0019-9567/78/0019-0265$02.00/0INFECTION AND IMMUNITY, Jan. 1978, p. 265-271Copyright © 1978 American Society for Microbiology

Vol. 19, No. 1Printed in U.S.A.

Immunoelectron Microscopic Localization ofLipopolysaccharides in the Cell Wall of Bacteroides oralis and

Fusobacterium nucleatumG. DAHLEN,1 H. NYGREN,l* AND H.-A. HANSSON2

Department of Oral Microbiology, Institute ofMedical Microbiology,' and Department of Histology,2University of Goteborg, Goteborg, Sweden

Received for publication 8 June 1977

Lipopolysaccharides (LPS) have been extracted and purified from two anaer-

obic gram-negative bacteria: Bacteroides oralis and Fusobacterium nucleatum.Chemical analysis of the preparations showed a great proportion of neutralsugars, mainly glucose, in LPS of B. oralis. In rabbits, LPS of B. oralis inducedboth immunoglobulin M and G antibodies in contrast to LPS of F. nucleatum,to which only immunoglobulin M antibodies were produced. An immunohisto-chemical method with horseradish peroxidase-labeled antibodies was used tolocalize LPS antigens at the ultrastructural level. An electron-dense reactionproduct, representing an immune complex consisting of bacterial surface antigensand specific rabbit immunoglobulin labeled with peroxidase, was surroundingthe cell wall, whereas appropriate controls were negative. The findings of thepresent study show that LPS of Bacteroides are probably bound to a complex,including glucans, in the outer membrane of the cell wall. LPS of Fusobacteriumresemble LPS of other gram-negative bacteria.

The outer membrane of the cell wall in gram-negative microorganisms consists of a macro-molecular complex of lipopolysaccharides (LPS)with biological activities collectively known asendotoxic. The anaerobic gram-negative bacte-ria Bacteroides and Fusobacterium have gath-ered increasing interest as possible pathogenicorganisms in many infections (2). These bacteriaare frequently found in association with manyoral infections, e.g., in periodontal disease (25)and periapical inflammation of teeth (G.Sundqvist, Ph.D. thesis, Umea University,Umea, Sweden, 1976), in which endotoxin mayparticipate as a pathogenetic factor. Bacteroidesstrains differ from other gram-negative bacteria,including Fusobacterium, in their chemical com-position of LPS. They do not contain heptoseor 2-keto-3-deoxyoctulosonic acid (KDO) (13);they have no typical lipid A (T. Hofstad, per-sonal communication); and the endotoxic activ-ity is low (12; G. Dahlen and T. Hofstad, Scand.J. Dent. Res., in press). Another difference con-cerning LPS of Bacteroides strains in contrastto Fusobacterium has appeared to be the con-tent of glucans in LPS preparations, when ex-traction with the phenol-water method is used(11). By immunoelectron microscopic technique,it should be possible to obtain an indicationwhether these glucans occur in the cytoplasmor as a part of the cell wall. In the present study

(i) the chemical composition, (ii) the immuno-genicity, and (iii) the ultrastructural localizationof LPS in the cell wall of Bacteroides and Fu-sobacterium were investigated.

MATERIALS AND METHODSBacterial strains. Bacteroides oralis (strain Bact-

MC3) and Fusobacterium nucleatum (strain Fus-MC8), originally isolated from the root canal of amonkey tooth, were used.

Cultivation. The culture medium was describedby Dahlen and Hofstad (in press). The cultivationwas carried out in a fermentor with stirring at aconstant pH 7.2. The cells were harvested at the endof the logarithmic phase.

Isolation of LPS. LPS were extracted with 45%phenol for 15 min at room temperature (11) frompacked wet cells of strain Bact-MC3 and from crushed,defatted, and buffer-extracted cells of strain Fus-MC8(18). After dialyzing against water, the LPS prepara-tions were purified by centrifugation at 100,000 x g for1 h, treated with deoxyribonuclease (DNase, 0.05mg/ml, pH 7.0, Sigma Chemical Co., St. Louis, Mo.)and ribonuclease (RNase, 0.05 mg/ml, pH 7.0, Sigma),recentrifuged, and lyophilized.Chemical analysis. The procedures for paper

chromatography and quantitative estimations for pro-tein, carbohydrates, fatty acid esters, glucosamine, andKDO content were carried out according to Dahlenand Hofstad (in press).

Serological method. For estimation of specificantibody activity, indirect hemagglutination (HA) and

265

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/i

ai o

n 24

Jan

uary

202

2 by

179

.27.

73.2

34.

266 DAHLRN, NYGREN, AND HANSSON

inhibition of hemagglutination were used, accordingto the modification described by Hofstad (14).

Preparation and conjugation of antibodies.Rabbits were injected intravenously with.1 ml of anLPS solution from strain Bact-MC3 or Fus-MC8 inincreasing concentrations from 0.1 to 1.5 mg/ml threetimes a week in 3 weeks, and were then bled 1 weekafter the last injection. The serum was collected andfractioned by gel chromatography (Sephadex G-200,Pharmacia, Uppsala, Sweden). Each fraction was an-alyzed for specific anti-LPS antibodies. Pooled frac-tions were concentrated 10 times against 30% polyeth-ylene glycol. Immunoglobulin G and M (IgG; IgM)antibodies to Bact-MC3 LPS and IgM antibodies toFus-MC8 LPS were conjugated with peroxidase. Theantibody solution (200,ul) was mixed under continuousshaking with 0.25% glutaraldehyde (25 pl) and horse-radish peroxidase (12.5 mg, type VI, Sigma) at roomtemperature for 45 to 90 min (1). The antibodiesconjugated with peroxidase were separated from freeperoxidase on a Sephadex G-100 column (Pharmacia).The fractions containing protein were identified bytheir adsorption at 280 and 405 nm. The specificantibody activity after conjugation was found in pilotstudies to decrease to about 5% of the initial activity.

Fixation of bacteria. The bacterial cells wereseparated from the culture medium by centrifugationand were fixed in a cacodylate-buffered solution (0.1M, pH 7.2) containing either 4% freshly preparedformaldehyde with 0.1% picric acid (10) or 3% purifiedglutaraldehyde. After aldehyde fixation, the bacteriawere rinsed in 0.15 M cacodylate buffer for 12 h. Thebacteria used in the morphological experiments werepostfixed in 2% osmium tetroxide in Veronal-acetatebuffer directly after the glutaraldehyde fixation.Immunohistochemical analysis. The bacteria,

either freshly isolated or fixed in aldehyde only, asdescribed above, were incubated for 20 to 30 min inperoxidase-conjugated antibodies diluted 1:10 to 1:200with 0.15 M cacodylate buffer, rinsed thoroughly, andincubated in cacodylate buffer containing 0.05% 3.3-diaminobenzidine-0.01% hydrogen peroxidase for 45min. After rinsing, the bacteria were postfixed for 1 hin 2% osmium tetroxide, dehydrated in ethanol, andembedded in Epon. Control incubations comprised (i)Bact-MC3 cells in anti-Fus-MC8 LPS antibodies con-jugated with peroxidase and (ii) Fus-MC8 cells in anti-Bact-MC3 LPS antibodies conjugated with peroxi-dase. Furthermore, cells ofboth strains were incubatedwith (iii) goat anti-rabbit immunoglobulins conjugatedwith peroxidase, (iv) antiserum adsorbed with homol-ogous antigen after conjugation, (v) peroxidase, and(vi) diaminobenzidine without preincubation with an-tibodies.

Electron microscopy. Sections were cut on anLKB Ultrotome III ultramicrotome and examined ina Siemens Elmiskop IA electron microscope. Un-stained sections and sections stained with uranyl ace-tate and lead citrate were used.

RESULTSChemical analysis. LPS from Bact-MC3

contained 2.6% protein, 57.6% carbohydrates,5.5% fatty acid esters, and 2.0% hexosamine.

KDO was lacking. The main sugar was glucose,but small amounts of glucosamine, galactose,rhamnose, and ribose were also found. LPS fromFus-MC8 contained 5.9% protein, 29.7% carbo-hydrates, 22.7% fatty acid esters, 6.3% glucosa-mine, and 2.0% KDO. The main sugars wereheptose, glucose, glucosamine, and rhamnose.Immunogenicity. Gel chromatography of

antiserum revealed two separate peaks for titersof Bact-MC3 LPS (Fig. 1), on the basis of thehemagglutination activity of each fraction. Thepeaks corresponded to the elution pattern ofIgM and IgG antibodies, respectively. By con-trast, LPS of Fus-MC8 gave only an IgM re-sponse, as indicated by a single peak in Fig. 2.The antibody titers of the pooled- and concen-

trated-antibody fractions showed no significantcross-reaction to LPS between Bact-MC3 andFus-MC8 antibodies by the hemagglutinationtest or by hemagglutination inhibition.Ultrastructural localization of LPS. The

morphology of the outer membrane of the cellwall of Bact-MC3 showed the triple-laminarstructure typical of gram-negative bacteria (Fig.3a). In Fus-MC8, the outer membrane of the

!80 t1,

E- jbx'0~~~~~~~~~1 0 20310

ra c t o ,l ,lu n b e r

FIG. 1. Fractionated rabbit antiserum againstLPS from B. oralis strain Bact-MC3. The antibodytiter of each fraction (bars) determined with indirecthemagglutination is plotted together with the proteinconcentration curve (A2so).

64

3 2

1 6

84

2

FIG. 2. Fractionated rabbit antiserum againstLPS from F. nucleatum strain Fus-MC8. The anti-body titer of each fraction (bars) determined withindirect hemagglutination is plotted together withthe protein concentration curve (A24.o.

INFECT. IMMUN.

A__2

2 5 6

2 8

6 4

3 2

6

8

4

2

40

I10

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/i

ai o

n 24

Jan

uary

202

2 by

179

.27.

73.2

34.

LPS IN BACTEROIDES AND FUSOBACTERIUM

.*-%.N ..- . , ,4*S s - i - .-f ~~~~~.,-. --!

t.

-+ , :. I *',,.

I

3a l 3bFIG. 3. (a) Electron micrograph of cells of B. oralis. The trilaminated outer membrane of the cell wall

(arrow) and the cytoplasmic membrane enclose the granular and filamentous structures of the cell. Theinner cell waU is very thin and can only be seen in some areas (arrow head). Counterstained with uranylacetate and lead citrate. x160,000. (b) Electron micrograph of cells of F. nucleatum counterstained withuranyl acetate and lead citrate. The cell waU consists of a trilamellar outer membrane (arrow) that isseparated from the electron-dense inner ceU wall and the cytoplasmic membrane proper (arrow head).Numerous granules, vesicles, and filaments are seen within the cytoplasm. x200,000.

cell wall was distinctly stained by uranyl acetateand lead citrate (Fig. 3b). The outer membraneof the cell wall stripped off easily and piled upin double layers, or it was completely lost. Noconclusive difference in- ultrastructure of theouter membrane was found between the twostrains. The inner lamina of the cell wall ofBact-MC3, however, was thinner and less elec-tron dense than that of Fus-MC8. Incubation ofhomologous cells in anti-Bact-MC3 LPS, con-jugated with peroxidase, resulted in a granularirregular staining of the outer membrane of thecell wall (Fig. 4a). In some sites, the immunereaction produced larger dots, with a cross-sec-tional area 10 to 100 times that of the smallgranules (Fig. 4a, b, and c). The immunohisto-chemical reaction was restricted to the mem-brane of the cell wall. No other part of the cellreacted.

After formaldehyde fixation, the immunohis-tochemical reaction was diminished (Fig. 4b)and was almost abolished by glutaraldehyde fix-

ation (Fig. 4c). The IgG antibodies to Bact-MC3LPS (Fig. 4d) showed a distribution similar tothe IgM antibodies (Fig. 4a). In Fig. 5a, theouter membrane of the cell wall of an unfixedcell is seen together with the immunohistochem-ical reaction product, which is irregularly dis-tributed. The reaction is restricted to the outermembrane of the cell wall. Glutaraldehyde fix-ation (Fig. 5b) almost abolished the reaction.Control incubations (see i through vi in Materi-als and Methods) resulted in very slight activity(Fig. 4e).

DISCUSSIONThe high glucose content in the LPS of Bac-

teroides has been considered to be a contami-nation of intracellular glucans (11, 16; Dahlenand Hofstad, in press). Attempts to prepare glu-cogen-free LPS from different Bacteroidesstrains with different extraction methods havefailed in the sense that a high glucose contentcould not be avoided (T. Hofstad, K. Sveen, and

VOL. 19, 1978 267..l.

Af! "'. I

4

/ ..

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/i

ai o

n 24

Jan

uary

202

2 by

179

.27.

73.2

34.

268 DAHLEN, NYGREN, AND HANSSON

e

S.k~ v-.-4., #

A'af o~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

SX,I

4a

r&. - ,P

-.

INFECT. IMMUN.

.0 O..44b

*4

bz4'

4'.

-a

t

I'I,,, .-

*

il

*16

a.0&tev

4

-"t

.i.

4d4.

4',l/ 4b

,.0

- I 4cFIG. 4. A-E.

a

4e

-R

IL

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/i

ai o

n 24

Jan

uary

202

2 by

179

.27.

73.2

34.

LPS IN BACTEROIDES AND FUSOBACTERIUM

C

A'O I W. 4 v. - &

5a

5bFIG. 5. (a) Electron micrograph of F. nucleatum incubated in peroxidase-conjugated IgM antibodies to

LPS and further processed for immunoelectron microscopy. The reactivity is limited to the outer membraneof the cell wall, and there are no reaction products to be observed in the interior of the bacteria cell. Notethe granular distribution of the reaction products indicating that parts of the covering LPS were strippedoff during the process of the bacteria. The three lamellar structures of the cell wall are seen through some ofthe granules (arrow). X70,000. (b) Electron micrograph of a glutaraldehyde-fixed cell of F. nucleatumincubated in peroxidase-conjugated antibodies to LPS of B. oralis. There is no significant reaction to beobserved in the cell wall. x60,000.

G. Dahl6n, Acta Pathol. Microbiol. Scand. Sect.B, in press). This glucan content of LPS Bact-MC3 may explain the diverse immunologicalpattern resulting in both IgM and IgG antibodyproduction observed in the present study. It iswell known that antibodies against somatic an-tigen of Salmonella species and Escherichiacoli in animals (20) and humans (8, 21, 24) are19S macroglobulins. Somatic antigen refers tothe polysaccharide-containing 0-antigen, whichis supposed to be the main antigenic determi-nant of LPS endotoxin. It might thus be arguedthat the LPS preparation of Bact-MC3 containstwo immunologically different substances and

does not correspond immunologically to theendotoxin of Enterobacteriaceae and Fusobac-terium. Recently Hofstad (15) has succeeded inseparating LPS from Bacteroides fragilis intotwo different types responsible for agglutinatingand precipitating antibodies, respectively. It isnot yet clear whether or not these two typeshave different chemical compositions with re-spect to their glucose content. Both IgM andIgG antibodies conjugated with peroxidase werefound to be attached to the outer membrane ofthe cell wall of Bact-MC3 cells (Fig. 4c and d).This fmding indicates that, even if an antigeniccontamination is responsible for the diverse im-

FIG. 4. (a) Cells of B. oralis incubated in peroxidase-conjugated IgM antibodies to LPS. There is noreaction to be observed in this except in the outer cell wall. (Unfixed.) Note the granular distribution of thereaction products and the larger dots (arrow). x 70,000. (b) A cell of B. oralis fixed in formaldehyde beforeincubation with IgM antibodies to LPS as described in (a). There is a decrease in the reactivity indicatingpartial loss of antigenic activity due to the Formalin fixation. There is no reactivity in the cytoplasmicmembrane or in the cytoplasm. Note the larger dot of antibody reaction (arrow). x70,000. (c) Electronmicrograph ofa glutaraldehyde-fixed cell ofB. oralis incubated inperoxidase conjugated with IgM antibodiesto LPS. There is a faint reaction to be observed in the outer membrane of the cell wall. x 70,000. (d) Electronmicrograph of B. oralis freshly incubated in peroxidase-conjugated IgG antibodies to LPS. The distributionof reaction products is similar although less intense than that in (a). x 75,000. (e) Electron micrograph of aoralis cell treated with peroxidase-conjugated antibodies to LPS of F. nucleatum. There is hardly anyreactivity to be observed. x 75,000.

269VOL. 19, 1978

VP 48%, '4"1, r.s'

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/i

ai o

n 24

Jan

uary

202

2 by

179

.27.

73.2

34.

270 DAHLEN, NYGREN, AND HANSSON

munological response, it originates from thesame outer membrane as the LPS. This supportsthe theory that glucans are a part of the LPSmolecular complex.The cell wall of gram-negative bacteria con-

sists of a number of distinct layers with differingchemical compositions. The outer membrane,which contains LPS, appears to be triple layeredwhen examined with the electron microscope(6). Various Bacteroides species have beenfound to have this triple-layered outer mem-brane (4, 5, 27). The present study confirms thatthe cell walls of B. oralis and F. nucleatum alsohave these structures.The staining of the LPS complex of Fus-MC8

was not constant along the whole-cell periphery.In spite of the fact that the cells were harvestedin the logarithmic phase of growth, the outermembrane of the cell wall was lost in someareas. The explanation for this difference in sta-bility of the cell wall between strains Bact-MC3and Fus-MC8 may be the chemical differencesin LPS composition, but more probably it maybe the result of an artifact produced by centrif-ugation; Fus-MC8 is a rather filamentous cell,which possibly makes it more sensitive to me-chanical forces than the more coccoid Bact-MC3.

Location of the 0 antigen-containing LPS inthe cell wall of gram-negative bacteria has beenstudied by other methods (9, 22, 23, 26). Utilizingthe peroxidase-conjugated method of other stud-ies for localization of bacterial surface antigens(3, 19), the location of LPS in the outer cellmembrane was confirmed for both Bacteroidesand Fusobacterium. No immunohistochemicalreaction was observed inside the cells or in theinner parts of the cell wall. This does not excludethe possibility of antigenic sites inside the cellthat could not be demonstrated because theconjugated immunoglobulins could not pene-trate through the cell wall. However, it shouldbe noted that no reaction product was seenwithin cells undergoing lysis where penetrationshould have been facilitated. This indicates alow probability for the existence of intracellularantigens cross-reacting with LPS.The immunohistochemical reactions were

highly reproducible. The controls showed almostno reactivity. There was no cross-reaction be-tween the different antisera in either the indirecthemagglutination test or the immunohistochem-ical reaction. This demonstrates the high relia-bility of the method. Two different patternswere demonstrated with regard to the distribu-tion of the reaction product on Bact-MC3 cells.One of these, the fine, granular type, also ap-peared on the cell wall of Fus-MC8 (Fig. 4a and

5a). No difference was seen in the appearanceof the label when either peroxidase-conjugatedIgG or IgM antibodies against Bact-MC3 LPSwere used. The coarser, dotlike precipitates wereformed irrespective of the method of fixation.The reaction products were localized exclusivelyon the outer part of the trilaminated membraneof the cell wall. The appearance of the largerdots was reminiscent of the "blebs" on the cellwall of B. fragilis described by Kasper (17). Thegranular pattern of the reactive sites indicatesthat these sites are randomly distributed on thecell wall. Aldehyde fixation diminished the in-tensity of the immunohistochemical reaction,but the distribution was the same in freshlyisolated and aldehyde-fixed cells. This meansthat there was no redistribution due to lack offixation of the LPS. Edebo and Holme (7)showed that LPS are extracted from the cellwall even in saline and buffer solutions. In thefresh preparations used in the present study,LPS may have been partially extracted, but, asthe distribution of the reaction products was thesame in freshly prepared and fixed cells, thiswas not considered to be a significant source oferror.

ACKNOWLEDGMENTSThis work was supported by Goteborgs Tandliikare-Sal-

Iskap and by grants from the Faculty of Odontology, Goteborg,and Swedish Medical Research Council Project B 77-2543-09.

LiMRATURE CMD1. Avrameas, S. 1969. Coupling of enzymes to proteins with

glutaraldehyde. Use of conjugates for the detection ofantigens and antibodies. Immunochemistry 6:43-48.

2. Balows, A., R. M. Dehaan, V. R. Dowell, and L. B.Guze. 1974. Anaerobic bacteria. Role in disease.Charles C Thomas, Publisher, Springfield, Ill.

3. Berthold, P., D. Bratthall, and C.-H. Berthold. 1974.Immunoperoxidase staining of Streptococcus mutans.Arch. Oral Biol. 19:1227-1230.

4. Bladen, H. A., and J. F. Waters. 1963. Electron micro-scopic study ofsome strains ofBacteroides. J. Bacteriol.86:1339-1344.

5. Costerton, J. W., H. N. Damgaard, and K.-J. Cheng.1974. Cell envelope morphology of rumen bacteria. J.Bacteriol. 118:1132-1143.

6. Costerton, J. W., J. M. Ingram, and K.-J. Cheng.1974. Structure and function of the cell envelope ofgram-negative bacteria. Bacteriol. Rev. 38:87-110.

7. Edebo, L., and T. Holme. 1965. Preparation of biologi-cally active fractions of Salmonella typhimurium. ActaPathol. Microbiol. Scand. 63:228-234.

8. Evans, R. T., S. Spaeth, and S. E. Morgenhagen.1966. Bacterial antibody in mammalian serum to obli-gatorily anaerobic gram-negative bacteria. J. Immunol.97:112-119.

9. Fukushi, K., F. Ariji, J. Yamaguchi, and S. Oka.1966. Electron microscopic localization of endotoxin ofEscherichia coli by the use of ferritin-conjugated anti-body. J. Electron Microsc. 15:137-142.

10. Haglid, K., A. Hamberger, H.-A. Hansson, H. Hyden,L. Persson, and L. Ronnback. 1976. Cellular and

INFECT. IMMUN.

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/i

ai o

n 24

Jan

uary

202

2 by

179

.27.

73.2

34.

LPS IN BACTEROIDES AND FUSOBACTERIUM

subcellular distribution of the S-100 protein in rabbitand rat central nervous system. J. Neurosci. Res.2:175-191.

11. Hofstad, T. 1968. Chemical characteristics ofBacteroidesmelaninogenicus endotoxin. Arch. Oral Biol.13:1149-1155.

12. Hofstad, T. 1970. Biological activities of endotoxin fromBacteroides melaninogenicus. Arch. Oral Biol.15:343-348.

13. Hofstad, T. 1974. The distribution of heptose and 2-keto-3-deoxyoctonate in Bacteroidaceae. J. Gen. Microbiol.85:314-320.

14. Hofstad, T. 1974. Antibodies reacting with lipopolysac-charides from Bacteroides melaninogenicus, Bacte-roides fragilis, and Fusobacterium nucleatum in serumfrom normal human subjects. J. Infect. Dis. 129:349-351.

15. Hofstad, T. 1976. Purification of the 0 antigen of Bacte-roides fragilis ss. fragilis NCTC 9343 from phenol-water extracts by gel filtration and chromatography onDEAE-cellulose and hydroxylapatite. Acta Pathol. Mi-crobiol. Scand. Sect. B. 84:229-234.

16. Hofstad, T., and T. Kristoffersen. 1970. Chemical char-acteristics of endotoxin from Bacteroides fragilisNCTC 9343. J. Gen. Microbiol. 61:15-19.

17. Kasper, D. L. 1976. The polysaccharide capsule of Bac-teroides fragilis subspecies fragilis: immunochemicaland morphologic definition. J. Infect. Dis. 133:79-87.

18. Kristoffersen, T. 1969. Immunochemical studies of oralfusobacteria. 3. Purification of a group reactive precip-itinogens. Acta Pathol. Microbiol. Scand. 77:447-456.

19. Lai, C.-H., M. A. Listgarten, and B. Rosan. 1975.Immunoelectron microscopic identification and locali-

zation ofStreptococcus sanguis with peroxidase-labeledantibody: localization of surface antigens in pure cul-tures. Infect. Immun. 11:193-199.

20. Landy, M., and W. P. WVeidanz. 1964. Natural antibodiesagainst gram-negative bacteria, p. 275-290. In M. Landyand W. Braun (ed.), Bacterial endotoxins. Rutgers Uni-versity Press, New Brunswick, N.J.

21. Michael, J. G., and F. S. Rosen. 1963. Association of"natural antibodies" to gram-negative bacteria with they-macroglobulins. J. Exp. Med. 118:619-626.

22. Milner, K. C., R. L. Anacker, K. Fukushi, W. T.Haskins, M. Landy, B. Malmgren, and E. Ribi.1963. Symposium on relationship of structure of micro-organisms to their immunological properties. III. Struc-ture and biological properties of surface antigens fromgram-negative bacteria. Bacteriol. Rev. 27:352-368.

23. Shands, J. W. 1966. Localization of somatic antigen ongram-negative bacteria using ferritin antibody conju-gates. Ann. N.Y. Acad. Sci. 133:292-298.

24. Smith, R. T. 1960. Immunity in infancy. Pediatr. Clin.North Am. 7:269-293.

25. Socransky, S. S. 1970. Relationship of bacteria to theetiology ofperiodontal disease. J. Dent. Res. 49:203-222.

26. Ushijima, T. 1970. Morphology and chemistry of thebacterial cell wall. 1. The location of mucopeptide inthe cell wall of Bacteroides convexus and its chemicalcomposition. Jpn. J. Microbiol. 14:15-25.

27. Ushijima, T., K. Ueno, S. Suzuki, and U. Kurimoto.1971. Morphology and chemistry of the cell wall. II.Sugar composition and location of 0-antigen in the cellwall of Bacteroides convexus. J. Electron Microsc.20:32-39.

271VOL. 19, 1978

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/i

ai o

n 24

Jan

uary

202

2 by

179

.27.

73.2

34.