidehapten of the organism from which the materialoriginated-serologically specific, but nonanti-genic and nontoxic. The upper left diagram ofFig. 1 is the Schlieren pattern of the main com-

352

on March 25, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

VOL. 27, 1963 SURFACE ANTIGEN'S FROM GRAM-NNEGATIVE BACTERIA

UNTREATED ORIGINAL-( 7 MIN.)

1ii

90MIN. HYDROL.( 3 MIN.)

15 M N. HYD RCO L.( 19 miN ?

hhj

90 MIN. HYDROL.(19.6 MIN.)

45MNI. HYCL.(19MlNI

90 MIN. HYDROL.(115 MIN.)

FIG. 1. Ultracentrifugation of Salmonella enteritidis endotoxin subjected to progressive hydrolysis with0.1 x acetic acid. Times shown in parentheses indicate time after operating speed was reached (50,740 rev/min). Fro-m Ribi et al. (10).

ponent of an untreated aqueous ether endotoxin17 min after operating speed was reached. Thiswas a rather homogeneous material which,under the conditions of test, gave a sedimenta-tion constant of 10 S from which it was estimatedto have a molecular weight of about 1 million.When the endotoxin was subjected to 15 minof mild acid hydrolysis, neutralized, and cen-trifuged, the corresponding pattern, after 19min, was that shown at top center of the samefigure. Here there are two components of aboutequal area, the one on the left sedimenting muchmore slowly, with an S value of 1.6. These com-ponents could be separated and shown to consist

of hapten and potent endotoxin, respectively.As hydrolysis progressed, the slow-moving peakincreased as the endotoxin peak vanished. Bymeasuring the areas under the diagrams and tak-ing the disappearance of the endotoxin peak as100% depolymerization, we constructed thepicture shown in Table 1 (10). Here, figures inthe column headed "Depolymerization (%)" aretaken as a measure of the increasing proportionof hapten in the hydrolysate while the biologicaldata at the right of the table show the progressivedeterioration of activity.A second intriguing observation was that all

preparations of endotoxin which we examined

35-3

on March 25, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

TABLE 1. Relationship between degree of depolymerization and reduction in biological properties ofendotoxina during progressive hydrolysis with 0.1 N acetic acid (experiment I)b

Loss of biological activitydDuration of Area undertreatment haptene tion' Lethality for Resistance towith acid boundary BCG-treated infection Tumor damage Pyrogenicity

mice

mmin cm-12 W % W W %0 0 0 0 0 0 0

15 1.9 49 40 88 86 2345 3.7 95 95.7 98.3 99.7 94.290 3.8 98 99.1 99.3 99.9 99.2

360 3.9 100 99.4 99.6 99.9 99.9

a Salmonella enteritidis, aqueous-ether extract, lot Se 209.bFrom Ribi et al. (10).c Based on 100% depolymerization in the 360-min sample.d Based on 0% destruction in 0-time sample.

contained at the outset, before any deliberatehydrolysis, varying proportions of the nontoxic,serologically specific, small-particle-sized hapten(10). In the Ouchterlony plates diagrammed inFig. 2, antiserum is in the center wells with serial0.5-log dilutions of endotoxin arranged clockwisearound them. The endotoxins had been derivedfrom the same culture by three different methods.The higher concentrations of endotoxin alwaysshow two lines of precipitation, one which neveris far from the antigen well and one which formsclose to the center well. The latter, although theline is very distinct, is quantitatively the lessercomponent because it is diluted out earlier. InFig. 3, with the same scheme of dilutions, thecomponent forming lines nearer the serum wellsis seen to increase, at the expense of the othercomponent, with longer periods of hydrolysis(10). The parallel with the ultracentrifuge datais clear. These components have been separatedon Sephadex columns, and the one responsiblefor lines nearest the antigen wells has beenidentified with complete, endotoxic 0 antigen.Results of such clarity depend upon the use ofhighly refined materials which enable one toavoid complications from flagellar and envelopeantigens or minor antigens from the cytoplasm.

NATIVE HAPTEN

The question of interest was, of course, whatwas the origin of hapten in preparations ofendotoxin. Among the possibilities to be con-

6 0,O 2

40)""4

0 0o ~0 __l, o

Aqueous-ether Trichloroacetic acid Phenol-water

FIG. 2. Comparison in gel diffusion of endotoxinsderived from Salmonella enteritidis by differentprocedures. Key (ug/ml): 1 = 1000; 2 = 300; 3 =

100; 4 = 30; 5 = 10; 6 = 3. From Ribi et al. (10).

6 %.Q-, 2

04

Untreated endotoxin

o f

Subjected to twominutes hydrolysis

O MOo k

0

Subjected to fiveminutes hydrolysis

FIG. 3. Effect of brief acid hydrolysis on the geldiffusion pattern of endotoxin. Key (,ug/ml): 1 =

1000; 2 = 300; 3 = 100; 4 = 30; 5 = 10; 6 = 3.From Ribi et al. (10).

sidered were: (i) endotoxin was partially depoly-merized by at least several of the commonlyused methods for extracting it, (ii) all preparationsof endotoxin were equilibrium mixtures of haptenand complete antigen, (iii) endotoxin was partiallydecomposed by enzymes already present in the

354 MILNER ET AL. BACTERIOL. REV.

on March 25, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

VOL. 27, 1963 SURFACE ANTIGENS FROM GRAM-NEGATIVE BACTERIA

bacterial source, (iv) the presence of hapten was

evidence of a precursor in the synthesis ofcomplete antigen. It has not been possible toidentify any of these as the sole correct explana-tion, but some partial answers have been suppliedby the following experiments.

Starting materials from Escherichia coli were

chosen because we were studying that species inconnection with experiments reported elsewhere(2). Freshly harvested cells were disrupted in waterwith a modified pressure cell (Servall RefrigeratedCell Fractionator). The supernatant from a firstcentrifugation was collected as the protoplasmfraction, and the cell-wall fraction was thoroughly

washed. The cleanly separated cell walls were

extracted by the phenol-water method of West-phal, Lideritz, and Bister (27), a procedure thatgave better yields from this starting materialthan extraction with aqueous ether or trichloro-acetic acid. The results of two experiments are

shown in Table 2. Solids in the water phase ofthe extraction mixture constituted 26 or 27 % ofthe starting material. In both cases, the cell-wallresidue, recovered from phenol and interfaciallayers, contained most of the nitrogen, whereasmost of the hexose and total carbohydrate hadbeen extracted.

Table 3 contains data on the endotoxic prop-

TABLE 2. Chemical analysis of

Organism

E. coli0113(Ec 14series)

E. coli0111: B4(Ec 10series)

Preparaticn*

Cell wallsPhenol-water

extract ofcell walls

Residue

Cell wallsPhenol-water

extract ofcell walls

Residue

cell walls, extracts of cell walls, and extraction residues of Escherichia coli

Recovery

100

2670

100

1 2770

Nitrogen iPhosphorus

9.5

1.913

8.1

1.312

1.2

2.10.67

1.1

1.50.52

Hexose

5.1

180.72

11

290.81

Totalcarbohy-drate

8.7

351.6

17

402.0

Totalfattyacids

35

3334

32

1232

Hexos-amine

4.9

132.1

5.6

9.00.72

* All preparations nucleic acid-free.

TABLE 3. Endotoxic properties of cell walls, extracts of cell walls, and extraction residues of Escherichia coli

Lethality Nonspecific Pyoeiiy Sntst harmnOrganism Preparation for mice protection Pyrogenicity (SLnts) Shwartzman

(LDbo) (Ensio)

tng ig Ag Yg pgE. coli 0113 (Ec Cell walls 1.9 30 3.0 2.7 16

14 series) Phenol-waterextract ofcell walls 0.35 9.0 0.16 0.16 1.3

Residue >4.0 > 200 > 500 > 200 > 400

E. coli O11l:B4 Cell walls 1.1 13 9.0 63(Ec 10 series) Phenol-water

extract ofcell walls 0.20 2.5 0.21 0.47

Residue >4.0 200 210 > 200

3554

on March 25, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

MILNER ET AL.

FIG. 4. (upper left) Comparison in gel diffusion of phenol-water extracts of cell walls, endotoxin extractedwith trichloroacetic acid, and haptens from Escherichia coli (3 days of incubation). Key: TCA Tox., endo-toxin extracted from whole cells with trichloroacetic acid; Acid Hapt., hapten purified from product of acidhydrolysis of endotoxin; (Hapt.), "native hapten" extracted from protoplasm; Phen-C.W., endotoxin ex-tracted from washed cell walls by hot phenol-water.

FIG. 5. (upper right) Comparison in gel diffusion of phenol-water extracts of cell walls, endotoxin ex-tracted with trichloroacetic acid, and haptens from Escherichia coli (5 days of incubation). Key: same asFig. 4.

FIG. 6. (lower left) Comparison in gel diffusion of phenol-water extracts of cell walls, endotoxin extractedwith trichloroacetic acid, and haptens from Escherichia coli (5 days of incubation). Key: same as Fig. 4,but different antiserum.

FIG. 7. (lower right) Comparison in gel diffusion of acid hapten, endotoxin, and trichloroacetic acidextract of protoplasm front Escherichia coli. Key: A, acid hapten; B, trichloroacetic acid endotoxin fromwhole cells; C, trichloroacetic acid extract of protoplasm.

erties of the same fractions. The inertness of theresidues indicates that the active material hasbeen effectively extracted, but a fraction con-

taining only a quarter of the whole cell walls isnow, on the average, more than four times as

potent as the starting material. Perhaps the

greater degree of dispersion is responsible. Thesedata bear out our overall impression that pyro-genicity and Shwartzman reactions, for example,are affected by the state of dispersion of activematerials, whereas lethality for mice and pro-tection, in general, are not.

35)6 BACTERIOL. REV.

on March 25, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

VOL. 27, 1963 SURFACE ANTIGENS FROM G

In Fig. 4, phenol extracts of cell walls fromE. coli 0113, trichloroacetic acid endotoxin fromwhole cells, and toxin-free haptens from thesame culture are compared by Ouchterlony geldiffusion tests. The haptens at the lower rightshow only the one broad line which merges witha similar line in the trichloroacetic acid endotoxinat upper right. At top and left are three extracts

XRAM-NEGATIVE BACTERIA 357

from cell walls which show, in addition to thecharacteristic endotoxin lines near the antigenwells, thin lines of precipitation in the regionusually occupied by hapten. These do not mergewith the known hapten lines, however. The prep-aration at the top is the sediment from a centri-fuged extract, which shows very little of the pre-sumed haptenic component. Figure 5 is the same

TABLE 4. Chemical analysis of protoplasm, extracts of protoplasm, and extraction residues of Escherichia coli

PI eparation

ProtoplasmTrichloro-

aceticacidextractof proto-plasm

Residue

ProtoplasmTrichloro-

aceticacid ex-

tract ofproto-plasm

Residue

Recovery Nitrogen Phosphorus Hexose

100 14 2.2 3.7

4.384

100

6.374

3.515

12

2.614

* Symbols: + = nucleic acid present; 0

0.282.1

2.1

0.191.8

173.0

9.8

3312

Totalcarbohy-drate

10

3610

15

4616

free from nucleic acid.

TABLE 5. Endotoxic properties of protoplasm, extracts of protoplasm, and extraction residues ofEscherichia coli

Lethality NonspecificOrganismPreparation formice protpection Tumor damage Pyrogenicity ShwartzmanOrganism Preparation for mic protection (TD50) (FI40) (SPDbo)(Loss ) (Eoso) s)( n)(fl)

mg pgy pg g pg

E. coli Protoplasm >20 >200 140 380113 (Ec 14 Trichloroaceticseries) acid extract of

protoplasm > 200 > 250 > 200Residue > 20 > 200 > 1000 > 400

E. coli Protoplasm >20 >200 >500 300 >2000111 :B4 Trichloroacetic(Ec 10 acid extract ofseries) protoplasm > 200 > 200 > 500

Residue > 20 > 400 > 300 > 1000 > 400

Organism

E. coli0113(Ec 14series)

E. coli0111 :B4(Ec 10series)

Hexos-amine

2.0

181.3

2.2

10

0.84

Totalfattyacids

24

3625

22

8.320

Nucleicacid*

+

0

+

0 on March 25, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

8MILNER ET AL.

FIG. 8. Comparison in gel diffusion of trichloroacetic acid extract of protoplasm with whole protoplasm,acid haptens, endotoxins, and an inert polysaccharide from Escherichia coli. Key: A , E, F, and J, endotoxinsextracted fromt whole cells with trichloroacetic acid; B, inert polysaccharide obtained by reacting endotoxinwith lithium alumttinumt hydride; C and H, acid haptens; D, aluminum citrate-endotoxin complex; G, wholeprotoplasm; 1, trichloroacetic acid extract of protoplasm ("native hapten").

plate a few days later, and merger of the innerbands appears to have taken place at the lowerleft. When the same antigens were set up in thesame pattern with another antiserum, the haptenbands merged all around, as illustrated in Fig. 6.These unexplained differences emphasize the needfor examining many preparations and interpret-ing the results of gel diffusion studies conser-vatively. Although there was hapten in ex-tracts from washed cell walls, it seemed to bein smaller quantity than in extracts of wholecells. Therefore, the protoplasm was also exam-ined in similar fashion.Most procedures for breaking cells produce

enough dissolved endotoxin or fine fragmentsof cell walls to contaminate the protoplasmbeyond hope of separation. The success of thenext experiment depended upon an initiallyclean separation, which was achieved with amodified pressure cell. Portions of the protoplasmwere then extracted with trichloroacetic acid in

the cold and centrifuged; dialyzed supernatantswere lyophilized. Chemical data from two suchexperiments are given in Table 4. The majorpoints of interest are that the extracts wsereobtained in yields of 4.3 and 6.3'S and that eachcontained considerably less nitrogen, but morehexose, hexosamine, and total carbohydrate,than the starting material. All of these fractionswere essentially without endotoxic activity (Table5). With few exceptions, none of the host re-sponses was elicited by the highest dose tested.The trichloroacetic acid extract of proto-

plasm was tested by the Ouchterlonv gel diffu-sion method (Fig. 7). It is at the right ofthe serum well, with a trichloroacetic acidendotoxin at the top and an acid hapten at theleft. The reaction of identity seems clear amongthe fast-migrating components. A more complexpreparation (Fig. 8) gives some further informa-tion. The whole protoplasm (G) shows a majorand a minor component, both merging with the

35)8 BA\ClTERIOL. REV.

on March 25, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

VOiL. 27, 1963 SURFACE ANTIGENS FROM GRAM-NEGATIVE BACTERIA

FIG. 9. Ultracentrifugation of trichloroacetic acidextract of protoplasm from Escherichia coli. Solu-tion (1%) in distilled water, photographed 120min after operating speed was reached (50,740rev/min, 20 C). Sedimentation constant, 1.65 S.

known hapten lines. The significance of thisfinding is not yet clear. The trichloroacetic acidextract from the material in G (I) shows a singlecomponent that merges with hapten. In anultracentrifuge (Fig. 9), the extract had essen-tially the same sedimentation constant (1.65 S) asthe acid hapten or the hal)tenic component fromendotoxin; however, the single, very sharp peak,from a concentration of 1';", photographed 2 hrafter operating speed was reached, suggests aremarkable uniformity of particle size. The acidhapten (Fig. 1) showed a much broader peakunder the same conditions, indicating poly-dispersion. Paper-strip electrophoresis indicatedthat the extract from protoplasm and acid-hydrolyzed endotoxin contained essentially thesame sugars, but the analysis was admittedlyincomplete. In Fig 10, the infrared spectrum of atrichloroacetic acid extract of protoplasm iscompared with that of an acid hapten and of acomplete endotoxin. The salient feature here isthe similarity of all three.

Thus, there appears to be a substantial amountof hapten in the toxin-free protoplasm of thesecells. This "native" hapten is an organizedmolecule of considerable size in which theterminal sugars of the chains must already havebeen linked in the final order; otherwise, pre-cil)itation with specific antibody would not be

expected to occur. It seems reasonable to assumethat such a hapten, once formed, could then betransported to the cell wall where enzymatic,or even purely physical-chemical, processes couldorganize it into the larger complexes which aretoxic and fully antigenic. That alternativemechanisms may exist is inferred from the reportof Osborn et al. (5) that a cell envelope fractionof E. coli contained enzymes capable of catalyz-ing the incorporation of galactose and glucoseinto the deficient polysaccharides of mutantstrains when the necessary pyrimidine nueleo-sides were provided.

MICELLAR STRUCTURE IN CELL WALLS

In turning from data which seem to implythat endotoxic 0 antigens of the cell wall arepolymers of haptenic units, synthesized in someregion of the protoplasm internal to the cell wallproper, to a broader view of the possible struc-ture of the final aggregate and its relation-ship with other components of the cell wall, one

,WAVELENGTH IN MICRONS

FIG. 10. Infrared absorption spectra of endotoxin,trichloroacetic acid extract of protoplasm, and acidhapten from Escherichia coli.

3.)9

on March 25, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

MILNER ET AL.

FIG. 11. Cell wall of Histoplasma capsulatum

enters an area where knowledge is fragmentaryand speculation is invited. Weidel and co-work-ers (21, 23) have presented evidence for a three-layered cell wall in which an inner mucopolypep-tide "R layer," constituting only about 10% ofthe total wall substance, is principally responsiblefor maintaining shape and rigidity. This R layer,which is the substrate for lysozyme and T2 phageenzyme, is depicted as composed of polypeptideballs held in a rigid network by comblike struc-tures in which teeth of peptide chains are fixedto a backbone of amino sugar. Overlying this and

clinging tightly to it is the lipopolysaccharidelayer containing the endotoxic 0 antigen andphage-receptor sites. External to both is a thickamorphous layer of lipoprotein. Although an Rlayer has recently been demonstrated in S.gallinaruan (22), the Weidel conception has beenderived chiefly from studies of one partiallyrough culture of E. coli. This picture is not,therefore, necessarily representative of all gram-negative bacteria, nor has it been universallyaccepted.The finding of an analogous mucopolymer

360 BACTERIOL. REV.

on March 25, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

SURFACE ANTIGENS FROM GRAM-NEGATIVE BACTERIA

FIG. 12. Cell walls of Salmonella enteritidis. From Ribi, Milner, and Perrine (12)

ground structure in a microbial cell wall was

reported several years ago from the RockyMountain Laboratory (11). Figure 11 shows thewall of a disrupted cell of the fungus Histo-plasma capsulatum. There is a vertical tear inthe empty wall exposing, at lower right andcenter, a triangular portion of the inner surfaceof the wall which is a distinctly fibrous network.Finely granular material clings to this fibrousbase on the outer side of the wall, but it could beprogressively removed by selective acid hydrol--sis and mechanical agitation, exposing the fibrousbase on that side also. The specific antigen of thecell wall was evidently in the granular layer,because antigen aI)I)eared in the washings as thislayer became solubilized (16). The clean fibrousmembranes were of sufficiently orderly struc-ture to give distinct X-ray diffraction patterns,which left no doubt that the material was, infact, chitin.

Figure 12 is the familiar electron microgral)h

of cell walls from S. enteritidis (12). The finestructure appears generally amorphous exceptfor occasional rodlike elements, as in the area toright of center at the extreme top. Of course, ifWeidel and colleagues are correct, the netlikeground layer, which in this case should correspondto the chitin layer of the fungus, might have beendisintegrated by enzymatic action before such apicture could have been taken (22). Vigorouswashing produces little change in such cell wallpreparations, but extraction with aqueous etheror other agents known to remove endotoxinproduces some noticeable thinning of the walls.If the results obtained with E. coli apply to S.enteritidis (and work in progress indicates thatthey do), then about 75 "I of the material depictedin Fig. 12 is certainly immunologically inert,and we do not yet know what proportion of theother quarter mayr also be inert. However, theextracted endotoxin is capable of being organizedinto more orderly structures, as illustrated in

361VO)L. 27, 1963

on March 25, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

MILNER ET AL.

FIG. 13. Endotoxin from Bordetella pertussis

Fig. 13 and 14. These are dried films of purifiedendotoxin extracted from cell walls or wholecells of Bordetella pertussis by the phenol-waterprocedure (27; Malmgren, unpublished data).Schramm, Westphal, and Luderitz (17) andWeidel, Frank, and Martin (23) previously showedpictures of endotoxins in partly organizedstructures described as rodlets, strings of pearls,or sausages. In these electron micrographs, wesee a number of fibers of various lengths amid apreponderance of shorter rodlike structures.

Diameters of all formed elements are in the rangeof 50 to 100 A, numbers having special sig-nificance as dimensions of colloidal particles (8).This suggested a parallel with the structure of cel-lulose and of synthetic fibers (6, 7, 13). Of course,the parallel is far from exact, but it will serve tointroduce a concept which has been only brieflyreferred to elsewhere (10).

Cellulose consists entirely of glucose units puttogether by glycosidic linkages. These chains,furthermore, tend to orient themselves laterally

362 BACTERIOL. REV.

on March 25, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

VOL. 27, 1963 SURFACE ANTIGENS FROM GRAM-NEGATIVE BACTERIA

FIG. 14. Endotoxin from Bordetella pertussis showing fibrils.

in fibrils that have a definite structure and co-hesiveness as diagrammed in Fig. 15 (3). Thepolymer chains are bundled into elementaryfibrils which are found to have a constant diam-eter in the neighborhood of 50 to 100 A. In thedistribution of random chain lengths, it is be-lieved, some areas have the regular structure ofa erystal lattice and these crystalline areas areseparated by relatively amorphous areas. Uponhydrolysis the fibers come apart at the amor-phous areas and form the micelles 50 to 100 Ain diameter and 500 to 600 A in length shown inFig. 16 (14). Here, in a number of instances,one may see how the longer fibers are put to-gether from smaller units. The "molecularweight" of the individual rodlets (micelles) is ofthe order of a few millions.To be sure, cellulose has a more orderly struc-

ture than any postulated for endotoxin, but thesematerials resemble one another in that both are

built from chain molecules. Chain moleculeswith a tendency to form micelles may organizeinto electron microscopically amorphous gels aswell as into fibrils of uniform thickness. Figure 17(4) is a diagram of one conception of how mi-celles may be found in a hydrated condition in gels.We believe that the natural state of endotoxicO antigen in a bacterium is a gel and that "ele-mentary fibrils" are obtained only upon manip-ulation. The important point is that the differ-ences between these forms are occasioned notso much by the primary organization of chainmolecules as by the further orientation of mi-celles, which are common to both gels and fibers.In Fig. 18 (MIalmgren, unpublished data) we seethreadlike structures formed from otherwiseamorphous endotoxin, apparently in response toconditions of drying in a given area of a film.This concept of a micellar gel composed ofheterogeneous macromolecules, apparently amor-

363

on March 25, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

MILNER ET AL.

I/ '1 DFIG. 15. Schematic representation of crystallite

structure of cellulose (fringe micellar theory).From Howsmon and Sisson (3).

phous in its natural state, and lying external to,or partly in the interstices of, a rigid supportingmembrane, affords a striking parallel to some ofthe so-called hemicelluloses of plant cell walls.To proceed with the analogy requires great

oversimplification, but perhaps the gist of theargument is not farfetched. The hapten of S.enteritidis is known to consist largely of three

common hexose sugars, a methyl pentose, atleast one heptose, a dideoxyhexose, and hexosa-mine (26). Since the sugars do not fully accountfor the analysis and since phosphorus and somefatty acids are always found, it may be assumedthat these, and perhaps additional components,also have a place in the molecule. But, sincesugars are the major constituents, one may bejustified in thinking of the hapten as fundamen-tally similar to a chain of about 100 hexose units.This would give a molecular weight of justunder 20,000 and a length of about 500 A. A groupof 50 or 100 of these chains bundled together orlinked end-to-end and folded would constitutea colloidal particle of the size known to exist inseveral micellar systems and would also be anaggregate of the weight and dimensions that wehave frequently observed to be the smallestparticle with complete endotoxic and O-antigenicproperties. The fact that endotoxin is known toexist also in much larger particles, very irregularin size, would only mean that, in many prepara-tions, the fundamental micelles are not freedfrom the larger organizations (e.g., gels or fibrils)of which they may also be a part in situ or intowhich they may become aggregated at some stepof preparation.

If hapten is synthesized in the protoplasm asdiscrete chain molecules of rather uniform size(and some evidence to this effect has been pre-sented), then the problem of how complete anti-gen is formed from these parts may not be overlycomplicated. Enzymes may well be involved,but some transport mechanism for selectivelybringing hapten molecules toward the cell sur-face may be of greater significance, since theseunits may have an innate ability to combine,under the altered circumstances, in the mannerindicated earlier. One of us (E. Ribi, unpublisheddata) has constructed a laboratory model todemonstrate some of the unusual capacities ofchain molecules. Polyaminocaproic acid wasobtained by polymerization of the simple aminoacid. A drop of phenol containing such chainmolecules in a monodispersed state was depositedat the bottom of a container of water. As thephenol dissipated into the water phase, thewater-insoluble polymer diffused onto the watersurface forming the membrane depicted inFig. 19. Here a single amino acid composes astructure bearing strong points of resemblanceto both the chitinous cell-wall skeleton of a

364 BACTERIOL REV.

on March 25, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

VOL. 27, 1963 SURFACE ANTIGENS FROM GRAM-NEGATIVE BACTERIA

FIG. 16. Hydrolyzed cellulose showing elementary fibrils and micelles. From Ribi and Rdnby (14).

fungus depicted earlier and the R layer of bac-teria as conceived by Weidel and colleagues.Polypeptide balls form both a granular layer anda network of fibrils 50 to 100 A in diameter. Thus,simple colloidal-physical phenomena may accountfor some apparently complex reorganizations ofmatter.

Several theoretical implications of the viewof antigen synthesis presented here are susceptibleto experimental verification as soon as the nec-essary materials can be prepared. Attention isdirected to only one of these. The hapten of S.

enteritidis is known to contain at least five anti-genic determinants (the Kauffmann-White fac-tors 1, 9, 12,, 122, and 123), and the chemicalend groups responsible have been substantiallyidentified by the investigative teams associatedwith Staub and Westphal (19, 25). A straight-chain molecule has two ends, accounting foronly two determinants; therefore, a considerabledegree of branching has been assumed. But itmight be expected that a highly branched haptenmolecule would form a highly cross-linked poly-mer and thus a less-soluble material than many

365

on March 25, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

MILNER ET AL.

endotoxic 0 antigens are found to be. If haptenwere composed of straight chains, each contain-ing all the groups, one could suppose that thechains ended in different places, thus exposing allthe determinants in different chains in a solutionof hapten or at the ends of each micelle in a prep-aration of complete antigen. However, the avail-able evidence indicates that, even in acid- oralkali-treated endotoxin, the 9 and 12 componentsare contained on one molecule (1, 18). It will,

FIG. 17. Schematic representation of fringedmicelles in gel form. From Kratky and Porod (4).

therefore, be of great interest to determinewhether native hapten from the protoplasm andhaptens produced by hydrolysis from extractedantigens behave similarly in quantitative pre-cipitation tests with single-factor antisera.

CONCLUSIONS

For the present, the ability of extracted endo-toxins to form fibrils of certain dimensions andthe physical-chemical relationships betweenendotoxin and hapten suggest the possibility ofmicellar structure. Moreover, the concept ofendotoxins as aggregations of hapten into micellesand gels accounts very well for reported observa-tions on the particle sizes of endotoxins and oftheir hydrolytic and enzymatic degradation prod-ucts. Although the mechanism for the funda-mental injury produced by endotoxins remains

FIG. 18. Endotoxin from Bordetella pertussis. Effect of drying.

366 BACTERIOL. REV.

on March 25, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

VOL. 27, 1963 SURFACE ANTIGENS FROM GRAM-NEGATIVE BACTERIA

FIG. 19. Model membrane of polyaminocaproic acid

too speculative for discussion at this time, theexistence in mammalian hosts of enzymes whichare able to degrade these highly organized materi-als to their haptenic elements (15, 20) has sug-gested that it may be this very organizationwhich is fundamental to elicitation of theirbiological activity.The finding of what has been referred to as

"native hapten," apart from endotoxin in anenvironment where genetically controlled syn-thesis might readily take place, provides a new

material for study, one which may facilitatedemonstration of the final stages in the biosyn-thesis of endotoxin.

LITERATURE CITED

1. COHEN, H. H. 1957. Mutual linkage betweenthe components IX and XII in Salmonellatyphi 0-antigen. J. Immunol. 81:445-451.

2. FUKUSHI, K., R. L. ANACKER, W. T. HASKINS,M. LANDY, K. C. MILNER, AND E. RIBI.1963. Chemical and biological comparison

367

on March 25, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

MILNER ET AL.

of endotoxins extracted from different cul-tures of Enterobacteriaceae. Bacteriol.Proc., p. 55.

3. HOWSMON, J. A., AND W. A. SISSON. 1954.Structures and properties of cellulose fibers.B. Submicroscopic structure, p. 231-346. InE. Ott, H. M. Spurlin, and M. W. Grafflin[ed.], Cellulose and cellulose derivatives,part 1. Interscience Publishers, Inc., NewYork.

4. KRATKY, O., AND G. POROD. 1955. Form undGrosse der kristallinen Bereiche, p. 192-232.In H. A. Stuart [ed.], Die Physik der Hoch-polymeren, vol. 3. Springer-Verlag, Berlin.

5. OSBORN, M. J., L. ROTHFIELD, S. M. ROSEN,AND B. L. HORECKER. 1963. Biosynthesis ofthe cell wall lipopolysaccharide of Salmo-nella typhitnurium. Bacteriol. Proc., p. 117.

6. RANBY, B. G., AND E. RIBI. 1950. Uber denFeinbau der Zellulose. Experientia 6:12-14.

7. RIBI, E. 1950. Uber den Feinbau der Vis-coseseide. Arkiv Kemi 2:551-566.

8. RIBI, E. 1951. Submicroscopic structure offibres and their formation. Nature 168:1082-1083.

9. RIBI, E., W. T. HASKINS, M. LANDY, ANDK. C. MILNER. 1961. Symposium on bac-terial endotoxins. I. Relationship of chemi-cal composition to biological activity.Bacteriol. Rev. 25:427-436.

10. RIBI, E., W. T. HASKINS, K. C. MILNER,R. L. ANACKER, D. B. RITTER, G. GOODE,R.-J. TRAPANI, AND M. LANDY. 1962. Physi-cochemical changes in endotoxin associatedwith loss of biological potency. J. Bacteriol.84 :803-814.

11. RIBI, E., B. H. HOYER, AND G. GOODE. 1957.The fine structure of the cell wall of Histo-plasma capsulatum as revealed by electronmicroscopy and X-ray diffraction studies.Bacteriol. Proc., p. 105.

12. RIBI, E., K. C. MILNER, AND T. D. PERRINE.1959. Endotoxic and antigenic fractionsfrom the cell wall of Salmionella enteritidis.Methods for separation and some biologicactivities. J. Immunol. 82:75-84.

13. RIBI, E., AND A. NORLING. 1954. Elektronen-mikroskopische und rontgenographischeStudien an Superpolyamidfaiden. Beziehun-gen zwischen dem Feinbau der Superpoly-amidfaden und dem der Viscoseseide. ArkivKemi 7:417-426.

14. RIBI, E., AND B. G. RANBY. 1950. Zur elek-tronenmikroskopischen Praparation vonKolloiden. Experientia 6:27.

15. RUDBACH, J. A., AND A. G. JOHNSON. 1962.Changes in serologic reactivity of endotoxininduced by fraction IV, (Cohn) of normal

human serum. Proc. Soc. Exptl. Biol. Med.111:651-655.

16. SALVIN, S. B., AND E. RIBI. 1955. Antigensfrom yeast phase of Histoplasniia capsulatum.II. Immunologic properties of protoplasmvs. cell walls. Proc. Soc. Exptl. Biol. Med.90:287-294.

17. SCHRAMM, G., 0. WESTPHAL, AND 0. LtDERITZ.1952. Uber bakterielle Reizstoffe. III.Physikalisch-chemisches Verhalten eineshochgereinigten Coli-pyrogens. Z. Natur-forsch. 7b:594-598.

18. STAUB, A. M., AND G. PON. 1956. Etude im-munochimique sur les salmonelles. 1. Polyo-sides extraits de S. typhi, S. para typhi B etS. typhi m zriunt. Ann. Inst. Pasteur 90:441-457.

19. STAUB, A. M., R. TINELLI, 0. Lt-DERITZ, AND0. WESTPHAL. 1959. Etude immunochimiquesur les Salmonella. V. Role de quelquessucres, et en particulier des 3-6 did6soxy-hexoses, dans la specificite des antigeinesO du tableau de Kauffmann-White. Ann.Inst. Pasteur 96:303-332.

20. TRAPANI, R.-J., W. F. WARAVDEKAR, M.LANDY, AND M. J. SHEAR. 1962. In vitro in-activation of endotoxin by an intracellularagent from rabbit liver. J. Infect. Diseases110:135-142.

21. WEIDEL, W. 1951. Uber die Zellmembran vonEscherichia coli B. I. Praparierung derMembranen. Analytische Daten. Morpholo-gie. Verhalten der Membranen gegenuiberden Bakteriophagen der T-Serie. Z. Natur-forsch. 6b:251-259.

22. WEIDEL, W., H. FRANK, AND W. LEUTGEB.1963. Autolytic enzymes as a source of errorin the preparation and study of Gram-negative cell walls. J. Gen. Microbiol.30:127-130.

23. WEIDEL, W., H. FRANK, AND H. H. MARTIN.1960. The rigid layer of the cell wall ofEscherichia coli, strain B. J. Gen. Microbiol.22:158-166.

24. WESTPHAL, 0. 1957. Pyrogens, p. 115-220. InF. Springer [ed.], Polysaccharides in biology.Trans. 2nd Macv Conference. MadisonPrinting Co., Madison, N.J.

25. WESTPHAL, 0. 1960. Recentes recherches surla chimie et la biologie des endotoxines desbact6ries a Gram n6gatif. Ann. Inst. Pasteur98:789-813.

26. WESTPHAL, O., AND 0. LUDERITZ. 1961. Chemiebakterieller O-Antigene. Pathol. Microbiol.24:870-889.

27. WESTPHAL, O., 0. LtDERITZ, AND F. BISTER.1952. UVber die Extraktion von Bakterien mitPhenol/Wasser. Z. Naturforsch. 7b:148-155.

368 BACTERIOL. ]REV.

on March 25, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from