RESEARCH ARTICLE

Yang et al., Journal of Medical Microbiology 2022;71:001568

DOI 10.1099/jmm.0.001568 OPEN
ACCESS

Variations in oral microbiome and its predictive functions

between tumorous and healthy individuals
Jing Yang1†, Peng He2,3†, Mou Zhou4, Sheng Li5, Jing Zhang2,3, Xia Tao2,3, Anna Wang2,3 and Xinwei Wu2,3,*

Abstract
Introduction. The oral cavity is one of the largest reservoirs of microorganisms and many pathogenic bacteria have been shown
to be associated with the aetiology of oral cancers.
Gap Statement. Owing to the complexity of oral microbial communities and their unclear relationship with oral cancer, identi-
fication of specific bacteria which contribute to oral cancer is a key imperative.
Aim. To compare and investigate the variations in the composition of the bacterial microbiome and its functions between
patients with oral tumorous lesions and healthy subjects.
Methodology. Twenty-­seven samples from individuals with oral tumours (five oral benign tumours and 22 oral squamous cell
carcinomas) and 15 samples from healthy subjects were collected. Genomic DNA was extracted and the V3–V5 region of the
16S rRNA gene was sequenced. Subsequently, bioinformatic assessment was conducted using QIIME2, PICRUSt and linear
discriminant analysis effect size analyses (LEfSe).
Results. The oral microbiota was composed mainly of the phyla Proteobacteria (31.76 %, 35.00 %), Bacteroidetes (30.13 %,
25.13 %) and Firmicutes (23.92 %, 17.07 %) in tumorous and healthy individuals, respectively. Neisseria, Prevotella, Fusobacterium,
Streptococcus, Capnocytophaga, Veillonella, Haemophilus, Prevotella, Porphyromonas and Leptotrichia were the most abundant
genera. Alpha diversity in the tumour group was significantly greater than that in the healthy group (P<0.05). Differential analy-
sis of microbes between groups demonstrated a significantly higher number of Neisseria, Veillonella, Streptococcus, Leptotri-
chia, Lautropia, Sphingopyxis, Sphingobium, Tannerella, Actinomyces and Rothia in healthy controls compared with the tumour
group. However, the genera Treponema, Micrococcus, Pseudomonas, Janthinobacterium, Parvimos, Loktanella, Staphylococcus,
Acinetobacter, Catonella, Aggregatibacter and Propionibacterium were significantly higher in the tumour group. Pathways related
to cancers, cell motility, environmental adaptation, metabolism and signal transduction were enhanced in the tumour group,
while functions associated with immune system diseases, replication, repair and translation were significantly enhanced in the
healthy group.
Conclusion. Variations in the oral microbiota and its functions showed a correlation with oral tumours. The tumour group
showed an increased abundance of some multi-­drug-­resistant and periodontitis-­related pathogens. The significantly altered
microbiotas may serve as potential biomarkers or inform combination therapy for oral tumours.

Received 21 October 2021; Accepted 31 May 2022; Published 03 August 2022

Author affiliations: 1Department of Clinical Laboratory, Hospital of Stomatology, Sun Yat-­sen University, Guangdong Provincial Key Laboratory of
Stomatology, Guanghua School of Stomatology, Sun Yat-­sen University, Guangzhou 510055, Guangdong, PR China; 2Department of Microbiology,
Guangzhou Center for Disease Control and Prevention, Guangzhou 510440, Guangdong, PR China; 3Institute of Public Health, Guangzhou Medical
University & Guangzhou Center for Disease Control and Prevention, Guangzhou 510440, Guangdong, PR China; 4Department of Blood Transfusion,
General Hospital of Southern Theatre Command of PLA, Guangzhou 510010, Guangdong, PR China; 5Department of Acupuncture and Moxibustion, The
2nd Clinical Hospital of Guangzhou University of Chinese Medicine, Guangzhou 510120, Guangdong, PR China.
*Correspondence: Xinwei Wu, ​tomwu@​126.​com
Keywords: oral tumours; microbiome; microbial functions; NGS.
Abbreviations: ASV, amplicon sequence variant; FDR, false discovery rate; FEP, fibroepithelial polyp; LDA, linear discriminate analysis; LEfSe, linear
discriminant analysis effect size analysis; OSCC, oral squamous cell carcinoma; OSMF, oral submucosa fibrosis; PCoA, principal coordinates analysis;
TCS, two-­component sensing system.
All sequencing results were submitted to NCBI with PRJNA730970 (SRR14583850–SRR14583891).
†These authors contributed equally to this work
Three supplementary table is available with the online version of this article.
001568 © 2022 The Authors
This is an open-­access article distributed under the terms of the Creative Commons Attribution License.

1
 Yang et al., Journal of Medical Microbiology 2022;71:001568

Introduction
Head and neck cancers are among the most common cancers worldwide [1], accounting for approximately 5 % of all tumours
[2]. Approximately half of all head and neck cancers occur in the oral cavity [2]. An estimated 354864 newly diagnosed cases of
lip and oral cavity cancers were reported in 2018, and these sites ranked 18th among the most commonly encountered cancer
sites [1]. In the USA, oral cancer was regarded as the sixth most common cause of cancer-­related mortality [2] and 52260 new
cases of oropharyngeal cancers were diagnosed in 2020 [3]. Oral squamous cell carcinomas (OSCCs) are considered to be the
most common malignant oral tumours [1]. The financial cost for managing oral cancer is considered to be the highest among
all cancers [4]. In addition, oral carcinoma has a high prevalence in Southern Asia, because of widespread use of tobacco-­related
products in this region [5]. In 2015, an estimated 51765 newly diagnosed cases of oral or oropharyngeal cancers were reported
in China, and 23830 deaths were attributable to oral cancer [6].
In-­depth characterization of the causes of oral cancer can help inform preventive interventions. Chronic use of alcohol, tobacco
and betel quid chewing are the most important risk factors for oral cancer [7–9]. However, in a recent study by Bai et al., smoking
and alcohol consumption were not found to be prognostic factors for oral cancer [10], which suggested the possibility of addi-
tional aetiological factors for oral tumours. In the 1990s, Helicobacter pylori was identified as the main causative agent for the
development of gastric cancer [11]. Subsequent research revealed an association of human cancers with certain viruses, bacteria
and parasites, and several microbes were identified as strong risk factors for specific types of cancers [12–14]. The oral cavity
is considered as one of the largest reservoirs of different types of microorganisms, including different species of bacteria, fungi
and viruses [15]. Studies have suggested an aetiological role of many pathogenic bacteria in the causation of oral cancers [16].
Porphyromonas gingivalis and Fusobacteria species were more prevalent on the surface of OSCCs [17]. Al-­hebshi et al. suggested
a close association of Fusobacterium nucleatum and Pseudomonas aeruginosa with OSSC [18]. In addition, Chen et al. unravelled
the variations in the composition of the salivary microbiome between patients with OSCC and oral submucosa fibrosis (OSMF)
and further found considerable reduction in species and phylogenetic diversity in OSMF patients compared with those with OSCC
[19]. Recently, Yang et al. found a close association of the oral microbiota community dynamics with OSCC stage; in addition,
microbiota communities in OSCC patients from stage 4 showed significantly greater complexity compared to that from healthy
controls [16]. Moreover, they found a significant increase in the abundance of Fusobacteria with the progression of oral cancer;
however, the abundance of Streptococcus, Haemophilus, Porphyromonas and Actinomyces decreased with the progression of oral
cancer [16]. In another study, the abundances of Bacillus, Enterococcus, Parvimonas, Peptostreptococcus and Slackia were different
between patients with premalignant lesions and OSCC [20]. Schmidt et al. found significant overrepresentation of Streptococcus
and Rothia at cancer lesion sites compared with the contralateral part of the anatomical site containing normal tissues in patients
with oral cancer [21]; however, other studies found no association between the presence of individual subgingival pathogens and
risk of cancer [22]. Owing to the complexity of oral microbial communities and the conflicting results of their relationship with
oral cancer, identification of specific bacteria that contribute to oral cancer is a key imperative.
A wide range of technologies have been adopted by researchers to assess the correlation of various microbiota with oral cancer,
such as pan-­pathogen array [23], quantitative PCR [24], immunological staining [25], antibody detection [26] and fluorescence
in situ hybridization [27]. However, these methods either depend on bacterial culture or have low throughput. Of note, the vast
majority of the microbiota cannot be cultured [28]; therefore, these methods cannot elaborate on the profiles of the oral microbiota.
High-­throughput sequencing technology, which is independent of culture, can identify thousands of microorganisms by partial
sequencing of the 16S rRNA gene.
To date, there is limited literature on the microbial variations between oral cancerous and healthy subjects. Therefore, in this
study, we aimed to compare the characteristics of oral microbiotas between patients with oral tumours and healthy subjects.
In addition, we compared the variations in the microbiome and its functions in patients with oral tumours (including benign
tumours and OSCCs) and in healthy subjects by high-­throughput sequencing technology.

Methods
Tumour diagnosis was confirmed by histopathology. Patients who were diagnosed with oral cancer for the first time and had no
previous history of any type of tumour were included in the study. Patients who received antibiotics for ≥2 weeks before this study
were excluded from the study group. Healthy control samples were carefully examined clinically and specialized examination
was carried out to exclude the presence of any lesion; healthy controls had no history of other oral mucosal diseases or severe
systemic disorders. Based on a well-­defined clinical protocol, patients with a history of betel chewing, diabetes mellitus or other
systemic complications, such as heart or liver failure, were excluded from this study.
Written informed consent was obtained from all participants. All experiments were performed according to the approved guide-
lines and regulations of this institution.
The surface of benign tumour tissues, OSCC tissues and identical tissues from healthy subjects were sampled using two identical
sterile swabs. Disposable sterile nylon flocking swabs were used (cy-­98000; Hua Chen Yang) and then stored in oral swab

2
 Yang et al., Journal of Medical Microbiology 2022;71:001568

Fig. 1. Rarefaction curves and boxplots of α-diversity in the healthy and tumorous groups. The healthy controls had no history of other oral mucosal
diseases or severe systemic disorders; the tumours from the patients were confirmed by histopathology. Observed species, Shannon indices and
Faith’s PD diversity were calculated using QIIME2, and then plotted using ggplot2. The plateau rarefaction curves indicated that the quenching depth
covered the maximum of the diversity per sample. The index of phylogenetic diversity (P<0.001, Mann–Whitney U test) and observed species (P=0.001,
Mann–Whitney U test) indicated that the diversity of the bacterial communities in the tumorous group was significantly greater than that in the
healthy control group. The Shannon index showed higher bacterial communities in tumorous subjects compared with healthy individuals; however, the
difference was not statistically significant (P>0.1, Mann–Whitney U test).

preservation solution (Tris, EDTA and antiseptic) to prevent DNA degradation. Sterile swabs without sampling were used as
controls. All samples were stored at 4 °C and transferred to an ultra-­low-­temperature freezer (–80 °C) until processing.

DNA extraction and sequencing

Genomic DNA from the oral samples were extracted using the power soil DNA isolation kit (MoBio Laboratories) [29]. Each
identical swab was cut and placed into individual tubes. Then, lysis buffer was added to the tubes, which were heated to 65 °C for
10 min and shaken for 20 min in a MoBio 96-­well plate shaker (MoBio Laboratories). All steps were performed according to the
manufacturer’s instructions. Two parallel DNA samples from one patient were pooled into one tube to minimize any potential
DNA extraction bias. After amplifying the variable V3–V5 region of the 16S rRNA gene with primers 515F/907R, the amplicon
libraries were constructed. Finally, Illumina high-­throughput paired-­end sequencing was performed using an Illumina PE250
platform (Illumina).

Bioinformatics and statistical analysis

Bioinformatics analysis and assessment were conducted using the DADA2 pipeline in QIIME2 [30]. The reads were clustered
into amplicon sequence variants (ASVs) and taxonomy assignment was performed as previously published [31, 32]. Analysis
of alpha (α) microbial diversity was performed using QIIME2 and further plotted using the ggplot2 package in R software. A
Mann–Whitney U test or t-­test was conducted to detect significant differences in α-diversity. Principal coordinates analysis
(PCoA) was performed to assess the differences in microbial distribution between the groups. Permutational multivariate analysis
of variance (PERMANOVA) was applied to further compare the variability in genus abundance and compositions of microbiota
between the groups. The Vegan package 2.0–10 in R software (version 3.2) was used to conduct PCoA. Linear discriminant
analysis effect size analysis (LEfSe) was performed to compare the relative number of different taxa between the groups [33].
The functions of the microbial communities were predicted using PICRUSTt (v1.1.1) [34]. Then, STAMP (v2.1.3) was applied
to determine the functional variations between tumorous and healthy subjects. P values of <0.05 were considered indicative of
statistical significance.

3
 Yang et al., Journal of Medical Microbiology 2022;71:001568

Fig. 2. The most abundant phyla between tumorous and healthy control samples. The healthy controls had no history of other oral mucosal diseases
or severe systemic disorders; the tumours from the patients were confirmed by histopathology. A total of 32 phyla were identified in the two groups.
Among these, 10 were discovered in healthy subjects and 32 were discovered in patients with tumours. In this plot, only the five most abundant phyla
are shown, and the rest are classified as others.

Results
Sampling and sequencing
The details of the 42 swabs collected, which included 27 swabs from tumorous patients and 15 from healthy controls, are presented
in Table S1 (available in the online version of this article). There were no significant between-­group differences with respect
to age (51.41±15.37 versus 46.07±11.09 years, t=1.2975, P=0.2025, t-­test) or sex [Chi-­squared test (χ2)=0.0021605, P=0.9629].
After genomic DNA extraction and sequencing, 1853614 paired reads were obtained. Following quality control and filtration
with the DADA2 pipeline, 519 545 reads were produced with an average of 12370 reads per sample. A total of 1901 features were
identified in the samples (Table S2). No nucleotides were detected in controls. All sequencing results were submitted to NCBI
with numbers SRR14583850–SRR14583891.

Diversity and variations in microbial community between different groups

Different indices (e.g. richness, Shannon and Faith’s PD) were used to assess the α-diversity of the bacterial community (Fig. 1).
Plateau rarefaction curves indicated that the quenching depth covered the maximum of the diversity per sample and new ASVs
with additional sequencing would not be expected. The α-diversity indicated a significant difference between the two groups. The
phylogenetic diversity index indicated that the diversity of the bacterial communities in the tumorous group was significantly
greater than that in the healthy control group (P<0.001, Mann–Whitney U test). A similar trend was demonstrated with respect
to the observed species (P=0.001, Mann–Whitney U test). In addition, the Shannon index showed that there were higher bacterial
communities in tumorous subjects compared with healthy individuals; however, the difference was not statistically significant
(P>0.1, Mann–Whitney U test).

4
 Yang et al., Journal of Medical Microbiology 2022;71:001568

Fig. 3. Heatmap and hierarchical clustering of the top 30 genera in all samples. The dendrogram was created based on the Euclidean distance
and Ward distance. D2 clustering methods were performed in heatmap packages in R software. The genera Neisseria, Prevotella, Fusobacterium,
Streptococcus, Capnocytophaga, Veillonella, Haemophilus, Porphyromonas, Leptotrichia and Aggregatibacter were the most abundant in all the groups.
Hierarchical clustering suggested that the samples from the tumorous group were clustered differently from those collected from the healthy controls.

Taxonomic analysis at the phylum level

A total of 32 phyla were identified in the two groups (Table S3). Among these, 10 were discovered in healthy subjects and 32 were
discovered in tumorous patients. In particular, 22 phyla were only discovered in the tumorous group, including Acidobacteria,
Chloroflexi, Chlorobi, Planctomycetes, Thermi, Nitrospirae, Synergistetes, OP8, Verrucomicrobia, Gemmatimonadetes, OP3, WS1,
Microgenomates, Lentisphaerae, AC1, Elusimicrobia, WS3, TPD.58, Parcubacteria, TM7, NC10 and Fibrobacteres. Among the 32
phyla, the five most abundant phyla in the tumorous and healthy controls accounted for 99.35 and 91.89 % of the total microbes,
respectively (Fig. 2). In the healthy controls, Proteobacteria was the most predominant phylum, which accounted for 31.76 %.
The second most abundant phylum was Bacteroidetes. which accounted for 30.13 %. The next most commonly detected phyla
were Firmicutes at 23.92 %, Fusobacteria at 11.24 %, Actinobacteria at 2.31 % and others at 0.65 %. A similar order of phylum
abundance was revealed in the tumorous group. However, the abundance of Proteobacteria was higher in the tumorous group
compared with the healthy controls. The percentage of Proteobacteria was higher (35.00%), while that of Bacteroidetes was lower
(25.13 %) in the tumorous group. The phyla Firmicutes and Fusobacteria were slightly reduced (17.07 and 10.70, respectively),
while the percentage of Actinobacteria was higher (3.98 %) in the tumorous group. The top five phyla accounted for 91.88 % and
the others accounted for 8.12%, which was much higher than that in the healthy control group (0.65 %). This suggested that the
microenvironment in tumorous patients had denser cultures with different types of microbiota.

Taxonomic analysis at the genus level

At the genus level, a total of 207 different genera were identified. The top 30 genera among all the samples were selected and
are shown in Fig. 3. The genera Neisseria, Prevotella, Fusobacterium, Streptococcus, Capnocytophaga, Veillonella, Haemophilus,
Porphyromonas, Leptotrichia and Aggregatibacter were the most abundant in all the groups. Hierarchical clustering suggested that
the samples from the tumorous group were clustered differently from those collected from the healthy controls.

5
 Yang et al., Journal of Medical Microbiology 2022;71:001568

Fig. 4. Principal component analysis (PCoA) at the ASV level. PCoA was conducted using QIIME2, then plotted in R software using ggplot2. The ellipse
was plotted with a confidence level of 0.95. The spots that represent the tumorous samples show a more dispersed distribution pattern than those
from the healthy controls. Samples from the tumorous group were significantly different from those in the control group (PERMANOVA, pseudo-­
F=5.756921, P=0.001).

Neisseria was the most dominant genus in the healthy control group, and accounted for 21.23 %, followed by Prevotella (17.57 %),
Streptococcus (9.74 %), Veillonella (9.09 %) and Fusobacterium (6.70 %). However, in the tumorous group, the most abundant genera
were Prevotella (11.02 %), Fusobacterium (9.02 %), Capnocytophaga (6.62 %), Neisseria (5.42 %) and Haemophilus (4.93 %) (Fig. 3).
Of interest, the top 10 genera in the control group accounted for 84.77 %, while the corresponding percentage in the tumorous
group was significantly lower (55.08 %). The top 30 genera decreased dramatically from 98.43 to 78.56 %. This indicated that the
microbiota in the tumorous group was more evenly distributed compared with that in the healthy control group.

Microbial signatures associated with tumorous and control group

Beta diversity revealed the differences in the composition of the bacterial community between the two groups. PCoA at the genus
level was used to determine the distribution of microbiota between the two groups (Fig. 4). As indicated by the hierarchical
clustering (Fig. 3), the spots that represent the tumorous samples showed a more dispersed distribution pattern than those from the
healthy controls. Samples from the tumorous group were significantly different from those in the control group (PERMANOVA,
pseudo-­F=5.756921, P=0.001).
The LEfSe method was used to identify the differential expression of genera between the two groups (Fig. 5). The genera Neisseria,
Veillonella, Streptococcus, Leptotrichia, Lautropia, Sphingopyxis, Sphingobium, Tranerella, Actinomyces and Rothia were significantly
higher in the healthy control group. However, Treponema, Micrococcus, Pseudomonas, Janthinobacterium, Parvimos, Loktanella,
Staphylococcus, Acinetobacter, Catonella, Aggregatibacter and Propionibacterium were significantly higher in the tumorous group.

Differential oral microbial community function

PICRUSt and STAMP revealed significant functional divergences in several genes involved in different pathways (Figs 6 and 7).
Similar to the distribution of microbial taxa, the distribution of gene functions was significantly different between the tumorous
and healthy control groups [t-­test, false discovery rate (FDR) correction, corrected P<0.05]. The functions connected with cancers

6
 Yang et al., Journal of Medical Microbiology 2022;71:001568

Fig. 5. Distinct taxa identified between two groups using linear discriminant analysis effect size analyses (LEfSe). The results of linear discriminate
analysis (LDA) showed that the bacteria were significantly different between the tumorous and healthy control samples: (a) Cladogram indicating the
phylogenetic distribution of bacteria that were significantly increased; and (b) LDA scores showing significant bacterial differences between groups
at the genus level. The genera Neisseria, Veillonella, Streptococcus, Leptotrichia, Lautropia, Sphingopyxis, Sphingobium, Tranerella, Actinomyces and
Rothia were significantly higher in the healthy control group, whereas Treponema, Micrococcus, Pseudomonas, Janthinobacterium, Parvimos, Loktanella,
Staphylococcus, Acinetobacter, Catonella, Aggregatibacter and Propionibacterium were significantly higher in the tumorous group.

(corrected P=0.010), cell motility (corrected P=0.009), environmental adaptation (corrected P<0.001), metabolism (corrected
P<0.001) and signal transduction (corrected P<0.011) were significantly enhanced in the tumorous group compared with the
control healthy group. However, the pathways related to immune system diseases (corrected P=0.013), replication and repair
(corrected P=0.021), and translation (corrected P=0.028) were significantly enhanced in the control healthy group (Fig. 6).
On level 3, 27 different pathways were significantly different (Fig. 7). Among these, 17 were more abundant in the tumorous group
and 10 were were more abundant in the control healthy group (t-­test, Storey FDR correction, corrected P<0.05). The pathways
that were related to MAPK signalling, two-­component system, bacterial chemotaxis, flagellar assembly, sulphur relay systems
and cell motility were significantly increased in the tumorous group compared with the healthy control group. However, the
samples from the control group had more genes related to lysine biosynthesis and aromatic amino acids (phenylalanine, tyrosine
and tryptophan) biosynthesis.

Discussion
There are approximately 1000 species that are capable of achieving stable colonization in the oral cavity [35]. The oral microflora
is regarded as the second most complex microbiota in humans [36]. The development of high-­throughput DNA sequencing
technology at the molecular genetics level has facilitated in-­depth characterization of the most important aspects of the oral
microbiota. In addition, it has helped identify the oral microbiota associated with oral health and oral cancer [37]. However,
the results pertaining to the relationship between oral health and specific microbiota have been inconsistent, because of diverse
human lineages, such as related to different living conditions, diverse cultural backgrounds and different dietary habits [38]. Use
of different technology [39], diverse types of sample [40, 41] and various methods of sequencing [42] may have contributed to
the inconsistent results. In this study, the surfaces of tumorous lesions and healthy controls were sampled and assessed further.
Oral samples from tumorous and healthy individuals were collected. The V3–V5 region of the 16S rRNA gene was chosen for
sequencing and analysis [43].
Our results indicated greater α-diversity in species in the tumorous group compared with the healthy control group (Fig. 1),
which was consistent with previous studies [40, 44]. In a detailed large-­scale assessment of the salivary microbiota conducted by
Yoshihisa et al., the bacterial richness in the salivary microbiota was found to be significantly associated with poor oral health [38].
In general, good oral health was associated with lower diversity in the salivary microbiome [45]. A recent survey of in situ bacterial
effects indicated that patients with a high microbiome diversity had increased tumour infiltration of CD4+ lymphocytes, which
consequently prolongs the management of cancer [46]. Heshiki et al. examined the influence of the microbiome on the treatment

7
 Yang et al., Journal of Medical Microbiology 2022;71:001568

Fig. 6. The gene functions on level-­2 that were significantly increased in different groups. The functions were predicted by PICRUSt with default
settings, then the t-­test and Storey FDR correction were conducted using STAMP. The functions connected with cancers (corrected P=0.010), cell
motility (corrected P=0.009), environmental adaptation (corrected P<0.001), metabolism (corrected P<0.001) and signal transduction (corrected
P<0.011) were significantly enhanced in the tumorous group compared with the healthy control group. However, the pathways related to immune
system diseases (corrected P=0.013), replication and repair (corrected P=0.021), and translation (corrected P=0.028) were significantly enhanced in
the healthy control group.

outcomes of eight diverse malignancy types and found that a more diverse intestinal microbiome was associated with improved
systemic and antitumour immunity [47]. Preserving diversity is an important element of the survival of various microbiota [48].
Changes in the oral microenvironment in tumorous patients may contribute to the changes in diversity, which is more likely to
induce a favourable antitumour response. However, some authors reported a significant reduction in microbial diversity in cancer
patients compared with controls [39]. This discrepancy may be attributable to the different ethnic background or lifestyle-­related
factors in the study population. The relationship between microbial diversity and oral disease status needs to be identified.
In this study, Proteobacteria, Bacteroidetes, Firmicutes, Fusobacteria and Actinobacteria were the most abundant phyla (Fig. 2).
These results are similar to those of other studies [40]. However, the most abundant bacteria in these phyla by percentage were
not consistent with this study. Proteobacteria was the most abundant phylum detected in our study, while Bacteroidetes was the
most abundant phylum reported by Zhao et al. [40]. Firmicutes was the most abundant phylum reported in another study [21].
Neisseria was the most abundant genus in our study, while Prevotella was the most abundant genus reported in the previous study
[40] (Fig. 3). These differences may be attributable to the different genetic characteristics of the population or the different timing
of sampling; therefore, more accurate standardized inclusion criteria and protocols are required to achieve consistent results.
In our study, Neisseria, Prevotella, Streptococcus, Veillonella, Fusobacterium, Capnocytophaga and Haemophilus were the most
abundant genera; all these have been regarded as the common core oral microbiota in other studies [49]. Neisseria was the most
abundant genus in our healthy control group, while Prevotella was the most abundant in the tumorous group. Yoshihisa et al.
suggested that the predominance of the genus Prevotella in the salivary microbiota could be attributed to severe periodontal
disease, while the predominance of the genus Neisseria indicated healthy periodontal conditions [38]. Similar results were reported

8
 Yang et al., Journal of Medical Microbiology 2022;71:001568

Fig. 7. The functions on level-­3 that were significantly increased in different groups. The functions were predicted by PICRUSt with default settings,
then the t-­test and Storey FDR correction were conducted using STAMP. Twenty-­seven different pathways were significantly different. Among these,
17 were more abundant in the tumorous group and 10 were were more abundant in the control healthy group (t-­test, Storey FDR correction, corrected
P<0.05). The pathways that were related to mitogen-­activated protein kinase (MAPK) signalling, two-­component system, bacterial chemotaxis, flagellar
assembly, sulphur relay systems and cell motility were significantly increased in the tumorous group compared with the healthy control group.
However, the samples from the control group had more genes related to lysine biosynthesis and aromatic amino acids (phenylalanine, tyrosine and
tryptophan) biosynthesis.

by Rafael et al., who compared the salivary microbiome from oropharyngeal cancer (OPSCC), OCSCC and normal healthy
controls, and found a significant decrease in Neisseria in OPSCC and OCSCC [39]. Zhao et al. suggested that Neisseria may play
an important role in maintaining the stability of the oral bacterial community [40]. In addition, our study found significantly
greater abundance of Streptococcus in the healthy group. Previous studies have suggested that Streptococcus may be beneficial to
the host, because these bacteria may produce molecules that are inhibitory to pathogenic species [50]. Tagaino et al. suggested
that Streptococcus mitis, Streptococcus salivarius and Streptococcus mutans may have the ability to produce acetaldehyde from
ethanol or glucose [51]. Acetaldehyde is a carcinogen and was shown to contribute to sister chromatid exchanges, gene mutations
and DNA-­strand breaks [51, 52]. The production of acetaldehyde is affected by the oral microenvironment, such as pH, oxygen
concentration, metabolic substrates and even salivary secretions [51]. Therefore, the production of acetaldehyde in healthy
individuals should be studied in future research.
Of the top 30 genera, Sphingobium was almost only exclusively discovered in healthy individuals. Recent studies have suggested
its potential protective role in breast cancer, because of its ability to metabolize oestrogen [53, 54]. Aggregatibacter was more
abundant in tumorous patients in our study. Aggregatibacter actinomycetemcomitans is one of the key causative pathogens in
chronic periodontitis. In addition, it has been associated with aggressive periodontitis [55]. Long-­term chronic periodontal
disease increases the risk of oral cavity cancers, especially OSCC [56]. However, consistent chronic inflammation may not be
the only reason for carcinogenesis, because accumulation of genome instabilities, such as point mutations, deletions, inversions,
duplications, translocations and chromosome loss, is a key factor for carcinogenesis [57]. Teshima et al. discovered that Aggre-
gatibacter actinomycetemcomitans infection can cause breakage in the DNA double-­strand in host cells [57], which suggested that
Aggregatibacter actinomycetemcomitans may increase the risk of genome instabilities in host cells and subsequently increase the
risk of carcinogenesis. Leukotoxin (LtxA), which is produced by Aggregatibacter actinomycetemcomitans, can kill neutrophils,

9
 Yang et al., Journal of Medical Microbiology 2022;71:001568

lymphocytes and monocytes, allowing it to evade host immune surveillance [58]. Therefore, Aggregatibacter actinomycetemcomi-
tans may have a deleterious effect on the human immune system. In addition, it has been shown to be associated with systemic
complications, such as endocarditis, rheumatoid arthritis [59] and lung cancer [60, 61]. Some studies have found an association
between Aggregatibacter actinomycetemcomitans and development of pancreatic cancer and oesophageal cancer [62, 63]. There-
fore, further research is required to reveal its role in the development and progression of oral cancers. In addition, Acinetobacter
baumannii is one of the most common clinical multidrug-­resistant bacteria that is frequently isolated in wound infection following
surgical intervention for OSCC [64]. However, our study was conducted before the surgical procedure; therefore, higher numbers
of Acinetobacter baumannii may have existed before surgery, which might have contributed to its commonly reported isolation
following oral cancer surgery. Inhibiting Acinetobacter growth before a surgical procedure could be an effective method to reduce
multidrug-­resistant bacterial infection. The abundance of the genus Propionibacterium was significantly greater in the tumour
group compared with that in the healthy control group. Perera et al. discovered that rates were higher in fibroepithelial polyps
(FEPs) compared with OSCCs [65]. It would be of interest to explore the role of Propionibacterium in the aetiology of FEPs or
benign tumours.
PICRUSt and the STAMP method were used to assess the functions of various bacterial communities. Genes related to signal
transduction were upregulated in the tumorous group compared with the healthy controls (Fig. 6). Further analysis suggested
that the genes connected to two-­component sensing systems (TCSs) were more abundant in the tumorous group (1.46–1.08 %,
P=0.04, t-­test, FDR-­corrected). These regulatory systems are ubiquitous signalling units found in prokaryotes, which tailor the
response based on the local conditions. TCSs were shown to mediate bacterial acclimatization to various environmental changes
by coupling environmental signals to gene expression [66]. Upregulation of these systems may promote bacterial survival. The
correlation of MAPK pathway changes with the oral microbiota may inform novel combination therapies for oral tumours.
Of interest, bacterial chemotaxis and flagellar assembly, which are involved in cell motility, were significantly higher in the
tumorous group (Fig. 7). Williams et al. showed that H. pylori mutants with defective chemotaxis induced less inflammation
than the wild-­type H. pylori [67]. Wang et al. suggested an important role of bacterial chemotaxis in periodontitis-­related
inflammation [68]. Similar results were indicated in other reports [18]. Flagellin was shown to stimulate significant migration
of innate immune cells, such as macrophages, dendritic cells and neutrophils to colonized tumours [69]. The functions of
both genes demonstrate their antitumour ability, and the upregulation of these suggested antitumour immunological effects
of some bacteria in the oral microbiota. In addition, genes correlated with replication and repair were significantly reduced
in tumorous patients. Oral bacterial virulence factors produced by bacteria such as Aggregatibacter actinomycetemcomitans
(which was more abundant in the tumorous group, Fig. 5b) may induce single-­strand DNA breakages, produce replication
stress response, and ultimately lead to breakage to double-­strands and arrest cell cycle replication [70]. Inefficient repair of
damaged DNA could allow survival of cells with defective DNA, resulting in accumulation of multiple genome instabilities
and promoting tumorigenesis and progression [71]. In addition, the pathways related to immune system disease, replication or
repair, and translation were significantly overrepresented in controls. Lysine biosynthesis was upregulated in healthy controls
(Fig. 7). Because the bacterial biosynthesis of lysine is important in the construction of bacterial peptidoglycan cell walls [72],
the underrepresented lysine biosynthesis genes in tumours indicated that their microenvironment had compromised the
microbiota. Biosynthesis of three aromatic amino acids (phenylalanine, tyrosine and tryptophan) was upregulated in healthy
controls (Fig. 7). These aromatic amino acids are essential for protein synthesis [73], and dysmetabolism of aromatic amino
acids leads to cancers [74].
Some limitations of our study should be considered. First, the small sample size may have reduced the number of significantly
different genera between the tumorous and healthy groups. However, 21 genera were identified, which included some genera
that have been reported by other researchers. Second, consumption of alcohol and tobacco smoking are important aetiological
factors for oral cancers, and therefore these factors are likely to affect the oral microbiota. Because of the small sample size, we
did not discuss their impact. Therefore, large samples and a detailed comprehensive questionnaire survey on living habits might
be required in future studies. Third, we did not assess the variations in the significantly different genera during cancer progres-
sion from early to advanced stages. Further validation of the statistically different genera in various histopathological stages of
dysplastic oral cancer could provide valuable insights.
In conclusion, we observed a correlation between different variations in the oral microbiota and its functions with oral
tumours. The tumorous group exhibited significantly greater diversity, which may indicate poor oral health. The tumorous
group also showed significant overrepresentation of the abundance of some multi-­drug-­resistant and periodontitis-­related
pathogens. The significantly altered microbiota could be explored further as novel biomarkers and may help inform combina-
tion therapies for oral tumours. The results of this study expended our knowledge of oral bacterial communities in patients
with oral tumours.

Funding information
This work was supported by Rapid Pathogen Detections and Infectious Disease Alerts, the Key Medical Discipline in Guangzhou (2021-­2023-­11); Basic
Research Project of Key Laboratory of Guangzhou (2021BRP004)

10
 Yang et al., Journal of Medical Microbiology 2022;71:001568

Author contributions
J.Y. and X.W. conceived and designed research; M.Z., J.Z., X.T. and A.W. collected data and conducted research; P.H. and S.L. analysed and interpreted
data; P.H. and J.Y. wrote the initial paper; X.W. and P.H. revised the paper; X.W. had primary responsibility for final content. All authors read and approved
the final manuscript.
Conflicts of interest
The authors declare that they have no conflicts of interest.
Ethical statement
This work was approved by the Research Ethics Committee of the Hospital of Stomatology, Sun Yat-­sen University at District of Yuexiu, Guangzhou
City, China.

References 19. Chen MY, Chen JW, Wu LW, Huang KC, Chen JY, et al. Carcinogen-
1. Bray F, Ferlay J, Soerjomataram I, Siegel RL, Torre LA, et al. Global esis of male oral submucous fibrosis alters salivary microbiomes.
cancer statistics 2018: GLOBOCAN estimates of incidence and J Dent Res 2021;100:397–405.
mortality worldwide for 36 cancers in 185 countries. CA Cancer J 20. Lee W-­H, Chen H-­M, Yang S-­F, Liang C, Peng C-­Y, et al. Bacterial
Clin 2018;68:394–424. alterations in salivary microbiota and their association in oral
2. Kademani D. Oral cancer. Mayo Clin Proc 2007;82:878–887. cancer. Sci Rep 2017;7:16540.
3. Siegel RL, Miller KD, Jemal A. Cancer statistics, 2020. CA Cancer J 21. Schmidt BL, Kuczynski J, Bhattacharya A, Huey B, Corby PM, et al.
Clin 2020;70:7–30. Changes in abundance of oral microbiota associated with oral
4. Huber MA, Tantiwongkosi B. Oral and oropharyngeal cancer. Med cancer. PLoS One 2014;9:e98741.
Clin North Am 2014;98:1299–1321. 22. Mai X, Genco RJ, LaMonte MJ, Hovey KM, Freudenheim JL,
5. Fu J-­Y, Wu C-­X, Zhang C-­P, Gao J, Luo J-­F, et al. Oral cancer inci- et  al. Periodontal pathogens and risk of incident cancer in post-
dence in Shanghai ---- a temporal trend analysis from 2003 to menopausal females: the buffalo osteoperio study. J Periodontol
2012. BMC Cancer 2018;18:686. 2016;87:257–267.
6. Lei L, Zheng R, Peng K, Si L, Peng J, et al. Incidence and mortality 23. Banerjee S, Tian T, Wei Z, Peck KN, Shih N, et al. Microbial signa-
of oral and oropharyngeal cancer in China, 2015. Chin J Cancer Res tures associated with oropharyngeal and oral squamous cell carci-
2020;32:1–9. nomas. Sci Rep 2017;7:4036.
7. Lin X, Wu X, Gomaa A, Chen J, Wu L, et al. Analysis of risk factors 24. Komiya Y, Shimomura Y, Higurashi T, Sugi Y, Arimoto J, et  al.
for multiple primary oral squamous cell carcinoma: a cohort study. Patients with colorectal cancer have identical strains of Fusobac-
Clin Oral Investig 2020;24:3147–3155. terium nucleatum in their colorectal cancer and oral cavity. Gut
2019;68:1335–1337.
8. Guha N, Warnakulasuriya S, Vlaanderen J, Straif K. Betel quid
chewing and the risk of oral and oropharyngeal cancers: a 25. Gao S, Li S, Ma Z, Liang S, Shan T, et al. Presence of Porphyromonas
meta-­analysis with implications for cancer control. Int J Cancer gingivalis in esophagus and its association with the clinicopatho-
2014;135:1433–1443. logical characteristics and survival in patients with esophageal
cancer. Infect Agent Cancer 2016;11:3.
9. Hung L-­C, Kung P-­T, Lung C-­H, Tsai M-­H, Liu S-­A, et  al. Assess-
ment of the risk of oral cancer Incidence in a high-­risk population 26. Gao S-­G, Yang J-­Q, Ma Z-­K, Yuan X, Zhao C, et  al. Preoperative
and establishment of a predictive model for oral cancer incidence serum immunoglobulin G and A antibodies to Porphyromonas
using a population-­based cohort in Taiwan. Int J Environ Res Public gingivalis are potential serum biomarkers for the diagnosis and
Health 2020;17:E665. prognosis of esophageal squamous cell carcinoma. BMC Cancer
2018;18:17.
10. Bai X-­X, Zhang J, Wei L. Analysis of primary oral and oropharyn-
geal squamous cell carcinoma in inhabitants of Beijing, China-­a 27. Castellarin M, Warren RL, Freeman JD, Dreolini L, Krzywinski M,
10-­
year continuous single-­ center study. BMC Oral Health et  al. Fusobacterium nucleatum infection is prevalent in human
2020;20:208. colorectal carcinoma. Genome Res 2012;22:299–306.
11. Kim SS, Ruiz VE, Carroll JD, Moss SF. Helicobacter pylori in the 28. Vartoukian SR, Palmer RM, Wade WG. Strategies for culture of
pathogenesis of gastric cancer and gastric lymphoma. Cancer Lett “unculturable” bacteria. FEMS Microbiol Lett 2010;309:1–7.
2011;305:228–238. 29. Vo A-­TE, Jedlicka JA. Protocols for metagenomic DNA extraction
12. Rochford R, Korir A, Newton R. Viral-­associated malignancies in and Illumina amplicon library preparation for faecal and swab
Africa: are viruses “infectious traces” or “dominant drivers”? Curr samples. Mol Ecol Resour 2014;14:1183–1197.
Opin Virol 2016;20:28–33. 30. Bolyen E, Rideout JR, Dillon MR, Bokulich NA, Abnet CC, et  al.
13. Flavahan WA, Gaskell E, Bernstein BE. Epigenetic plasticity and the Reproducible, interactive, scalable and extensible microbiome
hallmarks of cancer. Science 2017;357:eaal2380. data science using QIIME 2. Nat Biotechnol 2019;37:852–857.
14. Doorakkers E, Lagergren J, Engstrand L, Brusselaers N. Helico- 31. Kaehler BD, Bokulich NA, McDonald D, Knight R, Caporaso JG, et al.
bacter pylori eradication treatment and the risk of gastric adeno- Species abundance information improves sequence taxonomy
carcinoma in a Western population. Gut 2018;67:2092–2096. classification accuracy. Nat Commun 2019;10:4643.
15. Baker JL, Bor B, Agnello M, Shi W, He X. Ecology of the oral micro- 32. Bokulich NA, Kaehler BD, Rideout JR, Dillon M, Bolyen E, et  al.
biome: beyond bacteria. Trends Microbiol 2017;25:362–374. Optimizing taxonomic classification of marker-­ gene amplicon
16. Yang C-­Y, Yeh Y-­M, Yu H-­Y, Chin C-­Y, Hsu C-­W, et al. Oral microbiota sequences with QIIME 2’s q2-­feature-­classifier plugin. Microbiome
community dynamics associated with oral squamous cell carci- 2018;6:90.
noma staging. Front Microbiol 2018;9:862. 33. Ye SH, Siddle KJ, Park DJ, Sabeti PC. Benchmarking metagen-
17. Bolz J, Dosá E, Schubert J, Eckert AW. Bacterial colonization of omics tools for taxonomic classification. Cell 2019;178:779–794.
microbial biofilms in oral squamous cell carcinoma. Clin Oral 34. Langille MGI, Zaneveld J, Caporaso JG, McDonald D, Knights D, et al.
Investig 2014;18:409–414. Predictive functional profiling of microbial communities using 16S
18. Al-­Hebshi NN, Nasher AT, Maryoud MY, Homeida HE, Chen T, et al. rRNA marker gene sequences. Nat Biotechnol 2013;31:814–821.
Inflammatory bacteriome featuring Fusobacterium nucleatum and 35. Sun J, Tang Q, Yu S, Xie M, Xie Y, et  al. Role of the oral micro-
Pseudomonas aeruginosa identified in association with oral squa- biota in cancer evolution and progression. Cancer Med
mous cell carcinoma. Sci Rep 2017;7:1834. 2020;9:6306–6321.

11
 Yang et al., Journal of Medical Microbiology 2022;71:001568

36. Deo PN, Deshmukh R. Oral microbiome: unveiling the fundamen- 56. Tezal M, Sullivan MA, Hyland A, Marshall JR, Stoler D, et al. Chronic
tals. J Oral Maxillofac Pathol 2019;23:122–128. periodontitis and the incidence of head and neck squamous cell
37. Chattopadhyay I, Verma M, Panda M. Role of oral microbiome carcinoma. Cancer Epidemiol Biomarkers Prev 2009;18:2406–2412.
signatures in diagnosis and prognosis of oral cancer. Technol 57. Teshima R, Hanada K, Akada J, Kawano K, Yamaoka Y. Aggre-
Cancer Res Treat 2019;18:1533033819867354. gatibacter actinomycetemcomitans infection causes DNA double-­
38. Yamashita Y, Takeshita T. The oral microbiome and human health. strand breaks in host cells. Genes Cells 2018;23:264–273.
J Oral Sci 2017;59:201–206. 58. Sampathkumar V, Velusamy SK, Godboley D, Fine DH. Increased
39. Guerrero-­Preston R, Godoy-­Vitorino F, Jedlicka A, Rodríguez-­Hilario A, leukotoxin production: characterization of 100 base pairs within
the 530 base pair leukotoxin promoter region of Aggregatibacter
González H, et  al. 16S rRNA amplicon sequencing identifies micro-
actinomycetemcomitans. Sci Rep 2017;7:1887.
biota associated with oral cancer, human papilloma virus infection and
surgical treatment. Oncotarget 2016;7:51320–51334. 59. Mukherjee A, Jantsch V, Khan R, Hartung W, Fischer R, et al. Rheu-
matoid arthritis-­associated autoimmunity due to Aggregatibacter
40. Zhao H, Chu M, Huang Z, Yang X, Ran S, et  al. Variations in oral
actinomycetemcomitans and its resolution with antibiotic therapy.
microbiota associated with oral cancer. Sci Rep 2017;7:11773.
Front Immunol 2018;9:2352.
41. Wang L, Yin G, Guo Y, Zhao Y, Zhao M, et al. Variations in oral micro-
60. Turner E, Hastie T, Sundaresan PD. Pulmonary infection masquer-
biota composition are associated with a risk of throat cancer. Front
ades as malignancy in a patient with periodontitis. BMJ Case Rep
Cell Infect Microbiol 2019;9:205.
2021;14.
42. Al-­Hebshi NN, Nasher AT, Idris AM, Chen T. Robust species
61. Matzumura-­Kuan M, Jennings J. Aggregatibacter actinomycetem-
taxonomy assignment algorithm for 16S rRNA NGS reads: appli-
comitans infection mimicking lung cancer: a case report. Scand J
cation to oral carcinoma samples. J Oral Microbiol 2015;7:28934.
Infect Dis 2014;46:669–672.
43. Sun DL, Jiang X, Wu QL, Zhou NY. Intragenomic heterogeneity of 62. Fan X, Alekseyenko AV, Wu J, Peters BA, Jacobs EJ, et al. Human
16S rRNA genes causes overestimation of prokaryotic diversity. oral microbiome and prospective risk for pancreatic cancer: a
Appl Environ Microbiol 2013;79:5962–5969. population-­based nested case-­control study. Gut 2018;67:120–127.
44. Li Y, Tan X, Zhao X, Xu Z, Dai W, et  al. Composition and function 63. Kawasaki M, Ikeda Y, Ikeda E, Takahashi M, Tanaka D, et al. Oral
of oral microbiota between gingival squamous cell carcinoma and infectious bacteria in dental plaque and saliva as risk factors in
periodontitis. Oral Oncol 2020;107:104710. patients with esophageal cancer. Cancer 2021;127:512–519.
45. Takeshita T, Kageyama S, Furuta M, Tsuboi H, Takeuchi K, et  al. 64. Belusic-­Gobic M, Zubovic A, Predrijevac A, Harmicar D, Cerovic R,
Bacterial diversity in saliva and oral health-­related conditions: the et al. Microbiology of wound infection after oral cancer surgery. J
Hisayama Study. Sci Rep 2016;6:22164. Craniomaxillofac Surg 2020;48:700–705.
46. Sims TT, El Alam MB, Karpinets TV, Dorta-­Estremera S, Hegde VL, 65. Perera M, Al-­Hebshi NN, Perera I, Ipe D, Ulett GC, et  al. Inflam-
et  al. Gut microbiome diversity is an independent predictor of matory bacteriome and oral squamous cell carcinoma. J Dent Res
survival in cervical cancer patients receiving chemoradiation. 2018;97:725–732.
Commun Biol 2021;4:237.
66. Ibrahim IM, Puthiyaveetil S, Allen JF. A Two-­component regulatory
47. Heshiki Y, Vazquez-­Uribe R, Li J, Ni Y, Quainoo S, et al. Predictable system in transcriptional control of photosystem stoichiometry:
modulation of cancer treatment outcomes by the gut microbiota. redox-­dependent and sodium ion-­dependent phosphoryl transfer
Microbiome 2020;8:28. from cyanobacterial histidine kinase hik2 to response regulators
48. Bhushan B, Yadav AP, Singh SB, Ganju L. Diversity and functional Rre1 and RppA. Front Plant Sci 2016;7:137.
analysis of salivary microflora of Indian Antarctic expeditionaries. 67. Williams SM, Chen YT, Andermann TM, Carter JE, McGee DJ,
J Oral Microbiol 2019;11:1581513. et al. Helicobacter pylori chemotaxis modulates inflammation and
49. Zaura E, Keijser BJF, Huse SM, Crielaard W. Defining the healthy bacterium-­gastric epithelium interactions in infected mice. Infect
“core microbiome” of oral microbial communities. BMC Microbiol Immun 2007;75:3747–3757.
2009;9:259. 68. Wang J, Qi J, Zhao H, He S, Zhang Y, et al. Metagenomic sequencing
50. Abranches J, Zeng L, Kajfasz JK, Palmer SR, Chakraborty B, et al. reveals microbiota and its functional potential associated with
Biology of oral Streptococci. Microbiol Spectr 2018;6. periodontal disease. Sci Rep 2013;3:1843.
5 1. Tagaino R, Washio J, Abiko Y, Tanda N, Sasaki K, et al. Meta- 69. Duong MT-­Q, Qin Y, You S-­H, Min J-­J. Bacteria-­cancer interactions:
bolic property of acetaldehyde production from ethanol bacteria-­based cancer therapy. Exp Mol Med 2019;51:1–15.
and glucose by oral Streptococcus and Neisseria. Sci Rep 70. Sun Z, Xiong C, Teh SW, Lim JCW, Kumar S, et al. Mechanisms of
2019;9:10446. oral bacterial virulence factors in pancreatic cancer. Front Cell
52. Singh NP, Khan A. Acetaldehyde: genotoxicity and cytotoxicity in Infect Microbiol 2019;9:412.
human lymphocytes. Mutat Res 1995;337:9–17. 71. Negrini S, Gorgoulis VG, Halazonetis TD. Genomic instability--an
53. Parida S, Sharma D. The power of small changes: comprehen- evolving hallmark of cancer. Nat Rev Mol Cell Biol 2010;11:220–228.
sive analyses of microbial dysbiosis in breast cancer. Biochim 72. Hutton CA, Perugini MA, Gerrard JA. Inhibition of lysine biosyn-
Biophys Acta Rev Cancer 2019;1871:392–405. thesis: an evolving antibiotic strategy. Mol Biosyst 2007;3:458–465.
54. Lawani-­Luwaji EU, Alade T. Sphingomonadaceae: protective 73. Ennis MA, Ong A-­J, Lim K, Ball RO, Pencharz PB, et  al. Dietary
against breast cancer? Bull Natl Res Cent 2020;44:191. aromatic amino acid requirements during early and late gestation
55. Velusamy SK, Sampathkumar V, Ramasubbu N, Paster BJ, in healthy pregnant women. J Nutr 2020;150:3224–3230.
Fine DH. Aggregatibacter actinomycetemcomitans colonization 74. Sitkin SI, Tkachenko EI, Vakhitov TY. Metabolic dysbiosis of
and persistence in a primate model. Proc Natl Acad Sci U S A the gut microbiota and its biomarkers. Eksp Klin Gastroenterol
2019;116:22307–22313. 2016;12:6–29.

12