Abstract
Objectives
Clinical and radiographic measures are gold standards for diagnosing periodontitis but offer little information regarding the pathogenesis of the disease. We hypothesized that a comparison of gene expression signatures between healthy and diseased gingival tissues would provide novel insights in the pathobiology of periodontitis, and would inform the design of future studies.
Methods
Ninety systemically healthy non-smokers with moderate to advanced periodontitis (63 with chronic and 27 with aggressive periodontitis) each contributed with ≥2 “diseased” interproximal papillae [with bleeding on probing (BoP), pocket depth (PD) ≥4mm, and attachment loss (AL) ≥3mm)] and a “healthy” papilla, if available (no BoP, PD ≤4mm and AL ≤2mm). RNA was extracted, amplified, reverse-transcribed, labeled, and hybridized with AffymetrixU133Plus2.0 arrays. Differential expression was assayed in 247 individual tissue samples (183 from diseased and 64 from healthy sites) using a standard mixed-effects linear model approach, with patient effects considered random with a normal distribution, and gingival tissue status considered a two-level fixed effect. Gene ontology analysis summarized the expression patterns into biologically relevant categories.
Results
Transcriptome analysis revealed that a total of 12,744 probe sets were differentially expressed after adjusting for multiple comparisons (p<9.15×10-7). Of those, 5,295 were up-regulated and 7,449 down-regulated in disease when compared to health. Gene ontology analysis identified 61 differentially expressed groups (adjusted p<0.05) including apoptosis, antimicrobial humoral response, antigen presentation, regulation of metabolic processes, signal transduction, and angiogenesis.
Conclusions
Gingival tissue transcriptomes provide a valuable scientific tool for further hypothesis-driven studies of the pathobiology of periodontitis.
Keywords: Periodontitis, genomics, infection, gene expression, microarray
INTRODUCTION
A distinction between states of periodontal health and disease is feasible using a variety of diagnostic approaches. Among the common clinical variables, bleeding on probing (BoP) is considered to best reflect presence of an inflammatory infiltrate adjacent to the ulcerated epithelium of the periodontal pocket. 1, 2. Probing depth (PD) exceeding the typical depth of the healthy gingival crevice, in presence of BoP, is also used to signify periodontal pathology, while clinical attachment loss (AL) describes the cumulative exposure to destructive periodontitis. 3 Loss of periodontal tissue support can also be assessed radiographically. 4.
However, clinical and radiographic variables reflect poorly the underlying pathobiology of the various forms of periodontitis. The distinct histologic features of periodontal health and disease were first documented in a classic publication by Page and Schroeder. 5 These authors described the detailed morphologic characteristics of the gingival, sulcular, and pocket epithelium, the underlying connective tissue, and the types of resident and infiltrating blood cells in the initial, early, established and advanced periodontal lesion. In parallel, microbiologic approaches established common and distinct constituents of the periodontal microbiota in health and disease 6, 7 while biochemical approaches documented levels of cytokines, chemokines and other inflammatory mediators within the tissues and the gingival crevicular fluid. 8-10
A genomic tool that may add to the armamentarium of approaches to study the pathobiology of periodontitis is gene expression profiling, i.e., the systematic cataloging of messenger RNA sequences in a cell population, organ or tissue sample. In general, transcriptomes are a powerful means of generating comprehensive genome-level data sets on complex diseases and have provided enormous insights mostly in cancer research 11, 12, but also in other conditions such as muscular dystrophy 13, Alzheimer’s disease and dementia 14, 15, rheumatologic disorders 16, 17, and asthma. 18, 19
To our knowledge, a systematic transcriptome-based approach has not been applied so far in the study of periodontitis. Our group has initiated a series of studies to explore whether the currently recognized forms of periodontitis are characterized by distinct gene expression profiles in affected gingival tissues. 20 Our further goal is to explore the feasibility of a novel classification based on similarities in transcriptional profiles. The aim of this first report is to present a comprehensive description of the periodontal transcriptome in healthy and diseased gingival tissues.
MATERIAL AND METHODS
The study was approved by the Columbia University Institutional Review Board.
Subjects
Ninety subjects with moderate to severe periodontitis (63 with chronic and 27 with aggressive periodontitis) were recruited among those referred to the Columbia University College of Dental Medicine between November 2004 and April 2007. Eligible patients were (i) >13 yrs old; (ii) had ≥ 24 teeth; (iii) had no history of systematic periodontal therapy other than occasional prophylaxis, (iv) had received no systemic antibiotics or anti-inflammatory drugs for ≥ 6 months, (v) harbored ≥4 teeth with radiographic bone loss, (vi) did not have diabetes or any systemic condition that entails a diagnosis of “Periodontitis as a manifestation of systemic diseases” 21, (vii) were not pregnant, and (ix) were not current users of tobacco products or nicotine replacement medication. Signed informed consent was obtained prior to enrollment.
Clinical examination
All participants underwent a full-mouth examination of the periodontal tissues at six sites per tooth by a single, calibrated examiner. Variables recorded included presence/absence of visible dental plaque (PL), presence/absence of bleeding on probing (BoP), probing depth (PD), and attachment level (AL). Data were entered chair-side to a computer and stored at a central server.
Gingival tissue donor areas and tissue sample collection
Subsequently to clinical data entry, a specially developed software identified periodontally “diseased” and “healthy” tooth sites based on the clinical data. “Diseased” sites showed BoP, had interproximal PD>4mm, and concomitant AL≥3mm. “Healthy” sites showed no BoP, had PD≤4mm and AL≤2mm. Next, the software identified (i) maxillary “diseased” and “healthy” interdental papillae, based on the above criteria, and (ii) pairs of diseased interdental papillae with similar clinical presentation (PD and AL within 2mm of each other). A posterior maxillary sextant encompassing a pair of qualifying “diseased” interdental papillae was identified.
Periodontal surgery was performed at the identified sextant with no prior supra- or subgingival instrumentation. After local anesthesia, submarginal incisions were performed, mucoperiosteal flaps were reflected, and the portion of each interproximal gingival papilla that adhered to the root surface was carefully dissected. This section comprised the ulcerated epithelial lining of the interproximal periodontal pockets and the underlying connective tissue. After dissection, the gingival tissue specimens were thoroughly rinsed with sterile normal saline solution and transferred into Eppendorf tubes containing a liquid RNA stabilization reagent*. A minimum of 2 diseased papillae were harvested from each sextant and, whenever available, a healthy tissue specimen was obtained from an adjacent site. After collection of the specimens, pocket elimination/reduction periodontal surgery was completed according to standard procedures. All patients received additional periodontal therapy according to their individual needs.
RNA extraction, reverse transcription, in vitro cRNA synthesis
The tissue specimens were stored in a liquid RNA stabilization reagent* overnight at 4°C, snap-frozen and stored in liquid nitrogen. All further processing occurred simultaneously for gingival biopsies originating from the same donor. Specimens were homogenized in a liquid buffer†. After incubation with chloroform and centrifugation at 12,000g, RNA collected in the upper aqueous phase was precipitated by mixing with isopropyl-alcohol and additional centrifugation and washed in 75% ethanol. The extracted RNA was purified using a total RNA isolation kit‡, quantitated spectrophotometrically, and 7.5 micrograms of total RNA was reverse-transcribed using a one-cycle cDNA synthesis kit§. Synthesis of biotin-Labeled cRNA was performed using appropriate amplification reagents for in vitro transcription∥. The cRNA yield was determined spectrophotometrically at 260 nm. The cRNA was fragmented by incubation in fragmentation buffer at 94°C for 35min and stored at -80°C until hybridizations.
Gene Chip hybridizations
Human Genome arrays¶ were used including 54,675 probe sets to analyze more than 47,000 transcripts including 38,500 well-characterized human genes. Hybridizations, probe array scanning and gene expression analysis were performed at the Gene Chip Core Facility, Columbia University Genome Center. Each sample was hybridized once and each person contributed with 2 to 4 (median 3) samples.
Data analysis
Two statistical analyses packages were used throughout*,†. Expression data were normalized and summarized using the log scale robust multi-array analysis (RMA) 22 with default settings. Differential expression was assayed using a standard mixed-effects linear model approach, with patient effects considered random with a normal distribution, and gingival tissue status considered a two-level fixed effect (“healthy” vs. “diseased”). Statistical significance for each probe set was determined using both the Bonferroni criterion and q-value. 23 For each probe set, a fold-change was computed by dividing the raw expression values among “diseased” tissue samples by the raw expression values among “healthy” samples. Therefore, fold-change values represent relative RNA levels in “disease” vs. “health”.
Gene Ontology analysis was performed using ermineJ 24 with the Gene Score Resampling method. P-values were used as input to identify biologically-relevant groups of genes showing differential expression in health and disease. Gene symbols and descriptions were derived from the Gemma System (HG-U133_Plus_2_NoParents.an.zip) and downloaded from: http://www.bioinformatics.ubc.ca/microannots/.
Additional ontology analysis of all genes with a q-value of <0.05 was carried out using Pathway Express 25 in which the differentially expressed genes were mapped to the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways (http://www.genome.jp/kegg/).
Experimental details and results following the MIAME standards 26 are available at the Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo/) under accession number GSE 10334.
RESULTS
The mean age of the patients was 42 years (range 13-76; Table 1). Based on self-reported race/ethnicity, 37% of the patients were White, 21% Black, 32% of mixed race, and 76% Hispanic. According to the 1999 International Workshop for the Classification of Periodontal Diseases and Conditions criteria 27, 70% of the patients had chronic and 30% aggressive periodontitis. On average, study participants had 28 teeth present, 57 sites with PD≥5 mm, 54 sites with AL≥5 mm and 71% BoP. Among the 247 harvested gingival tissue samples (183 from diseased and 64 from healthy sites), 67% had PD≥5 mm and 62% AL≥5 mm (Table 2). No healthy gingival tissues samples were available from 26 subjects.
Table 1.
General characteristics of the study participants (n=90)
| Characteristic | (Mean±SD or %) | Range |
|---|---|---|
| Age (years) | 42±13 | 13 - 76 |
| Female | 50% | |
| Race | ||
| Black | 21% | |
| White | 37% | |
| Asian | 1% | |
| Mixed | 32% | |
| Other | 5% | |
| Declined to report | 4% | |
| Ethnicity | ||
| Hispanic | 76% | |
| Non-Hispanic | 23% | |
| Declined to report: | 1.1% | |
| Periodontal diagnosis | ||
| Chronic periodontitis | 70% | |
| Aggressive periodontitis | 30% | |
| Clinical Periodontal Variables | ||
| Number of teeth | 28±3 | 22 - 32 |
| Percent of sites with bleeding on probing (%) | 71±0.2 | 24 - 100 |
| Pocket depth (PD; mm) | 3.9±0.7 | 2.9 - 6.5 |
| Number of sites/subject with PD ≥ 5 mm | 57±25 | 12 - 156 |
| Clinical attachment level (AL; mm) | 4.1±0.9 | 2.7 - 6.5 |
| Number of sites/subject with AL≥ 5 mm | 54±30 | 10 - 150 |
Table 2.
Distribution of tissue samples according to pocket depth (PD) and clinical attachment levels (AL)
| % of tissue samples in specified PD range | % of tissue samples in specified AL range | ||
|---|---|---|---|
| 1-2 mm | 19% | 1-2 mm | 18% |
| 3-4 mm | 14% | 3-4 mm | 18% |
| 5 mm | 31% | 5 mm | 21% |
| ≥6 mm | 36% | ≥6 mm | 41% |
| Non-readable | 2% | ||
Transcriptome analysis revealed that 32,598 probes sets were differentially expressed between healthy and diseased tissue samples at q<0.05. Of those, 51% were up-regulated and 49% down-regulated in disease when compared to health. Applying the Bonferroni correction for 54,675 comparisons, a total of 12,744 probe sets were differentially regulated (p<9.15×10-7; 5,295 up-regulated and 7,449 down-regulated in disease when compared to health). The complete list of differentially regulated probe sets can be viewed on Online (Supplemental Table 1).
Fold-changes in expression ranged between 5.73 and 3.89 (all p-values<1.1×10-16) for the top 50 probe sets with increased expression in diseased relative to healthy tissue samples (Fig. 1A), and between 4.35 and 2.13 (all p-values<1.1×10-16) for the top 50 probe sets with decreased expression in diseased samples [inverse values of the strongest (0.23) and weakest (0.47) fold change values quoted; Fig. 1B].
Figure 1.
Visualization of the top 50 probe sets with increased expression in diseased, relative to healthy gingival tissue (A) and of the top 50 probe sets with decreased expression in diseased, relative to healthy gingival tissue (B). Gingival tissue samples are grouped according to clinical periodontal status with diseased tissues on the left (red horizontal bar) and healthy tissues on the right (green bar). The color of each pixel represents gene expression level with darker colors indicating lower relative expression values. Columns correspond to individual tissue samples and rows correspond to probe sets. Fold change (FC) describes the ratio of mean expression in diseased tissue over the mean expression in healthy tissue. Note that multiple probe sets map to a single gene. Due to space limitations, only one gene symbol and gene name per probe set are identified. A complete list of gene symbols and names per probe set is provided in the Online Supplement Table 1.
Gene ontology analysis identified 61 differentially expressed groups at p<0.05 including apoptosis, antimicrobial humoral response, antigen presentation, regulation of metabolic processes, signal transduction, and angiogenesis (Table 3). Four selected differentially regulated pathways by Pathway Express analysis are illustrated in Fig. 2 (MAPK signaling pathway, Fig. 2A; cytokine-cytokine receptor interaction, Fig. 2B; cell adhesion molecules, Fig. 2C; and apoptosis, Fig. 2D). The top 50 pathways identified by this analysis are listed in Table 4.
Table 3.
Gene Ontology groups differentially expressed in diseased and healthy gingival tissues at p<0.05
| Group Name | ID | p-value | # of probes | # of genes |
|---|---|---|---|---|
| Induction of apoptosis | GO:0006917 | 1.26E-09 | 360 | 152 |
| Negative regulation of cell proliferation | GO:0008285 | 1.76E-05 | 373 | 170 |
| Protein metabolic process | GO:0019538 | 5.19E-05 | 441 | 180 |
| Negative regulation of apoptosis | GO:0043066 | 1.00E-04 | 368 | 169 |
| Regulation of cellular process | GO:0050794 | 1.36E-04 | 360 | 163 |
| Antimicrobial humoral response (sensu Vertebrata) | GO:0019735 | 2.50E-04 | 145 | 87 |
| Antimicrobial humoral response | GO:0019730 | 2.53E-04 | 142 | 85 |
| Regulation of Ras protein signal transduction | GO:0046578 | 3.10E-04 | 275 | 99 |
| Cell motility | GO:0006928 | 3.57E-04 | 382 | 180 |
| Antigen processing and presentation of peptide antigen via MHC class I | GO:0002474 | 4.04E-04 | 163 | 52 |
| Taxis | GO:0042330 | 5.74E-04 | 189 | 110 |
| Lipid biosynthetic process | GO:0008610 | 6.29E-04 | 211 | 98 |
| Positive regulation of apoptosis | GO:0043065 | 9.02E-04 | 321 | 138 |
| Chemotaxis | GO:0006935 | 9.72E-04 | 194 | 114 |
| Rho protein signal transduction | GO:0007266 | 9.92E-04 | 260 | 96 |
| Protein kinase cascade | GO:0007243 | 1.11E-03 | 305 | 117 |
| Enzyme linked receptor protein signaling pathway | GO:0007167 | 1.16E-03 | 250 | 90 |
| Protein complex assembly | GO:0006461 | 1.18E-03 | 372 | 146 |
| Regulation of growth | GO:0040008 | 1.53E-03 | 217 | 98 |
| Rrna metabolic process | GO:0016072 | 2.16E-03 | 98 | 53 |
| Induction of programmed cell death | GO:0012502 | 2.21E-03 | 263 | 114 |
| Lymphocyte activation | GO:0046649 | 2.41E-03 | 103 | 52 |
| Regulation of apoptosis | GO:0042981 | 2.63E-03 | 342 | 143 |
| Anti-apoptosis | GO:0006916 | 2.75E-03 | 326 | 145 |
| Localization of cell | GO:0051674 | 3.28E-03 | 267 | 130 |
| MAPKKK cascade | GO:0000165 | 4.21E-03 | 170 | 72 |
| Transmembrane receptor protein tyrosine kinase signaling pathway | GO:0007169 | 4.37E-03 | 338 | 127 |
| Blood vessel morphogenesis | GO:0048514 | 4.57E-03 | 147 | 59 |
| Cellular defense response | GO:0006968 | 4.70E-03 | 184 | 95 |
| Angiogenesis | GO:0001525 | 5.68E-03 | 147 | 62 |
| Antigen processing and presentation of peptide antigen | GO:0048002 | 5.72E-03 | 153 | 49 |
| Tissue development | GO:0009888 | 5.74E-03 | 214 | 121 |
| Dephosphorylation | GO:0016311 | 5.79E-03 | 377 | 164 |
| Endocytosis | GO:0006897 | 6.32E-03 | 339 | 144 |
| Response to DNA damage stimulus | GO:0006974 | 7.99E-03 | 388 | 182 |
| Anatomical structure formation | GO:0048646 | 8.97E-03 | 139 | 56 |
| Macromolecule complex assembly | GO:0065003 | 9.55E-03 | 333 | 127 |
| Regulation of programmed cell death | GO:0043067 | 0.014246 | 239 | 99 |
| Actin filament-based process | GO:0030029 | 0.014402 | 272 | 99 |
| Cell growth | GO:0016049 | 0.014682 | 281 | 124 |
| Actin cytoskeleton organization and biogenesis | GO:0030036 | 0.01542 | 345 | 125 |
| Negative regulation of signal transduction | GO:0009968 | 0.0155 | 168 | 65 |
| Ectoderm development | GO:0007398 | 0.017542 | 137 | 78 |
| RNA metabolic process | GO:0016070 | 0.017803 | 215 | 93 |
| Ribosome biogenesis and assembly | GO:0042254 | 0.021545 | 110 | 59 |
| Positive regulation of transcription from RNA polymerase II promoter | GO:0045944 | 0.021767 | 139 | 45 |
| Phospholipid metabolic process | GO:0006644 | 0.02178 | 122 | 64 |
| Positive regulation of transcription, DNA-dependent | GO:0045893 | 0.021971 | 309 | 113 |
| R-rna processing | GO:0006364 | 0.021976 | 106 | 57 |
| Cytoskeleton organization and biogenesis | GO:0007010 | 0.028137 | 258 | 96 |
| Epidermis development | GO:0008544 | 0.030145 | 119 | 71 |
| Cytokine and chemokine mediated signaling pathway | GO:0019221 | 0.034265 | 45 | 25 |
| Regulation of cell growth | GO:0001558 | 0.03479 | 282 | 130 |
| Cell migration | GO:0016477 | 0.036679 | 227 | 92 |
| Carboxylic acid transport | GO:0046942 | 0.036845 | 76 | 37 |
| Protein amino acid dephosphorylation | GO:0006470 | 0.037447 | 343 | 146 |
| DNA replication | GO:0006260 | 0.037785 | 257 | 125 |
| Cellular lipid metabolic process | GO:0044255 | 0.038116 | 313 | 144 |
| Membrane invagination | GO:0010324 | 0.048082 | 206 | 85 |
Number of probe sets and number of genes refer to the number of probe sets and genes represented in each ontology group. Analysis was carried out using on ermineJ 23.
Figure 2.
Ontology analysis of selected pathways. (A) MAPK signaling pathway; (B) cytokine-cytokine receptor interaction; (C) cell adhesion molecules; (D) apoptosis. Genes shown in red are over-expressed and genes shown in blue under-expressed in diseased gingival tissues when compared to healthy tissues. Genes in green are unchanged at the p<0.05 significance level.
Table 4.
Ontology analysis of the top 50 differentially expressed pathways in diseased and healthy gingival tissues
| Impacted pathway a | Impact Factor b | Input genes / Pathway genes (%) c | p-value |
|---|---|---|---|
| Antigen processing and presentation | 42.2 | 39.0 | 0.102874871 |
| Cell adhesion molecules (CAMs) | 40.5 | 53.0 | 6.38E-05 |
| B cell receptor signaling pathway | 16.1 | 68.3 | 8.03E-07 |
| Adherens junction | 14.0 | 63.6 | 3.30E-06 |
| Leukocyte transendothelial migration | 13.1 | 56.0 | 2.02E-05 |
| Natural killer cell mediated cytotoxicity | 11.9 | 51.9 | 5.12E-05 |
| Circadian rhythm | 11.8 | 41.7 | 0.402801166 |
| Focal adhesion | 10.8 | 50.8 | 9.47E-05 |
| Renal cell carcinoma | 10.6 | 60.9 | 7.58E-05 |
| Regulation of actin cytoskeleton | 10.4 | 49.0 | 1.61E-04 |
| Phosphatidylinositol signaling system | 10.0 | 35.1 | 0.689660919 |
| Colorectal cancer | 8.8 | 55.3 | 4.88E-04 |
| T cell receptor signaling pathway | 8.3 | 52.7 | 0.001630673 |
| MAPK signaling pathway | 8.1 | 47.3 | 9.90E-04 |
| Tight junction | 6.9 | 49.6 | 0.004079315 |
| VEGF signaling pathway | 6.5 | 52.9 | 0.004965142 |
| GnRH signaling pathway | 6.3 | 50.5 | 0.006525075 |
| Fc epsilon RI signaling pathway | 6.3 | 52.0 | 0.00785814 |
| Cytokine-cytokine receptor interaction | 6.2 | 43.4 | 0.008217367 |
| Small cell lung cancer | 6.1 | 51.2 | 0.007222158 |
| Wnt signaling pathway | 5.7 | 46.3 | 0.010884856 |
| Chronic myeloid leukemia | 5.7 | 51.3 | 0.010359917 |
| Notch signaling pathway | 5.7 | 53.2 | 0.010958518 |
| Glioma | 5.7 | 50.0 | 0.029273605 |
| Alzheimer”s disease | 5.6 | 63.6 | 0.011980061 |
| Epithelial cell signaling in H. pylori infection | 5.5 | 50.7 | 0.019515233 |
| Insulin signaling pathway | 5.4 | 47.4 | 0.012739089 |
| Jak-STAT signaling pathway | 5.3 | 45.1 | 0.021944981 |
| Pancreatic cancer | 5.1 | 49.3 | 0.027534648 |
| Prostate cancer | 4.8 | 48.8 | 0.022130731 |
| ErbB signaling pathway | 4.8 | 48.3 | 0.027500726 |
| Toll-like receptor signaling pathway | 4.7 | 46.7 | 0.041239323 |
| Long-term depression | 4.7 | 47.4 | 0.05266026 |
| ECM-receptor interaction | 4.2 | 46.0 | 0.057512646 |
| Dorso-ventral axis formation | 4.1 | 53.6 | 0.044417147 |
| Melanogenesis | 4.1 | 45.1 | 0.053503025 |
| Endometrial cancer | 4.1 | 50.0 | 0.046475053 |
| TGF-beta signaling pathway | 3.8 | 44.0 | 0.085106723 |
| Axon guidance | 3.7 | 43.8 | 0.09158817 |
| Ubiquitin mediated proteolysis | 3.6 | 48.9 | 0.081701274 |
| Type I diabetes mellitus | 3.4 | 43.2 | 0.131289041 |
| Adipocytokine signaling pathway | 3.3 | 44.4 | 0.121801571 |
| Apoptosis | 3.3 | 42.9 | 0.190763082 |
| Huntington”s disease | 3.2 | 50.0 | 0.114527344 |
| Neurodegenerative Disorders | 3.1 | 50.0 | 0.114527344 |
| Melanoma | 2.9 | 43.7 | 0.177477033 |
| Calcium signaling pathway | 2.9 | 40.6 | 0.198877948 |
| Gap junction | 2.9 | 42.4 | 0.178991356 |
| Thyroid cancer | 2.9 | 41.9 | 0.374566966 |
| Long-term potentiation | 2.8 | 43.5 | 0.190074619 |
Analysis carried out by means of Pathway Express 24 using all genes with FDR <0.05, mapped to Kyoto encyclopedia of Genes and Genomes (KEGG) pathways (http:www.genome.jp/kegg/) and ranked according to Impact Factor.
The impact factor identifies the relatively most affected pathways by considering and integrating the proportion of differentially regulated genes, the perturbation factors of all pathway genes, as well as the consistency of the propagation of these perturbations throughout the pathway.
Ratio of the number of regulated genes in the pathway over the total number of genes currently mapped to the pathway.
DISCUSSION
To the best of our knowledge, this is the first study to systematically describe the transcriptomes of healthy and diseased gingival tissues in patients with destructive periodontal diseases. The primary aim of this report is to provide a comprehensive description that will serve as an information resource for investigators interested in the pathobiology of periodontitis. Clearly, the presented gene expression data need to be subjected to additional verification steps before their exact biological significance is fully appreciated. These may include confirmation by independent techniques on the mRNA level such as real time RT-PCR, and by proteomic analyses. Therefore, at this point, the presented data are not meant to provide unequivocal evidence for the involvement of any particular gene in the disease process, but rather to identify broad consortia of genes and pathways that are likely differentially expressed in states of gingival health and disease.
Our study has several strengths relevant to its ambition to serve as a high quality research resource. First, we have involved a relatively large sample of well-characterized, patients with periodontitis that were free of confounding exposures such as systemic disease, medications and smoking. Second, our gene expression data are generated by a large number of arrays representing strictly defined clinical conditions and multiple sites per subject. Third, our gingival tissue samples were obtained prior to any therapeutic manipulation of the gingival tissues. Lastly, by allowing direct access to our raw data, we enable independent investigators to conduct focused analyses targeting individual genes and pathways of particular interest to them.
We would like to draw attention to a number of points that will facilitate a correct interpretation of our findings: it must be realized that the reported transcriptomes represent the composite gene expression of a variety of cells that constitute and populate the healthy and diseased gingival tissues, including epithelial cells, connective tissue fibroblasts and infiltrating cells. Although the assayed tissue samples were deemed to be “diseased” or “healthy” based on accepted clinical signs of gingival inflammation the extent of the inflammatory infiltrate, the degree of vascularization and the epithelial/connective tissue content of each gingival tissue sample were unknown and likely variable. In future studies, use of cell-capture techniques may facilitate the study of homogeneous cell subpopulations, and may generate data that can be directly comparable to those stemming from well-defined in vitro systems, such as the recently reported transcriptional profiles of cultured oral epithelial cells challenged by specific periodontal pathogens and commensals 28-30, or the in vivo regulation of specific proteins in rodent junctional and pocket epithelia. 31 With respect to the clinical status of the obtained gingival tissue samples, it must be noted that the transcriptomes of healthy and intact gingival tissues of periodontitis patients may not necessarily be identical to those of healthy sites in subjects that have not experienced destructive periodontitis. Consequently, since our data are based exclusively on a cohort of patients with periodontitis, our findings cannot identify “susceptibility genes”. Furthermore, the observed heterogeneity in expression among diseased tissue samples even for genes that were, on average, undisputedly differentially regulated between health and disease may reflect varying states of disease activity among clinically homogeneous sites. There are additional potential explanations for this heterogeneity, such as differential bacterial colonization patterns across diseased sites. Future analyses from our group will incorporate data on bacterial colonization patterns and will be informative in this regard. Lastly, while a potential effect of infiltration anesthesia on gene expression is conceivable, there is little reason to expect differential anesthesia-mediated effects in diseased versus healthy samples, and thus a systematic bias in the reported comparisons.
In this first report, we will not proceed with an in-depth discussion of specific differentially regulated pathways in health and disease but will rather provide examples that underscore the utility of the expression data. At first glance, one can view the transcriptome findings as largely confirmatory of anticipated differences based on earlier histologic or biochemical analyses. For example, the vast majority of the top genes with increased expression in disease as compared to health are indeed immunoglobulin-related genes. However, genes far less readily associated with periodontitis were also observed to be least expressed in disease or, alternatively, most expressed in health (e.g. desmocollin 1, arylacetamide deacetylase-like 2, guanylate cyclase C). Likewise, the most expressed chemokine in disease (by 3.85-fold, p<10-18) was CXCL6 (granulocyte chemoattractant protein 2, GCP-2), a molecule known to be involved in inflammatory bowel diseases 32 but not earlier associated with gingival inflammation. Lastly, confirming and extending recent preliminary findings 33, our data showed that matrix metalloproteinases 7, 13, 3, 1, 9, 14, 2 and 28 and their inhibitors TIMP-3 and TIMP-2 are significantly up-regulated in diseased tissues. The above examples illustrate the utility of transcriptional data in guiding future focused studies of the pathobiology of periodontitis.
Supplementary Material
ACKNOWLEDGEMENTS
Supported by NIH grant DE015649 to Dr. Papapanou (National Institutes of Health, Bethesda, Maryland, USA). Dr. Pavlidis was supported by NIH grant GM076990 and a Michael Smith Foundation for Health Career Investigator Award (Vancouver, Canada); Dr. Handfield was supported by NIH grant DE16715; Dr. Kebschull was partly supported by a stipend from Neue Gruppe Wissenschaftsstiftung, Wangen/Allgäu, Germany.
Technical assistance was provided by Mr. Jun Yang, Division of Periodontics, Section of Oral and Diagnostic Sciences, Columbia University College of Dental Medicine.
Sources of support
Supported by NIH grant DE015649 to Dr. Papapanou
Footnotes
(RNAlater, Ambion, Austin, TX)
RNAlater, Ambion, Austin, TX, USA
Trizol; Invitrogen Life Technologies, Carlsbad, CA, USA
RNeasy; Qiagen, Valencia, CA, USA
GeneChip Expression 3′ amplification one-cycle cDNA synthesis kit; Affymetrix, Santa Clara, CA, USA
GeneChip Expression 3′-Amplification Reagents for IVT labeling kit; Affymetrix
Human Genome U-133 Plus 2.0 arrays; Affymetrix
R version 2.3.1 for Linux OS
SAS for PC version 9.1; SAS Institute, Cary, NC, USA
The authors declare that they have no conflicts of interest.
Summary of key findings
Gene expression signatures differentiate between healthy and diseased gingival tissues and may provide novel insights in the pathobiology of periodontitis
REFERENCES
- 1.Listgarten MA. Periodontal probing: what does it mean? J Clin Periodontol. 1980;7:165–176. doi: 10.1111/j.1600-051x.1980.tb01960.x. [DOI] [PubMed] [Google Scholar]
- 2.Harper DS, Robinson PJ. Correlation of histometric, microbial, and clinical indicators of periodontal disease status before and after root planing. J Clin Periodontol. 1987;14:190–196. doi: 10.1111/j.1600-051x.1987.tb00966.x. [DOI] [PubMed] [Google Scholar]
- 3.Hefti AF. Periodontal probing. Crit Rev Oral Biol Med. 1997;8:336–356. doi: 10.1177/10454411970080030601. [DOI] [PubMed] [Google Scholar]
- 4.Bragger U. Radiographic parameters: biological significance and clinical use. Periodontol 2000. 2005;39:73–90. doi: 10.1111/j.1600-0757.2005.00128.x. [DOI] [PubMed] [Google Scholar]
- 5.Page RC, Schroeder HE. Pathogenesis of inflammatory periodontal disease. A summary of current work. Lab Invest. 1976;34:235–249. [PubMed] [Google Scholar]
- 6.Listgarten MA, Helldén L. Relative distribution of bacteria at clinically healthy and periodontally diseased sites in humans. J Clin Periodontol. 1978;5:115–132. doi: 10.1111/j.1600-051x.1978.tb01913.x. [DOI] [PubMed] [Google Scholar]
- 7.Socransky SS, Haffajee AD, Dzink JL. Relationship of subgingival microbial complexes to clinical features at the sampled sites. J Clin Periodontol. 1988;15:440–444. doi: 10.1111/j.1600-051x.1988.tb01598.x. [DOI] [PubMed] [Google Scholar]
- 8.Bowers MR, Fisher LW, Termine JD, Somerman MJ. Connective tissue-associated proteins in crevicular fluid: potential markers for periodontal diseases. J Periodontol. 1989;60:448–451. doi: 10.1902/jop.1989.60.8.448. [DOI] [PubMed] [Google Scholar]
- 9.Mathur A, Michalowicz B, Castillo M, Aeppli D. Interleukin-1 alpha, interleukin-8 and interferon-alpha levels in gingival crevicular fluid. J Periodontal Res. 1996;31:489–495. doi: 10.1111/j.1600-0765.1996.tb01414.x. [DOI] [PubMed] [Google Scholar]
- 10.Dongari-Bagtzoglou AI, Ebersole JL. Increased presence of interleukin-6 (IL-6) and IL-8 secreting fibroblast subpopulations in adult periodontitis. J Periodontol. 1998;69:899–910. doi: 10.1902/jop.1998.69.8.899. [DOI] [PubMed] [Google Scholar]
- 11.Chung CH, Bernard PS, Perou CM. Molecular portraits and the family tree of cancer. Nat Genet. 2002;32(Suppl):533–540. doi: 10.1038/ng1038. [DOI] [PubMed] [Google Scholar]
- 12.Quackenbush J. Microarray analysis and tumor classification. N Engl J Med. 2006;354:2463–2472. doi: 10.1056/NEJMra042342. [DOI] [PubMed] [Google Scholar]
- 13.Haslett JN, Kunkel LM. Microarray analysis of normal and dystrophic skeletal muscle. Int J Dev Neurosci. 2002;20:359–365. doi: 10.1016/s0736-5748(02)00041-2. [DOI] [PubMed] [Google Scholar]
- 14.Colangelo V, Schurr J, Ball MJ, Pelaez RP, Bazan NG, Lukiw WJ. Gene expression profiling of 12633 genes in Alzheimer hippocampal CA1: transcription and neurotrophic factor down-regulation and up-regulation of apoptotic and pro-inflammatory signaling. J Neurosci Res. 2002;70:462–473. doi: 10.1002/jnr.10351. [DOI] [PubMed] [Google Scholar]
- 15.Haroutunian V, Katsel P, Schmeidler J. Transcriptional vulnerability of brain regions in Alzheimer’s disease and dementia. Neurobiol Aging. 2007 doi: 10.1016/j.neurobiolaging.2007.07.021. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Thornton S, Sowders D, Aronow B, et al. DNA microarray analysis reveals novel gene expression profiles in collagen-induced arthritis. Clin Immunol. 2002;105:155–168. doi: 10.1006/clim.2002.5227. [DOI] [PubMed] [Google Scholar]
- 17.van der Pouw Kraan TC, van Baarsen LG, Rustenburg F, Baltus B, Fero M, Verweij CL. Gene expression profiling in rheumatology. Methods Mol Med. 2007;136:305–327. doi: 10.1007/978-1-59745-402-5_22. [DOI] [PubMed] [Google Scholar]
- 18.Burke W. Genomics as a probe for disease biology. N Engl J Med. 2003;349:969–974. doi: 10.1056/NEJMra012479. [DOI] [PubMed] [Google Scholar]
- 19.Izuhara K, Saito H. Microarray-based identification of novel biomarkers in asthma. Allergol Int. 2006;55:361–367. doi: 10.2332/allergolint.55.361. [DOI] [PubMed] [Google Scholar]
- 20.Papapanou PN, Abron A, Verbitsky M, et al. Gene expression signatures in chronic and aggressive periodontitis: a pilot study. Eur J Oral Sci. 2004;112:216–223. doi: 10.1111/j.1600-0722.2004.00124.x. [DOI] [PubMed] [Google Scholar]
- 21.Lindhe J, Ranney RR, Lamster IB, et al. Consensus report: Periodontitis as a manifestation of systemic diseases. Ann Periodontol. 1999;4:64. [Google Scholar]
- 22.Irizarry RA, Bolstad BM, Collin F, Cope LM, Hobbs B, Speed TP. Summaries of Affymetrix GeneChip probe level data. Nucleic Acids Res. 2003;31:e15. doi: 10.1093/nar/gng015. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Storey JD, Tibshirani R. Statistical significance for genomewide studies. Proc Natl Acad Sci U S A. 2003;100:9440–9445. doi: 10.1073/pnas.1530509100. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Lee HK, Braynen W, Keshav K, Pavlidis P. ErmineJ: tool for functional analysis of gene expression data sets. BMC Bioinformatics. 2005;6:269. doi: 10.1186/1471-2105-6-269. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Khatri P, Voichita C, Kattan K, et al. Onto-Tools: new additions and improvements in 2006. Nucleic Acids Res. 2007;35:W206–211. doi: 10.1093/nar/gkm327. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Brazma A, Hingamp P, Quackenbush J, et al. Minimum information about a microarray experiment (MIAME)-toward standards for microarray data. Nat Genet. 2001;29:365–371. doi: 10.1038/ng1201-365. [DOI] [PubMed] [Google Scholar]
- 27.Caton JG, Armitage GC, editors. 1999 International Workshop for a Classification of Periodontal Diseases and Conditions. Ann Periodontol; Oak Brook, Illinois: 1999. pp. 1–112. [DOI] [PubMed] [Google Scholar]
- 28.Handfield M, Mans JJ, Zheng G, et al. Distinct transcriptional profiles characterize oral epithelium-microbiota interactions. Cell Microbiol. 2005;7:811–823. doi: 10.1111/j.1462-5822.2005.00513.x. [DOI] [PubMed] [Google Scholar]
- 29.Hasegawa Y, Mans JJ, Mao S, et al. Gingival epithelial cell transcriptional responses to commensal and opportunistic oral microbial species. Infect Immun. 2007;75:2540–2547. doi: 10.1128/IAI.01957-06. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Handfield M, Baker HV, Lamont RJ. Beyond good and evil in the oral cavity: insights into host-microbe relationships derived from transcriptional profiling of gingival cells. J Dent Res. 2008;87:203–223. doi: 10.1177/154405910808700302. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Ekuni D, Firth JD, Putnins EE. Regulation of epithelial cell growth factor receptor protein and gene expression using a rat periodontitis model. J Periodontal Res. 2006;41:340–349. doi: 10.1111/j.1600-0765.2006.00881.x. [DOI] [PubMed] [Google Scholar]
- 32.Gijsbers K, Van Assche G, Joossens S, et al. CXCR1-binding chemokines in inflammatory bowel diseases: down-regulated IL-8/CXCL8 production by leukocytes in Crohn’s disease and selective GCP-2/CXCL6 expression in inflamed intestinal tissue. Eur J Immunol. 2004;34:1992–2000. doi: 10.1002/eji.200324807. [DOI] [PubMed] [Google Scholar]
- 33.Kubota T, Itagaki M, Hoshino C, et al. Altered gene expression levels of matrix metalloproteinases and their inhibitors in periodontitis-affected gingival tissue. J Periodontol. 2008;79:166–173. doi: 10.1902/jop.2008.070159. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.