Abstract
Free full text
Discovery and validation of potential bacterial biomarkers for lung cancer
Abstract
Microbes are residents in a number of body sites, including the oral and nasal cavities, which are connected to the lung via the pharynx. The associations between oral diseases and increased risk of lung cancer have been reported in previous prospective studies. In this study, we measured variations of salivary microbiota and evaluated their potential association with lung cancer, including squamous cell carcinoma (SCC) and adenocarcinoma (AC). A three-phase study was performed: First, we investigated the salivary microbiota from 20 lung cancer patients (10 SCC and 10 AC) and control subjects (n=10) using a deep sequencing analysis. Salivary Capnocytophaga, Selenomonas, Veillonella and Neisseria were found to be significantly altered in patients with SCC and AC when compared to that in control subjects. Second, we confirmed the significant changes of Capnocytophaga, Veillonella and Neisseria in the same lung cancer patients using quantitative PCR (qPCR). Finally, these bacterial species were further validated on new patient/control cohorts (n=56) with qPCR. The combination of two bacterial biomarkers, Capnocytophaga and Veillonella, yielded a receiver operating characteristic (ROC) value of 0.86 with an 84.6% sensitivity and 86.7% specificity in distinguishing patients with SCC from control subjects and a ROC value of 0.80 with a 78.6% sensitivity and 80.0% specificity in distinguishing patients with AC from control subjects. In conclusion, we have for the first time demonstrated the association of saliva microbiota with lung cancer. Particularly, the combination of the 16S sequencing discovery with qPCR validation studies revealed that the levels of Capnocytophaga and Veillonella were significantly higher in the saliva from lung cancer patients, which may serve as potential biomarkers for the disease detection/classification.
Introduction
Lung cancer is considered a terminal illness, with a five-year survival rate of about 11% [1]. It is the most common cause of cancer-related deaths in North America and worldwide [2,3]. Tobacco smoking represents one of the main risk factors of lung cancer, since tobacco contains carcinogens that may induce cell transformation [4]. However, studies have also demonstrated that lung cancer involves immune responses [5], viral infections [6], and other factors that are not related to tobacco [6,7].
Recent studies have reported the identification of potential biomarkers in saliva, including oral bacteria, for cancer detection [8-13]. Particularly, salivary microbiota has been associated with oral squamous cell carcinoma and pancreatic ductal adenocarcinoma [12,13], suggesting that salivary microbiota may serve as an informative source for discovering non-invasive biomarkers of cancer diseases. Earlier studies have also shown serological evidence of an association between Chlamydia pneumoniae infection and lung cancer, and Chlamydia pneumoniae infection may serve as a biomarker for increased risk of lung cancer [14,15]. Most salivary bacteria are derived from the oral cavity, although some may also arise from the esophagus and the upper respiratory tract. Survival and growth of bacteria in the oral cavity are dependent on the oral environment, which can be affected by the composition of the sputum, and habits such as cigarette smoking. These factors are related to lung cancer occurrence and progression.
Based on these considerations, as well as previously reported involvement of bacteria in lung cancer tumorigenesis [14-16], we investigated the global variations of salivary microbiota in lung cancer patients. We quantified bacterial flora composition of saliva samples obtained from lung cancer patients and controls by sequencing V3 and V6 of 16S rDNA, using the Illumina HiSeq 2000 sequencer. This approach can identify more than 500 prevalent human bacterial species, with up to 100,000 bacterial sequences per sample. This high sensitivity and comprehensive analyses provide a new approach to investigating the relationship between lung cancer and bacterial composition.
Non-small cell lung cancer (NSCLC) accounts for more than 80% of lung cancer cases, most of which are either squamous cell carcinoma (SCC, ~30-35%) or adenocarcinoma (AC, ~50%). Therefore, we mainly enrolled patients with SCC or AC for the present study. To discover candidate bacterial biomarkers for SCC or AC, we performed the sequencing analysis of 10 samples from SCC patients, 10 samples from patients with AC and 10 samples from control subjects (n=30), and identified bacteria at the genus level that showed significant differences between both lung cancer groups and the control group. The results were subsequently confirmed using qPCR to quantify the relative abundance of bacteria. Finally, we further validated these potential bacterial biomarkers on new patient/control cohorts including a total of 41 lung cancer patients and 15 contro
Materials and methods
Sample collection
The study was approved by the Medical Ethics Committee at the Changzhou Second People’s Hospital (CSPH), Nanjing Medical University, and all saliva samples were collected according to the approved protocol. Written informed consent was obtained from all participants. Since smoking may affect the growth of bacteria in the oral cavity, only subjects with a smoking history of over 10 years were selected. Saliva samples were obtained after cancer detection and before treatment. None of the subjects, including the patients and the controls, manifested other diseases related to salivary bacteria, such as diabetes, immune dysfunction, herpes viral infections, or oral mucosal ulcers. Patient information is summarized in Table 1.
Table 1
Group | Subjects under study | Tumor stage | |||||
---|---|---|---|---|---|---|---|
| |||||||
n | Male | Age | T4 | T3 | T2 | ||
Discovery | SCC | 10 | 9 | 67.4±8.99 | 6 | 3 | 1 |
AC | 10 | 8 | 61.3±15.13 | 7 | 3 | 0 | |
Controls | 10 | 8 | 59.20±9.84 | N/A | |||
Validation | SCC | 13 | 12 | 69.8±6.03 | 12 | 1 | 0 |
AC | 28 | 24 | 63.8±11.78 | 24 | 2 | 2 | |
Controls | 15 | 10 | 57.5±9.01 | N/A |
Sequencing analysis of bacterial flora of saliva samples
DNA content in saliva was isolated by SDS lysis, following phenol extraction to remove proteins in saliva. The DNA content of samples recovered by ethanol precipitation was amplified using PCR, as previously described [17]. The amplified samples were sequenced using the Illumina HiSeq 2000, and V3 and V6 of 16S rDNA were analyzed. To ensure high quality readings, we adopted stringent conditions to process the sample sequences. In brief, the following processing steps were performed: 1) all readings were assigned to corresponding samples by allowing one mismatch to the sample barcode and two mismatches to the adjacent PCR primer; 2) the readings were subsequently de-noised by the PyroNoise algorithm [18]; 3) readings containing ambiguous nucleotides or a homopolymer longer than 8 base pairs (bp) were removed, as were sequences shorter than 200 bp or longer than 1000 bp; 4) the readings were aligned using a nearest alignment space termination (NAST)-based sequence aligner to a custom reference based on the SILVA alignment [19], and sequences that did not align to the anticipated region of the reference alignment were discarded; 5) chimeric sequences identified by the UCHIME algorithm were removed [20]; and 6) readings were classified using a Bayesian classifier with the Ribosomal Database Project (RDP). Sequences of mitochondria or unknowns (those readings that could not be classified at the kingdom level) were removed. Finally, all of the effective readings were clustered into operational taxonomic units (OTUs) based on 97% sequence similarity, using the mothur program [21]. The taxonomy profiling of samples at different taxonomic levels (phylum, class, order, family, genus) were created using QIIME [22].
Quantitative PCR
The candidate bacterial biomarkers identified from the sequencing analysis were further confirmed by qPCR on the same set of saliva samples used in the discovery phase (n=30). Specific PCR primers to detect 16S rDNA were designed and shown in Table 2. These primers were chosen from previously published literature [23] or by searching the RDP database. Verified potential biomarkers were further validated on a new patient (n=41) and control (n=15) cohort using qPCR.
Table 2
Name | Taxonomy | Sequences |
---|---|---|
Bacteroidales | Class | F WTYATTGGGTTTAAAGGG |
R GGTAAGGTTCCTCGCGTA | ||
Neisseria | Genus | F CTGTTGGGCARCWTGAYTGC |
R GATCGGTTTTRTGAGATTGG | ||
Capnocytophaga | Genus | F TGGWCAATGGTCGGAAGACTG |
R CCGCTACACTACACATTCCA | ||
Selenomonas | Genus | F ACRCGTAGRCAACCTGCCG |
R CGATCCGAAGACCTTCTTCAC | ||
Veillonella | Genus | F CGGGTGAGTAACGCGTAATCA |
R CCAACTAGCTGATGGGACGC | ||
Common | For all bacteria | F ATTAGATACCCTGGTAGTCC |
R CCCCGTCAATTCATTTGAGT |
Statistical analyses
Data analysis was performed with the GraphPad Prism and MedCalc software. Receiver operating characteristic (ROC) analysis was used to estimate the performance of potential biomarkers and evaluate if combining biomarkers improves sensitivity and specificity of disease detection.
Results
Differential salivary microflora profiles between lung cancer and control subjects
Our deep sequencing analysis indicated that the abundance of salivary bacteria is different between lung cancer and control groups. This can be seen at all taxonomic levels (Figures 1, ,22 and and3,3, Supplemental Figures 1, 2, 3 and 4). At the order-level, the most significantly altered ones included Flavobacteriales, Burkholderiales, Campylobacterales, Spirochaetales (more abundant in lung cancer) and Bacteroidales (less abundant in lung cancer) (Table 3). At the family-level, Veillonellaceae is significantly over-expressed whereas Lachnospiraceae is under-expressed in lung cancer patients (Table 4). More importantly, at the genus level, Capnocytophaga, Selenomonas, and Veillonella were found to be more abundant in both SCC and AC patients whereas Neisseria was less abundant in both SCC and AC patients than the controls (Table 5). On the other hand, the levels of Steptococcus, and Porphyromonas were only significantly lower in patients with SCC and Prevotella was not significantly changed in either cancer groups. All the sequencing results at different taxonomy levels (Phylums, Class, Order, Family and Genus) are shown in Supplemental Figures 1, 2, 3 and 4.
Table 3
Taxonomy/Order | Fold change (SCC vs CTRL) | P value (SCC vs CTRL) | Fold change (AC vs CTRL) | P value (AC vs CTRL) |
---|---|---|---|---|
Bacteroidales | 0.63 | 0.03 | 0.52 | 0.006 |
Flavobacteriales | 2.58 | 0.02 | 2.81 | 0.009 |
Fusobacteriales | 1.1 | 0.39 | 0.79 | 0.19 |
Burkholderiales | 9.01 | 0.02 | 4.76 | 0.03 |
Neisseriales | 0.49 | 0.02 | 0.76 | 0.17 |
Campylobacterales | 2.71 | 0.04 | 1.32 | 0.3 |
Pasteurellales | 0.81 | 0.36 | 1.16 | 0.38 |
Spirochaetales | 5.76 | 0.08 | 3.88 | 0.12 |
Clostridiales | 2.41 | 0.0002 | 1.66 | 0.01 |
Table 4
Taxonomy/Family | Fold change (SCC vs CTRL) | P value (SCC vs CTRL) | Fold change (AC vs CTRL) | P value (AC vs CTRL) |
---|---|---|---|---|
Actinomycetaceae | 0.45 | 0.08 | 0.38 | 0.04 |
Coriobacteriaceae | 0.25 | 0.07 | 0.21 | 0.03 |
Flavobacteriaceae | 1.96 | 0.11 | 2.4 | 0.01 |
Streptococcaceae | 0.36 | 0.01 | 0.72 | 0.16 |
Lachnospiraceae | 0.18 | 0.05 | 0.11 | 0.02 |
Veillonellaceae | 2.46 | 0.006 | 1.8 | 0.02 |
Table 5
Taxonomy/Genus | Fold change (SCC vs CTRL) | P value (SCC vs CTRL) | Fold change (AC vs CTRL) | P value (AC vs CTRL) |
---|---|---|---|---|
Porphyromonas | 0.24 | 0.005 | 0.46 | 0.1 |
Prevotella | 0.74 | 0.2 | 0.63 | 0.08 |
Capnocytophaga | 2.22 | 0.04 | 2.42 | 0.03 |
Neisseria | 0.44 | 0.009 | 0.63 | 0.04 |
Streptococcus | 0.36 | 0.007 | 0.72 | 0.16 |
Selenomonas | 34.7 | 0.01 | 10.15 | 0.04 |
Veillonella | 2.24 | 0.008 | 1.85 | 0.02 |
qPCR verification of the candidate bacterial biomarkers
To verify the finding from sequencing analysis, we used qPCR to quantify the levels of the identified bacteria genera in the same cancer and control subjects used for sequencing analysis. A common primer pair that can amplify most bacterial 16S rDNA was used for qPCR and subsequently the qPCR results based on the common primer pair were used to normalize the levels of individual bacterial rDNA in each control and cancer samples (Table 2). qPCR analysis showed significantly (p<0.05) higher levels of Capnocytophaga and Veillonella but lower levels of Neisseria in cancer samples than the control samples (Figure 4, Table 6). This confirms the sequencing analysis results and supports the feasibility of using qPCR to quantify the levels of selected bacterial clusters. Due to its very low level, Selenomonas was not detectable by qPCR in some of saliva samples, so it was not included in subsequent qPCR validation studies on independent patient populations. We also verified the alterations of Bacteroidales in SCC and AC patients with qPCR, and the results were similar to those obtained from the deep sequencing analysis (Figure 4, Table 6).
Table 6
Fold change (SCC vs CTRL) | P value (SCC vs CTRL) | Fold change (AC vs CTRL) | P value (AC vs CTRL) | |
---|---|---|---|---|
Neisseria | 0.4 | 0.004 | 0.55 | 0.04 |
Capnocytophaga | 3.18 | 0.04 | 3.78 | 0.04 |
Veillonella | 2.94 | 0.00006 | 2.39 | 0.0003 |
Bacteroidales | 0.69 | 0.09 | 0.6 | 0.03 |
Validation of potential bacterial biomarkers in new patient cohorts
To validate the three potential bacterial markers (Neisseria, Capnocytophaga and Veillonella), we used qPCR to quantify their levels in a new patient/control cohort, including 41 cancers (13 SCC, 28 AC) and 15 controls. The qPCR results indicated that the levels of Capnocytophaga and Veillonella were significantly higher in both SCC and AC groups, consistent with our qPCR verification analysis of previous 30 samples (Figure 5). However, the validation result of Neisseria was not as promising as those found in the qPCR verification study. Although the levels of Neisseria in the AC patients were significantly lower (p=0.009) than those in the controls, the comparison of SCC with control groups did not show significance difference.
As shown in Figure 6, the ROC analysis was performed to evaluate pre-clinical utility of these potential biomarkers and to assess if combining biomarkers may improve the sensitivity and specificity. The overall performance of the three potential biomarkers in detecting SCC and AC is summarized in Table 7. The ROC values of Veillonella were determined to be 0.81 for SCC and 0.68 for AC, and the ROC values of Capnocytophaga were 0.79 for SCC and 0.81 for AC. ROC value or sensitivity can be improved by combining Veillonella with Capnocytophaga. However, the ROC values significantly decreased when adding the third potential biomarker, Neisseria, for both SCC and AC.
Table 7
p value (SCC vs CTRL) | p value (AC vs CTRL) | ROC value (SCC vs CTRL) | Cut-off (SCC vs CTRL) | Sensitivity (SCC vs CTRL) | Specificity (SCC vs CTRL) | ROC value (AC vs CTRL) | Cut-off (AC vs CTRL) | Sensitivity (AC vs CTRL) | Specificity (AC vs CTRL) | |
---|---|---|---|---|---|---|---|---|---|---|
Neisseria | 0.260 | 0.009 | 0.62 | ≤1.5% | 30.80% | 100% | 0.78 | ≤2.1% | 57.10% | 93.30% |
Capnocytophaga | 0.020 | 0.004 | 0.79 | ≥0.1% | 92.30% | 53.30% | 0.81 | ≥0.7% | 71.40% | 80.00% |
Veillonella | 0.003 | 0.030 | 0.81 | ≥7.1% | 76.90% | 86.70% | 0.68 | ≥8.3% | 42.90% | 93.30% |
Capnocytophaga + Veillonella | - | - | 0.86 | - | 84.60% | 86.70% | 0.80 | - | 78.60% | 80.00% |
Capnocytophaga + Veillonella + Neisseria | - | - | 0.68 | - | 53.80% | 86.70% | 0.54 | - | 35.70% | 86.70% |
Discussion
Our intensive sequencing analysis demonstrated that the salivary microbiota profiles are significant altered in the patients with lung cancer when compared to those from control subjects. Detailed analysis of the deep sequencing results showed that the abundance of salivary bacteria varied between lung cancer and control groups at multiple taxonomic levels. Particularly, at the genus-level, Capnocytophaga, Selenomonas, and Veillonella were found to be more abundant in both SCC and AC patients whereas Neisseria was less abundant in both SCC and AC patients than the controls (Table 5). In addition, the abundance of Steptococcus, and Porphyromonas were both significantly lower in patients with SCC. Based on these results, we performed qPCR verification of Capnocytophaga, Selenomonas, Veillonella, and Neisseria in the same group of SCC and AC patients used for deep sequencing analysis and further validated the significant changes of Capnocytophaga, Veillonella and Neisseria in an independent patient/control cohort.
Based on ROC analysis, Veillonella and Capnocytophaga are more valuable biomarkers than Neisseria because they exhibit higher sensitivity/specificity and their levels are significantly higher in both cancer types than the control subjects (instead of being lower in cancer groups). In fact, if we combine Veillonella with Capnocytophaga, the ROC value reached 0.86 for SCC with a sensitivity of 84.6% and a specificity of 86.7%. Although the ROC value (0.80) became slightly lower than Capnocytophaga itself when combining these two biomarkers for AC, the sensitivity was improved (78.6%) and specificity reached 80%.
The intrinsic relationship between salivary microbiota and lung cancer is currently under investigation. Indeed, previous studies have provided critical insights on how bacteria may be related to abnormal growth of mammalian cells. Bacterium-produced toxins can disturb the cell cycle, resulting in altered cell growth [15,24,25]. This is caused by alterations to the genes that control normal cell division and apoptosis [26,27]. It is well known that gut microbiota is involved in the pathogenesis of inflammatory bowel disease due to the disruption of immune tolerance [28]. Similarly, salivary microbiota may also affect lung cells by inducing long-term immune response.
In our study, although more research is needed to establish the relationship of Neisseria, Steptococcus, and Porphyromonas to lung cancer, our findings clearly show that there is significant association of salivary Capnocytophaga and Veillonella with lung cancer (both SCC and AC). In fact, there have been increasing interests in studying the link between oral bacteria and respiratory infection [29-33]. An association between oral conditions such as periodontal disease and several respiratory conditions has been reported previously [33]. Oral periodontopathic bacteria can be aspirated into the lung to cause aspiration pneumonia. The teeth may also serve as a reservoir for respiratory pathogen colonization and subsequent nosocomial pneumonia. Once established in the mouth, these pathogens may cause lung infection with the following possible mechanisms: 1. aspiration of oral pathogens into the lung to cause infection; 2. periodontal disease-associated enzymes in saliva may modify mucosal surfaces to promote adhesion and colonization by respiratory pathogens, which are then aspirated into the lung; 3. periodontal disease-associated enzymes may destroy salivary pellicles on pathogenic bacteria to hinder their clearance from the mucosal surface; and 4. cytokines originating from periodontal tissues may alter respiratory epithelium to promote infection by respiratory pathogens [33]. In patients with Cystic fibrosis (CF), there is a high prevalence of oropharyngeal anaerobic bacteria in sputum. Recently, Rivas Caldas et al. demonstrated that the same Pseudomonas aeruginosa clonal types were present in both saliva and sputum samples of CF patients. This suggest that the oral cavity is a possible reservoir for lung infection [32]. Haemophilus pittmaniae (H. pittmaniae), a member of the human saliva microbiota, was identified by mass spectrometry and 16S rRNA gene sequencing in the sputum specimen collected from a patient with a massive fibrotic form of siderosis. H. pittmaniae was found to be responsible for the worsening of the patient’s chronic respiratory failure, and the patient’s condition rapidly improved when he was treated with oral amoxicillin, an antibiotic therapy guided by the in vitro susceptibility pattern of the H. pittmaniae isolate. These results suggest that H. pittmaniae is a possible new pathogen responsible for respiratory tract infection in patients with chronic lung diseases [30]. In addition, post-operative pulmonary infection often appears to result from aspiration of pathogens colonizing the oral cavity. In order to investigate if impaired periodontal status and pathogenic oral bacteria significantly contribute to development of aspiration pneumonia following neurosurgical operations, Bagyi et al. compared a matched cohort of 18 patients without postoperative lung complications to a cohort of 5 patients who developed pneumonia within 48 hours after brain surgery. They found that the number and severity of coexisting periodontal diseases were significantly greater in patients with postoperative pneumonia in comparison to the control group. Therefore, dental examination of the salivary microbiota may be needed in order to identify patients at high risk of developing postoperative respiratory infections [29].
Although Capnocytophaga and Veillonella are significantly associated with both SCC and AC, Veillonella appears to be a better biomarker for SCC whereas Capnocytophaga serves as a better biomarker for AC (Table 7). Veillonella was previously isolated from the lower airways of lung cancer patients, suggesting that Veillonella species may be related to lung cancer [34]. Normally found in the oropharyngeal tract, Capnocytophaga species have been reported to be involved in lung cancer, as they have been shown to be involved in the formation of lung abscesses [35] and lower respiratory tract infections [36]. These findings, including ours, seem to suggest that either these bacteria induce long-term immune response/infection to the organ or cancer growth environment favors the growth of these bacteria in the airway or oropharyngeal tracts. As discussed earlier, Neisseria was found at significantly lower levels in both SCC and AC patients than control subjects. It should be noted that similar results were also reported in pancreatic cancer saliva samples [12], suggesting that Neisseria may be related to inhibited growth of cancer cells.
As a summary, our studies have provided insights on the potential role of oral cavity as a reservoir of bacterial pathogens in lung cancer, and the discovery of potential bacterial biomarkers may lead to a non-invasive method for helping detect/classify the disease. Nevertheless, the exact role of these salivary bacteria in lung cancer occurrence and progression is largely unknown and certainly warrants further investigation.
Acknowledgements
We thank Wendy Aft, Elliot Razi, and Ethan Razi for editing. This study was supported by the Changzhou International Science and Technology Collaboration Program (#CZ20110021). The funders have no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
References
Articles from American Journal of Cancer Research are provided here courtesy of e-Century Publishing Corporation
Citations & impact
Impact metrics
Article citations
Unveiling microbial dynamics in lung adenocarcinoma and adjacent nontumor tissues: insights from nicotine exposure and diverse clinical stages via nanopore sequencing technology.
Front Cell Infect Microbiol, 14:1397989, 27 Aug 2024
Cited by: 0 articles | PMID: 39258251
Cancer and the Microbiome of the Human Body.
Nutrients, 16(16):2790, 21 Aug 2024
Cited by: 0 articles | PMID: 39203926 | PMCID: PMC11357655
Review Free full text in Europe PMC
Oncogene-addicted solid tumors and microbiome-lung cancer as a main character: a narrative review.
Transl Lung Cancer Res, 13(8):2050-2066, 06 Aug 2024
Cited by: 0 articles | PMID: 39263011 | PMCID: PMC11384476
Review Free full text in Europe PMC
Chronic periodontitis and risk of lung cancer: a nationwide cohort study.
Front Oncol, 14:1413590, 28 Jun 2024
Cited by: 0 articles | PMID: 39015494 | PMCID: PMC11250509
Saliva‑microbiome‑derived signatures: expected to become a potential biomarker for pulmonary nodules (MCEPN-1).
BMC Microbiol, 24(1):132, 20 Apr 2024
Cited by: 0 articles | PMID: 38643115 | PMCID: PMC11031921
Go to all (176) article citations
Data
Data behind the article
This data has been text mined from the article, or deposited into data resources.
BioStudies: supplemental material and supporting data
Similar Articles
To arrive at the top five similar articles we use a word-weighted algorithm to compare words from the Title and Abstract of each citation.
Variations of oral microbiota are associated with pancreatic diseases including pancreatic cancer.
Gut, 61(4):582-588, 12 Oct 2011
Cited by: 361 articles | PMID: 21994333 | PMCID: PMC3705763
Microbiota Biomarkers for Lung Cancer.
Diagnostics (Basel), 11(3):407, 27 Feb 2021
Cited by: 27 articles | PMID: 33673596 | PMCID: PMC7997424
Alterations of salivary microbial community associated with oropharyngeal and hypopharyngeal squamous cell carcinoma patients.
Arch Microbiol, 202(4):785-805, 12 Dec 2019
Cited by: 20 articles | PMID: 31832691
Oral Microbiome Characteristics in Patients With Autoimmune Hepatitis.
Front Cell Infect Microbiol, 11:656674, 19 May 2021
Cited by: 14 articles | PMID: 34094998 | PMCID: PMC8170700
The oral microbiome and human health.
J Oral Sci, 59(2):201-206, 01 Jan 2017
Cited by: 144 articles | PMID: 28637979
Review