Microarray analysis of heart and skeletal muscle in KIKO mice. (A) Array clustering based on mean average deviation shows sample clustering by tissue. Within tissues, samples tend to cluster by genotype. Color coding bars: Top (genotype): red, WT samples; black, KIKO samples. Left side (tissue): red, skeletal muscle samples; black, heart samples. (B) Number of differentially expressed genes at P < 0.005. Three-hundred and twenty-one probes (194 downregulated, 127 upregulated) were differentially expressed in skeletal muscle; 174 (106 downregulated, 68 upregulated) in heart of KIKO mice versus WT; (C) Heatmap depicting fold changes in cardiac and skeletal muscle samples in KIKO mice versus controls. Shades of red represent upregulation, shades of green downregulation. Genes and samples are clustered by similarity. Most genes have tissue-specific changes, whereas a subset of genes present changes across tissues. Color coding bar: yellow, skeletal muscle samples (KIKO versus average of WT); orange: heart samples (KIKO versus average of WT); (D and E) Gene ontology analysis of differentially expressed genes in heart and skeletal muscle samples from KIKO versus WT mice. Top categories include ‘sterol’ and ‘lipid metabolic process’ in skeletal muscle and ‘regulation of muscle contraction’ in heart. Bars represent the –Log of the over-representation P-value, as calculated by the DAVID algorithm (http://david.abcc.ncifcrf.gov/). Red and green represent the proportion of up and downregulated genes, respectively. The number at the end of the bar represents the number of probes differentially expressed in each category; (F) Real-time quantitative PCR (qPCR) confirmation on 14 among the top differentially expressed genes. Fold changes are expressed in log2. Error bars: standard error. qPCR confirmed tissue specificity (e.g. Myh4 only downregulated in heart, Scd2, Slc2a5, Acly only upregulated in skeletal muscle).

Genetic and pharmacologic modulation of the PPARγ pathway affects frataxin levels in vitro. (A and B) Pgc1a levels are correlated with frataxin levels in neural precursor cells from mouse models and in cells from patients with FRDA. (A) qPCR quantification of Fxn (white bars) and Pgc1a (blue bars) in neural precursor cells from the subventricular zone (SVZ-NPC) in wild-type (WT, n = 3), KIKI (n = 3) and KIKO (n = 3) mice, and in lymphoblasts and fibroblasts from FRDA patients (n = 4) and normal controls (n = 3). KIKI mice express ∼75% of normal frataxin levels, and KIKO ∼30%. Pgc1a mRNA expression levels are reduced in all these cell lines, in a degree which is proportional to Fxn downregulation. Bars represent the average of six replicates, error bars represent the standard error. P < 0.05 for all the comparisons; (B) Relative Pgc1a levels are strongly correlated with relative Fxn mRNA levels (r² = 0.9, P < 0.001); qPCR quantification in fibroblasts from controls (open circles, n = 3) and FRDA patients (filled circles, n = 4). (C and D) PGC1A downregulation via siRNAi reduces frataxin protein levels in control fibroblasts, and further in FRDA fibroblasts. Western blotting analysis (C) and quantification (D) of PGC-1A (top) and FXN (bottom) protein 72 h after transfection with siRNAi^Pgc1a. Control and FRDA are fibroblasts from healthy controls and patients, respectively. All experiments were performed on n = 4 FRDA cell lines and n = 3 control cell lines. Bars represent the average of six replicates, error bars represent the standard error. P < 0.05 for all comparisons.

C2C12 (A) and HL-1 (B) cells were transfected with shRNAi for frataxin. Expression levels of frataxin versus cells transfected with control shRNAi were assessed at 12, 24, 32, 48 and 72 h after transfection (top panel). Expression levels of nine genes were assessed at 24 and 48 h after transfection (bottom panel). Red bars: upregulated genes; green bars: downregulated genes. Fold changes are expressed in log2. Error bars: standard error. Asterisk: P < 0.05, one-sample t-test. (C) Representative western blots (top) and blot quantification (bottom) for Frataxin and Srebp1 after transfection with control shRNAi (left) and shRNAi^Fxn (right). (D) shRNAi^Fxn transfection reduces Pgc1a expression as measured by a luciferase assay in 293 cells after 24 h. (E) FRDA patients have a significantly lower insulin sensitivity index: boxplots representing the insulin sensitivity index S_I [×10⁻⁵ min⁻¹/(μU/ml)] in healthy controls and FRDA patients, matched for age and body mass index. n = 14 in each group, **P < 0.005 by Wilcoxon rank sum test. (F) In FRDA patients, S_I was significantly inversely correlated with the number of GAA repeats on the smaller FRDA allele (r = −0.55, P < 0.05). (G) Physical impairment scores are not significantly different between insulin-resistant and insulin-sensitive patients; the bar graph depicts the S_I (left) and the physical impairment scores [on a scale from 0 (normal) to 42 (most severe impairment), right] of the 50% (n = 7) most insulin sensitive (light gray bars) and of the 50% (n = 7) most insulin resistant patients (dark gray bars); P = 0.5 for comparison between the two groups.

Diabetes in FRDA and type-2 diabetes may have a final common pathway. In normal conditions (left panel), lipid breakdown and biosynthesis are tightly controlled by the two master regulators Pgc-1a and Srebp-1. In type-2 diabetes (center), increased dietary lipids provided to the skeletal muscle can overload TCA and divert long-chain-CoA species into lipid precursors than can be converted in lipids, leading to an increased lipid intramyocellular content, which in turn leads to reduced glucose transport, reduced glycogen synthesis and ultimately to insulin resistance. In FRDA (right), deficits of TCA enzymes occur, as well as decreased activity of complexes I–II–III of the respiratory chain. An ineffective TCA and mitochondrial OXPHOS would lead to TCA overflow, favoring a metabolic shift toward the redirection of citrate from the TCA cycle and electron transport chain to processing into fatty acids. In mitochondrial diseases and in FRDA in particular, the oxidative stress can contribute to the diabetes pathogenesis at both (i) skeletal muscle and liver (insulin targets) and (ii) the beta-cell level, leading to overt diabetes mellitus. TCA, tricarboxylic acid cycle; IMLs, intra-myocellular lipids.