Porphyromonas spp., Fusobacterium spp., and Bacteroides spp. dominate microbiota in the course of macropod progressive periodontal disease

Macropod progressive periodontal disease (MPPD) is a necrotizing, polymicrobial, inflammatory disease commonly diagnosed in captive macropods. MPPD is characterized by gingivitis associated with dental plaque formation, which progresses to periodontitis and then to osteomyelitis of the mandible or maxilla. However, the underlying microbial causes of this disease remain poorly understood. In this study, we collected 27 oral plaque samples and associated clinical records from 22 captive Macropodidae and Potoroidae individuals that were undergoing clinical examination at Adelaide and Monarto Zoos in South Australia (15 healthy, 7 gingivitis and 5 periodontitis-osteomyelitis samples). The V3-V4 region of the 16S ribosomal RNA gene was sequenced using an Illumina Miseq to explore links between MPPD and oral bacteria in these animals. Compositional differences were detected between the microbiota of periodontitis-osteomyelitis cases compared to healthy samples (p-value with Bonferroni correction < 0.01), as well as gingivitis cases compared to healthy samples (p-value with Bonferroni correction < 0.05) using Permutational Multivariate Analysis of Variance (PERMANOVA). An overabundance of Porphyromonas, Fusobacterium, and Bacteroides taxa was also identified in animals with MPPD compared to healthy individuals using linear discriminant analysis effect size (LEfSe; p =  < 0.05). An increased abundance of Desulfomicrobium also was detected in MPPD samples (LEfSe; p < 0.05), which could potentially reflect differences in disease progression. This is the first microbiota analysis of MPPD in captive macropods, and these results support a polymicrobial pathogenesis of MPPD, suggesting that the microbial interactions underpinning MPPD may be more complex than previously documented.

www.nature.com/scientificreports/ of disease (gingivitis and periodontitis-osteomyelitis) compared to the healthy condition. This will add to our understanding of the microbiology of MPPD in macropods and may lead to improved preventative measures or treatment of the disease.

Results
Clinical descriptions. After examination, 27 samples from 22 animals were classified as healthy (n = 15), gingivitis (n = 7) or periodontitis-osteomyelitis cases (n = 5). Supplementary 1 provides further clinical details on the two most severe cases of MPPD sampled.
Microbiota profiling. The relative abundances of bacteria at the phylum, genus, and species taxonomic levels were explored in 27 animals belonging to healthy, gingivitis and periodontitis-osteomyelitis categories. In total, 1178 OTUs were detected, assigned to 28 phyla, 186 families, 438 genera and 181 species (Fig. 1). Proteobacteria and Actinobacteria were the most abundant phyla in healthy samples, forming 80.8% of the entire microbiota (Supplementary 2). We also explored microbiota identified at the class and genus levels in healthy, gingivitis, and periodontitis-osteomyelitis samples ( Fig. 2  MPPD is associated with compositional shifts in oral microbiota, but not diversity.. We examined if shifts in alpha diversity were apparent between health and diseased animals using the Shannon's Diversity Index (Fig. 3). There were no significant shift in alpha diversity between healthy, gingivitis, and periodontitisosteomyelitis sample groups (Fig. 3).
In contrast, the composition (beta diversity) of oral microbiota in periodontitis-osteomyelitis samples were distinguishable from the healthy and gingivitis samples using Principal Coordinate Analysis (PCoA) plot of Bray Curtis Values (Fig. 4). In a PCoA, periodontitis-osteomyelitis samples grouped to the exclusion of healthy samples on the first PCoA component of a 3D PCoA plot (Fig. 4). Supplementary 7 presents the formula PCoA components. The first component explains 31% of variance in the microbiome data (Fig. 4). The first component separated periodontitis-osteomyelitis samples from healthy and gingivitis samples, as periodontitis-osteomyelitis samples had the lowest negative coefficient values for this component (Supplementary 7). Further, the microbial composition of healthy animals was significantly different from those suffering from both periodontitis-osteomyelitis (PERMANOVA with Bonferoni correction; 0.0002) and gingivitis (p = 0.0323) ( Table 2). Further, the composition of oral micorbiota in animals with periodontitis-osteomyelitis was not significantly different from those with gingivitis (p = 0.0833) ( Table 2), suggesting that there may be compositional similarities between these two stages of disease. Together, these results suggest that a distinct oral microbial communities are associated with PD and gingivitis in macropods compared to healthy animals.  In the comparison of gingivitis and healthy samples, Porphyromonas (logarithmic LDA score = 4.743 and p-value = 0.001), Fusobacterium (logarithmic LDA score = 4.648 and p-value = 0.031), and Bacteroides (logarithmic LDA score = 4.666 and p-value = 0.044) were the dominant taxa that were overrepresented in gingivitis samples. In contrast, Neisseria, a genera within the Proteobacteria (the most abundant phyla in healthy  In the comparison of periodontitis-osteomyelitis and healthy samples, Porphyromonas (logarithmic LDA score = 4.743 and p-value = 0.001), Fusobacterium (logarithmic LDA score = 4.648 and p-value = 0.031), and Bacteroides (logarithmic LDA score = 4.666 and p-value = 0.044) were the dominante species that were overrepresented   We also investigated differences in taxa abunance in animals suffering from gingivitis or periodontitis-osteomyelitis. Eighteen genera were significantly different in abundance in both the gingivitis versus healthy and periodontitis-osteomyelitis versus healthy sample comparisons (Table 1, Supplementary 6). Of particular note, the abundance of Desulfomicrobium increased in both gingivitis and periodontitis-osteomyelitis in comparison to healthy samples (Table 1), with the marked increase in periodontitis-osteomyelitis samples potentially having additive effects on disease progress (logarithmic LDA score = 3.564 and p-value = 0.002) (Supplementary 5).
We confirmed these resulst by additional performing Pearson correlations on species abundance and disease. Pearson correlations revealed highly significant, positive correlations (> 72%, p < 0.01) between the relative abundance of Porphyromonas, Fusobacterium, Bacteroides, and Desulfomicrobium in oral macropod samples (Table 3). In contrast, there was negative correlation between the abundance of Neisseria genus with Porphyromonas, Fusobacterium, Bacteroides, and Desulfomicrobium genera ( Table 3). Signature of Porphyromonas, Fusobacterium, and Bacteroides in MPPD was robust and consistent in different sexes, species, and zoos (Supplementary 8).

Discussion
Periodontal diseases are commonly reported clinical disorders in animals. Here, for the first time, the oral microbiota of healthy and MPPD-affected captive macropods were explored using 16S rRNA gene seuqencing. This study characterised the healthy macropod oral microbiota and identified differences in bacterial composition and taxa abundances between healthly samples and the different stages of MPPD, thus providing novel information about this polymicrobial disease and its prospective pathogenesis.
There are similarities and differences between PD in human and MPPD in macropods. In humans, PD is initiated by dental plaque and exhibits progression from reversible gingivitis to irreversible periodontitis. However, Table 2. Permutational multivariate analysis of variance (PERMANOVA) test using Bray-Curtis index showed significant differences in microbiota comparison of periodontitis-osteomyelitis group against healthy group as well as gingivitis against healthy group. www.nature.com/scientificreports/ unlike PD in humans, MPPD commonly progresses to necrotising osteomyelitis of the mandible or maxilla, with the formation of sequestra and proliferation of subperiosteal bone subsequently leading to bone deformity in the jaw (the characteristic 'lumpy jaw'). The progression to osteomyelitis, suppurative inflammation and necrosis of adjacent soft tissues observed in macropods is rare in humans 3,28 . In humans, early colonisers, such as Streptococcus spp. and Fusobacterium spp. provide a foundation that can allow late colonising bacteria to attach, which includes many anaerobic Gram-negative bacteria [4][5][6]29 . In the present study, we found that Porphyromonas, Fusobacterium, and Bacteroides are the most abundant genera in gingivitis samples. These genera also dominated the microbiome in periodontitis-osteomyelitis cases (the advanced stages of disease). Also, we observed a significant and positive correlation between the relative abundances of Porphyromonas and Fusobacterium genera and disease, similar to human PD. In contrast to humans, Streptococcus did not appear to be a major component of plaque microbiota in macropods. It has been well-documented that Fusobacterium nucleatum can enhance the attachment and invasion of Porphyromonas gingivalis or Aggregatibacter actinomycetemcomitans to human gingival epithelial cells 29 . In another study, Porphyromonas gingivalis entry into gingival epithelial cells was modulated by Fusobacterium nucleatum 30 . Additionally, it has been proposed that Fusobacterium spp. can inhibit the initial host innate immune response 29 . Scanning electron microscopy has shown that P. gingivalis and F. nucleatum can form consortia and penetrate Ca9-22 cells within 30-60 min after infection (early colonisation) 30 . Altogether, we suggest that Fusobacterium may be involved in early colonisation in MPPD, enhancing adhesion and invasion of species within the Porphyromonas genus that are likely to be either P. gulae or loveana. This interplay has been observed in murine alveolar bone loss and arthritis onset 31 .

Gingivitis
Microbiota profiling using 16S rRNA gene pyrosequencing identified the Gram-negative bacterial genera Fusobacterium, Bacteroides, and Porphyromonas as the dominant taxa in MPPD, as noted in previous culture and DGGE-based studies 3,8,17,18,20 . These findings support the hypothesis that Porphyromonas and Bacteroides may have a bigger role in disease pathogenesis than has historically been proposed. The importance of pathogens other than F. necrophorum in MPPD was first noted in 1977, and recent DGGE-based molecular studies, confirmed that Porphyromonas and Bacteroides are important genera in MPPD pathogenesis 3,20,24 . However, caution should be applied, as bacterial abundance may not necessarily correlate with pathogenicity.
Abundance of Desulfomicrobium also increased in MPPD samples (LEfSe test, p < 0.05), which could potentially have additive effects on disease progress. Desulfomicrobium orale has also been isolated from subgingival plaque of human patients with PD 32 , identified as a human oral pathogen 33 . This is the first time that another significant genera associated with PD in humans has been reported in MPPD, albeit at low abundance when compared to Porphyromonas, Bacteroides and Fusobacterium.
The dominance of the Proteobacteria phylum found in the oral microbiota of healthy macropods was similar to another study of the salivary microbiota in Tammar wallabies 25 . This contrasts with previous reports on the oral microbiota of other marsupials, such as Tasmanian Devils, which have a similar proportion of Proteobacteria, Bacteroidetes, Fusobacteria, and Firmicutes phyla (at around 20% of each) and koalas, where Bacteroidetes and Firmicutes were in the top three phyla 27,34 . In macropods, the highly abundant Proteobacteria phylum is composed of many species associated with gingival health in comparison to either gingivitis or periodontitis-osteomyelitis 3 . In particular, Pasteurellaceae and Moraxellaceae were the two major families found to be abundant in healthly samples 3 . An increased abundance of Gram-positive non-sporulating rods such as Corynebacterium sp. and Actinomyces sp. in healthy compared to diseased samples has also been reported in culture-based studies 18,23 . In line with those studies, species in the genera Corynebactrium, Actinomyces, Streptococcus, Lautropia, Leptotrichia and Capnocytophaga are associated with the healthy oral cavity in human microbiota studies 35 . It has also been noted that some disease-associated genera have overlapping species that are also present in healthy samples 35 .
Despite being the first study to profile the microbiota of MPPD using 16S rRNA gene pyrosequencing, this study had some limitations. In microbiota studies, sequencing depth normally includes the family, genus and sometimes species level, but there can be many OTUs that cannot be identified at the lowest level. The current study was based on a single zoological collection of animals at two sites, and the availability of MPPD cases was a limiting factor on sample size. Subsequently, there was an unequal number of healthy, gingivitis and periodontitis-osteomyelitis cases available for study, with only a limited number of gingivitis and periodontitis-osteomyelitis www.nature.com/scientificreports/ samples. Additionally, the low sample size meant that different species of macropods were included together in the study and the yellow-footed rock wallaby (YFRW) was overrepresented. Larger sample sizes, and comparison between macropod species and zoological collections, would be beneficial in future studies.

Conclusion
Despite individual-to-individual variation, bacterial communities likely undergo largely conserved changes during PD 36 . For the first time, we have profiled the shift in oral microbiota of captive macropods at different stages of MPPD, as well as characterised the healthy gingival microbiome. Porphyromonas, Fusobacterium, Bacteroides, and Desulfomicrobium may play key roles in this disease, as they appear at higher prevalence in MPPD cases compared with healthy animals. Overall, these results support a polymicrobial pathogenesis of MPPD and suggest that the diversity of bacteria involved, and the interactions between them, may be more complex than has been documented previously. Improving our understanding of the pathogenesis of MPPD is key to the development of more effective preventative and therapeutic measures.

Materials and methods
Ethics. All experiments and experimental protocols were approved by Zoos South Australia Animal Welfare and Ethics Committee, Australia. All experiments were performed in accordance with relevant guidelines and regulations of Zoos South Australia as well as School of Animal and Veterinary Sciences, The University of Adelaide. The study was carried out in compliance with the ARRIVE guidelines (https:// arriv eguid elines. org). Medical records (including dental charts and diagnostic imaging where available) were used to classify oral health status at the time of sampling. Animals were classified similarly to Antiabong et al. 24 as healthy, gingivitis and periodontitis-osteomyelitis cases, based predominantly on dental examination, with gingivitis and periodontitis-osteomyelitis cases respresenting early and advanced cases of MPPD, respectively. Individuals with gingivitis had gross swelling, redness of the gums, bleeding on swabbing, minor gingival recession and/or early periodontal pocket formation. Periodontitis-osteomyelitis was defined here as periodontitis with or without more progressive disease associated with soft tissue and/or bone involvement. Individuals were classified with periodontitis-osteomyelitis if, in addition to gingivitis, they had severe gingival recession, deep periodontal pockets, tooth mobility, bone necrosis, and/or other more severe lesions. Healthy animals were classified by the absence of the above gross lesions. For MPPD cases, the sample analysed was that collected from the affected tooth. Additional findings such as plaque in healthy animals (in the absence of gingivitis), other disease conditions present, episodes of previous dental disease and pouch status were also recorded. DNA extraction. The periodontal plaque samples were thawed at 37 °C in an anaerobic chamber (Coy, Grasslakes, Michigan, USA) and then vortexed for 20 s. DNA extraction was performed using the QIAGEN DNeasy Blood and Tissue Kit (Qiagen Hilden, Germany) spin-column protocol, modified to improve DNA yield by using 400 µL of sample, 40 µL of Proteinase K, 400 µL buffer AL and 400 µL ethanol. The elution step was performed twice, using 70µL and then 30 µL of elution buffer. Sample DNA concentrations and quality were tested using Nanodrop 1000 spectrophotometry (Thermo Fisher, Waltham, Massachusetts, USA). Concentrations were also measured with a Qubit Fluorimeter, following the assay preparation instructions from Qubit dsDNA HS Assay Kits (Thermo Fisher).
Illumina Miseq 16S rRNA gene sequencing. Published primers were used for the amplification of the 550 bp V3-V4 region of the bacterial 16S rRNA gene. PCR products were visualised following electrophoresis in agarose and staining with Gel Red™ to confirm positive yield for each sample. Samples were submitted to the South Australian Health & Medical Research Institute DNA Sequencing Facility for 16S Microbiota library preparation and sequencing. Library preparation followed the Illumina library preparation protocol, with the following primers: forward CCT ACG GGNGGC WGC AG, reverse GAC TAC HVGGG TAT CTA ATC C. Sequencing was carried out by Illumina Miseq V3 SBS Chemistry targeting machine. Amplicons were sequenced as paired reads with the length of 300 bp (2 × 300 bp). www.nature.com/scientificreports/ Microbiota profiling. Adapter trimming, fixed length trimming, merging paired reads and filtering based on the number of reads (to remove the samples with low coverage) were performed to obtain high quality sequence reads with enough depth for microbiota profiling and comparison as previously described 37 . CLC Microbial Genomics Module (QIAGEN) Version 11 was used to assign taxonomy to the reads from different samples. To this end, reads were clustered using representative sequences of pseudo-species called OTUs (operational taxonomic units). The OTU clustering tool clusters the reads and reduces the read collection in each sample to representative sequences (cluster centroids) that are 97% similar to any member of the cluster they represent. The SILVA database was used as the reference of 16S rRNAs 38 . The number of reads assigned to each OTU and the relative abundance of each OTU was calculated 39 .
Statistical analysis. Linear discriminant analysis (LDA) effect size (LEfSe) 40 algorithm, a test for highdimensional biomarker discovery in metagenomic data, was used to find differentially abundant bacteria between periodontitis-osteomyelitis versus healthy as well as gingivitis versus healthy samples. The following criteria were used for selection of bacteria with statistically different relative abundance: alpha value = 0.05 and threshold of absolute logarithmic LDA score > 2. Analysis was performed in the Galaxy platform (https:// hutte nhower. sph. harva rd. edu/ galaxy/). LEfSe algorithm benefits from a range of tests: (1) Non-parametric factorial Kruskal-Wallis (KW) sum-rank test to find bacteria with statistical significant differences between groups; (2) a set of pairwise tests among subclasses using the (unpaired) Wilcoxon rank-sum test; and (3) LDA for estimating the effect size of each differentially abundant bacteria as well as feature selection (dimension reduction) 40 .
Alpha diversity estimates the diversity within samples. Alpha diversity was performed based on Shannon Index. Kruskal-Wallis H test was used for measuring the statistical significance of alpha diversity. Kruskal-Wallis H assesses whether the values originate from the same distribution or whether their distribution is different depending on the group they belong to. This test is a nonparametric alternative to ANOVA. A significant p-value for the Kruskal-Wallis test means that at least one group has a different distribution. However, Kruskal-Wallis does not report which pairs have different distributions. Mann-Whitney U test was used to performs a pair-wise test to specifically find which pairs of groups follow different distributions.
Beta diversity measures the change in diversity between groups. Beta diversity was calculated in using CLC Microbial Genomics Module in two steps: (1) estimating the distance between each pair of samples; and (2) performing Principal Coordinate Analysis (PCoA) on the distance matrices. Bray-Curtis measurement was used to calculate the distance matrices.
PERMANOVA test 41 , also known as non-parameteric MANOVA, measures the effect of size and significance on beta diversity in comparisons of gingivitis group versus healthy and periodontitis-osteomyelitis versus healthy samples. PERMANOVA obtains its significance from a permutation test. The number of permutations was set to 99,999. Analysis was performed by CLC Microbial Genomics Module.