The murine lung microbiome in relation to the intestinal and vaginal bacterial communities, ARTYKUŁY NAUKOWE ...

[ Pobierz całość w formacie PDF ]
//-->Barfodet al. BMC Microbiology2013,13:303RESEARCH ARTICLEOpen AccessThe murine lung microbiome in relation to theintestinal and vaginal bacterial communitiesKenneth Klingenberg Barfod1,2*†, Michael Roggenbuck3†, Lars Hestbjerg Hansen3, Susanne Schjørring1,Søren Thor Larsen2, Søren Johannes Sørensen3and Karen Angeliki Krogfelt1AbstractBackground:This work provides the first description of the bacterial population of the lung microbiota in mice.The aim of this study was to examine the lung microbiome in mice, the most used animal model for inflammatorylung diseases such as COPD, cystic fibrosis and asthma.Bacterial communities from broncho-alveolar lavage fluids and lung tissue were compared to samples taken fromfecal matter (caecum) and vaginal lavage fluid from female BALB/cJ mice.Results:Using a customized 16S rRNA sequencing protocol amplifying the V3-V4 region our study shows that themice have a lung microbiome that cluster separately from mouse intestinal microbiome (caecum). The mouse lungmicrobiome is dominated byProteobacteria, Firmicutes, Actinobacteria, BacteroidetesandCyanobacteriaoverlappingthe vaginal microbiome. We also show that removal of host tissue or cells from lung fluid during the DNA extractionstep has an impact on the resulting bacterial community profile. Sample preparation needs to be considered whenchoosing an extraction method and interpreting data.Conclusions:We have consistently amplified bacterial DNA from mouse lungs that is distinct from the intestinalmicrobiome in these mice. The gut microbiome has been extensively studied for its links to development of disease.Here we suggest that also the lung microbiome could be important in relation to inflammatory lung diseases.Further research is needed to understand the contribution of the lung microbiome and the gut-lung axis to thedevelopment of lung diseases such as COPD and asthma.BackgroundStudies of the lung microbiome by culture independenttechniques and its impact on lung immunity is a rela-tively new field and may contribute to new advances inunderstanding respiratory diseases [1]. Healthy humanlungs have up until recently been considered to be sterileby culture-based techniques, but now new evidence haveidentified microbial communities both in healthy humansand in those with disease [2-4]. The human microbiomeproject [5] did not originally include the lungs, but re-cently the Lung HIV Microbiome Project has publishedthe first results in this field [6,7]. Investigations intolung microbiology and lung immunity in humans is lim-ited largely because of technical, ethical considerations* Correspondence:kkb@nrcwe.dk†Equal contributors1Statens Serum Institut, Artillerivej 5, 2300 Copenhagen S, Denmark2National Research Centre for the Working Environment, Lersø Parkallé 105,2100 Copenhagen O, DenmarkFull list of author information is available at the end of the articleand small samples sizes, whereas the use of animal modelscan provide novel information useful in investigations intothe importance of lung microbiome in the development oflung immunology. Effective utilization and developmentof animal models have recently been identified as one ofthe most important challenges in future lung microbiomeresearch by the NIH [8]. Whereas many studies havefocused on the gut microbiome and its impact on amongothers lung immunity and asthma, little work has beenperformed to examine the contribution of the lung micro-biome on the pathogenesis of pulmonary diseases. Espe-cially in inflammatory lung diseases such as asthma andCOPD, the local microbiome may play an important rolein the pathogenesis. The technical challenges related tothe novel culture-dependent techniques include consistentextraction of useful DNA, the development of PCR methodsand sampling methods for the less abundant bacterial loadof the lungs.© 2013 Barfod et al.; licensee BioMed Central Ltd. This is an open access article distributed under the terms of the Creativereproduction in any medium, provided the original work is properly cited.Barfodet al. BMC Microbiology2013,13:303Page 2 of 12We hypothesized that the problems with getting bac-terial DNA from lungs was due to the presence of hostDNA in the extractions. In this study, we have investi-gated the bacterial community from lungs of 20 miceusing rDNA amplicon 454 pyrosequencing. We alsoperformed a conventional cultivation study of 10 mousebronchoalveolar lavage (BAL) fluids on different agarplates. Sampling methods and DNA extraction proto-cols were investigated systematically: one BAL samplestill containing mouse cells (BAL-plus) and one BALsample, where the mouse cells were removed (BAL-minus) by cytospin. The bacterial communities in BALsamples were compared using DNA extractions fromwashed lung tissue, caecum samples and vaginal flushing.We chose to include vaginal samples for two majorreasons. The vaginal microbiome of BALB/c has notpreviously been described and could have influence onmicrobial“priming”and transfer from mother to pup.In this study, it also serves a reference sample from adifferent mucoid epithelium than lung. The bacteria wereclassified by their sequence into Operational TaxonomicUnits (OTU). An OTU is an approximation to taxonomyderived from classical cultivation techniques.We demonstrate the use of this methodology and de-scribe an uncultivable lung and vaginal microbiome inmice that are diverse and distinct from caecal micro-biome. Our results provide a basis for further studiesinto the lung microbiome in culture negative BAL fluidsin mouse models of inflammatory lung diseases sug-gested by descriptive human studies.tip of the left lung after the BAL procedure. Tissueswere snap-frozen in liquid nitrogen.Vaginal fluid samples were performed by inserting asterile pipette tip into the vaginal space flushing 3 timesback and forth with 30μLpyrogenfree infusion saline(0.9%) (Fresenius Kabi, Denmark) and frozen at−80°C.As the last procedure, the caecum samples were takenfrom the animals. With a sterile scissor the caecum wascut open and approximately 50 mg of caecal matter wasremoved using sterile plastic loops directly into cryo tubesand snap frozen in liquid nitrogen. All protocols wereapproved by the Danish Animal Experiments Inspectorate.Bacterial identification by culturingMouse BAL fluids, 200μLper mouse, were cultivatedon general growth media blood agar 5% (SSI, Denmark)and Chocolate Agar (SSI, Denmark) for fastidious bac-teria and incubated at 37°C for 24 hours. Another set ofplates with selective media was incubated under microaerophilic conditions (5%CO2, 3%H2, 5%O2and 87%N2)at 37°C for 48 hours [11]. The bacterial colonies weresubjected to routine identification by the Vitek2 system(Bio Mérieux, France).DNA extraction and PCRMethodsMice and sample collectionBALB/cJ female mice, reared together (Taconic M&B,Ry, Denmark), 7 weeks old, body weight 18–22 g, wererandomly distributed and housed 10 animals per cage(425 × 266 × 150 mm) with tap water and food (Altrominno 1324 Brogaard Denmark) provided ad libitum. Light/dark cycles were at 12 hours and room temperatureand relative humidity was kept at 19-22°C and 40-60%,respectively. Animals were handled by the same twoanimal technicians and conditioned in our animal facilityfor two weeks before use.The BAL procedure was performed as previouslydescribed with minor modifications [9]. We insertedsterile tube (Insyte, BD, Denmark) for each mouse andlungs were flushed two times with 0.8 mL pyrogenfreesaline (0.9%)(Fresenius Kabi, Denmark) and the recov-ered fluids were pooled (LF-plus). For the BAL sampleswithout mouse cells (BAL-minus) the BAL fluid wasspun at 400 g for 10 min a 4°C collecting the super-natant. All the BAL samples were frozen at−80°C.Lung tissue was collected using one, chlorine [10] andheat treated sterile scissors, per animal cutting the distalIsolation of bacterial DNA from frozen BAL or vaginalsamples was done using Qiagen spin protocol (Qiagen,DNA mini kit Denmark) for body fluids with the follow-ing modifications: Tubes were thawed and centrifuged at16.000 g for 5 min to spin down all the bacteria. Thesupernatant was discarded and the bacterial pellet wasresuspended with 450μLlysis buffer. Forty-fiveμLpro-teinase K and add 0.3 mL 0.1 mm zirconium/silica beads(Techum, Sweden) were added. Proceed with bead beat-ing step using TissueLyser (Qiagen, Denmark) for 6 minat 30 Hz. [12]. Lysis was performed by incubating inheat block at 56°C for 10 min. and then at 95°C for7 min. Proceed with protocol for body fluids from step 5.At the elution step, the AE buffer is preheated to 65°Cand DNA elution is performed with 100 ul with 3 minutesincubation at room temp before final spin. Isolation ofbacterial DNA from frozen caecal or tissue was doneusing Qiagen spin protocol for detection of pathogensfrom stool (Qiagen, DNA mini stool kit Denmark) withthe following modifications: Add 1.4 ml of the ASL bufferand perform bead beating, lysing and eluding as describeabove for body fluids. For tissues samples, chlorine [10]and heat sterilized 3 mm steel bead (Qiagen, Denmark)was added to the samples along with the zirconium/silica beads for extra tissue disruption.16S sequencingAmplicon libraries of the 16S rRNA gene of caecum, BALand vaginal samples were prepared with two PCR reactions.Barfodet al. BMC Microbiology2013,13:303Page 3 of 12In the first PCR, a 466 bp long fragment covering thevariable region V3 and V4 of the 16S rRNA gene, wasamplified with AccuPrime™ Pfx DNA Polymerase andthe bacteria and archaea specific primers 341 F and 806R(Table 1). The reaction started with an initialization at94°C for 2 min, followed by 44 cycles of denaturation at94°C for 20 sec, annealing at 56°C for 30 sec. and elong-ation at 68°C for 40 sec. The reaction was completedwith a final elongation at 68°C for 5 min. Due to the lowDNA (<0.5 ng ×μL-1) concentration in the samples weneeded to increase the cycle number above the standardof 30–35. This adjustment highly increased the risk ofamplifying contamination from extraction buffer andother experimental used liquids. To minimize this possi-bility we chose the lowest cycle number with a clearamplification band in the agarose gel and no signalsof negative controls from BAL procedure for DNApurification.In the second PCR the adaptors were attached to theamplicon library elongating the fragment towards 526 bpwith the primer TitA_341F and TitB_806R. The samereaction conditions of PCR I were applied in PCR IIwith a reduced cycle number of 15.Initially we tried to apply the same procedure forthe lung tissue samples but unspecific bands after gel-electrophoresis made it impossible to select the correctfragment size. To overcome this problem we chose theprimer 27 F and 1492R amplifying the entire 16S rRNAgene which appeared to be more specific. The PCR I con-ditions were the same as mentioned above except that theannealing temperature was reduced to 55°C and the cyclenumber to 40. In this perspective the Tag-PCR reactionwith TitA_341F and TitB_806R provided the selection forV3 and V4 as well as attaching the adaptors to theamplicons.Statistical analysis and bioinformaticsonly sequences with a minimum length of 250 bp. Chi-meras were removed by UCHIME [18]. The operationaltaxonomic units (OTU) were pickedde novoand clus-tered at 97% sequence similarity. The taxonomy wasassigned using RDP classifier (bootstrap threshold 0.8)greengenes as reference database [19].For statistical analysis, raw data were transferred intothe open source statistical program“R”[20]. The non-parametric Wilcoxon test (W) evaluated variations ofalpha diversity between two variables. We used thenon-parametric Kruskal-Wallis-test when comparing morethan 2 variables (KW). Dissimilarities in OTUs abundancebetween the samples were explained by KW and thesample clustering of the OTU count based Bray-Curtisdistance metric were examined by the analysis of simi-larity (anosim).ResultsTo determine the airway bacterial microbiota of theBALB/cJ mouse model based on 16S rDNA gene se-quencing, we have compared sequences found in thelungs with three different approaches, to sequences foundin corresponding vaginal and caecal samples.Over all sequence quality and results from all sample typesThe 16S rRNA gene sequences obtained from one half aplate of a 454 - Roche - Titanium pyrosequencing runwere quality filtered, trimmed and split into the corre-sponding animal samples with the Qiime pipeline ver-sion 1.6.0 using the default settings [17]. We consideredWe generated a total of 908256 sequences. After qualityfiltering and chimera check, 27% of sequences were re-moved and 660319 sequences were further processed forOTU picking (sequences ranged between 3530 up to 31638per animal sample). Thede novoOTU clustering revealed6487 OTUs. The OTU table was randomly subsampled toavoid differences based on sequencing effort leaving 3318OTUs for further analysis (Rarefaction curve are shown inAdditional file 1: Figure S5).We found a total of 19 bacterial phyla in the samplesanalysed. The most dominant (>0.5% abundance) phylaobserved wereAcidobacteria, Actinobacteria, Bacteroidetes,Firmicutes, ProteobacteriaandTM7.The difference in bac-terial composition at the phylum level between samplingsites is shown in Figure 1A.Table 1 PrimersPrimer27 F341806TitA_341FTitB_806R1492RSequence5′AGAGTTTGATCMTGGCTCAG-3′5′-CCTAYGGGRBGCASCAG-3′5′-GGACTACNNGGGTATCTAAT-3′5′-CGTATCGCCTCCCTCGCGCCATCAG-TAG-CCTAYGGGRBGCASCAG-3′5′-CTATGCGCCTTGCCAGCCCGCTCAG-GGACTACNNGGGTATCTAAT-3′5′-GGTTACCTTGTTACGACTT-3′Reference[13][14,15][14,15][16][16][13]Barfodet al. BMC Microbiology2013,13:303Page 4 of 12Figure 1Community composition. (A)Distribution of Phyla between sample types. LF-plus bronchoalveolar lavage (BAL) fluids and LF-minus isBAL where the mouse cells have been removed. LT is lung tissue and VF is vaginal flushing,(B)Venn diagram of identified shared and unique generafrom each sampling site. All the lung type samples are considered here as one. (complete list shown in Additional file 3: Table S4),(C)The PcoA plot isgenerated of the Bray-Curtis dissimilarity metric based on OTU counts and explains the largest variance between all samples (PCoA plot 1vs 3 andPCoA plot 2 vs. 3 are attached in Additional file 4: Figure S4),(D)Heat map of even subsampled OTU table. The dendrogram is two sited hierarchalclustered by abundance dissimilarity and the data are log transformed. Shown are only taxa, which counted for at least 0.5% of the generatedsequences. The x-axis clusters the animal samples and the y-axis the taxonomical information. * marks Vaginal subcluster S1 and ** subcluster S2.In Additional file 2: Table S2 we have listed all the bac-teria that were found, which were unique for the lungsamples and which were shared between sampling sites.The bacterial sequences of the lung samplesIf we only look at the lung samples, the most dominantlung phyla found wereProteobacteria, Firmicutes, Acti-nobacteria, BacteroidetesandCyanobacteria.Additionallywe observedFusobacteriaandCyanobacteriain the lungand vaginal samples.In order to highlight phyla variations in the lung com-munity compared to vaginal and caecal communities, wefirst we took the three lung sample types: bronchoalveo-lar lavage fluids (BAL-plus), and BAL-minus, where themouse cells have been removed by a spin protocol andfinally lung tissue from the distal tip of the lung andconsidered them as one ecological community. In thislung community profile,Actinobacteria,andProteobacteriawere clearly more abundant than in the caecum commu-nity (KW, p < 0.0001).Then, looking at the differences between the threelung sample types,Firmicutesappeared (KW, p < 0.05)more abundant in lung tissue (57%) than in BAL samples(20%). TheSR1bacteria were found only in BAL-minusand Lung tissue samples, butTenericuteswas observedin all samples, except in the vaginal samples. Otherphyla observed below 0.5% abundance wereChloroflexi,Deinococcus-Thermus, Fibrobacteres, Gemmatimonadetes,OD1, OP10, Planctomycetes, Verrucomicrobia,andWS3.Comparing lung sampling methods we also found asignificant variation forActinobacteriaandCyanobacteria,which were largely abundant in both type of BAL commu-nities relative to the lung tissue samples (KW, p < 0.05).At phylum level, the composition of the lung tissueBarfodet al. BMC Microbiology2013,13:303Page 5 of 12samples appeared to be very similar to the vaginal sam-ples except for a larger abundance ofCyanobacteriainvaginal samples (KW, p < 0.05).Bacterial sequences of the caecumLooking at the caecum samples, they contained moreFirmicutesandBacteroidetesKW, p < 0.0001) than thelung samples andAcidobacteriaandCyanobacteriawereabsent. The phylumBacteroidetes(29%) appeared to bethe second most abundant after theFirmicutes(59%).The vaginal and the caecal communities only hadRumi-nococcusin common, a genus that was not observed inthe lung microbiota. Three genera were found in caecalsamples alone;Robinsoniella, ParasutterellaandRamli-bacter.The low numbers of genera detected in the caecalsamples is due to the depth of taxonomic informationobtained for these particular OTU sequences towardsthe consensus lineage of the database.Overlapping generaFor an overview comparison between the different sampletypes, we have merged the results found in the differentlung communities and displayed the overlapping genera-wit hcaecum and vagina in a venn diagram. This diagramreflects 255 identified genera (summarized in Additionalfile 3: Table S4), that covers 76% of the sequences fromBAL-plus, 68% from BAL-minus, 66% of vaginal and lungtissue community and 27% of sequences assigned to thecaecum community (Figure 1B).Lung samples, vaginal and caecum samples shared the12 core generaBacteroides, Barnesiella, Odoribacter,Alistipes, Mucispirillum, Lactobacillus, Streptococcus,Peptoniphilus, Roseburia, Anaerotruncus, Oscillibacter,Pseudomonas.We observedParabacteroides, Eubacterium,Marvinbryantia, Butyricicoccus, Papillibacter, Bosea, Anae-roplasma,lung and caecum. The pulmonic and vaginalcommunity shared 103 genera (Additional file 3: Table S4).AdditionallyAkkermansiawas also found in the lung butonly in one caecum sample in the raw data set.Variability in community composition between samplesobtained from the same sampling site (Beta_diversity)To make a sample to sample comparison and illustrate thevariation between our mice we have performed a principlecoordinate analysis (PCoA) based on the Bray-Curtisdissimilarity between OTU count metric PCoA plot(Figure 1C), which explains the largest variance betweenall samples (Additional PCoA 2 and 3 are found inAdditional file 4: Figure S4).The caecal samples cluster together at a significant dis-tance from lung and vaginal communities, confirmed bythe analysis of similarity, anosim (R = 0.673, p = 0.001)The dissimilarity between the three lung communitieswas found to be little due to strong cluster overlap (anosim,R = 0.09, p = 0.05) when comparing only the lung distances.We found large variation within the vaginal samplesresulting in a division into subcluster 1 (S1), containinganimal vaginal sample 8, 5 and 2, and subcluster 2 (S2),vaginal sample 3,4,6,9 and 10 (anosim, R = 0.72, p = 0.001).The separation is clearly shown in PCoA1 (Figure 1C) andPCoA3 (Additional file 4: Figure S4). Those samples thatgrouped into S1 were found to be less similar to caecumand lung communities, whereas samples grouping into S2appeared more closely related to the lung microbiota.A more detailed description of the taxa responsible fordistinguishing bacterial communities in the lung, caecumand vagina is demonstrated using a heatmap dendrogram(Figure 1D).We removed from the subsampled OTU table all ob-servations accounting for less than 0.5% of the generatedsequences to visualize the taxa with main impact on thecommunity profile. This method provides maximal taxo-nomic resolution of each individual animal sample anddirectly reflects the PCoA plots since both analyses arebased on OTU count dissimilarities.For the caecum samples, 27% could be assigned to ataxonomic genus as mentioned before and the sequencesbelonged toAlistipes(16%)Anaeroplasma(1.5%) and a22 genera listed in Additional file 3: Table S4. Weobserved a better taxonomic resolution on the familylevel, were 77% of the reads were successful assigned.The three major families in the caecum wereLachnospira-ceae(33.8%),Ruminococcaceae (15.3%)andPorphyromo-nadaceae(7.9%).Vaginal samples within S1 contained between 56-97%ofStreptococcus,while vaginal samples within S2 onlyhad 0.2–10% of the gram-positive bacterium, explainingwhy here appears to be such a distinction between the S1and S2 groups. In addition toStreptococcus,notablecontributions fromAcinetobacter(6.2%),Sphinogmonas(3.3%),Enterococcus(3.1%), andPolaromonas(1.8%) werealso observed in the vaginal community.All lung samples had representative sequences fromgenera includingStaphylococcus(8.3%)Massilia(2.6%),Corynebacterium(2.2%),Pseudomonas(2.53%),Strepto-coccus(2.3%) andSphingomonas(1.7%) without signifi-cant variation (KW, p > 0.05).Even though the beta diversity measure indicated thatthere were minimal differences between the lung com-munities sampled using different methods, six majorgenera varied significantly (KW, p < 0.05).Acinetobacter,Pelomonas,andSchlegellawere more abundant in theBAL-plus samples in comparison to the BAL-minus orthe lung tissue samples.Arcobacter,andPolaromonaswerehighly associated with BAL-minus, whereasBrochothrixwas only found in the lung tissue samples. [ Pobierz całość w formacie PDF ]

The murine lung microbiome in relation to the intestinal and vaginal bacterial communities, ARTYKUŁY NAUKOWE ...

Linki