Generation of frataxin I151F knock-in mice
FXNI151F/wt heterozygous mice (C57BL/6J-Fxnem10(T146T,I151F)Lutzy/J) were obtained from the Jackson Laboratory (Stock Number 31922). To generate these mice, CRISPR/Cas9 was used to introduce the ATC→TTC mutation into codon I151 of murine Fxn. The resulting I151F substitution is equivalent to the human pathogenic I154F missense mutation (Figure 1A). An additional silent mutation is present in codon T146 (ACC→ACT). This change was deliberately co-introduced with the desired mutation in order to destroy the guide dependent protospacer adjacent motif (PAM) recognition site. Inter-crossing of these mice resulted in progeny of all possible genotypes. Of 180 mice born, 48 (26,7%) were WT mice (FXNwt/wt homozygous for wild type FXN), 91 (50,5%) were HET mice (FXNI151F/wt, heterozygous for wild type FXN and FXNI151F), and 41 (22,8%) were FXNI151F mice (FXNI151F/I151F , homozygous for the FXNI151F allele) (Figure 1D). Therefore, progeny was within the expected mendelian ratios indicating that FXNI151F mice were viable.
Frataxin protein levels are markedly reduced in FXNI151F mice
To determine the effect of the I151F mutation on frataxin content, we measured the levels of this protein in cerebrum, cerebellum and heart from 21 and 39-week-old WT, HET and FXNI151F mice. By western blot (using monoclonal antibody Ab219414), we observed that the content of mature frataxin in HET mice was approximately 50% of that observed in WT mice, while less than 6% of mature frataxin content was observed in FXNI151F mice (Fig. 2A and supplemental Figure 1B). We also analyzed the frataxin content in homogenates from spinal cord, dorsal root ganglia (DRG), liver, pancreas and skeletal muscle. These analyses confirmed that the I151F mutation causes a marked loss of frataxin content in all mouse tissues (Supplemental Figure 1). In order to exclude the possibility that the decreased frataxin signal in the western blot was caused by a loss of epitope recognition by the monoclonal antibody employed, three additional antibodies were used to detect frataxin in cerebrum homogenates. These were two polyclonal antibodies raised against full-length frataxin (Mbs8245785 and Ab175402), and a monoclonal antibody (Ab113691) raised against a peptide within aa 1-120 from human frataxin (thus not including I154). Also, we overexpressed murine frataxin in HEK293 cells in order to identify the migration pattern of mature, intermediate and frataxin proteoforms. These results are shown in supplemental figure 2A. Three major bands can be overserved in overexpressing cultures frataxin: the slower migrating band has an apparent molecular weight of 26.6 kDa, and corresponds to the precursor form (theoretical MW 22.9 kDa); the intermediate form has an apparent MW of 18.1 kDa (theoretical MW 18.6 kDa); the mature form has an apparent MW of 15.3 kDa (theoretical MW 14.4 kDa). Regarding frataxin detection in cerebrum homogenates with the different antibodies used, all of them were able to detect a band migrating at 15 kDa in WT samples. This band disappeared in I151F samples, confirming that the mature frataxin content in FXNI151F mice is very low. The Mbs8245785 and Ab113691 antibodies recognized bands in the 20-26 kDa range, but these bands may be caused by cross-reaction of the antibody with unknown proteins, as they were not detected by antibodies Ab219414 and Ab175402. To further confirm the frataxin decrease in FXNI151F mice, we used a mass spectrometry-based targeted proteomics SRM approach to detect frataxin in tissues collected from WT and FXNI151F mice. This approach was based on the detection of proteotypic peptides from frataxin in trypsin-digested tissue homogenates, which were analyzed in a liquid chromatograph coupled to a triple quadrupole mass spectrometer (proteotypic peptides are those peptides in a protein that are most likely to be observed by mass spectrometry and which are unique to that protein). We selected the four peptides with higher ESS score in peptideatlas server (ESS provides an estimation of the proteotypicity of each peptide) and for each one of these we selected the three most frequently observed transitions. These transitions are indicated in Supplemental Figure 2B. All these peptides are present in both mature, precursor and intermediate forms of frataxin. Using this approach, we were able to detect a signal in heart homogenates eluting at minute 22.7, corresponding to the 2+ to y7+ transition of the LGGDLGTYVINK peptide. We could not detect this signal in cerebrum or cerebellum homogenates, probably because frataxin is less abundant in the nervous system than in heart (as observed in figure 2A). None of the other peptides analyzed were detected. It is worth mentioning that frataxin is not easily detected by mass spectrometry as according to peptide atlas frataxin has been observed 386 times, while mitochondrial aconitase has been observed 26660 times. In order to confirm that the signal observed in heart homogenates corresponded to the LGGDLGTYVINK peptide, an isotopically labeled heavy version of the peptide was purchased and included in the analysis as an internal standard. This heavy form presented the same retention time as the light form. Moreover. the light form was not detected in heart from FXNI151F mice (Supplemental figure 2C), further confirming that frataxin was markedly decreased in these animals. Due to its low intensity, we were not able to accurately quantify the residual frataxin present in mutant mice, but we could estimate that it was below 5% of WT values. This value is consistent with the western blot data and confirms that frataxin is markedly decreased in FXNI151F mice. These results also indicate that intermediate or precursor proteoforms are not detected in neither WT or FXNI151F mice.
Insoluble intermediate frataxin proteoforms are not detected in FXNI151F mice
The presence of an insoluble intermediate frataxin proteoform has been reported in cells overexpressing human I154F frataxin . Results presented in the previous section suggest that this form may not be present in FXNI151F mice, as the denaturing conditions used for protein extraction in these assays should solubilize such intermediate insoluble forms. Nevertheless, in order to confirm this point, we analyzed the potential presence of this insoluble form under the same conditions used to detect human insoluble FXNI154F. To this purpose, we prepared tissue homogenates under native conditions, and we separated the soluble protein fraction from the insoluble protein fraction by centrifugation. After solubilizing the insoluble fraction in a buffer containing 4% SDS (a concentration high enough to solubilize the previously reported I154F insoluble intermediate frataxin form) both fractions were loaded on SDS-PAGE gels and frataxin proteoforms detected by western blot (antibody Ab219414). Membranes were overexposed in order to detect any potential intermediate or high molecular weight forms of frataxin. In these membranes, we could detect the presence of some high molecular weight bands in WT homogenates (Figure 2B). Nevertheless, these high molecular weight forms represented less than 1% of the chemiluminescent signal from the mature form, and were also decreased in homogenates from FXNI151F mice. Therefore, we can exclude the presence of an insoluble intermediate frataxin proteoform in mice harboring the I151F mutation.
FXNI151F mice present decreased weight gain
Body weight of WT and FXNI151F mice was measured every two weeks from birth. Weight gain was similar in WT and FXNI151F mice until 10 weeks of age. From that age on, weight gain was lower in FXNI151F mice than in WT mice. Therefore, significant weight differences were observed between WT and FXNI151F mice from 15 weeks of age onward (Figure 3A and B). The relative weight difference between WT and FXNI151F mice was progressive and was similar in both females and males. At 39 weeks of age, FXNI151F mice presented on average a 23% decrease in weight when compared with WT mice (Figure 3C and D). HET mice did not present any weight loss (Supplemental Figure 3A).
FXNI151F mice exhibit neurological deficits
To determine whether frataxin deficiency caused by the I151F mutation impacted the behavior of FXNI151F mice, we subjected these mice to several motor behavioral tasks. WT mice were also analyzed as a control. The number of mice used for each analysis is summarized in Supplemental Table 1. The motor coordination ability was assessed on a rotarod treadmill. This analysis was performed every two weeks, from 15 to 39 weeks of age. As shown in figure 4A, FXNI151F mice showed decreased coordination ability compared with WT. Differences were statistically significant from week 23 onward. Forelimb strength was tested using a hanging-wire test, where mice were assessed for their ability to hang on a horizontal wire by their forepaws. Analyses were performed every six weeks (starting at week 21). FXNI151F mice fell off the wire quicker than WT mice. Statistically significant differences were obtained for 27-, 33-, and 39-week-old mice (Figure 4B). Locomotor activity tests were performed using an open-field beam-breaker activity monitor. These analyses were performed every six weeks (starting at week 21). FXNI151F mice exhibited significantly reduced average velocity, ambulatory distance (total distance covered by the mice within a specific time), and number of crossings than WT mice (Figure 4C). Finally, we assessed gait ataxia using paw print analysis every six weeks (starting week 21). Significant differences between FXNI151F and WT mice were detected in 39-week-old mice, but not before. At this age, FXNI151F mice displayed reduced hind and front limb stride length compared with WT. This suggests that FXNI151F mice present a progressive ataxic gait (Figure 4D). Overall, the results presented in this section indicate that FXNI151F mice present progressive neurologic defects that are not observed before 23 weeks of age. Finally, in order to explore the effect of the I151F mutation in heterozygosis, we analyzed the performance of HET mice in an open field test. No significant differences were observed in the open field test between WT and HET mice, neither in velocity, distance travelled, nor number of crossings (Supplemental Figure 3B).
Analysis of frataxin-related proteins by targeted proteomics
We were interested in analyzing the biochemical consequences of frataxin deficiency in nervous and cardiac tissues before (21-weeks) and after the onset of neurological defects (39-weeks). With this purpose, we decided to focus on several proteins or pathways that have been related to frataxin. In this regard, frataxin deficiency has been described to cause loss of iron-sulfur containing proteins , decreased function of the OXPHOS system , changes in the content of superoxide dismutases , and decreased content of the pyruvate dehydrogenase component PDH A1 . Also, frataxin has been reported to interact with components of the iron-sulfur biosynthesis machinery , the OXPHOS system , and with the mitochondrial chaperone GRP75 [26, 34]. Therefore, in order to explore these pathways in FXNI151F mice, we set up a targeted proteomics method to analyze the content of proteins related to these pathways in tissues collected from WT and FXNI151F mice. We used an SRM-based targeted proteomics approach in which proteotypic peptides from each protein were detected in a LC-triple quadrupole mass spectrometer. This approach allows the quantitation of several proteins in multiple samples with high reproducibility. The proteins analyzed are listed in Table 1. They consisted of two components of the tricarboxylic acid cycle (CS and ACO2), several components of the OXPHOS system (SDHA, SDHB, CY1, QCR2, COX2, ATPA, and ATPB), two mitochondrial chaperones (HSP60 and GRP75), three components of the Pyruvate dehydrogenase complex (PDHA1, DLAT and DLDH), and two superoxide dismutases (SOD1 and SOD2). Three glycolytic enzymes were also included in the analysis (GAPDH, PKM, and Enolases A, B, G) in order to analyze a potential imbalance between respiratory and glycolytic pathways and/or between mitochondrial and cytosolic pathways. Two of the proteins analyzed contain iron-sulfur clusters (ACO2 and SDHB), while two of them (CYC and CY1) contain heme groups. The relative content of these 21 proteins was analyzed in cerebrum, cerebellum, and heart from WT and FXNI151F mice, sacrificed at 21 or 39 weeks of age (in the case of the enolases, ENOG only in nervous systems and ENOB only in heart). Figure 5A-C summarizes the most relevant findings while Supplemental Figure 4 contains the results from all the proteins analyzed. It can be appreciated that the most affected proteins were ACO2 and the two components of the OXPHOS complex II (SDHA and SDHB). These proteins showed a marked decrease in cerebellum and cerebrum from FXNI151F mice, both at 21 and 39 weeks. In heart, we could only observe loss of complex II in 21-week-old mice, but not in 39-week-old mice, while ACO2 content was not altered in heart at any age. Some other components of the OXPHOS system presented minor changes: the complex III components QCR2 and CY1 were decreased in cerebrum (21 weeks), cerebellum (21 and 39 weeks), and heart (21 weeks); ATPA and ATB were increased in 39-week-old cerebellum and heart. A second group of proteins showing changes in their content were the antioxidant enzymes SOD1 and SOD2. In heart, both enzymes were induced in 21-week-old mice, while their levels decreased in 39-week-old mice. Some induction of SOD2 was also observed in cerebrum at 21 and 39 weeks. Regarding PDH components, induction of both PDHA1 and DLAT was found in 39-week-old heart. In contrast, DLDH (the E3-component of the PDH complex) was decreased in 39-week-old heart and cerebellum. No major changes were observed in the content of the mitochondrial chaperones analyzed, nor of the glycolytic enzymes. The net representation shown in Supplemental Figure 4 allows a comparison of the relative content of the proteins (ratio FXNI151F /WT) at 21 and 39 weeks of age. Overall, these results indicate that frataxin deficiency does not cause a general loss of mitochondrial proteins. Instead, it causes specific changes in the content of certain proteins. Also, the net representation indicates how these changes evolve as mice age. In this regard, cerebrum and cerebellum do not experience major differences in the FXNI151F /WT protein ratio between 21-week-old and 39-week-old mice. Therefore, biochemical alterations precede the neurological defects described in the previous section. In contrast, the heart shows differences in the FXNI151F /WT protein ratio between 21- and 39-week-old mice, suggesting that the consequences of frataxin loss in this tissue are progressive. Finally, we analyzed the content of some representative proteins in cerebrum, cerebellum, and heart from HET mice. The proteins analyzed were CS, ACO2, SDHB, ATPB, SOD2, DLAT, GRP75, and PKM. This group includes representative proteins that experienced changes in the mutant mice (ACO2, SDHB, ATPB and SOD2), and also proteins that did not experience any changes, but are representative of different pathways/functions. These proteins were analyzed in cerebrum, cerebellum, and heart from 39-week-old mice by targeted proteomics. None of these proteins presented relevant changes in their content comparing HET and WT animals (results shown in Supplemental Figure 3). A slight decrease in the amounts of CS and SDHB was observed in the cerebellum of HET mice, but these changes were lower than those observed in FXNI151F mice and/or were not observed in cerebrum and heart.
FXNI151F mice present lower aconitase activity
In order to validate the results of the targeted proteomics analysis, we decided to measure the content, activity, and/or expression levels of those proteins most markedly altered in the proteomic analysis. We first focused on ACO2, a mitochondrial iron-sulfur enzyme, which converts citrate to isocitrate in the tricarboxylic acid cycle. As this enzyme is one of the most abundant iron-sulfur containing proteins, it is commonly used as an indicator of the status of iron-sulfur centers in the cell. Therefore, we analyzed its content by western blot and its activity by spectrophotometry. These analyses were performed in cerebrum, cerebellum, and heart homogenates from 21-week and 39-week-old mice. ACO2 is expected to account for most of the aconitase activity in these tissues as it represents between 80 and 90 % of all aconitase content according to the PaxDb database. Western blot analysis confirmed the results observed in the targeted proteomic analysis, as the results obtained were similar to those obtained in the previous section: a marked decrease was observed in the nervous system, while in heart tissue, no significant differences were observed between WT and FXNI151F mice (Figure 6). Regarding activity, we measured the ratio between aconitase and citrate synthase activities (ACO/CS ratio), which was used as a control for non-iron-sulfur mitochondrial enzyme. We observed a significantly decreased ACO/CS activity ratio in the three tissues tested, with more marked decreases in cerebrum and cerebellum than in heart (Figure 6). Overall, the results obtained confirm that aconitase content and activity is more affected in the nervous system than in heart. They also suggest that decreased aconitase activity is mostly due to decreased ACO2 protein content, although the presence of some inactivated protein cannot be excluded (as in some of the conditions analyzed, loss of activity is slightly higher than loss of protein content).
FXNI151F mice present decreased content of OXPHOS complexes I and II
The targeted proteomics analysis also revealed changes in the content of several components of the OXPHOS system, notably complex II. In order to confirm these results with an alternative approach, we used western blot to analyze the content of five components of the complexes I to V (one protein per complex) in cerebrum, cerebellum, and heart from 21- and 39-week-old mice. These proteins were NDUB8 from complex I, SDHB (complex II), QCR2 (complex III), COX1 (complex IV), and ATPA (complex V). None of the complex I components had been included in the targeted proteomics analysis as they could not be properly quantified. The results obtained are shown in Figure 7. It can be observed that FXNI151F mice present a marked loss of NDUB8 from complex I and SDHB from complex II in all tissues analyzed (except heart at 39 weeks), and that this decreased content was already observed at 21 weeks of age. The results from SDHB are similar to those found in the targeted proteomics analysis, confirming the validity of the data obtained. Regarding complexes III to V, a small decrease in QCR2 content (from complex III) could be appreciated in cerebellum, as previously observed in the targeted proteomics approach. No changes in ATPA content were observed by western blot analysis. Overall, the results obtained by the targeted proteomics approach and the western blot analysis indicate that FXNI151F mice present a deficiency in complex I and II components of the OXPHOS system, especially in the nervous system, while the components of the other complexes do not present major differences.
mRNA levels from NDUFB8, SDHA, SDHB, and ACO2
In order to discern whether the decreased content of NDUFB8, SDHA, SDHB, and ACO2 proteins was caused by lowered gene expression or by other mechanisms, the mRNA levels of these genes were analyzed by qPCR (Figure 8). We performed this analysis in cerebellum and heart tissues from 21- and 39-week-old mice. In cerebellum, no changes were found in the mRNA levels of these genes when comparing WT and FXNI151F mice. This result indicates that changes in protein content in this tissue are caused by posttranscriptional mechanisms. Regarding heart, a decrease in SDHB and NDUFB8 mRNA levels was observed in 21-week-old animals. Therefore, the decreased content of these proteins in heart can be attributed to a lowered gene expression. Indeed, in 39-week-old mice, the mRNA levels of both proteins are restored, as observed from protein content. In contrast, SDHA expression is not altered in 21-week-old mouse heart (despite lower protein content). Therefore, loss of SDHA in heart is also caused by posttranscriptional mechanisms. Overall, these results suggest that the mechanisms causing alterations in protein content may be tissue specific, and include transcriptional and posttranscriptional mechanisms.
Biosynthesis of lipoic acid is not compromised in FXNI151F mice
We also decided to analyze the content of protein-bound lipoic acid, a prosthetic group required for PDH activity. This analysis had two objectives: first, to complement the results from the targeted proteomics analysis, where several components of the PDH complex were analyzed; second, to further analyze the status of iron-sulfur clusters in FXNI151F mice, as synthesis of this cofactor requires lipoate synthase, which is an iron-sulfur containing enzyme. Western blot analysis of cerebrum, cerebellum, and heart lysates using antibodies raised against lipoate revealed the presence of two bands, at 70 and 50 kDa apparent molecular weight (Figure 9). The 70 kDa band (which is more intense) corresponds to lipoic acid bound to DLAT (the E2 component of PDH), while three lipoic acid containing proteins migrate at 50 kDa. These are the E2 component of alpha-ketoglutarate dehydrogenase (Dihydrolipoyllysine succinyltransferase, DLST), the E2 component of the branched chain alpha-ketodehydrogenase complex, and the PDH-binding component X (a structural subunit of the PDH complex) . As indicated in Figure 9, no significant differences were observed in DLAT-bound lipoic acid (70 kDa) between WT and FXNI151F mice. Decreased content in the 50kDa band was observed in 21-week-old cerebrum, but not in the other tissues analyzed. This band was also decreased in 39-week-old mice, although here differences did not reach statistical significance. Overall, these results indicate that protein-bound lipoic acid biosynthesis (which requires an iron-sulfur enzyme) is not compromised in FXNI151F mice.