Characteristics of Microbiome in Lung Adenocarcinoma Tissue from Patients in Southwestern China

Background: Increasing evidences have unveiled the connection between microbiome and lung cancer. This study aims to identify the characteristics of microbial communities in the lung cancer tissues from patients in southwestern China, and to compare the distinct microbial genes at different clinical stages of lung cancer for uncovering potential immunotherapy targets. Methods: Forty samples of were performed by 16S rRNA gene sequencing. The subjects were according (T N and smoke status. To identify the taxa composition of each sample, Operational Taxonomic Units (OTUs) were classied on the Effective Tags with 97% identity. The linear discriminant analysis effect size (LEfSe) method was utilized to compare relative abundances of all bacterial taxa between non-metastasis group and metastasis group. The Shannon index of the 97% identity OTUs was calculated to evaluate alpha diversity. Beta diversity measurement was calculated using Principal Co-ordinates Analysis (PCoA). were identied plays important role both in lymph node metastasis and tumor progression, which could provide specic immunotherapy strategy for lung cancer.


Introduction
Microbes function in majority of the biochemical activity on the earth, and they also in uence physiological homeostasis in the human body, including metabolism and immunity. Many studies have uncovered that the abundance and composition of microbial communities vary from different body habitats, such as gut, skin and vagina, which have strong links to healthy conditions as well as human diseases [1]. In the immune system, the recognition of symbiotic microbes is the key to awake immune protection. However, the host-protective immune responses may also arise counter-protective outcomes like in ammation and neoplasia [2]. Helicobacter pylori is the best-known bacterial pathogen of gastric cancer (GC). By local immune activation, H. pylori induces pathological changes and apoptosis of stomach epithelial cells and results in tumorigenesis [3].
Although the lung was considered sterile, emerging evidences have suggested that distinct microbiome exists in the lung and was associated with malignant transformation. By molecular techniques, D'Journo XB rst tested the microbial gene in the distal airways of resected lung specimens and found that CMV was in a high rate of the samples [4]. Lee SH et al. collected bronchoalveolar uids of lung masses patients and analyzed 16S rRNA. 288 genera were identi ed and the results suggested that the bacterial communities were differential between patients with lung cancer and with benign lung nodules [5].
Xinmin Yan et al. assessed the diversity of salivary microbiota and explored their relationship with lung cancer, suggesting that Capnocytophaga and Veillonella could serve as potential biomarkers for lung cancer because of their high proportion in the saliva from lung cancer patients [6].
The aforementioned studies focused on the tests of saliva, sputum specimens or bronchoalveolar uids, which may be bene t for lung cancer screening while have limited re ections of the composition and abundance of microbial communities within the lung cancer tissue. K. Leigh Greathouse et al. detected the 16S rRNA gene sequencing on the lung tissue from 33 controls and 143 cancer cases, demonstrating a lower alpha diversity in the controls and identi ed Acidovorax enriched in smokers [7].
Indeed, there has been few studies exploring the correlation between microbial communities and clinicopathology of the lung cancer. One study analyzed the lung microbiota in 165 non-malignant lung tissue samples from cancer patients and found that the lung microbiota is distinct from the microbial communities in other tissues. Besides, advanced stage patients have higher concentrations of genus Thermus while patients with metastases have higher Legionella [8]. Another study also suggested that differential genera of microbiome existed in different histologic types and metastasis conditions of bronchial washing uid (BWF) and sputum specimens from lung cancer patients [9]. Furthermore, many researchers have demonstrated that the biological patterns, genomic signatures as well as metastasis mechanisms are differential in various clinicopathology of lung cancer [10,11].
Inspired by these clues, we hypothesized that there are certain correlations between the taxa composition of microbiome in the lung cancer tissues and clinicopathological stages of lung carcinoma, including smoke status, tumor stages (T), lymph node and distant metastasis condition. Since 85% lung cancer cases are non-small-cell lung cancer (NSCLC), of which adenocarcinoma (ADC) is more common than other types and have an increasing trend of incidence rate, we selected ADC patients as our subjects in this study. We tend to identify the characteristics of microbial communities in the ADC tissues from patients in southwestern China. In addition, we compare the differential microbial genes at different clinical stages of lung cancer to uncover potential microbiome functions on the tumor progression.

Subjects and Samples collection
This study was approved by the Medical Ethics Committee of the First A liated Hospital of Kunming Medical University. Forty patients diagnosed with lung adenocarcinoma were recruited, of which 20 are males and 20 are females, with a mean age of 60.1 ± 8.1 years old (45-77 years). All the subjects were classi ed into 3 groups (T stage group, N stage group and Clinical stage group), according to T stage (T1 and T2), N stage (N0: without lymph node metastasis; N+: with lymph node metastasis) and clinical stage (I, II, III), respectively ( Table 1). The diagnosis was con rmed by histology according to the 2015 World Health Organization Classi cation of Tumor of the Lung. The staging was assigned by the 8th edition of the Cancer Joint Staging Manual of the American Joint Commission on Cancer (AJCC). Since antibiotics would affect the microbiome, patients who had used oral or systematic antibiotics in the past three months were not included in our study. Besides, all the enrolled patients were injected preoperative antibiotics before surgery. Finally, a total of forty lung cancer samples were collected and all the lung cancer tissues were immediately stored at -80 C then sent to the laboratory for DNA extraction.

DNA Isolation and 16S rRNA ampli cation
Genomic DNA of bacteria in the lung cancer samples were extracted by CTAB method. DNA purity were detected with 1% agarose gel and the DNA concentration was diluted to 1 ng/µL in sterile water. The diluted genomic DNA was used as templates and the 16S rRNA genes of V4 region were ampli ed using Phusion® High-Fidelity PCR Master Mix with GC Buffer (New England Biolabs) for ampli cation e ciency. PCR products were subject to vertical electrophoresis on 2% agarose gels and a colloid recovery kit (Qiagen, Valencia, CA) was applied to recover the target bands. Libraries were generated by the TruSeq DNA PCR-Free Sample Preparation Kit (Illumina, San Diego, USA), and the concentrations were quantitated with a Qubit uorometer and Q-PCR. Finally, the quali ed library were sequenced by NovaSeq6000 (Illumina).

Analysis of sequence data
The original Raw Tags was obtained by splicing the reads using FLASH (v1.2.7), then ltered to acquire Clean Tags using Qiime (Version 1.9.1). To identify the taxa composition of each sample, Operational Taxonomic Units (OTUs) were classi ed on the Effective Tags with 97% identity using Usearch (Uparse v7.0.1001) software. The presentative sequence of each OUT was annotated by RDP Classi er against the Silva (SSU123)16S rRNA database using con dence threshold of 80%, obtaining taxonomic classi cation at the Phylum, Class, Order, Family, Genus and Species level. Multiple sequence alignment was performed using MUSCLE3.6(Version 3.8.31)to further explore the phylogenetic relationships among different OTUs. Shannon index was performed by Qiime to descript alpha diversity. To determine beta diversity, Unifrac distance was calculated by Qiime and analyzed by principal coordinate analysis (PCoA) using R (Version 2.15.3). Microbiome abundance and diversity between different groups was calculated by t-test or Wilcoxon rank sum test and drawn by R. Linear discriminant analysis (LDA) effect size (LEfSe) analyses were performed with the online LEfSe tool (http://huttenhower.sph.harvard.edu/lefse/). The LDA threshold score was 4.

Results
Baseline data of all the subjects The subjects were grouped according to TNM stages, clinical stage and smoke status. No signi cant difference was found in age, gender and smoking status between different groups ( Verrucomicrobia (3.3%±4.2%). On the genus level, 21 genera were identi ed (> 1% average relative abundance) and the top 10 genera are shown in Fig. 1.

Analysis of microbial diversity within groups
Shannon index was calculated to evaluate the bacterial diversity within different groups and the result showed no signi cant difference (Fig. 2).

Analysis of microbial composition
To compare the composition of microbial communities in different groups, we calculate and analyzed the Weighted Unifrac distances, which was presented as PCoA plot. The result showed that there was dominant separation (p < 0.001, Fig. 3a) and signi cant difference (Fig. 3b) between N0 and N + group.
Most of the samples in the T1 group were more uniform and clustered distinctly away from the T2 group (Fig. 3c) and the comparisons were signi cant (p < 0.001, Fig. 3d). The separations were also found in clinical I, II and III group (Fig. 3e) and signi cant differences were found in either two groups. However, based on the smoke status, no signi cant difference was observed between smokers and non-smokers.

Analysis of differentially abundant taxa
We identi ed 224 genera in total, of which the common genera were shown as heatmap in differential stages and smoke status (Fig. 4). The top genera included Moraxella, Bi dobaterium, Lactobacilus, Blautia, Akkermansia, Alistipes and Faecalibaculum. To further identify speci c species in every group, we use LEfSe method to nd the differentially abundant taxa at each level. Firstly, in the N0 and N + group, we found 9 differential species including two classes Actinobacteria and Erysipelotrichia, three families Bi dobateriaceae, Muribaculaceae and Erysipelotrichales, two orders Bifodobacteriales and Erysipelotrichales, and one genus Bi dobacterium (Fig. 5a). On the cladogram, we can clearly see that the main speci c species mainly belong to the two classes Actinobacteria and Erysipelotrichia (Fig. 5b). All the abundance of these species was higher in N + group. The relative abundance level of genus Bi dobacterium was 10.78%±11.59% in the N0 group and 20.15%±13.44% in the N + group (p < 0.05). In addition, in the T1 and T2 group, the LEfSe result identi ed 4 phylum and 10 genera. The dominant phylum were Bacteroidetes, Actinobacteria, Verrucomicrobia and Proteobacteria (Fig. 6a). The differential genera were Moraxella, Dolosigranulum, Corynebacteriaceae and Citrobacter in the T2 group and Bi dobacterium, Alistipes, Akkermansia, Blautia, Lactobacillus as well as Facelibacterium in the T1 group (Fig. 6b).

Discussion
People considered that the lung wasn't sterile until some researchers demonstrated that chronic obstructive pulmonary disease (COPD) was associated with lung microbiome [12]. Afterwards, evidences were unveiled between microbiome and lung cancer [4,5,8]. However, due to the di culty of getting lung cancer tissue samples, previous studies were always focus on the sample tissues from BWF and sputum specimens. In addition, as the oral cavity is connected with lung by air and upper respiratory tract, the oral microbiome is likely to affect and contaminate the lung microbiome [13]. Herein, studies directly on the cancer tissues are extremely pressing and important. Some researchers described associations between speci c microbial diversity patterns and epidemiological exposures as well as associations between stages of disease with microbial composition, raising interesting mechanistic hypotheses [7][8][9]. In this study, based on stages and smoking status, we analyzed microbial diversity, composition and abundance on the tissue samples of lung adenocarcinoma. Because some investigators found that administration of antibiotics may re ect the in ammatory effects of repeat infections and changes in the composition of the lung microbiome [14], we excluded the subjects who had used oral or systematic antibiotics in the past three months to minus the antibiotic's effects.
Smoke is considered associated with both SCC and AD lung cancer while the association was stronger in SCC [15]. K. Leigh Greathouse et al. identi ed a speci c taxa, Acidovorax, which was enriched in smokers with SCC [7]. Additionally, in a study of non-malignant lung tissue (n = 152), they observed a signi cant increase in alpha diversity with higher number of pack years of smoking [8]. In our AD samples, we did not nd differential microbial diversity and composition between smokers and non-smokers, which may contribute to the different carcinogenesis of AD and SCC. Nevertheless, because the sample size was not big enough (30 in non-smoker group and 10 in smoke group), we need expand the sample size to further validate these results.
In our stage groups, we found no difference in the alpha diversity, indicating that the microbial diversity was uniform in our early AD tissues. Some researchers indicated that the non-malignant lung tissues had higher microbiota alpha diversity than the paired tumors, which also been previously shown for tumor/non-malignant samples from colorectal cancer patients [8,16]. Unfortunately, we did not collect normal lung tissue to con rm whether difference exists between normal and lung tissues. It will be important to explore if the microbiota in lung tissue plays a role in non-malignant area and in uence tumor progression or on the contrary.
In the lymph node metastasis and non-metastasis group, we identi ed two classes Actinobacteria and Erysipelotrichia and one genera Bi dobacterium (Actinobacteria) by LEfSe analysis. Bi dobacterium is common commensal in human oral, intestinal tract and vagina and is always believed to be nonpathogenic bacteria and bene cial to the host. However, increasing evidences demonstrated that Bi dobacterium is possible causing invasive human infection [17]. Especially, patients with positive Bi dobacterium blood culture always accompanied by gastrointestinal disease and/or impaired immunity [18]. In our study, the higher abundance of Bi dobacterium in the lymph node metastasis group may be related to the impaired immunity of cancer patients, which causing proliferation of this bacteria in the blood and lymph and further inducing lymph node metastasis of lung cancer cells. To verify this hypothesis, we need collect blood samples detecting abundance of Bi dobacterium from lung cancer patients in our future study. Interestingly, one study demonstrated that Bi dobacterium could enhance host antitumor immunity by increasing CD8 + T cell and promote anti-PD-L1 e cacy, which provides us unique carcinogenesis and immunotherapy strategy of lung cancer [19].
In the tumor stage groups, the LEfSe result identi ed 4 phylum and 10 genera, of which genus Bi dobacterium, Alistipes, Akkermansia, Lactobacillus and Facelibacterium were existed mainly in the T2 group. The abundance of Bi dobacterium was also the highest among the dominant genera, suggesting the possible pathogenic mechanism between Bi dobacterium with tumor progression. Akkermansia specializes in mucin degradation and functions between mucosal layer of large intestine and host immunity. Decreased Akkermansia is associated with some disease like IBD and acute appendicitis [20]. Besides, some studies believe that Akkermansia muciniphila has an anti-tumor effect, especially in gastroenteric tumor [21]. Fecal microbiota transplantation (FMT) of Akkermansia could modulate e cacy of PD-1 immunotherapy [22]. However, in our study, we nd decreased Akkermansia in the T1 group, which is contrary to its protective function in the gastrointestinal tract thus the pathogenic mechanism of Akkermansia needs to be explored in following study. Alistipes, belongs to Bacteroidetes phylum, has contrasting function on human metabolism and disease. It can protect host from diseases like liver brosis and colitis as well as promoting the incidence of colorectal cancer and metal depression [23]. Like Akkermansia, the abundance of Alistipes is higher in T2 group. These results suggest that microbiome in the lung tissues may promote the tumor progression despite some of which have protective function in intestinal diseases. Alternatively, tumor progression could also affect the microenvironment and microbiota of a larger surrounding area.
Some researchers identi ed speci c species in the advanced stage of lung cancer, indicating that the differential microbiome function in the tumor development [8]. We also observed differential bacterial composition and abundance in the clinical stage group. Unfortunately, the sample size in the III stage group was only 5 and no tumor tissues of IV stage were collected, which could not represent the microbiome status in the advanced AD. We will enroll more cancer patients with advanced stage to demonstrate our guess.
In conclusion, by 16S RNA sequencing, we identi ed dominant species of lung cancer tissue in different groups of AD patients. Bi dobacterium plays important role both in lymph node metastasis and tumor progression, which could provide speci c immunotherapy strategy for lung cancer.   Comparison of bacterial community structure (a) PCoA plot of N0 and N+ group. (b) Weighted Unifrac distance between N0 and N+ group with signi cant difference. (c) PCoA plot of T1 and T2 group. (d) Weighted Unifrac distance between T1 and T2 group. Signi cant difference was found between the two groups. (e) PCoA plot of clinical I, II and III group. (f) Weighted Unifrac distance between clinical I, II and III group. The differences were signi cant in every two groups. *p<0.05, **p<0.01,***p<0.001.