Periodontal Disease at theBioﬁlm–Gingival Interface

S. Offenbacher,* S.P. Barros,* R.E. Singer,* K. Moss,

†

R.C. Williams,* and J.D. Beck

†

Background:

A molecular epidemiologic study providedthe opportunity to characterize the biology of the bioﬁlm–gin-gival interface (BGI) in 6,768 community-dwelling subjects.

Methods:

Disease classifications and multivariable modelswere developed using clinical, microbial, inflammatory, andhost-response data. The purpose was to identify new clinicalcategories that represented distinct biologic phenotypesbased upon DNA checkerboardanalyses of eightplaque bac-teria,serumimmunoglobulinG(IgG)titersto17bacteria,andthe gingival crevicular fluid (GCF) levels of 16 inflammatorymediators. Five BGI clinical conditions were defined usingprobing depths (PDs) and bleeding on probing (BOP) scores.Subjects with all PDs

3 mm were grouped as BGI-healthy(14.3% of sample) or BGI-gingivitis (BGI-G, 15.1%). Subjectswith one or more PDs

‡

4 mm [deep lesion (DL)] were dividedinto low BOP (18.0%), moderate BOP (BGI-DL/MB, 39.7%),and severe BOP (BGI-DL/SB, 12.9%).

Results:

Subjects with BGI-G had increased levels of

Campylobacter rectus

–speciﬁc serum IgG levels (

0.01),and those with BGI-DL/SB had increased IgG levels to

Porphyromonas gingivalis

(

0.0003) and

C. rectus

(

0.01). BGI-DL/SB subjects had an excessive GCF interleukin(IL)-1

and prostaglandin E

response and an enhancedchronic inﬂammatory response with signiﬁcant increases inGCF IL-6 and monocyte chemotactic peptide-1. Within BGI-DL/SB subjects, more severe pocketing and BOP were asso-ciated with higher levels of GCF IL-1

, not higher microbialcounts or plaque scores.

Conclusions:

New BGI classiﬁcations create categorieswith distinct biologic phenotypes. The increased titers of

C.rectus

IgGamong68.5%oftheBGI-Gsubjectsandelevated

P. gingivalis

titers among BGI-DL/MB and BGI-DL/SB subjects (63.8% and 75.7%, respectively) are strongly supportiveofthemicrobialspeciﬁcityofpathogenesisforBGIcategories.

J Periodontol 2007;78:1911-1925.

KEY WORDS

Gingival crevicular ﬂuid; microbiology; periodontal disease;proteomics.

espite advances in our under-standing of the microbiology of the bioﬁlm, as well as the hostcellular and molecular mechanisms of pathogenesis, diagnostic categories forperiodontal classiﬁcation remain basedalmost exclusively upon medical anddental history and clinical signs andsymptoms of disease.

1-5

Unfortunately,current disease classifications that arebased upon clinical signs do not providemuch insight into the underlying sub-clinical biology of the process that in-volves a complex interaction of thebiofilm with the host inflammatory andimmune responses.Emergingmedicalmodelsofnosology(disease classification), such as that pro-posed by Casanova and Abel,

offer aconceptual framework for exploring dis-ease deﬁnitions that recognize the in-terplay of host and microbial biologicsystems. The application of these con-ceptstodeﬁneperiodontaldiseasewouldincorporate characterizing biologic sys-temsthatincludeuniqueindividualexpo-sures,e.g.,bacterialinfections,smoking,and hyperglycemia of diabetes

that in-teract with the genetic repertoire of theindividualhosttocreateabiologicpheno-type (a cascading cellular and molecularprocess that includes the mechanism of pathogenesis), which, over time, resultsin a clinical phenotype (the clinical pre-sentation of disease). It was suggestedby Casanova and Abel

that subtle ge-netic immunodeﬁciencies underlie mostmicrobial-based diseases in natural hu-man conditions. These investigators

* Center for Oral and Systemic Diseases and Department of Periodontology, School of Dentistry, University of North Carolina at Chapel Hill, Chapel Hill, NC.† DepartmentofDentalEcology,SchoolofDentistry,UniversityofNorthCarolinaatChapelHill.doi: 10.1902/jop.2007.060465

J Periodontol • October 2007

1911

suggested that opportunistic infections pose a tre-mendous diversity of potential bacterial exposuresthat, in combination with subtle genotype variationsinthehost,canresultinaspeciﬁcbiologicphenotypethat induces disease. Many types of underlying bio-logicphenotypesareprobablebecausethegeneticdi-versity in humans and the microbiota of the bioﬁlmpermits perhaps hundreds of different combinationsof microbes/host response. Despite this possible de-gree of complexity, there also may be shared path-ways with similar cellular and molecular responsesthatleadtocommonelementswithinthebiologicphe-notypesandsimilarclinicalpresentations.AspointedoutbyBaelumandLopez,

theunderlyingdiversityof biologic phenotypes that are currently ‘‘subclinical’’may not manifest as a variety of clinical phenotypesusingexistingdiseasedefinitions.Thus,thechallengefacingusscientifically,andthegoalofthisstudy,wastorefineexistingperiodontaldiseaseclassificationsina manner that will enable us to uncover and delineatethe various types of biologic phenotypes.The strategy we used in this study was to supple-ment clinical signs with diagnostic biomarker tests,includingmicrobial,inflammatory,andhostresponseassessments, to characterize the biologic phenotypeof each individual. We systematically examined howclinical signs could be grouped to segregate differentclinical disease classifications that differ from eachother biologically but share a similar biologic pheno-type within each group parsing a logical transitionfrom health to severe disease. Thus, in this study wesought to create new nosological classifications thatallowed the clinical definitions to reflect the biologicphenotype. The long-term rationale for attemptingtocreatenewclinicaldefinitionsthatshareacommonunderlying biologic phenotype is to improve the con-sistency of diagnosis and prediction of therapeuticresponse. Furthermore, by creating homogeneousgroups that share a common biologic phenotype,weshouldbeabletobetteridentifythehostandmicro-bial genetic contributions to the clinical conditions.In previous publications,

8-12

we provided clinical,microbial, and host-response characteristics of alarge community-dwelling adult population, catego-rizing subjects using traditional disease definitionsbased upon clinical signs. In this report, we soughtto identify those clinical signs that would enable ustosegregatediscreteclustersofindividualsthatcouldbe considered new, clinically relevant diagnostic cat-egories or syndromes that also reflected what washappening biologically at the biofilm–gingival inter-face(BGI).Wedevelopedmultivariablemodelstoas-certain which exposures or biologic characteristicsseemed to best describe the extent or severity of thecondition within each diagnostic category. In this re-port, we provide a new analysis that has exploredtheclinical,microbial,andhost-responsecharacteris-tics of this population. In addition, we present a de-tailed proteomic analysis of 16 cytokines within thegingival crevicular fluid (GCF) that was performedon a representative subset of 180 subjects. We com-bine all of these findings to demonstrate that newdisease categories can be defined that display signif-icantlydifferentpatternsofmicrobialburdenandhostresponsethatdefinethebiologyattheBGIinamannerthat would have been indistinguishable using tradi-tional disease classifications.

MATERIALS AND METHODS

Analytical Approach

The periodontal tissue–bioﬁlm interface is borderedonthegingivaltissuesidebytheepithelium(sulcular,pocket, and junctional) and the subgingival plaquewithinthepocket.Linearprobingdepth(PD)providesa reasonable approximation of the pocket surfaceareawhenconsideredatsixsitespertoothonallteethand expressed as summary or extent scores. Thedeeper the probeable pocket, the better the environ-ment for anaerobic subgingival bioﬁlm growth, andtherelationshipbetweenhigherperiodontalpathogencounts and increasing PD is well established.

Thus,PD can define one clinical attribute of the BGI thatshould influence the subgingival microbial composi-tion and growth density. When the pocket epitheliumis ulcerated or friable because of inflammatorychanges,bleedingonprobing(BOP)istheassociatedclinical sign. Thus, these two clinical signs (PD andBOP) were used to define current clinical disease sta-tus at the BGI. Although recession is an importantcomponent of attachment loss, it does not contributetotheanatomicalboundariesoftheBGI.Furthermore,attachment level, a measurement that is made rela-tive to a dental anatomical landmark, such as the ce-mento-enamel junction (CEJ),is also independent of the boundaries of the BGI and reflects historic attach-ment loss rather than current status.To determine the composition and magnitude of thebiofilmload,wedeterminedthelevelofpathogenspresent within the periodontal pocket using DNAcheckerboard methods in addition to traditional clini-cal measurements of plaque. The local host inflam-matory response is represented by measuring theconcentrationofbiochemicalmediatorsofinflamma-tion within the GCF. The patterns of these inflamma-tory mediators reveal the specific qualitative andquantitative aspects ofthe innate and chronic inflam-matory response. As one might expect, the concen-tration of inflammatory biomarkers within the GCFgenerally increased with greater microbial burden,evenamongsubjectswithoutanyclinicalsignsofdis-ease. However, by adjusting for the level of microbial

Disease at the Bioﬁlm–Gingival Interface

Volume 78 • Number 10

1912

burden using plaque scoresand microbial counts, wecandeterminewhetherthelevelofinflammationrela-tivetothelevelofbiofilmexposureforagivenBGIdis-ease group is in excess compared to the BGI-healthy(BGI-H)groupresponseand,therefore,representsanexcessive or hyperinflammatory response. Finally,the acquired immune response to the existing biofilmwas determined by measuring the serum antibodylevel to each pathogen. The immunoglobulin G(IgG) level to specific bacteria of the biofilm is an im-portantparameterbecause itimplicatesspecificbac-teria as being potentially etiologic as related to eachbiologic and clinical phenotype and reflects a sys-temic levelofexposureoftheorganism, albeitimper-fectly. The purpose of this investigation was to createclinical definitions of disease at the BGI that repre-sented different patterns of microbial, inflammatory,and immunologic parameters.

Periodontal Measurements

Periodontal examinations were performed on 6,768individuals living in four United States communities.This cross-sectional assessment of individuals alsoincluded detailed medical history data and was de-scribed in detail in several publications.

13-15

Full-mouth periodontal examinations were performed bycalibrated examiners on all teeth, including third mo-lars. Clinical assessments included plaque index,

gingivalindex(GI),

PDs(roundingdowntothenear-est millimeter), CEJ measurements, and attachmentlevels (computed from PD and CEJ) and were deter-minedinmillimetersatsixsitespertooth,aswasBOP(yes/no) at each site. The extent of BOP was ex-pressed as a percentage of all sites.

Measurements of Bacterial Plaque and Serum IgG Antibody

Levels of bacteria within plaque samples and serumantibodytiterstospeciﬁcorganismsweredeterminedonarandomsampleof1,673subjectsusingmethodsdescribed recently by Beck et al.

One plaque sam-plewasusedfromeachsubject,samplingthesubgin-gival mesio-buccal site of the maxillary right ﬁrstmolar, and assayed for 18 organisms by DNA wholechromosomal checkerboard arrays. For these ana-lyses, the levels of eight periodontal pathogens andthetiterswereusedfordiseasemodeldevelopment,in-cluding

Porphyromonas gingivalis

Treponema denti- cola

Tannerellaforsythia

(previously

T.forsythensis

Campylobacter rectus

Prevotella intermedia

Actino- bacillusactinomycetemcomitans

,and

Fusobacterium nucleatum

. Levels of organism were expressed ascounts using known microbial standards. SerumIgG levels were determined using immunochecker-board arrays using these same seven organisms plus

Parvimonas micra

(previously

Peptostreptococcus micros

Micromonas micros

Eikenella corrodens

Capnocytophagaochracea

Veillonellaparvula

Strep- tococcusoralis

Streptococcussanguinis

Streptococcus intermedius

Selenomonas noxia

, and

Actinomyces viscosus

as whole-cell antigens and were expressedas nanograms per milliliter of IgG-speciﬁc titers. Hu-man IgG at three concentrations using membrane-bound protein A was used on each membrane as aninternal standard. For the purposes of this analysis, thelevel of the organisms and the titer were consideredcontinuous variables.

Statistical Analyses

Clinical signs were expressed as extent scores

andusedascontinuousvariablestoexamineassociationswith biologic variables to create composite biologicphenotypes. Using extent PD, CAL (clinical attach-ment level), or GI scores as continuous variables torank individuals based upon disease severity did notcreate any consistent patterns in the levels of speciﬁcbacteria,theconcentration ofGCFinﬂammatoryme-diators,orantibodytiters.OnlyplaquelevelsandGCFinterleukin(IL)-1

levelswereweaklypositivelyasso-ciated with extent of disease using PD, GI, or CAL ascontinuous variables. Traditional definitions of peri-odontal status based upon attachment loss definitionsresulted in creating subgroups that did not differ inbiomarker level and were non-informative in con-structing a homogeneous biologic phenotype. Datamining with self-assembling hierarchical clusteringmethods resulted in biomarker subgroups that con-tained a diverse range of periodontal disease levelsusing traditional disease definitions or extent scores.However, the extent BOP and PD scores were associ-atedpositivelywithvariousbiomarkersandwereusedinitially to define clinical states. Most importantly, weiteratively selected the clinical definitions and thensoughttodeterminewhethertherewereunderlyingbi-ologic phenotypes with these defined categories. Inother words, we intentionally reverse engineered def-initionsofclinicalconditionsthatwerebaseduponbi-ologic data that defined the BGI. The final definitionsreflect cut-points in BOP and PD that resulted in sep-aratebiologicphenotypesandcreatedgradientlevelsof disease. Further subdivision of disease categoriesdid not result in different biologic phenotypes and,therefore, were used as aggregate categories withoutfurther subdivision. For example, segregating subjects with or without attachment loss within each of theBGIcategoriesdidnotresultintwogroupswithdif-fering levels of GCF mediator or microorganismcounts. In all multivariate models that included bio-logic phenotype markers, the use of indices, suchas GI, were supplanted by BOP in that BOP explainedagreaterpartofthevarianceofthefinaldiseasemodelthan GI, indicating that GI did not significantly im-provetheoverallmodel,i.e.,notstatisticallysignificant

J Periodontol • October 2007

Offenbacher, Barros, Singer, Moss, Williams, Beck

1913