To the best of our knowledge, this is the first study to examine the association between subgingival bacterial colonization patterns and gingival tissue gene expression in human periodontitis. Our data demonstrate that the variable bacterial content of the periodontal pocket correlates with distinct gene expression signatures in the adjacent gingival tissues. Importantly, even though we examined colonization patterns by only a limited number of bacterial species, we found that the variable subgingival bacterial load by several -but clearly not all- species correlated significantly with tissue gene expression. In other words, and to paraphrase both Anton van Leeuwenhoek and George Orwell, our data indicate that all subgingival "animalcules" are not "equal" in this respect.
In a recent publication [10
], we presented transcriptomic data from a subset of patients involved in the present report (90 patients and 247 arrays out of the total of 120 patients and 310 arrays included here) and compared gene expression profiles of clinically healthy and diseased gingival tissues in patients with periodontitis. We documented substantial differential gene expression between states of gingival health and disease that was reflected both by genes that were a priori
anticipated to be variably expressed based on current knowledge (e.g., several inflammatory, immune function- and apoptosis-related genes), but also by genes that are not readily associated with gingival inflammation (e.g., the transcription factor POU2AF1, the sperm associated antigen 4 which appears to be associated with apoptosis (own unpublished data), the cell adhesion-mediating protein desmocollin 1, and the signaling lymphocytic activation molecule family member 7). In the present study, we sought to investigate whether the bacterial content of the periodontal pocket is also a determinant of gene expression in the adjacent gingival tissues in order to enhance our understanding of the host-bacterial interactions that take place in the interface between the plaque biofilm and the periodontal pocket. We realize that the above question can ideally be addressed in a longitudinal prospective rather than a cross-sectional study. Thus, although our analyses considered bacterial colonization as the independent exposure and tissue gene expression as the outcome, it is impossible to rule out reverse causation, i.e., that the qualitative characteristics of the gingival tissue are the determinants of bacterial colonization. However, given that periodontitis is a bacterially-induced infection, the former approach is reasonable in the discussion of the observed correlations between colonization patterns and tissue gene expression signatures. We also want to draw the reader's attention to the fact that, despite our inferences on each particular bacterial species' effect on the gingival tissue transcriptome, we have not studied individual mono-infections. Therefore, any properties ascribed to a particular species with respect to its ability to regulate genes in the gingival tissues cannot be entirely segregated from concomitant synergistic or antagonistic effects of other covarying bacteria among the ones studied or, most importantly, of the several hundreds of cultivable and uncultivable species that are known to colonize the periodontal pocket and were not investigated in this work [36
]. Instead, the differential gene expression in the gingival tissues should more appropriately be attributed to the aggregate effect of the mixed microbial burden, and the specific investigated bacteria may simply serve as a surrogate for this mixed microbial burden to which they contribute. It must be further recognized that the gingival tissue transcriptomes are also influenced by a plethora of additional factors beyond those of bacterial origin, including biologically active host-derived molecules and tissue degradation byproducts, that could not be accounted for in our study.
In view of the above, and because the transcriptomic profiles analyzed originate from a mixed cell population comprising gingival epithelial cells, connective tissue fibroblasts and infiltrating cells, our data are not directly comparable with observations from the aforementioned in vitro
studies of mono-infections of oral epithelial cell lines. Nevertheless, our data corroborate and extent data from these experimental settings. For example, ontology analysis of epithelial cell pathways differentially regulated after infection with F. nucleatum
] identified MAPK signaling and regulation of actin cytoskeleton among the impacted pathways. Likewise, in line with observations by Handfield et al. [11
], apoptotic mitochondrial changes, the second highest differentially regulated ontology group according to levels of A. actinomycetemcomitans
was ranked 96th
according to subgingival levels of P. gingivalis
. Indeed, A. actinomycetemcomitans
is known to exert strong pro-apoptotic effects on various cell types encountered in inflamed gingival tissues, such as gingival epithelial cells [37
] or invading mononuclear cells [38
], attributed in part to its potent cytolethal distending toxin [39
]. On the other hand, P. gingivalis
was shown to inhibit apoptosis in primary gingival epithelial cells by ATP scavenging through its ATP-consuming nucleoside diphosphate kinase [40
]. In contrast, other in vitro
studies involving oral epithelial cells (for review see [41
]) reported apoptotic cell death induced by P. gingivalis
at very high (up to 1:50,000) multiplicities of infection [42
], which arguably exceeds the in vivo
burden in the periodontal pocket.
Thus, our data indicate presence of pro-apoptotic alterations in the gingival tissues in A. actinomycetemcomitans
-associated periodontitis, while the effects of P. gingivalis
appear to be primarily mediated by other pathways. Interestingly, our data corroborate a recent study that explored the hyper-responsiveness of peripheral blood neutrophils in periodontitis and demonstrated a significantly increased expression of several interferon-stimulated genes [43
]. As shown in Table , interferon alpha inducible protein-6 was among the top commonly up-regulated genes in gingival tissue lesions according to levels of A. actinomycetemcomitans
, P. gingivalis
and C. rectus
, and tissue-infiltrating neutrophils are a conceivable source for these transcripts.
In general, the magnitude of the differential expression of host tissue genes according to levels of A. actinomycetemcomitams
(with a total of 68 genes exceeding an absolute fold change of 2 when comparing tissue samples in the upper and lowest quintiles of subgingival colonization; Additional File 1
) was more limited than that of bacteria in the 'red complex' (488 genes for P. gingivalis
, 521 genes for T. forsythia
, 429 genes for T. denticola
; Additional Files 2
) or C. rectus
(450 genes; Additional File 8
The null hypothesis underlying the present study, i.e., that variable subgingival bacterial load by specific bacteria results in no differential gene expression in the adjacent pocket tissues, was rejected by our data. Indeed levels of only 2 of the 11 species investigated appeared to correlate poorly with differential gene expression in the tissues: A. naeslundii, whose levels were statistically associated with differential expression of only 8 probe sets out of the approximately 55,000 analyzed, and E. corrodens with <1% of the probe sets being differentially regulated between pockets with the highest versus the lowest levels of colonization. In contrast, 15-17% of the examined probes sets were differentially expressed according to subgingival levels of the "red complex" species and C. rectus, whose levels were the most strongly correlated with gingival tissue gene expression signatures among all investigated species.
Importantly, the above associations between bacterial colonization and gingival tissue gene expression signatures were confirmed in analyses adjusting for clinical periodontal status, although they were expectedly attenuated. In other words, the difference in the tissue transcriptomes between periodontal pockets with high versus low levels of colonization by the particular species identified as strong regulators of gene expression cannot solely be ascribed to differences in the clinical status of the sampled tissues [10
] which is known to correlate well with bacterial colonization patterns [31
]. Instead, our analyses based on either statistical adjustment or restriction to 'diseased' tissue samples consistently demonstrate that, even among periodontal pockets with similar clinical characteristics, the subgingival colonization patterns still influence the transcriptome of the adjacent gingival tissues. Assuming a generally positive correlation between gene and protein expression [44
], this finding is conceptually important as it suggests that the 'phenotype' of the periodontal pocket, and by extension its potential to experience additional periodontal tissue breakdown and/or an unfavorable or favorable treatment response, is also dependent on its bacterial content, and not merely on the traditional clinical parameters (non-specific plaque accumulation, bleeding on probing, probing depth and attachment level). It also provides biology-founded ammunition in favor of the controversial argument that microbial diagnostics have a place in the decision-making and therapeutic management of patients with periodontitis [46
Finally, we emphasize that the subject sample involved in the present study included both chronic and aggressive periodontitis patients and subjectsbelonging to various race/ethnicity groups. It is conceivable that the typeof disease and race/ethnicity-related charactersitics may be additional determinants of the gingival tissue transcriptome and/or may act asmodifiers of the association between bacterial colonization patterns andtissue gene expression. We intend to explore these possibilities insubsequent reports.