Microbiota in the Throat and Risk Factors for Laryngeal Carcinoma

Abstract

The compositions and abundances of the microbiota in the ecological niche of the human throat and the possible relationship between the microbiota and laryngeal cancer are poorly understood. To obtain insight into this, we enrolled 27 laryngeal carcinoma patients and 28 subjects with vocal cord polyps as controls. For each subject, we simultaneously collected swab samples from the upper throat near the epiglottis (site I) and tissue samples from the vestibulum laryngis to the subglottic region (site II). The microbiota of the throat were fully characterized by pyrosequencing of barcoded 16S rRNA genes. We found 14 phyla, 20 classes, 38 orders, 85 families, and 218 genera in the throats of enrolled subjects. The main phyla were Firmicutes (54.7%), Fusobacteria (14.8%), Bacteroidetes (12.7%), and Proteobacteria (10.6%). Streptococcus (37.3%), Fusobacterium (11.3%), and Prevotella (10.6%) were identified as the three most predominant genera in the throat. The relative abundances of 23 bacterial genera in site I were significantly different from those in site II (P < 0.05). The relative proportions of 12 genera largely varied between laryngeal cancer patients and control subjects (P < 0.05). Collectively, this study outlined the spatial structure of microbial communities in the human throat. The spatial structure of bacterial communities significantly varied in two anatomical sites of the throat. The bacterial profiles of the throat of laryngeal cancer patients were strongly different from those of control subjects, and several of these microorganisms may be related to laryngeal carcinoma.

INTRODUCTION

Laryngeal carcinoma (LC) is the second most common cancer of the respiratory tract, and the overall 5-year relative survival rate is approximately 64.2% (1, 2). Treatments for laryngeal cancer include surgery, radiotherapy, chemotherapy, and biological agents, which can be used either individually or in combination (3, 4). The risk of this type of cancer is greatly increased by cigarette smoking and alcohol consumption. However, the involvement of other agents, such as genetics, nutrition, life-style, and local microbial ecology conditions, remains uncertain (5–7). It has been reported that human papillomavirus 16 and Helicobacter pylori may be risk factors for laryngeal carcinoma (8, 9).

An emerging concept in the understanding of pathogenesis is that microbial communities may act as a unit and impact the outcomes of multiple diseases (10). Each person is an assemblage of both human cells and a huge amount of symbiotic species, and these microbial members play critical roles in the maintenance of human health (11). Recently, a number of diseases have been found to be linked to alterations in host-associated microbial communities, including obesity, malnutrition, and disorders of the skin, mouth, and alimentary system (12–17). The human oral cavity, throat, and respiratory tract are connected anatomical sites. These important and distinct ecological niches are colonized by numerous microorganisms. The dominant members in the oral cavity are the Firmicutes, Fusobacteria, and Bacteroidetes (18), and the major bacteria in the lung are Pseudomonas aeruginosa, Burkholderia cepacia complex, and Achromobacter xylosoxidans (19). In our previous pilot study, we explored partial profiles of microbiota in the microenvironment of the larynx (20); however, the biogeography of bacterial communities across the throat and the differences between laryngeal cancer patients and the healthy population have not been thoroughly investigated.

The purposes of this study were to determine the spatial structure of microbial communities in the human throat and to identify specific bacterial communities that may be associated with laryngeal cancer. We applied a high-throughput barcoded 16S rRNA gene pyrosequencing approach to fully characterize the composition and structure of microbiota in the throats of 27 laryngeal cancer patients and 28 control subjects.

MATERIALS AND METHODS

Study design and subjects.

A cross-sectional clinical study was conducted at the Eye, Ear, Nose, and Throat Hospital of Fudan University (Shanghai, China). Twenty-seven patients with laryngeal squamous cell carcinoma (LSCC) and 28 control subjects were enrolled from January 2011 to May 2012. In the laryngeal cancer group, 27 patients (25 males and 2 females) had histopathological confirmation of LSCC and underwent total laryngectomy. Ten of 27 LSCC patients were 60 years of age or younger, and the other 17 patients were >60 years of age. Tumor location and tumor stage were assigned according to the International Union Against Cancer TNM classification system (21). Among of them, glottic tumor was the most common type (16 patients), followed by supraglottic tumor (11 patients). None of patients had subglottic tumor. Eleven patients were classified as having early-stage disease (T1 and T2), while 16 patients had advanced laryngeal cancer (T3 and T4). For the control group, patients (24 males and 4 females) who were diagnosed with vocal cord polyps were selected as controls. Twelve of 28 subjects were 60 years of age or younger, and the other 16 subjects were >60 years of age. These subjects were underwent laryngoscopy surgery. Although they were not healthy, these subjects can be considered meaningful controls (22, 23). Prior to sample collection, exclusion criteria for both LSCC patients and control subjects were as follows: (i) use of hormones or antibiotics during the previous 3 months, (ii) an active bacterial or viral infection in another part of the body, or (iii) seropositivity for human immunodeficiency virus (8). The study protocol and informed consent documents were reviewed and approved by the Ethics Committee of the Eye, Ear, Nose, and Throat Hospital of Fudan University (Shanghai, China). All participants were informed and signed a written consent form according to the committee's regulations.

Sample collection and procedures.

In this study, swab and tissue samples were collected simultaneously from each enrolled subject in a laminar flow operation room to avoid contamination. Swab samples were taken from an anatomical site in the upper portion of the throat, near the epiglottis region (site I). These samples were collected according to Human Microbiome Project (HMP) protocols (http://hmpdacc.org/doc/HMP_Clinical_Protocol.pdf and http://www.ncbi.nlm.nih.gov/projects/gap/cgi-bin/GetPdf.cgi?id=phd003190.2). Tissue samples included tumor tissue and normal tissue adjacent to the tumor site from laryngeal cancer patients and control tissue from patients with vocal cord polyps. These tissue samples were collected from site II, which is below the epiglottis in the throat (including the vestibulum laryngis, tuberculum epiglotticum, vestibular fold, glottis, and subglottis). The tumor tissue samples were confirmed to be cancer tissues by histopathology. The normal tissues adjacent to the tumor sites were taken at least one centimeter away from the tumor sites (negative margins) of LSCC patients and were confirmed to be cancer free. The control mucosa tissue samples taken from the patients with vocal cord polyps were also confirmed to be cancer free. All of the swab and tissue specimens were stored separately at −80°C prior to further analyses.

The collected samples were divided into five groups: con-I, con-II, LC-III, LC-IV, and LC-V (Table 1). The samples from control subjects were assigned to group con-I, consisting of swab samples, and group con-II, consisting of mucosa tissue samples. The samples from patients with laryngeal cancer were assigned to group LC-III for swab samples, group LC-IV for tumor tissues from tumor sites, and group LC-V for normal tissues collected from sites adjacent to the tumor sites.

TABLE 1.

Five groups of samples from two anatomic sites of LSCC patients and control subjects

Anatomic sitec	No. of samples
	LSCCa	Controlb	Total
Site I
con-I		28	28
LC-III	27		27
Site II
con-II		28	28
LC-IV	27		27
LC-V	27		27
Total	81	56	137

^aLSCC samples were collected from patients with laryngeal squamous cell carcinoma.

^bControl samples were collected from patients with vocal cord polyps.

^cSite I, upper portion of throat near the epiglottis; site II, site below the epiglottis in the throat, including the vestibulum laryngis, tuberculum epiglotticum, vestibular fold, glottis, and subglottis; con-I, swab samples from control subjects; LC-III, swab samples from LSCC patients; con-II, mucosa tissue samples from control subjects; LC-IV, tumor tissues from tumor sites in LSCC patients; LC-V, normal tissues adjacent to tumor sites in LSCC patients.

DNA isolation and 16S rRNA gene amplification.

Whole genomic DNA was extracted from the samples with a DNeasy kit (catalog no. 69504; Qiagen, Germany), according to the manufacturer's instructions. A detailed version of the protocol was described previously (24). A lytic enzyme cocktail (Sigma-Aldrich, USA) mix and 0.1-mm-diameter stainless zirconia/silica beads (BioSpec, OK) were used based on the extraction protocol to maximize the quantity of DNA. Briefly, genomic DNA was extracted from swab samples and tissue samples separately. The swabs were thawed on ice before analysis and vortexed vigorously for 5 min to resuspend the cells. A total of 1,000 μl of Tris-EDTA (TE) buffer (pH 8.0) was added, and 500 μl of each sample was transferred into a bead-beating tube and kept on ice. The enzyme cocktail was added to each sample and 41 μl of TE buffer (pH 8.0) for a final volume of 100 μl per sample. Samples were digested by incubation at 37°C for 1 h. After the beads were added to each sample, bead beating was performed for 1 min at 2,100 rpm by using a Mini-Beadbeater-96 instrument (BioSpec, OK). Aliquots of crude lysates from each sample were transferred into new, sterile microcentrifuge tubes, and 50 μl of proteinase K and 500 μl of buffer AL were added. Samples were mixed by vortexing for 15 s and then incubated at 56°C for 10 min. After this step, 50 μl of 3 M sodium acetate (pH 5.5) and 500 μl 100% ethanol were added to each sample. The tissue samples were thawed on ice for 5 min, 200 μl of buffer ATL was added to the tube, and the samples were vortexed for 15 s. Twenty microliters of proteinase K was added to the tube, and the tube was vortexed well. After incubation of the sample at 56°C overnight, 59 μl of enzyme cocktail and 41 μl of TE buffer were added to each sample. The beads were added to each sample, and bead beating was performed. Five hundred microliters of AL buffer was added, and the sample was vortexed for 15 s and then incubated for 10 min at 70°C. The supernatant was transferred into a new 1.5-ml microcentrifuge tube, leaving the beads behind. Five hundred microliters of 100% ethanol was added to the tube, and the tube was vortexed for 15 s. Purification of genomic DNA was done with the Qiagen DNeasy kit according to the manufacturer's instructions (25). Purified DNA was stored temporarily at −80°C until further use.

Universal primers (Escherichia coli primer pair 27F-534R) were used for PCR amplification of the V1 to V3 hypervariable regions of the microbial 16S rRNA genes (Sangon, Shanghai, China) (20). PCR was performed under the following conditions: 95°C of denaturation for 2 min, followed by 30 cycles of 95°C for 20 s (denaturing), 56°C for 30 s (annealing), and 72°C for 5 min (elongation).

Sequence processing and data analyses.

Amplicon pyrosequencing was performed according to standard 454/Roche GS-FLX protocols. DNA samples were transformed into single-stranded template DNA (sstDNA) libraries according to the instructions of the GS DNA Library Preparation kit (Roche Applied Science). Libraries of sstDNA were clonally amplified in a bead-immobilized form with the GS emPCR kit (Roche Applied Science) and sequenced with a 454 Genome Sequencer FLX Titanium platform. Pyrosequencing was conducted at the Chinese National Human Genome Sequencing Center (Shanghai, China). The sequencing reads in this study included two portions that were from the swab samples and the tissue samples. The data for the tissue samples partially included previously reported raw data (20).

Only sequence reads that reached the following quality standards were analyzed further: (i) barcode sequences that were totally matched, (ii) sequences that were at least 200 bp in length, (iii) sequences with an average quality score of ≥20, (iv) sequences with a maximum of 2 matching errors to forward primer sequences, (v) sequences that had <2 ambiguous bases, and (vi) sequences that had a <7-bp homopolymer run (20, 25). A strategy to reduce noise was carried out by using mothur, based on the PyroNoise algorithm reported previously by Quince et al. (26).

The relative abundances of bacterial taxa in each sample between the laryngeal carcinoma and control groups were calculated and analyzed by using the Mann-Whitney test or Student's t test adjusted by the false discovery rate (FDR) (27). The similarity of operational taxonomic units (OTUs) determined by using UCLUST software was 97%. Sequence assignment was based on Ribosomal Database Project (RDP) criteria (28, 29), and the sequences were assigned to the hierarchical taxa at a cutoff of 80%. Principal coordinate analysis (PCoA) was conducted and analyzed with UniFrac software, and principal coordinates were visualized with R software and the heatmap.2 package. Based on UniFrac sample distances, the P value for statistical significance was computed with the R package CrossMatch. The OTU percentage matrix was scaled to a mean equal to zero and the variance unit. The sample distance was estimated with the R package vegan. Redundancy analysis (RDA) was applied to visualize the different bacterial communities of the laryngeal cancer and control groups based on the relative abundance of OTUs. OTUs that had at least 10% variability in their values were explained by the canonical axis, and the P value was analyzed by the Mann-Whitney test (27).

Nucleotide sequence accession number.

All of the sequences reported in this paper have been deposited in the GenBank Sequence Read Archive (accession no. SRP032176).

RESULTS

Overall sequencing output.

Twenty-seven LSCC patients and 28 control subjects were enrolled in this study. Five groups were assigned to the selected swab and tissue samples for these subjects (Table 1). The clinical characteristics of these cases are presented in Table S1 in the supplemental material. In total, 137 samples were collected from the enrolled subjects. After filtering of the sequences using read depth and quality requirements, we obtained complete variable regions 1 and 2 (V1-V2) of 16S rRNA genes. About 84,207 reads were detected for the control group (con-I and con-II), and 148,590 reads were acquired from the laryngeal cancer group (LC-III, LC-IV, and LC-V). After determining the OTUs at the level of 97% similarity, 7,943 OTUs were identified in the throats of the entire enrolled population. This data set consisted of high-quality sequences with an average length of 320 bp and 1,699 reads per sample, which were classified according to RDP criteria.

Microbiota composition and abundance in the throat.

Overall, there were 14 phyla, 20 classes, 38 orders, 85 families, and 218 genera identified in the throats of enrolled subjects. The phyla in the throat whose proportions were >0.01 (1%) were as follows: Firmicutes (54.7%), Fusobacteria (14.8%), Bacteroidetes (12.7%), Proteobacteria (10.6%), and Actinobacteria (6.2%). The genera whose proportions were >0.01 were Streptococcus (37.3%), Fusobacterium (11.3%), Prevotella (10.6%), Neisseria (7.0%), Veillonella (3.7%), Gemella (3.6%), Actinomyces (3.0%), Leptotrichia (3.0%), Rothia (2.7%), Parvimonas (2.2%), Peptostreptococcus (1.7%), Granulicatella (1.6%), Haemophilus (1.2%), and Porphyromonas (1.0%) (Fig. 1). A summary of the data, including the classes, orders, and families whose percentages were >0.01 in these communities, is presented in Fig. S1 in the supplemental material. It should be noted that the principal compositions of the bacterial communities in the throat were similar to those of the oral microbiota; however, the relative abundances of these species in the oral cavity were substantially different from those in the throat (30, 31).

FIG 1.

Relative abundances of bacterial communities in the human throat. (A) The main bacterial communities in the throat at the phylum level; (B) the major bacterial communities in the throat at the genus level.

Spatial profiles of microbiota in two different sites of the throat.

We characterized the spatial structure of microbial communities in two anatomic sites of the throat. There were 11 phyla, 16 classes, 28 orders, 59 families, 106 genera, and 2,075 OTUs of bacterial communities shared between site I and site II (see Fig. S2 in the supplemental material). However, the spatial structure of microbial communities in site I was significantly different from that in site II (P < 0.001) by PCoA (Fig. 2A). Moreover, we investigated the microbiota profiles in the laryngeal cancer patient and control subject groups separately in site I and site II. The bacterial community compositions and abundances were considerably discordant between site I and site II in LSCC patients (Fig. 2B) and in control subjects (Fig. 2C) by PCoA (P < 0.001 for both).

FIG 2.

Spatial diversities of microbial communities estimated by PCoA. The OTUs were determined for each sample, and the corresponding data matrix was subjected to PCoA. Each point is representative of a single sample and is colored according to disease status. The plot of the principal coordinate scores was generated by using UniFrac metrics. Based on UniFrac sample distance, the P value for statistical significance was computed by using the R package CrossMatch. Site I was in the upper portion of the throat near the epiglottis region, and site II was below the epiglottis portion in the throat. (A) In the enrolled population, the structures of the bacterial community in site I were significantly different from those in site II (P < 0.001). (B) The microbial structures are discordant in the samples from the two sites of laryngeal cancer subjects (P < 0.001). (C) Profiles of the bacterial community altered between the two sites in control subjects (P < 0.001).

The characteristics of microbiota in site I and site II were depicted at the levels of phylum and genus. The prevailing phyla in site I were as follows: Firmicutes (56.6%), Actinobacteria (11.9%), Fusobacteria (11.1%), Proteobacteria (10.4%), and Bacteroidetes (9.5%). The prevalent genera in site I were Streptococcus (38.8%), Neisseria (9.2%), Prevotella (8.1%), Veillonella (7.3%), Actinomyces (7.0%), Leptotrichia (6.4%), Fusobacterium (4.7%), Rothia (4.4%), Gemella (3.4%), and Granulicatella (2.8%). The major phyla in site II were as follows: Firmicutes (53.4%), Fusobacteria (17.2%), Bacteroidetes (14.8%), Proteobacteria (10.8%), and Actinobacteria (2.5%). The main genera in site II were Streptococcus (36.3%), Fusobacterium (15.8%), Prevotella (12.4%), Neisseria (5.6%), Gemella (3.7%), Parvimonas (3.6%), Peptostreptococcus (2.7%), Haemophilus (1.8%), Rothia (1.5%), Porphyromonas (1.3%), and Veillonella (1.2%). Based on bacterial composition and abundance, the complete linkage clustering heat map and the matrix heat map were combined for a correlogram analysis of the two different sites for all subjects (Fig. 3 and 4; see also Fig. S3 and S4 in the supplemental material). The relative abundances of six phyla and 23 genera in site I and site II were significantly different from those in the enrolled subjects. Among these bacterial communities, some with proportions of >1% were specified and presented. The phylum Actinobacteria was more prevalent in site I, but the phyla Fusobacteria and Bacteroidetes were more abundant in site II. The genera Veillonella, Leptotrichia, Actinomyces, Rothia, and Granulicatella were more prevalent in site I. However, the genera Fusobacterium, Prevotella, Parvimonas, Peptostreptococcus, and Porphyromonas were more prevalent in site II (see Table S2 in the supplemental material).

FIG 3.

Correlogram depicting the main abundant bacterial taxa in control population. This heat map summarizes the relative abundances of bacterial genera in subjects with vocal cord polyps and complete linkage clustering of samples based on the genus composition and abundance in communities. Group con-I consisted of swab samples from the upper throat (site I), and group con-II consisted of tissue samples of vocal cord polyps (site II). The bar in the top left corner indicates colors corresponding to the relative abundances of genera per sample.

FIG 4.

Correlogram presenting the main abundant bacterial taxa in laryngeal cancer patients. This heat map summarizes the relative abundances of bacterial genera in LSCC subjects and complete linkage clustering of samples based on genus composition and abundance in communities. Group LC-III consisted of swab samples from the upper throat of LSCC subjects (site I). Group LC-IV consisted of tumor tissues at tumor sites (site II), and group LC-V consisted of normal tissues adjacent to tumor sites (site II). The bar in the top left corner indicates colors corresponding to the relative abundances of genera of per sample.

Comparison of bacterial communities in LSCC patients and control subjects.

We compared the bacterial communities in LSCC patients and control subjects at the phylum, genus, and species levels. Three phyla and 12 genera differed in abundance between the laryngeal cancer patients and control subjects based on analyses by Student's t test and the FDR method (P < 0.05) (Table 2). To further compare the biodiversities of the laryngeal cancer patients and control subjects, RDA was applied to investigate the bacterial communities based on the relative abundance of OTUs (27). We found that the relative abundances of eight species varied between the cancer patients and control subjects (see Table S3 in the supplemental material).

TABLE 2.

Differences in bacterial communities in the throats of LSCC patients and control subjects

Bacterial taxon	% of reads for group		P valuec	Corrected P valued
	LSCC subjectsa	Controlsb
Phyla
Fusobacteria	17.8	10.3	0.007	0.023
Actinobacteria	4.6	8.6	0.006	0.023
Spirochaetes	0.3	0.1	0.006	0.023
Genera
Streptococcus	31.4	46	0.003	0.025
Fusobacterium	14.5	6.8	0.004	0.024
Prevotella	12.7	7.6	0.003	0.024
Parvimonas	3.2	0.8	0.001	0.012
Peptostreptococcus	2.6	0.5	<0.001	0.008
Rothia	1.7	4.1	0.01	0.044
Dialister	0.7	0.1	0.012	0.047
Catonella	0.5	0.1	0.001	0.012
Selenomonas	0.4	0.1	0.003	0.024
Peptostreptococcaceae incertae sedis	0.4	0.03	0.002	0.024
Treponema	0.3	0.1	0.006	0.032
Burkholderia	0.02	0.001	0.007	0.033

^aLSCC subjects were patients with laryngeal squamous cell carcinoma.

^bControls were patients with vocal cord polyps.

^cP values were determined by Student's t test.

^dCorrected P values were adjusted by the false discovery rate (FDR) method.

The composition and structure of bacterial communities of laryngeal cancer patients were different from those of control subjects. The phyla Fusobacteria and Spirochaetes were more prevalent in the throats of LSCC patients, but the phylum Actinobacteria was rare in LSCC patients (P < 0.05). The relative abundances of Fusobacterium, Prevotella, Parvimonas, Peptostreptococcus, Dialister, Catonella, Selenomonas, Peptostreptococcaceae incertae sedis, Treponema, and Burkholderia increased in laryngeal cancer patients (P < 0.05). However, the relative proportions of Streptococcus and Rothia decreased in LSCC patients (P < 0.05) (Table 2).

The abundances of eight species were altered considerably in LSCC patients and control subjects by RDA. The abundances Streptococcus, Fusobacterium, and Prevotella as key species changed between the two groups of subjects, which were well fitted by the sample scores on the canonical axis. Fusobacterium nucleatum, Fusobacterium sp. oral taxon, Prevotella intermedia, Prevotella tannerae, Prevotella oris, and Prevotella nigrescens were more prevalent in the throats of patients with laryngeal cancer; these bacteria may be risk factors in laryngeal cancer. However, the relative abundances of Streptococcus parasanguinis and the Streptococcus sp. oral taxon were reduced in LSCC patients (P < 0.05) (see Table S3 in the supplemental material).

Furthermore, we compared the compositions and structures of microbial communities of three tissue groups (con-II, LC-IV, and LC-V) from site II of control subjects and LSCC patients. Streptococcus had lower abundances in groups LC-IV and LC-V than in the con-II group (P = 0.005) (see Tables S4 and S5 in the supplemental material). Prevotella, Peptostreptococcus, and Catonella had higher proportions in group LC-V than in the con-II group (P < 0.05) (see Table S5 in the supplemental material). However, no discordance was found between groups LC-IV and LC-V (P > 0.05). Thus, the distributions of bacterial populations in group con-II varied compared to those in groups LC-IV and LC-V in site II.

DISCUSSION

In this study, next-generation DNA sequencing technology was exploited to characterize the composition and structure of microbiota in the throat. The results demonstrated that the throat harbors a vast array of bacterial communities. The rank abundances of bacterial species in communities differed widely between site I and site II, even though the two sites were anatomically connected. The structures of throat microbial communities varied between LSCC patients and control subjects, and several bacteria were associated with laryngeal cancer.

The microbial community in the throat exhibited a discordant structure compared with those in the oral cavity and the lung. The HMP investigated the microbiome of the oral cavity, containing 9 different sites, which included the throat (32). The results revealed that the predominant phylum in the oral cavity is Firmicutes (57%), followed by Actinobacteria (20%), Bacteroidetes (14%), and Proteobacteria (7%). The main genera were Streptococcus (34%), Rothia (12%), Veillonella (11%), Prevotella (10.5%), Gemella (6.9%), and Actinomyces (6.7%) (31). Our findings indicated that the relative proportions of the phyla Fusobacteria, Actinobacteria, and Proteobacteria varied at sites of the oral cavity and the throat. The relative abundances of the genera Veillonella, Gemella, Actinomyces, Rothia, Neisseria, Peptostreptococcus, Leptotrichia, Lactobacillus, and Porphyromonas in the oral cavity were discordant from those in the throat. We also recognized that the microbiota structures of the throat were highly discordant from those of the ecological niche of lung. A wide range of communities in the throat that were detected in our study were not found in the lung; the vast majority of communities in the lung are P. aeruginosa (76.5%), A. xylosoxidans (9.2%), and B. cepacia complex (7.6%) communities (19). Besides, our previous pilot study examined the bacterial structures in the larynx, but the characteristics of the closely related microenvironments in the throat were unknown (20). In this study, we optimized the methods of 16S RNA gene processing and statistical analyses to fully investigate the microbiota at a site near the epiglottis and at a site below the epiglottis in the throat. We highlighted that the relative abundances of communities in the two close sites of throat were different. These patterns of multiple microbes showing a high degree of variation probably suggest that disparate anatomic locations possess their own specific bioecological characteristics (32, 33).

The bacterial structures in the throats of laryngeal cancer patients were significantly different from those in the throats of control subjects. In this study, we demonstrated that the relative abundances of Fusobacterium and Prevotella increased in the throats of LSCC patients, but the relative abundance of Streptococcus decreased. Fusobacterium is a proinflammatory pathogen that can induce inflammation and promotes tumorigenesis by modulating the tumor immune microenvironment via the expansion of CD11b⁺ cells (34). A high abundance of Fusobacterium was found in patients with colorectal adenomas, colorectal cancer, and pancreatic cancer (35–39). The potential tumor-promoting mechanisms of Fusobacterium nucleatum are that it attaches to the cell surface by FadA binding to E-cadherin and decreases phosphorylation of β-catenin. β-Catenin then translocates into the nucleus and stimulates the lymphocyte enhancer factor (LEF), T-cell factor (TCF), and NF-κB oncogenes. This bacterium invades the cell via internalization of E-cadherin by clathrin; stimulates the interleukin-10 (IL-10), IL-8, IL-6, IL-1β, tumor necrosis factor alpha (TNF-α), and COX-2 inflammatory genes; and ultimately induces tumorigenesis (40, 41). Prevotella can adhere to epithelial cell lamellipodia and construct bacterial biofilm via its special affinity (42). This bacterium has a strong effect on shaping the host's metabolic and immune system activities (40). Highly elevated Prevotella levels in colorectal cancer patients, which appeared to induce an elevation in the level of mucosal regulatory T cells that overproduce IL-17 in the mucosa of cancer patients, were reported previously (43). Streptococcus species recognize receptors of α-amylase, proline-rich protein, and polysaccharides in the host and bind to other species via the autoinducer 2 (AI2) receptor. These bound communities then build a microbial biofilm through CiaRH, PsrP, and SpxB (44, 45). By generating the products of saccharolytic short-chain organic acid from carbohydrates, streptococcal species decrease the pH and change the local environment (46). In addition, Streptococcus, as an important constructive player, can interact with Fusobacterium and other microbes to develop a balanced microbial community. This behavior is conducted via repressing the phosphoenolpyruvate-dependent phosphotransferase system (PTS) and increasing oxidative stress protection (46). Therefore, these bacterial communities may be risk factors for induction of carcinoma or protective factors in defense against tumor via these mechanisms.

These microbial populations may be involved in cancer disease as a complex assemblage. The current comprehensive understanding of the pathogenesis of several human diseases is that the microbial community is a unit of pathogenicity (10, 47). Microorganisms collectively construct a mixed community, and this community is considered to be an integrated assemblage of populations that coexist and interplay in a given niche. Each unit occupies an effective role that contributes to the overall community and maintains the ecological balance of the biota. Multiple bacterial combinations can play a role in disease causation, suggesting that the outcome of disease could be the consequence of the collaborative activities of the polymicrobial community (48, 49). Therefore, based on the community-as-pathogen hypothesis, several bacterial communities may be more related to certain forms of the disease (10, 47, 49). According to previous research and our present findings, we postulated that microorganisms aggregate and colonize on the mucosa of the throat and that the microbiota construct bacterial biofilms on the surface of the epithelial cells. Highly structured and spatially arranged microbial communities interact with host cells, and each species plays a specific role in maintaining the balance of this ecological niche. Altered compositions and abundances of microbial communities may be associated with laryngeal carcinoma.

In sum, this study explored the microbial profiles in the throats of laryngeal cancer patients and control subjects. Future studies are required to determine the roles of these bacterial communities in the development of laryngeal cancer.