Superinfection and cure of infected cells as mechanisms for hepatitis C virus adaptation and persistence

Supe in ec ion and cu e o in ec ed cells as mechanisms o hepa i is C i us adap a ion and pe sis ence Ruian Ke a,b,1 , Hui Li c,d,1 , Shuyi Wang c,d , Wenge Ding c,d , Ruy M. Ribei o b,e , Elena E. Gio gi b , Tanmoy Bha acha ya ,g , Richa d J. O. Ba na d h , Bea ice H. Hahn c,d,2 , Geo ge M. Shaw c,d , and Alan S. Pe elson b,g,2 a Depa men o Ma hema ics, No h Ca olina S a e Uni e si y, Raleigh, NC 27695; b Theo e ical Biology and Biophysics G oup, Los Alamos Na ional Labo a o y, Los Alamos, NM 87545; c Depa men o Medicine, Uni e si y o Pennsyl ania, Philadelphia, PA 19104; d Depa men o Mic obiology, Uni e si y o Pennsyl ania, Philadelphia, PA 19104; e Labo a o y o Bioma hema ics, Facul y o Medicine, Uni e si y o Lisbon, 1600-276 Lisbon, Po ugal; Theo e ical Di ision, Los Alamos Na ional Labo a o y, Los Alamos, NM 87545; g San a Fe Ins i u e, San a Fe, NM 87501; and h Depa men o In ec ious Disease, Me ck & Co., Inc., Kenilwo h, NJ 07033 Con ibu ed by Bea ice H. Hahn, June 8, 2018 (sen o e iew Ma ch 28, 2018; e iewed by Alison P. Gal ani and Ma in A. Nowak) RNA i uses exis as a gene ically di e se quasispecies wi h ex ao - dina y abili y o adap o ab up changes in he hos en i onmen . Howe e , he molecula mechanisms ha con ibu e o hei apid adap a ion and pe sis ence in i o a e no well s udied. He e, we p obe hepa i is C i us (HCV) pe sis ence by analyzing clinical samples aken om subjec s who we e ea ed wi h a second- gene a ion HCV p o ease inhibi o . F equen longi udinal i al load de e mina ions and la ge-scale single-genome sequence analyses e ealed apid an i i al esis ance de elopmen , and su p isingly, dynamic u no e o dominan d ug- esis an mu an popula ions long a e ea men cessa ion. We i ed ma hema ical models o bo h he i al load and he i al sequencing da a, and he esul s p o ided s ong suppo o he c i ical oles ha supe in ec ion and cu e o in ec ed cells play in acili a ing he apid u no e and pe sis ence o i al popula ions. Mo e b oadly, ou esul s highligh he impo ance o conside ing i al dynamics and compe i ion a he in acellula le el in unde s anding apid i al adap a ion. Thus, we p opose a heo e ical amewo k in eg a ing i al and molecula mechanisms o explain apid i al e olu ion, esis ance, and pe sis- ence despi e an i i al ea men and hos immune esponses. i us e olu ion | i us pe sis ence | hepa i is C i us | phylodynamic modeling | ma hema ical modeling Many RNA i uses, including global pa hogens o majo medical impo ance such as hepa i is C i us (HCV), HIV, and ebola i us, exis wi hin in ec ed hos s as la ge popula ions o gene - ically ela ed i al a ian s commonly e e ed o as a quasispecies (1). The di e se na u e o he i al quasispecies allows i al pop- ula ions o e ol e apidly o adap o ab up changes in he hos en- i onmen (2); examples include escape om adap i e immune p essu es (3, 4) and de elopmen o esis ance o an i i als (5, 6). Despi e he impo ance o i al adap a ion, ou unde s anding o how he i al quasispecies esponds o selec ion p essu e and he unde - lying molecula mechanisms suppo ing i al adap a ion in i o is limi ed by a lack o equen longi udinal i al sequence da a gen- e a ed by me hods ha e ain linkage ac oss genes and genomes (7). P e iously, much a en ion has been de o ed o es ima ing he e ec i eness o an i i als agains i al a ian s in he quasispe- cies (8–10); howe e , he mechanisms ha d i e an adap i e/ i e a ian (e.g., a d ug- esis an mu an ) a an ini ially low equency o a high equency a e less well s udied. Theo e ical wo k sugges s ha i e a ian s a a low equency canno ex- pand unless he e a e su icien numbe s o a ge cells, ha is, “ eplica ion space”(5, 11–14). This is analogous o he concep o compe i i e elease in ecological “niche” heo y, ha is, a species is no able o expand unless he e is a niche aca ed by o he species (15). P e ious models ha e in gene al assumed ha , once a a ian in ec s a cell, he cell is occupied by he a ian o i s emaining li e ime, and hus he eplica ion space is p o ided by gene a ion o new a ge cells (5, 8, 13, 16–18). Unde his amewo k, i al a ian s compe e o in ec ing newly gene a ed a ge cells, and hus, he expansion o a low- equency i e mu an is de e mined by how quickly in ec ed cells die and a e eplaced by new a ge cells. This amewo k has been suc- cess ully applied o unde s and HIV d ug esis ance (6, 13, 16, 18). Howe e , he sou ce o eplica ion space o o he apidly e ol ing i uses, and in pa icula HCV, is unclea . HCV is a posi i e-s and RNA i us belonging o he la i- i idae, a amily o medically impo an single-s and RNA i uses. HCV in ec s ∼80–170 million people wo ldwide (19). Ch onic in ec ion can lead o ci hosis and hepa ocellula ca ci- noma. Cu en ea men s o HCV in ec ion in ol e combina- ions o di ec ac ing an i i als (DAAs), and hey ha e achie ed Signi icance Vi al popula ions exhibi an ex ao dina y abili y o su i e ab up changes in hos en i onmen by apidly gene a ing adap i e mu a ions. Howe e , ou unde s anding o how i al popula ions espond o selec ion p essu e and he unde lying molecula mechanisms suppo ing i al adap a ion in i o is limi ed. He e, we epo a se o clinical da a sampled om subjec s ch onically in ec ed by hepa i is C i us (HCV). The da a show apid expansion and u no e o d ug- esis an i uses ollowing ea men wi h an HCV p o ease inhibi o . By i ing ma hema ical models o he da a, we p opose ha supe in ec- ion and cu e o in ec ed cells play c i ical oles in acili a ing he apid expansion and u no e o i al popula ions. Ou esul s highligh he impo ance o conside ing in acellula i al com- pe i ion in unde s anding apid i al adap a ion. Au ho con ibu ions: R.K., H.L., R.M.R., G.M.S., and A.S.P. designed esea ch; R.K., H.L., S.W., W.D., R.M.R., E.E.G., T.B., R.J.O.B., B.H.H., G.M.S., and A.S.P. pe o med esea ch; R.K., H.L., S.W., W.D., R.M.R., E.E.G., T.B., R.J.O.B., B.H.H., G.M.S., and A.S.P. analyzed da a; and R.K., H.L., R.M.R., G.M.S., and A.S.P. w o e he pape . Re iewe s: A.P.G., Yale Cen e o In ec ious Disease Modeling and Analysis; and M.A.N., Ha a d Uni e si y. Con lic o in e es s a emen : R.J.O.B. is an employee and sha eholde o Me ck & Co., Inc. This wo k was unded by a g an om Me ck Sha p & Dohme Co p., a subsidia y o Me ck & Co., Inc. ( o G.M.S.). A.S.P. consul ed o Me ck Sha p & Dohme Co p., a sub- sidia y o Me ck & Co., Inc. This open access a icle is dis ibu ed unde C ea i e Commons A ibu ion-NonComme cial- NoDe i a i es License 4.0 (CC BY-NC-ND). Da a deposi ion: The sequences epo ed in his pape ha e been deposi ed in he GenBank da abase (accession nos. MF282014–MF284666). 1 R.K. and H.L. con ibu ed equally o his wo k. 2 To whom co espondence may be add essed. Email: [email protected] o [email p o ec ed]. This a icle con ains suppo ing in o ma ion online a www.pnas.o g/lookup/suppl/doi:10. 1073/pnas.1805267115/-/DCSupplemen al. Published online July 9, 2018. www.pnas.o g/cgi/doi/10.1073/pnas.1805267115 PNAS | ol. 115 | no. 30 | E7139–E7148 MEDICAL SCIENCES Downloaded by gues on No embe 30, 2020 ema kably high cu e a es (20–22). I onically, because o his clinical e ec i eness, combina ion ea men s canno gene ally be used o p obe he molecula mechanisms ha con ibu e o i us pe sis ence. Howe e , his in o ma ion may be c i ical o designing e ec i e accina ion s a egies o HCV and he apeu ic egimens and accines o o he membe s o he la i i idae amily. He e, we analyzed clinical samples collec ed om a phase I/IIa mono he apy ial (23) o a second-gene a ion HCV p o ease inhibi o , MK-5172 (also called g azop e i ) (22, 24–26), ha is ex emely po en bu suscep ible o d ug esis ance de elopmen h ough mul iple i al gene ic pa hways o escape. Measu ing i al loads and analyzing i al popula ions longi udinally by single-genome sequencing (7) allowed us o ollow he popula ion dynamics and e olu iona y esponses o HCV o d ug p essu e and i s emo al in i o a high esolu ion. We analyzed his unique da ase by combining ma he- ma ical modeling and phylogene ic analysis [ e med “phylody- namic”analysis (27)]. We p o ide s ong e idence ha cu e and supe in ec ion o in ec ed cells con ibu e essen ially o he epli- ca ion space, allowing esis ance mu an expansion and e olu ion. Resul s S udy Subjec s and Single-Genome Sequence Analysis. We s udied eigh pa icipan s om a dose- anging mono he apy ial o MK-5172 (23). Fi e we e ea ed wi h MK-5172 (50 o 800 mg) once daily o 7 d (subjec s 1–5, geno ypes 1a, n=4; 1b, n=1) and h ee we e ea ed wi h placebo (subjec s 6–8, geno ypes 1a, n=2; 1b, n=1) (SI Appendix,TableS1). All o he pa icipan s we e ollowed o 7–8wk a e ea men cessa ion. We measu ed i al loads longi udinally and analyzed i al popula ions using single-genome sequencing (7). Samples aken immedia ely be o e ea men ini ia ion, desig- na ed as day 0, we e analyzed in all eigh subjec s o geno ypic complexi y and baseline d ug esis ance mu a ions (DRMs). A each sampling imepoin , ∼100 i al genomes we e sequenced in each subjec (Fig. 1 and SI Appendix,Figs.S1–S7). Day 0 se- quences o each subjec showed pa e ns o i us di e si y ypical o ch onic in ec ion wi h b oad geno ypic he e ogenei y (SI Ap- pendix, Table S1). Maximum pai wise di e si y and mean pai wise di e si y anged om 1.17 o 5.64% and 0.53 o 3.45%, e- spec i ely. No d ug esis ance-associa ed mu a ions we e iden i- ied in he p e ea men sequences, al hough a Q80K subs i u ion was ound in all sequences in subjec s 2 and 3. Q80K is a common polymo phism ha con e s no esis ance o MK-5172. In he i e subjec s ea ed wi h MK-5172, i al load dec eased o e 5 logs and wen below (o close o) he limi o quan i ica ion (=25 IU/mL) du ing he 7-d ea men pe iod (Fig. 1Aand SI Appendix,Figs.S1–S4). A he end o his pe iod, i al load was oo low o pe mi sequencing analyses. By day 27, 20 d a e ea men was s opped, he i al load ebounded o le els (10 6 o 10 7 IU/mL) simila o he baseline le el be o e ea men in all ea ed subjec s and could be quan i ied as ea ly as day 14 (7 d a e s opping ea men ) in subjec s 1 and 4. Follow-up single-genome se- quencing analyses we e conduc ed on samples om days 14 o 62. A e he apy was s opped, he i al popula ions ha eme ged ca ied d ug esis ance mu a ions. Su p isingly, he dynamics o mu an i us popula ions exhibi ed s ingen popula ion bo le- necks and apid u no e o he dominan esis an mu an s du ing he ollow-up pe iod, e en in he p esence o ela i ely s able i al load (e.g., subjec 1 be ween days 27 and 56; Fig. 1). Ex emely apid and con inuous u no e o he esis an mu an s, a e he apy cessa ion, was obse ed in subjec s 1 and 4 (Fig. 1 and SI Appendix, Fig. S3). In con as o hese d ama ic changes in plasma i al RNA ( RNA) load in subjec s ea ed wi h MK-5172, he h ee subjec s who ecei ed placebo had no changes in plasma RNA le els and no sequences con aining DRMs. Vi al sequences om hese indi iduals we e ob ained on days 0 and 27 o 34 (SI Appendix, Figs. S5–S7). Below, we desc ibe he e olu iona y pa - e ns in subjec 1 in de ail. The pa e ns in o he ea ed subjec s a e desc ibed in SI Appendix,SI Me hods and Resul s. In subjec 1, none o 103 single genome sequences on day 0 con ained d ug esis ance-associa ed mu a ions (Fig. 1C). Con- e sely, on day 14 (i.e., 7 d a e he apy was s opped), all 111 single-genome sequences analyzed con ained one o mo e d ug esis ance-associa ed mu a ions. Fi y-six o he 111 sequences con ained single A156T o V mu a ions lis ed in colo on he igh - hand side o Fig. 1Cand wo sequences con ained a single R155G mu a ion. These 58 sequences con aining single d ug esis ance mu a ions we e widely dispe sed h oughou he phylogene ic ee o p e ea men sequences (Fig. 1B). This indica es ha nea ly all o hese d ug- esis an sequences ep esen ed dis inc i al lineages ha aced back 14 d ea lie o in ec ed hepa ocy es ha ca ied A156T o A156V mu a ions in widely di e gen gene ic back- g ounds. In ma ked con as , 51 o 111 sequences (46%) con ained a R155W-plus-A156G double mu a ion, in a nea ly homogeneous gene ic backg ound whe e sequence di e si y anged om 0 o 1 nucleo ide subs i u ion o e a span o 2,210 bp (<0.1% di e si y). This esul sugges s ha ∼46% o hepa ocy es, eleasing HCV i- ions in o he ci cula ion a 14 d a e he ini ia ion o MK- 5172 ea men (and 7 d a e i s discon inua ion), we e in ec ed by HCV genomes ha had e ol ed om a single R155W-plus- A156G d ug- esis an i al sequence lineage. In u n, his lineage mos likely emana ed om a single o a ew p oduc i ely in ec ed hepa ocy es ha exis ed be o e ea men ini ia ion (Fig. 1C). This in e ence is d awn based o he iden i y o nea iden i y (1 mu a ion in 2,210 n ) o he gene ic backg ound in which he R155W-plus- A156G d ug esis ance mu a ion expanded. These indings a e consis en wi h model p edic ions o he expec ed equency o a single mu a ion o any se o wo mu a ions in an HCV RNA quasispecies in he absence o selec ion (5, 28). In addi ion, one sequence wi h Y56H-plus-D168V and one wi h Y56H-plus-D168N double mu a ions we e also de ec ed. A day 27 when he plasma i al load had e u ned o baseline (>10 6 RNA/mL), he e was a s iking con ac ion in bo h he R155W-plus-A156G and A156T o V popula ions and a ema kable expansion o he disc e e homo- geneous Y56H-plus-D168V and Y56H-plus-D168N double-mu an popula ions (al eady de ec ed a day 14). A single dominan Y56H- plus-D168N double-mu an popula ion and h ee dominan Y56H- plus-D168V double-mu an popula ions oge he ep esen ed 70% o he sequences. This inding is all o he mo e ema kable gi en ha , by his ime, s eady-s a e plasma RNA load had e u ned o i s p e ea men se poin o >10 6 RNA copies pe mL and d ug he apy had been s opped 20 d p e iously. Two addi ional mono- phyle ic lineages, each con aining he double-mu an D168E-plus- F169I, we e p esen on day 27. By day 56, he D168E-plus-F169I mu an lineages ( i s de ec ed a day 27) expanded o comp ise 79% o he sequences, whe eas he p e iously dominan Y56H-plus- D168N and Y56H-plus-D168V lineages con ac ed and now comp ised only 12% o he sequences. Fou wild- ype sequences, ha is, sequences wi h no known esis ance mu a ions, we e de ec ed o he i s ime pos ea men a day 56. Fi ing a Baseline Vi al Dynamic Model o he Vi al Load and Sequence Da a. The apid i al load decline du ing ea men and con inuous u no e o esis an mu an s a e ea men cessa ion obse ed in he ea ed subjec s aise in iguing ques- ions, such as how do esis ance clones expand so apidly and how can esis an mu an s domina e he i al popula ion o such a long pe iod a e ea men cessa ion wi hou being eplaced by he non esis an i us? To add ess hese ques ions, we i s con- s uc ed a “baseline”mul is ain HCV model, simila o a p e ious s anda d i al dynamic model ha inco po a es compe i ion o a ge cells (5) (Me hods). The s ains in he model a e g ouped acco ding o he sha ed d ug esis ance mu a ions in he clinical samples as desc ibed in he sec ion abo e and as colo -coded in Fig. 1 and SI Appendix, Figs. S1–S7 (see SI Appendix, Table S2, o he mu an s modeled). In his model, we assume ha in ec ed cells a e los a a pe -capi a a e δand ha a ge cells p oli e a e E7140 | www.pnas.o g/cgi/doi/10.1073/pnas.1805267115 Ke e al. Downloaded by gues on No embe 30, 2020 ollowing a logis ic g ow h e m o eplace hem. Cu e and su- pe in ec ion o in ec ed cells a e no included in his model. We i ed his model o he da a om all i e ea ed subjec s. This model desc ibes bo h he i al load and he sequence da a well (SI Appendix,Fig.S8). Howe e , he mean loss a e o in- ec ed cells, δ, es ima ed ac oss all i e ea ed subjec s was 0.68 d −1 wi h a SD o 0.09 d −1 (see SI Appendix,TablesS3–S7, o he bes - i alues o δ). In ou model, he loss o in ec ed cells A C B Fig. 1. Sequen ial plasma i us load and sequences om subjec 1. (A) Time cou se o ea men wi h MK-5172 (shaded a ea, days 1–7), i al load de e - mina ions (blue solid do s), and i al sequence analyses (open ci cles a days 0, 14, 27, and 56). (B) A maximum-likelihood (ML) phylogene ic ee o all i al sequences sampled om subjec 1 om all ime poin s. T ee ips a e colo coded acco ding o he known esis an mu a ions hey bea . (C) ML phylogene ic ees o i al sequences sampled om subjec 1 a each ime poin . Ke e al. PNAS | ol. 115 | no. 30 | E7141 MEDICAL SCIENCES Downloaded by gues on No embe 30, 2020 co esponds o dea h o hese cells. A la ge dea h a e is needed o explain he apid second-phase i al load decline du ing ea men and apid u no e o dominan mu an s a e ea men . How- e e , such a la ge dea h a e o in ec ed cells (0.68 d −1 )doesno seem physiological, as i implies a hal -li e o in ec ed hepa ocy es o 1 d. In s a k con as , he dea h a e o in ec ed cells has been es ima ed as 0.14 d −1 om IFN he apy (29, 30) and p e iously used by Guedj e al. (31) and Lau e al. (21) o explain i al de- clines seen wi h o he DAAs. To con i m he necessi y o a high loss a e in his model, we e i ed he model ixing δa 0.14 d −1 and ound his al e na i e pa ame e iza ion explains he da a poo ly (SI Appendix,Fig.S9). These esul s s ongly sugges ha o he mechanisms (in addi ion o dea h o in ec ed cells) mus play essen ial oles in d i ing he pa e ns seen in he da a. Po en ial Roles o In ec ed Cell Cu e and In acellula Vi al Compe i ion Th ough Supe in ec ion. We hypo hesize ha cu e o in ec ed cells is he p ima y cause o he loss o in ec ed cells du ing he apid second-phase decline obse ed in ou da a and o he s udies (5, 8, 32). In addi ion, we hypo hesize ha he con inuous u no e o dominan mu an s in he absence o ea men is mos ly d i en by in acellula compe i ion among i al s ains, as a esul o supe in ec ion ( a he han compe i ion o newly gene a ed a ge cells). P e iously, cu e o in ec ed cells has been demons a ed in in i o s udies (33–37). In i o, he apid second-phase i al declines unde ea men we e sugges ed o be a ibu able o cu e o in ec ed cells (32, 38). Fo supe in- ec ion, in i o expe imen s showed ha mul iple i uses can en e he same cell, al hough limi ing hos esou ces necessa y o i al eplica ion and/o ansla ion may es ic ac i e p oduc ion o mul iple i uses (39, 40). In e es ingly, a ecen s udy demon- s a ed ha a i e HCV s ain can en e al eady-in ec ed cells and ou compe e he esiden s ain (41), sugges ing ha mul iple HCVs can en e a cell and compe e o in acellula esou ces. To es hese hypo heses, we ex ended he baseline model by inco po a ing in ec ed cell cu e and supe in ec ion, and ixing he a e o in ec ed cell dea h a 0.14 d −1 as es ima ed unde IFN he apy (29, 30) (Me hods). He e, we assumed ha in ec ed cells a e cu ed and become a ge cells again unde MK-5172 ea men a pe -capi a a e kcu e ·ð−log10ð1−«iÞÞ, whe e kcu e is a a e cons an and «iis he d ug e icacy agains he i h s ain (see Me hods o de ail). To model he in acellula i al com- pe i ion due o supe in ec ion, we assume, o simplici y, ha once an in ec ed cell is supe in ec ed, he i e s ain can ou - compe e he less i s ain in acellula ly, and he cell becomes a cell in ec ed by he i e s ain a a e k supe . We used h ee a ia ions o he ex ended model o es he im- po ance o cu e and supe in ec ion in explaining he kine ic pa e ns in he da a: (i) we allowed cu e bu no supe in ec ion (deno ed as he “cu e”model), o (ii) supe in ec ion bu no cu e (deno ed as he “supe in ec ion”model), o (iii) bo h p ocesses (deno ed as he “ ull”model) (see SI Appendix, Figs. S10 and S11 and Fig. 2, espec i ely, o model i s and SI Appendix,TablesS3–S7 o he bes - i pa ame e alues). To s a is ically es he impo - ance o cu e o /and supe in ec ion in explaining he clinical da a, we pe o med model selec ion using he co ec ed Akaike in o ma ion c i e ion (AICc) (Table 1). The esul s sugges ha cu e o in ec ed cells is needed o explain he apid second- phase i al load decline du ing ea men seen in all i e sub- jec s (SI Appendix,Fig.S12). I also helps o explain he apid selec ion o he esis an mu an s seen a day 14 in subjec s 1 and 4 (compa e he i s be ween solid lines and dashed lines o da a in SI Appendix,Fig.S13). Based on he bes - i pa ame e s in he bes model o each subjec (SI Appendix, Table S8), we es ima ed he mean a e o cu e o cells in ec ed by baseline i uses unde MK-5172 ea men [kcu e ·ð−log10ð1−«1ÞÞ], whe e «1is he d ug e icacy agains he baseline i uses) o be 0.56 d −1 . On a e age, i akes abou 1.9 d o cu e a cell unde MK-5172 ea men (SI Appendix, Table S8) and he loss o in ec ed cells is p ima ily h ough cu e a he han dea h o in ec ed cells (as- sumed o occu a a e o 0.14 d −1 ). The AICc esul s sugges ha in acellula compe i ion h ough supe in ec ion is needed o explain he apid and con- inuous u no e o dominan esis an s ains a e ea men s ops in subjec s 1 and 4 (SI Appendix, Fig. S14). This is because a e ea men cessa ion, d ug is elimina ed and d ug-induced cu e is no longe possible. As a esul , new a ge cells become a ailable a a much slowe a e han he a e du ing ea men . Howe e , supe in ec ion allows i e s ains o en e al eady- in ec ed cells and compe e in acellula ly, g ea ly inc easing he a e a which a i e s ain ises o a high equency. In con as , wi hou supe in ec ion and in acellula compe i ion, a i e s ain is p edic ed o inc ease in equency a a much lowe a e (see he changes in he equencies o he mu an in g een du ing he pe iod o days 30–60 in SI Appendix, Fig. S14A). We pe o med unce ain y analysis in he bes model o each subjec using likelihood p o iling (42) (SI Appendix, Tables S3– S7). These analyses indica ed ha he da a allow us o es ima e accu a ely he e ec i eness o MK-5172 agains he baseline i- us, « 1 , he a e o i al clea ance, c, and he cu e a e k cu e in all i e subjec s, and he alue o supe in ec ion cons an k supe in subjec s 1 and 4. In addi ion, we explo ed whe he models wi h al e na i e assump ions can explain he da a. This included a model assuming a cons an a e o a ge cell gene a ion and models assuming a DAA-independen cu e o in ec ed cells (SI Appendix,SI Me hods and Resul s). Fi ing esul s show ha he ull model p esen ed in he main ex is he bes model, and he conclusions abou he ole o supe in ec ion and cu e o in ec ed cells a e obus agains hese model a ia ions. Compensa o y Mu a ions, Clonal In e e ence o HCV-Resis an Mu an s. To unde s and be e he e olu iona y dynamics o he i al s ains conside ed in ou model, we analyzed how he s ains a di e en ime poin s a e ela ed o each o he . He e, we de ine s ains pheno ypically, ha is, each “s ain”consis s o a g oup o ela ed i al sequences in he clinical samples sha ing d ug e- sis ance mu a ions. We calcula ed he gene ic dis ances be ween sequences belonging o di e en g oups a he di e en sampling ime poin s (see Me hods and SI Appendix, Figs. S15–S17). We hen combined he mu a ional pa e n de i ed om he sequence da a wi h he esul s om i ing he dynamical models o gene a e a diag am ha summa izes he e olu iona y dynamics o he d ug- esis an s ains o e he pe iod o he s udy (Fig. 3). The diag am e eals how he HCV popula ion esponds o MK-5172 ea men . Fi s , i al load ebounded apidly in sub- jec s 1 and 4 o high le els ha pe mi ed sequencing by day 14, p esumably because o selec ion o he esis an mu an s bea ing A156T/V o R155G/W in subjec 1 and A156T/V in subjec 4. We es ima ed ha hese mu an s a e highly esis an , ye ha e high i ness cos s (Fig. 3), which is consis en wi h esul s shown in p e ious in i o s udies (24–26). Because o hei sho gene ic dis ance o he wild- ype i uses (SI Appendix, Figs. S15 and S17), hey a e likely p esen a low equencies be o e ea men . Du - ing and sho ly a e ea men , hey ise o high equencies ansien ly be o e being eplaced by o he esis an mu an s wi h highe i ness while d ug is clea ed. Second, mu an s wi h he Y56H mu a ion a e obse ed ansien ly in subjec s 1, 3, and 4 be ween days 27 and 34. Ou analysis sugges s ha he ansien appea ance o hese mu an s is due o hei in e media e e- sis ance and i ness alues. These mu an s ga e way o i e s ains wi h esis an mu a ions a posi ion 168 (e.g., D168E, D168A, o D168Y; see e s. 24 and 25) ha domina ed he i al popula ion in subjec s 1–4 a he end o ollow-up pe iod, ha is, 7–8wka e ea men cessa ion. These dynamics esemble he classical pa - e n, e med “clonal in e e ence,”as sugges ed by popula ion gene ic models o he e olu ion o asexual o ganisms (43, 44). E7142 | www.pnas.o g/cgi/doi/10.1073/pnas.1805267115 Ke e al. Downloaded by gues on No embe 30, 2020 The i nesses o he esis an mu an s obse ed a la e ime poin s a e p edic ed o be close o o highe han hose o he baseline i uses, which emained a low equencies in subjec s 1–4 (Fig. 3). A likely explana ion o he la e appea ance o hese high- i ness a ian s is ha mu a ion a posi ion 168 on he baseline i us backg ound esul ed in low i ness (and hus hese mu an s a e a low equency be o e ea men ). Howe e , g ow h o hese low- i ness esis an mu an s gene a ed compensa o y mu a ions ha inc eased hei i ness, and allowed hei equency o in- c ease and s abilize a a high le el in he popula ion. Las , we obse e ha d ug-sensi i e i uses, ha is, i uses wi h no known d ug- esis an mu a ions, eappea ed in he se- quences a he las ime poin o ollow-up in subjec s 1, 4, and 5. S a is ical analysis shows ha hese i uses a e likely o be de- scendan s o d ug-sensi i e i uses p esen be o e ea men a he han e e sions om esis an a ian s (SI Appendix,SI Me hods and Resul s). This sugges s ha he d ug-sensi i e i u- ses we e supp essed o low equencies unde ea men and only g ow o high equencies long a e ea men cessa ion. Discussion He e, we ha e analyzed he e olu iona y dynamics o he HCV popula ion in esponse o 7-d ea men wi h he p o ease inhibi o g azop e i (MK-5172) in i e subjec s using single-genome se- quencing and phylodynamic modeling. HCV- esis an a ian s un- de wen apid expansion a e d ug ea men wi h su p isingly as sequen ial u no e o dominan esis an mu an s in he 7- o 8-wk ollow-up pe iod, a dis inc ea u e o i al pe sis ence. Using ma hema ical models, we showed ha supe in ec ion and cu e o in ec ed cells a e likely o be c ucial mechanisms d i ing he ex- emely apid expansion and u no e o esis an a ian s obse ed in he clinical da a. Al hough supe in ec ion and cu e o in ec ed cells ha e been demons a ed in i o (33–37, 41), ou wo k sug- ges s ha hesep ocessesoccu andplay impo an oles in d i ing i al adap a ion and pe sis ence in i o. We es ima ed he a e o cu e o cells in ec ed by wild- ype i uses unde g azop e i o be 0.56 d −1 on a e age. Fu he , i akes abou 1.9 d on a e age o cu e a cell unde MK-5172 ea men . This high a e o in ec ed cell cu e unde ea men is consis en wi h he clinical da a in his s udy and om p e ious s udies (8, 30, 32, 45), whe e apid second-phase i al load de- clines ha e been obse ed. P e ious ma hema ical models sug- ges ed ha he second-phase decline e lec s bo h he dea h a e o in ec ed cells and he a e o in acellula RNA deg ada ion (31, 32, 38). Thus, his apid cu e o in ec ed cells is likely o be a esul o he apid loss o in acellula HCV RNAs unde ea - men . Ou model shows ha , due o he apid cu e, d ug-sensi i e i uses we e a a much lowe equency han esis an mu an s AB CD E 0 102030405060 1234567 Time (days) Log10 copies/mL 0 102030405060 0.0 0.2 0.4 0.6 0.8 1.0 Time (days) Mu an equencies 0 102030405060 1234567 Time (days) Log10 copies/mL 0 102030405060 0.0 0.2 0.4 0.6 0.8 1.0 Time (days) Mu an equencies 0 102030405060 1234567 Time (days) Log10 copies/mL 0 102030405060 0.0 0.2 0.4 0.6 0.8 1.0 Time (days) Mu an equencies 0 102030405060 1234567 Time (days) Log10 copies/mL 0102030405060 0.0 0.2 0.4 0.6 0.8 1.0 Time (days) Mu an equencies 0 102030405060 1234567 Time (days) Log10 copies/mL 0 102030405060 0.0 0.2 0.4 0.6 0.8 1.0 Time (days) Mu an equencies Fig. 2. Bes i o he “ ull”model (lines) o he clinical da a (ci cles) om i e subjec s ea ed wi h MK-5172. (A–E) Schema ic diag ams o he e olu iona y dynamics in subjec s 1–5, espec i ely. A–E, he da a and simula ion esul s o i al loads a e shown in open ci cles and black lines, espec i ely, on he Le ; he da a and simula ion esul s o mu an equencies a e shown in colo ed open ci cles and lines, espec i ely, on he Righ . The colo coding o each mu an conside ed is shown in SI Appendix, Table S2. Ke e al. PNAS | ol. 115 | no. 30 | E7143 MEDICAL SCIENCES Downloaded by gues on No embe 30, 2020 when ea men s opped. Tha coupled wi h he apid expansion o esis an mu an s and gene a ion o compensa o y mu a ions led o he disappea ance o a delay in he eappea ance o d ug-sensi i e i us. Supe in ec ion has been demons a ed in i o indi ec ly by he iden i ica ion o ecombinan o ms o HCV in pa ien s (46–49). Al hough ecombina ion be ween HCV genomes is appa en ly a e (50, 51), his does no mean ha supe in ec ion o hepa o- cy es is in equen , because he o ma ion o dis inc in acellula HCV eplica ion complexes and a p ocessi e RNA polyme ase may p e en ecombina ion. In line wi h a ecen in i o s udy showing ha a i e HCV s ain can en e al eady-in ec ed cells and ou compe e he esiden s ain (41), ou esul s sugges ha supe in ec ion and in acellula compe i ion occu equen ly in i o, especially when di e en s ains exhibi a di e ence in in acellula eplica ion i ness. We a gue ha supe in ec ion and in acellula compe i ion a e c i ical o he es ablishmen and expansion o adap i e mu an s and hus o he pe sis ence o HCV unde selec ion p essu es. Fo i uses ha la gely ansmi by cell- o-cell in ec ion, such as HCV (52–55), ansmission is o en limi ed om an in ec ed cell o a ew neighbo ing cells. When mos o he cells a e in ec ed by esiden i al a ian s, wi hou supe in ec ion, he p obabili y ha an adap i e mu an inds an unin ec ed cell is low. Thus, adap i e mu an s would likely be los , and o hose mu an s ha do es- ablish in ec ion, he a e a which hey expand would be low. In con as , wi h supe in ec ion and in acellula compe i ion, an adap i e mu an can en e all neighbo ing cells o compe e wi h any esiden i us, which subs an ially inc eases he p obabili y o es ablishmen and he a e a which he adap i e mu an expands. While p e ious heo e ical wo k discussed he ole o supe in- ec ion in i al e olu ion a he hos le el (56, 57), ou wo k highligh s he impo an ole ha supe in ec ion plays in acili- a ing i al adap a ion a he cellula le el. By combining phylogene ic analysis and ma hema ical modeling, we p o ided an in eg a ed unde s anding o he HCV e olu iona y and popula ion dynamics. Fi s , he HCV wi hin-hos e olu ion o HCV esembles he classical pa e n o he e olu ion o asexual o ganisms as sugges ed by popula ion gene ic models, e med clonal in e e ence (43, 44). This emphasizes he use ulness o popula ion gene ic models in unde s anding i al e olu iona y dynamics. Sec- ond, ou esul s sugges ha compensa o y mu a ions a e likely o be gene a ed a e ea men , allowing o s abiliza ion and pe sis- ence o esis an mu an s in he i al popula ion long a e ea - men cessa ion. This highligh s he need o main ain a high le el o adhe ence o combina ion he apies o DAAs. T ea men in e - up ion would allow apid expansion o i al mu an s esis an o one o mo e DAAs and possible gene a ion o compensa o y mu- a ions ha s abilize he esis an mu a ions a a high equency, and hus c ea ing oppo uni y o ansmission o esis an mu an s. In si ua ions whe e ea men is in e up ed equen ly, mul iple mu- a ions may accumula e such ha he e ol ed i us may become esis an o all DAAs in he combina ion he apy (12). Al hough beyond he scope o his wo k, a couple o p edic ions o ou s udy can be u he es ed expe imen ally. Fi s , ou models sugges ha supe in ec ion and in acellula compe i ion a e impo - an mechanisms acili a ing he apid expansion and u no e o esis an mu an s in i o a e ea men cessa ion. Al hough es ing his p edic ion in i o is no possible, in acellula i al compe i ion expe imen s as in e . 41 can be pe o med o compe e pai s o labo a o y i uses wi h he esis an mu a ions iden i ied in his s udy. The esul s would econ i m he occu ence o supe in ec ion and de e mine he ela i e i ness o hose mu an s a he in acellula scale. Second, ou analysis sugges s ha compensa o y mu a ions a e in ol ed in s abilizing he esis an mu a ions a amino acid 168 seen in subjec s 1–4. This can be es ed by i s pe o ming whole-genome sequencing on he i al isola es and hen iden i ying he sha ed mu- a ions o he i uses wi h a mu a ion a posi ion 168 in each subjec . These sha ed mu a ions would be candida e compensa o y mu a ions o he mu a ion a amino acid 168. In oducing one o a combina ion o hese mu a ions in o i al eplicons, and hen es ing and com- pa ing he eplica ion i ness o he eplicons (as shown in e s. 24– 26), would be a means o iden i y compensa o y mu a ions. To summa ize, he esul s in ou s udy elucida e impo an , ye p e iously un ecognized, mechanisms ha we sugges ope a e in i o, d i e wi hin-hos i al e olu ion, and allow he adap a ion and pe sis ence o HCV in he ace o d ug p essu e and immune esponse. A p e ious amewo k equen ly used in i al dynamic models (5, 8, 13, 16–18) assumed ha i uses compe e o he eplica ion space needed o adap a ion o changing hos en i- onmen s only h ough in ec ion o newly gene a ed a ge cells (Fig. 4A). Consequen ly, he a e o adap a ion was limi ed by he a e o a ge cell gene a ion. In con as , ou amewo k emphasizes ha bo h supe in ec ion and cu e o in ec ed cells (in addi ion o in ec ion o new a ge cells) can be impo an mechanisms con ibu ing o he eplica ion space needed o mu an i us expansion (Fig. 4B). This mechanism can g ea ly accele a e he a e o wi hin-hos i al adap a ion beyond he a e se by he gene a ion o a ge cells. This amewo k may be impo an in elucida ing he con ibu ions and mechanisms o HCV con ainmen and/o elimina ion media ed by he hos im- mune sys em ha occu na u ally o ollowing accina ion (58). Mo e b oadly, ou wo k poin s owa d he impo an ole mo- lecula mechanisms play in acili a ing i al e olu ion and pe - sis ence. F equen cell- o-cell in ec ion and high mul iplici y o in ec ion, ha is, mechanisms ha p o ide he means o in- acellula compe i ion, ha e been epo ed ecen ly o HIV, HCV, in luenza, and o he i uses (54, 59–61). We a gue ha ou amewo k and he conside a ion o i al compe i ion a bo h he in acellula and hos le el in pa icula , is c i ical o he unde s anding o he e olu iona y and adap i e dynamics o i al popula ion a he wi hin-hos le el (62, 63). Thus, i opens he doo o he nex gene a ion o i al dynamic models and quan- i a i e amewo ks o unde s and i al e olu ion, in e p e clin- ical da ase s, and p edic ea men ou comes. Table 1. Summa y o he model cha ac e is ics and he i ing esul s (i.e., AICc sco es) o each model o each subjec Model Model cha ac e is ics Fi ing esul s (AICc) k cu e k supe Subjec 1 Subjec 2 Subjec 3 Subjec 4 Subjec 5 To al Baseline model wi h δ=0.14 d −1 0.0 0.0 −16.2 −22.1 −8.1 5.0 −24.9 −66.3 Cu e model Fi ed 0.0 −32.8 −53.2 −47.6 −42.6 −86.5 −262.7 Supe in ec ion model 0.0 Fi ed −12.0 −17.8 −9.5 −5.1 −21.2 −65.6 Full model Fi ed Fi ed −63.8 −51.8 −44.5 −90.6 −83.6 −334.3 Bolded AICc sco es deno e he bes model i among all models o he i e subjec s. E7144 | www.pnas.o g/cgi/doi/10.1073/pnas.1805267115 Ke e al. Downloaded by gues on No embe 30, 2020 Me hods Expe imen al Model and Subjec De ails. Fo his s udy, Me ck Sha p and Dohme Co po a ion p o ided p eexis ing, deiden i ied human blood plasma specimens collec ed om a mul iins i u ional expe imen al d ug ea men p o ocol (MK-5172-004) designed o es he an i i al ac i i y o he HCV p o ease inhibi o g azop e i /MK-5172 (ClinicalT ials.go iden i ie NCT00998985). The specimens we e collec ed, s o ed, and in- en o ied wi h s udy subjec s’in o med consen and app o al o u u e use in he analysis o he e ec s o g azop e i /MK-5172 on HCV eplica- ion, al hough he specimens we e no collec ed speci ically o he e- sea ch p ojec desc ibed he e. Based on hese s ipula ions, he esea ch desc ibed in he cu en epo was deemed by he Ins i u ional Re iew Boa d o he Uni e si y o Pennsyl ania no o cons i u e human subjec s in esea ch. Single-Genome Sequencing. The pa ial NS2 and comple e NS3 and NS4A gene sequences om all eigh subjec s we e gene a ed using he single- genome sequencing me hod p e iously desc ibed (50). The p ime s used o he nes ed PCR included he ollowing: (i) geno ype 1a: i s - ound sense p ime 1aNS2.F2, 5′-ACCCGRCTTTGGTATTTGACATCACC-3′(nucleo ides 2983– 3008, H77); i s - ound an isense p ime 1aNS4B.R3, 5′-TATTGTATCCCACT- GATGAAGTTCCACAT-3′(nucleo ides 5634–5662, H77); second- ound sense p ime 1aNS2.F3, 5′-AAAGTGCCCTACTTYGTGCGCGT-3′(nucleo ides 3063–3085, H77); and second- ound an isense p ime 1aNS4B.R4, 5′-AGGGCCTTCTGCTT- GAACTGCTC-3′(nucleo ides 5517–5539, H77); (ii) geno ype 1b: i s - ound sense p ime 1bNS2.F2, 5′-GCCCGTCGTCTTYTCTGACATGGA-3′(nucleo ides 3257–3280, H77); i s - ound an isense p ime 1bNS4B.R3, 5′-TTCCA- CATGTGCTTCGCCCA-3′(nucleo ides 5622–5641, H77); second- ound sense p ime 1bNS2.F3, 5′-TCATCACCTGGGGGGCAGACA-3′(nucleo ides 3289–3309, H77); and second- ound an isense p ime 1bNS4B.R4, 5′-CGAGCGCCTTCTGCTT- GAATTG-3′(nucleo ides 5520–5541, H77). Vi al Load and Sequence Di e si y Analysis. The RNA measu emen was done by Me ck and Co., Inc., using he Roche Cobas TaqMAN 2.0 assay wi h a lowe limi o quan i ica ion o 25 IU/mL and a limi o de ec ion o 10 IU/mL. A o al o 2,755 NS2–NS3–NS4 genomes we e gene a ed and analyzed om he eigh subjec s. The median numbe o sequences analyzed pe ime poin was 107 (mean, 106; ange, 43–144) (SI Appendix,TableS1). Sequences alignmen s we e ini ially made wi h clus alW and hen hand- checked using Geneious o co ec codon alignmen s. The maximum, minimum, median, and mean di e si y o each sequence se was calcu- la ed based on Hamming dis ance (SI Appendix,TableS1). Fo each sub- jec , phylogene ic ees we e gene a ed by maximum-likelihood me hods using PhyML. The combined phylogene ic ee and he ee om each ime poin we e oo ed wi h consensus sequences om day 0. AB D C E Fig. 3. E olu iona y dynamics o HCV be o e and a e ea men wi h MK-5172. (A–E) Schema ic diag ams o he e olu iona y dynamics in subjec s 1–5, espec i ely. Each mu an is deno ed as a colo ed ci cle, and he colo coding is shown a he le -hand side o he yaxis in each panel. The numbe s wi hin he ci cles deno e he i ness o he mu an ela i e o he baseline s ain in he absence o ea men acco ding o he bes - i pa ame e alues in he bes model o each subjec . The xaxis shows he ime when sequence da a a e aken. The size o he ci cle is scaled acco ding o he equency o he mu an in a gi en sample. Solid a ows show he s ain (whe e he a ow s a s; say, s ain a) o which he majo i y o he sequences in a mu an s ain o in e es (whe e he a ow ends; say, s ain b) a e mos closely ela ed, and he numbe s on each a ow show he numbe o sequences in s ain b ha a e mos ly closely ela ed o s ain a and he o al numbe o sequences in s ain b. Ke e al. PNAS | ol. 115 | no. 30 | E7145 MEDICAL SCIENCES Downloaded by gues on No embe 30, 2020 Cons uc ion o a Baseline Mul is ain HCV Model. We cons uc a i al dynamic model o wi hin-hos in ec ion by mul iple s ains o HCV based on p e ious published models (5, 29, 64). In his model, we keep ack o ndi e en HCV s ains. The o dina y di e en ial equa ions (ODEs) o he model a e as ollows: dT d =ρT·T·0 B B B @ 1− T+P n i=1 Ii+N Tmax 1 C C C A −d·T−X n i=1 β·T·Vi, dIi d =β·T·Vi−δ·Ii, dVi d =ð1−«iÞ· i·p·Ii−c·Vi, «i=D·expð−w·maxð −7,0ÞÞ EC50,i+D·expð−w·maxð −7,0ÞÞ. Ta ge cells (T), p esumably a subse o hepa ocy es, die a a pe -capi a a e, d. Exis ing a ge cells can p oli e a e, and his p oli e a ion is modeled using a lo- gis ic e m, ρT·T·ð1−ðT+Pn i=1Ii+NÞ=Tmax Þ,asinRonge al.(5),whe eρTis a p oli e a ion a e cons an , I i a e cells in ec ed wi h i us s ain i,Nis he con- cen a ion o hepa ocy es ha a e no a ge cells, and T max is he li e ca ying capaci y. Ta ge cells a e in ec ed by he i h s ain o HCV (Vi)a a eβ·Vi.In ec ed cells die a pe -capi a a e, δ. D ug-sensi i e i uses, ep esen ed by s ain i=1, a e p oduced om in ec ed cells a a e ppe cell in he absence o ea men . We assume ha he i h s ain has a i ness o i ela i e o he d ug-sensi i e i uses ( 1 =1). This di e ence in i ness is e lec ed only in he di e ences in i al p o- duc ion in ou model. Thus, he p oduc ion a e o he i h s ain is i·pin he absence o ea men . Unde ea men , we assume ha i al p oduc ion o he i h s ain is educed by a ac o «i,whe e«i=1 co esponds o a 100% e ec i e d ug. The d ug e ec i eness « i is modeled as a unc ion o he d ug concen a ion and he EC 50 alue o each s ain (EC 50,i )acco ding oanE max model (65). We assume ha he d ug concen a ion, D, s ays cons an du ing he i s 7 d o ea men , since d ug concen a ion eaches i s maximal concen a ion wi hin 2–4h(23).A e ea men s ops on day 7, we assume he d ug concen a ion declines exponen- ially a a e, w, ob ained p e iously (23). All i uses a e clea ed a pe -capi a a e c. The alues o he pa ame e s a e shown in SI Appendix, Table S9. Cons uc ion o Models wi h Cu e and Supe in ec ion o In ec ed Cells. We cons uc ed a model inco po a ing cu e and supe in ec ion o in ec ed cells by ex ending he basic model. The ODEs o his model a e as ollows: dT d =ρT·T·0 B B B @ 1− T+P n i=1 Ii+N Tmax 1 C C C A −d·T−X n i=1 β·T·Vi+X n i=1 kcu e ·ð−log10ð1−«iÞÞ ·Ii, dIi d =β·T·Vi−δ·Ii+ksupe ·X n j=1 Mi,j−kcu e ·ð−log10ð1−«iÞÞ ·Ii, dVi d =ð1−«iÞ· i·p·Ii−c·Vi, A B Fig. 4. Concep ual amewo ks o i us e olu ion unde an i i al p essu e. Resis an mu an s expand h ough occupying/compe ing o a ailable eplica ion space. (A) P e ious modeling mos ly assumed ha , once a cell is in ec ed by a i us, he cell emains in ec ed un il dea h. The eplica ion space a ises h ough gene a ion o new a ge cells. Unde his amewo k, esis an mu an s expand h ough in ec ion o newly gene a ed a ge cells, and he a e o he inc ease o mu an equency is mos ly se by he a e a which in ec ed cells die and a e eplaced by newly gene a ed cells. (B) Ou esul s sugges a concep ual amewo k whe e he eplica ion space a ises om mul iple sou ces. In he p esence o po en an i i als, he le el o in acellula HCV RNAs dec eases, leading o cu e o in ec ion in some cells. Replica ion space hus a ises om bo h newly gene a ed cells and cu ed cells. In addi ion, supe in ec ion makes eplica ion space a ailable by allowing esis an i uses o en e an al eady-in ec ed cell and compe e o in acellula esou ces. Thus, cu e and supe in ec ion o cells allow esis an mu an expansion and u no e o occu a a much as e a e han he a e se by he dea h and eplacemen o in ec ed cells only. E7146 | www.pnas.o g/cgi/doi/10.1073/pnas.1805267115 Ke e al. Downloaded by gues on No embe 30, 2020 «i=D·expð−w·maxð −7,0ÞÞ EC50,i+D·expð−w·maxð −7,0ÞÞ, Mi,j=Δ·β·Ij·ViΔ≥0 Δ·β·Ii·VjΔ<0, whe e  Δ=ð1−«iÞ· i−1−«j· j. To model cu e o in ec ed cells, we assumed ha he a e o cu e o in ec ed cells is linea ly dependen on he log 10 o he e icacy o he d ug, log10ð1−«iÞ, and he a e cons an is k cu e . This assump ion is based on a p e ious inding wi h he p o ease inhibi o elap e i ha he a e o second phase o i al load decline inc eased linea ly wi h −log10ð1−«Þ(32). The supe in ec ion we ack he e includes he in ec ion o a cell al eady in ec ed wi h s ain iwi h a i us o s ain j, and he subsequen con e sion o he cell in o a cell ha p oduces i us o s ain j. We assume, o simplici y, ha , i a i e i us, s ain i, en e s a cell in ec ed wi h a less i s ain, s ain j (modeled using he e m β·Ij·Vi), i can ou compe e s ain j, and hen he cell is con e ed o a cell p oduc i ely in ec ed by s ain i. We u he assume he a e o his con e sion is linea ly dependen on he i ness di e ence (Δ) o he wo s ains, Δ=ð1−«iÞ· i−ð1−«jÞ· j, and k supe is a cons an o he e iciency o his con e sion. ACKNOWLEDGMENTS. We hank Ge ald Lea n o assis ance wi h DNA sequence compila ions. We g a e ully acknowledge Me ck Sha p & Dohme Co p., a subsidia y o Me ck & Co., Inc., o p o iding he samples analyzed in his wo k. Po ions o his wo k we e pe o med unde he auspices o he US Depa men o Ene gy unde Con ac DE-AC52-06NA25396. This wo k was unded by Na ional Ins i u es o Heal h G an s R01-OD011095, R01-AI028433, R01-AI078881 ( o A.S.P.), R01-AI116868 ( o R.M.R.), and U19-AI088791 and P30-AI45008 ( o G.M.S.). The con en o his manusc ip is solely he esponsibili y o he au ho s and does no necessa ily ep e- sen he o icial iews o he Na ional Ins i u es o Heal h o o Me ck Sha p & Dohme Co p., a subsidia y o Me ck & Co., Inc. 1. Domingo E, Sheldon J, Pe ales C (2012) Vi al quasispecies e olu ion. Mic obiol Mol Biol Re 76:159–216. 2. Lau ing AS, F ydman J, Andino R (2013) The ole o mu a ional obus ness in RNA i us e olu ion. Na Re Mic obiol 11:327–336. 3. Wei X, e al. (2003) An ibody neu aliza ion and escape by HIV-1. Na u e 422:307–312. 4. E ickson AL, e al. (2001) The ou come o hepa i is C i us in ec ion is p edic ed by escape mu a ions in epi opes a ge ed by cy o oxic T lymphocy es. Immuni y 15: 883–895. 5. Rong L, Daha i H, Ribei o RM, Pe elson AS (2010) Rapid eme gence o p o ease in- hibi o esis ance in hepa i is C i us. Sci T ansl Med 2:30 a32. 6. Rosenbloom DI, Hill AL, Rabi SA, Siliciano RF, Nowak MA (2012) An i e o i al dy- namics de e mines HIV e olu ion and p edic s he apy ou come. Na Med 18: 1378–1385. 7. Keele BF, e al. (2008) Iden i ica ion and cha ac e iza ion o ansmi ed and ea ly ounde i us en elopes in p ima y HIV-1 in ec ion. P oc Na l Acad Sci USA 105: 7552–7557. 8. Adiwijaya BS, e al. (2010) A mul i- a ian , i al dynamic model o geno ype 1 HCV o assess he in i o e olu ion o p o ease-inhibi o esis an a ian s. PLoS Compu Biol 6:e1000745. 9. F idell RA, e al. (2011) Geno ypic and pheno ypic analysis o a ian s esis an o hepa i is C i us nons uc u al p o ein 5A eplica ion complex inhibi o BMS- 790052 in humans: In i o and in i o co ela ions. Hepa ology 54:1924–1935. 10. Ke R, e al. (2014) Modelling clinical da a shows ac i e issue concen a ion o da- cla as i is 10- old lowe han i s plasma concen a ion. J An imic ob Chemo he 69: 724–727. 11. Zhang YY, Summe s J (2000) Low dynamic s a e o i al compe i ion in a ch onic a ian hepadna i us in ec ion. J Vi ol 74:5257–5265. 12. Ke R, Lo e do C, Qi H, Sun R, Lloyd-Smi h JO (2015) Ra ional design and adap i e managemen o combina ion he apies o hepa i is C i us in ec ion. PLoS Compu Biol 11:e1004040. 13. Alexande HK, Bonhoe e S (2012) P e-exis ence and eme gence o d ug esis ance in a gene alized model o in a-hos i al dynamics. Epidemics 4:187–202. 14. Guedj J, Rong L, Daha i H, Pe elson AS (2010) A pe spec i e on modelling hepa i is C i us in ec ion. J Vi al Hepa 17:825–833. 15. Lloyd-Smi h JO (2013) Vaca ed niches, compe i i e elease and he communi y ecol- ogy o pa hogen e adica ion. Philos T ans R Soc Lond B Biol Sci 368:20120150. 16. Bonhoe e S, Nowak MA (1997) P e-exis ence and eme gence o d ug esis ance in HIV-1 in ec ion. P oc Biol Sci 264:631–637. 17. Ciupe SM, Ribei o RM, Nelson PW, Dusheiko G, Pe elson AS (2007) The ole o cells e ac o y o p oduc i e in ec ion in acu e hepa i is B i al dynamics. P oc Na l Acad Sci USA 104:5050–5055. 18. Pe elson AS, Nelson PW (1999) Ma hema ical analysis o HIV-1 dynamics in i o. SIAM Re 41:3–44. 19. Thomas DL (2013) Global con ol o hepa i is C: Whe e challenge mee s oppo uni y. Na Med 19:850–858. 20. Feld JJ, e al.; ASTRAL-1 In es iga o s (2015) So osbu i and Velpa as i o HCV ge- no ype 1, 2, 4, 5, and 6 in ec ion. N Engl J Med 373:2599–2607. 21. Lau G, e al. (2016) E icacy and sa e y o 3-week esponse-guided iple di ec -ac ing an i i al he apy o ch onic hepa i is C in ec ion: A phase 2, open-label, p oo -o - concep s udy. Lance Gas oen e ol Hepa ol 1:97–104. 22. Sulkowski M, e al. (2015) E icacy and sa e y o 8 weeks e sus 12 weeks o ea men wi h g azop e i (MK-5172) and elbas i (MK-8742) wi h o wi hou iba i in in pa ien s wi h hepa i is C i us geno ype 1 mono-in ec ion and HIV/hepa i is C i us co-in ec ion (C-WORTHY): A andomised, open-label phase 2 ial. Lance 385:1087–1097. 23. B aina d DM, e al. (2010) Sa e y and an i i al ac i i y o MK-5172, a no el HCV NS3/ 4a p o ease inhibi o wi h po en ac i i y agains known esis ance mu an s, in ge- no ype 1 and 3 HCV-in ec ed pa ien s. Hepa ology 52:706A–707A. 24. Howe AY, e al. (2014) Vi ologic esis ance analysis om a phase 2 s udy o MK- 5172 combined wi h pegyla ed in e e on/ iba i in in ea men -nai e pa ien s wi h hepa i is C i us geno ype 1 in ec ion. Clin In ec Dis 59:1657–1665. 25. Shimakami T, e al. (2011) P o ease inhibi o - esis an hepa i is C i us mu an s wi h educed i ness om impai ed p oduc ion o in ec ious i us. Gas oen e ology 140: 667–675. 26. Summa V, e al. (2012) MK-5172, a selec i e inhibi o o hepa i is C i us NS3/4a p o ease wi h b oad ac i i y ac oss geno ypes and esis an a ian s. An imic ob Agen s Chemo he 56:4161–4167, and e a um (2014) 58:4995. 27. G en ell BT, e al. (2004) Uni ying he epidemiological and e olu iona y dynamics o pa hogens. Science 303:327–332. 28. Ribei o RM, e al. (2012) Quan i ying he di e si ica ion o hepa i is C i us (HCV) du ing p ima y in ec ion: Es ima es o he in i o mu a ion a e. PLoS Pa hog 8: e1002881. 29. Neumann AU, e al. (1998) Hepa i is C i al dynamics in i o and he an i i al e icacy o in e e on-alpha he apy. Science 282:103–107. 30. Snoeck E, e al. (2010) A comp ehensi e hepa i is C i al kine ic model explaining cu e. Clin Pha macol The 87:706–713. 31. Guedj J, e al. (2013) Modeling shows ha he NS5A inhibi o dacla as i has wo modes o ac ion and yields a sho e es ima e o he hepa i is C i us hal -li e. P oc Na l Acad Sci USA 110:3991–3996. 32. Guedj J, Pe elson AS (2011) Second-phase hepa i is C i us RNA decline du ing elap e i -based he apy inc eases wi h d ug e ec i eness: Implica ions o ea men du a ion. Hepa ology 53:1801–1808. 33. Bligh KJ, McKea ing JA, Rice CM (2002) Highly pe missi e cell lines o subgenomic and genomic hepa i is C i us RNA eplica ion. J Vi ol 76:13001–13014. 34. Lin K, Kwong AD, Lin C (2004) Combina ion o a hepa i is C i us NS3-NS4A p o ease inhibi o and alpha in e e on syne gis ically inhibi s i al RNA eplica ion and acil- i a es i al RNA clea ance in eplicon cells. An imic ob Agen s Chemo he 48: 4784–4792. 35. Lohmann V, Ho mann S, He ian U, Penin F, Ba enschlage R (2003) Vi al and cellula de e minan s o hepa i is C i us RNA eplica ion in cell cul u e. J Vi ol 77:3007–3019. 36. Randall G, G akoui A, Rice CM (2003) Clea ance o eplica ing hepa i is C i us e- plicon RNAs in cell cul u e by small in e e ing RNAs. P oc Na l Acad Sci USA 100: 235–240. 37. Vliegen I, Paeshuyse J, Zhong W, Ney s J (2015) In i o combina ions con aining Tegobu i a e highly e icien in cu ing cells om HCV eplicon and in delaying/ p e en ing he de elopmen o d ug esis ance. An i i al Res 120:112–121. 38. Guedj J, Neumann AU (2010) Unde s anding hepa i is C i al dynamics wi h di ec - ac ing an i i al agen s due o he in e play be ween in acellula eplica ion and cellula in ec ion dynamics. J Theo Biol 267:330–340. 39. Schalle T, e al. (2007) Analysis o hepa i is C i us supe in ec ion exclusion by using no el luo och ome gene- agged i al genomes. J Vi ol 81:4591–4603. 40. Tsche ne DM, e al. (2007) Supe in ec ion exclusion in cells in ec ed wi h hepa i is C i us. J Vi ol 81:3693–3703. 41. Webs e B, O M, G eene WC (2013) E asion o supe in ec ion exclusion and elimi- na ion o p ima y i al RNA by an adap ed s ain o hepa i is C i us. J Vi ol 87: 13354–13369. 42. Bolke BM (2008) Ecological Models and Da a in R (P ince on Uni P ess, P ince on). 43. Desai MM, Fishe DS (2007) Bene icial mu a ion selec ion balance and he e ec o linkage on posi i e selec ion. Gene ics 176:1759–1798. 44. Ge ish PJ, Lenski RE (1998) The a e o compe ing bene icial mu a ions in an asexual popula ion. Gene ica 102-103:127–144. 45. Rong L, e al. (2013) Analysis o hepa i is C i us decline du ing ea men wi h he p o ease inhibi o danop e i using a mul iscale model. PLoS Compu Biol 9: e1002959. 46. Hedskog C, e al. (2015) Cha ac e iza ion o hepa i is C i us in e geno ypic e- combinan s ains and associa ed i ological esponse o so osbu i / iba i in. Hepa ology 61:471–480. 47. Iles JC, e al. (2015) Cha ac e iza ion o hepa i is C i us ecombina ion in Came oon by use o nonspeci ic nex -gene a ion sequencing. J Clin Mic obiol 53:3155–3164. 48. Mo el V, e al. (2011) Gene ic ecombina ion o he hepa i is C i us: Clinical impli- ca ions. J Vi al Hepa 18:77–83. 49. Raghwani J, e al. (2012) O igin and e olu ion o he unique hepa i is C i us ci cu- la ing ecombinan o m 2k/1b. J Vi ol 86:2212–2220. 50. Li H, e al. (2012) Elucida ion o hepa i is C i us ansmission and ea ly di e si ica ion by single genome sequencing. PLoS Pa hog 8:e1002880. 51. S odda d MB, e al. (2015) Iden i ica ion, molecula cloning, and analysis o ull- leng h hepa i is C i us ansmi ed/ ounde geno ypes 1, 3, and 4. MBio 6:e02518. Ke e al. PNAS | ol. 115 | no. 30 | E7147 MEDICAL SCIENCES Downloaded by gues on No embe 30, 2020