The Consequences of DNA Lesions for Mitochondrial DNA Maintenance

The consequences of DNA lesions for mitochondrial DNA maintenance

Josefin M. E. Forslund

Department of Medical Biochemistry and Biophysics Umeå 2021 This work is protected by the Swedish Copyright Legislation (Act 1960:729) Dissertation for PhD ISBN: 978-91-7855-542-0 (print) ISBN: 978-91-7855-543-7 (pdf) ISSN: 0346-6612 New Series Number 2133 Cover photo: Mitochondrial DNA and DNA lesions (Josefin M. E. Forslund) Electronic version available at: http://umu.diva-portal.org/ Printed by: Cityprint i Norr AB Umeå, Sweden 2021

Till morfar Lennart, för att du alltid visste att jag skulle bli doktor.

Table of Contents

Abstract ...... iii Abbreviations ...... v Publication list ...... vii Author’s contribution ...... viii Introduction ...... 2 DNA replication ...... 3 Nuclear DNA ...... 5 Mitochondrial DNA ...... 6 Mitochondria and energy production ...... 6 MtDNA molecule and replication machinery ...... 8 MtDNA replication ...... 11 Mitochondrial dysfunction and disease ...... 13 MtDNA damage ...... 14 MtDNA damage tolerance ...... 14 Human PrimPol ...... 16 PrimPol’s polymerase activity and translesion synthesis ...... 17 Priming activity of PrimPol ...... 18 PrimPol’s functions in vivo ...... 19 PrimPol’s role in mtDNA replication ...... 20 Ribonucleotide incorporation ...... 21 Origins of ribonucleotides ...... 21 dNTP pools and Ribonucleotide reductase ...... 23 Removal of incorporated ribonucleotides ...... 24 Consequences of single ribonucleotide incorporation in nuclear DNA ...... 26 Detection of ribonucleotide incorporation in DNA ...... 27 Ribonucleotide incorporation in mitochondrial DNA ...... 28 Aims of thesis ...... 30 Results and discussion ...... 31 Part I – The role of PrimPol in mtDNA damage tolerance ...... 31 Mitochondrial replisome stalls at oxidative DNA damage (paper I) ...... 31 MtDNA replication is reinitiated by PrimPol after damage (paper II) ...... 35 Part II – Ribonucleotide incorporation in mtDNA ...... 42 Incorporation of ribonucleotides in yeast mtDNA (paper III) ...... 42 Incorporation of ribonucleotides by human mtDNA replication machinery (paper IV) ... 51 Conclusions and future outlook ...... 58 Part I ...... 58 Part II ...... 58 Final conclusions ...... 60 Acknowledgement ...... 61 References ...... 64

Abstract

Eukaryotic cells have their own energy-producing organelles called mitochondria. The energy is stored in the adenosine triphosphate (ATP) molecule and is produced via the oxidative phosphorylation process inside the mitochondria. Thirteen of the essential proteins required for this process are encoded on the mitochondrial DNA (mtDNA). To ensure sufficient energy production it is therefore important to maintain mtDNA integrity. MtDNA maintenance is dependent on several factors, which include the replicative DNA polymerase. In humans, the main mitochondrial polymerase is DNA polymerase gamma (Pol γ), whereas in S. cerevisiae the homolog is called Mip1. Defects in the mitochondrial DNA polymerase and mtDNA replication in general cause mitochondrial dysfunction, reduced energy production and, in humans, mitochondrial diseases.

DNA damage and non-standard nucleotides are frequently forming obstacles to the DNA replication machinery. One of the proteins that assists the nuclear replication machinery in dealing with DNA damage is the primase-polymerase PrimPol, performing either translesion DNA synthesis or alternatively priming replication restart after DNA damage. More recently, PrimPol was also identified inside the mitochondria. We therefore investigated the potential role of PrimPol to assist the mtDNA replication machinery at the site of mtDNA damage. Our results suggest that PrimPol does not work as a conventional translesion DNA polymerase at oxidative damage in the mitochondria, but instead interacts with the mtDNA replication machinery to support restart after replication stalling.

Stalling of DNA replication can also occur at wrongly inserted nucleotides. In this study, we pay extra attention to ribonucleotides, which are non- standard nucleotides in the context of DNA. Ribonucleotides (rNTPs) are normally building blocks for RNA but are occasionally utilized by DNA polymerases during DNA replication. Ribonucleotides are more reactive compared to dNTPs as they have an additional hydroxyl group (-OH). Their presence in the genome can lead to replication stress and genomic instability. In nuclear DNA, ribonucleotides are efficiently removed by the Ribonucleotide Excision Repair (RER) pathway and failure to remove them leads to human disease (e.g., Aicardi-Goutières syndrome). We investigated if ribonucleotides are removed from mtDNA and if not, how

iii

the replication machinery can tolerate the presence of ribonucleotides in the mtDNA.

By using several yeast strains with altered dNTP pools, we found that the RER pathway is not active in mitochondria. Instead, mitochondria have an innate tolerance to ribonucleotide incorporation in mtDNA and under normal cellular conditions mature human mtDNA contains ~50 ribonucleotides per genome. We show that this ribonucleotide tolerance is the result of human Pol γ’s remarkable abilities to 1) efficiently bypass ribonucleotides in the DNA template and 2) proficiently discriminate against the incorporation of free ribonucleotides during mtDNA replication. Pol γ’s discrimination capability against free ribonucleotides comes with a price. In the presence of high rNTP levels, Pol γ is inhibited in DNA synthesis and could eventually lead to frequent replication stalling. Together, these studies are in line with our hypothesis that ribonucleotides in mtDNA can be tolerated, with the consequence that mtDNA replication is in particular vulnerable to imbalances in rNTP/dNTP ratios.

In summary, this study shows that we cannot simply extrapolate our knowledge of nuclear DNA replication stress management to the mtDNA maintenance, highlighting the need to study the molecular mechanism by which the mtDNA replication machinery is able to cope with DNA lesions to prevent loss of mtDNA integrity and disease development.

Abbreviations mtSSB Mitochondrial single stranded binding protein 2D-AGE Two-Dimensional Agarose Gel Electrophoresis 6-4 PPs Pyrimidine (6-4) pyrimidine photoproducts 8-oxoguanine 8-oxo-G AEP Archeo-Eukaryotic Primase AP Abasic site ATP Adenosine triphosphate BER Base Excision repair CoA Coenzyme A CPEO Chronic Progressive External Ophthalmoplegia CPDs Cyclobutene Pyrimidine Dimers CSB Conserved sequence block CTNAs Chain-termination nucleotide analogues DNA Deoxyribonucleic acid DGUOK Deoxyguanosine kinase dRP 5´-deoxyribose phosphate dsDNA Double stranded DNA Exo- Exonuclease deficient FADH2 Reduced Flavin Adenine Dinucleotide G4 G-quadruplex HEK293 Human Embryonic Kidney cell line HIV Human Immunodeficiency virus HSP Heavy-strand promoter HU Hydroxyurea HydEn-Seq Hydrolytic End sequencing IFNα Interferon alpha KOH Potassium hydroxide LSP Light-strand promoter MEF Mouse Embryonic Fibroblast mtDNA Mitochondrial DNA

NADH Reduced Nicotinamide Adenine Dinucleotide NaOH Sodium hydroxide NCR Non-coding region nDNA Nuclear DNA NER Nucleotide Excision Repair NRTIs Nucleoside reverse transcriptase inhibitors nt Nucleotide OH Hydroxyl OH H-strand replication origin OL L-strand replication origin OXPHOS Oxidative phosphorylation PEO Progressive External Ophthalmoplegia Pol α DNA polymerase alpha Pol β DNA polymerase beta

Pol δ DNA polymerase delta Pol ε DNA polymerase epsilon Pol γ DNA polymerase gamma PolDIP2 DNA polymerase delta interacting protein 2 POLRMT Mitochondrial RNA polymerase RACE Rapid Amplification of cDNA Ends RER Ribonucleotide excision repair RITOLS Ribonucleotide incorporation throughout the lagging strand RNA Ribonucleic acid RNR Ribonucleotide reductase ROS Reactive oxygen species RPA Replication protein A SAMHD1 Sterile alpha motif and HD-domain containing protein 1 SDM Strand-displacement model ssDNA Single stranded DNA TAS Termination associated sequences TFAM Transcription factor A mitochondrial TK2 Thymidine kinase 2 TLS Translesion synthesis Top1 Topoisomerase 1 UV Ultraviolet WT Wild type ZnF Zinc finger

Nucleotides: dNs Deoxynucleosides dNMP Deoxyribonucleoside monophosphate dAMP Deoxyadenosine monophosphate dCMP Deoxycytidine monophosphate dGMP Deoxyguanine monophosphate dTMP Deoxythymidine monophosphate rNMP Ribonucleoside monophosphate rAMP Adenosine monophosphate rCMP Cytidine monophosphate rGMP Guanine monophosphate rUMP Uridine monophosphate dNTP Deoxyribonucleotide or deoxyribonucleoside triphosphate dATP Deoxyadenosine triphosphate dCTP Deoxycytidine triphosphate dGTP Deoxyguanine triphosphate dTTP Deoxythymidine triphosphate rNTP Ribonucleotide or ribonucleoside triphosphate rATP Adenosine triphosphate rCTP Cytidine triphosphate rGTP Guanosine triphosphate rUTP Uridine triphosphate ddNTP Dideoxynucleotide ddCTP 2´-3´-dideoxycytidine

Publication list

Paper I “Oxidative DNA damage stalls the human mitochondrial replisome.” Stojkovič G, Makarova AV, Wanrooij PH, Forslund J, Burgers PM and Wanrooij S. Sci. Rep. 6, 28942 (2016). Doi: 10.1038/srep28942

Paper II “PrimPol is required for replication reinitiation after mtDNA damage.” Torregrosa-Muñumer R*, Forslund JME*, Goffart S, Pfeiffer A, Stojkovič G, Carvalho G, Al-Furoukh N, Blanco L, Wanrooij S and Pohjoismäki JLO. PNAS. 114 (43) 11398-11403 (2017). Doi:10.1073/pnas.1705367114

Paper III “Ribonucleotides incorporated by the yeast mitochondrial DNA polymerase are not repaired.” Wanrooij PH, Engqvist MKM*, Forslund JME*, Navarrete C, Nilsson AK, Sedman J, Wanrooij S, Clausen AR and Chabes AR. PNAS. 114 (47) 12466-12471 (2017). Doi:10.1073/pnas.1713085114

Paper IV “The presence of rNTPs decreases the speed of mitochondrial DNA replication.” Forslund JME, Pfeiffer A, Stojkovič G, Wanrooij PH and Wanrooij S. PLoS Genetics, 14(3):e1007315 (2018). Doi:10.1371/journal.pgen.1007315

*These authors contributed equally

The original publications have been reproduced with the permission from each journal.

Related publications not included in this thesis: “Known unknowns of mammalian mitochondrial DNA maintenance.” Pohjoismäki JLO, Forslund JME, Goffart S, Torregrosa-Muñumer R and Wanrooij S. Bioessays 40(9) (2018).

vii

Author’s contribution

Paper II is included in the thesis of our collaborator Dr. Rubén Torregrosa- Muñumer (University of Eastern Finland). Dr. Rubén Torregrosa- Muñumer performed all two-dimensional neutral/neutral DNA gel- electrophoresis and the majority of the cell culture experiments. The author of this thesis performed the majority of the in vitro experiments and generation of PrimPol-inducible cell lines. Additional experiments were performed either by group members of Associate Professor Sjoerd Wanrooij’s laboratory or of Professor Blanco’s laboratory (Autonoma University of Madrid). The author of this thesis and Dr. Rubén Torregrosa- Muñumer share first authorship since they contributed equally to the work.

Paper I, III and IV are not included in another doctoral thesis.

viii

Introduction

Proper information storage, such as medical records or ancient books, has been important to keep data safe and accessible. To share necessary information and key events to the next generations, a suitable storage technique is required to protect the information over a long period of time. In recent times the development of storage methods has exploded in parallel with the development of computers. However, the most fundamental information for almost all living organisms has been stored for billions of years inside a molecule called deoxyribonucleic acid (DNA). DNA is often called “the code of life” since it contains all the instructions needed to build an organism.

In 1985, Francis Crick proposed “the central dogma” which describes how the genetic information is safely stored and accessible in the DNA via the ribonucleic acid (RNA) molecules to be able to produce proteins needed for cellular function (Fig 1). Unfortunately, as many other types of information, DNA can be harmed by both exogenous and endogenous processes. Damage to the DNA, such as lesions or breaks, can lead to modifications in the code and, in worst case, lost information. A defence against harmful factors as well as proper repair systems are crucial to keep the information well protected.

Figure 1. The central dogma of biology. All our genetic information is stored inside DNA molecules. The genome is replicated by DNA polymerases which ensures that the genetic information is duplicated and inherited to offspring cells. The genetic information is utilized in the transcription process when RNA polymerases produce messenger RNA molecules. The RNA molecules are translated into functional proteins by the ribosomes.

DNA replication The DNA molecule consists of two complementary polynucleotide chains, which are annealed to each other in opposite direction to form an antiparallel double stranded helix. The DNA molecule is built by adding monomeric deoxyribonucleoside triphosphates (deoxyribonucleotides, dNTPs). A dNTP consists of a five-carbon deoxyribose sugar with nitrogen-base and three phosphate groups attached. One of four different bases can be attached; adenine, cytosine, guanine or thymine (dATP, dCTP, dGTP or dTTP, respectively) (Fig 2A). Once the dNTP is built into the DNA strand, two phosphate groups leave the dNTP which becomes a deoxyribonucleoside monophosphate (dNMP).

To ensure safe inheritance of the genetic information to progeny cells, accurate duplication of the DNA molecule is required. This duplication process is called DNA replication and is catalysed by DNA polymerases. During DNA replication, monomeric dNTPs are linked to the DNA growing chain via a phosphodiester bond. This bond is created through a nucleophilic attack from the 3´-hydroxyl (3´-OH) group of the growing chain on the α-phosphate of the incoming dNTP1 (Fig 2B). The nucleotides are linked to each other via a covalent bond between the sugar and the phosphate groups. During the formation of this phosphodiester bond, two phosphate groups leave the reaction and the newly added dNTP is transformed into dNMP. New dNTPs can only be linked on the end where a free 3´-OH group is available, called the 3´-end. The opposite end, with a free 5´-phosphate group, is called the 5´-end.

The addition of dNTPs on the growing chain is not random since the complementary dNTP is added according to the specific base sequence on the template strand. The force holding the two antiparallel strands together is facilitated via hydrogen bonds between bases in the different strands. The two-ring bases, called purines, are always paired with a single-ring base, called pyrimidine. In other words; A will base pair with T and C will base pair with G2. The template strand (called Crick strand) will always have a 3´ to 5´ direction, since the new strand (called Watson strand) grows from the 5´-end to the 3´-end. The usage of one strand as template, ensure that two identical copies of the genetic information are produced. The two strands are held together via the hydrogen bond created between one base in each strand.

Figure 2. The molecular basis of DNA replication. (A) The building blocks for DNA synthesis are deoxynucleoside triphosphates (dNTPs). A dNTP consists of a sugar ring with five carbons (pentose ring). A nitrogenous base (either adenine, cytosine, guanine or thymine) is attached on the 1´carbon (green). On the 3´ carbon, a hydroxyl group is attached while on the 5´carbon, a triphosphate group is found (yellow). (B) A new DNA strand is synthesized from the 5´-end to 3´- end using the complementary strand as template. The triphosphate group of the incoming dNTP is nucleophilic attacked by the 2´-OH group of the 3´-end and a phosphodiester bond is formed to the incoming dNTP. The two phosphate groups leave the incoming dNTP, which is transformed into a dNMP. The type of dNTP used is dependent on the complementary base of the template strand. Hydrogen bonds are formed between the bases in the complementary strands and the base pairing holds the two strands together creating a double stranded helix.

The DNA polymerases cannot start the replication by de novo synthesis of DNA. The replication process is initiated by the formation of RNA primers at dedicated origins within the DNA. The DNA polymerases extend from the RNA primer and the replication is started. The replication usually proceeds on both strands simultaneously but in opposite direction, creating a fork-like structure. One strand, called the leading strand, will be

replicated continuously while the opposite strand, called lagging strand, needs the addition of many RNA primers. The lagging strand will be replicated discontinuous in separate sections which are called Okazaki fragments3. The Okazaki fragments are later joined together to form a uniform lagging strand.

For the cell to make use of the stored information, the DNA is transcribed into RNA molecules by the RNA polymerases. In contrast to DNA polymerases, RNA polymerases instead use ribonucleotides (rNTPs) as building blocks for synthesising the messenger RNAs (mRNAs). Eventually, the mRNAs get translated by ribosomes into polypeptide chains and folded into functional proteins.

In eukaryotic cells, two sets of DNAs can be found; the chromosomal nuclear DNA (nDNA) inside the nucleus and the mitochondrial DNA (mtDNA) located in the mitochondrial organelles.

Nuclear DNA The nDNA encodes for the majority of all proteins produced in the eukaryotic cells. It is located in the cell nucleus with the nuclear envelope enclosing the genome, separating it from the cytosol and other organelles. To be able to fit the almost two-meter-long DNA inside the nucleus, the genome is bound to specialized proteins which folds and tightly pack the DNA into compact but highly organised chromosomes.

During cell division, DNA is duplicated making two identical copies which are divided into the two cells. The replication process needs to be accurate, but at the same time very efficient. For this reason the process is strictly regulated at several levels to ensure accurate duplication of the DNA. Briefly, the chromosomal DNA of eukaryotes are replicated by three major replicative polymerases; DNA polymerase α, δ and ε. The replication process is initiated at all origins of replication at the same time via the Pol α-RNA primase complex. The protein complex synthesise a short RNA:DNA hybrid primer, which can be utilized to start the replication. The replication will continue in a bidirectional manner where the leading strand will be synthesised continuously by Pol ε and the lagging strand is synthesised discontinuously by Pol δ4,5.

Mitochondrial DNA In contrast to the nDNA, which has two copies of each chromosome, the mtDNA is found in multiple copy (few hundred up to thousands) inside the mitochondria. The human mtDNA is a small circular molecule of only 16 kb and strictly maternally inherited. On the mtDNA molecule, genes encoding for 22 tRNAs and 2 rRNAs can be found, as well as genes for 13 proteins. These proteins are all part of the mitochondrial electron transport chain which is crucial for cellular energy production. All the other proteins needed for mitochondrial function are encoded by genes found on the nDNA. This includes all proteins involved in the mtDNA replication, transcription and translation machineries.

Mitochondria and energy production The mitochondria were first named after their appearance in 1898 by Dr. Carl Benda; the Greek word “mito” means thread and “chondria” means granule6. The mitochondria are dynamic organelles found inside eukaryotic cells which are enclosed by double membranes. The outer and the inner membranes have different functions and are separated by an intermembrane space in between. In the outer membrane, transport proteins are transferring large proteins between the cytosol and the mitochondrion. The outer membrane is also permeable for very small molecules which diffuse into the intermembrane space. The inner membrane on the other hand consists of a double phospholipid layer which is not permeable except for ions. The inner membrane surrounds the mitochondrion matrix and only a very specific set of proteins and molecules can go through the inner membrane. The inner membrane is not smooth; it forms invaginations into the matrix which is called the cristae. The infoldings increase the area of inner membrane which also increase surface where the main function of mitochondria can take place7 (Fig 3).

Mitochondria have several important functions inside the cell, such as regulation of innate immunity and programmed cell death. However, the primary function of the mitochondria is to produce the energy-carrier molecule adenosine triphosphate (ATP). ATP can be produced when glucose is converted to pyruvate in anaerobic glycolysis, however, this process only harvests a small fraction of the energy stored within the glucose molecule. To withdraw more energy, pyruvate is transported into the mitochondrion and oxidized into acetyl Coenzyme A (CoA) which continues through the citric acid cycle. During the citric acid cycle, NADH

and FADH2 molecules are produced which are the main carriers of high energy electrons. NADH and FADH2 transport the electrons to the electron transport chain which is embedded in the inner membrane of the mitochondrion. Here, the oxidative phosphorylation process takes place. Briefly, the high-energy electrons together with O2, drives four large protein complexes to pump out protons to the intermembrane space. A proton gradient is achieved which will drive the membrane-bound enzyme ATP synthase. ATP synthase facilitate the conversion of ADP + Pi to ATP (Fig 3).

Figure 3. The mitochondrion with a simplified overview of the respiratory chain. The mitochondrion is encapsuled by two membranes; the outer membrane and the inner membrane. The four complexes (I, II, III and IV) are embedded in the inner membrane and are passing electrons - (e ) between each other through the NADH and FADH2 molecules. Protons are pumped out into the intermembrane space and a proton gradient is produced. The gradient drives the ATP synthase to produce ATP from ADP+Pi. The circular mitochondrial DNA can be found in the matrix at the inner membrane packed in nucleoids.

Most of the proteins needed for the large complexes in the electron transport chain are encoded on the nDNA, produced in the cytosol and imported into the mitochondrion. However, some of the more hydrophobic proteins in the large complexes are encoded on the mtDNA and produced directly inside the mitochondria. The electron transport chain and oxidative phosphorylation process are therefore completely dependent on an intact mtDNA for normally cellular function and ATP production.

MtDNA molecule and replication machinery It is believed that the presence of mitochondrial organelles once originated from α-proteobacteria taken up into an archaebacterium. During evolution, the endosymbiosis between the α-proteobacteria and archaebacteria developed into eukaryotic cells8. Most of the α-proteobacteria genes were either transferred to the nDNA or lost, while few are still kept on the relatively small mtDNA molecule. The ancestral α-proteobacteria later developed into energy producing mitochondria with several copies of small mtDNA molecules. The mtDNA is packed into nucleoids by the Transcription factor A mitochondrial (TFAM) protein, where each nucleoid contains one or two mtDNA molecules9,10.

The two strands of the mtDNA can be separated from each other using a cesium chloride gradient. This is due to the difference in base composition and thereby having different buoyant densities in the cesium chloride gradient. The heavy (H) strand is rich in guanines, while the light (L) strand has less guanines. The size of the mtDNA varies between eukaryotic cells; for example, the human mtDNA is only 16.6 kb, while the Saccharomyces cerevisiae (S. cerevisiae) mtDNA is much larger (85.8 kb). Even though the size of the human mtDNA is small, it encodes for 13 proteins, 2 rRNAs and 22 tRNAs. The 13 proteins found on the mtDNA are only a small portion of the 90+ proteins which shape the respiratory chain, however, the mtDNA encoded subunits are essential for OXPHOS function. Due to the significance of mtDNA in energy production, both the maintenance and the replication of mtDNA are regulated and essential processes in the cell.

The mtDNA is replicated by a unique core replication machinery that only works inside the mitochondria. In contrast to the nuclear replication, most of both mtDNA strands are replicated by one DNA polymerase, called 11 polymerase γ (Pol γ AB2, from now on called only Pol γ) . Pol γ is a heterotrimer protein with one large catalytic subunit (A) and two processivity subunits (B2).

The 140 kDa large catalytic A subunit contains two structural domains responsible for the Pol γ enzymatic activities12. First, the polymerase domain with 5´-3´ DNA polymerase activity which synthesizes the new mtDNA strand during replication. Second, the exonuclease domain with 3´-5´ exonuclease activity which can proofread wrongly inserted nucleotides.

Figure 4. The minimal human mitochondrial replisome. The human mtDNA is replicated by the heterotrimer DNA polymerase γ AB2 (Pol γ). MtDNA Pol γ synthesis the new DNA strand in 5´ to 3´direction. The hexamer Twinkle helicase assist Pol γ by unwinding the double stranded DNA, which creates a fork-like structure. The tetramer mitochondrial single stranded binding protein (mtSSB) bind to the single stranded DNA to avoid the formation of secondary structures.

The catalytic subunit A is dependent on the presence of two 55 kDa B subunits, which increase both the DNA binding and nucleotide binding of Pol γ A subunit and thereby increasing the processivity of the polymerase13. The yeast homolog to Pol γ is the mtDNA polymerase Mip1 which is similar to Pol γ A subunit, but lacks the B accessory subunits14. Pol γ also has 5´-deoxyribose phosphate (dRP) lyase activity which potentially is required for mitochondrial DNA repair15.

Pol γ has several features which distinguish it from nDNA polymerases. For example, Pol γ belongs to the family A polymerases together with the Escherichia coli DNA polymerase I and T7 DNA polymerase16. Pol γ has therefore more similarity to bacterial and bacteriophage polymerases than with the nDNA replicative polymerases. Also, Pol γ is sensitive to several nucleotide analogues used during for example HIV treatment. These treatments often effect mtDNA replication, giving side effects with mitochondrial dysfunction phenotypes16.

During mtDNA replication, Pol γ is dependent on assistance from several proteins to be able to duplicate the mtDNA. The helicase Twinkle, which is similar to the T7 primase/helicase, forms a hexamer 17. Twinkle catalyze unwinding of DNA duplex in ATP- or UTP-dependent manner with 5´-3´

direction. Twinkle assist Pol γ by unwinding double stranded DNA (dsDNA) in the mtDNA replication fork, facilitating continuous replication by Pol γ. During the unwinding, single stranded DNA (ssDNA) is formed and can spontaneously fold into secondary structures such as hairpins. The ssDNA is bound by the mitochondrial single stranded binding protein (mtSSB) to resolve these secondary structures to ensure efficient DNA replication by Pol γ18. MtSSB also stimulate the DNA unwinding activity of Twinkle via direct protein-protein interaction. Pol γ, Twinkle and mtSSB together forms the minimal mtDNA replisome which can be reconstituted in vitro19 (Fig 4).

The mitochondrial RNA polymerase (POLRMT) is essential for the gene expression and mtDNA transcription inside the mitochondria. The structure of POLRMT is similar to the T7 RNA polymerase, however, the transcription mechanism differs20. Briefly, transcription is initiated from two independent promoters on the H- and L-strand called heavy strand promoter (HSP) and light strand promoter (LSP). The transcription is initiated by POLRMT together with several other essential proteins21. Due to the lack of mtDNA introns, the transcription by POLRMT will generate long polycistronic precursor RNAs which are later processed into the different tRNA, rRNA and mRNAs.

POLRMT is also required for initiation of mtDNA replication; it provides primers needed for the replication to start at both the H- and the L- strand22,23. POLRMT is therefore part of the mtDNA replication machinery since it functions as the replicative primase of mtDNA replication.

Other proteins involved in mtDNA replication and maintenance In addition to the mtDNA replication machinery, several other factors have been identified as being involved in mtDNA replication and maintenance. Some examples are the other DNA polymerases found inside the mitochondria such as PrimPol and Pol β (See section “Human PrimPol” and for review24). Also, several DNA end-processing proteins are necessary for normal mtDNA replication such as MGME1, FEN1 and DNA225,26 as well as DNA ligase 3 for proper ligation during DNA repair27. Among other DNA repair proteins, we can also find RNase H1 (for more details see section “Ribonucleotide incorporation in mtDNA”) and several other DNA damage recognition proteins.

MtDNA replication Although the mtDNA is gene-rich, one region of about 1 kb does not contain any coding elements. Instead, this non-coding region (NCR) seem to be the control center of both replication initiation and transcription. The NCR, contains both transcriptional promoters (HSP and LSP) as well as the H-strand replication origin (OH) approximately 200 bp downstream of the LSP (Fig 5). The NCR also includes three conserved sequence blocks (CSB1, CSB2 and CSB3) located between the LSP and OH. POLRMT initiates transcription from LSP, after which it encounters the guanine-rich CSB2 region. The RNA transcript has the possibility to form a stable hybrid G-quadruplex (G4) structure bound to the displaced H-strand28,29. This R-loop formation can terminate the transcription after which the replication machinery can use the RNA as a primer. However, before this occurs the R-loop requires RNase H1 processing to generate a 3´-ends that is accessible for Pol γ to initiate replication30. The replication is then initiated from OH via a pre-terminated RNA transcript generated from LSP. The detail molecular mechanisms that regulate the switch from transcription to replication is not yet elucidated31,32.

After replication initiation at OH, the majority of replication forks terminate after 650 bp at a conserved region called the termination- associated sequence (TAS)33. The pre-terminated DNA strands forms the so called 7S DNA, which remains hybridized to the template DNA. About 5% of the replication events continue the duplication process beyond TAS and will copy the entire mtDNA molecule. The mtDNA replication mechanisms differs from the nDNA replication mechanism, possibly due to the dissimilar genome size and circularity of the mtDNA. The mode of mtDNA replication has been under intense debate and the proposed mtDNA replication models are only briefly described below.

The first model to be presented was the strand-displacement model (SDM), where both the H- and L-strands are continuously replicated, but in an asymmetric manner34,35. The replication proceeds unidirectional and initially only a new H-strand is synthesized. According to SDM, no Okazaki-like fragments are formed on the L-strand as in the case of nDNA replication, instead the displaced strand is covered by mtSSB until replication of the second DNA strand is initiated.

Figure 5. Overview of the mtDNA molecule and non-coding region (NCR). The NCR contains both the H- and L-strand promoters (HSP and LSP) as well as the H-strand origin of replication (OriH). Also, three conserved sequence boxes (CSB1-3) and termination-associated sequence (TAS) are located in the NCR. Preterminated replication forms the 7S DNA and are bound to the displaced H-strand, called the D-loop.

A second alternative proposed replication mode, called RITOLS (ribonucleotide incorporation throughout the lagging strand) is for most similar to the SDM mode of replication. However, with the exception of what covers the displaced strand: long RNA transcripts in the RITOLS mode and mtSSB in SDM replication36,37.

After the replication machinery has reached approximately two thirds of the mtDNA molecule, the L-strand origin (OL) gets exposed and folded into a stem-loop. POLRMT can synthesize a 25 nt primer from a poly-T stretch at the stem-loop which can be used by Pol γ to start the synthesis

of the L-strand38. The replication of both the H- and L-strand continue in opposite directions until a full circle is synthesized.

In contrast, the third suggested replication model, called strand-coupled model, resembles more the conventional replication observed in nDNA replication. Here, the model suggest that the replication of the L-strand is started at several positions and synthesized discontinuously39.

Mitochondrial dysfunction and disease Malfunction of the mitochondria can lead to a broad variety of disorders such as neuromuscular disease or cancer40,41. Since the mitochondrial organelles are responsible for the energy production, tissues with high energy demand are usually affected. This includes skeletal muscles, heart muscles, liver and brain, but other tissues can also show signs of mitochondrial dysfunction.

Although the mitochondria have many functions inside the cell, failure to assemble the respiratory chain seem to be one of the most common reasons for mitochondrial disease42. Defects in mtDNA replication, transcription and translation, can all result in respiratory chain dysfunction and inefficient energy production. These defects can arise from genetic mutations in the both the nuclear or mitochondrial genome.

All the genes needed for the mitochondrial replication, transcription or translation are encoded on the nDNA. Mutations in POLG (Pol γ) or TWNK (Twinkle) genes can lead to a defective mtDNA replication machinery and mtDNA instability. POLG- or TWNK-mutations are associated with a large variation in symptoms and often lead to the accumulation of multiple mtDNA deletions, progressive external ophthalmoplegia (PEO) and mitochondrial myopathy (for review see43). Other examples of nuclear gene alterations that cause mtDNA instability affect protein that are responsible for the dNTP supply to the mitochondria44,45.

Since the mtDNA encodes for 13 of the required proteins of the respiratory chain, mutations in the mtDNA sequence can also cause failed OXPHOS. Mutations on the mtDNA can be genetically inherited or can arise spontaneous due to exogenous or endogenous DNA damaging factors. The human mtDNA can be found in multiple copies inside the cells and often these copies are genetically identical. However, in some patients,

heteroplasmic mixtures can be observed, where a pathogenic mtDNA mutation co-exists with a wild-type mtDNA copy. The mutated mtDNA copies or the deleted molecules need to reach a certain threshold before causing mitochondrial dysfunction46.

MtDNA damage Common for both the nDNA and the mtDNA is the constant exposure to DNA damaging agents. In mitochondria, endogenous or exogenous stresses can lead to mutations or stalled mtDNA replication and cause mitochondrial dysfunction.

Endogenous factors produced within the cell can be harmful for the DNA. In mitochondria, the main intrinsic stress is the OXPHOS which, besides ATP, also produces reactive oxygen species (ROS). MtDNA might in particular be vulnerable to ROS oxidation, as the mtDNA is in close vicinity to the respiratory chain where ROS is produced. The ROS oxidation can lead to the formation of DNA lesions such as 8-oxoguanine (8-oxo-G) and abasic site (AP). In the 8-oxo-G lesion one additional oxygen atom is placed on the 8th carbon in the guanine base, while the AP lacks any base at all. These DNA damages can lead to mutagenesis and complete replication stalling47.

Exogenous factors, such as irradiation (e.g., ultraviolet UV irradiation) or genotoxic chemicals, can also result in DNA lesions. These chemicals include several intercalating agents (e.g., ethidium bromide) or drugs used in cancer therapy (e.g., cisplatin). Other type of drugs, such as nucleoside reverse transcriptase inhibitors (NRTIs) used in HIV treatment, can give drug side effects that lead to mitochondrial toxicity48. The human Pol γ is especially sensitive to these nucleoside analogues used in the antiviral treatment most likely because of Pol γ structural resemblance with the viral HIV reverse transcriptase.

MtDNA damage tolerance DNA damage tolerance in cells is important to ensure replication progression since long-term stalled replication forks can collapse and cause genomic instability. The nDNA, is armed with at least three distinct pathways to restart stalled replication forks: translesion synthesis, template switching (fork reversal) and repriming. How mitochondria overcome a replication fork stalling event and prevent harmful mtDNA instability

remains to be elucidated. There are however several damage tolerance mechanisms in the mitochondria, that prevent replication stalling, some of which are similar to the mechanisms found in the nucleus.

The best characterized repair pathway in the mitochondria is the base excision repair (BER)49,50. The mtDNA BER pathway is responsible to remove oxidative damage to the mitochondrial genome. The general idea is that most components of the mitochondrial BER machinery such as the glycosylases, the AP endonuclease and the ligase are shared with the nuclear BER pathway. It is still is unclear which DNA polymerases performs gap filling after the removal of oxidative mtDNA damage, but Pol γ and/or Pol β are the most likely candidates for this task inside the mitochondria.

Cells can also use specialized translesion polymerases to traverse DNA damage if lesion is not repaired or bypassed by the replicative DNA polymerases51. Translesion synthesis (TLS) polymerases exhibit a more flexible active site, making the binding pocket less tight compared to the replicative polymerases. The TLS polymerases can synthesis relatively efficiently over the DNA damage, but are often mutagenic. Several of the cellular TLS polymerases are reported to have a mitochondrial localization, such as PrimPol and Pol β24. The mitochondrial location of these proteins remains to be validated, until additional experiments can elucidate the yet unknown mitochondrial import mechanisms, or alternatively are able to assign a specific function to these polymerases in mtDNA maintenance.

Some more specific DNA damages, such as incorporation of dideoxynucleotides (ddNTPs) cannot be resolved by normal translesion synthesis. Instead, the replication can be restarted by repriming downstream of the DNA damage. In the nucleus the restart is thought to initiate from a primer that is synthesized by PrimPol. The finding that PrimPol localizes to the mitochondria suggests that the mitochondrial replication forks can similarly be restarted by a PrimPol dependent mechanism, however this still needs to be experimentally addressed.

Human PrimPol Recently a new DNA repair protein was identified which localize to both nucleus and mitochondria52. The novel protein, called PrimPol, has both primase and polymerase activity and is thought to be involved in many processes of both nDNA and mtDNA maintenance. PrimPol belongs to the archeo-eukaryotic primase (AEP) superfamily of primases which includes primases from Archaea and Eukaryotes. Orthologues of PrimPol can be found in many vertebrates, primitive eukaryotes (e.g., fungi and algae) and plants, but are absent in some species for example Drosophila melanogaster and S. cerevisiae.

Figure 6. Schematic overview of human PrimPol protein. (A) The human PrimPol have two major domains: the archeo-eukaryotic primase (AEP) domain and the Zink-finger (ZnF). The AEP domain contains all the catalytic residues essential for the polymerase and primase activities (Ia, Ib, I, II and III). In the ZnF domain, four residues are conserved which are needed for coordinating a Zink ion and stabilize PrimPol on single stranded DNA. The RPA-binding motif is located in the C-terminal of PrimPol. (B) A linker connects the AEP and the ZnF domain.

The human variant of PrimPol is a 560 amino acid protein with two major domains. The N-terminal part of PrimPol is an AEP-like domain with a catalytic core which consists of two ⍺/β modules (ModN and ModC) Fig 6. The crystal structure of the AEP domain bound to the template DNA revealed that PrimPol lacks the characteristics to fold into the typical right- hand polymerase structure. Instead, ModN and ModC surrounds the 3´- end of the primer without the typical resemblance to a right-hand structure. ModN interacts with the template strand via two important motifs (Ia and Ib)53. ModC contains three conserved motifs; I, II and III. Motif I and III shape the binding site for the divalent metal ion while motif II is needed for the binding of the incoming nucleotide. All three motifs are essential for both the DNA polymerase and primase activities54.

PrimPol’s polymerase activity and translesion synthesis When comparing PrimPol with replicative polymerases during DNA synthesis, it was clear that PrimPol possess much lower processivity54. This suggested that PrimPol’s polymerase activity is not likely to compete with replicative DNA polymerases during normal replication conditions, as replicative polymerase replicate both faster and with a higher accuracy. Recently, an interaction partner for PrimPol, called PolDIP2, was identified as a processivity factor. PolDIP2 is can stimulate both PrimPol’s DNA binding and polymerase activity. Although PrimPol DNA synthesis processivity increases substantial with PolDIP2, it is still a magnitude less processive when compared to replicative DNA polymerases55,56.

The processivity of PrimPol is dependent on the metal ion coordinated in the catalytic site. PrimPol can utilize both magnesium (Mg2+) and manganese (Mn2+) ions, but has lower DNA binding affinity and polymerase activities with magnesium. Manganese instead enhanced the DNA binding and thereby the polymerase activity with several folds57,58. However, the presence of manganese promoted a more error-prone PrimPol DNA synthesis and can induced a template-independent terminal- transferase activity54. The increased mutation rate and template- independent polymerization could potentially be harmful to the cell since PrimPol lacks 3´-5´ exonuclease proofreading ability. The lower activity with magnesium could therefore be an important mechanism to regulate promiscuous activity of PrimPol inside the cells.

Instead of participating in the bulk DNA replication, PrimPol’s polymerase activity is mainly important for TLS over DNA lesions. For example, PrimPol can in vitro bypass UV-induced damage (e. g., pyrimidine 6-4 pyrimidine photoproducts) and oxidative lesions (e.g., 8- oxo-G)52,59, but in an error-prone fashion. Translesion synthesis of more bulky DNA lesions, such as abasic sites (AP) and cyclobutene pyrimidine dimer (CPDs e.g., T-T dimers), makes PrimPol adapt a more pseudo- translesion polymerase state60. PrimPol has the ability to loop out the primer and reanneal it to a position downstream of the lesion. This primer translocation would generate shorter products than expected as this bypass skips several bases when copying the DNA template. The pseudo- translesion activity and primer reannealing is especially enhanced in the presence of manganese. The general thought is that, in vivo, both the TLS and the pseudo-TLS activity of PrimPol could help the replication machinery to overcome DNA lesions to complete the replication process.

Priming activity of PrimPol The second domain of PrimPol is located in the C-terminal and contains a conserved zinc finger (ZnF) motif with a zinc ion coordinated by three cysteines and one histidine. The ZnF domain of PrimPol is not essential for primer elongation, however indispensable for de novo synthesis (priming), with a critical role for two conserved amino acid residues C419 and H42654,60.

Figure 7. Summary of PrimPol functions. The human PrimPol has two activities; polymerase and primase activity. The polymerase activity (left) can perform normal 5´-3´ DNA synthesis, but with low processivity. PrimPol can also bypass lesions using translesion synthesis or pseudo-translesion synthesis. The bypass of lesions is often error-prone and result in mutations. The primase activity (right) of PrimPol can be used for repriming downstream of lesions, G-quadruplexes (G4s) or chain termination nucleotide analogues (e.g., ddCTP). The repriming result in single stranded DNA gaps which requires additional repair.

During a priming event, the ZnF-domain binds and stabilize PrimPol on single stranded DNA54. The stabilization allows PrimPol to synthesis the first dinucleotide on the ssDNA and start the de novo synthesis of DNA primers. Priming by PrimPol seems to be DNA sequence specific, since the ZnF-domain preferentially recognizes the 3´-GTCC-5´ DNA sequence

to initiate the priming process52,61. Although PrimPol has RNA priming ability, it has unexpectedly strong preference for dNTPs over NTPs52,59 and can therefore also synthesis DNA primers instead of only RNA primers.

Since the TLS function of PrimPol is limited compared with other TLS polymerases, the repriming activity of PrimPol seems to be a more efficient DNA damage tolerance mechanism. Also, certain type of DNA lesions cannot be resolved by TLS, for example at incorporation of chain- termination nucleotide analogues (CTNAs) or AP sites. Indeed, PrimPol can reprime downstream of both CTNAs and AP sites62, but also downstream of G-quadruplex structures63.

Considering both the primase and polymerase activity, PrimPol has the possibility to act as a DNA repair protein during many different types of DNA damage in the cells. Some of these functions are summarized in Fig 7.

PrimPol’s functions in vivo Several groups have extensively demonstrated PrimPol’s role in nDNA maintenance in vivo. Under normal conditions, PrimPol is not essential for nuclear replication since PrimPol knock out mouse are viable52. But DNA fiber analysis of PrimPol silenced cells showed a slower fork progression in S-phase which suggest a role of PrimPol during nDNA replication60. With both a primase and a polymerase activity, PrimPol makes an ideal candidate to assist the replication fork during nDNA replication stress.

Ultraviolet (UV) irradiation of human cells, leads to chromatin recruitment of PrimPol with an accumulation of PrimPol at the site of DNA damage60,64. The recruitment to the chromatin is facilitated via PrimPol’s interaction with the replication protein A (RPA). In the C-terminal of PrimPol, two motifs have been identified as RPA interaction sites65(Fig 6). Also, RPA clearly stimulate PrimPol in vitro by increasing the binding of PrimPol to a long single stranded DNA temple61.

PrimPol deficient cells are defective in replication restart and therefore show an increased level of γH2AX foci upon UV irradiation64,66. This suggest that PrimPol plays an important role during the recovery after UV damage. PrimPol also is an important player in the recovery phase after stalled replication induced by dNTP depletion (hydroxyurea HU) or

CTNA treatment60,62,64. PrimPol has also been suggested to play a role in double strand break repair via the Rad51 dependent homologous recombination pathway67. Taken together, these findings show the importance of PrimPol in nDNA maintenance, especially during replication stress.

PrimPol’s role in mtDNA replication Interestingly, subcellular fractionation revealed that PrimPol not only localize to the nucleus but also the mitochondrial compartment. Although PrimPol is not needed under normal conditions for nDNA replication, two- dimensional DNA gel electrophoresis analysis showed that silencing of PrimPol affected mtDNA replication which also resulted in a decreased of the mtDNA copy number52. Contradictory to this, another report, showed that PrimPol knock out cells have an increased mtDNA copy number and decreased replication68.

When first discovered, exome sequencing demonstrated mutations in the PrimPol gene (CCDC111) in patients with high myopia (problems with eye vision)69. In vitro, the Y89D mutation was reported to decrease both polymerase and primase activity as well as decreased DNA binding and decreased nucleotide affinity70. The same mutation was later found also in patients with chronic progressive external ophthalmoplegia syndrome (CPEO), a mitochondrial disorder associated with mtDNA deletions71. However, the Y89D mutation was also found in some healthy individuals and the role of PrimPol Y89D during the development of mtDNA deletions is still debated72.

When crosslinked, PrimPol could be pulled down with the mtDNA replication proteins Pol γ and Tfam52. This, together with the mutations found associated with mitochondrial disease, suggest that PrimPol might play an important role in mtDNA maintenance. However, the exact contribution of PrimPol in mtDNA replication and mtDNA damage tolerance still needs to be elucidated.

Ribonucleotide incorporation During the past twenty years it has become clear that genomic instability may result from not only exogenous stress, but also from endogenous processes such as misincorporation of non-canonical nucleotides. For example, DNA polymerases can occasionally misincorporate ribonucleotides (rNTPs) instead of dNTPs during DNA synthesis73. Ribonucleotides normally build up the different RNA molecules found inside the cells, but are very similar to dNTPs. There are only two differences: the thymine base is exchanged for an uracil base in RNA, and rNTPs have a OH-group on the second carbon atom of the furanose ring (Fig 8). The 2´-OH group is very reactive and thereby decreases the stability of RNA. If ribonucleotides are incorporated into DNA as ribonucleoside monophosphates (rNMPs), the 2´-OH of the rNMP can attack the sugar-phosphate backbone of the DNA, resulting in increased genomic instability (Fig 8). The rNMPs can also change the structure of the DNA which can affect both protein binding and replication74,75. Incorporated rNMPs have been detected in the genome of many different organisms such as bacteria76, yeast73 and mammalians77-80. Quantifications of incorporated rNMPs, suggest that rNMPs are the most frequently- incorporated non-canonical nucleotide in the yeast genome73.

Origins of ribonucleotides Replication by DNA polymerases generally requires a short RNA primer to initiate DNA synthesis. The DNA Pol α-primase is responsible for synthesizing a short primer to initiate replication, where the first 7-10 nucleotides are usually rNMPs. Due to the size of the nDNA, a large amount of RNA primers will be generated during DNA duplication, especially since a new Okazaki fragment must be initiated every 200 nucleotides on the lagging strand. The RNA primers are consecutive rNMPs left in the DNA and would be by far the most abundant source of incorporated ribonucleotides. However, these RNA primers are normally efficiently or partly removed during Okazaki fragment maturation81.

Another major souce of rNMPs in the DNA is the incorporation by the replicative DNA polymerases or from translesion DNA polymerases involved in DNA repair and gap filling73,82. Replicative DNA polymerases seem to have evolved to efficiently avoid ribonucleotide incorporation, since they discriminate efficiently against rNTPs.

Figure 8. Incorporated ribonucletides can cause strand break. (A) Comparison between a dNTP and a rNTP. The rNTP has a hydroxyl (OH) group on the second carbon. The four different bases are the same except for the thymine which is instead an uracil for the rNTP. (B) Under alkaline conditions, the OH-group of an incorporated rNMP can attack the sugar-phosphate backbone which result in a strand break.

Structural studies of the active site revealed that many DNA polymerases use a “steric gate” to discriminate between dNTPs and rNTPs (for example see73,83,84). This steric gate is usually a bulky amino acid side chain, such as phenylalanine or tyrosine, which clashes against the 2´-OH group on an incoming rNTP. The incoming rNTP will therefore be blocked from entering the active site, and dNTPs are favored as substrate. For translesion polymerases, the active site is often larger and therefore less selective against rNTPs compared to the replicative polymerases. Single rNMPs in nDNA could therefore also arise from the action of translesion polymerases.

In vitro experiments have shown that the rNMP insertion frequency varies considerably between DNA polymerases, and the incorporation pattern will therefore be different between DNA strands as well as between genomes85. For example, in S. cerevisiae the incorporation frequency for

the three replicative polymerases Pol α, δ and ε varied several folds, given that these enzymes incorporate one rNMP per 625, 5000 and 1250 nucleotides, respectively73. Structural data has made it possible to modify the active site of many DNA polymerases to create steric gate mutants. For example, in Pol ε, the modification of M644 has been used to generate two steric gate mutants with opposite effects: M644L which incorporates 3- fold less, and M644G which incorporates 11-fold more ribonucleotides compared to WT86.More recent work has also suggested that there could be more than just a steric barricade for the incoming rNTP. Studies of DNA polymerases from bacteria suggest that also a polar filter could prevent extensive rNTP incorporation87. dNTP pools and Ribonucleotide reductase Based on in vitro experiments it was apparent that the discrimination ability of the DNA polymerases is very high and the polymerases are therefore unlikely to use rNTPs as substrate. However, the incorporation frequency of the DNA polymerases is not only influenced by the intrinsic steric gate, but also by the dNTP and rNTP concentrations available during the replication process. Since the rNTPs are normally building blocks for RNA, they are needed in a high amount inside the cell. In both mammalian and yeast cells, dNTP concentrations were found to be much lower than rNTP concentrations (for example 36- to 190-fold lower in unsynchronized S. cerevisiae cells, depending on the rNTP-dNTP pair73,88,89). The high excess of rNTPs over dNTPs in the cell increases the probability that DNA polymerases incorporate rNTPs during replication. The ratio of rNTPs to dNTPs is especially high in non-proliferating cells, or outside of S-phase when dNTP levels are low.

The dNTPs are required during the S-phase of the cell cycle when the chromosomal DNA is duplicated. Additionally, dNTPs are required for DNA repair as well as replication of mtDNA even outside of the S-phase90. A constant dNTP supply is therefore needed for the cell to be able to both proliferate and maintain DNA integrity. However, it is crucial for the cell to have a strict regulation of dNTP levels, since altered dNTP pools are known to cause both genome instability and increased mutagenesis91-93. In mammalian cells, dNTPs are produced and regulated via two different pathways: the de novo pathway and the salvage pathway94.

In de novo dNTP synthesis, ribonucleoside diphosphates (NDPs) are reduced in the cytosol into deoxyribonucleoside diphosphates (dNDPs) by the ribonucleotide reductase (RNR) enzyme. The dNDPs are then phosphorylated into dNTPs95. The de novo pathway produces the majority of the dNTPs during the S-phase in proliferating cells when high dNTP pools are required for genome duplication. The peak in activity in S-phase is achieved through multiple levels of control, including the expression levels of the small subunit of RNR that peak in S-phase96. Also, some organisms such as S. cerevisiae, are completely dependent on the de novo pathway since they lack the salvage pathway. Furthermore, RNR is allosterically regulated to maintain a suitable overall concentration of dNTPs, as well as a balance between the four different dNTPs97.

In non-dividing cells when the activity of the de novo pathway is low, dNTPs are mainly produced by the salvage pathway that contributes to a low but constant level of dNTPs throughout the cell cycle. The salvage pathway utilizes deoxynucleosides (dNs) and deoxynucleotides derived either from the turnover of dNTPs or uptake from outside of the cell. The dNs are phosphorylated to dNTPs with the help of nucleoside and nucleotide kinases95. The salvage synthesis of dNTPs is balanced by the constant turnover (breakdown and build-up) of dNTPs by various enzymes, including the dNTP triphosphohydrolase SAMHD1 (Sterile alpha motif and HD-domain containing protein 1) that hydrolyzes dNTPs to dNs and thus limits cellular dNTP pools especially outside of S-phase98,99.

In contrast to the dNTPs, rNTP levels are kept relatively constant throughout the cell cycle. The ratio between dNTPs and rNTPs will therefore mainly depend on dNTP concentrations. Also, it implies that changed regulation of dNTP production might influence the ribonucleotide incorporation frequency in DNA in vivo.

Removal of incorporated ribonucleotides Many replicative DNA polymerases exhibit an exonuclease activity which proofread and efficiently removes incorrect bases. However, the DNA polymerases cannot or to a very low degree remove incorporated rNMPs using their proofreading ability84,100,101. To avoid the detrimental consequences from the high amount of incorporated rNMPs, efficient repair pathways are present in the nucleus.

Single rNMPs Single rNMPs incorporated into nuclear DNA are removed by the efficient Ribonucleotide Excision Repair (RER) pathway. The RER pathway is dependent on the RNase H2 enzyme which recognizes and cleaves one DNA strand on the 5´ side of a rNMP. The replicative Pol δ, together with the RFC clamp loader and PCNA, can displace the 5´-end strand, creating a flap containing the single 5´-terminal rNMP. The flap is resected by FEN1, the resulting gap filled by a DNA polymerase, and ligated by DNA ligase 1102,103.

Stretches of rNMPs The RNA primers formed during replication initiation contribute to stretches of several consecutive rNMPs in the genome. These RNA:DNA hybrids in the Okazaki fragments or R-loops can also to some extent be repaired by the RER pathway104,105. Stretches of a minimum of four consecutive rNMPs are also recognized and cleaved by the RNase H1 enzyme, contributing to their removal106,107.

Other ribonucleotide processing pathways Upon RER defects, other repair pathways can to some extent complement the defect in rNMP repair in the nucleus. For instance, in budding yeast ribonucleotides can in absence of RER be processed by Topoisomerase 1 (Top1)108,109. However, the incision made by Top1 generates a cyclic 2´- 3´-phosphate end which requires the action of additional repair factors to be removed. Furthermore, Top1-mediated rNMP removal can only remove parts of the incorporated rNMPs in the genome and frequent formation of double stranded breaks is detected in RER-deficient cells110, indicating that the RER-pathway would be the preferred repair pathway for single rNMPs. Finally, in RER deficient background, 2-5 base pair deletions increase in repeat sequences when introducing a Pol 2 steric gate mutant (Pol 2 M644G) with increased rNMP incorporation frequency86. Several of these deletions are not caused by the rNMPs themselves, but rather the Top1-dependent repair involving end processing and gap-filling111. The Top1-mediated rNMP removal thus increase the genome instability86,102,109.

Consequences of single ribonucleotide incorporation in nuclear DNA It is not only the reactive nature of the 2´-OH group of rNMPs that contributes to harmful consequences for the genome; products of wrongly processed rNMPs can also lead to genomic instability and replication stress. Studies in different organisms have demonstrated why a functional RNase H2 and RER pathway are important and conserved between species.

Direct effects on DNA replication It is clear from many studies that incorporated rNMPs can directly cause adverse effects during DNA replication. The cause of this increased replication stress can have several origins. Incomplete rNMP repair can result in nicks or rNMPs in the template can hinder the progress of DNA polymerases and the replisome, preventing timely completion of DNA replication (for some examples see73,76,84). The replicative polymerases have also varied abilities to bypass rNMP, which suggest that the effect of incorporated rNMPs is also dependent on the type of polymerase encountering it112.

Human and mouse In humans, loss-of-function mutations in the subunits of RNase H2 are associated with the inflammatory disease Aicardi-Goutières syndrome113. The patients have a constitutive interferon alpha (IFNα)-mediated inflammatory response connected to problems in RNA:DNA processing. Complete RNase H2 deficiency has not been reported in patients and homozygous RNase H2 knock-out mice died during embryonic development78, demonstrating that RNase H2 is essential. RNase H2 knockout embryos exhibited large chromosomal rearrangements, indicating that rNMPs might lead to double strand DNA breaks. The mice also showed signs of replication stress such as reduced cell proliferation and activation of a p53-dependent DNA damage response77,78.

Yeast In contrast to the knockout mice, RER-deficient yeast is viable. However, they show both genomic instability and checkpoint activation86. The exact mechanism behind rNMP-induced genomic instability is only partly understood but the specific consequences are many. As described previously, yeast strains with RER deficiencies have deletions on the nDNA when introducing a steric gate polymerase86. Also, signs of replication stress were observed such as activated S-phase checkpoint and

accumulation in S-phase. This resulted in reduced growth as well as high sensitivity to replication inhibitors86.

Bacteria The E. coli replication machinery is inhibited by embedded rNMPs as well as a high rNTP/dNTP ratio during replication. The E. coli replicative polymerase III incorporates one rNMP per 2300 nucleotides76. They are mainly removed by the RNase HII enzyme that, in contrast to mammalian RNase H2, only consists of a single subunit. Also, other repair pathways are likely to serve as back-up for RNase HII, such as the NER pathway114,115.

Incorporated ribonucleotides with positive functions Although incorporated rNMPs lead to several negative consequences, later studies have also suggested some beneficial functions of single rNMPs in the DNA. For instance, they may function as an imprint or mark to guide the mismatch repair system to the newly-synthesized strand116,117. Additionally, ribonucleotides can be used during DNA repair such as gap- filling during non-homologous end joining at double strand breaks82. Repair DNA polymerases can also incorporate ribonucleotides opposite oxidative damage such as 8-oxo-G, which might function as a fast way for the polymerase to bypass the damage and rescue stalled replication (for example see118).

In summary, the number of ribonucleotides found in the DNA is dependent on several factors. The intrinsic ability of the replicative and translesion DNA polymerases to discriminate between dNTPs and rNTPs during replication will determine the baseline ribonucleotide incorporation frequency. The incorporation frequency of the polymerases will be influenced by the dNTP and rNTP levels present during replication. Furthermore, once the ribonucleotides are incorporated, the removal pathways with RNase H2 and Topoisomerases I, can remove the rNMPs to a certain degree. In the presence of a functional RER pathway in the nucleus, nDNA will be nearly free from incorporated ribonucleotides after the DNA duplication is completed.

Detection of ribonucleotide incorporation in DNA The 2´-OH group on rNMPs can be exploited in detection of rNMP incorporation in DNA. This 2´-OH group is very sensitive to alkaline treatment, such as high concentration of sodium hydroxide (NaOH) or

potassium hydroxide (KOH). Upon alkaline treatment, the 2´-OH group undergoes hydrolysis which generates a nick on the DNA strand via a nucleophilic attach on the phosphate backbone (Fig 8). Alternatively, nicking at the position of the rNMP can be achieved through treatment with RNase H2. Both types of treatment can thus be used to generate single-strand DNA breaks at the position of the rNMP and visualize the rNMP abundancy in the DNA.

For long, detection of rNMPs was mainly achieved through gel-based methods such as Southern blot. Later, four laboratories developed different sequencing methods to be able to map the exact positions of rNMPs both in the nDNA and mtDNA119-122. One of the sequencing methods, called Hydrolytic End sequencing (HydEn-Seq), was developed in the Kunkel laboratory which was used as a tool to nicely demonstrate the division of labor during nDNA replication119.

Ribonucleotide incorporation in mitochondrial DNA Interestingly, nDNA and mtDNA differ significantly in rNMP content. The nDNA is almost completely free of rNMPs, while mature mammalian mtDNA contains one rNMP per 500-900 nucleotides80,123. Incorporated rNMPs were first described in mammalian mtDNA in the early 70’s124. These rNMPs are evenly distributed over the mature mtDNA and are not remnants of replication primers or replication intermediates.

Similar to the nDNA, the rNMPs can originate from different processes in mtDNA. Stretches of rNMPs, such as RNA primers, are synthesized by the POLRMT during initiation of mtDNA replication at both OH and OL. These RNA primers are removed by the RNase H1 together with additional DNA-processing proteins30,125,126. Interestingly, loss of RNase H1 does not lead to severe problems for the nDNA, probably because compensatory effects from RNase H278. In contrast, mutations or complete knockout of RNase H1 leads to extensive mitochondrial phenotypes such as encephalomyopathy or embryonic lethality127,128.

Since high amounts of single rNMPs are present in mtDNA, it was suggested that the replicative Pol γ was incorporating more rNMPs compared to nDNA polymerases. Contradictory to this hypothesis, Pol γ was found to incorporate around one rNMP every 2000 nucleotides, much less than nDNA polymerases80,102. Instead, the difference in rNMP content between the two genomes is presumably due to the difference in rNMP

removal. As described previously, single rNMPs in nDNA are mainly removed by the RER pathway with the RNase H2 being the key enzyme. So far, RNase H2 has not been identified inside the mitochondria, which suggests that no efficient repair of rNMPs is present inside the mitochondria. The single rNMPs present in the mtDNA is therefore probably a result of incorporation by Pol γ and the absence of a proper repair. Even other mitochondrially-localised polymerases such as PrimPol or Pol β may contribute to rNMP incorporation in mtDNA24.

The mtDNA normally contains a certain level of single rNMPs, but several studies have identified increased rNMP content during mitochondrial disease. For example, fibroblast mtDNA from patients with mutations in either of the mitochondrial nucleoside kinases (thymidine kinase 2 TK2 or deoxyguanosine kinase DGUOK) had higher rNMP content than controls80. Also, mice with decreased dGTP and dTTP pools due to defects in the mitochondrial inner membrane protein MPV17 exhibit increased rGMP incorporation129. Oppositley, decreased rNMP incorporation in mtDNA was observed in mice with increased dNTP pools due to knockout of SAMHD1123. These studies suggest that changes in mitochondrial dNTP pools affect rNMP incorporation frequency in mtDNA. However, single rNMPs can be bypassed by Pol γ without any major problem130. It is therefore unlikely that increase of rNMPs in mtDNA would be as harmful as they are in the nDNA. Whether the increased rNMPs are the cause behind the observed mitochondrial disease or not, still needs to be investigated. Although the rNMPs in mtDNA were identified almost five decades ago, there are still many open questions regarding their importance or their effect on mtDNA replication, transcription and maintenance.

Aims of thesis

The present study aims to understand how DNA damage and replication stress affects mtDNA maintenance. More specifically, the study aims to investigate the pathological mechanisms and negative consequences on mtDNA replication when bypassing or stalling at DNA damage such as oxidated bases or ribonucleotides.

Previous studies have shown a translesion primase-polymerase, called PrimPol, to be present inside the mammalian mitochondria. This study aims to understand the precise role of PrimPol in mtDNA metabolism.

The specific aims are:

1) To investigate the influence of oxidative stress on the mtDNA replication replisome and the potential role of PrimPol as a translesion synthesis polymerase during oxidative stress (paper I).

2) To understand the function of PrimPol in mitochondria and in mtDNA damage tolerance (paper II).

3) To elucidate the ribonucleotide incorporation pattern of yeast mitochondrial DNA polymerase Mip1 and how dNTP pool imbalances change the ribonucleotide incorporation frequency in the yeast mitochondrial genome (paper III).

4) To understand by which mechanisms the mitochondrial replication machinery tolerates embedded ribonucleotides in the mtDNA (paper IV).

Fulfilling these aims provides valuable insight into what mitochondrial mechanisms are essential to preserve mtDNA integrity under DNA stress. As outlined here, this is a relevant question not only to understand the early development and progress of mitochondrial disorders but also for other more common conditions that involve loss of mtDNA integrity such as neurodegeneration.

Results and discussion

Part I – The role of PrimPol in mtDNA damage tolerance DNA replication is not a continuous process as the replication fork can slow down or pause. Protein-DNA complexes or impaired proteins can arrest the replication fork, as well as DNA secondary structures or DNA damage. The nDNA replication has several mechanisms to tolerate both secondary structures and DNA damage during the replication process. The first option is a translesion DNA polymerase which can synthesize across the DNA damage to carry on the replication fork progression. Another option is a DNA primase which can synthesize a new primer downstream of a secondary structure or DNA damage from which the replication machinery can continue the synthesis. One of the main proteins responsible for repriming and translesion synthesis (TLS) is the primase- polymerase PrimPol. Since PrimPol is localized both in the nucleus and mitochondria, both the repriming and TLS activity could potentially also be important for mitochondrial DNA (mtDNA) damage tolerance. In the first part of this study, we investigated the role of PrimPol during mtDNA maintenance and DNA damage control.

Mitochondrial replisome stalls at oxidative DNA damage (paper I)

Oxidative DNA lesions stalls human mitochondrial Pol γ in vitro A significant amount of ROS is generated during the oxidative phosphorylation process inside the mitochondria. These ROS can oxidize DNA and lead to DNA damage such as 8-oxo-G or AP. Although mtDNA is frequently exposed to oxidative stress, little is known about how this effects mtDNA replication. To investigate the effect of oxidative DNA lesions on the mtDNA replication machinery, we used recombinant purified proteins on synthetic DNA templates containing an 8-oxo-G or an AP site. We choose to perform the reactions using two different dNTP levels which we expect to be physiological relevant inside the mitochondria. A dNTP composition which we call “normal” resembles dNTP concentrations measured in cycling cells and a dNTP composition with “low” dNTP concentrations similar to resting cells. Both dNTP concentrations are relevant as mtDNA,

in contrast to nDNA, is also replicated outside the S-phase, when dNTP concentrations are significantly lower. In our experiments, human wild type (WT) Pol γ alone shows only moderate DNA synthesis stalling when encountering 8-oxo-G in the template at normal dNTP concentrations. On the other hand, a template containing an abasic site (AP) completely blocks Pol γ DNA synthesis. As expected, at low dNTP concentrations the WT Pol γ pausing increased at 8-oxo-G and no bypass of AP site was visible. These results are in line with previous findings52,131. To understand if the stalling was dependent on the proofreading ability of Pol γ, we also tested an exonuclease- deficient (D274A exo-) variant of Pol γ. The exo- Pol γ bypassed the 8- oxo-G with higher efficiency compared to WT and could also to some extent bypass the AP site, although with low efficiency when bypass is compared to the efficiency on a control DNA template. According to the SDM replication model, most of the ssDNA will be coated with the abundant mtSSB protein. However, in the in vitro experiments, addition of mtSSB did not influence the bypass efficiency of an AP site by Pol γ. To further address the bypass efficiency by the whole mtDNA replisome, we combined Pol γ with the mtDNA helicase Twinkle. We constructed a minicircle template with a primer containing an overhang suitable for Twinkle loading. This allows continuous Twinkle unwinding of the double-stranded DNA when it is coupled to DNA synthesis by Pol γ. MtSSB was added to coat the single stranded DNA formed during Twinkle induced DNA unwinding. We show that the minimal mtDNA replisome could, at normal dNTP levels, bypass 8-oxoG during rolling circle replication showing a moderate pause at the site of DNA damaged. At low dNTP levels, the bypass efficiency of the mtDNA replisome decreases. In contrast, the AP site completely blocked the mtDNA replisome at all reaction conditions tested.

PrimPol does not stimulate Pol γ to bypass oxidative damage Our data suggest that the mtDNA replisome will have difficulties to bypass oxidative damage in the mtDNA. A complete block of replication, as in the case of an AP site, would be deleterious for the mtDNA replication. This raised the question how the mtDNA replication could persist when the DNA is exposed to ROS. In the nucleus, translesion DNA polymerases lend a hand to the replication fork to ensure bypass of DNA damage. One such example is the translesion DNA primase/polymerase PrimPol, which has the ability to assist nDNA replication machinery to bypass oxidative

DNA damage. PrimPol also localizes to the mitochondria which raised the question if PrimPol could function as a translesion synthesis polymerase to assist the mtDNA replisome to overcome oxidative DNA damage. To test this, we purified recombinant PrimPol and optimized the reaction conditions. PrimPol alone is not very processive and does not synthesize longer stretches of DNA at low and normal dNTP concentrations with Mg2+ as a cofactor. We therefore included, in addition to normal and low dNTP levels, a “high” dNTP concentration (200 μM) when testing PrimPol DNA synthesis. We found that, at all of the dNTP concentrations tested, PrimPol can bypass the 8-oxo-G damage, which argues for a role in the bypass of oxidative DNA damage. Unexpectedly, when combining PrimPol and Pol γ, PrimPol did not improve the Pol γ bypass at 8-oxoG site. However, the primer usage was higher when adding both PrimPol and Pol γ, which indicates that Pol γ is able to extend DNA formed by PrimPol. In contrast, PrimPol could assist Pol γ at bypassing the abasic site at high dNTP concentrations. However, at lower and more physiological relevant dNTP concentrations the presence of PrimPol did not make Pol γ bypass of damaged bases more efficiently. We conclude that PrimPol cannot help Pol γ when stalled at 8-oxo-G, but can increase AP bypass at high dNTP levels. However, we argue that this is unlikely relevant in vivo since the high dNTP levels tested here is beyond levels measured in cells.

PrimPol and its interaction with other components of mtDNA replisome We observed that PrimPol alone did not possess a high DNA synthesis ability and speculated that PrimPol might need additional factors to increase the processivity. We therefore investigated potential interactions between PrimPol and other components of the mtDNA replisome. The Pol γ B subunit function as a processivity factor for the catalytic Pol γ A subunit, however Pol γ B did not increase PrimPol’s DNA synthesis efficiency and is unlikely to serve as a processivity factor for PrimPol. In line with previous findings61,132 we observed an inhibition of PrimPol DNA synthesis upon addition of mtSSB. It is possible that mtSSB is essential to prevent undesired PrimPol DNA synthesis or priming inside the mitochondria.

One of the major participants in mtDNA replication is the DNA helicase Twinkle. To test whether PrimPol and Twinkle interacts, we combined the

two proteins on a DNA substrate with 40 nt primer overhang to allow Twinkle loading on the substrate. Interestingly, at high dNTP concentration, an increase of PrimPol DNA synthesis could be observed by the addition of Twinkle. The stimulation was dependent on Twinkle loading to the DNA but the stimulatory effect was similar with non- damaged or damaged DNA substrates. The enhancement of PrimPol activity by Twinkle is therefore not specifically to increase PrimPol’s translesion activity. Also, we tested if PrimPol in context of the complete mtDNA replisome could support bypass of DNA damage. At high, probably non-physiologically relevant dNTP concentrations, PrimPol was able to increase the overall DNA synthesis of the mtDNA replication machinery. However, this increase in DNA synthesis was not specific for the DNA damage since it was also observed on undamaged DNA templates.

Proper mtDNA maintenance is essential to retain the cellular energy production. Damage to the DNA, e.g., oxidative damage tested in this study, contribute to both stalled replication and mtDNA instability. Although DNA repair of oxidative damaged nucleotides exist inside the mitochondria, some of the damaged bases will remain after mtDNA replication. As demonstrated here, PrimPol itself does not possess high processivity during DNA synthesis and demands high, not physiological relevant, dNTP levels to be able to help mtDNA replication over oxidative bases. We hypothesise that PrimPol needs assistance of other proteins to increase the DNA synthesizes processivity, this could potentially also increase the translesion synthesis ability of PrimPol. In agreement with this theory, another mitochondrial localized PrimPol processivity factor was identified, called PolDIP255,56. Whether PolDip2 stimulation of PrimPol is essential for mtDNA replication machinery still needs to be addressed.

Our results suggest that PrimPol does not work as a conventional translesion polymerase at oxidative damage in the mitochondria. However, several findings argue for a role of PrimPol during mtDNA replication. First, PrimPol was found to localize to the mitochondrial compartment52. However, it is still unclear how and when PrimPol gets recruited to the mitochondria since PrimPol lacks the conventional mitochondrial targeting signal. In the nucleus, PrimPol gets recruited to the damage site via RPA interaction after for example UV exposure59,65, and it is possible that most PrimPol only gets imported into the mitochondria during a specific type of mitochondrial or mtDNA replication stress. Second, we

and others have reported interactions between PrimPol and proteins involved in mtDNA replication52. In our study we show a PrimPol stimulation depending on the loading of Twinkle helicase to DNA. If this stimulation is important in vivo is still an open question.

Finally, even though PrimPol knock-out cells are viable, the cells show signs of mtDNA stress with decreased mtDNA copy number also these cells have more trouble to recovery mtDNA levels after mtDNA depletion inductions (e.g., ethidium bromide treatment)52. This suggest that, under normal circumstances, PrimPol is not required for functional mtDNA replication, but plays a key role when mtDNA replication is put under pressure from different types of stress. In this described study, we only focused on the polymerase activity of PrimPol. There is also the possibility that PrimPol, instead of functioning as a conventional translesion polymerase inside the mitochondria, rather works as a translesion primase. In paper II, we therefore studied the second reported activity of PrimPol; re-priming after replication stalling and the specific role of this PrimPol activity in mtDNA maintenance.

MtDNA replication is reinitiated by PrimPol after damage (paper II) One of the main causes of mitochondrial dysfunction is OXPHOS deficiency. The OXPHOS system is dependent on replication and transcription of the mtDNA. Stalling of the mtDNA replication machinery can lead to pathological mtDNA rearrangements, mtDNA depletion and in the end OXPHOS deficiency. The accumulation of arrested replication forks can have several origins; DNA synthesis can be blocked by DNA lesions or due to a defect in one of the proteins in the replication machinery133,134. Similarly, incorporation of chain-terminating nucleoside analogues, such as 2´-3´-dideoxycytidine (ddCTP), can block DNA replication since they lack the 3´-OH group and cannot be extended (Fig 9). The mtDNA is especially sensitive to ddCTP since the replicative Pol γ is the only polymerase in mammalian cells which efficiently incorporates ddCTP during replication. Also, the ddC-ends cannot be resolved by Pol γ since the exonuclease proofreading activity is unable to remove incorporated dideoxyribonucleoside monophosphates135. The nucleotide analogue ddCTP is therefore an ideal tool to specifically study stalling of the mtDNA replication fork without effecting the nDNA replication process.

The translesion synthesis PrimPol activity cannot overcome the ddC- induced stalling as PrimPol cannot extend a primer with a 5´ dideoxyribonucleoside monophosphate end. Instead, a repriming event is needed downstream of the nucleoside analogue to restart the DNA replication. Although PrimPol has both primase activity and localise to the mitochondria, no clear role for PrimPol during repriming of mtDNA has been demonstrated. Therefore, we examined the significance of PrimPol’s repriming activity during mtDNA replication arrest and restart.

Figure 9. Dideoxynucleotides can block DNA synthesis. (A) Comparison between deoxynucleotides (dNTPs) and dideoxynucleotides (ddNTPs); ddNTP lack the hydroxylgroup (OH) on the 3´ carbon. (B) Simplified overview of normal DNA synthesis of the newly synthesized strand. The 3´-OH group of the primer end performs a nucleophilic attack on the α-phosphate of the incoming dNTP. The pyrophosphate group leaves the incoming dNTP and a covalent phosphodiester bond is formed. The base of the incoming dNTP will base pair with corresponding base on the template strand. The polymerization proceeds in the 5´to 3´ direction. (C) If a ddNTP gets incorporated into the new strand (becoming ddNMP), a OH-group will be absent from the 3´- end and the DNA polymerization reaction will be impossible.

Repriming by PrimPol after stalled mtDNA replication To understand the specific role of PrimPol during mtDNA replication, we used PrimPol -/- knock out MEF cells. We isolated the mtDNA from both PrimPol +/+ and PrimPol -/- MEFs and analyzed the DNA using two- dimensional agarose gel electrophoresis. The DNA was separated according to both size and shape allowing us to analyze mtDNA replication intermediates. Under normal conditions, no difference in mtDNA content or replication intermediates was observed between the PrimPol +/+ and PrimPol -/- cells. This is in line with previous studies showing that cells lacking PrimPol is still viable and not essential under normal cellular conditions52.

Next, we treated the cells with ddC which is converted to ddCTP when taken up by the cells. Inside the cells, the ddCTP gets incorporated into the mtDNA by Pol γ and cause mtDNA replication stalling. As expected, a significant decrease in mtDNA copy was observed in both cell lines during ddC treatment. Interestingley, when removing the ddC treatment, the copy number recovery was much slower in PrimPol -/- compared to PrimPol proficient cells. A distinct difference in replication intermediates was also observed between the two cell lines; PrimPol +/+ MEFs showed a substantial increase of replication intermediates after ddC treatment, which was absent in PrimPol -/- MEFs. The most pronounced increase was seen for the so-called bubble arc which are proposed to represent replication intermediates that are double-stranded (DNA) in nature. This suggests that the ddC treatment leads to an increased replication initiation in the control region of the mtDNA. However, the PrimPol -/- MEFs did not show any increase of replication intermediates when treated with ddC. This suggest that the increased reinitiation of mtDNA replication in PrimPol +/+ MEFs after ddC treatment is dependent on PrimPol. To further address PrimPol’s involvement in mtDNA replication, we also generated stable inducible cell lines (Flp In T-REx HEK293) overexpressing the human variant of either the WT PrimPol or a primase dead mutant (PrimPol-CH). To exclude effect on the nDNA, we artificially targeted PrimPol to the mitochondria, by addition of a mitochondrial targeting signal to the coding sequence. When PrimPol WT was overexpressed, a distinct accumulation of double- stranded mtDNA intermediates was detected. When we overexpressed the primase-deficient PrimPol-CH, we did not observe the increase of double- stranded intermediates. This indicates that the increase of dsDNA intermediates is fully dependent on the primase activity of PrimPol.

We reconstructed the ddC treatment scenario in vitro using recombinant purified Pol γ and PrimPol on a circular single stranded DNA template. Due to limited processivity, PrimPol alone could not produce long DNA fragments when using a normal DNA primer, while Pol γ efficiently synthesised full-length 7 kb DNA products. When combining Pol γ with PrimPol, a significant increase of DNA products is observed, also when we including ddCTP in the reactions (Fig 10A). We also used a primer with ddC on the 3´-end which Pol γ is unable to extend. Addition of PrimPol restarted the replication of the ddC-primer by synthesizing a new primer downstream of the ddC containing primer. As the DNA fragments produces were distinct in size, PrimPol seems to synthesizes primers at preferred sequences. We investigated this preferred PrimPol priming sites using 5´-RACE. We found that the main priming site on the M13mp18 plasmid was sequence specific (5´-GTCC-3´) which was previously reported to be the preferred sequence for PrimPol priming52. Priming at this position was independent on the damaged primer, since it was also seen on templates lacking a DNA primer when adding Pol γ together with PrimPol (Fig 10B).

Figure 10. PrimPol can assist replication by Pol γ in the presence of ddCTP. (A) Recombinant purified human Pol γ and PrimPol were used for in vitro DNA synthesis on 7 kb single stranded M13mp18 template. PrimPol enhanced the DNA synthesis of Pol γ in the presence of ddCTP. (B) Pol γ is unable to synthesis DNA in the absence of a DNA primer. With PrimPol present in the reaction, 7 kb full-length products are visible. PrimPol synthesis de novo DNA primers which can be extended by Pol γ. (C) Schematic presentation of the preferred priming sequence (3´-GTCC-5´) of PrimPol on the M13mp18 plasmid.

The in vitro data is in line with our results seen in vivo that PrimPol can reinitiate mtDNA replication by repriming after DNA damage. Since PrimPol itself is not incorporating or inhibited by the free ddCTP, it can assist Pol γ by synthesising new DNA primers. The primers can then be utilized by Pol γ which will overcome the replication arrest and the DNA synthesis is rescued. However, the rescue is not permanent as long as ddCTP is present, since Pol γ can repeat the incorporation of ddCTP, leading to repeated replication stalling. The mtDNA copy number will therefore decrease even in the presence of PrimPol, which is seen in PrimPol +/+ MEFs. On the contrary, when ddCTP is removed from the medium, the recovery is enhanced by PrimPol to restore the mtDNA copy levels again.

PrimPol reinitiate the mtDNA replication after UV-induced damage The mtDNA stalling phenotype in combination with increased replication intermediates observed after ddC treatment has previously been reported after UV-B exposure136. We therefore next investigated PrimPol’s role in reinitiation of mtDNA replication after UV induced DNA damage. UV-B treatment of either PrimPol +/+ and PrimPol -/- MEFs did not affect the mtDNA copy number. Analysis of replication intermediates of PrimPol +/+ cells revealed increased replication initiation similar to the phenotype observed after ddC treatment. In the PrimPol -/- cells on the other hand, the increase of replication intermediates was absent. This suggest that UV treatment cause an increase of replication initiation which is dependent on the presence of PrimPol. How the replication initiation by PrimPol is regulated is still unknown, but from our results we suggest that it is not dependent on the mtDNA copy as we did not observe a decrease in copy number in UV treated cells.

The priming by PrimPol followed by restart of mtDNA replication downstream of the DNA damage, gives rise to new questions. As a new primer downstream of the damage will generate a gap between the damaged 3´-end and the new 5´-end of the synthesised primer. The gap needs to be repaired (filled) to retain the genome integrity. The repair could potentially be facilitated by several mechanisms. The gapped molecules could be turned over in a higher degree to remove damaged mtDNA molecules. In nDNA, gaps can be repaired using RAD51 mediated homologous recombination137. However, RAD51 mitochondrial localization and homologous recombination in mtDNA repair has

remained controversial and more studies are needed to address how these gaps are repaired inside the mitochondria.

The preferred priming sequence 3´-GTCC-5´ is relatively abundant on the mtDNA which in theory would give PrimPol sufficient positions to prime the mtDNA replication. However, in vivo, mtSSB will be present to coat the single stranded DNA and will likely prevent PrimPol from promiscuous priming on the mtDNA during normal circumstances. Our overexpression of PrimPol to abnormal artificially high levels might overcome the inhibition of mtSSB. Even though the relevance of overexpressed PrimPol can be discussed, we see a decrease of single stranded DNA intermediates when overexpressing PrimPol. This suggests that there is some PrimPol priming on the lagging strand of mtDNA.

Figure 11. PrimPol can rescue mtDNA replication after ddCTP-induced stalling. (A) In the presence of ddCTP, Pol γ incorporates ddCTP into the 3´-end of the new DNA strand. (B) Pol γ is unable to remove or extend from the 3´-ddC end and further replication is blocked. (C) PrimPol reprimes downstream of the lesion in a sequence dependent manner (3´-GTCC-5´). (D) Pol γ extends from the new DNA primer made by PrimPol and replication is restarted.

Even though both the primase and polymerase activity of PrimPol can restart replication upon DNA stress, the rescue comes with a price. First, the DNA synthesis of PrimPol is, as many other translesion polymerases, more error-prone compared to the replicative polymerases. Excess amount of PrimPol could therefore result in increased mutations or indels. Second, promiscuous priming by PrimPol could potentially lead to increased single stranded gap formation in the DNA. In vivo, it will therefore be important for the cell to regulate PrimPol activity to avoid harmful effects of PrimPol’s priming and polymerase activity.

In summary, we have studied the mechanistic details and the role of PrimPol in mtDNA maintenance. Under normal conditions, PrimPol is not required for efficient mtDNA replication. But when the mtDNA replication fork stalls at DNA damage, e.g., chain-terminating nucleoside analogues, PrimPol is needed for restart of replication. The restart is facilitated by the primase activity of PrimPol which synthesize new primers in a sequence dependent manner. Pol γ can extend these primers and the mtDNA replication process is restarted (Fig 11). The reinitiation by PrimPol is especially important for recovery of mtDNA copy number after DNA damage. Our findings suggest that PrimPol plays an important role in mtDNA maintenance and is needed to protect the mtDNA genome against various of types of DNA insults to avoid mtDNA depletion.

Part II – Ribonucleotide incorporation in mtDNA

Incorporation of ribonucleotides in yeast mtDNA (paper III) Ribonucleotides (rNMPs) incorporated into the DNA contribute significantly to genomic instability due to the reactive 2´-OH group. Incorporated rNMPs would be by far the most frequent DNA lesion present in the nuclear genome if the ribonucleotide excision repair (RER) pathway would not be present. The RER pathway removes the rNMPs with high efficiency, leaving almost no rNMPs in the nuclear genome102. In contrast, the RER pathway has not been described inside the mitochondria and rNMPs were found in mtDNA already in the early 1970s124. Later, other studies using sequencing methods confirmed rNMPs in both yeast and human mtDNA80,119-121. This implies that there is no efficient repair of incorporated rNMPs in mitochondria, but no strong evidence has so far been presented. The rNMP incorporation in yeast nDNA have been studied extensively73, but the incorporation of rNMPs in mtDNA is still partly unexplored.

To understand the incorporation frequency and repair of rNMPs into the two genomes, we used yeast strains with altered dNTP pools that either contained or lacked a functional RER pathway. With these yeast strains as a tool, we challenged the replication process and investigated the difference in rNMP incorporation between the nuclear and mitochondrial genome. dNTP pools determine the rNMP incorporation pattern in nuclear DNA In a normal S. cerevisiae WT strain, the four dNTPs are not present in equimolar concentrations. When Nick McElhinny with colleagues measured the dNTP levels in logarithmically growing cells, dTTP was the nucleotide with highest concentration, while dGTP was the lowest (dTTP > dATP > dCTP > dGTP73). A stable supply and balance between the four dNTPs are needed for both replication and DNA repair. The ribonucleotide reductase (RNR) is the key regulator of dNTP levels and is tightly regulated on multiple levels. This regulation ensures that dNTP pools are sufficient for both replication during S-phase and DNA repair throughout the cell cycle. Allosteric regulation of RNR determines the overall dNTP concentration and the dNTP pool balance in the cell. In budding yeast, the allosteric specificity site of RNR can be mutated to induce specific dNTP pool imbalances92. Imbalanced, decreased or even increased dNTP levels

are known to contribute to genomic instability91,92. Alterations in the dNTP levels also change the dNTP/rNTP ratio, which could potentially result in increased rNMP incorporation into the DNA.

To investigate the rNMP incorporation frequency during imbalanced dNTP pools, we used several S. cerevisiae strains with point mutations in the allosteric specificity site of RNR. More specifically, the point mutations generated amino acid substitutions in the Rnr1 subunit to give distinct dNTP pool imbalances (for summary of strains see Fig 12). To identify and map the location of the incorporated rNMPs in these yeast strains, we used a 5´-DNA end-mapping technique called HydEn-Seq119. The HydEn-Seq method takes advantage of the fact that incorporated rNMPs are sensitive to an alkaline environment. Alkaline treatment of the DNA will introduce strand breaks at the positions of incorporated rNMPs, leading to smaller DNA fragments which can be sequenced. The sequencing method allows us to identify where in the genome the rNMPs were located and measure the relative frequency of each of the four rNMPs in terms of what percentage (%) of the total incorporated rNMPs they represent.

The sequencing result revealed that strains with mutations in RNR and imbalanced dNTP pools did not have a changed rNMP incorporation pattern in their nDNA. Not even extremely imbalanced dNTP pools, for example those found in the rnr1-Q288A or rnr1-Y285A strains, lead to a different rNMP incorporation pattern compared to the wildtype RNR1 strain. Our results therefore highlight the extreme efficiency of the RER pathway in removing incorporated rNMPs from nDNA.

To be able to detect increased rNMP incorporation in nDNA, we crossed the RNR mutant strains with strains deficient in RER (rnh201Δ). The deletion of RNH201 did not appreciably change the dNTP pool imbalances in the RNR mutants. In contrast to the RNH201 strains, deletion of RNH201 and inactivation of RER resulted in altered rNMP frequency in the strains with imbalanced dNTP pools. Sequencing result showed an inverse relationship between the relative rNMP frequency to the concentration of the corresponding dNTP. For example, the rnr1-Y285A rnh201Δ strain that displayed a 20-fold increase in dCTP compared to RNR1 rnh201Δ, showed a decreased rCMP frequency in its nuclear genome (10 % of all rNMPs were rCMP, compared to 35% in RNR1 rnh201Δ).

Figure 12. Summary of S. cerevisiae strains with point mutations in Rnr1 used in this study (only RNH201+ strains are shown for simplicity). The dNTP pools of each strain were measured and normalized against rNTP pools. Data adapted from Wanrooij et al 2017 PNAS.

The same inverse relationship was observed for other RNR-mutant strains in the rnh201Δ background. However, the rnr1-D57N rnh201Δ with balanced but increased dNTP pools did not show a change in the relative rNMP frequency. Collectively, these results suggest that the rNMP incorporation in nDNA is dependent on the overall dNTP levels in the cell, but that the rNMPs are normally efficiently removed by RER. rNMPs in mtDNA are not repaired, but determined by overall dNTP levels Next, we compared the relative rNMP frequencies for the mtDNA. In contrast with the nDNA where rNMP frequencies were only found to be altered in RER-defective strains, the mtDNA rNMP frequencies in RNH201 and rnh201Δ strains were equally affected by dNTP pool imbalances. For example, the rnr1-Q288A and rnr1-Q288A rnh201Δ strains both have very low dCTP levels, which generated a higher rCMP incorporation frequency (see example in Fig 13). Also, both strains have a significant decrease of rGMP and rAMP in the mtDNA, which is explained by the elevated levels of dGTP and dATP. Therefore, similarly to the nDNA, the relative rNMP frequencies in mtDNA were inversely related to

the dNTP pool. When comparing the rNMP incorporation frequency between the nDNA and mtDNA, a positive correlation was seen for the strains without a functional RER pathway (rnh201Δ strains), suggesting that dNTP pool imbalances affect the rNMP content of both genomes to a similar extent. However, the positive correlation was only observed for the relative frequencies for rCMP, rAMP and rGMP, while the frequency for rUMP varied too little to draw any conclusions. As expected, no correlation was seen in the RER proficient strains, since the RER pathway would efficiently remove all incorporated rNMPs in the nDNA.

Figure 13. Comparison of rNMP incorporation frequency between RNR1 WT and rnr1- Q288A strains. (A) Summary of dNTP pools in RNR1 WT and rnr1-Q288A strain. The number above the bars indicate the fold change relative to RNR1 WT. For example, the dCTP level is dramatically decreased in rnr1-Q288A while the dGTP levels are 25-fold increased. The arrows indicate the trend of the change (decrease or increase). (B) Summary of relative rNMP incorporation frequencies in mtDNA of RNR1 WT and rnr1-Q288A strains. For the rnr1-Q288A strain, with the decreased dCTP level, show a higher incorporation of rCMP in the mtDNA. Reversely, the increased dGTP pool resulted in decreased rGMP incorporation. The data is adapted from Wanrooij et al 2017 PNAS.

The rnr1-D57N rnh201Δ strain with elevated but balanced dNTP pools did not have a different relative rNMP incorporation pattern in mtDNA compared with WT. The result could either indicate that the dNTP increase seen in rnr1-D57N rnh201Δ does not affect rNMP incorporation, or, alternatively, it could be explained by a limitation in the assay where rNMPs are only measured relative to each other. Since the dNTP pool is balanced in the rnr1-D57N rnh201Δ strain, the incorporation of all rNMPs could give the same relative pattern as in WT but at lower total levels. To be able to discern between these two possibilities and measure rNMP incorporation quantitatively, we linearized the mtDNA using PmeI restriction enzyme and used the defined ends to normalize the mtDNA rNMPs against the chromosome ends as an internal control. To further elevate the dNTP pools, we deleted two negative regulators of RNR: the transcriptional repressor Crt1 and the protein inhibitor Sml1, generating a triple mutant strain (rnr1-D57N crt1Δ sml1Δ).

As expected, the deletion of Crt1 and Sml1 increased the dNTP pools by 10- to 15-fold in the rnr1-D57N crt1Δ sml1Δ strain and reduced the absolute rNMP incorporation in the mtDNA. Also, the deletion of RNase H2 in this strain did not increase the rNMP incorporation, which is further evidence for no repair of rNMPs by RER in the mitochondria. For nDNA, the rNMPs increased only in the rnh201Δ strain and no change was seen in the rnr1-D57N crt1Δ sml1Δ strain with higher dNTP pools. The HydEn- Seq data was further supported by Southern blot analysis of alkaline- or RNase H2-treated mtDNA isolated from the different strains.

Overall, our data suggest that the imbalance in overall dNTP pools caused by these RNR mutations are distributed equally between the nucleus and mitochondria. This supports the idea that the transport of cytosolic dNTP pools into the mitochondria is probably not regulated and is not able to modulate the balance of the mitochondrial dNTP pool. Also, we conclude that the rNMP incorporation in mtDNA is not affected by the absence of RNase H2 and an efficient rNMP repair pathway is missing inside the mitochondria. We cannot exclude that other processes would influence or repair incorporated rNMPs to a limited extent, but the RER pathway is not repairing rNMPs in yeast mtDNA.

Yeast mtDNA polymerase incorporates high levels of rNMPs into mtDNA Interestingly, the HydEn-Seq result demonstrated a clear difference in the relative incorporation frequencies of the four rNMPs between the nuclear and mitochondrial DNA. In the nDNA, rCMP and rAMP were most frequent, while the mtDNA had most rAMP and rGMP. We normalized the rNMP frequencies against the base composition of each genome, but the mtDNA and the nuclear genome still differed by having most of rGMP or rCMP, respectively. The disparity between the two genomes could arise from different dNTP pools in the two compartments. However, it is challenging to accurately measure the mitochondrial dNTP pool balance with current techniques, so we could not address this issue directly. Another possibility is that the replicative polymerases in nucleus and mitochondria have different discrimination against the ribonucleotides, and will therefore leave different incorporation patterns in the two genomes.

To understand the incorporation pattern in the mtDNA we expressed and purified recombinant yeast mtDNA polymerase Mip1 from E. coli. We first measured the overall rNMP incorporation frequency of WT and exonuclease-deficient (D171A E173A; exo-) Mip1 in a primer extension assay on a 3 kb single-stranded template. To keep the reactions as close to in vivo conditions as possible, we used physiologically relevant dNTP concentrations 73. Reaction products with and without rNTPs were alkaline treated using NaOH. Products containing rNMPs will hydrolyze during the alkaline treatment and the resulting smaller fragments will migrate faster on gel. The shift visible for the rNTP-containing reactions indicates that Mip1 incorporates rNMPs during the replication reaction. The reactions with only dNTPs present are not sensitive to alkaline treatment (compare lane 1-2 with 3-4 in Fig 14).

The median length of the products was determined for all reactions and overall rNTP incorporation frequency was calculated according to previous equation102. During the in vitro replication assay Mip1 incorporated approximately 1 rNMP per 600 nucleotides. In comparison, the yeast DNA polymerase ε and δ incorporate 1 rNMP per 640 and 720 nt, respectively, in the same experimental set up102. Mip1 exo- showed the same incorporation frequency as the WT protein, which suggest that the exonuclease activity does not proofread and remove any of the incorporated rNMPs. Previous studies of yeast replicative polymerases also report no or very inefficient proofreading of rNMPs84,101.

Figure 14. Stable rNMP incorporation by Mip1 7 kb. (A) Schematic overview of replication assay with yeast mitochondrial DNA polymerase Mip1 on 7 kb single stranded M13mp18 template. Radiolabeled α 32P-dCTP was included to be able to follow the reactions. The full-length products are alkaline-treated using NaOH to measure rNMP incorporation frequency. (B) Replication assay was performed using recombinant purified Mip1 wild type (WT) or exonuclease deficient (exo-). Reactions with or without rNTPs were treated with NaOH and separated on a denaturing alkaline agarose gel. The median lengths of each reaction were measured and the overall rNMP incorporation frequency was calculated for Mip1 WT and exo-.

To go into more detail of Mip1’s discrimination ability, we performed a single nucleotide insertion assay using primer extension reactions with Mip1 exo− on a shorter template (a 15-mer primer annealed to a 34-mer template). The single dNTP or rNTP corresponding to the +1 base in the template was added at physiological concentrations (Fig 15A and also73). The products were separated on a denaturing polyacrylamide gel to visualize the incorporation of the dNTP or its corresponding rNTP. Products with dNMP inserted will migrate faster compared with products containing rNMP (Fig 15B). The discrimination factor was determined by quantifying the band intensities for each dNTP or rNTP according to the equation previously described (Equation 173).

The single-nucleotide insertion assay revealed that Mip1 discriminates best against rATP, but least against rCTP. The incorporation pattern which we observed for Mip1 does therefore not explain the high rGMP levels in the mtDNA of the yeast strains, since Mip1 did not incorporate particularly more rGTP than the three other rNTPs. Whether this is due to a limitation in the discrimination assay or explained by a lower dGTP pool inside the mitochondria in vivo, still remains to be addressed.

Figure 15. Single nucleotide discrimination assay. (A) Discrimination against rNMPs is measured on a 34 nt template with a 15 nt primer annealed. A single dNTP or the complementary rNTP is added to the reaction in physiological relevant concentrations. The DNA polymerase will synthesis one nucleotide (+1) during the reaction. Figure adapted from Nick McElhinny et al 2010 PNAS. (B) Example of result from single nucleotide discrimination assay. The rNMP (ribo) product has reduced mobility in the gel compared with the corresponding dNMP (deoxy) product. A weak deoxy product can be visible in the rNTP-containing reactions due to impurities of the added rNTP. The example gel was performed by JME Forslund.

% Product extended Conc NTP Conc Pol Time !"# X X !"# X !"# % Product extended$!"# Conc dNTP Conc Pol$!"# Time$!"#

Equation 1. Calculation of rNMP discrimination factor from single nucleotide discrimination assay. The percentage (%) of extended product was measured for dNTP- and NTP-containing reactions. The discrimination factor was calculated following the equation which includes the percentage of the extended products, the ratio between dNTP and rNTP concentration, the ratio between of polymerase concentrations and the reaction times used.

We conclude that during the replication, Mip1 incorporates ribonucleotides to a similar extent as the yeast nuclear replicative polymerases, but the rNMPs are not efficiently removed as they are from nDNA. Without a functional repair pathway, the mtDNA would therefore contain more rNMPs compared to the nDNA. This observation could contribute to our understanding of why some patients with mutations in enzymes responsible for regulating the total dNTP pool only show mitochondrial phenotypes138. For example, while mutations in TK2 only

affect the dNTP pool inside the mitochondria, mutations in the small subunit of RNR modulate the total dNTP pools in the cell. However, mutations in either of these enzymes lead to mitochondrial dysfunction without having a major impact on nDNA. One explanation could be that the diseases often target postmitotic tissues where nDNA is no longer replicated, while mtDNA replicate outside of the S-phase. Nevertheless, dNTPs are still expected to be required during repair of nDNA even in the postmitotic tissue and not only required for mtDNA maintenance.

The increased rNMP incorporation in mtDNA from these patients could explain the distinct mitochondrial phenotypes observed since it is hypothesized that the increased rNMP levels could potentially lead to mtDNA instability. In line with this possibility, petite frequency, an indicator of mtDNA instability, was decreased in strains with increased dNTP pools and reduced rNMP incorporation. Also, rNMP sequencing of primary fibroblasts from patients with imbalanced dNTP pools revealed an increase of rNMPs in the mtDNA80. In contrast, removal of rNMPs did not appreciably improve the mtDNA stability in mice123 which argues for a more complicated pathological mechanism in patients with imbalanced dNTP pools. Whether it is decreased dNTP pools affecting the mtDNA replication or an increase of embedded rNMPs in mtDNA that cause mtDNA stability still remains unclear.

In summary, rNMPs are constantly incorporated by the replicative polymerases in both nucleus and mitochondria during the normal replication process. Our data show that the size of the dNTP pool influences the rNMP incorporation pattern and that imbalanced dNTP pools change the amount of rNMPs in the DNA. The incorporation frequency of a specific ribonucleotide is inversely correlated with the concentration of the corresponding dNTP. In the nucleus, we show that decreased dNTP pools do not increase rNMP incorporation since the RER pathway is efficiently removing the rNMPs. In contrast, we show that the mitochondria lack any repair pathway and decreased dNTP pools leads to increased rNMPs in mtDNA (Fig 16). The increased rNMP levels in mtDNA argue for an equal transmission of total cellular dNTP pool imbalances between the nucleus and the mitochondria. Taken together, this could partly explain why total cellular dNTP pool imbalances can cause only mtDNA defects in certain human diseases.

Figure 16. The influence of imbalanced dNTP pools on rNMP incorporation in nuclear and mitochondrial DNA. During normal DNA replication, a certain number of rNMP gets incorporated based on the discrimination ability of the replicative DNA polymerases. Changes in dNTP pools will lead to changed rNMP incorporation frequency in an inverse matter. With increased dNTP levels, the rNMP incorporation of the DNA polymerases will decrease. Decrease of one or several dNTPs will increase the rNMP incorporation in the DNA. In nDNA, the incorporated rNMPs will be repaired by the highly efficient ribonucleotide excision repair (RER) pathway and no rNMPs will be left. In contrast, the mitochondrial DNA (mtDNA) lack a efficient repair pathway and the increased rNMP incorporation due to decreased dNTP levels will result in a higher rNMP amount in the mtDNA compared to nDNA.

Incorporation of ribonucleotides by human mtDNA replication machinery (paper IV) During the DNA duplication process ribonucleotides are constantly incorporated into the DNA by the DNA replicative polymerases. The nDNA is protected from the harmful effects by the RER pathway which removes the incorporated ribonucleotides. Failure to remove the incorporated ribonucleotides leads to genome instability and has been associated with the human disease Aicardi-Goutières syndrome113. In contrast, the mammalian mtDNA is rich in ribonucleotides and so far, no ribonucleotide removal repair pathway has been described inside the mitochondria. The mtDNA replication machinery will therefore encounter embedded rNMPs during replication and needs to be more tolerant to ribonucleotides in the template compared to nuclear replicative DNA polymerases. The exact mechanism by which these embedded rNMPs are tolerated inside the mitochondria is still not fully understood. We therefore aimed to investigate the molecular mechanism by which the human

mtDNA replication machinery deals with embedded rNMPs in the DNA template during the replication process.

Human Pol γ can efficiently bypass single rNMPs with high fidelity Several studies reported on rNMPs incorporation in mtDNA in different species80,124,139. To visualize the rNMP content in mtDNA, we first isolated mtDNA from mouse liver using a sucrose gradient followed by phenol- extraction. To confirm the presence of ribonucleotides in mammalian mtDNA, we treated the mtDNA with NaOH which fragments leaving behind DNA nicks at the rNMP positions. Using Southern blot, we detected the mtDNA fragments and could visualize the rNMP incorporation in mouse liver mtDNA (see Fig 17A for example). We probed both strands and six different genomic locations and concluded that mouse mtDNA is rich in embedded rNMPs, independent of mtDNA region. The embedded rNMPs were evenly distribution, in line with previous findings80.

The experimental confirmation that mtDNA contains several embedded rNMPs per DNA molecule, lead us to investigate the impact of these rNMPs on mtDNA replication. During one round of replication, the human replicative mtDNA polymerase Pol γ, will encounter approximately one rNMP every per 500-900 nucleotides. Pol γ was previously shown to have the ability to bypass single rNMPs in the template, however, in these experiments only the exonuclease deficient (exo-) Pol γ was tested130. We therefore purified the wild WT Pol γ and performed primer extension reactions on synthetic DNA templates that contained single rNMPs. WT Pol γ did not stall and did not have reduced polymerization activity on any of the four rNMP (rA, rG, rC, rU) templates tested when compared with the control dNMP containing templates.

Since we did not see any rNMP inhibition on Pol γ activity, we asked if the 2.5-fold protein excess could mask a mild stalling or reduction in synthesis. To challenge the polymerase, we instead performed the assay with a large excess of DNA template over polymerase, to create the so- called “single hit condition”. The DNA products observed in single hit conditions arise from one round of synthesis, as the probability that the polymerase rebinds to the same template molecule is neglectable due to the excess of DNA template.

Figure 17. (A) Southern blot analysis of alkaline-treated mtDNA from mouse liver. The linearized full-length mouse mtDNA runs at approx. 16 kb in the denaturing agarose gel. Upon alkaline treatment, the mtDNA gets hydrolyzed at the positions of rNMPs and the mtDNA is separated into shorter fragments, indicating the presence of rNMPs in the DNA. (B) Schematic view of the DNA substrate used in rNMP bypass assays. A fluorescence tag (TET) is attached to the 5´-end of the 25 nt primer. The primer is annealed to either a dNMP (control) containing template strand or a template strand containing one or several consecutive rNMPs. (C) Primer extension reaction by DNA polymerase γ on template containing either dGMP template or template containing one up to three rNMPs. Pol γ can bypass one rNMP with same efficiency as for the control template (dGMP). The bypass efficiency decreases with increasing numbers of rNMPs in the template.

The termination probability was calculated using the band intensities in positions -2 up to +4 around the rNMP (rNMP position 0). Under single hit condition, only a very moderate increase of termination probability was observed for Pol γ at position -2 and -1 relative to the rNMP. To address the fidelity of Pol γ bypass of rNMPs, we also sequenced the products from the primer extension assay. Sequencing showed that Pol γ has high fidelity on the rNMP containing template and the correct base is inserted opposite rNMPs as often as on a control dNMP containing DNA template.

Taken together, we conclude that Pol γ can bypass single rNMPs with high efficiency (see fig 17B compare dG and rG template). This is in line with previous bypass studies of the Pol γ exo- variant as well with the reported

reverse transcriptase activity of Pol γ130,140. Interestingly, the measured Pol γ termination probability around the rNMP was lower compared to the termination probability of nuclear replicative polymerases84,100,112. This suggest that Pol γ, unlike nuclear replicative polymerases, has evolved and adapted to the rNMP-rich mtDNA and can efficiently bypass rNMPs with high fidelity. It is therefore hard to imagine that a single rNMPs, incorporated from Pol γ itself, would serve as an obstacle for mtDNA replication and cause any substantial replication stalling or mutagenesis. It needs to be mentioned, however, that a single rNMPs could still be a source of genomic instability due to the reactive OH group. In contrast to single rNMPs, consecutive rNMPs in the DNA were observed by us and others130 to cause Pol γ stalling (Fig 17B). A stretch of incorporated rNMPs could originate from unremoved RNA primers which can happen when there is a deficiency in RNase H1. Consecutive rNMPs can therefore be expected to be more harmful to the mtDNA replication machinery compared to the single rNMPs incorporated by Pol γ.

Human Pol γ is exceptionally good at discriminating against ribonucleotides Since the majority of ribonucleotides in mtDNA are not consecutive rNMPs, but single rNMPs, Pol γ is thought to be the main source for these embedded ribonucleotides. In vitro, on a short 70 nucleotide template, Pol γ incorporates approximately one rNMP per 2000 dNMP bases80. However, the use of such short DNA template could potentially influence the incorporation frequency since it has a defined sequence and others have shown that the sequence context can affect the incorporation frequency86. To elucidate the incorporation frequency of Pol γ without this sequence bias, we measured the stable ribonucleotide incorporation frequency on a 7 kb long DNA template. To mimic the in vivo situation, we used two different dNTP concentrations which resembles conditions at different stages of the cell cycle; “normal” and “low”89,141,142. The “normal” dNTP concentrations mimic the levels in S-phase, while “low” dNTP concentrations resemble the dNTP levels outside the S-phase or non- dividing cells. As an experimental control we included the S. cerevisiae nuclear replicative DNA polymerase δ which had already previously been tested in the same experimental set up102.

We performed primer extension reactions on the 7 kb M13 ssDNA template in the presences or absence of rNTPs. The DNA products were treated with NaOH and separated on a denaturing alkaline agarose gel. In

the absence of rNTPs, Pol γ synthesized long products which were not affected by the alkaline treatment. In contrast, addition of rNTPs generated DNA products sensitive to alkaline treatment and resulting in shorter DNA fragments, which indicates incorporation of rNMPs. The median length of the products was quantified and the incorporation frequency was calculated as described previously102. At “normal” dNTP levels, Pol γ incorporates 1 rNMP per 2300 nt which is close to the incorporation frequency measured using a short template (1 rNMP per 2000 nt80). At “low” dNTP levels the incorporation increased to 1 rNMP per 1400 nt, which confirms previous observation in yeast that decreased dNTP levels will increase the rNMP incorporation by DNA polymerases143. In comparison, Pol δ incorporated rNMPs with a higher frequency, 1 rNMP per 650 and 440 nt at “normal” and “low” dNTP levels, respectively. Interestingly, the yeast mtDNA homolog Mip1 incorporates far more rNMPs than human Pol γ (compare 1 rNMP per 600 nt with 1 rNMP per 2300 nt143). We conclude that human Pol γ can discriminate against rNTPs to a higher extent than yeast DNA polymerases.

During normal replication of double-stranded mtDNA, Pol γ is assisted by the mtDNA helicase Twinkle and the mitochondrial single stranded binding protein mtSSB. We therefore investigated if Twinkle and mtSSB effect Pol γ rNMP incorporation. For this purpose, we used a single- stranded mini-circle DNA template for continuous unwinding by Twinkle coupled to leading-strand replication by Pol γ. Similar to the incorporation on the 7 kb template, Pol γ inserted 1 rNMP per 2200 and 1600 nt at “normal” and “low”, respectively. We conclude that Pol γ’s ability to incorporate rNMPs is not influenced by the other mitochondrial replication factors tested here. Contradictory, the rNMP levels measured in vivo were 2- to 5-fold higher than the observed rNMP frequency in vitro80. The dissimilarity between the in vivo and in vitro rNMP frequencies could be explained by several things. The estimated mitochondrial dNTP concentrations could be lower than expected, or alternatively, the estimated rNTP concentrations could be higher than measured. Another explanation could be that some of the incorporated rNMPs arise from other DNA polymerases present in the mitochondria. In fact, two other polymerases were recently found inside the mitochondria; the DNA polymerase β and PrimPol which could potentially contribute to the rNMP frequency in vivo52,144.

Incorporated rNMPs are not repaired by the exonuclease activity of Pol γ The unusually low ribonucleotide incorporation frequency observed for Pol γ could be explained by the intrinsic Pol γ proofreading ability. We therefore purified the exonuclease deficient (exo-) Pol γ and compared the incorporation frequency with WT Pol γ. In the primer extension assay, no difference was seen between the WT and exo- Pol γ. This indicated that the exonuclease activity of Pol γ cannot repair incorporated rNMPs during replication, which is in agreement with our previous findings of the yeast mitochondrial Mip1. We confirmed the in vivo relevance of this finding by isolating mtDNA from liver of WT PolgA +/+ mice and proofreading deficient mice PolgA D275A/D275A. Alkaline treatment of the mtDNA did not reveal any difference in rNMP content between the PolgA +/+ and PolgA D275A/D275A mice. The in vivo and in vitro data are therefore consistent and we conclude that it is unlikely that the exonuclease activity of Pol γ effects the amount of rNMPs in mtDNA.

Free rNTPs reduce the replication speed of Pol γ Overall, DNA synthesis is not impaired when Pol γ encounters rNMPs in the DNA template. However, we detected a reduction in replication speed when reactions were performed in the presence of free rNTPs. Especially when low dNTP concentrations were used in the reaction, Pol γ showed a strong free-rNTPs dependent inhibited of DNA synthesis. In the absence of rNTPs pol γ synthesizes detectable full-length products (3 kb) after 30 min, while in the presence of rNTPs these full-length DNA products were only visible after 90 min. Interestingly, the exo- Pol γ did not show any signs of free-rNTPs dependent inhibition at the time points tested. We hypothesis that the free rNTPs present forces WT Pol γ into a polymerase- exonuclease idling mode which would slow down the replication speed.

Changed rNTP/dNTP ratio reduce the replication speed of Pol γ We further investigated the inhibition of Pol γ at different dNTP concentrations and a constant level of physiological relevant rNTP levels. The free-rNTPs dependent inhibition of Pol γ was not seen at high dNTP concentrations, however, was pronounced when low dNTP levels were used in the reaction. Since an imbalance in the dNTP pools is known to affect mtDNA replication, we further challenged Pol γ by lowering all four dNTPs individually. The limitation of one dNTP led to decreased replication, especially in the presence of rNTPs. Lowering pyrimidines concentrations (dTTP, dCTP) had a stronger effect compared to reducing the purine concentrations (dATP, dGTP). Taken together, this suggest that

limiting one dNTP could enhance the free-rNTPs dependent Pol γ stalling. The proper rNTP/dNTP ratio is therefore crucial for efficient replication by Pol γ.

While rNMPs in the nDNA are efficiently repaired, the mitochondria lack an efficient repair pathway and the mtDNA is therefore rich in single rNMPs. So far it has not been fully understood how the mtDNA replication machinery can deal with these rNMPs. Interestingly, a 3- to 4-fold increase of single rNMPs was reported in fibroblasts derived from patients with imbalanced mitochondrial dNTP pools80. It was therefore thought that the increased rNMPs would be part of the pathogenic mechanism in these patients. However, our results show that Pol γ can replicate passed rNMPs in the template with high efficiency and fidelity, which argues against a harmful effect of the levels of rNMPs identified in these patients. Instead, we demonstrate that Pol γ has an exceptional good discrimination ability against rNTPs, but are inhibited by high levels of free rNTPs. The inhibition was especially pronounced at low and imbalanced dNTP concentrations. Taken together, our in vitro findings suggest that embedded rNMPs in mtDNA will not lead to significant mtDNA stalling, but an imbalanced ratio between the free rNTPs and dNTPs could reduce the mtDNA replication speed. This could potentially be important during mitochondrial replication outside of the S-phase where the cellular dNTP levels are low. Also, in patients with imbalanced mitochondrial dNTP pools, for example in TK2 defective patients, an additional free-rNTP dependent inhibition of Pol γ could lead to mtDNA replication stalling, mtDNA depletion and potentially cause mtDNA disease.

Conclusions and future outlook

Part I Several recent studies have attempted to identify new proteins involved in mtDNA replication and repair. The nuclear primase and polymerase protein PrimPol was also found to be mitochondrially localized. In the first part of this thesis, I show results which suggest that PrimPol does not function as a conventional translesion polymerase inside mitochondria, in contrast to what has been suggested as a PrimPol-function in the nucleus. Instead, PrimPol uses its primase activity to assist the mtDNA replication machinery during replication stalling events, which could be caused by chain-termination analogues or UV-induced DNA damage. The restart of mtDNA replication after DNA damage can be crucial to avoid mtDNA depletion and mitochondrial dysfunction.

The data obtained in this PhD thesis and by others suggest that PrimPol does not participate in mtDNA replication during normal cellular conditions, since its contribution under these circumstances is likely to be limited by its low processivity and the requirement of high dNTP levels. However, further studies are needed to elucidate the importance of PrimPol during mtDNA replication, especially in the presence of the recently discovered PrimPol-processivity factor PolDip2, which also localizes to the mitochondria. Even though the composition of the core of mtDNA replication machinery is now well established, it is most probable that the role of additional proteins during mtDNA replication has remained undiscovered. Furthermore, additional research is needed to understand the molecular mechanisms of mtDNA maintenance during mtDNA replication stress.

Part II Ribonucleotides are occasionally incorporated into the genome by DNA polymerases. In the second part of this thesis, we demonstrate that the rNMP incorporation into DNA is influenced by the total cellular dNTP and rNTP pools. In the nDNA, these incorporated rNMPs are efficiently removed by the RER pathway, but we show that there is no efficient repair of rNMPs inside yeast mitochondria. Quantitative measurements of mitochondrial dNTP pools are still challenging, which hampers the understanding of mitochondrial disorders that arise from dNTP pool

imbalances. The rNMP incorporation measurements in mtDNA as we describe in our study could potentially be used in the future to provide estimates of mitochondrial dNTP pools.

In contrast to the nDNA polymerases, human Pol γ can bypass single rNMPs with high efficiency and is not inhibited by template rNMPs to the same extent as other polymerases. Pol γ might have adapted to the rNMP- rich nature of the mtDNA and can tolerate rNMPs to a certain level. Nevertheless, the presence of ribonucleotides in the DNA does not only affect the replicative polymerases. The reactive 2´-OH group of the rNMP could potentially cause DNA strand breaks and genomic instability. Also, it is still not known how other proteins, for example the transcription machinery, would deal with these rNMPs in the mtDNA. To address the specific effects of rNMPs in mtDNA, the rNMP levels need to be modulated without alteration of the cellular dNTP pools. This could potentially be achieved by modification of the Pol γ steric gate, to generate a mtDNA polymerase that incorporates more or less rNMPs than wildtype Pol γ. In this way, we could study the rNMP effect on the mtDNA replication and maintenance without secondary influences from impaired nDNA replication.

It is possible that ribonucleotides can serve as backup building blocks for DNA polymerases during DNA synthesis. This can be especially important for mtDNA, as replication inside the organelle also occurs outside S-phase when dNTP levels are relatively low. The use of ribonucleotides as alternative building blocks for DNA could avoid prolonged replication stalling and prevent replication fork collapse. This could in particular be useful for the S. cerevisiae mtDNA replication polymerase Mip1, which we show in paper III has low ribonucleotide discrimination ability. It might however not be relevant during mammalian mtDNA replication, as paper IV shows that the human Pol γ incorporated much fewer rNMPs compared to its yeast homolog Mip1. Also, the high discrimination ability of Pol γ inhibits the speed of DNA synthesis by Pol γ at higher rNTP/dNTP ratios. This could potentially contribute to the mtDNA depletion seen in patients with decreased dNTP pools, but a deeper understanding of the molecular mechanism is needed to confirm or discard this hypothesis.

Although ribonucleotides are generally considered a deleterious DNA lesion, incorporation of ribonucleotides does not necessarily have a

negative impact. Interestingly, recent studies show that incorporated ribonucleotides can also have an important function during DNA replication events. For example, ribonucleotides could regulate initiation of mating-type switching in S. pombe or function as signals during mismatch repair in S. cerevisiae116,117,145. In summary, my PhD project that demonstrates the lack of mitochondrial rNMP repair and the replication tolerance for rNMPs incorporated in mtDNA makes it tempting to speculate that rNMPs incorporated in mtDNA remain because of a yet-to- be-identified cellular function. However, recent work of Wanrooij PH with colleagues report very few rNMPs in mtDNA of SAMHD1-knock out mice123. Since the mice are healthy with less rNMPs-containing mtDNA, it is unlikely that the rNMPs have a cellular function under normal conditions. There is still a possibility that rNMPs have a function under certain stress conditions, such as inflammation or DNA damage, however a potential cellular role will remain hypothetical until a function has been identified.

Final conclusions At last, we conclude from this work that the mtDNA replication system possesses several mechanisms to deal with DNA lesions. This includes intrinsic mechanisms of Pol γ, such as a high discrimination ability against ribonucleotides and efficient bypass of ribonucleotides and oxidative damage in the template DNA. Also, additional proteins assist the core mtDNA replication machinery such as the primase-polymerase PrimPol. The results of this study extend the knowledge of the consequences that DNA lesions have on mtDNA synthesis and mtDNA integrity. These valuable insights provide further understanding of mtDNA maintenance, mitochondrial disease and could ultimately contribute to assist future therapeutical directions.

Acknowledgement

Since I probably don’t get the opportunity to arrange a big party and have a really long speech this crazy Covid-year, this is my chance to thank all the people who supported me during my PhD!

First of all, I would like to thank Sjoerd Wanrooij for everything! How do I even summarize my PhD and all wonderful years under your supervision? I think for you to fully understand the importance, I probably have to compare it to football. When I started in your lab, I had the opportunity to play in a team way beyond my talent which was terrifying. But over the years I learned that it is not a one-man show, there is no ideal player, no Zlatan that can win the match on his own. In the good and the bad times, doing a PhD in your lab has been all about team work. However, the team would not be winning without a great coach. Every time I have been one step too late or missed a shot, you always encouraged me to keep going. You celebrated my moments of joy and helped me move on from the moments I wish I could forget. The match is now coming to its end and I have one final shot before the referee blows the whistle. I am convinced that your supervision and support has helped me becoming a better player and develop my qualities so that the ball will finally hit the back of the net. Thank you for all your support!

Thank you Claes Gustafsson for being my opponent and taking your time to scrutinize my work, it means a lot to me! And thanks to all the committee members! I am looking forward to many interesting discussions.

Also, many thanks to my two co-supervisors; Erik Johansson and Nasim Sabouri. Thank you for all input on my work and advises over the years. Thanks to my two examinators Sven Carlsson and Anders Hofer for interesting discussions about my results.

Thanks to Paulina, my ribo-partner-in-crime; I will never forget the late nights purifying proteins together with you. I will forever be grateful for all the things I have learned from you, all our valuable discussions, all jokes and laughter we had together. You are a true inspiration!

To all the former and current members of S Wanrooij’s lab, it has been a pleasure to gotten to know all of you and spent time in the lab together!

Especially thanks to Sonja for all your help in the lab and interesting language lessons. Thanks to Mara for all your input, you are probably the smartest person I know! I hope your pasta forever stays away from ketchup! J Gorazd, my lab bench neighbor, I would not have managed my PhD without your technical and kinetic support! Kazu, thank for all your help and I wish you the best when you arrive back in Japan! Rafil, long time no see, but thanks for all the bets you started in the lab! May the force be with you J Annika, it is sad that you had to leave Umeå but thanks for the years together in the lab, the badminton sessions and the mitochondria-gingerbread baking! Xin, I will never forget the chloride gas experiments you did, see you hopefully soon at innebandy! And to the new people; Tran, Andi, Namrata and Mama - welcome to Umeå and I hope you will enjoy the time in the lab as much as I did. Thanks to all the other former members and students that have spent time in our lab.

I am very thankful for all the years I have spent at MedChem and all the people working here. It has been such an inspiring and friendly atmosphere! Many thanks to all the administrative staff that helped me over the years; Ingrid, Clas, Anna S, Jenny, Matilda, Erlinda and Christina. Ingrid, I hope that you really enjoy your retirement! Thanks to Elisabeth for all the help with dishes and autoclaves!

Anna Å, tack för att du välkomnade mig till MedChem för 9 (?!) år sedan under min Masterutbildning. Tack för allt värdefullt du lärt mig under dessa år, men framför allt tack för att du är min vän!

Thanks to all the people I have gotten to know at MedChem! Especially to all the members in Paulinas and Nasims groups for the joint group meetings we had together. Thanks to Anna Karin, Göran and Phong for always being so friendly and contributing to long chats! And thanks to Annelie for being such a great support in the beginning of my PhD- journey, we miss you in Umeå! To all the people who I played innebandy with during the last couple of years; it was intense and dangerous, but a lot of fun!

To all PhD fellows I had teaching with; some moments were very stressful, but we also had so much fun! A special thanks to you Lena for being such a big support in everything!

To all the collaborators who made this work possible – thank you all! Especially thanks to Jaakko Pohjoismäki, Rubén Torregrosa- Muñumer and all the people in Luis Blanco’s group. Also, many thanks to Henna Tyynismaa and her lab for the hospitality during my research visit!

Jag vill också passa på att tacka min gamla Biomedicin-vän Faranak. Tack för att du är en så underbar människa och gjorde alla jobbiga stunder under studietiden mycket lättare! Tack också till Jessica och Tove för underbart sällskap på Masterprogrammet. Utan er hade jag aldrig tagit mig så här långt. Och tack till Sandra för alla år på gymnasiet och tiden som rumskompisar i Umeå. Jag är evigt tacksam att ha dig som vän! Tack till alla andra vänner genom åren för ert stöd, ingen nämnd ingen glömd J

Tusen och åter tusen tack till min älskade familj! Tack mormor Ulla för att du alltid finns där och alltid tror på mig, du har en stor del i mitt hjärta! Mina underbara systrar Lisa, Anna och Elin – ni har inspirerat mig att gå min egen väg och ni har alltid stöttat mig i med- och motgång, jag vore ingenting utan er!! Mamma och pappa – tack för allt ni gjort för mig och för allt er stöd, tack för att ni är de bästa föräldrarna man kan ha! En massa kärlek till er! Marcus, jag kommer aldrig kunna tacka dig tillräckligt för allt du är för mig. Du är min klippa som jag lutar mig mot när det stormar, min stora trygghet i livet! Och till sist, Axel och Otto, ni har lärt mig vad som egentligen är viktigt i livet. Att leva i nuet tillsammans med er är det bästa som finns! Älskar er av hela mitt hjärta ♥

References

1 Kornberg, A. & Baker, T. A. DNA replication. 2nd edn, (W.H. Freeman, 1992). 2 Watson, J. D. & Crick, F. H. Molecular structure of nucleic acids; a structure for deoxyribose nucleic acid. Nature 171, 737-738, doi:10.1038/171737a0 (1953). 3 Kainuma-Kuroda, R. & Okazaki, R. Mechanism of DNA chain growth. XII. Asymmetry of replication of P2 phage DNA. J Mol Biol 94, 213-228, doi:10.1016/0022-2836(75)90079-0 (1975). 4 Pursell, Z. F., Isoz, I., Lundstrom, E. B., Johansson, E. & Kunkel, T. A. Yeast DNA polymerase epsilon participates in leading-strand DNA replication. Science 317, 127-130, doi:10.1126/science.1144067 (2007). 5 Nick McElhinny, S. A., Gordenin, D. A., Stith, C. M., Burgers, P. M. & Kunkel, T. A. Division of labor at the eukaryotic replication fork. Mol Cell 30, 137-144, doi:10.1016/j.molcel.2008.02.022 (2008). 6 Brenda, C. On spermatogenesis of vertebrates and higher invertebrates, Part II: The histogenesis of sperm. 393-398 (Veit & Comp., 1898). 7 Giacomello, M., Pyakurel, A., Glytsou, C. & Scorrano, L. The cell biology of mitochondrial membrane dynamics. Nat Rev Mol Cell Biol 21, 204-224, doi:10.1038/s41580-020-0210-7 (2020). 8 Williams, T. A. & Embley, T. M. Changing ideas about eukaryotic origins. Philos Trans R Soc Lond B Biol Sci 370, 20140318, doi:10.1098/rstb.2014.0318 (2015). 9 Kukat, C. et al. Super-resolution microscopy reveals that mammalian mitochondrial nucleoids have a uniform size and frequently contain a single copy of mtDNA. Proc Natl Acad Sci U S A 108, 13534-13539, doi:10.1073/pnas.1109263108 (2011). 10 Kaufman, B. A. et al. The mitochondrial transcription factor TFAM coordinates the assembly of multiple DNA molecules into nucleoid-like structures. Mol Biol Cell 18, 3225-3236, doi:10.1091/mbc.e07-05-0404 (2007). 11 Ropp, P. A. & Copeland, W. C. Cloning and characterization of the human mitochondrial DNA polymerase, DNA polymerase gamma. Genomics 36, 449-458, doi:10.1006/geno.1996.0490 (1996). 12 Lee, Y. S., Kennedy, W. D. & Yin, Y. W. Structural insight into processive human mitochondrial DNA synthesis and disease-related polymerase mutations. Cell 139, 312-324, doi:10.1016/j.cell.2009.07.050 (2009). 13 Lim, S. E., Longley, M. J. & Copeland, W. C. The mitochondrial p55 accessory subunit of human DNA polymerase gamma enhances DNA binding, promotes processive DNA synthesis, and confers N-ethylmaleimide resistance. J Biol Chem 274, 38197-38203, doi:10.1074/jbc.274.53.38197 (1999). 14 Viikov, K., Valjamae, P. & Sedman, J. Yeast mitochondrial DNA polymerase is a highly processive single-subunit enzyme. Mitochondrion 11, 119-126, doi:10.1016/j.mito.2010.08.007 (2011). 15 Longley, M. J., Prasad, R., Srivastava, D. K., Wilson, S. H. & Copeland, W. C. Identification of 5'- deoxyribose phosphate lyase activity in human DNA polymerase gamma and its role in mitochondrial base excision repair in vitro. Proc Natl Acad Sci U S A 95, 12244-12248, doi:10.1073/pnas.95.21.12244 (1998). 16 Lim, S. E., Ponamarev, M. V., Longley, M. J. & Copeland, W. C. Structural determinants in human DNA polymerase gamma account for mitochondrial toxicity from nucleoside analogs. J Mol Biol 329, 45-57, doi:10.1016/s0022-2836(03)00405-4 (2003). 17 Spelbrink, J. N. et al. Human mitochondrial DNA deletions associated with mutations in the gene encoding Twinkle, a phage T7 gene 4-like protein localized in mitochondria. Nat Genet 28, 223- 231, doi:10.1038/90058 (2001). 18 Mignotte, B., Barat, M. & Mounolou, J. C. Characterization of a mitochondrial protein binding to single-stranded DNA. Nucleic Acids Res 13, 1703-1716, doi:10.1093/nar/13.5.1703 (1985). 19 Korhonen, J. A., Pham, X. H., Pellegrini, M. & Falkenberg, M. Reconstitution of a minimal mtDNA replisome in vitro. EMBO J 23, 2423-2429, doi:10.1038/sj.emboj.7600257 (2004). 20 Gustafsson, C. M., Falkenberg, M. & Larsson, N. G. Maintenance and Expression of Mammalian Mitochondrial DNA. Annu Rev Biochem 85, 133-160, doi:10.1146/annurev-biochem-060815- 014402 (2016). 21 Falkenberg, M., Larsson, N. G. & Gustafsson, C. M. DNA replication and transcription in mammalian mitochondria. Annu Rev Biochem 76, 679-699, doi:10.1146/annurev.biochem.76.060305.152028 (2007).

22 Shadel, G. S. & Clayton, D. A. Mitochondrial DNA maintenance in vertebrates. Annu Rev Biochem 66, 409-435, doi:10.1146/annurev.biochem.66.1.409 (1997). 23 Wanrooij, S. et al. Human mitochondrial RNA polymerase primes lagging-strand DNA synthesis in vitro. Proc Natl Acad Sci U S A 105, 11122-11127, doi:10.1073/pnas.0805399105 (2008). 24 Krasich, R. & Copeland, W. C. DNA polymerases in the mitochondria: A critical review of the evidence. Front Biosci (Landmark Ed) 22, 692-709, doi:10.2741/4510 (2017). 25 Uhler, J. P. & Falkenberg, M. Primer removal during mammalian mitochondrial DNA replication. DNA Repair (Amst) 34, 28-38, doi:10.1016/j.dnarep.2015.07.003 (2015). 26 Holt, I. J. Mitochondrial DNA replication and repair: all a flap. Trends Biochem Sci 34, 358-365, doi:10.1016/j.tibs.2009.03.007 (2009). 27 Cotner-Gohara, E. et al. Human DNA ligase III recognizes DNA ends by dynamic switching between two DNA-bound states. Biochemistry 49, 6165-6176, doi:10.1021/bi100503w (2010). 28 Xu, B. & Clayton, D. A. A persistent RNA-DNA hybrid is formed during transcription at a phylogenetically conserved mitochondrial DNA sequence. Mol Cell Biol 15, 580-589, doi:10.1128/mcb.15.1.580 (1995). 29 Wanrooij, P. H. et al. A hybrid G-quadruplex structure formed between RNA and DNA explains the extraordinary stability of the mitochondrial R-loop. Nucleic Acids Res 40, 10334-10344, doi:10.1093/nar/gks802 (2012). 30 Posse, V. et al. RNase H1 directs origin-specific initiation of DNA replication in human mitochondria. PLoS Genet 15, e1007781, doi:10.1371/journal.pgen.1007781 (2019). 31 Chang, D. D., Hauswirth, W. W. & Clayton, D. A. Replication priming and transcription initiate from precisely the same site in mouse mitochondrial DNA. EMBO J 4, 1559-1567 (1985). 32 Agaronyan, K., Morozov, Y. I., Anikin, M. & Temiakov, D. Mitochondrial biology. Replication- transcription switch in human mitochondria. Science 347, 548-551, doi:10.1126/science.aaa0986 (2015). 33 Nicholls, T. J. et al. Linear mtDNA fragments and unusual mtDNA rearrangements associated with pathological deficiency of MGME1 exonuclease. Hum Mol Genet 23, 6147-6162, doi:10.1093/hmg/ddu336 (2014). 34 Robberson, D. L. & Clayton, D. A. Replication of mitochondrial DNA in mouse L cells and their thymidine kinase - derivatives: displacement replication on a covalently-closed circular template. Proc Natl Acad Sci U S A 69, 3810-3814, doi:10.1073/pnas.69.12.3810 (1972). 35 Clayton, D. A. Replication of animal mitochondrial DNA. Cell 28, 693-705, doi:10.1016/0092- 8674(82)90049-6 (1982). 36 Yasukawa, T. et al. Replication of vertebrate mitochondrial DNA entails transient ribonucleotide incorporation throughout the lagging strand. EMBO J 25, 5358-5371, doi:10.1038/sj.emboj.7601392 (2006). 37 Reyes, A. et al. Mitochondrial DNA replication proceeds via a 'bootlace' mechanism involving the incorporation of processed transcripts. Nucleic Acids Res 41, 5837-5850, doi:10.1093/nar/gkt196 (2013). 38 Fuste, J. M. et al. Mitochondrial RNA polymerase is needed for activation of the origin of light- strand DNA replication. Mol Cell 37, 67-78, doi:10.1016/j.molcel.2009.12.021 (2010). 39 Holt, I. J., Lorimer, H. E. & Jacobs, H. T. Coupled leading- and lagging-strand synthesis of mammalian mitochondrial DNA. Cell 100, 515-524, doi:10.1016/s0092-8674(00)80688-1 (2000). 40 Copeland, W. C. & Longley, M. J. Mitochondrial genome maintenance in health and disease. DNA Repair (Amst) 19, 190-198, doi:10.1016/j.dnarep.2014.03.010 (2014). 41 Gorman, G. S. et al. Mitochondrial diseases. Nat Rev Dis Primers 2, 16080, doi:10.1038/nrdp.2016.80 (2016). 42 Ylikallio, E. & Suomalainen, A. Mechanisms of mitochondrial diseases. Ann Med 44, 41-59, doi:10.3109/07853890.2011.598547 (2012). 43 Suomalainen, A. & Isohanni, P. Mitochondrial DNA depletion syndromes--many genes, common mechanisms. Neuromuscul Disord 20, 429-437, doi:10.1016/j.nmd.2010.03.017 (2010). 44 Mandel, H. et al. The deoxyguanosine kinase gene is mutated in individuals with depleted hepatocerebral mitochondrial DNA. Nat Genet 29, 337-341, doi:10.1038/ng746 (2001). 45 Saada, A. et al. Mutant mitochondrial thymidine kinase in mitochondrial DNA depletion myopathy. Nat Genet 29, 342-344, doi:10.1038/ng751 (2001). 46 Hayashi, J. et al. Introduction of disease-related mitochondrial DNA deletions into HeLa cells lacking mitochondrial DNA results in mitochondrial dysfunction. Proc Natl Acad Sci U S A 88, 10614-10618, doi:10.1073/pnas.88.23.10614 (1991). 47 Cooke, M. S., Evans, M. D., Dizdaroglu, M. & Lunec, J. Oxidative DNA damage: mechanisms, mutation, and disease. FASEB J 17, 1195-1214, doi:10.1096/fj.02-0752rev (2003).

48 Young, M. J. Off-Target Effects of Drugs that Disrupt Human Mitochondrial DNA Maintenance. Front Mol Biosci 4, 74, doi:10.3389/fmolb.2017.00074 (2017). 49 Pinz, K. G. & Bogenhagen, D. F. Efficient repair of abasic sites in DNA by mitochondrial enzymes. Mol Cell Biol 18, 1257-1265, doi:10.1128/mcb.18.3.1257 (1998). 50 Szczesny, B., Tann, A. W., Longley, M. J., Copeland, W. C. & Mitra, S. Long patch base excision repair in mammalian mitochondrial genomes. J Biol Chem 283, 26349-26356, doi:10.1074/jbc.M803491200 (2008). 51 Goodman, M. F. & Woodgate, R. Translesion DNA polymerases. Cold Spring Harb Perspect Biol 5, a010363, doi:10.1101/cshperspect.a010363 (2013). 52 Garcia-Gomez, S. et al. PrimPol, an archaic primase/polymerase operating in human cells. Mol Cell 52, 541-553, doi:10.1016/j.molcel.2013.09.025 (2013). 53 Rechkoblit, O. et al. Structure and mechanism of human PrimPol, a DNA polymerase with primase activity. Sci Adv 2, e1601317, doi:10.1126/sciadv.1601317 (2016). 54 Keen, B. A., Jozwiakowski, S. K., Bailey, L. J., Bianchi, J. & Doherty, A. J. Molecular dissection of the domain architecture and catalytic activities of human PrimPol. Nucleic Acids Res 42, 5830- 5845, doi:10.1093/nar/gku214 (2014). 55 Guilliam, T. A., Bailey, L. J., Brissett, N. C. & Doherty, A. J. PolDIP2 interacts with human PrimPol and enhances its DNA polymerase activities. Nucleic Acids Res 44, 3317-3329, doi:10.1093/nar/gkw175 (2016). 56 Kasho, K. et al. A unique arginine cluster in PolDIP2 enhances nucleotide binding and DNA synthesis by PrimPol. Nucleic Acids Res, doi:10.1093/nar/gkab049 (2021). 57 Zafar, M. K., Ketkar, A., Lodeiro, M. F., Cameron, C. E. & Eoff, R. L. Kinetic analysis of human PrimPol DNA polymerase activity reveals a generally error-prone enzyme capable of accurately bypassing 7,8-dihydro-8-oxo-2'-deoxyguanosine. Biochemistry 53, 6584-6594, doi:10.1021/bi501024u (2014). 58 Tokarsky, E. J., Wallenmeyer, P. C., Phi, K. K. & Suo, Z. Significant impact of divalent metal ions on the fidelity, sugar selectivity, and drug incorporation efficiency of human PrimPol. DNA Repair (Amst) 49, 51-59, doi:10.1016/j.dnarep.2016.11.003 (2017). 59 Bianchi, J. et al. PrimPol bypasses UV photoproducts during eukaryotic chromosomal DNA replication. Mol Cell 52, 566-573, doi:10.1016/j.molcel.2013.10.035 (2013). 60 Mouron, S. et al. Repriming of DNA synthesis at stalled replication forks by human PrimPol. Nat Struct Mol Biol 20, 1383-1389, doi:10.1038/nsmb.2719 (2013). 61 Martinez-Jimenez, M. I., Lahera, A. & Blanco, L. Human PrimPol activity is enhanced by RPA. Sci Rep 7, 783, doi:10.1038/s41598-017-00958-3 (2017). 62 Kobayashi, K. et al. Repriming by PrimPol is critical for DNA replication restart downstream of lesions and chain-terminating nucleosides. Cell Cycle 15, 1997-2008, doi:10.1080/15384101.2016.1191711 (2016). 63 Schiavone, D. et al. PrimPol Is Required for Replicative Tolerance of G Quadruplexes in Vertebrate Cells. Mol Cell 61, 161-169, doi:10.1016/j.molcel.2015.10.038 (2016). 64 Wan, L. et al. hPrimpol1/CCDC111 is a human DNA primase-polymerase required for the maintenance of genome integrity. EMBO Rep 14, 1104-1112, doi:10.1038/embor.2013.159 (2013). 65 Guilliam, T. A. et al. Molecular basis for PrimPol recruitment to replication forks by RPA. Nat Commun 8, 15222, doi:10.1038/ncomms15222 (2017). 66 Bailey, L. J., Bianchi, J., Hegarat, N., Hochegger, H. & Doherty, A. J. PrimPol-deficient cells exhibit a pronounced G2 checkpoint response following UV damage. Cell Cycle 15, 908-918, doi:10.1080/15384101.2015.1128597 (2016). 67 Vallerga, M. B., Mansilla, S. F., Federico, M. B., Bertolin, A. P. & Gottifredi, V. Rad51 recombinase prevents Mre11 nuclease-dependent degradation and excessive PrimPol-mediated elongation of nascent DNA after UV irradiation. Proc Natl Acad Sci U S A 112, E6624-6633, doi:10.1073/pnas.1508543112 (2015). 68 Bailey, L. J., Bianchi, J. & Doherty, A. J. PrimPol is required for the maintenance of efficient nuclear and mitochondrial DNA replication in human cells. Nucleic Acids Res 47, 4026-4038, doi:10.1093/nar/gkz056 (2019). 69 Zhao, F. et al. Exome sequencing reveals CCDC111 mutation associated with high myopia. Hum Genet 132, 913-921, doi:10.1007/s00439-013-1303-6 (2013). 70 Keen, B. A., Bailey, L. J., Jozwiakowski, S. K. & Doherty, A. J. Human PrimPol mutation associated with high myopia has a DNA replication defect. Nucleic Acids Res 42, 12102-12111, doi:10.1093/nar/gku879 (2014). 71 Kasamo, K., Nakamura, M., Daimou, Y. & Sano, A. A PRIMPOL mutation and variants in multiple genes may contribute to phenotypes in a familial case with chronic progressive external ophthalmoplegia symptoms. Neurosci Res 157, 58-63, doi:10.1016/j.neures.2019.07.006 (2020).

72 Li, J. & Zhang, Q. PRIMPOL mutation: functional study does not always reveal the truth. Invest Ophthalmol Vis Sci 56, 1181-1182, doi:10.1167/iovs.14-16072 (2015). 73 Nick McElhinny, S. A. et al. Abundant ribonucleotide incorporation into DNA by yeast replicative polymerases. Proc Natl Acad Sci U S A 107, 4949-4954, doi:10.1073/pnas.0914857107 (2010). 74 Chiu, H. C. et al. RNA intrusions change DNA elastic properties and structure. Nanoscale 6, 10009-10017, doi:10.1039/c4nr01794c (2014). 75 DeRose, E. F., Perera, L., Murray, M. S., Kunkel, T. A. & London, R. E. Solution structure of the Dickerson DNA dodecamer containing a single ribonucleotide. Biochemistry 51, 2407-2416, doi:10.1021/bi201710q (2012). 76 Yao, N. Y., Schroeder, J. W., Yurieva, O., Simmons, L. A. & O'Donnell, M. E. Cost of rNTP/dNTP pool imbalance at the replication fork. Proc Natl Acad Sci U S A 110, 12942-12947, doi:10.1073/pnas.1309506110 (2013). 77 Hiller, B. et al. Mammalian RNase H2 removes ribonucleotides from DNA to maintain genome integrity. J Exp Med 209, 1419-1426, doi:10.1084/jem.20120876 (2012). 78 Reijns, M. A. et al. Enzymatic removal of ribonucleotides from DNA is essential for mammalian genome integrity and development. Cell 149, 1008-1022, doi:10.1016/j.cell.2012.04.011 (2012). 79 Pizzi, S. et al. Reduction of hRNase H2 activity in Aicardi-Goutieres syndrome cells leads to replication stress and genome instability. Hum Mol Genet 24, 649-658, doi:10.1093/hmg/ddu485 (2015). 80 Berglund, A. K. et al. Nucleotide pools dictate the identity and frequency of ribonucleotide incorporation in mitochondrial DNA. PLoS Genet 13, e1006628, doi:10.1371/journal.pgen.1006628 (2017). 81 Reijns, M. A. M. et al. Lagging-strand replication shapes the mutational landscape of the genome. Nature 518, 502-506, doi:10.1038/nature14183 (2015). 82 Nick McElhinny, S. A. & Ramsden, D. A. Polymerase mu is a DNA-directed DNA/RNA polymerase. Mol Cell Biol 23, 2309-2315, doi:10.1128/mcb.23.7.2309-2315.2003 (2003). 83 Astatke, M., Ng, K., Grindley, N. D. & Joyce, C. M. A single side chain prevents Escherichia coli DNA polymerase I (Klenow fragment) from incorporating ribonucleotides. Proc Natl Acad Sci U S A 95, 3402-3407, doi:10.1073/pnas.95.7.3402 (1998). 84 Clausen, A. R., Zhang, S., Burgers, P. M., Lee, M. Y. & Kunkel, T. A. Ribonucleotide incorporation, proofreading and bypass by human DNA polymerase delta. DNA Repair (Amst) 12, 121-127, doi:10.1016/j.dnarep.2012.11.006 (2013). 85 Brown, J. A. & Suo, Z. Unlocking the sugar "steric gate" of DNA polymerases. Biochemistry 50, 1135-1142, doi:10.1021/bi101915z (2011). 86 Nick McElhinny, S. A. et al. Genome instability due to ribonucleotide incorporation into DNA. Nat Chem Biol 6, 774-781, doi:10.1038/nchembio.424 (2010). 87 Johnson, M. K., Kottur, J. & Nair, D. T. A polar filter in DNA polymerases prevents ribonucleotide incorporation. Nucleic Acids Res 47, 10693-10705, doi:10.1093/nar/gkz792 (2019). 88 Traut, T. W. Physiological concentrations of purines and pyrimidines. Mol Cell Biochem 140, 1- 22, doi:10.1007/BF00928361 (1994). 89 Ferraro, P., Franzolin, E., Pontarin, G., Reichard, P. & Bianchi, V. Quantitation of cellular deoxynucleoside triphosphates. Nucleic Acids Res 38, e85, doi:10.1093/nar/gkp1141 (2010). 90 Bogenhagen, D. & Clayton, D. A. Mouse L cell mitochondrial DNA molecules are selected randomly for replication throughout the cell cycle. Cell 11, 719-727, doi:10.1016/0092- 8674(77)90286-0 (1977). 91 Chabes, A. & Stillman, B. Constitutively high dNTP concentration inhibits cell cycle progression and the DNA damage checkpoint in yeast Saccharomyces cerevisiae. Proc Natl Acad Sci U S A 104, 1183-1188, doi:10.1073/pnas.0610585104 (2007). 92 Kumar, D. et al. Mechanisms of mutagenesis in vivo due to imbalanced dNTP pools. Nucleic Acids Res 39, 1360-1371, doi:10.1093/nar/gkq829 (2011). 93 Hanahan, D. & Weinberg, R. A. Hallmarks of cancer: the next generation. Cell 144, 646-674, doi:10.1016/j.cell.2011.02.013 (2011). 94 Hofer, A., Crona, M., Logan, D. T. & Sjoberg, B. M. DNA building blocks: keeping control of manufacture. Crit Rev Biochem Mol Biol 47, 50-63, doi:10.3109/10409238.2011.630372 (2012). 95 Mathews, C. K. Deoxyribonucleotide metabolism, mutagenesis and cancer. Nat Rev Cancer 15, 528-539, doi:10.1038/nrc3981 (2015). 96 Chabes, A. & Thelander, L. Controlled protein degradation regulates ribonucleotide reductase activity in proliferating mammalian cells during the normal cell cycle and in response to DNA damage and replication blocks. J Biol Chem 275, 17747-17753, doi:10.1074/jbc.M000799200 (2000).

97 Thelander, L. & Reichard, P. Reduction of ribonucleotides. Annu Rev Biochem 48, 133-158, doi:10.1146/annurev.bi.48.070179.001025 (1979). 98 Franzolin, E. et al. The deoxynucleotide triphosphohydrolase SAMHD1 is a major regulator of DNA precursor pools in mammalian cells. Proc Natl Acad Sci U S A 110, 14272-14277, doi:10.1073/pnas.1312033110 (2013). 99 Rampazzo, C. et al. Regulation by degradation, a cellular defense against deoxyribonucleotide pool imbalances. Mutat Res 703, 2-10, doi:10.1016/j.mrgentox.2010.06.002 (2010). 100 Goksenin, A. Y. et al. Human DNA polymerase epsilon is able to efficiently extend from multiple consecutive ribonucleotides. J Biol Chem 287, 42675-42684, doi:10.1074/jbc.M112.422733 (2012). 101 Williams, J. S. et al. Proofreading of ribonucleotides inserted into DNA by yeast DNA polymerase varepsilon. DNA Repair (Amst) 11, 649-656, doi:10.1016/j.dnarep.2012.05.004 (2012). 102 Sparks, J. L. et al. RNase H2-initiated ribonucleotide excision repair. Mol Cell 47, 980-986, doi:10.1016/j.molcel.2012.06.035 (2012). 103 Rydberg, B. & Game, J. Excision of misincorporated ribonucleotides in DNA by RNase H (type 2) and FEN-1 in cell-free extracts. Proc Natl Acad Sci U S A 99, 16654-16659, doi:10.1073/pnas.262591699 (2002). 104 Qiu, J., Qian, Y., Frank, P., Wintersberger, U. & Shen, B. Saccharomyces cerevisiae RNase H(35) functions in RNA primer removal during lagging-strand DNA synthesis, most efficiently in cooperation with Rad27 nuclease. Mol Cell Biol 19, 8361-8371, doi:10.1128/mcb.19.12.8361 (1999). 105 Cerritelli, S. M. & Crouch, R. J. Ribonuclease H: the enzymes in eukaryotes. FEBS J 276, 1494- 1505, doi:10.1111/j.1742-4658.2009.06908.x (2009). 106 Drolet, M. et al. Overexpression of RNase H partially complements the growth defect of an Escherichia coli delta topA mutant: R-loop formation is a major problem in the absence of DNA topoisomerase I. Proc Natl Acad Sci U S A 92, 3526-3530, doi:10.1073/pnas.92.8.3526 (1995). 107 Nowotny, M., Gaidamakov, S. A., Crouch, R. J. & Yang, W. Crystal structures of RNase H bound to an RNA/DNA hybrid: substrate specificity and metal-dependent catalysis. Cell 121, 1005-1016, doi:10.1016/j.cell.2005.04.024 (2005). 108 Sekiguchi, J. & Shuman, S. Site-specific ribonuclease activity of eukaryotic DNA topoisomerase I. Mol Cell 1, 89-97, doi:10.1016/s1097-2765(00)80010-6 (1997). 109 Williams, J. S. et al. Topoisomerase 1-mediated removal of ribonucleotides from nascent leading- strand DNA. Mol Cell 49, 1010-1015, doi:10.1016/j.molcel.2012.12.021 (2013). 110 Aguilera, A. & Garcia-Muse, T. R loops: from transcription byproducts to threats to genome stability. Mol Cell 46, 115-124, doi:10.1016/j.molcel.2012.04.009 (2012). 111 Kim, N. et al. Mutagenic processing of ribonucleotides in DNA by yeast topoisomerase I. Science 332, 1561-1564, doi:10.1126/science.1205016 (2011). 112 Watt, D. L., Johansson, E., Burgers, P. M. & Kunkel, T. A. Replication of ribonucleotide- containing DNA templates by yeast replicative polymerases. DNA Repair (Amst) 10, 897-902, doi:10.1016/j.dnarep.2011.05.009 (2011). 113 Crow, Y. J. et al. Mutations in genes encoding ribonuclease H2 subunits cause Aicardi-Goutieres syndrome and mimic congenital viral brain infection. Nat Genet 38, 910-916, doi:10.1038/ng1842 (2006). 114 Marey, I. et al. Non Random Distribution of DMD Deletion Breakpoints and Implication of Double Strand Breaks Repair and Replication Error Repair Mechanisms. J Neuromuscul Dis 3, 227-245, doi:10.3233/JND-150134 (2016). 115 Vaisman, A. et al. Investigating the mechanisms of ribonucleotide excision repair in Escherichia coli. Mutat Res 761, 21-33, doi:10.1016/j.mrfmmm.2014.01.005 (2014). 116 Lujan, S. A., Williams, J. S., Clausen, A. R., Clark, A. B. & Kunkel, T. A. Ribonucleotides are signals for mismatch repair of leading-strand replication errors. Mol Cell 50, 437-443, doi:10.1016/j.molcel.2013.03.017 (2013). 117 Ghodgaonkar, M. M. et al. Ribonucleotides misincorporated into DNA act as strand-discrimination signals in eukaryotic mismatch repair. Mol Cell 50, 323-332, doi:10.1016/j.molcel.2013.03.019 (2013). 118 Crespan, E. et al. Impact of ribonucleotide incorporation by DNA polymerases beta and lambda on oxidative base excision repair. Nat Commun 7, 10805, doi:10.1038/ncomms10805 (2016). 119 Clausen, A. R. et al. Tracking replication enzymology in vivo by genome-wide mapping of ribonucleotide incorporation. Nat Struct Mol Biol 22, 185-191, doi:10.1038/nsmb.2957 (2015). 120 Koh, K. D., Balachander, S., Hesselberth, J. R. & Storici, F. Ribose-seq: global mapping of ribonucleotides embedded in genomic DNA. Nat Methods 12, 251-257, 253 p following 257, doi:10.1038/nmeth.3259 (2015).

121 Ding, J., Taylor, M. S., Jackson, A. P. & Reijns, M. A. Genome-wide mapping of embedded ribonucleotides and other noncanonical nucleotides using emRiboSeq and EndoSeq. Nat Protoc 10, 1433-1444, doi:10.1038/nprot.2015.099 (2015). 122 Daigaku, Y. et al. A global profile of replicative polymerase usage. Nat Struct Mol Biol 22, 192- 198, doi:10.1038/nsmb.2962 (2015). 123 Wanrooij, P. H. et al. Elimination of rNMPs from mitochondrial DNA has no effect on its stability. Proc Natl Acad Sci U S A 117, 14306-14313, doi:10.1073/pnas.1916851117 (2020). 124 Grossman, L. I., Watson, R. & Vinograd, J. The presence of ribonucleotides in mature closed- circular mitochondrial DNA. Proc Natl Acad Sci U S A 70, 3339-3343, doi:10.1073/pnas.70.12.3339 (1973). 125 Uhler, J. P. et al. MGME1 processes flaps into ligatable nicks in concert with DNA polymerase gamma during mtDNA replication. Nucleic Acids Res 44, 5861-5871, doi:10.1093/nar/gkw468 (2016). 126 Al-Behadili, A. et al. A two-nuclease pathway involving RNase H1 is required for primer removal at human mitochondrial OriL. Nucleic Acids Res 46, 9471-9483, doi:10.1093/nar/gky708 (2018). 127 Cerritelli, S. M. et al. Failure to produce mitochondrial DNA results in embryonic lethality in Rnaseh1 null mice. Mol Cell 11, 807-815, doi:10.1016/s1097-2765(03)00088-1 (2003). 128 Reyes, A. et al. RNASEH1 Mutations Impair mtDNA Replication and Cause Adult-Onset Mitochondrial Encephalomyopathy. Am J Hum Genet 97, 186-193, doi:10.1016/j.ajhg.2015.05.013 (2015). 129 Moss, C. F. et al. Aberrant ribonucleotide incorporation and multiple deletions in mitochondrial DNA of the murine MPV17 disease model. Nucleic Acids Res 45, 12808-12815, doi:10.1093/nar/gkx1009 (2017). 130 Kasiviswanathan, R. & Copeland, W. C. Ribonucleotide discrimination and reverse transcription by the human mitochondrial DNA polymerase. J Biol Chem 286, 31490-31500, doi:10.1074/jbc.M111.252460 (2011). 131 Hanes, J. W., Thal, D. M. & Johnson, K. A. Incorporation and replication of 8-oxo- deoxyguanosine by the human mitochondrial DNA polymerase. J Biol Chem 281, 36241-36248, doi:10.1074/jbc.M607965200 (2006). 132 Guilliam, T. A. et al. Human PrimPol is a highly error-prone polymerase regulated by single- stranded DNA binding proteins. Nucleic Acids Res 43, 1056-1068, doi:10.1093/nar/gku1321 (2015). 133 Wanrooij, S., Goffart, S., Pohjoismaki, J. L., Yasukawa, T. & Spelbrink, J. N. Expression of catalytic mutants of the mtDNA helicase Twinkle and polymerase POLG causes distinct replication stalling phenotypes. Nucleic Acids Res 35, 3238-3251, doi:10.1093/nar/gkm215 (2007). 134 Goffart, S. et al. Twinkle mutations associated with autosomal dominant progressive external ophthalmoplegia lead to impaired helicase function and in vivo mtDNA replication stalling. Hum Mol Genet 18, 328-340, doi:10.1093/hmg/ddn359 (2009). 135 Lim, S. E. & Copeland, W. C. Differential incorporation and removal of antiviral deoxynucleotides by human DNA polymerase gamma. J Biol Chem 276, 23616-23623, doi:10.1074/jbc.M101114200 (2001). 136 Torregrosa-Munumer, R., Goffart, S., Haikonen, J. A. & Pohjoismaki, J. L. Low doses of ultraviolet radiation and oxidative damage induce dramatic accumulation of mitochondrial DNA replication intermediates, fork regression, and replication initiation shift. Mol Biol Cell 26, 4197- 4208, doi:10.1091/mbc.E15-06-0390 (2015). 137 Gonzalez-Prieto, R., Munoz-Cabello, A. M., Cabello-Lobato, M. J. & Prado, F. Rad51 replication fork recruitment is required for DNA damage tolerance. EMBO J 32, 1307-1321, doi:10.1038/emboj.2013.73 (2013). 138 Copeland, W. C. Defects in mitochondrial DNA replication and human disease. Crit Rev Biochem Mol Biol 47, 64-74, doi:10.3109/10409238.2011.632763 (2012). 139 Yang, M. Y. et al. Biased incorporation of ribonucleotides on the mitochondrial L-strand accounts for apparent strand-asymmetric DNA replication. Cell 111, 495-505, doi:10.1016/s0092- 8674(02)01075-9 (2002). 140 Murakami, E. et al. Characterization of novel reverse transcriptase and other RNA-associated catalytic activities by human DNA polymerase gamma: importance in mitochondrial DNA replication. J Biol Chem 278, 36403-36409, doi:10.1074/jbc.M306236200 (2003). 141 Gandhi, V. V. & Samuels, D. C. A review comparing deoxyribonucleoside triphosphate (dNTP) concentrations in the mitochondrial and cytoplasmic compartments of normal and transformed cells. Nucleosides Nucleotides Nucleic Acids 30, 317-339, doi:10.1080/15257770.2011.586955 (2011).

142 Nick McElhinny, S. A., Kissling, G. E. & Kunkel, T. A. Differential correction of lagging-strand replication errors made by DNA polymerases {alpha} and {delta}. Proc Natl Acad Sci U S A 107, 21070-21075, doi:10.1073/pnas.1013048107 (2010). 143 Wanrooij, P. H. et al. Ribonucleotides incorporated by the yeast mitochondrial DNA polymerase are not repaired. Proc Natl Acad Sci U S A 114, 12466-12471, doi:10.1073/pnas.1713085114 (2017). 144 Sykora, P. et al. DNA Polymerase Beta Participates in Mitochondrial DNA Repair. Mol Cell Biol 37, doi:10.1128/MCB.00237-17 (2017). 145 Vengrova, S. & Dalgaard, J. Z. RNase-sensitive DNA modification(s) initiates S. pombe mating- type switching. Genes Dev 18, 794-804, doi:10.1101/gad.289404 (2004).