Improved bacterial RNA-seq by Cas9-based depletion of ribosomal RNA reads.

Imp o ed bac e ial RNA-seq by Cas9-based deple ion o ibosomal RNA eads GIANLUCA PREZZA, 1 TOBIAS HECKEL, 2 SASCHA DIETRICH, 2 CHRISTINA HOMBERGER, 1 ALEXANDER J. WESTERMANN, 1,3,4 and JÖRG VOGEL 1,3,4 1 Helmhol z Ins i u e o RNA-based In ec ion Resea ch (HIRI), Helmhol z Cen e o In ec ion Resea ch (HZI), Wü zbu g, 97080, Ge many 2 Co e Uni Sys ems Medicine, Uni e si y o Wü zbu g, Wü zbu g, 97080, Ge many 3 Ins i u e o Molecula In ec ion Biology (IMIB), Uni e si y o Wü zbu g, Wü zbu g, 97080, Ge many ABSTRACT A majo challenge o RNA-seq analysis o gene exp ession is o achie e su icien co e age o in o ma i e non ibosomal ansc ip s. In euka yo ic samples, his is ypically achie ed by selec i e oligo(dT)-p iming o messenge RNAs o exclude ibosomal RNA ( RNA) du ing cDNA syn hesis. Howe e , his s a egy is no compa ible wi h p oka yo es in which unc ion- al ansc ip s a egene ally no polyadenyla ed. Too e come his, we adop ed DASH (deple ion o abundan sequences by hyb idiza ion), ini ially de eloped o euka yo ic cells, o imp o e bo h he sensi i i y and dep h o bac e ial RNA-seq. DASH uses he Cas9 nuclease o emo e unwan ed cDNA sequences p io o lib a y ampli ica ion. We epo he design, e alua ion, and op imiza ion o DASH expe imen s o s anda d bac e ial sho - ead sequencing app oaches, including so wa e o au oma ed guide RNA (gRNA) design o Cas9-media ed clea age in bac e ial DNA sequences. Using hese gRNA pools, we e ec i ely emo ed RNA eads (56%–86%) in RNA-seq lib a ies om wo di e en model bac e ia, he G am-nega i e pa hogen Salmonella en e ica and he anae obic gu commensal Bac e oides he aio aomic on. DASH wo ks obus ly, e en wi h subnanog am amoun s o inpu RNA. I s e iciency, high sensi i i y, ease o implemen a ion, and low cos (∼$5 pe sample) ende DASH an a ac i e al e na i e o RNA emo al p o ocols, in pa icula o ma e i- al-cons ained s udies whe e con en ional ibodeple ion echniques ail. Keywo ds: bac e ial RNA-seq; DASH; ibosomal RNA; Cas9; CRISPR; Salmonella;Bac e oides INTRODUCTION The ad en o high- h oughpu RNA sequencing (RNA- seq) has e olu ionized he ield o bac e ial RNA biology (C ouche and Thomson 2010; Hö e al. 2018). RNA-seq has shown ha bac e ial ansc ip omes, once belie ed o be simple in e ms o s uc u e and egula ion, can be almos as complex as hei euka yo ic coun e pa s (So ek and Cossa 2010), and helped o ealize ha bac e- ia amply use pos - ansc ip ional con ol o egula e gene exp ession (Hö e al. 2018). In RNA-seq, he expe imen al s eps p io o sequencing a e uni e sal and consis mainly o RNA ex ac ion, enzyma ic diges ion o genomic DNA, deple ion o ibosomal RNA ( RNA), and con e sion o he emaining RNA pool in o complemen a y DNA (cDNA) lib a ies. The emo al o RNA ( ypically ∼90% o he o al cellula RNA) is impo an as i much inc eases co e age o messenge RNA (mRNA) and egula o y non- coding RNA. A s aigh o wa d me hod o a oid RNA eads in euka yo ic RNA-seq s udies ha nesses oligo(dT) oligonucleo ides ha anneal o he poly(A) ail o mRNAs o selec i ely p ime e e se ansc ip ion (RT). Howe e , p oka yo ic ansc ip s a e no no mally polyadenyla ed (D ey us and Regnie 2002), which necessi a es he de el- opmen o al e na i e RNA emo al s a egies. The mos popula me hods o RNA deple ion om p o- ka yo ic samples ollow a “pull-ou ”s a egy whe eby RNA molecules a e deple ed om a sample wi h comple- men a y oligonucleo ides coupled o magne ic beads. This s a egy unde lies se e al comme cial eady- o-use ki s, es- pecially he popula MICROBExp ess and RiboMinus ki s (The mo Fishe Scien i ic) as well as he Ribo-Ze o echnol- ogy (Illumina). While i gene ally achie es excellen esul s o modelbac e iasuchasEsche ichiacoli, hese ki s a eas- socia ed wi h bo h, high cos ($60–80 pe sample) and a 4 These au ho s con ibu ed equally o his wo k. Co esponding au ho s: alexande .wes e mann@uni-wue zbu g. de, [email p o ec ed] A icle is online a h p://www. najou nal.o g/cgi/doi/10.1261/ na. 075945.120. F eely a ailable online h ough he RNA Open Access op ion. © 2020 P ezza e al. This a icle, published in RNA, is a ailable unde a C ea i e Commons License (A ibu ion-NonComme cial 4.0 In e na ional), as desc ibed a h p://c ea i ecommons.o g/licenses/ by-nc/4.0/. METHOD RNA (2020) 26:1069–1078; Published by Cold Sp ing Ha bo Labo a o y P ess o he RNA Socie y 1069 limi ed e iciency o nonmodel species such as he as majo i y o hose >1000 di e en bac e ia in he human mic obio a. In addi ion, Ribo-Ze o—long conside ed he gold s anda d o bac e ial RNA deple ion (Giannoukos e al. 2012; Pe o a e al. 2017)—was ecen ly discon in- ued. This has enewed e o s in he communi y o de elop in-house solu ions o RNA deple ion (Kim e al. 2019; Cul ine e al. 2020). A ailable al e na i e bac e ial RNA emo al s a egies use RNase H-media ed diges ion o RNA:DNA hyb ids (Huang e al. 2020), size-selec ion o mRNAs h ough liquid ch oma og aphy (Cas o e al. 2013), o selec i e ex- clusion o RNAs om cDNA con e sion wi h “no -so- an- dom”hexame s (Hi akawa e al. 2011; Chugani e al. 2012); he la e is also a ailable as pa o he Uni e sal P oka yo ic RNA-Seq ki (NuGen). O e all, hese me hods achie e good o excellen (∼70%–99%) RNA deple ion a es. Howe e , each o hese p o ocols deple es RNA a he RNA le el, ha is, p io o mul iplexing. This limi s he applicabili y o hese echniques o s udies wi h low- inpu ma e ial o high- h oughpu analysis including sin- gle-cell RNA-seq. Some ecen lib a y p epa a ion al e na- i es such as RNA ag-seq (Shishkin e al. 2015) ba code RNA samples and pool hem o join RNA deple ion, he eby educing cos . Howe e , since ba coding a he RNA le el equi es high inpu amoun s, i oo, is li le sui - ed o low-inpu bac e ial RNA-seq. Ins ead o emo ing RNA om he inpu sample, RNA agmen s migh as well be emo ed a he cDNA le el, ha is, ollowing lib a y p eampli ica ion and mul iplexing. An ea ly s udy diges ed RNA-de i ed cDNAs wi h a dou- ble s and-speci ic DNase a e mel ing and eannealing (Yi e al. 2011). Howe e , his me hod equi es subs an ial op imiza ion o he eannealing condi ions. Mo eo e , he p o ocol is ye o be combined wi h mul iplexing o cDNA. In o he wo ds, i s ill equi es nanog am amoun s o inpu RNA, while i s e iciency o RNA ead deple ion is in e io o he abo e-desc ibed me hods (Yi e al. 2011; Giannoukos e al. 2012). A sec- ond s udy in oduced cDNA-le el deple ion o RNAs h ough p obe-di- ec ed deg ada ion (PDD; [A che e al. 2014]), in which DNA p obes a e annealed o RNA-de i ed cDNAs and deg aded wi h double s and- speci ic DNase. While p omising, his me hod equi es ci cula iza ion o he cDNA agmen s, hampe ing i s in oduc ion in common RNA-seq p o ocols. Recen ly, p og ammed DNA clea - age by he CRISPR-associa ed nucle- ase Cas9 has been in oduced as a no el echnology o deple e wi h high sequence speci- ici y “unwan ed” agmen s om euka yo ic cDNA lib a - ies. In his so-called DASH (deple ion o abundan sequences by hyb idiza ion) app oach (Gu e al. 2016), a pool o single-guide RNAs (sgRNAs) is used o di ec Cas9 clea age o RNA-de i ed cDNA molecules du ing lib a y p epa a ion. Since he clea ed agmen s a e no ampli ied in he subsequen PCR s ep, in ac non- RNA agmen s become en iched (Fig. 1). To illus a e he powe o DASH, a pool o iled ( oughly e e y 50 bases) sgRNAs agains human mi ochond ial RNAs educed he co esponding cDNA eads by mo e han 1000- old, while i concomi an ly inc eased co e age o non iboso- mal ansc ip s by ∼2.4- old. DASH has also been used o inc ease co e age o nonabundan ansc ip s in human small RNA-seq lib a ies by a ge ing adap e di- me s and issue-speci ically highly exp essed mic oRNAs (Ha digan e al. 2019) and o deple e hemoglobin an- sc ip s om pola bea pe iphe al blood RNA samples be o e long- and sho - ead sequencing (By ne e al. 2019). Simila o DASH o euka yo ic cDNA, a he mos a- ble Cas9 a ian was ecen ly used o clea e E.coli 16S RNA sequences in cDNA du ing PCR-media ed lib a y ampli ica ion (Schmid e al. 2019). The p omise o DASH o p o ide a gene ic app oach no wi hs anding, he me hod is ye o be ully es ablished o bac e ial ansc ip omics. In his s udy, we e alua e DASH o bac e ial sho - ead RNA-seq o wo in ensely s udied bac e ia: Salmonella en e ica se o a Typhimu ium (hence o h, Salmonella), which is a majo model species o bo h, bac e ial RNA biology and pa ho- genesis; and Bac e oides he aio aomic on as an example o an abundan human mic obio a species. The p esen ed sgRNA design so wa e and op imized we -lab p o ocols FIGURE 1. O e iew o he bac e ial DASH wo k low. Schema ic o he p inciple behind DASH-media ed emo al o RNA-de i ed cDNA agmen s om sequencing lib a ies. The in- di idual s eps in he bac e ial DASH pipeline a e indica ed in he zoom-in a he bo om. P ezza e al. 1070 RNA (2020) Vol. 26, No. 8 bea g ea po en ial o e icien , sensi i e, and economic emo al o unwan ed RNA sequences o RNA-seq analy- sis o any bac e ium o in e es . RESULTS P oo -o -concep o bac e ial DASH and op imiza ion o Cas9 eac ion condi ions The well-s udied G am-nega i e bac e ium Salmonella has been he subjec o many RNA-seq s udies (K oge e al. 2013; Wes e mann e al. 2016). Unless deple ed, Salmonella RNA ypically ep esen s ∼95% o all cDNA eads in a lib a y (Fig. 2A; Be in e al. 2019). Salmonella RNA is ansc ibed om se en ibosomal ope ons ( n) ac oss he ch omosome, gi ing ise o each se en homo- logs o he 16S (genes sA-G) and 23S ( lA-G), and eigh copies o 5S RNA ( he nD locus ca ies wo 5S RNA-en- coding genes). Gene ally speaking, he sequences o he RNA genes a e highly conse ed ac oss all loci, bu hey do show indels and poin mu a ions. To deple e ibosomal agmen s in cDNA om Salmonella o al RNA, we ini ially designed a pool o sgRNAs a ge ing he consensus sequence o each RNA gene. To his end, we w o e a Py hon sc ip ha iden i ies SpCas9 a ge si es wi hin RNA genes based on he ol- lowing selec ion c i e ia: (i) pe ec complemen a i y o a leas i e ou o he se en (16S and 23S) o six ou o he eigh (5S) RNA copies, (ii) a GC con en o 35%–70%, (iii) a p edic ed low endency o o m seconda y s uc u es, and (i ) ollowed by he “NGG”p o ospace adjacen mo- i (PAM) o ecogni ion by he SpCas9 nuclease (Jinek e al. 2012). The sc ip aimed a iden i ying one a ge si e ha sa is ied he se c i e ia on a e age e e y ∼50 bp along he a ge space, as his densi y p o ed su icien in euka yo ic DASH (Fig. 2B; see Ma e ials and Me hods o de ails on sgRNA design; Gu e al. 2016). As a esul , ou sc ip p oposed 113 sgRNAs in o al. We hen o de ed DNA oligonucleo ides, each comp ising a T7 p omo e , a single a ge si e, and he i s pa o he SpCas9 sgRNA sca old. To hese sense oligonucleo ides, we added a uni- e sal, pa ially o e lapping e e se DNA oligonucleo ide con aining he emaining po ion o he sgRNA sca old. The esul ing annealing p oduc s we e illed up wi h he co esponding nucleobases, gi ing ise o he dou- ble-s anded DNA empla es ha we e subjec ed o in i o ansc ip- ion wi h T7 RNA polyme ase o yield he sgRNA pool. Salmonella cDNA was p eampli ied in wo PCR cycles and subsequen ly incuba ed wi h he in i o- ansc ibed sgRNA pool and SpCas9 nuclease o 2 h a 37°C. We used a mola sgRNA: Cas9:cDNA a io o 1000:100:1—as in e ed as op imal in he o iginal DASH p o ocol (Gu e al. 2016). Following diges ion, we emo ed Cas9 wi h a silica-based column pu i i- ca ion ki and p oceeded wi h 16 cy- cles o PCR o ampli y he unclea ed cDNA agmen s. Sequencing o he esul ing lib a y e ealed ha , e en a e Cas9 clea age, ∼83% o he ob- ained Salmonella eads de i ed om RNA, hus a me e ∼13% educ ion o e he con ol lib a y (Fig. 2C). Using he same sgRNA pool, we hen es ed whe he di e en eac ion condi ions would inc ease deple ion e iciency. Howe e , highe Cas9 con- cen a ions only made RNA deple- ion less e icien , sugges ing ha — among he concen a ions es ed— he abo e sgRNA:Cas9 a io o B AC FIGURE 2. Ini ial DASH un on Salmonella o al RNA. (A) Composi ion o o al RNA ex ac ed om exponen ially g owing Salmonella as deduced om RNA-seq eads. (B) Conse a ion o Salmonella RNAs genes and loca ion o he iden i ied sgRNA a ge si es. The pe -base con- se a ion o each RNA gene ac oss all Salmonella RNA homologs is plo ed. Abo e each con- sensus sequence ( ep esen ed as a g ay ec angle), he designed sgRNAs a ge si es a e indica ed by a black iangle. (C) Pilo un o DASH-media ed RNA deple ion in Salmonella. (Le ) The ac ion o ibosomal-de i ed eads o e he o al numbe o mapped eads o a con- ol and a DASH- ea ed lib a y. (Righ ) Same da a a e exp essed as e iciency o DASH-medi- a ed RNA deple ion. RNA deple ion in bac e ia ia DASH www. najou nal.o g 1071 1000:100 was op imal o bac e ial DASH, oo (Fig. 3A). Howe e , p eincuba ion o Cas9 wi h he sgRNA pool, mean o allow o mo e ime o he ibonucleop o ein complex o o m p io o addi ion o he cDNA subs a e, did inc ease RNA deple ion e iciency o ∼20% (Fig. 3B). Based on his inding, a Cas9:sgRNA p eincuba ion s ep was included in all u he eac ions. Maximizing sgRNA densi y The abo e-desc ibed ini ial ∼20% deple ion o RNA eads om Salmonella o al RNA lib a ies was a a c y om he >99% deple ion p e iously epo ed o human RNA sam- ples (Gu e al. 2016). Inspec ion o he ead leng h dis ibu- ion ob ained om ou “DASHed”Salmonella lib a ies (1000:100 sgRNA:Cas9 a io) con i med e icien deple- ion o RNA-de i ed eads >50 n (Fig. 3C). Howe e , sho e RNA eads we e e en en iched o e he un ea ed con ol sample. Fo compa ison, he DASH ea men ha dly a ec ed he ead leng h dis ibu ion o non iboso- mal eads (Fig. 3D). This sugges ed ha sho ibosomal cDNA agmen s e aded Cas9 clea age, p esumably because hey we e less likely o con ain a ull-leng h sgRNA a ge sequence. To es whe he co e ing mo e si es wi hin he RNA se- quences would mo e e icien ly emo e hose e ac o y sho RNA eads, we modi ied ou design ool o ob ain a sgRNA pool wi h he maximal numbe o a ge si es (Fig. 4A). This new sgRNA pool a ge ed all copies o he RNA genes indi idually a he han jus hei consensus sequences. Disca ding sgRNAs wi h po en ial o - a ge e ec s, his inal pool consis ed o 797 sgRNAs. Wi h he same mola a io o 1000:100:1 (sgRNA:Cas9:cDNA) as abo e, his pool b ough RNA deple ion e iciency o 38% (Fig. 4B). Gi en he much la ge numbe o sgRNAs in his new pool, we specula ed ha highe mola excess o Cas9:sgRNA o e cDNA could u he imp o e deple- ion e iciency. Indeed, changing he mola a io o 35,000:3500:1 (sgRNA:Cas9:cDNA) u he inc eased he deple ion e iciency, up o 56% (Fig. 4B). Consequen ly, he numbe o de- ec ed (RPKM > 1) Salmonella an- sc ip s a he se sequencing dep hs o ∼5–12 M eads inc eased om 2724 in he un ea ed sample o 3875 in he DASH- ea ed lib a y (Supplemen al Fig. S1A). To assess whe he any sgRNA in his la ge pool showed signs o o - a ge ing, we compa ed gene-wise ead coun s be ween he DASH- ea - ed and cogna e un ea ed lib a ies. Excluding RNA, ead coun s o all gene ic ea u es showed e y high co ela ion be ween he wo lib a ies (Fig. 4C). Likewise, ead co e age ac oss RNA genes dec eased upon DASH, libe a ing mo e in o ma i e eads ha mapped o mRNAs (e.g., dnaK) o egula o y small RNA (sRNA) sequences (PinT, In R, ChiX; Fig. 4D). Toge he , his demons a es ha ou design pipeline indeed se- lec s sgRNAs wi h e y li le o - a - ge ing and so enables speci ic emo al o RNA eads. Inc easing he sensi i i y o DASH One key ad an age o RNA deple- ion a he cDNA le el is ha i does no lowe he amoun o s a - ing ma e ial o lib a y p epa a ion. The e o e, we es ed whe he DASH B A C D FIGURE 3. Op imiza ion o DASH eac ion condi ions. (A) RNA deple ion e iciency upon DASH clea age wi h a ying amoun s o Cas9 and he sgRNA pool. Cas9 and sgRNA amoun s a e indica ed as mola excess o e a single agmen o he cDNA lib a y. (B) Impac o p ein- cuba ion o sgRNA and Cas9 p io o DASH on RNA deple ion e iciency. (C) Leng h dis ibu- ion o he mapped po ions o eads aligning o RNAs a e DASH compa ed wi h an un ea ed con ol. F equency alues a e exp essed as ac ion (%) o he o al numbe o eads mapping o RNAs. (D) Same as in Cbu o eads aligning o non- RNA genes. P ezza e al. 1072 RNA (2020) Vol. 26, No. 8 could imp o e he sensi i i y o ou RNA-seq p o ocol by sys ema ically dec easing he quan i y o inpu RNA om ∼800 ng o ∼0.4 ng (Fig. 5A). Using a ixed 1000:100:1 mo- la a io o sgRNA:Cas9:cDNA, we obse ed e icien RNA deple ion wi h each o he ou amoun s es ed, albei e i- ciency a ied om ∼30% o ∼50% (Fig. 5A). Impo an ly, he numbe o de ec ed gene ic ea u es (which is a mo e obus eadou ) wi h RPKM > 1 was s able in he wo highes RNA inpu amoun s and dec eased only in he lowes one (Supplemen al Fig. S1B, op), likely due o he s ochas ic loss o low-abun- dance ansc ip s. Fo high-abun- dance ansc ip s (RPKM > 25), DASH inc eased he numbe o de ec ed genes i espec i e o inpu amoun (Supplemen al Fig. S1B, bo om). In he o iginal euka yo ic DASH p o- ocol(Gue al.2016), heCas9enzyme is emo ed a e he clea age eac ion by pu i ica ion o e a column, which uns he isk o losing cDNA as well. He e, we implemen ed diges o he Cas9 p o ein by p o einase K (Ha di- gan e al. 2019) as anal e na i e ocol- umn pu i ica ion in bac e ial DASH. T ea men wi h p o einase K did no a ec he emo al o RNA eads (Fig. 5A), bu esul ed in inc eased cDNA yields a e he pos -DASH PCR ampli- ica ion (Supplemen al Table S1). Las ly, we es ed whe he DASH wo ks wi h a lib a y p epa a ion ki ha is op imized o low inpu sam- ples. Using he Taka a SMARTe S anded To al RNA-Seq ki wi h 1 ng o al Salmonella RNA as inpu , we success ully emo ed mo e han hal o he RNA eads (Fig. 5B). Con- comi an ly, and simila o he abo e li- b a ies gene a ed wi h he NEBNex ki , he p opo ion o mRNA and sRNA eads inc eased by app oxi- ma ely nine old (Fig. 5C). Howe e , DASH ea men led o a sligh in- c ease in he ac ion o eads <12 n in leng h, which should no be a majo conce n since hese eads a e ypi- cally il e ed ou du ing ead p ocess- ing (Supplemen al Fig. S2A). Taken oge he , his demons a es ha combining DASH wi h a lib a y con- s uc ion p o ocol op imized o mi- nu e RNA amoun s enables obus RNA-seq analysis o low-inpu sam- ples ex ac ed om as ew as ∼1000 bac e ia, and po en- ially e en ewe . DASHing Bac e oides he aio aomic on RNA To add ess gene alizabili y o bac e ial DASH, we selec ed a phylogene ically dis an species. The Bac e oidia ep e- sen a i e and human in es inal mic obio a membe B. B A C D FIGURE 4. Imp o ed DASH e iciency by maximizing sgRNA a ge si e densi y. (A) Wo k low o he so wa e de eloped o he design o high-densi y sgRNA pools. The numbe o sgRNAs passing each s ep is indica ed o he igh .(B) RNA deple ion e iciency using he 797 sgRNA pool in di e en a ios o Cas9 and sgRNA. Cas9 and sgRNA amoun s a e indica ed as mola excess o e a single agmen o he cDNA lib a y. (C) Co ela ion o ansc ip abundances in he con ol e sus DASHed (3500:35,000 excess o Cas9:sgRNA) lib a ies. Red do s (n= 22) ep esen RNA ansc ip s, blue ones (n= 1511) he gene ic ea u es wi h a leas 15 eads in he con ol and 150 eads in he DASH lib a y, and g ay do s (n= 3469) he gene ic ea u es below his h eshold. The eg ession line and co ela ion coe icien we e compu ed o he blue do s only. (D) Sequence ead co e age o a ep esen a i e RNA locus and he dnaK, pinT, in R, and chiX genes in he con ol and DASH lib a ies (3500:35,000 excess o Cas9: sgRNA). RNA deple ion in bac e ia ia DASH www. najou nal.o g 1073 he aio aomic on ha bo s i e RNA ope ons in i s genome, each ca ying one 23S, 16S, and 5S gene copy. Using ou cus om sc ip , we designed 651 sgRNAs a ge ing all B. he aio aomic on RNA ope ons indi idually (same selec- ion c i e ia as o he op imized Salmonella RNA deple- ion; see Fig. 4A). These sgRNAs we e es ed in pa allel wi h he wo di e en lib a y cons uc ion ki s and sgRNA: Cas9:cDNA mola a ios used o Salmonella samples (Figs. 4A, 5), on nanog am amoun s o B. he aio aomic on RNA (Fig. 6A). We obse ed deple ion o RNA eads up o 86% o he s anda d lib a y p epa a ion ki (NEB) and up o 76% o he low-inpu p o ocol (Taka a), wi h a co espond- ing inc ease in co e age o non- RNA ansc ip s (Fig. 6B,D; Supplemen al Fig. S2B). As be o e, he e was a high co e- la ion wi hin abundance o gene ic ea u es be ween he un ea ed and he DASH samples (Fig 6C), a guing o neg- ligible i any o - a ge ing by he sgRNAs. DISCUSSION Fi s sys ema ic e alua ion o DASH o bac e ial ansc ip omics This s udy epo s he adap a ion o he DASH echnology o p oka yo ic RNA samples. Th ough Cas9-media ed clea age o RNA-de i ed cDNA agmen s p io o lib a y ampli ica ion, ou p o ocol inc eases co e age o non ibosomal ansc ip s in Salmonella and Bac e oides o al RNA samples by ∼12- o ∼3.8- old, e- spec i ely. Only a ew changes o he no mal lib a y p epa a ion p o ocol we e necessa y o implemen ou DASH p o ocol in o a s anda d Illu- mina sho - ead sequencing pipeline. Since i emo es RNA agmen s a e RT, DASH o e s he majo ad- an age ha an ini ial ampli ica ion o he cDNA lib a y can be pe o med be o e Cas9 clea age. This inc eases he o e all amoun o cDNA and min- imizes s ochas ic agmen loss. Fo his eason, ou app oach eaches good deple ion (∼50%–80%) e en wi h minu e RNA amoun s well below he lowe ecommended limi o he common RNA deple ion ki s and echniques (Supplemen al Table S2). Impo an ly, he ecommended mini- mal inpu amoun o he “gold s an- da d”Ribo-Ze o ki was 500 ng RNA, whe eas ou DASH app oach wo ks obus ly o ∼400 pg o inpu RNA (Fig. 5A). Gi en ha we ha e succeed- ed in combining DASH wi h s a e-o - he-a lib a y p epa a ion ki s used in euka yo ic single- cell ansc ip omics, we expec o be able o u he educe he necessa y amoun o s a ing ma e ial in he u u e. Ob iously, his would open bac e ial RNA-seq o many ex- ci ing a eas o mic obiology; o gi e jus one example, i would allow one o pe o m gene exp ession p o iling on bac e ia eco e ed om insec gu s. While he up on in es men o pu chasing all DASH eagen s is high (Supplemen al Table S4), we es ima e a cos o $3–7 pe sample o RNA-seq lib a ies, which is >10- old lowe han o comme cial RNA deple ion ki s (Supplemen al Table S2). In his ega d, bac e ial DASH will emain compe i i e e en wi h a e y ecen ly eleased new Ribo-Ze o ki (“Ribo-Ze o Plus,”Illumina ca alog num- be : 20037135), which despi e now using enzyma ic RNA deple ion ins ead o RNA pull-ou , s ill uns a ∼$80 pe sample. Wha is mo e, DASH bea s po en ial o u he cos educ ion, o example, h ough in-house p oduc ion o he Cas9 p o ein o he T7 RNA polyme ase o in i o ansc ip ion. Al hough op imized on Salmonella RNA samples, he condi ions es ablished he e enabled us o success ully un DASH on a phylogene ically dis an bac e ium, B. he aio- aomic on. This a gues ha ou p o ocol is applicable o o- al RNA om di e se bac e ial species and, po en ially, e en oo ganismsbeyond hebac e ialkingdom.Inp inciple, he B A C FIGURE 5. DASH o low-inpu RNA samples. (A) DASH e iciency o s eadily dec easing inpu RNA amoun s and o di e en Cas9 emo al me hods. DASH was pe o med on 1/5 h o he cDNA esul ing om e e se ansc ip ion. Cas9 and he sgRNA pool we e used in a 1000 and 10,000 excess o e a single agmen o he cDNA lib a y, espec i ely. The indica ed RNA amoun s co espond o he e e se ansc ip ion inpu , while he cDNA amoun s e e o wha was used o DASH. (B) Deple ion e iciency o DASH (3500:35,000 excess o Cas9: sgRNA) when combined wi h di e en lib a y p epa a ion ki s. The indica ed RNA amoun s co espond o he e e se ansc ip ion inpu . (C) RNA class dis ibu ion o sequencing eads in he con ol and DASH samples shown in panel B. P ezza e al. 1074 RNA (2020) Vol. 26, No. 8 modula i y o ou DASH app oach should allow o he de- sign o combined sgRNA pools a ge ing di e en species o ibodeple ion o samples de i ed om mixed popula- ions, such as me a ansc ip omic samples o RNA mix u es isola ed om in ec ed hos cells and issues. Compa ison o p e ious DASH p o ocols The o iginal desc ip ion o DASH on euka yo ic samples (Gu e al. 2016) epo ed a educ ion o he a ge ed ag- men s by 99%, subs an ially highe han wha we achie ed he e in bac e- ia. How can his di e ence be ex- plained? Bo h he o iginal DASH (Gu e al. 2016) and a ecen ly upda ed p o ocol (Dyne man e al. 2020) we e combined wi h long- ead sequencing whe e a e age inse size was ∼300 n . Simila ly, he p e ious DASH-like expe imen wi h he mos able Cas9 was applied o an E.coli cDNA lib a y wi h an a e age inse size o 300–400 n (Schmid e al. 2019). In con as , we desc ibe he applica ion o DASH o a s anda d Illumina RNA-seq pipe- line wi h a maximal ead leng h o 75 n . Ob iously, he longe he inse s, he highe he numbe o a ge able agmen s, which biases DASH o- wa d longe eads (Fig. 3C). Howe e , sho - ead sequencing is he s anda d in he ield o bac e ial RNA-seq and we he e o e p edic ou DASH e - sion o be pa icula ly use ul o any ansc ip omics app oach ha in- ol es bac e ia. Addi ionally, ou im- p o ed DASH p o ocol omi s he mul iple phenol/chlo o o m ex ac- ion o column pu i ica ion s eps o euka yo ic DASH (Gu e al. 2016; Dyne man e al. 2020). Ins ead, we e- mo e Cas9 wi h a simple p o einase K ea men p io o u he lib a y am- pli ica ion. Since his minimizes he isk o cDNA loss om o ganic ex ac- ion o silica column pu i ica ion, ou p o ocol will be pa icula ly sui able o low-inpu samples. By using a he mos able Cas9, Quake and colleagues ecen ly dem- ons a ed ha DASH could be pe - o med simul aneously wi h cDNA lib a y ampli ica ion (Schmid e al. 2019). Howe e , he he mos able Cas9 a ian used in he s udy equi es a complex, 6 n - long PAM, which d ama ically educes he numbe o pos- sible sgRNA si es wi hin he RNA sequence space. Using ou Py hon sc ip wi h his hexame ic PAM, a mos 115 sgRNAs (as compa ed o 797 sgRNAs o SpCas9) could be designed o Salmonella RNA. This is abou he di e - ence be ween ou ini ial (Fig. 3) and he inal (Fig. 4) sgRNA pools, which ansla es in a h ee old di e ence in RNA ead emo al. The e o e, al hough he classical Cas9 om S ep ococcus pyogenes equi es clea age and PCR ampli ica ion o occu subsequen ly (no in pa allel), B A C D FIGURE 6. DASH-media ed emo al o Bac e oides he aio aomic on RNA. (A) RNA deple- ion e iciency om a B. he aio aomic on cDNA lib a y. DASH was combined wi h he di e en lib a y p epa a ion ki s and using he indica ed a ios o Cas9 and sgRNA o e cDNA ag- men s. The indica ed RNA amoun s co espond o he e e se ansc ip ion inpu . (B) RNA class dis ibu ion o eads in con ol and DASH (3500:35,000 excess o Cas9:sgRNA) samples. (C) Co ela ion o ansc ip abundances in he con ol and DASH lib a ies shown in panel B. Red do s (n= 15) ep esen RNA ansc ip s, blue ones (n= 2474) he gene ic ea u es wi h a leas 15 eads in he con ol lib a y and 150 eads in he DASH sample, and g ay do s (n= 2658) he ea u es below his cu o . The eg ession line and co ela ion coe icien we e com- pu ed o he blue do s only. (D) Sequence ead co e age o a ep esen a i e RNA locus and BT_3550 (encoding a pu a i e long-chain a y acid-CoA ligase) and he gibS and BTnc201 egula o y RNA genes (bo om) in he con ol and DASH lib a y o panels Band C. RNA deple ion in bac e ia ia DASH www. najou nal.o g 1075 i has an ad an age o e he mos able Cas9 wi h espec o RNA ead deple ion in sho - ead lib a ies. sgRNA design ool As pa o his wo k, we de eloped a Py hon sc ip o de- signing sgRNAs a ge ing he RNAs o a selec ed species wi h known e e ence genome (including anno a ions o ibosomal genes) ha ou pu s he sequences o he DNA oligonucleo ides needed as empla es o in i o- an- sc ibe he cus omized sgRNA pool. Impo an ly, howe e , ou so wa e can also be ed wi h manually en e ed coo di- na es o RNA genes, which will be impo an o o gan- isms ha lack a comple e ansc ip ome anno a ion, such as many ele an mic obio a membe s and impo an en i- onmen al bac e ia. Based on he assump ion ha maximizing sgRNA densi- y imp o es deple ion e iciencies, ou pipeline p edic s all possible sgRNAs and il e s ou only hose sequences wi h ex eme GC con en (<30% o >80%) o p edic ed o - a - ge e ec s. Howe e , ou algo i hm does no emo e guides wi h low p edic ed on- a ge ac i i ies, as we pos u- la e ha —as long as ee Cas9 molecules a e no he a e- limi ing ac o —indi idual sgRNAs wi h low on- a ge ac- i i y would no nega i ely impac ibosomal deple ion e - iciency by he en i e pool. In his espec , ou so wa e di e s om he many CRISPR design ools ha ha e been de eloped o genome edi ing (Liu e al. 2020) and sea ch o he “bes ”sgRNA pe each a ge gene/locus. Pe spec i e Fu he op imiza ion o he DASH app oach could include es ing al e na i e Cas nucleases (Gona opoulos- Pou na zis e al. 2020; Wessels e al. 2020), o example, high- ideli y Cas e sions and enzymes wi h al e ed PAM p e e ence o ele a ed he mos abili y (Schmid e al. 2019). Mo eo e , a be e unde s anding o he minimal sgRNA densi y o sa u a ed deple ion e iciency could help o educe bo h, he cos o he sgRNA empla e pool and he amoun s o Cas9 and sgRNA pe eac ion. Among he es ed in e als, we iden i ied a Cas9:sgRNA a io o 1:10 as op imal; howe e , e alua ing mo e e ined, in e media e a ios in he u u e could esul in mo e e i- cien a ge clea age. Finally, i is likely ha mul iple ounds o DASH on he same sample (Dyne man e al. 2020) lead o mo e e icien deple ion. MATERIALS AND METHODS RNA isola ion Bac e ial RNA was isola ed om an in i o cul u e o S.en e ica se o a Typhimu ium s ain SL1344 (S ocke e al. 1983) g own in Lennox b o h (LB) medium o an op ical densi y a 600 nm (OD 600 ) o 2.0 o om a cul u e o B. he aio aomic on VPI-5482 g own in TYG medium o an OD 600 o 0.5. To his end, cells we e ha es ed and o al RNA ex ac ed using he TRIzol eagen (In i ogen) acco ding o he manu ac u e ’s ecommenda ions. To emo e con amina ing genomic DNA, samples we e u he ea ed wi h 0.25 U o DNase I (Fe men as) pe 1 µg o RNA o 45 min a 37°C, ollowed by phenol–chlo o o m ex ac ion and e hanol p ecipi a ion o he RNA ansc ip s. RNA quali y was checked on an Agilen 2100 Bioanalyze (Agilen Technologies). sgRNA design and syn hesis Ta ge sequences wi hin Salmonella RNA genes we e iden i ied and selec ed wi h wo e sions o a cus om Py hon sc ip . In he i s e sion (Figs. 2–4), all RNA copies we e aligned wi h MUSCLE (Edga 2004) and he consensus sequence o each gene was gene a ed so ha all posi ions iden ical in a leas six (16S, 23S) o se en (5S) RNA copies we e main ained. All 20-n po en ial gRNA a ge s we e iden i ied by sea ching bo h s ands o he p esence o he “NGG”PAM and hen il e ed o emo e hose si es wi h an ex eme GC con en (i.e., GC < 35% o >70%) o s ong p edic ed seconda y s uc u es (MFE < −5, as compu ed wi h RNA old [Lo enz e al. 2011]). Wi hin he emaining pool, gRNAs we e hen selec ed o be ∼50 n dis an om each o he , s a ing om he one closes o he 5′end o he RNA sequence. The esul ing 113 sequences we e pu chased om IDT as a unique oligo pool, each wi h he ollowing s uc u e (5′ o 3′): T7 p omo e (TTCTAATACGACTCACTATA) + gRNA sequence + sca old (GTTTTAGAGCTAGAAATAGC). Since ac i i y o he T7 p omo e is enhanced when wo G’s a e p esen a he ansc ip- ion s a si e, one o wo G’s we e added immedia ely a e he T7 p omo e in oligos de i ed om gRNAs s a ing wi h a single o no G, espec i ely. The second e sion o he Py hon sc ip (Figs. 4, 5) designed all possible gRNA a ge sequences, independen o he conse a- ion and s uc u edness o each egion. The only il e ing c i e ia we e GC con en be ween 30%–80% and a low p edic ed o - a - ge p obabili y, de ined as he absence o sequences in he Salmonella ch omosome o plasmids ha aligned o he gRNA wi h up o h ee misma ches (iden i ied wi h Bow ie [Langmead e al. 2009]), ollowed by a alid PAM. The esul ing 979 sequenc- es we e pu chased om IDT as an “oPools Oligo Pool”wi h a sim- ila s uc u e han abo e, excep ha he sca old was GTTTTA GAGCTAGAAATAGCAAGTTAAAATAAGGCTAGTCCGTTATCA ACTTGAAAAAGTGGCACCGAGTCGGTGCTTTTTT and was ol- lowed by a sequence o ill-in eac ions (ACGATGTCGCAG AGTATGCC). The p ime used o illing-in was 5′-GGCA TACTCTGCGACATCGT-3′. Design o he B. he aio aomic on pool was done as abo e, esul ing in 651 sequences. The sc ip is eely a ailable on h ps://gi hub.com/gp ezza/DASH_ RNA_deple ion. dsDNA empla es o in i o ansc ip ion we e gene a ed in a ill-in eac ion pe o med wi h he KAPA HiFi Ho S a ReadyMix (KAPA Biosys ems). The i s pool eac ion was p imed wi h 5′-AA AAGCACCGACTCGGTGCCACTTTTTCAAGTTGATAACGGACT AGCCTTATTTTAACTTGCTATTTCTAGCTCTAAAAC-3′and con- sis ed o dena u a ion a 95°C o 3 min, annealing, and ex ension om 95°C o 30°C a 0.1°C/sec wi h 10 sec pause e e y 10°C and a inal ex ension a 72°C o 1 min. The second pool was illed-in P ezza e al. 1076 RNA (2020) Vol. 26, No. 8 wi h he 5′-GGCATACTCTGCGACATCGT-3′p ime and dena- u a ion a 95°C o 3 min, annealing a 60°C o 20 sec, and ex en- sion a 72°C o 1 min. sgRNA pools we e in i o- ansc ibed om 300 ng o column- pu i ied dsDNA empla e wi h he MEGAsho sc ip T7 T ansc ip ion Ki (The mo Fishe Scien i ic) and hen pu i ied wi h he Mona ch RNA Cleanup Ki (NEB). SpCas9 p o ein was pu chased om NEB (M0386M). cDNA lib a y gene a ion, Cas9 clea age, lib a y ampli ica ion, and Illumina sequencing Bac e ial o al RNA was agmen ed a 94°C o 2.75 min using he NEBNex Magnesium RNA F agmen a ion Module (NEB), de- phospho yla ed a he 3′end, phospho yla ed a he 5′end and decapped using 10 U T4-PNK ± 40 nmol ATP and 5 U RppH, e- spec i ely (NEB). A e each s ep, RNA was pu i ied wi h he Zymo RNA Clean & Concen a o ki (Gu e al. 2016). cDNA lib a - ies we e gene a ed wi h he NEBNex Mul iplex Small RNA Lib a y P ep Ki (NEB) and p eampli ied wi h wo cycles o PCR. Following pu i ica ion wi h he Oligo Clean & Concen a o ki (Zymo Resea ch), DASH ea men was pe o med simila o Gu e al. (2016). Speci ically, he pu i ied cDNA lib a y was incuba ed wi h he Cas9-sgRNA complex o 2 h a 37°C a he indica ed mo- la a ios. Whe e men ioned, Cas9 and he sgRNA pool we e p e- incuba ed a 37°C o 15 min be o e addi ion o he cDNA. A e he diges , Cas9 was emo ed om he eac ion by column pu i- ica ion wi h he Oligo Clean & Concen a o ki (Zymo Resea ch) o ea men wi h 0.8 U (∼20 µg) p o einase K (NEB) o 15 min a 37°C, ollowed by hea -inac i a ion (15 min a 95°C). The esul ing DASHed samples we e PCR ampli ied o 12–24 cycles o selec o non ibosomal, undiges ed cDNAs and pu i ied wi h MagSi- NGS p ep Plus beads (S einb enne Labo sys eme). Al e na i ely, cDNA lib a ies we e gene a ed om bac e ial o- al RNA using he Taka a SMARTe S anded To al RNA-Seq Ki 2 wi h 4 min RNA agmen a ion a 94°C and i e cycles o PCR o cDNA lib a y p eampli ica ion. A e column pu i ica ion, DASH was pe o med as desc ibed abo e. The esul ing DASHed sam- ples we e column pu i ied, PCR ampli ied wi h Taka a’s SeqAmp DNA Polyme ase o 18 cycles and u he pu i ied wi h AMPu e XP beads (Beckman Coul e ). Sequencing o lib a ies, spiked wi h 5% PhiX con ol lib a y, was pe o med in single-end mode on he Illumina Nex Seq 500 pla - o m wi h he Mid Ou pu Ki 2.5 (75 cycles). A summa y o all se- quenced samples and he espec i e eac ion condi ions is epo ed in Supplemen al Table S3. Demul iplexed FASTQ iles we e gene a ed wi h bcl2 as q2 2.20.0.422 (Illumina). The sequencing da a is cu en ly being up- loaded a NCBI Gene Exp ession Omnibus (h p://www.ncbi.nlm .nih.go /geo) unde he accession numbe GSE147155. Da a analysis Reads we e immed o NEBNex o Illumina T uSeq (Taka a ki ) adap e sequences using Cu adap e sion 2.5 wi h de aul pa- ame e s and he –nex seq- im=20 swi ch o handle wo colo se- quencing chemis y. Reads ha we e immed o leng h 0 we e disca ded. P ocessed eads we e mapped o he Salmonella (NC_016810.1, NC_017718.1, NC_017719.1, NC_017720.1) o Bac e oides (NC_004663.1, NC_004703.1) e e ence sequences. We modi ied he NC_016810.1 Salmonella ch omosome anno a- ion o include an upda ed sRNA anno a ion (Hö e al. 2020). The B. he aio aomic on sRNA anno a ion s ems om D Ryan, L Jenniches, S Reicha d , e al. (in p ep.). Mapping was pe o med wi h READemp ion e sion 0.4.3 (Fö s ne e al. 2014) wi h he a - gumen –a80(NEB samples) o –a80–R(Taka a samples) and wi h segemehl 0.2.0 (Ho mann e al. 2009). Gene quan i ica ion was done wi h he READemp ion subcommand gene_quan i wi h a gumen s –a–o10. Co e age plo s we e gene a ed wi h he subcommand co e age and isualized wi h IGV (Robinson e al. 2011). Read leng h dis ibu ion was analyzed wi h SAM ools (Li e al. 2009). RNA deple ion e iciency was de ined as 100 −DASH RNA eads % ∗100 no DASH RNA eads % . SUPPLEMENTAL MATERIAL Supplemen al ma e ial is a ailable o his a icle. ACKNOWLEDGMENTS We hank Elena Ka zowi sch om he Co e Uni SysMed a he Uni e si y o Wü zbu g o excellen echnical suppo . This wo k was suppo ed by he In e disziplinä e Zen um ü Klinische Fo schung (IZKF) a he Uni e si y o Wü zbu g (p ojec Z-6). G.P. was suppo ed by a g an o he Ge man Excellence Ini ia i e o he G adua e School o Li e Sciences, Uni e si y o Wü zbu g. Recei ed Ap il 18, 2020; accep ed May 1, 2020. REFERENCES A che SK, Shi okikh NE, P eiss T. 2014. Selec i e and lexible deple- ion o p oblema ic sequences om RNA-seq lib a ies a he cDNA s age. BMC Genomics 15: 401. doi:10.1186/1471-2164-15-401 Be in V, Pena anda C, Bandyopadhyay N, Yang R, Abi ua A, Bha acha yya RP, Fan A, A aham R, Li ny J, Sho esh N, e al. 2019. Hyb idiza ion-based cap u e o pa hogen mRNA enables pai ed hos -pa hogen ansc ip ional analysis. Sci Rep 9: 19244. doi:10.1038/s41598-019-55633-6 By ne A, Supple MA, Volden R, Laid e KL, Shapi o B, Vollme s C. 2019. Deple ion o hemoglobin ansc ip s and long- ead se- quencing imp o es he ansc ip ome anno a ion o he pola bea (U sus ma i imus). F on Gene 10: 643. doi:10.3389/ gene .2019.00643 Cas o TL, Sey e N, Ramos RT, Ba bosa S, Ca alho RD, Pin o AC, Ca nei o AR, Sil a WM, Pacheco LG, Downson C, e al. 2013. Ion o en -based ansc ip ional assessmen o a Co ynebac e ium pseudo ube culosis equi s ain e eals dena u - ing high-pe o mance liquid ch oma og aphy a p omising RNA deple ion me hod. Mic ob Bio echnol 6: 168–177. doi:10.1111/ 1751-7915.12020 Chugani S, Kim BS, Pha a asukol S, B i nache MJ, Choi SH, Ha wood CS, G eenbe g EP. 2012. S ain-dependen di e si y in RNA deple ion in bac e ia ia DASH www. najou nal.o g 1077