,/ Clin Periodontol IWS: 25: 134-144Printed in Denmark . All right.t reserved

Copyright © Munksguard 1998

juuTof

Clinical perindontcioyy

Microbial complexes insubgingival plaque

S.S.Socransky\ A.D.Haffaiee\M.A.Cugini\ C.Smith^ andR. L.Kent Jr.2Departments of, 'Periodontology, ^Biostalistics,Forsyth Dental Cenier, Boston, MA, USA

Socran.sky SS. Haffajee AD. Cugini MA, Smith C, Kent Jr. RL: Microbialcomplexes in subgingival plaque. J Clin Periodontol 1998; 25: 134-144.© Munksgaard, 1998.

Abstract. It has been recognized for some time thai bacterial species exist incomplexes in subgingival plaque. The purpose of the present investigation was toattempt to define such communities using data from large numbers of plaquesamples and different clustering and ordination techniques. Subgingival plaquesamples were taken from the mesial aspect of each tooth in 185 subjects (meanage 51 ± 16 years) with (o^l60) or without («^25) periodontitis. The presenceand levels of 40 subgingival taxa were determined in 13.261 plaque samples usingwhole genomic DNA probes and checkerboard DNA-DNA hybridization. Clinicalassessments were made at 6 sites per tooth at each visit. Similarities between pairsof species were computed using phi coefficients and species clustered using an aver-aged unweighted linkage sort. Community ordination was performed using prin-cipal components analysis and correspondence analysis. 5 major complexeswere consistently observed using any of the analytical methods. One complexconsisted of the tightly related group: Bacteroides forsythus. Porphvromonasgingivalis and Treponema denticoht. The 2nd complex consisted of a tightly re-lated core group including members of the Fusohaderium nucleatum/periodoniicumsubspecies, Prevotella intermedia. Prevotella nigrescens and Peptostreptucoccusmicros. Species associated with this group included: Eubacterium nodatum.Campylobacter rectus, Campylobacter .showae. Streptococcus consteltatus andCampylobacter graellis. The 3rd complex consisted of Streptoeoeeus sanguis. S.oralis, S. mitls. S. gordonii and S. intermedius. The 4th complex was comprisedof 3 Capnocytophaga species, Campylobacter concisus, Eikenella eorrodens andActinobaciiius actinomycetemcomitans serotype a. The 5th complex consisted ofVeiltonelta parvula and Actitwmyces odontotylicus. A. actinotnycetemcomitans sero-type b, Selenomonas twxia and Actinomyces naeslundii genospecies 2 {A. viscosusjwere outliers with little relation to each other and the 5 major complexes. The1st complex related strikingly to clinical measures of periodontal disease particu-larly pocket depth and bleeding on probing.

Key words: microbiology; bacteria; subgingivalplaque; ecology: periodontal

Accepted for publication 14 May 1997

The complexity of the subgingivalmicrobiota has been recognized sincethe 1st microscopic examination of thisecosystem by Van Leeuwenhoek in 1683(Tal, 1980). Since that time, numerousstudies have evaluated the compositionof plaque using light and electronmicroscopy, cultural techniques andmore recently immunologic or DNAprobe techniques. All techniques re-inforce Van Leeuwenhoek's initial ob-servation that subgingival plaques arecomprised of a large complex mixtureof bacterial species. Indeed it has beenestimated that 400 or more species re-

side in this area. Cursory examinationof a series of plaque samples suggeststhat there is little order in the micro-biota that colonizes gingival sulci orperiodontal pockets. However, examin-ation of sections of plaque by light andelectron microscopy indicates a surpris-ing degree of order in colonization pat-terns. For example, early supragingivalplaque demonstrated columnar ar-rangement of morphologically distinctbacterial species from the tooth surfaceto the outer surface of the plaque(Listgarten et al, 1975), Subgingivalplaque was frequently characterized by

a zone of gram negative and/or motilespecies located adjacent to the epitheliallining of the pocket while gram positiverods and cocci appeared to be forminga tightly adherent band of organisms onthe enamel or root surface (Listgarten1976, Listgarten 1994).

Cultural, immunologic or DNAprobe assessments of plaque have dem-onstrated that certain species frequentlyoccur together in subgingival plaquesamples. For example, Porphyromonasgingivalis is almost always observed insamples that arc harboring Bacteroidesforsythus. It has been speculated that B.

Mierobial complexes 135

forsythus in some fashion precedescolonization by P. gingivalis since B.forsythus is detected more frequently byitself (Gmur et al. 1989). Other com-plexes that have been observed includeP. gingivalis and Treponema denticola(Simonson et al. 1992b) and Fusobae-teriuin nucleatwn and Prevotella inter-media (Ali et al. 1994).

Understanding the relationshipamong bacterial species is useful inunderstanding the biology of subgingi-val plaque and in planning strategiesfor its control. Knowledge of the eco-logical relationships among bacterialspecies can direct and focus investi-gations on critical bacterial interac-tions.

Technological developments permitthe evaluation of large numbers of bac-terial species in large numbers of plaquesamples from a wide range of subjects.Such procedures provide a data base inwhich associations among bacterialspecies can be precisely examined. As-sociations between a pair of species isoften performed using contingencytables, sometimes 2x2 and sometimeswith more levels of each species. Clusteranalysis has been useful in describingclosely related species when more thana few pairs of species are examined.Community ordination is a procedurethat attempts to indicate closely relatedspecies within a community and thendemonstrate the relatedness among dif-ferent communities of species within theecosystem of interest. The purpose ofthe present investigation was to usecluster analysis and community ordi-nation techniques to examine relation-ships among bacterial species in subgin-gival plaque samples, and relate thecomplexes to clinical parameters ofperiodontal disease.

Material and MethodsSubject population

185 subjects ranging in age from 20-87years who were considered to be peri-odontatly healthy (n=25) or with evi-dence of prior attachment loss (n^ 160)were selected for study. All subjects hadal least 20 teeth. Exclusion criteria in-cluded pregnancy, periodontal therapyor antibiotics in the previous 3 months,any systemic condition which mighthave affected the progression or treat-ment of periodontitis and the need forpre-medication for monitoring or ther-apy. No subject with localized juvenileperiodontitis, rapidly progressive peri-

odontitis or acute necrotizing ulcerativegingivitis was included in the study.

Ciinicai monitoring

Subjects were screened for suitabilityand if accepted, were asked to sign in-formed consent forms. All subjects wereclinically monitored at baseline and sub-jects with periodontitis were monitoredat 3 month intervals post therapy. Plaqueaccumulation (0/1). overt gingivitis {0/1).bleeding on probing (0/1). suppuration(0/1). probing pocket depth and probingattachment level were measured at 6 sitesper tooth (mesiobuccal. buccal. disto-buccal. distolingual. lingual and mesiol-ingual) at all teeth excluding third mo-lars at each visit. The baseline clinicalfeatures of the 185 subjects are presentedin Table 1.

iMicrobioiogicai assessment

Subgingival plaque samples were takenfrom the mesio-buccal aspect of eachtooth in each subject at each monitor-ing visit. Counts of 40 subgingival spe-cies were determined in each plaquesample using the checkerboard DNA-DNA hybridization technique (Socran-sky et al. 1994). After the removal ofsupragingival plaque. subgingivalplaque samples were taken with individ-ual sterile Gracey curettes from the me-sial aspect of each tooth. The sampleswere placed in separate Eppendorftubes containing 0.15 ml TE (10 mMTris-HCI, 1 mM EDTA. pH 7.6). 0.15ml of 0.5 M NaOH was added to eachsample and the sample boiled in a waterbath for 5 min. The samples were neu-tralized using 0.8 ml 5 M ammonium

acetate. The released DNA was placedinto the extended slots of a Minislot(Immunetics, Cambridge MA) and thenconcentrated onto a nylon membrane(Boehringer Mannheim) by vacuumand fixed to the membrane by exposureto ultraviolet light followed by bakingat 120°C for 20 min. The Minislot de-vice permitted the deposition of 28 dif-ferent plaque samples in individual"lanes" on a single 15X15 cm nylonmembrane as well as 2 control lanescontaining 10̂ or lO** cells of each testspecies. The membrane with fixed DNAwas placed in a Miniblotter 45 {Im-munetics, Cambridge MA), with the"lanes" of DNA at 90° to the channelsof the device. A 30x45 "checkerboard"pattern was produced with 5 of theprobe lanes remaining empty to permitaccurate localization. Each channel wasused as a hybridization chamber forseparate DNA probes. Signals were de-tected by chemiluminescence.

DNA-DNA hybridizationThe membranes were prehybridized at42T for 1 hr in 50% formamide. 5XSSC(lXSSC-150 mM NaCl, 15 mM Nacitrate, pH 7.0), 1% casein (Sigma, St.Louis MO). 5 X Denhardt's reagent, 25mM sodium phosphate (pH 6.5) and 0.5mg/ml yeast RNA (Boehringer Mann-heim). Digoxigen in-labeled, whole chro-mosomal DNA probes were preparedusing a random primer technique (Eein-berg& Vogelstein 1983). The probes andhybridization buffer were placed in indi-vidual lanes of the Miniblotter and thewhole apparatus placed in a sealed plas-tic bag. Membranes were hybridizedovernight at 42°C in a hybridizing solu-tion containing 45% formamide.

Table L Baseline elinieal eharacteristies of subject group («= 185)

Mean (±SD) Range

age (years)no. missing teeth% malesmean pocket depth (mm)mean attaehment level (mm)% sites with:

plaqueredBOPsuppurationpocket depth <4 mmpocket depth 4-6 mmpocket depth >6 mmattachment level <4 mmattachment level 4-6 mmattachment level >6 mm

5I±163.1±3.l

433.l±0.72.9±l.2

69 ±3072 ±3049 ±371.7±5.474 ±2023±i7

3±672 ±2423±i85 ±9

20-870-8

1.7-6.60.8-6.8

0-1000-1000-1000 ^ 9

11-1000-760-577-1000-670-64

136 Socransky cl al.

F/^, I. Dendrogram of a cluster analysis of 32 subgingival taxa. Thesimilarity between pairs of species was computed using a phi coef-ficient, the coefficients were scaled and then sorted using an averageunweighted linkage sort. Clusters were formed with a threshold levelof 60"/i, The color coding used to delineate the groups in this figurewere employed in all figures.

Fig. 2. Community ordination of 32 subgingival taxa using corre-spondence analysis. The relationships among species were evaluatedusing the levels of ihe species at each of lhe sampled sites. Correspon-dence analysis was performed as described by Ludwig & Reynolds(1988) and the species were plotted along the first (.v-axis) and second(_}'-axis) axes. The colors indicaie ta.\a within the same group. A. naes-lundii genospecies 2 [A. viseosus). A. actintmnceteincomitans serotype 5b and S. no.xia were not part of any complex.Fig. j . Community ordination of 32 subgingival taxa using correspondence analysis. The analysis was performed as described in Fig. 2:however, the r-axis has been added. This presentation demonstrates that species that appeared to be closely associated in Fig, 2 such as F.nodatum and T deniicola were widely separated in the 3-dimensional presentation.

Fig. 4. Community ordination of 32 subgingival taxa using principal components analysis. The relationships among species were evaluatedusing the levels of the species at each of the sampled sites. Principal components analysis was performed as described by Ludwig &. Reynolds(1988) and the species were plotted along the first (.v-axis) and second (r-axis) principal components.Fig. 5. Community ordination of 32 subgingival taxa using principal components analysis. The analysis was performed as described in Fig.4; however, the third principal component (r-axis) has been added. This presentation demonstrated that species that appeared to be closelyassociated in Fig, 4 such as F. nodatum and T. denticola were widely separated in the 3 dimensional presentation.

5XSSC. ixDenhardfs reagent. 20 mMNa phosphate (pH 6,5). 0.2 mg/ml yeastRNA. 20 ng/ml o^i labeled probe. lO'V̂ .dextran sulfate and VA^ casein. Mem-branes were washed at low stringency toremove loosely bound probe and then athigh stringency (68°C. O.lxSSC, 0,1%

SDS, 20 min. twice) in a Disk Wisk ap-paratus (Schleicher and Schuell. KeeneNH).

Dcleclion and cnumeraliini

To detect hybrids, membranes wereblocked and then incubated with a

1:25,000 dilution of anti-digoxigeninantibody conjugated with alkalinephosphatase using the modification de-scribed by Engler-Bium et al. <1993).After washing, the rnembranes were in-cubated in Lumiphos 530 (Lumigen,Southfield, MI) for 45 min at 37X,

Microbial complexes 137

% of sites70

554020

15 —

10

B

Pg-

Pg*

forsythus -ve B. forsythus +veTd- Td* Td- Td+

64.2

1.1

5.1

1.3

7.8

3.1

7.3

10.1

Observed

Expected

None Pg Td Bf BfPg

BfTd

All

Fig. 6. Bar chart of the observed and expected frequencies of mettibers of the red complex(fi, forsvihus (Bf). T. denticola (Td) and P. gingivulis (Pg)) individually and in dilTerentcotTtbinations, The observed frequencies of the 3 species were summarized in a 2X2X2 contin-gency table and conligural frequency analysis (von Eye 1990) was performed to cotiipute theexpected frequencies. The inset presents the % of sites in each of the 8 cells.

% of sites50 —

40 —

30 =

22 —

1 5 -

8 —

P.

Pm~

Pm+

intermedia -ve P. intermedia +veFnv-42.3

2.0

Fnv+6.2

5.4

Fnv—7.6

2.4

Fnv*7.9

26.0

Observed

Expected

None

Fig. 7. Bar chart of ihe observed and expected frequencies of representative metnbers of theorange complex {P. micros (Pm). F. nucleatum ss vincentii (Fnv) and P. intertnedia (Pi/)individually and in different combinations. Observed and expected frequencies were conipntedas described for Fig, 6,

placed in a film cassette with ReflectionNEF filtn (Dupont, Boston MA) for 1h at 37°C and then developed. Twolanes in each run contained standardsat different concentrations. The sensi-tivity of this assay was adjusted to per-mit detection of 10"̂ ceils ofa given spe-

cies by adjusting the concentration ofeach DNA probe. This procedure wascarried out in order to provide the samesensitivity of detection for each species.Failure to detect a signal was recordedas zero, although conceivably, counts inthe I to 1000 range could have been

present. Signals were evaluated visuallyby comparison with the standards forthe test species. They were recorded as:0. not detected; 1,<IO^ cells; 2. - 1 0 ^ 3,10- to 10^ 4, -10"; 5 >10^ cells.

Statistical analysis

Microbiological data available for eachsubject included the level of each of 40test species from up to 28 plaquesamples (mean 25,2 samples) at eachvisit. The prevalence (the number ofsites at which the species was detected)of each species was also computed foreach subject at each visit. A total of13,261 plaque samples were evaluated.Data for 32 of the species examinedwere used in the following analyses. Thespecies were chosen on the basis of fre-quency of detection; species detectedin< 5% of sites were omitted. Two bytwo tables were set up for each of the496 pairs of species in which the pres-ence or absence of each species wassummarized. A phi coefficient was com-puted to assess association betweeneach pair of species and scaled to rangebetween 0 and lOO'Mi, The resultingsimilarities were clustered using an aver-age unweighted linkage sort (Sneath &Sokal, 1973). For comparative pur-poses, other sitnilarity coefficients wereemployed including the Bray Curtis,Mahalanobis (/- and the correlation co-efhcients.

Community ordination was per-formed using principal componentsanalysis and correspondence analysis(Ludwig & Reynolds 1988). The dataemployed were the ranks representingdifferent counts of the test species at the13,261 sampled sites. The data wereplotted as described by Ludwig & Reyn-olds (1988).

In order to examine the robustness ofthe observed relationships, subsets ofthe data were analyzed using the clusterand community ordination procedures.These subsets included use of first visitdata only and samples from sites withinspecified pocket depth ranges including0-3 mm, 4-6 mm and >6 mm as wellas pre- and post-therapy data.

Associations among species within acomplex were examined using config-ural frequency analysis. Significance ofdifferences between observed and ex-pected values within cells were deter-mined as described by von Eye (1990).

The relationship between the clinicalparameters and individual species and/or complexes was examined after aver-

138 Socransky et al.

aging data for a chosen parameterwithin the subject and then averagingacross subjects. For example, the levelsof a species or complex were averagedfor all non bleeding on probing sites ina subject as well for all bleeding onprobing sites in that subject prior to av-eraging within BOP categories acrosssubjects. Significance of differences forclinical parameters were sought usingthe Kruskal-Wallis test.

ResultsAssociations among bacterial species asdetermined by cluster analysis

Fig. I presents a dendrogram of the re-sults of the cluster analysis of all 13,261samples using ^ coefficients and an av-erage unweighted linkage sort. 5 clus-ters were formed with >60% similarityand included 29 of the 32 taxa evalu-ated. The red cluster consisted of P. gin-givalis. B. forsythus and T. denticola.The orange cluster consisted of F. nu-eleatum subspecies, P. intermedia andP. nigre.scens, Peptostreptoeoccus microsand Campylobacter rectus, Campvlo-bacter showae, Campylobacter gracilis,E. noddttim and S. eonstetlatus. The 3Capnocytophaga species. Catnpylo-bacter conclsus. Eikenella corrodens andActinobacillus actinomycetetncomitansserotype a formed the green cluster.while a group of streptococci made upthe yellow cluster. Streptococcus mitis.Streptococcus sanguis and Streptococ-cus oralis were most closely relatedwithin this group. Actinotnvces odonto-lyticus and Veillonella parvula formedthe purple cluster. Actinotnyces naeslun-dii genospecies 2 (ActinomycesvLseosus), Selenomonas noxia and A.aetinomycetemcotnitans serotype b didnot cluster with other species.

Associations among bacterial species asdetermined by community ordination

The results of community ordinationusing correspondence analysis are pre-sented in Fig. 2, This analysis re-inforced the relationships demonstratedin Fig. 1 and showed the relationshipsamong the different microbial complex-es. For example, the red complex con-sisting of P. gingivalis. B. forsythus andT. denticola was closely associated withthe orange complex that included F. nu-cleatum subspecies. P. intermedia andP. nigre.scens, P- micros, E. noiiatumand S. constellatus and 3 of theCampylobacter species. Fig. 2 also dem-

Proportion of Sites exhibitingthe Red complex

1 - 4

5 - 8

9-12

Fig. 8. Area plots depicting the percentage of sites colonized by tiiembers of the orange andred complexes. The data are based on examination of 13,261 subgingival plaque samples. Theleft hand panel indicates the percentage of sites colonized by 0, 1-4, 5-8 and 9-12 membersof the orange complex. The right hand panel presents ihe same data, but overlaid with thepercentage of sites in each of those categories that exhibited 1, 2 or 3 members of the redcomplex.

1 - 4 5-8 9-12

Proportion of Sites exhibitingthe Orange complex

Fig. 9. Area plots depicting the % of sites colonized by members of the red and orangecomplexes. The left hand panel indicates the % of sites colonized by 0, 1, 2 or 3 members ofthe red complex. Tbe right hand panel presents the same data, but overlaid with the percentageof sites in each of those categories tbat exhibited 0, 1-4, 5-8 and 9-12 members of the orangecomplex.

onstrated the relationships among theCapnocytophaga species, most of theStreptococcus species, E. eorrodens andC coneisus. These species were closelyrelated to each other and somewhat re-lated to the orange complex. A. actino-mycetemcomitans serotype b appearedto be closely related to C ochracea, butexamination of the 3 dimensional plotof the correspondence analysis (Fig. 3)indicated that these 2 species were dis-tant from one another. The relationshipbetween A. odontolyticus and V. parvulasuggested by cluster analysis was con-firmed by correspondence analysis.Further, the outlier status of S. noxia,A. actinomycetemcomitans serotype b

and A. naeslundii genospecies 2 {_A.viscosus) was evident (Figs. 2, 3),

Figs. 4, 5 present principal compon-ents analysis of the same data and indi-cate similar groupings of species. Thisanalysis reinforces the distinction of thered {P. gingivalis. B. forsythus and T.deniicola) and orange clusters as well asthe separation of these 2 groups fromthe green and yellow complexes.

Associations among species withincomplexes

The associations between species in thered complex are summarized in Fig, 6.64"/j of sites harbored none of the spe-

Microbial complexes 139

% sites colonized

1 0 0 -

7 5 -

5 0 -

2 5 -

n

1 0 0 -

7 5 -

5 0 -

2 5 -

n_

P. gingivalisp < 0,0001

B. forsyttiusp < 0.0001 g

F. nuc. ss vincentiip < 0,0001

p. intermediap < 0.0001

24116458

S. sanguis

,393174

4. viscosus

<3 3 4 5 6-7>7 3 4 5 6-7>7 <3 3 4 5 6-7>7POCKET DEPTH (MM)

Fig. 10. Bar charts of the mean (±SEM) of the % of sites colonized by 6 subgingival speciesat selected pocket depths. The percentage of sites colonized by each species al each pocketdepth category was computed for each subject and then averaged across subjects. Significanceof differences among pocket depth categories was tested using the Kruskal-Wallis test. Thetotal number of subjects was 185. The number of subjects exhibiting the different pocketdepths is shown above the bars in the A, nae.slundii genospecies 2 {A. vi.seosus) panel, whilethe number of sites in each pocket depth category is shown above the bars for S.

the same data together with the percen-tage of sites in each category that alsoharbored 1, 2 or 3 metnbers of the redcomplex. It is readily apparent thatmembers of the red complex are rarelyfound in the absence of members of theorange complex. With increasing colon-ization by the orange complex, moresites were colonized by increasing num-bers of the red complex. Fig. 9 presentsthe same data but with the left handpane! depicting the % of sites colonizedby members of the red complex. Only36% of sites harbored one or more ofthe red complex species compared with72% of sites which harbored membersof the orange complex (Fig 8). Theright hand panel indicates that a largeproportion of sites in which the redcomplex was not detected was colon-ized by members of the orange com-plex. Taken together the data suggestthat species in the orange complex pre-cede colonization by species of the redcomplex.

% sites colonized100 -| P. gingivalis

75 -

50 -

25 -

F. nuc. ss vincentii -. S. sanguis

B. forsytiius P. intermedia A. viscosus

3 4 5 6-7>7 <3 3 4 5 6-7>7POCKET DEPTH (MM)

Fig. II. Stacked bar charts of the % of sites colonized by different levels of 6 subgingivalspecies at selected pocket depths. The number of subjects and sites providing the data werepresented in Fig, 10,

cies in this complex, while \0"A> of sitesharbored all 3 species. These percen-tages were strikingly elevated as deter-mined by configural frequency analysis.In contrast, the individual species weredetected less frequently than expected.The same analysis was repeated usingrepresentative members of the orangecomplex (Fig. 7). In a fashion similar tothe red complex, the % of sites harbor-ing none or all of the species were sig-nificantly higher than expected, whilethe individual species and pairs of spe-

cies were detected less frequently thanexpected. Similar results were obtainedfor the yellow and green complexes(data not shown).

Relationship among complexes

The relationship between the red andorange complexes is illustrated in Figs,8. 9. The left hand pane! of Fig, 8 de-picts the "/i of sites colonized by 0. 1-4,5-8 and 9-12 members of the orangecomplex. The right hand panel presents

Reiationsiiip between bacterial speciesand ciinicai parameters

The relationship of the species in thedifferent complexes to pocket depth wasexamined. Species in the red complexexhibited a very strong relationshipwith pocket depth (Figs, 10. II), For ex-ample. B. forsythus and P. gingivalis(and T. denticola, data not shown) in-creased in prevalence and numbers withincreasing pocket depth. Similarly, allspecies in the orange complex showeda significant association with increasingpocket depth as exemplified by P. inter-media and F. nucleatum ss vincentii. A.naeslundii genospecies 2 {A. viscosus)and S. .sanguis provide examples of theremaining species which showed nostatistically significant relationship withpocket depth.

Since species in the red complex wereso strongly related to pocket depth, therelationship between individual andcombinations of species in this complexand pocket depth was examined further(Fig, 12), Sites with none of the speciesexhibited the shallowest mean pocketdepth, while sites harboring all 3showed the deepest. It is interesting tonote that sites harboring P. gingivalisalone or in combination with the other2 species exhibited the deepest meanpocket depths. The red complex and theindividual species in that group werealso strongly associated with bleedingon probing (Fig. 13).

140 Socran.skr el al.

Mean PD (mm)

4.5

3.4

Fig. 12. Bur chart of mean (±SEM) pocket depth atsites harboring none, all or different eomhinations ofthe speeies in the red eomplex {B. forsythus (Bfj. T.dcnticota (Tdj and P. gingivalis (Pg/). The averagepocket depth for each of the microbial combinationswas compuied within it subject and then averagedacross subjects.

Td Bf Pg Bf+Td Pg+Td

Counts

No-BOP

Av purple yellow green orange red

p<0.01

Pg Td Bf

Fi^. ti. Bar chart of the mean (±SEM) % of speciesin each complex at sites that bled or did noi bleed onprobing. The "/i< of species in a complex present at eachsite was computed. The values were averaged for eachBOP category witbin each subject and then averagedacross subjects. The only significant difference betweenbleeding and non bleeding sites was observed for tbered complex. The individual members of that complexdiffered significantly as dcpicied in the right panel.

A. naeslundii 2IA viscosus)

V.panulu

C, gtaciUi

S. constellitus

Fig. 14. Diagrammatic representation of the relation-ships of species within microbial complexes and be-tween the microbial complexes. This diagram wasbased on the results of multiple cluster and communityordination analyses using the entire data base as wellas subsets of data.

Microbial compte.xes 141

Discussion

The goal of the present investigationwas to attempt to understand the na-ture of the microbial complexes thatexist in subgingival plaque. Any rep-resentation of these complexes whetherby cluster analysis or community ordi-nation techniques suffers the limitationthat one is attetiipting to represent mul-ti-dimensional relationships in two orthree dimensions. Thus, different pres-entations of relationships are bound tosuggest somewhat different associationsamong species. In spite of these reser-vations, the associations observed werequite robust, in that different similaritycoefficients, different methods of com-munity ordination or subsets of thedata, such as flrst visit data only or datafrom sites with different baseline pocketdepths, provided essentially identicalgroupings. The only exceptions werethe occasional movement of E. nodittwninto the red complex. A. actinomyce-temcomitans serotype a joining A. acti-nomycetetncomitans serotype b and A.naeslundii genospecies 2 (A. viscosus)Joining the purple complex. Fig, 14 isan attempt to summarize the complexesand the relationships among complexesobserved in multiple analyses of the fulldata base and subsets of the base. Thepotential for human error in interpreta-tion of the multi-dimensional data ishigh: however. Fig. 14 may serve as apoint of departure for further evalu-ation of subgingival microbial relation-ships.

One concern in interpreting the datawould be the sensitivity and specificityof the DNA probes employed. The sen-sitivity of the assay was set to 10" cellsof a species by adjusting the concen-tration of each probe in the hybridiza-tion buffer. The specificity of the probeswas examined using a recently acquiredStorm Fluorimager (Molecular Dy-namics, Sunnyvale, CA, USA) and At-tophos (Amersham Life Science, Ar-lington Heights. Illinois, USA) insteadof Lumiphos 530 (Lumigen, Southfield.Michigan, USA) in the detection step.Over 92% of all probe: heterologous spe-cies reactions did not exhibit cross-reac-tions under the conditions of the assay.Some probes such as those to B. for-sythus, P. gingivalis and T. denticola didnot exhibit cross-reactions with any het-erologous species tested. When cross-re-actions were observed they were alwayswithin genera and were quite limited.For example, of the 6 Campylobacter

species evaluated, only C .showae andC. rectus cross-reacted but at a homo-logous:heterologous ratio > 100:1.There were no cross-reactions amongthe 3 Capnocytophaga species. Cross-re-actions within the Sireplococci wereminimal always exceeding 100:1 forhomologousiheterologous species, al-though S. oralis. S. mitis and S. sanguisshowed no cross-reactions with 5, in-tertnedius and S. constellatus. Thus, themajority of the observed relationshipswere unlikely to be due to cross-reac-tions of the DNA probes althoughsome influence cannot be entirely ruledout. A second concern might have beenthe fact that pre- and post-therapysamples were included for periodontitissubjects, but only first visit data wereused for periodontally healthy subjects.It must be emphasized that the relation-ships depicted in this manuscript wereobserved when only 1st visit data wereemployed; i.e., each subject contributedapproximately equal numbers ofsamples. The authors chose to presentthe entire data base since it includedpre- and post-therapy samples adding adegree of robustness to the observed re-lationships.

The relationship of P. gingivalis. T.denticola and B. forsythus (red complex)appears in one guise or another in theFigures provided in this manuscript aswell as when subsets of data were exam-ined from pre or post therapy visits orfor sites subset in different pocket depthranges (data not shown). Aspects of thiscomplex have been described in theliterature (Gmur et al. 1989. Simonsonet al, 1992ab). Gmur et al. (1989) de-scribed a strong relationship between B.forsythus and P. gingivalis in subgingi-val plaque samples from pockets of dif-ferent depths in adult subjects. Bothspecies were detected more frequentlyand in higher numbers in deeper peri-odontal pockets. In addition, P. gingi-valis was never detected in the absenceof B. forsythus. Simonson et al. (1992a)found a strong association between P.gingivalis and T. denticota in plaquesamples taken from subgingival sites in74 Fijians, 74 Colotnbians and 73Americans stationed in the Sinai desert,Pederson et al, (1994) demonstrated asimilar relationship in a continental USpopulation. Members of the red com-plex were found together in high num-bers in lesions of adult periodontitis(Hosaka et al, 1994. Umeda et al.1996); in particular, in sites with deeperpockets or more advanced lesions (Si-

monson et al. 1992b, Kojima et al.1993, Wolff et al. 1993. Ah et al. 1994,Kamma et al, 1995, Kigure et al. 1995).In particular, Kigure et al. (1995) usingimmunohistochemical techniques, pro-vided graphic demonstration of the re-lationship between T. denticola and P.gingivalis in biopsies of subgingivalplaque, epithelial and connective tissuesfrom different pocket depths in humanperiodontitis subjects. They demon-strated that both species were predomi-nant in pockets >4 mm. In 4 6 mmpockets, T. denticola was detected at thesurface layer of the plaque, while P. gin-givalis cells were detected in the layerbeneath. In deeper pockets, the speciesco-existed in large numbers. Otherstudies demonstrated a reduction in thespecies of this complex after scaling androot planing (Simonson et al. 1992b,Haffajee et al. 1997). Possible mechan-isms of pathogenicity of species in thered complex have been reviewed (Soc-ransky & Haffajee 1991, Haffajee &Socransky 1994); however, it is of inter-est that this trio has been shown to pro-duce proteolytic enzymes such as thosesought in the "BANA" test (Loesche etal, 1992) and in the proposed SK013peptidase test (Seida et al. 1992). Thebiological basis of the associationamong P. gingivalis, T. denticola and B.forsythus is not known. However, it hasbeen shown that members of this com-plex coaggregate strongly in vitro (Grcnier 1992a, Onagawa et al, 1994.Yao et al. 1996) and one species of thecomplex may produce growth factorsrequired by another in that complex(Grenier 1992b, Nihus et al. 1993).

A second complex that was observedin multiple analyses was the orangecomplex consisting of the F. nudeatumsubspecies, F. periodonticwn. P micros.P. intermedia. P. nigrescens. Streptococ-cus constellatus, E. nodatum. C. showae.C gracilis and C rectus. The species inthis group were closely associated withone another and this complex appearedclosely related to the red complex.Other studies have shown an associ-ation between members of this com-plex. For example, Ali et al. (1994)found that P. intermedia was always de-tected in the presence of F, nudeatum insubgingival plaque samples from deeppockets in a group of adult peri-odontitis subjects. P. micros and C. rec-ttis were significantly elevated insamples from mobile teeth comparedwith non-mobile teeth (Grant et al.1995), Von Troil-Lindcn et al. (1995)

142 Socranskv el at.

found that P. intermedia. C rectus andP. micros were significantly elevated insaliva samples from subjects with ad-vanced periodontitis compared withsamples from subjects with initial or noperiodontitis. Further, treatment thatincluded systemically administered me-tronidazole decreased levels of thesespecies and improved periodontal sta-tus. Members of the orange complexhave been associated with infections innon-periodontal sites. Sundqvist (1992)sampled 65 infected root canals. Themost frequently detected organisms in-cluded E. nudeatum. P. intermedia. P.micros. P. anaerobius. Eubacterium sp.and C. rectus. Strong associations werefound between F. nucleatum and P.micros. E. tnideatum and C rectus andP. intermedia and P. micros. Summanenet al. (1995) found that the most com-monly detected anaerobes in subcutane-ous abscesses of intravenous drug userswere E. nucleatunt. P. tnicros. P. inter-media. P. nigrescens, A. odontotyticusand V. parvula while E. nucleatum, P.intermedia. P. nigrescens. P. micros andEuhacterium species were most com-monly detected in 46 cases of anaerobicempyema (Civen et al. 1995). P. inter-media has been shown to stimulate thegrowth of 5. constellatus in vitro and thecombination of these species often pro-duced fatal pulmonary infections in amouse model system (Shinzato & Saito1994). One possible mechanism for co-aggregation between E. nodatum andstrains of E. nudeatum was described byGeorge and Falkler (1992). The coag-gregation was thought to involve a pro-tein receptor on E. nucleatum and aheat-stable protein or polysaccharidecomponent on E. nodatum.

While species within complexes wereclosely associated, the complexes them-selves seem to have specific relation-ships with one another (Fig. 12). Spe-cies such as A. naeslundii genospecies 2(A. viscosu.s). and members of the yel-low and green complexes were less com-monly associated with members of thered and orange complexes than witheach other. The 2 species in the purplecomplex were strongly related to eachother and to a lesser extent to membersof the orange, green and yellow com-plexes. The reasons for these relation-ships among complexes is unclear butit might be speculated that antagonisticrelationships (Grenier 1996) may existor that environments selective for onegroup may be less hospitable to a sec-ond group of organisms.

On observing Fig. 14. it is temptingto speculate that the relationships ob-served mimic to some extent the mi-crobial succession patterns that mayexist in developing plaque. It seemslikely that A. naeslundii genospecies 2{A. vLscosus) and the strcptococcal spe-cies are early colonizers followed byCapnocytophaga sp,, C. concisus, and E.corrodens. A. odontotyticus and V. par-vula may be bridging species leading tothe orange complex and ultimately, atsome sites, to the red complex. Thispattern of colonization is part specu-lation and in part based on data aboutthe composition of plaques in variousclinical conditions (Moore & Moore1994), It is interesting to note that atreatment such as SRP has a profoundeffect on the species of the red complexand virtually no effect on the majorityof other species except for an increasein A. naeslundii genospecies 2 {A.viscosu.s) and some of the streptococcalspecies (Haffajee et al. 1997),

The red complex showed the strong-est relationship with the clinical par-ameters considered most meaningful inperiodontai diagnosis. For example, theindividual species in the complex as wellas ihe complex itself related verystrongly with pocket depth and bleed-ing on probing. As described earlier inthe Discussion, other investigators havenoted the relationship between pocketdepth and members of the red complex(Gmur et al. 1989, Kigure et al. 1995.Simonson et al, 1992b). The orangecomplex also related to pocket depth,although this relationship and the re-lationship to other clinical parameterswas less striking.

Knowledge of the associations be-tween subgingival species maybe usedto interpret and guide periodontal ther-apy. The recognition that an inter-re-lated complex of species such as the redcomplex exists suggests that therapiesthat affect 1 of these species may influ-ence the colonization of the other spe-cies in this group. In addition, it mightbe speculated that altering the orangecomplex might prevent subsequentcolonization of species in the red com-plex providing the therapist with a sec-ond route to control this group of or-ganisms. A. actinomycetemcomitansserotype b related poorly to members ofthe red and orange complexes sug-gesting that therapies effective againstone set of pathogens may not necess-arily be effective against others. This in-deed appears to be the case for SRP

Further, other studies have suggestedthat different therapies affect differentsegments of the microbiota and the usean incorrect therapy may be ineffectivein lowering the target species and mightadversely affect other members of themicrobial community (Haffajee et al.1996),

The data of the present investigationsuggest the nature of some of the mi-crobial complexes in subgingivalplaque. Although the number of plaquesamples was quite large and the numberof species evaluated was reasonably ex-tensive, the investigation represents aninitial attempt at evaluating inter-re-lationships among subgingival species.While some of the associations depictedmay not be completely accurate, certainassociations were seen repeatedly usingdifferent analytical techniques and werein accord with data in the hterature.Further, certain complexes, and mem-bers within the complexes, relatedstrongly to clinical parameters of in-flammation and periodontal destruc-tion. The data provide a framework forunderstanding the complex ecology ob-served in plaque and could be used toguide approaches to diagnosis and ther-apy of periodontal diseases.

Acknowledgments

This work was supported in part by re-search grants DE-04881 and DE-10977from the National Institute of DentalResearch.

Zusammenfassung

Mit<rohiette Komplexe in der subgingivalenPlaqueEs wurde seit einigcr Zeit erkannt, daB bak-terielle Spezies in der subgingivalen Plaquein Komplexen existieren, Der Zweck der vor-liegenden Unlersuchiing war es, die Defini-tion solcher Gemeinschaften unter Verwen-dung der Daten einer groBen Anzahl vonPlaqueproben sowie verschiedener Cluster-und Ordnungstecbniken zu versuchen, Bei185 Personen (Altersdurchshniit 5]±f6 Jah-re) wurden von der Mesialflache eines jedenZahnes mit (/i=160) oder ohne Parodontitis('1 = 25) subgingivale Plagueprobcn genom-men, Unter Verwendung von Ganz-Gcnom-DNA-Sonden und der Schacbbrett-DNA-DNA-Hybridisierung wurden das Vorkom-men und der Anteil von 40 subgmgivalenTaxa bei 13261 Plaqueproben bestimmt. Zujedem Termin wurden an 6 Stellen pro Zahnklinischen Messungen durchgefuhrt. Ahn-lichkeite zwischen Speziespaaren wurden un-ter Verwendung von Phi-Koeffizienten er-

Microbial complexes 143

rechnet und die Speziesclusterung erfolgteunter Verwendung einer ungewichtctenDurchscbnitt-ortentierten Sortierung, Mit al-ien analytischen Methoden wurden 5 Haupt-komplexe bestandig beobachlet. Ein Kom-plex bestand aus der eng verbundenen Grup-pe: Bacieroldes forsythus, Porphyrtmionoxgingivalis und Treponema denticolu. Derzweite Komplex bestand aus etner eng ver-bundenen Zentralgruppe tiiit den Subspeziesvon Fuxobucleriwn nuelealitm/periotlurUicum,Prevolt'llu intermediu, Prevotella niigrescensund Peptostreptococctts micros. Spt-zies. diemit dieser Gruppe assoziiert waren, sind fol-gende: Eubacterium nodatuni, Campylobaclerrectus. Campyhbucter shonue, Sireptococeusconstellalus und Catnpylohaeter graeilis. Derdritte Komplex bestand aus Streptococcussanguis. S. oralis, S. mitis, S. gordonii und 5,intermedius. Der vicne Komplex umfajiie die3 Capnocytophaga Spezies, Campylobacterconeisus. Eikenella corrodens und Aeiinoba-eillus actinomyeelemcomitans Serotyp a, Derfunfte lComplex bestand aus I'eillonelta par-vula und Actinomyces odontolyticus. A. acti-nomycetemccmitans Serotyp b. Selenomonasnoxia und Actinomyees naeslundii Gcnospe-zies 2 {A. viscosus) lagen auCerhalb der Clu-ster mit geringer Verbindung zueinandcr undzu den 5 Hauptkomplexen, Der crste Kom-ptex hatte eine auffallende Verbindung zuden klinischen Parametern der Parodontaler-krankung. insbesondere Taschentiefe undSondlerungsblutung.

Resume

Contple.xes iiiierobiens dans la plaque sous-gingivaleII est reconnii depuis quelques temps que lesespeces bacteriennes existent sous forme decomplexes dans la plaque sous-gingivale, Lebut de la presents etude emit de tenter de de-finir ce genre de communautes en utilisanlles donnees fournies par un grand nombred'echantillons de plaque et differentes nie-thodes de groupement el de classement, Desechantillons de plaque sous-gingivalc ont etepreleves du cote mesial de chaque dent chez185 sujets (age moyen 51 ±16 aiisl avec (« =160) ou sans (7J = 25 ) parodontite. La presen-ce et les nivcaux de 40 taxons sous-gingivauxont ete determines dans 13261 echantillonsde plaque, en utilisant des sondes d'ADN ge-nomique entier et un hybridation ADN-ADN en damier. Les mensurations cliniquesont ete faites dans 6 sites par dent a chacunedes visiles, Les similarites entre les pairesd'especes ont ete calculees en utilisant lescoefficients phi e( les especes ont ete groupeesa I'aide d'un tri moyen des liaison.s sans pon-deration (average unweighted linkage sort).Le classement des communautes a ete fait aTaide de l'analyse des composantes principii-les et de Taiialyse des correspondances, 5complexes principaux ont ete reguherementobserves avec toules les methodes d'analyse,L"un des complexes elait constitue par legroupe etroitement apparente: Baeteroidesforsythus, Forphyromonas gingivalis et Trepo-

nemu denlieola. La second complexe etaitconstitue par un noyau elroitement apparen-te comprenant des membres des sous-especesFusobacterium nucleatum/periodonticum, Pre-voiella intermedia, Prevotella nigrescens etPeptostreplococcus micros. Les especes asso-ciees avec ce groupe comprenaient: Eubacie-rium nodatuni, Campylobacter recttts. Campy-lobaeier showae. Strepioeoccus eonsleltalus etCampylobacter graeilis. Le troisieme com-plexe etait compose de Slreptocoecus sanguis,S. oralis, S. mitis, S. gordonii et S. interme-dius. Le quatrieme cotiiplexe etait constituepar 3 especes de Capnoeytophuga, Canipylo-baeter eoncisus, Eikenella eorrodens et Aeti-nobacillus aetinomycetemeomitans serotype a.Le cinquieme compiexe erait compose parVeillonella parvula et Avtinomyces odoniolyli-cus. A la peripherie, A. actinomycetemcomi-lans serotype b, Selenomonas noxia et Aeti-nomyces naeslundii genospecies 2 (A. visco-sus), avaient Ires peu de relation entre eux etavec les 5 complexes principaux, Le premiercomplexe avait une relation frappante avecles mesures cliniques de la maladie parodon-tale. en particulier la profondeur des pocheset !e saignement au sondage.

References

Ali. R, W, Skatig. N,. Nilsen. R. & BakkenV, (1994) Microbial associations of 4 puta-tive periodontal pathogens in Sudaneseadult periodontitis patients determined byDNA probe analysis. Jourtia! of Periodon-tology 65. 1053-1057,

Civen. R,. Jousimies-Somer. H.. Marina. M,.Borenstein. L.. Shah. H, & Feingold. M,(1995) Retrospective review of cases of an-aerobic empyema and update of bacter-iology. Clinical Infectious Diseases 20,(suppl 2). S224~S229,

Engler-Blum, R,. Meier, M,, Frank. J.. &Muller. G. A, (1993) Reduction of back-ground problems in non radioactiveNorthern and Southern blot analyses en-ables higher sensitivity than -'-P-based hy-bridizations. Analytical Bioihem 210. 235-244,

Feinberg. A, P. & Vogelstein. B, (1983) Atechnique for radiolabeling DNA restric-tion endonuclease fragments to high speci-fic activity. Analytical Biochemistry 132, 6-13.

George. K. S. & Falkler. W, A, Jr. (1992) Co-aggregation studies of the Eubaeteriumspecies. Oral Microbiology and Immu-nology. 7, 285-290.

Gmur, R,. Strub, J. R. & Guggenheim, B.(1989) Prevalence o{ Bacleroides forsythusand Bacteroides gingivalis in subgingivalplaque of prosthodontically treated pa-tients on short recall. Journal of Peri-odonial Research 24, 113-120.

Grant. D, A,, Grant. D, A,. Flynn. M, J, &Slots, J, (1995) Periodontal microbiota ofmobile and non-mobile teeth. Journal ofPeriodontology 66, 386-390,

Grenier, D, (1992a) Demonstration ofa bi-

modal coaggregation reaction betweenPorphyromonas gingivalis and Treponi-madcnlieota. Oral Microbiology and Immu-nology 7. 280-284,

Grenier. D, (1992b) Nutritional interactionsbetween two suspected periodontopathog-ens, Treponemu dentieola and Porphyro-monas gingivalis. Infection and Immunity60.5298-5301,

Grenier, D, (1996) Antagonistic effect of oralbacteria towards Treponema dentieola.Journal of Clinical Mierobiology 34. 1249-1252,

Haffajee, A, D . Cugini. M, A.. Dibart. S..Smith, C . Kent. R, L, Jr, & Socransky S.S,(1997) The effect of SRP on the clinicaland microbiological parameters of peri-odontal diseases. Journal of Clinical Peri-odonlology 24. 324-334,

Haffajee. A. D. & Socransky, S, S. (1994) Mi-crobia! etiological agents of destructiveperiodontal diseases. In: Socransky. S,S &Haffajee. A, D . ed. Microbiology and Im-munology of periodontal diseases, Period-ontology 2000. 5. 78-111,

Hosaka, Y. Saito. A.. Nakagawa. T. Seida.K,. Yamada. S, & Okuda. K, (1994) Effectof initial therapy on dynamics of immuno-globulin G levels to some periodontopath-ic bacteria in serum and gingival crevicularfluid. Bulletin ofTokvo Denial College 35.207-216,

K.amma. J. J,, Nakou, M, & Manti, F. A.(1995) Predominant microflora of severe,moderate and minimal periodontal lesionsin young adults with rapidly progressiveperiodontitis. Journal of Periodontal Re-.seareh 30, 66- 72,

Kigure, T. Saito, A.. Seida, K,. Yamada. S,.lshihara, K, & Okuda. K. (1995) Distri-bution of Porphyromonas gingivalis andTreponema denticola in human subgingivalplaque at different periodontal pocketdepths examined by immunohistochemicalmethods. Journal of Periodontal Research30, 332-341.

Kojima. T, Yasui. S. & Ishikawa. L (1993)Distribution of Porphyromonas gingivalisin adult periodontitis patients. Journal ofPeriodontology 64, 1231-1237,

Listgarten. M, A, (1976) Structure of the mi-crobial flora associated with periodonta!health and disease in man. Journal ofPeriodontology 4tl. 1-18.

Listgarten, M, A, (1994) The structure ofdental plaque. Periodontology 2000 5. 52-65,

Listgarten, M, A,, Mayo, H. E, & Tremblay.R. (1975) Development of dental plaqueon epoxy resin crowns in man, A light andelectron microscopic study. Journal ofPeriodontology 4^. 10-26,

Loesche. W, J,. Lopatin, D E.. Giordano. J,.Alcoforado. G- & Hujoel, PP (1992) Com-parison of the benzoyl-DL-argininc-naph-thylamide (BANA) test. DNA probes, andimmunological reagents for ability to de-tect anaerobic periodontal infections dueto Porphyromonas gingivalis, Treponemadenticola and Baeteroides forsythus.

144 Socransky el al.

Journal of Clinical Microhiotogv 30, 427433,

Ludwig. J, A, & Reynolds. J, F, (1988) Stat-istical ecology. A primer on methods andcomputing. John Wiley & Sons, NewYork, pp, 223 256.

Moore. W. E, C & Moore. L. V H, (1994)The bacteria of periodontal diseases. Peri-odontology 2000 5. 66 77.

Nilius. A, M., Spencer. S. C, & Simonson, L,G. (1993) Stimulation of in vitro growthof Treponema denticota by extracellulargrowth factors produced by Porphyromon-(is gingivatis. Journal of Dentat Re.searvh72. 1027-1031.

Onagawa. M,,, Ishihara. K. & Okuda. K,(1994) Coaggregation between Porphyro-monas gingivatis and Treponemci denticola.Bulletin of Tokvo Dentat Cottege 35. 171-181,

Pederson. E, D, Miller. J, W. Matheson. S..Simonson. L, G,. Chaduick. D, E,. Covili.P, J,, Turner. D, W,. Lamberts. B. L, &Morton. H- E, (1994) Trypsin-like activitylevels of Treponema denticota and Porphy-romonas gingivatis in adults with peri-odontitis. Journal of Clinical Periodonto-logy 21, 519-525,

Seida. K.., Saito. A.. Yamada. S.. Ishihara.K.. Naito. Y, & Okuda. K.- (1992) A sensi-tive enzymatic method (SK-Ol 3) for detec-tion of Treponema denticola. Porphyro-monas gingivati.s and Bacteroide.sforsythusin subgingival plaque samples. Journal ofPeriodontal Research 11. 86 91,

Shinzato, T, & Saito. A, (1994) A mechanismof palhogenicity of "Streptococcus millerigroup" in pulmonary infection: synergywith an anaerobe. Journal of MedicalMicrobiology 4(i, 118-123.

Simonson. L, G.. McMahon. K, T. Childers.D. W. & Morton. H. E. (1992a) Bacieria!synergy of Treponema denticola and Por-phyrotnonas gingivatis in a multinationalpopulation. Oral Microhiotogy and Itmmi-notogyl. 111-112,

Simonson, L, G,. Robinson. P. J,. Pranger. R,J,. Cohen. M. E. & Morton. H, E. (1992b)Treponema denticota and Porphyromonasgingivatis as prognostic markers followingperiodontal treatment. Journal of Period-ontologv 63. 270-273,

Sneath R H, A, & Sokal R, R. Numericaltaxonomy. The principles and practice ofnumerical classification, San Francisco:WH Freeman & Co,, 1973,

Socransky. S, S, & Haffajee A. D, (1991) Mi-crobial mechanisms in the pathogenesis ofdestructive periodontal diseases: a criticalassessment. Journal of Periodontat Re-search 26. 195-212.

Socransky. S, S,. Smith. C . Martin. L,,Paster. B, J,. Dewhirst. F, E, & Levin. A.E, (f994) Checkerboard DNA-DNA hy-bridization, Biotechuques 17. 788-792,

Summanen. P, H.. Talan. D, A,. Strong. C ,McTeague, M,. Bennion. R,. Thompson. J,E, j r . Vaisanen. M, L.. Moran. G,. Winer.M, & Finegold. S. M, (1995) Bacteriologyof skin and soft-tissue infections: compari-son of infections in intravenous drug usersand individuals with no history of intra-venous drug use, Ctinicat Infectious Di.';-eases 20 (suppl 2). S279-S282,

Sundqvist. G, (1992) Associations beiweenmicrobial species in dental root canal ln-feclions. Oral Microbiotogy and Immu-nology 7.257-262.

Tal. M, (1980) Periodontal disease and oralhygiene, De.scribed by Antonic van

Leeuwenhoek, Journal of Periodontotogv51. 668-669.

Umeda. M,. Tominaga. Y,. He. T. Yano. K..Walanabe. H, & Ishikawa, I. (1996) Mi-crobial flora in the acute phase of peri-odontitis and the effect of local adminis-tration of minocycline, Journat of Period-ontology 67. 422-427,

von Eye. A, (1990) Introduction to conftguratfrequency anatysis. The .search for typesand antitypes in cross-classifications. Cam-bridge University Press. Cambridge. UK.pp 16-26.

von Troil-Linden. B. Torrko. H.. Aialuusa.S,, Jousimies-Sonier. H. & Asikainen. S,(1995) Salivary levels of suspected peri-odontal pathogens in relation to peri-odontal status and treatment. Journal ofDental Research 74. 1789 1795,

Wolff. L. F.. Aeppli. D, M,. Pihistrom, B,.Anderson. L., Stoltenberg, J,. Osborn. J,.Hardie. N,. Shelburne. C, & Fischer. G.(1993) Natural distribution of 5 bacteriaassociated with periodontai disease.Journal of Clinical Periodontologv 20. 699-706,

Yao. E, S.. Lamont. R. J,. Leu. S, P & Wein-berg. A. (1996) Interbacterial bindingamong strains of pathogenic and commen-sal oral bacterial species. Oral Micro-biotogy antt Immunology 11. 35-41,

Address:

Signnmd S. SocranskyFor.syth Dentat Center140 The FenwayBoston MA. 02115USA