Alteration of Gut Microbiome in Lung Cancer Patients- A Pilot Study

Lung cancer is the leading cause of cancer death. Better understanding of factors and pathways involved in lung cancer is needed to improve diagnose and treatment strategies. Recent studies have provided insights into the possible correlation between intestinal dysbiosis and cancer development. Although the immunological relationship between gut and lung had been suggested by many researches, however, to date, no study had investigated the characterization of gut microbiome in treatment naïve lung cancer patients, whether it is distinct from that of health individuals and contribute to the onset and development of lung cancer remain unclear. In this study, we investigated whether gut microbiome of lung cancer patients (LC, n=28) is altered compare with that of matched healthy individuals (HC, n=19) by high throughout sequencing of the V3-V4 regions of 16S rDNA in their fecal samples. We also identied microbiota signatures specic for different histological types of lung cancer, including squamous cell carcinoma (SCC), adenocarcinoma (ADC) and small cell lung cancer (SCLC). 16S rDNA sequencing results showed that the gut microbiome of lung cancer patients is characterized by decreased relative abundance of Prevotella, and increased bacteria groups such as Actinomyces, and Streptococcus, etc. We also detected an obvious structural shift in gut microbiome between ADC and SCLC patients. However, given the limited number of study cases, the association between microbial dysbiosis and lung cancer is not clearly understood, future studies involving larger cohorts and metagenomics, or metabolomics, may elucidate the correlations between gut microbiota and lung cancer development. which used the nonparametric factorial Kruskal–Wallis and Wilcoxon rank sum test to identify key OTUs for separating different treatment groups at a signicance level of 0.05. Metastats analysis was performed based on the raw count data matrix to nd out the taxa statistically different between HC and LC samples.


Background
Lung cancer is the leading cause of cancer death [1]. There are estimated 1.82 million new cases of lung cancer globally, which constitutes nearly 13% of all newly diagnosed cancer cases annually [2]. More than one-third of lung cancer worldwide occurring in China, where 733,000 new cases of lung cancer are diagnosed, and about 591,000 Chinese people died from it each year [3]. From a histopathological point of view, lung cancer is categorized into two main types: non-small cell lung cancer (NSCLC) and small cell lung cancer (SCLC). NSCLC accounts for more than 80% of lung cancer cases, most of which are either squamous cell carcinoma (SCC) or adenocarcinoma (ADC). In east China, ADC (60.0%), SCC (25.6%), and SCLC (8.5%) are the major histologic subtypes of newly diagnosed lung cancer according to the clinical pro les and trend analysis in a tertiary care hospital during 2011-2015. Better understanding of factors and pathways involved in lung cancer is urgently needed to improve treatment strategies.
As a heterogeneous disease, many factors are involved in the onset and development of lung cancer.
Although smoking is considered as an important factor, only 10-15% of smokers develop cancer [4], which highlights other in uences, such as the involvement of microbial communities. Recent studies showed that changes in the lung microbiome may be relevant for progression and exacerbations in lung cancer [5,6]. By comparison of microbiome in bronchoalveolar lavage uid of patients with lung cancer with benign mass like lesions, Lee et al. found that the genera Veillonella and Megasphaera are more abundant in lung cancer patients, which may serve as potential biomarkers for the disease detection/classi cation [7]. A recent research, in which surgical lung tissue samples were used, concluded that the microbiota of the lung cancer is unique, with the genus Thermus more abundant in tissue from advanced stage (IIIB, IV) patients, while Legionella is higher in patients who develop metastases [8]. These studies provide insights into the possible correlation between microbiota and lung cancer development.
However, lung microbiota does not seem to be the only microbial factor contributing to the development of lung cancer. The in uence of gut microbiota on lung immunity has been vastly explored and several studies have linked changes in the gut microbiome with lung diseases [9,10]. For example, gut microbiota is responsible for inducing lung in ammatory reaction against bacterial challenge and enhancing neutrophils in ltration through TLR4 in mice [11,12]. Antibiotic treatment can cause overgrowth of particular fungal species in the gut and promote allergic airway in ammation via fungiinduced prostaglandin E2 [13]. Conversely, the lung microbiota also in uences the gut microbiota through the blood stream. Acute lung injury (ALI) can disrupt the lung microbiota, induces a transient translocation of bacteria into blood and causes an acute increase of bacterial load in cecum [14]. These studies emphasized the important role of gut-lung axis in development of diseases [15]. However, to date, no study had investigated the characterization of gut microbiome in treatment naïve (newly diagnosed, untreated) lung cancer patients, whether it is distinct from that of health individuals and contribute to the onset and development of lung cancer remain unclear.
In this pilot study, we investigated whether gut microbiome of lung cancer patients (LC, n = 28) is altered compare with that of matched healthy individuals (HC, n = 19) by high throughout sequencing of the V3-V4 regions of 16S rDNA in their fecal samples. We also identi ed microbiota signatures speci c for different histological types of lung cancer, including SSC, ADC, and SCLC.

Subjects and sample collection
The recruitment of participants and the process of sample collection are depicted in Fig. 1 Collection Tubes, which were pre-lled with Stool DNA Stabilizer for collection (Stratec, Germany), then frozen and stored at -80 ºC for further use. All subjects were examined clinically before sampling and were subsequently divided into three groups (Table 1) The participants with the following diseases were excluded: cardiovascular disease, diabetes mellitus, liver cirrhosis, irritable bowel syndrome, in ammatory bowel disease, infections with known active bacteria, fungi, or virus. Those who abused drug or alcohol in the last year, or used antibiotics, probiotics, prebiotics, synbiotics in the month, or received chemotherapy before collection of the fecal sample were also excluded. DNA extraction, polymerase chain reaction (PCR) and pyrosequencing Total DNA was extracted from 100 mg of fecal samples using E.Z.N.A. ® Stool DNA kit (Omega Bio-tek, Inc.) according to the manufacturer's instructions. A Nanodrop 2000 Spectrophotometer was used to evaluate the purity and concentration of isolated DNA. All samples were diluted to have a uniform DNA concentration of 10 ng/μl. The polymerase chain reaction (PCR) was performed to amplify the V3and V4 region of the 16S rRNA gene using the primers 341F (5′-CCTAYGGGRBGCASCAG-3′) and 806R (5′-GGACTACNNGGGTATCTAAT-3′); template DNA was absent in the negative control. PCR products were monitored on a 2% agarose gel. After amplicons extraction, samples were puri ed using AXYGEN gel extraction kit (Qiagen) and quanti ed by Quant-iT PicoGreen dsDNA Assay Kit on Microplate reader (BioTek FLx800). Sequencing and data analysis were subsequently performed on an Illumina MiSeq platform by Personal Biotechnology Co., Ltd. (Shanghai, China). The taxa classi cation and statistical analysis were conducted as described in previous studies.

Statistics and analysis
Illumina MiSeq sequences obtained after quality control analysis were used in the present analysis, which were uploaded to QIIME (Quantitative Insights Into Microbial Ecology, v1.8.0) for further study. The operational taxonomy units (OTUs) of representative sequences at a similarity cutoff of 97% and their relative abundance (α-diversity) were used to calculate ACE and Chao1 index by UCLUST. The abundance and diversity of the OTUs (β-diversity) were examined using principal component analysis (PCA), Principal coordinates analysis (PCoA) and nonmetric multidimensional scaling (NMDS) with weighted and unweighted UniFrac analysis in R software. The statistical signi cance of the separation among groups was assessed by the linear discriminant analysis effect size (LEfSe) method based on linear discriminant analysis scores exploited by Curtis Huttenhower (http://huttenhower.sph.harvard.edu/galaxy/), which used the nonparametric factorial Kruskal-Wallis and Wilcoxon rank sum test to identify key OTUs for separating different treatment groups at a signi cance level of 0.05. Metastats analysis was performed based on the raw count data matrix to nd out the taxa statistically different between HC and LC samples.
For the analyses of clinical data, the non-parametric t-test between HC and LC groups was performed with the assistance of GraphPad Prism 6 (Graph Pad Software, La Jolla, CA, USA). Results were considered to be statistically signi cant with P < 0.05. The datasets are publicly available (Accession number PRJNA 681703).

Characteristics of participants
A total of 50 participants enrolled in the Department of Medical Oncology, The Second A liated Hospital of Dalian Medical University (Dalian, China). Participants were excluded after fecal sample collection because of antibiotic use, received chemotherapy, or combined with other diseases such as diabetes mellitus (Fig. 1). Finally, a total of 28 LC patients were remained, they were mainly males (83%) with a median age of 65 years old (Table 1) and with an average smoking index of 606.79±127.1. Average smoking index was the number of cigarettes per day × years of smoking. 19 healthy controls (HC) were included for age and gender matching, with an average smoking index of 350.00±72.1 (P > 0.05 compared with LC group) (details in Table 1). More than 64% of the enrolled LC patients were smokers, with a high smoking index about 606.79±127.1. Among them, 28.57% (8/28) of the patients were diagnosed as SCLC, and 71.43% (20/28) were NSCLC. Among the NSCLC patients, 21.43% (6/28) were diagnosed as squamous cell carcinoma and 50.00% (14/28) were adenocarcinoma.
The overall structure of gut microbiome in LC patients By sequencing of the 16S ribosomal RNA gene, we found that The HC and LC groups share about 2368 same Operational Taxonomic Units (OTUs), but there are 372 OTUs were obtained speci cally in the HC group, and 202 were obtained speci cally in the LC group (Fig. 2a). We observed signi cant decrease in alpha diversity of gut microbiota in LC group, which expressed by the ACE and Chao1 index ( Fig. 2b and c). Whereas the Shannon diversity index and the Simpson index did not show signi cant differences between two groups ( Supplementary Fig. S1). Changes in the relative abundance of gut microbes in LC patients were observed not only on the phylum level, but also on the levels of order, class, and family ( Supplementary Fig. S2). At genus level, signi cant decreased abundance of Prevotella and elevated abundance of Bacteroides, and Ruminococcus etc. were detected in LC patients (Fig. 2d).
The beta diversity metrics from the control and LC individuals also showed strong grouping pattern.
Although signi cant inter-individual variation exist among patients and the healthy controls, the fecal microbiota of the two groups still separated clearly according to community composition using Principal component analysis (PCA, Supplementary Fig. S3), and unweighted/weighted UniFrac Principal coordinates analysis (PCoA) (Fig. 2e). These differences were also observed by the two-dimensional Nonmetric Multidimensional Scaling (NMDS) based on unweighted/weighted UniFrac (Fig. 2f). Most of the samples from each group clustered together as evaluated by Hierarchical clustering based on Weighted UniFrac by the method of Unweighted pair-group method with arithmetic means (UPGMA, Supplementary Fig. S4). Especially, when analyzed by the method of Partial Least Squares Discriminant Analysis (PLSDA), we observed a signi cant separation between the LC patients and HCs (Fig. 2g).

Altered microbiota composition in LC patients
The alteration of gut microbiota in LC patients was further proved by the LEfSe approach, which identi ed the key phylotypes responsible for the difference between the two groups. Actinobacteria, Bacilli, Ruminococcus, Streptococcus, and Mycobacteriaceae, etc, which were most abundant in the LC group, and Prevotella, Bacteroidetes, and Dialister, which were most abundant in the HCs, were the dominant phylotypes that contributed to the difference between the intestinal microbiota of LC patients and HCs ( Fig. 3a and b). The signi cantly elevated relative abundance of Actinomyceae, Streptococcus, and Ruminococcus, and decreased abundance of Prevotellaceae were observed in the gut microbiome of most of the LC patients, which suggested a highly consistence among different individuals ( Supplementary Fig. S5).
Using the method of Metastats, we found a signi cant decrease in the abundance of Bacteroidetes (phylum), Bacteroidia (class), Bacteroidales (order) and elevated abundance of Firmicutes (phylum), Bacilli (class), Actinomycetales (order), Bacillales (order), Lactobacillales (order) in gut of the LC patients ( Supplementary Fig. S6). On the family level, signi cant elevation of the relative abundance of Streptococcaceae, Actinomycetaceae, decreased abundance of the Prevotellaceae and Veillomellaceae were observed in gut of LC patients compared with HCs (Fig. 4a). These changes may mainly due to the changing of genus such as the Streptococcus, Actimomyces, and Prevotella etc (Fig. 4b). In addition, we also detected elevation of genera such as Ruminococcus, Rothia, Bacillus, Peptostreptococcus, Mycoacterium, etc, and decreased abundance of Dialister in gut of LC patients, which are consistent with LEfSe analysis results.

Comparison of gut microbiome in LC patients with speci c histological types
We further performed a detailed comparison of the gut microbiome in lung cancer patients according to different histological types, including adenocarcinoma (Group A, n=14), squamous cell carcinoma (Group B, n=6), and SCLC (Group C, n=8). Only the index of alpha diversity (ACE) in SCLC patients is signi cantly lower than control (Fig. 5a). Other groups showed no signi cant differences compared either with HC or other groups, although the average indexes of each of the speci c histological type are obviously lower than that of the HCs (Fig. 5a and Supplementary Fig. S7). The beta diversity analysis by PCoA and NMDS showed no obvious separation between groups (Fig. 5b and c). When analyzed by the method of PLSDA, we observed a separation between group A and C, which suggested that the gut microbiome of SCLC patients may differ from that of LC patients with adenocarcinoma (Fig. 5c). The taxonomy-based comparison at the genus level showed that, Bi dobacterium, Clostridium, and Prevotella, etc, are the dominant phylotypes in group A, but were signi cantly reduced in the other two groups. In group B, the dominant genera including Ruminococcus, Lachnospira, and Lactobacillus, etc, which are less abundant in group A and C. in addition, the genera of Streptococcus, Anaerotruncus, and Bacillus, etc are more abundant in group C when compared with group A and B (Supplementary Fig. S8).

Discussion
Alterations of the gut microbiota in uence the incidence and progression of not only gastric carcinogenesis [16,17], but also extra-intestinal cancers, such as breast and hepatocellular carcinoma, presumably through in ammatory and metabolic circuitries [18,19]. Meanwhile, gut microbiota was also found contribute to the acute lung injury [20], the exacerbation of chronic obstructive pulmonary disease (COPD) [21,22], and the development of asthma [23,24], which highlighted its important role in affecting the respiratory system. Actually the hypothesis of "gut-lung" axis has been raised 20 years ago, when study found that gut-derived injurious factors can reach to the lung and systemic circulation via the intestinal lymphatics [25]. Gut microbiota was found to be responsible for inducing lung in ammation against bacteria in mice and enhancing neutrophils in ltration through activation of tool like receptor 4 (TLR4) [26]. Such immune transmission from gut to lung has been proved by many studies. One viewpoint supports that exacerbation of chronic lung diseases occur as an uncontrolled and inappropriate in ammatory response to bacteria colonizing damaged airways due to an ineffective Peyer's patch-derived T lymphocyte response [20,27,28]. These studies above strongly suggested a possible correlation between intestinal dysbiosis a microbial imbalance or maladaptation inside the gut [29], and the development of lung cancer. However, to our knowledge, no study has explored this possibility yet.
Here we performed a pilot study to evaluated the changing of gut microbiota in lung cancer patients. Although based on a small number of cases, we found a statistically signi cant decrease of alpha diversity in gut microbiota of lung cancer patients, compared with that of the healthy controls, which was consist with previous discovery that non-malignant lung tissues have higher microbiota alpha diversity than the paired tumors [9]. The decline in both the bacterial diversity and richness was found in gut of patients with chronic in ammation such as in ammatory bowel disease (IBD) and colonrectal cancer, especially in patients with conventional adenoma [26]. Reduced respiratory microbiome was also shown associated with greater emphysema and increased immune cell in ltration in COPD patients [27]. These observations seem in line with the hygiene hypothesis that diverse microbes play an essential role in establishing the immune networks of a host, while in patients with various noncommunicable in ammatory diseases, such as asthma, these regulatory networks seemed to be underrepresented and poorly developed [28].
We also observed that alterations of the gut microbial structure in lung cancer patients were characterized by the signi cant decrease of Bacteroidetes (phylum), Bacteroidia (Class), Bacteroidales (Order), and increase of Firmicutes (phylum), Bacilli (Class), Actinomycetales (Order), and Bacillales (Order), which were mainly contributed by the reduction of bacteria genera such as Prevotella and Dialister, and the increase in Ruminococcus,Streptococcus, Rothia, Bacillus, Actinomyces, Peptostreptococcus, etc. These changes are not consistent with previous studies in sliva, bronchoalveolar lavage uid, or lung tissues in patients with lung cancer, except that the genus of Streptococcus was previously found increased in lower airway of lung cancer patients [30]. But interestingly, they are highly agreed with the observations in patients with colon cancer that the class of Bacilli, genera of Streptococcus, Actinomyces, Peptostreptococcus, and etc are enriched [31]. Importantly, we found a signi cantly increased ratio of Bacteroides/Prevotella, which was also proved in patients with colorectal cancer when compared with normal individuals [26]. Taken together, these results indicate a state of dysbiosis in the gut microbiome of patients with lung cancer.
As the major genus that was found signi cantly reduced in LC patients' gut, Prevotella had drawn many of our attention. Prevotella strains are classically considered commensal bacteria due to the extensive presence in the healthy human body and the rare involvement in infections. Prevotella can stimulate epithelial cells to produce cytokines such as IL-8, IL-6 and CCL20, which can promote mucosal Th17 immune responses and neutrophil recruitment [32]. Increased Prevotella abundance is associated with augmented Th17-mediated mucosal in ammation, which is in line with the marked capacity of Prevotella in driving Th17 immune responses through activation of TLR2. These studies suggested an immunestimulating activity of this genus. Congruously, Prevotella abundance was found reduced within the lung microbiota of patients with asthma and COPD [33] (Hilty et al., 2010), which further highlighted its important role in affecting lung cancer development.
On the other side, the increasing of certain bacteria in LC patients suggested a deleterious role of them in the development of cancer. Species of Streptococcus are found to increase in patients with lung cancer. Compared to healthy controls, the NSCLC patients presented signi cantly higher frequencies of Th1 and Th17 cells reacting toS. salivarius and S. agalactiae, in the PB, LC, and GI tract [34]. The order of Actinomycetales had been as potential colorectal cancer driver bacteria [26]. Among them, the lamentous Gram-positive anaerobic genus Actinomyces, was found to cause Actinomycosis, which is a rare and slowly progressive infectious disease that can affect a variety of organ systems including the lung [35]. Actinomyces was also found signi cantly associated with the carcinoma-in-adenoma group [36]. Besides, Peptostreptococcus species, P. stomatis was con rming known to be associated with CRC [37]. P. anaerobius, which is increased in human colon tumors compared with nontumor tissues, can enhance AOM-induced tumorigenesis in mice by activating TLR2/4-ROS-cholesterol axis [38]. In addition, species of Atopobium, A. parvalum was found positively correlates with pediatric IBD disease severity [39]. However, the effects of these bacteria on lung cancer development are still unclear, which deserves in-deep study.
In addition, we compared the gut microbiome of lung cancer patients according to different histological types based on collected samples, including SSC, ADC, and SCLC. Although the alpha diversity among groups showed no difference, we detected an obvious separation of beta diversity between patients with ADC and SCLC. Given the limited number of study cases, further large scale studies on the characterization of gut microbiome in LC patients with different histological types are necessary.

Conclusions
In conclusion, our results showed that the gut microbiome of lung cancer patients altered signi cantly compared with healthy individuals, which was characterized by a large reduction of Prevotella spp. However, the association between microbial dysbiosis and lung cancer is not clearly understood, future studies involving larger cohorts and metagenomics, or metabolomics, may elucidate the correlations between gut microbiota and lung cancer development.

Declarations
Ethics approval and consent to participate This study protocol was approved by the Ethics Committee of the Second A liated Hospital of Dalian Medical University, Dalian, China. After receiving a written description of the aim of this study, all participants gave written informed consent prior to enrollment.

Consent for publication
In submitting an article to any of the journals published by International Microbiology I certify that; I am authorized by my co-authors to enter into these arrangements.
I warrant, on behalf of myself and my co-authors, that: the article is original, we have to clarify that since the research content of this study had been previously submitted to an academic conference (IASLC 19th World Conference on Lung Cancer, Sep 23-26, 2018) in the form of abstract [40], but our detail data were not published in this abstract. I am/we are the sole author(s) of the article and have full authority to enter into this agreement and in granting rights to International Microbiology are not in breach of any other obligation; the article contains nothing that is unlawful, libellous, or which would, if published, constitute a breach of contract or of con dence or of commitment given to secrecy; I/we have taken due care to ensure the integrity of the article. To my/our -and currently accepted scienti c -knowledge all statements contained in it purporting to be facts are true and any formula or instruction contained in the article will not, if followed accurately, cause any injury, illness or damage to the user. I, and all co-authors, agree that the article, if editorially accepted for publication, shall be licensed under copyright owner.

Availability of data and material
All data generated or analyzed during this study are included in this published article or available from the corresponding author upon reasonable request.   The recruitment of participants and the process of sample collection.   The speci c bacterial groups that are signi cantly changed in LC patients compared with HCs detected by the MetaStat method. a Families, b Genera.