Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Reduced diversity and altered composition of the gut microbiome in individuals with myalgic encephalomyelitis/chronic fatigue syndrome

Reduced diversity and altered composition of the gut microbiome in individuals with myalgic... Background: Gastrointestinal disturbances are among symptoms commonly reported by individuals diagnosed with myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS). However, whether ME/CFS is associated with an altered microbiome has remained uncertain. Here, we profiled gut microbial diversity by sequencing 16S ribosomal ribonucleic acid (rRNA) genes from stool as well as inflammatory markers from serum for cases (n =48) and controls (n = 39). We also examined a set of inflammatory markers in blood: C-reactive protein (CRP), intestinal fatty acid-binding protein (I-FABP), lipopolysaccharide (LPS), LPS-binding protein (LBP), and soluble CD14 (sCD14). Results: We observed elevated levels of some blood markers for microbial translocation in ME/CFS patients; levels of LPS, LBP, and sCD14 were elevated in ME/CFS subjects. Levels of LBP correlated with LPS and sCD14 and LPS levels correlated with sCD14. Through deep sequencing of bacterial rRNA markers, we identified differences between the gut microbiomes of healthy individuals and patients with ME/CFS. We observed that bacterial diversity was decreased in the ME/CFS specimens compared to controls, in particular, a reduction in the relative abundance and diversity of members belonging to the Firmicutes phylum. In the patient cohort, we find less diversity as well as increases in specific species often reported to be pro-inflammatory species and reduction in species frequently described as anti-inflammatory. Using a machine learning approach trained on the data obtained from 16S rRNA and inflammatory markers, individuals were classified correctly as ME/CFS with a cross-validation accuracy of 82.93 %. Conclusions: Our results indicate dysbiosis of the gut microbiota in this disease and further suggest an increased incidence of microbial translocation, which may play a role in inflammatory symptoms in ME/CFS. Keywords: Myalgic encephalomyelitis, Chronic fatigue syndrome, Inflammation, Lipopolysaccharides, Microbiome, Microbial translocation, Beta-diversity Background have been the basis of the widely used Fukuda diagnostic Myalgic encephalomyelitis (ME), also known as chronic criteria [1]. Many ME/CFS patients also report gastro- fatigue syndrome (CFS), or ME/CFS, is a debilitating intestinal (GI) symptoms, including but not limited to illness of unknown etiology with no widely accepted irritable bowel syndrome (IBS) [2–6]. Intestinal discomfort therapy. Primary symptoms reported by patients are fa- is also indicated in a survey of drug use by individuals with tigue, muscle and/or joint paint, sore throat, headaches, CFS compared to controls, which found significantly more unrefreshing sleep, and post-exertional malaise and use of antacids, H2 blockers, and proton pump inhibitors in the ME/CFS cohort [7]. The prevalence of bowel symptoms has led to attempts * Correspondence: [email protected] 1 to treat the disease by probiotic oral or rectal supplements. Department of Molecular Biology and Genetics, Cornell University, Ithaca, NY, USA Borody et al. [8] reported improvements in a majority of Full list of author information is available at the end of the article © 2016 The Author(s). Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Giloteaux et al. Microbiome (2016) 4:30 Page 2 of 12 patients at 4 weeks following bacteriotherapy comprised self-reported good health, 34 ME/CFS patients and 7 con- of rectal infusion of 13 enteric bacteria, though the trols self-reported gastrointestinal disturbances such as number with a sustained response was not well docu- constipation, diarrhea, or intestinal discomfort. Many ME/ mented. In two small studies, marginal improvement in CFS patients are able to identify an acute, often flu-like, certain symptoms was reported following oral probiotic illness that immediately preceded the onset of the dis- therapy [9, 10]. ease eventually diagnosed as ME/CFS, while others are Two reports suggest altered gut microbiota in ME/CFS unaware of an initiating event and consider their onset patients. Using culture-based methods, Sheedy et al. [11] to be gradual. Among the 48 ME/CFS patients in the described higher levels of D-lactic acid producing Entero- study, 19 indicated a gradual and 25 stated a sudden onset. coccus and Streptococcus spp. in ME/CFS patients vs. con- ME/CFS subjects completed the SF-36 form (Additional trols. More recently, Norwegian ME/CFS patients and file 1: Figure S1) and Bell’s Disability scale (Table 1). healthy controls were found to exhibit differences in gut In comparison to other studies in which patients diag- microbiota composition through a 16S rRNA gene se- nosed with ME/CFS also filled out the SF-36 form, our quencing study [12]. It is well documented that gut study population fell within the same ranges on the eight microbiota can be significant with respect to patho- subscales of the SF-36 (Additional file 1: Figure S1). logical intestinal conditions such as ulcerative colitis (UC), Crohn’s disease (CD) [13], and systemic diseases Measurements of levels of microbial translocation such as diabetes [14]. Because of the frequent occur- markers indicate microbial translocation rence of GI disturbances, as well as these prior reports We quantified plasma levels of hsCRP, lipopolysaccha- of abnormalities, we investigated the diversity and com- rides (LPS) as a marker of microbial translocation (MT) position of the gut microbiota of ME/CFS patients in and plasma intestinal fatty acid binding protein (I-FABP) comparison to healthy individuals. as a marker for gastrointestinal tract damage in both Along with GI symptoms, individuals with ME/CFS ap- groups. The distribution of plasma hsCRP, LPS and I-FABP pear to have both immune activation and immune dys- is showninFig.1.Levels of hsCRP were higher in theME/ function. Many of the common symptoms reported by CFSpopulationincomparisontohealthycontrols(1.38 ME/CFS patients are characteristic of inflammatory ill- and 1.21 mg/L, respectively), but the difference was not nesses [15]. Most reports concerning cytokine levels in statistically significant (P = 0.15, Fig. 1a, Table 2). ME/CFS patients vs. controls are somewhat limited in ME/CFS patients had significantly higher plasma LPS scope and discordant, but several recent papers with a levels than healthy individuals (median ME/CFS— 51-plex cytokine assay indicate abnormal immune sig- 119.43 pg/mL vs. controls—74.74 pg/mL, P < 0.0005, natures in plasma and cerebrospinal fluid [16, 17]. Fig. 1b and Table 2). The median plasma I-FABP level Abnormal immune activation can be caused by trans- was 341.9 pg/mL in the ME/CFS group and 301 pg/mL in location of microbes from an inflamed gut [18]. A prior the healthy group. Though the median I-FABP levels in report indicated increased IgA and IgM to lipopolysac- the ME/CFS group was higher than that of the healthy charide (LPS) in serum of CFS patients [19]. We therefore group, the difference was not statistically significant assayed plasma levels of LPS and LPS-binding protein, as (P = 0.27, Fig. 1c, Table 2). well as the LPS/LBP receptor sCD14 [20]. We also exam- To obtain further information concerning chronic LPS ined the levels of C-reactive protein, an inflammatory stimulation in vivo, we also measured plasma sCD14 levels marker, and I-FABP as a marker for gastrointestinal tract and plasma LBP, which is produced by gastrointestinal and integrity [21]. Table 1 Characteristics of the study population Objective molecular markers for diagnosis of ME/CFS are lacking. We examined the levels of plasma markers Controls (n = 39) ME/CFS (n = 49) and microbiota composition in the diseased vs. healthy Gender Female 30 38 subjects in order to determine whether the data, taken Male 9 11 together, could predict ME/CFS vs. healthy status. Age Mean ± SD 45.5 ± 9.9 50.2 ± 12.6 Median (range) 48 (20–61) 51 (19–71) Results BMI Mean ± SD 27.1 ± 6.1 25.5 ± 4.9 Study population characteristics Median (range) 26 (17–47) 24.5 (16–40) Subjects with ME/CFS were established patients of a ME/ CFS specialist, Susan Levine, M.D. and fit the Fukuda diag- Bell’s disability scale 10–20 NA 15 nostic criteria [1]. This study began before the criteria for 30–40 21 systemic exertion intolerance disease (SEID) were estab- 50–60 9 lished [22], but most, perhaps all, also fit the description of >60 4 SEID. Of the 48 patients and 39 control participants who Giloteaux et al. Microbiome (2016) 4:30 Page 3 of 12 Fig. 1 Microbial translocation, gastrointestinal tract damage, and evidence for direct LPS stimulation in vivo in ME/CFS: plasma levels of hsCRP (a), LPS (b), I-FABP (c), sCD14 (d), and LBP (e) determined in our cohorts of controls and ME/CFS diagnosed individuals. p values were calculated by the Wilcoxon-Mann-Whitney U test hepatic epithelial cells. Thus, increased LPS in the circula- Stool microbiota of ME/CFS patients exhibit reduced tion promote hepatic synthesis of LBP, a plasma protein diversity and different composition than healthy controls that increases the binding of LPS to CD14. sCD14 and The hypervariable V4 region of 16S rRNA genes was se- LBP concentrations in both groups are shown in Fig. 1. For quenced from fecal samples of individuals with ME/CFS the ME/CFS cohort the median plasma sCD14 concentra- (n = 48) and healthy individuals (n = 39). A total of tion was 1.97 ug/mL, and the median LBP plasma concen- 8,534,117 high-quality and classifiable reads were gener- tration was 17.68 ug/mL. These values were significantly ated from all samples, with an average of 98,093 ± 29,231 different from the plasma sCD14 and LBP concentra- reads per sample. Binning sequences using a pairwise tions of the healthy volunteers (1.36 ug/mL; P < 0.0005 identity threshold of 97 %, we obtained an average of and 12.32 ug/mL; P < 0.0005, respectively) (Fig. 1d, e, 1330 ± 423 operational taxonomic units (OTUs) per Table 2). sample. The sequence-based rarefaction curves based on Next, we analyzed the associations among biomarker the Phylogenetic Diversity (PD) metric were nearly asymp- measurements in the ME/CFS population. As can be totic and a Wilcoxon rank-sum test demonstrated a statis- seen in Fig. 2a, b plasma LPS levels correlated posi- tical difference in the diversity of ME/CFS and healthy tively with levels of sCD14 and LBP (r= 0.347, P< 0.01 individuals (P = 0.004, W = 1268) (Fig. 3a). and r = 0.487, P< 0.01, respectively), consistent with We examined the number of “observed species,” i.e., the stimulation of sCD14 production by LPS in vivo. In number of 97 % ID OTUs observed in 32,223 sequences, addition, we found a strong significant correlation be- the estimators of community evenness (Shannon H), and tween plasma sCD14 and hsCRP and sCD14 and LBP; richness (Chao1 and PD) in the two group of samples. high levels of sCD14 were associated with high levels of ME/CFS samples had a significant overall lower microbial hsCRP (r = 0.507, P <0.01) and LBP (r =0.578, P <0.01) diversity, with lower evenness (H = 5.33 ± 0.93 vs. 5.92 ± (Fig. 2c, d). We also analyzed whether enterocyte dam- 0.93, P = 0.004), and lower richness (observed species, age (i.e., I-FABP levels) was associated with the pro- 1204 ± 351 vs. 1486 ± 456; Chao1, 2363 ± 704 vs. 2918 ± posed microbial translocation markers LPS, sCD14, and 885, P = 0.002; PD, 61.6 ± 16.7 vs. 73.4 ± 19.04, P =0.004) LBP. We found no relationship between I-FABP and (Fig. 3b). LPS levels (r = −0.125; P =0.278), I-FABP and sCD14 To evaluate overall differences in beta-diversity between levels (r = −0.117; P = 0.310), or I-FABP and LBP levels the microbiomes, we applied Principal Component Ana- (r = −0.08; P =0.488). lysis (PCoA) to weighted and unweighted UniFrac distance Giloteaux et al. Microbiome (2016) 4:30 Page 4 of 12 Table 2 Plasma levels of markers of inflammation (hsCRP), healthy individuals, this corresponded to 46 and 45 % re- microbial translocation (LPS, sCD14, and LBP) and gastrointestinal spectively of the rarified 16S rRNA sequences. Also, Pro- damage (I-FABP) in ME/CFS and healthy individuals teobacteria made up the next largest represented phylum Analytes Control ME/CFS (3.6 %), with Verrucomicrobia and Actinobacteria in rela- hsCRP (mg/L) tively low relative abundance (2.1 and 1.6 %, respectively). At the phylum level, the abundance of the Bacteroidetes n 34 46 was comparable (52 %) in both datasets (Fig. 4a). ME/CFS Median 1.21 1.38 samples showed lower relative abundance of Firmicutes Quartiles 0.38–2.69 0.70–2.71 (35 %) (Fig. 4a) and higher relative abundance of Proteo- Range 0.27–5.09 0.23–19.3 bacteria (8 %), due almost entirely to a twofold increase in LPS (pg/mL) the Proteobacteria family Enterobacteriaceae (6 vs. 3 % for n 39 49 ME/CFS and healthy individuals, respectively) (Fig. 4b). Within the Firmicutes, at the family level, Ruminococca- Median 74.74 119.43 ceae were lower in the ME/CFS samples (16 vs. 11 % in Quartiles 54.34–99.54 66.21–144.41 ME/CFS and healthy individuals respectively) (Fig. 4b), Range 32.21–187.32 34.32–279.30 whereas Lachnospiraceae were similar among both data- I-FABP (pg/mL) sets (16 % for both healthy and ME/CFS samples). Some n 33 44 differences were noted between cases and controls in Median 234.40 255.85 family members of the Bacteroidetes, i.e., Bacteroidaceae (35 vs. 43 %), Rickenellaceae (3 vs. 4 %), and Prevotella- Quartiles 159.75–412.90 171.275–450.70 ceae (3.2 vs. 0.7 %). Finally, within the Actinobacteria, Range 14–1067.9 23.2–909.9 Bifidobacteriaceae were lower in the ME/CFS samples sCD14 (ug/mL) (1 vs. 0.5 %). n 39 49 At the OTU level, 40 OTUs were found to be signifi- Median 1.36 1.97 cantly different between groups after multiple testing cor- Quartiles 1.09–1.67 1.71–2.32 rection. The majority of them belonged to the Firmicutes phylum, including members of the Ruminococcaceae Range 0.62–2.42 0.95–3.02 family such as Oscillospira spp. (q = 0.016), Faecalibac- LBP (ug/mL) terium prausnitzii (q = 0.014), and Ruminococcus spp. n 37 49 (q = 0.014) and members of the Lachnospiraceae, i.e., Median 12.32 17.68 Coprococcus spp. (q = 0.014). Other OTUs included Quartiles 10.22–13.73 13.04–22.56 members of the Actinobacteria such as Eggerthella Range 8.65–18.76 7.06–34.52 lenta (q = 0.014) and Collinsella aerofaciens (q = 0.014). These significant differences were further confirmed by LEfSe analysis, which uses linear discriminant analysis (LDA) coupled with effect size measurements to identify metric matrices generated for the sample set. Within the bacterial taxa whose sequences are differentially abundant microbial community cluster, there appears to be no clear between ME/CFS and healthy individuals. In addition to difference in beta-diversity between the ME/CFS group detecting significant features, LEfSe also ranks features by and healthy group using both weighted (Additional file 2: effect size, which put features explaining most of the bio- Figure S2a) and unweighted (Additional file 2: Figure S2b) logical difference at top (Segata et al. 2011). LEfSe identified UniFrac distance matrices. None of the other parameters 24 discriminative features (genus level, LDA score >2) tested, i.e., sex, BMI, or clinical data revealed clustering whose relative abundance varied significantly among fecal (data not shown). Because beta-diversity clustering as mea- samples taken from the ME/CFS and healthy groups (Fig. 5). sured by UniFrac shows how dissimilar overall community ME/CFS microbiota were enriched with an unclassified structure is between samples, the samples may not cluster member of the Desulfohalobacteriaceae and genera from in a manner that reflects differences detected at the OTU the Firmicutes phylum, i.e., Oscillospira, Lactococcus, Anae- level, or the overall alpha diversity within groups. rotruncus and Coprobacillus and Eggerthella, a member of The overall microbial composition for ME/CFS and con- the Actinobacteria phylum (P < 0.05, Fig. 5). Eighteen gen- trols differed at the phylum and family levels (Fig. 4a, b), era were enriched in the control group compared to the although none of these differences were statistically sig- ME/CFS group (Fig. 5) with members mainly belonging to nificant after multiple test correction. The two largest the Firmicutes phylum. We observed that members of phyla represented in each dataset of healthy and ME/CFS- the Ruminococcaeae and Bifidobacteriaceae, i.e., Fae- afflicted individuals were Firmicutes and Bacteroidetes. In calibacterium and Bifidobacterium, respectively, were Giloteaux et al. Microbiome (2016) 4:30 Page 5 of 12 Fig. 2 Correlation between plasma levels of LPS and sCD14 (a), plasma levels of LPS and LBP (b), plasma levels of sCD14 and LBP (c), and plasma levels of hsCRP and sCD14 (d) in the ME/CFS population. Spearman’s rank test was used to determine correlations Fig. 3 Rarefaction curves and confusion matrix. a Rarefaction curves for the microbiota of healthy individuals and ME/CFS patients (each group was rarefied to the number of sequences of the less-sequenced sample, i.e., 32223 sequences). The p value was calculated by the Wilcoxon rank-sum test and b comparison of alpha diversity indexes in ME/CFS and healthy individuals Giloteaux et al. Microbiome (2016) 4:30 Page 6 of 12 Fig. 4 Composition of the gut microbiome of healthy individuals and ME/CFS patients. Relative abundance of phylum-level (a) and family-level (b) gut microbial taxa significantly increased in healthy individuals (P =0.03 with the highest proportion of samples correctly classified and 0.04, respectively). when genus-level taxa along with data from the inflamma- tory markers were used in the analysis. With 97 % ID Classifying subjects into patients vs. controls from OTUs used in the analysis, 82 % of the samples could be inflammatory markers and microbiome data correctly classified (standard deviation of 0.14). With Using a machine learning approach, samples were mostly OTUs collapsed at the species level, the average accuracy successfully classified into healthy and ME/CFS groups, was 0.80 with a standard deviation of 0.11. Collapsing Fig. 5 Histogram of the LDA scores computed for genera differentially abundant between ME/CFS and healthy individuals. ME/CFS-enriched genera are indicated with a positive LDA score, and genera enriched in healthy individuals have a negative score. The LDA score indicates the effect size and ranking of each differentially abundant taxon Giloteaux et al. Microbiome (2016) 4:30 Page 7 of 12 taxonomy to the genus level, individuals with ME/CFS were IgA and IgM to LPS of Gram-negative enterobacteria [19]. classified correctly and separately from the healthy group Our data supports the hypothesis of increased MT in the with an average success rate of 0.82 ± 0.12. The receiver ME/CFS group as evidenced by (i) significantly raised operating curves, the AUC ROC value for the ME/CFS levels of plasma LPS and (ii) significantly higher levels of samples (0.89), and the confusion matrix are presented sCD14 and LBP, as indicators of direct LPS stimulation. in Fig. 6. The feature importance scores for the genus- Increased gut permeability and increased LPS levels have level analysis, which shows the relative importance of been also described in patients with liver diseases, alcoholic, clinical values and microbial abundances, are available and nonalcoholic steatohepatitis [23], during chronic HIV in Additional file 3: Table S1. Additionally, processing infection [24], and in inflammatory bowel disease (IBD) microbial sequencing data without including BMI and [25, 26] suggesting that an activation of pro-inflammatory blood inflammatory marker levels results in 70, 75, and and endotoxin-signaling cascades could be important 72 % classification accuracy for genus, species, and OTU- for disease progression in ME/CFS. Consequently, high level data respectively (confusion matrices available in plasma LPS levels in ME/CFS could result from an in- Additional file 4: Figure S3). creased production of endotoxin upon changes in the gut microbiota. Furthermore, we observed that sCD14 Discussion levels positively correlated with levels of LPS, LBP, and Our analysis of the microbiome in cases suggests that hsCRP. If there is damage to the gut mucosa, microbial gastrointestinal tract of ME/CFS patients is a pro-inflam- translocation could increase, altering antimicrobial regula- matory environment. This environment might cause dam- tors and dysregulating the innate immune system. age to the intestinal epithelium, thus augmenting microbial As a marker, LPS is limited to particular microbes, as translocation (MT) and subsequently triggering an im- it is only present in Gram-negative bacteria. sCD14 is mune response. It has been previously documented that produced primarily by macrophages and hepatocytes in disruption of mucosal barrier function occurs in ME/CFS response to LPS but is also stimulated by other bacterial as demonstrated by the increased serum concentrations of and viral agents [27]. LBP functions as a co-factor along Fig 6 Receiver operating characteristic curves (a) for controls and ME/CFS patients determined using the inflammatory markers and sequencing datasets (even sampled at 32,233 sequences) and a supervised learning approach with randomForest algorithm and (b) confusion matrix for random forest analysis (values are presented as percentage) and ROC area under the curve (AUC) value for 97 % OTUs collapsed at the genus level. Mean AUC ROC value for five times repeated, 10-fold cross validation Giloteaux et al. Microbiome (2016) 4:30 Page 8 of 12 with sCD14 and is constitutively synthesized in hepato- these compounds as electron acceptors [43, 44] to gener- cytes to recognize LPS released to the bloodstream but ate energy and foster their own growth in the gut. We did various inflammatory factors such as IL22, IL-6, and not collect information concerning diet of patients and TNF-α can induce its expression [28, 29]. Nevertheless, thus do not know whether this factor might have affected we found significantly more patients with elevated levels the composition and/or metabolism of the colonic micro- of these biomarkers in comparison to the healthy group, biota in our cohorts. suggesting that more MT occurs in people affected by We observed significantly lower levels of the genus ME/CFS. Faecalibacterium, a member of the Ruminococcaceae in Using both aerobic and anaerobic culturing methods, the ME/CFS population. For example, Faecalibacterium Butt and colleagues were the first to present evidence of prausnitzii, which produces an anti-inflammatory pro- altered fecal microbiota in ME/CFS patients compared tein [45], is reduced in ME/CFS cases relative to con- to healthy individuals [30]. Subsequently, using culture trols. This genus is also depleted in IBD [13, 38] and methods and metabolite analysis, Sheedy et al. [11] ob- ulcerative colitis [46] and has been shown to have anti- tained information concerning the fecal microbiome in inflammatory properties both in vitro and in vivo [37]. patient and health cohorts. Both found that D-lactic Faecalibacterium belongs to a group of producers of acid-producing Enterococcus and Streptococcus species butyrate, a short chain fatty acid known to have anti- were strongly over-represented in ME/CFS patients and inflammatory properties and to protect the intestine that among anaerobic bacteria, Prevotella was a bacterial [40]. Individuals with IBD and IBS [47] exhibit a lack of genus found to be in excess in subjects with ME/CFS. butyrate-producing bacteria and lower levels of butyr- Recently, a study used high-throughput 16S rRNA ate in their gut [48, 49] which modulates different pro- gene sequencing to investigate the presence of specific cesses including hormone and cytokine secretion (e.g., alterations in the gut microbiota of ME/CFS patients leptin, IL-10) and activation of immune/inflammatory from Belgium and Norway [12]. The authors amplified responses [50–52]. the V5 and V6 hypervariable 16S rRNA regions and se- We also found a decrease in Bifidobacterium, previously quenced the amplicons using a Roche FLX 454 sequen- observed in IBS [53–57], IBD [58], and type II diabetes cer, which resulted in an average of only 6000–7000 [59]. Bifidobacteria are a group of lactic acid-producing reads/sample. In contrast, we amplified the V4 hypervar- bacteria that are widely used as probiotics and as tar- iable region of the 16S rRNA gene, sequenced amplicons gets for prebiosis [60]. Treatment with Bifidobacterium using the MiSeq Illumina platform, obtained an average infantis 35624 was reported to reduce CRP levels in a of many more reads/sample (98,000), and compared the cohort of ME/CFS patients [61]. resulting sequences to a different database, the Green- We have employed a supervised machine learning ap- genes non-redundant reference database [31]. Our ana- proach to help prediction of disease state based on the lysis showed that within-sample diversity is lower in the microbiome sequence datasets [62]. Using this approach, ME/CFS specimens compared to controls. The same indi- we were able to classify unlabeled samples with some de- ces in the Fremont et al. [12] study did not differ between gree of accuracy, as demonstrated by the high AUC ROC ME/CFS and healthy subjects [12], likely due to the lower value obtained (0.8928) at the genus level. This method read number they obtained. Lower richness has also been has been recently used in several microbiome surveys to observed in unhealthy or inflammatory states [32, 33] and accurately place individuals into an IBD/healthy category has been associated with IBD, necrotizing enterocolitis [63], including ulcerative colitis (ROC AUC = 0.9225) and [34], and greater abdominal discomfort levels in patients colonic (ROC AUC = 0.8787) or ileal Crohn’s disease with food intolerances [35, 36]. (ROC AUC = 0.9699) [64]. Such an approach could there- Regardless of disease state, bacteria belonging to the fore serve as a complement to other non-invasive diagno- Firmicutes, Bacteroidetes, Proteobacteria, and Actino- ses of symptoms or as an initial diagnosis to determine if bacteria phyla represented the vast majority of sequences the subject likely has ME/CFS. Because this is a relatively identified. We observed reduced levels of members of the small cohort, to move to a formal diagnostic clinical ap- dominant phylum Firmicutes, also noted repeatedly in plication, a large cohort of ME/CFS and healthy con- Crohn’s disease patients [13, 37, 38]. Proteobacteria were trols would be needed to verify that classification would more abundant in ME/CFS patients than in controls, retain its accuracy with independent sample handling observed as well in inflammatory bowel disease (IBD) pa- and sequencing. tients [39, 40]. In an inflamed gut, infiltrating macrophages and neutrophils release sulfur- and nitrogen-derived Conclusions metabolites such as tetrathionate and nitrate [41–44]. Taken together, our results suggest an ongoing damage to Opportunistic members of the Proteobacteria can take the gut mucosa, leading to increased microbial transloca- advantage of the host inflammatory response by using tion in ME/CFS, which in turn could alter antimicrobial Giloteaux et al. Microbiome (2016) 4:30 Page 9 of 12 regulators and disregulate the innate immune system. in plasma samples by commercially available enzyme- Differences between the gut microbiomes of healthy indi- linked immunosorbent assays (ELISA). Plasma sCD14 viduals and patients with ME/CFS were identified in terms was quantified using the Quantikine Human sCD14 of relative abundance of specific genera. There is no single Immunoassay(R&DSystems,Minneapolis,MN),and precise alteration of the gut microbiota in all ME/CFS pa- plasma LBP was measured by LBP soluble ELISA kit tients we examined, but our data converges to support the (Hycult Biotechnology, Uden, The Netherlands) according concept of a less diverse and unstable community of bac- to the manufacturers’ protocols. Plasma bacterial endo- teria in the disorder. It highlights the association of specific toxin, i.e., LPS, was measured from heparinized blood bacterial taxa with ME/CFS, and the identification of the samples (Brandtzaeg) using the Limulus Amebocyte Lysate underlying role of this altered commensal gut microbiota (LAL) assay (Lonza Group Ltd, Allendale, NJ). The method could lead to novel diagnostic and therapeutic strategies uses a chromogenic endpoint assay yielding data as endo- that would improve clinical outcome. Future studies may toxin units (EU/ml). Briefly, 100 μl of each plasma sample also reveal additional molecular markers that could be was diluted in 200 μlof β-G-Blocker (Lonza Group Ltd, combined with gut microbiome information to enhance Allendale, NJ) to eliminate the possibility of false positives. the sensitivity and specificity of ME/CFS diagnostic assays. Samples were further diluted with 100 μl of pyrogen-free The cause of ME/CFS is unknown, but gut dysbiosis water to give a final dilution of 1:4. All dilutions were pre- could be contributing to some of the symptoms and their pared in pyrogen-free tubes. Samples were then placed in a severity. Developing therapeutic interventions aimed at re- water bath at 85 °C for 15 min to inactivate inhibitory ducing local inflammation, restoring gastrointestinal tract plasma proteins. Results of LPS measured were expressed immunity and integrity and modifying the intestinal micro- in picograms per milliliter (1 EU/ml = 100 pg/ml). Levels biome may ameliorate ME/CFS symptoms in a number of of intestinal fatty acid binding protein (I-FABP), a marker affected patients. associated with enterocyte damage, were assayed using an ELISA (Hycult Biotechnology, Uden, the Netherlands) ac- Methods cording to the manufacturer’s instructions. All the samples Human subjects and sample/data collection were run in duplicate. All work involving human subjects was approved by the Cornell University Institutional Review Board. Fecal sam- Statistical analysis of plasma markers levels ples were collected at home by participants in 15-ml con- We initially performed a Shapiro-Wilk test to check if ical tubes containing RNAlater (Life Technologies, Grand the data was normally distributed [66]. In case of viola- Island, NY) and refrigerated prior to shipment. Upon ar- tion of normality, data was log transformed and checked rival at Cornell University, the samples were divided into again for normality. Both parametric independent samples aliquots and stored at −80 °C until processing. Blood t test and a non-parametric Wilcoxon-Mann-Whitney U samples were drawn into EDTA and heparin tubes from test were used to determine the significance of differences an antecubital vein with subjects in the seated position. in each subject group. Values of P < 0.05 were considered Samples were shipped by overnight courier from New statistically significant. All data from the biomarkers levels York City to Cornell University (Ithaca). Upon receipt, determination were processed and analyzed in SPSS samples were centrifuged at 4000 r.p.m. for 30 min to Statistics Version 21 (Armonk, NY: IBM Corp). pellet blood cells, and plasma was stored at −80 °C until further analyses. BMI, age, and gender of subjects were DNA extraction, 16S rRNA gene sequencing recorded. ME/CFS subjects completed the Short Form Metagenomic DNA was isolated from an aliquot of 36 Health Survey (SF-36.org) and Bell’s Disability Scale ~100 mg from each fecal sample using the PowerSoil-htp [65]. Potential controls from the same geographic area DNA isolation kit (MoBio Laboratories Ltd, Carlsbad, CA), as cases were screened by the physician for suitability which involves both chemical and physical lysis of the cells. as healthy controls. As indicated in Table 1, mean and We amplified 16S rRNA genes (V4 hypervariable region) median ages of cases and controls were within 5 years from bulk DNA using the 515F and 806R primers as and the female to male ratios were similar. previously described [67] prior to sequencing. Duplicate PCR reactions of samples and extraction blanks consisted Plasma level determination of hsCRP, sCD14, LBP, LPS, of 2.5X HotMasterMix (5-Prime, Inc., Gaithersburg, MD), and I-FABP 10–100 ng DNA template, and 0.05 μMofeach primer. High-sensitive C-reactive protein (hsCRP) was measured DNA amplification of samples and extraction blanks was from unhemolyzed EDTA plasma using a Chemilumines- performed on a 96-well plate with a minicycler PTC 200 cence immunoassay on an Immulite 2000 (Siemens (MJ Research) starting with 3-min denaturation at 94 °C, Medical Solutions Diagnostics, Deerfield, IL). Markers followed by 25 cycles consisting of denaturation (45 s at for microbial translocation, sCD14 and LBP, were measured 94 °C), annealing (60 s at 50 °C), extension (90 s at 72 °C), Giloteaux et al. Microbiome (2016) 4:30 Page 10 of 12 and a final extension at 72 °C for 10 min. Samples were A machine learning approach was used to identify vari- randomly distributed on the plate with no grouping for ables discriminating the two groups of samples (feature sample type. The replicate PCR reactions were combined selection). For these analyses, we used either 97 % OTUs, and purified using a magnetic bead system (Mag-Bind or taxa abundances based on combining OTUs at the spe- EZPure, Omega Bio-Tek, Norcross, GA). PCR amplicons cies and genus levels. Classification of samples as healthy were quantified using the QuantiT PicoGreen dsDNA controls or ME/CFS was carried out by using a random Assay Kit (Invitrogen, Carlsbad, CA). Aliquots of amplicons forest approach with supervised learning [73] and area (at equal masses) were combined for a final concentration under the curve (AUC) calculation to optimize feature of approximately 15 ng/μl. Extraction blanks showed no (e.g., abundance of a particular genus) selection, imple- amplification. All amplicons were then sequenced on a mented in the software package R. Scripts, required pack- single run using the Illumina MiSeq 2x250 bp platform at ages, and instructions for processing data are available on Cornell Biotechnology Resource Center Genomics Facility. https://gist.github.com/walterst/2222618976a66b3fc8dd. In Quality filtering and analysis of the 16S rRNA gene addition to the taxonomic abundance data, levels of in- sequence data were performed with QIIME 1.9.0 as flammatory markers (BMI, sCD14, LBP, LPS, and I-FABP) previously described [68]. Briefly, matching paired-end were included in the analysis. Average accuracies were raw sequences (mate-pairs) were merged using the fastq- calculated with five repeats of 10-fold cross validation, join command in the ea-utils software package (http:// which is intended to predict the accuracy of the model, code.google.com/p/ea-utils), and merged sequences with and indicate over-fitting if significantly different than less than a 200-bp overlap were filtered out of the dataset. the full dataset results, by subsampling the data and The remaining merged sequences were quality filtered testing this training subsample against the remaining and assigned to samples based on their barcodes using the data reference set. default parameters of QIIME. Sequences were assigned to 97 % ID OTUs by comparing them to a non-redundant Additional files reference database of near-full length sequences [31]. All OTUs that were observed fewer than two times, i.e., sin- Additional file 1: Figure S1. 36-Item Short Form Health Survey (SF-36) gletons, were removed from the analysis. The OTU table profiles from studies reporting SF-36 scores for individuals with a ME/CFS was rarefied to the sequence count of the sample with the diagnosis. (PDF 2514 kb) lowest sequence depth, 32,223 sequences per sample, and Additional file 2: Figure S2. Principal Coordinate Analysis (PCoA) plot of healthy controls versus subjects with ME/CFS. Distances were calculated used in all subsequent analyses. For statistical compari- with weighted UniFrac (a) and unweighted UniFrac (b). Data were evenly sons of healthy individuals to those afflicted with ME/ sampled at 32223 sequences per sample. (PDF 2631 kb) CFS, p values obtained with the Wilcoxon-Mann-Whitney Additional file 3: Table S1. Feature Importance Scores for genus-level U test were corrected for multiple comparisons using the supervised learning. The feature importance score is the percentage increase in error rate when the given feature is permuted while other values remain false discovery rate of Benjamini and Hochberg, imple- constant. As there are only two categories, the increased error rate is equal mented in the QIIME pipeline. We used both the weighted for both categories. (XLSX 42 kb) and unweighted UniFrac distance metrics as measures of Additional file 4: Figure S3. Confusion matrices for random forest between-sample (beta) diversity and applied principal coor- analysis of microbial sequencing data (values are presented as %) and ROC area under the curve (AUC) values at the genus (a), species (b) and OTU (c) dinates analysis (PCoA) to visualize patterns of diversity. level. (PDF 1170 kb) Within-samples (alpha) diversity was calculated using three Additional file 5: Table S2. Per-sample metadata mapping file used different measures (1) ChaoI index [69]; (2) Shannon Index throughout the QIIME pipeline. (XLSX 61 kb) [70]; and (3) Phylogenetic Diversity [71]. LEfSe analysis and machine learning Abbreviations hsCRP, high sensitivity C-reactive protein; IBD, inflammatory bowel disease; Linear discriminant effect size analysis (LEfSe) on filtered IBS, irritable bowel syndrome; I-FABP, intestinal fatty acid binding protein; datasets [72] was performed at the genus level to find fea- LBP, lipopolysaccharide-binding protein; LPS, lipopolysaccharides; ME/CFS, tures (genera) differentially represented between healthy myalgic encephalomyelitis/chronic fatigue syndrome; MT, microbial translocation; sCD14, soluble CD14 and ME/CFS groups. LEfSe combines the standard tests for statistical significance (Kruskal-Wallis test and pair- wise Wilcoxon test) with linear discriminate analysis. It Acknowledgements We thank the subjects for providing samples and information for the study ranks features by effect size, which put features that ex- and Lin Lin for technical assistance. plain most of the biological difference at top. LEfSe ana- lysis was performed under the following conditions: the α Funding value for the factorial Kruskal-Wallis test among classes This work was supported by grant 1R21AI101614 from NIH NIAID to M.R.H. was 0.05 and the threshold on the logarithmic LDA score and R.E.L. The funders had no role in study design, data collection, analysis for discriminative features was 2.0. and interpretation, decision to publish, or preparation of the manuscript. Giloteaux et al. Microbiome (2016) 4:30 Page 11 of 12 Availability of data and materials imbalances in human inflammatory bowel diseases. Proc Natl Acad Sci U S The sequence data supporting the results of this article are available in the A. 2007;104(34):13780–5. European Bioinformatics Institute Sequence Read Archive under accession 14. Qin J, Li Y, Cai Z, Li S, Zhu J, Zhang F, et al. A metagenome-wide number PRJEB13092. The mapping file used for the QIIME pipeline is association study of gut microbiota in type 2 diabetes. Nature. 2012; available on Additional file 5: Table S2. 490(7418):55–60. 15. Komaroff AL, Buchwald DS. Chronic fatigue syndrome: an update. Annu Rev Med. 1998;49:1–13. Authors’ contributions 16. Hornig M, Gottschalk G, Peterson DL, Knox KK, Schultz AF, Eddy ML, et al. LG designed the experiments, processed the samples, conducted the Cytokine network analysis of cerebrospinal fluid in myalgic experiments, and performed the statistical analysis with SPSS. SML recruited, encephalomyelitis/chronic fatigue syndrome. Mol Psychiatry. 2016;21(2):261– diagnosed, and sampled the blood from the subjects. LG and JKG performed the sequence analysis using QIIME. WAW performed the supervised learning 17. Hornig M, Montoya JG, Klimas NG, Levine S, Felsenstein D, Bateman L, et al. machine analysis. RL and MH contributed to study design. LG, JKG, WAW, RL, Distinct plasma immune signatures in ME/CFS are present early in the and MH performed analysis and writing. All authors read and approved the course of illness. Sci Adv. 2015;1(1). final manuscript. 18. Vyboh K, Jenabian MA, Mehraj V, Routy JP. HIV and the gut microbiota, partners in crime: breaking the vicious cycle to unearth new therapeutic Competing interests targets. J Immunol Res. 2015;2015:614127. The authors declare that they have no competing interests. 19. Maes M, Mihaylova I, Leunis JC. Increased serum IgA and IgM against LPS of enterobacteria in chronic fatigue syndrome (CFS): indication for the Ethics approval and consent to participate involvement of gram-negative enterobacteria in the etiology of CFS and for The institutional review board of Cornell University approved the study, and the presence of an increased gut-intestinal permeability. J Affect Disord. patients and controls gave written informed consent after the study protocol 2007;99(1–3):237–40. was fully explained. All consented to blood draw, to provide a stool sample 20. Schumann RR. Function of lipopolysaccharide (LPS)-binding protein (LBP) and to the availability of the stored samples for additional bioassays. and CD14, the receptor for LPS/LBP complexes: a short review. Res Immunol. 1992;143(1):11–5. Author details 21. Pelsers MMAL, Hermens WT, Glatz JFC. Fatty acid-binding proteins as Department of Molecular Biology and Genetics, Cornell University, Ithaca, plasma markers of tissue injury. Clinica Chimica Acta. 2005;352(1–2):15–35. NY, USA. Department of Microbiology, Cornell University, Ithaca, NY, USA. 22. Medicine II. Beyond myalgic encephalomyelitis/chronic fatigue syndrome: Private Practice, New York, NY, USA. redefining an illness. Washington, D.C: The National Academies Press; 2015. 23. Zhu L, Baker SS, Gill C, Liu W, Alkhouri R, Baker RD, et al. Characterization Received: 29 February 2016 Accepted: 11 May 2016 of gut microbiomes in nonalcoholic steatohepatitis (NASH) patients: a connection between endogenous alcohol and NASH. Hepatology. 2013; 57(2):601–9. 24. Openshaw PJ. Crossing barriers: infections of the lung and the gut. Mucosal References Immunol. 2009;2(2):100–2. 1. Fukuda K, Straus SE, Hickie I, Sharpe MC, Dobbins JG, Komaroff A. The 25. Wellmann W, Fink PC, Benner F, Schmidt FW. Endotoxaemia in active chronic fatigue syndrome: a comprehensive approach to its definition and Crohn’s disease. Treatment with whole gut irrigation and 5-aminosalicylic study. International Chronic Fatigue Syndrome Study Group. Ann Intern acid. Gut. 1986;27(7):814–20. Med. 1994;121(12):953–9. 26. Wyatt J, Vogelsang H, Hubl W, Waldhoer T, Lochs H. Intestinal permeability and 2. Riedl A, Schmidtmann M, Stengel A, Goebel M, Wisser AS, Klapp BF, et al. the prediction of relapse in Crohn’s disease. Lancet. 1993;341(8858):1437–9. Somatic comorbidities of irritable bowel syndrome: a systematic analysis. 27. Anas AA, Hovius JW, van 't Veer C, van der Poll T, de Vos AF. Role of CD14 J Psychosom Res. 2008;64(6):573–82. in a mouse model of acute lung inflammation induced by different 3. Whitehead WE, Palsson O, Jones KR. Systematic review of the comorbidity lipopolysaccharide chemotypes. PLoS One. 2010;5(4):e10183. of irritable bowel syndrome with other disorders: what are the causes and 28. Wolk K, Witte E, Hoffmann U, Doecke WD, Endesfelder S, Asadullah K, et al. implications? Gastroenterology. 2002;122(4):1140–56. IL-22 induces lipopolysaccharide-binding protein in hepatocytes: a potential 4. Wojczynski MK, North KE, Pedersen NL, Sullivan PF. Irritable bowel syndrome: systemic role of IL-22 in Crohn’s disease. J Immunol. 2007;178(9):5973–81. a co-twin control analysis. Am J Gastroenterol. 2007;102(10):2220–9. 29. Wan Y, Freeswick PD, Khemlani LS, Kispert PH, Wang SC, Su GL, et al. Role of 5. Aaron LA, Burke MM, Buchwald D. Overlapping conditions among patients lipopolysaccharide (LPS), interleukin-1, interleukin-6, tumor necrosis factor, with chronic fatigue syndrome, fibromyalgia, and temporomandibular and dexamethasone in regulation of LPS-binding protein expression in disorder. Arch Intern Med. 2000;160(2):221–7. normal hepatocytes and hepatocytes from LPS-treated rats. Infect Immun. 6. Komaroff AL, Buchwald D. Symptoms and signs of chronic fatigue 1995;63(7):2435–42. syndrome. Rev Infect Dis. 1991;13 Suppl 1:S8–11. 30. Butt HL, Dunstan R, N.R. M, T. R. Bacterial colonosis in patients with 7. Jones JF, Nisenbaum R, Reeves WC. Medication use by persons with chronic persistent fatigue. In Proceedings of the AHMF international clinical and fatigue syndrome: results of a randomized telephone survey in Wichita, Kansas. scientific conference Sydney, Australia: AHMF; 2001. Health Qual Life Outcomes. 2003;1:74. 31. Brodie EL, Desantis TZ, Joyner DC, Baek SM, Larsen JT, Andersen GL, et al. 8. Borody TJ, Nowak A, Finlayson S. The GI microbiome and its role in chronic Application of a high-density oligonucleotide microarray approach to study fatigue syndrome: a summary of bacteriotherapy. ACNEM J. 2012;31(3):3–8. bacterial population dynamics during uranium reduction and reoxidation. 9. Sullivan A, Nord CE, Evengard B. Effect of supplement with lactic-acid Appl Environ Microbiol. 2006;72(9):6288–98. producing bacteria on fatigue and physical activity in patients with chronic 32. Chang JY, Antonopoulos DA, Kalra A, Tonelli A, Khalife WT, Schmidt TM, fatigue syndrome. Nutr J. 2009;8:4. et al. Decreased diversity of the fecal microbiome in recurrent Clostridium 10. Rao AV, Bested AC, Beaulne TM, Katzman MA, Iorio C, Berardi JM, et al. A difficile-associated diarrhea. J Infect Dis. 2008;197(3):435–8. randomized, double-blind, placebo-controlled pilot study of a probiotic in 33. Willing BP, Dicksved J, Halfvarson J, Andersson AF, Lucio M, Zheng Z, et al. emotional symptoms of chronic fatigue syndrome. Gut Pathog. 2009;1(1):6. A pyrosequencing study in twins shows that gastrointestinal microbial 11. Sheedy JR, Wettenhall RE, Scanlon D, Gooley PR, Lewis DP, McGregor N, et profiles vary with inflammatory bowel disease phenotypes. al. Increased d-lactic acid intestinal bacteria in patients with chronic fatigue Gastroenterology. 2010;139(6):1844–54. e1. syndrome. In Vivo. 2009;23(4):621–8. 12. Fremont M, Coomans D, Massart S, De Meirleir K. High-throughput 16S 34. McMurtry VE, Gupta RW, Tran L, Blanchard EE, Penn D, Taylor CM, et al. rRNA gene sequencing reveals alterations of intestinal microbiota in Bacterial diversity and Clostridia abundance decrease with increasing myalgic encephalomyelitis/chronic fatigue syndrome patients. Anaerobe. severity of necrotizing enterocolitis. Microbiome. 2015;3:11. 2013;22:50–6. 35. Fujimoto T, Imaeda H, Takahashi K, Kasumi E, Bamba S, Fujiyama Y, et al. 13. Frank DN, St Amand AL, Feldman RA, Boedeker EC, Harpaz N, Pace NR. Decreased abundance of Faecalibacterium prausnitzii in the gut microbiota Molecular-phylogenetic characterization of microbial community of Crohn’s disease. J Gastroenterol Hepatol. 2013;28(4):613–9. Giloteaux et al. Microbiome (2016) 4:30 Page 12 of 12 36. Hippe B,Remely M,Bartosiewicz N, RiedelM, Nichterl C,Schatz L, etal. 57. Maukonen J, Satokari R, Matto J, Soderlund H, Mattila-Sandholm T, Saarela M. Abundance and diversity of GI microbiota rather than IgG4 levels correlate with Prevalence and temporal stability of selected clostridial groups in irritable abdominal inconvenience and gut permeability in consumers claiming food bowel syndrome in relation to predominant faecal bacteria. J Med Microbiol. intolerances. Endocr Metab Immune Disord Drug Targets. 2014;14(1):67–75. 2006;55(Pt 5):625–33. 37. Sokol H, Pigneur B, Watterlot L, Lakhdari O, Bermudez-Humaran LG, 58. Swidsinski A, Ladhoff A, Pernthaler A, Swidsinski S, Loening-Baucke V, Ortner M, Gratadoux JJ, et al. Faecalibacterium prausnitzii is an anti-inflammatory et al. Mucosal flora in inflammatory bowel disease. Gastroenterology. 2002; commensal bacterium identified by gut microbiota analysis of Crohn 122(1):44–54. disease patients. Proc Natl Acad Sci U S A. 2008;105(43):16731–6. 59. Wu X, Ma C, Han L, Nawaz M, Gao F, Zhang X, et al. Molecular characterisation of the faecal microbiota in patients with type II diabetes. Curr Microbiol. 2010; 38. Martinez-Medina M, Aldeguer X, Gonzalez-Huix F, Acero D, Garcia-Gil LJ. 61(1):69–78. Abnormal microbiota composition in the ileocolonic mucosa of Crohn’s 60. Furrie E, Macfarlane S, Kennedy A, Cummings JH, Walsh SV, O'Neil DA, et al. disease patients as revealed by polymerase chain reaction-denaturing Synbiotic therapy (Bifidobacterium longum/Synergy 1) initiates resolution of gradient gel electrophoresis. Inflamm Bowel Dis. 2006;12(12):1136–45. inflammation in patients with active ulcerative colitis: a randomised 39. Sartor RB. Key questions to guide a better understanding of host- controlled pilot trial. Gut. 2005;54(2):242–9. commensal microbiota interactions in intestinal inflammation. Mucosal 61. Groeger D, O'Mahony L, Murphy EF, Bourke JF, Dinan TG, Kiely B, et al. Immunol. 2011;4(2):127–32. Bifidobacterium infantis 35624 modulates host inflammatory processes 40. Sartor RB. Genetics and environmental interactions shape the intestinal beyond the gut. Gut Microbes. 2013;4(4):325–39. microbiome to promote inflammatory bowel disease versus mucosal 62. Knights D, Costello EK, Knight R. Supervised classification of human homeostasis. Gastroenterology. 2010;139(6):1816–9. microbiota. FEMS Microbiol Rev. 2011;35(2):343–59. 41. Davies CA, Rocks SA, O'Shaughnessy MC, Perrett D, Winyard PG. Analysis of 63. Papa E, Docktor M, Smillie C, Weber S, Preheim SP, Gevers D, et al. Non-invasive nitrite and nitrate in the study of inflammation. Methods Mol Biol. 2003;225: mapping of the gastrointestinal microbiota identifies children with inflammatory 305–20. bowel disease. PLoS One. 2012;7(6):e39242. 42. Thiennimitr P, Winter SE, Winter MG, Xavier MN, Tolstikov V, Huseby DL, et al. 64. Walters WA, Xu Z, Knight R. Meta-analyses of human gut microbes Intestinal inflammation allows Salmonella to use ethanolamine to compete associated with obesity and IBD. FEBS Lett. 2014;588(22):4223–33. with the microbiota. Proc Natl Acad Sci U S A. 2011;108(42):17480–5. 65. Bell DS. The doctor’s guide to chronic fatigue syndrome. Reading, Mass: 43. Winter SE, Lopez CA, Baumler AJ. The dynamics of gut-associated microbial Addison-Wesley; 1995. communities during inflammation. EMBO Rep. 2013;14(4):319–27. 66. Shapiro SS, Wilk MB. An analysis of variance test for normality (complete 44. Winter SE, Winter MG, Xavier MN, Thiennimitr P, Poon V, Keestra AM, et al. samples). Biometrika. 1965;52(3–4):591–611. Host-derived nitrate boosts growth of E. coli in the inflamed gut. Science. 67. Caporaso JG, Lauber CL, Walters WA, Berg-Lyons D, Lozupone CA, Turnbaugh PJ, 2013;339(6120):708–11. et al. Global patterns of 16S rRNA diversity at a depth of millions of sequences 45. Quevrain E, Maubert MA, Michon C, Chain F, Marquant R, Tailhades J, et al. per sample. Proc Natl Acad Sci U S A. 2011;108 Suppl 1:4516–22. Identification of an anti-inflammatory protein from Faecalibacterium 68. Goodrich JK, Waters JL, Poole AC, Sutter JL, Koren O, Blekhman R, et al. prausnitzii, a commensal bacterium deficient in Crohn’s disease. Gut. 2016; Human genetics shape the gut microbiome. Cell. 2014;159(4):789–99. 65(3):415–25. 69. Chao A. Non-parametric estimation of the number of classes in a 46. Machiels K, Joossens M, Sabino J, De Preter V, Arijs I, Eeckhaut V, et al. A population. Scandinavian Journal of Statistics. 1984;11:265–70. decrease of the butyrate-producing species Roseburia hominis and 70. Shannon CE. A mathematical theory of communication. Bell SystTech J. Faecalibacterium prausnitzii defines dysbiosis in patients with ulcerative 1948;27:379–423. colitis. Gut. 2014;63(8):1275–83. 71. Faith DP. Conservation evaluation and phylogenetic diversity. Biol Conserv. 47. Pozuelo M, Panda S, Santiago A, Mendez S, Accarino A, Santos J, et al. 1992;61:1–10. Reduction of butyrate- and methane-producing microorganisms in patients 72. Segata N, Izard J, Waldron L, Gevers D, Miropolsky L, Garrett WS, et al. with irritable bowel syndrome. Sci Rep. 2015;5:12693. Metagenomic biomarker discovery and explanation. Genome Biol. 2011; 48. Varela E, Manichanh C, Gallart M, Torrejon A, Borruel N, Casellas F, et al. 12(6):R60. Colonisation by Faecalibacterium prausnitzii and maintenance of clinical 73. Liaw A, Wiener M. Classification and regression by randomForest. R News. remission in patients with ulcerative colitis. Aliment Pharmacol Ther. 2013; 2002;2:18–22. 38(2):151–61. 49. Sokol H, Seksik P, Furet JP, Firmesse O, Nion-Larmurier I, Beaugerie L, et al. Low counts of Faecalibacterium prausnitzii in colitis microbiota. Inflamm Bowel Dis. 2009;15(8):1183–9. 50. Saemann MD, Bohmig GA, Osterreicher CH, Burtscher H, Parolini O, Diakos C, et al. Anti-inflammatory effects of sodium butyrate on human monocytes: potent inhibition of IL-12 and up-regulation of IL-10 production. FASEB J. 2000; 14(15):2380–2. 51. Segain JP, Raingeard de la Bletiere D, Bourreille A, Leray V, Gervois N, Rosales C, et al. Butyrate inhibits inflammatory responses through NFkappaB inhibition: implications for Crohn’s disease. Gut. 2000;47(3):397–403. 52. Vinolo MA, Rodrigues HG, Nachbar RT, Curi R. Regulation of inflammation by short chain fatty acids. Nutrients. 2011;3(10):858–76. 53. Kassinen A, Krogius-Kurikka L, Makivuokko H, Rinttila T, Paulin L, Corander J, et Submit your next manuscript to BioMed Central al. The fecal microbiota of irritable bowel syndrome patients differs significantly from that of healthy subjects. Gastroenterology. 2007;133(1):24–33. and we will help you at every step: 54. Tana C, Umesaki Y, Imaoka A, Handa T, Kanazawa M, Fukudo S. Altered • We accept pre-submission inquiries profiles of intestinal microbiota and organic acids may be the origin of symptoms in irritable bowel syndrome. Neurogastroenterol Motil. 2010; � Our selector tool helps you to find the most relevant journal 22(5):512–9. e114-5. � We provide round the clock customer support 55. Malinen E, Rinttila T, Kajander K, Matto J, Kassinen A, Krogius L, et al. � Convenient online submission Analysis of the fecal microbiota of irritable bowel syndrome patients and healthy controls with real-time PCR. Am J Gastroenterol. 2005;100(2):373–82. � Thorough peer review 56. Matto J, Maunuksela L, Kajander K, Palva A, Korpela R, Kassinen A, et al. � Inclusion in PubMed and all major indexing services Composition and temporal stability of gastrointestinal microbiota in irritable � Maximum visibility for your research bowel syndrome—a longitudinal study in IBS and control subjects. FEMS Immunol Med Microbiol. 2005;43(2):213–22. Submit your manuscript at www.biomedcentral.com/submit http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Microbiome Springer Journals

Reduced diversity and altered composition of the gut microbiome in individuals with myalgic encephalomyelitis/chronic fatigue syndrome

Download PDF
12 pages

Loading next page...
 
/lp/springer-journals/reduced-diversity-and-altered-composition-of-the-gut-microbiome-in-sNeIWbxoE0

References (80)

Publisher
Springer Journals
Copyright
Copyright © 2016 by The Author(s).
Subject
Biomedicine; Medical Microbiology; Bioinformatics; Microbial Ecology; Microbiology; Microbial Genetics and Genomics; Virology
eISSN
2049-2618
DOI
10.1186/s40168-016-0171-4
pmid
27338587
Publisher site
See Article on Publisher Site

Abstract

Background: Gastrointestinal disturbances are among symptoms commonly reported by individuals diagnosed with myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS). However, whether ME/CFS is associated with an altered microbiome has remained uncertain. Here, we profiled gut microbial diversity by sequencing 16S ribosomal ribonucleic acid (rRNA) genes from stool as well as inflammatory markers from serum for cases (n =48) and controls (n = 39). We also examined a set of inflammatory markers in blood: C-reactive protein (CRP), intestinal fatty acid-binding protein (I-FABP), lipopolysaccharide (LPS), LPS-binding protein (LBP), and soluble CD14 (sCD14). Results: We observed elevated levels of some blood markers for microbial translocation in ME/CFS patients; levels of LPS, LBP, and sCD14 were elevated in ME/CFS subjects. Levels of LBP correlated with LPS and sCD14 and LPS levels correlated with sCD14. Through deep sequencing of bacterial rRNA markers, we identified differences between the gut microbiomes of healthy individuals and patients with ME/CFS. We observed that bacterial diversity was decreased in the ME/CFS specimens compared to controls, in particular, a reduction in the relative abundance and diversity of members belonging to the Firmicutes phylum. In the patient cohort, we find less diversity as well as increases in specific species often reported to be pro-inflammatory species and reduction in species frequently described as anti-inflammatory. Using a machine learning approach trained on the data obtained from 16S rRNA and inflammatory markers, individuals were classified correctly as ME/CFS with a cross-validation accuracy of 82.93 %. Conclusions: Our results indicate dysbiosis of the gut microbiota in this disease and further suggest an increased incidence of microbial translocation, which may play a role in inflammatory symptoms in ME/CFS. Keywords: Myalgic encephalomyelitis, Chronic fatigue syndrome, Inflammation, Lipopolysaccharides, Microbiome, Microbial translocation, Beta-diversity Background have been the basis of the widely used Fukuda diagnostic Myalgic encephalomyelitis (ME), also known as chronic criteria [1]. Many ME/CFS patients also report gastro- fatigue syndrome (CFS), or ME/CFS, is a debilitating intestinal (GI) symptoms, including but not limited to illness of unknown etiology with no widely accepted irritable bowel syndrome (IBS) [2–6]. Intestinal discomfort therapy. Primary symptoms reported by patients are fa- is also indicated in a survey of drug use by individuals with tigue, muscle and/or joint paint, sore throat, headaches, CFS compared to controls, which found significantly more unrefreshing sleep, and post-exertional malaise and use of antacids, H2 blockers, and proton pump inhibitors in the ME/CFS cohort [7]. The prevalence of bowel symptoms has led to attempts * Correspondence: [email protected] 1 to treat the disease by probiotic oral or rectal supplements. Department of Molecular Biology and Genetics, Cornell University, Ithaca, NY, USA Borody et al. [8] reported improvements in a majority of Full list of author information is available at the end of the article © 2016 The Author(s). Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Giloteaux et al. Microbiome (2016) 4:30 Page 2 of 12 patients at 4 weeks following bacteriotherapy comprised self-reported good health, 34 ME/CFS patients and 7 con- of rectal infusion of 13 enteric bacteria, though the trols self-reported gastrointestinal disturbances such as number with a sustained response was not well docu- constipation, diarrhea, or intestinal discomfort. Many ME/ mented. In two small studies, marginal improvement in CFS patients are able to identify an acute, often flu-like, certain symptoms was reported following oral probiotic illness that immediately preceded the onset of the dis- therapy [9, 10]. ease eventually diagnosed as ME/CFS, while others are Two reports suggest altered gut microbiota in ME/CFS unaware of an initiating event and consider their onset patients. Using culture-based methods, Sheedy et al. [11] to be gradual. Among the 48 ME/CFS patients in the described higher levels of D-lactic acid producing Entero- study, 19 indicated a gradual and 25 stated a sudden onset. coccus and Streptococcus spp. in ME/CFS patients vs. con- ME/CFS subjects completed the SF-36 form (Additional trols. More recently, Norwegian ME/CFS patients and file 1: Figure S1) and Bell’s Disability scale (Table 1). healthy controls were found to exhibit differences in gut In comparison to other studies in which patients diag- microbiota composition through a 16S rRNA gene se- nosed with ME/CFS also filled out the SF-36 form, our quencing study [12]. It is well documented that gut study population fell within the same ranges on the eight microbiota can be significant with respect to patho- subscales of the SF-36 (Additional file 1: Figure S1). logical intestinal conditions such as ulcerative colitis (UC), Crohn’s disease (CD) [13], and systemic diseases Measurements of levels of microbial translocation such as diabetes [14]. Because of the frequent occur- markers indicate microbial translocation rence of GI disturbances, as well as these prior reports We quantified plasma levels of hsCRP, lipopolysaccha- of abnormalities, we investigated the diversity and com- rides (LPS) as a marker of microbial translocation (MT) position of the gut microbiota of ME/CFS patients in and plasma intestinal fatty acid binding protein (I-FABP) comparison to healthy individuals. as a marker for gastrointestinal tract damage in both Along with GI symptoms, individuals with ME/CFS ap- groups. The distribution of plasma hsCRP, LPS and I-FABP pear to have both immune activation and immune dys- is showninFig.1.Levels of hsCRP were higher in theME/ function. Many of the common symptoms reported by CFSpopulationincomparisontohealthycontrols(1.38 ME/CFS patients are characteristic of inflammatory ill- and 1.21 mg/L, respectively), but the difference was not nesses [15]. Most reports concerning cytokine levels in statistically significant (P = 0.15, Fig. 1a, Table 2). ME/CFS patients vs. controls are somewhat limited in ME/CFS patients had significantly higher plasma LPS scope and discordant, but several recent papers with a levels than healthy individuals (median ME/CFS— 51-plex cytokine assay indicate abnormal immune sig- 119.43 pg/mL vs. controls—74.74 pg/mL, P < 0.0005, natures in plasma and cerebrospinal fluid [16, 17]. Fig. 1b and Table 2). The median plasma I-FABP level Abnormal immune activation can be caused by trans- was 341.9 pg/mL in the ME/CFS group and 301 pg/mL in location of microbes from an inflamed gut [18]. A prior the healthy group. Though the median I-FABP levels in report indicated increased IgA and IgM to lipopolysac- the ME/CFS group was higher than that of the healthy charide (LPS) in serum of CFS patients [19]. We therefore group, the difference was not statistically significant assayed plasma levels of LPS and LPS-binding protein, as (P = 0.27, Fig. 1c, Table 2). well as the LPS/LBP receptor sCD14 [20]. We also exam- To obtain further information concerning chronic LPS ined the levels of C-reactive protein, an inflammatory stimulation in vivo, we also measured plasma sCD14 levels marker, and I-FABP as a marker for gastrointestinal tract and plasma LBP, which is produced by gastrointestinal and integrity [21]. Table 1 Characteristics of the study population Objective molecular markers for diagnosis of ME/CFS are lacking. We examined the levels of plasma markers Controls (n = 39) ME/CFS (n = 49) and microbiota composition in the diseased vs. healthy Gender Female 30 38 subjects in order to determine whether the data, taken Male 9 11 together, could predict ME/CFS vs. healthy status. Age Mean ± SD 45.5 ± 9.9 50.2 ± 12.6 Median (range) 48 (20–61) 51 (19–71) Results BMI Mean ± SD 27.1 ± 6.1 25.5 ± 4.9 Study population characteristics Median (range) 26 (17–47) 24.5 (16–40) Subjects with ME/CFS were established patients of a ME/ CFS specialist, Susan Levine, M.D. and fit the Fukuda diag- Bell’s disability scale 10–20 NA 15 nostic criteria [1]. This study began before the criteria for 30–40 21 systemic exertion intolerance disease (SEID) were estab- 50–60 9 lished [22], but most, perhaps all, also fit the description of >60 4 SEID. Of the 48 patients and 39 control participants who Giloteaux et al. Microbiome (2016) 4:30 Page 3 of 12 Fig. 1 Microbial translocation, gastrointestinal tract damage, and evidence for direct LPS stimulation in vivo in ME/CFS: plasma levels of hsCRP (a), LPS (b), I-FABP (c), sCD14 (d), and LBP (e) determined in our cohorts of controls and ME/CFS diagnosed individuals. p values were calculated by the Wilcoxon-Mann-Whitney U test hepatic epithelial cells. Thus, increased LPS in the circula- Stool microbiota of ME/CFS patients exhibit reduced tion promote hepatic synthesis of LBP, a plasma protein diversity and different composition than healthy controls that increases the binding of LPS to CD14. sCD14 and The hypervariable V4 region of 16S rRNA genes was se- LBP concentrations in both groups are shown in Fig. 1. For quenced from fecal samples of individuals with ME/CFS the ME/CFS cohort the median plasma sCD14 concentra- (n = 48) and healthy individuals (n = 39). A total of tion was 1.97 ug/mL, and the median LBP plasma concen- 8,534,117 high-quality and classifiable reads were gener- tration was 17.68 ug/mL. These values were significantly ated from all samples, with an average of 98,093 ± 29,231 different from the plasma sCD14 and LBP concentra- reads per sample. Binning sequences using a pairwise tions of the healthy volunteers (1.36 ug/mL; P < 0.0005 identity threshold of 97 %, we obtained an average of and 12.32 ug/mL; P < 0.0005, respectively) (Fig. 1d, e, 1330 ± 423 operational taxonomic units (OTUs) per Table 2). sample. The sequence-based rarefaction curves based on Next, we analyzed the associations among biomarker the Phylogenetic Diversity (PD) metric were nearly asymp- measurements in the ME/CFS population. As can be totic and a Wilcoxon rank-sum test demonstrated a statis- seen in Fig. 2a, b plasma LPS levels correlated posi- tical difference in the diversity of ME/CFS and healthy tively with levels of sCD14 and LBP (r= 0.347, P< 0.01 individuals (P = 0.004, W = 1268) (Fig. 3a). and r = 0.487, P< 0.01, respectively), consistent with We examined the number of “observed species,” i.e., the stimulation of sCD14 production by LPS in vivo. In number of 97 % ID OTUs observed in 32,223 sequences, addition, we found a strong significant correlation be- the estimators of community evenness (Shannon H), and tween plasma sCD14 and hsCRP and sCD14 and LBP; richness (Chao1 and PD) in the two group of samples. high levels of sCD14 were associated with high levels of ME/CFS samples had a significant overall lower microbial hsCRP (r = 0.507, P <0.01) and LBP (r =0.578, P <0.01) diversity, with lower evenness (H = 5.33 ± 0.93 vs. 5.92 ± (Fig. 2c, d). We also analyzed whether enterocyte dam- 0.93, P = 0.004), and lower richness (observed species, age (i.e., I-FABP levels) was associated with the pro- 1204 ± 351 vs. 1486 ± 456; Chao1, 2363 ± 704 vs. 2918 ± posed microbial translocation markers LPS, sCD14, and 885, P = 0.002; PD, 61.6 ± 16.7 vs. 73.4 ± 19.04, P =0.004) LBP. We found no relationship between I-FABP and (Fig. 3b). LPS levels (r = −0.125; P =0.278), I-FABP and sCD14 To evaluate overall differences in beta-diversity between levels (r = −0.117; P = 0.310), or I-FABP and LBP levels the microbiomes, we applied Principal Component Ana- (r = −0.08; P =0.488). lysis (PCoA) to weighted and unweighted UniFrac distance Giloteaux et al. Microbiome (2016) 4:30 Page 4 of 12 Table 2 Plasma levels of markers of inflammation (hsCRP), healthy individuals, this corresponded to 46 and 45 % re- microbial translocation (LPS, sCD14, and LBP) and gastrointestinal spectively of the rarified 16S rRNA sequences. Also, Pro- damage (I-FABP) in ME/CFS and healthy individuals teobacteria made up the next largest represented phylum Analytes Control ME/CFS (3.6 %), with Verrucomicrobia and Actinobacteria in rela- hsCRP (mg/L) tively low relative abundance (2.1 and 1.6 %, respectively). At the phylum level, the abundance of the Bacteroidetes n 34 46 was comparable (52 %) in both datasets (Fig. 4a). ME/CFS Median 1.21 1.38 samples showed lower relative abundance of Firmicutes Quartiles 0.38–2.69 0.70–2.71 (35 %) (Fig. 4a) and higher relative abundance of Proteo- Range 0.27–5.09 0.23–19.3 bacteria (8 %), due almost entirely to a twofold increase in LPS (pg/mL) the Proteobacteria family Enterobacteriaceae (6 vs. 3 % for n 39 49 ME/CFS and healthy individuals, respectively) (Fig. 4b). Within the Firmicutes, at the family level, Ruminococca- Median 74.74 119.43 ceae were lower in the ME/CFS samples (16 vs. 11 % in Quartiles 54.34–99.54 66.21–144.41 ME/CFS and healthy individuals respectively) (Fig. 4b), Range 32.21–187.32 34.32–279.30 whereas Lachnospiraceae were similar among both data- I-FABP (pg/mL) sets (16 % for both healthy and ME/CFS samples). Some n 33 44 differences were noted between cases and controls in Median 234.40 255.85 family members of the Bacteroidetes, i.e., Bacteroidaceae (35 vs. 43 %), Rickenellaceae (3 vs. 4 %), and Prevotella- Quartiles 159.75–412.90 171.275–450.70 ceae (3.2 vs. 0.7 %). Finally, within the Actinobacteria, Range 14–1067.9 23.2–909.9 Bifidobacteriaceae were lower in the ME/CFS samples sCD14 (ug/mL) (1 vs. 0.5 %). n 39 49 At the OTU level, 40 OTUs were found to be signifi- Median 1.36 1.97 cantly different between groups after multiple testing cor- Quartiles 1.09–1.67 1.71–2.32 rection. The majority of them belonged to the Firmicutes phylum, including members of the Ruminococcaceae Range 0.62–2.42 0.95–3.02 family such as Oscillospira spp. (q = 0.016), Faecalibac- LBP (ug/mL) terium prausnitzii (q = 0.014), and Ruminococcus spp. n 37 49 (q = 0.014) and members of the Lachnospiraceae, i.e., Median 12.32 17.68 Coprococcus spp. (q = 0.014). Other OTUs included Quartiles 10.22–13.73 13.04–22.56 members of the Actinobacteria such as Eggerthella Range 8.65–18.76 7.06–34.52 lenta (q = 0.014) and Collinsella aerofaciens (q = 0.014). These significant differences were further confirmed by LEfSe analysis, which uses linear discriminant analysis (LDA) coupled with effect size measurements to identify metric matrices generated for the sample set. Within the bacterial taxa whose sequences are differentially abundant microbial community cluster, there appears to be no clear between ME/CFS and healthy individuals. In addition to difference in beta-diversity between the ME/CFS group detecting significant features, LEfSe also ranks features by and healthy group using both weighted (Additional file 2: effect size, which put features explaining most of the bio- Figure S2a) and unweighted (Additional file 2: Figure S2b) logical difference at top (Segata et al. 2011). LEfSe identified UniFrac distance matrices. None of the other parameters 24 discriminative features (genus level, LDA score >2) tested, i.e., sex, BMI, or clinical data revealed clustering whose relative abundance varied significantly among fecal (data not shown). Because beta-diversity clustering as mea- samples taken from the ME/CFS and healthy groups (Fig. 5). sured by UniFrac shows how dissimilar overall community ME/CFS microbiota were enriched with an unclassified structure is between samples, the samples may not cluster member of the Desulfohalobacteriaceae and genera from in a manner that reflects differences detected at the OTU the Firmicutes phylum, i.e., Oscillospira, Lactococcus, Anae- level, or the overall alpha diversity within groups. rotruncus and Coprobacillus and Eggerthella, a member of The overall microbial composition for ME/CFS and con- the Actinobacteria phylum (P < 0.05, Fig. 5). Eighteen gen- trols differed at the phylum and family levels (Fig. 4a, b), era were enriched in the control group compared to the although none of these differences were statistically sig- ME/CFS group (Fig. 5) with members mainly belonging to nificant after multiple test correction. The two largest the Firmicutes phylum. We observed that members of phyla represented in each dataset of healthy and ME/CFS- the Ruminococcaeae and Bifidobacteriaceae, i.e., Fae- afflicted individuals were Firmicutes and Bacteroidetes. In calibacterium and Bifidobacterium, respectively, were Giloteaux et al. Microbiome (2016) 4:30 Page 5 of 12 Fig. 2 Correlation between plasma levels of LPS and sCD14 (a), plasma levels of LPS and LBP (b), plasma levels of sCD14 and LBP (c), and plasma levels of hsCRP and sCD14 (d) in the ME/CFS population. Spearman’s rank test was used to determine correlations Fig. 3 Rarefaction curves and confusion matrix. a Rarefaction curves for the microbiota of healthy individuals and ME/CFS patients (each group was rarefied to the number of sequences of the less-sequenced sample, i.e., 32223 sequences). The p value was calculated by the Wilcoxon rank-sum test and b comparison of alpha diversity indexes in ME/CFS and healthy individuals Giloteaux et al. Microbiome (2016) 4:30 Page 6 of 12 Fig. 4 Composition of the gut microbiome of healthy individuals and ME/CFS patients. Relative abundance of phylum-level (a) and family-level (b) gut microbial taxa significantly increased in healthy individuals (P =0.03 with the highest proportion of samples correctly classified and 0.04, respectively). when genus-level taxa along with data from the inflamma- tory markers were used in the analysis. With 97 % ID Classifying subjects into patients vs. controls from OTUs used in the analysis, 82 % of the samples could be inflammatory markers and microbiome data correctly classified (standard deviation of 0.14). With Using a machine learning approach, samples were mostly OTUs collapsed at the species level, the average accuracy successfully classified into healthy and ME/CFS groups, was 0.80 with a standard deviation of 0.11. Collapsing Fig. 5 Histogram of the LDA scores computed for genera differentially abundant between ME/CFS and healthy individuals. ME/CFS-enriched genera are indicated with a positive LDA score, and genera enriched in healthy individuals have a negative score. The LDA score indicates the effect size and ranking of each differentially abundant taxon Giloteaux et al. Microbiome (2016) 4:30 Page 7 of 12 taxonomy to the genus level, individuals with ME/CFS were IgA and IgM to LPS of Gram-negative enterobacteria [19]. classified correctly and separately from the healthy group Our data supports the hypothesis of increased MT in the with an average success rate of 0.82 ± 0.12. The receiver ME/CFS group as evidenced by (i) significantly raised operating curves, the AUC ROC value for the ME/CFS levels of plasma LPS and (ii) significantly higher levels of samples (0.89), and the confusion matrix are presented sCD14 and LBP, as indicators of direct LPS stimulation. in Fig. 6. The feature importance scores for the genus- Increased gut permeability and increased LPS levels have level analysis, which shows the relative importance of been also described in patients with liver diseases, alcoholic, clinical values and microbial abundances, are available and nonalcoholic steatohepatitis [23], during chronic HIV in Additional file 3: Table S1. Additionally, processing infection [24], and in inflammatory bowel disease (IBD) microbial sequencing data without including BMI and [25, 26] suggesting that an activation of pro-inflammatory blood inflammatory marker levels results in 70, 75, and and endotoxin-signaling cascades could be important 72 % classification accuracy for genus, species, and OTU- for disease progression in ME/CFS. Consequently, high level data respectively (confusion matrices available in plasma LPS levels in ME/CFS could result from an in- Additional file 4: Figure S3). creased production of endotoxin upon changes in the gut microbiota. Furthermore, we observed that sCD14 Discussion levels positively correlated with levels of LPS, LBP, and Our analysis of the microbiome in cases suggests that hsCRP. If there is damage to the gut mucosa, microbial gastrointestinal tract of ME/CFS patients is a pro-inflam- translocation could increase, altering antimicrobial regula- matory environment. This environment might cause dam- tors and dysregulating the innate immune system. age to the intestinal epithelium, thus augmenting microbial As a marker, LPS is limited to particular microbes, as translocation (MT) and subsequently triggering an im- it is only present in Gram-negative bacteria. sCD14 is mune response. It has been previously documented that produced primarily by macrophages and hepatocytes in disruption of mucosal barrier function occurs in ME/CFS response to LPS but is also stimulated by other bacterial as demonstrated by the increased serum concentrations of and viral agents [27]. LBP functions as a co-factor along Fig 6 Receiver operating characteristic curves (a) for controls and ME/CFS patients determined using the inflammatory markers and sequencing datasets (even sampled at 32,233 sequences) and a supervised learning approach with randomForest algorithm and (b) confusion matrix for random forest analysis (values are presented as percentage) and ROC area under the curve (AUC) value for 97 % OTUs collapsed at the genus level. Mean AUC ROC value for five times repeated, 10-fold cross validation Giloteaux et al. Microbiome (2016) 4:30 Page 8 of 12 with sCD14 and is constitutively synthesized in hepato- these compounds as electron acceptors [43, 44] to gener- cytes to recognize LPS released to the bloodstream but ate energy and foster their own growth in the gut. We did various inflammatory factors such as IL22, IL-6, and not collect information concerning diet of patients and TNF-α can induce its expression [28, 29]. Nevertheless, thus do not know whether this factor might have affected we found significantly more patients with elevated levels the composition and/or metabolism of the colonic micro- of these biomarkers in comparison to the healthy group, biota in our cohorts. suggesting that more MT occurs in people affected by We observed significantly lower levels of the genus ME/CFS. Faecalibacterium, a member of the Ruminococcaceae in Using both aerobic and anaerobic culturing methods, the ME/CFS population. For example, Faecalibacterium Butt and colleagues were the first to present evidence of prausnitzii, which produces an anti-inflammatory pro- altered fecal microbiota in ME/CFS patients compared tein [45], is reduced in ME/CFS cases relative to con- to healthy individuals [30]. Subsequently, using culture trols. This genus is also depleted in IBD [13, 38] and methods and metabolite analysis, Sheedy et al. [11] ob- ulcerative colitis [46] and has been shown to have anti- tained information concerning the fecal microbiome in inflammatory properties both in vitro and in vivo [37]. patient and health cohorts. Both found that D-lactic Faecalibacterium belongs to a group of producers of acid-producing Enterococcus and Streptococcus species butyrate, a short chain fatty acid known to have anti- were strongly over-represented in ME/CFS patients and inflammatory properties and to protect the intestine that among anaerobic bacteria, Prevotella was a bacterial [40]. Individuals with IBD and IBS [47] exhibit a lack of genus found to be in excess in subjects with ME/CFS. butyrate-producing bacteria and lower levels of butyr- Recently, a study used high-throughput 16S rRNA ate in their gut [48, 49] which modulates different pro- gene sequencing to investigate the presence of specific cesses including hormone and cytokine secretion (e.g., alterations in the gut microbiota of ME/CFS patients leptin, IL-10) and activation of immune/inflammatory from Belgium and Norway [12]. The authors amplified responses [50–52]. the V5 and V6 hypervariable 16S rRNA regions and se- We also found a decrease in Bifidobacterium, previously quenced the amplicons using a Roche FLX 454 sequen- observed in IBS [53–57], IBD [58], and type II diabetes cer, which resulted in an average of only 6000–7000 [59]. Bifidobacteria are a group of lactic acid-producing reads/sample. In contrast, we amplified the V4 hypervar- bacteria that are widely used as probiotics and as tar- iable region of the 16S rRNA gene, sequenced amplicons gets for prebiosis [60]. Treatment with Bifidobacterium using the MiSeq Illumina platform, obtained an average infantis 35624 was reported to reduce CRP levels in a of many more reads/sample (98,000), and compared the cohort of ME/CFS patients [61]. resulting sequences to a different database, the Green- We have employed a supervised machine learning ap- genes non-redundant reference database [31]. Our ana- proach to help prediction of disease state based on the lysis showed that within-sample diversity is lower in the microbiome sequence datasets [62]. Using this approach, ME/CFS specimens compared to controls. The same indi- we were able to classify unlabeled samples with some de- ces in the Fremont et al. [12] study did not differ between gree of accuracy, as demonstrated by the high AUC ROC ME/CFS and healthy subjects [12], likely due to the lower value obtained (0.8928) at the genus level. This method read number they obtained. Lower richness has also been has been recently used in several microbiome surveys to observed in unhealthy or inflammatory states [32, 33] and accurately place individuals into an IBD/healthy category has been associated with IBD, necrotizing enterocolitis [63], including ulcerative colitis (ROC AUC = 0.9225) and [34], and greater abdominal discomfort levels in patients colonic (ROC AUC = 0.8787) or ileal Crohn’s disease with food intolerances [35, 36]. (ROC AUC = 0.9699) [64]. Such an approach could there- Regardless of disease state, bacteria belonging to the fore serve as a complement to other non-invasive diagno- Firmicutes, Bacteroidetes, Proteobacteria, and Actino- ses of symptoms or as an initial diagnosis to determine if bacteria phyla represented the vast majority of sequences the subject likely has ME/CFS. Because this is a relatively identified. We observed reduced levels of members of the small cohort, to move to a formal diagnostic clinical ap- dominant phylum Firmicutes, also noted repeatedly in plication, a large cohort of ME/CFS and healthy con- Crohn’s disease patients [13, 37, 38]. Proteobacteria were trols would be needed to verify that classification would more abundant in ME/CFS patients than in controls, retain its accuracy with independent sample handling observed as well in inflammatory bowel disease (IBD) pa- and sequencing. tients [39, 40]. In an inflamed gut, infiltrating macrophages and neutrophils release sulfur- and nitrogen-derived Conclusions metabolites such as tetrathionate and nitrate [41–44]. Taken together, our results suggest an ongoing damage to Opportunistic members of the Proteobacteria can take the gut mucosa, leading to increased microbial transloca- advantage of the host inflammatory response by using tion in ME/CFS, which in turn could alter antimicrobial Giloteaux et al. Microbiome (2016) 4:30 Page 9 of 12 regulators and disregulate the innate immune system. in plasma samples by commercially available enzyme- Differences between the gut microbiomes of healthy indi- linked immunosorbent assays (ELISA). Plasma sCD14 viduals and patients with ME/CFS were identified in terms was quantified using the Quantikine Human sCD14 of relative abundance of specific genera. There is no single Immunoassay(R&DSystems,Minneapolis,MN),and precise alteration of the gut microbiota in all ME/CFS pa- plasma LBP was measured by LBP soluble ELISA kit tients we examined, but our data converges to support the (Hycult Biotechnology, Uden, The Netherlands) according concept of a less diverse and unstable community of bac- to the manufacturers’ protocols. Plasma bacterial endo- teria in the disorder. It highlights the association of specific toxin, i.e., LPS, was measured from heparinized blood bacterial taxa with ME/CFS, and the identification of the samples (Brandtzaeg) using the Limulus Amebocyte Lysate underlying role of this altered commensal gut microbiota (LAL) assay (Lonza Group Ltd, Allendale, NJ). The method could lead to novel diagnostic and therapeutic strategies uses a chromogenic endpoint assay yielding data as endo- that would improve clinical outcome. Future studies may toxin units (EU/ml). Briefly, 100 μl of each plasma sample also reveal additional molecular markers that could be was diluted in 200 μlof β-G-Blocker (Lonza Group Ltd, combined with gut microbiome information to enhance Allendale, NJ) to eliminate the possibility of false positives. the sensitivity and specificity of ME/CFS diagnostic assays. Samples were further diluted with 100 μl of pyrogen-free The cause of ME/CFS is unknown, but gut dysbiosis water to give a final dilution of 1:4. All dilutions were pre- could be contributing to some of the symptoms and their pared in pyrogen-free tubes. Samples were then placed in a severity. Developing therapeutic interventions aimed at re- water bath at 85 °C for 15 min to inactivate inhibitory ducing local inflammation, restoring gastrointestinal tract plasma proteins. Results of LPS measured were expressed immunity and integrity and modifying the intestinal micro- in picograms per milliliter (1 EU/ml = 100 pg/ml). Levels biome may ameliorate ME/CFS symptoms in a number of of intestinal fatty acid binding protein (I-FABP), a marker affected patients. associated with enterocyte damage, were assayed using an ELISA (Hycult Biotechnology, Uden, the Netherlands) ac- Methods cording to the manufacturer’s instructions. All the samples Human subjects and sample/data collection were run in duplicate. All work involving human subjects was approved by the Cornell University Institutional Review Board. Fecal sam- Statistical analysis of plasma markers levels ples were collected at home by participants in 15-ml con- We initially performed a Shapiro-Wilk test to check if ical tubes containing RNAlater (Life Technologies, Grand the data was normally distributed [66]. In case of viola- Island, NY) and refrigerated prior to shipment. Upon ar- tion of normality, data was log transformed and checked rival at Cornell University, the samples were divided into again for normality. Both parametric independent samples aliquots and stored at −80 °C until processing. Blood t test and a non-parametric Wilcoxon-Mann-Whitney U samples were drawn into EDTA and heparin tubes from test were used to determine the significance of differences an antecubital vein with subjects in the seated position. in each subject group. Values of P < 0.05 were considered Samples were shipped by overnight courier from New statistically significant. All data from the biomarkers levels York City to Cornell University (Ithaca). Upon receipt, determination were processed and analyzed in SPSS samples were centrifuged at 4000 r.p.m. for 30 min to Statistics Version 21 (Armonk, NY: IBM Corp). pellet blood cells, and plasma was stored at −80 °C until further analyses. BMI, age, and gender of subjects were DNA extraction, 16S rRNA gene sequencing recorded. ME/CFS subjects completed the Short Form Metagenomic DNA was isolated from an aliquot of 36 Health Survey (SF-36.org) and Bell’s Disability Scale ~100 mg from each fecal sample using the PowerSoil-htp [65]. Potential controls from the same geographic area DNA isolation kit (MoBio Laboratories Ltd, Carlsbad, CA), as cases were screened by the physician for suitability which involves both chemical and physical lysis of the cells. as healthy controls. As indicated in Table 1, mean and We amplified 16S rRNA genes (V4 hypervariable region) median ages of cases and controls were within 5 years from bulk DNA using the 515F and 806R primers as and the female to male ratios were similar. previously described [67] prior to sequencing. Duplicate PCR reactions of samples and extraction blanks consisted Plasma level determination of hsCRP, sCD14, LBP, LPS, of 2.5X HotMasterMix (5-Prime, Inc., Gaithersburg, MD), and I-FABP 10–100 ng DNA template, and 0.05 μMofeach primer. High-sensitive C-reactive protein (hsCRP) was measured DNA amplification of samples and extraction blanks was from unhemolyzed EDTA plasma using a Chemilumines- performed on a 96-well plate with a minicycler PTC 200 cence immunoassay on an Immulite 2000 (Siemens (MJ Research) starting with 3-min denaturation at 94 °C, Medical Solutions Diagnostics, Deerfield, IL). Markers followed by 25 cycles consisting of denaturation (45 s at for microbial translocation, sCD14 and LBP, were measured 94 °C), annealing (60 s at 50 °C), extension (90 s at 72 °C), Giloteaux et al. Microbiome (2016) 4:30 Page 10 of 12 and a final extension at 72 °C for 10 min. Samples were A machine learning approach was used to identify vari- randomly distributed on the plate with no grouping for ables discriminating the two groups of samples (feature sample type. The replicate PCR reactions were combined selection). For these analyses, we used either 97 % OTUs, and purified using a magnetic bead system (Mag-Bind or taxa abundances based on combining OTUs at the spe- EZPure, Omega Bio-Tek, Norcross, GA). PCR amplicons cies and genus levels. Classification of samples as healthy were quantified using the QuantiT PicoGreen dsDNA controls or ME/CFS was carried out by using a random Assay Kit (Invitrogen, Carlsbad, CA). Aliquots of amplicons forest approach with supervised learning [73] and area (at equal masses) were combined for a final concentration under the curve (AUC) calculation to optimize feature of approximately 15 ng/μl. Extraction blanks showed no (e.g., abundance of a particular genus) selection, imple- amplification. All amplicons were then sequenced on a mented in the software package R. Scripts, required pack- single run using the Illumina MiSeq 2x250 bp platform at ages, and instructions for processing data are available on Cornell Biotechnology Resource Center Genomics Facility. https://gist.github.com/walterst/2222618976a66b3fc8dd. In Quality filtering and analysis of the 16S rRNA gene addition to the taxonomic abundance data, levels of in- sequence data were performed with QIIME 1.9.0 as flammatory markers (BMI, sCD14, LBP, LPS, and I-FABP) previously described [68]. Briefly, matching paired-end were included in the analysis. Average accuracies were raw sequences (mate-pairs) were merged using the fastq- calculated with five repeats of 10-fold cross validation, join command in the ea-utils software package (http:// which is intended to predict the accuracy of the model, code.google.com/p/ea-utils), and merged sequences with and indicate over-fitting if significantly different than less than a 200-bp overlap were filtered out of the dataset. the full dataset results, by subsampling the data and The remaining merged sequences were quality filtered testing this training subsample against the remaining and assigned to samples based on their barcodes using the data reference set. default parameters of QIIME. Sequences were assigned to 97 % ID OTUs by comparing them to a non-redundant Additional files reference database of near-full length sequences [31]. All OTUs that were observed fewer than two times, i.e., sin- Additional file 1: Figure S1. 36-Item Short Form Health Survey (SF-36) gletons, were removed from the analysis. The OTU table profiles from studies reporting SF-36 scores for individuals with a ME/CFS was rarefied to the sequence count of the sample with the diagnosis. (PDF 2514 kb) lowest sequence depth, 32,223 sequences per sample, and Additional file 2: Figure S2. Principal Coordinate Analysis (PCoA) plot of healthy controls versus subjects with ME/CFS. Distances were calculated used in all subsequent analyses. For statistical compari- with weighted UniFrac (a) and unweighted UniFrac (b). Data were evenly sons of healthy individuals to those afflicted with ME/ sampled at 32223 sequences per sample. (PDF 2631 kb) CFS, p values obtained with the Wilcoxon-Mann-Whitney Additional file 3: Table S1. Feature Importance Scores for genus-level U test were corrected for multiple comparisons using the supervised learning. The feature importance score is the percentage increase in error rate when the given feature is permuted while other values remain false discovery rate of Benjamini and Hochberg, imple- constant. As there are only two categories, the increased error rate is equal mented in the QIIME pipeline. We used both the weighted for both categories. (XLSX 42 kb) and unweighted UniFrac distance metrics as measures of Additional file 4: Figure S3. Confusion matrices for random forest between-sample (beta) diversity and applied principal coor- analysis of microbial sequencing data (values are presented as %) and ROC area under the curve (AUC) values at the genus (a), species (b) and OTU (c) dinates analysis (PCoA) to visualize patterns of diversity. level. (PDF 1170 kb) Within-samples (alpha) diversity was calculated using three Additional file 5: Table S2. Per-sample metadata mapping file used different measures (1) ChaoI index [69]; (2) Shannon Index throughout the QIIME pipeline. (XLSX 61 kb) [70]; and (3) Phylogenetic Diversity [71]. LEfSe analysis and machine learning Abbreviations hsCRP, high sensitivity C-reactive protein; IBD, inflammatory bowel disease; Linear discriminant effect size analysis (LEfSe) on filtered IBS, irritable bowel syndrome; I-FABP, intestinal fatty acid binding protein; datasets [72] was performed at the genus level to find fea- LBP, lipopolysaccharide-binding protein; LPS, lipopolysaccharides; ME/CFS, tures (genera) differentially represented between healthy myalgic encephalomyelitis/chronic fatigue syndrome; MT, microbial translocation; sCD14, soluble CD14 and ME/CFS groups. LEfSe combines the standard tests for statistical significance (Kruskal-Wallis test and pair- wise Wilcoxon test) with linear discriminate analysis. It Acknowledgements We thank the subjects for providing samples and information for the study ranks features by effect size, which put features that ex- and Lin Lin for technical assistance. plain most of the biological difference at top. LEfSe ana- lysis was performed under the following conditions: the α Funding value for the factorial Kruskal-Wallis test among classes This work was supported by grant 1R21AI101614 from NIH NIAID to M.R.H. was 0.05 and the threshold on the logarithmic LDA score and R.E.L. The funders had no role in study design, data collection, analysis for discriminative features was 2.0. and interpretation, decision to publish, or preparation of the manuscript. Giloteaux et al. Microbiome (2016) 4:30 Page 11 of 12 Availability of data and materials imbalances in human inflammatory bowel diseases. Proc Natl Acad Sci U S The sequence data supporting the results of this article are available in the A. 2007;104(34):13780–5. European Bioinformatics Institute Sequence Read Archive under accession 14. Qin J, Li Y, Cai Z, Li S, Zhu J, Zhang F, et al. A metagenome-wide number PRJEB13092. The mapping file used for the QIIME pipeline is association study of gut microbiota in type 2 diabetes. Nature. 2012; available on Additional file 5: Table S2. 490(7418):55–60. 15. Komaroff AL, Buchwald DS. Chronic fatigue syndrome: an update. Annu Rev Med. 1998;49:1–13. Authors’ contributions 16. Hornig M, Gottschalk G, Peterson DL, Knox KK, Schultz AF, Eddy ML, et al. LG designed the experiments, processed the samples, conducted the Cytokine network analysis of cerebrospinal fluid in myalgic experiments, and performed the statistical analysis with SPSS. SML recruited, encephalomyelitis/chronic fatigue syndrome. Mol Psychiatry. 2016;21(2):261– diagnosed, and sampled the blood from the subjects. LG and JKG performed the sequence analysis using QIIME. WAW performed the supervised learning 17. Hornig M, Montoya JG, Klimas NG, Levine S, Felsenstein D, Bateman L, et al. machine analysis. RL and MH contributed to study design. LG, JKG, WAW, RL, Distinct plasma immune signatures in ME/CFS are present early in the and MH performed analysis and writing. All authors read and approved the course of illness. Sci Adv. 2015;1(1). final manuscript. 18. Vyboh K, Jenabian MA, Mehraj V, Routy JP. HIV and the gut microbiota, partners in crime: breaking the vicious cycle to unearth new therapeutic Competing interests targets. J Immunol Res. 2015;2015:614127. The authors declare that they have no competing interests. 19. Maes M, Mihaylova I, Leunis JC. Increased serum IgA and IgM against LPS of enterobacteria in chronic fatigue syndrome (CFS): indication for the Ethics approval and consent to participate involvement of gram-negative enterobacteria in the etiology of CFS and for The institutional review board of Cornell University approved the study, and the presence of an increased gut-intestinal permeability. J Affect Disord. patients and controls gave written informed consent after the study protocol 2007;99(1–3):237–40. was fully explained. All consented to blood draw, to provide a stool sample 20. Schumann RR. Function of lipopolysaccharide (LPS)-binding protein (LBP) and to the availability of the stored samples for additional bioassays. and CD14, the receptor for LPS/LBP complexes: a short review. Res Immunol. 1992;143(1):11–5. Author details 21. Pelsers MMAL, Hermens WT, Glatz JFC. Fatty acid-binding proteins as Department of Molecular Biology and Genetics, Cornell University, Ithaca, plasma markers of tissue injury. Clinica Chimica Acta. 2005;352(1–2):15–35. NY, USA. Department of Microbiology, Cornell University, Ithaca, NY, USA. 22. Medicine II. Beyond myalgic encephalomyelitis/chronic fatigue syndrome: Private Practice, New York, NY, USA. redefining an illness. Washington, D.C: The National Academies Press; 2015. 23. Zhu L, Baker SS, Gill C, Liu W, Alkhouri R, Baker RD, et al. Characterization Received: 29 February 2016 Accepted: 11 May 2016 of gut microbiomes in nonalcoholic steatohepatitis (NASH) patients: a connection between endogenous alcohol and NASH. Hepatology. 2013; 57(2):601–9. 24. Openshaw PJ. Crossing barriers: infections of the lung and the gut. Mucosal References Immunol. 2009;2(2):100–2. 1. Fukuda K, Straus SE, Hickie I, Sharpe MC, Dobbins JG, Komaroff A. The 25. Wellmann W, Fink PC, Benner F, Schmidt FW. Endotoxaemia in active chronic fatigue syndrome: a comprehensive approach to its definition and Crohn’s disease. Treatment with whole gut irrigation and 5-aminosalicylic study. International Chronic Fatigue Syndrome Study Group. Ann Intern acid. Gut. 1986;27(7):814–20. Med. 1994;121(12):953–9. 26. Wyatt J, Vogelsang H, Hubl W, Waldhoer T, Lochs H. Intestinal permeability and 2. Riedl A, Schmidtmann M, Stengel A, Goebel M, Wisser AS, Klapp BF, et al. the prediction of relapse in Crohn’s disease. Lancet. 1993;341(8858):1437–9. Somatic comorbidities of irritable bowel syndrome: a systematic analysis. 27. Anas AA, Hovius JW, van 't Veer C, van der Poll T, de Vos AF. Role of CD14 J Psychosom Res. 2008;64(6):573–82. in a mouse model of acute lung inflammation induced by different 3. Whitehead WE, Palsson O, Jones KR. Systematic review of the comorbidity lipopolysaccharide chemotypes. PLoS One. 2010;5(4):e10183. of irritable bowel syndrome with other disorders: what are the causes and 28. Wolk K, Witte E, Hoffmann U, Doecke WD, Endesfelder S, Asadullah K, et al. implications? Gastroenterology. 2002;122(4):1140–56. IL-22 induces lipopolysaccharide-binding protein in hepatocytes: a potential 4. Wojczynski MK, North KE, Pedersen NL, Sullivan PF. Irritable bowel syndrome: systemic role of IL-22 in Crohn’s disease. J Immunol. 2007;178(9):5973–81. a co-twin control analysis. Am J Gastroenterol. 2007;102(10):2220–9. 29. Wan Y, Freeswick PD, Khemlani LS, Kispert PH, Wang SC, Su GL, et al. Role of 5. Aaron LA, Burke MM, Buchwald D. Overlapping conditions among patients lipopolysaccharide (LPS), interleukin-1, interleukin-6, tumor necrosis factor, with chronic fatigue syndrome, fibromyalgia, and temporomandibular and dexamethasone in regulation of LPS-binding protein expression in disorder. Arch Intern Med. 2000;160(2):221–7. normal hepatocytes and hepatocytes from LPS-treated rats. Infect Immun. 6. Komaroff AL, Buchwald D. Symptoms and signs of chronic fatigue 1995;63(7):2435–42. syndrome. Rev Infect Dis. 1991;13 Suppl 1:S8–11. 30. Butt HL, Dunstan R, N.R. M, T. R. Bacterial colonosis in patients with 7. Jones JF, Nisenbaum R, Reeves WC. Medication use by persons with chronic persistent fatigue. In Proceedings of the AHMF international clinical and fatigue syndrome: results of a randomized telephone survey in Wichita, Kansas. scientific conference Sydney, Australia: AHMF; 2001. Health Qual Life Outcomes. 2003;1:74. 31. Brodie EL, Desantis TZ, Joyner DC, Baek SM, Larsen JT, Andersen GL, et al. 8. Borody TJ, Nowak A, Finlayson S. The GI microbiome and its role in chronic Application of a high-density oligonucleotide microarray approach to study fatigue syndrome: a summary of bacteriotherapy. ACNEM J. 2012;31(3):3–8. bacterial population dynamics during uranium reduction and reoxidation. 9. Sullivan A, Nord CE, Evengard B. Effect of supplement with lactic-acid Appl Environ Microbiol. 2006;72(9):6288–98. producing bacteria on fatigue and physical activity in patients with chronic 32. Chang JY, Antonopoulos DA, Kalra A, Tonelli A, Khalife WT, Schmidt TM, fatigue syndrome. Nutr J. 2009;8:4. et al. Decreased diversity of the fecal microbiome in recurrent Clostridium 10. Rao AV, Bested AC, Beaulne TM, Katzman MA, Iorio C, Berardi JM, et al. A difficile-associated diarrhea. J Infect Dis. 2008;197(3):435–8. randomized, double-blind, placebo-controlled pilot study of a probiotic in 33. Willing BP, Dicksved J, Halfvarson J, Andersson AF, Lucio M, Zheng Z, et al. emotional symptoms of chronic fatigue syndrome. Gut Pathog. 2009;1(1):6. A pyrosequencing study in twins shows that gastrointestinal microbial 11. Sheedy JR, Wettenhall RE, Scanlon D, Gooley PR, Lewis DP, McGregor N, et profiles vary with inflammatory bowel disease phenotypes. al. Increased d-lactic acid intestinal bacteria in patients with chronic fatigue Gastroenterology. 2010;139(6):1844–54. e1. syndrome. In Vivo. 2009;23(4):621–8. 12. Fremont M, Coomans D, Massart S, De Meirleir K. High-throughput 16S 34. McMurtry VE, Gupta RW, Tran L, Blanchard EE, Penn D, Taylor CM, et al. rRNA gene sequencing reveals alterations of intestinal microbiota in Bacterial diversity and Clostridia abundance decrease with increasing myalgic encephalomyelitis/chronic fatigue syndrome patients. Anaerobe. severity of necrotizing enterocolitis. Microbiome. 2015;3:11. 2013;22:50–6. 35. Fujimoto T, Imaeda H, Takahashi K, Kasumi E, Bamba S, Fujiyama Y, et al. 13. Frank DN, St Amand AL, Feldman RA, Boedeker EC, Harpaz N, Pace NR. Decreased abundance of Faecalibacterium prausnitzii in the gut microbiota Molecular-phylogenetic characterization of microbial community of Crohn’s disease. J Gastroenterol Hepatol. 2013;28(4):613–9. Giloteaux et al. Microbiome (2016) 4:30 Page 12 of 12 36. Hippe B,Remely M,Bartosiewicz N, RiedelM, Nichterl C,Schatz L, etal. 57. Maukonen J, Satokari R, Matto J, Soderlund H, Mattila-Sandholm T, Saarela M. Abundance and diversity of GI microbiota rather than IgG4 levels correlate with Prevalence and temporal stability of selected clostridial groups in irritable abdominal inconvenience and gut permeability in consumers claiming food bowel syndrome in relation to predominant faecal bacteria. J Med Microbiol. intolerances. Endocr Metab Immune Disord Drug Targets. 2014;14(1):67–75. 2006;55(Pt 5):625–33. 37. Sokol H, Pigneur B, Watterlot L, Lakhdari O, Bermudez-Humaran LG, 58. Swidsinski A, Ladhoff A, Pernthaler A, Swidsinski S, Loening-Baucke V, Ortner M, Gratadoux JJ, et al. Faecalibacterium prausnitzii is an anti-inflammatory et al. Mucosal flora in inflammatory bowel disease. Gastroenterology. 2002; commensal bacterium identified by gut microbiota analysis of Crohn 122(1):44–54. disease patients. Proc Natl Acad Sci U S A. 2008;105(43):16731–6. 59. Wu X, Ma C, Han L, Nawaz M, Gao F, Zhang X, et al. Molecular characterisation of the faecal microbiota in patients with type II diabetes. Curr Microbiol. 2010; 38. Martinez-Medina M, Aldeguer X, Gonzalez-Huix F, Acero D, Garcia-Gil LJ. 61(1):69–78. Abnormal microbiota composition in the ileocolonic mucosa of Crohn’s 60. Furrie E, Macfarlane S, Kennedy A, Cummings JH, Walsh SV, O'Neil DA, et al. disease patients as revealed by polymerase chain reaction-denaturing Synbiotic therapy (Bifidobacterium longum/Synergy 1) initiates resolution of gradient gel electrophoresis. Inflamm Bowel Dis. 2006;12(12):1136–45. inflammation in patients with active ulcerative colitis: a randomised 39. Sartor RB. Key questions to guide a better understanding of host- controlled pilot trial. Gut. 2005;54(2):242–9. commensal microbiota interactions in intestinal inflammation. Mucosal 61. Groeger D, O'Mahony L, Murphy EF, Bourke JF, Dinan TG, Kiely B, et al. Immunol. 2011;4(2):127–32. Bifidobacterium infantis 35624 modulates host inflammatory processes 40. Sartor RB. Genetics and environmental interactions shape the intestinal beyond the gut. Gut Microbes. 2013;4(4):325–39. microbiome to promote inflammatory bowel disease versus mucosal 62. Knights D, Costello EK, Knight R. Supervised classification of human homeostasis. Gastroenterology. 2010;139(6):1816–9. microbiota. FEMS Microbiol Rev. 2011;35(2):343–59. 41. Davies CA, Rocks SA, O'Shaughnessy MC, Perrett D, Winyard PG. Analysis of 63. Papa E, Docktor M, Smillie C, Weber S, Preheim SP, Gevers D, et al. Non-invasive nitrite and nitrate in the study of inflammation. Methods Mol Biol. 2003;225: mapping of the gastrointestinal microbiota identifies children with inflammatory 305–20. bowel disease. PLoS One. 2012;7(6):e39242. 42. Thiennimitr P, Winter SE, Winter MG, Xavier MN, Tolstikov V, Huseby DL, et al. 64. Walters WA, Xu Z, Knight R. Meta-analyses of human gut microbes Intestinal inflammation allows Salmonella to use ethanolamine to compete associated with obesity and IBD. FEBS Lett. 2014;588(22):4223–33. with the microbiota. Proc Natl Acad Sci U S A. 2011;108(42):17480–5. 65. Bell DS. The doctor’s guide to chronic fatigue syndrome. Reading, Mass: 43. Winter SE, Lopez CA, Baumler AJ. The dynamics of gut-associated microbial Addison-Wesley; 1995. communities during inflammation. EMBO Rep. 2013;14(4):319–27. 66. Shapiro SS, Wilk MB. An analysis of variance test for normality (complete 44. Winter SE, Winter MG, Xavier MN, Thiennimitr P, Poon V, Keestra AM, et al. samples). Biometrika. 1965;52(3–4):591–611. Host-derived nitrate boosts growth of E. coli in the inflamed gut. Science. 67. Caporaso JG, Lauber CL, Walters WA, Berg-Lyons D, Lozupone CA, Turnbaugh PJ, 2013;339(6120):708–11. et al. Global patterns of 16S rRNA diversity at a depth of millions of sequences 45. Quevrain E, Maubert MA, Michon C, Chain F, Marquant R, Tailhades J, et al. per sample. Proc Natl Acad Sci U S A. 2011;108 Suppl 1:4516–22. Identification of an anti-inflammatory protein from Faecalibacterium 68. Goodrich JK, Waters JL, Poole AC, Sutter JL, Koren O, Blekhman R, et al. prausnitzii, a commensal bacterium deficient in Crohn’s disease. Gut. 2016; Human genetics shape the gut microbiome. Cell. 2014;159(4):789–99. 65(3):415–25. 69. Chao A. Non-parametric estimation of the number of classes in a 46. Machiels K, Joossens M, Sabino J, De Preter V, Arijs I, Eeckhaut V, et al. A population. Scandinavian Journal of Statistics. 1984;11:265–70. decrease of the butyrate-producing species Roseburia hominis and 70. Shannon CE. A mathematical theory of communication. Bell SystTech J. Faecalibacterium prausnitzii defines dysbiosis in patients with ulcerative 1948;27:379–423. colitis. Gut. 2014;63(8):1275–83. 71. Faith DP. Conservation evaluation and phylogenetic diversity. Biol Conserv. 47. Pozuelo M, Panda S, Santiago A, Mendez S, Accarino A, Santos J, et al. 1992;61:1–10. Reduction of butyrate- and methane-producing microorganisms in patients 72. Segata N, Izard J, Waldron L, Gevers D, Miropolsky L, Garrett WS, et al. with irritable bowel syndrome. Sci Rep. 2015;5:12693. Metagenomic biomarker discovery and explanation. Genome Biol. 2011; 48. Varela E, Manichanh C, Gallart M, Torrejon A, Borruel N, Casellas F, et al. 12(6):R60. Colonisation by Faecalibacterium prausnitzii and maintenance of clinical 73. Liaw A, Wiener M. Classification and regression by randomForest. R News. remission in patients with ulcerative colitis. Aliment Pharmacol Ther. 2013; 2002;2:18–22. 38(2):151–61. 49. Sokol H, Seksik P, Furet JP, Firmesse O, Nion-Larmurier I, Beaugerie L, et al. Low counts of Faecalibacterium prausnitzii in colitis microbiota. Inflamm Bowel Dis. 2009;15(8):1183–9. 50. Saemann MD, Bohmig GA, Osterreicher CH, Burtscher H, Parolini O, Diakos C, et al. Anti-inflammatory effects of sodium butyrate on human monocytes: potent inhibition of IL-12 and up-regulation of IL-10 production. FASEB J. 2000; 14(15):2380–2. 51. Segain JP, Raingeard de la Bletiere D, Bourreille A, Leray V, Gervois N, Rosales C, et al. Butyrate inhibits inflammatory responses through NFkappaB inhibition: implications for Crohn’s disease. Gut. 2000;47(3):397–403. 52. Vinolo MA, Rodrigues HG, Nachbar RT, Curi R. Regulation of inflammation by short chain fatty acids. Nutrients. 2011;3(10):858–76. 53. Kassinen A, Krogius-Kurikka L, Makivuokko H, Rinttila T, Paulin L, Corander J, et Submit your next manuscript to BioMed Central al. The fecal microbiota of irritable bowel syndrome patients differs significantly from that of healthy subjects. Gastroenterology. 2007;133(1):24–33. and we will help you at every step: 54. Tana C, Umesaki Y, Imaoka A, Handa T, Kanazawa M, Fukudo S. Altered • We accept pre-submission inquiries profiles of intestinal microbiota and organic acids may be the origin of symptoms in irritable bowel syndrome. Neurogastroenterol Motil. 2010; � Our selector tool helps you to find the most relevant journal 22(5):512–9. e114-5. � We provide round the clock customer support 55. Malinen E, Rinttila T, Kajander K, Matto J, Kassinen A, Krogius L, et al. � Convenient online submission Analysis of the fecal microbiota of irritable bowel syndrome patients and healthy controls with real-time PCR. Am J Gastroenterol. 2005;100(2):373–82. � Thorough peer review 56. Matto J, Maunuksela L, Kajander K, Palva A, Korpela R, Kassinen A, et al. � Inclusion in PubMed and all major indexing services Composition and temporal stability of gastrointestinal microbiota in irritable � Maximum visibility for your research bowel syndrome—a longitudinal study in IBS and control subjects. FEMS Immunol Med Microbiol. 2005;43(2):213–22. Submit your manuscript at www.biomedcentral.com/submit

Journal

Microbiome – Springer Journals

Published: Jun 23, 2016

There are no references for this article.