World Workshop on Oral Medicine VII: Targeting the oral microbiome Part 2: Current knowledge on malignant and potentially malignant oral disorders.

OBJECTIVE
The World Workshop on Oral Medicine VII chose the oral microbiome as a focus area. Part 1 presents the methodological state of the science for oral microbiome studies. Part 2 was guided by the question: What is currently known about the microbiome associated with oral squamous cell carcinoma and potentially malignant disorders of the oral mucosa?


MATERIALS AND METHODS
A scoping review methodology was followed to identify and analyse relevant studies on the composition and potential functions of the oral microbiota using high-throughput sequencing techniques. The authors performed searches in PubMed and EMBASE. After removal of duplicates, a total of 239 potentially studies were identified.


RESULTS
Twenty-three studies on oral squamous cell carcinoma, two on oral leukoplakia and four on oral lichen planus were included with substantial differences in diagnostic criteria, sample type, region sequenced and sequencing method utilised. The majority of studies focused on bacterial identification and recorded statistically significant differences in the oral microbiota associated with health and disease. However, even when comparing studies of similar methodology, the microbial differences between health and disease varied considerably. No consensus on the composition of the microbiomes associated with these conditions on genus and species level could be obtained. Six studies on oral squamous cell carcinoma had included in silico predicted microbial functions (genes and/or pathways) and found some similarities between the studies.


CONCLUSIONS
Attempts to reveal the microbiome associated with oral mucosal diseases are still in its infancy, and the studies demonstrate significant clinical and methodological heterogeneity across disease categories. The immense richness and diversity of the microbiota clearly illustrate that there is a need for additional methodologically comparable studies utilising deep sequencing approaches in significant cohorts of subjects together with functional analyses. Our hope is that following the recipe as outlined in our preceding companion paper, that is Part 1, will enhance achieving this in the future and elucidate the role of the oral microbiome in oral squamous cell carcinoma and potentially malignant disorders of the oral mucosa.

The aim of this scoping review was to present and critically evaluate the published literature on sequence-based determination of the oral microbiome and its potential functions in association with OSCC and OPMDs. Together with Part 1 of this review, this report is designed to provide encouragement for oral medicine specialists to explore this exciting field.

| Research question
This review was guided by the question, "What is currently known about the association between the oral microbiome and OSCC and OPMDs, as determined by non-specific sequence-based identification?" Identifying the microbiota must be accompanied by investigations on properties of the microorganisms involved, and it is in the combination of these approaches that potential aetiopathogenic candidates can be selected and further tested in hypothesis-driven investigations. Therefore, we employed a scoping review methodology, including studies both on identification and on functional properties in order to map the relevant literature, identify key concepts and gaps in research, and trace a path for future research (Daudt, van Mossel, & Scott, 2013). The format for this scoping review is based on the framework originally outlined by Arksey and O'Malley (2005) and further refined by Levac, Colquhoun, and O'Brien (2010).
Our framework includes the following five key phases: (a) identifying the research question, (b) identifying relevant studies, (c) study selection, (d) data characterisation and (e) data summary and synthesis.

| Identifying relevant studies
We included all studies published until October 2018 that aimed to determine the oral microbial composition (bacteriome, mycobiome and virome), as well as metagenomic, transcriptomic, and microbial proteomic and metabolomic studies of the microbiota from patients with OSCC and OPMDs. The oral mucosal diseases included in our search were as follows: OSCC, oral leukoplakia, oral erythroplakia, Oral dysplasia (oral dysplasia OR oropharyngeal dysplasia OR ((("mouth"[MeSH Terms] OR "mouth"[All Fields] OR "oral"[All Fields]) AND ("epithelial"[All Fields] AND "dysplasia"[All Fields]) OR "epithelial dysplasia"[All Fields])))) OR ((("oropharynx"[MeSH Terms] OR "oropharynx"[All Fields] OR "oropharyngeal"[All Fields]) AND ("epithelial"[All Fields] AND "dysplasia"[All Fields]) OR"epithelial dysplasia"[All Fields]))) AND English [lang] oral lichen planus, palatal lesions in reverse smokers, actinic cheilitis, oral discoid lupus erythematosus and oral submucous fibrosis. Oral verrucous carcinoma and oral epithelial dysplasia, a histopathological diagnosis associated with an increased risk of malignant transformation, were also included in our searches. Initial literature searches revealed very few studies, and for this reason, we decided not to exclude papers based on poorly described diagnostic criteria.
Since the purpose of this review was to reveal what is known about associations of the oral microbiome with respect to OSCC and OPMDs based on non-specific sequencing, we excluded studies that only utilised culture methods and those that identified a limited number of species. Likewise, we excluded studies focusing on a single or a few microbial properties as well as metabolomic studies that did not allow differentiation between human and microbial metabolites.

| Study selection
We searched the electronic databases PubMed and EMBASE for peer-reviewed papers using a combination of both MeSH terms (translated into Emtree terms for EMBASE search) and free-text forms (Table 1). Individual search strings were constructed for each of the above-mentioned oral mucosal diseases and for oral epithelial dysplasia, using synonyms and derivatives, except for leukoplakia and erythroplakia that were combined in one search, because preliminary searches retrieved the same papers for both diseases.
To focus the searches on the microbiota, the search strings for the mucosal conditions were combined with a microbiome search string using search terms related to identification by non-specific sequencing and microbial properties ( Table 1). The term "oral" was not included because comparison between searches with and without this term had shown more papers when "oral" was omitted.

| Data characterisation
All citations deemed relevant after title and abstract screening were procured for subsequent review of the full-text article. A form was developed to confirm relevance and to extract study characteristics such as publication year, country, number of subjects, sample type and technology. In addition, for identification studies, genes sequenced, number of clones/reads identified, taxonomic detection level, main microbial results and study limitations were recorded, whereas for properties studies, type of study, main results and study limitations were recorded (Table S1). The included papers were subdivided into "identification studies" for the articles dealing with identification of the microbiota (bacteriome, mycobiome and virome) and "properties studies" for the articles related to metagenomics, transcriptomics, and microbial proteomics and metabolomics. Data from identification studies were compiled into a Microsoft Excel for Mac (Microsoft Corporation) spreadsheet. Data from properties studies were extracted in essay form and incorporated in the results section. whole-genome sequences of the species identified were retrieved from public databases and annotated genes compared to give a picture of potential properties.  Table 3.

| The microbiota associated with oral squamous cell carcinoma
Twenty-three non-specific sequence-based identification studies of the microbiota were included, 14 of which were published in 2017-2018 (Table 2)   Sanger sequencing identified a very limited number of clones due to methodological limitations, whereas those using second-generation sequencing ranged from a few thousand to over 100,000 filtered reads per sample. Studies including a large number of subjects and clones/filtered reads per sample tended to report a higher number of significant differences in the microbiome between health and disease as compared to those studies including a few subjects and clones/filtered reads per sample (Table 3). All studies but one (Bebek et al., 2012) could identify to genus level and many to species level.
Finally, the studies had been performed in geographically widespread locations.
In the clear majority of bacteriome studies, statistically significant differences were found between health and disease ( Table 3).
The reader is referred to the individual papers for detailed information on the statistics employed and to the guidelines on statistical analysis of microbial composition data outlined by Odintsova, Tyakht, and Alexeev (2017).
Of the studies using tissue/swab samples, the one from Bebek et al. (2012) was omitted from the comparison, as reads were identified to phylum and family level only. At the genus level, Parvimonas was found to be significantly increased in tissue/swab samples in the study by Wang et al. (2017) Table 3).
The 14 studies that aimed to identify the bacterial part of the microbiome in OSCC were also divided into two groups based on sample type: tissue/swab samples in one group and saliva/oral rinse/mouthwash samples in another group. The genera that were significantly increased in OSCC tissue/swab samples ranged in number from 3 to 23 (Table 3) Six studies that analysed tissue/swab samples included identification to species level, in which cases the diversity was even more pronounced. Among the most abundant species in these studies, Fusobacterium nucleatum was the only species that was reported to be significantly enriched in OSCC in more than one study (Al-Hebshi et al., 2015Zhao et al., 2017). All other species were only found to be enriched in a single study. Comparisons of saliva samples showed similar results.
Only two mycobiome studies of OSCC, both using tissue samples, were retrieved (Mukherjee et al., 2017;Perera et al., 2017). Perera et al. (2017) reported three fungal species to be significantly elevated in OSCC compared to control patients with fibroepithelial polyps. Mukherjee et al. (2017), who detected to genus level only, found two genera enriched in OSCC compared to healthy sites.
None of these genera were shared with the genera reported by Perera et al. (2017).
The single virome study found no virus to be specifically associated with tongue squamous cell carcinoma samples, including highrisk HPV .
The two studies describing the microbiome associated with oral leukoplakia differed in sample type (swab and saliva), which hampers a reliable comparison. The results of these studies differed in the genera found to be associated with OSCC; Haemophilus in the study by Hu et al. (2016) and Fusobacterium, Leptotrichia and Campylobacter in the study by Amer et al. (2017) were found to be significantly enriched in oral leukoplakia (Table 3).
All four studies on oral lichen planus used different sample types for elucidation of the bacteriome. One study ) did not include a control group, but described also the mycobiome. No single genus or species among the predominant taxa was significantly elevated in oral lichen planus in two or more of the studies including a control group (Table 3).

| Potential contribution of the microbiome in oral carcinogenesis
Six studies identifying the bacterial taxa associated with OSCC by and Al-Hebshi et al. (2017) found that the genes associated with the synthesis of three amino acids were decreased, although they were not the same amino acids. Zhao et al. (2017) found that pathways related to amino acid metabolism were significantly decreased in OSCC, while Yang, Yeh et al. (2018) reported a decrease in amino acid metabolism with increased OSCC staging. There were also, however, conflicting results: Perera et al. (2018) found that genes encoding membrane molecules were enriched in OSCC tissue samples, whereas Zhao et al. (2017) found that pathways related to membrane transport were decreased in swabs from OSCC. In addition, genes associated with glycolysis/glycogenesis were more abundant in the control groups of two studies Perera et al., 2018), whereas Yang, Yeh et al. (2018) found carbohydrate-related metabolism to increase with OSCC staging. The use of uniform diagnostic criteria and definitions of the diseases were not considered as an inclusion criterion due to the few studies identified in pilot searches. Although the tissue samples analysed in the included studies were histopathologically verified, the diagnostic criteria for HNSCC/OSCC and leukoplakia were poorly described. In the case of oral lichen planus, the majority of the studies used the diagnostic criteria outlined by the WHO (Kramer, Lucas, Pindborg, & Sobin, 1978). This is a potential source of variations in microbiological results alongside the methodological variables related to the microbiological analyses.

| D ISCUSS I ON
The OSCC studies clearly demonstrate the influence of the many methodological variables on the identified microbiome, as  (Table S1) identification to genus level. This may provide an overall insight into the potential differences between health and disease. However, species-level or even subspecies-level identification is preferred when seeking aetiopathogenic microorganisms. Recent versions of next-generation sequencing equipment are able to sequence two to three variable regions, which enables identification to species level for most genera. Overall, next-generation sequencing has revealed a much higher number of species in all types of samples than was previously known, and has drawn attention to new taxa. Due to the tremendous species richness within and diversity between samples, it is likely that the differences between health and disease revealed by studies based on a small number of individuals are due to inter-individual variation rather than disease relatedness. Thus, caution must be taken when interpreting the results from microbiome studies.
Four studies on OSCC/HNSCC included over 100 patients and similar numbers of controls (Börnigen et al., 2017;Hayes et al., 2018;Lee et al., 2017;Yang, Yeh et al., 2018). They all tested saliva/mouthwash/oral rinse samples and sequenced either the V3 or the V3-V4 region. Hayes et al. (2018) and Börnigen et al. (2017) investigated patients with HNSCC, and while Hayes et al. (2018) failed to find any significant difference between healthy and diseased sites, Börnigen et al. (2017) reported the genus Dialister to be significantly increased in HNSCC. On the species level, the two studies also differed (Table 3). Lee et al. (2017) and Yang, Yeh et al. (2018) both tested patients with OSCC patients, but no genera were shared among those significantly enriched in disease (Table 3). Despite extended comparability, no uniform picture of the oral microbiota could be drawn from these four large-scale studies.
The two studies on the mycobiome associated with OSCC confirmed the picture of divergent results. Although it is well known that infection with high-risk HPV increases the risk for oral cancer, the only virome study on OSCC did not associate any virus, including high-risk HPV, with tongue squamous cell carcinoma. The same was the case with the studies on the bacteriome associated with oral leukoplakia and oral lichen planus (Table 3).
Collectively, the microbiome studies demonstrate a highly complex diversity in the oral microbiome associated with oral mucosal diseases. Variations in single species most often cannot discriminate between health and disease for those conditions that involve the resident microbiota. Therefore, comparisons of complexes of microorganisms or community-level comparisons are now being included in the analysis. By analogy to periodontal disease, the contribution of the resident oral microbiota to aetiopathogenesis represents a concerted action of many species in a dysbiotic microbiota; however, even discriminative complexes of bacterial taxa differ from one study to another (Table 3). This indicates that identification of the microbiota alone is not enough to reveal the significance of the resident microbiota in relation to oral mucosal diseases. The in silico analyses of the functional properties of the OSCC-associated microbiota revealed by 16S rRNA gene sequencing did, in fact, report some similarities especially related to pro-inflammatory properties Börnigen et al., 2017;Perera et al., 2018).
Thus, despite discrepancies in the microbiota, functional similarities in the microbiome associated with OSCC could be found. Likewise, a study using metatranscriptomics reported that although there was highly variable metabolic gene expression of the individual species, the metabolism of the entire microbial community associated with periodontal disease was similar among periodontitis patients (Jorth et al., 2014). In addition to sequence-based identification of the  (Clavel, Lagkouvardos, & Hiergeist, 2016;Gupta et al., 2017;Nash et al., 2017;Virgin, 2014;Zhang et al., 2017).
Nevertheless, the microbial differences between health and disease have led to experimental treatment protocols involving probiotics and faecal transplantation. A number of systematic reviews conclude that faecal microbiota transplantation induces remission in some patients with active ulcerative colitis (Costello et al., 2017;Paramsothy et al., 2017). The results on probiotics are more equivocal (Ghouri et al., 2014), maybe due to random choice of species. However, a recent systematic review highlighted that three trials using the same combination of probiotics, tested on a total of 319 patients, caused remission in active ulcerative colitis (Derwa, Gracie, Hamlin, & Ford, 2017 and Candida  with variable results. The oral medicine specialist is therefore encouraged to enter the field of microbiome research. There is both ample opportunity and a clear need for oral medicine specialists who have access to valuable study populations to collaborate with researchers in microbiology, gene sequencing, bioinformatics and biostatistics, to unravel the potential contribution of the oral microbiota as a predictor and aetiological factor in oral mucosal diseases. This review illustrates the difficulties in comparing studies using diverse methodologies. In order to draw solid conclusions, more studies using the same methodology as well as multicentre studies are preferred, as they may inform on the influence of ethnic and environmental variations. It has been repeatedly mentioned that association studies do not indicate aetiological significance of the microorganisms that differ between health and disease. This is unequivocally true. It is also evident that the choice of methods greatly influences the results. Nevertheless, association studies can be valuable for identifying species or microbial compositions of potential importance that can be further studied for their value as prognostic markers of disease or their aetiological and contributory significance in hypothesis-driven studies. In that respect, the oral medicine specialists are encouraged to establish longitudinal cohort studies, which are ideal, albeit cumbersome, to determine both aetiopathogenic relationships and prognostic markers. A recent study involving two US cancer cohorts showed that enrichment and depletion of specific oral bacterial species in mouthwash samples taken before the cancer diagnosis were associated with higher risk of oesophageal cancer development (Peters et al., 2017). Such studies can lead to development of rapid diagnostic tests. The methods are already available, for example the microarray technique (HOMINGS (Human Oral Microbe Identification using Next Generation Sequencing) (Mougeot et al., 2016) and PathoChip (specific cancer-associated species) (Banerjee et al., 2017)). Finally, identification of the microbiome by targeting specific genes cannot stand alone. Metagenomics and transcriptomics adding knowledge of microbial functional properties are crucial methods for an increased understanding of the oral microbial community in health and disease. As pointed out in Part 1, such studies should follow a standardised protocol evolved in collaboration with researchers with microbiological, sequencing, bioinformatics and statistical skills. The contribution of the oral medicine specialist is essential for accurate diagnosis of the patients based on clear diagnostic criteria.

| CON CLUS IONS
High-throughput sequencing has made it possible to determine the microbial composition as well as the functional properties of a sample quickly and at relatively low costs. Of the oral mucosal diseases, OSCC is the only one where it has been applied to any appreciable extent. The studies have not led to a common concept of the microbiota associated with health and disease, probably due to inaccurate diagnosis, numerous methodological variations, a high species richness and an immense diversity between samples. There is clearly a need for more and comparable designed microbiome studies of oral mucosal diseases. As pointed out in Part 1 of this review, such studies should follow a standardised protocol evolved in collaboration with experts in microbiology, sequencing, bioinformatics and biostatistics. The oral medicine specialist is essential for initiation of studies as he/she has access to relevant study populations.

ACK N OWLED G EM ENTS
We thank Thomas Clavel (University clinic Aachen, Germany) for critical reviewing of the manuscript and fruitful discussions. The