In vivo and in vitro mechanistic characterization of a clinically relevant PolγA mutation

Mutations in POLG, encoding POLγA, the catalytic subunit of the mitochondrial DNA polymerase, cause a spectrum of disorders characterized by mtDNA instability. However, the molecular pathogenesis of POLG-related diseases is poorly understood and efficient treatments are missing. Here, we generated a POLGA449T/A449T mouse model, which reproduces the most common human recessive mutation of POLG, encoding the A467T change, and dissected the mechanisms underlying pathogenicity. We show that the A449T mutation impairs DNA binding and mtDNA synthesis activities of POLγ in vivo and in vitro. Interestingly, the A467T mutation also strongly impairs interactions with POLγB, the homodimeric accessory subunit of holo-POLγ. This allows the free POLγA to become a substrate for LONP1 protease degradation, leading to dramatically reduced levels of POLγA, which in turn exacerbates the molecular phenotypes of PolgA449T/A449T mice. Importantly, we validated this mechanism for other mutations affecting the interaction between the two POLγ subunits. We suggest that LONP1 dependent degradation of POLγA can be exploited as a target for the development of future therapies.


Introduction
POLG, encoding the catalytic subunit of mitochondrial DNA-specific polymerase gamma (POL), contains the highest number of deleterious mutations of the human coding genome, associated with a huge spectrum of clinical, molecular and biochemical phenotypes. Autosomal dominant and recessive mutations have been reported, and among the recessive ones, by far the most common is a missense mutation changing A467 into a T, in an area of the protein which has no obvious function.
Nevertheless, the consequences of this mutation are devastating in humans, including severe, juvenile neurodegeneration of the spinal cord, brainstem, neocortex, particularly in the occipital pole, associated with ataxia, refractory epilepsy, toxicity to valproate, cognitive regression and eventually global neurological impairment and death. Status epilepticus is a common end conclusive outcome of the homozygous patients. The A467T is the most frequent POL mutation in Scandinavian and Northern European Countries, together with another change, the W748S change, which seems to be part of the Finnish disease heritage. In both cases, no obvious alteration of relevant known function of the primary structure of the protein seems to be affected, neither the proofreading activity present in the N-terminus of the protein, downstream from the mitochondrial targeting sequencing, nor the polymerase domain, which is confined to the C-terminus of the sequence. The intermediate region, where the two deleterious mutations are located, is generically defined as the "linker" region between the proofreading and the polymerase domain, and seems to play a role, not better specified, in binding the second component of POL, the 55 kDa subunit PolB, which binds homodimerically POLA, conferring to the whole trimeric mtDNA replisome a marked increase in polymerase processivity. Therefore, the mechanistic abnormalities by the two most common recessive mutations affecting the activity of this important enzyme are virtually unknown. Notably, the A467T mutation, when associated with an allelic null mutation leads to an even more devastating condition named Alpers-Huttenlocher syndrome, which affects babies suffering of refractory epilepsy due to severe spongiotic atrophy of the brain, and hepatic failure. As already mentioned, in addition to the A467T and W748S, over 300 mutations have been described in POLG (Human DNA Polymerase Gamma Mutation Database: https://tools.niehs.nih.gov/polg/) leading to a spectrum of diseases, which has been tentatively classified in the following categories: (i) the already mentioned Alpers-Huttenlocher syndrome (AHS), (ii) myocerebrohepatopathy spectrum (MCHS), which presents with developmental delay, lactic acidosis, myopathy and hepatic impairment; (iii) the typical spectrum of the A467T and W748S mutations (in either homozygosity or in combination), including myoclonic epilepsy myopathy sensory ataxia (MEMSA), comprising spinocerebellar ataxia with epilepsy (SCAE), frequently associated with sensory ataxia neuropathy with dysarthria and ophthalmoplegia (SANDO), and, (iv) finally, autosomal dominant and recessive progressive external ophthalmoplegia (ad and arPEO) (Rahman & Copeland, 2019). The plethora of terms defining the different conditions of POL mutations reflects the huge spectrum of syndromic presentations associated with abnormalities of this essential enzyme. Consequently, it is difficult to rigidly classify these syndromes as often symptoms overlap, the same mutation can be associated to more than one presentation, and each phenotype can be the consequence of the combination of different allelic mutations. In several cases, particularly in syndromes associated with progressive external ophthalmoplegia, abnormalities of the mtDNA integrity lead to the accumulation of multiple mtDNA deleted species; in other cases, such as the A467T mutation itself, these large scale rearrangements are rare, but whenever necropsy examination has been performed, depletion of mtDNA was found in critical areas, particularly in specific regions of the brain (Tzoulis, Tran et al., 2014). Again, the mechanistic details leading to the generation of these molecular lesions is poorly understood.
As already mentioned, DNA polymerase  (POL) is a heterotrimer with one catalytic POLA subunit and two POLB accessory subunits (Gustafsson, Falkenberg et al., 2016). The POLG gene codes for the 140 kDa POLA subunit that harbors DNA polymerase, 3-5 exonuclease, and 5-deoxyribose phosphate lyase activities (Longley, Prasad et al., 1998), whereas POLG2 encodes the 55 kDa POLB, which stabilizes the interactions with template-DNA, thereby increasing processivity (Lim, Longley et al., 1999). POLγ is the only DNA polymerase required for mtDNA replication in mammalian mitochondria and, at the replication fork, it works in concert with the TWINKLE DNA helicase (Spelbrink, Li et al., 2001). The mitochondrial single-stranded DNAbinding protein (mtSSB) stimulates mtDNA synthesis by increasing the helicase activity of TWINKLE and the DNA synthesis activity of POL (Korhonen, Gaspari et al., 2003).
Four mutations (A467T, W748S, G848S and the T251I-P587L allelic pair) account for ~50% of all mutations identified in patients with POLG-related diseases, with ~75% of patients carrying at least one of these mutant alleles (Uusimaa, Gowda et al., 2013). POLG mutations may lead to mtDNA instability, causing either multiple deletions or depletion 2 . However, there is no obvious genotypephenotype correlation, and, as mentioned above, the same mutation can often lead to mtDNA deletions, mtDNA depletion or both. A prototypical example is highlighted by the homozygous mutation A467T mutation, which has been associated with a range of phenotypes, from childhoodonset fatal AHS to MEMSA, ANS and SANDO (Rahman & Copeland, 2019). In addition, the age of onset and the progression of POLG-related disease in patients with the same POLG mutations is astonishingly variable and can span several decades. For instance, the onset of disease spans >70 years in compound heterozygous patients carrying the T251I-P587L mutations on one allele (DeBalsi, Longley et al., 2017) and the G848S mutation on the other, and it spans at least four decades of life in homozygous A467T patients (Rajakulendran, Pitceathly et al., 2016, Tzoulis, Engelsen et al., 2006. The A467T is the most common pathogenic variant of POLG (de Vries, Rodenburg et al., 2007, Ferrari, Lamantea et al., 2005, Horvath, Hudson et al., 2006, Nguyen, Sharief et al., 2006, and it seems to severely impair POLγ activity (Chan, Longley et al., 2005).
The proposed, but not firmly established, mechanisms include (i) reduction of the affinity of the enzyme for deoxynucleotide triphosphates (dNTPs), (ii) lowering of the catalytic activity, and (iii) reduction of the affinity for the POLG2-encoding accessory subunit. All this would lead to stalling of the replication fork and depletion or instability of mtDNA. Nevertheless, mechanistic evidence of these effect remains largely hypothetical. It is therefore crucial for the understanding of the pathogenic mechanisms encompassing this huge area of neurodegenerative disorders, and the dissection of the very function of several structures and interactions of POLA, to better understand the mechanistic processes leading to POL impairment in the presence of a change that seems not to compromise known essential function of the protein, but that nevertheless retains an obviously relevant physiological function and has an essential impact in the medicine associated with POL impairment.
Here, we generated a Polg A449T/A449T knockin mouse by CRISPR/Cas9 technology, which reproduces the A467T human mutation. By using a combination of cellular, in vivo and in vitro techniques, we show that the mutation not only impacts on polymerase activity, DNA binding and interaction with PolγB, but also makes the protein exquisitely susceptible to degradation by the LONP1 protease, the most important soluble protease of the mitochondrial matrix. Our results thus reveal a novel pathogenic mechanism for POLG-related diseases, which in turn suggests new avenues for development of future therapies.

Generation and characterization of Polg A449T/A449T mutant mice
To investigate the molecular pathogenesis of POLG-related disorders, we generated a Polg A449T/A449T homozygous knockin mouse, corresponding to the human A467T mutation, by CRISPR/Cas9 technology (Supplementary Figure 1). Three-month old Polg A449T/A449T homozygous animals did not show any gross phenotype compared to wild-type (WT) littermates, including similar body weight curve and rotarod performance (not shown). However, a significant reduction in treadmill motor endurance was detected ( Figure 1A). Although whole body metabolism was similar in KI and controls by CLAMS analysis, a significant reduction in spontaneous rearing movements was observed in Polg A449T/A449T mutants (Figure 1B-C and Supplementary Figure 2). Post-mortem hematoxylin and eosin staining did not show any gross abnormality in any tissue.

POLA is reduced in Polg A449T/A449T tissues
We analysed the effects of the A449T mutation on POLA and POLB protein levels.
Immunoblotting revealed strong reduction of PolA A449T amount, as low as 50%, in all tissues examined, including brain, heart, skeletal muscle, liver and kidney (Figure 2A-B). In contrast, POLB levels were unchanged in most tissues, albeit a mild upregulation and downregulation in brain and heart, respectively, was observed (

Reduced mtDNA content and impaired replication in Polg A449T/A449T tissues and MEFs
Since mutations in POLG are associated with mtDNA instability in human patients, we next investigated mtDNA content and integrity in several tissues, including liver, skeletal muscle, brain, kidney and heart from both Polg A449T/A449T vs. WT littermates ( Figure 3A). MtDNA copy number was significantly reduced in the skeletal muscle of Polg A449T/A449T (80±4%, p˂0.01) compared to WT littermates, without accumulation of multiple deletions ( Figure 3B  To investigate in detail the effects on mtDNA replication we generated mouse embryonic fibroblasts (MEFs) from Polg A449T/A449T and WT cells. The mtDNA content was similar in the two genotypes (Supplementary Figure 3K). We then investigated mtDNA replication in MEFs, using 5-ethynyl-2´-deoxyuridine (EdU) staining in junction with an anti-DNA antibody to label replicating and total mtDNA. Interestingly, we observed a significantly increased fraction of replicating mtDNA molecules in Polg A449T/A449T vs. WT MEFs ( Figure 3C-F), indicating that more mtDNA foci were engaged in replication. Next, we used ethidium bromide (EtBr) to deplete mtDNA content, and found that after removal of EtBr, mtDNA content recovered to pre-treatment values within 3 days in WT MEFs, whereas no recovery at all was observed in the mutant cells ( Figure 3G), strongly indicating severely impaired mtDNA replication in stress conditions of Polg A449T/A449T mouse mitochondria.
Given the mild reduction in mtDNA copy number in the mutant mice, we decided to challenge them with a single injection of carbon tetrachloride (CCl 4 ), which induces acute liver damage, triggering liver cell division to repopulate the necrotic areas. Two days after the injection, both WT and Polg A449T/A449T showed extensive areas of necrosis (approximatively 35% of the liver), which was reduced to 6±0.46 % in WT mice after four days, whereas it was still above 10% in Polg A449T/A449T mice (10±1.15%, p˂0.05) ( Figure 3H-I). This result clearly indicates that cell replication is impaired in Polg A449T/A449T , likely due to lack of bioenergetic supply by impaired mitochondria in stress conditions. Since the antiepileptic drug valproic acid (VPA) is known to induce acute liver failure in patients with the A467T mutation in POL (Saneto, Lee et al., 2010, Stewart, Horvath et al., 2010, we treated our mice with VPA by daily oral gavage (300 mg/kg) for one week or in food pellets (1.5% VPA) for two month. Both WT and Polg A449T/A449T did not show any sign of hepatic failure or histological damage. However, whilst WT-treated mice showed increased mtDNA copy number compared to non-treated WT mice (124±9.7%, p˂0.05) after chronic treatment, the mtDNA content in Polg A449T/A449T -treated mice was similar to non-treated mutants (94.9±4.7%, p=0.4889), suggesting reduced mtDNA replication in response to VPA challenge ( Figure 3J).

Polg A449T/A449T mitochondria have reduced 7S DNA and accumulate replication intermediates
We then investigated mtDNA replication in the tissues of the mutant and control mice by Southern blot. Normally, about 95% of all replication events are prematurely terminated, generating a 650 nucleotide-long molecule, called 7S DNA (Bogenhagen & Clayton, 1978, Doda, Wright et al., 1981, Nicholls & Minczuk, 2014. In Polg A449T/A449T but not in WT littermate, the 7S DNA levels were significantly reduced in skeletal muscle and kidney, and a similar trend was also present in the other analysed tissues, except for the heart, ( Figure 4A To better investigate the mechanistic details of mtDNA replication, we then performed in organello replication experiments in isolated liver mitochondria ( Figure 4C), by pulse-labelling with -32 P-dATP. Although no obvious differences were detected in mtDNA replication rates between Polg A449T/A449T and WT mice ( Figure 4C), the signal due to long but incomplete mtDNA molecules was much more intense in the Polg mutant compared to WT samples, thus suggesting accumulation of replication intermediates (RIs) in the mutant vs. controls. Accordingly, we applied twodimension agarose gel electrophoresis (2D-AGE), which resolves DNA molecules based on size and shape, allowing a snapshot of the RIs. Notably, Polg A449T/A449T mice displayed an overall accumulation of the different types of RIs compared to WT animals ( Figure 4D  These novel data clearly demonstrate that the A449T mutation impairs mtDNA replication in both cultured cells and in vivo.

PolA A449T protein has reduced affinity for DNA which is partially rescued by PolB subunit
To further document the stalling phenotype of the A449T mutant in vitro, we expressed and purified WT and mutant PolA as recombinant proteins. First, we used an electrophoretic mobility shift assay (EMSA) to measure the binding of Pol to a primed DNA template. When alone, POLA A449T bound DNA ≈ 3.4 times more weakly than POLA WT ( Figure 5A and Supplemantary Figure 5A) and remained substantially lower than the WT also after the addition of POLB ( Figure   5B and Supplementary Figure 5B).

POLA A449T has reduced polymerase activity which is partially rescued by POLB
Next, we investigated POLA activities using a short DNA template annealed to a radioactively labelled primer. By performing the experiment across a range of dNTP concentrations, we could analyze both polymerase and exonuclease function. The exonuclease activity can digest the labelled primer, whereas the polymerase activity can elongate the primer and synthesize an additional short, 15 nucleotide stretch of DNA. As expected, at lower dNTP levels, POLA wt displayed 3'-5' exonuclease activity, but at higher concentrations, it switched to polymerase activity ( Figure 5C).
Addition of POLB reduced exonuclease activity and favored DNA synthesis even at lower dNTP concentrations ( Figure 5D). The mutant POLA A449T was completely inactive in isolation, most likely due to its inability to efficiently bind primed DNA ( Figure 5C). Nevertheless, addition of POLB restored the polymerase activities of POLA A449T , to levels similar to those observed with POLA wt (Figure 5D), whereas exonuclease activity was reduced also in POLA wt as a consequence of predominant polymerase activity measured in vitro ( Figure 5D).
To further challenge the system, we performed a DNA synthesis assays using a long circular ssDNA template of 3000 nts ( Figure 5E). POLA A449T displayed a clearly slower DNA synthesis rate compared to the POLA WT , even in the presence of the POLB subunit ( Figure 5F). In order to monitor the effects of the A449T mutation on replication of dsDNA, we used a template containing a 4 kb long dsDNA region with a free 3′-end acting as a primer ( Figure 5G). Addition of the TWINKLE DNA helicase was required to unwind the DNA and the reaction was stimulated by mtSSB ( Figure 5G). This reaction is absolutely dependent on POLB and once initiated, very long stretches of DNA can be formed. In this rolling circle replication assay, POLA A449T showed reduced polymerase DNA synthesis rate ( Figure 5H) compared to POLA WT , at all concentrations tested (Supplementary Figure 5C), demonstrating that POLA A449T has reduced polymerase activity.
This in vitro result is in perfect agreement with the stalling phenotype seen in vivo. A similar effect was obtained with the human POLA A467T ( Figure 5I).
Analysis of incorporated radiolabelled nucleotides over time indicated that the in vitro replication rates with 10 M dNTPs, were reduced to about 60% for POLA A449T compared to POLA WT (3.5 fmol/min vs 5.5 fmol/min) (Supplementary Figure 5D and 5E). Interestingly, the reduction was more pronounced with human POLA A467T compared to human POLA WT (1.4 fmol/min vs 5.3 fmol/min), than for the mouse equivalents, which could explain the more severe phenotype observed in patients (Supplementary Figure 5D and 5E).

POLA is unstable in absence of POLB
The amount of POLA A449T was reduced in the Polg A449T/A449T , which could also contribute to impaired mtDNA replication. To better understand the impact of the A449T mutation on protein stability, we performed a thermofluor stability assay and monitored temperature-induced unfolding of POLA WT and POLA A449T , both in absence and in presence of POLB ( Figure 6A and 6B). The stability assay revealed no major differences in the fluorescence profile between POLA WT and POLA A449T from 37°C upwards, but the fluorescence signal of POLA A449T was already higher than the WT at 25°C, clearly indicating that the mutant protein was already partially unfolded even at <37°C temperatures ( Figure 6A and 6B). Interestingly, the presence of the POLB had a dramatic stabilizing effect, by increasing the unfolding temperature of about 10 °C for both proteins ( Figure 6A and 6B). These data suggest that also POLA WT is partially unstable in the absence of POLB. Accordingly, a recent report demonstrated that human POLB-knockout cells showed severe decrease in POLA levels (Do, Matsuda et al., 2020).

POLA A449T has reduced affinity for POLB
We hypothesized that the A449T mutation could impair interactions with POLB and thus destabilize POLA A449T . To address this possibility, we investigated POLA A449T interactions with POLB by performing size-exclusion chromatography. At 1:1 molar ratio of POLA and POLB (calculated as a dimer), POL WT and POL migrated as a single peak, corresponding to a stable complex between the two proteins ( Figure 6C), as confirmed by SDS-PAGE ( Figure 6D). In contrast, POLA A449T and POLB showed an additional peak, corresponding to unbound POLB ( Figure 6C and 6E). The resolution of the chromatography cannot separate free POLA from the POL holoenzyme. Thus, the A449T mutation significantly reduces the interaction between POLA and POLB subunits. This observation is in agreement with data for human POLA A467T (Chan et al., 2005).

POLB protects POLA against LONP1 degradation
Next, we investigated if free, partially unfolded POLA could be a target for protein degradation. In mitochondria, the LONP1 protease degrades misfolded proteins and is linked to regulation of mtDNA copy number (Bezawork-Geleta, Brodie et al., 2015). We therefore decided to investigate if POLA was a target for LONP1.
We first used siRNA interference against LONP1, POLA and POLB in HeLa cells. Interestingly, LONP1 knockdown caused a robust increase in POLA levels ( Figure 6F and 6G), whereas POLB was unaffected ( Figure 6F), supporting the idea that POLA is a specific target of LONP1 degradation. In agreement with a stabilizing effect of POLB, knockdown of Polg2 mRNA also caused a reduction of POLA levels ( Figure 6F and 6H). Both POLB and POLA knockdown resulted in an increase of LONP1 ( Figure 6F). To further support our findings, we evaluated the steady state levels of PolA in a mouse model with a tissue-specific Lonp1 knockout (6I and 6J).
Notably, PolA levels were increased in heart samples of Lonp1 -/compared to control littermates.
Collectively, these results support that LONP1 specifically targets POLA both in cells and in vivo.
To investigate if POLA is a direct target for LONP1 degradation, we performed a size-exclusion chromatography with recombinant protein to assess if POLA can form a complex with LONP1. To ensure that PolA was not degraded by LONP1 during the experiment, we used the mutant LONP1 S855A , which traps substrates without degrading them (Kereiche, Kovacik et al., 2016). As shown in Figure 7A, we observed a co-elution of LONP S855A and POLA, revealing an interaction between these two proteins.
We also monitored LONP1 dependent degradation of POLA and POLB in vitro. We followed the reactions over time and used another well-characterized LONP1 substrate, TFAM, as a positive control (Lu, Lee et al., 2013, Matsushima, Goto et al., 2010. The TFAM levels were reduced by 50% in about 3 minutes ( Figure 7B). We observed no LONP1-dependent degradation of POLB, confirming that the accessory subunit is not a substrate of LONP1 ( Figure 7C, lanes7-10). In contrast, both isolated POLA WT and POLA A449T were efficiently degraded, with a 50% reduction in about 20 minutes ( Figure 7C, lanes 2-5, 7D, lanes 2-5 and 7E). The slower degradation time compared to TFAM could in part be explained by the size difference between the two substrates, with POLA being about 6-fold larger. LONP1 is an ATP dependent enzyme, and no degradation of PolA was therefore observed in the absence of ATP ( Figure 7C, lanes 1, 6 and 11).
Next, we examined POLA in complex with POLB. Interestingly, the presence of POLB completely blocked POLA wt degradation ( Figure 7C, lanes 12-15 and Figure 7E). In contrast, POLB was unable to efficiently block degradation of POLA A449T and the levels of the mutant protein decreased significantly over the time of the experiment ( Figure 7D, compare lanes 7-10 with 12-15 and Figure 7E). We also used the human WT and A467T mutant versions of POLA, and observed similar results ( Figure 7F, 7G and 7H). We conclude that impaired interaction between POLB and POLA A449T leads to increased LONP1 dependent degradation of POLA A449T . This observation could explain the lower levels of POLA A449T observed in vivo.
To validate our model, we also analysed two additional POLA mutations. The mouse version of POLA W748S (PolA W725S ), which also displays reduced interactions with PolB (Supplementary Figure 6A, 6B and 6C) and human POLA D274A , which has no effect on POLB interactions (Macao, Uhler et al., 2015). As expected, POLA W725S but not POLA D274A was degraded in presence of POLB (Supplementary Figure 6D, 6E and 6F). Overall, these data provide in vitro evidence that POLB affects POLA folding and protects the protein from degradation. Our data also suggest that other POLA mutations affecting the interactions with POLB or vice versa (e.g. mutations in POLB affecting the interaction with POLA ) may be subjected to LONP1 degradation.

Discussion
Mutations in POLG are a relatively common cause of a spectrum of mitochondrial disease. The substantial lack of relevant in vivo models has hampered our understanding of the pathogenesis of these POLG-related disorders. Here we developed a mouse model for human POLG A467T and study the molecular pathogenesis of this common mutation in vivo. We complemented this analysis with detailed biochemical characterization of the corresponding events in vitro.
The homozygous Polg A449T/A449T mice displayed a mild decrease of mtDNA in skeletal muscle, impaired treadmill performance and reduced spontaneous movements. The mice also displayed reduced 7S DNA levels in several tissues, as has been previously described in mouse knock-out models for other components of the mitochondrial replication machinery, including POLB (Di Re, Sembongi et al., 2009); mtSSB (Ruhanen, Borrie et al., 2010); and TWINKLE (Milenkovic, Matic et al., 2013). The obvious explanation for the mild phenotype observed in the Polg A449T/A449T compared to that of the patients is that compensatory mechanisms may be more efficient in mice than in humans, since the essential biochemical features are qualitatively identical between the two organisms. For instance, one possibility is that in the mouse the stabilization by POLB is more effective than in humans. In addition, in humans the A467T mutation has been associated to profound depletion of mtDNA in brain and other tissues 10 , while in mice the depletion is very mild, possibly because of more successful replication events, as suggested by reduced 7S DNA levels and increased number of replicating nucleoids in MEFs.
Experiments carried out on Polg A449T/A449T MEFs showed an increased number of actively replicating nucleoids as well as impaired mtDNA synthesis upon stress-induced mtDNA depletion with EtBr. The observation was in agreement with a reduced capacity of Polg A449T/A449T mice to recover from the CCl 4 -induced liver necrosis, demonstrating that effective mtDNA replication is necessary for liver regeneration. Similar observations have been made earlier in a knockout mouse model of mitochondrial topoisomerase I (TOP1mt) (Khiati, Baechler et al., 2015). In the Polg A449T/A449T mice, we observed an accumulation of RIs, as demonstrated both by in organello replication assay and 2D-AGE analysis, suggesting a replicative fork stalling phenotype. Notably, we did not observe any deletions in mouse mtDNA. Our analysis of POLA A449T in vitro revealed impaired interactions with POLB and reduced replication efficiency, which were in agreement with the effects observed in vivo and previous in vitro analysis of A467T mutation in human POLA (Chan et al., 2005). Interestingly, a comparison between the mouse POLA A449T and human POLA A467T proteins, revealed similar, but more pronounced replication defects for the human polymerase, which may explain why the A467T mutation causes more severe phenotypes, including multiple mtDNA deletions.
As demonstrated here, a reduction of POLA A449T protein levels also contributes to the pathogenesis of phenotypes observed in the mouse. Using a thermofluor stability assay, we found that POLA (WT and mutant) was structurally unstable at physiological temperatures, but strongly stabilized in complex with POLB ( Figure 6B). In support to this notion, isolated POLA was shown to be efficiently degraded by LONP1, a protease that recognizes unfolded mitochondrial proteins, and Our findings clearly indicate that LONP1 degrades POLA in vitro and in vivo. However, we cannot exclude that other proteases may also contribute to this process in vivo. For instance, ClpXP, another AAA+ protease of the mitochondrial matrix (Goard & Schimmer, 2014) can also degrade free POLA in vitro and the reaction is blocked by addition of POLB. However, in contrast to LONP1, POLA levels are not changed in a mouse knock-out model for the ClpXP-protein (unpublished). The rapid degradation of free POLA could be of physiological relevance, since on its own, the protein displays higher exonuclease activity, which can potentially disturb mtDNA replication ( Figure 5C).
In conclusion, we here describe in detail the in vivo and in vitro features of a common mutation of POLA, shedding light on the pathogenesis of an intriguing and previously poorly understood condition. Furthermore, we suggest a new mechanism contributing to the pathogenesis of POLGrelated diseases. Mutations in POLG or POLG2 that cause weaker interactions within the POL holoenzyme will lead to LONP1-dependent degradation of POLA, resulting in protein depletion in vivo. We speculate that interventions aimed at increasing POLA stability, either by increasing interactions with POLB or by reducing LONP1, may have therapeutic value in affected patients.
We believe that the Polg A449T/A449T mouse model developed here will be a valuable tool for studies of new therapeutic interventions and help clarify the role of mitochondrial proteases in mtDNA maintenance disorders. All the authors contributed to writing the manuscript.

Declaration of Interests
The authors have no competing interests to disclose  animals measured in the CLAMS™ system. Data are presented as mean ± SEM. *p<0.05; Student's t-test. Each symbol represents a biological replicate.

C. Spontaneous ambulatory activity (horizontal movement counts) of 3 month-old WT and
Polg A449T/A449T animals measured in the CLAMS™ system. Data are presented as mean ± SEM.
Student's t-test. Each symbol represents a biological replicate.

Generation of Polg A449T/A449T mice
Polg A449T/A449T mice were generated by a double-nickase CRISPR/Cas9 D10A mediated gene editing of mouse Polg gene in exon 7 (c.1345G>A / p.A449T). For a detailed representation see (Supplementary Figure 1A). The selected sgRNAs (Table 1) were cloned into plasmid pSpCas9(BB)-PX330 (Addgene #42230), using the BbsI site. The resulting constructs were used as a template to amplify by PCR the gRNA (spacer + scaffold) preceded by a T3 promoter to allow subsequent in vitro transcription. The in vitro transcription was carried using the MEGAscript T3 Transcription Kit (Life Technologies). The same kit was used to produce Cas9 D10A mRNA using as template the plasmid pCAG-T3-hCasD10A-pA (Addgene #51638). The 140bp ssDNA homology direct repair (HDR) donor (Table 1) was acquired from IDT. Cas9 D10A mRNA, gRNAs and HDR donor were microinjected into fertilized FVB/NJ one-cell embryos (Core Facility for Conditional Mutagenesis, Milan). Genotyping of Polg A449T/A449T mice was performed by PCR (primers Polg_A449T_Fw + Polg_A449T_Rv, Table 1), followed by a restriction digestion with PvuII. WT allele produces a fragment of 769bp, which is cleaved in the A449T allele producing two fragments of 490bp + 279bp (Supplementary Figure 1B). The PCR is carried using GoTaq DNA polymerase (Promega, UK) and the following PCR conditions: 95 °C for 30 s, 63.7 °C for 30 s, and 72 °C for 1 min, 35 cycles.

Animal work
All animal experiments were carried out in accordance with the UK Animals (Scientific Procedures) Act 1986 (PPL: P6C20975A) and EU Directive 2010/63/EU. The mice were kept on FVB/NJ background, and wild-type littermates were used as controls. The animals were maintained in a temperature-and humidity-controlled animal care facility with a 12-h light/12-h dark cycle and free access to water and food, and they were monitored weekly to examine body condition, weight, and general health. All mice were sacrificed by cervical dislocation at 3 months of age for subsequent analysis.
Lonp1 gene targeting (Lonp1 +/tm1a(EUCOMM)Hmgu/Ieg , project number HEPD0936_3_B11) was carried out as part of the The European Conditional Mouse Mutagenesis Program (EUCOMM), on the C57BL/6NTac genetic background. We generated the heart and skeletal muscle specific Lonp1 knockout mice by mating Lonp1 fl/fl animals with transgenic mice expressing cre recombinase under the control of muscle creatine kinase promoter (Ckmm-cre) (Larsson, Wang et al., 1998)

Treadmill
A standard treadmill apparatus (Panlab) was used to measure motor endurance according to the number of falls in the motivational air puff during a gradually accelerating program with speed initially at 6.5 m/min and increasing by 0.5 m/min every 3 min. The test was terminated by exhaustion, defined as >10 air puffs activations/min.

Comprehensive laboratory animal monitoring system (CLAMS)
Mice were individually placed in the CLAMS™ system of metabolic cages and monitored over a 48-h period. Data were collected every 10-min. The parameters analyzed were: ambulatory and rear movements, VO2 (volume of oxygen consumed, ml/kg/h), VCO2 (volume of carbon dioxide produced, ml/kg/h), RER (respiratory exchange ratio) and heat (kcal/h).

Pharmacological treatments
In VPA-treated mice, VPA (Sigma) was administrated by daily oral gavage (300mg/kg in water) or added to a standard diet at 1.5% (1.5g-VPA/ 1kg-Food) and administered for 60 days, starting at 8 weeks of age.
In CCl 4 experiments, mice received a single IP injection of CCl 4 (1 mL/kg body weight diluted 1/10 in olive oil (Sigma). Mice were sacrificed after 2 or 4 days. For histology analysis, Livers samples were fixed in 10% formalin and embedded in paraffin. Sections of 4-µm were stained with H&E according to standard protocols. The quantification of necrotic areas was done with ImageJ by dividing the necrotic areas around the central veins by total area of the section. Five different regions of the slide were analyzed and average value obtained. Total RNA was extracted from the indicated tissues using the TRIzol Reagent (Thermofisher)

DNA and RNA extraction
following the manufacture protocol.

Real-time quantitative PCR
For mtDNA relative quantification, SYBR Green real-time qPCR was performed using primers specific to a mouse mtDNA region in the COI gene. Primers specific to RNaseP, a single copy gene taken as a nuclear gene reference. All primers are listed in Table 1. Approximately 25 ng of DNA was used per reaction.
For the quantification of mRNA levels, cDNA was retrotranscribed from total RNA extracted using the Omniscript RT kit (Qiagen). For mitochondrial transcripts CoI and Nd4, specific primers (Table   1) were used as described above with SYBR Green chemistry. Expression was calculated using the ∆∆Ct analysis using Gapdh as reference.
Specific Gene Expression TaqMan assays (Invitrogen) were used for Polg and Polg2. Expression was calculated using the ∆∆Ct analysis using B2m as reference.

Long-range PCR
MtDNA was amplified from 50 ng of total DNA with the primers (LongR_mtDNA_Fw and LongR_mtDNA_Rv, Table 1) using PrimeSTAR GXL DNA polymerase (TAKARA, Japan) and following PCR conditions: 98 °C for 10 s, 68 °C for 13 min, 35 cycles.

Cell cultures
Polg A449T/A449T and control mouse embryo fibroblasts (MEFs) were prepared from individual E12.5 embryos and were cultured in complete DMEM with high glucose and 10% fetal bovine serum.
MEFs were seeded in six-well plates at 20% confluence. Cells were incubated with or without 100ng/mL EtBr for 5 days and DNA samples were collected every 24h. At day 5, cells new medium without EtBr was added and cells were allowed to recover for an additional 8 days. Again, DNA samples were collected every 24h. MtDNA quantification was performed as described above. After three days, cells were harvested, washed with PBS and used for Western Blotting as described above.
Briefly, cells seeded in 24-well plate were fixed in 5% paraformaldehyde (PFA) in PBS at 37°C for 15 minutes (min) and incubated with 50 mM ammonium chloride in PBS for 10 min at room temperature (RT). After three washes in PBS, cells were permeabilized using 0.1% Triton X-100 in PBS for 10 min, washed 3 times with PBS, and then blocked in 10% FBS in PBS for 20 min at RT.
Cells were then incubated with indicated primary antibodies for 2 hours in 5% FBS/PBS, washed in 5 % FBS in PBS and incubated with secondary Alexa Fluor conjugated antibodies in 5% FBS/PBS for 1 hour at RT. We used the following antibodies: TOM20 (1:1000) was from Abcam (ab232589) processed once with the "smooth" function in Fiji and nucleus was removed. Images were then manually thresholded, 'smoothed' and number of particles were obtained using the "Analyze particles" plugin in Fiji with a minimum area of 0.1 μm 2 . The representative images in figure 3 were processed once with the "smooth" function in Fiji.

Biochemical analysis of MRC complexes
Liver and muscle samples stored in liquid nitrogen were homogenized in 10mM of potassium phosphate buffer (pH=7.4), and the spectrophotometric activity of respiratory chain complexes I, II, III and IV, as well as citrate synthase, was measured as described (Bugiani, Invernizzi et al., 2004).

BNGE and in-gel activity
For blue native gel electrophoresis (BNGE) analysis, skeletal muscle and liver mitochondria were isolated as previously described (Fernandez-Vizarra, Lopez-Perez et al., 2002). Samples were resuspended in 1.5 M aminocaproic acid, 50 mM Bis-Tris/HCl (pH 7) and 4 mg of dodecyl maltoside/mg of protein, and incubated for 5 min on ice before centrifuging at 20,000 × g at 4°C.
5% Coomassie G250 was added to the supernatant. 100 μg was separated by 4%-12% gradient BNGE and further subjected to a Complex I in-gel activity (IGA), as previously described (Calvaruso, Smeitink et al., 2008). To allow for cI activity to appear, gels were incubated between 1.5 and 24 h in cI-IGA reaction buffer.

COX/SDH histochemical analysis
For histochemical analysis, skeletal muscle samples were frozen in isopentane pre-cooled in liquid nitrogen. Eight-micrometer-thick sections were stained for COX, SDH and combined COX/SDH activity as described (Sciacco & Bonilla, 1996).

Southern Blot
Three micrograms of total DNA isolated from each tissue were restricted using the restriction enzyme BlpI according to manufacturer's instructions (New England Biolabs). Products were separated on 0.8% agarose gels (Invitrogen Ultrapure) and dry-blotted overnight onto nylon membrane (GE Magnaprobe). Membranes were hybridized with radiolabeled probes overnight at 65°C in 0.25 M phosphate buffer (pH 7.6) and 7% SDS, then washed for 3 × 20 min in 1× SSC and 0.1% SDS and imaged using a phosphorimager (GE Healthcare) and scanned using an Amersham Typhoon 5 scanner. For primer sequences used for producing probes, see Table 1.

In Organello Replication
Labeling of mtDNA in isolated organelles was performed as previously described (Reyes, Kazak et al., 2013).
Briefly, isolated liver was minced and homogenized in 4 ml/g of tissue in Sucrose-Tris-EDTA blotting and hybridization were carried as described above.

2D-AGE
For two-dimensional gels, DNA was extracted from fresh liver-isolated mitochondria purified by sucrose gradient followed by phenol-chloroform extraction. POLγA and POLγB were added as indicated in the figures and reactions were incubated at RT for 10 min before separation on a 6% Native PAGE gel in 0.5 X TBE for 35min at 180V. Bands were visualized by autoradiography.
For K d analysis, band intensities representing unbound and bound DNA were quantified using Multi Gauge V3.0 software (Fujifilm Life Sciences). The fraction of bound DNA was determined from the background-subtracted signal intensities using the expression: bound/(bound+unbound). The fraction of DNA bound in each reaction was plotted versus the concentration of POLγA or POLγA-B2. Data were fit using the "one site -specific binding" algorithm in Prism 8 (Graphpad Software) to obtain values for K d .

Coupled exonuclease-polymerase assay
DNA polymerization and 3′-5′ exonuclease activity were assayed using the same primer-template as described above for EMSA. The reaction mixture contained 10 fmol of the DNA template, 25 mM Tris-HCl [pH 7.8], 10% glycerol, 1 mM DTT, 10 mM MgCl 2 , 100 µg/ml BSA, 60 fmol of POLγA, 120 fmol of POLγB and the indicated concentrations of the four dNTPs. The reaction was incubated at 37°C for 15 min and stopped by the addition of 10 µl of TBE-UREA-sample buffer (BioRad).
The samples were analysed on a 15% denaturing polyacrylamide gel in 1 X TBE buffer.

Rolling circle in vitro replication assay
A 32 P 5′-labeled 70-mer oligonucleotide [5′-42(T)-ATCTCAGCGATCTGTCTATTTCGTTCAT-3′] was hybridized to a single-stranded pBluescript SK(+) followed by one cycle of polymerization using KOD polymerase (Novagen) to produce a ∼3-kb double-stranded template with a preformed replication fork. Reactions of 20µl were carried out containing 10 fmol template DNA, 25 mM Incorporation of [α-32 P]-dCTP was measured by spotting 5 μl aliquots of the reaction mixture (after the indicated time points at 37°C) on Hybond N+ membrane strips (GE Healthcare Lifesciences).
The membranes were washed (3 × with 2 × SSC and 1 × with 95% EtOH) and the remaining activity was quantified using Multi Gauge V3.0 software (Fujifilm Life Sciences). A dilution series of known specific activity of [α-32 P]-dCTP was used as a standard.

Thermofluor assay
The fluorescent dye Sypro Orange (Invitrogen) was used to monitor the temperature-induced unfolding of wild-type and mutant POLγA. Wild type and mutant proteins were set up in 96-well PCR plates at a final concentration of 1.6 μM protein and 5× dye in assay buffer (50 mM Tris-HCl pH 7.8, 10 mM DTT, 50 mM MgCl2 and 5 mM ATP). Differential scanning fluorimetry was performed in a C1000 Thermal Cycler using the CFX96 real time software (BioRad). Scans were recorded using the HEX emission filter (560-580 nm) between 4 and 95°C in 0.5°C increments with a 5 s equilibration time. The melting temperature (Tm) was determined from the first derivative of a plot of fluorescence intensity versus temperature (Matulis, Kranz et al., 2005). The standard error was calculated from 3 independent measurements.

LONP1 proteolysis assay
Protease activity of purified LONP1 on POLA was measured in a 15 l reaction volumes containing 0.5 g of LONP1 wild-type and 0.55 g of POLA (in presence or absence of 0.22 ug of PolB). When having both POLA and POLB in the same reaction, a preincubation in ice for 10 min is made before adding LONP1 to the reaction. Samples were incubated at 37°C for 0-90 min in