Patent Grant
11390924
This application is a U.S. national stage patent application of International Patent Application No. PCT/GB2014/000234, filed Jun. 13, 2014, which claims the priority of GB Application No. 1310691.9, filed Jun. 14, 2013, the entire contents of each of which are incorporated herein by reference.
The present invention relates to an assay for use in a method of determining the oral health status of a canine animal by identifying certain bacteria present or absent in a sample taken from the animal, and applying the information set out herein for each identified bacteria to statistical models in order to determine the oral health status of the animal.
Periodontal disease is a significant problem in dogs, affecting 56-60% of the adult population. It is an inflammatory disease of the supporting tissues of the teeth with tissue damage eventually leading to tooth loss if left untreated. The aetiological agent of periodontitis is dental plaque, a complex biofilm of bacteria suspended in a matrix of bacterial exudates and secreted products. The activity of some bacteria induces a host immune response and results in inflammation of the gingival tissue referred to as gingivitis (G). If the disease progresses, tissue damage becomes more severe leading to an increasing loss of the periodontal ligament surrounding the tooth and is referred to as early stage periodontitis (PD1). Further progression to advanced periodontitis (PD3-PD4) is characterized by significant destruction of the periodontal ligament and supporting tissues including bone. If left untreated the condition is painful for a prolonged period before eventual tooth loss occurs. There is also some evidence in the human field that periodontitis is a risk factor for numerous other conditions, most notably cardiovascular disease.
Assessment of canine oral health is usually made by performing a clinical examination of the dog's mouth while the animal is under anaesthesia, which is a time consuming and resource intensive process which is not without risk to the animal. Physical characteristics of the teeth and gums are used to determine the levels of gingivitis and periodontitis. If other indicators could be reliably used to determine oral health status there may be benefits both in terms of the need to anaesthetise dogs and the ways in which dogs can be used for oral health research.
Traditional methods used to diagnose periodontitis rely on clinical indicators such as signs of inflammation, probing depths, extent and pattern of loss of clinical attachment and bone and other symptoms including the amount of observable plaque and calculus. Such an examination is costly and requires highly trained professionals to examine patients closely. In the case of dogs thorough examination usually requires the attention of a veterinary dentist. Also routine dental maintenance, scaling and inspection of any diseased areas usually require a general anaesthetic to be applied, further complicating the procedure and increasing the resources required.
It is accepted that bacteria present in human dental plaque are the aetiological agent of periodontal disease; though the specific organisms involved in the initiation of disease and the basis of the subsequent events thereafter are unclear. A working hypothesis is that specific antigens or enzymes produced by bacteria in the plaque biofilm initiate activation of the host inflammatory response, the latter being the main pathological agent of periodontal disease.
The initial stages of disease are observed clinically as red and inflamed gums, defined as gingivitis. Without treatment by removal of the plaque biofilm, gingivitis may progress to early periodontitis. The earliest stage of periodontitis (PD1) is characterised by initial tissue breakdown and loss of up to 25% attachment of the periodontal ligament surrounding the tooth root. In humans, this switch from gingivitis to periodontitis appears to be restricted to 10-15% of the population. The onset of periodontitis is defined by irreversible tissue destruction and if left untreated will progress to extreme periodontitis (PD3-PD4). This is characterised by extensive (50-75%) destruction of the periodontal ligament, gum recession and breakdown of supporting tissues eventually leading to the loss of the tooth. The periodontal disease process can be inhibited in the early stages (PD1) by dental scaling and polishing of the periodontal pocket to remove the source of inflammation (dental plaque) with subsequent regular plaque removal by tooth brushing. As such, increasing the understanding of the early stages of disease, (gingivitis through to PD1) in pet dogs where non-surgical interventions may be effective would be desirable.
The diversity of bacterial species found in the canine oral microbiome has been reported using culture independent molecular methods from 51 dogs. Based on full length 16S rDNA Sanger sequencing, 353 taxa were identified; of these 80% were novel and only 16.4% were shared with the human oral microbiome. This indicates a clear difference between the bacterial populations in human versus canine mouths.
Dewhirst et al., (PLoS One, Vol 7, 2012) describes a study to identify the major species of bacteria present in the canine oral microbiome. This paper describes the major bacteria present in the canine oral microbiome. However, without information linking the different species and genera of bacteria found to specific health conditions (whether healthy, gingivitis or periodontitis) then knowledge about which bacteria are present is not informative with regards to predicting the health state of the animal concerned.
Sturgeon et al., (Vet. Microbiol., Vol. 162, 2013) describe a study that used pyrosequencing of 16S rRNA gene to study oral samples from six healthy dogs. This paper lists the genera found in the six healthy dogs. Since the authors did not test dogs that were not healthy (with gingivitis or periodontitis, for example) it is not possible to tell if any of the genera that they identified are especially characteristic of health. In the absence of these data, simply knowing that these genera are present in some healthy samples is not informative with regard to predicting the health state of the animal concerned.
Thus, there is a clear need to identify particular bacterial species in canine plaque that are significantly associated with health, gingivitis and mild periodontitis.
Therefore, the present invention provides an assay for use in a method for determining the oral health status of a canine animal, the method comprising an assay, wherein the assay comprises means for identifying at least two bacteria selected from the list consisting of
Parvimonas
Peptostreptococcus
Moraxella
Filifactor
Schwartzia
Treponema
Capnocytophaga
Atopobium
Phascolarctobacterium
Globicatella
Prevotella
Curtobacterium
Granulicatella
Solobacterium
provided that at least one of the bacteria is selected from the list consisting of
Parvimonas
Filifactor
Schwartzia
Capnocytophaga
Atopobium
Phascolarctobacterium
Globicatella
Curtobacterium
Granulicatella
Solobacterium
In a sample taken from the animal and applying the information set out in
The assay may also further comprise means for identifying at least one bacteria selected from the list consisting of;
Leucobacter
Odoribacter
Selenomonas
Actinomyces
Propionivibrio
Xenophilus
Corynebacterium
Escherichia
Lautropia
Leptotrichia
in a sample taken from the animal.
Alternatively or additionally, the assay may also comprise means for identifying at least one bacteria selected from the list consisting of;
Staphylococcus
Tannerella
Arcobacter
Catonella
Chryseobacterium
Fusobacterium
The bacteria identified may be of a particular species or are of a genus. Where a genus is identified, any number of members of that genus can be indicative of a particular oral health status when used in the assay of the invention; the sum of all members of the genus may be used in the predictive models. Where a species is identified, different species from the same genus cannot be assumed to have the same predictive value.
The inventors surveyed the oral microbiota of a sufficiently large canine cohort, at great enough depth to identify significant changes in bacterial taxa (phyla, genera and species) between dogs with healthy gingiva and those with gingivitis or mild periodontitis (PD1), and found important links between certain bacteria and different health states.
In this way, the inventors have developed a more relevant conscious testing methodology whereby the oral health status of the animal is assessed by molecular markers for disease state, such as bacterial species. Such a test allows much higher numbers of animals to be assessed and eliminate the need for general anaesthetics.
The development of such a method also has applications for home-owned pets. A conscious test for oral health state enables more frequent monitoring of a pets oral health and provides encouragement for the use of preventative measures such as oral care treats and tooth brushing.
The oral health status of the canine animal may be classified as healthy, gingivitis or periodontitis. The models used to reliably predict the oral health status can predict health/not health (referred to herein as H/not H), where “not health” means gingivitis or periodontitis, or disease/not disease (also referred to herein as P/not P), where “not disease” means health or gingivitis. In this way, models can be combined in order to predict whether an animal has gingivitis, is healthy or has periodontitis.
The sample from the animal to be tested in the animal may be dental plaque, gingival crevicular fluid saliva, or a mixture of any of these. As an advantage over present methods for determining the oral health status of an animal, such samples can be obtained non-invasively and without the need for anaesthetic or expert veterinary care. The sample may be used in the assay of the invention immediately, or it may be processed and stored for future analysis.
The means to identify the two or more bacteria present in the sample may be Quantitative PCR, sequencing or antibody binding. Fluorescent in situ hybridisation may be used. Methods of extracting DNA or protein from bacteria are well known to one skilled in the art.
The bacteria disclosed herein each have 16S DNA sequences that are used to identify them, as is known in the art. Such sequences are publically available and, as such, enable the design of primers by the skilled person. The COT (canine oral taxon) numbers associated with each species or genus described herein enables its identification through sequencing. The species may, for example, be identified through www.ncbi.nlm.nih.gov/nuccore, carrying out a text search for “canine oral taxon” which produces the list of known species, and their associated COT number. Sequences associated with each species are given, enabling, for example, primer design for species identification within a sample from the animal.
Methods of extracting bacteria from a sample are well known to one skilled in the art, as are techniques for extracting DNA from bacterial cells.
Sequencing techniques are well known in the art, including primer design, PCR techniques, sequencing techniques, and antibody assays, such as ELISAs. Antibodies to bacteria-specific proteins can be generated by the skilled person and used to detect certain bacteria in a sample by routine methods.
The assay may comprise means for identifying from 2 to 20 bacteria, or suitably, from 3 to 10, or 4 to 12 bacteria. Alternatively, the assay may comprise means to identify from 2 to 100 bacteria, 5 to 50 bacteria, or 10 to 30 bacteria selected from the lists of species and genera as set out above.
The step to identify the bacteria in order to determine the oral health status of the animal may comprise determining the presence or absence of two or more of a bacteria selected from the list consisting of;
Parvimonas
Peptostreptococcus
Moraxella
Filifactor
Schwartzia
Treponema
Capnocytophaga
Atopobium
Phascolarctobacterium
Globicatella
Prevotella
Curtobacterium
Granulicatella
Solobacterium
provided that at least one of the bacteria is selected from the list consisting of;
Parvimonas
Filifactor
Schwartzia
Capnocytophaga
Atopobium
Phascolarctobacterium
Globicatella
Curtobacterium
Granulicatella
Solobacterium
in a sample from the animal. The presence or absence may be referred to as a ‘binary’ test. The presence or absence may be determined of one or more further bacteria selected from the list consisting of;
Leucobacter
Odoribacter
Selenomonas
Actinomyces
Propionivibrio
Xenophilus
Corynebacterium
Escherichia
Lautropia
Leptotrichia
in a sample from the animal.
Alternatively or additionally the presence or absence may be determined of one or more further bacteria selected from the list consisting of;
Staphylococcus
Tannerella
Arcobacter
Catonella
Chryseobacterium
Fusobacterium
in a sample from the animal, in order to predict the oral health status of an animal.
The presence of bacteria associated with disease can give an indication that the animal has gingivitis or periodontitis. The absence of a bacteria associated with disease is a good indication that the dog has good oral health. The presence of the health associated bacteria can help to determine how healthy is the mouth of the animal, although is less strong of an indicator than the presence of a disease associated bacterial species or genera. A binary test (determining presence or absence) can involve identifying just one incidence of a bacterial species or there may be a threshold, in that a particular bacterial species or genus is not considered present until the count for that particular species/genera reaches at least 3, or at least 5 or at least 7, or at least 9.
A bacterial count may be determined by the number of times its sequence information is identified in a sample, by qPCR or by colony count. By counts, it is meant an absolute number, rather than a proportional number.
Alternatively or additionally the step to determine the oral health status may comprise determining the proportion of total plaque bacteria of two or more of a bacteria species or genera selected from the list consisting of
Parvimonas
Peptostreptococcus
Moraxella
Filifactor
Schwartzia
Treponema
Capnocytophaga
Atopobium
Phascolarctobacterium
Globicatella
Prevotella
Curtobacterium
Granulicatella
Solobacterium
provided that at least one of the bacteria is selected from the list consisting of
Parvimonas
Filifactor
Schwartzia
Capnocytophaga
Atopobium
Phascolarctobacterium
Globicatella
Curtobacterium
Granulicatella
Solobacterium
in a sample from the animal.
The proportion of total plaque bacteria may be determined, at a further one or more bacteria selected from the list consisting of
Leucobacter
Odoribacter
Selenomonas
Actinomyces
Propionivibrio
Xenophilus
Corynebacterium
Escherichia
Lautropia
Leptotrichia
in a sample from the animal
Alternatively or additional, the proportion of total plaque bacteria may be determined of a further one or more bacteria selected from the list consisting of
Staphylococcus
Tannerella
Arcobacter
Catonella
Chryseobacterium
Fusobacterium
in a sample from the animal
By proportion it is meant the percentage of total bacteria within the sample that is formed by a particular bacterial species or genus.
Alternatively or additionally the step to identify the bacteria in order to determine the oral health status may comprise determining the number of counts of two or more of a bacteria species or genera selected from the list consisting of
Parvimonas
Peptostreptococcus
Moraxella
Filifactor
Schwartzia
Treponema
Capnocytophaga
Atopobium
Phascolarctobacterium
Globicatella
Prevotella
Curtobacterium
Granulicatella
Solobacterium
provided that at least one of the bacteria is selected from the list consisting of
Parvimonas
Filifactor
Schwartzia
Capnocytophaga
Atopobium
Phascolarctobacterium
Globicatella
Curtobacterium
Granulicatella
Solobacterium
in a sample from the animal.
The number of counts may be determined of a further one or more bacteria selected from the list consisting of
Leucobacter
Odoribacter
Selenomonas
Actinomyces
Propionivibrio
Xenophilus
Corynebacterium
Escherichia
Lautropia
Leptotrichia
in a sample from the animal.
Alternatively or additional, the number of counts may be determined of a further one or more bacteria selected from the list consisting of
Staphylococcus
Tannerella
Arcobacter
Catonella
Chryseobacterium
Fusobacterium
in a sample from the animal.
Information such as the (i) presence or absence; (ii) the proportion of total bacteria; or (iii) number of counts of two or more bacterial species or genus can be used individually or in combination in order to predict the oral health status of an animal, as shown by the examples, below.
The present invention also relates to a method of determining the oral health status of a canine animal, the method comprising the use of the assay of the invention to identify at least two bacteria as herein described in a sample taken from the animal and applying the information set out in
Statistical models that are able to be used in order to generate such predictions are well known in the art, using the data included herein, in tables 1 to 5. The data show the number of times a certain species or genus was found in the oral cavity of animals, and whether each animal in which it was found has been classed as being in good oral health, having gingivitis or periodontitis.
This information, combined with information regarding a particular animal obtained from the assay of the invention, and the use of statistical models known to the skilled person (examples of which are described herein) can lead to a prediction that is up to 83% accurate.
Currently, without using methods (that involve invasive procedures and anaesthetics), a canine animal is predicted to have a 50/50 chance of suffering from periodontitis, or not (i.e. being P/not P). Thus, the present invention shows a clear improvement and benefit over current invasive, time consuming and expensive methods.
Once the information on presence/absence and/or proportion and/or counts of two or more bacteria has been obtained, one or more known statistical models can be used to predict the oral health status of the canine animal.
Models are shown in Examples 2 to 12. Further models are known to one skilled in the art. The information obtained can be analysed to produce predictive values to enable the oral health status of an animal (health, gingivitis or periodontisis/disease) to be determined.
By determining the oral health status it is meant that the current status is determined, such as whether the mouth of the animal is in good oral health, has gingivitis or periodontitis at the time of taking the sample. The oral health status prediction may also encompass a prediction on whether a healthy mouth is likely to develop gingivitis or periodontitis, or whether gingivitis is likely to develop into periodontitis, based on the bacterial species or genera that are present/absent and the levels or counts of such bacteria. Thus, the assay of the present invention allows reliable prediction of the future oral health of the animal as well as providing information on the current oral health of the canine animal.
In practice, the method of the invention involves the (non-invasive) collection of a sample from the oral cavity of a canine animal, the use of the assay of the invention to identify the presence or absence of the herein described bacteria in the sample. The information can then be compared to that presented in
The present invention also provides a method of improving or maintaining the oral health of an animal, the method comprising determining the oral health status of a canine animal using the assay according to the invention and providing to the animal a foodstuff or supplement which is formulated to improve or maintain oral health, depending on the oral health status that has been determined by way of the assay.
Such food products are known in the art, such as those containing active ingredients to improve oral health or those designed to remove plaque by abrasion, analogous to regular tooth-brushing. The amount or frequency of the foodstuff or supplement can be determined depending on the result of the assay. The predicted future health of the animal can also be taken into account when determining how often such oral care foodstuffs should be provided.
The present invention also provides an assay for determining the oral health status of a canine animal, the assay comprising means for identifying at least two bacteria selected from the list consisting of Anaerovorax.sp.COT-124, Actinobaceria.sp.COT-376, Actinomyces.sp., Actinomyces.sp.COT-252, Aquaspirillum.sp.COT-091, Bergeyella.zoohelcum.COT-186, Brachymonas.sp.COT-015, Campylobactersp.COT-011, Capnocytophaga.canimorsus, Capnocytophaga.sp.COT-339, Capnocytophaga.sp.COT-362, Cardiobacterium.sp.COT-176, Desulfovibrionales.sp.COT-009, Escherichia.coli, Fusobacterium.sp.COT-169, Fusobacterium.sp.COT-189, g.Atopobium, g.bacterium.cp04.17, g.Capnocytophaga, g.Corynebacterium, g.Ottowia, g.Parvimonas, g.Prevotella, g.Schwartzia, Globicatella.sp.COT-107, LachnospiraceaeXIVa[G-2].sp.COT-062, LachnospiraceaeXIVa[G-4].sp.COT-099, Lautropia.sp.COT-060, Leucobacter.sp.COT-288, Parvimonas.sp.COT-101, PeptostreptococcaceaeXI[G-1].sp.COT-006, PeptostreptococcaceaeXI[G-1].sp.COT-071, PeptostreptococcaceaeXI[G-3].sp.COT-034, PeptostreptococcaceaeXI[G-3].sp.COT-307, PeptostreptococcaceaeXIII[G-2].sp.COT-077, Porphyromonadaceae[G-1].sp.COT-184, Porphyromonas.sp.COT-290, Prevotella.sp.COT-282, Prevotella.sp.COT-298, Prophyromonas.sp.COT-239, Proprionibacterium.sp.COT-296, Proprionibacterium.sp.COT-321, Proprionibacterium.sp.COT-365, Spirochaeta.sp.COT-314, Spirochaeta.sp.COT-379, Streptococcus.sp.cluster2789, Capnocytophaga.canimorus.COT-235, Chloroflexi[G-1].sp.COT-306, Desulfovibrio.sp.COT-070, Erysipelotrichaceae[G-3].sp.COT-302, Filifactor.sp.COT-064, Filifactor.sp.COT-163, Fusobacterium.sp.COT-236, g.bacterium.cp04.17, g.Parvimonas, g.Phascolarcto.bacterium, g.Solobacterium, g.Veillonellaceae.bacterium, Gemella.palaticanis.COT-089, Helcococcus.sp.COT-069, Helcococcus.sp.COT-140, LachnospiraceaeXIVa[G-6].sp.COT-161, Parvimonas.sp.COT-101, PeptostreptococcaceaeXI[G-1].sp.COT-004, PeptostreptococcaceaeXI[G-1].sp.COT-258, PeptostreptococcaceaeXI[G-3].sp.COT-104, PeptostreptococcaceaeXI[G4].sp.COT-019, PeptostreptococcaceaeXI[G-4].sp.COT-019, PeptostreptococcaceaeXI[G-6].sp.COT-067, PeptostreptococcaceaeXIII[G-1].sp.COT-030, PeptostreptococcaceaeXIII[G-2].sp.COT-077, Porphyromonas.crevioricanis.COT-253, Porphyromonas.macacae.COT-192, Porphyromonas.sp.COT-181, Porphyromonas.sp.COT-361, Prevotella.sp.COT-195, Proprionibacterium.sp.COT-296, Selenomonas.sputigena.COT-342, Streptococcus.minor.COT-116, Xenophilus.sp.COT-174, Bacteroidia[G-4].sp.COT-387, Bacteroides.denticanoris.COT-183(Prevotellasp?), Desulfomicrobium.orale.COT-008, Filifactor.villosus.COT-031, g.Moraxella, g.Peptostreptococcus, Porphyromonas.gulael.COT-052, Porphyromonas.gulaell.COT-052, Treponema.denticola.COT-197, Frigovirgula.sp.COT-058, Moraxella.sp.COT-017, g.Filifactor, g.Treponema, Bacteroides.heparinolyticus.COT-310, Bacteroides.sp.COT-040, Bacteroides.tectus.COT-039, Clostridiales[F-2.G-1].sp.COT-100P0005, ClostridialesIII[G-3].sp.COT-388, Neisseria.sp.COT-049, Neisseria.sp.COT-090, Neisseria.weaveri.COT-269, Neisseria.zoodegmatis.COT-349, Pasteurella.canis.COT-273, Pasteurellaceae.sp.COT-271, Pasteurelladogmatis.COT-092, Synergistales[G-1].sp.COT-180, Synergistales[G-1].sp.COT-244, Filifactor.sp.COT-064, Peptostreptococcus.sp.COT-033, Treponema.sp.COT-200, Treponema.sp.COT-351, Treponema.sp.COT-359, Treponema.sp.COT-198, Streptococcus anginosus COT-117, Peptostreptococcus_XI_sp COT-067, Frigovirgula_sp_COT-007, g. Odoribacter, in a sample taken from the animal and determining the oral health status of the animal.
Should the assay or method of the invention determine the likelihood that the oral health status is “P” (periodontosis/disease), the animal may be referred for a more thorough dental check to determine the next course of action.
If the assay determines that the oral health status is “not H” (i.e. not healthy), a further assay may be carried out to determine whether the status is Phot P. Alternatively or additionally, diet changes (as described above), dental chews (such as Dentastix®), tooth brushing or a dental check may be recommended for the animal. A further test in 1 month, 3 months, 6 months or 12 months may also be recommended.
If the assay determines the likelihood of ‘health’, a further assay may be recommended in 1 month, 3 months, 6 months or 12 months and for the oral health regime (e.g. diet, dental chews, tooth brushing) to be maintained.
The present invention also provides a method of determining the efficacy of an oral care product in a canine animal, the method comprising determining the oral health status of a canine animal using the assay of the invention; providing to the animal an oral care product for a certain period of time; and determining the oral health status of the animal using the assay of the invention after the period of time has elapsed.
Such a method enables the progress of the improvement of oral health to be monitored, without any invasive or risky procedures to the animal. More than one time point may be used, for example weekly, monthly, every two, three, four, five or six months, or annually. The amount of improvement of the oral health status may give an indication of the efficacy of the oral care product used in the method. Such oral care products or procedures may include foodstuffs, supplements, dental chews, tooth brushing, amongst others.
Such a method allows product efficacy to be tested in a scenario that more closely mimics the real life environment and would have the following impact on the ability to test oral care products. Larger scale trials could be performed, as the choice of animals would not be limited to those in research facilities. The access to animals outside of research facilities also allows longer term trials to be performed allowing the long term benefits of oral care products to be assessed. Products could be tested for their ability to remove plaque and reduce gingivitis rather than for only the ability to prevent the formation of plaque or development of gingivitis. The method also allows the measuring of prevention and the measuring of treatment. Comparison trials between products are able to be performed easily and with no detriment or stress to the animal.
The invention will now be described with reference to the following non-limiting figures and examples in which;
Subgingival plaque samples were collected from 223 dogs with healthy gingiva, gingivitis and mild periodontitis with 72 to 77 samples per health status. DNA was extracted from the plaque samples and subjected to PCR amplification of the V1-V3 region of the 16S rDNA. Pyrosequencing of the PCR amplicons identified a total of 274 operational taxonomic units (OTUs) after bioinformatic and statistical analysis. Porphyromonas was the most abundant genus in all disease stages, particularly in health along with Moraxella and Bergeyella. Peptostreptococcus, Actinomyces, and Peptostreptococcaceae were the most abundant genera in mild periodontitis. Logistic regression analysis identified species from each of these genera that were significantly associated with health, gingivitis or mild periodontitis. Principal component analysis showed distinct community profiles in health and disease. The species identified show some similarities with health and periodontal disease in humans but also major differences. In contrast to human, healthy canine plaque was found to be dominated by Gram negative bacterial species whereas Gram positive anaerobic species predominate in disease.
Sampling Strategy and Study Cohort
The study was approved by the WALTHAM® Centre for Pet Nutrition ethical review committee, and run under licensed authority in accordance with the UK Animals (Scientific Procedures) Act 1986. Owner consent was obtained and an owner survey was completed for all dogs included in the study.
Dental assessments, scoring and subgingival plaque sampling were performed by a single veterinary dentist (L. Milella) to avoid variation in scoring. The study cohort comprised client owned pet dogs presented at a veterinary referral dental clinic. Only dogs under anaesthetic for routine dental treatment or treatment for non-periodontal complications e.g. broken teeth or other non-infectious conditions were screened for inclusion in the study. No dogs were anaesthetised solely for the collection of plaque samples. The periodontal health status of each dog was obtained following the Wiggs & Lobprise scoring system (Wiggs & Lobprise, 1997) and plaque samples taken from dogs regarded as having healthy teeth and gums, gingivitis or mild periodontitis (PD1, <25% attachment loss). Dogs were excluded from the study if they had: 1) Significant veterinary oral care within the preceding 3 months; 2) Regular dental care at home i.e. dogs whose teeth are regularly brushed; 3) Systemic or oral antibiotic treatment any of the previous 3 months and 4) Evidence of any extra-oral bacterial infections in the past month. Breeds thought to exhibit an alternative early onset/aggressive form of periodontitis, were also excluded. These breeds were Greyhounds, Yorkshire Terriers, Maltese and Toy Poodles.
Sub-gingival plaque samples were collected using a sterile periodontal probe and placed in 350 μl TE buffer (50 mM Tris pH 7.6, 1 mM EDTA pH 8.0 & 0.5% Tween 20) prior to storage at −20° C.
Healthy dogs were sampled subgingivally at eighteen sites, targeting the teeth most often affected by PD (upper 103-108 bilaterally and lower 404, 408 and 409 bilaterally), to support plaque volumes in the absence of periodontal pockets. Periodontally diseased dogs were sampled for subgingival plaque at up to twelve diseased sites (103, 104, 108, 404, 408, 409 bilaterally) during their normal periodontal treatment. Information on dog age, breed, size, sex and neuter status was collated.
DNA Extraction & Amplification of 16S rDNA
DNA was extracted from the plaque samples using an Epicentre Masterpure Gram Positive DNA Purification Kit, according to the manufacturer's instructions with an additional overnight lysis. Plaque samples were centrifuged at 5000×g for 10 minutes and the cell pellet resuspended in 150 μl of TE buffer. Following vortexing, 1 μl Ready-Lyse Lysozyme (Epicentre, UK) was added and the lysis mix incubated overnight at 37° C. for 18 hrs overnight. After DNA extraction the DNA pellet was suspended in TE buffer (10 mM Tris-CI and 0.5 mM pH 9.0 EDTA) and quantified and the purity ascertained using a NanoDrop ND1000 spectrophotometer (NanoDrop Technologies Inc).
The V1-V3 region of the 165 rDNA was amplified from subgingival plaque DNA extractions using Extensor Hi-Fidelity PCR Enzyme Mix (AB-0792, Thermo, UK) in a 96-well format. A mix of two universal forward primers was used; FLX_27FYM (CGTATCGCCTCCCTCGCGCCATCAG AGAGTTTGATYMTGGCTCAG) at 9.5 pmol/μl and FLX_27F_Bif (CGTATCGCCTCCCTCGCGCCATCAG AGGGTTCGATTCTGGCTCAG) at 0.5 pmol/μl (where italics represent FLX Titanium Primer A and bold represents 165 sDNA primer sequence). The latter was included to ensure representation of the genus Bifidobacter, a lower concentration was chosen due to the low representation of this genus in previous studies of canine plaque. The DNA was to be sequenced from the reverse primer thus 20 different 7mer MID tags were included in the reverse primer (CTATGCGCCTTGCCAGCCCGCTCAGXXXXXXXTY ACCGCGGCTGCTGG) where italics represent FLX Titanium Primer B, X represents MID sequence and bold represents 16S sDNA reverse primer sequence.
Library Preparation
Library preparation and sequencing was performed by Beckman Coulter Genomics, UK. The 16S rDNA amplicons were purified with Agencourt AMPure XP beads (Beckman Coulter Inc, UK), quantified using the Quant-iT™ PicoGreen® dsDNA Assay Kit (Invitrogen, UK) then pooled into groups of 20 samples prior to Emulsion PCR. Libraries were then sequenced on a Roche Genome Sequencer FLX Titanium System™ using the FLX Titanium B primer only with a target of {tilde over ( )}15,000 unidirectional reads per sample.
Sequence Processing and Analysis
The standard flowgram files (SFF) for each of the 223 samples were initially filtered by selecting reads with at least 360 flows and truncating long reads to 720 flows. Reads were filtered and denoised using the AmpliconNoise software (version V1.21; Quince et al., 2009, 2011). For the initial filtering step, reads were truncated when flow signals dropped below 0.7, indicative of poor quality. A maximum of 20,000 reads per sample were used with exception of a few samples due to the computational demands of the denoising algorithm. Subsequently reads were denoised in three stages; 1) Pyronoise to remove noise from flowgrams resulting from 454 sequencing errors (PyronoiseM parameters −s 60, −c 0.01), 2) Seqnoise to remove errors resulting from PCR amplification (SeqNoiseM parameters −s 25, −c 0.08), 3) Perseus to detect and remove chimeras resulting from PCR recombination. The denoised sequences were then clustered using QIIME, a pipeline for performing microbial community analysis that integrates many third party tools which have become standard in the field. The QIIME script pick_otus.py, which utilises the Uclust software program, was used to cluster sequences with >98% identity. Uclust was run with modified parameters, with gap opening penalty set to 2.0 and gap extension penalty set to 1.0 and —A flag to ensure optimum alignment.
Representative sequences of all observed OTUs that passed the filtering criteria for sequence abundance (see statistical analysis section) across health states were searched against the Canine Oral Microbiome Database (COMD) using BLASTN of NCBI-BLAST 2.2.27+. The COMD sequence database contained 460 published 16S DNA sequences obtained from canine oral taxa (Genbank accession numbers JN713151-JN713566 & KF030193-KF030235). Additionally, representative sequences were searched against the 376,437 sequences in the Silva SSU database release 108. For each representative sequence the best BLAST hit in the COMD database was chosen as the reference sequence. If the alignment did not meet the cut-off criteria of 98% sequence identity and 98% sequence coverage the best hit from the Silva database was chosen. The assignments were checked for redundancies (two or more OTUs assigned to the same reference sequence). Redundancies were resolved by keeping the taxonomy for the OTU with the better match and assigning the next best match to the other OTU.
A multiple sequence alignment (MSA) was constructed by aligning each reference sequence to the Greengenes core set (revision May 3, 2011) with PyNAST using the script align_seqs.py of the Qiime pipeline. The MSA was filtered using the filter_alignment.py script of the Qiime pipeline. The MSA was converted to Phylip interleaved format using ClustalW 2.1. A maximum likelihood tree of 1000 bootstrap replicates was inferred with PhyML 3 revision 20120412. A GTR model of nucleotide substitution was chosen and the proportion of invariant sites was estimated from the data. Evolutionary rates were approximated by a discrete gamma model of eight categories. The tree was visualised and combined with abundance and significance data in iTOL.
A second tree with a reduced amount of taxa was inferred at the genus level. For this purpose all species of the same genus were collated into a single taxon. The 165 sequence of the most abundant species of that genus was used for tree inference using the methods described above. If no genus information was present, taxa forming a Glade in the full tree were grouped together and the new taxon was named e.g. “Actinomyces Glade A”. Abundance information was added up for all members of each summarised taxon and plotted on the tree using iTOL. The tree was complemented with information on the number of original taxa summarised and the number of significant taxa. See table 3 in which taxa were grouped together.
Statistical Analysis
Health and disease associations: The most abundant OTUs (>0.05%) were analysed using logistic regression analyses (Generalised linear model with a binomial distribution and logit link) for proportions, using the count for the OTU out of the total number of sequences and health status was included as a fixed effect. OTUs were classified in a single group of “rare” taxa if either they were present in each health status group at an average proportion below 0.05% or were present in less than two samples. The 0.05% cut-off was selected based on statistical analysis of data from mock communities containing 17 known species sequenced on five separate 454 runs. The mock communities were analysed for presence and absence of species using a false positive rate of 0.3% (i.e. finding species that were not included in the mock community) and false negative rate of 1.7% (i.e. the failure to identify the species that were known to be present) and aimed for a coefficient of variation of <20% (data not shown). As the data were of very low proportions {tilde over ( )}0.1%, a permutation test (1000 repeats) was used to test the robustness of the logistic regression analysis assumptions. The effect of health status for the true data was then compared to the effects from the permutations to give a more robust p-value for the overall effect of health status. The permutation p-values were adjusted according to the false discovery method of Benjamini and Hochberg (1995).
Principal component analysis (PCA) was performed on the log 10 (proportions+0.00003 to allow for zeros) to determine if variability of the most abundant OTUs (>0.05% of total reads) was associated with health status, gender, size and age.
Gram-stain status: The OTUs excluding rares were classified as gram positive or gram negative based on literature searches using the genus name. The number of sequences gram positive or gram negative were then analysed by logistic regression for proportions (adjusting for the total number of sequences and allowing for estimation of over dispersion) with health status as a fixed effect.
Oxygen Requirement: The non-rare OTUs were classified as aerobic, anaerobic or facultative anaerobe based on literature searches using the genus name. The number of sequences of aerobic, anaerobic and facultative anaerobe were then analysed (separately) by logistic regression for proportions (adjusting for the total number of sequences and allowing for estimation of over dispersion) with health status as a fixed effect. All health statuses were significantly different Shannon diversity Index: a linear model was used to analyse the data, with health status as a fixed effect and weighting the variability by health status as there were significant differences in the variability of indexes between health statuses.
Species richness: a linear model was used to analyse all OTUs including the rare sequences with health status as a fixed effect, the total number of sequences as a covariate (to adjust for the differing number of sequences between samples), and weighting the variability by health status (as there were significant differences in the variability of indexes between health statuses.
All analyses were performed in SIMCA-P version 10 (Umetrics).
Results
Study Cohort
Subgingival plaque bacterial communities were sampled from a total of 223 dogs; 72 with healthy gingiva, 77 with gingivitis and 74 with mild periodontitis. Dog size and age are putative risk factors for periodontitis and therefore sample associated metadata was also obtained (see Table 1). The majority of samples were collected from small, medium and large dogs with giant dogs represented at a much lower frequency. As expected the mean age of dogs increased with disease stage and significant differences (p<0.001) were observed in the mean ages of dogs in health compared to gingivitis and gingivitis compared to mild periodontitis using a two-tailed T-test with unequal variance. Lesser significance was observed in health versus gingivitis (p<0.05).
Sequence Quality
The 223 canine subgingival plaque samples were analysed by 454-pyrosequencing of the 3′ end of the V1-V3 region and a total of 6,071,129 sequence reads were obtained that passed the sequencing providers initial sequence quality filter. After Pyronoise, Seqnoise and chimera removal using Perseus the number of sequence reads was reduced to 3,110,837. The final number of sequence reads per sample ranged from 2,801 to 30,050 with a median number of reads of 11,682, 12,674 and 15,111 from healthy, gingivitis and mild periodontitis samples respectively.
Bacterial Composition of Canine Plaque
The resulting 3,110,837 sequences were assigned to 6,431 operational taxonomic units (OTUs) using U-Clust within QIIME and a cut-off of ≥98% sequence identity. OTUs were classed and grouped as rare if either they were present in each health status group at an average proportion below 0.05% or were present in less than two samples (see methods for rationale). This reduced the number of OTUs analysed to 274 plus the rare group.
Taxonomic assignment of each of the 274 OTUs resulted in 222 (81%) and 30 (11%) mapping to sequences within COMD (Dewhirst et al., 2012) and Silva respectively with ≥98% identity. The remaining 22 OTUs (8%) shared between 91.4% and 97.7% identity to sequences within the Silva database. The majority of the sequences belonged to seven phyla; Firmicutes (28.5%), Bacteroidetes (26.5%), Proteobacteria (17.36%), Actinobacteria (15.3%), Fusobacteria (3.7%), Spirochaetes (1.9%) and TM7 (1.1%). There were also a further five phyla identified; Synergistetes (0.9%), Chloroflexi (0.7%), SRI (0.4%), Tenericutes (0.09%) Elusimicrobia (0.04%) and a small proportion of the sequences were of unknown phylogeny (0.08%). The rare group accounted for the remaining 3.4% of the sequence reads.
The abundance of each of the 99 genera in plaque samples from healthy dogs and those with gingivitis and mild periodontitis are depicted by the green, orange and red outer bars respectively and the grey bar indicates the number of species (OTUs ≥98% sequence identity) in that genus. Of the 274 species identified, the 26 most abundant accounted for approximately 50% of the sequence reads (see Table 2). Porphyromonas cangingivalis COT-109 (OTU #2517) was the most abundant taxa representing 7.4% of the total number of sequence reads. Moraxella sp. COT-396 (4266) and Actinomyces canis COT-409 (OTU #6029) were the next most abundant representing 3.47% and 3.23% of the sequence reads respectively. Five other species each represented between 2% and 2.8% of the population; Bergeyella zoohelcum COT-186 (OTU #2232), Peptostreptococcus sp. COT_033 (OTU #6027), Peptostreptococcaceae sp. COT_004 (OTU #5570), Porphyromonas gulae COT-052 (OTU #2678), Porphyromonas gingivicanis COT-022 (OTU #5364). A further 18 OTUs represented between 0.85% and 2% of the population and the remaining 248 OTUs ranged in relative abundance from 0.01% to 0.81%
Porphyromonas cangingivalis COT-109
Moraxella sp. COT-396
Actinomyces canis COT-409
Bergeyella zoohelcum COT-186
Peptostreptococcus sp. COT-033
Peptostreptococcaceae sp. COT-004
Porphyromonas gulae COT-052
Porphyromonas gingivicanis COT-022
Filifactor villosus COT-031
Actinomyces sp. COT-083
Actinomyces sp. COT-252
Neisseria shayeganii COT-090
Fusobacterium sp. COT-189
Porphyromonas canoris COT-108
Porphyromonas gulae COT-052
Corynebacterium freiburgense COT-403
Peptostreptococcaceae sp. COT-077
Clostridiales sp. COT-028
Fusobacterium sp. COT-169
Pasteurellaceae sp. COT-080
Capnocytophaga sp. COT-339
Erysipelotrichaceae sp. COT-311
Peptostreptococcaceae sp. COT-135
Lachnospiraceae sp. COT-036
Moraxella sp. COT-018
Capnocytophaga cynodegmi COT-254
Associations with Health and Disease
Logistic regression analysis identified that 90 of the 274 OTUs had a statistically significant effect of health status after randomisation and multiplicity correction. Of these, 54 showed a significant difference between health and gingivitis, 73 showed a significant difference between gingivitis and mild periodontitis and 87 showed a significant difference between health and mild periodontitis (
Bacteroides sp. COT-040
Odoribacter denticanis COT-
Bergeyella zoohelcum COT-
Capnocytophaga canimorsus
Capnocytophaga cynodegmi
Capnocytophaga sp. COT-
Capnocytophaga sp. COT-
Cloacibacterium sp. COT-320
Chloroflexi bacterium COT-
Helcococcus sp.
Peptococcus sp. COT-044
Filifactor alocis COT-001
Filifactor sp. COT-164
Clostridiales bacterium
Clostridiales bacterium
Clostridiales bacterium
Schwartzia sp. COT-063
Lautropia sp. COT-175
Brachymonas sp. COT-015
Comamonas sp. COT-270
Neisseria shayeganii
Cardiobacterium sp.
Moraxella sp. COT-396
Pasteurella canis COT-273
Treponema denticola
Synergistales bacterium
Synergistales bacterium
Synergistales bacterium
Actinomyces suimastitidis
Porphyromonas cangingivalis
Porphyromonas sp. COT-290
Saccharofermentans
Conchiformibius steedae
Synergistales bacterium
Capnocytophaga sp.
Catonella sp. COT-257
Synergistales bacterium
Prevotella sp. COT-372
Schwartzia sp. COT-213
Treponema sp. COT-351
Bacteroidia bacterium
Parvimonas sp. COT-101
Lautropia sp. COT-060
Pasteurella dagmatis COT-
Corynebacterium mustelae
Corynebacterium canis
Campylobacter sp. COT-011
Escherichia coli COT-277
Clostridiales bacterium
Acholeplasmatales bacterium
Globicatella sp. COT-107
Parvimonas sp. COT-035
Moraxella sp. COT-328
Neisseria canis AY426973
Desulfovibrionales
bacterium COT-009
Leptotrichia sp. COT-345
Aquaspirillum sp. COT-091
Stenotrophomonas sp.
Actinomyces hyovaginalis
Clostridiales bacterium
Of the most abundant health associated species, Moraxella sp. COT 396 (QIIME OTU #4266, 6.61%), Bergeyella zoohelcum COT-186 (OTU #2232, 5.48%), Neisseria shayeganii COT-090 (OTU #2233, 3.28%) and Pasteurellaceae sp. COT-080 (OTU #1678, 2.37%) were significantly more abundant in health and gingivitis than in mild periodontitis and were also significantly more abundant in health than gingivitis (See table 3). Capnocytophaga sp. COT-339 (OTU #4929, 2.96%), Stenotrophomonas sp. COT-224 (OTU #564, 1.1%) were also significantly more abundant in health than in gingivitis and mild periodontitis but the relative abundance in gingivitis and mild periodontitis were not significantly different. Again, Porphyromonas cangingivalis COT-109 (OTU #2517, 10.47%) and Capnocytophaga cynodegmi COT-254 (OTU #5514, 1.24%) were significantly more abundant in health and gingivitis than in mild periodontitis but the relative abundance in health and gingivitis did not significantly differ.
The most abundant disease associated species included Peptostreptococcaceae sp. COT-004 (OTU #5570, 3.5%) and Lachnospiraceae sp. COT-036 (OTU #6025, 1.7%) which were significantly more abundant in mild periodontitis than health and gingivitis and did not significantly differ in their relative abundance in health and gingivitis. Clostridiales sp. COT-028 (OTU #1916, 2.2%), Peptostreptococcaceae sp. COT-135 (OTU #368, 1.4%), Peptostreptococcaceae sp. COT-077 (OTU #2463, 2.7%), Peptococcus sp. COT-044 (OTU #4905, 1.2%), Peptostreptococcaceae sp. COT-019 (OTU #908, 1.2%) and Peptostreptococcaceae sp. COT-030 (OTU #4774, 1.1%) were also significantly more abundant in mild periodontitis than in health and gingivitis but were also more abundant in gingivitis than health. Corynebacterium canis COT-421 (OTU #149, 1.1%) was significantly more abundant in mild periodontitis and gingivitis than health but gingivitis and mild periodontitis samples were not significantly different.
Principal component analysis was used to investigate correlations of OTUs with health status, age, size and gender. The first component explained 14.7% and the second component 9.5% of the variability in the OTU log10 proportions (see
Gram-Stain Status and Oxygen Requirements
The probable Gram-stain status was determined by literature searches followed by logistic regression analysis of proportions of Gram positive or Gram negative non-rare OTUs; this showed that health, gingivitis and mild periodontitis groups were significantly different. Samples from dogs with mild periodontitis had a significantly higher proportion of Gram positive OTUs than those from dogs with gingivitis (P<0.001) and healthy gingiva (P<0.001). Gingivitis samples had a significantly higher proportion of Gram positive OTUs than samples from the healthy group (P=0.003; see
The probable oxygen requirements were also determined by literature searches and analysed by logistic regression for proportions of aerobes, facultative anaerobes and anaerobes. Clear differences in oxygen requirements were observed between the bacterial population in healthy, gingivitis and mild periodontitis samples. Samples from dogs with healthy gingiva had significantly higher proportions of aerobes than gingivitis and periodontitis samples (P=0.006 & P<0.001 respectively) and gingivitis samples had a significantly higher proportion of aerobes than samples from dogs with mild periodontitis (P<0.001; see
Species Richness and Diversity
A linear model was used to compare the number of operational taxonomic units (OTUs) including rare OTUs in health, gingivitis and mild periodontitis. This showed that all health status were significantly different (see
A linear model was used to analyse the Shannon diversity index data and showed that all health status were significantly different (see
As a whole the most predominant phyla observed were the Bacteroidetes, Proteobacteria, Firmicutes and Actinobacteria from which 26 of the most abundant species made up 49.8% of all sequences. The proportions of these phyla shifted between disease stages with the Proteobacteria and Bacteroidetes being most abundant in plaque from the healthy cohort and the Firmicutes being more abundant in the mild periodontitis cohort. Comparisons with the human oral microbiota become most striking at the genus & species level. Whilst Streptococcus spp. are abundant in healthy humans they are rare in dogs; the lack of cariogenic Streptococcus spp. is presumably the reason dental caries is a rarely observed disease in dogs. Of note in this respect is the pH of canine saliva (pH 8.5), which is considerably more alkaline than that of human saliva (pH 6.5 to 7.5). It is possible that this difference in pH contributes to the lack of Streptococci in the dog oral cavity along with the lower level of sugars in the diet. The latter would be consistent with the recent observation that the human oral microflora evolved to a more cariogenic nature following the introduction of processed sugars to the diet during the industrial revolution (Adler et al. 2013). In healthy dogs Porphyromonas cangingivalis Canine Oral Taxon (COT)-109, Moraxella sp. COT-396 and Bergeyella zoohelcum COT-186 were the most abundant species. The latter two are also abundant in human health but the abundance of a Porphyromonad in healthy dogs is in contrast to the human oral microbiome where P. gingivalis has been synonymous with the red complex and human periodontal disease. The abundance of Porphyromonas, Moraxella and Bergeyella in healthy dogs was also observed in a recent 454 pyrosequencing study of 6 dogs.
With respect to canine periodontal disease the Actinomyces, Peptostreptococcaceae and Porphyromonas species predominated. The most abundant species being P. cangingivalis COT-109 (again), Peptostreptococcus sp. COT-033, Actinomyces sp. COT-374, Peptostreptococcaceae XI [G-1] sp. COT-004 and Peptostreptococcaceae XIII [G-2] sp. COT-077. Fusobacterium and Treponema spp associated with human periodontal disease where present but at lower abundance and only one Treponema spp. (T. denticola) was significantly associated with mild periodontitis (Griffen et al., 2012). This difference in the apparent importance of Treponemes in the disease state may be as a result of an earlier stage of Periodontitis being surveyed in this study than the human one. It is also accentuated by the large number of different Treponeme species identified in our analysis (16) leading to fragmentation of the abundance. Indeed, if grouped at the genus level the Treponeme species make up 2.15% of the total in disease.
Relatively few species were associated solely with gingivitis (Leptotrichia sp. COT-345, Neisseria canis AY426973 and an uncultured Capnocytophaga sp. HM333068). The abundance of health associated species did not always follow a linear reduction in abundance in gingivitis through to PD1, for many their abundance was also relatively high in gingivitis. This was also true for mild periodontitis associated species making it challenging to differentiate a health/gingivitis associated species from a health associated species or a gingivitis/periodontitis associated species from a periodontitis associated species. Presumably certain health associated species can compete in the gingivitis environment but not in periodontitis and vice versa for periodontitis associated species.
In human plaque Gram-positive bacteria have traditionally been regarded as health associated and anaerobic Gram-negative bacteria as disease associated. Griffen's recent survey noted that this may be an over simplification with at least one Gram-positive bacteria (Filifactor alocis) being abundant in human disease (Griffen et al., 2012). Our observations in dog are in contrast to those from the human oral microbiome with Gram negative species being significantly more abundant in healthy plaque samples and Gram positives significantly more abundant in periodontitis plaque samples. The lack of Streptococci in dog results in the health associated species being dominated by Gram-negative aerobes. In contrast to health, the abundance of periodontitis associated Firmicutes, particularly Peptostreptococcaceae spp., means that Gram-positive anaerobes predominate in the periodontitis associated species. The environmental pressures that drive selection of species presumably include nutrient sources, oxygen stress, pH and immunological factors. We hypothesise that the major health associates may be aerobic early colonisers that are able to metabolise salivary carbohydrates such as mucins and proline rich proteins. With chronic gingivitis and periodontitis, uncontrolled inflammation and bacterial activity result in the destruction of gingival tissue leading to anaerobic periodontal pockets containing protein rich gingival crevicular fluid and immunological agents. This may then drive the rise in abundance of proteolytic anaerobic Clostridiales and Peptostreptococcaceae and Porphyromonads known for their ability to resist host defenses and utilise host oxidative immune responses (Mydel et al., 2006). The ability of the Gram negative anaerobe Porphyromonas cangingivalis to predominate in all three health states suggests that it is both metabolically flexible enough to colonise in health and able to compete against other Porphyromonas spp. in the disease environment. Its ability to prosper in health which is traditionally considered to be a more aerobic environment is interesting given that Porphyromonads are strict anaerobes.
The data set used consisted of 454 sequencing data from 223 dog (H=72, G=77 & PD1=74) sub-gingival plaque samples pooled from multiple teeth of the same health state from the same individual.
Health State
The primary output of interest was the classification of cases into one of three health categories. These were:
In the predictive modelling process, in addition to this 3-way classification, several alternative outputs were also investigated for the purpose of simplifying the models, for understanding which bacterial markers were indicative of each classification and for possible use in ‘two-stage’ predictive models. The primary set of outputs used was:
An output of secondary interest was the average Gingivitis score obtained from sampled teeth. Although the individual teeth receive a discrete score taking the values 0, 1, 2 or 3, the average score is a continuous variable ranging from 0 to 3, which has an impact on the types of predictive models used and the approach to assessing the performance of the models. The Test Set Performance was assessed in terms of the regression R-squared value (the percentage of the variability in the output variable explained by the model). Since the R-squared can be difficult to interpret, the Residual Standard Deviation, which provides an indication of the mean distance between the observed and predicted values, was also calculated.
In addition to the actual (continuous) mean Gingivitis score, the score rounded to the nearest integer value was also used as an output in some models. This was treated as purely categorical, which allows the performance to be assessed in the same way as for the Health State models. Since the initial results of this exercise were not very promising, no further optimisation of these models was performed.
Inputs/Predictors
Several sets of predictors were used in the predictive modelling. The following predictor variants were used:
Counts: These are the original count variables, comprising prevalence counts for individual bacterial species and genera.
Proportions: These are the counts transformed to proportions by dividing by the total count for each animal (‘row-wise’ transformation).
Binary (count>1): These are binary presence/absence scores obtained using counts less than or equal to one to define “absent” and counts greater than one as “present”.
Binary (count>7): These are binary presence/absence scores obtained using counts less than or equal to seven to define “absent” and counts greater than seven as “present”.
Relative Counts & Proportions: These are obtained by dividing the counts and proportions by the mean value for the whole predictor (‘column-wise’ transformation).
Relative Counts & Proportions Truncated: These are obtained by taking the Relative values and “truncating” any values greater than 1 by setting them equal to 1. This transformation results in values ranging from 0 to 1 (with, typically, a large number of truncated values).
Relative Counts & Proportions 3-level Categorical: These are categorical variables obtained by taking the Relative values and setting them into three categories using threshold values of 0.0005 (0.05% of the mean) and 0.1 (10% of the mean). These were then used as categorical predictors in the models.
Based on the results from the above separate modelling exercises, a further set of models was fitted using a subset of predictors identified as those selected by the earlier models. These are listed in table 4. These predictors were drawn from the Count, Proportion and Binary predictor sets, but omitted the Relative predictors. This allowed the models to select any combination of these predictors. For a subset of output variables, a further set of 3-level categorical predictors (those selected by the earlier models) were added to this set to allow the creation of models that included a combination of continuous/binary and categorical predictors.
Table 4—Subset of ‘Best’ Predictors
This list shows the subset of ‘best’ predictor variables identified from the early-stage modelling and used in later stages to allow the creation of models including more than one predictor type. The variable prefixes identify the variable type and should be interpreted as follows:
Predictive Modelling
Standard Model Types
A standard set of models (as known to the person skilled in the art) was applied to each combination of predictors and output variables. The classification model types used were:
The model types used for the prediction of Gingivitis score (a continuous output variable) were:
Because of the limited success of the attempts a two-stage approach to produce models capable of performing a (H/G/P) classification was investigated. This involved developing a set of models trained for simpler two-way classification and combining some of the best-performing two-way models to produce a final three-way (H/G/P) classification, in the following combinations:
These results provide evidence that it is possible to use the bacterial species found in canine sub-gingival plaque to diagnose oral health state.
Misclassification Matrix
Classification Functions
Misclassification Matrix
Tree Structure
Misclassification Matrix
Classification Functions
Misclassification Matrix
Tree Structure
Misclassification Matrix for 2-stage Model A
This example shows the identification of two or more bacteria as claimed using a binary test, combined with the information give n in tables 1 to 5 and the use of statistical models gives a reliable prediction of the health status of an animal.
This example shows the identification of two or more bacteria as claimed using a proportional test, combined with the information give n in tables 1 to 5 and the use of statistical models gives a reliable prediction of the health status of an animal.
This example shows that predictive models based on known bacteria are improved once at least one novel bacteria identified as part of the present invention is included in the analysis.
Misclassification Matrix
Classification Functions
Misclassification Matrix
Classification Functions
Misclassification Matrix
Tree Structure
Supplementary Information
PMML Files for Selected Models
The following PMML files show the deployment code for each of models 1 to 4.
| Number | Date | Country | Kind |
|---|---|---|---|
| 1310691 | Jun 2013 | GB | national |
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/GB2014/000234 | 6/13/2014 | WO | 00 |
| Publishing Document | Publishing Date | Country | Kind |
|---|---|---|---|
| WO2014/199115 | 12/18/2014 | WO | A |
| Number | Name | Date | Kind |
|---|---|---|---|
| 1466578 | Isabel | Aug 1923 | A |
| 5407661 | Simone | Apr 1995 | A |
| 5513663 | Van et al. | May 1996 | A |
| 6197305 | Friedman et al. | Mar 2001 | B1 |
| 6414036 | Ninkov | Jul 2002 | B1 |
| 6495176 | McGenity et al. | Dec 2002 | B1 |
| 6574504 | Mazaury et al. | Jun 2003 | B1 |
| 6623767 | Morice | Sep 2003 | B1 |
| 6652892 | McGenity et al. | Nov 2003 | B2 |
| 7691418 | Rossel | Apr 2010 | B2 |
| 7879377 | Dahl et al. | Feb 2011 | B2 |
| 7910139 | Bombardelli | Mar 2011 | B2 |
| 20030228400 | Dahl et al. | Dec 2003 | A1 |
| 20070134402 | Feder | Jun 2007 | A1 |
| 20080026083 | Reynolds | Jan 2008 | A1 |
| 20100061944 | Marshall-Jones et al. | Mar 2010 | A1 |
| 20110300197 | McGenity | Dec 2011 | A1 |
| 20130330681 | Sacks | Dec 2013 | A1 |
| Number | Date | Country |
|---|---|---|
| 2461499 | Feb 1981 | FR |
| 2622453 | May 1989 | FR |
| H08109118 | Apr 1996 | JP |
| 20030074055 | Sep 2003 | KR |
| Entry |
|---|
| What's Living in your Mouth? Cascades Center For Dental Health—Accessed Dec. 11, 2018 at https://cascadesdental.com/whats-living-in-your-mouth/ (provided by Applicant's Representative, attached to interview agenda). |
| Culham et al., Journal of Veterinary Dentistry, vol. 15, No. 4, pp. 165-168 (1998). |
| Elliott et al., Journal of Clinical Microbiology, vol. 43, No. 11; 2005 (of record). |
| Dewhirst et al., PlosOne, vol. 7, No. 4, e36067, pp. 1-12; published Apr. 27, 2012 (of record). |
| Culham et al., Journal of Veterinary Dentistry, vol. 15, No. 4, pp. 165-168 (1998) (of record). |
| Riggio et al., Veterinary Microbiology, vol. 150, pp. 394-400 (2011) (Year: 2011). |
| Dewhirst et al., PlosOne, vol. 7, No. 4, e36067, pp. 1-12; published Apr. 27, 2012 (of record). (Year: 2012). |
| Culham et al., Journal of Veterinary Dentistry, vol. 15, No. 4, pp. 165-168 (1998) (of record). (Year: 1998). |
| Sturgeon et al., Veterinary Microbiology, vol. 162, pp. 891-898 (electronically published Nov. 20, 2012) (Year: 2012). |
| Riggio et al., Veterinary Microbiology, vol. 150, pp. 394-400 (2011) (of record). (Year: 2011). |
| Elliott et al., Journal of Clinical Microbiology, vol. 43, No. 11; 2005 (of record). (Year: 2006). |
| Sturgeon et al., Veterinary Microbiology, vol. 162, pp. 891-898 (Nov. 20, 2012) (of record). (Year: 2012). |
| D'Aiuto, et al., “Periodontitis: from local infection to systemic diseases”, Int J Immunopathol Pharmacol. 18(3 Suppl), Jul.-Sep. 1-11, 2005. |
| Dewhirst, et al., “The Canine Oral Microbiome”, Plos One, vol. 7, No. 4, Apr. 19, 2012, e36067. |
| Elliott, et al., “Cultivable oral microbiota of domestic dogs”, Journal of Clinical Microbiology, American Society for Microbiology, US, Nov. 1, 2005, 5470-5476. |
| Kortegaard, et al., “Periodontal disease in research beagle dogs—an epidemiological study: Paper”, Journal of Small Animal Practice, 49, 2008, 610-616. |
| Kyllar, et al., “Prevalence of dental disorders in pet dogs”, 2005, 496-505. |
| Lawson, et al., “Peptostreptococcus canis sp. nov., isolated from subgingival plaque from canine oral cavity”, Anaerobe, vol. 18, No. 6, Dec. 1, 2012, 597-601. |
| Sturgeon, et al., “Metagenomic analysis of the canine oral cavity as revealed by high-throughput pyrosequencing of the 165 rRNA gene”, Veterinary Microbiology, vol. 162, No. 2-4, Mar. 1, 2003, 891-898. |
| “Periodontal (Gum) Disease, Causes, Symptoms, and Treatments”, US Department of Health and Human Services, Jan. 2006, pp. 1-16. |
| Appendino, et al., “Antibacterial galloylated alkylphloroglucinol glucosides from myrtle (Myrtus communis)”, Medicinal & Aromatic Plants Abstracts, National Institute of Science Communication and Information, vol. 28, No. 3, Jun. 2006, XP018018747, 5 pgs. |
| Aydin, et al., “Determination of nutritional and physical properties of myrtle (Myrtus communis L.) fruits growing in Turkey”, Journal of Food Engineering, vol. 79, No. 2, Nov. 1, 2006, pp. 453-458; XP005844950. |
| Casey, et al., “Development of a robust microtitre plate-based assay method for assessment of bioactivity”, Microbial methods 58 (Sep. 2004) 327-334 (Abstract only, 2 pgs). |
| Dard-E-Dandaan, “Qaraabaadeen Azam wa Akmal (20th Gentry AD)”, Matba Siddiqi, Delhi, Matba Mustafai, Delhi, 1909 AD, p. 690 (accessed Dec. 29, 2012). |
| Fenno et al., “The opdB Locus Encodes the Trypsin-Like Peptidase Activity of Treponema denticola Infection and Immunity”, Infection and Immunity, Oct. 2001, p. 6193-6200, vol. 69, No. 10 (11 pgs). |
| Hennet, et al., “Evaluation of the Logan & Boyce Plaque Index for the Study of Dental Plaque Accumulation in Dogs”, Res. Vet. Sci., Apr. 2006, 80(2):175-180. |
| Kuhn, et al., “Comparison of Biofilms Formed by Candidaalbicans and Candidaparapsilosis on Bioprosthetic Surfaces”, Infection and Immunity. 70 (2): 878-888, Feb. 2002 (20 pgs). |
| McEldowney, et al., “Variability of the Influence of Physioichemical Factors Affecting Bacterial Adhesion ot Polystyrene Substrata”, Appl Environ Microbiol, 52:460-465, Jan. 1986. |
| Murad, Khazaain-al-Advia, vol. 1 (20th century AD), Nadeem Yunus Printer, Sheikh Mohd Basheer Sons, Lahore 1911 AD, p. 371 (Accessed online Dec. 29, 2012). |
| Pratten, et al., “Response of a single species biofilms and microcosm dental plaques to pulsing with chlorhexidine”, Journal of Antimicrobial Chemotherapy 42 453-459, Oct. 1998. |
| Romani, et al., “Identification and quantitation of polyphenols in leaves of Myrtus communis L”, Chromatographia, Wiesbaden DE, vol. 49, No. 1-2, Jan. 1, 1999 pp. 17-20. |
| Allaker et al., Prevalence of Porphyromonas and Prevotella species in the dental plaque of dogs, The Veterinary Record, (1997), 140, 147-148. |
| Davis et al., A Cross-Sectional Survey of Bacterial Species in Plaque from Client Owned Dogs with Healthy Gingiva, Gingivitis or Mild Periodontitis, PLOSOne, vol. 8, issue 12, e83158, Dec. 2013, 12 pages. |
| Isogai et al., Ecology of Genus Porphyromonas in Canine Periodontal Disease, J. Vet. Med., Series B, vol. 46, No. 7, (1999), 476-483. |
| McDonald et al., “Characterising the Canine Oral Microbiome by Direct Sequencing of Reverse-Transcribed rRNA Molecules,” PLOS One, vol. 11, issue 6, Jun. 2016, e0157046, 17 pages. |
| Number | Date | Country | |
|---|---|---|---|
| 20160138089 A1 | May 2016 | US |
