Correction of half the cardiomyocytes fully rescue Friedreich ataxia mitochondrial cardiomyopathy through cell-autonomous mechanisms

Abstract

Friedreich ataxia (FA) is currently an incurable inherited mitochondrial neurodegenerative disease caused by reduced levels of frataxin. Cardiac failure constitutes the main cause of premature death in FA. While adeno-associated virus-mediated cardiac gene therapy was shown to fully reverse the cardiac and mitochondrial phenotype in mouse models, this was achieved at high dose of vector resulting in the transduction of almost all cardiomyocytes, a dose and biodistribution that is unlikely to be replicated in clinic. The purpose of this study was to define the minimum vector biodistribution corresponding to the therapeutic threshold, at different stages of the disease progression. Correlative analysis of vector cardiac biodistribution, survival, cardiac function and biochemical hallmarks of the disease revealed that full rescue of the cardiac function was achieved when only half of the cardiomyocytes were transduced. In addition, meaningful therapeutic effect was achieved with as little as 30% transduction coverage. This therapeutic effect was mediated through cell-autonomous mechanisms for mitochondria homeostasis, although a significant increase in survival of uncorrected neighboring cells was observed. Overall, this study identifies the biodistribution thresholds and the underlying mechanisms conditioning the success of cardiac gene therapy in Friedreich ataxia and provides guidelines for the development of the clinical administration paradigm.

Introduction

Friedreich ataxia (FA) is a rare neurodegenerative disease characterized by spinocerebellar and sensory ataxia associated with cardiac hypertrophy and diabetes (1). Cardiac dysfunction is a major medical concern in FA as it is responsible for 59% of premature death of patients (2). Moreover, cardiac anomalies have been shown to be prevalent in 85% of pediatric FA patients (3). FA is mainly caused by a guanine-adenine-adenine repeat (GAA)_n expansion within the first intron of the frataxin gene (FXN) (4). The GAA expansion causes the heterochromatinization of the locus leading to reduced transcription (4). FXN is a highly conserved mitochondrial protein regulating the biosynthesis of iron–sulfur clusters (Fe-S) (5). Fe-S clusters are prosthetic groups crucial for many biological functions, including the mitochondrial respiratory chain and iron metabolism (6). The consequence of frataxin deficiency are Fe-S cluster deficit, impairment of Fe-S enzymes such as succinate dehydrogenase (SDH), mitochondrial dysfunction and iron overload, leading to cellular dysfunction and death (7). Several therapeutics approaches targeting the secondary consequences of FXN deficiency such as iron chelators, antioxidants and activators of the mitochondria biogenesis have been evaluated without much success in clinic (8). More promising therapeutic approaches aiming at increasing steady-state FXN levels are also being developed by modifying the locus through epigenetic modulation, stabilizing the protein, protein replacement or gene therapy (9–18). Previously, we have demonstrated the therapeutic efficiency of adeno-associated virus (AAV)-mediated cardiac FXN gene transfer to prevent and rapidly reverse the cardiac phenotype in the Mck model, a conditional knock-out model that recapitulates most features of the FA cardiomyopathy (7,19). This was achieved following a single high-dose intravenous injection of AAVrh10 vector, resulting in transduction of >90% of the cardiomyocytes and high expression level of FXN protein (up to 24-fold the endogenous level). While free from functional toxicity in the mouse studies, high level of FXN overexpression is probably not required since asymptomatic carriers express around half of the normal endogenous level. It is also not desirable for long-term clinical safety that can hardly be assessed in small animal model. Furthermore, despite major progress in AAV vector production (20), engineering of new AAV serotype with optimized cardiotropism and neurotropism (21) and more efficient delivery procedure in large animal model (22), it is unlikely to achieve full transduction coverage of the heart in clinical setting. To advance this promising therapeutic strategy towards the clinic, it is therefore crucial to identify the therapeutic thresholds for the vector biodistribution and expression in the heart.

Here, the primary objective was to assess the level of cardiac function rescue following the administration of decreasing doses of the AAVrh10.CAG-hFXN-HA vector in the FA Mck model. The study was carried out at two different time points, upon early and advanced cardiac dysfunction, to investigate the effect of disease progression on the therapeutic outcome. Secondary objectives were to characterize the vector pharmacokinetic and pharmacodynamics in the heart. Through correlation and regression analyses, we have defined the minimal cardiac biodistribution, i.e. vector copies number (VCN) per cell and percentage of cardiomyocytes rescued, to correct completely or partially the cardiac function. Furthermore, we have demonstrated the lack of cross-correction between treated and untreated neighboring cells, in particular for the rescue of Fe-S cluster enzyme activity, mitochondria proliferation and iron metabolism. Finally, plasmatic concentration of GDF15, recently shown to be associated with the integrated mitochondrial stress response (23,24), increased in MCK mice with disease progression and was rescued following gene therapy proportionally to the vector heart biodistribution. Overall, these findings provide a comprehensive framework for the successful preclinical development of FA cardiac gene therapy in mouse, as well as quantifiable therapeutic thresholds for the design of appropriate therapeutic vector and cardiac delivery protocol in large animal model.

Results

Correction of cardiac dysfunction in a dose-dependent manner, both when treated upon early cardiac dysfunction or heart failure

To determine the lowest therapeutic dose of AAVrh10-CAG-hFXN-HA upon early or late cardiac dysfunction, Mck mice were injected, at either 5 or 7 weeks of age, with decreasing doses of vector. Following the first doses evaluation at 5 weeks (5.4 × 10¹³ to 1 × 10¹² vg/kg), 7 weeks Mck mice were treated subsequently with a more narrowed and relevant dose range (2.5 × 10¹³ to 2.5 × 10¹² vg/kg).

At 5 weeks, the left ventricle (LV) systolic function is modestly impaired, while at 7 weeks the LV is substantially dilated and hypertrophied leading to one-third of the normal cardiac blood output (CO) (7). In the proof-of-concept study (19), a single IV injection of 5.4 × 10¹³ vg/kg at 7 weeks was sufficient to fully reverse the cardiac phenotype by 14 weeks of age. Here, mice were treated with one of six doses, ranging from the 5 × 10¹³ vg/kg down to 1 × 10¹² vg/kg, decreasing by 2 to 2.5-fold increments. Mice treated at 5 and 7 weeks were monitored until 12.5 and 15.5 weeks of age, i.e. 7 and 8 weeks post-injection, respectively. These endpoints were selected because they are well beyond the median survival of untreated Mck mice (9.5 weeks), but allow histological and molecular analyses of the lowest dose group, whose survival was poorly improved (Figs 1A and 2A). As the phenotype of untreated Mck mice is very robust and reproducible, historical data were plotted for echocardiography, survival and body weight (BW). Echocardiography evaluations prior to treatment confirmed the same phenotype in animals of the present studies as in historical data (Supplementary Material, Tables S1 and S2).

Figure 1

Dose-dependent rescue of survival, cardiac function, hypertrophy and fibrosis in Mck mice treated at 5 weeks of age, at early stage of cardiac dysfunction. (A) Survival of Mck mice treated with decreasing doses of AAVrh10.CAG-hFXN-HA vector from 5 × 10¹³ down to 1 × 10¹² vg/kg at 5 weeks of age. Control mice were treated with vehicle (NaCl). For untreated Mck mice, historical data were plotted. Compared to untreated Mck, the survival of Mck mice treated with any of the evaluated doses, besides 1 × 10¹² vg/kg, was significantly increased (log-rank Mantel–Cox statistical test, P < 0.05). Compared to WT mice, only the survival of Mck mice treated at 2.5 × 10¹² and 1 × 10¹² vg/kg was significantly decreased. (B) BW. Statistical analyses are reported in the Supplementary Material, Table S1. (C–E) Longitudinal echocardiography analysis of LV SF (C) LV mass normalized to BW (D) and CO normalized to BW (E). Data are represented as mean ± SD and statistical analyses reported in the Supplementary Material, Table S1. See also Supplementary Material, Fig. S1 for mice individual LV SF and correlation analyses between dose-LV SF and dose-CO/BW. (F) Histological analysis of fibrosis and cell infiltrates by H&E or WGA staining. Representative imaging of the LV anterior wall. Scale bar, 50 μm. See also Supplementary Material, Figs S3 and S4 for extended fibrosis analysis.

Figure 2

Dose-dependent rescue of survival, cardiac function, hypertrophy and fibrosis in Mck mice treated at 7 weeks of age, upon heart failure. (A) Survival of Mck mice treated with decreasing doses of AAVrh10.CAG-hFXN-HA vector from 2.5 × 10¹³ down to 2.5 × 10¹² vg/kg. Control mice were treated with vehicle (NaCl). For untreated Mck mice, historical data were plotted. Compared to untreated mice, the survival of Mck mice treated with any of the evaluated doses, was significantly increased (log-rank Mantel–Cox statistical test, P < 0.05). Compared to WT mice, only the survival of Mck mice treated at 5 × 10¹² and 2.5 × 10¹² vg/kg was significantly decreased. (B) BW. Data are represented as mean ± SD and statistical analyses reported in the Supplementary Material, Table S2. (C–E) Longitudinal echocardiography analysis of LV SF (C), LV mass normalized to BW (D) and CO normalized to BW (E). Data are represented as mean ± SD and statistical analysis reported in the Supplementary Material, Table S2. See also Supplementary Material, Fig. S2 for mice individual LV SF and correlation analyses between dose-LV SF and dose-CO/BW. (F) Histological analysis of fibrosis and cell infiltrates by H&E or WGA staining. Representative imaging of the LV anterior wall. See also Supplementary Material, Figs S5 and S6 for extended fibrosis analysis.

When treated at 5 weeks, all mice injected with 5 × 10¹³ and 2.5 × 10¹³ vg/kg, as well as four out of five mice treated at 1 × 10¹³ vg/kg, survived up to 12.5 weeks and showed normal BW and echocardiography parameters (Fig. 1A–E and Supplementary Material, Table S1). Lower dose groups only showed partial or no rescue, in a dose-dependent manner. Nonetheless, meaningful therapeutic effects were also achieved at 5 × 10¹² vg/kg. At this dose, the cardiac function and hypertrophy were stabilized, while the incidence and severity of heart failure-associated comorbidities, including arrhythmia, hydrothorax and muscle wasting, were decreased. Altogether, these results indicate that 1 × 10¹³ vg/kg corresponded to the pivotal therapeutic dose between partial and complete functional rescue. Noteworthy, the heterogeneity of the therapeutic outcome among each dose group was low in the two highest dose groups but increased substantially in the lower dose groups, as illustrated by the LV shortening fraction (SF) (Supplementary Material, Fig. S1B). As demonstrated below, this was in line with the vector biodistribution and its relative heterogeneity among each dose group, especially when approaching the therapeutic thresholds.

On the other hand, Mck mice treated at 7 weeks of age, presenting with severe LV dilatation and heart failure, showed a full therapeutic rescue only when treated with 2.5 × 10¹³ vg/kg (Fig. 2C–E and Supplementary Material, Table S2). Five out of six mice treated at this dose were normalized for cardiac function, BW growth and survival. The remaining mouse in this dose group displayed subnormal but stabilized cardiac function and compensated hypertrophy. As expected, mice treated with lower doses showed decreasing survival and poorer therapeutic rescue. Strikingly, all mice treated with 1 × 10¹³ vg/kg survived up to 15 weeks, i.e. increased lifespan by 57%, despite presenting severely altered cardiac function (16–17% LV SF). Noteworthy, this meaningful therapeutic effect was achieved at a dose 5-fold lower than the one used in the initial study (19). Here, 2.5 × 10¹³ vg/kg was the pivotal therapeutic dose and similarly the heterogeneity of the therapeutic outcome increased substantially in the lower dose groups (Supplementary Material, Fig. S2B).

Interestingly, a similar trend was observed for heart fibrosis. Above these pivotal doses, there were only few or no fibrotic patches and cell infiltrates in both Mck mice treated at 5 weeks (Fig. 1F and Supplementary Material, Fig. S3) and 7 weeks of age (Fig. 2F and Supplementary Material, Fig. S5). Mice treated at these pivotal doses or below displayed increasing amount of fibrosis, either interstitial or as patches, as revealed by hematoxylin and eosin (H&E) staining and confirmed by the labelling of extracellular matrix proteoglycans with wheat germ agglutinin (WGA) staining (Supplementary Material, Figs S4A and S6A). The relative heart surface labelled with WGA, compared between wild-type (WT) and treated Mck mice, confirmed this trend but showed statistical difference only at the lowest doses (Supplementary Material, Figs S4B and S6B). Furthermore, H&E revealed the subcellular disorganization of cardiomyocytes in untreated and treated Mck mice, in the regions of the heart sections circled with blue dotted line (Supplementary Material, Figs S3 and S5). Observations at high magnification revealed that the disorganization of the myofilaments, labelled by eosin with a typical fibrillary aspect, while the endoplasmic reticulum, labelled by hematoxylin, appeared substantially dilated. Mck mice treated at the pivotal therapeutic dose or above displayed few or none of these subcellular alterations, while mice treated with decreasing dose showed increasing level of alteration.

Here, the minimal therapeutic doses were identified as 1 and 2.5 × 10¹³ vg/kg, respectively for Mck mice treated upon early and advanced cardiac dysfunction. This difference is most likely explained by the rapid progression of the disease past 7 weeks and the slow expression kinetic of the vector in vivo. Indeed, we have shown previously that the FXN protein level reaches its peak of expression in the heart, and therefore its maximum therapeutic effect, only 10 days post-injection (19).

Vector biodistribution in heart correlated strongly with vector expression and the percentage of cardiomyocytes transduced

In order to investigate the biological significance of these pivotal doses and the underlying mechanisms of the therapeutic effect, we first quantified the vector biodistribution and frataxin expression in the heart and then analyzed their correlation. The distribution of the VCN per diploid genome, human FXN mRNA level and protein concentration ([hFXN]) were quantified specifically at the level of the apex, middle and base of the heart. In both 5- and 7-week-old treated animals, the VCN, FXN mRNA and [hFXN] appeared to decrease proportionally with the dose administrated (Fig. 3A–C and Supplementary Material, Fig. S7A–C) but did not reach statistical significance due to the low number of mice per group and their relative heterogeneity. Thus, the relationship between dosing and VCN can be described by linear regression, but with moderate stringency (Fig. 3D and Supplementary Material, Fig. S7D).

Figure 3

Analysis of vector biodistribution and pharmacokinetic in the heart of Mck mice treated at 5 weeks of age. (A–C) Vector biodistribution reported as VCN per diploid genome (A), mRNA level normalized to 18S (reported as 2^-deltaCt × 10⁶) (B) and tissue concentration of human FXN protein ([hFXN]) (C). Individual values, dose group range and average level are represented. Dotted line corresponds to the minimum [hFXN] associated with full rescue of heart function. One-way analysis of variance (ANOVA) with false discovery rate (FDR) 5% statistical test: ns, non-significant; ^*q < 0.05; ^**q < 0.01; ^***q < 0.0001. (D) Correlation analysis between dose and VCN. Spearman non-parametric test: R = 0.87, P < 0.0001. Linear regression analysis: R² = 0.95. (E) Correlation between vector biodistribution and mRNA level. Spearman non-parametric test: R = 0.89, P < 0.0001. Linear regression analysis of log-transformed values: R² = 0.85. (F) Correlation between vector biodistribution and [hFXN]. Spearman non-parametric test: R = 0.91, P < 0.0001. Linear regression analysis of log-transformed values: R² = 0.72. Individual values, best fit curve and 95% confidence intervals are plotted. See also Supplementary Material, Fig S7A–F for the similar analysis of Mck mice treated at 7 weeks.

The VCN measured at the apex, middle and base of the heart, was similar for each individual mouse, despite an apparent trend towards higher VCN in the middle of the heart (Supplementary Material, Figs S8A and S9A). In line with these results, the vector expression was very similar at these different levels, when measured by quantitative reverse-transcriptase PCR (qRT-PCR) in a subset of samples (Supplementary Material, Fig. S8B). Furthermore, the vector biodistribution and expression strongly correlated and can be described robustly by logarithmic regression (Fig. 3E–F and Supplementary Material, Fig. S7E–F). The vector pharmacokinetics, i.e. ratio [hFXN]/VCN, was very similar in 5 and 7 weeks treated mice (i.e. 705 ± 501 and 634 ± 350, respectively) and not statistically different following t-test analysis (P = 0.63). Hence, the vector pharmacokinetic appears highly predictable in the mouse heart, independently of the age of treatment or disease severity at treatment and sacrifice.

Overall, the [hFXN] ranged between 2 and 10 927 ng of FXN per mg of total protein (Supplementary Material, Tables S3 and S4). Noteworthy, the mouse FXN protein concentration ([mFXN]) is on average 147 ± 42 ng/mg in the heart of naive adult WT mice (n = 6), 60 ± 11 ng/mg in healthy heterozygote mice (n = 6) and 9 ± 3 ng/mg in Mck mice (n = 6). Thus, the levels of transgenic FXN protein ranged between 0.01–74-fold the normal endogenous level. Rescue of cardiac function was systematically associated with [hFXN] ≥ 84 ng/mg (Supplementary Material, Table S3). Importantly, these numbers correspond to average values reflecting unequal expression of frataxin among the cardiomyocytes, some with low or no expression while others expressing high levels of FXN (Fig. 4A). Nonetheless, the [hFXN] therapeutic threshold identified here is within the physiological range and suggest that only a moderate increase in FXN is necessary while substantial overexpression (>70-fold) is not detrimental to the cardiac function or heart histology (Fig. 1C–F and Supplementary Material, Fig. S3).

To quantify the percentage of cardiomyocytes transduced and expressing the therapeutic FXN protein, heart sections were immunolabelling for both human FXN and dystrophin (Fig. 4A–B). In addition, the heart surface rescued for Fe–S enzyme activity was quantified on adjacent heart sections by histoenzymatic staining of SDH activity (Fig. 4C–D and Supplementary Material, Fig. S7G–H). As expected, the percentage of FXN positive cardiomyocyte and the cardiac surface positive for SDH activity decreased with the dose administrated (Fig. 4B and D and Supplementary Material, Fig. S7H). Importantly, the percentage of heart surface positive for SDH activity was homogenous throughout the heart volume in mice treated with the highest doses (Supplementary Material, Figs S8C and S9B). However, at lower doses, this was less homogeneous with higher SDH enzyme activity in the middle of the heart, in line with the VCN analysis (Supplementary Material, Fig. S8A). In contrast, the distribution of the SDH activity (SupplementaryMaterial, Fig. S8D) and FXN labelling (Supplementary MaterialFig. S8E) was overall very homogenous along short axis of the heart. In line with the expected rescue of Fe–S enzyme following FXN re-expression, there were strong correlations between SDH and FXN positive cardiomyocytes (Fig. 4E) and with VCN (Fig. 4F–G and Supplementary Material, Fig. S7I).

Figure 4

Cell distribution of human FXN protein and Fe–S cluster enzyme activity are correlated in the heart of Mck mice treated at 5 weeks. (A) Representative fluorescent microscopy imaging of heart tissue sections immunolabelled for dystrophin (red), human FXN (hFXN, green) and stained with 4′,6-diamidino-2-phenylindole (DAPI) (blue). Constant time of acquisition for hFXN signal (1000 ms). Scale bar, 50 μm. The corresponding dose and VCN are reported above each image. (B) Percentage of cardiomyocytes positive for hFXN in the LV. Individual values, dose group range and average level are represented. One-way ANOVA with FDR 5% statistical test: ns, ^*q < 0.05, ^**q < 0.01, ^***q < 0.0001. See also Supplementary Material, Fig. S8E for separate quantification in septum, anterior and posterior LV walls. (C) Histoenzymatic staining for SDH activity on adjacent sections. Scale bar, 2.5 mm. (D) Quantification of the cardiac surface stained for SDH activity. Individual values, dose group range and average level are represented. One-way ANOVA with FDR 5% statistical test: ns, ^*q < 0.05, ^**q < 0.01, ^***q < 0.0001. See also Supplementary Material, Fig. S8C for separate quantification in the apex, middle and base of the heart and Supplementary Material, Fig. S8D for separate quantification in septum, anterior and posterior LV walls and right ventricle (RV). (E) Correlation analysis between the percentage of cardiomyocytes positive for hFXN and heart surface for SDH enzymatic activity. Spearman correlation test: R = 0.83, P < 0.0001. Linear regression analysis: R² = 0.70. Individual datapoints, best fit curve and 95% confidence intervals are plotted. (F–G) Correlation analysis between VCN and the percentage of cardiomyocytes positive for hFXN or heart surface positive for SDH enzymatic activity. Individual datapoints, best fit curve and 95% confidence intervals are plotted. (F) Spearman correlation test: R = 0.67, P = 0.0028. Sigmoid regression analysis: R² = 0.64. (G) Spearman correlation test: R = 0.82, P < 0.0001. Sigmoid regression analysis: R² = 0.87. See also Supplementary Material, Fig. S7G–I for similar analysis in Mck mice treated at 7 weeks.

Figure 5

The rescue of the cardiac dysfunction and hypertrophy are predicted by the vector biodistribution, cell distribution of human FXN protein and rescue of Fe–S cluster enzyme activity. Mck mice treated at 5 weeks of age and sacrificed at 12 weeks. (A–C) Correlation analysis between VCN and LV SF, LV mass to BW and cardiac output normalized to BW. (A) Spearman non-parametric test: R = 0.69, P = 0.0027. Sigmoid regression analysis: R² = 0.80. (B) Spearman non-parametric test: R = −0.90, P < 0.000. Sigmoid regression analysis: R² = 0.93. (C) Spearman non-parametric test: R = 0.75, P = 0.0008. Sigmoid regression analysis: R² = 0.64. (D–F) Correlation analysis between VCN and gene expression of biomarkers indicative LV volume/pressure overload (Nppa), calcium handling impairment (Serca2a) and maladaptive transcriptional switch (ratio Myh7Myh6). mRNA levels are normalized to 18S and reported as 2^-deltaCt × 10⁶. (D) Spearman correlation test: R = −0.81, P = 0.0004. Sigmoid regression analysis: R² = 0.96. (E) Spearman correlation test: R = 0.86, P < 0.0001. Sigmoid regression analysis: R² = 0.86. (F) Spearman correlation test: R = −0.71, P = 0.0037. Sigmoid regression analysis: R² = 0.91. (G–I) Correlation analysis between heart surface positive for SDH activity and LV SF, LV mass normalized to BW and cardiac output normalized to BW. (G) Spearman non-parametric test: R = 0.70, P = 0.0002. Sigmoid regression analysis: R² = 0.65. (H) Spearman non-parametric test: R = 0.88, P < 0.0001. Sigmoid regression analysis: R² = 0.78. (I) Spearman non-parametric test: R = −0.67, P = 0.0005. Sigmoid regression analysis: R² = 0.67. Individual datapoints, best fit curve and 95% confidence intervals are plotted. Grey area represents the range of normal value for control mice. See also Supplementary Material, Fig. S10 for similar analysis of Mck mice treated at 7 weeks.

Treatment of 50% of cardiomyocytes is sufficient to obtain cardiac function correction, independently of the cardiomyopathy progression

Next, we conducted a series of correlative and regression analyses between the cardiac vector biodistribution and the heart function and morphological outcome, to identify the underlying therapeutic thresholds.

In both 5- and 7-week-old treated mice, the heart vector biodistribution correlated strongly with the efficiency of cardiac function rescue. This is illustrated by the robust correlation between VCN and LV SF, LV mass or cardiac output (Fig. 5A–C and Supplementary Material, Fig. S10A–C). In line with these results, the gene expression of cardiac biomarkers of pressure/volume overload natriuretic peptide type A (Nppa), calcium handling (Serca2a) and maladaptive transcriptional switch (ratio Myh7/Myh6) were also strongly correlated to the VCN (Fig. 5D–F). Strikingly, the VCN threshold for which complete rescue was achieved falls within the 0.2–0.4 VCN range, while partial correction is associated with VCN of 0.1–0.2. Furthermore, these thresholds were similar in 5- and 7-week-old treated Mck mice despite the differences in the disease progression and severity at the time of treatment.

Figure 6

Rescue of iron metabolism is predicted by vector biodistribution and is mediated by cell-autonomous rescue Fe–S enzyme activity. Microscopy imaging of heart tissue sections from control and Mck mice treated at 5 weeks and sacrificed at 12 weeks, as well as untreated 9-week-old Mck mice. (A) DAB-enhanced Perls staining of Fe³⁺ deposits. The corresponding dose and VCN are reported above. Scale bar: upper row, 1 mm; lower row, 25 μm. (B) Quantification of heart surface labelled with Perls staining. Individual values, dose-group range and average level are represented. One-way ANOVA with FDR 5% statistical test: ns, ^*q < 0.05, ^**q < 0.01, ^***q < 0.0001. (C–D) Correlation analysis between heart surface positive for Perls staining with VCN or heart surface positive for SDH enzymatic activity. Individual datapoints, best fit curve and 95% confidence intervals are plotted. (C) Spearman non-parametric test: R = −0.88, P < 0.0001). One phase regression analysis: R² = 0.81. (D) Spearman non-parametric test: R = −0.94, P < 0.0001. One phase regression analysis: R² = 0.67. See also Supplementary Material, Fig. S11 for similar analysis in Mck mice treated at 7 weeks. (E) Co-staining of heart frozen sections for SDH enzymatic activity and transferrin receptor 1 (TrfR1) in WT, untreated and treated Mck at the dose of 5 × 10¹²vg/kg. Nuclear DNA labelling with DAPI. Scale bar, 25 μm. See also Supplementary Material, Figs S12 and S13 for histological analysis of mitochondria bioenergetic and dynamic.

Remarkably, the therapeutic threshold of 0.2–0.4 VCN was associated with 45–50% of cardiomyocytes positive for FXN expression (Fig. 4F) and with 45–65% of cardiomyocytes positive for SDH activity (Fig. 4G and Supplementary Material, Fig. S7I). This was independently cross-validated by the direct correlation between the rescue of SDH activity and the rescue efficiency of cardiac function and hypertrophy (Fig. 5G–I and Supplementary Material, Fig. S10D–F). Moreover, meaningful therapeutic rescue of cardiac function and survival was achieved with as little as 30–40% of cardiomyocytes treated, in both mice treated at 5 and 7 weeks. Interestingly, cardiac output appears to benefit more easily from low cardiac biodistribution and more linearly from its increase in comparison to LV SF or LV mass.

In summary, the cellular and molecular thresholds conditioning the therapeutics effects are independent of the severity of the cardiac dysfunction and were identified as (1) the percentage of treated cardiomyocytes, i.e. ≥50%, (2) the corresponding VCN, i.e. 0.2–0.4 and (3) moderate increase of FXN protein level, i.e. >84 ng/mg.

Cell-autonomous rescue of mitochondrial hallmarks associated with FXN deficiency

Beyond the impairment of Fe-S enzyme activity, such as SDH, secondary hallmarks of FXN deficiency are increased cellular iron uptake and deposits, as well as altered mitochondria proliferation, autophagy and bioenergetics (7,25,26). Cardiac iron deposits are detected by Perls staining and electron microscopy starting 7 weeks of age in untreated Mck mice (7). In 5- and 7-week-old treated Mck mice, the prevention and clearance of Perls positive iron deposits in cardiomyocytes appeared to be dose dependent (Fig. 6A–B and Supplementary Material, Fig. S11A–B). Moreover, the amount of iron deposits was inversely correlated with the VCN (Fig. 6C and Supplementary Material, Fig. S11C) and the rescue of SDH activity (Fig. 6D and Supplementary Material, Fig. S11D).

Figure 7

Gdf15 plasmatic concentration correlates with disease progression, decreases shortly after AAVrh10.CAG-hFXN-HA administration and correlates with vector biodistribution in the heart of treated Mck mice. (A) Untreated Mck mice (grey, n = 6–11) and their control littermates (white, n = 5–11) were sampled for plasma between 3 and 9 weeks of age. Gdf15 concentration ([Gdf15]) was measured by ELISA assay. Data are reported as individual values, group range and mean level. Mann–Whitney statistical test’s P-values are reported. (B) [Gdf15] in control and Mck mice treated at 7 weeks and sacrificed 1 week later. Control mice injected with NaCl 0.9% (white, n = 10), Mck treated injected with 5 × 10¹³ vg/kg of AAVrh10.CAG-hFXN-HA (Hatched, n = 10) and Mck mice injected with NaCl 0.9% (grey, n = 7). Data are reported as individual values, group range and mean level. Mann–Whitney statistical test’s P-values are reported. See also Supplementary Material, Fig. S14 for molecular analysis of vector biodistribution and expression, analysis of cardiac dysfunction biomarkers and ATF4 pathway. (C–D) Correlation analysis between [Gdf15] and cardiac vector biodistribution, i.e. VCN (C) or heart surface rescued for SDH activity (D). Spearman non-parametric test R = −0.93, P < 0.0001 and R = −0.88, P = 0.0016, respectively. Modelling by one-phase decay (C) and sigmoid regression (D) with best fit curve and 95% confidence intervals: R² = 0.93 and 0.86, respectively.

Deficiency in Fe-S biosynthesis results in a switch of the iron regulatory protein IRP1 to its Iron Response Element (IRE) form, thereby regulating protein essential for iron metabolism, such as the iron transporter TrfR1 (27). As illustrated by the immunofluorescent labelling of TrfR1 on heart sections, untreated 9-week-old Mck mice displayed much higher signal than WT control mice (Fig. 6E). High level of TrfR1 colocalized systematically with cardiomyocytes-lacking SDH activity. Furthermore, cardiomyocytes with residual SDH activity displayed low TrfR1 labelling, while SDH negative cardiomyocytes displayed variable levels of TrfR1 immunolabelling. Treated Mck mice displayed identical colocalization patterns, independently of the dose administered or the efficiency of cardiac function rescue (Fig. 6E). These first results suggested a cell-autonomous effect, i.e only cells transduced with the AAVrh.10.CAG vector, expressing the therapeutic hFXN protein and therefore rescued for SDH activity, are rescued for iron metabolism.

To confirm the cell-autonomous nature of this therapeutic effect, we have investigated the colocalization relationship between SDH enzymatic activity, mitochondrial bioenergetic and mitochondria dynamic. Impairment of mitochondria bioenergetic and NAD+/NADH metabolism was evaluated indirectly by assessing the level of pan-lysine acetylation (AcetylK) (26,28). This post-translation modification was previously shown to be substantially elevated in untreated Mck mice heart, particularly on mitochondrial proteins, increasing with disease progression (26). Here, this was confirmed independently by western blot analysis on total heart protein extracts from WT controls and untreated Mck mice, between 3 and 9 weeks of age (Supplementary Material, Fig. S12A) and by immunofluorescent labelling of heart sections from untreated 9-week-old Mck mice and WT control mice (Supplementary Material, Fig. S12B). Untreated Mck mice displayed high level of mitochondria AcetylK, as revealed by the filamentous mitochondrial network in SDH-negative cardiomyocytes. In contrast, residual SDH-positive cardiomyocytes displayed near normal level of AcetylK labelling. Treated Mck mice displayed the same pattern of labelling (Supplementary Material, Fig. S12B) with high AcetylK signal colocalized systematically with SDH-negative cardiomyocytes, while SDH-positive cardiomyocytes always displayed low signal levels (Supplementary Material, Fig. S12B). Strikingly, identical co-localization patterns were observed also for prohibitin (Supplementary Material, Fig. S13A), a marker of mitochondria biomass and for the autophagosome receptor Sqstm1 (Supplementary Material, Fig. S13B).

Altogether, these results demonstrate a biodistribution-dependent and cell-autonomous rescue of iron metabolism and mitochondria homeostasis in the heart, when using SDH activity as a proxy for cell expression of the hFXN therapeutic transgene.

Importantly, the cell-autonomous effects were observed systematically in all treated mice, independently of the dose, VCN, percentage of cardiomyocytes treated (FXN and SDH positive) and the level of cardiac function rescue. The direct clinical implication would be that the heart function is rescued only when enough cardiomyocytes are directly transduced and collectively generate sufficient mechanical force. This is illustrated by the mouse treated with the therapeutic pivotal dose 1 × 10¹³ vg/kg imaged in the Supplementary Material, Figs S12A and S13A–B. While this mouse had fully normalized cardiac function, the associated vector biodistribution was in the lower range of this dose group. Indeed, the VCN was 0.17 while this group values ranged from 0.16 to 0.86. Similarly, the percentage of heart surface positive for SDH was 39%, while this group ranged from 33 to 80%. In line with the correlation analysis reported above, this mouse illustrates nicely the cell-autonomous mechanism underlying the therapeutic rescue of cardiac gene therapy.

Dose response rescue of the integrated stress response and the GDF15 plasma concentration

Previously, the integrated stress response (ISR) was shown to be activated in the Mck mouse heart, increasing with disease progression (29) and rescued just 1 week after the gene therapy treatment of 7-week-old Mck mice (19). More recently, the ISR has been shown to be a common cellular stress response of mitochondrial failure mediated by ATF4 and results in increase gene expression and plasma level of growth differentiation factor 15 ([GDF15]) (24,30,31).

Plasmatic [GDF15] was significantly elevated in untreated Mck mice, as soon as 5 weeks of age, upon early LV systolic dysfunction and increased with the cardiomyopathy progression (Fig. 7A). To assess the effect of the treatment on GDF15, 7-week-old Mck mice were injected with 5 × 10¹³ vg/kg of vector and sacrificed 1 week later. Significant cardiac vector biodistribution (VCN = 2.4 ± 0.9) and expression were confirmed (Supplementary Material, Fig. S14A–B). Importantly, just 1 week after treatment, there was no or very limited rescue of the cardiac gene expression for biomarkers of pressure/volume overload, maladaptive transcriptional switch and calcium handling (Supplementary Material, Fig. S14C). In contrast, gene expression of Atf4 and its target genes, including Asns, Mthfd2, Ddit3, Trib3 and Gdf15, were substantially down-regulated 1 week after treatment (Supplementary Material, Fig. S14D). In line with these results, plasmatic [GDF15] was also partially rescued just 1 week after treatment, suggesting an early and direct effect mediated by FXN re-expression and Fe-S biosynthesis rescue. Furthermore, Mck mice treated with decreasing doses of vector at 5 weeks of age, displayed [Gdf15] level directly correlating with the vector biodistribution and cardiac surface rescue for SDH activity, which also support the direct and cell-autonomous relationship between FXN, Fe-S and ISR-driven expression of GDF15 (Fig. 7D–E). Thus, plasmatic [GDF15] is a direct readout of the mitochondrial stress response associated with FXN cardiac depletion in the Mck mouse model and potentially also in a recently developed inducible mouse model that showed similar involvement of the ISR (32). In consequence, [GDF15] could be a useful biomarker for the preclinical evaluation and development of candidate drugs in these mouse models.

Discussion

To advance the preclinical development of cardiac gene therapy for FA, we have investigated the biodistribution thresholds conditioning the partial or complete rescue of the heart function in the Mck mouse model. Previously, the proof-of-concept studies have been conducted in mouse models with high dose of AAV vector, resulting in >90% of cardiomyocyte transduced and >8-fold the endogenous FXN level (11,18,19). However, the same dose and biodistribution efficiency is unlikely to be replicated in adult patients, with the current vector manufacturing and delivery methods. Therefore, it appears crucial to define the minimal therapeutic threshold as amenable objectives in order to design efficiently relevant clinical administration paradigms. Here, we demonstrate that the therapeutic outcome was mainly dependent on the vector cardiac biodistribution and not so much on the cellular level of FXN protein re-expressed. Indeed, full rescue of the cardiac function and hypertrophy was achieved systematically with around half the cardiomyocytes treated throughout the heart and 0.2–0.4 VCN. Moreover, partial correction and meaningful therapeutic effects on survival and heart failure associated comorbidities were also achieved with as low as 30–40% of treated cardiomyocytes and 0.1–0.2 VCN. Surprisingly, these therapeutic thresholds are not altered by the severity of the cardiac dysfunction at the time of treatment. Furthermore, the vector pharmacokinetic was overall very reproducible and unaffected by the age or severity of cardiac and cellular dysfunction at the time of treatment or sacrifice. Although no obvious functional or histological toxicity could be observed in the heart, up to 72-fold the endogenous level, the toxic threshold and mechanism associated with FXN overexpression will have to be fully investigated to define safety guidelines.

Furthermore, we have investigated the interplay between the vector biodistribution, rescue of mitochondrial hallmarks of FXN deficiency and the progressive rescue of the cardiac function. Here, we have shown a direct and linear relationship between the percentage of cardiomyocytes positive for FXN labelling, the rescue of Fe-S enzymes activity (SDH) and the rescue of iron metabolism and mitochondria homeostasis. Altogether, our results rule out cross-correction between transduced and untransduced cardiomyocytes for the rescue of mitochondrial hallmarks of FXN deficiency and support a cell-autonomous therapeutic mechanism. Only cells transduced with the vector express FXN protein are corrected. Noteworthy, this also excludes the horizontal transfer of mitochondria or FXN protein between cardiomyocytes, as a possible therapeutic strategy. This appears as a crucial consideration for the clinical trial design of cardiac gene therapy or other similar therapeutic strategies, as their success might also depend directly on the percentage of cardiomyocytes treated throughout patients’ heart, enabling sufficient mechanical force to be generated and to restore normal heart contractility.

Finally, we have investigated the ISR in relation to the vector biodistribution and the therapeutic outcome. While the mechanistic relationship between FXN deficiency and the ISR remains to be elucidated, the ISR activation has been shown in several FXN knock-out or siRNA mouse models (29,32,33), as well as in vitro, following FXN knock-down in human cell line (34). In the current study, GDF15, which is a target gene of the ISR and a cytokine involved in the systemic regulation of the body metabolism (35), displayed elevated cardiac gene expression and plasma concentration in untreated Mck mice. Strikingly, GDF15 plasma concentration increased with the cardiomyopathy progression and was rescued almost completely just 1 week after the administration of high dose of vector. While the ISR activation and GDF15 overexpression are secondary pathological events, they appear closely related to the rescue of Fe-S cluster biosynthesis and mitochondria function. Indeed, the cardiac expression of all ISR target genes measured was rescued just 1 week after treatment, before any significant correction of the cardiac phenotype at the functional or molecular levels. Hence, GDF15 could be leveraged as a valuable read-out, in these mouse models, to infer the therapeutic efficacy and cardiac biodistribution for other therapeutic agents, besides gene therapy, such as protein replacement (16) or drug targeting downstream mitochondrial dysfunction and stress (36,37). To date, there is no published data regarding [GDF15] levels in FA patients.

Overall, we have elucidated the biodistribution thresholds and the cellular mechanisms mediating the dose-dependent therapeutic effect of cardiac gene therapy in the Mck mouse model recapitulating FA cardiomyopathy. This provides a comprehensive framework for the preclinical development of FA cardiac gene therapy protocols in small and large animal model, and therefore the successful design of clinical trial. Beyond gene therapy, these therapeutic thresholds and the cell-autonomous effect are considerations that most likely apply to other therapeutic strategies for FA such as synthetic mRNA delivery (12), artificial transcription factor (11), protein replacement (15,16), RNA (13) or DNA (10) modifiers.

Materials and Methods

Detailed methods are given in the Supplementary Material information.

Animal procedures

Mck mice with a specific deletion of Fxn gene in cardiac and skeletal muscle were described previously (7). All animal procedures were approved by the local and national ethical committee for Animal Care and Use. Gender-balanced Mck mice at 5 and 7 weeks were injected intravenously with AAVrh10 vector encoding human FXN or NaCl 0.9%. Echocardiography were performed as described previously (38) by an operator blinded to genotype and treatment.

Histochemistry

The 5 μm thick frozen heart sections were stained with H&E or WGA conjugated with Alexa 488 nm, DAB 3,3′-Diaminobenzidine (DAB)-enhanced Perls labelling, SDH histoenzymatic activity or immunofluorescent labelling, as previously described (19).

Correlative DNA, RNA and protein analysis

RNA was extracted with Tri-Reagent® [Molecular Research Center, Cat. No. TR118] from heart frozen sections, 100 μm thick, sampled exclusively at the apex, middle or basis of the heart. DNA was extracted from heart samples following RNA isolation, following the manufacturer’s instructions. Proteins were extracted from heart tissue section 2–3 mm thick as described previously (19). VCN per diploid genome were quantified by qPCR, gene expression by qRT-PCR as described previously (19).

Enzyme-linked immunoassay (ELISA) analysis

ELISA assay was performed in duplicate, using the SimpleStep Human Frataxin ELISA Kit (ABCAM, ab176112), Mouse Frataxin ELISA kit (ABCAM, ab199078) and Mouse GDF-15 DuoSet ELISA kit (R&Dsystems, DY6385), following manufacturer instructions.

Statistical, correlation and regression analysis

Unless otherwise specified, data are reported as mean ± standard deviation (SD). Statistical analyses were carried out using GraphPad Prism 6 (GraphPad Software, La Jolla, California, USA) and methods are described in the figure legends.

Acknowledgements

We thank Shankaranarayanan Pattabhiraman, Hugues Jacob, Doulay Dembele for scientific discussions.

Conflict of Interest statement. H.P. is a scientific co-founder and advisor of AAVlife SAS. B.B. is currently employed by Adverum Biotechnologies, although the work was done previous to employment. All other authors declare no competing financial interests.

Funding

Lefoulond-Delalande Foundation, Institut de France (post-doctoral research fellowships years 2013 and 2016); the Friedreich Ataxia Research Alliance (general research grant year 2014 and Keith Michael Andrus Cardiac Research Award year 2016); the French Muscular Dystrophy Association AFM-Téléthon (Research grant year 2014); AAVLIFE SAS (sponsored research agreement years 2014–2016). Grant [ANR-10-LABX-0030-INRT], a French State fund managed by the Agence Nationale de la Recherche under the frame program Investissement d’Avenir [ANR-10-IDEX-0002-02].

Author notes

†Current address: 1035 O'Brien Drive, Menlo Park, California 94025, USA.