Comparative study on the esophageal microflora of different tissue types in patients with esophageal squamous cell carcinoma

: Objective: This study aimed to compare the differences in microbiota between the postoperative tissues and esophageal mucosa tissues of patients with esophageal squamous cell carcinoma (ESCC). Methods : Seventy-two patients who had ESCC and diagnosed in Taihe Hospital were selected from July 2018 to July 2019 to participate in this work. Then, 27 postoperative tissues and 45 mucosa samples of ESCC were collected. The sequence V 4 hypervariable region was amplified, and Illumina MiSeq sequencing was performed to analyze the differences between the two groups. Results: Results revealed that the Shannon and Chao1 indices of the postoperative esophageal cancer tissue group (Group A) were significantly higher ( P <0.05) than those of the esophageal mucosa group (Group B). The Simpson index of Group A was higher than that of Group B, but the difference was not significant ( P >0.05). The beta diversity of the two groups was also not significantly different ( P >0.05). LEfSe analysis showed that the abundance of Megasphaera , Actinobacteria , Enterobacteriaceae , and Enterobacteriales in Group A was significantly higher than that in Group B, but the abundance of Mogibacteriaceae in Group B was significantly higher than that in Group A ( P <0.05). At the phylum level, Actinobacteria and Verrucomicrobiae were more abundant in the postoperative tissue group than in the esophageal mucosa group. The abundance of Fusobacteriia , SR1 , and Spirochaetes in the postoperative tissue group was significantly lower than that of the esophageal mucosa group( P <0.05). Conclusion: The source of the sample should be considered in studies on the esophageal flora. More relevant findings were obtained from postoperative tissues than from the normal mucosa.


Introduction
Esophageal carcinoma (EC) ranks seventh in terms of the global incidence of malignant tumors and sixth in terms of mortality [1]. This type of cancer mainly includes two pathological types, namely, esophageal squamous cell carcinoma (ESCC) and esophageal adenocarcinoma (EAC) [2]. The incidence of EC in China ranks among the top five in the world [3], the sixth in terms of the incidence of malignant tumors in the country, and fourth in terms of death [4]. ESCC accounts for over than 90% of all EC cases [5].
As part of the tumor microenvironment, microorganisms may participate in tumor development by inducing chronic or persistent inflammation [6]. However, the microecological composition of each part is not uniform, and different parts of the gastrointestinal tract may have specific microecological communities [7]. Gender, obese body type, age, food, host genetic background, environment, antibacterial drugs, and other factors affect microbial structures [8][9][10][11][12].
Moreover, different methods of material extraction may affect research results on the digestive tract flora.
The microflora in the digestive tract is related to the occurrence and development of ESCC [13]. Changes in the esophageal flora should be studied, or specific bacterial changes should be detected; such studies may be beneficial to the early diagnosis, evaluation, and favorable prognosis judgment of ESCC [14][15][16]. Sampling methods for research on the flora causing esophageal diseases include saliva collection, oropharyngeal swabs, esophageal mucosal swabs, endoscopic biopsies, endoscopic mucosal resection specimens, surgical biopsies after esophageal surgeries, esophageal string tests (ESTs), and new Cytosponge devices [17][18][19][20][21]. Studies on the esophageal flora of patients with ESCC remain in infancy, and the most suitable type of samples for this disease is unknown. This study aimed to identify an ideal method for studies on the esophageal flora of patients with ESCC by comparing differences in the bacterial flora between surgical tissues and endoscopic esophageal mucosa tissues.

Sample source
A total of 72 patients who had ESCC and diagnosed via digestive endoscopy and thoracic surgery at Taihe Hospital from July 2018 to July 2019were selected for this research and divided into two groups. Group A was composed of 27 esophageal cancer postoperative tissue samples, and Group B was assigned 45 esophageal mucosa samples. Gender and age were not significantly different between the two groups ( Table 1).
The following inclusion criteria of ESCC were applied : age ≥18 years; pathological diagnosis of ESCC; without metabolic diseases (such as diabetes), hyperlipidemia, or other infectious diseases; good general condition; no intake of antibiotics, acid suppressants, or probiotics within the past 2 months; balanced diet and no special dietary habits; and no serious liver, kidney, and immunodeficiency diseases. The exclusion criteria were as follows: use of drugs affecting the microecology of the esophagus in the past 2 months; complications of metabolic or infectious diseases; presence of tumors other than ESCC; incomplete data; and not considered suitable for inclusion by the researchers.
The inclusion criteria of the healthy control group were established with reference to the literature [22,23]: the enrolled persons were ≥18 years old and were generally in good condition; endoscopy showed no abnormal esophagus lesions and pathologically proved normal squamous epithelium, the exclusion criteria were the same as ESCC.
The study protocol was reviewed and approved by the Taihe Hospital Ethics Committee, and all patients' written consent were obtained before they were allowed to participate in this research. Moreover, this study was conducted in accordance with the provisions of the Helsinki Declaration.

Sample collection
Endoscopy center specimens were obtained, gastroscopy was performed 6-8 hours after fasting, and warm water was used for gargling before examination. After the esophageal tumor lesions were found, four to eight specimens were collected with sterile biopsy forceps for examination. Two specimens were marked, placed in sterile cryopreservation tubes, and frozen in liquid nitrogen. The remaining tissues were routinely stored in formalin and sent to the pathology department. Surgical specimens were cryopreserved and sent to the pathology department, and appropriate samples were chosen for follow-up studies in accordance with the inclusion criteria.

DNA extraction
The DNA of the sample was extracted with a Power Maxkit ( MoBio Laboratories, Carlsbad, CA, USA) byusing sodium dodecyl sulfate lysate freeze-thaw method. The purity and quantity of the DNA were determined with NanoDrop (Thermo Fisher Scientific, Waltham, MA, USA) , and the sample was frozen at −20 °C for later use.

Bioinformatic analysis and statistical analysis
Quantitative insights into microbial ecology (QIIME version 1.8.0) and R package (www.r-project.org, version 3.2.0) were used to analyze the data. Alpha diversity indices, including Chao1, Shannon, Simpson, and abundance-based coverage estimator, were calculated.
The abundance and uniformity of OTUs were compared, and UniFrac distance was calculated [24].
Principal component analysis (PCA) and nonmetric multidimensional scaling (NMDS) plots were drawn for the beta analysis of the sample flora structure. The t-test and Monte Carlo permutation test were used to draw box plots and compare differences in the UniFrac distance between groups. Permutational multivariate analysis of variance (PERMANOVA) was performed to evaluate the markers of community structure differentiation among groups [25].
Vegan from R package, MEGAN [26,27], and Graphical Phylogenetic Analysis (GraPhlAn) were used to visualize the groups and abundances [28].Venn diagrams were generated using the Venn Diagram module of the R package to visualize common and unique OTUs between samples or groups. The Kruskal method in the R-package was adopted to compare the differences in classification levels within or between groups.
Linear discriminant analysis (LDA) was combined with a Kruskal-Wallis test through lineardiscriminant analysis effect size (LEfSe) analysis [29] to screen out biomarkers, i.e., species with remarkable differences between groups. Random forest analysis was performed using the default settings of the random forest module in R to compare differences between groups.
Microbial functions were predicted in accordance with PICRUSt [30].

Alpha diversity analysis
The Shannon and Chao indices of postoperative tissue (Group A) samples were significantly higher than those of esophageal mucosa group tissue (Group B) samples (P<0.05).
The Simpson index of Group A was higher than that of Group B, but the difference was not significant (P>0.05).These findings indicated that the diversity of microbial flora in postoperative tissues was higher than that in the esophageal mucosa group ( Fig.1 A 1-3).

Beta diversity analysis
PC1 and PC2 represented the potential factors influencing the deviation of the microbial composition of the two groups. For the two groups, PC1=19.14%, P PC1 =0.67, and P PC2 =0.42, suggesting that the bacterial compositions in the two groups had no significant difference (Fig.1B).
NMDS analysis was more stable for data with a complex structure than for data with a simple structure, and the overall flora of the two groups could notbe clearly distinguished. This result demonstrated that the overall composition of the flora of the two groups exhibited no significant difference (Fig. 1C).

Differential LEfSe analysis
The abundance of Megasphaera, Actinobacteria, Enterobacteriaceae, and Enterobacteriales in the esophageal postoperative tissue (Group A) was higher than that in the esophageal mucosal tissue (Group B), but the abundance of Mogibacteriaceae in the esophageal mucosa tissue group (Group B) was higher than that of the esophageal tissue (Group A).The difference was statistically significant (P<0.05, Fig. 2 A and B).

Structural analysis of the two groups of the esophageal flora
Our study only showed the results at phylum and genus levels and didnot include an analysis of the microbial flora composition at other levels.  Campylobacter, Actinobacillus, and Treponema in the former was significantly lower than that in the latter (P<0.05, Table 3).

Random forest classification tree genus classification effect and ROC curve (genus level)
The top 60 species were selected forthe random forest method to establish a model. The error rate was 22.59% (Fig.3A). The receiver operating characteristic (ROC) curve confirmed that our model was reliable and could effectively distinguish between the two groups of samples (AUC=0.86, Fig. 3B).

Comparison of phenotype classification basedon BugBase
The Gram-positive bacteria in the postoperative tissue group exhibited an advantage over the Gram-positive bacteria in the mucosal tissue group. By contrast, the Gram-negative bacteria in the postoperative tissue group were significantly weaker than those in the mucosal tissue group(P<0.05, Figure S2).The two groups were similar underthe following conditions: aerobic, anaerobic, presence of mobileelements, facultatively anaerobic, forms-biofilms, potentiallypathogenic, and stress-tolerant conditions, and the difference was not significant (P>0.05, Table 4).

Discussion
The human microbiota (HM) includes trillions of bacteria, archaea, fungi, and viruses that interact with the human body [32]. Normal HM plays a role in human nutrition, drug metabolism, maintenance of the integrity of the intestinal mucosal barrier, immunomodulation, and protection against pathogens [33]. Changes in the microbial community composition are related to many diseases, including tumors [34,35]. Bacteria were first found in tumors over a century ago.
Different tumor types have unique flora, but the characterization of tumor microbiomes is often challenging because of their low biomass [36].The microbiota, as a part of the tumor microenvironment, plays an important role in tumorigenesis and metastasis. However, the microecology of the different body parts has not been unified.
The number of bacteria in the digestive tract is 10 times the total number of human cells [30]. Most of the bacteria have a specific spatial distribution and are not cultivable. The flora in the oral cavity, esophagus, and rectum are diverse in terms of type and quantity [37]. The composition of the gastrointestinal flora in the same body undergoes remarkable changes [35].
The abundance of the flora in the upper digestive tract is higher than that in the lower digestive tract [38], and the flora of the left and right colons show remarkable differences [39,40]. Thus, the specificity of the tissue flora should be considered in other related studies.
Different sampling methods may affect the results of studieson the flora. Some scholars have explored approaches to obtain a reasonable sampling method. The most commonly used sample types on intestinal flora studies are fecal and mucosal biopsy specimens, but these two methods have limitations [41]. Araujo-Perez F et al. [42] found that the bacterial diversity of ananal swab is higher than that of rectal biopsy. Kotar T et al. [43] and Fair K et al. [44] obtained similar conclusions and confirmed that the sampling time of the anal swab influences sequencing results. An oral swab is more practical and cost efficient than tissue biopsy in studies on oral flora in mice. Thus, swabbing maybe the preferred sampling method [45]. Budding A E et al. [41] believed that the anal swab results of flora canbe reproduced whether sample collection is performed at home or in outpatient clinics; thus, anal swab is suitable for clinical diagnosis.
The esophagus contains many types of bacteria, and an abundant flora can be found between the oropharynx and the stomach. Some esophageal flora in the stomach are similar to those in the oral cavity. Hence, the different parts of the esophagus have no specific flora [18,46]. Studies have been performed to enhance the sampling methods of esophageal flora. Liu et al. [50] believed that the swabs and biopsies of patients with ESCC have similar microbial profiles.
However, Gall et al. [18] and colleagues believed that the amount of DNA recovered from amucosal chip brush is greater than that from mucosal samples in esophageal adenocarcinoma.
Okereke I et al. [51] studied Barrett's esophagus and confirmed that swabs obtained from the oropharynx or an endoscope cannot replace the biopsies of esophageal mucosa. Further research [19] also verified that mucosal biopsy should be used for analysis on the esophageal flora.
Alpha diversity can reflect the diversity of a microbial community. The Chao1 index describes the richness of a community and reflects the number of microbial members, such as OTUs, in a community. The Shannon and Simpson indices reflect the uniformity of a community and the abundance among members. We found that the Chao1 and Shannon indices of the postoperative tissue group increased compared with those of the mucosal tissue group. Although the Simpson index of the postoperative tissue group was higher than that of the mucosal tissue group, but the difference between the two groups was not significant, suggesting that the postoperative tissue flora was richer than the mucosal tissue flora. Increased and good uniformity was noted, i.e., the abundance of the bacteria between the groups was slightly different. Beta diversity was used to examine the similarity of the colony structure of different samples, and good clustering was found in the sample set. The two groups of bacteria, i.e., P PC1 and P PC2 , had>0.05, indicating that the diversity of the two groups exhibited no significant difference.
LEfSe analysis revealed that the flora of the two groups included considerably different species. Megasphaera, Actinobacteria, Enterobacteriaceae, and Enterobacteriales were more abundant in the postoperative tissues of the esophagus than in the mucosal tissues.
Mogibacteriaceae was more abundant in the mucosal tissue group than in the postoperative tissue group. The bacterial species of thetwo groups were compared at the phylum and genus levels. The predominant phyla in the postoperative tissue group were Actinobacteria and Verrucomicrobiae.
The dominant phyla in the mucosal tissue group were Fusobacteriia, SR1, and Spirochaetes. Lastly, future studies should involve a larger sample type and size to determine the best approach for assessing esophageal samples. In this study, specific bacteria were detected, and the role of these species in ESCC should be studied further in future studies.

References:
The authors declare that they have no conflict of interest. The distances from the response sample are similar. The greater the distance between two points, the greater the difference in the community structure between the two groups.
In the plot, blue represents the esophageal mucosa group, and red corresponds to the esophageal cancer postoperative tissue group.