All materials for the maintenance of fibroblasts were purchased from Merck (formerly Sigma-Aldrich) unless otherwise stated.

Patient skin biopsies were performed following informed consent, with the approval of the United Kingdom National Research Ethics Service (South Central-Berkshire). Human fibroblast cultures were established from skin explants following standard procedure. Cells were maintained in high-glucose Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% FCS, 100U/mL penicillin and 100 μg/mL streptomycin, 2 mM L-glutamine and 50 μg/mL uridine at 37°C in a humidified incubator with 5% CO₂. Media was changed every 3–4 days. Cells were routinely cultured in T25/T75 tissue culture flasks and split when reached ∼80–90% confluency. Prior to the experiments, cells were tested for possible Mycoplasma contamination using the EZ-PCRTM Mycoplasma Test Kit (Geneflow) following the manufacturer’s instructions.

Fibroblasts were seeded into six well plates at a density of 20,000 cells per well. Cells were left to attach overnight. The next day (day 0) the media was changed to either high-glucose maintenance media or glucose free DMEM (Thermo Fisher Scientific, Gibco) supplemented with 10% FCS, 100U/mL penicillin and 100 μg/mL streptomycin, 1 mM sodium pyruvate, 50 μg/mL uridine and 10 mM galactose. Media was changed on day 3. On day 7 the cells were trypsinized and counted manually using a haemocytometer. The number of cells remaining in galactose media was expressed as the percentage of the cells counted in high-glucose media (Mortiboys, 2006).

For experimental purposes regular foetal calf serum was replaced with dialysed foetal calf serum throughout the experimental procedure (dFCS, Thermo Fisher Scientific). Cells were treated in high-glucose DMEM media (Merck) supplemented with 100U/mL penicillin and 100 μg/mL streptomycin (Merck), 2 mM L-glutamine (Merck) and either 10% or 0.1% dialysed foetal calf serum (Thermo Fisher Scientific), no uridine. Control and patient derived fibroblasts were seeded into 96-well optical clear bottom tissue culture treated plates (Perkin Elmer) at a density of 5 × 10² cells/well for culture with 10% dFCS and 3 × 10³ cells/well for reduced serum 0.1% dFCS culture. Fibroblasts were supplemented with all possible combinations of deoxynucleosides (A, T, G, C, AT, AC, AG, CT, CG, GT, ATC, ATG, AGC, TGC, ATGC, Sigma-Aldrich), each at either 50 or 200 µM concentration for 7 days. Deoxynucleosides were replenished on day 3. Fibroblasts cultured in 0.1% dFCS were pre-conditioned in low serum medium for 3 days (d-3) to stop nuclear replication and therefore mimic affected post-mitotic tissue. On day 7 the plates were stained and imaged as described above.

As previously, for experimental purposes regular foetal calf serum was replaced with dialysed serum throughout the experimental procedure (dFCS, Thermo Fisher Scientific). Cells were treated in high-glucose DMEM media (Merck) supplemented 100U/mL penicillin and 100 μg/mL streptomycin (Merck), 2 mM L-glutamine (Merck) and 0.1% dialysed foetal calf serum (Thermo Fisher Scientific), no uridine.

Control and POLG, TWNK and DGUOK deficient fibroblasts were seeded into 8×T25 (per cell culture) tissue culture treated flasks in high-glucose DMEM media containing 10% dFCS at a density of 2 × 10⁶ cells/flask (d −14). The next day, the media was replaced by high-glucose DMEM containing 0.1% dFCS (d −13). High-glucose DMEM with 0.1% dialysed serum was used for the remainder of the experimental procedure. 72 h later (d −10) 7xT25 flasks/cell culture were treated with 20 µM 2′-3′-dideoxycytidine (ddC, Sigma-Aldrich) for 10 days. Media with ddC was changed on d −7 and d −4. 1×T25 flask/cell culture was kept as untreated control. After 10 days the cells from the untreated flasks and 1× ddC treated flask were harvested as reference pellets (d0). The remaining six flasks/cell culture were split into 3× untreated (DMSO control) and 3× treated. The treated flasks were supplemented with the combination of all four deoxynucleosides (ATGC) each at a concentration of 50 µM. Media was changed every 3–4 days. A pair of treated and untreated flasks were harvested after 7 days (d7), 14 days (d14) and 21 days (d21) of treatment.

The experimental procedure was repeated with each cell culture on 4 separate occasions. On two occasions the cells were supplemented with regular deoxyribonucleosides. These samples were used for subsequent quantification of mtDNA content. On two occasions the deoxynucleoside supplement mix consisted of 90% regular and 10% heavy labelled deoxynucleosides 13C A, 13C C, 15N13C T and 15N G were purchased from Silantes. These samples were used to quantify the incorporated deoxyribonucleosides by mass spectroscopy.

Total cellular DNA was isolated using the GenElute Mammalian Genomic DNA Miniprep Kit (Sigma-Aldrich) following manufacturer’s instructions.

Hydrolysis was carried out in 50 µL reactions that contained 50 ng DNA, 100 mM NaCl, 20 mM MgCl₂, 20 mM Tris pH 7.9, 1000 U/mL Benzonase, 600 mU/mL Phosphodiesterase I, 80 U/mL Alkaline phosphatase, 36 μg/mL EHNA hydrochloride and 2.7 mM deferoxamine for 2 h at 37°C. Following lyophilisation, nucleosides were resuspended in 50 µL buffer A (10 mM ammonium acetate, pH 6).

For the analysis by HPLC–QQQ mass spectrometry, a 1,290 Infinity UHPLC was fitted with a Zorbax Eclipse plus C18 column, (1.8 µm, 2.1 mm 150 mm; Agilent) and coupled to a 6495a Triple Quadrupole mass spectrometer (Agilent Technologies) equipped with a Jetstream ESI-AJS source. The data were acquired in dMRM mode using positive electrospray ionisation (ESI1). 5 μL of the samples were injected per run. The gradient used to elute the nucleosides started by a 5 min isocratic gradient composed of 100% buffer A and 0% buffer B (100% methanol) with a flow rate of 0.4 mL/min and was followed by the subsequent steps: 5–8 min, 94.4% A; 8–9 min, 94.4% A; 9–16 min 86.3% A; 16–17 min 0% A; 17–21 min 0% A; 21–24.3 min 100% A; 24.3–25 min 100% A. The AJS ESI settings were as follows: drying gas temperature 230°C, the drying gas flow 14 lmin-1, nebulizer 20 psi, sheath gas temperature 400°C, sheath gas flow 11 lmin-1, Vcap 2,000 V and nozzle voltage 0 V. The iFunnel parameters were as follows: high pressure RF 110 V, low pressure RF 80 V. The fragmentor of the QQQ mass spectrometer was set to 380 V and the delta EMV set to +200.

The raw mass spectrometry data was analysed using the MassHunter Quant Software package (Agilent Technologies, version B.08.01). The transitions and retention times used for the characterization of nucleosides and their adducts are summarised in Supplementary Table S1. For the identification of compounds, raw mass spectrometry data was processed using the dMRM extraction function in the MassHunter software.

In addition to using high throughput imaging to estimate mitochondrial DNA copy number, we confirmed key findings by using real-time quantitative PCR, using primers and probes specific for mitochondrial DNA and the single copy nuclear gene, B2M, assayed simultaneously using a PE7500 real-time PCR instrument (Applied Biosystems, Thermo Fisher Scientific, Waltham, MA, United States) (primer and probe sequences available online at https://github.com/StochasticBiology/mtdna-nucleosides).

Normality for the distribution of each sample in each condition was assessed using a Shapiro-Wilk test. Differences from baseline were assessed with Mann-Whitney U test where the distribution differed from normality and t-tests were used where there was no evidence against normality, IBM SPSS statistics version 28 was used to carry out the analysis. Error bars are standard errors unless otherwise stated to account for differences between runs and cell lines when amalgamating data, linear mixed models (LMMs) were used for some analyses, assigning random effects to run (and cell line where appropriate). These additional statistical analysis were performed in R (R C Team, 2022) using libraries ggplot2 (Wickham, 2016), ggrepel (S K, 2021), readxl (Wickham H, 2023) nlme (Pinheiro et al., 2022) and lme4 (Bates et al., 2015). Code and data for this analysis is freely available online at https://github.com/StochasticBiology/mtdna-nucleosides.

Patient and control fibroblast cultures used in this study are summarised, with a short clinical description in Table 2. We selected three controls from our bank of 14 anonymised control fibroblast cultures comprising five children with normal cytogenetics undergoing diagnostic skin biopsy for karyotyping and nine healthy consented adults aged 18–81 years. Control 1 was selected being from the paediatric age range and Controls 2 and 3 are adults who are near to the median for most of the readouts we measure. We studied two patients who were both compound heterozygotes for mutations in POLG, encoding the catalytic subunit of the mitochondrial DNA polymerase gamma. Patient POLG1 presented at age 2, mitochondrial DNA depletion in liver (26% mtDNA content) and reduced mtDNA levels in muscle (64% mtDNA content). Patient POLG2 presented as an infant and profound mtDNA depletion was reported in both liver (9% mtDNA content) and muscle (9% mtDNA content). We investigated the mitochondrial DNA content of patient fibroblasts using high-content imaging of live cells labelled with the fluorescent nucleic acid stain PicoGreen (Ashley et al., 2005) (Figure 1A), a method we have validated against qPCR (Uusimaa et al., 2014). We compared the mtDNA content of patient cells with fibroblasts derived from a paediatric (Control 1) and an adult (Control 2) control. The mtDNA content of POLG1 cells was reduced by ∼20% when compared to the control cells, while POLG2 cells represent an extreme case and are almost completely devoid of mtDNA and of mitochondrially encoded proteins (Figure 1B; Supplementary Figures S1A–C). Two additional fibroblast lines included were derived from a patient with a homozygous mutation in the mitochondrial helicase Twinkle (TWNK) and one with defects in deoxyguanosine kinase (DGUOK), an essential component of the mitochondrial nucleotide salvage pathway. We included patient TWNK in this study because to our knowledge the response of TWNK deficient cells have never been investigated in deoxynucleoside supplementation studies. Additionally, we choose patient DGUOK because these cells are expected to respond to supplementation (see Table 1 for references). When measured with PicoGreen, TWNK and DGUOK fibroblasts had a lower mtDNA content compared to Control 1 cells, which was a statistically significant (all p < 0.001, Mann-Whitney U test) but small magnitude effect (Figure 1C). We measured mosaic mtDNA depletion by quantifying the proportion of Rho 0 cells, those void of mtDNA, in each culture. Mosaic mtDNA depletion was mild in POLG1 cells and profound in POLG2 cells (Figure 1D); the proportion of profoundly mtDNA depleted cells was sometimes as high as 80% in culture POLG2 (Figure 1D). Not surprisingly, when investigated by Western blot analysis respiratory chain components were significantly reduced in POLG2 cells (Supplementary Figure S1A). TWNK and DGUOK cells had no detectable increase in Rho 0s when compared to the controls (Figure 1E).

Mitochondrial membrane potential of control and patient cells was measured alongside mtDNA content by high-content imaging using the vital dye TMRM that accumulates in the matrix compartment of the organelle. The histogram of TMRM signal measured on a cell-by-cell basis suggested an increase in depolarised mitochondria in both POLG1 and POLG2 cells when compared to Control 2 (Figure 2A), entirely consistent with the appearance by eye (Figure 1A; Supplementary Figure S1B). The TMRM signal in TWNK and DGUOK fibroblasts were similar to Control 2 cells (Figure 2B), with the DGUOK cells being slightly more depolarised. Comparisons of cell growth under energetic stress is a useful surrogate for mitochondrial function in severely compromised cells (Mortiboys, 2006). To investigate the respiratory function of the depolarised POLG1 and POLG2 cultures, we cultured these cells under energetic stress in glucose-free DMEM supplemented with 10 mM galactose. The cells were counted after 7 days and counts in the galactose culture were expressed as the proportion of cells cultured in high-glucose medium (Figure 2C). Because of their profound mtDNA depletion POLG2 cells are highly dependent on glucose, and most cells died in galactose, while the number of cells remaining in culture was reduced to ∼50% in POLG1 (Figure 2C).

We supplemented our patient cells with exogenous deoxynucleosides and subsequently measured mtDNA content and membrane potential using high-content imaging as previously (Diot et al., 2015) to establish any potential benefit to the cells. For the remainder of this study, we used the vital dye MitoTracker Red CMXRos, rather than TMRM to measure mitochondrial membrane potential. We previously validated PG/TMRM (Uusimaa et al., 2014), but not PG/MitoTracker Red CMXRos (henceforth “CMXRos”) staining as a readout for mtDNA content. We therefore compared CMXRos with TMRM. Like TMRM, CMXRos is a red-fluorescent dye that stains mitochondria in live cells and its accumulation is dependent upon membrane potential. It has the advantage of being suitable for fixation and manifests less run-to-run variability than TMRM. Measurements of mtDNA copy number (summed area of PicoGreen mtDNA signal) under different treatments within a culture counterstained with TMRM correlated well with PicoGreen signal counterstained with CMXRos (Supplementary Figure S2). Previous studies (Table 1) used deoxynucleoside monophosphate supplements that resulted in mtDNA copy number increase in DGUOK as well as POLG mutant cells at concentrations of 200 µM or higher without reporting any toxicity up to 800 µM. The potential benefit of nucleoside supplementation has not been investigated in TWNK mutant cells to our knowledge. Based on published data listed in Table 1, we supplemented both cycling (10% dFCS) and quiescent (0.1% dFCS) fibroblasts with individual (A, T, G, C) or combinations of deoxynucleosides (AT, GC, AG, TC, AC, GT, ATG, ATC, AGC, TGC, ATGC) at 200 µM concentration for 7 days. In line with previous investigators (Table 1), we used dialysed serum to maintain low exogenous nucleotide levels in the culture media. Previous authors suggested that supplementing Poly γ deficient cells with a combination of all four nucleotides is beneficial (Bulst et al., 2009). Given that our knowledge of nucleoside and nucleotide cycles and balances within cells is limited, especially in MDS (Ashley et al., 2007; Nikkanen et al., 2016) we wanted to explore the effects of more combinations of deoxynucleoside supplements. While the mtDNA content of POLG1, TWNK and even Control 1 fibroblasts doubled following supplementation with certain combinations (Figures 3A, B) we noted that simultaneously the cell numbers were lower for both Control 1 and POLG1 in 10% dFCS (see scattergrams in Supplementary Figure S3 showing that for 10% FCS cell numbers vs. mtDNA content the ungrouped R ² ranges from 0.43 to 0.84). To visually represent the potential negative correlation between increased mtDNA content and low cell numbers we plotted the percentage change in mtDNA content and the percentage change in nuclear (cell) count on the same graph for each condition (Figures 3A, B). The treatments that simultaneously increased mtDNA content and reduced cell numbers included the combination of all four nucleosides favoured by previous investigators (Bulst et al., 2009). As well as significantly increasing mtDNA content across patients and controls (log transformed summed area of mtDNA nucleoids compared with baseline, all p < 0.001 with LMM, grouping by cell line), ATGC caused a ∼40% reduction in cell numbers in both control and patient cells and a concomitant reduction in mitochondrial membrane potential (MMP) (Figures 3C, D). This indicates that increased mtDNA content was associated with impaired mitochondrial function. In both POLG1 and TWNK fibroblasts the worst treatment combinations were ATG, TGC, GC, GT and G alone resulting in a reduced cell count by 50%–80% (Figures 3A, B). The treatment regime was similarly detrimental to healthy control cells. Supplementation of deoxy pyrimidines alone or in combination (C, CT or T) did not reduce mitochondrial membrane potential or cell numbers and were hence harmless to the cells. Of these, CT significantly increased mtDNA content in all patients and controls (log transformed summed area of mtDNA nucleoids compared with baseline, POLG2 p < 0.01, the rest all p < 0.001 with LMM), as did T or C alone in DGUOK cells (p < 0.01 with LMM). Supplementation of quiescent cells cultured in 0.1% dialysed serum with high doses (200 µM) of deoxynucleoside supplements also resulted in significant reduction in cell numbers (Supplementary Figure S4). However, the magnitude of the increase in mtDNA and of the reduction in cell counts was less than in cycling cells.

POLG2 cells in which Rho 0 cell counts reached 80% (Figure 1D) did not improve as much in response to any nucleoside combinations at 200 µM (not shown). As with POLG1 and control fibroblasts, supplementation with T, C and CT improved both cell counts and MMP in 10% FCS on POLG2 cells (not shown), but mtDNA content was only significantly increased in CT (p < 0.01 as above). CT also improved cell numbers but not MMP or mtDNA content in 0.1%FCS (not shown). The response of these cells to pyrimidine supplementation is consistent with their profoundly low respiratory chain activity and hence likely inability to synthesize them (Supplementary Figure S1; Figures 2A, C).

Given these results, we chose to proceed with a treatment regime using a lower deoxynucleoside concentration of 50 µM based on recently published promising results (Blazquez-Bermejo et al., 2019).

Supplementation of cycling patient cells (10% dFCS) with 50 µM deoxynucleosides eliminated the toxic effects seen with 200 µM treatment (Figures 4A, B). To recapitulate the nucleotide levels of terminally differentiated cells (Bradshaw and Samuels, 2005) representing the most affected tissues in MDS, we followed the lead of previous authors by reducing the level of FCS in our medium to put the cells into a stationary growth phase. In replicating cells, most nucleotide components are synthesized by the de novo pathway, rather than the salvage pathway, used by stationary cells. Because the phase of the cell cycle substantially alters mtDNA content (Mandel et al., 2001), previous fibroblast models have used serum starvation (Magnusson et al., 2003). In the absence of serum, less than 1% of cells in a primary fibroblast culture pass through S phase (Taanman et al., 2003). Therefore, we proceeded to investigate the effect of nucleoside supplementation at 50 µM concentration in MDS fibroblasts cultured in 0.1% serum. MtDNA content in patient POLG1 cells was increased most consistently by the combination ATGC without a decrease in cell counts (Figure 5A). This treatment did not significantly improve cellular growth in POLG2 cells (Figure 5B). While all combinations of nucleoside supplements increased mtDNA content in POLG2 cells, due to their very low mtDNA content these percentage changes are misleading in this case. Therefore, we quantified the changes in the percentage of Rho cells in POLG2 cells following treatment (Supplementary Figure S5). Supplementation at 50 µM reduced the proportion of profoundly depleted Rho zeroes in POLG2 cells in both 10% and 0.1% FCS (Supplementary Figure S5). The mild reduction in Rho 0s with deoxypyrimidine supplementation in 0.1% FCS is again consistent with their profoundly low respiratory chain complexes and hence likely inability to synthesize them (Supplementary Figure S1). TWNK cells had increased mtDNA content in response to ATGC, however, cell numbers were reduced by ∼20% (Figure 5C). DGUOK fibroblasts responded well to most combinations of deoxynucleoside supplementation by increased mtDNA content, the most promising combination being ATGC (Figure 5D) in line with previous publications (Table 1). The number of cells analysed, and the p-value for the increase/decrease in mtDNA or mean CMXRos intensity over baseline in the experiments illustrated in Figures 3–5 is shown in Supplementary Figure S3.

As well as finding that cell numbers were reduced in many of the treatments that included 200 µM deoxyguanosine (G any), we found that a substantial minority of G supplemented cells had increased area of their mitochondrial reticulum as well as increased mtDNA content, likely representing a proportion of cells in S phase (Figures 6A–D Supplementary Figure S6A). This was not, or was less apparent with 50 µM supplements (data not shown). We previously showed that the integrated density (the summed product of average signal intensity and nucleoid area) for PicoGreen signal is quantitative for mtDNA in wild type cycling cells (Ashley et al., 2005), and others that nuclear signal (Hoechst) correlates with phases of the cell cycle in high throughput imaging studies (Frolich et al., 2020). Hence, we validated our use of the integrated density of the PicoGreen nuclear signal as a semiquantitative measure of nuclear DNA using EdU (Roy et al., 2018) to pulse label S phase nuclei (Supplementary Figures S6B–C). The EdU labelling thus reflects our estimates of cell ploidy determined by PicoGreen labelling (Supplementary Figure S6D). Cells with nuclei with a high PicoGreen signal were apparent in baseline in 0.1% FCS in both patient and control (Figure 6A). This and the increased nuclear DNA content (Figure 6C) suggest that the proportion of these in S–G2 phase, rose significantly in POLG cultures supplemented with G containing combinations.

Next, we investigated whether supplementation with ATGC could rescue mtDNA recovery following depletion, to further support our high-content imaging results. Cells were cultured in 0.1% dFCS high-glucose media for 3 days to induce quiescence, as previously (Taanman et al., 2003). This was followed by 10 days treatment with 20 µM ddC to deplete the cells of mtDNA. Following ddC treatment Control, POLG1, TWNK and DGUOK fibroblasts constantly cultured in 0.1% dFCS media were left to recover for 21 days either untreated or supplemented with 50 µM ATGC. Treated and untreated cell pellets were collected at the end of ddC treatment (d10) and at 14 (d24) and 21 (d31) days into recovery. Total cellular DNA was extracted from the samples and subjected to qPCR analysis to determine mtDNA copy number (normalised to baseline) at each time point (Figure 7). Following 10 days ddC treatment Control 3, POLG1, TWNK and DGUOK cells had and mtDNA content of ∼50% compared to untreated samples (Figure 7). The plots shown in Figure 7 are strongly suggestive of a treatment effect with ATGC. However, as we only have two replicates for each time course, individual t-tests between treatment and control only yielded p < 0.05 in two instances (control cells and TWNK, day 14). Statistical power here suffers from splitting the data by timepoint. In a preliminary attempt to take the data together, we used two-way ANOVA to block time effects and query the influence of treatment on mtDNA. Here, treatment effects were identified as significant (p < 0.01) in all cases except POLG (p = 0.40). It must be noted that the individual datapoints here are not fully independent and a more detailed statistical treatment would capture the time dependence more fully.

While mtDNA content reached pre-depletion levels by day 14 in the treated group, untreated cells were approaching similar mtDNA levels only at day 21. ATGC supplementation rescued mtDNA recovery in TWNK and DGUOK cells. MtDNA levels returned to 100% in TWNK cells and even exceeded 100% in DGUOK cells in the supplemented groups as opposed to the untreated cells which steadily remained at ∼50% without any sign of recovery.

The depletion recovery experiment was repeated with each cell culture on 4 separate occasions. On two occasions the cells were supplemented with regular deoxyribonucleosides. These samples were used for subsequent quantification of mtDNA content. On two occasions the deoxynucleoside supplement mix consisted of 90% regular and 10% heavy labelled deoxynucleosides 13C A, 13C C, 15N13C T and 15N G. These samples were used to quantify the labelled exogenous nucleotides incorporated into DNA using Mass Spectroscopy (Figure 8A). We used the rate of incorporation of exogenous labelled nucleosides into DNA to probe the endogenous nucleoside pools under conditions favouring mtDNA synthesis. Figure 8A shows the rate of labelling of total DNA extracted from patient POLG2 fibroblasts (see Supplementary Figure S7 for the remaining three patients). We previously used short term pulse labelling of MDS fibroblasts with radiolabelled deoxynucleotides to show that the specific activity of newly synthesized mtDNA reflects the intra-mitochondrial nucleotide pool (Ashley et al., 2007). The ratio of labelled to unlabelled DNA in the whole cell DNA reflects the relative availability of the limiting nucleosides for DNA replication. These were most striking in the POLG2 cells derived from the POLG patient with the more severe mtDNA depletion, reflecting profound defect. The specific activity suggests that the dTTP pool was the lowest followed closely by dGTP, then dCTP and dATP was the highest in all the cultures (Supplementary Figure S7). Compared to the dNTP pools in the control (Figure 8B), the dATP pool was particularly low in POLG2, the severely affected patient, and the dCTP pool was also lower than the control. The milder patient POLG1 did not appear to have any deficiencies in the dNTP pool which appeared to be comparable to the control. Consistent with the expected impaired synthesis the dGTP specific activity of patient DGUOK was significantly higher than control (p = 0.014) but the dTTP incorporation suggested that the dTTP pool was also compromised (p = 0.005). As previously (Ashley et al., 2007), the reciprocal of the specific activity reflects the size of the pool of nucleosides available for DNA labelling. We therefore plotted the reciprocal of (i) the average specific activity normalised to control and (ii) the normalised specific activity that is highest in each patient culture (Figure 8C which should reflect the limiting nucleotide) and hence reflect the DNA synthesis rate (Figure 8B). This nicely validates the qPCR data showing rates of recovery from mtDNA depletion (Figures 7, 8B) suggesting that the labelled DNA reflects mtDNA synthesis and that the pools of available nucleosides is related to intramitochondrial availability. Because previous investigators found an excess of misincorporations at dTTP positions in the reference sequence of POLG iPS cells compared to controls (Hamalainen et al., 2019), we sequenced fibroblast mtDNA (average depth 5,000 reads). Sequence analysis of the POLG1 and POLG2 non-supplemented fibroblasts recovering from depletion by ddC did not show an excess over other misincorporations or over controls (not shown).

We supplemented cycling and quiescent fibroblasts from patients with 15 combinations of deoxynucleosides (A, T, G, C, AT, AC, AG, CT, CG, GT, ATC, ATG, AGC, TGC, ATGC) at high (200 µM) and low (50 µM) concentrations for 7 days. By using high-content imaging to measure changes in mtDNA content upon treatment we were able to quantify additional parameters of cellular wellbeing such as cell numbers, mitochondrial membrane potential and morphology, all linked to phenotypic severity ((Ashley et al., 2008) and Figures 1, 3). This technique enabled us to document a significant reduction in cell numbers (up to <90%) coupled with reduced membrane potential following supplementation at high deoxynucleoside concentrations in cycling cells, especially in the combinations that were most effective at restoring mtDNA content at 200 µM concentration (Figure 3). The most detrimental combinations of nucleosides were those that included deoxyguanosine, the only exceptions being the combinations AG and AGC. Furthermore, a substantial reduction in cell numbers was also evident in quiescent non-replicating fibroblasts treated at 200 μM, although to a lesser extent (up to 50%), indicating a combination of cytotoxicity and a prolongation in cell doubling time. This is entirely consistent with propensity of abnormally high dNTP levels to impair the fidelity of nuclear replication (Pai and Kearsey, 2017) and likely to cause cell-cycle arrest. The toxicity of supplemented deoxyguanosine was previously documented in Jurkat cells causing apoptosis via increasing dGTP as well as necrosis by depleting ATP (Batiuk et al., 2001). Furthermore, high-content imaging enabled us to investigate the effects of supplements at the level of single cells. We were thus able to group cells by nuclear DNA content and infer that the significant (p < 0.001) increase in nuclear DNA content in the cells treated with high-concentrations of deoxyguanosine containing supplements (G any) compared with untreated cultures was consistent with slowing of S/G2 phase, but not in cells treated with the other combinations (Figure 6; Supplementary Figure S5).

Due to the toxic effects of nucleoside supplements at high concentration (200 µM) we proceeded to supplementing cycling and quiescent fibroblasts with the 15 different combinations of nucleosides, reducing the concentration to 50 µM. We show that at 50 µM neither cell counts nor membrane potential are affected negatively, except for TGC, GT, GC combinations (Figure 4), that were the most detrimental at high concentrations as well (Figure 3). Hence the reduced concentration alleviated the toxic effects. We also found that T and T in combination with C were beneficial to fibroblasts from patient POLG1, when grown in 10% FCS, and T alone to profoundly depleted cells POLG2 (Figures 2–4). Neither benefited TWNK or control cells. The combination of deoxypyrimidines (CT) is used in commercially available MT1621 (Zogenix) (Amtmann et al., 2023) which was granted FDA Breakthrough Therapy and PRIME designations in 2018, for use in thymidine kinase deficiency. Our results suggest that this combination could be explored further for the treatment of POLG patients. In quiescent fibroblasts, a combination of all four nucleosides (ATGC) resulted in the highest increase in mtDNA content in POLG1, TWNK and DGUOK cells (all p < 0.001, Figure 5). Additionally, ATGC supplementation following mtDNA depletion with ddC significantly improved the recovery of mtDNA in these patient cells (Figure 7).

Exogenous isotopically labelled nucleosides were incorporated in cellular DNA, providing strong evidence that nucleosides can be taken up by the cell. Because the cells with the most profound mtDNA depletion, that have the slowest doubling time, acquire the highest proportion of label (Figure 8C; Supplementary Figure S7), we also have supporting evidence that exogenous nucleosides can be incorporated into mtDNA, and hence improve mtDNA copy number (Figure 7). It is well known that nucleotide imbalance, either excess or deficiency of nucleotides, can stall nuclear DNA replication (Pai and Kearsey, 2017) and derange nuclear DNA repair, check points and chromosome segregation. Increased de novo synthesis provides the nucleotides for the rapid pool expansion required for DNA replication, whilst allosteric regulation of ribonucleotide reductase (RNR) ensures that the synthesised dNTP pools are balanced. Cytoplasmic deoxyguanosine pools cannot be phosphorylated and will be converted to guanine by purine nucleoside phosphorylase (PNP, Supplementary Figure S7). Guanine is a substrate for the salvage pathway enzyme hypoxanthine-guanine phosphoribosyl transferase (HPRT) and is converted to GMP by transfer of a phosphoribosyl moiety from phosphoribosyl pyrophosphate (PRPP). The increase in GMP pools will inhibit the first committed step of purine de novo synthesis catalysed by glutamine phosphoribosylpyrophosphate amidotransferase (amidoPRT) (Kelley et al., 1975). PRPP is a driver of de novo synthesis by relieving inhibition on amidoPRT and reduced PRPP pools will contribute to reduced flux through the synthetic pathway. An unexpected result of incubation with G may therefore be that cytoplasmic dGTP pools are restricted, leading to inhibition or pausing of DNA replication and cell division in cycling cells. A mechanism of cytoplasmic dGTP pool restriction following G supplementation is consistent with the finding that adding all four nucleosides to the growth medium rather, than rescuing cell division, also led to a significant decrease in cell numbers, albeit to a lesser extent than supplementation with G alone. By contrast, supplementation with the purine deoxynucleoside A alone would be expected to have less of an effect on de novo purine synthesis as unlike G, A can be phosphorylated to the nucleotide, and this is indeed the case. Deoxyadenosine (which escapes this pathway) is converted to hypoxanthine by the sequential action of adenosine deaminase and PNP. Hypoxanthine is salvaged by HPRT to form IMP, consuming PRPP in the process and impacting on de novo purine synthesis (Supplementary Figure S7). IMP is able to replenish both adenine and guanine ribonucleotide pools, which may be of benefit to dividing cells. However, altering the balance between ATP and dATP pools may inhibit RNR activity and the capacity to expand dNTP pools needed for DNA replication (Hofer et al., 2012). The addition of pyrimidine deoxynucleosides to G in cycling cells caused a marginal further reduction in cell numbers compared to G alone, suggesting that dysregulation of de novo purine synthesis is, however, not the only mechanism for the reduction in cell numbers in the cycling cell model. It therefore seems likely that nuclear replication and/or mitosis was stalled in the large G-any treated cells.

MtDNA synthesis can be uncoupled from the cell cycle because it uses dNTPs from the salvage pathway outside of S phase. Hence, mtDNA content can continue to increase, even if nuclear replication is stalled. It is conceivable that other processes such as inhibited mitophagy could also contribute to the increase in mtDNA content. Despite increasing mtDNA copy number, the G-containing supplements generally decreased mitochondrial membrane potential, suggesting that the additional mtDNA may be ineffective or damaged. Hence, many of the combinations of supplements in the G-any group are probably toxic both to nuclear and mtDNA synthesis.