Organ-Level Protein Networks As a Reference for the Host Effects of the Microbiome

Home , ATPAF1, CELA1, CELA3B, DHX16, HEBP2, NUBPL

Downloaded from genome.cshlp.org on October 8, 2021 - Published by Cold Spring Harbor Laboratory Press

Resource Organ-level protein networks as a reference for the host effects of the microbiome

Robert H. Mills,1,2,3,4 Jacob M. Wozniak,1,2 Alison Vrbanac,3 Anaamika Campeau,1,2 Benoit Chassaing,5,6,7,8 Andrew Gewirtz,5 Rob Knight,3,4 and David J. Gonzalez1,2,4 1Department of Pharmacology, University of California, San Diego, California 92093, USA; 2Skaggs School of Pharmacy and Pharmaceutical Sciences, University of California, San Diego, California 92093, USA; 3Department of Pediatrics, and Department of Computer Science and Engineering, University of California, San Diego, California 92093, USA; 4Center for Microbiome Innovation, University of California, San Diego, California 92093, USA; 5Center for Inflammation, Immunity and Infection, Institute for Biomedical Sciences, Georgia State University, Atlanta, Georgia 30303, USA; 6Neuroscience Institute, Georgia State University, Atlanta, Georgia, USA; 7INSERM, U1016, 75014 Paris, France; 8Université de Paris, 75006 Paris, France

Connections between the microbiome and health are rapidly emerging in a wide range of diseases. However, a detailed mechanistic understanding of how different microbial communities are influencing their hosts is often lacking. One method researchers have used to understand these effects are germ-free (GF) mouse models. Differences found within the organ systems of these model organisms may highlight generalizable mechanisms that microbiome dysbioses have throughout the host. Here, we applied multiplexed, quantitative proteomics on the brains, spleens, hearts, small intestines, and colons of conventionally raised and GF mice, identifying associations to colonization state in over 7000 proteins. Highly ranked associations were constructed into protein–protein interaction networks and visualized onto an interactive 3D mouse model for user-guided exploration. These results act as a resource for microbiome researchers hoping to identify host effects of microbiome colonization on a given organ of interest. Our results include validation of previously reported effects in xenobiotic metabolism, the innate immune system, and glutamate-associated proteins while simultaneously providing organism-wide context. We highlight organism-wide differences in mitochondrial proteins including consistent increases in NNT, a mitochondrial protein with essential roles in influencing levels of NADH and NADPH, in all analyzed organs of conventional mice. Our networks also reveal new associations for further exploration, including protease responses in the spleen, high-density lipoproteins in the heart, and glutamatergic signaling in the brain. In total, our study provides a resource for microbiome researchers through detailed tables and visualization of the protein-level effects of microbial colonization on several organ systems.

[Supplemental material is available for this article.]

The gut microbiome is emerging as a critical component of human have been performed, but these studies have generally highlighted health. It has been shown that the microbial communities coloniz- a select few organ tissues. Protein-level studies have shown varying ing our bodies play important roles in the immune development responses to colonization along different regions of the gastroin- of infants (Milani et al. 2017) and the regulation of the innate im- testinal (GI) tract (Lichtman et al. 2016), changes in drug metabo- mune system (Thaiss et al. 2016). Further, a dysbiosis of the gut lizing proteins in livers and kidneys (Kuno et al. 2016), and microbiome has been correlated with many diseases including in- differences in circulating fatty acids from an analysis of serum flammatory bowel disease (IBD) (Sartor and Wu 2017), diabetes and livers (Kindt et al. 2018). They have also shown that microbial (Tilg and Moschen 2014), obesity (Bouter et al. 2017), cardiovascu- colonization alters posttranslational modifications, including his- lar disease (Ahmadmehrabi and Tang 2017), and mental health tone acetylation and methylation in liver, colon, and adipose tis- disorders (Nguyen et al. 2018). Microbial production or modifi- sue (Krautkramer et al. 2016), as well as lysine acetylation in the cation of metabolites such as bile acids, choline derivatives, gut and liver (Simon et al. 2012). An important transcriptomic vitamins, and lipids provide some insight into the underlying study revealed a strong connection between colonization and in- host-microbe interactions in these diseases (Nicholson et al. creased Nnt, a mitochondrial protein that has functions in redox 2012). However, many mechanisms mediating these disease states homeostasis and biosynthetic pathways through the generation remain unknown. of NADH and NADP+ (Mardinoglu et al. 2015). The authors found Germ-free (GF) mouse models, wherein a mouse is raised Nnt transcripts increased in several conventional mouse tissues, in- without any exposure to microbes, have been an invaluable tool cluding sections of the small intestine, colon, and liver, which cor- for assessing causal effects in microbiome research (Bhattarai and related with significant alterations in host amino acid levels and Kashyap 2016). GF models also provide an opportunity to under- glutathione metabolism (Mardinoglu et al. 2015). Other related stand the fundamental effects of microbial colonization at an or- studies found transcript differences in the brain (Diaz Heijtz ganismal scale. Systems level analyses of the tissues of GF mice

© 2020 Mills et al. This article is distributed exclusively by Cold Spring Harbor Laboratory Press for the first six months after the full-issue publication date (see Corresponding author: [email protected] http://genome.cshlp.org/site/misc/terms.xhtml). After six months, it is available Article published online before print. Article, supplemental material, and publi- under a Creative Commons License (Attribution-NonCommercial 4.0 Interna- cation date are at http://www.genome.org/cgi/doi/10.1101/gr.256875.119. tional), as described at http://creativecommons.org/licenses/by-nc/4.0/.

30:1–11 Published by Cold Spring Harbor Laboratory Press; ISSN 1088-9051/20; www.genome.org Genome Research 1 www.genome.org Downloaded from genome.cshlp.org on October 8, 2021 - Published by Cold Spring Harbor Laboratory Press

Mills et al. et al. 2011) and further highlighted the SPLEEN SMALL NNNTMTHFDHF 1 GI-dependent transcript effects of micro- PGGRMGRRMMCM 1

AKR1C1R1C 4PTPTP4A4 1AANONOO6 AGGGPPAPAT2 INTESTINE HMBMBS RRMM1 bial colonization in Myd88-deficient NUP15UP1 3 TDP2 CNNDNDP2 GGLLS AKR1C1KKR1KR 3 ABHD14BHDH B TYMTTYYMS STEEAPAPP3 FECH PPSAAATT1 GGLUGLLUD1 KCKCCNACNABAB2 DDHRHRSR 1 RRDDHDH11 ANXA1NXAN 3 mice (Larsson et al. 2012). RRM2 NR3CR3 1 ALLLDHDH1H11AA3 TTFRRC CTSE CTH AKAKKR1KR1C1C 2 MAOAOADHRRSS3 AASSNNS DHDHRRSS11 FFTLL1 SERPINARPIN 6 LLCT RDDH119 GALGALALKK1 GMGMM57M5771 GGCCCLLC ATP6V6V1V F Here, we sought to further detail the CCTSCTTTSSB HP CHIL3 MMAAOOB EEPPDGDX GAGAT1 HHINTT1 PRRRSSSSS1 GSSTTO1GSSTASTTAA4 SSORSOORD MECR TRYRY5 TBBXAASA 1 AAKKR1B1B7 PPAPA2 LLTF LLCNN2 ABCB1B1A ALALDDHH1A1A7 DDCCXCXR RRPLRPPLL117 ZZFFP6222 GGSSTSTATA1ALLDDDHH1H11AA1 CARRS NDDDUFFAF 8 protein effects of microbiota coloniza- MCG___155095050995CCPACPPAPA2 ALLOXLOX1X15 PPCKPCCKK1 TTREREH CCPPBPB1 NNGGP COOX7C CYCYP2P2C6C65AABBCD3 CCAAT RPS1RPS1PS 9 AMAMMY2Y22A2A2 PPNNLNLIP S100AS1100A9 PLBB1 WARA S CCPCPAPPAA1 UGUGGTT2T2B32B334 CE CCTRRL CYYYP2YPP22BB1B10 ACACADCADD1D11 RRDHDDHH7 tion occurring both inside and outside CEELAA2A2A CELAELELAA1 UUGGT22AA3 NDDUDUFSF 8 CCEEL MRPS3MRRPSR 4 AMYAAMM 1 ACP5 HCLSCLS1 IVDECCHHDHDC2 SERSSEERERPIRRP NANA6 AACECE2 ENNPEEP CEELALA3A3B ST6GAL6GA 1 SSCP2

SYSYYCN CES22G TNKS1BPKS1 1 STTARRDR 5 CYP2D2CYCYPYPP222DDD226 of the GI tract as a reference for micro- CTTRRB1 SARSARS2 APPOAOAA1DDPDPYPYPYD PHB UBAP2UB L IGHG M MAPA K13 AIFM2 HSD3BHS 3 UGDU H FFYB SUULTT1DT HSH1SD3S 33BB3 AARRAAF TOMM2OOMMOM 2 MAN2BAN2 2 YBX1 CEES1ES1E SPPPARPARCLR L1 CASPASP7 GM4936493 8 HCFCCFC1 XPNPEPPPNNNPE 2 CNKSRCNNKSN 3 SARNP AMMY2A52 CCN3 IGKV5-3GK 9 EPEPPHXPHHXX2 MYL4 ACOCOT7 IGHGGHG1 CECES2SS22E VVNVNNNN1PDZKPDZKDZD 1 PRKCPRKR Q biome researchers interested in a given CDC42SEC42 2 PVVALLB ABRAXASRAX 2 CCEESES2S2G FYB1 EECI1 DUDUUOOX2 SARTART1 CYP44V3IIYD CCEESES1S1F CMPKCMMPM 2 22100110C04RIK MS4AS4A1 MEPEP1A NNNT IGHA WWDRR1R 8 PPLD3 CDV3 NDDRGG1 CCMBCMMBL WBSCR1BSC 6 host protein, organ system, or protein HEBPEBP1 AACAACS LY6D ORCRCR 2SRSFRSFS 6 CDA REREEG33G network. Associations of highly ranked WASASF2 PIGIGR LLRGLRRGRG1 DDQX1 GMGMM57577771 FFCGRGRT AATGG 3 EEROO1O1L CCELA2CEELAE A proteins were constructed into protein– WIPI2 SOOORBRBSRSBSBRGRGASGAPA3 2 LYYZZ1 HEART TRYRY5CCELACEELLA1 IGJ PPLSS1FFRRK CTRCTRTRL ARHHGGEF10L CTTTTN PLIN4 PRRSSSSS1 protein interaction networks for the GPAM INA MAPREAPR 3MMZBZB1 TXXXNDNDDCC5AACCCTRCTTTRRR110 CPAPA2CTTRBTRRBRB1 GNPAT NEFM PLIN5 FKKBPBPP1P 1 brain, spleen, heart, small intestine, and CWF19LWF1 1 CCPBB1 MCG_1509MCGM GG__15_ 5 CHMP1HMP A PSMDSMDSM 9 SSCIIN HHEXXB SERPINA6 CTDSPTDS 1 HPGD IGKV17-12V17 7 colon, as well as a global network. While GPLD1 CHMP4HMPM B PSMBSM 5 FHL2 TMEM23EM2 6 B33GAGALAAMT5 MY22A2 2 APOA1 DHX8 SPPPLL2L A

RCN1 IGKV1-11KV1 7 NMTN 1 PRPF38A NNT ENNTPDP 5 GRB7 TRIMRIM2 APOE FDPS the brains and gastrointestinal tract of GF HP DHX16 SCAMPCAM 3 FSTL1 CENPEN V MAP2KAP2 3 TM9SFM9S 1 GM2052M205 1 SEC11EC1 A PBLD1 FBP1 CD82 PRMTRM 5 CST3 NTAN1 AKR1C19 ACAT2 mice have been characterized in several ZNF62NF6 2 ERO1ERO A ANP32NP3 E DENND4NN B CCA3 EMC3 TBCB ENPPNPP7 BTNLBTN 1 MPP1 ABHDBH 6 ACTR3 FBXL18 NGP IMPACT TDP2 MAST4 SSIAE CHORDCHOR 1 IGLCGLC1 ARHGDIRHG B IGKV9-120KV9 GK CYP44V2 SUCLAUCL 2 ESS2 CLYBCLY L SULT1CULT 2 NAALADLAALA 1 RCC1RCC L CORO1A UBE2Z C530008M17RIK studies, other organs such as the heart KDM5A FAM210A IGKV1-117 GM43738 IGKV6-15 TLK1 LSM14A NDUFB3 FNTA and spleen may be of interest given the RASSF4 HMBS SNRK IGKV1-135 PRRC2B SLC16A1 ATPAF1 SDCBP2 FCRL5 ORC2 TBCE IGHV1-22 MYH6 emerging roles of the microbiota in ath- KRT86 PURA IGKV5-39 GATAD2B COLON ISYNA1 TERF2IP PRMT3 CTDSPL PAFAHH1B1 GNAQ erosclerosis (Karlsson et al. 2012) and im- ACAT3 POGZ SELENBPLEN 1 SSNDD1 LHPP MAOA FFXRXR2 ARHGEF1HG 7 HEBPEBP1 SERPINARP 6 ASPA N UBAP2BBAP L mune development (Chung et al. 2012). PROSPRO 1 HYDIYD N CARD1ARD 9 FFXRXR1 BRAIN SLC26ALC26 3 MUPUP8 We hypothesized that applying methods HEBPEB 1IRAKRAK1 AABI1 NCKCK1 for improved accuracy in quantitative MUP1UP 1 NNNT NNT WASFWAS 2 UBL4 EIF3JEEIF3EIIF 2 proteomics (Ting et al. 2011) would fur- SERPINA3RPI K MMCCCCC 2 EEIF33K LIIFR IYD ACADAACAAC L S100A00 5 ABCFAABCAB FAR1RFGAPG 3 RPS2PS226 ther define the influence of the micro- IGHGGH 1 PRPRRPFF3F31 EIF1 SRP7SRP 2 CDADCAD 1 MTGM 1 ASNS RBPRB 2 CD59D5 A RNNASSES 4 SSF3A3AA1 CPTPT1B SRPRP668 GOLGAOOLG 2AUPA 1 WDFYDF 1 COOPOPZ1 biota in each of these organs. With this, RBBPBB 5 FAABPBPP1 TAF15 EEEEF1EFF11BB2 HPX FRASRA 1 CEELA3A3B ANG4 MMRTO4 OSTC HSHSSPAPA1A B APAPOPPOAOOAA4 SSNNRRPRPF AMAMY2M 2A2 2 SYYMPK RRPLPPLLPP0 FAABPBPP3 DDNNANAAJJJCC3 HSHSSPAA5 SSSCASC 1 GSTZT 1 SSYCCN MMTXMTTXTX1 RRPLPRPLRPLP2 HSPPAA2 we hope to reveal microbiota-induced GYS1 RASGRFSSG 1 HDLDLBP SCGCGN CTRBB1 REEG3EGG33G FAABBPBP5 MMTCMTCHCH1 HHSSSPSPAPPAA1A L HSPA1HSHSSPA A CCPACPPA2 TIMTIMIMMMM5M50 UBAU 5 TATAGAGLGLNLN2 DNNNANAJAJAJBB111 AMT CCEEL PPRRSRSSSS1 GGPSS1 CTRRL MCG_G_15_151509505 5 UUROUROD P4H4HB HSHSSPBSPB1 LMOMO7 changes that could underlie disease CCPACPPPAA1 CKCKMTKMKMMTT1GGPLPLP D1 DNDNAJNAAJJJCC1C 0 GMM57M55777 1 GOOLMOLM1 HEHEBEBBPP2 TXTXXNDXNDCD 5 PPNLLIL PKKRTT1ATP6APATP6APP66 2 MZBMZ 1 PNLLIPRRPRP2 NDUFA1DDUF LLDH2 HB CCPBB1 KKRTRTT7 LALAMAMBMMBB1 states. Our results validate a body of liter- CES2S22A CEEELAA2A A PKP2 MMDHH2 AALPALLPLPI FKKBPP111 CWF19LWWF1 1 NDDDUFB4 RASGRF2 AAMMYMY1 KKRRTRT1T 8 SSYYP CYYPPP2CC5C 5 KRRTRT10 MIA3 CDC42SECCDC42 2CUTC TC ature in the field, provide visualization GSSTTTMM5 GGSTGSSTOT 1 MEPEP1A CHL1 MAPKAPKAAPPKA 3 CYPP3AA1A 3 AANOANNO6 AAGRAGR2 GGPXX2 AACCEE2 PIITRRMR 1 SSORORRD TGMGM3 XPNPEPXPNPN 1 tools for contextualizing the organism- CLCLCA1PPP1R1PPPP1 B Increased in Conventional MGGAGAM SIS GALLM NNLN SLLC15LCC155A1 ABAA T wide effects of microbial colonization, SLLC33A3 1 Increased in Germ-Free DPPD 4 AAIFF1 CHIA1 CCPQ RGGSS10 NNCF2 FAHD1 CTDSPTDSTD 1 and identify several new host-microbiota GLOOD5 CD20D200 NAALADLAAALAAA 1 OOBSCSSCCN NIPPSNAPN 2 TRIM1RIM 4 MUP2UP22 IGHGGHG1 NEBN L associations for further investigation. IGKV5-3KV5 9 MYYOMOMM2 Low High EMC7 AW551985519 4 DHHRSH S7S C ADH6DH6A CALMALM1 THEMH 4 Significance Significance OCIADCIA 2 HPGD CAADM3 ARSB

Results Figure 1. Organ-specific protein networks modulated by microbial colonization. Proteins significantly increased in either GF or conventional animals were analyzed for interactions through STRINGdb. Edges Construction of protein–protein in each node represent the combined score accounting for all interaction sources. The edges are sized by interaction networks the combined score, with the minimum threshold being 0.4 (of a maximum confidence 1). Nodes represent gene names of significant proteins with a minimum statistical cutoff of |π| > 1. Red indicates a sig- To determine the protein-level conse- nificantly higher presence in conventional mice and gray indicates the opposite. The nodes are sized by π quences of microbial colonization, three the level of significance as assessed by -score. biological replicates of five different tissues (brain, small intestine, colon, spleen, heart) were analyzed from conventionally raised or GF mice. sis (Fig. 2; Supplemental Table S2). To identify overlapping mech- Multiplexed proteomic analysis of tissue homogenates resulted anisms occurring throughout all organs, we compiled all highly in the quantification of 7752 proteins overall, of which 4663 ranked proteins within each organ (|π| > 1) into a single protein were quantified across all samples. These 4663 proteins were network (Fig. 3). used for downstream analysis. A separate pilot study of the brain Colonization state of the mouse appeared to have larger im- tissue resulted in the quantification of 6203 proteins. pacts on organs in direct contact with the gut microbiota, namely, On a per-organ basis, we contrasted the protein abundances the small intestine and colon. The GI organs analyzed had an av- of tissue collected from conventional mice against tissue from erage of 210 proteins associated with colonization state while the GF mice. We ranked the association of each protein to coloniza- three organs outside of the GI tract averaged 52. The brain yielded tion state by accounting for both significance level and fold- the lowest number of associated proteins with only 22. GI tract or- change (Supplemental Table S1). Interaction networks were built gans also displayed a higher percentage of interconnectivity, with in order to identify groups of proteins whose abundance was mod- an average of 71% of associated proteins within GI organs having a ulated by microbial colonization. We first constructed organ-spe- moderate-confidence connection to another associated protein cific protein interaction networks containing all the proteins within the organ, while organs outside the GI tract averaged with a highly ranked (|π| > 1) association with the colonization 39% (Fig. 1). We hypothesize that the interconnectivity and state of a given organ (Fig. 1). Functional enrichments within these number of associations with colonization is related to the direct organ networks were then assessed by gene-set enrichment analy- contact of GI organs with microbiota. However, it is possible

2 Genome Research www.genome.org Downloaded from genome.cshlp.org on October 8, 2021 - Published by Cold Spring Harbor Laboratory Press

Microbiome-impacted organ protein networks

SPLEEN SMALL INTESTINE Disulfide bond Peptidase S1A Disulfide bond Trypsin-like Peptidase S1A,chymotrypsin-type peptidases GO:0005615~extracellular space Proteolysis GO:0007586~digestion Secreted Immunoglobulin-like domain Pancreatic secretion SM00406:IGv Protein digestion and Protein digestion and absorption absorption mmu04972:Pancreatic secretion Iron binding IPR013106:Immunoglobulin V-set 0 5 10 15 IPR007677:Gasdermin -Log₁₀(Adj. p-value) Necrosis Metabolism of xenobiotics by cytochrome P450 mmu00340:Histidine metabolism HEART Lipid metabolism beta-amyloid mmu00140:Steroid hormone biosynthesis binding PIRSF000097:aldo-keto reductase glycerol-3-P O-acyltransferase mmu00983:Drug metabolism - other enzymes HDL mmu00982:Drug metabolism - cytochrome P450 high-density lipoprotein particle NADP positive regulation of triglyceride GO:0016491~oxidoreductase activity biosynthetic process Signal 10 5 0 5 10 IG -Log₁₀(Adj. p-value) Perilipin Immunoglobulin V-set Intermediate filament head, DNA-binding domain Keratin, type I COLON Phospholipid/glycerol Pancreatic secretion acyltransferase SM00406:IGv 1.5 1.0 0.5 0.0 -Log₁₀(Adj. p-value) Heat shock protein 70 family Protein processing in ER digestion BRAIN Stress response Heat shock protein 70 Peptidase S1A, chymotrypsin-type ER chaperone complex endoplasmic reticulum lumen Regulation of Ras protein Secreted signal transduction IPR007110:Immunoglobulin-like domain mmu03320:PPAR signaling pathway Extracellular region IPR000463:Cytosolic fatty-acid binding Metalloprotease GO:0008237~metallopeptidase activity 0.0 0.2 0.4 0.6 0.8 Lipocalin/cytosolic fatty-acid binding protein domain -Log₁₀(Adj. p-value) IPR011038:Calycin-like IPR012674:Calycin GO:0005215~transporter activity Increased in Conventional 42024 Increased in Germ-Free -Log₁₀(Adj. p-value)

Figure 2. Functional enrichments associated with microbial colonization within each organ system. Proteins significantly increased in either GF or conventional animals within each organ were analyzed for functional enrichments using DAVID. All proteins identified within the experiment were used asa background. Displayed are bar plots showing the −Log10(adj. P-values) associated with selected functional groupings. Benjamini–Hochberg correction was applied to account for multiple hypothesis testing. The bars are plotted in red if they are associated with the proteins enriched among conventional mice and gray if they are associated with GF mice. that including a higher portion of intestinal tissue may have influ- this was the larger portion of proteins associated with GF status enced these results. within the small intestine (66%) than in the colon (42%) (Fig. 1). Other studies identified similar results at a pathway and individual protein level. After a literature search for protein or transcript Validation of protein networks through previously reported differences within GF models, we identified seven publications associations with related findings (Supplemental Table S3). In brief, changes Our networks provided support for previously reported broad-scale in xenobiotic metabolism were reported in the GI, liver, and effects on GI organs, as well as abundance shifts from specific tran- kidney, both at the transcriptional level (Fu et al. 2017) and scripts or proteins. As shown in a previous proteomic study the protein level (Kuno et al. 2016). Our networks also highlighted (Lichtman et al. 2016), the small intestine and colon displayed dis- changes in glutamate-related proteins which were previously tinct changes as a result of microbial colonization. One example of reported at the transcriptional level (El Aidy et al. 2013). We also

Genome Research 3 www.genome.org Downloaded from genome.cshlp.org on October 8, 2021 - Published by Cold Spring Harbor Laboratory Press

Mills et al.

CASP7 PRPF31 DHX8 SART1 DNAJC10 MZB1 CASP7 FSTL1 DNAJB11 SNRPF B3GALT5LCN2 SPARCL1 HSPA5 MIA3M YBX1 HSPA1B DNAJC10 MZB 1 LTF RCN1CCNN1 FABP1P4HB SYMPK HPX GOLM1 HSPA2 NUP153 DHX16PRPF38A CST3 DNAJB11 APOEAAPPOPOE REG3G HSPA1L HSPA5 LAMB1LALA APOA4 WDR18 SRSF6 GM5771 DNAJC33 HSPA1A ZFP622 SF3A1 HSPB1 9 LYZ1 NR3C1 MRTO4 HSPA1B ERO1LE 1L AAPOA1 HEBP2 EIF3K UBE2Z LMO7 FABP1 GPS1 TRY5 FRK P4HB CDA HP HDLBP RPS19 RPLP2 HSPA2 NUP153 EIF3J2 RPL17 FBXL188 FABP5 UBA5 RPLP00 2 SRP68 RPS26 FTL1 HSPA1L W PSMD9 PRSS1 LRG1 TXNDC5 APOA4 HSPA1A SRP72 OSTC HSPB1 HEXB PSME1 PSMB5 3 MAPKAPK3 AKR1C14C ACTR10ACCTCTR1R10 CELA11 AKR1C12 AKR1C13 10 PKP2 PAFAH1B1 RBP2P MAP2K3 POA1 CELA2A CELA3B CCPA2 MAPK13 DPYD EIF3K RRM2 Stress Response SYCN AKR1C19 MTHFD1 TYMS NEFM CTRL COPZ1 ARFGAP3 CMPK2 CPA1A 8 ORC2 RRM1 CPB1 CEL NCK1 GOLGA2 PNLIPRP2PNLIP CD59A CST3 PGRMC1 ABI1 REG3G MCG_15095 RETSAT ACTR3 KCNAB2K CTRB1 DGAT1 ALDH1A1 WASF2 GM5771 DN RDH11 TFRC KRT1 RDH7 SIS ATP6AP22 CTTN LYZ1 ANO6 KRT10 DHRS3 MAOA MGAMM AM ATP6V1F KRT18 11 ACE2 ALDH1A7 AMY1 AGPAT2 HCLS1 KRT7 HEBP2 AKR1B7 MAOB GPLD1 KRT86 XPNPEP2 TRY5 FRK DPP4 LCT CDA GPAM HP MEP1A LY6D GALM ALPI GNPAT7 TREH 4 FECH HMBS FABP5 GALK1 UGT2B34 VNN1 SORD DCXR CYP2D26

EPHX2 UGDH UROD PIGR SCP2 CYP2C65 CBR3 ABCD3 CAT CYP2B10 UGT2A3 GSTM5 GPX2 CYP2C55 LRG1 FTL1 CES1E T CTH ABAT GSTA1 CYP3A13 PRSS1 NNT LDHB GCLC NDUFA12NDUFB4 GSTO11 5 GSTA4 AMT GLS CNDP2 ALDH1A3 ALOX15 NDUFS8 NDUFB3 HEXB COX7C NDUFA8 GLUD1 SUCLA2 HINT1 PTP4A1 MDH2 CHMP4B CHMP1A ACAT2 PTPN23 SARS2 6 AACS AMY2A2 PCK1 Innate Immunity CARS MYL4 OBSCN WARS MYH6 TIMM50 TOMM22

FXR2FXR1PSAT1 ASNSPRMT3PRMT5 LHPPPPA2 FNTA IVD MCCC2 EMC3EMC7 WIPI2 ATG3 ARAF PHB FAHD1GSTZ1 MECRACADL FDPS

TREH CELA1 UGT2B34 PKP2 GALK1 CYP2D26 CELA2A CELA3B CPA2C CYP2C65 UGDH CBR3 SYCN CYP2B10 UGT2A3 GSTM5 CTRL GPX2 CYP2C55 CPA1A CES CEL GSTA1 CYP3A13 CPB1 GSTO11 GSTA4

PNLIPRP2PNLIP ALOX15 MCG_15095 Xenobiotic metabolism with Cytochrome P450s CTRB1 DG

SIS AN Other Functional Groups Digestive enzymes from pancreas 1. Spliceosome 9. Translation 2. Proteosome 10. Nucleotide Color Bar Graph Size 3. Antigen processing biosynthesis 4. Protein digestion 11. Keratin Brain Small Signiﬁcant in 5. Mitochondrial respiratory Intestine Increased in 1 Organ chain NADH dehydrogenase Colon Conventional 6. Mitochondrial glutamate metabolism Spleen Increased in Heart Signiﬁcant in 7. Neutrophil degranulation Germ-Free 5 Organs 8. Bile acid synthesis

Figure 3. Combined organ protein networks modulated by microbial colonization. Proteins significantly increased in either GF or conventional animals were analyzed for interactions through STRINGdb. Edges in each node represent the combined score accounting for all interaction sources. The edges are sized by the combined score, with the minimum threshold being 0.8 (of a maximum confidence 1). Nodes represent gene names of proteins with a highly ranked association (a minimum statistical cutoff of |π| > 1) within at least one organ. Nodes are sized by the number of organs with which the protein had a strong association. The level of association of each node to a particular organ is colored according to the fraction of the total |π|-score each organ contributes. Below each node is a bar plot of the π-scores for each organ within the node. Putative functional groupings within the network are highlighted. Select sections are highlighted in colored boxes and shown in 2× zoom.

4 Genome Research www.genome.org Downloaded from genome.cshlp.org on October 8, 2021 - Published by Cold Spring Harbor Laboratory Press

Microbiome-impacted organ protein networks report the regulation of innate immune proteins including the an- lated to mitochondrial glutamate metabolism and mitochondrial timicrobial peptide REG3G (Larsson et al. 2012) and the regulation respiratory chain NADH dehydrogenase, with most of the related of NNT (Mardinoglu et al. 2015), which will be discussed further proteins down-regulated in the small intestine of conventional below. mice (Fig. 3). In addition, proteins relating to the mitochondrial reduction of glutathione, GLUD1, and GLS had confirmed associa- Organ-specific network results tions with previous transcriptomic analysis (Supplemental Table S3). While this subnetwork was largely derived from small intes- Functional enrichment analysis of GI tract organs resulted in tine proteins, there was evidence that this system may also be af- stronger and more diverse associations with colonization status fected in the brain through AMT, which is involved in the than organs outside the GI tract. Several enrichments emerged mitochondrial metabolism of glycine. with potential links to redox shifts in the small intestine. These The protein networks identified generalized differences included disulfide bonds, oxidoreductase activity, NADP, and among all organs related to the innate and adaptive immune sys- xenobiotic metabolism through cytochrome P450s (Figs. 2, 3). tems. Proteins related to innate immunity tended to be increased Functional differences in the colon highlighted pancreatic secre- in conventional mice, while proteins associated with neutrophil tion, immunoglobulins, proteins of the heat shock protein 70 fam- degranulation were associated with GF mice (Fig. 3). REG3G, a pro- ily, digestion, and stress as the functions increased in conventional tein associated with toll-like receptor (TLR) signaling subsequent mice (Figs. 2, 3). GF colons had enrichments for transporter activ- to pathogen-associated molecular pattern (PAMP) activation in ity, Calycin, fatty-acid binding, metalloproteases, and peroxisome Paneth cells, was increased among conventional mice. Additional- proliferator-activated receptor (PPAR) signaling (Fig. 2). Together, ly, proteins related to antigen processing were moderately en- these results may indicate stress response as an important factor riched in GF mice. mediating host-microbiota interactions in the colon and redox We next evaluated protein relationships to GF status at an or- states influencing interactions of the small intestine. ganismal level by applying a compositionally aware multinomial Organs outside of the GI tract have been less studied in regard regression technique. This technique accounts for organ type to as- to their regulatory responses to the microbiome. Within the sess organism-wide protein associations through ranks. Our top- spleen, proteins increased in conventional mice were found to ranked protein associated with conventional status was NNT, be part of a highly connected functional protein network consist- while the protein with the strongest association with GF status ing primarily of pancreatic digestive enzymes (CELA2A, CELA3B, was IGKV5-39, a protein involved in immune response and immu- CELA1, CPA1, CPA2, CTRB1, CTRL, CTSE, TRY5, etc.), iron-bind- noglobulin production (Fig. 4A). As observed from the traditional ing proteins (FST1, TFRC, STEAP3, FECH, LTF, and HP) and innate statistical approach, NNT was significantly higher in conventional immune mediators (LCN2, S100A9, NGP, ITGAM, and CHIL3) mice within all organs included in the regression (Fig. 4B). IGKV5- (Figs. 1, 3). Many of the digestive enzymes have roles in the GI tract 39 was more strongly associated with the spleen, colon, and heart and were similarly associated to colonization state within the small than the small intestine (Fig. 4B). intestine and colon (Fig. 3). Iron-binding and immune proteins Next, we assessed the 150 top- and bottom-ranked proteins were similarly regulated in other organs (Fig. 3), but the differential associated with GF status from the multinomial regression and cre- regulation of LTF (Lactotransferrin) and FSLT1 (Follistatin-related ated a protein–protein interaction network (Fig. 4C). Of interest protein 1) were primarily restricted to the spleen. Given the biolog- was a cluster of proteins related to redox states in mitochondria. ical roles of the spleen, the influence of these proteins in mediating Several proteins including ECSIT, NDUFS4, NUBPL, NDUFA11, immune processes may be of significant interest. NDUFB6, NDUFA8, and ATPAF1 are all related to mitochondrial The networks associated with the brain and the heart dis- complex 1 of the electron transport chain. Other evidence of played fewer interconnected proteins. Only 34% and 27% of pro- the organism-wide impact of microbial colonization on mitochon- teins had a connection within the heart network and brain dria included shifts in mitochondrial ribosomal proteins (MRPS23, network, respectively. However, a group of primarily GF-associated MRPL50, MRPS17, MRPL49, and MRPL17) and proteins related to proteins that included APOA1 and APOE was found among the mitochondrial heat shock response (HSPD1, HSPE1, SOD2, heart proteins (Fig. 1). These results may suggest changes in lipid LONP1). The data presented here suggest that microbial coloni- profiles in the heart: increased high-density lipoproteins (HDL) zation may have organismal-level impacts on mitochondrial and chylomicrons in GF compared to conventional mice. The function. brain showed limited functional enrichments (Supplemental Table S2). However, both RASGRF1 and RASGRF2 were increased among conventional mice. These proteins may be of interest given Interactive 3D visualization of associations with colonization status implications in glutamatergic excitatory synaptic signaling and in To encourage user interaction with our data, we created a web- facilitating long-term potentiation leading to enhanced memory, based display of our results projected onto a 3D model of a mouse. learning, and synaptic plasticity (Drake et al. 2011; Schwechter This interactive display allows for user-guided exploration of the et al. 2013). protein and pathway-level associations with colonization status. To access the model, users should access the ‘ili web server (https Organism-wide network results ://ili.embl.de/), then drag-and-drop the provided texturization One prominent finding from our generalized network was a signif- file (Supplemental File S1) and a supplemental table for either icant increase in NNT in all conventional tissues. This protein is a the proteins (Supplemental Table S4) or pathways (Supplemental key regulator of generalized biosynthetic processes and is related to Table S5) identified in this study. With this tool, users can search glutamate synthesis (Mardinoglu et al. 2015). Statistically, NNT for proteins and pathways of interest and project the association was among the strongest relationships found within all organs an- scores onto all the organs analyzed in this study. alyzed (π = 3.8, 4, 2.3, 7, and 3.7 for small intestine, colon, spleen, For the protein-level visualization, each protein is listed by heart, and brain, respectively). We also identified subnetworks re- the protein name with any Gene Ontology (GO) molecular

Genome Research 5 www.genome.org Downloaded from genome.cshlp.org on October 8, 2021 - Published by Cold Spring Harbor Laboratory Press

Mills et al.

ABNNT the strong association with conventional status for NNT within all organs.

NNT The pathway-level visualizations help to summarize and explore the gene NNT set enrichment analyses. On a per-organ basis, an association with conventional or GF status for each functional category was calculated by comparing the statistical strength of gene set enrichments (Supplemental Table S2). These asso- IGKV5-39 ciation scores have been summarized in an ‘ili-compatible file in Supplemental Table S5. Pathway association scores IGKV5-39 ranged from 12.90 (highly associated with conventional status) to −7.25 (high- IGKV5-39 ly associated with GF status). We demon- strate the use of this tool to visualize the associations with “Oxidoreductase” in Figure 5B.

ATTP5P5F5 1 C WDDRDR4R 3 UTPUTTPT 6 Mitochondrial Discussion NONOOCC3C3L NUBPLNDNDDUUFB3 Complex 1 HEEEATATTRT 1 Our multiorgan analysis was utilized to UTTP1T 111 LSURFSUSURU 6 NDUFSNNDDDUF 4 NDDUUFSFS8 NDUNDNDUDUUFFA1FAFA1A11 IPPOO4 NDNDUNDUDUFUFAFFFAAFAF7 generate protein interaction maps and PDPDHDHB NDUFANDDUFU 3 NDDUFFAFA8 3D visualization tools that researchers NDDDUFUFAF 5 NDDDUFB6 PRRKAKAAAA 2 can use to understand organ-specific Splicing PDKDK1 PCCX SSSB MTHFDMTTHFT 1Innate RRAGRAR A and organism-wide changes that may un- SNRNP7SNRNRRNR 0 LRRRGG1 Immunity SNNNRRPRPF UGT2AGT2 3 derlie host-microbiome interactions. SODOD2 PRRSRSSS 1 MRRRPPL4P 9 REG3REEG3E G RPPPS2S2S 0 SFF3A33AA1 DHHXX1X16 CCDA From our networks, we can identify com- RPPPLL3L LMRMRRPPLPL1L 7 MRRPS17 HSSPPEPE1 HNRRNNPANNPA0 TTRYY5 SSIS mon themes found from previous SRSFSSRRRS 6 GGLA UGDGGDH MMRPL5MRRRP 0 NPPPCC2 RPL2R L2223AA-PSA-P 3 SFFF3A3AA2 GAAALLM analyses of GF tissue. For example, we RPPS2S2S26 HSSPDD1 GGALLT DCCPP1P A MRRPS23 Mitochondrial found changes in innate and adaptive im- RPS1RPPPS10 LONPOON 1 heat shock PSSMDD1 NFKBNFFFK 2 LCCT RPPPL3PLL373 A mune responses to microbial coloniza- Mitochondrial AARF5 PSMPSMMDD1D14 SYCN TRAAPPPC6B OSTC Ribosome YKYKT6 tion (Larsson et al. 2012), changes in the COOPPZP 1 TRAAPPPC1P 0 BCL2L1CCL2L 4 PSMESM 4 CEELAA3A B glutamine and glutamate pathway (El DDDOOSO T PPLDLD1 TSPPAPAN1A 4 ARHGAP1HG 5 CTRT B1 RANBPRAANBA 9 TRAPPC1AAPP 2 TERFER 2 Aidy et al. 2013; Mardinoglu et al. NNRARAS CCTTRT L KCKCNAC ABA 2 SUZ12 RRHOHOF 2015), and changes in xenobiotic degra- GOOLM1 CKKAPK P4 SYNESYYYN 2 NBEALBBEA 2 TBLL1XRR1 DOOK3 ARHGEFRHHGH 7 dation pathways (Kuno et al. 2016; Fu SPHKPPH 2 CAALMA MLM 4 HISH ST2HH2AAH 1 NOOOSS1 PTTKTK2 CAMKAAM 4 et al. 2017; Kindt et al. 2018). This study PPAP2PPPAPA B CCTTTN GRBGRRBR 7 DMMD DIAPP2 contributes to the field by moving toward NUP6NUUUP 2 Increased in LPL NUPUP1P101 7 ITTGAA1 Conventional an organism-wide understanding of host- SCCARB2 MYHMYYYH8 HSPA12PPA B TPPMPM1 Increased in microbiome interactions. Here, we incor- Germ-Free BAGAG3 porate new statistical and visualization Figure 4. Organism-level protein networks modulated by the microbiome. A multinomial regression tools for multiorgan analysis and include controlling for organ and microbial colonization state of the mice was used to assess proteins associated understudied organs from outside of the with colonization status. (A) Proteins ranked by regression coefficient; proteins with coefficients of the GI tract. As roles for the microbiome are greatest magnitude are most associated with colonization status. Proteins with positive coefficients are expanding into immunity (Thaiss et al. more abundant in conventional mice, while proteins with negative coefficients are more abundant in 2016), cardiovascular disease (Ahmad- GF mice. (B) Log abundance of NNT or IGKV5-39 over the entire proteome in each organ. (C) Protein–protein interaction networks from the top-ranked proteins from the multinomial regression as- mehrabi and Tang 2017), and mental sociated with both conventional and GF status when controlling for organ and mouse. The top 150 pro- health disorders (Nguyen et al. 2018), de- teins associated with both GF and conventional status were analyzed (300 proteins total), and proteins fining the influence of the microbiome with high confidence interactions (0.8) are shown. Nodes are sized by the absolute value of the regres- on these tissues may be of importance. sion coefficient and colored by association with GF (gray) or conventional (red) status. Putative functional groupings are indicated. Indeed, our networks provided several putative organism-level roles for the microbiome. function terms associated with the protein. With this, users are Our protein networks highlight the potential of organism- able to search for a protein of interest or identify proteins of inter- wide redox state changes being linked to the microbiome. This is est by molecular functions. The ‘ili-compatible table of association perhaps best highlighted in the distal increases of NNT associated scores (Supplemental Table S4) uses the π-statistic for convention- with microbial colonization. NNT is a mitochondrial protein with al/GF status. These scores range from 6.98 (significantly increased well-described roles in glutathione redox reactions through the in conventional mice) to −5.11 (significantly increased in GF conversion of NADPH to NADP+ (Ronchi et al. 2013). mice). An example use case is displayed in Figure 5A, which depicts Glutathione is interconverted with glutathione disulfide,

6 Genome Research www.genome.org Downloaded from genome.cshlp.org on October 8, 2021 - Published by Cold Spring Harbor Laboratory Press

Microbiome-impacted organ protein networks

A portance given the increasing literature regarding microbiota regulation of lipids and lipoproteins including HDL (Nakaya and Brain Spleen Colon Ikewaki 2018), as well as the microbial links to metabolites leading to atherosclerosis (Wang et al. 2011b). Our findings may indicate a mechanism in which microbiota influence the makeup of the Heart Small heart through changes in lipoproteins. Intestine The data of the spleen proteomes revealed networks of prote- ases with similar increases within the small intestine and colon in the presence of the microbiota. These networks also suggested potential links between the gut microbiota and pancreatic secretion. Protein-level Pancreatic secretion of enzymes is thought to be largely regulated Associations GF- Conventional- Associated Associated through circulating hormones (Singh and Webster 1978). The gut B microbiome has several potential links to pancreatitis and pancre- Brain Spleen atic cancer, which may be mediated through sensing of microbial Colon compounds such as lipopolysaccharide through TLR4 (Leal-Lopes et al. 2015). Here, we have observed a link between microbial colonization and increased pancreatic secretion of digestive proteins, Heart Small which may be of interest for further investigation. Intestine The organism-wide effects of microbial colonization illustrat- ed in our networks may have implications in several disease states. Dysregulation of complex 1 has been implicated in microbiome- related diseases, including ulcerative colitis (Haberman et al. 2019), and there is accumulating evidence of mitochondrial dys- Pathway-level GF- Conventional- Associations Associated Associated function in Crohn’s disease (Mottawea et al. 2016). This association between the microbiome and mitochondria are thought to Figure 5. Interactive 3D visualization of associations with colonization be mediated through three key microbiome metabolites: short- status. A 3D mouse model was generated for use on the web-based ‘ili plat- form (https://ili.embl.de/). (A) An example use case for the protein-level as- chain fatty acids, the urolithins, and lactate (Franco-Obregón sociation visualizations shown through plotting the π-score enrichment for and Gilbert 2017). While speculative, it is possible that the IBD conventional colonization status. (B) An example use case for the pathway microbiota may influence these interactions. Glutamatergic sig- level association visualizations shown through highlighting the enrichment naling has been suggested as a potential target for treating mood scores for “Oxidoreductase.” Pathway association scores were generated disorders (Zarate et al. 2010). Our identification of proteins in through –Log10(Benjaminin–Hochberg corrected P-values) of the conventional organs minus the GF organs (from Supplemental Table S2). the brain related to glutamatergic signaling and glutamate (i.e., RASGRF proteins and NNT) may be of relevance to the discussions surrounding utilization of the microbiome to treat mood disorders collectively representing the most abundant redox pair in the body (Mangiola et al. 2016). (Circu and Aw 2011). Abundances of this redox pair are often used There are likely many unknown mechanisms mediating host- as an indication of the general redox state (Circu and Aw 2011), microbiota interactions. Our detailed maps and visualization tools which is involved in a large variety of biological processes (Circu for understanding the organ-level impacts of microbial coloniza- and Aw 2011; Birben et al. 2012; Ray et al. 2012). Evidence of intes- tion give insight into the unique and common protein-level tinal redox differences in GF animals has been known since the changes occurring throughout mice. These networks suggest 1970s (Koopman et al. 1975; Celesk et al. 1976), and evidence of changes throughout the mice related to mitochondrial dysfunc- microbiota-dependent regulation of Nnt in the GI tract has been tion and redox states. Though fewer changes were found in organs previously described (Mardinoglu et al. 2015). Here, we find evi- apart from the GI tract, our protein networks of the spleen, brain, dence of increased NNT throughout the entire conventional and heart may provide insight for researchers establishing connec- mouse. Additionally, both the traditional and multivariate statisti- tions between the microbiota and diseases related to these organ cal methods used for the identification of organism-wide effects systems. We hope these networks and visualization tools may be found networks of proteins related to mitochondria, including mi- useful in the microbiome research community to help dissect tochondrial complex 1 proteins. Mitochondrial complex 1 activity the specific effects that microbial communities have in a given or- might also be linked to redox states as complex 1 activity was gan system, protein, or protein network of interest. In total, we shown to be dependent on glutathione transport into the mito- view our study as a step toward better understanding the role of chondria (Kamga et al. 2010). In summary, many of our protein the microbiota in health and disease. networks may have been influenced by redox states, including shifts in mitochondrial proteins, the innate immune processes such as neutrophil degranulation, and degradation of xenobiotics Methods through cytochrome P450s (Kramer and Darley-Usmar 2015). Though our strongest relationships to microbial colonization Gnotobiotic mice were found within the GI tract, which was, unsurprisingly, the fo- Three male GF C57BL/6 mice were kept under GF conditions in a cus of most previous GF organ analyses (Larsson et al. 2012; El Aidy Park Bioservices isolator in Georgia State University’s (GSU’s) GF et al. 2013; Mardinoglu et al. 2015; Lichtman et al. 2016), the anal- facility, and three male conventional C57BL/6 were kept in regular yses of the brain, spleen, and heart did yield several interesting re- housing at GSU’s animal facility. At 5 mo of age, mice were eutha- sults. Within the heart, our analyses indicated a relationship to nized and organs were collected followed by immediate snap-freez- HDL and the GF state. Our results within the heart may be of im- ing. All mice were bred and housed at Georgia State University,

Genome Research 7 www.genome.org Downloaded from genome.cshlp.org on October 8, 2021 - Published by Cold Spring Harbor Laboratory Press

Mills et al.

Atlanta, Georgia, USA, under institutionally approved protocols at room temperature and was quenched by adding 8 µL of 5% hy- (IACUC # A14033). droxylamine (Sigma-Aldrich). TMT-labeled peptides from each of the organ samples were acidified by adding 50 µL of 1% TFA and subsequently combined into a composite sample per TMT 10- Protein digestion and TMT labeling plex experiment. During the pilot experiment with the brains, All organ proteome methods were performed as previously de- the pooled samples were desalted and fractionated using a Pierce scribed (Lapek et al. 2018). Snap-frozen organs (stored at −80°C High pH Reversed-Phase Peptide Fractionation kit (Thermo Fisher beforehand) were suspended in PBS and homogenized using a Scientific) per the manufacturer’s instructions. For the larger study, Mini BeadBeater (Biospec). Our final analysis contained three samples were pooled per 10-plex experiment, desalted with C18 biological replicates of each colonization condition per organ, Sep-Paks, and further fractionated as described below. and no technical replicates were collected given strong correlation observed between biological replicates in our past work (Lapek Collection of LC-MS2/MS3 spectra for protein identification et al. 2018). Organ homogenates were lysed in 1 mL of buffer com- and quantification posed of 75 mM NaCl (Sigma-Aldrich), 3% sodium dodecyl sulfate (SDS, Thermo Fisher Scientific), 1 mM NaF (Sigma-Aldrich), 1 mM Data acquisition methods were performed as previously described beta-glycerophosphate (Sigma-Aldrich), 1 mM sodium ortho- (Lapek et al. 2018). Sample fractionation, excluding the brain ex- vanadate (Sigma-Aldrich), 10 mM sodium pyrophosphate periment, was performed by basic pH reverse-phase liquid chroma- (Sigma-Aldrich), 1 mM phenylmethylsulfonyl fluoride (PMSF, tography with concatenated fractions as previously described Sigma-Aldrich), and Complete Mini EDTA-free protease inhibitors (Wang et al. 2011a). Briefly, samples were resuspended in 5% for- (1 tablet per 10 mL, Roche) in 50 mM HEPES (Sigma-Aldrich), pH mic acid/5% acetonitrile and separated over a 4.6 mm × 250 mm 8.5 (Villén and Gygi 2008). To ensure full lysis, homogenates were C18 column (Thermo Fisher Scientific) on an Ultimate 3000 passed through a 21-gauge syringe 20 times. Insoluble debris was HPLC fitted with a fraction collector, degasser, and variable wave- then pelleted by centrifugation for 5 min at 14,000 rpm. length detector. The separation was performed over a 22% to 35%, Supernatants were transferred to new tubes, and an equal volume 60-min linear gradient of acetonitrile in 10 mM ammonium bicar- of 8 M urea in 50 mM HEPES, pH 8.5 was added to each sample. bonate (Thermo Fisher Scientific) at 0.5 mL/min. The resulting 96 Samples were then vortexed and further lysed through two 10- fractions were combined as previously described (Wang et al. sec intervals of probe sonication at 25% amplitude. 2011a). All fractions were dried under vacuum and resuspended Proteins were reduced with dithiothreitol (DTT, Sigma- in 5% formic acid/5% acetonitrile and analyzed by liquid chroma- Aldrich) and alkylated with iodoacetamide (IAA, Sigma-Aldrich) tography (LC)-MS2/MS3 for identification and quantitation. (Haas et al. 2006). Proteins were next precipitated via methanol- All LC-MS2/MS3 experiments were carried out on an Orbitrap chloroform precipitation (Wessel and Flügge 1984). Precipitated Fusion (Thermo Fisher Scientific) with an in-line Easy-nLC 1000 proteins were resolubilized in 300 µL of 1 M urea (Thermo Fisher (Thermo Fisher Scientific) and chilled autosampler. Home-pulled, Scientific) in 50 mM HEPES, pH 8.5. Proteins were digested in a home-packed columns (100 µm ID × 30 cm, 360 µm OD) were used two-step digestion process. First, 3 µg of LysC (Wako) was added for analysis. Analytical columns were triple-packed with 5 µm C4 to each sample, and samples were digested overnight at room tem- resin, 3 µm C18 resin, and 1.8 µm C18 resin (Sepax) to lengths perature. Next, 3 µg of trypsin was added, and samples were digest- of 0.5, 0.5, and 30 cm, respectively. Peptides were loaded at 500 ed for 6 h at 37°C. Digests were acidified with Pierce Trifluoroacetic bar and eluted with a linear gradient of 11% to 30% acetonitrile Acid (TFA, Thermo Fisher Scientific) to quench the digestion reac- in 0.125% formic acid over 165 min at a flow rate of 300 nL/ tion. Peptides were desalted with C18 Sep-Paks (Waters) as previ- min, with the column heated to 60°C. Nano-electrospray ioniza- ously described (Tolonen and Haas 2014). Concentration of tion was performed by applying 2000 V through a stainless-steel desalted peptides was determined using a Pierce Quantitative T-junction at the inlet of the microcapillary column. Colorimetric Peptide Assay (Thermo Fisher Scientific), and pep- The mass spectrometer was run in data-dependent mode, tides were aliquoted into 50-µg portions. Aliquots were dried un- where a survey scan was performed over 500–1200 m/z at a resolu- der vacuum and stored at −80°C until they were labeled with tion of 60,000 in the Orbitrap. Automatic gain control (AGC) was TMT reagents. set to 2 × 105 for MS1 with a maximum ion injection time of 100 Peptides were labeled with 10-plex TMT reagents (Thermo ms. The S-lens RF was set to 60, and centroided data was collected. Fisher Scientific) (Thompson et al. 2003; McAlister et al. 2014) as Top-N mode was used to select the most abundant ions in the MS1 previously described (Wang et al. 2011a). TMT reagents were recon- scan for MS2 and MS3 with N set to 10. stituted in dry acetonitrile (Sigma-Aldrich) at 20 µg/µL. Dried pep- The decision tree option was used for MS2 analysis, using tides were resuspended in 30% dry acetonitrile in 200 mM HEPES, charge state and m/z range as qualifiers. Ions carrying two charges pH 8.5, and 7 µL of the appropriate TMT reagent was added to pep- were analyzed from the m/z range of 600–1200, and ions carrying tides. Reagents 126 and 131 (Thermo Fisher Scientific) were used to three and four charges were selected from the m/z range of 500– label peptide aliquots composed of an equal concentration of every 1200. An ion intensity threshold of 5 × 104 was used. MS2 spectra sample within all mass spectrometry (MS) runs. These composite were obtained using quadrupole isolation at a 0.5 Th window samples acted as a reference within all mass spectrometry runs and fragmented using Collision Induced Dissociation with a nor- for data normalization purposes described later. Remaining re- malized collision energy of 30%. Fragment ions were detected agents were used to label samples in random order with no bias re- and centroided data collected in the linear ion trap using rapid garding animal of origin, organ, or colonization status. This scan rate with an AGC target of 1 × 104 and maximum ion injec- randomization was performed to prevent known batch-effects in tion time of 35 ms. mass spectrometry experiments (Brenes et al. 2019). The brain sam- MS3 analysis was performed using synchronous precursor se- ples were analyzed as a pilot experiment before the other organs. lection (SPS) enabled to maximize TMT quantitation sensitivity For brains, all procedures were performed as above, though the (McAlister et al. 2014). A maximum of 10 MS2 precursors was spec- TMT experiment was separate from other organs. Within the brain ified for the SPS setting, which were simultaneously isolated TMT experiment, one channel, 129C, consisted of a 50-µg average and fragmented for MS3 analysis. Higher-Energy Collisional of all peptides from brain samples. Labeling was carried out for 1 h Dissociation fragmentation was used for MS3 analysis with a

8 Genome Research www.genome.org Downloaded from genome.cshlp.org on October 8, 2021 - Published by Cold Spring Harbor Laboratory Press

Microbiome-impacted organ protein networks normalized collision energy of 55%. Resultant fragment ions (MS3) on the combined evidence scores, with thicker edges indicating were detected in the Orbitrap at a resolution of 60,000 with a low higher confidence. Per-organ networks were performed with a me- mass cut-off of 110 m/z. AGC for MS3 spectra was set to 1 × 105 dium minimum confidence (0.4) to visualize connectivity with a maximum ion injection time of 100 ms. Centroided data through maximizing potential connections, while combined or- were collected, and MS2 ions between the range of 40 m/z below gan networks utilized a high minimum confidence (0.8) to identi- and 15 m/z above the precursor m/z were excluded by SPS. fy putative functional groupings. Functional enrichment analysis was performed through the DAVID server (Huang da et al. 2009a,b) to identify significant Data processing and normalization groups of proteins per organ, split between GF and colonized Data were processed using Proteome Discover 2.1 (Thermo Fisher states. Parameters were set as previously described (Nicolay et al. Scientific). MS2 data were searched against UniProt mouse databas- 2015). Benjamini–Hochberg corrected P-values were reported for es (downloaded 7/2/2018 and 5/11/2017 for the brain analysis and the most significant groupings. Bar plots were visualized through other organs, respectively) using the SEQUEST algorithm (Eng GraphPad Prism (version 7.0b). et al. 1994). A decoy search was also conducted with sequences Songbird (Morton et al. 2019) was used to implement the in reverse order (Peng et al. 2003; Elias et al. 2005; Elias and Gygi multinomial regression analysis with organ, mouse, and coloniza- 2007). A precursor mass tolerance of 50 ppm (Beausoleil et al. tion status used in the regression formula. Model parameters were 2006; Huttlin et al. 2010) was specified and 0.6 Da tolerance for as follows: 10,000 epochs, batch size of 5, differential prior of 1.0, MS2 fragments. Static modification of TMT 10-plex tags on lysine learning rate of 0.001, gradient clipping size of 10, and proteins and peptide n-termini (+229.162932 Da) and carbamidomethyla- with >5 counts were included. As the brain tissue was processed tion of cysteines (+57.02146 Da) were specified. Variable oxidation as an independent pilot study, we could not include this data as of methionine (+15.99492 Da) was also included in search param- part of the multinomial regression analysis given the independent eters. Data were filtered to 1% peptide and protein level false dis- normalization used in each experiment. covery rates with the target-decoy strategy through Percolator (Käll et al. 2007; Spivak et al. 2009). 3D mouse model TMT reporter ion intensities were extracted from MS3 spectra The 3D mouse model was generated as described previously for quantitative analysis, and signal-to-noise values were used for (Quinn et al. 2019). In brief, MRI images were acquired at the quantitation. Spectra were filtered and summed as previously de- UCSD Center for Functional MRI from a euthanized 8-wk-old scribed (Lapek et al. 2017b). Data were normalized in a multistep female C57BL/6 mouse with a Bruker 7T/20 MRI scanner using process, whereby they were first normalized to the pooled stan- a quadrature birdcage transceiver. MRI parameters were as dards (TMT-126 and -131) for each protein and then to the median follows: 3D FLASH protocol with TE/TR = 6 ms/15 ms and matrix signal across the pooled standards from all experiments (Lapek size 128 × 64 × 156, field of view prescribed to match the body et al. 2017a). An average of these normalizations was used for size. InVesalius (https://link.springer.com/chapter/10.1007/978- the next step. As the brain samples did not contain a composite 3-319-27857-5_5) was used to trace individual organs in each bridge channel for normalization, raw signal/noise ratios were nor- MRI slice to generate the 3D model. The model was then processed malized by the average signal of the given protein divided by the with Blender (https://www.blender.org/) for smoothing. median of all protein averages. To account for slight differences Interactive display of associations can be done through the in amounts of protein labeled, these values were then normalized following steps: (1) accessing the ‘ili-web browser (https://ili to the median of the entire data set and reported as final normal- .embl.de) (Protsyuk et al. 2018); (2) uploading the 3D mouse mod- ized summed signal-to-noise ratios per protein per sample. el visualization file (Supplemental File S1); (3) uploading either the protein-based associations table (Supplemental Table S4) or the Statistical analysis pathway-based associations table (Supplemental Table S5). Pathway-level association scores were determined by compar- Bioinformatic analysis was performed in Python (version 3.5.1), ing the −Log (Benjamini–Hochberg corrected P-values) for each and records are available online in Jupyter Notebooks (https 10 functional enrichment in Supplemental Table S2. The statistical ://github.com/rhmills/Germ-free-organ-proteomics). To prevent strength of the functional enrichment in GF tissue was subtracted statistical artifacts generated due to the various methods of dealing from the statistical strength of the functional enrichment in con- with missing values, only proteins with quantification in all sam- ventional tissue. The following formula summarizes the calcula- ples were used. A Student’s t-test with unequal variance was per- tion: formed through the package, SciPy (https://www.scipy.org). For ranking purposes, we evaluated associations with colonization = − . − Association Score (( Log10(Adj P value for conventional status)) state through π-score, which accounts for both fold-change and − (− Log (Adj. P − value for GF Status))). P-value (Xiao et al. 2014). A statistical cutoff of |π| > 1 was chosen 10 based on previous work (Tran et al. 2019). This statistical measure corresponds to a significance level of α ∼ 0.05 and allowed for mod- Data access erate stringency while including an adequate number of proteins for protein network construction and functional enrichment The proteomic data generated in this study have been submitted analysis. to the Mass Spectrometry Interactive Virtual Environment Protein–protein interaction networks were created through (MassIVE) Repository (https://massive.ucsd.edu) under the study STRINGdb (Szklarczyk et al. 2015). Associations between proteins ID MSV000083874. Code is available through GitHub (https:// were determined through default settings, accounting for text github.com/rhmills/Germ-free-organ-proteomics) and as Supple- mining, experiments, databases, coexpression, neighborhood, mental Code. gene fusion, and co-occurrence. Connections were restricted to interactions between proteins within the query list only. Networks Competing interest statement were subsequently visualized through Cytoscape (version 3.5.1) (Shannon et al. 2003). Edges within protein networks were based The authors declare no competing interests.

Genome Research 9 www.genome.org Downloaded from genome.cshlp.org on October 8, 2021 - Published by Cold Spring Harbor Laboratory Press

Mills et al.

Acknowledgments ment accuracy in shotgun proteomics. Mol Cell Proteomics 5: 1326– 1337. doi:10.1074/mcp.M500339-MCP200 D.J.G. was supported by the Ray Thomas Edwards Foundation and Haberman Y, Karns R, Dexheimer PJ, Schirmer M, Somekh J, Jurickova I, ’ Braun T, Novak E, Bauman L, Collins MH, et al. 2019. Ulcerative colitis the UC Office of the President. R.H.M. was supported by UCSD s mucosal transcriptomes reveal mitochondriopathy and personalized Gastroenterology T32 Predoctoral fellowship. A.C. was supported mechanisms underlying disease severity and treatment response. Nat by the UCSD Microbial Sciences Initiative Graduate Research Commun 10: 38. doi:10.1038/s41467-018-07841-3 Fellowship and by the UCSD Graduate Training Program in Huang da W, Sherman BT, Lempicki RA. 2009a. Bioinformatics enrichment Cellular and Molecular Pharmacology through an institutional tools: paths toward the comprehensive functional analysis of large gene lists. Nucleic Acids Res 37: 1–13. doi:10.1093/nar/gkn923 training grant from the National Institute of General Medical Huang da W, Sherman BT, Lempicki RA. 2009b. Systematic and integrative Sciences, T32 GM007752. analysis of large gene lists using DAVID bioinformatics resources. Nat Author contributions: R.H.M., A.V., R.K., A.G., and D.J.G. con- Protoc 4: 44–57. doi:10.1038/nprot.2008.211 ceived and designed the experiments. R.H.M., B.C., and A.V. per- Huttlin EL, Jedrychowski MP, Elias JE, Goswami T, Rad R, Beausoleil SA, Villén J, Haas W, Sowa ME, Gygi SP. 2010. A tissue-specific atlas of formed the experiments. R.H.M., J.M.W., A.V., and A.C. mouse protein phosphorylation and expression. Cell 143: 1174–1189. analyzed the data. R.H.M., J.M.W., A.C., A.V., and D.J.G. wrote doi:10.1016/j.cell.2010.12.001 the manuscript. All authors read and approved the final Käll L, Canterbury JD, Weston J, Noble WS, MacCoss MJ. 2007. Semi-super- manuscript. vised learning for peptide identification from shotgun proteomics data- sets. Nat Methods 4: 923–925. doi:10.1038/nmeth1113 Kamga CK, Zhang SX, Wang Y. 2010. Dicarboxylate carrier-mediated glutathione transport is essential for reactive oxygen species homeostasis and References normal respiration in rat brain mitochondria. Am J Physiol Cell Physiol 299: C497–C505. doi:10.1152/ajpcell.00058.2010 Ahmadmehrabi S, Tang WHW. 2017. Gut microbiome and its role in cardio- Karlsson FH, Fåk F, Nookaew I, Tremaroli V, Fagerberg B, Petranovic D, vascular diseases. Curr Opin Cardiol 32: 761–766. doi:10.1097/HCO Bäckhed F, Nielsen J. 2012. Symptomatic atherosclerosis is associated .0000000000000445 with an altered gut metagenome. Nat Commun 3: 1245. doi:10.1038/ Beausoleil SA, Villén J, Gerber SA, Rush J, Gygi SP. 2006. A probability-based ncomms2266 approach for high-throughput protein phosphorylation analysis and Kindt A, Liebisch G, Clavel T, Haller D, Hörmannsperger G, Yoon H, site localization. Nat Biotechnol 24: 1285–1292. doi:10.1038/nbt1240 Kolmeder D, Sigruener A, Krautbauer S, Seeliger C, et al. 2018. The gut Bhattarai Y, Kashyap PC. 2016. Germ-free mice model for studying host-mi- microbiota promotes hepatic fatty acid desaturation and elongation in crobial interactions. Methods Mol Biol 1438: 123–135. doi:10.1007/978- mice. Nat Commun 9: 3760. doi:10.1038/s41467-018-05767-4 1-4939-3661-8_8 Koopman JP, Janssen FG, van Druten JA. 1975. Oxidation-reduction poten- Birben E, Sahiner UM, Sackesen C, Erzurum S, Kalayci O. 2012. Oxidative tials in the cecal contents of rats and mice. Proc Soc Exp Biol Med 149: stress and antioxidant defense. World Allergy Organ J 5: 9–19. doi:10 995–999. doi:10.3181/00379727-149-38942 .1097/WOX.0b013e3182439613 Kramer PA, Darley-Usmar VM. 2015. The emerging theme of redox bioener- Bouter KE, van Raalte DH, Groen AK, Nieuwdorp M. 2017. Role of the gut getics in health and disease. Biomed J 38: 294–300. doi:10.4103/2319- microbiome in the pathogenesis of obesity and obesity-related metabol- 4170.155591 ic dysfunction. Gastroenterology 152: 1671–1678. doi:10.1053/j.gastro Krautkramer KA, Kreznar JH, Romano KA, Vivas EI, Barrett-Wilt GA, .2016.12.048 Rabaglia ME, Keller MP, Attie AD, Rey FE, Denu JM. 2016. Diet-micro- Brenes A, Hukelmann J, Bensaddek D, Lamond AI. 2019. Multibatch TMT biota interactions mediate global epigenetic programming in multiple reveals false positives, batch effects and missing values. Mol Cell host tissues. Mol Cell 64: 982–992. doi:10.1016/j.molcel.2016.10.025 Proteomics 18: 1967–1980. doi:10.1074/mcp.RA119.001472 Kuno T, Hirayama-Kurogi M, Ito S, Ohtsuki S. 2016. Effect of intestinal flora Celesk RA, Asano T, Wagner M. 1976. The size pH, and redox potential of on protein expression of drug-metabolizing enzymes and transporters the cecum in mice associated with various microbial floras. Proc Soc in the liver and kidney of germ-free and antibiotics-treated mice. Mol Exp Biol Med 151: 260–263. doi:10.3181/00379727-151-39187 Pharm 13: 2691–2701. doi:10.1021/acs.molpharmaceut.6b00259 Chung H, Pamp SJ, Hill JA, Surana NK, Edelman SM, Troy EB, Reading NC, Lapek JD, Greninger P, Morris R, Amzallag A, Pruteanu-Malinici I, Benes CH, Villablanca EJ, Wang S, Mora JR, et al. 2012. Gut immune maturation Haas W. 2017a. Detection of dysregulated protein association networks depends on colonization with a host-specific microbiota. Cell 149: by high-throughput proteomics predicts cancer vulnerabilities. Nat 1578–1593. doi:10.1016/j.cell.2012.04.037 Biotechnol 35: 983–989. doi:10.1038/nbt.3955 Circu ML, Aw TY. 2011. Redox biology of the intestine. Free Radic Res 45: Lapek JD, Lewinski MK, Wozniak JM, Guatelli J, Gonzalez DJ. 2017b. 1245–1266. doi:10.3109/10715762.2011.611509 Quantitative temporal viromics of an inducible HIV-1 model yields in- Diaz Heijtz R, Wang S, Anuar F, Qian Y, Bjorkholm B, Samuelsson A, sight to global host targets and phospho-dynamics associated with pro- Hibberd ML, Forssberg H, Pettersson S. 2011. Normal gut microbiota tein Vpr. Mol Cell Proteomics 16: 1447–1461. doi:10.1074/mcp.M116 modulates brain development and behavior. Proc Natl Acad Sci 108: .066019 3047–3052. doi:10.1073/pnas.1010529108 Lapek JD, Mills RH, Wozniak JM, Campeau A, Fang RH, Wei X, van de Groep Drake NM, DeVito LM, Cleland TA, Soloway PD. 2011. Imprinted Rasgrf1 K, Perez-Lopez A, van Sorge NM, Raffatellu M, et al. 2018. Defining host expression in neonatal mice affects olfactory learning and memory. responses during systemic bacterial infection through construction of a Genes Brain Behav 10: 392–403. doi:10.1111/j.1601-183X.2011.00678.x murine organ proteome atlas. Cell Syst 6: 579–592.e4. doi:10.1016/j.cels El Aidy S, Derrien M, Merrifield CA, Levenez F, Dore J, Boekschoten MV, .2018.04.010 Dekker J, Holmes E, Zoetendal EG, van Baarlen P, et al. 2013. Gut bacte- Larsson E, Tremaroli V, Lee YS, Koren O, Nookaew I, Fricker A, Nielsen J, Ley ria-host metabolic interplay during conventionalisation of the mouse RE, Bäckhed F. 2012. Analysis of gut microbial regulation of host gene germfree colon. ISME J 7: 743–755. doi:10.1038/ismej.2012.142 expression along the length of the gut and regulation of gut microbial Elias JE, Gygi SP. 2007. Target-decoy search strategy for increased confi- ecology through MyD88. Gut 61: 1124–1131. doi:10.1136/gutjnl- dence in large-scale protein identifications by mass spectrometry. Nat 2011-301104 Methods 4: 207–214. doi:10.1038/nmeth1019 Leal-Lopes C, Velloso FJ, Campopiano JC, Sogayar MC, Correa RG. 2015. Elias JE, Haas W, Faherty BK, Gygi SP. 2005. Comparative evaluation of mass Roles of commensal microbiota in pancreas homeostasis and pancreatic spectrometry platforms used in large-scale proteomics investigations. pathologies. J Diabetes Res 2015: 284680. doi:10.1155/2015/284680 Nat Methods 2: 667–675. doi:10.1038/nmeth785 Lichtman JS, Alsentzer E, Jaffe M, Sprockett D, Masutani E, Ikwa E, Eng JK, McCormack AL, Yates JR. 1994. An approach to correlate tandem Fragiadakis GK, Clifford D, Huang BE, Sonnenburg JL, et al. 2016. The mass spectral data of peptides with amino acid sequences in a protein effect of microbial colonization on the host proteome varies by gastro- database. J Am Soc Mass Spectrom 5: 976–989. doi:10.1016/1044-0305 intestinal location. ISME J 10: 1170–1181. doi:10.1038/ismej.2015.187 (94)80016-2 Mangiola F, Ianiro G, Franceschi F, Fagiuoli S, Gasbarrini G, Gasbarrini A. Franco-Obregón A, Gilbert JA. 2017. The microbiome-mitochondrion con- 2016. Gut microbiota in autism and mood disorders. World J nection: common ancestries, common mechanisms, common goals. Gastroenterol 22: 361–368. doi:10.3748/wjg.v22.i1.361 mSystems 2: e00018. doi:10.1128/mSystems.00018-17 Mardinoglu A, Shoaie S, Bergentall M, Ghaffari P, Zhang C, Larsson E, Fu ZD, Selwyn FP, Cui JY, Klaassen CD. 2017. RNA-seq profiling of intestinal Bäckhed F, Nielsen J. 2015. The gut microbiota modulates host amino expression of xenobiotic processing genes in germ-free mice. Drug Metab acid and glutathione metabolism in mice. Mol Syst Biol 11: 834. doi:10 Dispos 45: 1225–1238. doi:10.1124/dmd.117.077313 .15252/msb.20156487 Haas W, Faherty BK, Gerber SA, Elias JE, Beausoleil SA, Bakalarski CE, Li X, McAlister GC, Nusinow DP, Jedrychowski MP, Wühr M, Huttlin EL, Villén J, Gygi SP. 2006. Optimization and use of peptide mass measure- Erickson BK, Rad R, Haas W, Gygi SP. 2014. MultiNotch MS3 enables

10 Genome Research www.genome.org Downloaded from genome.cshlp.org on October 8, 2021 - Published by Cold Spring Harbor Laboratory Press

Microbiome-impacted organ protein networks

accurate, sensitive, and multiplexed detection of differential expression integrated models of biomolecular interaction networks. Genome Res across cancer cell line proteomes. Anal Chem 86: 7150–7158. doi:10 13: 2498–2504. doi:10.1101/gr.1239303 .1021/ac502040v Simon GM, Cheng J, Gordon JI. 2012. Quantitative assessment of the im- Milani C, Duranti S, Bottacini F, Casey E, Turroni F, Mahony J, Belzer C, pact of the gut microbiota on lysine ε-acetylation of host proteins using Delgado Palacio S, Arboleya Montes S, Mancabelli L, et al. 2017. The first gnotobiotic mice. Proc Natl Acad Sci 109: 11133–11138. doi:10.1073/ microbial colonizers of the human gut: composition, activities, and pnas.1208669109 health implications of the infant gut microbiota. Microbiol Mol Biol Singh M, Webster PD. 1978. Neurohormonal control of pancreatic secre- Rev 81. doi:10.1128/MMBR.00036-17 tion. A review. Gastroenterology 74: 294–309. doi:10.1016/0016-5085 Morton JT, Marotz C, Washburne A, Silverman J, Zaramela LS, Edlund A, (78)90812-0 Zengler K, Knight R. 2019. Establishing microbial composition measure- Spivak M, Weston J, Bottou L, Käll L, Noble WS. 2009. Improvements to the ment standards with reference frames. Nat Commun 10: 2719. doi:10 percolator algorithm for peptide identification from shotgun proteo- .1038/s41467-019-10656-5 mics data sets. J Proteome Res 8: 3737–3745. doi:10.1021/pr801109k Mottawea W, Chiang CK, Mühlbauer M, Starr AE, Butcher J, Abujamel T, Szklarczyk D, Franceschini A, Wyder S, Forslund K, Heller D, Huerta-Cepas J, Deeke SA, Brandel A, Zhou H, Shokralla S, et al. 2016. Altered intestinal Simonovic M, Roth A, Santos A, Tsafou KP, et al. 2015. STRING v10: pro- microbiota-host mitochondria crosstalk in new onset Crohn’s disease. tein–protein interaction networks, integrated over the tree of life. Nat Commun 7: 13419. doi:10.1038/ncomms13419 Nucleic Acids Res 43: D447–D452. doi:10.1093/nar/gku1003 Nakaya K, Ikewaki K. 2018. Microbiota and HDL metabolism. Curr Opin Thaiss CA, Zmora N, Levy M, Elinav E. 2016. The microbiome and innate Lipidol 29: 18–23. doi:10.1097/MOL.0000000000000472 immunity. Nature 535: 65–74. doi:10.1038/nature18847 Nguyen TT, Kosciolek T, Eyler LT, Knight R, Jeste DV. 2018. Overview and Thompson A, Schäfer J, Kuhn K, Kienle S, Schwarz J, Schmidt G, Neumann systematic review of studies of microbiome in schizophrenia and bipolar T, Johnstone R, Mohammed AK, Hamon C. 2003. Tandem mass tags: a disorder. J Psychiatr Res 99: 50–61. doi:10.1016/j.jpsychires.2018.01 novel quantification strategy for comparative analysis of complex pro- 75: – .013 tein mixtures by MS/MS. Anal Chem 1895 1904. doi:10.1021/ Nicholson JK, Holmes E, Kinross J, Burcelin R, Gibson G, Jia W, Pettersson S. ac0262560 2012. Host-gut microbiota metabolic interactions. Science 336: 1262– Tilg H, Moschen AR. 2014. Microbiota and diabetes: an evolving relation- 63: – 1267. doi:10.1126/science.1223813 ship. Gut 1513 1521. doi:10.1136/gutjnl-2014-306928 Nicolay BN, Danielian PS, Kottakis F, Lapek JD, Sanidas I, Miles WO, Ting L, Rad R, Gygi SP, Haas W. 2011. MS3 eliminates ratio distortion in iso- 8: – Dehnad M, Tschöp K, Gierut JJ, Manning AL, et al. 2015. Proteomic baric multiplexed quantitative proteomics. Nat Methods 937 940. analysis of pRb loss highlights a signature of decreased mitochondrial doi:10.1038/nmeth.1714 Genes Dev 29: – Tolonen AC, Haas W. 2014. Quantitative proteomics using reductive dime- oxidative phosphorylation. 1875 1889. doi:10.1101/ 89: gad.264127.115 thylation for stable isotope labeling. J Vis Exp e51416. doi:10.3791/ Peng J, Elias JE, Thoreen CC, Licklider LJ, Gygi SP. 2003. Evaluation of mul- 51416 tidimensional chromatography coupled with tandem mass spectrome- Tran HQ, Mills RH, Peters NV, Holder MK, de Vries GJ, Knight R, Chassaing try (LC/LC-MS/MS) for large-scale protein analysis: the yeast B, Gonzalez DJ, Gewirtz AT. 2019. Associations of the fecal microbial proteome. J Proteome Res 2: 43–50. doi:10.1021/pr025556v proteome composition and proneness to diet-induced obesity. Mol Cell Proteomics 18: – Protsyuk I, Melnik AV, Nothias LF, Rappez L, Phapale P, Aksenov AA, 1864 1879. doi:10.1074/mcp.RA119.001623 Bouslimani A, Ryazanov S, Dorrestein PC, Alexandrov T. 2018. 3D mo- Villén J, Gygi SP. 2008. The SCX/IMAC enrichment approach for global phosphorylation analysis by mass spectrometry. Nat Protoc 3: 1630– lecular cartography using LC–MS facilitated by Optimus and ‘ili soft- 1638. doi:10.1038/nprot.2008.150 ware. Nat Protoc 13: 134–154. doi:10.1038/nprot.2017.122 Wang Y, Yang F, Gritsenko MA, Wang Y, Clauss T, Liu T, Shen Y, Monroe Quinn RA, Vrbanac A, Melnik AV, Patras KA, Christy M, Nelson AT, Aksenov ME, Lopez-Ferrer D, Reno T, et al. 2011a. Reversed-phase chromatogra- A, Tripathi A, Humphrey G, da Silva R, et al. 2019. Chemical impacts of phy with multiple fraction concatenation strategy for proteome profil- the microbiome across scales reveal novel conjugated bile acids. bioRxiv ing of human MCF10A cells. Proteomics 11: 2019–2026. doi:10.1002/ doi:10.1101/654756 pmic.201000722 Ray PD, Huang BW, Tsuji Y. 2012. Reactive oxygen species (ROS) homeosta- Wang Z, Klipfell E, Bennett BJ, Koeth R, Levison BS, Dugar B, Feldstein AE, sis and redox regulation in cellular signaling. Cell Signal 24: 981–990. Britt EB, Fu X, Chung YM, et al. 2011b. Gut flora metabolism of phos- doi:10.1016/j.cellsig.2012.01.008 phatidylcholine promotes cardiovascular disease. Nature 472: 57–63. Ronchi JA, Figueira TR, Ravagnani FG, Oliveira HC, Vercesi AE, Castilho RF. doi:10.1038/nature09922 2013. A spontaneous mutation in the nicotinamide nucleotide transhy- Wessel D, Flügge UI. 1984. A method for the quantitative recovery of pro- drogenase gene of C57BL/6J mice results in mitochondrial redox abnor- 63: – tein in dilute solution in the presence of detergents and lipids. Anal malities. Free Radic Biol Med 446 456. doi:10.1016/j.freeradbiomed Biochem 138: 141–143. doi:10.1016/0003-2697(84)90782-6 .2013.05.049 Xiao Y, Hsiao TH, Suresh U, Chen HI, Wu X, Wolf SE, Chen Y. 2014. A novel Sartor RB, Wu GD. 2017. Roles for intestinal bacteria, viruses, and fungi in significance score for gene selection and ranking. Bioinformatics 30: pathogenesis of inflammatory bowel diseases and therapeutic ap- – 152: – 801 807. doi:10.1093/bioinformatics/btr671 proaches. Gastroenterology 327 339.e4. doi:10.1053/j.gastro.2016 Zarate CJ, Machado-Vieira R, Henter I, Ibrahim L, Diazgranados N, .10.012 Salvadore G. 2010. Glutamatergic modulators: the future of treating Schwechter B, Rosenmund C, Tolias KF. 2013. RasGRF2 Rac-GEF activity mood disorders? Harv Rev Psychiatry 18: 293–303. doi:10.3109/ couples NMDA receptor calcium flux to enhanced synaptic trans- 10673229.2010.511059 mission. Proc Natl Acad Sci 110: 14462–14467. doi:10.1073/pnas .1304340110 Shannon P, Markiel A, Ozier O, Baliga NS, Wang JT, Ramage D, Amin N, Schwikowski B, Ideker T. 2003. Cytoscape: a software environment for Received September 13, 2019; accepted in revised form January 21, 2020.

Genome Research 11 www.genome.org Downloaded from genome.cshlp.org on October 8, 2021 - Published by Cold Spring Harbor Laboratory Press

Organ-level protein networks as a reference for the host effects of the microbiome

Robert H. Mills, Jacob M. Wozniak, Alison Vrbanac, et al.

Genome Res. published online January 28, 2020 Access the most recent version at doi:10.1101/gr.256875.119

Supplemental http://genome.cshlp.org/content/suppl/2020/02/11/gr.256875.119.DC1 Material

Creative This article is distributed exclusively by Cold Spring Harbor Laboratory Press for the Commons first six months after the full-issue publication date (see License http://genome.cshlp.org/site/misc/terms.xhtml). After six months, it is available under a Creative Commons License (Attribution-NonCommercial 4.0 International), as described at http://creativecommons.org/licenses/by-nc/4.0/.

Email Alerting Receive free email alerts when new articles cite this article - sign up in the box at the Service top right corner of the article or click here.

Advance online articles have been peer reviewed and accepted for publication but have not yet appeared in the paper journal (edited, typeset versions may be posted when available prior to final publication). Advance online articles are citable and establish publication priority; they are indexed by PubMed from initial publication. Citations to Advance online articles must include the digital object identifier (DOIs) and date of initial publication.

To subscribe to Genome Research go to: https://genome.cshlp.org/subscriptions