Rumen microbial community composition varies with diet and host, but a core microbiome is found across a wide geographical range

1 Scien i ic RepoR s | 5:14567 | DOi: 10.1038/s ep14567 www.na u e.com/scien i ic epo s Rumen mic obial communi y composi ion a ies wi h die and hos , bu a co e mic obiome is ound ac oss a wide geog aphical ange Gemma Hende son1, Fai h Cox1, Si a Ganesh1, A jan Jonke 1, Wayne Young1, Global Rumen Census Collabo a o s† & Pe e H. Janssen1 1AgResea ch Limi ed, G asslands Resea ch Cen e, Palme s on No h 4442, New Zealand. †A comp ehensi e lis o au ho s and a ilia ions appea a he end o he pape . Co espondence and eques s o ma e ials should be add essed o G.H. (email: [email p o ec ed]) o P.H.J. (email: pe e .janssen@ag esea ch. co.nz) Recei ed: 14 Ap il 2015 Accep ed: 01 Sep embe 2015 Published: 09 Oc obe 2015 OPEN Ruminan li es ock a e impo an sou ces o human ood and global g eenhouse gas emissions. Feed deg ada ion and me hane o ma ion by uminan s ely on me abolic in e ac ions be ween umen mic obes and a ec uminan p oduc i i y. Rumen and camelid o egu mic obial communi y composi ion was de e mined in 742 samples om 32 animal species and 35 coun ies, o es ima e i his was in luenced by die , hos species, o geog aphy. Simila bac e ia and a chaea domina ed in nea ly all samples, while p o ozoal communi ies we e mo e a iable. The dominan bac e ia a e poo ly cha ac e ised, bu he me hanogenic a chaea a e be e known and highly conse ed ac oss he wo ld. This uni e sali y and limi ed di e si y could make i possible o mi iga e me hane emissions by de eloping s a egies ha a ge he ew dominan me hanogens. Di e ences in mic obial communi y composi ions we e p edominan ly a ibu able o die , wi h he hos being less in luen ial. The e we e ew s ong co-occu ence pa e ns be ween mic obes, sugges ing ha majo me abolic in e ac ions a e non-selec i e a he han speci ic. Ruminan s a e one o he mos success ul g oups o he bi o ous mammals on he plane , wi h a ound 200 species ep esen ed by app oxima ely 75 million wild and 3.5 billion domes ica ed indi iduals wo ld- wide1. Ruminan s a e de ined by hei mode o plan diges ion, and ha e e ol ed a o es omach, he umen, ha allows pa ial mic obial diges ion o eed be o e i en e s he ue s omach. Ruminan s hemsel es do no p oduce he enzymes needed o deg ade mos complex plan polysaccha ides, and he umen p o ides an en i onmen o a ich and dense conso ium o anae obic mic obes ha ul il his me abolic ole. These umen mic obes e men eed o o m ola ile a y acids ha a e majo nu ien sou ces o he hos animal and con ibu e signi ican ly o uminan p oduc i i y. The hos also uses mic obial biomass and some un e men ed eed componen s once hese exi he umen o he emainde o he diges i e ac . Ruminan s ha e e ol ed a ious umen ana omies and beha iou s o h i e on a ange o plan species, and his lexibili y has enabled hem o occupy many di e en habi a s spanning a wide ange o clima es2. These we e also impo an ac o s in hei domes ica ion, allowing con e sion o human-indiges ible plan ma e ial in o eadily-accessible animal goods, especially dai y p oduc s, mea , and use ul ib es. Ruminan s ha e hus played a i al ole in sus aining and de eloping many human cul u es, as well as being used as d a animals and ha ing eligious and s a us alues. www.na u e.com/scien i ic epo s/ 2 Scien i ic RepoR s | 5:14567 | DOi: 10.1038/s ep14567 Rumen mic obes can be assigned o di e en unc ional g oups, such as celluloly ics, amyloly ics, p o eoly ics, e c., which deg ade he wide a ie y o eed componen s o u he me abolize some o he p oduc s o med by o he mic obes3. Fo example, me hanogens, he me hane- o ming a chaea, a e among hose ha me abolize hyd ogen o med by some e men a i e mic obes o o m me hane. The me hane gene a ed du ing his e men a ion con ibu es o global an h opogenic g eenhouse gas emis- sions4 and ep esen s a 2–12% loss o eed ene gy o he animal5. Di e ences in umen mic obial com- muni ies unde lie a ia ions in me hane o ma ion6 and he con e sion o eed o animal p oduc s7,8. The e o e, unde s anding hese communi ies is key o unde s anding uminal ans o ma ions o plan ma e ial o bo h undesi able and use ul uminan p oduc s. The aim o his s udy was o de e mine he composi ion o he mic obio a in umen and o egu samples om 742 indi idual animals om a ound he wo ld. The esul ing da ase allowed us o de e - mine ha die a y ac o s domina e o e hos species in de e mining mic obial communi y composi ion, iden i y he dominan mic obes and hei po en ial associa ions, and desc ibe he deg ee o simila i y o umen mic obial communi ies wo ldwide. Resul s and Discussion This is he la ges single s udy o examine mic obial communi ies ac oss a ange o uminan and camelid species, die s, and geog aphical egions. A s anda dised pipeline was used o p ocess samples in o de o minimise a ia ion in oduced by p ocessing s eps such as DNA ex ac ion o PCR ampli ica ion. This is impo an o de ec ing au hen ic pa e ns a he han ones in oduced by me hodological di e ences be ween di e en s udies9. The p ime s chosen ampli y, o he bes o ou cu en knowledge, he a ge gene egions om nea ly all known bac e ia, a chaea, and umen cilia es. Dominan umen mic obes. Despi e he ange o uminan s wi h di e en eeding s a egies and die s, simila umen bac e ia we e abundan a ound he wo ld (Fig. 1). The e was some a ia ion in bac e ial communi y composi ions in animals om di e en egions, likely o be caused by di e ences in die , clima e, and a ming p ac ices. The 30 mos abundan bac e ial g oups (G eengenes10 axonomy summa ised a he genus-le el) we e all ound in o e 90% o samples, and oge he comp ised 89.4% o all sequence da a (Supplemen a y Table 1) and we e simila o hose desc ibed in an ea lie me a-analysis o umen mic obial communi ies11. All 30 a e known umen-inhabi ing bac e ia. Because he samples came om a wide ange o uminan species, die s, and geog aphical loca ions, hese da a sugges ha new dominan bac e ia a e no likely o be ound in u u e s udies. The se en mos abundan bac e ial Figu e 1. O igins o samples and hei bac e ial and a chaeal communi y composi ions in di e en egions. Numbe s below pie cha s ep esen he numbe o samples o which da a we e ob ained. The mos abundan bac e ia and a chaea a e named in clockwise o de s a ing a he op o he pie cha . Fu he de ails o samples and communi y composi ion a e gi en in Supplemen a y Tables 1, 2, 3, and 4 and Supplemen a y Da a 1. Mmc. Me hanomassiliicoccales. The map was sou ced om Wikimedia Commons (h p://commons.wikimedia.o g/wiki/File:BlankMap-Wo ld- 2.png, o iginal uploade Roke, accessed May 2013). Pie cha s we e p oduced in Mic oso Excel and he composi e image gene a ed wi h Mic oso Powe Poin and Adobe Illus a o . h ps://c ea i ecommons.o g/licenses/by-sa/3.0/deed.en www.na u e.com/scien i ic epo s/ 3 Scien i ic RepoR s | 5:14567 | DOi: 10.1038/s ep14567 g oups comp ised 67.1% o all bac e ial sequence da a, we e de ec ed in all samples (Supplemen a y Fig. 1), and can be conside ed he “dominan ” umen bac e ia. They we e P e o ella, Bu y i ib io, and Ruminococcus, as well as unclassi ied Lachnospi aceae, Ruminococcaceae, Bac e oidales, and Clos idiales. These migh be conside ed a “co e bac e ial mic obiome” a he genus le el o highe , because hey a e p esen in a la ge selec ion o uminan s, so con i ming he sugges ion ha he e is a co e umen mic obiome9. Howe e , hese bac e ial g oups we e no equally abundan in all animal species (P ≤ 0.005; Supplemen a y Table 2). Wi h he excep ion o Bu y i ib io12, hese g oups a e no adequa ely ep e- sen ed by cha ac e ised cul u es13, and hei unc ions a e no well unde s ood. Inspec ion o he mos abundan and p e alen bac e ial ope a ional axonomic uni s (OTUs) in he da ase showed ha only 14% ell wi hin a named species, and 70% we e no e en wi hin a o mally ecognised genus (Fig.2a). When cul u ed isola es om as-ye unnamed species we e included in he analysis, he dominan OTUs we e be e (35%) bu s ill poo ly ep esen ed by cul u es ha belonged o po en ially he same species (Fig.2b). This s udy shows ha , while we appea o ecognize he dom- inan umen bac e ia, conside able mic obiological e o is s ill equi ed o unde s and hem. Some e o s ha e been made o isola e mo e cul u es and ga he mo e in o ma ion abou hese bac e ia13,14. Fo example, he genomes o P e o ella a . uminicola Tc2-24, umen bac e ium R-7, and o he isola es whose 16S RNA gene sequences a e simila o hose o dominan umen bac e ial OTUs (Fig.2b), ha e been sequenced as pa o he Hunga e1000 p ojec 15. Because he e is a lux o bo h liquids and solids h ough he umen16, mic obes mus ac i ely me ab- olize o gain ene gy and mul iply o coun e ac washou and so main ain popula ions in he umen17. The dominan bac e ia ound in his s udy a e he e o e likely o be esponsible o he majo i y o he Figu e 2. Dominan bac e ial and a chaeal ope a ional axonomic uni s (OTUs). Simila i ies (Supplemen a y Tables 8 and 9) o he 50 mos abundan and 50 mos p e alen bac e ial (77 unique OTUs, (a,b) and a chaeal (64 unique OTUs, c,d) OTUs o he mos closely ela ed ype (a,c) and cul u ed (b,d) s ains a e plo ed oge he wi h p e alence and abundance da a. Backg ound shading indica es nominal wi hin-species (da k g ey), wi hin-genus (mid g ey) and below genus (ligh g ey) simila i ies. P e alence indica es he pe cen age o samples ha an OTU occu s in. The size o each ci cle indica es he mean abundance o each OTU (Supplemen a y Tables 8 and 9). Bac e ial OTU abundances we e mul iplied by a ac o o 15 ela i e o a chaeal OTUs. Mbb. Me hanob e ibac e . www.na u e.com/scien i ic epo s/ 4 Scien i ic RepoR s | 5:14567 | DOi: 10.1038/s ep14567 ans o ma ion o inges ed eed in he umen and camelid o egu , especially o cellulose, hemicellulose, pec in, s a ch, uc an, o ganic acids, and p o ein, as hese a e he majo ene gy-yielding subs a es used o mic obial g ow h17. The e is also a con e gence o bac e ial communi y s uc u e in he umen and in he c op o he hoa zin, a bi d ha elies on a o egu e men a ion o inges ed lea es18. Thus mic obial communi y s uc u e seems o be d i en by he simila i y o o gan unc ion ex ending ac oss he umen, he camelid o egu , and he c op o his unusual bi d. Mo e e o s should go in o cha ac e izing he me abolism and oles o hese bac e ia ha a e he esponsible o he majo i y o eed e men a ion, wi h he aim o enhancing animal p oduc i i y and educing me hane emissions. Nea ly all he a chaea we e iden i ied as me hanogens known o be esiden s o he umen (Supplemen a y Table 3, Supplemen a y Tex 1), and hei ela i e abundances we e compa able o p e i- ous s udies19. The dominan a chaeal g oups we e ema kably simila in all egions o he wo ld (Fig.1). This uni e sali y and limi ed di e si y was also ecen ly no ed in su ey o a chaea in New Zealand umi- nan s20 and could make i possible o success ully mi iga e me hane emissions by de eloping s a egies, such as accines o small-molecule inhibi o s, ha a ge he ew dominan me hanogens. Membe s o he Me hanob e ibac e go schalkii and Me hanob e ibac e uminan ium clades we e ound in almos all samples, and we e he wo la ges g oups, accoun ing o 74% o all a chaea. Toge he wi h a Me hanosphae a sp. and wo Me hanomassiliicoccaceae-a ilia ed g oups, he i e dominan me h- anogen g oups comp ised 89.2% o he a chaeal communi ies (Supplemen a y Fig. 1), showing ha umen a chaea a e much less di e se han umen bac e ia. This likely e lec s he na ow ange o sub- s a es hey use. Me hanomic obium has p e iously been epo ed as abundan in uminan s in Asia19. In ou s udy, hey we e ound o comp ise > 5% o he a chaeal communi y o some Aus alian, B azilian, Chinese, No h Ame ican, and Sou h A ican ca le, as well as Sou h A ican sheep, showing hem o be widely dis ibu ed, bu no uni e sally p e alen . The i e dominan me hanogen g oups we e no equally abundan in all animal species g oups (P ≤ 0.005; Supplemen al Table 4). In con as o bac e ia, he umen a chaea a e be e ep esen ed by cul u es, wi h 58% o he mos abundan and p e alen OTUs alling wi hin a named species, and all bu 22% wi hin named gene a (Fig.2c). All o he la e we e membe s o Me hanomassiliicoccales, which is an o de o ela i ely poo ly-cha ac e ised me hanogens21 o which ep esen a i e cul u es o as-ye unnamed species and gene a a e a ailable (Fig.2d)22. The 50 mos abundan OTUs accoun ed o 74.5% o he a chaeal sequence da a, again indica ing a much lowe di e si y han in he bac e ia, whe e he 50 mos abundan OTUs made up only 11.0%. By assigning physiologies (Supplemen a y Table 5) o he sequence abundance in o ma ion (Supplemen a y Table 3), i can be concluded ha 77.7% o a chaea we e hyd ogeno ophic me hano- gens, while 22.1% had he abili y o g ow wi h hyd ogen plus me hyl g oups de i ed om me hanol o me hylamines. Me hanogens able o o m me hane om ace a e (Me hanosa cina spp. and Me hanosae a spp.) we e ex emely a e (< 0.015%; Supplemen al Da a 1), as expec ed based on hei gene al slow g ow h a es ha would no allow hem o be main ained in he umen unde no mal condi ions. Almos all p o ozoal sequence da a (> 99.9%) we e assigned o 12 genus-equi alen p o ozoal g oups (Supplemen a y Table 6). I was appa en ha he a iabili y o p o ozoa be ween and wi hin coho s o co-loca ed animals was much g ea e han ha o bac e ia and a chaea (Supplemen a y Fig. 2). I has been epo ed ha he e is s ong hos indi iduali y o umen p o ozoal communi y s uc u e9, and his is e iden in ou s udy. The gene a En odinium and Epidinium domina ed, occu ing in mo e han 90% o samples and ep esen ing 54.7% o p o ozoal sequence da a (Supplemen a y Fig. 1). Many o he p o ozoal gene a we e p esen in g ea e han 70% o he samples, indica ing a wide p e alence. Gene a such as Enoploplas on and Oph yoscolex had a wide han expec ed hos dis ibu ion. They a e conside ed o be mainly p esen in sheep and ca le, espec i ely23, bu we also ound Enoploplas on in ca le, dee , and eindee samples om wel e coun ies, and Oph yoscolex in bu alo, goa s, dee , sheep, and gi a e samples om 18 coun ies. Al hough di e en umen p o ozoa a e epo ed o ha e limi ed hos and geog aphical dis ibu ions, hos speci ici y has been ques ioned24. I seems likely ha u he in es iga ion will demons a e g ea e ubiqui y o he umen p o ozoa. E ec s o die and hos on mic obial communi y composi ion. Because he abundance o mic obial g oups a ied be ween animal species g oups and coho s (Supplemen a y Tables 2 and 4; Supplemen a y Fig. 2), we looked o ac o s ha migh unde lie his. Rumen and camelid o egu mic o- bial communi y s uc u e could be expec ed o be shaped by mo phological, physiological, and e en beha iou al cha ac e is ics ha e ol ed along wi h he a ied eeding s a egies in he a ious umi- nan lineages2. Indeed, adap a ion has esul ed in a di e si y o umen sizes and passage a es o umen con en s, allowing uminan species o exploi a ange o eed ypes. In addi ion o eed composi ion e ec s25, hese hos adap a ions migh also play a ole in egula ing umen mic obial communi y s uc- u e. Because ou da ase was om uminan s and camelids om di e en lineages consuming a ange o die s, hos and die e ec s on umen mic obial communi y s uc u e could be sepa a ed. To look a die and hos e ec s, we classi ied he die s based on o age and b owse o concen a e con en (Supplemen a y Table 7) and g ouped he animals acco ding o hei lineage (Supplemen a y Da a 1). Mic obial communi ies could clea ly be disc imina ed by bo h hos and die (Fig. 3a), wi h bac e ia being he main d i e s behind he obse ed di e ences (Fig.3b). This p obably e lec s hei mo e di e se me abolic capabili ies compa ed wi h he less e sa ile a chaea and p o ozoa. We in es- iga ed he pa e ns o mic obial abundances ac oss hos s and die s (Fig.3c, Supplemen a y Fig. 3–6). www.na u e.com/scien i ic epo s/ 5 Scien i ic RepoR s | 5:14567 | DOi: 10.1038/s ep14567 Figu e 3. E ec o hos species and die a y o age o concen a e a ios on mic obial communi ies. Die s we e g ouped (Supplemen a y Table 7) as o age-domina ed (F), mixed o age-concen a e (50–70% o age, FC), mixed concen a e- o age (50–70% concen a e, CF), o concen a e-domina ed (C). (a) Disc iminan analysis o mic obial communi ies in samples ( ep esen ed by poin s colou ed by animal and die ) e ealed ha bo h hos and die de e mined communi y composi ion. (b) Bi-plo ha shows mic obial g oups (iden i ied by colou s) unde lying he sepa a ion o samples in panel (a). Se e al bac e ial g oups s ongly disc imina e he samples by hos and die , indica ed by hei p esence owa ds he ou side o he bi-plo . A chaeal and p o ozoal g oups a e less disc imina o y, and so a e clus e ed nea e he cen e. (c) The hea map shows ha bac e ial abundances a e di e en ially associa ed wi h die and hos (colou key shows he associa ion sco e; see Supplemen a y Figs 3–5 o addi ional da a). (d) Unclassi ied Veillonellaceae, and (e) Fib obac e a e examples o bac e ia ha caused bo ines and cap ids o clus e sepa a ely om o he species in he hea map. The numbe o samples in each ca ego y is gi en in pa en heses in panels (c–e). *indica es unclassi ied bac e ia wi hin an o de o amily. www.na u e.com/scien i ic epo s/ 6 Scien i ic RepoR s | 5:14567 | DOi: 10.1038/s ep14567 Ruminococcus, one o he dominan bac e ia, was ela i ely e enly dis ibu ed, bu his was an excep ion. Fo many bac e ia, die was he majo ac o de e mining ela i e abundance. Bac e ial communi ies om o age- ed animals we e simila o each o he , hose om concen a e- ed animals we e simila o each o he , bu dis inc om hose in o age- ed animals, and hose om animals ed mixed die s we e in e media e be ween hese. Unclassi ied Bac e oidales and Ruminococcaceae we e mo e abundan in all animals ed o ages. Some as-ye poo ly cha ac e ised Bac e oidales a e pos ula ed o be able o deg ade cellulose, and hei genomes encode a b oad ange o plan polysaccha ide deg ading capa- bili ies26,27, which could explain hei pa e n o dis ibu ion. In con as , membe s o P e o ella and unclassi ied Succini ib ionaceae we e mo e abundan in animals ed die s con aining concen a e. Based on he physiologies o cul u ed ela i es28,29, hese a e p obably majo p oduce s o p opiona e and he p opiona e-p ecu so succina e, and so a e esponsible o he g ea e le els o p opiona e o med om concen a e- ich die s25. The abundance o only a ew o he majo bac e ial g oups was associa ed wi h hos lineage (Fig.3c). Fo example, unclassi ied Veillonellaceae we e p opo ionally mo e abundan in sheep, dee , and camelids (Fig.3d). This may be ela ed o di e ences in umen and camelid o egu sizes, ana omy, and eeding equencies compa ed o bo ines2. The ela i e abundances o se e al majo bac e ial g oups we e a ec ed by bo h hos and die (Fig. 3c). Unclassi ied Clos idiales we e mos abundan in bo ines ed o age and leas abundan in bo ines ed high concen a e die s, while in cap ids, ce ids, and camelids hese die di e ences we e a less p onounced. Bu y i ib io was mos abundan in umen samples om bo ines ed mixes o o age and concen a es. Fib obac e was mos abundan in bo ines ed o age. When concen a e was included in bo ine die s, he ela i e abundance o Fib obac e was dec eased, bu i was s ill mo e abundan han in o he animals. To examine i s dis ibu ion in mo e de ail, we compa ed Fib obac e abundances ac oss di e en uminan species and ound signi ican ly highe le els in bo ines compa ed o dee , sheep, o camelids (Fig.3e). These da a sugges ha Fib obac e is a ou ed in he bo ine umen and, gi en ha i is cellulose deg ade 30, may play an essen ial ole in he deg ada ion o plan ib e in ca le. O e all, die was a majo de e minan o bac e ial communi y s uc u e. This may be because physical and chemical cha ac e is ics o he eed de e mine he di e en mic obial niches a ailable. In con as o he pos -gas ic mammalian diges i e ac 31, and due o he shee olume o diges a and eed inpu , he e is p obably less shaping o he umen mic obial communi y by local hos biological ac o s such as he immune sys em, sec e ed an imic obial pep ides, hos -cell glycosyla ion, and hos -de i ed nu ien s. Associa ions be ween umen mic obes. The abundance pa e ns wi hin bac e ial, a chaeal, and p o ozoal communi ies in di e en hos s ed di e en die s showed ha ce ain mic obes exhibi ed pa al- lel pa e ns o ela i e abundance (Fig.3c, Supplemen a y Figs 2–6). We he e o e looked o co ela ions wi hin and be ween bac e ia, a chaea, and p o ozoa (Fig.4 and Supplemen a y Fig. 7), easoning ha speci ic associa ions should be seen ac oss die s, hos s, and geog aphy. Nega i e co ela ions o abun- dances o g oups we e obse ed wi hin he bac e ia, a chaea, and p o ozoa, including eplacemen e ec s be ween dominan g oups wi hin each o hese (Supplemen a y Tex 1 and Supplemen a y Fig. 7). Few s ong posi i e co ela ions we e ound wi hin bac e ia, a chaea, and p o ozoa. Fo example, he e was a s ong co ela ion be ween Veillonellaceae and he TG5 g oup, d i en by hei co-occu ence wi hin ce ids and cap ids. These mic obes may coope a e in he umen, o hey may sha e simila equi emen s and so ce ain hos s and die s would o e be e oppo uni ies o hei g ow h. This explana ion could also unde lie he s ong posi i e co ela ions obse ed be ween di e en g oups o me hylo ophic me h- anogens (Supplemen a y Fig. 7). They may be esponding o die s ich in me hyl g oups, such as eeds wi h high le els o pec ins o osmoly es such as be aine. The s onges co ela ion wi hin p o ozoa was a posi i e one be ween Dasy icha and Iso icha. These wo gene a o holo ichous p o ozoa display e y simila spec a o subs a e use, including use o plan soluble suga s and s o age ca bohyd a es24, again sugges ing ha co-occu ence may be due o exploi a ion o simila oppo uni ies. We also in es iga ed associa ions be ween bac e ia, a chaea, and p o ozoa. S ikingly, no s ong co - ela ions we e de ec ed be ween a chaea and p o ozoa (Supplemen a y Fig. 7). Me hanogens a e known o colonize p o ozoa, and his mu ualis ic ela ionship is belie ed o enhance me hane o ma ion in he umen32. The occu ence o speci ic symbioses be ween me hanogens and umen p o ozoa has been specula ed on, bu no con incingly demons a ed33. The lack o s ong co-occu ence pa e ns wi hin his s udy indica es ha hese undoub edly impo an associa ions a e p obably non-speci ic, o occu a a s ain le el. Fu he in es iga ion is equi ed o co obo a e his in e es ing inding, as mechanisms ha media e he coloniza ion o p o ozoa by a chaea emain o be elucida ed. These could ha e in e es ing e olu iona y aspec s i hey allow non-speci ic in e ac ions o o m o a e media ed by s ain-speci ic mechanisms ha con e di e en pa ne speci ici ies wi hin a chaeal o p o ozoal species. In con as , he e we e some posi i e associa ions be ween bac e ial and p o ozoal g oups. Mos no iceable we e he associa ions o Iso icha and Dasy icha wi h Fib obac e . Fib obac e we e epo ed o dec ease in abun- dance in animals whe e p o ozoa we e elimina ed34, indica ing ha he e may be a mu ually bene icial ela ionship be ween hese p o ozoa and Fib obac e , which a e su ace colonize s o plan ma e ial23. No s ong associa ions we e ound be ween he mos abundan bac e ia and a chaea (Fig.4). This was su p ising, since umen bac e ia deg ade eed and p oduce he subs a es ha me hanogens use o g ow h, mainly hyd ogen and me hyl g oups. In con as , he e we e dis inc posi i e associa ions be ween some less abundan bac e ia and a chaea. The s onges associa ion was be ween bac e ia such as he www.na u e.com/scien i ic epo s/ 7 Scien i ic RepoR s | 5:14567 | DOi: 10.1038/s ep14567 succina e-p oducing Succini ib ionaceae, he succina e-using Dialis e , and he amino-acid- e men ing Acidaminococcus, and me hanogens belonging o he Me hanomassiliicoccaceae, Me hanosphae a sp. A4, and Me hanob e ibac e bo isko eani. Succini ib io spp. deg ade pec in28, and me hanol is equi ed o g ow h o Me hanomassiliicoccaceae35 and Me hanosphae a36, explaining pa o his pa e n. O he associa ions we e be ween he me hylo ophic me hanogen Me hanosphae a sp. ISO3-F5 and di e - en bac e ia, including membe s o Lachnospi aceae. These associa ions may be based on he abili y o Lachnospi aceae o deg ade pec in and so p o ide me hanol as a subs a e o he me hylo ophs37. The associa ions be ween o he Me hanomassiliicoccaceae g oups and a ious unclassi ied membe s o Bac e oidales sugges he possibili y o ye u he me hanol-dependen me abolic in e ac ions. In con- as o a chaeal-p o ozoal in e ac ions, hese indings sugges ha some a chaeal-bac e ial in e ac ions a e speci ic, in e ing specialised mechanisms o pa ne ecogni ion o e y simila equi emen s o g ow h. The basis o hese associa ions emains o be de e mined. Howe e , he gene al lack o s ong associa ion pa e ns be ween p o ozoa and he majo bac e ia on he one hand, and he majo me h- anogen g oups on he o he , sugges s ha conse ed mechanisms may media e he in e ac ions be ween hyd ogen p oducing and hyd ogen consuming mic obes, allowing lexible in e ac ions. This may aid me hane mi iga ion esea ch, since in e e ing wi h hese po en ially uni e sal mechanisms could slow he a e o hyd ogen ans e and so slow me hane o ma ion38. I may also be ha he in e ac ions mainly occu ia pools o common me aboli es, especially whe e he end p oduc s o one g oup o m he subs a es o ano he . The esul s o his su ey showed ha he umen mic obial ecosys em is domina ed by a co e com- muni y composed o poo ly-cha ac e ised mic obes, especially amongs he bac e ia. Die had mo e in luence han animal species on umen o camelid o egu mic obial communi y composi ion. Rumen ecosys ems a e ypi ied by s ong me abolic in e ac ions be ween mic obes ha acili a e he e men- a ion o plan ma e ial o p oduc s use ul o bo h he hos and o he umen mic obes3,17,25,32. The ela i ely ew co-occu ence pa e ns seen in his s udy sugges ha hese mic obial in e ac ions do no ely on exclusi e associa ions, and could indica e conside able p omiscui y be ween membe s o in e ac - ing unc ional g oups. Analysis a me agenomic and me a ansc ip omic le els could in u u e unco e whe he common unc ional elemen s ha acili a e in e ac ions a e sha ed among mul iple species. I seems plausible ha unc ional edundancy among he mic obes9 means ha mul iple mic obial species can ul il he same unc ion, wi h di e en combina ions o mic obes being co-selec ed depending on Figu e 4. Associa ions be ween bac e ia and a chaea. The ne wo k is based on associa ion sco es compu ed ia egula ised canonical co ela ion analysis wi h an absolu e associa ion sco e g ea e han 0.15. The colou o he lines indica es he s eng h o he associa ion. The sizes o he diamonds and ci cles indica e he mean a e age abundance and mic obial g oups a e iden i ied by numbe s (Supplemen a y Tables 1 and 3). Mbb. Me hanob e ibac e , Mmc. Me hanomassiliicoccales, *indica es unclassi ied bac e ia wi hin a amily. www.na u e.com/scien i ic epo s/ 8 Scien i ic RepoR s | 5:14567 | DOi: 10.1038/s ep14567 he die . This lexibili y o umen mic obial communi y s uc u e would con e on he uminan hos he abili y o exploi a a ie y o di e en plan eeds. Me hods Geog aphical dis ibu ion and di e si y o gas oin es inal ac con en samples. A o al o 742 samples om 32 species o sub-species o uminan s and o he o egu e men e s in 35 coun ies and se en global egions we e selec ed o sequencing o mic obial ma ke genes (Fig.1, Supplemen a y Da a 1). The samples we e om ca le, bison, and bu alo (bo ines), sheep and goa s (cap ids), dee (ce - ids), and alpacas, llamas, and guanacos (camelids), including di e se b eeds o domes ic ca le, sheep, and goa s, and we e la gely made up o small coho s o ou o mo e co-loca ed indi iduals consuming he same die . We included o egu samples o camelids in his s udy, ecognizing ha hese o gans ha e a common unc ion bu e ol ed sepa a ely39. The use o animals, including wel a e, husband y, expe i- men al p ocedu es, and he collec ion o samples used o his s udy, was, whe e applicable, app o ed by named ins i u ional and/o licensing commi ees and pe o med in acco dance wi h app o ed ins i u- ional and egula o y guidelines (please e e o Supplemen a y Da a 1 o de ails o hese). Sample collec ion, DNA ex ac ion, ampli ica ion and p ocessing o samples o high- h oughpu sequencing. To minimise a ia ion in oduced by di e ing me hodologies, such as choice o sampling o DNA ex ac ion me hod40 and p ime -d i en gene ampli ica ion biases41, we used a s anda dised pipeline o p ocess samples (unless indica ed o he wise in Supplemen a y Da a 1). B ie ly, app oxima ely 20 g o whole (i.e., solid and liquid) mid- umen o camelid o egu con en s we e collec ed ia s omach ube, cannula, o pos mo em as p e iously desc ibed35. Samples we e imme- dia ely ozen, eeze-d ied, and hen cou ie ed o AgResea ch. F eeze-d ied samples we e homoge- nised in a co ee blende and DNA was ex ac ed om a ep esen a i e 30 mg subsample using he PCQI me hod40,42. We assessed he s uc u e o mic obial communi ies by sequencing egions o bac- e ial and a chaeal 16S RNA genes and cilia e p o ozoal 18S RNA genes in iplica e as desc ibed p e iously35,37 using p ime s comp ised o (5′ o 3′ ) a sequencing adap e (A o B), a sample-unique 12-base e o -co ec ing Golay ba code on one o each p ime pai , a wo-base linke , and a g oup-speci ic sequence a ge ing he ma ke gene. Fo bac e ia, he p ime s we e Ba515Rmod1 (adap e A-ba code-GT-CCGCGGCKGCTGGCAC) and Ba9F (adap e B-AC-GAGTTTGATCMTGGCTCAG). Fo a chaea, he p ime s we e A 915aF (adap e A-ba code-GT-AGGAATTGGCGGGGGAGCAC) and A 1386R (adap e B-CA- GCGGTGTGTGCAAGGAGC). Fo p o ozoa, he p ime s we e Reg1320R (adap e A-ba code-TC-AATTGCAAAGATCTATCCC) and RP841F (adap e B-AA-GACTAGGGATTGGARTGG). Linke A was CCATCTCATCCCTGCGTGTCTCCGACTCAG and linke B was CCTATCCCCTGTGTGCCTTGGCAGTCTCAG. Amplicons we e sequenced using 454 GS FLX Ti anium chemis y a Eu o ins MWG Ope on (Ebe sbe g, Ge many). Sample p ocess- ing and pipeline ep oducibili y con ols we e pe o med o iden i y a ia ion in oduced du ing sam- ple p ocessing (Supplemen a y Tex 1). Sequence da a a e a ailable om GenBank [accession numbe s PRJNA272135, PRJNA272136, and PRJNA273417]. Phylogene ic analysis o sequencing da a. Py osequence da a we e p ocessed and analysed using he QIIME so wa e package e sion 1.843. Sequences o e 400 bp in leng h wi h an a e age quali y sco e o e 25 we e assigned o a speci ic sample ia he ba codes. The numbe o bac e ial, a chaeal, and cilia e p o ozoal sequencing eads a ailable o analysis a e summa ised in Supplemen a y Da a 1. Sequence da a we e g ouped in o ope a ional axonomic uni s (OTUs) sha ing o e 97% (bac e ia – UCLUST44), 99% (a chaea - UCLUST) o 100% (cilia e p o ozoa – p e ix_su ix op ion in QIIME) sequence sim- ila i y. Sequences we e assigned o phylogene ic g oups by BLAST45. Bac e ial 16S RNA genes we e assigned using he G eengenes da abase e sion 13_510, a chaeal 16S RNA genes using RIM-DB e - sion 13_11_1322 and cilia e p o ozoal 18S RNA genes agains an in-house da abase46. Bac e ial and cilia e p o ozoal da a we e summa ised a he genus le el. A chaea we e summa ised a he species le el. Samples o which low ead numbe s we e ob ained o ha con ained high p opo ions o sequences om “exogenous” bac e ia (i.e., likely en i onmen al con aminan s such as S eno ophomonas) we e excluded om u he analyses (Supplemen a y Tex 1). The iden i y o he mos abundan and p e alen OTUs was de e mined using BLAST45 agains sequences om ype ma e ial and agains all sequences (excluding sequences om model o ganisms o en i onmen al samples) in he n da abase47. Belle ophon ( e sion 3, 200 bp window, Hube -Hugenhol z co ec ion48) was used o iden i y chime ic OTU sequences. Sequence simila i ies g ea e han 97% and 93% we e used as cu -o s o classi y OTUs a species- and genus le el, espec i ely. The a ionale o hese cu -o s was discussed by Ken e s e al.49. Simpli ied classi ica ion o die a y in o ma ion and o he ac o s. The ange o die s consumed by he animals om which he samples came was highly di e se and complex. Fo his eason, and whe e he in o ma ion was a ailable, die s we e ca ego ised in e ms o o age ype, o age plan , and o age o concen a e a io (Supplemen a y Table 7). Die s likely o con ain > 5% s a ch (e.g., whole o g ain c ops o maize, ba ley, whea , ice, as well as pea, po a o, so ghum, e c.) o > 5% pec in (e.g., bee s o legumes such as al al a and clo e ) we e also iden i ied. Animals ha had been ed hei espec i e die s www.na u e.com/scien i ic epo s/ 9 Scien i ic RepoR s | 5:14567 | DOi: 10.1038/s ep14567 o less han a wo-week pe iod we e no ed in Supplemen a y Da a 1. Fac o s such as gende , age, mod- i ica ions (e.g., cannula ion), ea men s (e.g., an ibio ics, d ench, su ge y), a ming condi ions, season, con ac wi h o he animals, and sample p ocessing s eps ha may a ec appa en mic obial commu- ni y composi ions (e.g., DNA ex ac ion me hod, sample ac ion used, sample s o age, e c.) we e also eco ded (Supplemen a y Da a 1). Whe e de ails we e no p o ided, la i ude, longi ude, and ele a ion we e es ima ed using h p://www.mapcoo dina es.ne /en. Clima e zones we e designa ed acco ding o he Köppen-Geige clima e classi ica ion scheme50. S a is ical analyses. The esul ing da ase allowed us o es ablish whe he animal o die a y ac o s ela e o umen and camelid o egu mic obial communi y composi ion, iden i y he dominan mic obes and hei po en ial associa ions, and desc ibe he deg ee o simila i y o umen and camelid o egu mic obial communi ies wo ldwide. S a is ical analyses o mic obial da a we e pe o med using GenS a o Windows51, R so wa e52, and QIIME43. P incipal coo dina e analysis o B ay-Cu is dissimila i y ma ices, analysis o a iance, spa se pa ial leas squa es disc iminan analysis (sPLS-DA, using a sPLS eg ession app oach), and canonical disc iminan analyses (CDA) o mic obial communi y composi ion da a in con ex o he me ada a (Supplemen a y Da a 1) we e used o iden i y impac s o ac o s such as hos lineage, die , e c. on umen and camelid o egu mic obial communi ies and o iden i y he g oups associa ed wi h hese ac o s. Pea son, Spea man, Spa CC53, and egula ised canonical co ela ion analyses (CCA) we e used o iden i y associa ions wi hin and be ween a chaeal, bac e ial, and p o o- zoal g oups. Associa ion sco es we e isualised as ele ance ne wo ks and clus e ed image maps (CIM, hea maps) ep esen ing he i s wo dimensions. González e al. p o ides a comp ehensi e o e iew o sPLS-DA, CCA and he co esponding ‘pai wise associa ions’, ne wo k and CIM echniques and hei applica ion54. Re e ences 1. Hackmann, T. J. & Spain, J. N. In i ed e iew: uminan ecology and e olu ion: pe spec i es use ul o uminan li es ock esea ch and p oduc ion. J. Dai y Sci. 93, 1320–1334 (2010). 2. Ho mann, R. R. E olu iona y s eps o ecophysiological adap a ion and di e si ica ion o uminan s: a compa a i e iew o hei diges i e sys em. Oecologia 78, 443–457 (1989). 3. Hunga e, R. E. The Rumen and i s Mic obes. (Academic P ess, 1966). 4. Ripple, W. J. e al. Ruminan s, clima e change and clima e policy. Na . Clim. Change 4, 2–5 (2014). 5. Johnson, D. E. & Wa d, G. M. Es ima es o animal me hane emissions. En i on. Moni . Assess. 42, 113–141 (1996). 6. Ki elmann, S. e al. Two di e en bac e ial communi y ypes a e linked wi h he low-me hane emission ai in sheep. PLoS ONE 9, e103171 (2014). 7. Jami, E., Whi e, B. A. & Miz ahi, I. Po en ial ole o he bo ine umen mic obiome in modula ing milk composi ion and eed e iciency. PLoS ONE 9, e85423 (2014). 8. Ca be y, C. A., Kenny, D. A., Han, S., McCabe, M. S. & Wa e s, S. M. E ec o pheno ypic esidual eed in ake and die a y o age con en on he umen mic obial communi y o bee ca le. Appl. En i on. Mic obiol. 78, 4949–4958 (2012). 9. Weime , P. J. Redundancy, esilience, and hos speci ici y o he uminal mic obio a: implica ions o enginee ing imp o ed uminal e men a ions. F on . Mic obiol. 6, 296 (2015). 10. McDonald, D. e al. An imp o ed G eengenes axonomy wi h explici anks o ecological and e olu iona y analyses o bac e ia and a chaea. ISME J. 6, 610–618 (2012). 11. Kim, M., Mo ison, M. & Yu, Z. S a us o he phylogene ic di e si y census o uminal mic obiomes. FEMS Mic obiol. Ecol. 76, 49–63 (2011). 12. Pailla d, D. e al. Rela ion be ween phylogene ic posi ion, lipid me abolism and bu y a e p oduc ion by di e en Bu y i ib io-like bac e ia om he umen. An onie an Leeuwenhoek 91, 417–422 (2007). 13. C ee ey, C. J., Kelly, W. J., Hende son, G. & Leahy, S. C. De e mining he cul u able accessibili y o he umen bac e ial mic obiome. Mic ob. Bio echnol. 7, 467–479 (2014). 14. Nyonyo, T., Shinkai, T., Tajima, A. & Mi sumo i, M. E ec o media composi ion, including gelling agen s, on isola ion o p e iously uncul u ed umen bac e ia. Le . Appl. Mic obiol. 56, 63–70 (2013). 15. McAllis e , T. A. e al. Ruminan Nu i ion Symposium: Use o genomics and ansc ip omics o iden i y s a egies o lowe uminal me hanogenesis. J. Anim. Sci. 93, 1431–1449 (2015). 16. Uden, P., Rounsa ille, T. R., Wiggans, G. R. & Van Soes , P. J. The measu emen o liquid and solid diges a e en ion in uminan s, equines and abbi s gi en imo hy (Phleum p a ense) hay. B . J. Nu . 48, 329–339 (1982). 17. Van Soes , P. J. Nu i ional Ecology o he Ruminan . 2 edn, (Co nell Uni e si y P ess, 1994). 18. Godoy-Vi o ino, F. e al. Compa a i e analyses o o egu and hindgu bac e ial communi ies in hoa zins and cows. ISME J. 6, 531–541 (2012). 19. Janssen, P. H. & Ki s, M. S uc u e o he a chaeal communi y o he umen. Appl. En i on. Mic obiol. 74, 3619–3625 (2008). 20. Seedo , H., Ki elmann, S. & Janssen, P. H. Few highly abundan ope a ional axonomic uni s domina e wi hin umen me hanogenic a chaeal species in New Zealand sheep and ca le. Appl. En i on. Mic obiol. 81, 986–995 (2015). 21. Bo el, G. e al. Compa a i e genomics highligh s he unique biology o Me hanomassiliicoccales, a The moplasma ales- ela ed se en h o de o me hanogenic a chaea ha encodes py olysine. BMC Genomics 15, 679 (2014). 22. Seedo , H., Ki elmann, S., Hende son, G. & Janssen, P. H. RIM-DB: a axonomic amewo k o communi y s uc u e analysis o me hanogenic a chaea om he umen and o he in es inal en i onmen s. Pee J 2, e494 (2014). 23. Williams, A. G. & Coleman, G. S. The Rumen P o ozoa. (Sp inge -Ve lag New Yo k Inc., 1992). 24. Deho i y, B. A. in Pa asi ic P o ozoa Vol. 3 (eds J. P. K eie & J. R. Bake ) Ch. 1, 1–42 (Academic P ess, Inc., 1993). 25. Russell, J. B. & Rychlik, J. L. Fac o s ha al e umen mic obial ecology. Science 292, 1119–1122 (2001). 26. Pope, P. B. e al. Me agenomics o he S alba d eindee umen mic obiome e eals abundance o polysaccha ide u iliza ion loci. PLoS ONE 7, e38571 (2012). 27. Naas, A. E. e al. Do umen Bac e oide es u ilize an al e na i e mechanism o cellulose deg ada ion? mBio 5, e01401–e01414 (2014). 28. B yan , M. P. & Small, N. Cha ac e is ics o wo new gene a o anae obic cu ed ods isola ed om he umen o ca le. J. Bac e iol. 72, 22–26 (1956).