Mammalian RNase H1 directs RNA primer formation for mtDNA replication initiation and is also necessary for mtDNA replication completion

Abstract The in vivo role for RNase H1 in mammalian mitochondria has been much debated. Loss of RNase H1 is embryonic lethal and to further study its role in mtDNA expression we characterized a conditional knockout of Rnaseh1 in mouse heart. We report that RNase H1 is essential for processing of RNA primers to allow site-specific initiation of mtDNA replication. Without RNase H1, the RNA:DNA hybrids at the replication origins are not processed and mtDNA replication is initiated at non-canonical sites and becomes impaired. Importantly, RNase H1 is also needed for replication completion and in its absence linear deleted mtDNA molecules extending between the two origins of mtDNA replication are formed accompanied by mtDNA depletion. The steady-state levels of mitochondrial transcripts follow the levels of mtDNA, and RNA processing is not altered in the absence of RNase H1. Finally, we report the first patient with a homozygous pathogenic mutation in the hybrid-binding domain of RNase H1 causing impaired mtDNA replication. In contrast to catalytically inactive variants of RNase H1, this mutant version has enhanced enzyme activity but shows impaired primer formation. This finding shows that the RNase H1 activity must be strictly controlled to allow proper regulation of mtDNA replication.


INTRODUCTION
The ribonuclease H (RNase H) family of enzymes are endonucleases that cleave the RNA component of RNA:DNA hybrids without sequence specificity, and are found in prokaryotes, archaea and eukaryotes (1). Two main groups of this endonuclease family, RNase H1 and RNase H2, are found in all three domains of life and are the only variants present in mammalian cells. The mammalian RNase H2 is an obligate heterotrimer consisting of a catalytic subunit, RNase H2A, complexed with two structural subunits, RNase H2B and RNase H2C. The RNase H2 enzyme is only present in the nucleus and has been reported to be important for removal of mis-incorporated ribonucleotides and RNA:DNA hybrids formed during DNA replication and repair (2)(3)(4). In contrast, RNase H1 is a monomeric enzyme that has been reported to have two isoforms created by alternate translation start sites in the encoding mRNA (5). The longer isoform contains a mitochondrial targeting sequence (MTS) directing import of RNase H1 into mitochondria, whereas the predicted shorter isoform lacks the MTS and is localized to the nucleus. In the nucleus, RNase H1 has been suggested to contribute to genome stability and to preserve telomere integrity by resolving R-loops (6)(7)(8), which are triple-stranded structures comprised of an RNA:DNA hybrid and a displaced DNA strand. It is well known that R-loops arise from different sources and can be resolved by several enzymatic activities in the nucleus such as topoisomerases and helicases (9)(10)(11)(12). RNase H1 has been implicated in the removal of R-loops formed during nuclear DNA replication (6), immunoglobulin class switch DNA recombination (13) or as by-products of rDNA transcription in the nucleoli (14). In mitochondria, RNase H1 is essential for mitochondrial DNA (mtDNA) replication (15) and has been linked to several aspects of the expression and replication of mtDNA. In the mouse, homozygous Rnaseh1 knockout causes severe mtDNA depletion and developmental arrest at embryonic day (E) 8.5 (15). A study of cultured mouse embryonic fibroblasts (MEFs) lacking RNase H1 reported that it is required for primer removal during mtDNA replication (16). In support of this model, an in vitro study of mtDNA replication initiation at the light-strand origin of replication (O L ) with recombinant replication factors revealed an important role for RNase H1 in RNA primer removal (17). However, RNase H1 can only partly remove primers as it leaves 1-3 ribonucleotides at the RNA-DNA junction (17), consistent with its known biochemical properties (18). A subsequent study of patient cell lines expressing mutant RNase H1 combined with in vitro biochemistry suggested that RNase H1 also has a role in primer maturation by generating the 3 RNA-ends that the mitochondrial DNA polymerase ␥ (POL␥ ) uses for replication initiation (19). In addition to these two proposed roles in mtDNA replication, i.e. primer maturation and removal, RNase H1 has also been reported to regulate transcription of 7S RNA, an abundant non-coding mtDNA transcript linked to replication primer synthesis (20). Furthermore, RNase H1 has been suggested to play a role in pre-rRNA processing via interaction with the P32 protein of the mitochondrial nucleoid (21). Constitutive and inducible knockout of RNase H1 in mouse liver causes severe liver degeneration due to mitochondrial dysfunction suggested to be caused by R-loop accumulation resulting in reduced transcription and replication of mtDNA (22). Although the exact role of RNase H1 in mammalian mitochondria in vivo needs further clarification, additional support for a critical role in mtDNA maintenance comes from studies of patients with mutations in RNASEH1 (23)(24)(25)(26). A range of pathogenic mutations have been found in RNASEH1 that directly reduces the catalytic function of the enzyme. Affected patients display typical adult-onset mitochondrial disease phenotypes, such as mitochondrial myopathy with exercise intolerance, muscle weakness with progressive external ophthalmoplegia (PEO) and ptosis, and central nervous system involvement with cerebellar atrophy and ataxia (23)(24)(25)(26). In these patients, genetic analysis of skeletal muscle biopsy specimens typically demonstrates rearrangements often accompanied by depletion of mtDNA (23)(24)(25)(26).
In this study, we have extensively characterized the in vivo role of mouse RNase H1 and report that it is essential both during embryonic development and for the function of differentiated organs, such as the heart and skeletal muscle. Loss of RNase H1 in the heart leads to mtDNA replication stalling with accumulation of linear deleted mtDNA fragments and mtDNA depletion. Ablation of RNase H1 also leads to low levels of mtDNA-encoded transcripts and deficient oxidative phosphorylation (OXPHOS), which is likely related to the marked mtDNA depletion as no transcription elongation or RNA processing defects are observed. We provide evidence that RNase H1 is necessary to restrict RNA replication primer formation to specific sites at the heavy-strand origin of replication (O H ) and O L in conditional knockout hearts of the mouse. Finally, we report a mitochondrial disease patient with altered mtDNA replication carrying a novel homozygous pathogenic mutation that makes RNase H1 more active in processing RNA:DNA hybrids than the wild-type enzyme. These findings demonstrate that the RNase H1 activity needs to be carefully balanced to promote accurate mtDNA replication.

Mouse models
Rnaseh1 knockout mice were generated by Taconic Artemis by targeting the gene in embryonic stem cells derived from C57BL/6N using BAC clones from C57BL/6J RPCI-23 BAC library. To generate the Rnaseh1 knockout allele, exon II of the Rnaseh1 locus was flanked by loxP sites. The puromycin resistance cassette was introduced as a selection marker and removed by mating Rnaseh1 +/loxP-pur mice with transgenic mice ubiquitously expressing Flp recombinase. To obtain germline knockout mice, Rnaseh1 +/loxP mice were mated with mice ubiquitously expressing Cre recombinase under the ␤-actin promotor to generate heterozygous knockout Rnaseh1 +/− mice. The tissue-specific conditional knockout of RNase H1 in heart and skeletal muscle was obtained by crossing Rnaseh1 +/loxP mice with transgenic mice expressing Cre recombinase under the control of the muscle creatinine kinase promoter (Ckmm-cre). Animals were housed in a 12-h light/dark cycle at 21 • C and fed with a normal chow diet ad libitum. Animal studies were approved by the animal welfare ethics committee and performed in compliance with National and European law. Mice at the age of 8, 16 and 24 weeks were used for biochemical analyses.

Cell lines
Human embryonic kidney cells (HEK293T) and patientderived fibroblast lines were used for in vitro studies. Cells were cultured in DMEM GlutaMax (Invitrogen) medium supplemented with 10% fetal bovine serum (FBS) and 1× PenStrep and grown in an incubator at 37 • C and 5% CO 2 . The patient fibroblast line carrying a homozygous c.424G > A (V142I) mutation in the H-domain of RNase H1 was obtained from a patient diagnosed at Karolinska University Hospital. This patient presented with PEO and muscular fatigue. A muscle biopsy specimen from the tibialis anterior muscle at 41 years of age showed a general reduction in the mitochondrial ATP production rate and a combined respiratory enzyme complex deficiency. Histochemical analysis revealed numerous ragged-red and COXnegative muscle fibers. This mutation has been previously reported (23). Control fibroblast lines C1-3 were obtained from healthy individuals at Karolinska University Hospital. The patient fibroblast line carrying a novel homozygous c.86A > G (Y29C) mutation in the HBD of RNase H1 was obtained from a patient diagnosed at the IRCCS Istituto delle Scienze Neurologiche di Bologna in Italy. The male proband and a sister had an identical phenotype of adult-onset PEO and ptosis, associated with peripheral neuropathy, sensorineural deafness and mitochondrial myopathy with abundant COX-negative fibers. Later in life, both siblings developed parkinsonism and eventually cognitive deterioration, and died in their 80s. In addition, we studied a patient fibroblast line with a novel loss-of-function mutation in MGME1, c.563 564delCAinsT (S188YfsX7). This patient was diagnosed at the IRCCS Istituto delle Scienze Neurologiche di Bologna in Italy and had earlyonset PEO and ptosis, with mild mitochondrial myopathy and MRI of the brain showed evidence of cerebellar atrophy. The control fibroblast lines, C4 and C5, were obtained from healthy individuals at IRCCS Istituto delle Scienze Neurologiche di Bologna in Italy. For all individuals recruited in Italy, informed consent was obtained for fibroblast cell line establishment and the study was approved by the institutional ethical board (Comitato Etico Interaziendale Bologna-Imola, #CE-BI 13036).

Isolation of mitochondria from mouse tissue
Mitochondria were isolated from mouse heart using differential centrifugation as previously described (27). Following dissection, fresh tissues were cut and washed with ice cold PBS and homogenized in mitochondrial isolation buffer consisting of 310 mM sucrose, 10 mM Tris-HCl, 1 mM EDTA and freshly added 1× Complete protease inhibitor cocktail (Roche). The homogenate was centrifuged at 1000 g for 10 min at 4 • C to remove the cell debris and nuclei. The supernatant was collected and spun at 10 000 g for 15 min at 4 • C to isolate mitochondria. Crude mitochondrial pellets were resuspended in mitochondria isolation buffer and protein concentration was determined with the Qubit 4.0 fluorometer (Invitrogen).

Isolation of mitochondria from cultured cells
Mitochondria were isolated from cultured cells harvested after reaching 80-90% confluency. Cells were washed two times in cold PBS. The cell pellet was resuspended in 8 ml of mitochondria isolation buffer consisting of 20 mM HEPES pH 7.6, 220 mM mannitol, 70 mM sucrose, 1 mM EDTA and freshly added 1× Complete protease inhibitor cocktail (Roche) and 2 mg/ml of fatty acid-free bovine serum albumin (BSA). To facilitate swelling, the cell suspension was incubated on ice for 30 min. Subsequently, homogenization was performed with 40 strokes by using a manual homogenizer. The cell homogenate was centrifuged at 800 g for 10 min at 4 • C to remove the cell debris. The supernatant was collected and centrifuged at 10 000 g for 10 min at 4 • C. The mitochondrial pellet was collected and resuspended in mitochondria isolation buffer without BSA and centrifuged at 10 000 for 10 min at 4 • C. Protein quantification was performed with the Qubit 4.0 fluorometer (Invitrogen).

Cellular fractionation
HEK293T cells were cultured until reaching 80-90% confluency. Cells were then harvested, pelleted by centrifugation at 800 g for 7 min at 4 • C and washed two times in cold PBS. Mitochondria-, cytosol-and nuclei-containing fractions were obtained by using the Abcam Cell Fractionation Kit according to the instructions. The fractions were further analyzed by SDS-PAGE and western blots. The antibodies used for this experiment are listed in Supplementary Table  S1.

Protease protection assay
For each treatment, 1 mg of isolated mitochondria (see isolation of mitochondria from cultured cells) from HEK293T cells were pelleted, resuspended and incubated for 30 min in 400 l mitochondria isolation buffer [20 mM HEPES pH 7.6, 220 mM mannitol, 70 mM sucrose, 1 mM EDTA with freshly added 1× Complete protease inhibitor cocktail (Roche) and 2 mg/ml of fatty acid-free bovine serum albumin (BSA)], hypotonic swelling buffer (10 mM HEPES-KOH, pH 7.4) or mitochondria isolation buffer supplemented with 1% Triton X-100. Each of the three aliquots of mitochondria in different buffers were further subdivided into two aliquots that contained proteinase K at a final concentration of 60 g/ml or no proteinase K. After 20 min of incubation on ice, protease K activity was inhibited with 2 l of 0.1 M phenylmethylsulfonyl fluoride (PMSF) added to each sample. Proteins were precipitated with 0.02% sodium deoxycholate and 12% trichloroacetic acid. Finally, samples were analyzed by SDS-PAGE and western blots. The antibodies used for this experiment are listed in Supplementary  Table S1.

DNA extraction and mtDNA quantification
Genomic DNA from mouse tissues or human cells was isolated using the Gentra Puregene Tissue Kit (Qiagen) according to instructions. Total DNA from E8.5 embryos was isolated by incubation in 24 mM NaOH, 0.2 mM EDTA at 96 • C for 40 min. The extracts were neutralized by adding 40 mM Tris (pH 7.8). DNA quantification was performed with the Qubit 4.0 fluorometer (Invitrogen). The mtDNA copy number was measured by qPCR in a QuantStudio 6 Flex Real-Time PCR System using the TaqMan™ Universal Master Mix II with UNG and TaqMan probes for Nd5, Atp6 and 16S. The 18S probe was used to normalize samples to the nuclear DNA content. The mtDNA copy number in patient muscle biopsies was quantified by qPCR as previously described (28). The mtDNA deletions in patient muscle biopsies were quantified by droplet digital PCR (ddPCR) with specific probes in MT-ND4 and MT-ND1 (29). The 7S DNA in patient muscle biopsies was quantified by ddPCR by amplifying two mtDNA regions, one within and one outside of the 7S DNA region (29).

Southern blot analyses
Murine genomic DNA (or mtDNA where indicated), 2 g (or more where indicated), was digested with one of the following enzymes: SacI, XhoI, BglII or EagI. Human genomic DNA was digested with the EcoRI and BamHI enzymes. Linearized DNA fragments were separated in 0.8% agarose gels and transferred to Hybond-N+ membranes (GE Healthcare). To release 7S DNA, samples were heated for 3 min at 93 • C prior to loading. Next, membranes were hybridized with ␣-[ 32 P]-dCTP-labelled dsDNA probes to detect total mtDNA (pAM1 probe for mouse, and CytB for human samples), 7S DNA, or nuclear 18S rDNA as a loading control. Radioactive signals were visualized using Phos-phorImager screens and a Typhoon 7000 FLA (GE Healthcare).

In organello replication assay
Freshly isolated mitochondria (1 mg) were washed two times in 0.5 ml of cold incubation buffer [25 mM sucrose, 75 mM sorbitol, 100 mM KCl, 10 mM K 2 HPO 4 , 0.05 mM EDTA, 5 mM MgCl 2 , 10 mM glutamate, 2.5 mM malate, 10 mM Tris-HCl pH (7.4)] supplemented with fatty acidfree BSA (1 mg/ml) and 1 mM ADP. Mitochondria were then pelleted at 8000 g for 4 min and resuspended in 0.5 ml of prewarmed incubation buffer supplemented with 1 mg/ml of fatty acid-free BSA (1 mg/ml), 1 mM ADP, 50 M each of dTTP, dCTP and dGTP and 20 Ci ␣-[ 32 P]-dATP. Next, samples were incubated for 2 h at 37 • C on a rotating wheel. After incubation mitochondria were pelleted at 8000 g for 4 min and washed three times with washing buffer [10% glycerol, 0.15 mM MgCl 2 , 10 mM Tris-HCl pH (6,8)]. mtDNA was extracted with a Gentra Puregene Tissue Kit, digested with Sac I HF enzyme, separated on a 0.8% agarose gel and transferred to Hybond-N+ membrane. The signal from the membrane was visualized by autoradiography. To control for mitochondrial amount used in the assay, a protein sample, taken before the mtDNA extraction, was resolved by SDS-PAGE and transferred to a polyvinylidene difluoride (PVDF) membrane, which was incubated with an antibody against voltage-dependent anion channel (VDAC).

2DNAGE
Fresh heart mitochondria were isolated and mtDNA was extracted using the Gentra Puregene Tissue Kit without RNase A treatment. The mtDNA was digested with BclI, precipitated and 3 g from each sample was loaded onto the first-dimension gels (0.4% agarose without ethidium bromide). Electrophoresis was carried out at 27 V for 18 h at room temperature. The lanes containing DNA were excised, turned 90 • counterclockwise and cast with 1% agarose (500 ng/ml ethidium bromide). Second dimension gels were separated at constant 260 mA for 6 h at 4 • C. Gels were transferred to nitrocellulose membranes by Southern blotting, and probed with PCR-generated DNA fragments detecting either the O H -containing (nt 12 034-16 180) or the O Lcontaining (nt 3102-7084) region.

Primer extension
Total DNA was isolated from the mouse heart tissue with the Gentra Puregene Tissue Kit without RNase A treatment. An 3 g aliquot of DNA was incubated for 1 h at 37 • C with or without RNase H2 (NEB). Primer extension was performed with 2 U Taq DNA polymerase (NEB) in 1× ThermoPol buffer, 200 M dNTPs and 1.5 pmol labelled primer. The primer, corresponding to nt 15 181-15 201, was 5 -end labelled with ␥ -[ 32 P]-ATP using PNK (NEB). The primer extension reaction was initiated for 5 min at 95 • C and then continued for 25 cycles with 30 s at 95 • C, 30 s at 58 • C, 45 s at 72 • C and 5 min at 72 • C. The reactions were terminated and precipitated with ethanol.

Analysis of mtDNA deletions using deep sequencing
Total or mitochondrial DNA was extracted from mouse heart or brain, libraries were prepared using TruSeq Nano DNA Library Preparation Kit, and sequenced using NovaSeq 6000. Analyses were performed following a previously described pipeline (30). To avoid false negatives at Nuclear Mitochondrial Segments (NUMTs), reads were directly aligned to the mouse mitochondrial genome (GRCm38) to identify deletions and duplications, and then gapped alignments, indicative of deletions/duplications, were clustered and visualized. Heteroplasmy levels were calculated by comparing the number of reads supporting the corresponding breakpoints to the total number of reads overlapping the breakpoints after removal of PCR duplicates.

HydEn-sequencing
Free 5 DNA ends of mtDNA were mapped by HydEn sequencing as previously described (31). Briefly, genomic DNA was isolated from mouse heart tissue with the Gentra Puregene Tissue Kit without RNase A treatment. HydEnseq was performed by hydrolysis of 1 g of genomic DNA with 0.3 M KOH for 2 h at 55 • C. After hydrolysis, DNA was denatured for 3 min at 85 • C, cooled on ice and treated with 10 U T4 Polynucleotide Kinase (3 phosphatase minus) in 1 mM ATP for 20 min at 37 • C and then treated for 20 min at 65 • C. DNA was purified using 1.8 volumes of CleanNGS beads and eluted in 14 l EB. A 13 l aliquot of DNA was denatured for 3 min at 85 • C and ligated to 1 M ARC140 using 10 U T4 RNA ligase in 25% PEG3350, 1 mM CoCl 3 (NH3) 6 and 20 M ATP overnight at room temperature. The following day, two washes were performed using 0.8 volumes of CleanNGS beads and eluted with 20 l EB in the first step and 13.8 l EB in the second step. DNA was denatured for 3 min at 85 • C and incubated for 5 min with 1xBF for T7 DNA polymerase, 0.2 mM dNTPs, 40 g/ml BSA and 1 M of the adapter consisting of ARC76/ARC77. Reactions were then treated with 4 U of T7 DNA polymerase for 5 min. Products were purified with 0.8 volumes of CleanNGS beads and amplified using primers ARC78-ARC84 and ARC160. All oligos are from IDT. Sequences are provided in Supplementary Table S1. The star (*) indicates a phosphorothioate bond. ARC140 contains a 5 -amino group instead of a 5 -OH group, in combination with a C6 linker. Purified libraries were then sequenced using the Illumina NextSeq500 instrument for 2 × 75 paired end reads. Reads were trimmed for quality and adapter sequence with cutadapt (-m 15 -q 10-match-readwildcards). Pairs with one or both reads shorter than 15 nts were discarded. Mate 1 of the remaining pairs was aligned to an index containing the sequence of all oligos used in the preparation of these libraries with bowtie sing bowtie 0.12.8 (-m1 -v2), and all pairs with successful alignments were discarded. Pairs passing this filter were subsequently aligned to the mm10 M. musculus reference mitochondrial genome. Single-end alignments were then performed using mate 1 of all unaligned pairs (-v2). The count of ends of all pairedend and single-end alignments were determined and these counts were converted to bedGraph format for visualization.

RT-qPCR and northern blot analyses
Total RNA from mouse heart was extracted using the TRIzol/chloroform extraction method. RNA was quantified with the Qubit 4.0 fluorometer (Invitrogen). RT-qPCR was used to determine Rnaseh1 and mitochondrial mRNA levels and northern blotting was used to assess 7S RNA levels. For the qPCR analysis, the RNA samples were treated with deoxyribonuclease (DNase) to eliminate potential DNA contamination. Next, the cDNA synthesis was performed with High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher) according to kit instructions. The qPCR was run in a QuantStudio 6 Flex Real-Time PCR System using TaqMan TM Universal Master Mix II, with UNG and Taqman probes, Rnaseh1, 12S, 16S, Atp6, Atp8, Cox1, Cox2, Cox3, CytB, Nd1, Nd2, Nd3, Nd4, Nd5, Nd6. ActinB was used for normalization. For northern blotting, total RNA (3 g) was separated on a neutral 10% polyacrylamide gel. Next, separated RNAs were transferred to Hybond-N + membrane and hybridized with a randomly ␣-[ 32 P]-dCTP-labelled dsDNA probe to detect 5.8S rRNA and a strand-specific oligonucleotide probe labeled with ␣-[ 32 P]-dATP to detect 7S RNA. The membrane was used to expose a PhosphorImager screen.

In organello transcription assay
Freshly isolated mitochondria (1 mg) from mouse heart were washed twice in 0.5 ml of cold incubation buffer (see the in organello replication assay) supplemented with 1 mg/ml of fatty acid-free BSA and 1 mM ADP. Next, mitochondria were resuspended in 1 ml of prewarmed incubation buffer supplemented with 1 mg/ml of fatty acid-free BSA, 1 mM ADP and 60 Ci ␣-[ 32 P]-UTP. Samples were further incubated 1.5 h at 37 • C on a rotating wheel. Mitochondria were pelleted, resuspended again in prewarmed incubation buffer containing 1 mg/ml of fatty acid-free BSA, 1 mM ADP and non-radioactive UTP and incubated for an additional 10 min at 37 • C. Finally, samples were washed three times in washing buffer (see the in organello replication assay). The mtRNAs were extracted using the TRIzol/chloroform extraction method. Samples were separated in 1% MOPS-formaldehyde agarose gel, transferred to Hybond-N+ membrane and the signal was visualized by autoradiography. To ensure equal input, aliquots of mitochondrial proteins, taken before mtRNA isolation, were separated by SDS-PAGE and transferred to a PVDF membrane which was hybridized with an antibody against the mitochondrial HSP60 protein.

Transcriptome and small RNA-seq analyses
Total RNA was extracted from control and Rnaseh1 knockout hearts from 24-week-old mice by using the Qiagen miRNeasy kit according to the kit instructions for transcriptome analysis. Mitochondrial RNA was extracted with the same kit from isolated heart mitochondria of 24-weekold mice for small RNA-seq analysis. For TruSeq, pairedend reads were aligned to the mouse genome (mm10) with HISAT2 (32) v2.10 using default parameters. Properly paired, primary alignments were retained and read pairs aligned to the mitochondrial genome were realigned with the -no-spliced-alignment parameter enabled to remove spurious mitochondrial splice junctions. Gene counts were produced with featureCounts (33) v1.6.3 (-C -p -B -P -s 2) using the Ensembl GRCm38 v94 annotation combined with an in-house mitochondrial annotation. Differential expression was analyzed with DESeq2 (34) v1.18.1 in R v3.5.1. For small RNA, single-end reads were aligned to the mouse genome (mm10) with bowtie2 (35) v2.3.4.1 using default parameters. Gene counts were produced with feature-Counts v1.6.3 (-s 1) using the Ensembl GRCm38 v94 annotation combined with an in-house mitochondrial annotation. Differential expression was analyzed with DESeq2 v1.18.1 in R v3.5.1. Strand-specific coverage profiles were generated with samtools (36) and bedtools (37) genomecov and normalized to total reads mapped to mitochondria and the generated data was visualised with Circos (38).

Western blot analyses
For western blot analysis, mitochondria from the mouse heart or total-cell lysates from skin-derived fibroblasts (10-25g) were resuspended in 4x Lämmli buffer (4% SDS, 20% glycerol, 120 mM Tris, 0.02% Bromophenol Blue), separated by SDS-PAGE (4-12% Bis-Tris Protein Gel, Invitrogen) and transferred to PVDF membranes. Immunoblotting was done according to standard procedures. Primary antibodies used for immunoblotting are listed in the Supplementary Table S1.

Protein purification
The coding sequence of human RNase H1 carrying a Cterminal 6 × His-tag without the predicted mitochondrial targeting sequence (residues 1-26) was cloned into a pBac-PAK9 vector (Clontech). The mutant RNase H1 versions, Y29C and V142I were generated by site-directed mutagenesis (Agilent Technologies). Recombinant wild-type and mutant human RNase H1 proteins were expressed in Sf9 insect cells as previously described (40). The proteins were purified over HIS-Select Nickel Affinity Gel (Sigma-Aldrich) pre-equilibrated with 10 mM imidazole in buffer A (25 mM Tris-HCl pH 7.5, 400 mM NaCl, 10% glycerol, 10 mM ␤-mercaptoethanol and 1× protease inhibitors), followed by elution with 250 mM imidazole in buffer A. The proteins were subsequently purified over HiTrap Heparin (GE Healthcare) followed by HiTrap SP (GE Healthcare). The HiTrap columns were equilibrated in 0.2 M NaCl purification buffer (25 mM NaHPO 4 pH 7.1, 10% glycerol, 1 mM DTT and 1× protease inhibitors), and the proteins were eluted in a 20-ml salt gradient (0.2-1.2 M NaCl in purification buffer). All other recombinant proteins were expressed and purified as described previously (41).

Nuclease activity assay
Nuclease activity reactions were performed on an 80 nt DNA template annealed to a 3 end labelled 52 nt chimeric oligonucleotide that contained 26 ribonucleotides followed by 26 deoxyribonucleotides (26RNA:26DNA). The reactions (20 l) contained 1× RNase H buffer (50 mM Tris-HCl pH 8.0, 75 mM KCl, 3 mM MgCl 2 and 10 mM DTT), 20 fmol substrate, and increasing concentrations of wildtype or mutant RNase H1 proteins (10-300 fmol). The reactions were incubated at 37 • C for 60 min and then stopped by the addition of 5 l stop buffer [10 mM Tris-HCl pH 8.0, 0.2 M NaCl, 1 mM EDTA, 660 g/ml glycogen (Roche), and 100 g/ml proteinase K (Ambion)] followed by incubation at 42 • C for 45 min. The samples were recovered by ethanol precipitation in the presence of 0.5-volume ammonium acetate (7.5 M) and dissolved in 10 l of TE buffer. The samples were then treated with 300 mM KCl or KOH for 2 h at 55 • C and stopped with 2× stop buffer (formamide with 10 mM EDTA, 0.025% bromophenol blue and 0.025% xylene cyanol).

In vitro transcription
All transcription reaction volumes were 25 l and con- NaCl, 1 mM EDTA, 100 g/ml glycogen (Roche) and 100 g/ml proteinase K (Ambion)] followed by incubation at 42 • C for 45 min. The transcripts were recovered by ethanol precipitation and the pellets were dissolved in 20 l gel loading buffer (98% formamide, 10 mM EDTA, 0.025% xylene cyanol FF and 0.025% bromophenol blue) and heated at 95 • C for 3 min. The samples were analyzed on 4% denaturing polyacrylamide gels (1× TBE and 7 M urea) followed by exposure on photo film. Low Molecular Weight DNA Ladder (NEB) was used as a size marker. g/ml glycogen (Roche) and 100 g/ml proteinase K (Ambion) followed by incubation at 42 • C for 45 min. The reactions were treated with KOH (300 mM) for 2 h at 55 • C. The samples were recovered by ethanol precipitation in the presence of 0.5 volumes ammonium acetate (7.5 M), dissolved in 10 l gel loading buffer (98% formamide, 10 mM EDTA, 0.025% xylene cyanol FF, and 0.025% bromophenol blue), and heated at 95 • C for 3 min. The products were analyzed on 6% denaturing polyacrylamide gels (1× TBE and 7 M urea) for samples with dCTP or 12% denaturing polyacrylamide sequencing gels (1× TBE and 7 M urea) for samples with ddCTP. The gels were exposed on photo film.

Oligonucleotides
For all in vitro studies, oligonucleotides were PAGE purified (Eurofins MWG Operon) with the exception of the primer extension oligonucleotide. Oligonucleotides were radiolabelled on the 3 -end. A 51 nt long 26RNA:25DNA oligonucleotide was radiolabelled to generate a 52 nt long 26RNA:26DNA oligonucleotide using terminal transferase (NEB) and ␣-[ 32 P]-dCTP, followed by 3 overhang removal using T4 DNA polymerase (NEB).

RNase H1 localizes to the mitochondrial matrix
RNase H1 has been reported to be localized to both the nucleus and mitochondria in mice and humans (5,15). However, when we performed subcellular fractionation of human embryonic kidney cells (HEK293T), we found that RNase H1 was present in the mitochondrial and cytosolic fractions but it was not detected in the nuclear fraction (Figure 1A). To determine the submitochondrial localization of RNase H1, we performed protease protection assays and found that RNase H1 was not degraded by proteinase K added to mitochondria with an intact or disrupted outer membrane. The outer membrane protein Tom20 and the intermembrane space protein AIF were degraded in mitochondria and mitoplasts, respectively. Upon lysis of the inner membrane by detergent treatment, RNase H1 and the mitochondrial matrix protein HSP60 were degraded by proteinase K (Figure 1B). These findings show that RNase H1 is present in the mitochondrial matrix. A recent highconfidence mitochondrial proteome of HEK293 cells provides independent support for mitochondrial localization of RNase H1 (42). It is important to point out that our findings cannot exclude that a minor portion of RNase H1 also may be localized to the nucleus.

RNase H1 is essential for embryo development and maintenance of adult differentiated tissues in mice
To study the in vivo role of RNase H1, we generated a conditional mouse knockout allele where exon II of the Rnaseh1 gene was flanked with loxP sites ( Figure 1C). Heterozygous knockout mice (Rnaseh1 +/-) were obtained by crossing Rnaseh1 +/loxP mice with β-actin-cre expressing mice. Intercrosses of Rnaseh1 +/mice produced no viable homozygous knockouts (genotyped offspring = 117; Rnaseh1 +/+ , n = 47; Rnaseh1 +/-,n = 70; Rnaseh1 -/-,n = 0). In line with a previous study (15), we observed that Rnaseh1 was essential for embryonic development. Homozygous knockout embryos (Rnaseh1 -/-) were small and poorly developed in comparison with wild-type (Rnaseh1 +/+ ) embryos at E9.5 ( Figure  1D). Furthermore, embryos analyzed at E8.5 had an almost complete depletion of mtDNA ( Figure 1E). To study the in vivo role of RNase H1 in a differentiated tissue, we proceeded to generate mice with disruption of Rnaseh1 in heart and skeletal muscle (Genotype: Rnaseh1 loxP/loxP ; +/Ckmmcre; hereafter denoted Rnaseh1 knockout mice). Reverse transcription PCR (RT-PCR) analyses showed that exon 2 was lacking in Rnaseh1 transcripts in the knockout hearts (Supplementary Figure S1A). And the real-time quantitative PCR (qPCR) showed that Rnaseh1 transcript levels were significantly reduced (Supplementary Figure S1B), consistent with degradation of the Rnaseh1 transcripts by nonsense-mediated RNA decay in Rnaseh1 knockout hearts. The Rnaseh1 knockout mice had reduced lifespan with maximal longevity of ∼24 weeks ( Supplementary Fig-ure S1C). From 8 weeks onwards, Rnaseh1 knockout animals developed a gradual increase in the heart to body weight ratio indicating progressive cardiomyopathy ( Figure  1F). These results establish that RNase H1 is essential for both embryonic development and normal heart function in mice.

mtDNA depletion and formation of linear deleted mtDNA in RNase H1 knockouts
Analysis of mtDNA levels by qPCR in hearts lacking RNase H1 revealed a marked decrease at 8 weeks of age (∼22-31% of wild-type levels), progressing to a profound reduction at 24 weeks of age (∼4% of wild-type levels) (Figure 2A). The mtDNA depletion observed by qPCR was confirmed by Southern blot analyses ( Figure  2B and Supplementary Figure S1D). Interestingly, linear deleted mtDNA fragments were present on Southern blots of RNase H1 knockout hearts ( Figure 2B

Defective mtDNA replication in Rnaseh1 knockout mice
The depletion of mtDNA and the aberrant replication product in Rnaseh1 knockout mice prompted us to further investigate the regulation of mtDNA replication. Analysis by in organello replication assays showed that linear deleted mtDNA fragments were generated as by-products during ongoing mtDNA replication in heart mitochondria lacking RNase H1 ( Figure 2C). To further investigate mtDNA replication, we assessed the steady-state levels of 7S DNA, the short nascent DNA molecule formed by premature replication termination at the end of the control region resulting in D-loop formation (46). Interestingly, we found a drastic reduction of 7S DNA levels in mouse hearts lacking RNase H1 even when Southern blots were overloaded to compensate for the mtDNA depletion ( Figure 2D). Moreover, in organello mtDNA replication assays of heart mitochondria from Rnaseh1 knockout mice showed no de novo synthesis of a distinct 7S DNA species ( Figure 2C). Next, we performed 2D neutral-neutral agarose gel electrophoresis (2DNAGE) to analyze mtDNA replication products initiated from O H and found increased levels of stalled mtDNA replication intermediates ( Figure 2E), consistent with generalized inhibition of mtDNA replication progression. The 2DNAGE analysis probed with a restriction fragment containing O H showed the presence of high molecular weight arcs resulting from blockage of the restriction sites, presumably due to unprocessed RNA primers still annealed to the DNA at the BclI restriction site immediately downstream of the light strand promotor (LSP) ( Figure 2E). We also observed site-specific stalling near O L in the absence of RNase H1 ( Figure 2E). To develop a global overview Welch's t-test; n = 7-9 animals per group; *P < 0.05; **P < 0.01; ***P < 0.001. (F) Heart-to-body weight ratio of conditional knockout mice (L/L, cre) and controls (L/L). Data are represented as mean ± SEM; Welch's t-test; n = 5-11 animals of each genotype per age group; *P < 0.05; **P < 0.01; ***P < 0.001. of putative mtDNA alterations in the Rnaseh1 knockout mice, we performed deep sequencing using the recently developed Mitochondrial Structural Alterations (MitoSAlt) high-throughput computational pipeline for analysis (30). We found no circular mtDNA molecules with deletions or mtDNA duplications in the Rnaseh1 knockout hearts (Figure 2F), whereas simultaneous analysis of Mgme1 knockout brains, used as a control, identified multiple, short duplications in the non-coding region (NCR), as previously described (30). Our results thus show that the marked mtDNA depletion in Rnaseh1 knockout mice is a direct consequence of mtDNA replication defects.

RNase H1 directs site-specific primer formation for initiation of mtDNA replication
We performed next-generation sequencing of Rnaseh1 knockout heart DNA to map the ends of the linear deleted mtDNA molecules. The linear deletions were shown to contain sequences corresponding to the major arc region of mtDNA ( Figure 3A), similar to the findings in MGME1 knockout mice (43). However, the regions close to O H and O L were comparatively underrepresented in the linear deleted mtDNA molecules from Rnaseh1 knockout hearts ( Figure 3A). This sharp decrease in sequence coverage observed at O L ( Figure 3A) is in line with the replication stalling we observed at O L on 2DNAGE ( Figure 2E). We proceeded to use hydrolytic DNA end sequencing (HydEn) (19,31) and found peaks indicating DNA ends located a few nucleotides downstream of O L in the Rnaseh1 knockout hearts ( Figure 3B). Analysis of HydEn-seq data in the D-loop region revealed peaks at O H in both wild-type and Rnaseh1 knockout hearts ( Figure 3D). However, additional peaks were found in the termination associated sequence (TAS) region in the Rnaseh1 knockout hearts (Figure 3D), pointing to novel 5 DNA ends. These peaks are in the region where a drop in mtDNA sequence coverage is observed (Figure 3A, C). As an independent means of detecting 5 DNA ends of the nascent H-strand, we performed primer extension assays of DNA from wild-type and Rnaseh1 knockout hearts. Primer extension with a primer complementary to the H-strand of the CytB gene (position 15181-15201) generated terminated products at different locations in the D-loop region in Rnaseh1 knockout hearts that were absent in control mice ( Figure 3E). These results are consistent with the presence of alternate sites for initia-

Severe mitochondrial dysfunction in Rnaseh1 knockout mice
Given reports that RNase H1 is important for removal of RNA:DNA hybrids formed during transcription of mtDNA (20,22), we monitored the effects of Rnaseh1 knockout on mitochondrial gene expression. Consistent with the progressive mtDNA depletion (Figure 2A), the steady-state levels of mitochondrial transcripts in hearts of Rnaseh1 knockout mice were slightly lower at 8 weeks and drastically lower at 24 weeks of age ( Figure 4A). TruSeq and small RNA sequencing showed a general decrease in both long and short mtDNA-encoded transcripts in 24-week-old Rnaseh1 knockout mice ( Figure 4B, C). No processing defects of mitochondrial transcripts were found ( Figure 4B). Northern blot analyses showed a similar age-dependent reduction in the steady-state levels of 7S RNA in the knockout hearts ( Figure 5A), correlating with the decrease of mtDNA levels (Figure 2A). The mitochondrial transcription elongation factor (TEFM) is required for the transition from transcription initiation to elongation, and in its absence a range of abnormally short 7S RNAs accumulate (47) (Figure 5A). In contrast, loss of RNase H1 does not appear to affect transcription elongation as only full length 7S RNAs were found ( Figure 5A). Consistent with these results, there is a gradual decline in the abundance of promoter-distal transcripts in heart-and skeletal-muscle-specific Tefm knock- out mice (47), but this pattern was not seen in Rnaseh1 knockouts ( Figure 4B, C) further arguing against a transcription elongation defect in the absence of RNase H1. In organello transcription assays also showed significantly decreased de novo transcription in isolated heart mitochondria of Rnaseh1 knockout mice at the age of 16 and 24 weeks ( Figure 5B). Loss of RNase H1 thus leads to decreased levels of mitochondrial transcripts, likely secondary to the decrease of mtDNA templates, without evidence for defective transcription elongation or impaired transcript processing. Consistent with the depletion of mtDNA and the global decrease in the levels of mtDNA-encoded transcripts, the steady-state levels of OXPHOS subunits were profoundly decreased in 24-week-old Rnaseh1 knockout mice (Supplementary Figure S3A). Histochemical analysis of the combined cytochrome c oxidase/succinate dehydrogenase (COX/SDH) enzyme activity in skeletal muscle showed profound COX deficiency in Rnaseh1 knockout mice (Supplementary Figure S3B). On blue native polyacrylamide gels (BN-PAGE), there was a profound decrease in complex I and IV (COX) activity in Rnaseh1 knockout hearts at the age of 24 weeks (Supplementary Figure S3C). As expected, the levels of mitochondrial transcription factor A (TFAM), which normally follow the levels of mtDNA (48,49), and the levels of the leucine-rich pentatricopeptide repeat protein (LRPPRC), which typically follows the mitochondrial mRNA levels (50), were both decreased in hearts of 24 weeks old Rnaseh1 knockout mice (Supplementary Figure  S3A). In contrast, the levels of mitochondrial RNA polymerase (POLRMT) were increased (Supplementary Figure  S3A), indicating that loss of RNase H1 leads to compensatory transcription increase, as seen in other mouse mutants affecting mtDNA expression (47,51,52).
In summary, our results demonstrate that loss of RNase H1 affects mtDNA replication in mice, which, in turn, leads Nucleic Acids Research, 2022, Vol. 50, No. 15 8759  to a severe decrease of mitochondrial transcript levels and profoundly impaired OXPHOS capacity.

A novel pathogenic mutation affecting the hybrid-binding domain of RNase H1
An important insight into the in vivo function and mechanism of action of RNase H1 comes from studies in patients with pathogenic mutations in RNASEH1 reducing enzyme activity. These patients often have mtDNA depletion similar to the findings in Rnaseh1 knockout mice (15) (Figure 2A). However, unlike Rnaseh1 knockout mice, patients with RNASEH1 mutations do not accumulate linear deleted mtDNA molecules but instead have high amounts of circular deleted mtDNA molecules in skeletal muscle (23)(24)(25)(26). This difference may be explained by the complete absence of RNase H1 in the knockout mice, whereas patients express an RNase H1 protein with impaired function. In contrast, most patients described with MGME1 variants carry loss-of-function mutations, potentially explaining the greater similarity in mtDNA phenotypes between patients and knockout mice (43,53). Patients with RNASEH1 mutations are typically homozygous or compound heterozygous for mutations in the catalytic (H) domain or compound heterozygous for mutations in the H-domain and the connection domain (CD) (Figure 6A). In a recently diagnosed pair of siblings, sister and brother, we found a homozygous c.86A > G mutation (Y29C) in the region of the RNASEH1 gene encoding the hybrid-binding domain (HBD) of RNase H1 ( Figure 6A). No mutations have previously been described in this domain and both patients developed very similar phenotypes with parkinsonism and cognitive deterioration, which have not previously been reported in patients with pathogenic RNASEH1 mutations. Also, classical mitochondrial phenotypes were observed in both siblings, i.e. ptosis, PEO, peripheral neuropathy, sensorineural deafness, and mitochondrial myopathy with abundant COX-negative fibers ( Figure 6B), multiple mtDNA deletions (Supplementary Figure S4A) but normal mtDNA copy number (Supplementary Figure S4B). Southern blot analyses in fibroblasts from the Y29C patient and another patient with a previously described V142I mutation of RNase H1 also showed normal mtDNA content ( Figure 6C) (23). However, in organello replication assays showed decreased de novo mtDNA synthesis in mitochondria of both Y29C and V142I patient-derived fibroblasts ( Figure 6D), demonstrating that malfunction of either the HBD or H-domain of RNase H1 impairs mtDNA replication. As previously reported for patient fibroblasts carrying the V142I amino acid substitution of RNase H1 (23), no linear deleted mtDNA was seen in fibroblasts derived from the Y29C patient (Figure 6C). As a control sample, we included fibroblasts from a patient affected by early-onset PEO and mitochondrial myopathy with sparse COX-negative fibers (Supplementary Figure S4C) caused by a novel loss-of-function mutation (S188YfsX6) in MGME1. This patient had linear deleted mtDNA molecules in fibroblasts ( Figure 6C) and skeletal muscle (Supplementary Figure S4A), similar to the findings in other patients lacking MGME1 (45). Interestingly, the steady-state levels of 7S DNA were increased in fibroblasts from the RNASEH1 patient with the V142I mutation, as previously described (23), and remained normal in the patient with the Y29C mutation of RNase H1 ( Figure 6E). A normal amount of 7S DNA was also observed in the skeletal muscle biopsy specimen (Supplementary Figure S4D). The patient with the S188YfsX6 mutation of MGME1 had increased levels of 7S DNA in both fibroblasts and skeletal muscle ( Figure 6E and Supplementary Figure S4D) consistent with findings in other patients with MGME1 mutations (53). In contrast to the findings in patients, Rnaseh1 knockout mice showed a complete lack of 7S DNA ( Figure 2C, D). The steady-state levels of RNase H1 protein were significantly reduced in fibroblasts from patients with the two different RNASEH1 mutations ( Figure 6F), showing that both the Y29C and V142I amino acid substitutions of RNase H1 impair not only protein function but also protein turnover.

Tightly regulated RNase H1 activity is critical for replication initiation in vitro
Based on the different disease phenotypes, we compared the activity of the RNase H1 variants in vitro. First, we characterized the RNase H1, RNase H1:Y29C and RNase H1:V142I variants on a defined RNA:DNA hybrid template. RNase H1:Y29C was more active than the wild-type enzyme ( Figure 7A), whereas RNase H1:V142I had a sig-  nificantly reduced cleavage activity, as previously described (19). We have previously demonstrated that transcripts from LSP need to be processed by RNase H1 to promote sitespecific initiation of mtDNA replication at O H in vitro and we therefore investigated whether the mutant versions of RNase H1 can process R-loops formed by the transcription machinery. Surprisingly, we found that the activity of RNase H1:Y29C was increased relative to the wild-type enzyme, whereas RNase H1:V142I had significantly reduced cleavage activity ( Figure 7A, B). However, when we employed an in vitro replication assay, fewer initiation products were formed by RNase H1:Y29C ( Figure 7C) despite its increased processing activity on RNA:DNA templates (Figure 7A, B). Our findings thus show that the catalytic activity of RNase H1 does not necessarily correlate with its ability to form replication primers, and that the HBD and H-domain must act in a coordinated way to generate the RNA primers necessary for initiation of mtDNA replication.

DISCUSSION
Mammalian RNase H1 has an essential role in mtDNA replication (15), but it has been unclear at which exact step it acts in vivo. Here, we demonstrate that RNase H1 function is crucial for site-specific restriction of replication primer formation to limit initiation of mtDNA synthesis to the replication origins ( Figure 8A). In addition, RNase H1 has a role during completion of mtDNA replication as demonstrated by the impaired replication fork progression and the formation of linear deleted mtDNA fragments in the absence of RNase H1 ( Figure 8B, C). According to the widely accepted strand-displacement model (SDM) of mtDNA replication (54), replication of the leading (H) strand commences at O H and proceeds unidirectionally until approximately two thirds of the H-strand has been synthesized. At this point, the region surrounding O L becomes single stranded and forms a stem loop structure where lagging (L) strand synthesis initiates. Replication initiation at O H depends on transcription from LSP that results in an R-loop that is processed by RNase H1 to form RNA primers (19). The 3 RNA ends that are formed after processing by RNase H1 are used by POL␥ to initiate DNA synthesis (19). In the absence of RNase H1, R-loop structures formed by transcription from LSP remain unprocessed and origin-specific replication initiation does not occur at O H . Instead, we report here that mtDNA replication  (B) RNase H1 is also required for replication completion of both the H and L strands (L strand replication completion is not shown). Loss of RNase H1 prevents digestion of the RNA primer residues attached to the 5 end of the nascent DNA strand. After completion of mtDNA replication the RNA will remain attached to the 5 end of the DNA and will prevent DNA ligation. (C) Formation of linear mtDNA molecules with deletions in Rnaseh1 knockout mice. Impaired ligation of the nascent H strand generates mtDNA molecules with a nick. In the next replication round, the nicked H strand becomes the template for L strand synthesis and replication of the L strand will therefore have an increased risk for premature termination near the TAS region. As a result, the ssDNA fragment spanning the minor mtDNA arc will be degraded leaving the linear, dsDNA fragment covering the major mtDNA arc.
is initiated at specific non-canonical sites downstream of O H and O L in the absence of RNase H1 ( Figure 8A). In agreement with our data, unspecific replication initiation from multiple sites unrelated to O H was observed in fibroblasts from a patient with a homozygous disease-causing mutation in RNase H1 (19) and another replication initiation site, referred to as ori-b was reported in study in MEFs lacking RNase H1 (16). Moreover, the requirement for RNase H1 in the replication primer maturation process is conserved in evolution as it also has been demonstrated at the ColE1 replication origin in bacteria (55).
We have previously shown that mtSSB plays a major role in controlling the replication/transcription switch at O H by promoting the formation and stabilization of the R-loop transcribed from LSP (52). In the absence of mtSSB, no R-loop is formed despite a strong increase in transcrip-tion from LSP, and mtDNA cannot be maintained at normal levels (52). These findings argue that mtSSB, in addition to its role in coating single-stranded mtDNA during replication, also provides a switch between transcription for primer formation and gene expression. In contrast, transcripts from LSP are terminated very close to the promoter in the absence of TEFM and both replication primer formation and transcription for gene expression are drastically impaired (47). Our data provide additional information on the mechanism for replication primer formation as we show that RNase H1 is necessary for generating the 3 RNAends of the replication primers in vivo. Moreover, RNAseq shows that RNase H1 is not important for removing R-loops, that potentially would block transcription elongation, or for ribosomal RNA processing, at variance with other reports (20)(21)(22). Given the potent strand displacement activity of POLRMT (56), we propose that this conserved structural feature may explain why POLRMT alone can be responsible for negotiating at least some R-loops formed as by-products of mitochondrial transcription.
In addition to the well documented role for RNase H1 in replication initiation, it is also likely important for mtDNA replication completion. When mtDNA replication has progressed around the circle, POL␥ will encounter and displace the 5 end where replication started to create a ∼100 nucleotides long DNA flap (57,58). Once replication is initiated, RNase H1 will cleave the remaining RNA primer hybridized to the template DNA. However, in the absence of RNase H1, the RNA primer will remain attached to the DNA flap displaced by POL␥ (16). Normally, the DNA flap is processed by MGME1, but this processing may be impaired if the RNA primer remains attached to the 5 end. It is important to recognize that impaired function of enzymes involved in determining the length and structure (POL␥ , RNase H1) or degradation (MGME1) of the DNA flap all cause the formation of linear deleted mtDNA molecules extending between the two origins of mtDNA replication (43)(44)(45)59). Collectively, published studies and the results we present here strongly argue for the model where the linear deleted mtDNA molecules are formed during mtDNA replication as the result of persistent DNA flaps (60) or RNA residues attached to the 5 end of the flaps that prevent ligation ( Figure 8B). Ligation is likely impaired at both origins because of persisting DNA flaps, which means that a fraction of all mtDNA replication events cannot be completed causing the formation of linear deletions ( Figure  8C). According to this model, the linear deleted mtDNA molecules are constantly formed but cannot undergo replication as they lack both origins of mtDNA replication.
Eukaryotic RNase H1 has an HBD, a CD and an Hdomain (61) ( Figure 6A). The HBD binds double-stranded RNA and RNA:DNA-hybrids (62)(63)(64)(65)(66) to ensure enzyme processivity. Deletion of the HBD generates an enzyme that has a more distributive cleavage activity, similar to the activity of the E. coli RNase H1 enzyme (65,66) that lacks the HBD. We show that a functioning HBD and H-domain are required for mammalian RNase H1 activity in vivo. Skinderived fibroblasts from mitochondrial disease patients carrying mutations in the HBD or H-domain have defects in de novo mtDNA replication. Interestingly, mutations in the HBD and H-domain have opposite effects on the RNase H1 cleavage activity in vitro, causing significantly increased or drastically decreased enzyme activity, respectively. These results show that the RNase H1 activity must be tightly regulated in vivo. The HBD of RNase H1 may therefore not only have a role in controlling enzyme processivity, as previously proposed (65,66), but may also be necessary to position the enzyme for site-specific primer processing for mtDNA replication initiation.
In conclusion, we demonstrate that a carefully balanced RNase H1 activity is essential for regulating two crucial steps of mtDNA replication in vivo, i.e. replication initiation and replication completion. Lack of RNase H1 activity leads to mtDNA replication defects and the formation of linear deleted mtDNA fragments but does not directly affect transcription or posttranscriptional RNA processing in mammalian mitochondria.

DATA AVAILABILITY
The mtDNA sequencing data, HydEn sequencing data and RNA sequencing data have been deposited in Gene Expression Omnibus under accession number GSE199771, GSE199198 and GSE199147, respectively.