Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 7-Day Trial for You or Your Team.

Learn More →

A non-endoscopic device to sample the oesophageal microbiota: a case-control study

A non-endoscopic device to sample the oesophageal microbiota: a case-control study Articles A non-endoscopic device to sample the oesophageal microbiota: a case-control study Daffolyn R Fels Elliott, Alan W Walker, Maria O’Donovan, Julian Parkhill, Rebecca C Fitzgerald Summary Lancet Gastroenterol Hepatol Background The strongest risk factor for oesophageal adenocarcinoma is reflux disease, and the rising incidence of 2017; 2: 32–42 this coincides with the eradication of Helicobacter pylori, both of which might alter the oesophageal microbiota. Published Online We aimed to profile the microbiota at different stages of Barrett’s carcinogenesis and investigate the Cytosponge as a November 11, 2016 minimally invasive tool for sampling the oesophageal microbiota. http://dx.doi.org/10.1016/ S2468-1253(16)30086-3 Methods In this case-control study, 16S rRNA gene amplicon sequencing was done on 210 oesophageal samples from This online publication has 86 patients representing the Barrett’s oesophagus progression sequence (normal squamous controls [n=20], non-dysplastic been corrected twice. The first corrected version appeared at [n=24] and dysplastic Barrett’s oesophagus [n=23], and oesophageal adenocarcinoma [n=19]), relevant negative controls, thelancet.com/gastrohep on and replicates on the Illumina MiSeq platform. Samples were taken from patients enrolled in the BEST2 study at five UK December 9, 2016, and the hospitals and the OCCAMS study at six UK hospitals. We compared fresh frozen tissue, fresh frozen endoscopic second on October 4, 2017 brushings, and the Cytosponge device for microbial DNA yield (qPCR), diversity, and community composition. See Comment page 4 Medical Research Centre Cancer Findings There was decreased microbial diversity in oesophageal adenocarcinoma tissue compared with tissue from Unit, Hutchison/MRC Research Centre, University of healthy control patients as measured by the observed operational taxonomic unit (OTU) richness (p=0·0012), Chao Cambridge, Cambridge, UK estimated total richness (p=0·0004), and Shannon diversity index (p=0·0075). Lactobacillus fermentum was enriched (D R Fels Elliott MD, in oesophageal adenocarcinoma (p=0·028), and lactic acid bacteria dominated the microenvironment in seven (47%) Prof R C Fitzgerald MD); of 15 cases of oesophageal adenocarcinoma. Comparison of oesophageal sampling methods showed that the Pathogen Genomics Group, Wellcome Trust Sanger Cytosponge yielded more than ten-times higher quantities of microbial DNA than did endoscopic brushes or biopsies Institute, Hinxton, UK using quantitative PCR (p<0·0001). The Cytosponge samples contained the majority of taxa detected in biopsy and (A W Walker PhD, brush samples, but were enriched for genera from the oral cavity and stomach, including Fusobacterium, Megasphaera, Prof J Parkhill PhD); Rowett Campylobacter, Capnocytophaga, and Dialister. The Cytosponge detected decreased microbial diversity in patients with Institute of Nutrition and Health, University of Aberdeen, high-grade dysplasia in comparison to control patients, as measured by the observed OTU richness (p=0·0147), Chao Aberdeen, UK (A W Walker); and estimated total richness (p=0·023), and Shannon diversity index (p=0·0085). Department of Histopathology, Cambridge Interpretation Alterations in microbial communities occur in the lower oesophagus in Barrett’s carcinogenesis, which University Hospital NHS Trust, Cambridge, UK can be detected at the pre-invasive stage of high-grade dysplasia with the novel Cytosponge device. Our findings are (M O’Donovan MD) potentially applicable to early disease detection, and future test development should focus on longitudinal sampling Correspondence to: of the microbiota to monitor for changes in microbial diversity in a larger cohort of patients. Prof Rebecca C Fitzgerald, Medical Research Centre Cancer Funding Cancer Research UK, National Institute for Health Research, Medical Research Council, Wellcome Trust, Unit, Hutchison/MRC Research Centre, University of Cambridge, The Scottish Government (RESAS). Cambridge CB2 0XZ, UK [email protected] Copyright © The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 license. Introduction with the eradication of Helicobacter pylori, which could alter Oesophageal adenocarcinoma is an aggressive malignancy the composition of microbiota and promote bacterial with poor outcomes that generally develops from a overgrowth. Furthermore, reflux disease is treated with premalignant columnar epithelium called Barrett’s antacid drugs such as proton-pump inhibitors, which have oesophagus. The incidence of oesophageal adenocarcinoma profound effects on gastric acidity and might affect the has increased by six times in developed countries during gastro-oesophageal microbiota. the past three decades. Both Barrett’s oesophagus and There is growing evidence linking abnormal changes in oesophageal adeno carcinoma are thought to develop in the microbiota, known as dysbiosis, with human cancer. response to chronic acid reflux in the lower oesophagus, One of the best described examples is colon carcinoma, which precipitates inflammation and mucosal injury over in which gastrointestinal microbiota have been shown to time. Reflux disease has increased with the obesity promote carcinogenesis in the setting of colonic 6–8 epidemic and altered eating habits in high-income inflammation. Studies have also linked Fusobacterium countries, and central adiposity might also influence nucleatum to colon carcinoma through an altered tumour 3 9,10 carcinogenesis through the release of adipokines. immuno-environment, but without associated colitis. Additionally, epidemiological evidence suggests that the The oesophagus has far fewer bacteria than the colon; rising incidence of oesophageal adenocarcinoma coincides nevertheless, alterations in the microbiota might occur in 32 www.thelancet.com/gastrohep Vol 2 January 2017 Articles Research in context Evidence before this study sequencing, and compares patients with oesophageal Epidemiological evidence suggests that the rising incidence of adenocarcinoma and Barrett’s oesophagus with healthy control oesophageal adenocarcinoma coincides with the obesity patients. We found decreased microbial diversity in oesophageal epidemic, gastro-oesophageal reflux disease, and eradication of adenocarcinoma tissue compared with controls, with Helicobacter pylori with antibiotics and acid suppression enrichment of acid-tolerant bacteria such as Lactobacillus treatment—all risk factors that are capable of altering the fermentum. The microbial diversity was reduced in the lower gastro-oesophageal microbiota. Three studies with small oesophagus regardless of whether cancerous or healthy numbers of patients have shown modest alterations in the oesophageal tissue was sampled within the same patients. microbiota in Barrett’s oesophagus and oesophagitis with We further translated our findings to the setting of early 16S rRNA gene amplicon sequencing. However, studies using detection, using the Cytosponge to sample the microbiota in culture-independent methods to profile the oesophageal Barrett’s oesophagus and high-grade dysplasia. We showed that microbiota in oesophageal adenocarcinoma or high-grade the Cytosponge collected high microbial DNA yield and dysplasia are lacking. One explanation for the scarcity of detected decreased diversity in the pre-invasive stage of oesophageal microbiota studies is the challenge of endoscopic high-grade dysplasia. sampling and low microbial DNA yield. Based on previous studies, Implications of all the available evidence we postulated that the novel Cytosponge could be an effective Alterations in microbial communities occur in the lower tool to sample the microbiota along the entire length of oesophagus in Barrett’s carcinogenesis, which are possible to the oesophagus. detect with the minimally invasive Cytosponge. Our findings are Added value of this study potentially applicable to early disease detection, and future test This study provides a comprehensive characterisation of the development should focus on longitudinal sampling of the microbiota at different stages of the Barrett’s oesophagus microbiota to monitor for changes in microbial diversity in a progression sequence using 16S rRNA gene amplicon larger cohort of patients. 11–13 reflux oesophagitis and Barrett’s oesophagus. However, Methods the role of the microbiota in Barrett’s carcinogenesis is Study design and participants not clearly defined and there is no clinical reference In this case-control study, endoscopic biopsies, brushes, standard at present for sampling the oesophageal Cytosponge samples, and throat swabs were collected microbiota. One of the challenges in studying the from patients with a diagnosis of non-dysplastic Barrett’s oesophageal microbiota is that endoscopy is an invasive oesophagus or high-grade dysplasia, and from control test that provides only a focal sampling of the microbiota patients with symptoms of reflux or dyspepsia enrolled in biopsy samples, and a slightly larger surface area with in the Barrett’s Oesophagus Screening Trial (BEST2) at endoscopic brushings. Minimally invasive methods for five UK hospitals. In patients with a diagnosis of sampling the oesophageal microbiota could be clinically oesophageal adenocarcinoma, tissue samples from the useful for detection and risk stratification of patients with tumour and matched normal squamous oesophagus Barrett’s oesophagus. were collected from six UK hospitals participating in Here we investigate the Cytosponge prototype Oesophageal Cancer Clinical and Molecular Stratification (Europlaz, Southminster, UK) as a non-endoscopic (OCCAMS) for the International Cancer Genome cell-sampling device that can collect a representative Consortium (ICGC). 14,15 sample of cells along the length of the oesophagus. The The patient inclusion criteria included age between device consists of a spherical mesh that is compressed 20 and 90 years with either normal endoscopy or within a gelatine capsule and attached to a string. Once endoscopic and histological documentation of Barrett’s swallowed, the capsule dissolves and the Cytosponge oesophagus or oesophageal adenocarcinoma. The expands in the patient’s stomach before being withdrawn exclusion criteria were current infection, recent antibiotic on a string through the patient’s mouth. We have treatment, previous chemotherapy treatment, and previously shown that this device is a safe, acceptable documented pathological findings unrelated to Barrett’s method for diagnosing Barrett’s oesophagus, with oesophagus or oesophageal adenocarcinoma. Ethical 14–16 promising accuracy and cost-effectiveness. The goal approval was obtained from the National Research Ethics of this study was to provide a comprehensive description Services Cambridgeshire Research Ethics Committee of the microbiota in the different pathogenic stages of on behalf of all hospital centres in the BEST2 trial oesophageal adenocarcinoma using 16S rRNA gene (REC 10/H0308/71) and the OCCAMS/ICGC trial amplicon sequencing and to test the feasibility of the (REC 07/H0305/52 and 10/H0305/51). Written informed Cytosponge to detect changes in the microbiota occurring consent was obtained from all participants before the in Barrett’s oesophagus and high-grade dysplasia. collection of samples and recording of clinical information. www.thelancet.com/gastrohep Vol 2 January 2017 33 Articles Procedures accession number ERP005191. The 16S rRNA gene For patients enrolled in the BEST2 study, matched amplicon sequence analysis was done with mothur. endoscopic biopsies were taken from an area of The MiSeq standard operating procedure was Barrett’s oesophagus and proximal normal squamous followed with the exception of chimera checking, oesophagus. Endoscopic brushings were taken from an which was done with chimera.perseus, and unique area of normal squamous oesophagus only. The sequences were removed using the split.abund and tissue samples collected in the OCCAMS/ICGC study remove.seqs commands before building the distance included endoscopic biopsies, endoscopic mucosal matrix. Contaminant operational taxonomic units resection specimens, and surgical biopsies after (OTUs) were defined as having greater proportional oesophagectomy. Sampling from oesophagectomy abundance in negative controls, alongside previous specimens was done with a sterile scalpel blade (cutting evidence that these OTUs were derived from genera down to submucosa) within 1 h of surgical resection. that are common contaminants, and these reads were All samples were flash frozen in liquid nitrogen and removed using remove.seqs (3434 OTUs defined as stored at –80°C except for the Cytosponge samples, contaminants from 5757 total OTUs). Comparisons which were preserved in BD SurePath liquid at 4°C. All between replicate samples (using Metastats as patients fasted overnight before endoscopy or surgery. implemented in mothur) showed no significant As part of routine perioperative procedure, patients differences so replicates were pooled using the Linux with oesophageal adenocarcinoma who underwent sed command to maximise the number of reads per oeso phagectomy received prophylactic intravenous sample, and samples with fewer than 550 reads or antibiotics at the time of surgery, up to 6 h before the Good’s coverage less than 95% were removed with research samples were obtained. The very close timing remove.groups. The Good’s coverage estimator is of this perioperative antibiotic exposure should not used to assess what proportion of the total OTUs greatly affect microbial community composition present within a given sample are detected in the profiles because 16S rRNA gene amplicon sequencing sequencing results, and thus gives an indication of detects both live and dead bacterial cells cross- how thorough the sampling has been at the chosen See Online for appendix sectionally (appendix p 1). sequencing depth. For measures of diversity that Cytosponge samples were vortexed and centrifuged are sensitive to the sequencing depth, random to pellet cellular debris (215 g for 5 min), and the subsampling was done at the lowest number of reads residual supernatant was used for microbial DNA per sample using the sub.sample command in mothur. extraction after further high-speed centrifugation Data were subsampled at 631 reads for the analysis of (14 000 g for 10 min). DNA was isolated from all tissue samples (median Good’s coverage 96·36%, oesophageal samples using the Precellys Soil DNA Kit range 92·08–99·37), 656 reads for matched tumour- (Peqlab, Southampton, UK). The 16S rRNA gene was normal pairs (97·41%, range 95·27–98·78), 631 reads amplified using primers for the V1-V2 region: 27F for the comparison of different sampling methods 5 ʹ AATGATACGGCGACCACCGAGATCTACAC (96·51%, range 91·13–99·84), and 19 303 reads for TATGGTAATT CC AGMGTTYGATYMTGGCTCAG the analysis of Cytosponge samples (99·83%, range and 338R 5ʹCAAGCAGAAGACGGCATACGAGAT 99·70–99·95). When determining the shared genera NNNNNNNNNNNN AGTCAGTCAG AA between Cytosponge samples, biopsies, and brushes, a GCTGCCTCCCGTAGGAGT, where Illumina adapter cutoff of 0·0001% proportional abundance was used to sequences are at the 5ʹ end, and the N string is a unique focus on the more abundant OTUs that are less likely barcode. Most samples had two barcoded replicates to to be susceptible to errors introduced by subsampling ensure reproducibility (labelled A or B in the (below 0·0001% there were less than 27 reads appendix p 5), and negative controls from every DNA supporting each OTU across all the samples). A second extraction step underwent additional PCR cycles to cutoff value of 0·1% was chosen arbitrarily to show the identify contaminant organisms. The reaction similarity between sample types and is a common 23,24 conditions were 98°C for 2 min, 25 cycles at 98°C for cutoff used in previous studies. Sequence identity 30 s, 50°C for 30 s, and 72°C for 90 s, and extension at was confirmed at the species level, where possible, by 72°C for 5 min. Negative controls were kit reagents or carrying out NCBI BLAST analysis on representative nuclease-free water, which underwent 45 amplification sequences using MegaBLAST. cycles. All samples were amplified in duplicate Quantitative PCR for the 16S rRNA gene was done with and pooled to minimise PCR bias and maximise SYBR Green I Master Mix (Roche, Mannheim, Germany) yield. The PCR products were concentrated using on the LC480 LightCycler 480 II (Roche, Mannheim, ethanol precipitation and quantified using a Qubit Germany), in triplicate. The reaction conditions were 2·0 Fluorometer before sequencing on the MiSeq 95°C for 5 min, 40 cycles at 95°C for 15 s, 60°C for 30 s, Illumina platform using 2 × 250 bp read length. and 72°C for 90 s, and a melt curve. The primer sequences For the European Nucleotide 16S rRNA gene amplicon sequencing data have been were 331F 5ʹTCCTACGGGAGGCAGCAGT and 797R Archive see http://www.ebi. ac.uk/ena/data/view/ERP005191 deposited in the European Nucleotide Archive under 5ʹGGACTACCAGGGTATCTAATCCTGTT. 34 www.thelancet.com/gastrohep Vol 2 January 2017 Articles Statistical analysis 24 patients with Barrett’s oesophagus, 12 (71%) of The non-parametric Kruskal-Wallis test and Dunns 17 patients with oesophageal adenocarcinoma (two did not multiple comparisons post-test were used for comparisons report whether or not they were taking acid-suppressant between diagnostic groups in Graphpad Prism (version 6). drugs), and 15 (75%) of 20 control patients. One patient Within Graphpad, the Wilcoxon signed rank test was used with Barrett’s oesophagus and two patients with for analyses involving matched tumour-normal pairs, and high-grade dysplasia reported taking a course of antibiotics the Friedman test and Dunns post-test were used for within the past month. analyses of matched samples from endoscopic biopsies, After filtering the sequencing data and removing brushes, and the Cytosponge. We used LEfSe, a contaminant sequences, the mean number of reads metagenomic biomarker discovery method, to identify for tissue samples was 6649 (SD 10421) and the microbial taxa that differed significantly between controls, median was 3064 (5688) and the proportion of reads that Barrett’s oesophagus, and oesophageal adenocarcinoma were subsampled for diversity analyses was 9·5% (for samples. LEfSe was used within mothur. All sequencing 631 reads cutoff) and 9·9% (for 656 reads cutoff). reads were included for the composition analysis using 14 tissue samples did not meet quality criteria and LEfSe. LEfSe ranks OTUs in the order that it considers were excluded from the analysis, leaving 16 control, these taxa to be most likely to explain differences between 17 Barrett’s oesophagus, and 15 oesophageal microbial communities using linear discriminant analysis adenocarcinoma samples. Five phyla accounted for the to estimate effect size. A full explanation of the statistical majority of sequencing reads in the dataset: Firmicutes approaches used in LEfSe can be found in the original (59·9%), Bacteroidetes (15·1%), Proteobacteria (12·8%), article by Segata and colleagues. Within mothur, Actinobacteria (5·8%), and Fusobacteria (5·4%). the Bray-Curtis calculator was used to describe the 1060 OTUs were identified and classified as belonging to dissimilarity between communities by taking into account 345 different genera. both the overlap in OTUs that are present between By LEfSe, at the phylum level, the Barrett’s samples samples and the proportional abundance of those OTUs contained a higher proportional abundance of in each sample. Using dissimilarity information calculated Proteobacteria (mean 18·6%, median 14·8%, SE 5·5%) with the Bray-Curtis calculator, the parsimony test and the compared with controls (mean 8·5%, median 8·2%, analysis of molecular variance (AMOVA) test were used to SE 5·5%; p=0·017) and oesophageal adeno carcinoma show significantly different clustering between microbiota samples (mean 7·6%, median 3·9%, SE 2·4%). The profiles from the different diagnostic groups. IBM SPSS control samples were enriched for several taxa at the Statistics (version 24) was used to analyse patient data family level, including the Gram-negative, anaerobic (ANOVA for mean age and Fisher’s exact test for sex, Veillonellaceae (p=0·012, overall proportional abundance ethnicity, and antacid usage). A significant p value was defined as less than 0·05. Control Non-dysplastic High-grade Oesophageal p value (n=20) Barrett’s dysplasia adenocarcinoma Role of the funding source oesophagus (n=23) (n=19) (n=24) The funders had no role in the study design, data analysis, interpretation of data, or writing of the report. The Age, years 57 (29–86) 68 (53–79) 65 (50–82) 70 (44–79) 0·001 corresponding author had full access to all the data in the Sex study and had final responsibility for the decision to Male 7 (35%) 16 (67%) 19 (83%) 15 (79%) 0·006 submit for publication. Female 13 (65%) 8 (33%) 4 (17%) 4 (21%) ·· White ethnicity 19 (95%) 24 (100%) 23 (100%) 16 (89%)* 0·089 Results Antacid usage 15 (75%) 22 (92%) 23 (100%) 12 (71%)* 0·011 To investigate whether the development of oesophageal Samples passing quality control/samples sequenced adenocarcinoma was associated with dysbiosis, we did Cytosponge 20/20 24/24 23/23 ·· 16S rRNA gene amplicon sequencing on tissue samples Brush (squamous only) 19/19 19/19 ·· ·· ·· from patients with oesophageal adenocarcinoma (n=19), Tissue 16/19 17/24 (Barrett’s ·· 15/19 (oesophageal ·· Barrett’s oesophagus (n=24), and healthy control patients oesophagus); adenocarcinoma); 15/24 (squamous)† 15/19 (squamous)† (n=19; table). Patients with oesophageal adenocarcinoma Total number of 58 91 23 38 ·· and Barrett’s oesophagus were older (p=0·001) and samples sequenced predominantly male (p=0·025) in comparison with the control patients, which is consistent with the known Data are n (%) or median (range) unless otherwise stated. ··=samples were not sequenced for all diagnostic categories for the following reasons: it was not safe or appropriate to have patients with cancer swallow the Cytosponge; we epidemiology of this disease. The number of patients compared brush sampling methods for areas of normal oesophagus in patients with no pathology or Barrett’s recruited at each participating hospital centre and oesophagus only; and tissue samples were not available for dysplasia. †Tissue samples from the area of Barrett’s additional clinicopathological data for patients with oesophagus or tumour and matched normal squamous oesophagus were taken in the same cases. *Not recorded, oesophageal adenocarcinoma is provided in the appendix ethnicity unknown for one patient and antacid usage unknown for two patients. (pp 3, 4). Acid-suppressant drugs were taken regularly Table: Patient demographics and oesophageal samples by 23 (100%) of 23 patients with dysplasia, 22 (92%) of www.thelancet.com/gastrohep Vol 2 January 2017 35 Articles Phylum Family Streptococcaceae Veillonellaceae p=0·012 Bacillales Incerrae Sedis XI Lactobacillaceae Firmicutes Carnobacteriaceae p=0·038 Lachnospiraceae p=0·012 Eubacteriaceae Peptostreptococcaceae Erysipelotrichaceae p=0·0021 Prevotellaceae Bacteroidetes Porphyromonadaceae Flavobacteriaceae Pasteurellaceae Helicobacteraceae Proteobacteria Neisseriaceae p=0·0175 Enterobacteriaceae Campylobacteraceae p=0·00038 Leptotrichiaceae Fusobacteria Fusobacteriaceae Micrococcaceae Actinobacteria Coriobacteriaceae p=0·01 Actinomycetaceae p=0·0019 Spirochaetes Spirochaetaceae Control SR1 SR1 family incertae sedis Barrett’s oesophagus TM7 TM7 family incertae sedis Cancer 50 0 0 1 10 20 30 40 50 Proportional abundance (%) Proportional abundance (%) Family Genus Veillonella p=0·002 Dialister Veillonellaceae Selenomonas Megasphaera p=0·0027 Carnobacteriaceae * Granulicatella p=0·037 Oribacterium Lachnospiraceae Catonella Moryella Erysipelotrichaceae Solobacterium p=0·012 Campylobacteraceae ‡ Campylobacter p=0·0004 Olsenella Coriobacteriaceae Atopobium p=0·0090 Actinomycetaceae Actinomyces p=0·0022 0 1 2 3 4 5 6 7 8 Proportional abundance (%) Figure 1: Proportional abundance of microbial taxa (A) Mean proportional abundance of the eight most prevalent phyla and 25 most prevalent families in tissue samples for healthy control patients (n=16), patients with Barrett’s oesophagus (n=17), and patients with oesophageal adenocarcinoma (n=15). Significant differences were calculated with linear discriminant analysis effect size (LEfSe), and error bars are standard error of the mean. (B) Mean proportional abundance of representative genera from significantly enriched families identified in (A). Only genera with overall proportional abundances greater than 0·1% are included and error bars are standard error of the mean. *p<0·05. †p<0·01. ‡p<0·001. 5·3%) and micro aerophilic Campylobacteraceae (p=0·0021, overall proportional abundance 0·4%), and the (p=0·00038, overall proportional abundance 0·2%), the Gram-positive, facultative anaerobic Carnobacteriaceae Gram-positive, anaerobic Lachnospiraceae (p=0·012, over- (p=0·038, overall proportional abundance 1·6%), all proportional abundance 1%) and Erysipelotrichaceae and Actinomycetaceae (p=0·0019, overall proportional 36 www.thelancet.com/gastrohep Vol 2 January 2017 Articles Family Barrett’s oesophagus CAM445 Cancer RS013 Streptococcaceae Cancer ST036 Prevotellaceae Barrett’s oesophagus NEW048 Pasteurellaceae Cancer PS003 Veillonellaceae Barrett’s oesophagus CAM470 Lactobacillaceae Barrett’s oesophagus PHH009 Leptotrichiaceae Cancer AH135 Micrococcaceae Cancer RS010 Bacillales Incertae Sedix XI Cancer ED041 Carnobacteriaceae Control CAM487 Helicobacteraceae Cancer AH169 Porphyromonadaceae Cancer AH085 Fusobacteriaceae Barrett’s oesophagus CAM042 Actinomycetaceae Cancer WG008 Lachnospiraceae Cancer ST041 Peptostreptococcaceae Cancer RS007 Other Barrett’s oesophagus CAM446 Barrett’s oesophagus NEW047 Control NOT039 Control CAM464 Cancer SH071 Cancer AH155 Barrett’s oesophagus CAM249 Barrett’s oesophagus CAM248 Barrett’s oesophagus CAM427 Barrett’s oesophagus CAM459 Barrett’s oesophagus NOT046 Control CAM197 Control CAM161 Cancer ST037 Barrett’s oesophagus UCL055 Control NOT025 Barrett’s oesophagus UCL058 Control NOT032 Control CAM202 Cancer RS006 Control CAM208 Control CAM460 Barrett’s oesophagus UCL038 Barrett’s oesophagus CAM458 Barrett’s oesophagus CAM449 Control CAM153 Control NOT031 Control NOT054 Control CAM216 Control CAM451 Control CAM424 Figure 2: Microbial community composition in Barrett’s oesophagus The oesophageal adenocarcinoma and control patient groups largely cluster away from each other in this Bray-Curtis cluster dendrogram (p=0·002, parsimony test), but there is no significant difference in clustering for Barrett’s oesophagus. Microbial composition is shown at the family level for each tissue sample. Data were sub-sampled at 631 reads per sample. abundance 0·8%; figure 1A). Significant genera within (median 0%, SE 0·007%) in Barrett’s oesophagus samples these families included Veillonella (p=0·002, overall and 0·004% (median 0%, SE 0·003%) in control samples proportional abundance 3·8%), Megasphaera (p=0·0027, (p=0·028). Sequence identity was confirmed where overall proportional abundance 0·3%), Granulicatella possible with NCBI BLAST. One Barrett’s oesophagus (p=0·037, overall proportional abundance 1·6%), sample contained a high proportional abundance of Actinomyces (p=0·0022, overall proportional abundance H pylori sequences (>99%). 0·8%), Solobacterium (p=0·012, overall proportional Typically, oesophageal adenocarcinoma samples abundance 0·3%), and Campylobacter (p=0·0004, overall clustered away from controls in a Bray-Curtis cluster pro portional abundance 0·2%; figure 1B). In oesophageal dendrogram (p=0·002, parsimony test), emphasising adenocarcinoma samples, the Gram-positive, anaerobic the difference in community structure (figure 2). The Coriobacteriaceae was enriched at the family level (p=0·01, Bray-Curtis algorithm describes the dissimilarity overall proportional abundance 1·9%), but there were no between communities by taking into account both the significant genera identified within this family. At the overlap in OTUs that are present and the proportional species level, the Gram-positive, facultative anaerobic abundance of those OTUs. Samples that have fewer Lactobacillus fermentum was enriched in oesophageal overlapping OTUs and OTUs with less similar adenocarcinoma with mean proportional abundance of proportional abundances will cluster separately, and this 0·6% (median 0·009%, SE 0·5%) compared with 0·01% differential clustering was further shown by principal www.thelancet.com/gastrohep Vol 2 January 2017 37 Articles coordinate analysis (p=0·001, AMOVA test; appendix p 1). Alpha diversity refers to the species diversity within a The microbial communities of seven of 15 oesophageal given environment and includes the number of species adenocarcinoma samples were dominated by the (richness) and the proportion of those species (evenness) Gram-positive order Lactobacillales. Of the seven patients within the microbial community. Three indices of alpha with a high proportion of acid-tolerant Lactobacillales, diversity, observed OTU richness (p=0·0012; figure 3A), six were taking antacid drugs. Five oesophageal the Chao estimate of total OTU richness (p=0·0004; adenocarcinoma samples had a high proportional figure 3B), and the Shannon diversity index (p=0·0075; abundance of Streptococcus spp (69–98%) and two figure 3C) showed that diversity was lower in oesophageal samples had a high proportional abundance of adenocarcinoma samples than in controls. In comparison Lactobacillus spp (87% and 92%). NCBI BLAST showed with Barrett’s oesophagus samples, the oesophageal that the representative species were Streptococcus adenocarcinoma samples showed a decrease in observed pneumoniae/mitis, Streptococcus salivarius/vestibularis, OTU richness and the Chao estimate, but not in the Streptococcus parasanguinis, Lactobacillus gasseri, and Shannon diversity index. 13 patients with oesophageal Lactobacillus helveticus/suntoryeus/gallinarum (/ indicates adenocarcinoma had matched normal squamous tissue where it was not possible to differentiate between sampled proximal to the tumour, and in these patients different species using the 16S rRNA gene regions there was no difference in OTU richness between the sequenced). Although abundant in the acidic stomach normal and tumour tissue (p=0·9065; figure 3D). environment, such a high proportional abundance of Similarly, there was no difference for the Chao estimate Lactobacillus spp was an unexpected finding in the (p>0·999) or the Shannon index (p=0·6355). Furthermore, oesophagus. When we examined matched healthy and there was no difference in overall bacterial abundance tumour tissue for the two patients with high Lactobacillus between matched normal squamous and tumour tissue spp, we found that this genus dominated the lower (p=0·782; figure 3E). These results suggested that the oesophagus regardless of disease state (appendix p 2). decreased microbial diversity was pervasive throughout Gram-positive rods were visualised in areas of ulceration the lower oesophagus in oesophageal adenocarcinoma in tumour PS003 with a high proportional abundance of and was independent of the absolute quantity of Lactobacillus spp (appendix p 2). oesophageal bacteria. A p=0·0012 B p=0·0004 C 150 ‡ 200 † 5 p=0·0075 * 4 0 0 Control Barrett’s Cancer Control Barrett’s Cancer Control Barrett’s Cancer oesophagus oesophagus oesophagus D E p=0·9065 p=0·7820 80 30 0 15 Normal squamous Tumour Normal squamous Tumour Sample diagnosis Sample diagnosis Figure 3: Microbial alpha diversity in oesophageal adenocarcinoma (A) Observed richness of bacterial operational taxonomic units (OTUs). (B) The Chao estimate of total OTU richness and (C) the Shannon diversity index are shown for tissue samples from healthy control patients (n=16), patients with Barrett’s oesophagus (n=17), and patients with oesophageal adenocarcinoma (n=15). Statistical significance was calculated with the Kruskal-Wallis test and Dunns multiple comparisons post test. Data were subsampled at 631 reads per sample. (D) Observed richness of bacterial OTUs for paired normal squamous and tumour tissue samples from 13 patients (26 samples), Wilcoxon signed rank test. Data were subsampled at 656 reads per sample. (E) Overall bacterial abundance using 16S rRNA gene qPCR in matched tumour and normal squamous tissue from 16 patients (32 samples), Wilcoxon signed rank test. Error bars represent standard deviation. *p<0·05. †p<0·01. ‡p<0·001. 38 www.thelancet.com/gastrohep Vol 2 January 2017 Number of OTUs Number of OTUs Cycle threshold (Ct) Chao estimate Shannon index Articles We did additional analyses to ensure that the phylum level, the Cytosponge samples contained a higher proportional abundance of Tenericutes in comparison differences noted were not due to sex, age, or acid with the other sample types using LEfSe (p = 4·7 × 10 suppression. We repeated the diversity analysis with ⁵, male patients only (excluding ten control patients, overall proportional abundance 0·2%). At the genus level, four patients with Barrett’s oesophagus, and three the Cytosponge samples contained greater proportional patients with cancer) and the results were consistent, abundances of Fusobacterium (p<0·0001, overall pro- with decreases in diversity in cancer samples compared portional abundance 2%), Megasphaera (p<0·0001, overall with controls for the observed OTU richness (p=0·0029), proportional abundance 1·8%), Campylobacter (p<0·0001, the Chao estimate of total OTU richness (p=0·0017), and overall proportional abundance 1·7%), Capnocytophaga the Shannon diversity index (p=0·0070). We repeated the (p=0·00058, overall proportional abundance 0·7%), and diversity analysis with patients aged 60 years and older Dialister (p<0·0001, overall proportional abundance 0·2%). (excluding nine control patients, two patients with In keeping with these findings, principal coordinate Barrett’s oesophagus, and two patients with cancer) and analysis with the Bray-Curtis algorithm showed that the the results showed a similar trend, which was significant for the observed OTU richness (p=0·0448), the Chao Biopsy Brush estimate of total OTU richness (p=0·0288), but not the Cytosponge Shannon diversity index (p=0·0892). We also did a 0·4 Throat swab subgroup analysis for age within each diagnostic subgroup (using median age as a cutoff within each 0·2 group) and there were no significant differences in diversity for younger versus older patients within any of 0·0 the subgroups. When we excluded patients who were not taking acid suppression (four control patients, –0·2 two patients with Barrett’s oesphagus, four patients with cancer) or unknown acid suppression status (one patient –0·4 with cancer) the results were similar, with decreased –0·4 –0·2 0·0 0·2 0·4 0·6 0·8 PC1=19·6% diversity in oesophageal adenocarcinoma samples compared with controls, as evidenced by the observed B C p<0·0001 p=0·0104 OTU richness (p=0·0065), the Chao estimate of total 35 150 * * OTU richness (p=0·0033), and the Shannon diversity index (p=0·0202). Given that the decrease in microbial diversity in oesophageal adenocarcinoma seemed widespread throughout the oesophagus, we questioned whether the Cytosponge could be a useful tool to sample the microbiota along the entire length of the oesophagus and upper gastrointestinal tract. 15 patients with Barrett’s 15 0 oesophagus and 16 control patients had Cytosponge D E samples that underwent 16S rRNA gene amplicon p=0·0156 p=0·5968 300 5 sequencing with matched endoscopic biopsies and brushes taken from an area of normal squamous oesophagus. 13 of these patients also had swabs of their posterior pharynx to analyse the similarities and differences between the oesophageal and oral microbiota. Overall, 1455 OTUs were identified and mapped to 100 381 genera. Using a cutoff of 0·0001% overall 1 proportional abundance (138 genera classified), 84·1% of 0 0 genera were shared between Cytosponge samples, Biopsy Brush Cytosponge Biopsy Brush Cytosponge biopsies, and brushes, and 83·6% were shared between Figure 4: Comparison of different methods to sample the oesophageal microbiota Cytosponge samples and throat swabs. A stricter cutoff (A) Principal coordinate analysis with the Bray-Curtis algorithm for matched endoscopic biopsies, brushes, and of 0·1% overall proportional abundance (41 genera Cytosponge samples (31 patients) and 13 throat swabs from a subset of these patients. The first axis (PC1) classified) showed supporting reads for 100% of genera accounts for 19·6% of the sample variance and the second axis (PC2) accounts for 6·3% of the variance. Data were in all sample types, suggesting an overlap in community subsampled at 631 reads per sample. (B) Overall bacterial abundance using 16S rRNA gene-based quantitative PCR in matched endoscopic biopsies, brushes, and Cytosponge samples (20 patients), Friedman test and Dunns membership between the oral cavity, oesophagus, and multiple comparisons post test. (C) The observed diversity of bacterial operational taxonomic units (OTUs), gastric cardia. (D) the Chao estimate of total OTU richness, and (E) the Shannon diversity index for matched endoscopic biopsies, Although most microbial taxa overlapped between brushes, and Cytosponge samples (31 patients), Friedman test and Dunns multiple comparisons post test. sample types, the proportional abundances differed. At the Data were subsampled at 631 reads per sample. *p<0·05. †p<0·01. ‡p<0·001. §p<0·0001. www.thelancet.com/gastrohep Vol 2 January 2017 39 Chao estimate Cycle threshold PC2=6·3% Shannon index Number of OTUs Articles A C p=0·0147 p=0·023 p=0·0085 * * 300 400 5 0 0 Control Barrett’s Dysplasia Control Barrett’s Dysplasia Control Barrett’s Dysplasia oesophagus oesophagus oesophagus Figure 5: Microbial alpha diversity in high-grade dysplasia detected with the Cytosponge (A) Observed richness of bacterial operational taxonomic units (OTUs), (B) the Chao estimate of total OTU richness, and (C) the Shannon diversity index for Cytosponge samples taken from normal squamous control patients (n=20), patients with Barrett’s oesophagus (n=24), and patients with high-grade dysplasia (n=23). Statistical significance was calculated with the Kruskal-Wallis test and Dunns multiple comparisons post-test. Data were sub-sampled at 19 303 reads per sample. Error bars represent standard deviation. *p<0·05. Cytosponge samples clustered away from the throat swabs, abundance 0·1%), and unclassified Prevotellaceae endoscopic biopsies, and brushes (p<0·001, AMOVA test; (p=0·047, overall proportional abundance 1·3%). figure 4A). The throat swabs clustered distinctly from all the other sample types as well (p<0·001). There was no Discussion difference in clustering between biopsies and brushes on Our sequencing data showed decreased microbial the principal coordinate analysis plot (p=0·459). diversity and altered community composition in As expected, because of increased sampling surface oesophageal adenocarcinoma. Interestingly, patients with area, quantitative PCR of overall bacterial abundance oesophageal adenocarcinoma appeared to have this showed the quantity of microbial DNA isolated from reduced diversity regardless of whether cancerous or Cytosponge samples was greater than that from matched normal oesophageal tissue was sampled. The genera that biopsies and brushes (20 patients, p<0·0001; figure 4B). were decreased in proportional abundance in oesophageal After subsampling to normalise for sequencing depth, adenocarcinoma included Gram-negative (Veillonella, there was a decrease in observed OTU richness Megasphaera, and Campylobacter) and Gram-positive (p=0·0104; figure 4C) and the Chao estimate of total OTU taxa (Granulicatella, Atopobium, Actinomyces, and richness (p=0·0156; figure 4D) in endoscopic brush Solobacterium). There was significantly increased samples, but no difference for the Shannon index proportional abundance for L fermentum in patients with (p=0·5968; figure 4E). oesophageal adenocarcinoma compared with control To translate our findings to the setting of early detection, patients and those with Barrett’s oesophagus, and there we tested the usefulness of the Cytosponge to detect was a high proportional abundance of acid-tolerant changes in microbial diversity in patients with high-grade Lactobacillales (Lactobacillus spp and Streptococcus spp) dysplasia (n=23). The mean number of reads for in a subset (seven [47%] of 15) of oesophageal Cytosponge samples was 40 753 (SD 9717) and the median adenocarcinoma samples. To translate our findings to was 40 821 (11714), and the proportion of reads that were the setting of early detection, we investigated the use of subsampled was 47% (for 19 303 reads cutoff). The the Cytosponge device for sampling the oesophageal observed OTU richness was decreased in high-grade microbiota in Barrett’s oesophagus and high-grade dysplasia compared with controls (p=0·0147; figure 5A), dysplasia. The Cytosponge had high microbial DNA yield as were the Chao estimate of total OTU richness (p=0·023; and detected significantly decreased diversity in patients figure 5B) and the Shannon index (p=0·0085; figure 5C). with high-grade dysplasia compared with control patients. There was decreased diversity in Barrett’s oesophagus, but Lactobacillales, which are lactic acid bacteria, are so this was significant only for the Shannon index. In named for their ability to produce lactate from the general, the Cytosponge samples showed homogeneous fermentation of carbohydrates and to survive under harsh results for microbiota composition between diagnostic acidic conditions. Their resilience to low pH might groups at the phylum and family levels, suggesting that enable Lactobacillus spp and Streptococcus spp to thrive in the fraction of microbiota sampled from the area of the tumour niche in a subset of patients with oesophageal Barrett’s oesophagus was diluted by the copious bacteria adenocarcinoma, and production of lactic acid by these sampled from the rest of the oesophagus, oral cavity, and bacteria could further acidify the microenvironment. stomach. Despite this, three genera were identified that Lactic acid fermentation can also produce noxious distinguished controls from the other sample types using by-products, such as hydrogen peroxide, that directly LEfSe: Dialister (p=0·027, overall proportional abundance inhibit the growth of competitor bacteria and enable 0·3%), Schlegelella (p=0·016, overall proportional Lactobacillales to dominate the lower oesophagus. Given 40 www.thelancet.com/gastrohep Vol 2 January 2017 Number of OTUs Chao estimate Shannon index Articles the altered microbial composition in oesophageal distinct clustering in principal coordinate analysis. adenocarcinoma samples, it would be interesting to Despite dilution from sampling the upper gastrointestinal correlate microbiota data with expression and activity of tract, it was still possible for the Cytosponge to detect Toll-like receptors (TLRs), particularly TLR2, given the a decrease in diversity and community composition increased proportional abundance of Gram-positive between normal squamous controls and high-grade genera in a subset of cancer samples. Other authors have dysplasia. Similar to the Cytosponge, Fillon and investigated TLR expression in Barrett’s carcinogenesis colleagues described a minimally invasive oesophageal and found overexpression of TLRs 1, 2, 4, 6, and 9 in string test to sample the microbiota in a paediatric 30–33 human oesophageal adenocarcinoma samples and population. The oesophageal string test detected a similar TLRs 1–3, 6, 7, and 9 in a rat reflux model. microbial composition to that in matched oesophageal Although the microbial community structure differed biopsies, but required the patients to remain in hospital significantly in oesophageal adenocarcinoma in our overnight with the string secured to their cheek. study, there was only a modest reduction in diversity in Alternatively, the Cytosponge is a convenient test that can Barrett’s oesophagus and no genera were identified be given in a general practitioner’s or family doctor’s that discriminated between controls and Barrett’s office with the supervision of a trained nurse, and takes 14,39 oesophagus, or between Barrett’s oesophagus and only 5–7 min to complete. The Cytosponge can also oesophageal adenocarcinoma. It is possible that very low provide histological data for inflammatory pathologies abundance genera might be difficult to detect in such as candidal oesophagitis, herpes oesophagitis, and oesophageal biopsies given the low microbial DNA yield, eosinophilic oesophagitis. Our initial results using the and notably some pathogens have been shown to cause Cytosponge are promising, and future test development overt disease while only accounting for a low proportional should focus on longitudinal sampling of the microbiota abundance of the total microbiota, such as Clostridium to monitor changes in microbial diversity over time in a 35 36 37 difficile, Citrobacter rodentium, and Fusobacterium spp. larger cohort of patients. Further research should also Similarly, Amir and colleagues were unable to identify examine the role of diet, dysphagia, and other external any taxa that differentiated between controls (n=15) and influences on the oesophageal microbiota. Barrett’s oesophagus (n=6), or oesophagitis (n=13) using Contributors DRFE designed and performed the experiments, analysed the data, and LEfSe. By contrast, Yang and colleagues reported that wrote the manuscript. AWW designed experiments, analysed 16S rRNA Gram-negative bacteria were significantly enriched in gene data, and critically revised the manuscript. MOD assessed histology Barrett’s oesophagus (n=10) and reflux oesophagitis and immunohistochemistry. JP managed the Wellcome Trust Sanger (n=12) compared with controls (n=12). The main Institute’s Pathogen Genomics Group’s 16S rRNA gene amplicon sequencing activities and provided critical resources and support. limiting factor of these microbiota studies is the relatively RCF supervised the project, provided funding, designed the study, and small sample size and substantial inter-individual critically revised the manuscript. variation in microbiota composition. Another limitation Declaration of interests is that although LEfSe is useful for biological We declare no competing interests. interpretation of metagenomic data, it does not correct Acknowledgments for multiple comparisons, so there is a risk of false This study is supported by Cancer Research UK as part of the International discovery (p value, α=0·05). The inclusion of appropriate Cancer Genome Consortium (RG66287), with infrastructure support from negative controls and replicate samples is also paramount the Cambridge Biomedical Research Centre (Cancer Research UK) and the Cambridge Experimental Medicine Centre (National Institute for Health for low microbial biomass samples to facilitate removal Research [NIHR]), a Programme Grant to RCF from the Medical Research of contaminant OTUs that might also lead to false Council (RG72746), an NIHR Professorship to RCF (RG67258), and a discovery, and this was a major strength of our study. United European Gastroenterology prize to DRFE. Funding for AWW, JP, We also imposed strict quality control criteria, resulting and 16S rRNA gene amplicon sequencing was provided by the Wellcome Trust (098051). AWW receives core funding support from the Scottish in the exclusion of 14 tissue samples with low sequencing Government Rural and Environmental Science and Analysis Service read numbers and Good’s coverage estimates. The (RESAS). We thank the Wellcome Trust Sanger Institute’s core sequencing difficulty in obtaining good quality sequencing data from teams for generating 16S rRNA gene data and the Human Research Tissue oesophageal samples highlights the potential use of the Bank for the sample collection, which is supported by the NIHR Cambridge Biomedical Research Centre. The following UK hospitals Cytosponge device, which samples a larger surface area. participated in sample collection through the OCCAMS collaboration Our results suggest that it is feasible to sample network: Cambridge University Hospitals NHS Foundation Trust, oesophageal microbiota using the Cytosponge, and the Cambridge; Edinburgh Royal Infirmary, Edinburgh; Royal Surrey County device detected the majority of genera present in Hospital National Health Service (NHS) Foundation Trust, Guildford; University of Southampton and University Hospital Southampton NHS endoscopic biopsies and brushes. The high microbial Foundation Trust, Southampton; St Thomas’ Hospital, London; and Wigan DNA yield collected by the Cytosponge reflects sampling and Leigh NHS Foundation Trust, Manchester. The following UK hospitals of the entire length of the oesophagus as well as the contributed samples as part of the BEST2 study: Cambridge University proximal stomach and oral cavity as it is withdrawn. Hospitals NHS Foundation Trust, Cambridge; Nottingham University Hospitals NHS Trust, Nottingham; University College Hospital London, The throat swabs showed similarities in community London; Newcastle Upon Tyne Hospital, Newcastle; and Hinchingbrooke membership between the oral cavity and oesophagus, Hospital, Huntingdon. but the proportional abundances differed, as shown by www.thelancet.com/gastrohep Vol 2 January 2017 41 Articles References 19 Kozich JJ, Westcott SL, Baxter NT, Highlander SK, Schloss PD. 1 Pohl H, Welch HG. The role of overdiagnosis and reclassification in Development of a dual-index sequencing strategy and curation the marked increase of esophageal adenocarcinoma incidence. pipeline for analyzing amplicon sequence data on the MiSeq Illumina J Natl Cancer Inst 2005; 97: 142–46. sequencing platform. Appl Environ Microbiol 2013; 79: 5112–20. 2 Picardo SL, Maher SG, O’Sullivan JN, Reynolds JV. Barrett’s to 20 Salter SJ, Cox MJ, Turek EM, et al. Reagent and laboratory oesophageal cancer sequence: a model of Inflammatory-driven upper contamination can critically impact sequence-based microbiome gastrointestinal cancer. Dig Surg 2012; 29: 251–60. analyses. BMC Biol 2014; 12: 87. 3 Singh S, Sharma AN, Murad MH, et al. Central adiposity is 21 Paulson JN, Pop M, Bravo HC. Metastats: an improved statistical associated with increased risk of esophageal inflammation, method for analysis of metagenomic data. Genome Biol 2011; 12: 1–27. metaplasia, and adenocarcinoma: a systematic review and 22 Good IJ. The population frequencies of species and the estimation of meta-analysis. Clin Gastroenterol Hepatol 2013; 11: 1399–412. population parameters. Biometrika 1953; 40: 237–64. 4 Nie S, Chen T, Yang X, Huai P, Lu M. Association of Helicobacter pylori 23 Zaura E, Keijser BJ, Huse SM, Crielaard W. Defining the healthy infection with esophageal adenocarcinoma and squamous cell “core microbiome” of oral microbial communities. BMC Microbiol carcinoma: a meta-analysis. Dis Esophagus 2014; 27: 645–53. 2009; 9: 259. 5 Amir I, Konikoff FM, Oppenheim M, Gophna U, H alf EE. 24 Li K, Bihan M, Methe BA. Analyses of the stability and core taxonomic Gastric microbiota is altered in esophagitis and Barrett’s esophagus memberships of the human microbiome. PLoS One 2013; 8: e63139. and further modified by proton pump inhibitors. Environ Microbiol 25 Morgulis A, Coulouris G, Raytselis Y, Madden TL, Agarwala R, 2013; 16: 2905–14. Schäffer AA. Database indexing for production MegaBLAST 6 Wu S, Rhee K-J, Albesiano E, et al. A human colonic commensal searches. Bioinformatics 2008; 24: 1757–64. promotes colon tumorigenesis via activation of T helper type 17 T cell 26 Nadkarni MA, Martin FE, Jacques NA, Hunter N. Determination of responses. Nat Med 2009; 15: 1016–22. bacterial load by real-time PCR using a broad-range (universal) probe 7 Arthur JC, Perez-Chanona E, Mühlbauer M, et al. and primers set. Microbiology 2002; 148: 257–66. Intestinal inflammation targets cancer-inducing activity of the 27 Segata N, Izard J, Waldron L, et al. Metagenomic biomarker microbiota. Science 2012; 338: 120–23. discovery and explanation. Genome Biol 2011; 12: R60. 8 Hu B, Elinav E, Huber S, et al. Microbiota-induced activation of 28 Shaheen NJ, Richter JE. Barrett’s oesophagus. Lancet 2009; epithelial IL-6 signaling links inflammasome-driven inflammation 373: 850–61. with transmissible cancer. Proc Natl Acad Sci USA 2013; 110: 9862–67. 29 Kleerebezem M, Hols P, Bernard E, et al. The extracellular biology 9 Kostic AD, Chun E, Robertson L, et al. Fusobacterium nucleatum of the lactobacilli. FEMS Microbiol Rev 2010; 34: 199–230. potentiates intestinal tumorigenesis and modulates the 30 Huhta H, Helminen O, Kauppila JH, et al. Toll-like receptor 9 tumor-immune microenvironment. Cell Host Microbe 2013; expression in the natural history of Barrett mucosa. Virchows Arch 14: 207–15. 2015; 467: 9–18. 10 Rubinstein MR, Wang X, Liu W, Hao Y, Cai G, Han YW. 31 Huhta H, Helminen O, Lehenkari PP, Saarnio J, Karttunen TJ, Fusobacterium nucleatum promotes colorectal carcinogenesis by Kauppila JH. Toll-like receptors 1, 2, 4 and 6 in esophageal modulating E-cadherin/β-catenin signaling via its fadA adhesin. epithelium, Barrett’s esophagus, dysplasia and adenocarcinoma. Cell Host Microbe 2013; 14: 195–206. Oncotarget 2016; 7: 23658–67. 11 Yang L, Lu X, Nossa CW, Francois F, Peek RM, Pei Z. 32 Verbeek RE, Siersema PD, Ten Kate FJ, et al. Toll-like receptor 4 Inflammation and intestinal metaplasia of the distal esophagus are activation in Barrett’s esophagus results in a strong increase in associated with alterations in the microbiome. Gastroenterology 2009; COX-2 expression. J Gastroenterol 2013; 49: 1121–34. 137: 588–97. 33 Kauppila JH, Takala H, Selander KS, Lehenkari PP, Saarnio J, 12 Gall A, Fero J, McCoy C, et al. Bacterial composition of the human Karttunen TJ. Increased Toll-like receptor 9 expression indicates upper gastrointestinal tract microbiome is dynamic and associated adverse prognosis in oesophageal adenocarcinoma. Histopathology with genomic instability in a Barrett’s esophagus cohort. PLoS One 2011; 59: 643–49. 2015; 10: e0129055. 34 Zaidi AH, Kelly LA, Kreft RE, et al. Associations of microbiota and 13 Blackett KL, Siddhi SS, Cleary S, et al. Oesophageal bacterial biofilm toll-like receptor signaling pathway in esophageal adenocarcinoma. changes in gastro-oesophageal reflux disease, Barrett’s and BMC Cancer 2015; 16: 52. oesophageal carcinoma: association or causality? 35 Buffie CG, Jarchum I, Equinda M, et al. Profound alterations of Aliment Pharmacol Ther 2013; 37: 1084–92. intestinal microbiota following a single dose of clindamycin results 14 Kadri SR, Lao-Sirieix P, O’Donovan M, et al. Acceptability and in sustained susceptibility to Clostridium difficile-induced colitis. accuracy of a non-endoscopic screening test for Barrett’s oesophagus Infect Immun 2012; 80: 62–73. in primary care: cohort study. BMJ 2010; 341: c4372. 36 Clare S, John V, Walker AW, et al. Enhanced susceptibility to 15 Ross-Innes CS, Debiram-Beecham I, O’Donovan M, et al. Citrobacter rodentium infection in microRNA-155-deficient mice. Evaluation of a minimally invasive cell sampling device coupled Infect Immun 2013; 81: 723–32. with assessment of trefoil factor 3 expression for diagnosing 37 Kostic AD, Gevers D, Pedamallu CS, et al. Genomic analysis Barrett’s esophagus: a multi-center case-control study. PLoS Med identifies association of Fusobacterium with colorectal carcinoma. 2015; 12: e1001780. Genome Res 2012; 22: 292–98. 16 Benaglia T, Sharples LD, Fitzgerald RC, Lyratzopoulos G. 38 Fillon SA, Harris JK, Wagner BD, et al. Novel device to sample the Health benefits and cost effectiveness of endoscopic and esophageal microbiome—the esophageal string test. PLoS One nonendoscopic cytosponge screening for Barrett’s esophagus. 2012; 7: e42938. Gastroenterology 2013; 144: 62–73. 39 Vaughan TL, Fitzgerald RC. Precision prevention of oesophageal 17 Cangelosi GA, Meschke JS. Dead or alive: molecular assessment of adenocarcinoma. Nat Rev Gastroenterol Hepatol 2015; 12: 243–48. microbial viability. Appl Environ Microbiol 2014; 80: 5884–91. 40 Paterson AL, Lao-Sirieix P, O’Donovan M, et al. Range of 18 Schloss PD, Westcott SL, Ryabin T, et al. Introducing mothur: pathologies diagnosed using a minimally invasive capsule sponge open-source, platform-independent, community-supported software to evaluate patients with reflux symptoms. Histopathology 2016; for describing and comparing microbial communities. published online July 15. DOI:10.1111/his.13039. Appl Environ Microbiol 2009; 75: 7537–41. 42 www.thelancet.com/gastrohep Vol 2 January 2017 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png The Lancet Gastroenterology & Hepatology Unpaywall

A non-endoscopic device to sample the oesophageal microbiota: a case-control study

Elliott, Daffolyn R Fels; Walker, Alan W; O'Donovan, Maria; Parkhill, Julian; Fitzgerald, Rebecca C
The Lancet Gastroenterology & Hepatology – Jan 1, 2017
Download PDF
11 pages

Loading next page...
 
/lp/unpaywall/a-non-endoscopic-device-to-sample-the-oesophageal-microbiota-a-case-1VBBtgUQM3

References (43)

Publisher
Unpaywall
ISSN
2468-1156
DOI
10.1016/s2468-1253(16)30086-3
Publisher site
See Article on Publisher Site

Abstract

Articles A non-endoscopic device to sample the oesophageal microbiota: a case-control study Daffolyn R Fels Elliott, Alan W Walker, Maria O’Donovan, Julian Parkhill, Rebecca C Fitzgerald Summary Lancet Gastroenterol Hepatol Background The strongest risk factor for oesophageal adenocarcinoma is reflux disease, and the rising incidence of 2017; 2: 32–42 this coincides with the eradication of Helicobacter pylori, both of which might alter the oesophageal microbiota. Published Online We aimed to profile the microbiota at different stages of Barrett’s carcinogenesis and investigate the Cytosponge as a November 11, 2016 minimally invasive tool for sampling the oesophageal microbiota. http://dx.doi.org/10.1016/ S2468-1253(16)30086-3 Methods In this case-control study, 16S rRNA gene amplicon sequencing was done on 210 oesophageal samples from This online publication has 86 patients representing the Barrett’s oesophagus progression sequence (normal squamous controls [n=20], non-dysplastic been corrected twice. The first corrected version appeared at [n=24] and dysplastic Barrett’s oesophagus [n=23], and oesophageal adenocarcinoma [n=19]), relevant negative controls, thelancet.com/gastrohep on and replicates on the Illumina MiSeq platform. Samples were taken from patients enrolled in the BEST2 study at five UK December 9, 2016, and the hospitals and the OCCAMS study at six UK hospitals. We compared fresh frozen tissue, fresh frozen endoscopic second on October 4, 2017 brushings, and the Cytosponge device for microbial DNA yield (qPCR), diversity, and community composition. See Comment page 4 Medical Research Centre Cancer Findings There was decreased microbial diversity in oesophageal adenocarcinoma tissue compared with tissue from Unit, Hutchison/MRC Research Centre, University of healthy control patients as measured by the observed operational taxonomic unit (OTU) richness (p=0·0012), Chao Cambridge, Cambridge, UK estimated total richness (p=0·0004), and Shannon diversity index (p=0·0075). Lactobacillus fermentum was enriched (D R Fels Elliott MD, in oesophageal adenocarcinoma (p=0·028), and lactic acid bacteria dominated the microenvironment in seven (47%) Prof R C Fitzgerald MD); of 15 cases of oesophageal adenocarcinoma. Comparison of oesophageal sampling methods showed that the Pathogen Genomics Group, Wellcome Trust Sanger Cytosponge yielded more than ten-times higher quantities of microbial DNA than did endoscopic brushes or biopsies Institute, Hinxton, UK using quantitative PCR (p<0·0001). The Cytosponge samples contained the majority of taxa detected in biopsy and (A W Walker PhD, brush samples, but were enriched for genera from the oral cavity and stomach, including Fusobacterium, Megasphaera, Prof J Parkhill PhD); Rowett Campylobacter, Capnocytophaga, and Dialister. The Cytosponge detected decreased microbial diversity in patients with Institute of Nutrition and Health, University of Aberdeen, high-grade dysplasia in comparison to control patients, as measured by the observed OTU richness (p=0·0147), Chao Aberdeen, UK (A W Walker); and estimated total richness (p=0·023), and Shannon diversity index (p=0·0085). Department of Histopathology, Cambridge Interpretation Alterations in microbial communities occur in the lower oesophagus in Barrett’s carcinogenesis, which University Hospital NHS Trust, Cambridge, UK can be detected at the pre-invasive stage of high-grade dysplasia with the novel Cytosponge device. Our findings are (M O’Donovan MD) potentially applicable to early disease detection, and future test development should focus on longitudinal sampling Correspondence to: of the microbiota to monitor for changes in microbial diversity in a larger cohort of patients. Prof Rebecca C Fitzgerald, Medical Research Centre Cancer Funding Cancer Research UK, National Institute for Health Research, Medical Research Council, Wellcome Trust, Unit, Hutchison/MRC Research Centre, University of Cambridge, The Scottish Government (RESAS). Cambridge CB2 0XZ, UK [email protected] Copyright © The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 license. Introduction with the eradication of Helicobacter pylori, which could alter Oesophageal adenocarcinoma is an aggressive malignancy the composition of microbiota and promote bacterial with poor outcomes that generally develops from a overgrowth. Furthermore, reflux disease is treated with premalignant columnar epithelium called Barrett’s antacid drugs such as proton-pump inhibitors, which have oesophagus. The incidence of oesophageal adenocarcinoma profound effects on gastric acidity and might affect the has increased by six times in developed countries during gastro-oesophageal microbiota. the past three decades. Both Barrett’s oesophagus and There is growing evidence linking abnormal changes in oesophageal adeno carcinoma are thought to develop in the microbiota, known as dysbiosis, with human cancer. response to chronic acid reflux in the lower oesophagus, One of the best described examples is colon carcinoma, which precipitates inflammation and mucosal injury over in which gastrointestinal microbiota have been shown to time. Reflux disease has increased with the obesity promote carcinogenesis in the setting of colonic 6–8 epidemic and altered eating habits in high-income inflammation. Studies have also linked Fusobacterium countries, and central adiposity might also influence nucleatum to colon carcinoma through an altered tumour 3 9,10 carcinogenesis through the release of adipokines. immuno-environment, but without associated colitis. Additionally, epidemiological evidence suggests that the The oesophagus has far fewer bacteria than the colon; rising incidence of oesophageal adenocarcinoma coincides nevertheless, alterations in the microbiota might occur in 32 www.thelancet.com/gastrohep Vol 2 January 2017 Articles Research in context Evidence before this study sequencing, and compares patients with oesophageal Epidemiological evidence suggests that the rising incidence of adenocarcinoma and Barrett’s oesophagus with healthy control oesophageal adenocarcinoma coincides with the obesity patients. We found decreased microbial diversity in oesophageal epidemic, gastro-oesophageal reflux disease, and eradication of adenocarcinoma tissue compared with controls, with Helicobacter pylori with antibiotics and acid suppression enrichment of acid-tolerant bacteria such as Lactobacillus treatment—all risk factors that are capable of altering the fermentum. The microbial diversity was reduced in the lower gastro-oesophageal microbiota. Three studies with small oesophagus regardless of whether cancerous or healthy numbers of patients have shown modest alterations in the oesophageal tissue was sampled within the same patients. microbiota in Barrett’s oesophagus and oesophagitis with We further translated our findings to the setting of early 16S rRNA gene amplicon sequencing. However, studies using detection, using the Cytosponge to sample the microbiota in culture-independent methods to profile the oesophageal Barrett’s oesophagus and high-grade dysplasia. We showed that microbiota in oesophageal adenocarcinoma or high-grade the Cytosponge collected high microbial DNA yield and dysplasia are lacking. One explanation for the scarcity of detected decreased diversity in the pre-invasive stage of oesophageal microbiota studies is the challenge of endoscopic high-grade dysplasia. sampling and low microbial DNA yield. Based on previous studies, Implications of all the available evidence we postulated that the novel Cytosponge could be an effective Alterations in microbial communities occur in the lower tool to sample the microbiota along the entire length of oesophagus in Barrett’s carcinogenesis, which are possible to the oesophagus. detect with the minimally invasive Cytosponge. Our findings are Added value of this study potentially applicable to early disease detection, and future test This study provides a comprehensive characterisation of the development should focus on longitudinal sampling of the microbiota at different stages of the Barrett’s oesophagus microbiota to monitor for changes in microbial diversity in a progression sequence using 16S rRNA gene amplicon larger cohort of patients. 11–13 reflux oesophagitis and Barrett’s oesophagus. However, Methods the role of the microbiota in Barrett’s carcinogenesis is Study design and participants not clearly defined and there is no clinical reference In this case-control study, endoscopic biopsies, brushes, standard at present for sampling the oesophageal Cytosponge samples, and throat swabs were collected microbiota. One of the challenges in studying the from patients with a diagnosis of non-dysplastic Barrett’s oesophageal microbiota is that endoscopy is an invasive oesophagus or high-grade dysplasia, and from control test that provides only a focal sampling of the microbiota patients with symptoms of reflux or dyspepsia enrolled in biopsy samples, and a slightly larger surface area with in the Barrett’s Oesophagus Screening Trial (BEST2) at endoscopic brushings. Minimally invasive methods for five UK hospitals. In patients with a diagnosis of sampling the oesophageal microbiota could be clinically oesophageal adenocarcinoma, tissue samples from the useful for detection and risk stratification of patients with tumour and matched normal squamous oesophagus Barrett’s oesophagus. were collected from six UK hospitals participating in Here we investigate the Cytosponge prototype Oesophageal Cancer Clinical and Molecular Stratification (Europlaz, Southminster, UK) as a non-endoscopic (OCCAMS) for the International Cancer Genome cell-sampling device that can collect a representative Consortium (ICGC). 14,15 sample of cells along the length of the oesophagus. The The patient inclusion criteria included age between device consists of a spherical mesh that is compressed 20 and 90 years with either normal endoscopy or within a gelatine capsule and attached to a string. Once endoscopic and histological documentation of Barrett’s swallowed, the capsule dissolves and the Cytosponge oesophagus or oesophageal adenocarcinoma. The expands in the patient’s stomach before being withdrawn exclusion criteria were current infection, recent antibiotic on a string through the patient’s mouth. We have treatment, previous chemotherapy treatment, and previously shown that this device is a safe, acceptable documented pathological findings unrelated to Barrett’s method for diagnosing Barrett’s oesophagus, with oesophagus or oesophageal adenocarcinoma. Ethical 14–16 promising accuracy and cost-effectiveness. The goal approval was obtained from the National Research Ethics of this study was to provide a comprehensive description Services Cambridgeshire Research Ethics Committee of the microbiota in the different pathogenic stages of on behalf of all hospital centres in the BEST2 trial oesophageal adenocarcinoma using 16S rRNA gene (REC 10/H0308/71) and the OCCAMS/ICGC trial amplicon sequencing and to test the feasibility of the (REC 07/H0305/52 and 10/H0305/51). Written informed Cytosponge to detect changes in the microbiota occurring consent was obtained from all participants before the in Barrett’s oesophagus and high-grade dysplasia. collection of samples and recording of clinical information. www.thelancet.com/gastrohep Vol 2 January 2017 33 Articles Procedures accession number ERP005191. The 16S rRNA gene For patients enrolled in the BEST2 study, matched amplicon sequence analysis was done with mothur. endoscopic biopsies were taken from an area of The MiSeq standard operating procedure was Barrett’s oesophagus and proximal normal squamous followed with the exception of chimera checking, oesophagus. Endoscopic brushings were taken from an which was done with chimera.perseus, and unique area of normal squamous oesophagus only. The sequences were removed using the split.abund and tissue samples collected in the OCCAMS/ICGC study remove.seqs commands before building the distance included endoscopic biopsies, endoscopic mucosal matrix. Contaminant operational taxonomic units resection specimens, and surgical biopsies after (OTUs) were defined as having greater proportional oesophagectomy. Sampling from oesophagectomy abundance in negative controls, alongside previous specimens was done with a sterile scalpel blade (cutting evidence that these OTUs were derived from genera down to submucosa) within 1 h of surgical resection. that are common contaminants, and these reads were All samples were flash frozen in liquid nitrogen and removed using remove.seqs (3434 OTUs defined as stored at –80°C except for the Cytosponge samples, contaminants from 5757 total OTUs). Comparisons which were preserved in BD SurePath liquid at 4°C. All between replicate samples (using Metastats as patients fasted overnight before endoscopy or surgery. implemented in mothur) showed no significant As part of routine perioperative procedure, patients differences so replicates were pooled using the Linux with oesophageal adenocarcinoma who underwent sed command to maximise the number of reads per oeso phagectomy received prophylactic intravenous sample, and samples with fewer than 550 reads or antibiotics at the time of surgery, up to 6 h before the Good’s coverage less than 95% were removed with research samples were obtained. The very close timing remove.groups. The Good’s coverage estimator is of this perioperative antibiotic exposure should not used to assess what proportion of the total OTUs greatly affect microbial community composition present within a given sample are detected in the profiles because 16S rRNA gene amplicon sequencing sequencing results, and thus gives an indication of detects both live and dead bacterial cells cross- how thorough the sampling has been at the chosen See Online for appendix sectionally (appendix p 1). sequencing depth. For measures of diversity that Cytosponge samples were vortexed and centrifuged are sensitive to the sequencing depth, random to pellet cellular debris (215 g for 5 min), and the subsampling was done at the lowest number of reads residual supernatant was used for microbial DNA per sample using the sub.sample command in mothur. extraction after further high-speed centrifugation Data were subsampled at 631 reads for the analysis of (14 000 g for 10 min). DNA was isolated from all tissue samples (median Good’s coverage 96·36%, oesophageal samples using the Precellys Soil DNA Kit range 92·08–99·37), 656 reads for matched tumour- (Peqlab, Southampton, UK). The 16S rRNA gene was normal pairs (97·41%, range 95·27–98·78), 631 reads amplified using primers for the V1-V2 region: 27F for the comparison of different sampling methods 5 ʹ AATGATACGGCGACCACCGAGATCTACAC (96·51%, range 91·13–99·84), and 19 303 reads for TATGGTAATT CC AGMGTTYGATYMTGGCTCAG the analysis of Cytosponge samples (99·83%, range and 338R 5ʹCAAGCAGAAGACGGCATACGAGAT 99·70–99·95). When determining the shared genera NNNNNNNNNNNN AGTCAGTCAG AA between Cytosponge samples, biopsies, and brushes, a GCTGCCTCCCGTAGGAGT, where Illumina adapter cutoff of 0·0001% proportional abundance was used to sequences are at the 5ʹ end, and the N string is a unique focus on the more abundant OTUs that are less likely barcode. Most samples had two barcoded replicates to to be susceptible to errors introduced by subsampling ensure reproducibility (labelled A or B in the (below 0·0001% there were less than 27 reads appendix p 5), and negative controls from every DNA supporting each OTU across all the samples). A second extraction step underwent additional PCR cycles to cutoff value of 0·1% was chosen arbitrarily to show the identify contaminant organisms. The reaction similarity between sample types and is a common 23,24 conditions were 98°C for 2 min, 25 cycles at 98°C for cutoff used in previous studies. Sequence identity 30 s, 50°C for 30 s, and 72°C for 90 s, and extension at was confirmed at the species level, where possible, by 72°C for 5 min. Negative controls were kit reagents or carrying out NCBI BLAST analysis on representative nuclease-free water, which underwent 45 amplification sequences using MegaBLAST. cycles. All samples were amplified in duplicate Quantitative PCR for the 16S rRNA gene was done with and pooled to minimise PCR bias and maximise SYBR Green I Master Mix (Roche, Mannheim, Germany) yield. The PCR products were concentrated using on the LC480 LightCycler 480 II (Roche, Mannheim, ethanol precipitation and quantified using a Qubit Germany), in triplicate. The reaction conditions were 2·0 Fluorometer before sequencing on the MiSeq 95°C for 5 min, 40 cycles at 95°C for 15 s, 60°C for 30 s, Illumina platform using 2 × 250 bp read length. and 72°C for 90 s, and a melt curve. The primer sequences For the European Nucleotide 16S rRNA gene amplicon sequencing data have been were 331F 5ʹTCCTACGGGAGGCAGCAGT and 797R Archive see http://www.ebi. ac.uk/ena/data/view/ERP005191 deposited in the European Nucleotide Archive under 5ʹGGACTACCAGGGTATCTAATCCTGTT. 34 www.thelancet.com/gastrohep Vol 2 January 2017 Articles Statistical analysis 24 patients with Barrett’s oesophagus, 12 (71%) of The non-parametric Kruskal-Wallis test and Dunns 17 patients with oesophageal adenocarcinoma (two did not multiple comparisons post-test were used for comparisons report whether or not they were taking acid-suppressant between diagnostic groups in Graphpad Prism (version 6). drugs), and 15 (75%) of 20 control patients. One patient Within Graphpad, the Wilcoxon signed rank test was used with Barrett’s oesophagus and two patients with for analyses involving matched tumour-normal pairs, and high-grade dysplasia reported taking a course of antibiotics the Friedman test and Dunns post-test were used for within the past month. analyses of matched samples from endoscopic biopsies, After filtering the sequencing data and removing brushes, and the Cytosponge. We used LEfSe, a contaminant sequences, the mean number of reads metagenomic biomarker discovery method, to identify for tissue samples was 6649 (SD 10421) and the microbial taxa that differed significantly between controls, median was 3064 (5688) and the proportion of reads that Barrett’s oesophagus, and oesophageal adenocarcinoma were subsampled for diversity analyses was 9·5% (for samples. LEfSe was used within mothur. All sequencing 631 reads cutoff) and 9·9% (for 656 reads cutoff). reads were included for the composition analysis using 14 tissue samples did not meet quality criteria and LEfSe. LEfSe ranks OTUs in the order that it considers were excluded from the analysis, leaving 16 control, these taxa to be most likely to explain differences between 17 Barrett’s oesophagus, and 15 oesophageal microbial communities using linear discriminant analysis adenocarcinoma samples. Five phyla accounted for the to estimate effect size. A full explanation of the statistical majority of sequencing reads in the dataset: Firmicutes approaches used in LEfSe can be found in the original (59·9%), Bacteroidetes (15·1%), Proteobacteria (12·8%), article by Segata and colleagues. Within mothur, Actinobacteria (5·8%), and Fusobacteria (5·4%). the Bray-Curtis calculator was used to describe the 1060 OTUs were identified and classified as belonging to dissimilarity between communities by taking into account 345 different genera. both the overlap in OTUs that are present between By LEfSe, at the phylum level, the Barrett’s samples samples and the proportional abundance of those OTUs contained a higher proportional abundance of in each sample. Using dissimilarity information calculated Proteobacteria (mean 18·6%, median 14·8%, SE 5·5%) with the Bray-Curtis calculator, the parsimony test and the compared with controls (mean 8·5%, median 8·2%, analysis of molecular variance (AMOVA) test were used to SE 5·5%; p=0·017) and oesophageal adeno carcinoma show significantly different clustering between microbiota samples (mean 7·6%, median 3·9%, SE 2·4%). The profiles from the different diagnostic groups. IBM SPSS control samples were enriched for several taxa at the Statistics (version 24) was used to analyse patient data family level, including the Gram-negative, anaerobic (ANOVA for mean age and Fisher’s exact test for sex, Veillonellaceae (p=0·012, overall proportional abundance ethnicity, and antacid usage). A significant p value was defined as less than 0·05. Control Non-dysplastic High-grade Oesophageal p value (n=20) Barrett’s dysplasia adenocarcinoma Role of the funding source oesophagus (n=23) (n=19) (n=24) The funders had no role in the study design, data analysis, interpretation of data, or writing of the report. The Age, years 57 (29–86) 68 (53–79) 65 (50–82) 70 (44–79) 0·001 corresponding author had full access to all the data in the Sex study and had final responsibility for the decision to Male 7 (35%) 16 (67%) 19 (83%) 15 (79%) 0·006 submit for publication. Female 13 (65%) 8 (33%) 4 (17%) 4 (21%) ·· White ethnicity 19 (95%) 24 (100%) 23 (100%) 16 (89%)* 0·089 Results Antacid usage 15 (75%) 22 (92%) 23 (100%) 12 (71%)* 0·011 To investigate whether the development of oesophageal Samples passing quality control/samples sequenced adenocarcinoma was associated with dysbiosis, we did Cytosponge 20/20 24/24 23/23 ·· 16S rRNA gene amplicon sequencing on tissue samples Brush (squamous only) 19/19 19/19 ·· ·· ·· from patients with oesophageal adenocarcinoma (n=19), Tissue 16/19 17/24 (Barrett’s ·· 15/19 (oesophageal ·· Barrett’s oesophagus (n=24), and healthy control patients oesophagus); adenocarcinoma); 15/24 (squamous)† 15/19 (squamous)† (n=19; table). Patients with oesophageal adenocarcinoma Total number of 58 91 23 38 ·· and Barrett’s oesophagus were older (p=0·001) and samples sequenced predominantly male (p=0·025) in comparison with the control patients, which is consistent with the known Data are n (%) or median (range) unless otherwise stated. ··=samples were not sequenced for all diagnostic categories for the following reasons: it was not safe or appropriate to have patients with cancer swallow the Cytosponge; we epidemiology of this disease. The number of patients compared brush sampling methods for areas of normal oesophagus in patients with no pathology or Barrett’s recruited at each participating hospital centre and oesophagus only; and tissue samples were not available for dysplasia. †Tissue samples from the area of Barrett’s additional clinicopathological data for patients with oesophagus or tumour and matched normal squamous oesophagus were taken in the same cases. *Not recorded, oesophageal adenocarcinoma is provided in the appendix ethnicity unknown for one patient and antacid usage unknown for two patients. (pp 3, 4). Acid-suppressant drugs were taken regularly Table: Patient demographics and oesophageal samples by 23 (100%) of 23 patients with dysplasia, 22 (92%) of www.thelancet.com/gastrohep Vol 2 January 2017 35 Articles Phylum Family Streptococcaceae Veillonellaceae p=0·012 Bacillales Incerrae Sedis XI Lactobacillaceae Firmicutes Carnobacteriaceae p=0·038 Lachnospiraceae p=0·012 Eubacteriaceae Peptostreptococcaceae Erysipelotrichaceae p=0·0021 Prevotellaceae Bacteroidetes Porphyromonadaceae Flavobacteriaceae Pasteurellaceae Helicobacteraceae Proteobacteria Neisseriaceae p=0·0175 Enterobacteriaceae Campylobacteraceae p=0·00038 Leptotrichiaceae Fusobacteria Fusobacteriaceae Micrococcaceae Actinobacteria Coriobacteriaceae p=0·01 Actinomycetaceae p=0·0019 Spirochaetes Spirochaetaceae Control SR1 SR1 family incertae sedis Barrett’s oesophagus TM7 TM7 family incertae sedis Cancer 50 0 0 1 10 20 30 40 50 Proportional abundance (%) Proportional abundance (%) Family Genus Veillonella p=0·002 Dialister Veillonellaceae Selenomonas Megasphaera p=0·0027 Carnobacteriaceae * Granulicatella p=0·037 Oribacterium Lachnospiraceae Catonella Moryella Erysipelotrichaceae Solobacterium p=0·012 Campylobacteraceae ‡ Campylobacter p=0·0004 Olsenella Coriobacteriaceae Atopobium p=0·0090 Actinomycetaceae Actinomyces p=0·0022 0 1 2 3 4 5 6 7 8 Proportional abundance (%) Figure 1: Proportional abundance of microbial taxa (A) Mean proportional abundance of the eight most prevalent phyla and 25 most prevalent families in tissue samples for healthy control patients (n=16), patients with Barrett’s oesophagus (n=17), and patients with oesophageal adenocarcinoma (n=15). Significant differences were calculated with linear discriminant analysis effect size (LEfSe), and error bars are standard error of the mean. (B) Mean proportional abundance of representative genera from significantly enriched families identified in (A). Only genera with overall proportional abundances greater than 0·1% are included and error bars are standard error of the mean. *p<0·05. †p<0·01. ‡p<0·001. 5·3%) and micro aerophilic Campylobacteraceae (p=0·0021, overall proportional abundance 0·4%), and the (p=0·00038, overall proportional abundance 0·2%), the Gram-positive, facultative anaerobic Carnobacteriaceae Gram-positive, anaerobic Lachnospiraceae (p=0·012, over- (p=0·038, overall proportional abundance 1·6%), all proportional abundance 1%) and Erysipelotrichaceae and Actinomycetaceae (p=0·0019, overall proportional 36 www.thelancet.com/gastrohep Vol 2 January 2017 Articles Family Barrett’s oesophagus CAM445 Cancer RS013 Streptococcaceae Cancer ST036 Prevotellaceae Barrett’s oesophagus NEW048 Pasteurellaceae Cancer PS003 Veillonellaceae Barrett’s oesophagus CAM470 Lactobacillaceae Barrett’s oesophagus PHH009 Leptotrichiaceae Cancer AH135 Micrococcaceae Cancer RS010 Bacillales Incertae Sedix XI Cancer ED041 Carnobacteriaceae Control CAM487 Helicobacteraceae Cancer AH169 Porphyromonadaceae Cancer AH085 Fusobacteriaceae Barrett’s oesophagus CAM042 Actinomycetaceae Cancer WG008 Lachnospiraceae Cancer ST041 Peptostreptococcaceae Cancer RS007 Other Barrett’s oesophagus CAM446 Barrett’s oesophagus NEW047 Control NOT039 Control CAM464 Cancer SH071 Cancer AH155 Barrett’s oesophagus CAM249 Barrett’s oesophagus CAM248 Barrett’s oesophagus CAM427 Barrett’s oesophagus CAM459 Barrett’s oesophagus NOT046 Control CAM197 Control CAM161 Cancer ST037 Barrett’s oesophagus UCL055 Control NOT025 Barrett’s oesophagus UCL058 Control NOT032 Control CAM202 Cancer RS006 Control CAM208 Control CAM460 Barrett’s oesophagus UCL038 Barrett’s oesophagus CAM458 Barrett’s oesophagus CAM449 Control CAM153 Control NOT031 Control NOT054 Control CAM216 Control CAM451 Control CAM424 Figure 2: Microbial community composition in Barrett’s oesophagus The oesophageal adenocarcinoma and control patient groups largely cluster away from each other in this Bray-Curtis cluster dendrogram (p=0·002, parsimony test), but there is no significant difference in clustering for Barrett’s oesophagus. Microbial composition is shown at the family level for each tissue sample. Data were sub-sampled at 631 reads per sample. abundance 0·8%; figure 1A). Significant genera within (median 0%, SE 0·007%) in Barrett’s oesophagus samples these families included Veillonella (p=0·002, overall and 0·004% (median 0%, SE 0·003%) in control samples proportional abundance 3·8%), Megasphaera (p=0·0027, (p=0·028). Sequence identity was confirmed where overall proportional abundance 0·3%), Granulicatella possible with NCBI BLAST. One Barrett’s oesophagus (p=0·037, overall proportional abundance 1·6%), sample contained a high proportional abundance of Actinomyces (p=0·0022, overall proportional abundance H pylori sequences (>99%). 0·8%), Solobacterium (p=0·012, overall proportional Typically, oesophageal adenocarcinoma samples abundance 0·3%), and Campylobacter (p=0·0004, overall clustered away from controls in a Bray-Curtis cluster pro portional abundance 0·2%; figure 1B). In oesophageal dendrogram (p=0·002, parsimony test), emphasising adenocarcinoma samples, the Gram-positive, anaerobic the difference in community structure (figure 2). The Coriobacteriaceae was enriched at the family level (p=0·01, Bray-Curtis algorithm describes the dissimilarity overall proportional abundance 1·9%), but there were no between communities by taking into account both the significant genera identified within this family. At the overlap in OTUs that are present and the proportional species level, the Gram-positive, facultative anaerobic abundance of those OTUs. Samples that have fewer Lactobacillus fermentum was enriched in oesophageal overlapping OTUs and OTUs with less similar adenocarcinoma with mean proportional abundance of proportional abundances will cluster separately, and this 0·6% (median 0·009%, SE 0·5%) compared with 0·01% differential clustering was further shown by principal www.thelancet.com/gastrohep Vol 2 January 2017 37 Articles coordinate analysis (p=0·001, AMOVA test; appendix p 1). Alpha diversity refers to the species diversity within a The microbial communities of seven of 15 oesophageal given environment and includes the number of species adenocarcinoma samples were dominated by the (richness) and the proportion of those species (evenness) Gram-positive order Lactobacillales. Of the seven patients within the microbial community. Three indices of alpha with a high proportion of acid-tolerant Lactobacillales, diversity, observed OTU richness (p=0·0012; figure 3A), six were taking antacid drugs. Five oesophageal the Chao estimate of total OTU richness (p=0·0004; adenocarcinoma samples had a high proportional figure 3B), and the Shannon diversity index (p=0·0075; abundance of Streptococcus spp (69–98%) and two figure 3C) showed that diversity was lower in oesophageal samples had a high proportional abundance of adenocarcinoma samples than in controls. In comparison Lactobacillus spp (87% and 92%). NCBI BLAST showed with Barrett’s oesophagus samples, the oesophageal that the representative species were Streptococcus adenocarcinoma samples showed a decrease in observed pneumoniae/mitis, Streptococcus salivarius/vestibularis, OTU richness and the Chao estimate, but not in the Streptococcus parasanguinis, Lactobacillus gasseri, and Shannon diversity index. 13 patients with oesophageal Lactobacillus helveticus/suntoryeus/gallinarum (/ indicates adenocarcinoma had matched normal squamous tissue where it was not possible to differentiate between sampled proximal to the tumour, and in these patients different species using the 16S rRNA gene regions there was no difference in OTU richness between the sequenced). Although abundant in the acidic stomach normal and tumour tissue (p=0·9065; figure 3D). environment, such a high proportional abundance of Similarly, there was no difference for the Chao estimate Lactobacillus spp was an unexpected finding in the (p>0·999) or the Shannon index (p=0·6355). Furthermore, oesophagus. When we examined matched healthy and there was no difference in overall bacterial abundance tumour tissue for the two patients with high Lactobacillus between matched normal squamous and tumour tissue spp, we found that this genus dominated the lower (p=0·782; figure 3E). These results suggested that the oesophagus regardless of disease state (appendix p 2). decreased microbial diversity was pervasive throughout Gram-positive rods were visualised in areas of ulceration the lower oesophagus in oesophageal adenocarcinoma in tumour PS003 with a high proportional abundance of and was independent of the absolute quantity of Lactobacillus spp (appendix p 2). oesophageal bacteria. A p=0·0012 B p=0·0004 C 150 ‡ 200 † 5 p=0·0075 * 4 0 0 Control Barrett’s Cancer Control Barrett’s Cancer Control Barrett’s Cancer oesophagus oesophagus oesophagus D E p=0·9065 p=0·7820 80 30 0 15 Normal squamous Tumour Normal squamous Tumour Sample diagnosis Sample diagnosis Figure 3: Microbial alpha diversity in oesophageal adenocarcinoma (A) Observed richness of bacterial operational taxonomic units (OTUs). (B) The Chao estimate of total OTU richness and (C) the Shannon diversity index are shown for tissue samples from healthy control patients (n=16), patients with Barrett’s oesophagus (n=17), and patients with oesophageal adenocarcinoma (n=15). Statistical significance was calculated with the Kruskal-Wallis test and Dunns multiple comparisons post test. Data were subsampled at 631 reads per sample. (D) Observed richness of bacterial OTUs for paired normal squamous and tumour tissue samples from 13 patients (26 samples), Wilcoxon signed rank test. Data were subsampled at 656 reads per sample. (E) Overall bacterial abundance using 16S rRNA gene qPCR in matched tumour and normal squamous tissue from 16 patients (32 samples), Wilcoxon signed rank test. Error bars represent standard deviation. *p<0·05. †p<0·01. ‡p<0·001. 38 www.thelancet.com/gastrohep Vol 2 January 2017 Number of OTUs Number of OTUs Cycle threshold (Ct) Chao estimate Shannon index Articles We did additional analyses to ensure that the phylum level, the Cytosponge samples contained a higher proportional abundance of Tenericutes in comparison differences noted were not due to sex, age, or acid with the other sample types using LEfSe (p = 4·7 × 10 suppression. We repeated the diversity analysis with ⁵, male patients only (excluding ten control patients, overall proportional abundance 0·2%). At the genus level, four patients with Barrett’s oesophagus, and three the Cytosponge samples contained greater proportional patients with cancer) and the results were consistent, abundances of Fusobacterium (p<0·0001, overall pro- with decreases in diversity in cancer samples compared portional abundance 2%), Megasphaera (p<0·0001, overall with controls for the observed OTU richness (p=0·0029), proportional abundance 1·8%), Campylobacter (p<0·0001, the Chao estimate of total OTU richness (p=0·0017), and overall proportional abundance 1·7%), Capnocytophaga the Shannon diversity index (p=0·0070). We repeated the (p=0·00058, overall proportional abundance 0·7%), and diversity analysis with patients aged 60 years and older Dialister (p<0·0001, overall proportional abundance 0·2%). (excluding nine control patients, two patients with In keeping with these findings, principal coordinate Barrett’s oesophagus, and two patients with cancer) and analysis with the Bray-Curtis algorithm showed that the the results showed a similar trend, which was significant for the observed OTU richness (p=0·0448), the Chao Biopsy Brush estimate of total OTU richness (p=0·0288), but not the Cytosponge Shannon diversity index (p=0·0892). We also did a 0·4 Throat swab subgroup analysis for age within each diagnostic subgroup (using median age as a cutoff within each 0·2 group) and there were no significant differences in diversity for younger versus older patients within any of 0·0 the subgroups. When we excluded patients who were not taking acid suppression (four control patients, –0·2 two patients with Barrett’s oesphagus, four patients with cancer) or unknown acid suppression status (one patient –0·4 with cancer) the results were similar, with decreased –0·4 –0·2 0·0 0·2 0·4 0·6 0·8 PC1=19·6% diversity in oesophageal adenocarcinoma samples compared with controls, as evidenced by the observed B C p<0·0001 p=0·0104 OTU richness (p=0·0065), the Chao estimate of total 35 150 * * OTU richness (p=0·0033), and the Shannon diversity index (p=0·0202). Given that the decrease in microbial diversity in oesophageal adenocarcinoma seemed widespread throughout the oesophagus, we questioned whether the Cytosponge could be a useful tool to sample the microbiota along the entire length of the oesophagus and upper gastrointestinal tract. 15 patients with Barrett’s 15 0 oesophagus and 16 control patients had Cytosponge D E samples that underwent 16S rRNA gene amplicon p=0·0156 p=0·5968 300 5 sequencing with matched endoscopic biopsies and brushes taken from an area of normal squamous oesophagus. 13 of these patients also had swabs of their posterior pharynx to analyse the similarities and differences between the oesophageal and oral microbiota. Overall, 1455 OTUs were identified and mapped to 100 381 genera. Using a cutoff of 0·0001% overall 1 proportional abundance (138 genera classified), 84·1% of 0 0 genera were shared between Cytosponge samples, Biopsy Brush Cytosponge Biopsy Brush Cytosponge biopsies, and brushes, and 83·6% were shared between Figure 4: Comparison of different methods to sample the oesophageal microbiota Cytosponge samples and throat swabs. A stricter cutoff (A) Principal coordinate analysis with the Bray-Curtis algorithm for matched endoscopic biopsies, brushes, and of 0·1% overall proportional abundance (41 genera Cytosponge samples (31 patients) and 13 throat swabs from a subset of these patients. The first axis (PC1) classified) showed supporting reads for 100% of genera accounts for 19·6% of the sample variance and the second axis (PC2) accounts for 6·3% of the variance. Data were in all sample types, suggesting an overlap in community subsampled at 631 reads per sample. (B) Overall bacterial abundance using 16S rRNA gene-based quantitative PCR in matched endoscopic biopsies, brushes, and Cytosponge samples (20 patients), Friedman test and Dunns membership between the oral cavity, oesophagus, and multiple comparisons post test. (C) The observed diversity of bacterial operational taxonomic units (OTUs), gastric cardia. (D) the Chao estimate of total OTU richness, and (E) the Shannon diversity index for matched endoscopic biopsies, Although most microbial taxa overlapped between brushes, and Cytosponge samples (31 patients), Friedman test and Dunns multiple comparisons post test. sample types, the proportional abundances differed. At the Data were subsampled at 631 reads per sample. *p<0·05. †p<0·01. ‡p<0·001. §p<0·0001. www.thelancet.com/gastrohep Vol 2 January 2017 39 Chao estimate Cycle threshold PC2=6·3% Shannon index Number of OTUs Articles A C p=0·0147 p=0·023 p=0·0085 * * 300 400 5 0 0 Control Barrett’s Dysplasia Control Barrett’s Dysplasia Control Barrett’s Dysplasia oesophagus oesophagus oesophagus Figure 5: Microbial alpha diversity in high-grade dysplasia detected with the Cytosponge (A) Observed richness of bacterial operational taxonomic units (OTUs), (B) the Chao estimate of total OTU richness, and (C) the Shannon diversity index for Cytosponge samples taken from normal squamous control patients (n=20), patients with Barrett’s oesophagus (n=24), and patients with high-grade dysplasia (n=23). Statistical significance was calculated with the Kruskal-Wallis test and Dunns multiple comparisons post-test. Data were sub-sampled at 19 303 reads per sample. Error bars represent standard deviation. *p<0·05. Cytosponge samples clustered away from the throat swabs, abundance 0·1%), and unclassified Prevotellaceae endoscopic biopsies, and brushes (p<0·001, AMOVA test; (p=0·047, overall proportional abundance 1·3%). figure 4A). The throat swabs clustered distinctly from all the other sample types as well (p<0·001). There was no Discussion difference in clustering between biopsies and brushes on Our sequencing data showed decreased microbial the principal coordinate analysis plot (p=0·459). diversity and altered community composition in As expected, because of increased sampling surface oesophageal adenocarcinoma. Interestingly, patients with area, quantitative PCR of overall bacterial abundance oesophageal adenocarcinoma appeared to have this showed the quantity of microbial DNA isolated from reduced diversity regardless of whether cancerous or Cytosponge samples was greater than that from matched normal oesophageal tissue was sampled. The genera that biopsies and brushes (20 patients, p<0·0001; figure 4B). were decreased in proportional abundance in oesophageal After subsampling to normalise for sequencing depth, adenocarcinoma included Gram-negative (Veillonella, there was a decrease in observed OTU richness Megasphaera, and Campylobacter) and Gram-positive (p=0·0104; figure 4C) and the Chao estimate of total OTU taxa (Granulicatella, Atopobium, Actinomyces, and richness (p=0·0156; figure 4D) in endoscopic brush Solobacterium). There was significantly increased samples, but no difference for the Shannon index proportional abundance for L fermentum in patients with (p=0·5968; figure 4E). oesophageal adenocarcinoma compared with control To translate our findings to the setting of early detection, patients and those with Barrett’s oesophagus, and there we tested the usefulness of the Cytosponge to detect was a high proportional abundance of acid-tolerant changes in microbial diversity in patients with high-grade Lactobacillales (Lactobacillus spp and Streptococcus spp) dysplasia (n=23). The mean number of reads for in a subset (seven [47%] of 15) of oesophageal Cytosponge samples was 40 753 (SD 9717) and the median adenocarcinoma samples. To translate our findings to was 40 821 (11714), and the proportion of reads that were the setting of early detection, we investigated the use of subsampled was 47% (for 19 303 reads cutoff). The the Cytosponge device for sampling the oesophageal observed OTU richness was decreased in high-grade microbiota in Barrett’s oesophagus and high-grade dysplasia compared with controls (p=0·0147; figure 5A), dysplasia. The Cytosponge had high microbial DNA yield as were the Chao estimate of total OTU richness (p=0·023; and detected significantly decreased diversity in patients figure 5B) and the Shannon index (p=0·0085; figure 5C). with high-grade dysplasia compared with control patients. There was decreased diversity in Barrett’s oesophagus, but Lactobacillales, which are lactic acid bacteria, are so this was significant only for the Shannon index. In named for their ability to produce lactate from the general, the Cytosponge samples showed homogeneous fermentation of carbohydrates and to survive under harsh results for microbiota composition between diagnostic acidic conditions. Their resilience to low pH might groups at the phylum and family levels, suggesting that enable Lactobacillus spp and Streptococcus spp to thrive in the fraction of microbiota sampled from the area of the tumour niche in a subset of patients with oesophageal Barrett’s oesophagus was diluted by the copious bacteria adenocarcinoma, and production of lactic acid by these sampled from the rest of the oesophagus, oral cavity, and bacteria could further acidify the microenvironment. stomach. Despite this, three genera were identified that Lactic acid fermentation can also produce noxious distinguished controls from the other sample types using by-products, such as hydrogen peroxide, that directly LEfSe: Dialister (p=0·027, overall proportional abundance inhibit the growth of competitor bacteria and enable 0·3%), Schlegelella (p=0·016, overall proportional Lactobacillales to dominate the lower oesophagus. Given 40 www.thelancet.com/gastrohep Vol 2 January 2017 Number of OTUs Chao estimate Shannon index Articles the altered microbial composition in oesophageal distinct clustering in principal coordinate analysis. adenocarcinoma samples, it would be interesting to Despite dilution from sampling the upper gastrointestinal correlate microbiota data with expression and activity of tract, it was still possible for the Cytosponge to detect Toll-like receptors (TLRs), particularly TLR2, given the a decrease in diversity and community composition increased proportional abundance of Gram-positive between normal squamous controls and high-grade genera in a subset of cancer samples. Other authors have dysplasia. Similar to the Cytosponge, Fillon and investigated TLR expression in Barrett’s carcinogenesis colleagues described a minimally invasive oesophageal and found overexpression of TLRs 1, 2, 4, 6, and 9 in string test to sample the microbiota in a paediatric 30–33 human oesophageal adenocarcinoma samples and population. The oesophageal string test detected a similar TLRs 1–3, 6, 7, and 9 in a rat reflux model. microbial composition to that in matched oesophageal Although the microbial community structure differed biopsies, but required the patients to remain in hospital significantly in oesophageal adenocarcinoma in our overnight with the string secured to their cheek. study, there was only a modest reduction in diversity in Alternatively, the Cytosponge is a convenient test that can Barrett’s oesophagus and no genera were identified be given in a general practitioner’s or family doctor’s that discriminated between controls and Barrett’s office with the supervision of a trained nurse, and takes 14,39 oesophagus, or between Barrett’s oesophagus and only 5–7 min to complete. The Cytosponge can also oesophageal adenocarcinoma. It is possible that very low provide histological data for inflammatory pathologies abundance genera might be difficult to detect in such as candidal oesophagitis, herpes oesophagitis, and oesophageal biopsies given the low microbial DNA yield, eosinophilic oesophagitis. Our initial results using the and notably some pathogens have been shown to cause Cytosponge are promising, and future test development overt disease while only accounting for a low proportional should focus on longitudinal sampling of the microbiota abundance of the total microbiota, such as Clostridium to monitor changes in microbial diversity over time in a 35 36 37 difficile, Citrobacter rodentium, and Fusobacterium spp. larger cohort of patients. Further research should also Similarly, Amir and colleagues were unable to identify examine the role of diet, dysphagia, and other external any taxa that differentiated between controls (n=15) and influences on the oesophageal microbiota. Barrett’s oesophagus (n=6), or oesophagitis (n=13) using Contributors DRFE designed and performed the experiments, analysed the data, and LEfSe. By contrast, Yang and colleagues reported that wrote the manuscript. AWW designed experiments, analysed 16S rRNA Gram-negative bacteria were significantly enriched in gene data, and critically revised the manuscript. MOD assessed histology Barrett’s oesophagus (n=10) and reflux oesophagitis and immunohistochemistry. JP managed the Wellcome Trust Sanger (n=12) compared with controls (n=12). The main Institute’s Pathogen Genomics Group’s 16S rRNA gene amplicon sequencing activities and provided critical resources and support. limiting factor of these microbiota studies is the relatively RCF supervised the project, provided funding, designed the study, and small sample size and substantial inter-individual critically revised the manuscript. variation in microbiota composition. Another limitation Declaration of interests is that although LEfSe is useful for biological We declare no competing interests. interpretation of metagenomic data, it does not correct Acknowledgments for multiple comparisons, so there is a risk of false This study is supported by Cancer Research UK as part of the International discovery (p value, α=0·05). The inclusion of appropriate Cancer Genome Consortium (RG66287), with infrastructure support from negative controls and replicate samples is also paramount the Cambridge Biomedical Research Centre (Cancer Research UK) and the Cambridge Experimental Medicine Centre (National Institute for Health for low microbial biomass samples to facilitate removal Research [NIHR]), a Programme Grant to RCF from the Medical Research of contaminant OTUs that might also lead to false Council (RG72746), an NIHR Professorship to RCF (RG67258), and a discovery, and this was a major strength of our study. United European Gastroenterology prize to DRFE. Funding for AWW, JP, We also imposed strict quality control criteria, resulting and 16S rRNA gene amplicon sequencing was provided by the Wellcome Trust (098051). AWW receives core funding support from the Scottish in the exclusion of 14 tissue samples with low sequencing Government Rural and Environmental Science and Analysis Service read numbers and Good’s coverage estimates. The (RESAS). We thank the Wellcome Trust Sanger Institute’s core sequencing difficulty in obtaining good quality sequencing data from teams for generating 16S rRNA gene data and the Human Research Tissue oesophageal samples highlights the potential use of the Bank for the sample collection, which is supported by the NIHR Cambridge Biomedical Research Centre. The following UK hospitals Cytosponge device, which samples a larger surface area. participated in sample collection through the OCCAMS collaboration Our results suggest that it is feasible to sample network: Cambridge University Hospitals NHS Foundation Trust, oesophageal microbiota using the Cytosponge, and the Cambridge; Edinburgh Royal Infirmary, Edinburgh; Royal Surrey County device detected the majority of genera present in Hospital National Health Service (NHS) Foundation Trust, Guildford; University of Southampton and University Hospital Southampton NHS endoscopic biopsies and brushes. The high microbial Foundation Trust, Southampton; St Thomas’ Hospital, London; and Wigan DNA yield collected by the Cytosponge reflects sampling and Leigh NHS Foundation Trust, Manchester. The following UK hospitals of the entire length of the oesophagus as well as the contributed samples as part of the BEST2 study: Cambridge University proximal stomach and oral cavity as it is withdrawn. Hospitals NHS Foundation Trust, Cambridge; Nottingham University Hospitals NHS Trust, Nottingham; University College Hospital London, The throat swabs showed similarities in community London; Newcastle Upon Tyne Hospital, Newcastle; and Hinchingbrooke membership between the oral cavity and oesophagus, Hospital, Huntingdon. but the proportional abundances differed, as shown by www.thelancet.com/gastrohep Vol 2 January 2017 41 Articles References 19 Kozich JJ, Westcott SL, Baxter NT, Highlander SK, Schloss PD. 1 Pohl H, Welch HG. The role of overdiagnosis and reclassification in Development of a dual-index sequencing strategy and curation the marked increase of esophageal adenocarcinoma incidence. pipeline for analyzing amplicon sequence data on the MiSeq Illumina J Natl Cancer Inst 2005; 97: 142–46. sequencing platform. Appl Environ Microbiol 2013; 79: 5112–20. 2 Picardo SL, Maher SG, O’Sullivan JN, Reynolds JV. Barrett’s to 20 Salter SJ, Cox MJ, Turek EM, et al. Reagent and laboratory oesophageal cancer sequence: a model of Inflammatory-driven upper contamination can critically impact sequence-based microbiome gastrointestinal cancer. Dig Surg 2012; 29: 251–60. analyses. BMC Biol 2014; 12: 87. 3 Singh S, Sharma AN, Murad MH, et al. Central adiposity is 21 Paulson JN, Pop M, Bravo HC. Metastats: an improved statistical associated with increased risk of esophageal inflammation, method for analysis of metagenomic data. Genome Biol 2011; 12: 1–27. metaplasia, and adenocarcinoma: a systematic review and 22 Good IJ. The population frequencies of species and the estimation of meta-analysis. Clin Gastroenterol Hepatol 2013; 11: 1399–412. population parameters. Biometrika 1953; 40: 237–64. 4 Nie S, Chen T, Yang X, Huai P, Lu M. Association of Helicobacter pylori 23 Zaura E, Keijser BJ, Huse SM, Crielaard W. Defining the healthy infection with esophageal adenocarcinoma and squamous cell “core microbiome” of oral microbial communities. BMC Microbiol carcinoma: a meta-analysis. Dis Esophagus 2014; 27: 645–53. 2009; 9: 259. 5 Amir I, Konikoff FM, Oppenheim M, Gophna U, H alf EE. 24 Li K, Bihan M, Methe BA. Analyses of the stability and core taxonomic Gastric microbiota is altered in esophagitis and Barrett’s esophagus memberships of the human microbiome. PLoS One 2013; 8: e63139. and further modified by proton pump inhibitors. Environ Microbiol 25 Morgulis A, Coulouris G, Raytselis Y, Madden TL, Agarwala R, 2013; 16: 2905–14. Schäffer AA. Database indexing for production MegaBLAST 6 Wu S, Rhee K-J, Albesiano E, et al. A human colonic commensal searches. Bioinformatics 2008; 24: 1757–64. promotes colon tumorigenesis via activation of T helper type 17 T cell 26 Nadkarni MA, Martin FE, Jacques NA, Hunter N. Determination of responses. Nat Med 2009; 15: 1016–22. bacterial load by real-time PCR using a broad-range (universal) probe 7 Arthur JC, Perez-Chanona E, Mühlbauer M, et al. and primers set. Microbiology 2002; 148: 257–66. Intestinal inflammation targets cancer-inducing activity of the 27 Segata N, Izard J, Waldron L, et al. Metagenomic biomarker microbiota. Science 2012; 338: 120–23. discovery and explanation. Genome Biol 2011; 12: R60. 8 Hu B, Elinav E, Huber S, et al. Microbiota-induced activation of 28 Shaheen NJ, Richter JE. Barrett’s oesophagus. Lancet 2009; epithelial IL-6 signaling links inflammasome-driven inflammation 373: 850–61. with transmissible cancer. Proc Natl Acad Sci USA 2013; 110: 9862–67. 29 Kleerebezem M, Hols P, Bernard E, et al. The extracellular biology 9 Kostic AD, Chun E, Robertson L, et al. Fusobacterium nucleatum of the lactobacilli. FEMS Microbiol Rev 2010; 34: 199–230. potentiates intestinal tumorigenesis and modulates the 30 Huhta H, Helminen O, Kauppila JH, et al. Toll-like receptor 9 tumor-immune microenvironment. Cell Host Microbe 2013; expression in the natural history of Barrett mucosa. Virchows Arch 14: 207–15. 2015; 467: 9–18. 10 Rubinstein MR, Wang X, Liu W, Hao Y, Cai G, Han YW. 31 Huhta H, Helminen O, Lehenkari PP, Saarnio J, Karttunen TJ, Fusobacterium nucleatum promotes colorectal carcinogenesis by Kauppila JH. Toll-like receptors 1, 2, 4 and 6 in esophageal modulating E-cadherin/β-catenin signaling via its fadA adhesin. epithelium, Barrett’s esophagus, dysplasia and adenocarcinoma. Cell Host Microbe 2013; 14: 195–206. Oncotarget 2016; 7: 23658–67. 11 Yang L, Lu X, Nossa CW, Francois F, Peek RM, Pei Z. 32 Verbeek RE, Siersema PD, Ten Kate FJ, et al. Toll-like receptor 4 Inflammation and intestinal metaplasia of the distal esophagus are activation in Barrett’s esophagus results in a strong increase in associated with alterations in the microbiome. Gastroenterology 2009; COX-2 expression. J Gastroenterol 2013; 49: 1121–34. 137: 588–97. 33 Kauppila JH, Takala H, Selander KS, Lehenkari PP, Saarnio J, 12 Gall A, Fero J, McCoy C, et al. Bacterial composition of the human Karttunen TJ. Increased Toll-like receptor 9 expression indicates upper gastrointestinal tract microbiome is dynamic and associated adverse prognosis in oesophageal adenocarcinoma. Histopathology with genomic instability in a Barrett’s esophagus cohort. PLoS One 2011; 59: 643–49. 2015; 10: e0129055. 34 Zaidi AH, Kelly LA, Kreft RE, et al. Associations of microbiota and 13 Blackett KL, Siddhi SS, Cleary S, et al. Oesophageal bacterial biofilm toll-like receptor signaling pathway in esophageal adenocarcinoma. changes in gastro-oesophageal reflux disease, Barrett’s and BMC Cancer 2015; 16: 52. oesophageal carcinoma: association or causality? 35 Buffie CG, Jarchum I, Equinda M, et al. Profound alterations of Aliment Pharmacol Ther 2013; 37: 1084–92. intestinal microbiota following a single dose of clindamycin results 14 Kadri SR, Lao-Sirieix P, O’Donovan M, et al. Acceptability and in sustained susceptibility to Clostridium difficile-induced colitis. accuracy of a non-endoscopic screening test for Barrett’s oesophagus Infect Immun 2012; 80: 62–73. in primary care: cohort study. BMJ 2010; 341: c4372. 36 Clare S, John V, Walker AW, et al. Enhanced susceptibility to 15 Ross-Innes CS, Debiram-Beecham I, O’Donovan M, et al. Citrobacter rodentium infection in microRNA-155-deficient mice. Evaluation of a minimally invasive cell sampling device coupled Infect Immun 2013; 81: 723–32. with assessment of trefoil factor 3 expression for diagnosing 37 Kostic AD, Gevers D, Pedamallu CS, et al. Genomic analysis Barrett’s esophagus: a multi-center case-control study. PLoS Med identifies association of Fusobacterium with colorectal carcinoma. 2015; 12: e1001780. Genome Res 2012; 22: 292–98. 16 Benaglia T, Sharples LD, Fitzgerald RC, Lyratzopoulos G. 38 Fillon SA, Harris JK, Wagner BD, et al. Novel device to sample the Health benefits and cost effectiveness of endoscopic and esophageal microbiome—the esophageal string test. PLoS One nonendoscopic cytosponge screening for Barrett’s esophagus. 2012; 7: e42938. Gastroenterology 2013; 144: 62–73. 39 Vaughan TL, Fitzgerald RC. Precision prevention of oesophageal 17 Cangelosi GA, Meschke JS. Dead or alive: molecular assessment of adenocarcinoma. Nat Rev Gastroenterol Hepatol 2015; 12: 243–48. microbial viability. Appl Environ Microbiol 2014; 80: 5884–91. 40 Paterson AL, Lao-Sirieix P, O’Donovan M, et al. Range of 18 Schloss PD, Westcott SL, Ryabin T, et al. Introducing mothur: pathologies diagnosed using a minimally invasive capsule sponge open-source, platform-independent, community-supported software to evaluate patients with reflux symptoms. Histopathology 2016; for describing and comparing microbial communities. published online July 15. DOI:10.1111/his.13039. Appl Environ Microbiol 2009; 75: 7537–41. 42 www.thelancet.com/gastrohep Vol 2 January 2017

Journal

The Lancet Gastroenterology & Hepatology – Unpaywall

Published: Jan 1, 2017

There are no references for this article.