Crosstalk between B lymphocytes, microbiota and the intestinal epithelium governs immunity versus metabolism in the gut

Transcription

1 a r t i l e s Crosstalk between B lymphoytes, mirobiota and the intestinal epithelium governs immunity versus metabolism in the gut Natalia Shulzhenko,0,, Andrey Morgun,0,, William Hsiao 2, Mihele Battle, Mihael Yao 4, Oksana Gavrilova 5, Marlene Orandle 6, Lloyd Mayer 7, Andrew J Mapherson 8, Kathy D MCoy 8,9, Claire Fraser-Liggett 2 & Polly Matzinger 20 Nature Ameria, In. All rights reserved. Using a systems biology approah, we disovered and disseted a three-way interation between the immune system, the intestinal epithelium and the mirobiota. We found that, in the absene of B ells, or of IgA, and in the presene of the mirobiota, the intestinal epithelium launhes its own protetive mehanisms, upregulating interferon-induible immune response pathways and simultaneously repressing Gata4-related metaboli funtions. This shift in intestinal funtion leads to lipid malabsorption and dereased deposition of body fat. Network analysis revealed the presene of two interonneted epithelial-ell gene networks, one governing lipid metabolism and another regulating immunity, that were inversely expressed. Gene expression patterns in gut biopsies from individuals with ommon variable immunodefiieny or with HIV infetion and intestinal malabsorption were very similar to those of the B ell defiient mie, providing a possible explanation for a longstanding enigmati assoiation between immunodefiieny and defetive lipid absorption in humans. The mammalian gut is a omplex eosystem with three main interating omponents: the intestinal epithelium with its neuronal onnetions, the gut-assoiated immune tissue and the ommensal mirobiota. These omponents have several bidiretional interations. The mirobiota, for example, are essential for the emergene of T ell subsets and the differentiation of gut B ells into IgA-produing plasma ells 7. Conversely, hosts that lak T and B ells, that make only IgM antibodies or that have defetive innate immune sensors show hanges in intestinal mirobiota 8 0 that sometimes lead to metaboli abnormalities and obesity 8. Between the ommensals and the intestinal epithelium, some dialogues indue the epithelium to produe speifi fuosylated glyans,2, whereas others inrease energy harvest from food. It has been proposed that trialogues may also govern gut metabolism 4, but there has been no diret evidene for this idea. Here we show that a trialogue does indeed exist. A defet in adaptive immunity indiretly influenes the balane between metaboli and immune funtions of the gut epithelium via a three-way onversation between the two host systems and the intestinal mirobiota. Normally, immune protetion in the gut results from a partnership between the immune system (supplying B ells, T ells and innate immune ells) and the epithelium (supplying antimirobial peptides and a muosal layer that hinders baterial invasion 2,5 ). To begin deiphering the immune system s effet on the homeostati funtions of the gut epithelium, we studied global gene expression in the jejunum of B ell defiient mie. In the presene of the mirobiota, the intestinal epithelium in these mie launhed its own defense mehanisms, ativating innate immune genes at the expense of metaboli ones primarily regulated by the transription fator Gata4. This reated a defet in fat absorption resulting in dereased body fat and leptin levels. The moleular features of the malabsorption found in the B ell defiient mie were also present in IgA-defiient mie, humans with ommon variable immunodefiieny (CVID) and humans with HIV infetion. These data support our previous suggestion that tissues take an ative role in their own defense 6,7. When the immune system funtions optimally, the intestinal epithelium an onentrate on its metaboli funtions. However, if the immune system is dysfuntional, the epithelium takes on some of the missing immune funtions at the expense of its metaboli ativity. This is also the first example, to our knowledge, of a trialogue (in mie and humans) in whih the adaptive immune system, the intestine and the mirobiota govern a homeostati metaboli funtion. Ghost Lab, T Cell Tolerane and Memory Setion, Laboratory of Cellular and Moleular Immunology, National Institute of Allergy and Infetious Diseases (NIAID), US National Institutes of Health (NIH), Bethesda, Maryland, USA. 2 Institute for Genome Sienes, University of Maryland Shool of Mediine, Baltimore, Maryland, USA. Department of Cell Biology, Neurobiology & Anatomy, Medial College of Wisonsin, Milwaukee, Wisonsin, USA. 4 Muosal Immunology Setion, Laboratory of Host Defenses, NIAID, NIH, Bethesda, Maryland, USA. 5 Mouse Metabolism Core Laboratory, National Institute of Diabetes and Digestive and Kidney Diseases, NIH, Bethesda, Maryland, USA. 6 Comparative Mediine Branh, NIAID, NIH, Bethesda, Maryland, USA. 7 Immunology Institute, Mount Sinai Medial Center, New York, New York, USA. 8 University of Bern, Bern, Switzerland. 9 Farnombe Family Digestive Health Researh Institute, MMaster University, Hamilton, Ontario, Canada. 0 Present address: College of Pharmay, Oregon State University, Corvallis, Oregon, USA. These authors ontributed equally to this work. Correspondene should be addressed to A.M. (anemorgun@hotmail.om), N.S. (natalia.shulzhenko@oregonstate.edu) or P.M. (pm@helix.nih.gov). Reeived 9 April; aepted 9 September; published online 20 November 20; doi:0.08/nm.2505 nature mediine VOLUME 7 NUMBER 2 DECEMBER

2 A r t i l e s 20 Nature Ameria, In. All rights reserved. RESULTS Gene expression in the gut of B ell defiient mie B ells are among the most prominent populations of immune ells in the small intestine s lamina propria, presumably beause of their role in protetion from pathogens. We started studying their role in intestinal homeostasis by examining gene expression in the jejunum of B ell defiient () mie. To exlude effets of partiular mutations or of unique bakground genes, we inluded mie arrying two different mutations preventing B ell development (µmt: arrying a deletion in the transmembrane domain of the IgM heavy hain; and JhKO: arrying a deletion in the J segment of the immunoglobulin heavy hain lous) on two different strain bakgrounds (B0.A and BALB/). To identify robust gene profiles, we used large sample sizes (27 mie per group, see Supplementary Fig. a) and ompared homozygous defiient mie to heterozygous littermates and also to wild-type (WT) non-littermates (Fig. a). After we exluded genes expressed in B ells, about 280 genes (298 geneprobes) showed expression differenes between both types of mie and their respetive ontrols (Fig. b, Supplementary Table and Supplementary Fig. b). Here we refer to this set of genes as the profile. By Gene Ontology enrihment analysis, most of the genes upregulated in B ell a defiient intestines were involved in immunity, inluding genes related to defense, inflammatory and interferon-induible responses (Fig. and Supplementary Fig. 2a). These results suggest that loss of one part of the immune system (B ells) leads to ompensatory hanges in other parts. The downregulated genes were mostly involved in various metaboli proesses (Fig. d), predominantly oxidation and redution (eletron transport, energy generation) and lipid, steroid and holesterol metabolism and transport. These hanges were not due to simple loss of intestinal 229 epithelium, as areful morphologial examination showed no differenes between b Arrays and ontrol mie (Supplementary Fig. 2b). This downregulation of metabolism-related gene programs suggests that the intestinal epithelium, despite normal morphology, annot optimally handle nutrients or perform metaboli funtions in the absene of B ells. Figure Dysregulation of gene expression in the small intestine of mie. (a) Diagram of the disovery of the profile (see details in Methods). (b) Heatmap of the differentially expressed genes ( profile) in the jejunum of mie and their orresponding ontrols (false disovery rate <0%). Eah line represents one gene probe; eah olumn represents one array; olor represents the differene between and the orresponding ontrol on that array, blue indiating lower and red higher expression in mie; gray olor indiates missing values; het, heterozygous. (,d) Enrihment of Gene Ontology ategories (Biologial Proess) for the up- () and downregulated (d) genes in mie. The top ten ategories are shown. Genes µmt versus het Littermates IgA is ruial for normal intestinal gene expression Although B ells might affet gut homeostasis in several ways 8,9, the most likely is via antibody prodution. We therefore studied the gene expression profile of AID and µs double knokout mie, whih arry a deletion in the gene for ativation-indued ytidine deaminase, and a deletion in the seretion-speifi portion of the gene enoding IgM. These mie have B ells but do not produe sereted antibodies 9,20. The orrelation (r = 0.8) between the 280-gene profile of the antibodydefiient and of the two types of mie (Fig. 2a) was nearly as good as the orrelation (r = 0.85) between the two strains themselves (Fig. 2b), showing that antibody defiieny aounts for most of the gene expression properties of the B ell defiient phenotype. Beause IgA is the major antibody lass in the gut, we next obtained the intestinal expression profiles of mie defiient for IgA by virtue of a deletion in Igh-2 (IgAKO). The omparison of gene expression alterations in and IgAKO mie (to their respetive ontrols) indiated that about 70% of the phenotype seen in mie seemed to be due to the absene of IgA (Fig. 2), despite a ~00-fold ompensatory inrease in intestinal IgM gene expression (Fig. 2d). Thus, the majority of the intestinal effets of B ells seem to be mediated by IgA, and only a few depend on other B ell funtions. Gene expression in the jejunum of µmt versus heterozygous littermates 0 gene probes 48 gene probes Exluding B ell origin genes 0 gene probes 89 gene probes B ell gene probes µmt JhKO versus versus WT WT Nonlittermates Upregulated in intestine Downregulated in intestine Up Down KO versus ontrol log 2 fold hange d Not different in nonlittermates (exluded) Immune response Defense response Response to stimulus Response to virus Inflammatory response Innate immune response Response to bioti stimulus Response to wounding Response to other organism Cellular defense response Oxidation redution Steroid metaboli proess Lipid metaboli proess Alohol metaboli proess Response to hemial stimulus Cofator metaboli proess Cellular lipid metab. proess Cholesterol metaboli proess Lipid transport Hormone metaboli proess Validation on nonlittermates 82 gene probes 6 gene probes Up Down Final profile Upregulated in log 0 P value Downregulated in log 0 P value VOLUME 7 NUMBER 2 DECEMBER 20 nature mediine

3 a r t i l e s 20 Nature Ameria, In. All rights reserved. Figure 2 Dysregulation of gene expression is present in the gut of antibody-defiient and IgA-defiient mie. (a) Gene expression ratios between and ontrol (WT and heterozygous) mie (y axis, n = 27 per group) and between antibody-defiient versus heterozygous ontrol mie (x axis; n = 9 per group). (b) Gene expression ratios in two strains of mie: B0A-µMT versus ontrol (WT and heterozygous) mie (y axis, n = 7 per group) and BALB/-JhKO versus WT ontrol mie (x axis; n = 0 per group). () Gene expression ratios between and ontrol (y axis; 27 per group) and between immunoglobulin A-defiient (IgA) and heterozygous ontrol (x axis; n = 0 per group) mie. Similarity of gene expression profile between the experiments was estimated using Pearson s orrelation. Eah symbol represents one gene from the profile established in Figure. The values on axes are log 2 fold hange knokout versus ontrol. (d) Levels of IgM transripts (probe intensities on arrays) in the jejunum of IgA KO and heterozygous ontrol mie (P < 0.00). Mirobiota are essential for the B ell effet Beause most of the antibodies in the gut are direted against the mirobiota 2, a plausible senario would be that IgA defiieny leads to general overgrowth of mirobes and/or greater mirobial aess to the jejunum of mie. We found no differene in total amounts of baterial DNA (Fig. a), but serum lipopolysaharide (LPS) onentrations were higher in mie ompared to ontrol mie (Supplementary Fig. 2), suggesting that antibody defiieny might hange the speies distribution within the mirobial ommunity. We therefore analyzed the mirobiota in the jejunum by sequening DNA oding for 6S baterial ribosomal RNA (a highly onserved gene) and omparing the ompositions of different taxa. To ontrol for non genotype-related fators (suh as age, environment and parentage) we performed pairwise omparisons of sex-mathed and ontrol littermates. To ontrol for potential seasonal hanges in parameters suh as lighting, and for different bathes of food and bedding, we did two independent experiments 4 months apart. Overall, the baterial omposition was similar among ontrol and mie (Supplementary Fig. a). However, there were moderate differenes in three families. mie harbored fewer Clostridiaea-family bateria and more of the Paraous genus (Rhodobateraeae family) and of a speifi operational taxonomi unit (OTU) omprising the Latoous genus subgroup F49WGKH02HS7CN (Streptooaeae family; Supplementary Data) (Fig. b). All these bateria were minority representatives of the ommunity in both and ontrol mie, with the most abundant one (Clostridiaea) making up only about 0.4% of the jejunal bateria (Supplementary Fig. b), suggesting that small hanges within the mirobial ommunity may promote substantial alterations in intestinal funtion. Alternatively, these hanges in baterial populations may be a side effet of antibody defiieny and have little role in the metaboli hanges we saw. To test these possibilities, we derived germ-free lines of mie (both µmt and JhKO), plus their orresponding germ-free ontrols, and ompared their jejunal gene expression profiles. The expression differenes between and WT mie disappeared under germfree onditions for all but six genes of the profile (Fig. ). All of the germ-free mie, regardless of their B ell status, resembled the onventionally reared WT ontrols (Fig. d) rather than the onventionally reared mie, suggesting that the mirobiota do indeed have a role in the deline of metaboli gut funtion that ours in the absene of B ells. We next asked whether the phenotype required preisely the mirobial populations of the mie or whether any mirobes would do. We did a four-way omparison by olonizing germ-free and heterozygous ontrol littermates with mirobiota from or heterozygous ontrol mie. After ~ weeks, the gene expression a Down in antibody KO Down in lgako Correlation r = 0.8 P < 0.00 Up in antibody KO Correlation r = 0.68 P < 0.00 Up in lgako b hanges of these olonized mie were governed by their genotypes rather than by the soure of olonizing mirobiota (Fig. e,f and Supplementary Fig. 4a). For example, the expression ratios between onventionally raised and ontrol mie versus those of the germ-free reonstituted mie were similar, whether the reonstituting mirobiota ame from ontrol (Fig. e) or (Fig. f) mie. Thus, mirobial interations are neessary for the expression of the phenotype, but the host s genotype governs the final outome. Gata4-related metaboli defets in mie To determine how B ells and the mirobiota regulate intestinal metabolism, we searhed the promoters of the downregulated genes in the profile for any highly represented transription fator binding sites. There were almost seven times more Gata4-binding sites among these promoters than among promoters of ~,000 jejunally expressed genes that did not differ between and ontrols or among promoters of all known mouse genes (Fig. 4a). Gata4 is a key player in intestinal gene regulation and funtion, where it is speifially expressed in epithelial ells 22,2. To ask how genes in the profile might behave in the presene of B ells but the absene of Gata4, we studied mie that speifially lak Gata4 expression only in the intestinal epithelium (Gata4KO vil ) by virtue of a deletion in the Gata4 gene generated by Cre reombinase expressed under a villin promoter 2. Approximately 60% of the gene expression hanges in mie were also present in Gata4KO vil mie (Fig. 4b and Supplementary Table 2), indiating that the majority of the gene expression hanges in the mie were due to impaired Gata4-dependent funtions. These hanges were speifi to Gata4, as deletions in two other intestinally ative transription fators (Ppara and Klf9) 24,25 did not lead to similar hanges (Supplementary Fig. 4b). Several genes that were onordantly regulated in Gata4KO vil and mie are involved in fat uptake and metabolism. Sl27a2 enodes a fatty aid transporter 26 ; Osbpl enodes an intraellular lipid reeptor 27 ; and Aaab, Clps, Agmo and Apo enode proteins involved in lipid oxidation, hydroxylation or atabolism. Mie laking Pdk4 are resistant to the diabetogeni effets of high-fat diets 280, and Gata4KO vil mie have impaired fat and, espeially, Up in µmt Down in µmt Down in JhKO d IgM mrna 0,000, IgAKO Correlation r = 0.85 P < Up in JhKO nature mediine VOLUME 7 NUMBER 2 DECEMBER

4 A r t i l e s 20 Nature Ameria, In. All rights reserved. a d profile up genes Total baterial DNA profile down genes b Clostridiaeae (family) Paraous (genus) Latoous subgroup (genus) 00 holesterol absorption 2. We therefore examined lipid uptake in the mie and found that their holesterol and fat absorption were indeed less effiient ompared to the orresponding heterozygous GF GF e Conventional mie Correlation r = 0.56 P < Colonized with ontrol mirobiota 0 00 f Conventional mie Conventional mie littermates (Fig. 4 and Supplementary Fig. 5a), though there was no differene in overall food intake (Supplementary Fig. 5b). We also ompared the amounts of fat arried by homozygous Rnf206 Serpinae Serpinaa Mpt Serpinan Serpina Correlation r = 0.5 P < 0.0 Germ-free mie Correlation r = 0.58 P < Colonized with mirobiota Figure Mirobiota are neessary for intestinal alterations in mie. (a) Amount of total baterial DNA (ng) per 0 ng of isolated total DNA from jejunum ontent of and ontrol mie (eah dot represents data from two mie of the same litter). (b) Amount of DNA (pg) for Clostridiaea (family), Paraoous (genus) and for an operational taxonomi unit orresponding to a subgroup of Latoous (genus) per 0 ng of isolated total DNA from jejunum ontent of and ontrol mie. Data represented as in a. () Gene expression ratios between and ontrol (WT and heterozygous) (y axis; n = 27 per group) and between germ-free versus germ-free ontrol (WT and heterozygous) mie (x axis; n = per group). Six genes that are differentially expressed between germ-free and germ-free ontrol mie are labeled. (d) Two-dimensional visualization of the gene expression of upregulated ( profile up) and downregulated ( profile down) genes in four groups of mie. Eah symbol represents one mouse. Values are summary metris of upregulated (y axis) and downregulated (x axis) genes of the profile. Ellipses represent two s.d. from the entroids for (dashed line) and (solid line) groups. GF, germ-free. (e) Gene expression ratios between onventional and ontrol (WT and heterozygous) mie (y axis; n = 27 per group) and between ex germ-free versus ontrol mie olonized with mirobiota from ontrol mie (x axis; n = 8 per group). (f) Gene expression ratios between onventional and ontrol (WT and heterozygous) mie (y axis; n = 27 per group) and between ex germ-free versus ontrol mie olonized with mirobiota from ontrol mie (x axis; n = 5 per group). Eah symbol in,e,f represents one gene from the profile, and the values on axes are log 2 fold hange for knokout versus ontrol. Figure 4 Dysregulation of Gata4-dependent funtions in mie. (a) Proportion of Gata4 binding sites in the promoters of downregulated genes from profile, promoters of all known mouse genes and of ~,000 nonvarying genes expressed in the jejunum (P < 0 4 ). (b) Gene expression ratios (log 2 ) between versus WT and heterozygous ontrol (y axis; n = 27 per group) and between Gata4KO vil versus heterozygous ontrol mie (x axis; n = 5 per group). Eah symbol represents one gene. () Intestinal holesterol absorption in and heterozygous littermates. Eah dot represents one mouse. Eah line represents one litter. (d) Proportion of perigonaldal fat (grams) in relation to total body weight (grams) in (left), Gata4KO vil (right) and their orresponding heterozygous ontrol mie. (e) Serum leptin levels in (left) and Gata4KO vil (right) mie and their orresponding heterozygous ontrol mie. (*P < 0.05; **P < 0.0). In d,e, for and ontrol mie, eah dot represents median value of eah genotype in a given litter and eah line represents one litter. For Gata4KO vil mie, eah dot represents one mouse. a Perentage Gata4 binding sites Downreg.genes All mouse genes Nonvarying genes d 0.8 ** Fat amount (% perigonadal fat / body weight) Perentage perigonadal fat / body weight b Down in Gata4KO Vil 0.5 Gata4KO Vil ** e Serum leptin (pg ml ) Correlation r = 0.6 P < Up in Gata4KO Vil Perentage holesterol absorbed Serum leptin (pg ml ) ,000 8,000 ** *,000 2,000 2,000 6,000,000 ** 0 Gata4KO Vil 588 VOLUME 7 NUMBER 2 DECEMBER 20 nature mediine

5 a r t i l e s a 20 Nature Ameria, In. All rights reserved. b Cd8 Cl5 Trg Cd Down in RagKO Correlation r = 0.6 (All genes) r = 0.7 (Gata4-dep. genes) Up in RagKO Whole tissue Up in B-ell KO Down in B-ell KO Correlation r = 0.8 P < Epithelium Figure 5 Inreased immune and dereased metaboli state of epithelium in mie. (a) Gene expression network of profile reonstruted from the data of ontrol mie (zoomed in to the entral part of the network; the whole network is in Supplementary Fig. 5a). Eah line represents a orrelation, and eah node represents a gene. White lines are positive and blak are negative orrelations; triangles are Gata4-dependent genes, squares are the T ell genes and irles are all other genes; red node fills indiate upregulation, and light blue fills are downregulation; green node outline is a subnetwork, yellow is a subnetwork 2, blue is both subnetworks, and unoutlined node symbols do not belong to either of the two subnetworks. (b) Gene expression ratios (log 2 ) between versus ontrol (y axis; n = 27 per group) and between RAG2-knokout (RagKO) versus ontrol (x axis; n = 8 per group) mie. Eah symbol represents one gene. Grey squares are T ell related genes. Pink diamonds are Gata4-dependent genes; blue are all other genes. () Gene expression ratios (log 2 ) between gut whole tissue of versus ontrol mie (y axis; n = 27 per group) and between intestinal epithelium of versus ontrol mie (x axis; n = per group). Eah symbol represents one gene. (d) Immunohistohemistry for Zbp in jejunum of and ontrol mie (left) and the results of semiquantitative blind evaluation (right; n = 4, *P < 0.05; bars represent s.d.); sale bars, 40 µm. d Zbp expression sore 2 0 * versus heterozygous B ell suffiient littermates on both regular and high-fat diets and by Gata4KO vil versus ontrol mie. Both types of defiient mie had signifiantly lower amounts of perigonadal fat than their ontrols, regardless of diet (Fig. 4d and Supplementary Fig. 5). We also assessed other body fat stores (inguinal, mesenteri) and general adiposity by nulear magneti resonane in an independent group of mie and found dereased amounts of fat ompared to heterozygous littermate ontrols (Supplementary Fig. 5e). Finally, beause fat stores orrelate with systemi leptin levels, we measured serum leptin onentrations in Gata4KO vil and mie and found lower amounts of leptin in both types of defiient mie than in ontrols (Fig. 4e and Supplementary Fig. 5d,e). Thus, many of the metaboli hanges seen in mie seem to be Gata4 dependent. Immune and metaboli gene networks in gut epithelium The profile onsists of two parts that are not obviously related: downregulated metaboli and upregulated immune genes (Fig. ). To searh for a physiologial link between them, we used a reently developed analytial tool for reonstrution of gene expression networks 2. Correlations in expression of the genes in the profile in a group of 9 normal mie (omprising both WT and heterozygous mie) revealed a network of 228 interonneted genes ontaining 688 onnetions and two major subnetworks (Fig. 5a and Supplementary Fig. 6a). The first subnetwork (77 genes, 56% downregulated), inorporated many of the downregulated metaboli genes, two-thirds of whih were Gata4 dependent, and was highly enrihed for genes governing lipid transport and other metaboli proesses (Supplementary Fig. 6b). The seond subnetwork (4 genes, 88% nature mediine VOLUME 7 NUMBER 2 DECEMBER

6 A r t i l e s a Treated/ontrol fold hange CpG E. oli IFN LPS Polyl:C (in vivo) Irf7 Gbp6 Zbp Aaab Crip Cyp27a Dbp Pdk4 Vat b Mie Mie 5 Correlation r = 0.64 Correlation r = 0.7 P < P < Down in CVID Up in CVID Down in HIV Up in HIV Humans Humans Normal host B lymphoyte or antibody-defiient host 20 Nature Ameria, In. All rights reserved. Figure 6 Balane between immune and metaboli proesses in vitro and in vivo in mie and humans. (a) Influene of different stimuli on immune (red) and metaboli (blue) gene expression in intestinal epithelial ells in vitro. MODE-K ells were inubated for 24 h with different stimuli. Gene expression assessed by real-time PCR is represented as median fold hange ompared to untreated ells in five independent experiments. For omparison, the right-most olumns show the fold hanges deteted in jejunum of mie ompared to ontrol mie. Supplementary Figure 9 shows P values for all omparisons. (b) Gene expression ratios between versus ontrol mie (y axis; n = 27 per group) and between individuals with CVID versus individuals with normal gut histology (x axis; n = per group; left) and between versus ontrol mie (y axis; n = 27 per group) and between HIV-infeted individuals versus healthy ontrols (x axis; n = per group; right). The Pearson orrelation between log 2 ratios for mie and humans is shown. GATA4-dependent genes are shown in pink. () A shemati diagram illustrating that, unlike the situation with upregulated) ontained mostly upregulated immune and inflammatory genes (Supplementary Fig. 6). These metaboli and immune subnetworks are onneted by five genes and are inversely regulated (Fig. 5a). This might reflet the ativity of two interating ell types (for example, epithelial and immune ells) or of two sets of genes within a single ell (for example, metaboli and immune genes in epithelium). We found an indiation for the first senario that pointed to T ells, as the immune subnetwork ontains a minigroup of highly onneted T ell speifi genes (Cdg, Gzma, Gzmb, Cd8a, Trg and Cl5; Fig. 5a), and flow ytometry analysis revealed an inreased proportion of CD8 + and CD4 + CD8 + T ells in the jejunum of mie (Supplementary Fig. 7), onsistent with an earlier report that µmt mie harbor elevated populations of highly ytotoxi intestinal CD8 + T lymphoytes. Downmodulation of intestinal metaboli funtions might thus be due to the negative effets of over-exuberant T ell ytotoxi responses. To test this possibility, we analyzed the profile in RAG2-knokout mie (whih lak both B and T ells) and found that the lak of T ells had little impat on the profile (exept for T ell related genes) and even less effet on Gata4-dependent genes (Fig. 5b). The alternative senario, that there might be a moleular swith between immune and metaboli proesses within the epithelial ells themselves, was suggested by our finding that a single gene (Gbp6, enoding an interferon-induible GTPase 4 ) strongly onnets the two subnetworks and orrelates positively with immune and negatively with metaboli subnetwork genes. To test this hypothesis, we isolated intestinal epithelial ells (with <5% intraepithelial T ells) and ompared their expression of the profile with that of the intat whole tissue. The epitheliumspeifi expression profile resembled that of the whole tissue (Fig. 5), with a orrelation (r = 0.8) similar to that between the two types of mie (r = 0.85, Fig. 2b). Protein amounts (of a small number of testable genes) followed the mrna levels. On western blots, interferon-indued GTPase abundane was elevated in epithelial ells ompared Mirobiota IgA GATA4 Epithelial ells B Adipose Dietary lipids Mirobiota GATA4 Epithelial ells Immune funtion Adipose Dietary lipids Metaboli funtion Absorption Deposition replete B ells, the balane between immune and metaboli proesses within gut epithelial ells in hosts laking B lymphoytes or IgA antibodies is skewed towards an immune response against mirobiota (gray, blak, green and blue ovoid objets) at the expense of GATA4-related metabolism, mainly lipid absorption and deposition. to heterozygous ontrols, whereas the metaboli proteins aetyl-coa ayltransferase and latase were suppressed (Supplementary Fig. 8). Histologially, expression of Zbp (Z DNA binding protein, a ytosoli baterial DNA sensor and ativator of interferon (IFN)-regulatory fators 5 ), one of the most onneted genes in the immune subnetwork, was also elevated in epithelium of versus heterozygous ontrol mie (Fig. 5d). Altogether, these data suggest that hanges in the intestinal epithelial ells themselves are primarily responsible for the immune and metaboli phenotype we see in the whole-tissue analysis. Mirobe effets on epithelium Beause the epithelial differenes between and ontrol mie were evident only when mirobes were present, we asked whether the mirobes at diretly on the epithelium or indiretly by ativating immune ells that, in turn, affet epithelial funtion. To test this, we added Esherihia oli and Toll-like reeptor ligands to the mouse epithelial ell line MODE-K and measured the expression of Gata4-dependent metaboli genes. Although most intestinal epithelial ell lines stop expressing many metaboli genes that are expressed in vivo, we found that the MODE-K ell line retains expression of a few of the Gata4-related metaboli genes in the profile. Treating MODE-K ells with heat-inativated E. oli, or with LPS, dereased the expression of most of the metaboli genes and simultaneously indued expression of immune genes (Fig. 6a). Poly I:C had an effet on half of the tested genes, and CpG treatment indued no signifiant hanges in most of the genes (Fig. 6a). To test the predition from the network analysis that IFN-dependent responses would downregulate metaboli genes, we also added a ombination of IFN-α and IFN-γ and found that this did indeed ause a diret inhibition of metaboli funtion (Fig. 6a and Supplementary Fig. 9). Therefore, in the absene of B ells or of IgA, the intestinal epithelium itself responds to mirobes, upregulating IFN-induible immune response pathways and simultaneously repressing Gata4-related metaboli funtions. 590 VOLUME 7 NUMBER 2 DECEMBER 20 nature mediine

7 a r t i l e s 20 Nature Ameria, In. All rights reserved. Immunodefiienies in humans Several features led us to the hypothesis that our results linking B ells, intestinal epithelium and mirobiota might be relevant to humans. First, the intestinal mirobiota are ruial in host energy uptake and fat metabolism and have been linked to human obesity (reviewed in ref. 6). Seond, the GATA4 gene is ~90% onserved between humans and mie, with similar tissue mrna expression levels (Supplementary Fig. 0). Third, two human immunodefiienies (CVID and seondary immunodefiieny aused by HIV infetion) have similar gastrointestinal syndromes with unexplained malabsorption, low weight gain and steatorrhea (fatty stools) 7 4. Although the CVID syndrome was first desribed ~60 years ago 42, the moleular pathogenesis of the intestinal defets in either immunodefiieny is still unlear 4,44. To determine whether the moleular mehanisms found in the mie might also operate in the human disease, we analyzed gene expression profiles of duodenal biopsies from three humans with CVID with gastrointestinal pathology and three immunodefiient subjets with normal duodenal histology. For HIV, we used reently published gene expression data omparing HIV-infeted individuals to healthy uninfeted people 45. The gene expression differenes between immunodefiient humans and their respetive ontrols were remarkably similar to the results from mie (Fig. 6b). Genes onordantly upregulated in immunodefiient mie and humans largely involved immune omponents, suh as T ell genes (CD8A, CD5, TNFSF0), interferon-indued genes (RSAD2, OAS2, IFI44, IFITM, HLA-C, HLA-DRB) and omplement (Supplementary Tables and 4), whereas the downregulated profile ontained mostly metaboli funtions suh as lipid and arbohydrate metabolism (APOC, PDK4, HPGD, FBP), oxidative redution (IYD) and transport of mironutrients, inluding vitamins (ABCC6, SLC46A, OPLAH). Finally, ~95% of the Gata4-dependent genes were onordantly regulated in both the mie and the human immunodefiienies (Fig. 6b), suggesting that these genes are diretly involved in the human malabsorption and impaired lipid deposition (Fig. 6). DISCUSSION Our data unover a three-way onversation in the small intestine between the epithelium, B ells and the mirobiota that moderates immunity and also influenes intestinal and systemi metaboli homeostasis. This trialogue affets the balane between immune IFN-related and metaboli GATA4-related funtions within epithelial ells in both humans and mie, and it seems to be involved in the previously unexplained malabsorption and onsequent malnutrition in people with primary and seondary immunodefiieny (from CVID and HIV, respetively). Although B ells have several funtions in immunity 8,9,46,47, we found that their effet on intestinal metabolism depends on their ability to serete IgA (the antibody normally found in muosal surfaes). Most intestinal IgA is direted against intestinal flora 2, and it an have several different effets. It restrits baterial aess to the epithelium and to intestinal immune ells 2,48, and it an hange the moleules expressed by bateria 49 or promote survival of speifi bateria 50. To isolate the relevant IgA funtion, we analyzed the mirobiota found in normal and B ell defiient mie. The total numbers of luminal bateria were similar, and there was no obvious pathogen among the bateria in mie. Rather, there were hanges in three minor ommensal baterial subsets. This left us with three possibilities. First, the phenotype was unrelated to the mirobiota. Seond, minor hanges in the mirobiota were suffiient to ause the phenotype. Or, third, hanges in the ativity, rather than the diversity of the intestinal mirobiota, were important. To test the first, we derived germ-free B ell defiient and littermate ontrol mie and found that the phenotype did indeed depend on the mirobiota. To test the seond, we transferred the mirobiota from and ontrol mie to germ-free mie and found that both populations of mirobes indued the B ell defiient phenotype in germ-free reipients, whereas neither indued the phenotype in WT germ-free mie. Together, these results suggested that mirobes were needed to reveal the B ell defiient phenotype, but no partiular mirobe was essential. However, in a preliminary study, we found that olonization with pure Latobaillus reuteri (one of the most ommon jejunal bateria in both and heterozygous ontrol mie) did not indue the phenotype (data not shown), suggesting that ontat with a partiular mirobe, or an interating group of mirobes, may be important. The higher levels of serum LPS that we found in mie suggest that the mirobes in these mie are in greater ontat with the gut tissue, perhaps diretly signaling the epithelium to the presene of the offending bateria. These results support the idea that disease does not neessarily require the presene of a lassial pathogen. Commensals may ause disease or not, depending on the host s genotype, as the funtional ativity of the same mirobe might differ depending on the host 49. Beause the absene of B ells resulted in the inreased expression of many immune genes, we asked whether the downregulation of metaboli funtion in epithelial ells was an indiret result of inreased ativity by ells of the immune system or a result of hanges of expression in the epithelium itself. By network analysis, we found that the epithelial ells have two interating gene networks, one governing lipid metabolism (mostly regulated by Gata4) and one governing innate immunity (mostly made up of IFN-dependent genes) that are inversely onneted via a small number of genes, with the main one being Gbp6, an IFN-induible gene suggested to have antibaterial funtion 4. Isolated epithelial ells from mie, tested immediately ex vivo, showed inreased immune funtion and dereased metaboli funtion at both the RNA and protein levels. Further, an epithelial ell line responded diretly to bateria and baterial omponents, as well as to IFNs, by upregulating immune genes and downregulating metaboli ones. Together, these data suggest that the intestinal epithelium responds to bateria, and to inflammatory signals, by upregulating its own immune funtions while downregulating its metaboli ones. The reverse an also our, as we saw that the intestinal ells of Gata4KO vil mie, whih have dereased metaboli funtion and gene expression, have simultaneously inreased their expression of immune genes. It is noteworthy that the onnetions between the epithelial immune and metaboli funtions are not speifi to B ell defiieny, as the gene expression network was reonstruted on the basis of the variability of gene expression seen in normal mie. Therefore, the interplay between immune and metaboli proesses within epithelial ells is a normal physiologial proess that beomes skewed in the small intestine of mie laking B ells. Immunodefiient humans with CVID or HIV have several features in ommon with the and IgAKO mie, suggesting that the underlying mehanism of the gastrointestinal disorders is similar in the two speies. First, as in mie, no pathogen has onsistently been assoiated with gastrointestinal syndrome in people with CVID or HIV 8,4. Seond, in some ases, HIV-infeted people have low populations of intestinal IgA-sereting plasma ells 4,5,52. Third, although many immunodefiient humans are routinely treated with IgG infusions (whih help to protet them from infetions), nature mediine VOLUME 7 NUMBER 2 DECEMBER 20 59

8 A r t i l e s 20 Nature Ameria, In. All rights reserved. suh infusions do not ameliorate the gastrointestinal problems in CVID 8,44. Similarly, two months of IgG injetions did not result in any normalization of the intestinal gene expression profile of mie (data not shown). Our finding that IgA is the major B ell produt maintaining the metaboli trialogue suggests that it might be useful to develop a pool of stable, intestinal IgA moleules to add to the urrent treatment protool. Fourth, there is a link through GATA4. The expression of GATA4-dependent genes is similarly affeted in mie and in CVID and HIV intestinal biopsies, suggesting that the moleular mehanism of malabsorption is similar in mie and humans. It has been proposed that these metaboli alterations might result from harmful ativity of ytotoxi T ells residing in the gut of immunodefiient people 8. Our results, however, indiate that T ells have little influene on the metaboli dysbalane; rather, epithelial ells themselves rediret their ativity from metaboli to immune funtion in response to hanges in the miroenvironment. Therefore, treatment for GI syndrome might need to be refoused from generi anti-inflammatory mediation to speifi modulation of GATA4-dependent ativity of epithelial ells. Finally, our study supports our previous proposal 6,7 that tissues take an ative role in their own immune protetion. In the presene of a funtional adaptive immune system, the intestinal epithelium an onentrate on its metaboli funtions. However, if the adaptive immune system is dysfuntional (in this ase beause it laks B ells), the intestinal epithelium takes on some of the missing immune funtions at the expense of its metaboli ativity (Fig. 6). It will be worthwhile to investigate further how often suh diversion is the basis for autoimmune disease. Methods Methods and any assoiated referenes are available in the online version of the paper at Aession odes. Gene expression files ontaining array data are available under the GSE294 superseries in the Gene Expression Omnibus (GEO) data repository. All mirobiome data are deposited in the National Center for Biotehnology Information (NCBI) Sequene Read Arhive under aession number SRP0025. Note: Supplementary information is available on the Nature Mediine website. Aknowledgments This researh was supported in part by the Intramural Researh Program of the NIAID, NIH. We thank T. Myers, Q. Su and A. Godinez of the NIAID miroarray faility for exellent tehnial support; R. Shwartz for providing funding for germ-free re-derivation; W. Strober for help with human subjets; C. Jones for tehnial support of baterial pyrosequening; T.D. Randall (Trudeau Institute) for AID/µS mie; S. Epstein (US Food and Drug Administration), J. Misplon (US Food and Drug Administration) and D.P. Huston (Texas A&M Health Siene Center) for IgAKO mie; D. Kaiserlian (Institut National de la Santé et de la Reherhe Médiale) and R. Blumberg (Harvard University) for providing the MODE-K ell line; R. Varma for help in experiments; Laboratory of Cellular and Moleular Immunology members and M. Sterman Dolnikoff, I. Shmulevih and A. Dzutsev for disussions and suggestions; Y. Kotliarov for help with graphis; S. Varma for help finding human gene homologues; J. Coursen for exellent tehnial support; B. Epstein for assistane with software programming; and the personnel of NIH mouse failities in buildings 4 and 6B. The Cininnati Mouse Metaboli Phenotyping Center is supported by NIH grant U24 DK05960 and National Gnotobioti Rodent Resoure Center at the University of North Carolina is supported by grant P40RR0860. AUTHOR CONTRIBUTIONS N.S. and A.M. oneived the original idea, designed the study, onduted most of the experiments, analyzed the data and wrote the manusript. W.H. analyzed mirobiome data and drafted the related part of the manusript. M.B. provided Gata4KO vil mie and did some Gata4-related experiments. M.Y. and L.M. olleted duodenal biopsies and performed linial evaluation of human subjets. O.G. performed and analyzed experiments related to lipid metabolism. M.O. performed histologial evaluation of mouse samples. A.J.M. and K.D.M. re-derived germ-free JhKO mie and olleted organs. C.F.-L. supervised mirobiome evaluation and drafted the related part of the manusript. P.M. supervised the whole study and wrote the manusript. COMPETING FINANCIAL INTERESTS The authors delare no ompeting finanial interests. Published online at Reprints and permissions information is available online at reprints/index.html.. Cerf-Bensussan, N. & Gaboriau-Routhiau, V. The immune system and the gut mirobiota: friends or foes? Nat. Rev. Immunol. 0, (200). 2. Hooper, L.V. & Mapherson, A.J. Immune adaptations that maintain homeostasis with the intestinal mirobiota. Nat. Rev. Immunol. 0, (200).. Umesaki, Y., Setoyama, H., Matsumoto, S., Imaoka, A. & Itoh, K. Differential roles of segmented filamentous bateria and lostridia in development of the intestinal immune system. Infet. Immun. 67, 5045 (999). 4. Ivanov, I.I. et al. Indution of intestinal T H 7 ells by segmented filamentous bateria. Cell 9, (2009). 5. Gaboriau-Routhiau, V. et al. The key role of segmented filamentous bateria in the oordinated maturation of gut helper T ell responses. Immunity, (2009). 6. Klaasen, H.L. et al. Apathogeni, intestinal, segmented, filamentous bateria stimulate the muosal immune system of mie. Infet. Immun. 6, 006 (99). 7. Round, J.L. & Mazmanian, S.K. Induible Foxp + regulatory T-ell development by a ommensal baterium of the intestinal mirobiota. Pro. Natl. Aad. Si. USA 07, (200). 8. Vijay-Kumar, M. et al. Metaboli syndrome and altered gut mirobiota in mie laking Toll-like reeptor 5. Siene 28, (200). 9. Fagarasan, S. et al. Critial roles of ativation-indued ytidine deaminase in the homeostasis of gut flora. Siene 298, (2002). 0. Suzuki, K. et al. Aberrant expansion of segmented filamentous bateria in IgAdefiient gut. Pro. Natl. Aad. Si. USA 0, (2004).. Umesaki, Y., Okada, Y., Matsumoto, S., Imaoka, A. & Setoyama, H. Segmented filamentous bateria are indigenous intestinal bateria that ativate intraepithelial lymphoytes and indue MHC lass II moleules and fuosyl asialo GM glyolipids on the small intestinal epithelial ells in the ex-germ-free mouse. Mirobiol. Immunol. 9, (995). 2. Bry, L., Falk, P.G., Midtvedt, T. & Gordon, J.I. A model of host-mirobial interations in an open mammalian eosystem. Siene 27, 808 (996).. Turnbaugh, P.J. et al. An obesity-assoiated gut mirobiome with inreased apaity for energy harvest. Nature 444, (2006). 4. Gordon, J.I., Hooper, L.V., MNevin, M.S., Wong, M. & Bry, L. Epithelial ell growth and differentiation. III. Promoting diversity in the intestine: onversations between the miroflora, epithelium, and diffuse GALT. Am. J. Physiol. 27, G565 G570 (997). 5. Treiner, E. et al. Seletion of evolutionarily onserved muosal-assoiated invariant T ells by MR. Nature 422, (200). 6. Matzinger, P. Tolerane, danger and the extended family. Annu. Rev. Immunol. 2, (994). 7. Matzinger, P. & Kamala, T. Tissue-based lass ontrol: the other side of tolerane. Nat. Rev. Immunol., (20). 8. Mizoguhi, A., Mizoguhi, E., Takedatsu, H., Blumberg, R.S. & Bhan, A.K. Chroni intestinal inflammatory ondition generates IL-0 produing regulatory B ell subset haraterized by CDd upregulation. Immunity 6, (2002). 9. Golovkina, T.V., Shlomhik, M., Hannum, L. & Chervonsky, A. Organogeni role of B lymphoytes in muosal immunity. Siene 286, (999). 20. Carragher, D.M., Kaminski, D.A., Moquin, A., Hartson, L. & Randall, T.D. A novel role for non-neutralizing antibodies against nuleoprotein in failitating resistane to influenza virus. J. Immunol. 8, (2008). 2. Mapherson, A.J. et al. A primitive T ell independent mehanism of intestinal muosal IgA responses to ommensal bateria. Siene 288, (2000). 22. Bosse, T. et al. Gata4 is essential for the maintenane of jejunal-ileal identities in the adult mouse small intestine. Mol. Cell. Biol. 26, (2006). 2. Battle, M.A. et al. GATA4 is essential for jejunal funtion in mie. Gastroenterology 5, (2008). 24. Bünger, M. et al. Genome-wide analysis of PPARα ativation in murine small intestine. Physiol. Genomis 0, (2007). 25. Simmen, F.A. et al. Dysregulation of intestinal rypt ell proliferation and villus ell migration in mie laking Kruppel-like fator 9. Am. J. Physiol. Gastrointest. Liver Physiol. 292, G757 G769 (2007). 26. Falon, A. et al. FATP2 is a hepati fatty aid transporter and peroxisomal very long-hain ayl-coa synthetase. Am. J. Physiol. Endorinol. Metab. 299, E84 E9 (200). 27. Lehto, M. et al. The OSBP-related protein family in humans. J. Lipid Res. 42, 20 2 (200). 592 VOLUME 7 NUMBER 2 DECEMBER 20 nature mediine

9 a r t i l e s 20 Nature Ameria, In. All rights reserved. 28. Hwang, B., Jeoung, N.H. & Harris, R.A. Pyruvate dehydrogenase kinase isoenzyme 4 (PDHK4) defiieny attenuates the long-term negative effets of a high-saturated fat diet. Biohem. J. 42, (2009). 29. Watshinger, K. et al. Identifiation of the gene enoding alkylglyerol monooxygenase defines a third lass of tetrahydrobiopterin-dependent enzymes. Pro. Natl. Aad. Si. USA 07, (200). 0. Berger, K., Winzell, M.S., Mei, J. & Erlanson-Albertsson, C. Enterostatin and its target mehanisms during regulation of fat intake. Physiol. Behav. 8, (2004).. Ahima, R.S. & Flier, J.S. Leptin. Annu. Rev. Physiol. 62, 4 47 (2000). 2. Skinner, J. et al. Construt and ompare gene oexpression networks with DAPfinder and DAPview. BMC Bioinformatis 2, 286 (20).. Nishiyama, Y. et al. Homeostati regulation of intestinal villous epithelia by B lymphoytes. J. Immunol. 68, (2002). 4. Degrandi, D. et al. Extensive haraterization of IFN-indued GTPases mgbp to mgbp0 involved in host defense. J. Immunol. 79, (2007). 5. Yanai, H., Savitsky, D., Tamura, T. & Taniguhi, T. Regulation of the ytosoli DNAsensing system in innate immunity: a urrent view. Curr. Opin. Immunol. 2, 7 22 (2009). 6. Turnbaugh, P.J. & Gordon, J.I. The ore gut mirobiome, energy balane and obesity. J. Physiol. (Lond.) 587, (2009). 7. Hughes, W.S., Cerda, J.J., Holtzapple, P. & Brooks, F.P. Primary hypogammaglobulinemia and malabsorption. Ann. Intern. Med. 74, (97). 8. Mannon, P.J. et al. Exess IL-2 but not IL-2 aompanies the inflammatory bowel disease assoiated with ommon variable immunodefiieny. Gastroenterology, (2006). 9. Agarwal, S. & Mayer, L. Pathogenesis and treatment of gastrointestinal disease in antibody defiieny syndromes. J. Allergy Clin. Immunol. 24, (2009). 40. Koh, J. et al. Steatorrhea: a ommon manifestation in patients with HIV/AIDS. Nutrition 2, (996). 4. Ullrih, R., Rieken, E.O. & Zeitz, M. Human immunodefiieny virus indued enteropathy. Immunol. Res. 0, (99). 42. Sanford, J.P., Favour, C.B. & Tribeman, MS. Absene of serum gamma globulins in an adult. N. Engl. J. Med. 250, (954). 4. Owens, S.R. & Greenson, J.K. The pathology of malabsorption: urrent onepts. Histopathology 50, (2007). 44. Malamut, G. et al. The enteropathy assoiated with ommon variable immunodefiieny: the delineated frontiers with elia disease. Am. J. Gastroenterol. 05, (200). 45. Lerner, P. et al. The gut muosal viral reservoir in HIV-infeted patients is not the major soure of rebound plasma viremia following interruption of highly ative antiretroviral therapy. J. Virol. 85, (20). 46. Crawford, A., Maleod, M., Shumaher, T., Corlett, L. & Gray, D. Primary T ell expansion and differentiation in vivo requires antigen presentation by B ells. J. Immunol. 76, (2006). 47. Wojiehowski, W. et al. Cytokine-produing effetor B ells regulate type 2 immunity to H. polygyrus. Immunity 0, 42 4 (2009). 48. Cong, Y., Feng, T., Fujihashi, K., Shoeb, T.R. & Elson, C.O. A dominant, oordinated T regulatory ell IgA response to the intestinal mirobiota. Pro. Natl. Aad. Si. USA 06, (2009). 49. Peterson, D.A., MNulty, N.P., Guruge, J.L. & Gordon, J.I. IgA response to symbioti bateria as a mediator of gut homeostasis. Cell Host Mirobe 2, 289 (2007). 50. Obata, T. et al. Indigenous opportunisti bateria inhabit mammalian gut-assoiated lymphoid tissues and share a muosal antibody-mediated symbiosis. Pro. Natl. Aad. Si. USA 07, (200). 5. Samurra, R.W. et al. Muosal plasma ell repertoire during HIV- infetion. J. Immunol. 69, (2002). 52. Kotler, D.P., Sholes, J.V. & Tierney, A.R. Intestinal plasma ell alterations in aquired immunodefiieny syndrome. Dig. Dis. Si. 2, 298 (987). nature mediine VOLUME 7 NUMBER 2 DECEMBER 20 59