The Gut Microbiome as a Promising Biomarker of Cancer Progression Among Female Never-smokers With Lung Adenocarcinoma

Background: The gut microbiome plays an important role in the immune system and has attracted attention as a new biomarker of various diseases, including cancer. As such, in this study, we examined the relationship between the gut microbiome and lung cancer progression. In addition, we assessed the correlation between the gut microbiome and epidermal growth factor receptor (EGFR) mutation status. Methods: Female never-smokers diagnosed with lung adenocarcinoma were consecutively enrolled between May 2018 and August 2019. Fecal samples were collected within 1 month before or after diagnosis and before administration of any lung cancer treatment. Principal coordinate analyses were performed using Bray-Curtis distance matrices to investigate the effects of clinical variables (age, body mass index, Tumor-Node-Metastasis stage, T category, N category, M category, primary tumor size, performance status, and EGFR mutation status) on the gut microbial community. A correlation analysis was also performed to determine the strength of association between the dominant taxonomy (comprising ≥ 1% of the relative abundance of bacterial DNA sequences) and clinical variables. Results: A total of 37 patients were enrolled. T category and primary tumor size were signicantly correlated with the gut microbial community (p=0.018 and 0.041, respectively). At the genus level, a signicant positive correlation was observed between the relative abundance of Faecalibacterium and both T category (correlation coecient, R=0.51, p=0.0013) and primary tumor size (R=0.37, p=0.024), whereas a signicant negative correlation was observed between the relative abundances of Fusicatenibacter and Bacteroides and T category (R=−0.35, p=0.034 and R=−0.32, p=0.05, respectively) and primary tumor size (R=−0.36, p=0.029 and R=−0.41, p=0.012, respectively). EGFR mutation status had no statistically signicant effect on the gut microbial community (p=0.11). However, the relative abundances of Bidobacterium and Faecalibacterium were signicantly higher in EGFR

between May 2018 and August 2019. Fecal samples were collected within 1 month before or after diagnosis and before administration of any lung cancer treatment. Principal coordinate analyses were performed using Bray-Curtis distance matrices to investigate the effects of clinical variables (age, body mass index, Tumor-Node-Metastasis stage, T category, N category, M category, primary tumor size, performance status, and EGFR mutation status) on the gut microbial community. A correlation analysis was also performed to determine the strength of association between the dominant taxonomy (comprising ≥1% of the relative abundance of bacterial DNA sequences) and clinical variables.
Results: A total of 37 patients were enrolled. T category and primary tumor size were signi cantly correlated with the gut microbial community (p=0.018 and 0.041, respectively). At the genus level, a signi cant positive correlation was observed between the relative abundance of Faecalibacterium and both T category (correlation coe cient, R=0.51, p=0.0013) and primary tumor size (R=0.37, p=0.024), whereas a signi cant negative correlation was observed between the relative abundances of Fusicatenibacter and Bacteroides and T category (R=−0.35, p=0.034 and R=−0.32, p=0.05, respectively) and primary tumor size (R=−0.36, p=0.029 and R=−0.41, p=0.012, respectively). EGFR mutation status had no statistically signi cant effect on the gut microbial community (p=0.11). However, the relative abundances of Bi dobacterium and Faecalibacterium were signi cantly higher in EGFR mutationnegative patients than EGFR mutation-positive patients (p=0.012 and 0.041), whereas the relative abundance of Blautia was signi cantly lower in EGFR mutation-negative patients (p=0.036).
Conclusions: This is the rst study identifying the gut microbiome as a promising biomarker of lung cancer progression. Further elucidation of the role of the gut microbiome in lung cancer progression could facilitate development of new treatments for lung cancer.

Background
Lung cancer is the most prevalent cancer in the world, with 1.76 million deaths attributed to this disease in 2018, according to data from the World Health Organization. Unfortunately, as many lung cancer patients are diagnosed in the later stages of disease, when surgery or chemotherapy are less effective, the 5-year survival rate is only 16% [1,2].
Although smoking is the most important cause of lung cancer, it is estimated that approximately 25% of lung cancer patients have never smoked [3,4]. Lung cancers in never-smokers are more common in women, and the histology is primarily adenocarcinoma [4]. Moreover, the prevalence of mutations in the epidermal growth factor receptor (EGFR) is reportedly high among female never-smokers with lung adenocarcinoma [5]. Although EGFR-mutation targeted therapy is often effective in these patients, the development of drug resistance is a serious problem [6]. The pathogenesis of lung cancer progression in never-smokers remains poorly understood, however, and thus should be elucidated in further studies.
The human microbiome plays an important role in the immune system, regulating various metabolic pathways under a wide range of conditions, including pulmonary diseases [7]. Several studies have suggested that the oral and lung microbiota contribute to the development of lung cancer [8][9][10]. However, the microbiota of the gut is the largest contributor to the human microbiome, weighing approximately 1.5 kg in total, and due to its prominence, the gut microbiome is attracting attention as a potential biomarker of diseases, including various cancers [11][12][13]. For example, Tanoue et al. reported that 11 speci c bacterial strains in the feces of healthy human donors synergistically induce the accumulation of CD8 + T cells, enhancing the anti-tumor e cacy of immune checkpoint inhibitors in a mouse model [14]. In a study involving 100 patients (60 with advanced non-small cell lung cancer and 40 with renal cell carcinoma) who received anti-programmed cell death 1 immunotherapy, metagenomic analysis of patient stool samples at disease diagnosis revealed a correlation between the clinical response to immune checkpoint inhibitors and the relative abundance of Akkermansia muciniphila [15]. It is also gradually elucidated that changes in the composition of the gut microbial community are linked to the development of respiratory diseases such as asthma and cystic brosis, and this cross-talk between the gut and lung constitutes what is known as the "gut-lung axis" [13,16,17]. Based on these data, it can be hypothesized that the gut microbiome impacts the progression of lung cancer.
Several recent studies comparing the gut microbiomes of lung cancer patients and healthy individuals also suggested that there is a relationship between the gut microbiome and lung cancer [18,19]. However, these studies were affected by two serious concerns. First, the patient group consisted of individuals with various histologic types of lung cancer. As the different histologic types of lung cancer develop via different mechanisms, lung cancer patients should be classi ed based on histology. Second, tobacco smoking affects the composition of the gut microbiome and is therefore an important potential confounding factor that must be considered when examining the relationship between the gut microbiome and lung cancer [20]. However, the abovementioned studies did not match the healthy individuals and lung cancer patients based on smoking history.
In the present study, we examined the relationship between the gut microbiome and cancer progression among female never-smokers with lung adenocarcinoma. In addition, we assessed the relationship between the gut microbiome and EGFR mutation status among these patients.

Study design
This retrospective study, performed at Kobe University Hospital, was approved by the institutional review board (approval no. B200044). Informed consent was obtained from all patients included in the study. All procedures performed were in accordance with the ethical standards of the institutional and national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Study population and data collection
Female never-smokers histologically diagnosed with lung adenocarcinoma were consecutively enrolled in the study between May 2018 and August 2019. All patients included in the study lived in the same prefecture (Hyogo, Japan) and reported no unusual gastrointestinal symptoms within the 3-month period prior to enrollment. Fecal and blood samples were collected within 1 month before or after the diagnosis of lung cancer, before administration of any lung cancer treatment (surgery, chemotherapy, or radiation).
Prior to fecal sampling, we determined whether antibiotics had been used by the patient. The eighth edition of the Tumor-Node-Metastasis (TNM) staging guideline for lung cancer was used to classify the extent of disease. Eastern Cooperative Oncology Group performance status (PS) and size of the solid component of the primary tumor were also assessed. EGFR mutation status was determined for all but one of the patients. Data from the enrolled patients were retrospectively analyzed in this study.

Fecal sample collection and bacterial DNA extraction
Fecal samples were placed in 15-ml vials containing 3 ml of guanidine thiocyanate (GuSCN) solution (TechnoSuruga Laboratory Co., Ltd., Shizuoka, Japan). The samples were then mixed by vortexing and refrigerated at 4°C. Bacterial DNA was extracted from the samples in GuSCN solution according to a previously described bead-beating method [21].
Ampli cation and sequencing of the 16S ribosomal RNA gene Bacterial DNA extracted from fecal samples was ampli ed by PCR. The V3-V4 regions of the bacterial 16S rRNA gene were ampli ed using the following primers: forward: 5'-TCGTCGGCAGCGTCAGATGTGTATAAGCGACAGCCTACGGGNGGCWGCAG-3', and reverse:5'-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGGACTACHVGGGTATCTAATCC-3'. Detailed descriptions of the primer set and PCR conditions were reported elsewhere [21]. After addition of the sequencing adapters, the amplicons were sequenced using an Illumina MiSeq platform (Illumina Inc., CA, USA) according to the manufacturer's instructions.
Paired-end FASTQ les obtained from MiSeq were trimmed and merged prior to picking operational taxonomic units (OTUs). OTU classi cation and diversity analyses were performed using the QIIME pipeline (version 1.9.1) [22]. All steps from FASTQ le trimming to analyses of gut microbiota diversity were performed automatically according to previously described methods [23]. OTUs were clustered against the SILVA 128 reference database at 97% similarity using the USEARCH algorithm [24,25]. The entire list of gut bacteria of each patient can be found in Supplementary le.

Data presentation and statistical analysis
The resulting data were imported into R software (version 3.5.1) using BIOM les exported by the QIIME pipeline. α-Diversity indexes were calculated using the estimate_richness function in the "phyloseq" R package. β-Diversity indexes, calculated based on Bray-Curtis distances using genus-level data, were generated using the vegdist function in the "vegan" R package. Principal coordinate analysis (PCoAs) was carried out using the dudi.pco function in the "ade4" R package. Covariates of gut microbiome βdiversity variation were identi ed by calculating the associations between continuous or categorical clinical variables and genus-level community coordinates using the env t function in the vegan R package. Correlations were analyzed using Spearman correlation analysis (cor function in "stats" R package). Heatmaps were created using the "superheat" R package, and other graphs were created using the R package "ggplot2". The statistical signi cance of differences was analyzed using JMP 9.0.2 software (SAS Institute Inc., Cary, NC, USA). Data are summarized as the mean (standard deviation) for continuous variables or number or percentage for categorical variables. In all cases, p values ≤0.05 were considered signi cant.

Patient characteristics
A total of 37 female never-smokers with histologically diagnosed lung adenocarcinoma were consecutively enrolled in this study between May 2018 and August 2019. The patients' clinical characteristics and laboratory data are summarized in Table 1. The average age and body mass index (BMI) were 69 years and 22.2 kg/m 2 , respectively. The Japanese National Health Survey conducted in 2017 reported an average BMI of 23.1 kg/m 2 for females aged 60 to 74 years. Thus, the average BMI of the patients included in this study was similar to that of average Japanese women of the same age group. Most patients (n = 27, 73%) had PS 0 disease, whereas 9 patients (24%) had PS 1, and only 1 patient (3%) had PS 2 disease. Therefore, PS was classi ed as 0 and ≥ 1 in the subsequent analysis. Table 1 also shows strati cation of patients by TNM staging and EGFR mutation status. The size of the solid component of primary tumors is described as "primary tumor size" and also shown in Table 1. All of the patients' laboratory data were within normal ranges. Table 1 Clinical characteristics and laboratory data for patients included in the study (n = 37). Alb, albumin; ALT, aspartate aminotransferase; AST, aspartate aminotransferase; BMI, body mass index; BUN, blood urea nitrogen; Cr, creatinine; CRP, C-reactive protein; EGFR, epidermal growth factor receptor; Hb, hemoglobin; LDH, lactate dehydrogenase; Plt, platelet; SD, standard deviation; T-bil, total bilirubin; T-Cho, total cholesterol; TNM, Tumor-Node-Metastasis; PS, performance status; TP, total protein; WBC, white blood cell.

Effects of antibiotics on the gut microbiome
We initially investigated whether the use of antibiotics before fecal sampling affected the gut microbiome of the patients included in the study. Although no patients reported using antibiotics within the 2-week period prior to fecal sampling, 4 patients used antibiotics within 3 months before sampling, and 3 of these patients used antibiotics within 1 month of sampling. A phylogenetic analysis found no similarities in the gut microbiome among patients who took antibiotics (Fig. 1). These data thus indicate that antibiotics use did not signi cantly affect the gut microbiome of patients enrolled in this study.

Effects of clinical variables on the gut microbial community
PCoA was performed using Bray-Curtis distance matrices with the env t function of the vegan R package to evaluate the effects of the clinical variables shown in Table 1 (age, BMI, PS, TNM stage, T category, N category, M category, primary tumor size, and EGFR mutation status) on the gut microbial community. This analysis identi ed T category and primary tumor size as signi cantly correlated with the composition of the gut microbiome. (p = 0.018 and p = 0.041, respectively) ( Table 2). The PCoA plot is shown in Fig. 2. Speci c gut bacteria signi cantly associated with tumor progression A correlation analysis was performed to determine the strength of association between the dominant taxonomy (comprising ≥ 1% of the relative abundance of bacterial DNA sequences) and six clinical variables (vectors shown in Table 2). The results are shown in Fig. 3. We particularly focused on clinical variables that signi cantly affected the gut microbial community (T category and primary tumor size) and examined speci c gut bacteria associated with both of these variables. At the genus level, a signi cant positive correlation was observed between the relative abundance of Faecalibacterium and both T category (correlation coe cient, R = 0.51, p = 0.0013) and primary tumor size (R = 0.37, p = 0.024), whereas a signi cant negative correlation was observed between the relative abundances of Fusicatenibacter and Bacteroides and T category (R = − 0.35, p = 0.034 and R = − 0.32, p = 0.05, respectively) and primary tumor size (R = − 0.36, p = 0.029 and R = − 0.41, p = 0.012, respectively). Scatter diagrams depicting these relationships are shown in Fig. 4. In addition, the relative abundances of Faecalibacterium and Fusicatenibacter were also correlated with TNM stage (R = 0.31, p = 0.06 and R = − 0.34, p = 0.039, respectively), although the former correlation was not statistically signi cant (Fig. 3).

Relationship between speci c gut bacteria and EGFR mutation status
Finally, we examined the association between the gut microbiome and EGFR mutation status. Although the PCoA indicated that EGFR mutation status did not have a statistically signi cant effect on gut microbial community (p = 0.11) ( Table 2), analysis of variance identi ed some bacteria at the genus level as signi cantly associated with EGFR mutation status. Speci cally, the relative abundances of Bi dobacterium and Faecalibacterium were signi cantly higher in EGFR mutation-negative patients than EGFR mutation-positive patients, whereas the relative abundance of Blautia was signi cantly lower in EGFR mutation-negative patients (Fig. 5).

Discussion
Research focusing on how the gut microbiome in uences distant organs, such as the lungs, is increasing.
The proposed "gut-lung axis" concept holds that dysbiosis of the gut microbiome or gut microbiota metabolites can modulate the immune system in the lungs and cause various respiratory diseases [13,16,17]. Indeed, metabolism of dietary ber by the gut microbiota was shown to affect allergic airway diseases in mice [26]. Another study suggested that the gut microbiome affects the development of the microbial community of the respiratory tract in infants with cystic brosis [27]. However, there are few reports of studies investigating the relationship between the gut microbiome and lung cancer.
Considerable recent attention has focused on the use of probiotics, living microorganisms that can improve the composition or function of the gut microbiota, in the context of treatments for various neoplasms, including lung cancer [28,29]. Probiotics are commonly found in yogurt, and an inverse association between yogurt intake and lung cancer risk was reported in a pooled analysis of 10 prospective cohorts involving 1,445,850 adults [30]. Based on these data, we considered it important to evaluate the relationship between the gut microbiome and lung cancer, as the resulting data could help to establish new lung cancer treatments targeting the gut microbiome.
In this study, we found that T category and primary tumor size are signi cantly associated with the gut microbial community among female never-smokers with lung adenocarcinoma. To our knowledge, this is the rst study to identify the gut microbiome as a promising biomarker for lung cancer progression. Our analyses also revealed a positive correlation between the relative abundance of Faecalibacterium and tumor progression, whereas a negative correlation was found between the relative abundances of Fusicatenibacter and Bacteroides and tumor progression. Previous studies have demonstrated that some speci c gut bacteria can affect immune cells in the tumor microenvironment [14,31]. This observation provides a potential clue for explaining the link between gut bacteria and lung cancer progression. For example, Faecalibacterium prausnitzii (sole known species within the genus Faecalibacterium), a common anaerobic bacterium usually representing more than 5% of the total gut bacterial population, upregulates regulatory T cells both in vivo and in vitro [32,33]. Generally, regulatory T cells suppress the activity of cytotoxic T cells, and it was reported that the number of tumor-in ltrating regulatory T cells is associated with worse recurrence-free survival in non-small cell lung cancer [34]. Based on these data, we speculate that Faecalibacterium plays a role in lung cancer progression by activating the function of regulatory T cells in the tumor microenvironment. Regarding Bacteroides, Vétizou et al reported that T-cell responses speci c for Bacteroides thetaiotaomicron or Bacteroides fragilis were associated with the e cacy of cytotoxic T-lymphocyte-associated antigen 4 (a negative regulator of T-cell activation) blockade in mice and humans [35]. This result suggests that Bacteroides can upregulate T cells in the tumor microenvironment and suppress tumor proliferation. It remains unclear whether Fusicatenibacter has an effect on immune cells because no detailed reports are available. Additional studies are needed to determine whether the speci c bacteria identi ed in this study (Faecalibacterium, Fusicatenibacter, and Bacteroides) actually affect immune cells in the tumor microenvironment of lung cancer.
Our study has some clinical strengths, such as eliminating the confounding effect of smoking. Smoking is known to affect the gut microbiome and increase the numbers of Bacteroides and decrease those of Firmicutes [20]. In a previous study comparing the gut microbiomes of lung cancer patients and healthy individuals, lung cancer patients had higher levels of Bacteroides and lower levels of Faecalibacterium (of the phylum Firmicutes) [18]. However, the percentage of ever-smokers in that study was higher among lung cancer patients than healthy controls (63.4% vs 43.9%). Thus, it is likely that the result of the study was in uenced by the effect of smoking. Although changes in the gut microbiome due to smoking may be concerned with carcinogenesis, that potential link should be examined in future prospective studies. By contrast, our study completely excluded the possible confounding effect of smoking, because only never-smokers were included. Second, this is the rst study to examine the relationship between EGFR mutation status and the gut microbiome. Regarding colorectal cancer, Burns et al previously reported that the tumor microenvironment microbial community is correlated with mutations in tumor DNA, and that this correlation can be used to predict mutated genes based on the microbiome [36]. Some bacteria identi ed in our study (Bi dobacterium, Faecalibacterium, and Blautia) were signi cantly correlated with EGFR mutation status. To the best of our knowledge, there are no reports describing the relationship between these bacteria and EGFR mutation status. With regard to the results of our study, we could not determine whether the changes in these bacteria occurred before or after the appearance of EGFR mutations. However, a previous study reported that Enterococcus faecalis can induce EGFR activation in human oral cancer cells via the production of a speci c signaling molecule [37]. Thus, it is possible to hypothesize that the bacteria identi ed in this study are also associated in development of EGFR mutations and that they may play a crucial role in developing new treatments for lung cancers with EGFR mutations.
The potential limitations of the present study are as follows. First, the number of included patients was small. Thus, larger studies are needed to con rm the results of our study. Second, although this study provides initial insights into the relationship between the gut microbiome and lung cancer progression, it was not possible to determine whether alterations in the gut microbiome contribute to lung cancer progression or whether lung cancer progression changes the gut microbiome. To elucidate this causeand-effect relationship, animal experiments or future longitudinal studies that incorporate repeatedly collected fecal samples will be necessary. Third, we used primary tumor size as a marker for cancer progression, but this parameter does not necessarily re ect the growth rate of the tumor. To evaluate cancer progression more accurately, for example, tumor volume doubling time is preferable. However, we could not measure tumor volume doubling time because of the study design.

Conclusions
In summary, our study identi ed the gut microbiome as a promising biomarker for cancer progression among female never-smokers with lung adenocarcinoma. By using probiotics or creating molecules that target microbial enzymes, we may expect to improve the e cacy of existing lung cancer therapies. Funding MT received grant support from Daiichi Sankyo Co, Ltd.

Availability of data and materials
The datasets used and/or analyzed during the current study are included in a supplementary le.
Ethics approval and consent to participate The procedures in this study related to ethics and patient consent were approved by the Clinical and Translational Research Center of Kobe University Hospital (permission number: B200044).

Consent for publication
Written consent forms were obtained from all participants. Figure 1 Phylogenetic analysis of the gut microbiome. Gray numbers are associated with patients who used antibiotics before fecal sampling within the indicated time period, and yellow numbers are associated Page 16/19 with patients who did not use antibiotics within the indicated time period.   Scatter diagrams depicting the relationships between speci c gut bacteria and T category or tumor size.

Figures
T category and primary tumor size were correlated with the relative abundance of Faecalibacterium (A and B), Fusicatenibacter (C and D), and Bacteroides (E and F). Each point represents a single sample and is colored by T category and primary tumor size. The vertical axis shows the number of each bacterial genus among 104 total operational taxonomic units. R, correlation coe cient.