J Periodontol • October 2007
Periodontal Disease at the Biofilm–Gingival Interface S. Offenbacher,* S.P. Barros,* R.E. Singer,* K. Moss,† R.C. Williams,* and J.D. Beck†
Background: A molecular epidemiologic study provided the opportunity to characterize the biology of the biofilm–gingival interface (BGI) in 6,768 community-dwelling subjects. Methods: Disease classifications and multivariable models were developed using clinical, microbial, inflammatory, and host-response data. The purpose was to identify new clinical categories that represented distinct biologic phenotypes based upon DNA checkerboard analyses of eight plaque bacteria, serum immunoglobulin G (IgG) titers to 17 bacteria, and the gingival crevicular fluid (GCF) levels of 16 inflammatory mediators. Five BGI clinical conditions were defined using probing 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%). Subjects with one or more PDs ‡4 mm [deep lesion (DL)] were divided into 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 (P = 0.01), and those with BGI-DL/SB had increased IgG levels to Porphyromonas gingivalis (P < 0.0003) and C. rectus (P < 0.01). BGI-DL/SB subjects had an excessive GCF interleukin (IL)-1b and prostaglandin E2 response and an enhanced chronic inflammatory response with significant increases in GCF IL-6 and monocyte chemotactic peptide-1. Within BGIDL/SB subjects, more severe pocketing and BOP were associated with higher levels of GCF IL-1b, not higher microbial counts or plaque scores. Conclusions: New BGI classifications create categories with distinct biologic phenotypes. The increased titers of C. rectus IgG among 68.5% of the BGI-G subjects and elevated P. gingivalis titers among BGI-DL/MB and BGI-DL/SB subjects (63.8% and 75.7%, respectively) are strongly supportive of the microbial specificity of pathogenesis for BGI categories. J Periodontol 2007;78:1911-1925. KEY WORDS Gingival crevicular fluid; microbiology; periodontal disease; proteomics.
* Center for Oral and Systemic Diseases and Department of Periodontology, School of Dentistry, University of North Carolina at Chapel Hill, Chapel Hill, NC. † Department of Dental Ecology, School of Dentistry, University of North Carolina at Chapel Hill.
D
espite advances in our understanding of the microbiology of the biofilm, as well as the host cellular and molecular mechanisms of pathogenesis, diagnostic categories for periodontal classification remain based almost exclusively upon medical and dental history and clinical signs and symptoms of disease.1-5 Unfortunately, current disease classifications that are based upon clinical signs do not provide much insight into the underlying subclinical biology of the process that involves a complex interaction of the biofilm with the host inflammatory and immune responses. Emerging medical models of nosology (disease classification), such as that proposed by Casanova and Abel,6 offer a conceptual framework for exploring disease definitions that recognize the interplay of host and microbial biologic systems. The application of these concepts to define periodontal disease would incorporate characterizing biologic systems that include unique individual exposures, e.g., bacterial infections, smoking, and hyperglycemia of diabetes, that interact with the genetic repertoire of the individual host to create a biologic phenotype (a cascading cellular and molecular process that includes the mechanism of pathogenesis), which, over time, results in a clinical phenotype (the clinical presentation of disease). It was suggested by Casanova and Abel6 that subtle genetic immunodeficiencies underlie most microbial-based diseases in natural human conditions. These investigators
doi: 10.1902/jop.2007.060465
1911
Disease at the Biofilm–Gingival Interface
suggested that opportunistic infections pose a tremendous diversity of potential bacterial exposures that, in combination with subtle genotype variations in the host, can result in a specific biologic phenotype that induces disease. Many types of underlying biologic phenotypes are probable because the genetic diversity in humans and the microbiota of the biofilm permits perhaps hundreds of different combinations of microbes/host response. Despite this possible degree of complexity, there also may be shared pathways with similar cellular and molecular responses that lead to common elements within the biologic phenotypes and similar clinical presentations. As pointed out by Baelum and Lopez,7 the underlying diversity of biologic phenotypes that are currently ‘‘subclinical’’ may not manifest as a variety of clinical phenotypes using existing disease definitions. Thus, the challenge facing us scientifically, and the goal of this study, was to refine existing periodontal disease classifications in a manner that will enable us to uncover and delineate the various types of biologic phenotypes. The strategy we used in this study was to supplement clinical signs with diagnostic biomarker tests, including microbial, inflammatory, and host response assessments, to characterize the biologic phenotype of each individual. We systematically examined how clinical signs could be grouped to segregate different clinical disease classifications that differ from each other biologically but share a similar biologic phenotype within each group parsing a logical transition from health to severe disease. Thus, in this study we sought to create new nosological classifications that allowed the clinical definitions to reflect the biologic phenotype. The long-term rationale for attempting to create new clinical definitions that share a common underlying biologic phenotype is to improve the consistency of diagnosis and prediction of therapeutic response. Furthermore, by creating homogeneous groups that share a common biologic phenotype, we should be able to better identify the host and microbial genetic contributions to the clinical conditions. In previous publications,8-12 we provided clinical, microbial, and host-response characteristics of a large community-dwelling adult population, categorizing subjects using traditional disease definitions based upon clinical signs. In this report, we sought to identify those clinical signs that would enable us to segregate discrete clusters of individuals that could be considered new, clinically relevant diagnostic categories or syndromes that also reflected what was happening biologically at the biofilm–gingival interface (BGI). We developed multivariable models to ascertain which exposures or biologic characteristics seemed to best describe the extent or severity of the condition within each diagnostic category. In this report, we provide a new analysis that has explored 1912
Volume 78 • Number 10
the clinical, microbial, and host-response characteristics of this population. In addition, we present a detailed proteomic analysis of 16 cytokines within the gingival crevicular fluid (GCF) that was performed on a representative subset of 180 subjects. We combine all of these findings to demonstrate that new disease categories can be defined that display significantly different patterns of microbial burden and host response that define the biology at the BGI in a manner that would have been indistinguishable using traditional disease classifications. MATERIALS AND METHODS Analytical Approach The periodontal tissue–biofilm interface is bordered on the gingival tissue side by the epithelium (sulcular, pocket, and junctional) and the subgingival plaque within the pocket. Linear probing depth (PD) provides a reasonable approximation of the pocket surface area when considered at six sites per tooth on all teeth and expressed as summary or extent scores. The deeper the probeable pocket, the better the environment for anaerobic subgingival biofilm growth, and the relationship between higher periodontal pathogen counts and increasing PD is well established.5 Thus, PD can define one clinical attribute of the BGI that should influence the subgingival microbial composition and growth density. When the pocket epithelium is ulcerated or friable because of inflammatory changes, bleeding on probing (BOP) is the associated clinical sign. Thus, these two clinical signs (PD and BOP) were used to define current clinical disease status at the BGI. Although recession is an important component of attachment loss, it does not contribute to the anatomical boundaries of the BGI. Furthermore, attachment level, a measurement that is made relative to a dental anatomical landmark, such as the cemento-enamel junction (CEJ), is also independent of the boundaries of the BGI and reflects historic attachment loss rather than current status. To determine the composition and magnitude of the biofilm load, we determined the level of pathogens present within the periodontal pocket using DNA checkerboard methods in addition to traditional clinical measurements of plaque. The local host inflammatory response is represented by measuring the concentration of biochemical mediators of inflammation within the GCF. The patterns of these inflammatory mediators reveal the specific qualitative and quantitative aspects of the innate and chronic inflammatory response. As one might expect, the concentration of inflammatory biomarkers within the GCF generally increased with greater microbial burden, even among subjects without any clinical signs of disease. However, by adjusting for the level of microbial
J Periodontol • October 2007
burden using plaque scores and microbial counts, we can determine whether the level of inflammation relative to the level of biofilm exposure for a given BGI disease group is in excess compared to the BGI-healthy (BGI-H) group response and, therefore, represents an excessive or hyperinflammatory response. Finally, the acquired immune response to the existing biofilm was determined by measuring the serum antibody level to each pathogen. The immunoglobulin G (IgG) level to specific bacteria of the biofilm is an important parameter because it implicates specific bacteria as being potentially etiologic as related to each biologic and clinical phenotype and reflects a systemic level of exposure of the organism, albeit imperfectly. The purpose of this investigation was to create clinical definitions of disease at the BGI that represented different patterns of microbial, inflammatory, and immunologic parameters. Periodontal Measurements Periodontal examinations were performed on 6,768 individuals living in four United States communities. This cross-sectional assessment of individuals also included detailed medical history data and was described in detail in several publications.13-15 Fullmouth periodontal examinations were performed by calibrated examiners on all teeth, including third molars. Clinical assessments included plaque index,16 gingival index (GI),17 PDs (rounding down to the nearest millimeter), CEJ measurements, and attachment levels (computed from PD and CEJ) and were determined in millimeters at six sites per tooth, as was BOP (yes/no) at each site. The extent of BOP was expressed as a percentage of all sites. Measurements of Bacterial Plaque and Serum IgG Antibody Levels of bacteria within plaque samples and serum antibody titers to specific organisms were determined on a random sample of 1,673 subjects using methods described recently by Beck et al.18 One plaque sample was used from each subject, sampling the subgingival mesio-buccal site of the maxillary right first molar, and assayed for 18 organisms by DNA whole chromosomal checkerboard arrays. For these analyses, the levels of eight periodontal pathogens and the titers were used for disease model development, including Porphyromonas gingivalis, Treponema denticola, Tannerella forsythia (previously T. forsythensis), Campylobacter rectus, Prevotella intermedia, Actinobacillus actinomycetemcomitans, and Fusobacterium nucleatum. Levels of organism were expressed as counts using known microbial standards. Serum IgG levels were determined using immunocheckerboard arrays using these same seven organisms plus Parvimonas micra (previously Peptostreptococcus micros or Micromonas micros), Eikenella corrodens,
Offenbacher, Barros, Singer, Moss, Williams, Beck
Capnocytophaga ochracea, Veillonella parvula, Streptococcus oralis, Streptococcus sanguinis, Streptococcus intermedius, Selenomonas noxia, and Actinomyces viscosus as whole-cell antigens and were expressed as nanograms per milliliter of IgG-specific titers. Human IgG at three concentrations using membranebound protein A was used on each membrane as an internal standard. For the purposes of this analysis, the level of the organisms and the titer were considered continuous variables. Statistical Analyses Clinical signs were expressed as extent scores19 and used as continuous variables to examine associations with biologic variables to create composite biologic phenotypes. Using extent PD, CAL (clinical attachment level), or GI scores as continuous variables to rank individuals based upon disease severity did not create any consistent patterns in the levels of specific bacteria, the concentration of GCF inflammatory mediators, or antibody titers. Only plaque levels and GCF interleukin (IL)-1b levels were weakly positively associated with extent of disease using PD, GI, or CAL as continuous variables. Traditional definitions of periodontal status based upon attachment loss definitions resulted in creating subgroups that did not differ in biomarker level and were non-informative in constructing a homogeneous biologic phenotype. Data mining with self-assembling hierarchical clustering methods resulted in biomarker subgroups that contained a diverse range of periodontal disease levels using traditional disease definitions or extent scores. However, the extent BOP and PD scores were associated positively with various biomarkers and were used initially to define clinical states. Most importantly, we iteratively selected the clinical definitions and then sought to determine whether there were underlying biologic phenotypes with these defined categories. In other words, we intentionally reverse engineered definitions of clinical conditions that were based upon biologic data that defined the BGI. The final definitions reflect cut-points in BOP and PD that resulted in separate biologic phenotypes and created gradient levels of disease. Further subdivision of disease categories did not result in different biologic phenotypes and, therefore, were used as aggregate categories without further subdivision. For example, segregating subjects with or without attachment loss within each of the BGI categories did not result in two groups with differing levels of GCF mediator or microorganism counts. In all multivariate models that included biologic phenotype markers, the use of indices, such as GI, were supplanted by BOP in that BOP explained a greater part of the variance of the final disease model than GI, indicating that GI did not significantly improve the overall model, i.e., not statistically significant 1913
Disease at the Biofilm–Gingival Interface
and affecting <5% of the total model variance. In a similar manner, all CAL and/or CEJ measurements were replaced by PD measurements. Thus, only PD and BOP were used to define the final BGI clinical categories. Differences between disease categories in patientbased characteristics, periodontal clinical measures, and GCF mediator levels were determined using general linear models for continuous variables and x2 tests for categorical variables. P values <0.05 were considered statistically significant. To determine whether there were overall differences among groups and to correct for multiple comparisons, the Hotelling T2 statistic was computed for levels of organisms and titers. Once an overall difference was established by Hotelling T2, group differences in levels and of each periodontal organism were based upon non-parametric analysis of rank scores using general linear models for overall significance and for group comparisons to the healthiest group (BGI-H). Group differences in IgG levels to specific organisms were computed using general linear models. To define which independent variables contribute to disease definition categories compared to BGI-H, we used logistic regression to create generalized logits models. Multivariate linear regression models were developed to define the extent BOP or extent PD ‡4 mm as the dependent variable within each disease category. GCF Analyses GCF was collected from the mesio-buccal aspect of each first molar or, if missing, an alternate site as reviewed in detail previously.20 Four GCF strips were eluted and analyzed separately for each subject. GCF analyte concentration data were pooled to provide a patient mean value in ng/ml for each analyte. The mean GCF levels of prostaglandin E2 (PGE2) and IL-1b were determined for 5,604 subjects. Sitespecific analyses of these data as related to clinical signs are described elsewhere.21 In this report, the patient pooled mean level of each mediator was used as a patient-level descriptor of whole-mouth periodontal inflammatory response. For a more comprehensive proteomic analysis of GCF mediators, we used a multiplex analytical platform as described previously.12 Multiple cytokine reagents permitted the analyses of 16 mediators simultaneously on each GCF sample. For these proteomic GCF analyses, a random sample of subjects was selected in each of the five BGI categories from the entire sample, matching frequency distribution on age (within 5 years), race, and gender. Smokers were omitted. A total of 180 subjects were selected for these analyses. Each GCF sample (four per patient at the disto-buccal aspect of each second premolar) was analyzed using cytokine profiling by multiplex immunoassay that permitted the simulta1914
Volume 78 • Number 10
neous measurement of 16 analytes in each GCF sample with the following mean minimum detection level in nanograms per milliliter: interferon gamma (IFN-g), 0.31; IL-17, 0.39; IL-1b, 0.27; IL-1 receptor antagonist (IL-1ra), 2.06; regulated upon activation in normal T cells expressed and secreted (RANTES), 1.08; monocyte chemotactic peptide 1 (MCP-1), 0.95; macrophage inflammatory protein 1 beta (MIP-1b), 2.12; epithelial-neutrophil activating peptide 78 (ENA-78), 2.71; IL-4, 1.75; IL-5, 0.33; IL-6, 0.36; IL-8, 0.39; tumor necrosis factor-alpha (TNF-a), 0.47; granulocytecolony stimulating factor (G-CSF), 0.57; IL-10, 0.13; and MIP-1a, 3.10. Means and SE were computed for the five BGI groups. Samples with mediator levels that were below the detection level or out of range exceeding the maximum standard were included in these analyses truncating the values using the lowest or highest standard value, respectively. The mean values for each mediator were adjusted for age, gender, and diabetes, and statistical differences between groups tested using general linear models with a P value <0.05 considered significant. RESULTS Exploratory Analyses to Create Definitions of Disease That Include Biologic Phenotype Data Because all subjects in this adult population (mean age, 62.4 [SD 5.6] years; range, 52 to 74 years) had some attachment loss and could be considered having periodontitis, we use the term BGI as a prefix to clarify that the clinical state is defined without consideration of attachment level or the position of the free gingival margin relative to the CEJ. Five levels of disease were defined based upon shallow (PD £3 mm) or deep (PD ‡4 mm) PDs in combination with low or high bleeding scores.9 Two shallow PD groups were defined: BGI-H was defined as all PDs £3 mm and <10% BOP (971 subjects); BGI-gingivitis (BGI-G; 1,023 subjects) was defined as individuals with all PDs £3 mm and ‡10% BOP. Three BGI deep lesion (DL) groups were created based upon whole-mouth PD and BOP scores. BGI-deep lesion/low bleeding (BGI-DL/LB; 1,217 subjects) was defined as one or more sites with PD ‡4 mm and BOP extent scores <10%; BGI-DL/moderate bleeding (BGI-DL/MB; 2,685 subjects) was defined as one or more sites with PD ‡4 mm and BOP extent scores between 10% and 50%; and BGI-DL/severe bleeding (BGI-DL/SB; 879 subjects) included subjects with one or more sites with PD ‡4 mm and BOP extent scores ‡50%. These five BGI categories created a gradient of disease severity while creating categories that differed by biologic phenotype at site-based and patient-based levels. The site-based analysis of the association between PD and BOP for each of the five BGI categories appears in Figure 1. The mean extent BOP score is
J Periodontol • October 2007
shown pooling all sites in all patients within each BGI disease level at each PD to create a PD–bleedingresponse curve for each category. These curves are similar to a dose-response function, whereby the dose is the level of the PD and the response is the mean bleeding score at that depth. Simply stated, the two categories BGI-H and BGI-G are subjects with shallow pockets with low BOP (<10%) or high BOP (‡10%), respectively. The BGI-DL/LB, BGI-DL/MB, and BGIDL/SB subjects have deeper pockets with low, moderate, and severe BOP, respectively. Although patient-based PD and BOP summary scores were used to create the five groups, the site-level bleeding status differed significantly among groups across various PDs, except for the low BOP groups BGI-H and BGI-DL/LB, which are indistinguishable for PDs £3 mm. For example, if one considers only 3-mm PD sites, 12.4% exhibited BOP in BGI-H, 58.6% had BOP among BGI-G, 10.7% had BOP among BGI-DL/LB, 45.5% had BOP in BGI-DL/MB, and 82.0% had BOP in BGI-DL/SB. This finding suggests that the underlying biology inducing the BOP response associated with 3-mm pockets is fundamentally different among subjects in the BGI-G, BGI-DL/MB, and BGI-DL/SB categories compared to the two lowest BOP groups (BGI-H and BGI-DL/LB). Alternatively, the underlying biology may be identical, but simply more prevalent at shallow sites among subjects with greater disease. Similarly, at 5-mm PDs, 31.8% of the sites exhibited BOP among subjects with BGI-DL/LB, which is significantly lower than the 67.2% BOP among BGI-DL/MB subjects and the 91.9% BOP among BGI-DL/SB subjects. Thus, combining PD and BOP measurements at
Figure 1. The percentage of sites with BOP are shown as a function of PD for each of the five BGI case definitions based upon PD and BOP scores for each subject. Data represent pooled site-based variables for each level for each BGI group: 971 BGI-H; 1,023 BGI-G; 1,217 BGI-DL/LB; 2,685 BGI-DL/MB; and 872 BGI-DL/SB.
Offenbacher, Barros, Singer, Moss, Williams, Beck
a site level to define a PD–BOP response curve creates logical clinical disease categories based upon clinical signs and suggests that the differences in bleeding responses seen between clinical phenotypes based on PD may reflect differing subject-level biologic phenotypes. Clinical Characteristics of BGI Classifications The subject characteristics for the five BGI category distributions with 14.3% as BGI-H, 15.1% as BGI-G, 18.0% as BGI-DL/LB, 39.7% as BGI-DL/MB, and 12.9% as BGI-DL/SB are shown in Table 1. There were significant BGI group differences with regard to race, gender, diabetes, education, smoking (history and pack-year), and body mass index (BMI). A variety of periodontal measurements, the GCF IL-1b and PGE2 levels, and the oral health behaviors are shown for the five BGI categories in Table 2. As expected, there are significant differences among all clinical measurements, and these values are provided for descriptive and comparative purposes. The mean extent of teeth with interproximal clinical attachment loss (ICAL) ‡3 mm was 23.9%, even among the healthiest BGI-H individuals, i.e., no clinically significant BOP or pocketing; it increased across the more diseased groups to 73.0% among the BGI-DL/SB group. Within the BGI-H group of 971 subjects, only 205 (3% of the total) had no sites with ICAL ‡3 mm. Within the 1,023 BGI-G subjects, only 161 (2.4% of the total) had no sites with ICAL ‡3 mm. Within each of the five BGI groups, those subjects without ICAL ‡3 mm were indistinguishable from those with one or more sites in terms of biologic phenotype and, therefore, were considered in aggregate. Similarly, the prevalence of disease using the Centers for Disease Control and Prevention (CDC) and Oral Conditions and Pregnancy (OCAP) definitions22,23 of disease are shown for comparison. The levels of GCF IL-1b and PGE2 also are lowest among those with BGI-H, increasing significantly among subjects with BGI-G, BGI-DL/LB, BGI-DL/MB, and BGI-DL/SB. Higher levels of these GCF mediators are seen in high BOP conditions (BGI-G and BGI-DL/SB). These five categories are clearly different from traditional definitions of health, gingivitis, mild periodontitis, moderate periodontitis, and severe periodontitis; however, they display a similar gradient with increasing severity and extent of clinical signs from BGI-H through BGI-DL/SB categories. However, there is an important exception to this trend with the BGI-DL/LB category, which has lower plaque (P = 0.03) and GI scores (P <0.001) but deeper pocketing compared to the healthiest group. The BGI-DL/LB category has an average of 5.69% sites with PD ‡4 mm, which is equivalent to an average of eight sites per subject with PD ‡4 mm, as computed using the mean number of teeth. 1915
Disease at the Biofilm–Gingival Interface
Volume 78 • Number 10
Table 1.
Subject Characteristics by BGI Level Subject Characteristic
BGI-H
BGI-G
BGI-DL/LB
BGI-DL/MB
BGI-DL/SB
P Value
Subjects (N [% total])
971 (14.3%)
1,023 (15.1%)
1,217 (18.0%)
2,685 (39.7%)
872 (12.9%)
Age (years; mean [SE])
62.2 (0.2)
62.4 (0.2)
62.4 (0.2)
62.4 (0.1)
62.8 (0.2)
Race: African American
314 (32.3%)
257 (25.1%)
140 (11.5%)
314 (11.7%)
275 (31.5%)
Race: white
657 (67.7%)
766 (74.9%)
1,077 (88.5%)
2,371 (88.3%)
597 (68.5%)
Gender: male
261 (26.9%)
378 (37.0%)
586 (48.2%)
1,352 (50.4%)
516 (59.2%)
Gender: female
710 (73.1%)
645 (63.1%)
631 (51.9%)
1,333 (49.7%)
356 (40.8%)
Diabetes: yes
105 (10.9%)
164 (16.1%)
128 (10.6%)
357 (13.4%)
180 (20.9%)
Diabetes: no
855 (89.1%)
853 (83.9%)
1,081 (89.4%)
2,316 (86.6%)
680 (79.1%)
Education: basic
141 (14.5%)
170 (16.7%)
86 (7.1%)
305 (11.4%)
215 (24.7%)
Education: intermediate
401 (41.3%)
440 (43.1%)
489 (40.2%)
1,239 (46.2%)
344 (39.5%)
Education: advanced
428 (44.1%)
410 (40.2%)
641 (52.7%)
1,139 (42.5%)
311 (35.8%)
Smoker: current
102 (10.6%)
88 (8.6%)
179 (14.7%)
324 (12.1%)
147 (17.0%)
Smoker: former
355 (36.8%)
342 (33.5%)
560 (46.1%)
1,142 (42.6%)
345 (39.9%)
Smoker: never
509 (52.7%)
590 (57.8%)
476 (39.2%)
1,215 (45.3%)
372 (43.1%)
<0.0001
Pack-years (mean [SE])
11.3 (0.7)
9.2 (0.6)
16.2 (0.6)
13.7 (0.4)
17.2 (0.7)
<0.0001
BMI (kg/m2; mean [SE])
28.5 (0.2)
28.8 (0.2)
27.7 (0.2)
28.6 (0.1)
29.3 (0.2)
<0.0001
0.20
<0.0001
<0.0001
<0.0001
<0.0001
The subject characteristics for the five BGI levels defining periodontal status using only PD and BOP extent scores. These disease classifications do not incorporate attachment level measurements. Basic education includes up through high school; intermediate education is some college; and advanced is some graduate education.
Compared to the BGI-DL/MB and BGI-DL/SB groups, the BGI-DL/LB group showed a significant trend for a greater prevalence of reported brushing two or more times per day, flossing daily, and visiting the dentist within the last year, a finding that is consistent with the lower plaque and GI scores. Biologic Phenotype of BGI Classifications The levels of periodontal biofilm organisms for each BGI group appear in Table 3 and the IgG titers appear in Table 4. The group differences in bacterial level and IgG titers are significant overall by the Hotelling T2 test (P <0.0001), which considers multiple comparisons. Relative to the BGI-H category, BGI-G had significantly higher total counts and orange complex counts, as well as specific elevations in P. intermedia, Prevotella nigrescens, and C. rectus (Table 3). Within this group, the dominant microbial shift is the significant twofold increase in C. rectus. The total bacterial counts and the total red and orange complex counts are significantly lower among BGI-DL/LB subjects compared to the BGI-H group, consistent with the clinical presentation of lower plaque scores and lower mean GI scores but the trend for higher BOP and PDs 1916
(Tables 2 and 3). All three of the red complex bacteria are significantly lower in the BGI-DL/LB group compared to the BGI-H group, as are the counts of C. rectus, F. nucleatum, and A. actinomycetemcomitans. The BGI-DL/MB subjects had significantly higher total, red complex, and orange complex microbial counts compared to BGI-H. Among the BGI-DL/MB subjects, the dominant red complex organisms were T. forsythia and T. denticola, which were elevated twofold and 1.8-fold, respectively. Among the BGI-DL/ SB subjects, the total bacterial count was elevated 2.5-fold compared to BGI-H subjects. This increase extended across the entire biofilm, including organisms of the red and orange complexes and A. actinomycetemcomitans. A similar, but not identical, pattern for the BGI group differences in serum IgG titers to the plaque bacteria are shown in Table 4, with a significant increase in total IgG levels in the two most severe disease categories. Compared to the BGI-H group, the BGI-G group had increased titers to C. rectus and a trend for higher V. parvula titers. The BGI-DL/LB group had significantly elevated titers to S. sanguinis and A. viscosus compared to BGI-H. Stronger IgG
Offenbacher, Barros, Singer, Moss, Williams, Beck
J Periodontol • October 2007
Table 2.
Periodontal Clinical Measurements (mean [SE]), GCF IL-1b and PGE2 Levels (mean ng/ml [SE]), and Oral Health Behaviors by BGI Category All PD £3 mm
Periodontal Measure
1+ Sites With PD ‡4 mm
BOP £10%
BOP >10%
BOP £10%
>10% BOP <50%
BOP ‡50%
BGI-H
BGI-G
BGI-DL/LB
BGI-DL/MB
BGI-DL/SB
P Value
23.9 0.29 0.16 4.6
23.5 0.48 0.35 26.1
19.4 0.99 1.15 71.3
(0.2) (0.02) (0.02) (0.40)
<0.0001 <0.0001 <0.0001 <0.0001
2.69 (0.02) 22.61 (0.32) 12.97 (0.23)
<0.0001 <0.0001 <0.0001
Teeth (N) Plaque score GI score Extent BOP
19.7 0.34 0.33 3.3
(0.2) (0.02) (0.02) (0.38)
PD Extent PD ‡4 mm Extent PD ‡5 mm
1.47 (0.01) – –
1.52 (0.01) – –
AL Extent AL ‡3 mm Extent AL ‡4 mm Extent of teeth with ICAL ‡3 mm Extent of teeth with ICAL ‡3 mm and PD ‡5 mm
1.38 12.5 4.2 23.9
1.41 13.6 4.9 27.1
(0.03) (0.66) (0.53) (0.84) –
18.5 0.54 0.56 30.1
(0.2) (0.02) (0.02) (0.36)
(0.03) (0.64) (0.52) (0.81) –
(0.2) (0.01) (0.01) (0.34)
1.84 (0.01) 5.69 (0.27) 1.61 (0.20) 1.67 21.41 9.6 39.7 6.5
(0.03) (0.59) (0.47) (0.75) (0.40)
(0.1) (0.01) (0.01) (0.23)
2.00 (0.01) 8.65 (0.18) 3.38 (0.13) 1.76 23.4 11.0 44.9 11.7
(0.02) (0.40) (0.32) (0.50) (0.30)
2.93 48.7 30.9 73.0 31.7
(0.03) (0.70) (0.56) (0.88) (0.50)
<0.0001 <0.0001 <0.0001 <0.0001 <0.0001
CDC definition* (n [%])* Healthy Moderate Severe
796 (82.0%) 175 (18.0%) 0 (0.0%)
832 (81.3%) 191 (18.7%) 0 (0.0%)
412 (33.9%) 654 (53.7%) 151 (12.4%)
701 (26.1%) 1,403 (52.3%) 581 (21.6%)
62 (7.1%) 361 (41.4%) 449 (51.5%)
<0.0001
OCAP definition† (n [%]) Health Mild Moderate/severe
613 (63.1%) 358 (36.9%) 0 (0.0%)
219 (21.4%) 804 (78.6%) 0 (0.0%)
0 (0.0%) 1,043 (85.7%) 174 (14.3%)
0 (0.0%) 1,905 (71.0%) 780 (29.1%)
0 (0.0%) 390 (44.7%) 482 (55.3%)
<0.0001
122.5 (4.3) 218.1 (5.3)
141.4 (2.9) 234.7 (3.6)
GCF mediators (ng/ml) IL-1b PGE2 Oral health behaviors Brush two or more times/day (n [%]) Floss daily (n [%]) Dental visit £1 year (n [%])
104.1 (4.8) 198.9 (6.3)
148.7 (4.7) 249.0 (6.3)
194.7 (5.3) 254.4 (6.8)
<0.0001 <0.0001
755 (78.3)
725 (71.4)
890 (73.4)
1,822 (68.1)
516 (59.4)
<0.0001
689 (71.5) 736 (76.5)
619 (61.5) 745 (73.3)
929 (76.6) 1,049 (86.6)
1,693 (63.3) 2,173 (81.3)
387 (44.6) 531 (61.4)
<0.0001 <0.0001
AAP = American Academy of Periodontology; AL = attachment loss; – = not applicable. * CDC/AAP definition of periodontal disease was derived independently during a 2003 meeting by a Working Group on Surveillance Systems for Periodontal Infections meeting. This three-level definition is as follows: severe periodontitis: at least two interproximal sites (not on same tooth) with CAL ‡6 mm and at least one interproximal site with PD ‡5 mm; moderate periodontitis: at least two interproximal sites with CAL 4 or 5 mm (not on same tooth) and no interproximal sites with CAL ‡6 mm; and healthy/gingivitis: individuals not meeting the above definitions. † In OCAP, periodontal disease was defined as a three-level variable: health, mild disease, or moderate-severe disease. Mothers with periodontal health were defined as having all PDs £4 mm, with no 3- or 4-mm sites showing BOP. Moderate-severe periodontitis subjects had 15 or more sites with PDs >4 mm; the remaining subjects were placed in the mild category.
responses are seen among the BGI-DL/MB group that displayed elevated titers for the total biofilm, red complex organisms (T. forsythia and T denticola), C. rectus, P. micra, A. actinomycetemcomitans, and S. sanguinis compared to the BGI-H group. In the BGI-DL/SB group, the total microbial counts increased 2.5-fold compared to the BGI-H group (Table 3). This was accompanied by a 27% increase in IgG level titer against the total biofilm. However, two bac-
terial IgG responses are rather dramatic, with P. gingivalis titers increasing 3.1-fold and C. rectus IgG levels increasing 1.8-fold (followed by A. actinomycetemcomitans titers, 1.4-fold), compared to BGI-H. BGI-DL/ SB also demonstrated moderately increased titers to T. denticola, P. intermedia, P. micra, and C. ochracea. To determine whether there were different underlying biologic phenotypes for each of the four BGI disease categories, we combined the patient background 1917
Disease at the Biofilm–Gingival Interface
Volume 78 • Number 10
Table 3.
Level of Selected Periodontal Organisms by BGI Category Median Counts ·103 (interquartile range) BGI-H (n = 186)
IgG Antibody Total biofilm counts
BGI-DL/LB (n = 261)
BGI-DL/MB (n = 562)
BGI-DL/SB (n = 218)
27.3 (11.8 to 77.9) 37.8 (17.9 to 108.1)† 22.0 (9.7 to 44.9)†‡ 40.3 (16.4 to 186.7)†‡ 67.1 (26.4 to 335.3)†‡
P. gingivalis T. forsythia T. denticola Red biofilm counts P. intermedia P. nigrescens C. rectus F. nucleatum Orange biofilm counts
BGI-G (n = 204)
1.3 1.0 2.1 5.2
(0.0 (0.0 (0.0 (1.9
to to to to
4.0) 3.9) 8.0) 15.8)
1.1 1.5 3.0 7.1
3.7 2.7 2.0 5.3 21.4
(0.0 (0.0 (0.0 (0.0 (6.3
to to to to to
16.9) 9.9) 9.3) 22.7) 61.5)
6.3 4.6 4.0 9.0 29.1
A. actinomycetemcomitans
1.6 (0.0 to 5.8)
(0.0 (0.0 (0.0 (2.5
to to to to
3.2) 4.8) 9.8) 16.5)
(0.0 to 18.9) (0.0 to 17.3)† (0.8 to 13.6)† (0.0 to 34.0) (11.2 to 90.1)†
2.4 (0.2 to 7.8)
†
†‡
0.7 0.5 1.2 3.6
(0.0 (0.0 (0.0 (1.3
to to to to
2.3) 2.4)† 5.7)† 10.1)†
1.5 2.0 3.8 8.1
(0.0 (0.0 (0.0 (2.7
to to to to
4.4) 7.1)†‡ 15.3)† 25.5)†‡
2.3 3.3 7.0 13.6
(0.2 (0.8 (1.4 (4.6
2.7 1.1 1.2 2.5 15.1
(0.0 (0.0 (0.0 (0.0 (5.3
to to to to to
10.4) 6.1 7.2) 5.8 5.7)† 3.9 9.9 11.4)† 31.5)†‡ 29.4
(0.0 (0.0 (0.0 (0.0 (9.1
to to to to to
23.8) 24.7)†‡ 18.3)†‡ 65.2)†‡ 140.6)†‡
10.1 8.1 6.8 14.8 48.8
(0.0 to 43.5)†‡ (0.5 to 34.4)†‡ (1.8 to 29.3)†‡ (0.6 to 93.8)†‡ (16.2 to 247.1)†‡
1.1 (0.0 to 4.0)†
2.7 (0.3 to 8.5)†
to to to to
6.9) 14.7)†‡ 24.5)†‡ 50.6)†‡
3.5 (1.0 to 10.8)†‡
P Value* <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001
* P values are for group differences as determined using non-parametric one-way analysis of rank scores. † Statistically significant compared to BGI-H at P <0.05. ‡ Statistically different compared to BGI-H at P <0.01.
characteristics, plaque levels, GCF inflammatory response, microbial plaque composition, and serum IgG response (Table 5) and developed generalized logits models. Compared to BGI-H, BGI-G had signifi-
cantly increased plaque, and the subjects tended to be white, male, and diabetic. Plaque, race, gender, and diabetes were significant in all four BGI disease models, with plaque, gender, and diabetes having
Table 4.
Biofilm-Specific Serum IgG Levels (ng/ml [SD]) by BGI Level IgG Antibody
BGI-H
BGI-G
BGI-DL/LB
BGI-DL/MB ‡
BGI-DL/SB †‡
P Value*
Total biofilm IgG
1,067 (53)
1,140 (51)
1,143 (50)
1,204 (32)
P. gingivalis T. forsythia T. denticola Red biofilm IgG
53.8 37.8 28.9 122.1
(9.1) (2.8) (1.6) (11.3)
50.7 37.8 31.0 120.2
(8.8) (2.7) (1.5) (10.9)
60.9 40.1 28.1 130.6
(8.4) (2.6) (1.5) (10.5)
69.4 46.6 33.4 150.9
(5.5) (1.7)†‡ (1.0)†‡ (6.9)‡
168.4 45.5 36.5 251.8
(9.5)†‡ (2.9) (1.6)†‡ (11.7)†‡
<0.0001 0.009 0.003 <0.0001
P. intermedia P. nigrescens C. rectus F. nucleatum P. micra Orange biofilm IgG
82.2 138.2 25.7 16.5 88.3 351.2
(5.8) (8.6) (2.4) (1.3) (6.4) (18.6)
74.5 153.6 33.1 19.7 96.6 377.7
(5.7) (8.4) (2.3)‡ (1.3) (6.2) (18.0)
77.2 128.1 25.5 14.2 104.6 350.2
(5.4) (8.0) (2.2) (1.2) (5.9) (17.3)
83.1 143.7 33.1 17.0 108.8 386.3
(3.5) (5.3) (1.4)†‡ (0.8) (3.9)†‡ (11.3)
108.0 144.0 47.4 16.3 112.1 428.2
(6.1)‡ (9.0) (2.5)†‡ (1.4) (6.7)†‡ (19.4)†‡
0.0005 0.27 <0.0001 0.04 0.03 0.02
A. actinomycetemcomitans E. corrodens C. ochracea V. parvula S. intermedius S. oralis S. sanguinis S. noxia A. viscosus Other biofilm IgG
122.8 18.8 31.6 11.0 92.6 61.6 65.0 92.5 76.0 574.4
(7.8) (1.6) (2.0) (1.1) (9.7) (7.1) (6.9) (11.9) (7.8) (33.2)
136.0 21.6 34.7 14.1 91.4 76.0 69.0 108.7 79.3 631.8
(7.6) (1.5) (2.0) (1.1)‡ (9.4) (6.9) (6.7) (11.6) (7.5) (32.2)
120.5 15.3 30.3 12.6 108.4 68.1 88.4 95.0 104.7 643.8
(7.3) (1.5) (1.9) (1.0) (9.0) (6.7) (6.4)‡ (11.1) (7.2)†‡ (30.8)
146.7 18.8 34.3 12.0 112.0 77.1 87.0 69.9 92.2 651.0
(4.8)†‡ (1.0) (1.2) (0.7) (5.9) (4.3) (4.2)†‡ (7.2) (4.7) (20.2)‡
177.2 21.8 42.0 13.6 94.8 81.7 78.1 70.7 80.3 662.4
(8.2)†‡ (1.6) (2.1)†‡ (1.1) (10.1) (7.5) (7.2) (12.5) (8.1) (34.7)
<0.0001 0.02 0.0006 0.23 0.20 0.24 0.02 0.03 0.03 0.33
Bold text highlights statistically significant findings. * P values are for group differences as determined by least square procedures. † Statistically significant compared to BGI-H at P <0.01. ‡ Statistically different compared to BGI-H at P <0.05.
1918
1,353 (55.5)
0.003
Offenbacher, Barros, Singer, Moss, Williams, Beck
J Periodontol • October 2007
Table 5.
Generalized Logits Models for BGI-G, BGI-DL/LB, BGI-DL/MB, and BGI-DL/SB Using the BGI-H Level as a Reference Odds Ratio (95% confidence interval)* Variable
BGI-G
BGI-DL/LB
Extent PI >1 (10%)
1.15 (1.10 to 1.19)
†
1.05 (>1.00 to 1.09)
Race (white)
3.09 (2.11 to 4.52)†
Gender (male)
BGI-DL/MB †
BGI-DL/SB †
1.38 (1.32 to 1.45)†
2.32 (1.57 to 3.42)†
3.97 (2.80 to 5.61)†
2.79 (1.83 to 4.25)†
1.74 (1.30 to 2.33)†
2.46 (1.87 to 3.23)†
2.56 (1.99 to 3.29)†
4.25 (3.09 to 5.84)†
Diabetes
1.76 (1.17 to 2.66)†
1.54 (1.01 to 2.34)†
1.65 (1.14 to 2.41)†
2.22 (1.44 to 3.43)†
GCF IL-1b (0.1 ng/ml)
1.43 (1.24 to 1.66)†
1.22 (1.05 to 1.41)†
1.47 (1.29 to 1.69)†
1.80 (1.55 to 2.08)†
GCF PGE2 (0.1 ng/ml)
1.18 (1.10 to 1.27)†
1.05 (0.97 to 1.14)
1.11 (1.03 to 1.18)†
1.16 (1.07 to 1.26)†
C. rectus IgG (0.1 mg/ml)
1.44 (1.14 to 1.83)†
1.02 (0.79 to 1.32)
1.31 (1.05 to 1.64)†
1.45 (1.13 to 1.87)†
P. gingivalis IgG (0.1 mg/ml)
1.00 (0.90 to 1.11)
1.14 (1.04 to 1.25)†
1.17 (1.07 to 1.27)†
1.25 (1.14 to 1.37)†
1.22 (1.18 to 1.27)
PI = plaque index. * Also adjusted for age, education, number of teeth, smoking (current, former, never, and pack-years), and BMI. † Statistically significant at P <0.05.
larger effects in the BGI-DL/SB group. Smoking history or pack-years were not significantly different among these four BGI groups compared to BGI-H. Compared to BGI-H subjects, all four disease groups had significantly increased GCF levels of IL-1b. GCF PGE2 was significantly elevated in the BGI-G, BGI-DL/ MB, and BGI-DL/SB, groups compared to the BGI-H group. In all four BGI models, the serum IgG level supplanted the level of organism present. The BGI-G group had significantly increased serum levels of C. rectus IgG. The C. rectus–specific antibody was selected among all IgG titers as the single best predictor of BGI-G using logistic regression to select the best single model. For the BGI-DL/LB, BGI-DL/MB, and BGI-DL/SB groups, P. gingivalis was the best single predictor. P gingivalis followed by C. rectus were the two best IgG predictors of BGI-DL/MB or BGI-DL/ SB using logistic regression to identify the best two predictors. For this reason, the IgG titers for both organisms are included in these models; however, the P. gingivalis titer does not contribute significantly to BGI-G, and the C. rectus titer does not contribute significantly to the BGI-DL/LB model. Collectively, these data strongly support the concept that BGI-G is associated with an increased C. rectus titer without significant P. gingivalis exposure and that BGI-DL/MB and BGI-DL/SB are characterized by the exposure of C. rectus and the additional microbial antigenic exposure of P. gingivalis. This statement is supported further by the fact that two independent methods (IgG titer and DNA-based microbial counts) both selected for these two microbes. Furthermore, in these composite phenotypes that define the BGI disease states,
it is clear that within each diagnostic category there seem to be different biologic aspects related to the inflammatory and antibody response that we collectively refer to as a biologic phenotype. The relatively large sample size of this molecular epidemiologic study enabled us to describe those biologic phenotypes within each of the disease categories (Table 5) and to define those that contribute to more severe disease within each BGI group (Table 6). Using multivariable linear regression models to explain the extent of BOP and extent PD ‡4 mm as dependent variables, the relative contribution of each biologic parameter was examined adjusting for other factors. With each 10% increase in plaque extent scores above 1, there was a significant 1.2% increase in extent BOP scores in the BGI-G subjects. The extent of plaque had significant, but smaller, effects on bleeding scores among BGI-DL/LB, BGI-DL/MB, and BGI-DL/SB subjects. Increasing plaque levels were associated with greater extent of PDs (‡4 mm) only among the BGI-DL/LB and BGI-DL/MB groups, and the effects of plaque on BOP were much stronger than the effects on PDs. Plaque scores had no effect on the extent PD ‡4 mm among BGI-DL/SB subjects. Greater BOP and extent PD ‡4 mm scores were found among African American BGI-DL/SB subjects (negative b coefficient), whereas white BGI-DL/LB subjects were more likely to have higher BOP levels. Diabetes did not significantly increase the extent of BOP or extent PD ‡4 mm within disease groups. Former smoking history significantly inhibited BOP among BGI-DL/MB subjects, and current smoking had a major impact on extent PD ‡4 mm among BGI-DL/ 1919
Disease at the Biofilm–Gingival Interface
Volume 78 • Number 10
Table 6.
Multivariable Regression Models* for Extent of BOP and PD ‡4 mm Within Each BGI Disease Level Extent PD ‡4 mm
Extent BOP Variable
BGI-G
BGI-DL/LB
BGI-DL/MB
BGI-DL/SB
BGI-DL/LB
Extent PI >1 1.21 (<0.0001) 0.18 (<0.0001) 0.50 (<0.0001) 0.74 (0.001) 0.16 (0.03) (per 10%) Race (white) -0.24 (0.92) 0.95 (0.04) 0.70 (0.50) -8.86 (<0.0001) 0.01 (0.99) Diabetes Smoking: never Smoking: former Smoking: current Smoking: pack-years BMI
3.46 (0.11) Ref
0.51 (0.18) Ref
-0.45 (0.59) Ref
1.53 (0.38) Ref
-0.38 (0.54) Ref
BGI-DL/MB
BGI-DL/SB
0.31 (<0.0001) 0.27 (0.30) 0.29 (0.68) -0.25 (0.65) Ref
-8.01 (0.0004) 3.63 (0.08) Ref
-3.30 (0.11)
0.15 (0.61)
-1.50 (0.04)
-1.24 (0.49)
-0.39 (0.43)
-2.00 (0.59)
0.44 (0.33)
-2.26 (0.055)
0.62 (0.82)
0.99 (0.18)
-0.05 (0.95)
-0.12 (0.13)
0.02 (0.91)
0.04 (0.92)
0.61 (<0.0001) 0.56 (<0.0001) 0.50 (0.28)
-0.08 (0.54)
0.03 (0.22)
0.01 (0.80)
0.19 (0.15)
0.02 (0.70)
0.21 (0.11)
0.23 (0.40)
0.64 (0.004)
0.80 (0.0002) 0.47 (0.01)
1.05 (0.0001)
0.19 (0.002)
0.33 (0.04)
0.29 (0.02)
0.14 (0.14)
0.70 (<0.0001)
GCF IL-1b 1.51 (0.01) (0.1 ng/ml) P. gingivalis IgG 0.31 (0.57) (0.1 mg/ml)
0.26 (0.58)
1.13 (0.60)
4.14 (<0.0001) 8.84 (0.005)
0.15 (<0.0001) 0.23 (0.16)
0.35 (0.0008)
Bold values highlight statistically significant findings. PI = plaque index; Ref = reference group to which other categories are compared. Table shows only significant independent variables, parameter estimates (b coefficient), and (P values). * Also adjusted for age, gender, smoking, education, number of teeth, GCF PGE2, and C. rectus IgG titer.
MB and BGI-DL/SB subjects. BMI had a significant effect on extent PD ‡4 mm within the BGI-DL/MB group. Diabetes and BMI had a non-significant trend toward increasing the extent PD ‡4 mm in BGI-DL/SB. Increased GCF levels of IL-1b significantly increased BOP and extent PD ‡4 mm across all disease categories. For example, for each 1 ng/ml increase in GCF IL-1b level, there was a 10.5% increase in extent PD ‡4 mm among BGI-DL/SB, which, on average, was 12 more sites with PD ‡4 mm. The BGI-DL/LB subjects had greater extent PD ‡4 mm with increasing GCF IL-1b. Although GCF PGE2 levels were higher among all three disease groups (Table 5), the levels were not statistically significantly associated with increased disease expression within these diseased groups. Serum titers to C. rectus, although predictive of the presence of BGI-G (Table 5), did not significantly predict the severity of the condition, nor did they contribute to the extent of BOP or PD ‡4 mm among any BGI level. In contrast, the IgG levels to P. gingivalis were increased with higher BOP scores among the BGIDL/MB and BGI-DL/SB groups, as well as increased extent PD ‡4 mm. The increased P. gingivalis IgG titers had the largest effect on increasing PDs in the BGI-DL/SB subjects. 1920
The GCF levels of 16 inflammatory mediators among the subset of 180 individuals within the five BGI groups (randomly selected, but balanced on age, race, and gender) appear in Table 7. These mean values are adjusted for the extent of PD ‡4 mm, race, gender, age, diabetes, and BMI. Thus, these may reflect subject-level differences in the qualitative and quantitative nature of the inflammatory response, independent of the level of disease or microbial burden present. Several mediators were not significantly different among groups, including IL-10, -4, -5, -8, -17, and -1ra, MIP-1a and -1b, IFN-g, and ENA-78. The adjusted means confirm that GCF IL-1b showed a trend for increases across diseased groups, showing statistically significant increases among the BGIDL/MB and BGI-DL/SB groups using a multiplex analytical platform. GCF PGE2 levels showed similar increases, with a significant 1.75-fold increase among the BGI-DL/SB group (P = 0.006; data not shown). These data also demonstrated that there were additional inflammatory responses within the BGI-DL/ SB group, with a 4.5-fold increase in MCP-1 and a 6.2-fold increase in IL-6. The 2.1-fold increase in TNF-a should be interpreted with caution because there were no overall statistical differences for the group (P = 0.29), nor were there overall group differences
Offenbacher, Barros, Singer, Moss, Williams, Beck
J Periodontol • October 2007
Table 7.
Detailed Proteomic GCF Inflammatory Mediator Analysis of Subset of Subjects From BGI Groups GCF Mediators (ng/ml; mean [SE])
BGI-H (n = 60)
BGI-G (n = 60)
BGI-DL/LB (n = 10)
BGI-DL/MB (n = 20)
BGI-DL/SB (n = 30)
P Value*
IL-10
1.40 (0.12)
1.52 (0.12)
1.48 (0.28)
1.35 (0.20)
1.54 (0.22)
0.90
IL-1b
†
†‡
253.6 (40.4)
342.3 (39.6)
291.5 (91.4)
437.9 (66.7)
IL-4
1.07 (0.32)
1.50 (0.32)
0.73 (0.73)
1.80 (0.53)
2.06 (0.57)
0.46
IL-5
0.06 (0.03)
0.07 (0.03)
0.08 (0.06)
0.13 (0.04)
0.05 (0.05)
0.68
IL-6
0.90 (0.64)
1.93 (0.63)
2.22 (1.44)
1.74 (1.05)
5.62 (1.12)†‡
0.02
IL-8
1,205 (140)
1,211 (137)
1,155 (316)
1,430 (231)
1.78 (0.39)
2.12 (0.38)
2.35 (0.88)
2.75 (0.64)
3.70 (0.68)†
0.29
†
0.17
TNF-a
518.6 (70.9)
1,698 (245)
0.03
0.53
G-CSF
82.30 (15.0)
108.1 (14.7)
65.7 (33.9)
126.9 (24.7)
148.7 (26.3)
MIP-1a
25.6 (3.94)
32.42 (3.87)
27.8 (8.93)
28.6 (6.51)
40.6 (6.92)
0.37
MIP-1b
29.3 (5.17)
38.6 (5.07)
29.81 (11.70)
27.15 (8.53)
41.2 (9.07)
0.47
-0.13 (0.36)
0.33 (0.35)
0.28 (0.81)
0.12 (0.59)
2.45 (0.63)†‡
0.02
MCP-1
3.18 (1.59)
7.12 (1.56)
6.97 (3.59)
5.61 (2.62)
14.44 (2.78)†‡
0.02
IFN-g
2.03 (0.73)
2.12 (0.72)
2.18 (1.66)
3.56 (1.21)
4.60 (1.29)
0.56
IL-17
0.28 (0.11)
0.55 (0.11)
0.30 (0.25)
0.40 (0.18)
0.66 (0.20)
0.29
ENA-78
84.9 (9.5)
92.1 (9.4)
57.8 (21.6)
58.9 (15.7)
86.7 (16.7)
0.30
IL-1ra (/100)
21.5 (0.97)
23.6 (0.95)
21.6 (2.20)
24.2 (1.60)
25.6 (1.70)
0.25
RANTES
Significant values are highlighted in bold. Data have been adjusted for extent PD ‡4 mm, race, gender, age, diabetes, and BMI. * P values are for group differences as determined by mean least squares. † Statistically elevated compared to BGI-H at P <0.05. ‡ Statistically elevated compared to BGI-H at P <0.01.
in G-CSF. Although there was a statistically significant increase in RANTES in BGI-DL/SB, this GCF mediator was detectible in only 1.7% of all 720 GCF samples analyzed. DISCUSSION These data strongly support the concept that defining periodontal status using clinical measurements of PD and BOP levels that reflect the integrity of the BGI can create diagnostic subgroups of individuals who have differing biologic phenotypes, even after considering traditional clinical risk factors for periodontal disease. The BGI categories make sense clinically because they logically create a natural gradient in disease severity that can be identified easily using simple clinical measurements. Furthermore, because the categories display different biologic phenotypes with regard to the underlying qualitative and quantitative aspects of the microbial and inflammatory response, these
groupings eventually may be shown to provide useful categories for studying response to treatment. For example, in this United States population, the levels of plaque, gender, diabetic status, and race were significant contributors to the presence of BGI disease expression, but the impact of these were not uniform across different BGI categories. Subjects with deep lesions, but low BOP, have less plaque than BGI-H subjects but more biochemical inflammation, raising the question as to whether anti-inflammatory strategies may be more efficient therapeutic approaches than antimicrobial treatments in these subjects. Levels of specific bacteria are undoubtedly significant contributors to the presence or absence of disease in these BGI models, but the nature of the microbial challenge and the host innate and acquired immune response to the oral biofilm also seem to be critical to disease presentation. Increased expression of IL-1b and PGE2, which are important biochemical mediators of periodontal tissue destruction,20,21 is 1921
Disease at the Biofilm–Gingival Interface
an essential characteristic of patients with disease at the BGI. This is true even for the BGI-G subjects, suggesting that this represents an inflammatory state that may not be innocuous. The level of GCF mediators of the innate immune response is high, even adjusting for the level of plaque or the level of organism present, compared to subjects with BGI-H; for that reason, we use the term ‘‘excessive’’ or ‘‘hyperinflammatory’’ as a descriptor. In essence, this represents a shift in the patient’s dose-response relationship whereby low levels of microbes (dosage) might not trigger IL-1b release (response) in a BGI-H subject, but the same level of microbial challenge will trigger a robust IL1b response in BGI-DL subjects. For example, consider that the periodontal sulcus of a typical BGI-H subject has a total microbial count of ;27,300 bacteria (median) and a mean GCF level of IL-1b of 104 ng/ ml. When one does the calculation (molecular weight IL-1b: ;10,162 Da), there are ;1 trillion (1012) molecules of IL-1b within the GCF for every bacterium present within the pocket. However, if one compares the response in the mildly upregulated BGI-DL/LB group, there are total counts of ;22,000 bacteria with a mean GCF level of IL-1b of 122 ng/ml. This computes to almost 1.4 times as many IL-1b molecules per bacterium for this BGI-DL/LB group and >1.6 trillion IL-1b molecules per single bacterium in the pocket for BGI-DL/SP. Therefore, it is not just an increase in concentration of IL-1b within the pocket, it is an increase in concentration relative to the microbial load. We are unsure as to the mechanisms that underlie this upregulated inflammatory response because many explanations are plausible. Nonetheless, this observation in a large community-dwelling population provides a robust demonstration that subjects with disease have a high innate inflammatory response locally at the BGI, which is in excess of that modeled by plaque scores or levels of organisms as determined by DNA checkerboard methods. It also is possible that the excessive inflammatory response may be attributable to the presence of unidentified, perhaps uncultivable, microorganisms that were not represented by the plaque scores or the checkerboard analyses in this investigation. However, this study also indicated that there is a characteristic systemic antibody response to selected organisms within the biofilm, in addition to the high microbial counts that discriminate between BGI disease classifications, thereby implicating specific organisms in pathogenesis among certain subjects within these BGI groups. Microbial and Seroreactive Phenotypes of Shallow and Deep BGI Lesions This investigation demonstrated that BGI-G is associated with an acquired immune response to C. rectus. 1922
Volume 78 • Number 10
Among seropositive BGI-G subjects, 68.5% had elevated titers to C. rectus (above the median level of BGI-H subjects). This is a well-known orange complex periodontal pathogen that was associated with gingivitis and early periodontitis in previous studies by Tanner et al.24 and is related closely phylogenetically, i.e., 96% homology, to the other gastrointestinal pathogen Helicobacter pylori.25 The association of this pathogen with high bleeding scores in shallow pockets raises the possibility that C. rectus may induce ulceration of the pocket epithelium and therefore, BOP, in a manner similar to that elicited by the ulcerogenic H. pylori on the gut epithelium. The level of organism is correlated highly to the serum antibody response, but the serum IgG response is associated more strongly with disease, probably reflecting the fact that this organism is evading host defenses and, thereby, eliciting local inflammation and BOP; this may result in a greater systemic exposure and antibody response. Collectively, these microbial and IgG responses provide two independent lines of evidence to strongly implicate the role of C. rectus in gingival inflammatory changes that occur in the shallow gingival pocket. In a similar manner, the P. gingivalis IgG response emerges as a dominant player in the more advanced BGI deep lesion of BGI-DL/MB and BGI-DL/SB, with 63.8% and 75.7% of subjects having elevated titers, respectively, compared to the BGI-H median. Increases in IgG titer to P. gingivalis were associated with greater bleeding scores and extent of deep pocketing in subjects with BGI-DL/MB and BGI-DL/SB disease, with GCF IL-1b levels enhancing BOP among BGI-DL/SB. The presence of elevated titers to C. rectus as the second best IgG predictor (following P. gingivalis titers) for the BGI-DL/MB and BGI-DL/SB categories suggested that the role of C. rectus in mediating the shallow lesion also may persist in these deep lesion categories. Smoking, a well-established traditional risk factor,26 also plays a major role in disease as an ‘‘inhibitor’’ of bleeding scores in BGI-DL/MB and a strong ‘‘enhancer’’ of greater pocketing (extent PD ‡4 mm). Our findings also are consistent with an earlier report by Saito et al.,27 regarding the role of obesity as a contributing factor to periodontal disease expression. One BGI classification seems to mirror a common clinical presentation of a patient with deep pockets, but minimal BOP, e.g., a ‘‘healthy or stable’’ periodontal recall patient. BGI-DL/LB subjects have pocketing, but minimal clinical inflammation. This clinical presentation is consistent with the observation that these subjects have lower plaque scores and greater dental healthcare use patterns. Although we have no data regarding specific periodontal treatment history, many in this group might have had periodontal care
J Periodontol • October 2007
included in their dental treatments. These subjects are quite similar to the BGI-H group, except they have deeper pockets and show higher IL-1b levels within the GCF and higher P. gingivalis titers, despite having lower counts of most organisms, including P. gingivalis and C. rectus (Table 3). This BGI-DL/LB group also is distinguished by the lack of increased C. rectus titers, a trait that is seen among all other disease categories. The low C. rectus counts and serum IgG antibody in this group are consistent with low BOP scores, especially compared to the BGI-G group. Furthermore, the BGI-DL/LB group had the highest ratio of IgG/bacteria for P. gingivalis and C. rectus compared to all other BGI groups; this suggested that the antibody response may be more effective in dampening the microbial counts and clinical inflammatory response, i.e., low BOP, seen in this group of patients compared to BGI-DL/MP and BGI-DL/SP. It also is possible that the increase in the (IgG/bacteria) ratio may be due to these subjects having more periodontal therapy. Although these data are cross-sectional in nature, in a clinical setting, these patients with pocketing but with minimal inflammation might be suggestive of a more stable clinical phenotype, i.e., less bleeding and, therefore, less likely to show future disease progression. Perhaps the most intriguing findings from these molecular epidemiologic analyses are the characteristics of the biologic phenotype that is associated with the most severe form of disease, the BGI-DL/SB condition that affects 12.9% of this population. Disease extent of PD ‡4 mm within this BGI-DL/SB group is not influenced by the level of supragingival plaque. BGI-DL/SB subjects have an excess innate acute inflammatory response characterized by higher PGE2 and especially IL-1b, which is even greater than that of BGI-G or BGI-DL/MB subjects. These BGI-DL/SB subjects have much higher P. gingivalis IgG responses (and subgingival P. gingivalis load) with more extensive disease compared to BGI-DL/MB and have differing qualitative inflammatory responses across a broader spectrum of inflammatory mediators (Table 7). In addition to increased levels of GCF IL-1b and PGE2, the BGI-DL/SB condition is associated with an increased expression of several key mediators of chronic inflammation that are not elevated among other subjects with BGI-deep lesions, indicating that the nature of the inflammatory response is qualitatively different, not just quantitatively different, among these subjects. For example, increased levels of IL-6 are critical in that this molecule regulates the transition from acute to chronic inflammation and induces the synthesis of MCP-1.28 These individuals have much greater levels of RANTES and MCP-1 (C-C chemokines) that enhance monocytic recruitment and activation and promote dendritic cell formation.
Offenbacher, Barros, Singer, Moss, Williams, Beck
RANTES was elevated significantly among BGI-SP subjects; however, this increase was not a common event because only 1.7% of GCF samples examined had detectible levels of RANTES. Although further characterization seems warranted, these findings indicate that the underlying inflammatory characteristics of BGI-DL/SB are quite different, displaying a coordinated enhanced expression of IL-6 and MCP-1. The combination of IL-1b and -6 and PGE2 is capable of inducing bone and connective tissue destruction. This is consistent with the finding that these subjects have approximately five times the amount of interproximal bone loss compared to BGI-DL/MB patients, as calculated from the difference in extent of teeth with ICAL £3 mm plus PD ‡5 mm. The levels of these mediators, as determined by this survey method, should be interpreted with caution; the numbers of subjects (180) and GCF samples (720) represented a relatively small nested case-control sample of the population. Analyses of additional subjects should reveal additional patterns. Smoking also seems to have a much greater impact on disease levels among the BGI-DL/SB subjects, contributing to a 2.6-fold greater extent PD ‡4 mm levels compared to those with less periodontal disease (BGI-MP). It is not clear whether smoking further upregulates the local inflammatory response or whether the effects of smoking are worsened in the presence of an upregulated immunoregulatory state. However, limiting analyses to never smokers suggested that the same increased inflammatory response is present (data not shown). In any event, both effects are important because those who are exposed to smoking and excessive inflammation have greater disease expression than subjects with either exposure alone, i.e., interaction effects are significant. It is not known whether the excessive local production of inflammatory mediators among the BGI groups are attributable to differences in underlying genotype,29 epigenotype,30 or the presence of an unknown, perhaps non-cultivable organism that was not identified by the methods used in this study. However, establishing distinct clinical groups that provide unique biologic phenotypes is an important step toward establishing linkages with the underlying single nucleotide polymorphisms patterns of key inflammatory or transcriptional regulatory genes or the possible epigenetic modifications of chromatin structure that may arise from microbial or environmental factors.29,30 In other words, specific biofilm organisms or exposures may have the capacity to alter the ‘‘inflammatory set point’’ of the local tissues in certain individuals via epigenetic mechanisms.29,30 Alternatively, the broad activation of these innate mediators of chronic inflammation and bone resorption (IL-6 and MCP-1) may reflect the increased load, and 1923
Disease at the Biofilm–Gingival Interface
perhaps tissue invasion, of P. gingivalis. Further experiments are needed to address these issues. Despite the large size and richly comprehensive nature of this molecular epidemiologic dataset, certain limitations regarding the general applicability of these findings should be emphasized. The population is a sample of four United States communities of older adults with a mean age of 62.4 years. With advancing age, the clinical and biologic phenotype associated with periodontal conditions might change and the effects of possible treatments in many individuals might be more apparent. Thus, the applicability of these BGI definitions to other populations may not yield similar biologic phenotype characteristics. Furthermore, considering the age range of the population, it may not include the most aggressive forms of disease that might have resulted in tooth loss and edentulism. We included numbers of teeth in all relevant models to, in part, adjust for the effect of tooth loss due to disease. Although this cross-sectional study cannot establish temporality, the BGI-G condition is consistent with an inflammatory gingival lesion in shallow pockets or perhaps an early gingival lesion, even in the presence of extensive loss of attachment with recession. However, it is not known whether this shallow lesion develops into a deep lesion over time, remains stable, or migrates toward the root apex retaining shallow PDs. The gingival and alveolar dimensions that define the hard and soft tissue biomass, or the ‘‘periodontal biotype,’’ may be important in determining whether inflammation results in pocketing or recession. In fact, the time-dependent relationships or transitions within or between any of these disease categories are unknown during disease progression or in response to treatment. Finally, our goal of achieving homogeneity within each disease group, although clearly better than traditional definitions, is not perfect. For example, certain subjects within the BGI-DL/SB group clearly showed selective elevations in A. actinomycetemcomitans counts and titers, rather than P. gingivalis or P. gingivalis plus C. rectus. Patients within the BGIDL/MB group also demonstrated higher counts of spirochetes, and certain subjects also had significantly elevated IgG titers to these organisms. Thus, patient-level diagnostics and further refinement in our nosological identification schema are indicated. Nonetheless, we suggest that this approach for assessing disease groupings using clinical signs that are common to dental practices seems to enable the partial characterization of underlying biologic phenotypes. To assign a patient to one of the five BGI categories, one only needs to identify whether the patient has one or more PDs ‡4 mm (rounding down to the nearest millimeter) and compute the percentage of total sites that bleed upon probing, i.e., 1924
Volume 78 • Number 10
(number of sites with BOP/total number of sites in mouth) · 100, and then classify by referring to the shallow versus deep lesion pocketing and percentage BOP groupings described above. We do not know whether patients move from one BGI category to another category as disease progresses or in response to treatment. Although treatment responses cannot be predicted from these data, the fact that these BGI classifications and biologic phenotypes differ in response to microbial load and inflammatory response suggests the likelihood that treatment responses may vary among differing phenotypes. This is an important question for future studies. Furthermore, creating clinical definitions that reflect the biology of pathogenesis ultimately will enable us to better define the microbial and host genotypes associated with disease and personalize diagnostics and therapeutics. CONCLUSIONS We provided evidence for the creation of new disease definitions that enable us to identify four distinct biologic phenotypes of BGI disease. All four definitions of disease are influenced by the presence of plaque, race, gender, and diabetes. The BGI-G lesion is characterized by an increased GCF level of IL-1b and PGE2 and exposure to C. rectus. The BGI-DL/LB condition is associated with elevated GCF IL-1b and P. gingivalis IgG titers, but lower counts of most organisms. BGI-DL/MB and BGI-DL/SB are associated with an elevated innate inflammatory phenotype (increased IL-1b and PGE2) and exposure to C. rectus and P. gingivalis. In addition, BGI-DL/SB is associated with an excessive innate inflammatory response for the level of organism present that is characterized by even higher levels of IL-1b and an increased expression of mediators of chronic inflammation (IL-6 and MCP-1, as well as the less frequent elevation in RANTES). Thus, this study provides strong molecular epidemiologic evidence that periodontal conditions reflect an excessive inflammatory response relative to the level of the microbial burden presented by the oral biofilm, as measured by plaque scores and subgingival DNA checkerboard analyses, and that there is microbial specificity in disease presentation and pathogenesis. ACKNOWLEDGMENTS The authors thank Dr. Ken Kornman, president and chief scientific officer, Interleukin Genetics, Waltham, MA, for his helpful comments. This work was supported by National Institutes of Health/National Institute of Dental and Craniofacial Research grants DE-13079 and RR-00046. REFERENCES 1. Williams RC. Periodontal disease. N Engl J Med 1990; 322:373-382.
J Periodontol • October 2007
2. Armitage GC. Periodontal diagnoses and classification of periodontal diseases. Periodontol 2000 2004;34:9-21. 3. American Academy of Periodontology. Diagnosis of periodontal diseases (position paper). J Periodontol 2003;74:1237-1247 (erratum 2004;75:779). 4. Armitage GC. Classifying periodontal diseases – A longstanding dilemma. Periodontol 2000 2002;30:9-23. 5. Offenbacher S. Periodontal diseases: Pathogenesis. Ann Periodontol 1996;1:821-878. 6. Casanova JL, Abel L. The human model: A genetic dissection of immunity to infection in natural conditions. Nat Rev Immunol 2004;4:55-66. 7. Baelum V, Lopez R. Focus. Defining and classifying periodontitis: Need for a paradigm shift? Eur J Oral Sci 2003;111:2-6. 8. Beck JD. Risk revisited. Community Dent Oral Epidemiol 1998;26:220-225. 9. Offenbacher S. Commentary: Clinical implications of periodontal disease assessments using probing depth and bleeding on probing to measure the status of the periodontal-biofilm interface. J Int Acad Periodontol 2005;7(4 Suppl.):157-161. 10. Beck JD, Offenbacher S. Relationships among clinical measures of periodontal disease and their associations with systemic markers. Ann Periodontol 2002;7: 79-89. 11. Beck JD, Eke P, Heiss G, et al. Periodontal disease and coronary heart disease: A reappraisal of the exposure. Circulation 2005;112:19-24. 12. Offenbacher S, Barros SP, Champagne MEC, et al. Diagnosis of periodontal disease at the periodontal tissue interface: Biological correlates. Oral Biosci Med 2005;213:1-6. 13. Elter JR, Champagne CM, Offenbacher S, Beck JD. Relationship of periodontal disease and tooth loss to prevalence of coronary heart disease. J Periodontol 2004; 75:782-790. 14. Beck JD, Elter JR, Heiss G, Couper D, Mauriello SM, Offenbacher S. Relationship of periodontal disease to carotid artery intima-media wall thickness: The atherosclerosis risk in communities (ARIC) study. Arterioscler Thromb Vasc Biol 2001;21:1816-1822. 15. Elter JR, Offenbacher S, White RP, Beck JD. Third molars associated with periodontal pathology in older Americans. J Oral Maxillofac Surg 2005;63:179-184. ¨ e H, Silness J. Periodontal disease in pregnancy. I: 16. Lo Prevalence and severity. Acta Odontol Scand 1963;21: 532-551. ¨ e H. Periodontal disease in pregnancy. II: 17. Silness J, Lo Correlation between oral hygiene and periodontal condition. Acta Odontol Scand 1964;22:121-135. 18. Beck JD, Eke P, Lin D, et al. Associations between IgG antibody to oral organisms and carotid intima-medial thickness in community-dwelling adults. Atherosclerosis 2005;183:342-348.
Offenbacher, Barros, Singer, Moss, Williams, Beck
19. Carlos J, Wolfe M, Kingman A. The extent and severity index: A simple method for use in epidemiologic studies of periodontal disease. J Clin Periodontol 1986;13: 500-505. 20. Champagne CM, Buchanan W, Reddy MS, Preisser JS, Beck JD, Offenbacher S. Potential for gingival crevice fluid measures as predictors of risk for periodontal diseases. Periodontol 2000 2003;31:167-180. 21. Zhong Y, Slade GD, Beck JD, Offenbacher S. Gingival crevicular fluid interleukin-1b, prostaglandin E2 and periodontal status in a community population. J Clin Periodontol 2007;34:285-293. 22. Page RC, Eke PI. Case definitions for use in population-based surveillance of periodontitis. J Periodontol 2007;78:1387-1399. 23. Offenbacher S, Boggess KA, Murtha AP, et al. Progressive periodontal disease and risk of very preterm delivery. Obstet Gynecol 2006;107:29-36 (erratum 2006; 107:1171). 24. Tanner A, Maiden MF, Macuch PJ, Murray LL, Kent RL Jr. Microbiota of health, gingivitis, and initial periodontitis. J Clin Periodontol 1998;25:85-98. 25. Vandamme P, Falsen E, Rossau R, et al. Revision of Campylobacter, Helicobacter, and Wolinella taxonomy: Emendation of generic descriptions and proposal of Arcobacter gen. nov. Int J Syst Bacteriol 1991;41: 88-103. 26. Heasman L, Stacey F, Preshaw PM, McCracken GI, Hepburn S, Heasman PA. The effect of smoking on periodontal treatment response: A review of clinical evidence. J Clin Periodontol 2006;33:241-253. 27. Saito T, Shimazaki Y, Sakamoto M. Obesity and periodontitis. N Engl J Med 1998;13; 339:482-483. 28. Kaplanski G, Marin V, Montero-Julian F, Mantovani A, Farnarier C. IL-6: A regulator of the transition from neutrophil to monocyte recruitment during inflammation. Trends Immunol 2003;24:25-29. 29. Kornman K, Duff G, Reilly P. Re: A critical assessment of interleukin-1 (IL-1) genotyping when used in a genetic susceptibility test for severe chronic periodontitis. Greenstein G, Hart TC (2002;73:231-247). J Periodontol 2002;73:1553-1556. 30. Bobetsis YA, Barros SP, Lin DM, et al. Bacterial infection promotes DNA hypermethylation. J Dent Res 2007; 86:169-174. Correspondence: Dr. Steven Offenbacher, Center for Oral and Systemic Diseases, University of North Carolina at Chapel Hill School of Dentistry, CB #7455, Dental Research Center Rm. 222, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599-7455. Fax: 919/9667537; e-mail:
[email protected]. Submitted November 22, 2006; accepted for publication April 25, 2007.
1925