Population-level genome sequencing reveals distinct Mycobacterium tuberculosis intrahost mutational trajectories in simian immunodeficiency virus co-infected and antiretroviral treated non-human primates

Whole g enome sequencing o f Mycobact erium tubercul osis (M tb) pop ula tions from clini c a l samples has incr easingly ide n t ified genes under going selection wi thin an d betw een hos ts tha t driv e d iff eren t ial in f ect ion and t r ea t men t outc omes. How ev er , the i n tr a hos t Mtb mut ational landsc ape—especia lly in the c o n text of human i mmunode ficiency vi rus c oinf e ction and an tir et r ovir al ther apy (AR T)—r emain s less clea r , as do the po t ential impacts of such mut ations on Mtb inf ec tion dynamics. Her e, w e perf ormed whole g enome sequencin g of Mtb popula t ions isolat ed from appr o xima t ely 480 inf ec t ed tis sues acr oss 20 non-human prima t es, includ ing animals c o-inf ected with simian i m munode fici ency vir us (SIV) with or without vi r ologic a l suppr ession by AR T . W e iden tified 11 6 mut a tions tha t emer g ed during inf e ction, includ ing those tha t ar e ov er r epr es ent ed within indi vidual tissu es and a subset that ar e shar ed acr oss tissue s during Mtb dissemina tion. W e furth er find di ff er e n tia l mut a tion tr ajecto ries acr oss tr eatmen t gr oups, with higher mut a tion r at e and bact er ial ou tgr owth in SIV-inf ect ed hos ts and increase d pr ev alence of o xi da ti v e damag e-asso cia ted mut a t ions in c oin f ec t ed anima ls on A R T . Finall y , w e demons t rate a c ommon p atte r n o f m u tat i o n i n M t b l i p i d m e t abolism an d polyk etid e s yn thas e g enes and iden ti fy a subset of NHP - deriv ed mut a t ions that ha v e also in dependen t ly arisen i n human clinic al isol a t e s. T og ether , our popula ti on-based sequencing unc ov e r s Mtb div er si fi c a ti on during early inf ec tion, c aptu r es discret e bact e rial dissemination ev e n ts a nd inf er s di f f e r entia l immune pr essur es f aced by Mtb i n th e set ting of SIV-Mtb c oin f ect ion and AR T suppr ession. Importance T uber culosis r emains a leading c ause of d ea th w orl dwide, espec ially in p eople living with HIV . How HIV inf ection and a n ti r etr ovi r al ther apy impacts Mycobac t eriu m tuberculosis (Mtb) in tr ahost ev o luti on r emains un clea r . Using whole g enome sequ encin g fr om hundr eds o f inf ec t ed tissu es fr om non-human pr ima t es, w e find th a t simian immun ode fi ciency v irus c o- inf ec t ed ho s ts and those r ece iving a n tir e tr ovir al ther apy e x act d iff e r en t i m mune pr essur es on Mtb leading t o diff er e n t mut a t ion r a t es and types of DNA damag e tha t ar e incurr ed. Mt b mut a tions w e r e en riched in g enes i n v olv ed in lipid met abol isms and so me of these a r e also .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint 3 seen in human TB str ains. This w or k highligh ts th e r ol e of immune pr ess ur e t o al t er bact e rial pa thw a y s tha t ma y enable Mtb adapta tion t o th e hos t. Backgr ound T uber culosis (TB) r emains the t op s i ngle inf ect ious c ause of mort ali ty worldwide, with people l iving with HIV (PLHIV) bearing an outsiz ed c o s t o f the ongoing epidemic 1 . Studies ha ve shown signific an t impr ov emen t in pr ev enting TB mort ality in this po pula tion t hr ough the widespr ead use of an t ir etr ovir a l therapy (AR T)—especially when st arted early—but ind ividual s on long-t erm AR T s till r et ain an elev at ed r isk o f dev e loping of ac tive TB dis ease despi t e vi r ologi c c ontr ol and st able CD4 c ounts 2–5 . Though s tudies ha v e ide n ti fied d y sfu nctional adapti v e and inna t e immune r esponses assoc iated with HIV /Mtb c oinf e ction (e. g., r eview ed in 6 ) , t he mechanisms acc oun ting f or inc r eas ed TB disease p r ogr ession in PLHIV r emain incomplet ely understo o d . T o model early Myc obact erium tuberculosis (M tb) inf ec tion out c omes r el a ted to HIV c oinf ect ion, a r ece n t study in non-human prima t es (NHP) es t abl ished chr oni c simian immunode ficiency virus (SIV) inf e ct ion fi r s t—includ ing a subset o f an imals with virologic suppr ession using AR T—be f or e cha ll enging animals with a g eneti c ally barc oded isog enic libr a ry of Mtb 7 . This w o rk f ound that SIV-positive animals developed mor e sev e r e TB disease, in cluding higher bact erial burden and gr ea t er i nflamma tion; and tha t early AR T during SIV inf ection c oul d signific a n tly r educe th ese eff ects. Ho w ev er , PET -CT i ma ging , pa thology scoring and sequencing of bar c oded Mtb str ains also showed tha t AR T did not r es t o r e Mtb di ssemina tion t o le v els obser v ed in non-SIV inf ect ed c o n tr o ls —especially to e xtr apulmonary si t es —su g g es ting tha t AR T animals e xperienced con t inuing def e cts in cer t ain aspe cts o f ear ly M tb con tr ol 7 . How ev e r , th e impact of SIV and AR T on th e bac t eri al intr ahos t e v olu tiona ry lands c ape i n this s tudy r emained unclear . Whole g enome sequencing of Mtb c lini c al s tr a ins is in cr easingly iden t if yi ng g enes and pa thw a y s under going selection in hos ts that dri v e di f f e r ential in f e ction and trea tment out c omes 8–10 . These mut a tional signa tur es ha v e been pow erfu l t oo ls f or s t udying long t erm c o- ev oluti on of Mtb wi th human hos ts 10 and identif ica tion of nov e l mut a ti o ns tha t c ontribu t e t o .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint 4 the emer g ence o f antibio tic r esi s t anc e 11,12 , man y of which ha v e be en tr ans la t ed int o d iagnos tic s f or det ec tion of multi- and e xtremely- drug r esis t a n t inf ec tions. Furt her , se quencing of i n tr aho st Mtb popula t ions c an ide n ti fy unfix ed an tibio tic-r esi s t anc e mut a tions tha t emer g e during tr ea tment 13–15 and c an ser v e as a r e cor d of ho s t immune pr essure acting on Mtb 16,17 . How ev er , sequencing is oft en limi t ed t o anat omic al ly accessib le sit e s such the airw a y , with s tudie s sug g es ting ther e is gr eat er Mtb di v ersity p r ese n t in th e human lung and extr apulmonary t issues c ompar ed t o sputum 18,19 . In this s tudy , w e h ypothe siz ed tha t s equencing Mtb g enomes fr om ana to mic ally d iv er se tissues in NHP s w ould c ompr ehensi vely ide n ti fy i n tr ahost mut a t ions, uncov er distinc t immune pr essur es in the cont e xt of SIV c o inf ect ion wi th and withou t AR T and a lso highligh t bio logic pr ocesses that ma y be i mport an t f or Mtb adapt a tion t o the host milieu. Her e, w e perf ormed whole g enome sequencing of Mtb p opula ti ons isol a ted fr om ov er 480 inf ect ed t issues f r om 20 non-human pri ma t es, finding 116 M tb mut a tions that ar ose within the h os t. These mut a tions show tha t the Mtb mut a t ion r a te i s signific a n tly incr eased in SIV-inf ect ed NHP s, while the pr oporti on of o xid a ti v e damag e-associa ted mut a ti ons is incr eas ed in anim als on AR T c ompar ed t o n on-SIV c on tr o ls. W e fur ther sho w tha t Mtb lipid met abolism and b i os yn theti c g enes ar e enriched f or int erho s t mut a t ions, which w e pr opose r eflec t oppor tuniti es f or Mtb adaption t o hos t nutr ient condit ions.

Whole genome sequencing defines t h e in tra hos t M tb mutati onal lan dscape In this w ork, w e cur at ed a set o f Mt b g enomic DNA samples isola t ed fr o m a prior non- human prima t e (NHP) s tudy 7 , which i ncluded naï v e or chr oni c ally SIV-inf ect ed animals (with or without AR T) tha t w er e cha lleng ed wi th a bar c oded libr a ry of Mtb. In prior w ork, the viable Mtb popula ti ons from all in f ec t ed t issues (including lung , thor acic lymph node and e xtr apulmonar y sit es) w er e plat ed for g enomic DNA e xtr acti on and ampli c on-based seq uencing w as used t o de fine th e bar c oded Mtb str ains that w er e inf ec ting each tissue 7 . In this s tudy , w e perf ormed whole g enome sequencing of this DN A fr om a subset of 482 in tr aho s t Mtb popula t ions is olat e d .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint 5 fr om 20 NH P s (Supplemen t ary T able 1), including 7 tha t w e r e sole ly inf ec t ed with Mtb (TB); 4 chr onic a lly in f ec t ed with SIV prior to Mtb challeng e (SIV+TB); and 9 SIV-inf ect ed but v ir ally suppr essed with A R T (SIV+AR T+TB) only 3 da y s aft e r inf ecti on (Figur e 1A). Mtb sequen ces wer e mapped t o a r e cen tl y g ener at ed Mtb Er dman r ef ere nce g enome 20 , which w as deri v ed f r om a r epr es en ta tive colon y f r om the l ibr ary us ed to inf ec t thes e NHP s. Using the haplotype mut a t ion c a lling t ool Mu t ect2, w e d e fined o v er 100,000 single nucleotide polymorphisms (SNP s) and small ins ertion and delet ions ( indels) acr oss all samples, with the v as t majority a t low f r equency and li k ely r ep r ese n t sequ encing artif ac ts ( Supplemen t ary Figure 1A). Subsequen t fil t e ring of low c over ag e sit es and c onside ring only v aria n t s a t >10% pr ev alenc e within tissues (see Methods) y ielde d 143 unique, high -c onfidence c and ida t e mut ations th at w er e not pr ese n t in our ref er en ce g enome (Supplemen t ary T able 2). As some v arian ts c oul d ha v e arisen during libr ary c ons t ruct io n in vitr o (e. g., see ‘B C1’ e x a mple in F igur e 1B), w e further c ompar ed the tissu es that c o n t ain ed a g enomic v arian t ag ains t pr evious ly g ener a t ed barc oding da t a fr om the same samples t o down select 116 mut a tions that w er e lik el y in viv o-deri v ed (se e Methods), as they ar e only f ound in a subset of sit es inf ec t ed by the same bar c oded Mtb s tr a in (e. g., BC2 , BC3 and BC4 in Figur e 1B ). This l e ft 27 mut a ti ons th a t we c an not rul e ou t ar e pr e- e xis ting in th e in f ect ing libr a ry (Supplemen t ary T able 2), s o w e el ected t o e x clude these fr om downs tr eam analy ses. Of the in tr aho s t mut a tion s, 57% w er e pr edicted t o chang e pr o t ein c od ing (53 missense mut a tions, 11 fr ameshift indels and 2 in-fr ame in t ernal delet ions), 31 w er e pr ed ict e d t o b e s ynon ymous and 19 mut a tions w ere l oc at ed wi thin int er g eni c regions (Figur e 1B). Cumula tively , in tr ahost mut a t ions o ccurred acr o ss the M tb g enome, but c ert a in g eno mic r egions harbo r ed mor e mut a tions than w ould be e x pect ed giv en c omplet e ly r andom s ampling (Figur e 1C). Specifi c all y , w e obser v ed loc al mut at ional ma xima betw een nucleotides 1 –1.2 Mb, 1.8–1.9 Mb and 3.5–3.7 Mb, with the highes t m utat i onal densit y f ound betw een 1.8–1.9 Mb, a r egion tha t includes se v er al poly k etide s y n thase g enes ( pk s7 , pk s 8 and p k s11 ). .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint 6 SIV-inf ect ed and an t iretroviral trea t ed NHPs harbor Mtb with divergen t m ut at iona l traject or ies W e ne xt c alculat ed r el a ti v e Mtb muta tion r at es ac r oss t r eatmen t gr oups using the t o t a l number of unique mut a tions identif ied in each animal (see Methods ). Giv en the e x act Mt b g ener a tion time in NHP s is unknow n (though it is e xpect ed t o be signif ic antly slow er than in a x enic cultu r e 16 ) and th a t mut a tion r at es ma y be furt her a f f e ct ed by SIV c oi nf ection and AR T , w e elected t o use a fi x ed g ener a t ion time f or our c alcu la t ions in or de r t o d e fi ne a r ela ti v e mut a t io n rate p e r u n i t t i m e a c r oss c ond itions ra ther than deri v e absolu t e r at es per g ener a tion. In Figur e 2A, w e f ound tha t SIV-c oinf ect ed NHP s (but not those on AR T) had signific antly highe r mut a tional r a tes c ompar ed t o Mtb o nly animals. Ther e ar e sev er a l poss ibl e dri v er s of inc r eased r ela t iv e muta tion ra t e. Firs t, in SIV-inf ect ed animals, ther e w as a s t atisti c al t r end (p=0.09) o f incr eased Mtb r epli c ation o v er the cour se of inf e ction (Figur e 2B), as ev a lua t ed by quanti fying chr omosomal equiv alen ts (CE Q )—t he number of Mtb g enomes in tissue homog ena tes r epr ese n ting the t o t al li v e and dead bact eria bur d en ov er the c our se of in f ect ion. Further , there is signific a n tly mor e li v e Mtb (CFU) in SIV c oinf ec t ed animals a t necr ops y (Figur e 2C), su g g es ting decr eased c apaci ty t o c lear Mtb, w hich ma y also c o n tr ibute to gr eat er r et e n tion of mut ant s tr ains in these immunoc ompr omised hos ts. Sec ond, inc r eased mut ation rate s co u l d a l s o b e d r i ve n b y e l e v a ted DNA d amag e due t o hos t immune a t t ack, with prior w ork sug g es ting tha t the majority of in tr ah os t Mtb mut a tions in la t e n tl y in f ec t ed NHP s and in sputu m of acti v e TB p a tie n ts ar e muta tion s associat ed o xida t ive damag e 16,17 —namely , C>T /G> A tr an sitions and A>G/C>A tr ans v er sions r e fle cting cy t osine deamina tion 21 and 8-o x o-guanine f orma tion 22 , r espect iv e ly . If our obser v e d o xida ti v e mut a t ions ar e associat ed with hos t immune atta c k , we wo u l d h ypothesiz e tha t t issues mark ed by o xida tiv e mut a tions also e x er t gr ea t e r immun e pr essur e on Mtb, leading t o bet te r bact eria l c ontr ol. T o t es t th is, w e fir st down-selec t ed 54 ‘ origin’ tissues (Figur e 3, bo x ed no des) wher e g enomic mut a tions lik e ly ar o se (s ee Method s), as mut a t ions f ound in some tissu es could ha v e hi t chhik e d thr ough Mtb dissemina tion. W e find tha t origin si t es with o xidativ e mut ati ons ha v e signifi c an tl y low er Mtb bur dens c ompar ed t o si t e s associa ted with non-o xida tive mut a tions, c onsi s ten t wit h .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint 7 incr eased immune pr essur e (Figur e 2 D). With this lin k t o ho s t damag e, w e ne xt c ompar ed the pr ev alence of o xidati v e mut a tions a cr oss tr e a tmen t condit ions, finding t ha t SIV+AR T+TB (but not SIV+TB) animals tr ended (p = 0. 09) t ow ar ds mor e o xida tiv e mut a t io ns c ompar ed t o Mtb only animals (Figur e 2E). This si gnal w as driv en by a majority of AR T anim als harboring elev at ed fr equenc ies of C>T /G>A and C>A/ G>T mut a tions (Figur e 2F). Further , ther e is signif ic a n t c orrela t ion betw e en the p r ev alence of o xida t iv e damag e-associa ted mut ations with in vi v o Mtb mut a tion ra t es in SIV+AR T+TB anima ls, but no t f o r th e TB on ly o r SIV+TB animals (Figur e 2G) , sug g es ting tha t Mtb is e xper iencing incr eased o xida t iv e damag e in a s ubset of SIV+AR T+TB animals. This incr eased o xida tive damag e ma y be r ec or ded as earlier ap pear ance of mut a tion s a t origin s it es; and c ons ist en t with th is, w e find tha t lung gr anulomas with in tr ahost mut a t ion s w er e det ec t ed signifi c antly ear lier b y PE T-C T in the SIV+AR T+TB ani mals c ompar ed t o the TB only gr oup (Supplemen t ary Figur e 1B ). T og ether , the da t a sug g es t tha t de s pit e similar bac t er ial out c omes (Figur e 2A-C), AR T vir al suppr ession in some hos ts d oes n ot r est o r e the h os t en vir onmen t t o tha t of n on-SIV a nimals, ins t ead f av oring an immu ne milieu assoc iat e d incr eased o xid a ti v e immune pr essur e . In traho s t mutatio ns unc over dissemin ation even ts wi thin barc o ded Mtb dis semination network s B y int egr a ting our whole g enome s equencing with p r evious barc oding da t a, w e w ere able t o link 79 i n tr ahost mut a t ions with the Mtb barc oded s tr ain that h arbor s th ese v a riants (see Methods). Of these, 67 mut a tio ns w er e f ound in a single tissue (c o nsidering all sampled lung , lymph node and e xtr apulmon ary sites) while 13 v arian ts w e r e s har ed acr oss tissu es (Figur e 3), sug g es ting they lik e ly aros e once and w e r e spr ead thr ough subs equen t dissemina t ion ev en ts. As only some tissues c o lonized by the same bar c oded Mtb s tr ai n ma y harbor a giv en in tr ahost mut ation, w e used in v iv o v ariants as addit ional ‘barc odes’ wi th which t o inf er th e or der o f dis semina tion ev e n ts in eac h Mtb dis semina tion netw or k. As an annot at ed e x ample, w e identif ied 4 in viv o v aria n ts o ccu rring in the same Mtb barc oded strain (Q25_C G TCAG T C ) backgr ound in NHP 232 18 (Fi gur e 3, bot t om righ t). B y mapping the in tr ahos t v ariants pr ese n t and absen t in these tissues, w e ar e able t o inf e r a t least 3 distinc t disse mina tion ev e n ts: f ir s t , .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint 8 tw o tissues r ep r ese n t dissemina t ion of the wildtype par e n t str ain; sec ond, a T C CC AC> T deletion a t posi tion 182022 (r ed r ing) emer g ed and w as then d issemina ted to 1 2 tissues; and th ir d, a new 277532_ C>T v arian t (bla ck r i ng) ar ose on t op of the T CCC AC> T deletion and both mut a tions w e r e then diss emina t ed to 7 other si t es. Finall y , tw o v a riants —one (2554553_ C>T , or ang e ring) th a t appea r ed in an oth erwise wi ldtype backgr ound and one (2806057_G>T , blue ring) tha t appear ed on t op of the T CCC AC> T deletion—emer g ed independen tly in s ingle tissues lik ely af t er these pr ior d issemina ti on ev en ts. Similar br anched netw ork s wer e in f e r able acr oss othe r animals (e. g., NH P 22517 (gr ey), NHP 3351 9 (gr een), NHP 24018 (gr een) and NHP 22117 (y ellow)), suppo rting a model where Mtb dissemina t ion is s truc tur ed as p ar allel subnetw o rk s of diss emina tion r a ther than a single r adia tion ev ent. Furthe r , w e find 7 i ns t ances (Figur e 3, arr ow s) whe r e we c an inf er directio nal dissemina tion (i.e., wher e a mut ation appear s unfix ed a t an or igin si t e and then hi t chhik e s as a fix ed mut ation t o downs tream tissue s). Of these, 5 d issemina tio n ev e n ts i n v olv e Mtb spr eading fr om and t o lung si t es (Figur e 3, r e d arr ow s), while 2 c ases in v o lv e disse mina tion t o lung si t e s fr om a thor acic lymph node (Figur e 3 , blue arr ow s), sug g es ting tha t Mtb can in princip le r eturn t o the lung aft er d issemina tio n t o oth er ana tomic al si t es. Mtb mut a tions in NHPs and huma ns are enriched in l ipi d met abol ism and bios yn t hetic processes W e ne xt c orr elat ed th e NHP in tr ahos t mut a tions ag ains t two tr e a tmen t-naïv e human s tudies that als o ide n ti fied intr ahos t Mtb mut ations: L iu et al. 17 (321 mut a ti ons f ound in 25 7 g enes in sputum fr om non-HIV inf ec t ed p a tie n ts with a cti v e TB); and Lie berman et al. 18 ( 5 18 in tr ahost mut a t ions in 403 g enes fr o m an aut ops y s tudy o f HIV-c oinf ec t e d individua ls). In all 3 da t asets, in tr aho s t mut a tions w e r e mos tly f ound once in diff er e n t g en es, which pr ecluded looking at the r atio of non-s ynon ymous t o s ynon ymous mut a t ions on a g e ne basis; and only th e ir on tr anspor t e r , irtA, w as mut a t ed i n all thr e e da t ase ts (Figur e 4A). Despit e la ck of on e-t o-one g ene ov erlap, ther e w as a signifi c a n t enrichment of intr ahos t mut a t io ns in My c obr osw e r 23 - pr edic t ed l ipid met abol ism g enes ( Figur e 4B) in the NHP da t aset (15% of mut a ti ons) an d .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint 9 Lieberman et al. (9% of mut a t ions) d a t asets, with a st atisti c al t r end (p=0. 05) f or Liu et al. (10% of mut a tion s). Further , w e find tha t in tr a-NHP mut a t ed g enes had signif ic a n tly mor e pr edic t e d in t er ac tions than e xpected by chanc e (p = 0.005) using the STRING da t abase 24 and th a t the put a ti v e g en e int er ac tions c lust ered by My coBr ow ser func tional ca t egories, with lipi d met abolism g enes (especial ly pk s g e nes) centr ally l oca t ed (Supplemen t a r y Figur e 1C). In t erPr o analy sis identif ied sever al pr o t ein domains tha t w er e enrich ed in the NHP da t ase t (Supplemen t ary Figur e 2), which inc l uded lip id met abolism domains ( e. g., acyltr ansf er ase (FDR< 1x 10 -5 )) and po lyk etid e s ynthase-associat ed domains (FDR < 3 x 10 -3 ). The g enes associat e d with these t e rms w er e highly ov erlap ping and included pk s7, pk s8, pk s12, mbtD , pk s1 and pk s2 , which ar e inv olved in the bios y n th esis of d iv er s e cel l w all lipid s tru ct ur es and se c ondar y met abolit es. As in NHP s, w e did obse r v e enrichmen t of po lyk etid e domains t erms in the human da t asets as w ell (Supplemen t a ry Figu r e 3), sug g es ting these bios ynthetic g enes as a whole ar e c ommon t ar g ets of mut a t ion in v iv o. In traho s t mutatio ns are enriched wit h in gran ulomas a nd are associ at ed wi t h dissemina tion W e ne xt assessed whether in vivo deriv ed mut ations c an conf er i n trahos t fi tness bene fi ts dur ing inf e ction, f o cusing on mut a tion s th a t chang e pr o t ein sequence (missens e, indels), which ar e mor e lik e ly t o impact g ene funct ion and Mtb fitness c ompar ed t o s ynon ymous mut a tions. T o do this, w e c onsid er ed M tb fi tness acr os s t w o ana t omic a l sc a les: either lo c al ly within an ind ividual t issue or s y st emic a lly through dissemina tion t o multip l e tissues. Fir st, t o assess ind ividual tiss ue fitn ess, w e c ompar ed the proport i on of wild type a llel es t o non-s ynon ymous and s ynon ymous mut a tions, f inding tha t the pr ev alence o f non- s ynon ymous mut a tions w er e signif ican tly higher than tha t o f s ynon ymous alleles (Figur e 4C), sug g es ting tha t a subset of these mut a ti ons ma y be loc a lly ben e f icial f o r Mtb. Sec ond, t o s tud y dissemina tion, w e c ompar ed th e fr ac tion of tissue s inf e ct ed by a bar c o de d Mtb s tr ain tha t also c ont ained a non-s ynon ymous or s ynon ymous g enomic v arian t, with the h ypothesis th a t dissemina tion-pr omoting mut a tions will be enrich ed within dis semina t e d netw ork s. Her e, we .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint 10 did not obser v e a st atistic a l di ff er en ce betw een s yn on ymous and non-s ynon ymous mut a tions, though ther e is a wide r ang e of dissemina tion in bo th gr oups (Figur e 4D). Int er e s tingly , mut a tions th a t wer e asso ciat ed wi th highly disseminat ed Mtb s tr ains ar e dis tin ct fr om those tha t ar e enrich ed in ind ividual tissue s, which ma y su g g es t tha t addi tional f act or s beyond lo c a l fitness c o n tribu t e t o d issemina tion. NHP derived Mtb mut atio ns also arise in huma n clinical s trai ns W e finally ass essed whethe r i n tr ahost NHP mut a tions c an be found acros s human Mtb clini c al str ains. T o do th is, w e lev er aged a r ece n t ph ylog ene tic anal y sis tha t mapped ov er 55,000 Mtb cl ini c al g enomes t o an ance s tral g enome 25 and l ook ed f or identic a l mut a ti ons as thos e obser v ed in ou r NHP da t aset. W e find 22 mut a tions (12 non-s ynon ymous, 9 s ynon ymous and 3 in t e r g enic ) th a t w er e also pr ese n t in a t least one human Mtb clin ic al str ain (T able 1). M os t o f these mut a tions appear in a single clinic al is olat e, but some mut a tions (e. g., pk s7 fr ameshift a t c odon 1174 and non-s ynon ymous mut a tions in r v0785, r v0987 and r v3594) ar ose independently acr os s diff e r en t inf ec tions and w er e suc cessfully tr ansmit t ed betw e en peopl e (e. g., a missense mut a tion in rnj, a gene whose l oss has been ass ociat ed with drug toler ance 26 , ar ose once but w as subsequen t ly s har ed acr oss 10 individuals ). W e al so find an int er g eni c mut a tion 25 nucleo tides upstr eam of the E SX-1 r egula t or whiB6, whic h is an ar ea of high mut a tional d iv er s ity ac r oss c lini c al str ains 25 . Finally , w e show tha t intr a-NHP mut a tions tha t ma t ch mut a tions in clin ic a l s tr ains are signifi c an tl y enri ched f or C>T /G>A mut a tions (Figur e 4E), further support ing a model in which i m mune o xida tive damag e is dri v er o f Mtb div e r sif ica tion in the hos t.

De fin ing the in tr ahost Mtb mut a tio nal r eper t oire within trea tmen t-nai v e hos ts ma y inf orm the ev olu tiona ry pr e ssur es s haping Mtb adapt a t ion t o the host and also r e c o r d how immune en vir onmen ts ma y dif f er in individuals who ha ve othe r c omorbi dities such as HIV c o- .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint 11 inf ect ion. Her e, sequencing of Mtb fr om inf ect ed NHP hos ts finds th a t SIV c oinf ection w a s associat ed with signifi c antly incr ease d Mtb mut a t ional d iv er s ity c ompar ed t o M tb onl y animals, lik ely du e t o tw o non-mutually e x clus iv e mechanisms associat ed with immunosuppr ession: fir st, gr ea t er over all Mtb r epli c ation c ap acity in SIV-inf ect ed h os ts th a t c r e a t e oppo rtuni ties f or mut a tions and s ec ond, in cr eased r e t ention o f dele t eri ous spont aneous mut a tions tha t ma y otherwise be c lear ed by the host. Con v er sel y , incr eased o xidative immune pr essur e is not l ik ely driving mut a tion, as ch r onic SIV coi nf ection is known t o impair immune c ontr ol in th is NHP model 28,29 and w e obser v e f ew er o xi da tive damag e associat ed mut ations in SIV animals r elative t o Mtb on ly c ontr ols, a f inding tha t h as also been repor t ed in HIV-positive pa tie n ts 17,18 . Further , w e also ev aluat ed th e impact of AR T on Mtb e v olut ion with in SIV-inf ect ed hos ts. Prior bar coding analy ses sug g es t ed th a t SIV+AR T+TB animals r emained mor e permissiv e f or Mt b dissemina tion t o e xtr apulmonary si tes, sugg es ting a lev el of c on t inuing immune dy s function despit e vir o logic con t r ol 7 . Her e, w e f i nd Mtb from a majority of A R T animals had an inc r eased pr oporti on of o xida t iv e damag e asso cia ted mut a t ions relativ e to TB only animals, and tha t t his signal w as posi tively c or r el a t ed with higher ov er all mut a t ional ra t es. W e speculat e tha t a subse t of SIV+AR T+TB ani mals c ould be moun ting a dy sr egulat ed heigh tened infla m ma t ory response t o Mtb inf ecti on. Pr evious w or k has shown that a lv eo lar macr ophag e s fr om HIV-neg a tiv e individuals on pre-e xposur e AR T pr oph yla xis moun t an alt ered immune r esponse t o subsequent Mtb e xposur e 30 and as w e did not ha v e samples fr om an SIV -n egat i ve A RT o n l y g r oup, w e c annot d iscern if a n ti r etr ovir al e xpos ur e alone or a c ombina ti on of SIV a nd AR T is r espons ible f or this sh ift in Mtb o xidati v e damag e. Our findings also c o ntex tualiz e bac terial ev o luti on and dissemina t ion wi thin the ho s t. Fir s t, as gr anulomas ar e typ ic a lly c o l oniz ed by a single bacterial f ounder and tha t di ff er e n t Mtb isog enic s tr ains ha v e het e r og eneous dissemina t ion c apaci ty e v en wi thin the same hos t 31 , we w er e ab le t o in c orpor at e intr ahost mut a t ions as se c onda ry ‘barc od es’ t o fur ther s tud y otherwise cl onally barc oded s tr ains and inf er c omple x dissemination pat t erns wher e Mtb is lik ely spr ead ing in par all el subne twork s s tr ains originating fr om di ff er e n t t issues. This is in c ontr as t t o a single ev e n t di ssemina ti on model, wher e all futu r e dissemin a t ed si t es a r e seede d fr om a single early lesion. W e do n ot e that a majority of the dissemination netw or k s harbor .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint 12 mut a tions a t a single tissue, which c ould be c ons ist ent with a single ev en t model, but giv en ther e ar e f ew in tr ahost mut a tion s det ect ed i n g ener al, w e c anno t rule out par alle l dissemina tion in these c ases as w ell. Sec ond, our g enomic mapping f ound tha t in tr aho s t mut a t ions w er e enr iched in Mtb lipid c at abol ism and bios y n thet ic g e nes (Figur e 5), whi ch ma y re fle ct th e import ance of bet a- o xida tion o f host lip ids f or M tb nut rien t acquisi tion du ring inf ec tion 32 . While bet a-o xida tio n supports c arbon acqu isit ion, it a lso in cr eases reducti v e s tr es s by g ener a t in g NADH , which c an be c ount er ac t ed thr ough sev er al mechanisms. Fir s t, NADH is c onsumed thr o ugh r espir a t ion, an d w e obser v e mut a tions in put ative nitrite ef flux pr o t eins narK3 and narU , which m a y f acilit ate the use of ni trate a s a t erminal elec tron accept or in the cont e xt of micr oa er ophili c c ond itions 33 . Sec ond, NADPH equiv alen ts ar e util ized f or b ios yn th esis of c ell en v e lope componen ts and lipid s t or ag e molecules by la r g e modular polyk etide s y n thase (p k s). The pk s en zymes ar e r esponsible f or pr oducing k ey cel l w all c o mponen ts (e. g., dim y c ocer osa t e s, m y c ocer osic ac ids, lipoar abinomannan, mannos yl-ß-1 phosphom y c ok et ides, sulf o lipid and the s ider opho r e m y c obactin) using l ipid pr ecur s ors 34 ; and recent w o rk has show n that p h thio cer ol dim y c ocer osat e pr oduct ion is r eq uir ed f o r optimal Mtb gr owth o n pr opionat e, whic h pr esumably acts t o r e-balance the r e ductiv e s t r ess fr om odd-chain l ipid d e gr ada tion 35 . Her e, w e find signi fic a n t enrichment o f i n tr ah os t mut ations—both in NHP and hu man da t asets—within pk s g enes, which ma y be bene ficia l to Mtb in the human popula tion; f or ex ample, w e iden tifie d a fr ameshift in pk s7 tha t also appe ar s t o be independently mut at ed 7 times acr oss human clini c al str ains and has b een tr ansmi t t ed bet w een 24 indiv iduals (T able 1 ). In t e r es t ingly , while pk s7 kn ock out str ains hav e be en re port ed to be a tt enu a t ed in aer oso l inf ect ion in C57BL/6 mice 36 , ou r dat a would sug g es t thi s en z yme c an be dispensabl e in h uman inf ecti on an d tr ansmission, which w ould be c ons i s t e n t with r e cen t w or k showing p ks7 insert ion mut an t s r et aining the abili ty t o i n tr a v enou sly i nf ect a g enetic a lly diver se panel of mice 37 . T og ether , these mut a tional s igna tur es h ighligh t th e met abolic fle x ibili ty o f Mtb in balanc ing ener gy and r edo x needs under ho s t-imposed s tr e sses. .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint 13 Limit at ions t o t he s tudy Sev er al limit a t ions t o this s tudy r e main. Fir s t, our da t aset li k ely unde res tima t e s true in tr ahost mut a tiona l div e r sit y , as la te-emer ging mut a tions (which will be a t low er pr ev a lenc e c ompar ed t o earl y arising mut a tions) ar e unlik ely t o be sampled f or the sequencing depth used in this study . Further , some v ariants will also ha v e b een fi lt e r ed out by our 10% pr ev alence thr eshold, which w as imposed to reduce f alse posi ti v e c al ling of low-fr equency sequen cing err or s. Final ly , highly r epeti tive g enes of th e PE/PP E f amilies ar e di f f icult to uniquely map using short r ead sequencing t echnol ogy , so w e ha v e mask ed these r egions in ou r r ef erence g enome 20 and subsequen t anal y ses. Sec ond, our ana ly ses ar e r e s tri ct ed to an ear ly in f ect ion time poi n t when the bac t er ial popula ti on siz e remains siz eable but mut a tional d iv er s ity ma y be limit ed c ompar ed t o human inf ect ion with Mtb, wher e inf e ctions of mon ths t o y ear s enable Mtb t o accumula t e in tr aho st mut a tions. On the opposi t e end, human longitudinal s tudies ha v e s hown tha t an tibio ti c tr ea tment (and pr esumably immune a tt ack leading to l a t e n t inf ec tion) will shr ink the in vi v o airw a y mut an t pool 13–15 , which will l ik ely el imina t e subsets of mut ations tha t w er e ben e fi cia l early—but not l a t e—in in f ect ion. Further , this NHP s tudy do es not c aptur e popu la t ion bot tlen eck s o ccurr ing during in t er-h os t tr ansmission, wher e mut a tions t ha t ma y be bene ficial within a ho s t al so mus t suc cessful ly access the a irw a y , sur v iv e e xpo sur e t o th e outs ide en vir onmen t and tr ansmit inf ec tion to a new hos t. This high bar f or sp r ea d betw een hosts ma y e xplain wh y the v as t majority of mut a tions f ound acr oss ov er 55,000 Mtb clini c al s tr ains e xi s t at t erminal br anches within single indiv iduals 27 and tha t 81% of in tr a-NHP mut a tions f ound here did not ha v e a dir e ct c ount erpar t in t he human clinic a l s tr ain c o llect ion. Finally , while g ene-lev el dat a sug g es ts ongoing Mtb mut a tion lipid and met abolic g enes , it is di f f icult t o inf er whether a gi v en missense or s ynon ymous mut ation is pos iti v ely or neg a tiv ely impacting g ene function and ther eby nomina ting pa thw a y s as essen tia l or non- essen tia l f or host adapta tion. Additi onally , as the e x act l ipid subs tr at es o f man y par alogues of g enes in v olved in lipid met abolism and bios yn thesis r emain unknown, futur e w ork will b e .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint 14 needed t o d e fin e the nutr ient div ersi ty of Mtb host micr oen vir onments—especiall y a t r e la t iv e ly inaccessib le anat omic si t es bey ond sp utum—tha t pr omot e and r estri ct suc cess fu l Mtb in f ect ion.

Sequencing o f Mtb genomic po pul ati ons from in f ect ed NHP tissues W e cur a t ed a set o f Mtb g enomic DNA samples tha t w er e pr evi ously g ene rate d fo r bar c ode s equencing 7 . In or der t o c om par e Mtb e v olut ion across similar inf e ction time sc al es, w e chose t o ana ly z e samples f r om the 20 NHP s fr om the original tha t sur vi v e d 10-12 w eek s of Mtb inf ect ion, though w e did sequence tissues fr om 3 addi tional SIV c oin f e ct ed NHP s r each ed a humane endpoin t much earlier . Wh ole g enome sequencing w as perf or med on the Illumina Nov aSeq S4 300 cy cle platf orm acr oss 4 sequen cing runs (see Supplemen t a ry T able 2 f or metric s and sequencing a ccession c odes). R e ads w er e f ilter ed f or good sequencin g quality r eads using fa st p 38 and then mapped t o the r ec en tly g ener at ed ba r c od ed Mtb Er dman r ef e r ence g enome (GenBank Access ion CP172229.1) 20 , which w as obt ained fr om an al iquot of the libr a ry used to inf ect NHP s acr oss s t udies. Af t er alignmen t, w e applied Mut e ct2 f rom the GA TK t oolki t ( h t tps:// g a tk.br oadins t itu t e.or g /hc/ e n-us ) t o iden t ify al l mut a tiona l chang es (single nucleotid e polymorphisms and insertion/de letio ns) a t an y fr eque ncy r e la ti v e t o the ref er en ce g enome. The functiona l eff ect of each mut a t ion (e. g. in t er g en ic, s ynon ymous, missense, fr ameshift) w as a lso pr edic t ed using SnpE f f based on our recen t Er dman annot ation 20 . The Mut ect2 ou tput y ielded a pr e liminary l is t of o v er 100,000 det ect ed g enomi c chang es acr oss al l samples, which wer e mainly f ound a t low pr ev a lence (<10% mut an t allel es .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint 15 v er sus wi ldtype sequenc es) and dri ven primary by a subset of samples. T o fi ler these v aria n ts, w e enacted 3 QC s teps: firs t, w e us ed the i n ter quar tile r ang e (IQR) method t o de f ine ou tlie r samples with a number of Mu t ec t2 varian ts a t 1.5x gr eat er than the IQR a cr oss all the samples fr om each sequen cing ba tch. This yielde d 46 out lier samples that w er e remov ed fr om subsequen t anal y ses (Supplemen t al Figur e 1A, ‘ QC1’). Not ably , the v ast majority mut a tion s in these outl ier s w er e enr iched f or C to A tr ans v er sions (Supplemen t al Figur e 1B), a mut a tional chang e tha t has been pr ev iously ass ociat ed with DNA damag e artif acts arising during libr ar y pr epar a tion due t o eleva t ed he a t 39 . Sec ond, w e filt e r ed r emaining v ari an ts t o remov e low- c onfidenc e c a lls at si t es with subst a ndar d sequencing depth, which inc l uded only inc luded 1) sequencing runs wi th an a ver ag e r ea d depth abov e 50x; 2) g enomic v aria n ts th a t s a t in r egion s wher e ther e w as ov er a ll sequenc ing depth tha t e x ceeded 50% of the a v erag e sequencing depth of the sample; and 3) v ariants whos e pr ev alenc e w as abo v e 10% (r e la t ive t o all r eads a t that position ) t o acc ou n t f o r r ar e sequen cing err or s (Supplement al Figur e 1A, ‘ QC2’). Thir d, due to dif f icult y in uniquely ass igning short r eads t o r epetit iv e r egions in t he Mtb g enome, w e r easoned th a t the r e wil l be s ys t ema t ic mismapping ev en ts yield ing the sa me e x act ‘mut a tion’ (in both chr omosomal position and type of DNA base chang e) being f o und in samples ev en acr oss diff eren t animals. Filt ering o ut the e x act same v arian ts f ound in mor e than 1 animal, yielded a c andidat e li s t of 217 v ariants f ound acr oss 125 tissu es (Suppl emen t al Figur e 1A, ‘ QC3 s t ep’). Finall y , w e visuall y inspec t ed the r ead pi leups g ener a ted by eac h mapping file using Sam t ools and c onfi rmed tha t the d iff er en t r eads w ere mapping acr oss the sit e o f int eres t an d tha t v aria n ts w ere not clu s t e ring a t the 5’ and 3 ’ ends of r eads. This pr oduced our fina l v alidat ed t able of 143 uniqu e hig h c onfid ence d e nov o mut a t ions f or fur ther ana ly s is (Supplemen t ary T able 1). Definin g pre-e xis ting a nd i n tra hos t-de rived Mtb mut a tions As our r ef er e nce Er dman g enome was deri v ed fr om a single barc oded c lone with in a lar g er barc oded l ibr ary , w e c annot e xclude that some mut a tions iden t ifi ed by sequencing c ould ha v e arisen in vitr o is spec ific t o diff er en t barc oded s tr a in. T o c onser v a ti v ely separ at e in vivo .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint 16 deriv ed mut a ti ons v er sus those that ma y ha v e been p r e-e xis ting i n the lib r ary , w e c ompiled th e Mtb bar c od ed s tr ain s pr ese n t i n eac h tissue sample from prior barc ode amplic on sequenc ing 7 and c ompar ed this t o the Mutect2 det ected v ar ian ts pr es en t in each tis sue. As diagr amed in Figur e 1B , w e c onside r ed mut a t ions i n viv o-deri v ed i f they satisfied 1 o f th e 2 f ollowing criteria. Fir s t, w e fir s t lo ok ed f or tiss ues tha t c ont ained a g enomic v arian t and a si ngle bar c oded Mtb s tr ain, which allow ed us t o link a g enomic v ariant t o a barc oded Mt b s tr ain g en ome. W e then iden t ifi ed all tissues c ont a ining tha t bar c ode d Mtb s tr ain and assessed the pr esence/ absence of the g enomic v arian t of i n t e r est in those tissu es. If w e obser v ed a p atte r n wher e some ti ssues had only the wild t ype al lele while other s had the g enomic v ariant, we r easoned the det ec t ed mut a t ion lik e l y origina t ed dur ing inf ection. This sit ua tion is diagr ammed as ‘BC2’ in Figur e 1B. Sec ond, when w e c annot link a g enomic v arian t t o a single bar c od ed Mt b s tr ain (often the c as e when a muta tion is found in a tissue that w as in f ec t ed by mor e th an one barc oded Mtb s tr ain), w e look ed f or the g enomic v a rian t of int er e s t acr oss all tissues c ont aining ev e ry p ossibl e bar c oded M tb s tr ain it c ould b e link e d t o. If w e obser v ed bo th pr esen ce of wild typ e Mtb and mut an t alle les in every poten t ial ba r c ode netw ork, w e also conside r ed this v ar ian t as l ik el y emer ging in viv o. This situation i s dia gr a med as ‘BC3’ and ‘BC4’ in Fi gur e 1B. If a v ariant does no t sa t isfy th e cri t e ria abov e, w e c annot ru le ou t the m ut a ti on could ha v e emer g ed in vitr o , so it is consid er ed ‘pr e-e xis ting ’ and e x cluded fr om further analy ses. An e x ample of this is ‘BC1’ in Figur e 1 B , wher e a mut a tion w as pr esent i n all tissues th a t w as inf ec t ed that barc oded str ain. Definin g mut a n t origin tissues a nd t i ming of or igin emergence T o de fine mut a ti on ‘ origin’ sit es, w e fir s t id en t ifie d tissues c ont aining a si ngle bar c oded Mtb s tr ain. If si t es c o n t ain an unfix ed mut an t al lele (i.e., <75% of a ll reads a t tha t si t e are mut an t), then we c onside r ed th ese tissues as harbor ing both wild typ e (i.e., par e n t al ) and mut an t alle les. These ar e lik ely the origin si t es where the muta tion or iginally ar ose. In the .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint 17 par en t s tudy , animal lungs w er e imag ed using PET -C T on a mon thly basis at 4-, 8, and 1 2- w eek s pos t inf ec tion t o r egist er the lo c a ti o n and pr es ence o f newly det ec t ed lu ng lesions 7 . Her e, w e c ompar ed the time of PET-C T det ecti on of the lung-specific origin si t es (PET-C T r egis tr a tion w a s not done f or thor acic lymph node an d e xtr apulmonary si t es) a cr oss t r eat men t gr oups. Mut at ion ra t e qu antificat ion Similar t o p r evious w o rk us ing c ol on y based sequenc ing fr om inf ec t e d NHP s 16 , w e c alcul a t ed a Mtb mut ation ra t e f or e ach animal by dividing th e number of unique muta tiona l ev en ts (af t er normalizing f or number of tissues sampled) by the number of bact erial g ener a tions (per unit of inf ec tion ti me) occurring in the hos t. Her e, su m med the number o f diff er e n t mut a ti ons across al l tissues in a NHP t o e s tima te t he mut ation r at e o f the en t ir e Mtb popula ti on within one animal using t he f ollowing equa ti on: μ = m 4i354i35[ N4i35 ∗ 4i35( t4i354i354i35 g )] / D In this equation, μ r ep r esents the t o tal mut a tion ra t e a cr oss an animal, m is the number of uniqu e mut a t ional ev ents de t ec t e d by sequenc ing acr oss all tiss ues in the an imal, N is the Mtb g enome siz e (set a t 4,400,000 bp), t is the t ot al inf ec tion t ime of each NHP (da y s of inf ect ion multipl ied by 24 hour s), g is the Mtb in viv o g ener a t ion time (arb itr arily used 18 hour s f or a single doubl ing) and D is th e tot al number of tissues that w e r e s e quenced per animal. Please not e that as the e x act Mtb div ision r a te within a giv en animal is unknown, the numeric al mut a tion r at e is not absolute; it is in s t ead the rela tive mut a ti on r at e per fix ed un it o f inf ecti on time in this s tudy . Bact erial burden qua n ti ficatio n acros s animals W e obt ained pr eviously published t is sue-lev el li v e bact e rial bur den (CFU) and t ot al ( liv e and dead) bacterial chr omosomal eq uiv alents (CE Q ) measur emen ts f or t he animals tha t w er e included in this s tudy 7 . T o t al CFU an d CE Qs w er e summed f or a ll sampl ed tissu es fr om each animal f or s t atisti c al comparisons. .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint 18 Bact erial barc o ding network visual iz a tion Using pr eviously publ ished amplicon bar c oding dat a 7 fr om the tis sues seq uenced in th is s tudy , w e g ener at ed netw ork plo ts by c on vert ing a t abl e o f g enomic varian ts pr ese n t in a ll sequenced tissu es int o a netw ork plo t using the R pack ag e, igr aph. Specific all y , each tissue tha t c ont ained a giv en barc oded Mtb str ain w as initiall y plott ed as nodes given the same c olor . A ring w as added in a di ff er e n t c o lor to a ti ssue/node if it c ont ained a uni que g enomic v aria nt aft er s equencing such that d iff eren t v arian ts within the same bar coded d i ssemina tion ne tw ork ar e r epr es ent ed by diff er e n t c olo red rings. Using the pr esence or absence of g enomic mut a tions shar ed ac r oss ti ssues, w e inf erred dissemina t ion subnetw o rk s with th e f ew e st dissemina tion e v ents. In the c ase of netw ork s f o r which w e a r e unable to or der th e ev e n ts of dissemina tion, we arb itr aril y repr es en t ed dissemination as a s ingle star-lik e netw or k. If a g enomic v arian t w as f ound in mor e than one tissu e with a netw or k, w e manually r e-or der e d c onnec ting lines betw een th ese nod es t o f orm sub-netw ork s. W e finally import ed the igr aph bar c ode netw or k s int o Adobe Illustr a t or t o visuall y impr ov e line de f init i on, upda t e c olo r s f or visualiza tion and sc a le f or siz e. The e dg elis t r epresenting the c onnec tions betw een tis sues in f o r all visual iz ed barc oded netw o rk s is provided as an R objec t in Supplementary Dat a File 1. No t e: some tissues harb or ed mult iple g en o mic mut a tions bu t igr aph is limit ed to adding a single r ing , so w e added these v ar ian ts and th e sit e o f origin si t es (as bo x e s) in Adobe Illus tr at or . The pr esence of g en omic mut a ti ons and origin si t es wi thin the networ k s ar e i ndic at ed by met ada ta c olumns in Supplemen t a ry Dat a File 1 . STRING and In t erpro a nalyses The lis t o f all g enes with in v iv o d eri v ed mut a t ions from our NHP da t a, Li u et al. 17 and Liberman et al. 18 w ere anal y z ed o n the STRING da t abase ( h t tps:// s tri ng-db.or g ) to d ef i n e put a ti v e int er act ions. All int er act ing nodes w ere then colored by th eir p r edic t ed 23 functi ona l c at egories. Finally , pr o t ein domain en richmen t analysis (Int erPr o) w as performed on the STRING da t abase f or a ll g enes. .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint 19 Iden tif ication of i n tr a-NHP mut ations in human Mtb cl inical strains W e obt ained the lis t of g enomic mut a tions identif ied acr oss appr o xi ma t ely 55,000 publicly a v ai lable Mtb c lini c al iso late sequences fr om a r ece n t publi c at ion 27 , which mapped mut a tions ag ains t a r e c onstruc t ed a nces tr al Mtb g enome and c on s truc t e d a ph ylog enetic tree t o ide n ti fy independe n t mut a t ional ev en ts and sha ring of mut a ti ons ac r oss indiv iduals. W e sear ched f or NHP deriv ed intr ag eni c mut a tions that also show ed up with the same DNA mut a tion (e. g. C>T) a t the same c odon position in the human Mtb cli nic al is olat e dat a. F or in t e r g enic sequen ces, w e used BLAST t o sear ch f or iden t ic a l s tr e t ches o f g enomic sequence ar ound the mut a tion si t e and then i den ti fied the r e lev an t g enomic posit i on of the mut a tion in the ancestr al Mtb g enome used t o map the human clini c al iso lat es. Dat a ava ilab ili ty The g enomic mappin g , quality c on tr o l and v arian t det ec tion c od e and met hods ar e full y described at the f ol lowing Github li nk: h t tps:// github.c om/F ortune-Lab/ Mtb-NHP -g enotyping . All sequencing da t a has been deposited on the Sequencing R ead Ar chiv e, with accession c od es and link s f or ind ividual samples pr ovi ded in Supplemen t a ry T able 1. All seq uencing da t a are als o a v ailable under B iopr ojec t Accession PRJNA1432 747. All met ada t a associat ed with the d a t aset s in this w or k ar e depos ited on F ai r domhub a t th e f ollowing lin k: h t tps://f air domhub.or g / s tudies/1401 Acknowledgments This pr oject has been funded in part with f eder al funds f r om the Na tion al Ins titu t e o f Aller gy & I nf ectious Dis eases, Na tio nal Ins titu t es of Health under con tract 75N93019 C000 71 and NIH, R01 AI134 195. The fund er s had no r ole in s tudy design, da t a c olle ction and in t e rpr eta ti on, or the de cision t o sub mit the w ork f o r publi c ation. .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint 20

1. W orld Health Or g aniz a t ion. Glo bal T u berculosis R eport 2024 . h t tps://www . who.in t/public a t ions/m/it em/t op-findings-and-messag es -in-t he-2024 -r eport. 2. Gupt a, A., W ood, R., K aplan, R., Bekk er , L.-G. & La wn, S. D . T uber culosis inc idence r at es during 8 y ear s of f o llow-up of an an ti retr ovir al trea tmen t c oho rt in Sou th Af ric a: c omparison with r a t es in the c ommunity . PLoS One 7 , e3415 6 (2012). 3. Lawn, S. D ., My er , L., Bekk er , L.-G. & W ood, R. Bur den of tuber cu losis in an an tir e tr ovir al tr ea tment pr ogr amme in sub- Sahar an Afric a: impact on tr eatmen t ou t c ome s and implic a t ions f or tuber cu losis c o n tr o l. AIDS 20, 1 605–161 2 (2006). 4. St erling , T . R. et al. Risk F act or s f or T uber culosi s Aft e r Highly Activ e Antir e trovir al Ther apy Initia t ion in the Uni t ed St at es and Ca nada: Implic a tions f o r T uber culos is Sc r eening. The Journal of Inf ect ious Diseases 204 , 893–901 (2011). .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint 21 5. Vir dee, A. K. et al. Risk o f T uber culos i s Aft er Achieving HIV Vir ologic a l Suppr ession on An tir etr ov ir al Ther apy: A Danish Na ti onwide Pr ospectiv e Coho rt Study . Cli nical Inf ectio us Diseases 80, 854–86 3 (2025). 6. Bell, L. C. K. & Nour sadeghi, M. P a thog enesis of HIV-1 and My c obac t erium tuber culo sis c o- inf ect ion. Nat R ev Microbio l 16, 80–90 (2018). 7. Diedrich, C. R. et al. An t ir etr ov ir al tre a tmen t does no t pr e v en t e xtr apulmo nary tube r culos is during SIV /Mtb c oinf ecti on in mac aqu es. JCI Insight 11, (2026). 8. St anley , S., Liu, Q. & F ortune, S. M. M y c obac t er ium tuber culos is func tional g enetic diver sit y , alt ered drug sensit ivit y , and pr ec ision medicine. F ron t Cell I nf ect Microbi ol 12 , 1007 958 (2022). 9. Sw eeney , M. I., Carr anz a, C. E. & T obi n, D . M. Unde r s t anding Myc obacteriu m tuber culosis thr ough its g enomic di v er si ty and ev o lution. PLoS P athog 21, e101295 6 (20 25). 10. Goig , G. A. et al. E c ology , global div ersity and ev o luti onary mechanisms in t he My c obac t er ium tuber culos is comple x. Nat R ev Microbiol ht tps://doi.or g /10.1038 / s415 79- 025-0 1159 -w (20 25) doi:10.10 38/ s4 1 579-0 25- 0115 9- w . 11. W alk er , T . M. et al. The 2021 WHO cata l ogue of Myc obacterium tuber cu losi s c omple x mut a tions assoc iat ed with d rug r esistance: A g enotypic anal y sis. La ncet Microbe 3 , e265– e273 (2022). .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint 22 12. Genome-wide associa tion s tudie s of global My c obac t er ium tuber culo sis re sis t anc e t o 13 an timicr obia ls in 10,228 g enomes iden ti fy new r esi s t anc e mechanisms. PLo S Biol 20, e3001755 (20 22). 13. T r auner , A. et al. The within-hos t pop ula tion dynamics of Myc obac t er ium tuber culosi s v ary with tr e a tmen t e ffi c acy . Genome Biol 18, 71 (20 17). 14. Nimmo , C. et al. Dynamics of within-h os t Myc obac t erium tube r culos is diversity and het er or e sist ance dur ing tr eatmen t. EBioMedicine 55, 102747 (2020). 15. V ar g as, R. et al. In-hos t popula t ion dy namics of My c oba ct e rium tuber cu losi s c omple x during activ e d isease. Elif e 10, (2021). 16. F or d, C. B. et al. Use of whole g enome sequencing t o es t ima t e th e mut a ti o n rate o f My c obac t er ium tuber culos is during lat ent inf e ction. Na t Genet 43, 482–486 (2011). 17. Liu, Q. et al. My c obac t er ium tuber cul osis clin ic a l isol a tes c a rry mut ational signa tur es o f hos t immune en vir onmen ts. Sci Adv 6 , eab a4901 (202 0). 18. Lieberman, T . D . et al. Genomic div ers ity in au t ops y samples r eveals within- hos t dissemina tion of HIV-associa t ed M y cobact er ium tuber culos is. Nat Med 22, 1470–14 74 (2016). 19. Mor eno-Molina, M. et al. Genomic an aly ses o f Myc obac t erium tuberculosis fr om human lung r esecti ons r ev ea l a high fr equ en cy of pol y clona l inf ection s. Nat Comm un 12, 2716 (2021). .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint 23 20. Marin, M. G. et al. Complet e g enome sequence of a virul en t ba r c od ed Mycobact er ium tuber culo sis str . Er dman c ommonly used f or non-human prima t e inf e ction s tudies. Microbiol R esour An nounc e0123224 (2025) doi:10.112 8/ mr a.01 232 -24. 21. K r eutz er , D . A. & Essigmann, J . M. Oxi diz ed, deamina t ed c yt os ines are a so ur ce of C --> T tr ansitions in vi v o. Proc Natl Acad Sci U S A 95, 35 78–358 2 (1998). 22. K urthk oti, K. et al. A dis ti nct ph ysiolog ic al role o f MutY in mut a t ion pr e v en t i on in m y c obact er ia. Microbio logy (R ea ding ) 156 , 88–93 (20 10). 23. K apopoulou, A., Lew , J . M. & Cole, S. T . The My c oBr ow se r por t al: a c ompr eh ensiv e and manually annot at ed r es our ce f o r m y cobact er ial g enomes. T uberculosis (E di nb) 91, 8–13 (2011). 24. Szklar c z yk, D . et a l. The STRING da t abase in 2023: pr ot ein-pr otein assoc ia t i on netw orks and functiona l enri chmen t anal y ses f or any sequenced g enome of int er e s t. Nucl eic Acids R es 51, D638–D646 (202 3). 25. Culviner , P . H. et al. E v olu tion o f Myc o bact erium tube r culos is tr anscr iption regula tion is associat ed with incr eased tr ansmissio n and drug r esist ance. Pr epri n t at h t tps://doi.or g /10.1101/2 025. 05.01 . 651750 (20 25). 26. Martini, M. C. et al. Loss of RNase J l e ads t o multi-drug t ol er ance and accu mula tion of h ighly s tructu r ed mRNA fr agmen ts in My c ob act erium tube r culos is. PLoS P athog 18 , e10107 05 (2022). .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint 24 27. Culviner , P . H. et al. E v olu tion o f My c obac t er ium tuber culos is tr anscription r egula tion is assoc ia ted with in cr ease d tr ansmission and drug r esi s t ance. b i oRxiv 2025.05 .01.6 5175 0 (2025) doi:10.1 10 1/2025 .05.0 1.651 750. 28. Diedrich, C. R. et al. SIV and My cobac t erium tuberculosi s s yner gy within th e gr anuloma acceler at es th e r eac tiv ation p a tt ern o f lat ent tuberculosis. PLoS P ath og 16, e1008413 (2020). 29. Diedrich, C. R. et al. R eac tiv ation of l at ent tuberculosis in cynomolgus mac aques inf ec t ed with SIV is associat ed with early pe rip her al T cell deple tion and no t virus lo ad. PLoS One 5 , e9611 (2010). 30. Corr ea-Macedo , W . et al. Alv eolar macr ophag es fr om per sons living with HIV show impair ed epig enetic r esponse t o Myc obac t er iu m tuber culosis. J C lin In ves t 131 , (2021). 31. Martin, C. J . et al. Digit a lly Ba r c od ing My c obac t er ium tuber culos is R ev ea ls In Viv o Inf ec tion Dynamics in the Ma c aque Model of T uber culosi s. mBio 8 , (2017). 32. v an der Klug t, T ., v an den Big g elaar , R. H. G. A. & Saris, A. Hos t and bact eria l lipid met abolism during tuber culos is inf e ct ions: possibi lit ies t o s yner gise ho s t- a nd bact er ia- dir ec t ed ther ap ies. Crit R ev Microbiol 1–21 (2024) doi:10.10 80/1 04084 1X.2 024.237 0979 . 33. Cook, G. M., Har ds, K ., Vilchè z e, C., H artman, T . & Berney , M. Ener g etics of R espir a ti on and Oxida tiv e Phosphoryl a ti on in Myc oba ct eria. M icrobiol Spectr 2 , (2014). 34. Quadri, L. E. N. Bios yn thesis of m y c ob act eria l lip ids by poly k etide s y n thases and bey ond. Crit R ev Biochem Mol Biol 49, 179–211 (2014). .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint 25 35. Mulholland, C. V . et al. Pr opionat e prev en ts loss of the PDIM virul ence l ipi d in My c obac t er ium tuber culos is. Nat M icrobiol 9 , 1607–1618 (2024). 36. R ousseau, C. et al. Virulence at t en ua t ion of two Mas-lik e po lyk etide s yn tha se mut an ts of My c obac t er ium tuber culos is. Microbi ology (R ead ing ) 149 , 1837–18 47 (200 3). 37. Smith, C. M . et al. Hos t-pa thog en g en etic int er act ions unde rlie tuber cu losis susceptibi lity in g enetic al ly di v er se mice. Elif e 11, (2022). 38. Chen, S. Ultr af as t one-pass F A ST Q da t a pr epr ocess ing , quality c o n tr o l, and deduplica ti on using f as tp. Imet a 2 , e107 (2023). 39. Cos t ello , M. et al. D isc o v ery and charact eri z a t ion of a rtif ac tual mut a t ions i n deep c over ag e t ar g et ed c aptur e sequencing dat a du e t o o xidati v e DNA damag e during sample pr epar a tion. Nucleic Acids R es 41, e67 (201 3). FIGURE LEGENDS Figure 1. Whole genome sequencing identifies intrahost mutations. (A) Study design in which naïv e, SIV inf ected NHP s (with and without suppressive AR T init iat ed a t 3 d a y s pos t inf ec tion) w er e chal leng ed with a barc oded M t b libr ary . All bac t eria wer e isolat ed from inf ect ed tissues, g enomic DNA e xtr act ed f or who le g e nome sequencing and in tr ahost mut ations ide n ti fied. ( B) T op: crit er ia f or d e fin ing in vivo deri ved mut a tions, wher e e v er y r ow (Tissu es A-D) r epr esents sit es inf ec t ed by a barc oded M tb s tr ai n (c olumns BC1-4). Tissues c ont ainin g a g enomic mut a tion ar e c o lor ed gr een whi le whi t e indi c at es p r esence of wild type M t b. Bot tom: Mut a ti ons ar e gr ouped by their p r edi ct ed impact on p r ot e in coding. (C) T op: f or each animal (lines), the g enomic posit ions with de t ec t ed mut a t ions in an y tissue a r e plo t t ed (d ots). B ott om: the number s of mut a t ions (f r om all N HP s) obser v ed within 100 kb g enomic window s ar e plo t t ed (y ellow). Gr ey shad ing indica tes the 9 5% c onfidence le v el o f the number o f e xpect ed mut ations .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint 26 in a giv en 100 kb window when samp ling 116 r andom sit es in the Mtb g en ome acr oss 1000 simula tions. Figure 2. Mutational rates and profiles across SIV coinfection and ART . (A ) R ela ti v e mut a t ional r a t es w ere c alcu lat ed f or each animal per unit of inf ection time. (B) The t ot al ch r omosomal equiv alents (CE Qs) fr om all inf ect e d tissues in each NHP w as c ompar ed acr oss tr eatment gr oups. (C) The t ot al c ol on y f orming units (CFU) in each NHP a t necr ops y w as c ompar ed acr oss gr oups. (B and C) St a ti s ti c al t esting w as perf ormed using a K rusk al W allis t est with Dunn’ s c orrecti on. (D) The t o t al Mtb CFUs a t necr ops y f or origin s it es c o n t ain ing an o xida tive- (C>T /G>A and C>A/G>T) or non-o xida tiv e da mag e mut a tion (all o ther SNP s) w e r e c ompar ed us ing a Mann-Whitney U t es t. (E) The pe r cent of a ll mut ations obser v ed in a t r ea t men t gr oup a t t ribu t e d t o a dis tinc t type of g enomic mut a tio n. P -v alue w as de fined by a Fisher ’ s E x act t es t of o xida t ive mut a tions v e r sus non-o xida ti v e mut ations in the AR T gr oup c ompar ed t o t he TB only gr oup. (F) The per cent of all muta tions in each NHP (dot) separ at ed by type o f muta tion. ( G) Corr e la t ion betw een the number of o xid a ti v e damag e-associa t ed mut a t ions pe r NHP (normaliz ed by number of tissues sampled) with eac h animal’ s Mtb r ela ti v e mut a t ion r a te (one-t ailed P ear so n c orrela t ion). Figure 3 . In vivo variants uncover discrete dissemination events within barcoded Mtb dissemination networks . Color ed no des denote ti ssues shar ing the same Mtb s tr ain as defined fr om prior bar c ode sequ encing in r ef 7 . Rings indic a te Mtb mut a tions prese n t in tha t t issue, with each c ol or repr ese n ting a di ff er e n t g enomic mut a tion. Tissues with one bar c oded M tb str ain and a mixtur e of wi ldtype and mut ant alle les lik e ly ser v e as ‘ origin’ sit es, which ar e d enoted as bo x es. Inf err ed dissemina t ion ev e n t s fr om origin si t es a r e ind ica t ed by arr ow s (r ed arr ows indica t e origin s it e is a lung tissue dissemina ting t o o ther lung tissue s, while blue a rr ows indica t e origin si t e is a thor acic lymph node dissemina t ing t o lung tis sues). .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint 27 Figure 4. Intr ahost mutations are enriched in lipid metabolism genes and confer Mtb fitness . (A) Ov erlap o f g enes with in tr aho s t mut a tions (s ynon ymous and non-s yn on ymous) in NHP s (95 g enes) and tw o human da t asets (257 and 403 g enes f or Liu e t al. 17 and L i eberman et al. 18 ). (B) Fr action of i n tr ahos t mut a t ions in g enes fr om each Myc obr ow se r c ategory is p lott ed. The enrichmen t o f i n tr ahost mut a ti ons in each fun ctiona l ca t egory w as c a lculat ed ag ains t the number of g enes ann ot at ed f or eac h c at egory i n the H37R v g enome using a Fisher ’ s e x act t e s t. (C) The pr oportion o f mut an t r eads ver sus all r eads w as c ompar ed a t ea ch origin tissue. (D) The per cent of tissues con t a ining a bar c oded Mtb s tr ain tha t also c a rried a g enomic v arian t of in t e r est w as c ompar ed. (C-D) St a tistical analy ses w e r e perf ormed using an unpair ed t-t e s t. (E) A Fisher ’ s Ex act t es t w as used to c omp ar e the pr opo rtion o f C-t o-T tr ansitions f or v arian t a llel es tha t w e r e and w e r e not f ound acr oss 55,000 hu man clinic al str ain g enomes. Figure 5. Model of intrahost mutations in metabolic and cell envelope pr ocesses . NHP -deriv ed Mtb mut a tions are gr ouped int o fun ctional pr o cesses using li t er atur e cura tion and conser v e d domain similarity t o micr ob ial h omologs. Genes in bla ck r epresen t al l i ntr ahos t deri v ed M tb mut a tions, while g enes in r ed are mut a tions assigned t o t he ‘pr e-deri v ed’ mut a tion gr oup. Genes mark ed with a s t e ris k s repr es en t fr ameshift mut a t ions; g enes in par en thes is repr ese nt s ynon ymous mut a tions and the r e maining ar e pr edict ed missens e mut a ti ons. Functiona l c at egories ar e c ol or ed as f ollow s: lipid met abolism and bios yn thesis (r ed), cell en v e lope (y ellow), nucl eotid e s yn th esis/r ep air (or ang e), met aboli t e tr anspo rt (gr een ), pr ot e i n homeos t asis/tr ansl a tion (blue), r ed o x pr ocesses/r esp ir a ti on (gr ey), v ir ulence (pu rple). The schema tic dep icts lipas es breaking down c omple x host and ba ct e rial li pids i n t o f at ty a cids, which c an then be act iv at ed f or degr ada tion (or in c orpor at ed i n to other molecules ) by c onjug a tion t o CoA thr ough f a tty ac id CoA lig ases (F A CL). Dif f er ent acy l-CoA subs tr a t es a r e then degr aded thr ough it era ti v e c y cles of bet a-o xida tion, y ielding acet yl-CoA tha t c an be used to driv e the T CA cy cle and r esp ir a ti on as w e ll as be ing incorpor at ed int o Mtb sec on dar y met abolit es. F a tty acid lig a t ion t o AMP via f a t ty acid AMP lig ases (F AAL) c an be dir ectly i n co r p o rate d b y p ks (and oth er b iosyn thesi s g enes) int o lar g e r l ipid-based s tru ctur es, man y of which ar e c omponen ts of th e cell envelope. Bet a-o xida tion o f lipids also y i elds NADH, leading t o .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint 28 incr eased r educt iv e s tress that needs t o be r e-balanced. This c an o ccur th rough c on v er sation o f NADH t o NAD PH , which is utiliz ed f or sec ondary met aboli t e s yn th esi s as w ell as NADH c onsumption during r espira tion, whic h c an bec ome impair ed due t o ho s t p r essur es lik e h ypo xia, acidifi c ation and inflamma tion. This ma y lead t o leak ag e of electr ons from the r espir at or y chain, g ener a ting r eacti v e o xy g en species a nd macr omolecule dama g e in Mtb. I n the c ont e xt of AR T tr ea tment, w e h ypothes iz e an augmen t ed inf lamma t ory en vir onment not seen in Mtb only animals, which ma y yield gr ea t er M tb damag e and incr eased det ec tion of o xida ti v e-damag e associat ed mut a t ions. Imag e w as pr epar ed using BioR end er . T able 1. NHP intrahost mutations are also observ ed in human Mtb clinical strains . Iden tic a l g enomic mut a tions in the NHP da t aset w er e ma t ched t o mut a t ions obs er v ed in ov er 55,000 clini c al isol a t es 27 . Clini c al s tr ain muta tions and th e number o f independ en t mut ations ev e n ts w er e pr ed ic t ed by comparing each clini c al s tr ain ag ains t a M t b ances tr al g enome r ec on s truc tion. Nons yn = non-s ynon ymous, Syn = s ynon ymous, IG = in t er genic. Pk s7 mut a tion s with an ‘ A ’ and ‘B’ indic at e di f f er e n t mut a tions lead ing t o a fr ameshift in t he same c odon, while w e find two di f f e r ent type s of i nt er g e nic whiB6 mut a ti ons a t the same posi tion. Supplementary Figur e 1. (A) The number of g enomic v arian ts (par e n theses) de t ected by Mut ect2 per sequ enced sample (dot s) be f or e and after quali ty c o n tr o l (QC) s t eps. (B) The time of PET -CT det ection (e ither 4-, 8- or 1 2-w eek s pos t in f ecti on) of a lung gr anuloma tha t ser v ed as an origin si t e f or an in vi v o mut a t ion is c ompar ed across gr oups. St a ti s ti cal t est compar ed the dis tr ibution o f 8 and 12-w eek lesions using a K rusk al- W allis t es t with Dunn’ s c orr ec tion. (C) STRING analy ses showing the p r edic ted int er act ions bet w een a ll g enes in t he NHP da t aset. On ly in t er ac ting nodes are shown and ar e c olo r ed by the ir Myc obr ow se r ca t egory . .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint 29 Supplementary Figure 2. All genes with intrahost mutations (all mutations = synonymous and non-synonymous mutations) for each dataset was analyzed for InterPro protein domain enrichment on the STRING database. Supplementary Table 1. List of genomic accession numbers for all sequenced Mtb populations from infected tissues in this study. All samples are also available under the Bioproject accession: PRJNA1432747. Supplementary Table 2. Final table of high confidence Mutect2-detected genomic mutations occurring in vivo. Supplementary Data File 1. An R object that provides data frames containing the tissues sharing disseminated barcoded Mtb strains. The name of each list item matches the following format: “ Animal ID_QtagBarcodeSequence” . Each table (using only the Node1 and Node2 columns) can be used as an edgelist by the igraph R package to generate the network structures shown in Figure 3. However , additional metadata columns in each table—i.e., columns indicating the diff erent variants associated with this barcoded strain (column names indicate genomic position associated with a mutation) and a final column indicating which tissues ser ve as origins for each mutation—will need to be incorporated to fully reproduce Figure 3. In the metadata columns, ‘Node1’ and ‘Node2’ ref ers to the tissues in the same row . .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint A B High Confidence Variants (143) 31 53 19 11 2 Pre-existing? (27) In vivo defined (116) 6 17 8 2 Missense Synonymous Frameshift Intergenic Truncation Tissue A Tissue B Tissue C Tissue D BC2 BC3 BC4 OR Tissue A Tissue B Tissue C Tissue D BC1 N = 7 N = 4 N = 9 TBSIV + TBSIV + ART + TB SIV infection ART (3 dpi) Barcoded Mtb Infection (bronchoscope) Isolation of Mtb and bacterial DNA from infected tissues 1) Population-level whole genome sequencing 2) Alignment to Mtb Erdman genome and SNP / indel detection (Mutect2) 3) QC analyses and in vivo variant filtering 5 months 10-12 weeks Genomic Analyses C Mtb genomic position (x 106 nt) # of mutations / 100kb window TB SIV+TB SIV+ART+TB 0.0 5.0 ×10 5 1.0 ×10 6 1.5 ×10 6 2.0 ×10 6 2.5 ×10 6 3.0 ×10 6 3.5 ×10 6 4.0 ×10 6 4.5 ×10 6 0 2 4 6 8 50 100 150 200 Mtb genomic position (nt) # variants / 100Mb window Mutation vs NHP and genomic window 21818 22218 22318 23218 23518 24018 33519 33919 34319 22417 22717 22817 23017 30018 30118 33419 22117 22317 22517 22617 SIM 0.0 5.0 ×10 5 1.0 ×10 6 1.5 ×10 6 2.0 ×10 6 2.5 ×10 6 3.0 ×10 6 3.5 ×10 6 4.0 ×10 6 4.5 ×10 6 0 2 4 6 8 50 100 150 200 Mtb genomic position (nt) # variants / 100Mb window Mutation vs NHP and genomic window 21818 22218 22318 23218 23518 24018 33519 33919 34319 22417 22717 22817 23017 30018 30118 33419 22117 22317 22517 22617 SIM TB 22417 22717 22817 23017 30018 30118 33419 SIV + TB 22117 22317 22517 22617 SIV + ART + TB 21818 22218 22318 23218 23518 14018 33519 33919 34319 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint Non-oxOxidative 1 10 100 1000 10000 100000 Fig 2D: Total origin CFU by mutation type Total CFUs 0.0036 C>T_G>AC>T_G>AC>T_G>AC>A_G>TC>A_G>TC>A_G>TG>C_C>GG>C_C>GG>C_C>GT>C_A>GT>C_A>GT>C_A>GA>C_T>GA>C_T>GA>C_T>GA>T_T>AA>T_T>AA>T_T>ADeletionDeletionDeletionInsertionInsertionInsertion 0.0 0.2 0.4 0.6 0.8 1.0% of all mutations / NHP Fig 2E: Fig 2E: Avg %mutation_by NHP B C D TB SIV ART 0.0 5.0×10-10 1.0×10-9 1.5×10-9 Mutation / bp / generation Fig 2A: Fig 2A: MutationRate per time_hrs 0.0221 0.4613 A Mutations (x10-10) / bp / unit time 5 Mtb SIV + Mtb SIV + ART + Mtb 0 15 10 p = 0.46 *p = 0.02 Mutations (x10-10) / bp / generation # Oxidative mutations / tissue 105 150 0.1 0.2 0.3 0.4 0.5 0.0 5.0×10-10 1.0×10-9 1.5×10-9 0.0 0.1 0.2 0.3 0.4 0.5 Mutation Rate / unit time Oxidative mutations / tissue #ox_mut vs. Mutation rate TB-ox SIV-on ART-ox Pearson r r 95% confidence interval R squared P value P (one-tailed) P value summary Significant? (alpha = 0.05) Number of XY Pairs Mutation Rate / unit time vs. TB-ox 0.1334 -0.6889 to 0.8055 0.01779 0.3878 ns No 7 Mutation Rate / unit time vs. SIV-on -0.4642 -0.9856 to 0.8971 0.2155 0.2679 ns No 4 Mutation Rate / unit time vs. ART-ox 0.7048 0.07652 to 0.9325 0.4968 0.0170 * Yes 9 Mutation Rate / unit time vs. TB-CT 0.1091 -0.7016 to 0.7967 0.01190 0.4079 ns No 7 Mutation Rate / unit time vs. SIV-CT 0.1014 -0.9525 to 0.9681 0.01029 0.4493 ns No 4 Mutation Rate / unit time vs. ART-CT 0.7051 0.07709 to 0.9325 0.4972 0.0169 * Yes 9 TB: r = 0.13; p = 0.39 SIV+TB: r = -0.46; p = 0.27 SIV+ART+TB: r = 0.70; p = 0.02 F TB-CEQ SIV ART 106 107 Total CEQ/NHP Fig 2B: Total NHP outcomes (all lesions): CEQ 0.0945 0.3197 SIV + ART + Mtb Total CEQ / NHP SIV + MtbMtb 107 106 p = 0.32 p = 0.09 TB-CFU SIV ART 103 104 105 106 107 Total CFU/NHP Fig 2C: Total NHP outcomes (all lesions): CFU 0.0066 0.5502 SIV + Mtb SIV + ART + MtbMtb 105 103 104 107 106 Total CFU / NHP *p T_G>A C>A_G>T G>C_C>G T>C_A>G A>C_T>G A>T_T>A Deletion Insertion G Oxidative damage 100 0 20 40 60 80% of all mutations TB SIV + TB SIV + ART + TB C>T / G>AC>A / G>TG>C / C>GT>C / A>GA>C / T>GA>T / T>ADeletionInsertion E 104 102 103 105 10 Oxidative mutationsNon -oxidative mutations Total CFU at origin site p = 0.09 *p < 0.01 1 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint Synonymous Non-synonymous Intergenic 0.0 0.2 0.4 0.6 0.8 Fig 4C: %Prevalance at origin sites % Alternate Alleles at tissue 0.0381 In vivo SynIn vivo NonIn vivo IG 0.0 0.2 0.4 0.6 0.8 1.0% of tissues within network Figure 4D: Fig 4D: Dissemination: % of network 0.3458 A C Lieberman: 403 Liu: 257 NHP: 95 40 1 16 15 Number of genes with mutations cell wall and cell processes conserved hypotheticals information pathways intermediary metabolism and respiration lipid metabolism virulence, detoxification, adaptation regulatory proteins PE/PPEunknown stable RNAs insertion seqs and phages 0.0 0.1 0.2 0.3Fraction of genes H37Rv NHP_all QL_all TL_all Intermediary Metabolism & Respiration p < 0.001 Cell wall / Cell ProcessesConserved Hypotheticals Information Pathways Lipid Metabolism PE/PPE Virulence, detoxification & adaptation Regulatory Proteins Fraction of genes in dataset 0 0.1 0.2 0.3 p = 0.05 p < 0.05 NHP in vivo derived Lieberman et al Liu et al H37RV p < 0.05 p < 0.001 p < 0.05 p = 0.02 B D E Percent variant alleles at each origin tissue Synonymous Non -synonymousIntergenic 0 20 40 60 80 *p < 0.04 rv1925 (fadD31) rv1648 rv0232 Cell wall process Lipid metabolism Intermediatory metabolism Information system TRUEFALSE TRUEFALSE 0 20 40 60 80 100 Clinical strain by % type of mutation % of all mutations C>T_G>A C>A_G>T G>C_C>G T>C_A>G A>C_T>G A>T_T>A Deletion Insertion TB SIV ART In clinical strains? 100 80 % of all mutations 0 20 40 60 Yes *p = 0.04 No A>C / T>G A>T / T>A Deletion Insertion C>T / G>A C>A / G>T G>C / C>G T>C / A>G Synonymous Non -synonymousIntergenic % barcode network with variant 0 20 40 60 80 rv1648 rv0154c (fadE2) 100 p = 0.35 rv1384 rv3665 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint Lipases ppe28* lipC lipJ rv2075 rv2565 (rv1835) (rv3479) lipG (lucA) FACL fadD3 1 FAAL (fadD34) Lipid / metabolite synthesis pks7*, pks8, pks11, pks12 , rv0232, pta, entC, tesB1, grcC1, rv1520, ptbB (pks1), (pks2), (rv1524), (mbtD) ubiA, (rv1520) Mycolic Acids PDIM / TDM MPM Transport irtA kdpC ctpI rv3728 rv3273 rv0987* rv0943 dppA dppB (ceoB) pitB corA (ctpI) (rv2508c) Cell Envelope otsB1 rv1648 rv0225 rv3594 PE_PGRS36 PE_PGRS43 (rv2402) (glgE) (murE) (glfT2) (PE_PGRS9) PPE16 (rv1230) (PPE42) (aftA) Rv1433* PE_PGRS1 rv1648 Redox / Respiration pntAa fbiA rv1812c rv2295 rv0785 rv0839* narK3 tcrA* rv3698 moaA2 (narU) + LAM Sulfolipids Isonitrile Lipopetides (Cu) Mycobactins (Fe) • Rewired (lipid) metabolism? • Biosynthetic electron sinks? • Mtb virulence (lipid) synthesis? • Redox imbalance? • Metabolic toxicity? • Host immune damage? Nucleotide dinP* ruvB* uvrD1 parE2 rnj upp parA parE2 (rv3202) recA* nrdF2 ercc3 rv2559c dnaN rv2435c Protein / translation pepC*, folC*, rv1002, rv0075, metE, rpsA, (ilvD), (rv0484), (carB), (rv2426), (rv3228) rv1112 Virulence sapM IG_whiB6 mycP2? Beta- oxidation fadE2* echA11 (fadE31) (echA9) Other metabolism rv0939.hpaG (catechol) (rv1026 (ppGpp?)) oxcA (oxalate) .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint Gene DNA change Protein Change Mutation # Events # Strains rv0785 C>CG NonSyn Ala561fs 2 2 rv0987 A>G NonSyn Ser710fs 3 3 pks7 CG>C NonSyn Ala1174fs A: 5 B: 2 A: 21 B: 3 rv0075 G>A NonSyn Gly175Ser 1 1 rv0943 G>A NonSyn Ala54Val 1 1 rv1648 A>G NonSyn Ser26Gly 1 3 pks8 C>T NonSyn Thr1274Met 1 1 lipJ G>A NonSyn Ala390Val 1 4 rv2075 C>A NonSyn Val415Phe 1 1 rv2565 C>T NonSyn Ser292Phe 1 1 rnj T>C NonSyn Arg361Gly 1 10 entC C>T NonSyn Ala270Val 1 8 rv3594 G>A NonSyn Arg215His 2 3 fadD34 C>T Syn Gly143Gly 1 1 Rv1520 T>C Syn Arg81Arg 1 1 Rv1835c C>T Syn Ala326Ala 2 2 Rv2402 G>A Syn Ala324Ala 1 1 Rv2426c T>G Syn Ala256Ala 3 12 Rv3228 G>A Syn Leu102Leu 2 6 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint fadE31 C>T Syn Asp190Asp 2 4 aftA C>T Syn Arg5Arg 5 27 glfT2 G>A Syn Asn229Asn 1 3 PE-PGRS38- pbpB C>A IG PGRS38 (-20) 1 1 Rv3268- Rv3269 A>C IG rv3269 (-110) 1 2 whiB6- Rv3863 C>T IG whiB6 (-25) C>T: 4 C>G:2 C>T:4 C>G:10 Table 1. NHP intrahost mutations are also observed in human Mtb clinical strains . Identical genomic mutations in the NHP dataset were matched to mutations observed in over 55,000 clinical isolates 27 . Clinical strain mutations and the number of independent mutations events were predicted by comparing each clinical strain against a Mtb ancestral genome reconstruction. Nonsyn = non-synonymous, Syn = synonymous, IG = intergenic. Pks7 mutations with an ‘ A ’ and ‘B’ indicate different mutations leading to a frameshift in the same codon, while we find two diff erent types of intergenic whiB6 mutations at the same position. .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.03.714442doi: bioRxiv preprint