An expansion of rare lineage intestinal microbes characterizes rheumatoid arthritis
- 8.2k Downloads
The adaptive immune response in rheumatoid arthritis (RA) is influenced by an interaction between host genetics and environment, particularly the host microbiome. Association of the gut microbiota with various diseases has been reported, though the specific components of the microbiota that affect the host response leading to disease remain unknown. However, there is limited information on the role of gut microbiota in RA. In this study we aimed to define a microbial and metabolite profile that could predict disease status. In addition, we aimed to generate a humanized model of arthritis to confirm the RA-associated microbe.
To identify an RA biomarker profile, the 16S ribosomal DNA of fecal samples from RA patients, first-degree relatives (to rule out environment/background as confounding factors), and random healthy non-RA controls were sequenced. Analysis of metabolites and their association with specific taxa was performed to investigate a potential mechanistic link. The role of an RA-associated microbe was confirmed using a human epithelial cell line and a humanized mouse model of arthritis.
Patients with RA exhibited decreased gut microbial diversity compared with controls, which correlated with disease duration and autoantibody levels. A taxon-level analysis suggested an expansion of rare taxa, Actinobacteria, with a decrease in abundant taxa in patients with RA compared with controls. Prediction models based on the random forests algorithm suggested that three genera, Collinsella, Eggerthella, and Faecalibacterium, segregated with RA. The abundance of Collinsella correlated strongly with high levels of alpha-aminoadipic acid and asparagine as well as production of the proinflammatory cytokine IL-17A. A role for Collinsella in altering gut permeability and disease severity was confirmed in experimental arthritis.
These observations suggest dysbiosis in RA patients resulting from the abundance of certain rare bacterial lineages. A correlation between the intestinal microbiota and metabolic signatures could determine a predictive profile for disease causation and progression.
KeywordsRheumatoid Arthritis Rheumatoid Arthritis Patient Random Forest Humanize Mouse Rheumatoid Arthritis Pathogenesis
body mass index
type II collagen
Kyoto Encyclopedia of Genes and Genomes
linear discriminant analysis
operational taxonomic unit
polymerase chain reaction
Permutational Multivariate Analysis of Variance
Rheumatoid arthritis (RA) is a systemic autoimmune disease characterized by inflammation of the synovial joints. Genome-wide association studies have shown that genetic factors contribute to RA susceptibility, with genes in the major histocompatibility complex (MHC) providing the strongest association and other genetic factors providing additional risk . However, low concordance of RA in monozygotic twins indicates involvement of other factors —perhaps an interaction between genetic and environmental factors—in the development of RA . An infectious etiology of RA has been proposed for decades, although conclusive evidence is lacking .
During the past decade, our understanding of the interaction between microbes and host has evolved from a passive commensal relationship to recognition that the gut microbiota is essential for maintaining immune homeostasis [5, 6]. Recent studies suggest that the aberrant immune response in RA may be associated with dysbiosis of the gut microbiota [7, 8, 9, 10]. Alterations of the normal gut microbiome can affect mucosal immunity with a consequent effect on extra-intestinal diseases like RA [8, 9], diabetes, and obesity [11, 12]. Differences in the abundance of certain commonly present gut commensals between RA patients and those with other rheumatologic diseases, as well as with healthy controls (HCs), suggests the gut microbiota has a possible association with RA [7, 9, 10]. A role for the gut microbiota in RA pathogenesis is further supported by the success of antibiotic treatment in some RA patients .
There is a growing realization that gut microbes and their products may affect the adaptive immune response. Introduction of segmented filamentous bacteria restores the presence of TH17 cells and contributes to the onset of arthritis in germ-free mice . Mice carrying RA-susceptible human leukocyte antigen (HLA) genes show a loss of sex- and age-dependent changes in the gut microbiota that is associated with a proinflammatory cytokine profile in the gut compared with mice with RA-resistant genes . The Human Microbiome Project and other studies have documented the diversity of the gut microbiome in healthy individuals and gut-related diseases [16, 17, 18, 19].
Association of the gut microbiota with various diseases has been reported, though the specific components of the microbiota that affect the host response leading to disease remain unknown. In this study, we show not only an association between RA and certain genera but also that the role of microbes in the pathogenesis of RA is biologically plausible. Our observations suggest that RA is characterized by an expansion of certain intestinal microbes that are present in low abundance in non-RA healthy individuals.
Characteristics of the study population
RA (n = 40)
Control (n = 32)
Age, years, mean (median)
BMI, mean (median), IQR
30.4 (30.1), 23.1–33.1
30.9 (30.6), 23.9–32.4
Disease activity parameters
Disease duration, months, mean (median)
DAS28, mean (median), IQR
3.2 (2.6), 0–4.3
HAQ, mean (median), IQR
0.6 (0.5), 0.1–1.1
ESR, mm/h, mean (median), IQR
18.5 (11), 5–23
CRP, mg/l, mean (median), IQR
12.7 (3.6), 3–8.6
Patient VAS pain, mm, mean (median), IQR
37.3 (28.0), 12.8–49.5
TJC-28, mean (median), IQR
4.2 (1), 0–6
SJC-28, mean (median), IQR
4.3 (2), 0–8
RF titer, kU/l, mean (median), IQR
113 (63), 15–164
ACPA titer, kAU/l, mean (median), IQR
110 (99), 16–250
At the time of enrollment, any household first-degree relatives (FDRs; n = 15) who consented and did not have any symptoms of inflammatory arthritis or other autoimmune diseases were also enrolled. Other controls (n = 17) included sex- and age-matched healthy individuals with no known history of autoimmune diseases. For convenience, FDRs + HCs are labeled as controls in the figures. Any patient or control on antibiotics, consuming probiotics, or having a known history of inflammatory bowel disease or other autoimmune diseases like diabetes and multiple sclerosis were excluded. All human studies were approved by the Institutional Review Board of Mayo Clinic. Written informed consent was received from all participants prior to inclusion in the study.
Sample collection, 16S sequencing, metabolomics, and bioinformatics processing
Fecal samples were frozen within 24 h of their receipt. Microbial DNA was extracted from fecal samples using the MoBio PowerSoil Kit with a bead-beating step. A polymerase chain reaction (PCR) was performed using 50 ng cDNA and 0.3 μM V3-V5 barcoded primers targeting 357 F and 926R with Kapa HiFi Hotstart Ready Mix (Kapa Biosystems). Samples were pooled to equal concentrations, then sequenced on one lane of MiSeq at the Mayo Genomics Facility using the MiSeq Reagent Kit v2 (500 cycles; Illumina Inc.), generating 20 M 2x250 reads. Pre-processed sequence files were then processed by IM-TORNADO .
Plasma samples were used for determining metabolites by mass spectrometry coupled with liquid chromatography in Mayo Metabolomics Core facility. These data were only available for patients with RA and FDRs. Methods for analysis of microbiome and metabolomics data are detailed in Additional file 1: Statistical analyses.
Staining for tight junction proteins
The human intestinal epithelial cell line CACO-2 (ATCC) was grown in vitro as per recommendations. Expression of the tight junction protein ZO-1 was measured by immunofluorescence using a purified anti-ZO-1 antibody (Life Technologies) as the primary antibody and fluorescein isothiocyanate (FITC)-conjugated anti-rabbit IgG (Jackson ImmunoResearch Laboratories) as the secondary antibody. Expression of ZO-1 was observed using confocal microscopy (Leica DM2500, LAS-AF) and the mean florescence intensity of ZO-1 expression was calculated using image J software.
Collagen-induced arthritis and treatment with Collinsella
Animal care and experiments were conducted in accordance with the institutional guidelines and after approval from the institutional animal care and use committee. The HLA-DQ8.AEo mice used in this study have been characterized and the collagen-induced arthritis model in the HLA-DQ8 transgenic mice has been described previously [21, 22]. Arthritis was induced in DQ8 mice (n = 18) and, 2 weeks later, mice (n = 10) were treated with Collinsella (109 bacteria suspended in 100 μl tryptic soy broth (TSB), ATCC25986 strain VPI 1003, cultured as per instructions) or with media every alternate day for 4 weeks during which time the onset and progression of arthritis was monitored. The arthritic severity of the mice was evaluated with a grading system of 0–3 for each paw as described previously . The mean arthritic score was determined using arthritic animals only.
To evaluate the T-cell response to Collinsella-primed dendritic cells (DCs), 10 days post-immunization, splenic CD4 T cells sorted from CII-primed DQ8 mice (200 μg of CII emulsified 1:1 in complete Freund's adjuvant (CFA) were cultured in vitro in the presence or absence of CII (50 μg/ml) and DCs (pre-cultured with bacteria or supernatant of the bacterial culture). T-cell proliferation was measured by routine 3H-thymidine incorporation . All experiments were done two to three times for reproducibility.
As gut permeability may be diet-dependent, all transgenic mice were kept on a standard diet. Changes in intestinal permeability were determined using 4-KDa FITC-labeled dextran. Mice were deprived of food for 3 h, then gavaged with FITC-labeled dextran (0.6 mg/g body weight). Mice were bled and serum collected 3 h later. FITC-dextran content of the sera was determined by using a microplate reader with an excitation of 490 nm and emission detection at 525 nm as reported previously .
rtPCR for cytokine and chemokine expression
RNA was extracted from CACO-2 cells using RNeasy columns (Qiagen) and cDNA was prepared using the SuperScript III First Strand Synthesis System (Invitrogen). Qiagen PAHS-073A RT2 Profiler PCR Array Human Th17 Response plates were used as per the manufacturer’s instructions. The data were analyzed as per the online resources of the manufacturer from their Data Analysis Center.
Colonization of Collinsella
All of the statistical analyses were performed in R-3.0.2 (R Development Core Teams). Details are given in Additional file 1: Statistical analyses.
Disease duration and seropositivity are associated with decreased microbial diversity
The gut microbiotas of patients with RA differ from those of FDRs and HCs
To determine if RA patients have a dysbiotic gut microbiota, we compared the 16S sequences of RA patients with controls (15 FDRs with no autoimmune disease and 17 randomly enrolled HCs; Table 1; Additional file 1: Figure S3). UniFrac analysis demonstrated that the microbiota of the FDRs was not significantly different from that of HCs (P > 0.1), and the average distance between FDRs and HCs was smaller than that between FDRs and RA patients (Additional file 1: Figure S4), indicating that disease status had larger effects than genetic and environmental factors. No significant correlation of the microbiota between FDRs and RA patients (P = 0.40) was observed. We thus pooled FDRs and HCs as a single control group to improve statistical power and identify consistent change.
Expansion of rare microbial lineages characterizes the RA gut microbiota
We applied PICRUSt  to infer the functional content of the microbiota. Among 26 KEGG (Kyoto Encyclopedia of Genes and Genomes) pathways tested (Additional file 2: Table S4), the amino acid metabolism pathway exhibited differences between RA patients and controls; specifically, a decrease in OTUs with amino acid metabolism capabilities was measured in RA patients compared with controls (unadjusted P = 0.03; Additional file 1: Figure S8).
Predictive modeling of the gut microbial profile for RA
The metabolome is associated with the intestinal microbiota in patients with RA
Collinsella enhances disease severity in humanized mice
Collinsella increases gut permeability by reducing the expression of tight junction protein in epithelial cells and induces expression of IL-17 network cytokines
For applicability to RA, a human intestinal epithelial cell line, CACO-2, was cultured in the presence or absence of Collinsella (Fig. 6d). Escherichia coli was used as a control. Our observations suggested a significant decrease in the expression of the tight junction protein ZO-1 in the presence of Collinsella as determined by staining and mean florescence intensity (Fig. 6d, e) while E. coli did not show a significant difference compared with the media control (Additional file 1: Figure S12). To determine if Collinsella lowers gut permeability, we compared gut permeability before and after administering media (n = 10) or Collinsella (n = 10) for 3 weeks. Collinsella administration led to a significant increase in gut permeability compared with pre-treatment (P < 0.05; Fig. 6f). No change in gut permeability was observed with E. coli (Additional file 1: Figure S12). We also determined if Collinsella induced mRNA expression of Th17 regulatory network cytokines in CACO-2 cells (n = 3; Fig. 6g). Compared with controls, culturing with Collinsella led to more than a twofold increase in the expression of interleukin (IL)-17A as well as RORα and chemokines CXCL1 and CXCL5, which are known to regulate production of IL-17 [30, 31]. Also, an increase in NFkB1 expression suggested activation of inflammatory pathways. These data suggest that an expansion of Collinsella may cause an increase in pro-inflammatory conditions with a loss of gut epithelial integrity.
Accumulating evidence suggests that RA is a multifactorial disease dependent on an interaction between genetic and environmental factors [32, 33]. The HLA-DRB1 “shared epitope” provides the highest genetic risk factor for RA patients . Environmental factors affecting mucosal surfaces (smoking and infections) have the most influence on RA pathogenesis .
There is growing speculation about the role of the gut in systemic autoimmune diseases. Humanized mice expressing an RA-susceptible HLA gene exhibited a dysbiotic fecal microbiome compared with mice carrying an RA-resistant gene . A recent study detected an abundance of the bacteria P. copri in fecal samples of patients with new-onset RA, suggesting a role for the gut microbiome in RA . Interestingly, an inverse relationship between the presence of P. copri and the presence of a shared epitope was observed, suggesting that this bacterium may contribute to pathogenesis in a subset of patients. We did not observe a difference in either the abundance of Prevotellaceae or P. copri or their associated OTUs between RA patients and controls (Additional file 1: Figure S7). In contrast to the previous study, however, all the patients in the present study were currently on a treatment regimen. Association of disease severity measures with the gut microbiota of patients showed rheumatoid factor levels and disease duration to be associated with the decreased species richness after adjusting for various drugs used for treatment. Together, the present and previous data suggest that cohorts with different patient characteristics, including disease stage (i.e., early versus established), activity, and geographical locations, may show different microbial associations. Microbial metabolites may provide a window to the functioning of the microbiota and assume universal importance.
Autoreactive T-cell responses and auto-antibody production leading up to the onset of RA occur much earlier than the clinical presentation of RA . Since random HCs possess different genetic factors to the RA patients, we enrolled FDRs as additional controls so the major effects observed would be driven by environmental factors or due directly to the disease process. This would help in elucidating the expansion or contraction of specific bacterial clades in RA patients. Interestingly, FDRs did not differ significantly from HCs in their fecal microbiota, suggesting that differences in certain taxa, such as those observed in the current study, may be dependent on disease state or factors other than genetics, although genetic factors may contribute to an altered state of the immune response. However, the sample size for FDRs was small, which may have limited the power of analysis.
Our data suggest that the differential microbial community structure between RA patients and controls was driven by differences in taxa, mainly the presence and abundance of rare and less abundant lineages. The predictive model suggested that microbes belonging to the phylum Actinobacteria play a significant role in RA pathogenesis as both Collinsella and Eggerthella were observed to predict the RA status. The abundances of Eggerthella and Collinsella were not significantly associated with the three commonly used drugs, methotrexate (MTX), prednisone, and hydroxychloroquine. These observations confirm a recent study that showed that dysbiosis in the gut microbiome in RA patients is restored partially after treatment with MTX . The abundance of Faecalibacterium showed a significant positive association with the use of hydroxychloroquine (P < 0.05), which does not explain the reduced Faecalibacterium in RA. Overall, these observations suggest that treatments are probably not the confounding factor for the identified associations. An important role for Collinsella was confirmed both in vivo using a humanized mouse model of arthritis and in vitro using human intestinal epithelial cells. A recent study showed shared sequences between Collinsella and DRB1*0401, suggesting that Collinsella might contribute to RA via molecular mimicry , further supporting the current observations that HLA class II molecules can present self-HLA-derived peptides  and mimicry with a pathogen can result in enhanced stimulation and inflammation in certain conditions.
One mechanism by which Collinsella contributes to disease pathogenesis is by increasing gut permeability as observed by the lower expression of tight junction proteins. Additionally, Collinsella influences the epithelial production of IL-17A and the chemokines CXCL1 and CXCL5, which may result in recruitment of neutrophils and activation of NFkB, which has been observed to be involved in the pathologic effects of a gut pathobiont . Recently, a multifactorial role of neutrophils has been suggested in RA . CXCL5 production by epithelial cells in response to Bacteroides fragilis is associated with an inflammatory response . Both CXCL1 and CXCL5 are increased in arthritis . IL-17A, a major cytokine involved in arthritis, upregulates CXCL1, which is known to cause increased cell migration, angiogenesis, and activation of the STAT-3 pathway . Induction of Th17 cytokines systemically by Collinsella would be informative about its role in arthritis but was not investigated in this study. Our data suggest that Collinsella contributes to hyper-permeability of the gut by reducing the expression of the tight junction protein ZO-1 directly, as well as by producing specific metabolites. In support of this, the abundance of Collinsella correlated strongly with high levels of beta-alanine, alpha-aminoadipic acid, and asparagine. Alpha-aminoadipic acid is a marker for autoimmunity and age-associated changes in human collagen [41, 42], while asparagine is a non-essential amino acid involved in the tricarboxylic acid cycle and blocking apoptosis . Currently, the source of asparagine is unknown in this study. Age-associated changes in collagen and blocking of apoptosis could be involved in the autoreactive response to collagen in patients, though these mechanisms need to be proved.
Eggerthella lenta is another organism that was detected with more abundance in RA patients, using multiple methods of analysis, but only rarely in controls. Eggerthella uses ornithine as substrate to generate energy, producing citrulline and carbamyol phosphate as byproducts. We did not observe any association between the presence of Eggerthella and citrulline levels in the sera of patients. However, it is unknown whether RA patients carry higher loads of this amino acid or citrullinated peptides in the gut. Based on the higher abundance of Eggerthella in patients in the present study, we predict that patients with RA may exhibit increased levels of citrulline in the gut available for citrullination, against which antibodies might be produced. Carbamyol phosphate is an enzyme that is involved in the pyrimidine pathway. This pathway is upregulated in RA patients and typically treated with pyrimidine synthesis inhibitors such as leflunomide . None of the patients in our study were on leflunomide. While these data provide tantalizing clues, the roles of these metabolites and rare taxa of the gut microbiome deserve further study.
The gut microbiota of RA patients exhibited decreased diversity with increased disease duration and seropositivity. This change in diversity stemmed from an expansion of rare lineages like Eggerthella and from a contraction of the common beneficial genera like Faecalibacterium. Faecalibacterium is one of the most abundant Firmicutes in the human gut that produces butyrate . Butyrate is required for epithelial proliferation and mucin synthesis and production, which helps maintain the integrity of the gut epithelial layer. A decreased abundance of Faecalibacterium with increased Collinsella may lead to an increase in epithelial permeability, causing microbial fragments and products to enter the sub-epithelial space and lamina propria. In the presence of these conditions, a change in abundance of any microbial clade that leads to an altered immune state may cause local inflammation in the gut as well as outside the gut. The Boruta feature selection algorithm and LEfse analysis also confirmed the significance of the differential presence of Eggerthella, Collinsella, and Faecalibacterium in RA patients compared with controls.
Elevated BMI has a significant impact on the gut microbiota of RA patients in this study. The distribution of BMI was not significantly different between the patients and controls, suggesting obesity is unlikely to have a major confounding effect on the differences in the gut microbiota between patients. Obesity and elevated BMI are associated with both the incidence [46, 47] and prognosis of RA . Therefore, restricting the sample of patients to those with healthy weight might have inappropriately limited the generalizability of our findings, though we cannot exclude the possibility of confounding by BMI in this study. Future prospective longitudinal studies are warranted to dissect the potential interactions of obesity and gut microbiota on the pathophysiology of RA.
One can envisage that in a healthy state, dynamic microbiome structures based on sex, diet, and other factors and driven by specific bacterial groups, maintain homeostasis that modulates the immune response. In contrast, this kind of microbial axis dynamism is lost in patients. Although specific molecular mechanisms remain largely unexplored, the results of this study suggest that susceptibility to RA could be triggered by gut dysbiosis and alterations in pathways in which rare lineages are involved. However, the study needs to be confirmed with a larger patient and FDR cohort. An interesting observation was the loss of sex-biased differences in RA patients, as a healthy human microbiota is sex-dependent . In the present study, there were not enough males in the HC group to perform an analysis of sufficient power to evaluate this factor. Our studies support previous data that showed a loss of sex bias in the fecal microbiota of genetically arthritis-susceptible humanized mice . Further, similar to the humanized mice, expansion of certain taxa was observed in RA patients.
Collectively, our data demonstrate that a dysbiotic gut microbiota in RA patients, characterized by a decrease in Faecalibacterium and expansion of C. aerofaciens and E. lenta, could trigger inflammatory conditions in the gut that depend on the production of chemokines and IL-17A and compromise the gut epithelium integrity. It is possible that the inflammatory conditions can be modulated by prebiotics or probiotics. The therapeutic potential of the only probiotics, lactobacilli, used as treatment for RA is inconclusive, with some studies in favor of their use while others did not show significant improvement with lactobacilli using the American College of Rheumatology (ACR) response criteria for RA [50, 51, 52, 53]. Our data suggest specific microbial clades may be viable targets for therapeutic manipulation by diet, probiotics, prebiotics, and/or beneficial gut commensals. Determining the functions of the microbial clades that expand or contract in RA will assist in developing effective means to target them.
All human studies were approved by the Institutional Review Board of Mayo Clinic and conducted in accordance with the Helsinki Declaration. Written informed consent was received from all participants prior to inclusion in the study. Animal care and experiments were conducted in accordance with and after approval from the Institutional animal care and use committee.
Availability of data and materials
Data can be accessed via BioProject PRJNA317370.
The authors thank Dr. Chella David for the transgenic mice and Julie Hanson for the maintenance of animals, and study coordinators Sharlene Allen and Kimberly Timm for enrolling patients and controls.
The work was supported by Mark E. and Mary A. Davis to Mayo Clinic, Mayo Clinic Center of Individualized Medicine and, in part, with funds from the National Institutes of Health, AR30752 to VT.
- 1.Raychaudhuri S, Sandor C, Stahl EA, Freudenberg J, Lee HS, Jia X, Alfredsson L, Padyukov L, Klareskog L, Worthington J, et al. Five amino acids in three HLA proteins explain most of the association between MHC and seropositive rheumatoid arthritis. Nat Genet. 2012;44:291–6.Google Scholar
- 2.Silman AJ, MacGregor AJ, Thomson W, Holligan S, Carthy D, Farhan A, Ollier WE. Twin concordance rates for rheumatoid arthritis: results from a nationwide study. Br J Rheumatol. 1993;32:903–7.Google Scholar
- 8.Zhang X, Zhang D, Jia H, Feng Q, Wang D, Liang D, Wu X, Li J, Tang L, Li Y, et al. The oral and gut microbiomes are perturbed in rheumatoid arthritis and partly normalized after treatment. Nat Med. 2015;21:895–905.Google Scholar
- 9.Scher JU, Sczesnak A, Longman RS, Segata N, Ubeda C, Bielski C, Rostron T, Cerundolo V, Pamer EG, Abramson SB, et al. Expansion of intestinal Prevotella copri correlates with enhanced susceptibility to arthritis. eLife. 2013;2:e01202.Google Scholar
- 12.Turnbaugh PJ, Hamady M, Yatsunenko T, Cantarel BL, Duncan A, Ley RE, Sogin ML, Jones WJ, Roe BA, Affourtit JP, et al. A core gut microbiome in obese and lean twins. Nature. 2009;457:480–4.Google Scholar
- 14.Wu HJ, Ivanov II, Darce J, Hattori K, Shima T, Umesaki Y, Littman DR, Benoist C, Mathis D. Gut-residing segmented filamentous bacteria drive autoimmune arthritis via T helper 17 cells. Immunity. 2010;32:815–27.Google Scholar
- 15.Gomez A, Luckey D, Yeoman CJ, Marietta EV, Berg Miller ME, Murray JA, White BA, Taneja V. Loss of sex and age driven differences in the gut microbiome characterize arthritis-susceptible 0401 mice but not arthritis-resistant 0402 mice. PLoS One. 2012;7:e36095.Google Scholar
- 16.Morgan XC, Tickle TL, Sokol H, Gevers D, Devaney KL, Ward DV, Reyes JA, Shah SA, LeLeiko N, Snapper SB, et al. Dysfunction of the intestinal microbiome in inflammatory bowel disease and treatment. Genome Biol. 2012;13:R79.Google Scholar
- 17.Arumugam M, Raes J, Pelletier E, Le Paslier D, Yamada T, Mende DR, Fernandes GR, Tap J, Bruls T, Batto JM, et al. Enterotypes of the human gut microbiome. Nature. 2011;473:174–80.Google Scholar
- 20.Jeraldo P, Kalari K, Chen X, Bhavsar J, Mangalam A, White B, Nelson H, Kocher JP, Chia N. IM-TORNADO: a tool for comparison of 16S reads from paired-end libraries. PLoS One. 2014;9:e114804.Google Scholar
- 25.Chen J, Bittinger K, Charlson ES, Hoffmann C, Lewis J, Wu GD, Collman RG, Bushman FD, Li H. Associating microbiome composition with environmental covariates using generalized UniFrac distances. Bioinformatics. 2012;28:2106–13.Google Scholar
- 26.Segata N, Izard J, Waldron L, Gevers D, Miropolsky L, Garrett WS, Huttenhower C. Metagenomic biomarker discovery and explanation. Genome Biol. 2011;12:R60.Google Scholar
- 27.Langille MG, Zaneveld J, Caporaso JG, McDonald D, Knights D, Reyes JA, Clemente JC, Burkepile DE, Vega Thurber RL, Knight R, et al. Predictive functional profiling of microbial communities using 16S rRNA marker gene sequences. Nat Biotechnol. 2013;31:814–21.Google Scholar
- 28.Breiman L. Random Forests. Machine Learning. 2001;45:5–32.Google Scholar
- 29.Kursa MB, Rudnicki WR. Feature selection with the Boruta package. J Stat Softw. 2010;36:13.Google Scholar
- 30.Nouailles G, Dorhoi A, Koch M, Zerrahn J, Weiner 3rd J, Fae KC, Arrey F, Kuhlmann S, Bandermann S, Loewe D, et al. CXCL5-secreting pulmonary epithelial cells drive destructive neutrophilic inflammation in tuberculosis. J Clin Invest. 2014;124:1268–82.Google Scholar
- 31.Yang XO, Pappu BP, Nurieva R, Akimzhanov A, Kang HS, Chung Y, Ma L, Shah B, Panopoulos AD, Schluns KS, et al. T helper 17 lineage differentiation is programmed by orphan nuclear receptors ROR alpha and ROR gamma. Immunity. 2008;28:29–39.Google Scholar
- 34.Nielen MM, van Schaardenburg D, Reesink HW, van de Stadt RJ, van der Horst-Bruinsma IE, de Koning MH, Habibuw MR, Vandenbroucke JP, Dijkmans BA. Specific autoantibodies precede the symptoms of rheumatoid arthritis: a study of serial measurements in blood donors. Arthritis Rheum. 2004;50:380–6.Google Scholar
- 38.Kim JM, Oh YK, Kim YJ, Oh HB, Cho YJ. Polarized secretion of CXC chemokines by human intestinal epithelial cells in response to Bacteroides fragilis enterotoxin: NF-kappa B plays a major role in the regulation of IL-8 expression. Clin Exp Immunol. 2001;123:421–7.CrossRefPubMedPubMedCentralGoogle Scholar
- 39.Pickens SR, Chamberlain ND, Volin MV, Gonzalez M, Pope RM, Mandelin 2nd AM, Kolls JK, Shahrara S. Anti-CXCL5 therapy ameliorates IL-17-induced arthritis by decreasing joint vascularization. Angiogenesis. 2011;14:443–55.Google Scholar
- 41.Wang TJ, Ngo D, Psychogios N, Dejam A, Larson MG, Vasan RS, Ghorbani A, O'Sullivan J, Cheng S, Rhee EP, et al. 2-Aminoadipic acid is a biomarker for diabetes risk. J Clin Invest. 2013;123:4309–17.Google Scholar
- 43.Zhang J, Fan J, Venneti S, Cross JR, Takagi T, Bhinder B, Djaballah H, Kanai M, Cheng EH, Judkins AR, et al. Asparagine plays a critical role in regulating cellular adaptation to glutamine depletion. Mol Cell. 2014;56:205–18.Google Scholar
- 47.Qin B, Yang M, Fu H, Ma N, Wei T, Tang Q, Hu Z, Liang Y, Yang Z, Zhong R. Body mass index and the risk of rheumatoid arthritis: a systematic review and dose-response meta-analysis. Arthritis Res Ther. 2015;17:86.Google Scholar
- 52.Hatakka K, Martio J, Korpela M, Herranen M, Poussa T, Laasanen T, Saxelin M, Vapaatalo H, Moilanen E, Korpela R. Effects of probiotic therapy on the activity and activation of mild rheumatoid arthritis--a pilot study. Scand J Rheumatol. 2003;32:211–5.Google Scholar
Open AccessThis 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.