Mitochondrial genome maintenance—the kinetoplast story

Abstract Mitochondrial DNA replication is an essential process in most eukaryotes. Similar to the diversity in mitochondrial genome size and organization in the different eukaryotic supergroups, there is considerable diversity in the replication process of the mitochondrial DNA. In this review, we summarize the current knowledge of mitochondrial DNA replication and the associated factors in trypanosomes with a focus on Trypanosoma brucei, and provide a new model of minicircle replication for this protozoan parasite. The model assumes the mitochondrial DNA (kinetoplast DNA, kDNA) of T. brucei to be loosely diploid in nature and the replication of the genome to occur at two replication centers at the opposing ends of the kDNA disc (also known as antipodal sites, APS). The new model is consistent with the localization of most replication factors and in contrast to the current model, it does not require the assumption of an unknown sorting and transport complex moving freshly replicated DNA to the APS. In combination with the previously proposed sexual stages of the parasite in the insect vector, the new model provides a mechanism for maintenance of the mitochondrial genetic diversity.


T he mitoc hondrial DNA of trypanosomes and its unique structure
In 1913, Robertson described a structure close to the base of the fla gellum, whic h she named the kinetonucleus (Robertson 1913 ).In 1924, Bresslau and Scremin detected DNA within the kinetonucleus, which later was named the kinetoplast (Steinert et al. 1958 , Burton andDusanic 1968 ).In transmission electron microscopy, the kinetoplast of Trypanosoma brucei appears to be a disc-sha ped, electr on-dense structur e within the cell's single mitochondrion (Fig. 1 ).The biochemical isolation of kinetoplasts revealed that they consist of two types of DNA molecules, the maxiand the minicir cles.Minicir cles in T. brucei are nonsupercoiled, 1 kb DNA molecules that are organized in an interlocked network.In Crithidia fasciculata , an insect parasite related to T. brucei, each minicircle is linked to approximately three other minicircles (Chen et al. 1995 ).The maxicircles are 23 kb in T. brucei and about 30 of these ar e interwov en into the minicir cle netw ork (Sha pir o 1993 , Chen et al. 1995, Cooper et al. 2019 ).Studies in C. fasciculata suggest that the electron-dense kinetoplast DN A (kDN A) structure is formed by sequential condensation e v ents involving a number of kDNA associated, basic histone H1-like proteins (KAP; Xu et al. 1996, Yaffe et al. 2021 ).Ov er all, this r esults in a DNA structur e with an estimated size of about 10 7 kDa (Schneider 2001 ).In situ , the interlocked minicircles are stretched out and are oriented side by side, resulting in the typical striated network structure observed in tr ansmission electr on micr oscopy (Fig. 1 ).Prior to replication, the T. brucei kinetoplast disc is ∼450 nm in diameter and ∼150 nm in height (Jakob et al. 2016 ), the latter corresponding to a ppr ox-imately half the circumference of a minicircle (Shapiro and Englund 1995 ).
The minicircles contribute to over 90% to the mass of the network and each minicircle encodes three to five guide RNAs, which ar e r equir ed for maxicircle tr anscript editing (Hajduk and Oc hsenr eiter 2010 , Aphasizhe v and Aphasizhe v a 2011 ).A r ecent study in T. brucei has identified 391 distinct minicircle molecules that differ gr eatl y in av er a ge copy number (Cooper et al. 2019 ).Despite the sequence diversity, the minicircles contain a conserved region of around 100-200 bp including the origin of replication (Chen and Donelson 1980 ).The leading-strand synthesis starts at the universal minicircle sequence (UMS; 12 bp), which is located within the conserv ed r egion (Ntambi andEnglund 1985 , Birk enme yer et al. 1987 ).Also, the first Okazaki fr a gment is synthesized at an invariant hexamer site in the conserv ed r egion (Ray 1989 ).Furthermore, the minicircles also harbor a bent structur e, whic h is caused by m ultiple A-tr acts of 5 bp length that ar e positioned in phase with the helical repeat (Marini et al. 1982 ).The function of the helical bend r egion r emains unclear but it has been hypothesized that it might be involved in the organization of the minicircles within the network (Sha pir o and Englund 1995 , Jensen andEnglund 2012 ).
In T. brucei ∼30 identical maxicircles encode 18 proteins and two ribosomal RNAs (12S and 9S) but no tRNAs .The protein coding genes are mostly involved in the respiratory chain and include the ATP-synthase, cytoc hr ome oxidase, NADH dehydr ogenase subunits, and ribosomal components (ribosomal subunits uS12 and uS3m) as r e vie wed in (Fea gin 2000, Sc hneider 2001 ).In addition, four open reading frames of unknown function are en-Figur e 1. T he kinetoplast of T. brucei .(A) Transmission electron micr oscopy ima ge of a thin section thr ough the basal body and kDNA network of T. brucei .In this orientation, the kDNA disc is cut orthogonal to its surface.BB, basal body; F, flagellum; TAC, tripartite attachment complex; and kDNA, mitochondrial genome.coded.A total of 12 of the protein-encoding genes are cryptogenes and they r equir e post-tr anscriptional modification thr ough insertion/deletion RNA editing in order to be translatable.Specificity of the RNA editing process comes from the minicircle-encoded guide RNAs, that serve as templates for proper insertion and deletion (Hajduk andOc hsenr eiter 2010 , Aphasizhe v a et al. 2020 ).
Electr on micr oscopy studies in combination with ethanolic phosphotungstic acid (E-PTA) have shown an asymmetric distribution of basic pr oteins ar ound the kDNA disc with a strong enrichment at the two opposing regions that are known as the antipodal sites (APS; Fig. 2 A; Melendy et al. 1988, Ferguson et al. 1992, Gluenz et al. 2007 ).It has been speculated that the very basic histone-like kDNA associated protein 4 (TbKAP4), TbKAP6, and the pol ymer ase β-PAK could contribute to the E-PTA staining (Xu et al. 1996, Saxo wsk y et al. 2003, Gluenz et al. 2007, Wang et al. 2014 ).While in vitro experiments with isolated kDN A netw orks demonstrate that the histone-like basic proteins are able to compact or change the network structure, the precise localization of these proteins has not been shown (Xu et al. 1996, Wang et al. 2014 ).The APS are thought to contain free minicircle replication intermediates and 14 of the 19 c har acterized minicircle r eplication factors (Table 1 ; Melendy et al. 1988, Ferguson et al. 1992, Johnson and Englund 1998 ).Furthermore, the heterogeneous E-PTA staining within the APS suggests the presence of subdomains with differ ent pr otein composition, whic h is supported by the differential localization of the mitoc hondrial topoisomer ase II and ligase k β in the APS (Downey et al. 2005, Gluenz et al. 2007 ).Although the APS have long been described, their composition, dimensions and dynamics remain poorly understood.In longitudinal kDNA sections, a thin rim of E-PTA staining further indicates the presence of basic proteins along the edge of the network disc (Gluenz et al. 2007 ).Such a confinement of the kDNA has been modeled as a potential driving force for minicircle network formation (Diao et al. 2012 ).Ho w e v er, similar to the APS, the identities of the basic pr oteins ar ound the disc r emain unknown.
Kinetoplast biology, including replication and segregation of the kDNA, has been studied in a number of different Kinetoplastea, including C. fasciculata , Leishmania tarentolae , Trypanosoma cruzi , T .evansi , T .equiperdum , T .gambiense , T .lewisi , T .mega , and more recentl y, pr edominantl y in T. brucei .In this r e vie w, we mostl y focus on T. brucei .

Replication of the kDNA
Replication of the kDNA begins prior to the nuclear S phase and leads to a gr adual incr ease in size of the two kDNA lobes e v entually forming two new kDNA discs (Fig. 2 ; Hoeijmakers andWeijers 1980 , Guilbride andEnglund 1998 ).In the current model, the minicircles are released from the network into the region between the kDNA disc and the mitochondrial membrane near the flagellar basal body (BB) called the kinetoflagellar zone (KFZ; Drew and Englund 2001 ).Here, they are replicated unidirectionally via theta intermediates (Englund 1979, Melendy et al. 1988, Sheline et al. 1989, Ryan and Englund 1989a,b , Abu-Elneel et al. 2001, Dr e w and Englund 2001 ).The two resulting daughter minicircles, which contain nicks and gaps from replication, are subsequently reattached to the network periphery at the APS in T. brucei .Only after r eattac hment to the network, the last remaining nicks and gaps ar e r epair ed, allowing the r eplication mac hinery to distinguish between replicated (nicked, gapped) and nonreplicated (covalently closed) minicircles.As gener all y accepted by the field, this mec hanism is thought to ensure that each minicircle is replicated only once during each generation (Englund 1979, Pérez-Morga and Englund 1993a,b , Guilbride and Englund 1998, Liu et al. 2009a ).The position where replicated minicircles are reattached to the kinetoplast netw ork w as sho wn to differ in T. brucei compared to other Kinetoplastea.While r eattac hment occurs at opposite or APS of the kinetoplast in T. brucei , the replicated circles are reattached to the network on the whole kinetoplast circumference in C. fasciculata , L. tarantolae , or T. cruzi (Guilbride and Englund 1998 ).Maxicircles, similar to the minicircles, replicate unidirectionally via theta intermediates.Ho w e v er, unlik e the minicircles, the y remain attached to the network during replication (Carpenter and Englund 1995 ).After replication the daughter networks are segregated, but initiall y r emain connected by the "nabelsc hn ur" that lik ely consists of maxicircles (Fig. 2 D; Gluenz et al. 2011 ).The segregation of the kDNA is mediated by the physical connection of the kDNA to the basal bodies via the tr ansmembr ane spanning tripartite attachment complex (TAC; Ogbadoyi et al. 2003 ).The TAC and its components were recently reviewed and are not in the focus of this manuscript (Schneider and Ochsenreiter 2018 ).

Minicircle replication-the known factors
In the following c ha pters, we discuss the properties of the proteins known to be involved in minicircle replication with a focus on components from T. brucei .We start with the release of minicircles into the KFZ, then move on to the replication and reattachment process.We have summarized the localization of each minicircle replication factor in Table 1 .We , furthermore , pro vide a supplementary table that summarizes the current knowledge on all kDNA r eplication and segr egation pr oteins including their localization and knockdown or overexpression phenotypes (Table S1, Supporting Information).

Release of the minicircles for replication
Uni v ersal minicircle binding proteins 1/2, UMSBP1, and UMSBP2 (Tb927.10.6070,Tb927.10.6060) The conserv ed r egion of the minicircles contains the origin of replication and the UMS-the binding site of the UMS-binding protein (UMSBP) in T. brucei and C. fasciculata .In C. fasciculata , the protein localizes in the KFZ and is likely involved in replication initiation (Tzfati et al. 1992, Avrahami et al. 1995, Abu-Elneel et al. 1999, Milman et al. 2007 ).In T. brucei , two UMSBP orthologues containing zinc-finger domains have been identified (Milman et al. 2007, Klebanov-Akop y an et al. 2018 ).UMSBP2 localizes at the telomeres in the nucleus and is essential for their structure and function (Klebanov-Akop y an et al. 2018 ).Although it has not been directly demonstrated, UMSBP1 is likely the ortholog with functions related to kDNA replication.The simultaneous depletion of both UMSBPs in T. brucei results in nuclear mitosis defects, inhibition of minicircle replication initiation, and inhibition of BB segregation (Milman et al. 2007 ).This is well in line with the two separate localizations of the orthologues.It has been suggested that the mitochondrial UMSBP functions as an origin-binding pr otein, whic h triggers the initiation of replication through the recruitment of other replication factors (Milman et al. 2007 ).Furthermore, interaction of UMSBP with histone H1-like proteins and the ability of that protein complex to decondense kDN A netw orks have been sho wn.These decondensed netw orks then become accessible for topological decatenation by topoisomerase II, resulting in the release of minicircle monomers (Kapeller et al. 2011 ).Furthermore, UMSBP binding to the UMS induces conformational changes in the Mitochondrial topoisomerase II, TOP2 (Tb927.9.5590) Based on the functional r equir ements, the most likely candidate for the release of minicircles from the kDN A netw ork is TOP2.
Ho w e v er, the depletion of TOP2 by RNAi does not lead to early defects in the release of minicircles and its localization is mostly at the APS, which is not compatible with the release of minicircles into the KFZ (Wang et al. 2000, Wang and Englund 2001, Downey et al. 2005, Kulikowicz and Shapiro 2006 ).The second function of TOP2 is the r eattac hment of minicircles to the network postreplication, which is supported by the functional studies that showed str ong incr ease of nic ked and ga pped fr ee minicir cles upon RN Ai depletion of the enzyme.Also, the localization of TOP2 to the APS supports its role in reattachment of minicircles (Wang and Englund 2001 ).Furthermore , T OP2 seems to be essential for the maintenance of the network structure, and therefore, might also be responsible for remodeling of the kDN A netw ork during and after replication (Lindsay et al. 2008 ).

Initiation of minicircle replication, synthesis
As described abo ve , the minicircles replicate via theta intermediates.While the light strand synthesis is continuous and starts at the UMS (Englund et al. 1982, Ray 1989, Ferguson et al. 1992 ), heavy strand synthesis is discontinuous (Kitchin et al. 1984 , Birk enme yer andRay 1986 ).
Mitochondrial RNA-binding protein 38, RBP38 (Tb927.8.2740) RBP38 contains a domain homologous to the antirestriction protein ArdC single-stranded DNA-binding domain, i.e. found in bacterial mobile elements (Krishnan et al. 2018 ).RBP38 binds the template strands of the UMS and the hexamer, and thus could be the minicircle replication origin recognition protein (Liu et al. 2006 ).
How this potential function relates to UMSBP1 (that also binds to the UMS), remains to be investigated.One hypothesis is that UMSBP1 is r equir ed to r ecruit RBP38 to the UMS, whic h in turn allows binding of a helicase and a topoisomerase to unwind the DNA helix.Such an unwinding process could start at the origin of replication and proceed along the whole minicircle (Shlomai 2004, Liu et al. 2006 ).Sur prisingl y, RBP38 localizes at the APS and not in the KFZ, where initiation of replication is thought to take place (Liu et al. 2006 ).
Mitochondrial helicases PIF1 and PIF5 (Tb927.11.6890,Tb927.8.3560) Nomenclature of PIF helicases originates from their discovery in the yeast petite integration frequency locus (PIF; Boulé and Zakian 2006 ).Trypanosoma brucei encodes eight helicases related to the PIF1 helicase of yeast.A total of six of them are localized in the mitochondrion (PIF1, 2, 4, 5, 7, and 8) while PIF6 is found in the nucleus and PIF3 localizes throughout the cytoplasm (Liu et al. 2009a ).A total of three of the PIF proteins are essential (PIF1, PIF2, and PIF8), and one (PIF5) shows a growth defect upon overexpression (Liu et al. 2009a ,b ).PIF1 and PIF5 function in minicircle replication.PIF1 localizes at the APS and its depletion leads to kDNA loss and accumulation of multiply interlocked, covalently closed minicircle dimers (termed fraction U), probably derived fr om late r eplication intermediates (Sundin and Varshavsky 1981, Liu et al. 2009b, 2010 ).The a ppear ance of a fraction U type minicir cle species w as pr e viousl y observ ed after depletion of TOP2 and POLIB b y RN Ai (Bruhn et al. 2010 , Jensen andEnglund 2012 ).PIF5, similar to PIF1, is localized to the APS, ho w e v er, its depletion by RNAi does not show any detectable phenotype.Overexpression of PIF5, on the other hand, leads to reduced cell growth and a moderate loss of kDNA.Furthermore, PIF5 was shown to function in the process of primer removal in the minicircle Okazaki fragments (Liu et al. 2009b ).
Primase 2, PRI2 (Tb927.1.4010) The T. brucei genome encodes two mitochondrial DNA primases named PRI1 and PRI2 (Hines andRay 2010 , 2011 ).Both enzymes have primase activity and are related to primases of eukaryotic viruses.PRI2 is a 129 kDa, very basic protein, i.e. conserved among the Kinetoplastea, except for its long N-terminal extension.RNAibased depletion of the enzyme leads to a loss of kDNA with maxiand minicircles being equally affected.Interestingly, RNAi also leads to an accumulation of covalently closed minicircles, and ther efor e, PRI2 seems to be dir ectl y involv ed in the initiation of minicircle replication.Whether the enzyme primes one or both strands at the replication origin, remains unknown (Hines and Ray 2011 ).Similar to the majority of the replication factors, PRI2 localizes at the APS, while initiation of replication is thought to take place in the KFZ (Hines and Ray 2011 ).

Pol ymer ases
In many of the well-studied eukaryotic model systems, the viral Pol γ is the re plicati ve enzyme for the mitochondrial genome.A P ol γ homologue , ho w e v er, is not encoded in the genome of T. brucei (Jensen and Englund 2012 ).Instead, the parasite uses se v en DNA pol ymer ases POLIA, IB, IC, ID, P ol κ, P ol β, and β-PAK to replicate and repair the mitochondrial genome.
Pol ymer ase IB, POLIB (Tb927.11.4690) POLIB localizes to two sites in the KFZ, similar to UMSBP.Depletion of POLIB by RNAi leads to a decrease of nicked and gapped free minicir cles, kDN A netw ork shrinkage and eventually kDNA loss.This phenotype suggests a function in minicircle replication.The enzyme potentially functions as part of a heterodimeric replicase synthesizing leading as well as lagging strand of the minicircles (Klingbeil et al. 2002, Bruhn et al. 2010 ).
Pol ymer ase IC, POLIC (Tb927.7.3990) In nonreplicating and early replicating kinetoplasts, POLIC localizes in the region of the KFZ, whereas in kinetoplasts that hav e been r eplicated, POLIC is pr edominantl y found at the APS (Concepción-Ace v edo et al. 2018 ).The differential localization is regulated by the N-terminal region of the protein (Miller et al. 2020 ).RNAi-based depletion of POLIC leads to network shrinkage with a faster loss of maxicircles than minicircles.Also, no loss of minicircle replication intermediates was observed, suggesting a possible role of POLIC in maxicircle replication (Klingbeil et al. 2002 ).Additionally, POLIC seems to also have a DNA polymerization independent role in the distribution of pr ogen y kDNA networks (Miller et al. 2020 ).
Pol ymer ase ID, POLID (Tb927.11.3260) POLID is gener all y distributed throughout the mitochondrion.In kinetoplast S phase, the protein is enriched at the kinetoplast disc and also at the APS (Klingbeil et al. 2002, Concepción-Ace v edo et al. 2012 ).RNAi-based depletion of POLID leads to a r a pid decline of mini-and maxicircles, e v entuall y leading to kDNA loss.Furthermore, POLID depletion causes a transient accumulation of covalently closed as well as nicked and gapped minicircle replication intermediates just prior to kDNA loss .T his beha vior suggests a role in minicircle replication (Chandler et al. 2008 ).Ho w ever, the decline of maxicircles during POLID RNAi is more rapid and complete than that of the minicircles, ther efor e, alr eady Jensen and Englund suggested that the effect on minicircle replication might be indirect and that POLID is required for maxicircle replication (Chandler et al. 2008 ).
Protein 93, p93 (Tb927.3.1180) This 93 kDa, basic pr otein was initiall y identified in a screen for mitoc hondriall y tar geted pr oteins with differ ential, S phasespecific expression (Li et al. 2007 ).It localizes to the APS during the S phase of the cell cycle.Depletion of p93 by RNAi causes an early loss of nicked and gapped minicircle replication intermediates, suggesting its involvement in minicircle replication (Li et al. 2007 ).The protein is conserved in the Kinetoplastea, but its precise role in the replication process remains elusive.
Mitochondrial topoisomerase IA, TOPIAmt (Tb927.10.1900) The mitochondrial TOPIAmt of T. brucei is related to bacterial topoisomerases IA and reverse gyrases that can relax negatively super coiled DN A (Scocca and Sha pir o 2008 ).The pr otein displays a dynamic localization at the kDNA disc, mostly at the APS but also in the KFZ.It seems to play a role in the resolution of late theta structures (Scocca and Shapiro 2008 ).

DN A r epair
Structure-specific endonuclease-1, SSE-1 (Tb927.10.340) Structure-specific endonuclease-1 (SSE-1) was originally purified fr om C. f asciculata and was thought to be inv olved in minicir cle primer r emov al (Engel andRay 1998 , 1999 ).Mor e r ecentl y, SSE-1 was also studied in T. brucei (Liu et al. 2005 ).It localizes to the APS as it does in C. fasciculata (Engel andRay 1999 , Liu et al. 2005 ).SSE-1 RNAi depletion in T. brucei leads to an increase in nicked and ga pped fr ee minicir cles and a delay in netw ork segregation.This is likely caused by a defect or delay in gap repair, and thus supports its function in primer r emov al (Liu et al. 2005 ).Mitochondrial DNA ligase homologs, LIG K-alpha and LIG K-beta (Tb927.7.610,Tb927.7.600)

Pol ymer ase beta
Final gap repair and covalent closure of nicks of newly replicated minicircles occur after r eattac hment to the kDN A netw ork and after all minicircles have been replicated.Trypanosoma brucei encodes two mitochondrial ligase genes.LIG k α and LIG k β are highly div er gent fr om other eukaryotic ligases (Downey et al. 2005 ).LIG k β localizes at the APS and is thought to repair most of the gaps together with Pol β, before the ne wl y r eplicated minicircles ar e r eattached to the network.Interestingly, it does not seem to colocalize with TOP2, suggesting that distinct replication centers might exist.LIG k α localizes throughout the kDNA and is thought to repair the final gaps at the end of kDNA replication together with Pol β-PAK (Downey et al. 2005 , Jensen andEnglund 2012 ).Depletion of LIG k α leads to decrease in kDNA size follo w ed b y asymmetric division of the network and the associated complete loss of the kDNA.
Pol ymer ase kappa, Pol κ (Tb927.11.8530) T his P ol κ related enzyme was first discov er ed in T. cruzi and is conserved in most Kinetoplastea.In T. cruzi , Pol κ localizes at the APS and at the BB proximal phase of the kDNA disc in the mitoc hondrion, whic h is very unusual when compared to orthologues from species outside the Kinetoplastea.In vitro studies show that Pol κ efficiently bypasses 8-oxoguanine lesions, supporting its potential role in DNA repair (Rajão et al. 2009 ).Furthermore in vivo ov er expr ession of the enzyme render the parasite less sensitive to gamma radiation suggesting a potential role of Pol κ in homologous recombination (Rajão et al. 2009 ).

Pol ymer ase IA, POLIA (Tb927.4.2950)
POLIA is distributed throughout the mitochondrion and the depletion of POLIA does not lead to any detectable phenotype different from the wild-type cells.Together with phylogenetic relationship of its POLA domain to pol ymer ase theta enzymes, which are mostl y involv ed in DNA r epair, a r ole in DNA r epair r ather than replication seems possible (Klingbeil et al. 2002 ).

Rea ttac hment of replicated minicircles at the APS
In most organisms, the RNA primers for the Okazaki fragments ar e immediatel y r emov ed after r eplication.The curr ent kDNA replication model suggests that after replication, the minicircles ar e mov ed fr om the KFZ to the APS and onl y ther e most primers ar e r emov ed and the ga ps between the Okazaki fr a gments ar e r e-paired (Ryan and Englund 1989a ,b ).Interestingly, at least one gap remains in the minicircles until after they are reattached to the growing disc.This last gap is repaired just prior to kDNA segregation and thus serves as a signal for the completion of replication.

Minicircle replication factor 172, MiRF172 (Tb927.3.2050)
MiRF172 is a large, basic protein that localizes to the APS (Amodeo et al. 2018 ).RNAi-based depletion of MiRF172 leads to an initial increase in nicked, gapped minicircles that are not reattached to the network (Amodeo et al. 2018 ).Ov er all, this leads to a decrease in network size , i.e .seen in mini-and maxicircle abundance.While the precise function remains unclear, MiRF172 is possibly involved in the r eattac hment of r eplicated minicircles to the growing network.Inter estingl y, the pr otein was initiall y discov er ed in a scr een for novel components of the kDNA segr egation mac hinery (the TAC) but was then found to r equir e both the kDNA itself and upstream components of the TAC for its localization and thus might provide a link between the two processes (Amodeo et al. 2018 ).

Uni v ersal minicircle binding protein UMSBP1
UMSBP1 not only binds to the UMS, but also to the hexamer (the sequence at the start of the first Okazaki fr a gment; Abu-Elneel et al. 1999 ).While binding the template strand of the UMS, UMSBP also binds the complementary strand of the hexamer.The gap flanking the first Okazaki fr a gment is one of the last minicircle gaps being repaired after re plication (Birk enme yer et al. 1987 , Ryan and Englund 1989a ,b ).This gap starts at the hexamer sequence and ends at the UMS.As suggested by Jensen and Englund, it is possible that UMSBP binds to the 5 terminus of the Okazaki fr a gment r egion and pr otects it fr om pr ematur e r epair (Jensen and Englund 2012 ).

Maxicircle replication-the known factors
Little is known about the replication of maxicircles.Similar to the minicircles, the y re plicate unidir ectionall y via theta intermediates but remain attached to the kDNA disc at all times (Carpenter and Englund 1995 ).They form homo-and heterocatenates with each other and the minicircles , respectively.J ust prior and at the beginning of the kDNA segr egation pr ocess, at least some of the maxicircles ar e concentr ated betw een the tw o separating discs and it is thought that they are eventually resolved through TOP2 activity.
Mitochondrial primase 1, PRI1 (Tb927.8.2550) RNAi against PRI1 causes loss of kDNA.The protein was localized pr edominantl y to the APS, but a weak signal was also observed in the KFZ (Hines and Ray 2010 ).Although, the localization at the APS is not indicative of the involvement in maxicircle replication, it is thought that PRI1 is a maxicircle-specific replication protein.This is based on the observation that RNAi targeting PRI1 causes maxicircles to be depleted faster than minicircles .Furthermore , the depletion of PRI1 does not lead to loss of minicircle replication intermediates (Hines and Ray 2010 ).

Mitochondrial helicase PIF2 (Tb927.11.6900)
PIF2 is a 115 kDa, basic protein localized to the KFZ in T. brucei (Table 1 ).The protein has ATP and Mg 2 + dependent helicase activity and is one of the few known proteins to be clearly involved in maxicir cle replication.RN Ai targeting PIF2 leads to maxicircle loss , while o v er expr ession of the protein causes a 3-to a 6-fold increase in maxicircle abundance (Liu et al. 2009a ).Abundance of the protein seems to be directly or indirectly regulated by the HslVU pr otease (see below).Inter estingl y, despite the r egulation by the HslVU protease, PIF2 abundance seems not to change during the cell cycle (Crozier et al. 2018 ).
Leucyl aminopeptidase metalloprotease 1, LAP1 (Tb927.8.3060) LAP1 is an M17 family leucyl aminopeptidase metalloprotease (LAP) with a basic pI of 9.8.The protein shows a kinetoplast S phase-specific localization and it colocalizes with maxicircles at the nabelschnur during kDNA division, while it is also present at the APS.RNAi depletion of the peptidase leads to an increase of cells with two nuclei and two kinetoplasts that are arrested prior to cytokinesis.Ov er expr ession on the other hand, r esults in loss of kDNA.While these experiments suggest a function of LAP1 in the kDNA segregation process, the details remain unknown (Peña-Diaz et al. 2017 ).

Other kDNA-associated proteins
Kinetoplast-associated protein 6, TbKAP6 (Tb927.10.8890) TbKAP6 is an HMG-box containing pr otein, whic h is homologous to CfKAP4 in C. fasciculata and TcKAP6 in T. cruzi (Xu et al. 1996, Cavalcanti et al. 2009, Wang et al. 2014 ).It localizes throughout the kDNA disc during the entire cell cycle.Functional studies using RNAi targeting TbKAP6 show a decrease in minicircle release from the network prior to replication, eventually leading to loss of mini-and maxicircles and to partially disorganized kDNA networks (Wang et al. 2014 ).Ov er expr ession on the other hand, leads to an increase in minicircle release, supporting the idea that the protein's function might involve the release process prior to replication initiation (Wang et al. 2014 ).Recombinant versions of TbKAP6 were shown to promote topoisomerase II-driven release of minicircles from isolated C. fasciculata networks in vitro (Wang et al. 2014 ).Based on the experiments in T .brucei , T .cruzi and C. fasciculata it is likely that the function of TbKAP6 is involving the minicircle release process prior to replication.

Tb927.2.6100
The highly basic protein (pI 10.7) encoded by the gene Tb927.2.6100 was identified by mass spectrometry of protein extr acts fr om isolated kDNA.The pr otein seems to be r estricted to Trypanosoma species and is not found, e.g. in Leishmania and Crithidia .It localizes at the kinetoplast and tandem affinity purification using a tagged version of Tb927.2.6100 identified mostly proteins of the mitochondrial ribosome and its assembly factors.Depletion of the protein by RNAi impairs growth, leading to small kDN As, kDN A loss and decrease in guide RNAs and maxicircle transcripts.Ho w ever, functional details of this protein remain unknown (Beck et al. 2013 ).

Regulation and control of the kDNA replication
The T. brucei cell cycle can be divided into three to some extent independent subcycles (nuclear, kinetoplast, and cytoskeletal subcycle; Wheeler et al. 2019 ).Howe v er, the kinetoplast S phase is somehow timed in coordination with the nuclear S phase (Woodw ar d and Gull 1990 ). Synthesis of kDNA initiates just before the start of the nuclear S phase and the replicated kDN A netw ork divides just before mitosis.We do not know m uc h about possible r egulation mec hanisms in the kinetoplast subcycle .Nonetheless , some discoveries about possible mechanisms regulating kDNA r eplication wer e made and ar e described in the following text.
Mitochondrial helicase homolog PIF8 (Tb927.7.1000) From the six mitochondrial helicases known, PIF8 is the smallest and most div er gent among them.It likel y does not hav e a helicase activity (Liu et al. 2009a, Wang et al. 2012 ).PIF8 mainly localizes at the phase of the kinetoplast, i.e. distal to the BB, but the localization pattern varies with different kDNA replication stages.Depletion of PIF8 led to moderate kDNA loss and only minor effects on cov alentl y closed minicircles, while nicked and gapped minicircle r eplication intermediates decr eased to ar ound 50% compar ed to noninduced cells .Furthermore , depletion of PIF8 b y RN Ai leads to disorganization of the kDNA structure (Wang et al. 2012 ).Altogether this could suggests that PIF8 has a function in organizing parts of the r eplication mac hinery during the replication process.
HslVU protease complex (Tb927.11.10240,Tb927.5.1520, and Tb927.11.12230) HsIVU is a bacterial-like ATP-dependent protease complex consisting of thr ee pr oteins in T. brucei ; o ne HslV homologue (Tb927.11.10240) and two HslU homologues (Tb927.5.1520,TbH-slU1 and Tb927.11.12230, TbHslU2;Li et al. 2008 ).In T. brucei , the thr ee pr oteins ar e distributed thr oughout the mitoc hondrion with an increased concentration at the kDNA (Li et al. 2008 ).The protease complex is essential for T. brucei and depletion leads to a 20fold increase in minicircles and a 3-fold increase in maxicircles.Additionall y, the cov alentl y closed, as well as nicked and gapped minicircle replication intermediates increase by 5-to 6-fold upon depletion of the enzyme complex.Furthermore, Hs1VU-depleted cells generate very large kDN A netw orks with a distorted structure.It has been suggested that HslVU is involved in controlling kDNA synthesis through the degradation of a positive regulator of DNA replication.Upon HslVU RNAi, this positive regulator may be stabilized, leading to kDNA ov er-r eplication and a significant increase in kDNA mass (Li et al. 2008 ).A target candidate for the trypanosome HslVU is potentially PIF2, as increased levels of PIF2 are detected upon HslVU knockdown (Li et al. 2008 ).Other targets of the T. brucei HslVU are currently unknown.
Puf nine target 1, PNT1 (Tb927.11.6550) PNT1 (Puf nine target 1) is a C11 cysteine peptidase , i.e .r equir ed exclusiv el y for the maintenance of the kinetoplast (Gr e wal et al. 2016 ).RNAi targeting PNT1 leads to a loss of kDNA, while its overexpression causes the formation of extra kDNAs (called ancillary kDNAs), that are not connected to the basal bodies.PNT1 is localized at the opposing ends of the two growing kinetoplasts in the area of the APS.It is not clear how and whether PNT1 regulates kDN A replication, but it w as sho wn that its activity is essential for par asite surviv al (Gr e wal et al. 2016 ).

UMSBP-redox-regulated binding
Redox pathwa ys ha ve been shown to contr ol bioc hemical pr ocesses such as transcription.Redox regulation was also observed for UMSBP in C. fasciculata and is the first example of a redoxregulated DNA synthesis protein (Onn et al. 2004, Shlomai 2010 ).
The active form of UMSBP is full y r educed, mor e pr ecisel y, the cysteine residues of the zinc finger domains were observed to be reduced.Oxidation of the -SH to S-S renders the protein inactive, not binding to the UMS because zinc is no longer bound after oxidation (Onn et al. 2004 ).Further, NADPH stimulates the reduction of UMSBP and as a consequence leads to increased binding to the UMS (Sela et al. 2008, Shlomai 2010 ).The active and inactive form of UMSBP fluctuate in a cell-cycle-dependent manner in C. fasciculata .A total of two peaks of activity were observed during the kinetoplast S phase, and ther efor e, it was suggested that this might be a mechanism for kDNA replication regulation (Sela et al. 2008 , Sela andShlomai 2009 ).

Division and segregation of the kDNA
The last step of the kinetoplast S phase is the division of the double-sized network.The kDNA is segregated by movements of the basal bodies that are attached to the kinetoplast through the TAC (Robinson andGull 1991 , Ogbadoyi et al. 2003 ).Simultaneously with the replication of the kDNA, the new flagellum starts to form.In the first step, the pro-BB's proximal face tilts to w ar ds the mitoc hondrial membr anes.Subsequentl y, the TAC is formed from the BB to w ar d the kDN A (Fig. 2 B and C).While the new BB is already attached to the kDNA, it rotates around the old BB, which is thought to help flagellar pocket biogenesis (Fig. 2 C).It also possibly contributes to the bilobed shape of the replicated kinetoplasts during late S phase and beginning of segregation.After replication, the kinetoplasts remain connected to each other b y the maxicir cles, which become visible as nabelschnur when the distance between the segregating networks increases (Fig. 2 D).
With increasing distance between the daughter networks, the nabelschnur becomes longer and thinner and r eac hes at least 1 μm (Gluenz et al. 2007(Gluenz et al. , 2011 ) ).The last step of kDNA division is the unlinking of the maxicir cles b y cleav a ge of the nabelsc hnur.As suggested by Jensen and Englund, this pr esumabl y is performed by TOP2 with some involvement of LAP1 (Jensen andEnglund 2012 , Peña-Diaz et al. 2017 ).After the division of the network, all the r emaining nic ks and ga ps ar e r epair ed, r esulting in two networks containing cov alentl y closed DNA circles only.

An updated model of mitochondrial genome replication in T. brucei -the loose-diploid model
The current model of kDNA replication suggests that the minicircles are released into the KFZ, and are replicated there .T hen the two daughter minicircles are separated and migrate to opposite sides of the kDN A netw ork, to be r eattac hed at the APS.While this model elegantly explains the maintenance of two complete sets of minicircles after replication, the localization of several proteins involved in this process, as well as the localization of the DNA replication intermediates is not congruent with the model.For example , T OP2 is suggested to be responsible for the release of the minicircles into the KFZ but is localized mainly at the APS.Se v er al pr oteins involv ed in r e plication initiation lik e RBP38, the helicases, and the primases, localize to the APS and not the KFZ and thus, their positioning is not consistent with the current model where the replication and segregation of the minicircles has to occur in the KFZ (Table 1 ).Also, especially the early replication intermediates ar e exclusiv el y found in a r egion that could be the APS but not in the middle of the kDNA disc.Furthermor e, the curr ent model also r equir es the pr esence of a completel y unknown sorting and tr ansport mec hanism that would ensur e pr oper separ ation and movement of the two daughter minicircles to the APS post replication.
Based on a comparison of the known minicircle replication factors and their localization we propose a new model (Table 1 , Fig. 3 , Supplementary movie).In such a model, the minicircles are released, r eplicated, and r eattac hed at the same lobe of the disc (Fig. 3 ).To allow maintenance of the essential minicircle genome by this replication mechanism, the kDN A netw ork must consist of a loose-diploid minicircle set (Fig. 3 ).Loose-diploid refers to the hypothesis that one set of essential minicircles is present, potentially in varying numbers, in each of the two lobes of the kinetoplast.We suggest that the entire minicircle replication occurs at the APS, where most of the re plicati ve enzymes are localized.The APS themselves might not be static entities and might move along the kDNA disc or alternativ el y the kDNA itself might move relative to the APS during replication as it has been proposed previously in (Liu and Englund 2007 ).The mechanism of replication via theta intermediates and the r emov al of Okazaki fr a gments and ga p r epair has been elucidated and described in great detail (Ryan andEnglund 1989a , Pér ez-Mor ga andEnglund 1993b ).We assume the replication process is completed faster for the leading strand minicircles than for the lagging strand minicircles, since fewer ga ps hav e to be mended in the leading str and.Consequentl y, the leading strand minicircle can be r eattac hed shortl y after r eplication, while gap repair for the lagging strand is still ongoing and would delay the r eattac hment of these minicircles .T his leads to sta gger ed distribution of the minicircles in the growing disc, allowing for an a ppr o ximate se paration betw een the tw o daughter minicircle sets, which is important for maintaining a bilobed minicircle distribution for the next generation.The loose-diploid model also provides an explanation for the observed dynamics in a population's minicircle r epertoir e ov er time since it assumes a certain sloppiness in the redistribution of the minicircles in the disc (Savill andHiggs 1999 , Cooper et al. 2019 ).
Maxicir cles, similar to minicir cles, ar e r eplicated unidir ectionally via theta intermediates .T hey, ho w ever, remain attached to the disc and concentrate between the two replicated discs just prior to kDNA segregation (Carpenter andEnglund 1995 , Gluenz et al. 2011 ).The function of this positioning remains unclear.One could speculate that this is a consequence of the minicircle replication process, but it might also fulfill another function.The kDNA disc is rather compact and transcription of the maxicircle genes might be impacted by this .T hus , the "exposure" during kDNA segregation could also provide a window of opportunity for transcription to occur and would allow for the spatial separation of replication and transcription.
Inter estingl y , T .brucei is thought to produce gametes for sexual r epr oduction in the saliv ary glands of the tsetse fly (Peacoc k et al. 2011(Peacoc k et al. , 2014 ) ).Through the detection of a gamete specific protein (HAP2) numerous intermediate meiosis stages with varying kDNA/nucleus ratios have recently been discovered (Peacock et al. 2021 ).These cells e v en contained small kDN A netw orks that could be result of a meiotic segregation event.What remains unknown, is whether trypanosomes inherit their kDNA uni-or bipar ental (Peacoc k et al. 2011(Peacoc k et al. , 2014 ) ).The loose-diploid replication model introduced above would argue for a possible biparental inheritance of kDNA upon fusion of gametes.It is possible that the Figur e 3. T he diploid model of minicircle replication in the kDNA.Minicircles are released from the disc into the KFZ, bind to some replication factors, and move to the APS (green), where replication is initiated and proceeds via theta intermediates.Replication leaves the minicircle created from the leading strand with a single nick, i.e. rapidly repaired at the APS before the minicircle is reattached to the growing kDNA network.The lagging strand minicircle is left with multiple nicks and gaps to be repaired.T herefore , it will remain in the APS for a prolonged period of time (where the nicks and ga ps ar e being r epair ed).As the disc is gr owing, the APS, and thus the site of minicircle r eattac hment, will mov e a wa y fr om the position wher e the leading strand minicircle was r eattac hed, leading to a spatial separation of the daughter minicircles .T he depiction in this figure focuses on the processes in the disc on the left (depicted in c y an).The same process also applies to the disc growing on the right (depicted in gra y).T he APS of the disc in focus, are depicted in green, the network bound minicircles of the diploid genome are depicted in cyan; before DNA synthesis, free minicircles are shown in black; leading strand daughter circles are then depicted in red, while lagging strand daughter circles are shown in yellow.For simplicity we omitted the replication of the maxicircles in this figure .kDN A w ould undergo a "meiotic segregation" during gamete formation.This would subsequently allow fusion of the networks during gamete fusion to obtain a diploid minicircle set in the pr ogen y.The diploid model in combination with biparental inheritance of kDN A w ould, thus provide a mechanism on how to maintain minicircle diversity during the life cycle of the parasite.

Similarities and differences to the current model
The major differences to the current model of kDNA replication in T. brucei and Crithidia lie (i) in the assumption that the the kDNA disc of a cell prior to replication already contains two complete sets of the essential minicircles (is loose-diploid) and (ii) that each set is localized in the one lobe of the disc and (iii) thus the two minicircle sets are replicated separ atel y.
Similar to the current models in T. brucei and Crithidia , the loosediploid model also assumes (i) that the release of minicircles occurs later all y into the KFZ wher e they encounter the r eplication machinery and (ii) that lagging strand minicircle progenies require extensiv e ga p r epair and thus accumulate outside the disc, while the leading strand minicircle progenies are reattached immediately after replication is finished, which is similar to what has been described for the Crithidia minicircles (Kitchin et al. 1985 ).As a consequence, the lagging strand progenies are positioned differ entl y in the growing disc (see Fig. 3 ).In addition to this passive process of daughter minicircle separation we also envision a more activ e mec hanism involving either an activ el y moving r eplication or r eattac hment mac hinery or an oscillating and rotating kDNA disc as has been suggested pr e viousl y for T. brucei and Crithidia , respectiv el y (Liu and Englund 2007 ).Since there is no life cell imaging data follo wing kDN A replication the actual dynamics of this pr ocess ar e difficult to e v aluate.

Where the loose-diploid model has shortcomings
The proposed model is based on the localization of mostly epitope tagged replication factors at the APS.Although tagging is a r eliable tec hnique in T. brucei we can not exclude that suc h pr o-teins are mislocalized or non functional due to the epitope tag.Ho w e v er, at least for the topoisomerase TOP2 and several of the pol ymer ases (P olIC, P olIB) mislocalization is unlik ely since the y were detected using antibodies targeting the endogenous protein.
We assume that there are two se parate re plication machineries that only replicate the set of minicircles in their lobe of the kDNA disc.This likely requires some structural organization.Although there is no experimental evidence, the TAC could provide a structur al element fr om whic h suc h a separ ation could be or ganized.The proper distribution of the minicircles in the kDNA lobe is k e y for the next generation.In addition to the passive mechanism of differ ential r eplication/r epair (see abov e) we assume that either the r eattac hment mac hinery is moving along/acr oss the disc or as proposed previously the kDNA disc itself is moving (Jensen and Englund 2012 ).Curr entl y, w e kno w of no mechanism explaining either of the mo vements .
In conclusion, we have summarized the current knowledge on 29 kDNA replication factors from T. brucei and Crithidia and suggest an updated model for T. brucei minicircle replication that does not r equir e the assumption of a curr entl y unknown sorting and transport complex moving fr eshl y r eplicated minicircles to the opposing ends of the kDNA disc.Instead, it depends on a loosely diploid structure of the kDN A netw ork and two replication centers at the APS.This is consistent with the majority of the localized replication factors (Table 1 ).In combination with the proposed sexual stages during the life cycle of the parasite, the new model would also provide a mechanism for maintenance of minicircle diversity.For future analyses of this replication model, it will be of great interest to assure wild-type-like expression levels of the tagged proteins of interest to decrease the risk of artifacts during detailed localization studies.

Figure 2 .
Figure 2. Ov ervie w of kDNA r eplication and segr egation in T. brucei .(A) Dorsov entr al vie w of the posterior region of a cell depicting the flagellar pocket, the basal bodies, and the mitochonrial membranes surrounding the kDNA at the beginning of kDNA replication.(B) and (C) Minicircles are released into the kinetoflagellar zone (KFZ), replicated, and reattached at the APS.Maxicircles are replicated, while remaining inside the growing kDNA network.Sim ultaneousl y to the kDNA replication process, the pro-basal body (pBB) tilts to face the mitochondrial membrane, and via the exclusion zone filaments (EZF), sets the base for growth of the new tripartite attachment complex (TAC).(C) and (D) During pBB maturation, the pBB rotates ar ound the matur e basal body (BB), and assembles the TAC proteins of the differentiated membranes (DM) and the unilateral filaments (ULF).(D) After completion of minicircle replication, the replicated kinetoplasts remain attached to each other by the maxicircles that accumulate between the two kDNA discs.Microtubule quartet (MTQ).The Images in this figure are stills from an animated model (see supplementary material).
and pol ymer ase beta-PAK, Pol and Pol -PAK (Tb927.5.2780,Tb927.5.2790)Pol β and Pol β-PAK are the first examples of mitoc hondrial pol ymerase β enzymes (Saxo wsk y et al.2002 , 2003 ).Pol β-PAK was named after its N-terminal pr oline-alanine-l ysine-ric h (PAK) extension of about 300 amino acids.Both enzymes are thought to be responsible for gap filling and repair of minicircles albeit at differ ent locations.Additionall y, they show 5 -deoxyribose phosphate lyase activity suggesting a possible role in base excision repair(Saxo wsk y et al. 2002 ).Pol β is localized at the APS, where it is likely responsible for gap repair on the lagging strand prior to r eattac hment of the circles to the network.Pol β-PAK on the other hand, is found throughout the kDNA disc, where it is potentially r equir ed for ga p r epair prior to network segregation(Saxo wsk y et al. 2003 ).The different localization of the two enzymes is correlating with a difference in biochemical properties.While Pol β has its highest activity at pH 8.0, Pol β-PAK maximum activity is shifted to w ar ds a more basic pH (pH 9.0), potentially reflecting the environment within the kDNA disc.

Table 1 .
Localization of proteins associated with minicircle replication.