Oral Microbiome and Gingival Transcriptome Profiles of Ligature-Induced Periodontitis

Abstract

This investigation evaluated the relationship of the oral microbiome and gingival transcriptome in health and periodontitis in nonhuman primates (Macaca mulatta). Subgingival plaque samples and gingival biopsies were collected from healthy sites and at sites undergoing ligature-induced periodontitis. Microbial samples were analyzed with 16S amplicon sequencing to identify bacterial profiles in young (3 to 7 y) and adult (12 to 23 y) animals. The gingival transcriptome was determined with a microarray analysis and focused on the expression level of 452 genes that are associated with the development of inflammation and innate and adaptive immune responses. Of the 396 total operational taxonomic units (OTUs) identified across the samples, 81.8% were detected in the young group and 99.5% in the adult group. Nevertheless, 58 of the OTUs composed 88% of the signal in adults, and 49 OTUs covered 91% of the OTU readouts in the young group. Correlation analyses between the microbiome members and specific gingival genes showed a high number of significant bacteria-gene correlations in the young healthy tissues, which decreased by 75% in diseased tissues. In contrast, these correlations increased by 2.5-fold in diseased versus healthy tissues of adult animals. Complexes of bacteria were delineated that related to specific sets of immune genes, differing in health and disease and in the young versus adult animals. The correlated gene profiles demonstrated selected pathway overrepresentation related to particular bacterial complexes. These results provide novel insights into microbiome changes with disease and the relationship of these changes to specific gene profiles and likely biologic activities occurring in healthy and diseased gingival tissues in this human-like periodontitis model.

Introduction

Cross-sectional human studies have documented qualitative/quantitative differences in the subgingival microbial biofilm ecology of healthy and periodontitis sites. The oral infection paradigm has identified species that appear to emerge with disease and has described the importance of microbial complexes contributing to a biofilm-based “food and virulence” web driving disease (Colombo et al. 2009; Griffen et al. 2012; Lopez et al. 2015). The model also emphasizes a geospatial organization of the microbiome (Valm et al. 2011) and incorporates a current focus on a dysbiotic microbiome that contributes to/“causes” a dysregulation of the host response resulting in disease (Ebersole et al. 2017). However, the dynamics of the interactions of the components of the microbiome and the armamentarium of host responses to this challenge at mucosal sites remain rather nebulous.

While the microbiomes in health and disease have been described in humans (Wade 2013; Jorth et al. 2014; Feres et al. 2016; Kilian et al. 2016), limitations of studies of human disease do not allow definitive interpretation of how early in an active disease process critical microbiome alterations occur, as well as the characteristics of local host responses that reflect and/or drive dysbiotic changes and progressing disease (Meyle and Chapple 2015; Kilian et al. 2016). Thus, there remains a need to detail the dynamics of these biologic changes. This study used a well-described ligature-induced periodontitis nonhuman primate model with an oral microbiome more similar to humans (Oz and Puleo 2011; Ocon et al. 2013; Colombo et al. 2017) to describe the differences in the subgingival microbiome coupled with alterations in the adjacent tissue immune response–related transcriptome coincident with periodontal lesions. The investigation also addressed relationships in young versus adult individuals to document characteristics that associate with an environment that appears less susceptible to initiation of tissue destruction in periodontitis, previously demonstrated in younger humans and nonhuman primates (Bimstein and Ebersole 1989; Ebersole et al. 2014).

Methods

Nonhuman Primate Model and Oral Clinical Evaluation

Macaca mulatta (n = 36; 19 females and 17 males) housed at the Caribbean Primate Research Center were used in these studies (Gonzalez et al. 2011). The animals were distributed by age into a young/adolescent group (Y/ADO; n = 18; 3 to 7 y) and an adult/aged group (AD/AG; n = 18; 12 to 23 y). The nonhuman primates were fed a supplemented monkey diet of 20% protein, 5% fat, and 10% fiber (8773, Teklad NIB Primate Diet Modified; Harlan Teklad). A protocol conforming to the ARRIVE guidelines and approved by the Institutional Animal Care and Use Committee of the University of Puerto Rico enabled clinical measures of periodontal health/disease, including probing pocket depth and bleeding on probing from full mouth measures with 4 sites per tooth (Ebersole et al. 2008). A ligature-induced periodontitis model was implemented, as we described previously (Ebersole et al. 2014), by tying 3-0 silk sutures around the necks of maxillary and mandibular premolar and first and second molar teeth in 3 quadrants in each animal. The untreated quadrant was used to obtain baseline healthy tissue and microbiome samples from each animal.

Tissue Sampling and Gene Expression Microarray Analysis

Individual buccal gingival samples from healthy or periodontitis-affected premolar/molar maxillary region tissues of each animal were taken and maintained frozen in RNAlater for microarray analysis (GeneChip Rhesus Gene 1.0 ST Array; Affymetrix; Gonzalez et al. 2011). Normalization of values across the chips was accomplished through Affymetrix RMA and the MAS 5 algorithms (Ferrin et al. 2018). For each gene, differences in expression were assessed across the groups through analysis of variance (SAS 9.3; SAS Institute). The values for the tissues were then compared between the age groups or by health versus periodontitis via a t test (P < 0.05). Correlations of microbiome and gingival tissue gene expression levels were determined through Pearson correlation coefficient analysis (P < 0.05). The data have been uploaded to http://www.ncbi.nlm.nih.gov/geo/info/submission.html.

Analysis of Oral Microbiome

Bacterial samples were obtained by a curette and analyzed with a MiSeq instrument (Kozich et al. 2013; Kirakodu et al. 2019) for the total composition of the microbiome from each sample (Schloss et al. 2009; Edgar et al. 2011). Sequences were clustered into phylotypes based on their sequence similarity, and these binned phylotypes were assigned to their taxonomic classification with the Human Oral Microbiome Database (version 13; http://www.homd.org/index.php?name=seqDownload&file&type=R). Raw data were deposited at the National Institutes of Health’s National Center for Biotechnology Information (BioProject PRJNA516659). Statistical differences of bacterial operational taxonomic units (OTUs) were determined with a t test (P < 0.05). Correlations of OTUs within the oral microbiome were determined with Pearson correlation coefficient analysis (P < 0.05). Correlations between the microbiome components and the gingival gene expression were determined only for matching samples derived from the same tooth in each animal. Matching samples with sufficient microbiome signals were compared for 6 (Y/ADO) and 11 (AD/AG) healthy samples and 16 (Y/ADO) and 37 (AD/AG) samples obtained throughout the disease process.

Results

In our previous study, younger animals demonstrated bleeding on probing (i.e., inflammation) similar to adult animals postligation, although we noted a significantly lower magnitude of tissue destruction (probing pocket depths) in the younger animals (Gonzalez et al. 2018). This suggested that the relationship of the oral microbiome to local tissue responses could reflect a more disease-oriented oral microbiome, even in absence of clinical changes in the periodontium. Of the total 396 OTUs identified, 81.8% were detected in the young group and 99.5% in the adult group (Appendix Table 1). Nevertheless, 58 OTUs comprised 88% of the overall adult reads, and 49 OTUs covered 91% of the readouts in young samples. The same OTUs comprised a dominant portion of the microbiome in both age groups (Fig. 1). However, significant differences in levels of individual OTUs were seen between age groups, in health and disease, emphasizing substantive differences in the ecology in the primates under varied conditions. Figure 2 presents patterns of the bacteria with significantly correlated relative abundance levels in young and adult samples. Multiple OTU-based complexes within the healthy and diseased plaque samples were identified. Differences are noted in the members of the healthy complexes between young and adult samples, with only 2 complexes showing about 60% similarity of the individual bacteria in the complexes. Moreover, the complexes with disease were quite distinct between the age groups.

Figure 1.

Comparison of relative abundance of operational taxonomic units (OTUs) in the oral microbiomes of (A) healthy or (B) disease samples from young and adult animals. The bars denote means of 6 young and 11 adult healthy samples and 16 young and 37 adult samples obtained throughout the disease process. The asterisk (*) denotes that the OTU was statistically different at least at P < 0.05.

Figure 2.

Comparison of relative abundance of individual bacterial members within the bacterial complexes identified in healthy samples (I-III) and disease samples (I-II) in (A) the young/adolescent group (Y/ADO; 3 to 7 y) and (B, C) the adult/aged group (AD/AG; 12 to 23 y). The pie slice for each bacterium/taxon reflects the relative amount within the specific complex. The overall percentage is the sum of the relative abundance of all members of the complex within the overall microbiome. The arrow denotes the starting position of the pie slices related to the legend. Operational taxonomic unit identifiers in black are positively correlated with other members in the complex, and those in red are negatively correlated.

Figure 3 depicts the magnitude of significant correlations between the relative abundance of individual bacterial OTUs and a set of 452 immune response genes (Appendix Table 2) in samples from health and disease in the young and adult animals. In young healthy tissues, there was a significantly greater number of positive and negative correlations (28.1%) as compared with the other groups, with similar numbers of genes in each direction. Many of these relationships were lost with disease in the young samples (7.5%) with a preponderance of positive correlations. In contrast, healthy adult tissue gene levels showed a low frequency of correlations with members of the microbiome (3.9%), which increased to 9.4% of the possible correlations in the adult diseased tissues. The majority of these bacterial-mRNA correlations in adult disease were positive.

Figure 3.

Distribution of the number of immune genes that were significantly correlated with the predominant operational taxonomic units in the young/adolescent (Y/ADO) and adult/aged (AD/AG) animals in healthy and diseased samples. The total available correlations (25,810) are identified as significantly correlated, positively or negatively, with the relative abundance of individual bacteria/taxa. The asterisk (*) denotes significantly greater than all other age or clinical samples at P < 0.05. The hash (#) denotes significantly greater than healthy adult samples at P < 0.05.

The frequency of significant correlations in the young healthy tissues ranged from 40 to 100 host genes for each dominant OTU, while with disease the frequency was lower at 1 to 80 gene correlations (Fig. 4A). In health, the positive and negative correlations with the number of host genes were routinely inversed for a particular OTU. The pattern of correlations was different with disease in the young animals, with nearly 41% of the bacteria showing a similar positive and negative correlation with different sets of genes. However, a group of OTUs (e.g., Bacteria_unclassified, Bacteroidetes_unclassified, C. morbi, Leptotrichia_unclassified, P. fusca, S. parasanguinus II, Streptococcus sp. 058, Treponema_unclassified) were all substantially skewed toward positive or negative correlations in the young diseased samples. In healthy adult samples, there was a limited number of OTUs with skewed correlation patterns (e.g., M. micronuciformis, P. denticola, P. oris, Prevotella sp. 313, Streptococcus_unclassified, V. dispar), all demonstrating many positive correlations with expressed host genes (Fig. 4B). Only Prevotella sp. 317 and P. piscolens were skewed toward negative correlations. However, with adult disease, a dramatic profile of bacteria and gingival gene expression was observed, with nearly 50% of the OTUs showing correlations with many of the immune genes. Of this group, two-thirds were skewed uniquely toward positive correlations.

Figure 4.

The number of host immune response genes that were significantly correlated, positively or negatively, with individual bacteria/taxa in healthy and disease sites from (A) young and (B) adult animal samples.

Complexes of bacteria were also significantly correlated with distinctive patterns of host genes (Appendix Table 3). These patterns of positive/negative relationships of the bacterial complex members differed in the young versus adult samples. In healthy young animals, complexes A and D overlapped their OTU composition, although the correlation with host gene expression was always opposite and the D complex contained species unique to this complex. Similarly, complexes B, E, F, and G demonstrated overlapping OTUs, although only complex F showed host gene response correlations opposite of the other related bacterial complexes. Finally, in the young animals, complexes C and G showed OTU overlap with opposing direction of gene expression correlations. In contrast, while complexes were noted in samples from disease sites, the number of OTUs in the complexes was more limited and showed limited similarities across the complexes. In the adult healthy samples, 6 complexes were identified. Generally, these showed more limited OTU numbers as compared with the complexes in the young animals and reflected positive correlations with the immune response genes. Only complex V showed multiple negative correlations with some OTU overlap with complex Q. More robust complexes were detected in the adult disease samples than in disease samples from the young animals. Also striking was the observation that the OTU members in the complexes from young and adult animals were substantially different.

In young healthy samples, nearly 80% of the immune response genes were related to ≥1 of the bacterial complexes, while only 60% of the genes were correlated in disease (Table). The pathways/functions of the immune response genes were similar for complexes A and C in health and I and K in disease. Furthermore, the remaining bacterial complexes related to the expression of genes that were more focused toward selected biologic pathways. In adults, only 30% of the genes related to bacterial complexes in health and 44% in disease, both substantially lower than in the young animals. Moreover, in the adult animals, the bacterial complexes each displayed rather unique host gene targets.

Table.

Significant Correlations Between Bacterial Complexes and Immune Pathways/Biological Functions Identified by Host Gene Expression Patterns.

Complex	Genes, n (%)	JAK / STAT	TLR	B Cell	IFNγ	IL	T Cell	p38 MAPK	ICC	TGFβ	PDGF	Apop / FAS	p53	IR	Cell Adh	Coag
Young/adolescent
A	92	+	+	+	+	+	+	+	+	+	+
B	90	+			+	+			+
C	58	+	+	+	+	+	+		+	+	+	+	+
D	59								+			+
E	20	+	+				+
F	29								+					+	+
G	12			+
Total	360 (79.6)
H	65	+		+			+		+
I	68		+	+	+	+	+		+			+
J	31			+
K	21		+	+		+	+							+
L	15								+
M	9			+		+			+
N	17					+
O	20			+		+						+
P	32		+				+		+
Total	279 (61.7)
Adult/aged
Q	21	+					+							+	+
R	23	+				+								+
S	53	+				+	+	+	+
T	8
U	25					+		+	+			+
V	9											+	+
Total	139 (30.8)
W	120	+	+	+	+	+			+		+	+			+	+
X	38					+	+
Y	30							+	+
Z	12					+
Total	200 (44.2)

Apop/FAS, apoptotic pathways; Cell Adh, cell adhesion genes; Coag, coagulation pathways; ICC, inflammation-mediated cytokine/chemokine pathways; IL, interleukin signaling pathway; IR, immune response pathways.

We then stratified the genes that were correlated with the bacterial complexes based on a principal role in innate immunity, inflammation, or cellular or humoral immune responses (Appendix Table 4). In the young animals, the number of positively and negatively correlated gene expression levels was generally similar in healthy and diseased tissues. In contrast, the number of immune response genes that were significantly correlated with the bacterial complexes was substantially skewed toward positive correlations in healthy and diseased samples. These data were also examined for the distribution of individual genes with an elevated frequency of correlations within different immune response categories (Appendix Fig. 1). In health and disease in the young animals, genes representing all 4 categories showed that approximately 60% exhibited levels that were primarily positively correlated with the bacterial complexes. In healthy adult samples, innate, cellular, and humoral immune responses were primarily positively correlated, while the inflammation-associated genes were negatively correlated with the microbiome components. In contrast, samples from diseased sites in the adults showed that the majority of the highly correlated genes were positively related to the dominant members of the microbiome across all immune categories.

Finally, we compared the components of the microbial complexes that were highly correlated with immune gene expression levels in young (Y/ADO) and adult (AD/AG) animals in health and disease (Appendix Fig. 2). Overlapping of OTUs that correlated among the bacteria and OTUs that correlated with host gene expression within the younger group of healthy samples was noted (Appendix Fig. 2A: 65% to 79% overlap). A similar relationship was seen with 2 of the 3 complexes in healthy adult samples (Appendix Fig. 2B: 79% and 86% overlap). In contrast, this type of bacteria-bacteria and bacteria-host association was not observed with OTUs and gene expression correlations in disease samples in either the young or adult samples. In these cases, there was <50% overlap between the OTU and gene expression correlations. Finally, Appendix Figure 2C shows that in health and disease, there was a limited overlap of the members of the bacterial complexes between the age groups, and half of the complexes in the adult group had no OTU overlap with complexes from the young animals.

Discussion

Studies have demonstrated colonization of the microbiome of young individuals and children with a microbial ecology that includes various purported oral pathobionts (Bimstein et al. 2002). These also exist in the oral cavity over many decades, apparently unrelated to driving a disease process. Under certain oral/systemic environmental conditions, a subset of the adult population by the fourth decade of life and beyond demonstrates emergence of the pathogens and a dysbiotic microbiome that is directly linked to biological and clinical features of periodontitis. The initiation and progression of periodontitis reflect a microbial symbiome transitioning to a pathobiome through local environmental factors, resulting in a chronic infection with the dysbiotic microbiome. However, the clinical features of periodontitis are driven by a dysregulation of the host response armamentarium, resulting in chronic inflammation and disrupted resolution of this inflammation. Nevertheless, studies of the human model of this disease are limited by the ability to clearly document the earliest transition of the microbiome, as well as the sequence of microbial and host response changes that presage clinical disease. A nonhuman primate model of ligature-induced periodontitis was used to shed light on these events and describe the microbiome-gingival transcriptome interactions in young and adult animals. The results demonstrated differences in the microbiome constituents of young versus adult animals in periodontal health. We also identified specific bacterial complexes in the young and adult animals, with levels of individual members highly correlated and related to specific quantitative profiles of gingival gene expression. These bacterial-gene relationships were greatest in healthy young tissues and were skewed toward a positive correlation supporting a controlled response to the normal bacterial burden. In contrast, with disease, these correlations of host responses to the increased quality/quantity of the microbiome were substantially decreased, indicating loss of this more regulated interaction. The profile of relationships in the adult group was quite different, with low correlations between the bacteria and host responses in healthy tissues that increased significantly with disease. Some evidence linking human gingival cellular and tissue responses to various oral bacteria has been provided for certain periodontal pathogens having a predilection for inducing a core subset of genes related to inflammatory responses (Kebschull and Papapanou 2011). Additionally, complexes of oral bacteria as microbial factors related to the extent/severity of periodontitis have been described (Dahlén et al. 2016). These types of findings were extended to link bacterial complexes with gingival crevicular fluid inflammatory mediator levels in periodontitis (Rescala et al. 2010); Teles et al. 2010). However, human disease is documented via a point-in-time clinical measure, leading to a lack of information regarding the details of the disease process or even if sites sampled were actually no longer demonstrating disease activity or progression. Additionally, various studies showed that selected bacteria were related to certain gene polymorphisms and could be used to classify different types of periodontitis (Divaris et al. 2012; Morelli et al. 2018). Our findings provide some insights to fundamental differences in how the age of the host affects the composition and responses to the oral microbiome. However, further analyses will be needed to provide insight into more specifics of the microbial and host response components that reflect the transition toward a dysbiotic and dysregulated response that hallmarks clinical disease.

We were able to discern individual members of the microbiome that showed an elevated frequency of correlations with individual host immune response genes in the gingival tissues. Of note was the substantial difference in correlation profiles of individual bacterial species/taxa and the host responses between young and adult animals. The frequency of significant correlations in the young healthy tissues was greater than that during disease in this group. Also, in health, particular species/taxa of bacteria exhibited an increased array of positive or negative correlations with the host response genes but generally not both. However, with disease, nearly half of the bacteria showed a similar positive and negative correlation with different sets of genes, with only a select few members of the microbiome showing skewed positive or negative correlations. In the adult healthy samples, this analysis showed a limited number of the species/taxa with elevated, generally positive correlations to host response genes. The disease process in adults resulted in nearly 50% of the OTUs showing correlations skewed positively with many of the immune genes. While there is a general lack of comparable data on periodontitis in the literature, the findings may support that the dysregulated host response pattern with disease in adults is driven by a complex of specific members of the oral microbiome. Thus, targeting these species may provide new insights into risk assessment or even virulence determinants that trigger the disease-related responses.

We identified a series of 78 immune response genes with a high frequency of correlations to the individual bacteria in the oral microbiome. In healthy and diseased tissues from young animals, the genes were associated with inflammation and innate, cellular, and humoral immune activities. While individual genes showed skewing of positive or negative correlations to the bacteria, there was no obvious pattern across these aspects of the host response. A similar analysis in healthy adult tissues showed fewer genes related to the bacteria (n = 31) and that inflammatory genes had a predilection for negative correlations to multiple bacterial species/taxa. With disease in the adults, the number of genes with an elevated number of correlations increased slightly, with the majority in all immune categories showing significant positive correlations with the bacterial levels in the microbiome. While there was a select number of genes that demonstrated the most frequent significant correlations with the microbiome components, focusing on any specific gene (or genes) tends to minimize the complex and interactive network that occurs in responses within the gingival tissues. In health, positively correlated genes were related to innate responses, including NK cell functions (e.g., IFNA2, IFIT1, SIGIRR, KKIR3DL0, KLRF1, IL21), and negatively correlated genes tended to associate with inflammatory responses (e.g., CASP1, NFKB2, NFKB1, ATG16L1). With disease, the positively correlated genes represented likely effects on T- and B-cell functions (e.g., TNFRSF13B, CCL18, SELPLG, IL18R1, IL18RAP), with few individual genes showing negative correlations to the bacteria. Also, of this group of host genes with a high number of bacterial correlations, there was actually little overlap between the age groups and in the healthly versus diseased tissues. Again the literature is limited regarding any comparative data; however, one view would be an emphasis on unique features of host tissues’ response to the changing microbial burden that is affected by the age of the individual.

Using this model, we were able to discern somewhat discrete bacterial complexes with multiple members that differed in the young and adult samples. However, the data demonstrated a predilection for these complexes to strongly correlate with host gene families associated with specific pathways and functions. Also of interest was that in these complexes, certain species/taxa were positively or negatively correlated with the same repertoire of host genes. An additional observation was that, in the young samples, there was a greater number of bacterial complexes, and they correlated to a significantly greater number of host gene expression levels than noted in the adult samples. The data also suggested that the correlated immune system gene representation was similar in health and disease in the young and health in adults; however, innate immune genes were overrepresented, and there was a decrease in cellular immunity–related gene expression in diseased tissues from adults. The findings are consistent with altered cellular immune response capacity at the level of gene expression and immune pathway functions in progressing periodontitis that have been reported in rodent models (Han et al. 2007) and humans (Abusleme and Moutsopoulos 2017; Dutzan et al. 2018).

Our findings with the nonhuman primate model of periodontitis extend observations in humans regarding microbiome changes with disease and demonstrate clear differences in bacterial complexes and associated gingival tissues’ responses in young and adult individuals. The interactions between individual members of the bacterial complexes and pathways/functions of host immune genes provide additional insight into the complexity of these relationships. Moreover, the details of these correlations may help clarify the differences in prevalence and severity of disease expression that increase with age.

Author Contributions

J. Ebersole, O.A. Gonzalez, contributed to conception, design, data acquisition, analysis, and interpretation, drafted and critically revised the manuscript; S. Kirakodu, J. Chen, R. Nagarajan, contributed to data analysis and interpretation, critically revised the manuscript. All authors gave final approval and agree to be accountable for all aspects of the work.