|Home | About | Journals | Submit | Contact Us | Français|
Frataxin (FXN) is a highly conserved mitochondrial protein. Reduced FXN levels cause Friedreich ataxia, a recessive neurodegenerative disease. Typical patients carry GAA repeat expansions on both alleles, while a subgroup of patients carry a missense mutation on one allele and a GAA repeat expansion on the other. Here, we report that selected disease‐related FXN missense mutations impair FXN localization, interaction with mitochondria processing peptidase, and processing.
Immunocytochemical studies and subcellular fractionation were performed to study FXN import into the mitochondria and examine the mechanism by which mutations impair FXN processing. Coimmunoprecipitation was performed to study the interaction between FXN and mitochondrial processing peptidase. A proteasome inhibitor was used to model traditional therapeutic strategies. In addition, clinical profiles of subjects with and without point mutations were compared in a large natural history study.
FXNI 154F and FXNG 130V missense mutations decrease FXN 81–210 levels compared with FXNWT, FXNR 165C, and FXNW 155R, but do not block its association with mitochondria. FXNI 154F and FXNG 130V also impair FXN maturation and enhance the binding between FXN 42–210 and mitochondria processing peptidase. Furthermore, blocking proteosomal degradation does not increase FXN 81–210 levels. Additionally, impaired FXN processing also occurs in fibroblasts from patients with FXNG 130V. Finally, clinical data from patients with FXNG 130V and FXNI 154F mutations demonstrates a lower severity compared with other individuals with Friedreich ataxia.
These data suggest that the effects on processing associated with FXNG 130V and FXNI 154F mutations lead to higher levels of partially processed FXN, which may contribute to the milder clinical phenotypes in these patients.
Friedreich ataxia (FRDA) affects about one in every 50,000 people in the United States. This slowly progressive ataxia frequently begins in the first decade of life and is associated with dysarthria, spasticity in the lower limbs, scoliosis, absence of lower limb reflexes, and loss of position and vibration sense.1, 2, 3 At present, there is no cure or effective treatment. FRDA is characterized by decreased expression of the mitochondrial FXN protein, from the frataxin gene on chromosome 9. FXN is important for proper mitochondrial function, but the mechanism by which decreased expression leads to disease pathology is not entirely known.4 FRDA is most commonly caused by an expansion of a GAA repeat tract in the first intron of the FXN gene on both alleles, and less commonly by a GAA repeat on one allele accompanied by a point mutation in the other FXN allele. In typical FRDA, the length of the shortest GAA expansion correlates with disease severity; longer GAA expansions result in earlier onset and a faster progression.5, 6, 7 The phenotype of patients who carry a GAA expansion on one allele and a missense mutation on the other allele cannot be predicted with certainty; these patients can have a mild or severe clinical outcome,8 creating a unique platform to understand clinical and genetic heterogeneity. In general, patients carrying G130V mutations have milder phenotypes in single case reports and small series. In contrast, individuals with mutations in W155R and R165C have much more severe phenotypes.9, 10, 11, 12
Upon entry into the mitochondria, FXN1–210 is processed by mitochondria processing peptidase (MPP) into FXN42–210, followed by FXN81–210. Many missense mutations in the C‐terminus end of FXN have been identified in FRDA, yet, few have been characterized in vivo or in situ. Some are proposed to disrupt mRNA expression (various splice site mutations such as c.165 + 1 G>A and c.384 −2 A>G),11, 12 translation initiation (c.1A>T, c.2T>C, c.2delT, c.3 g>T, c.3G>A), or protein folding (L106S).9, 13, 14, 15 These should produce little to no functional protein, and their associated phenotype should be severe in conjunction with a long GAA repeat on the other FXN allele. In contrast, R165C, W155R, G130V, and I154F mutations are suspected to produce stable protein. However, R165C and W155R lead to biochemical deficiencies in vitro.16, 17 The mechanism behind the disease‐causing features of G130V and I154F is less clear, having been suggested to reflect abnormal maturation or dysfunctional FXN in different models.9 In the present study, we have ascertained the features of these mutants in mammalian cell systems to understand from a cellular perspective how they might lead to dysfunction in FRDA.
Each FXN mutant was created using the pcDNA3.1 plasmid with wild‐type human FXN containing a C‐terminus hemagglutin (HA) tag (Addgene Plasmid #31895) and the Agilent QuikChange XL Site‐Directed Mutagenesis Kit.
Human embryonic kidney (HEK 293) cells were grown on coverslips and transfected via Lipofectamine 2000 reagent with 4 μg of DNA (2 μg FXN and 2 μg mito‐GFP). Twenty‐four hours after transfection, cells were fixed with 4% paraformaldehyde followed by treatment with blocking buffer containing 5% normal goat serum, 3% Triton X‐100, and 1% BSA. Primary antibody to the HA epitope was added at a 1:100 dilution overnight. Alexa Fluor 568 secondary antibody was added at a dilution of 1:100 and cells were imaged by confocal microscopy.
Following transfection of FXN‐mutant constructs, HEK 293 cells were lysed with buffer containing: 150 mmol/L sodium chloride, 1 mmol/L EDTA, 100 mmol/L Tris‐HCl, 1% Triton X‐100, 1% sodium deoxycholate, 0.1% sodium dodecyl sulfate, and protease inhibitor cocktail (Millipore #539134) 1:1000 at pH 7.4 for 1 h and centrifuged at 150g to collect whole cell lysates. Whole cell lysates were centrifuged at 100g followed by 150g to separate the soluble mitochondria fraction from the cytosolic fraction, and 100g to collect insoluble mitochondria pellet from soluble mitochondria fraction using a Thermo Scientific Mitochondria Isolation Kit for Mammalian Cells (#89874). The protein concentration of each fraction was determined using a BCA protein assay, and 4 μg of each fraction was loaded on a 12% NuPage gel for electrophoresis, followed by transfer to nitrocellulose membranes. Membranes were blocked with 3% milk for 1 h and incubated with primary HA antibody overnight at 4°C. Membranes were then incubated with secondary HRP‐conjugated antibody for 2 h and immunoreactive bands were visualized using luminol‐enhanced chemiluminescence (ECL) HPR substrate.
Twenty‐four hours after transfection, cells were lysed with buffer containing: 150 mmol/L sodium chloride, 1 mmol/L EDTA, 100 mmol/L Tris‐HCl, 1% Triton X‐100, 1% sodium deoxycholate, 0.1% sodium dodecyl sulfate, and protease inhibitor cocktail (Millipore #539134) 1:1000 at pH 7.4 for 1 h. For coimmunoprecipitation, 2 μg of MPP primary antibody was added to 800 μg of total lysate and rocked for 2 h at 4°C. The lysate and antibody solution was then added to washed Protein G Agarose beads overnight, rocking back and forth at 4°C. The following day the beads, lysate, and antibody solution were centrifuged at 14,000g and washed five times with IP lysis buffer containing: 150 mmol/L sodium chloride, 1 mmol/L EDTA, 100 mmol/L Tris‐HCl, 1% Triton X‐100, and 0.5% sodium deoxycholate at pH 7.4. Sample buffer (2X) was added to the beads and heated to 100°C for 5 min. The immunoprecipitated proteins were loaded on a 12% NuPage gel. Normal IgG primary antibody was used as a control as well as anti‐FXN primary antibody, followed by Trueblot secondary HRP‐conjugated antibody (Rockland #18‐8841‐31) to detect immunoreactive bands.
Transfected cells were treated with 10 μmol/L of MG132, cell‐permeable proteasome inhibitor, for 5 h. Following cell lysis, equal amounts of total cell lysate were loaded on a 12% NuPage gel.
Fibroblast cells from healthy controls and FRDA patients with point mutations were lysed in buffer (0.25 mol/L NaCl, 5 mmol/L EDTA, 50 mmol/L HEPES [pH 7.5], 0.1% NP‐40, 0.5 mmol/L DTT) supplemented with 0.1% protease inhibitor cocktail (Sigma Aldrich) and kept on ice for 20 min. The lysates were centrifuged at 20,000g for 10 min at 4°C. The clarified supernatants were transferred to fresh tubes and protein concentrations were determined by Bradford assay. A quantity of 75 μg of whole cell lysate were separated by SDS‐PAGE and transferred to a PVDF membrane. Immunoblotting was performed with antibodies against FXN (Santa Cruz Biotechnology) and GAPDH (Millipore), and the signals were detected by HRP‐mediated chemiluminescence. Densitometry was performed using Image J software (NIH), and the calculated signal ratio of FXN42–210 to FXN81–210 in each group is plotted. The bars represent the average signal for each group: CTRL= 5 fibroblast lines (n = 13), G130V =3 lines (n = 17), and Typical = 7 lines (n = 8). The asterisk indicates significant differences as determined by Student's t‐test (P < 0.05).
Image J Software was used to quantify FXN levels on western blots and is represented as mean ± SEM. Two‐tailed Student's t‐test was used to compare mutants to WT. Significance was set at P < 0.05. Image J software was also used to calculate Pearson's correlation coefficient for quantification of colocalization in immunofluorescence images.
Clinical measure results were derived from a long‐standing natural history study from 12 American and Australian sites.18 In this study, data is collected annually on clinical features of > 900 individuals with FRDA. Data from the baseline cross sectional visits were used in this study including overall medical history and scores on the Friedreich Ataxia Rating Scale (FARS) (a quantified neurological exam); Ataxia Staging scale (a disability score); the Timed 25‐Foot Walk (T25FW), scored as the reciprocal (a simple performance test of walking); 9‐Hole Peg Test (9HPT), scored as the reciprocal (a simple test of hand function); Contrast Letter Acuity test, the sum of the number of letters read on each of three Sloan charts (a quantitative test of vision); and an Activities of Daily Living (ADL) scale. All these measures capture progressive neurological dysfunction in FRDA. The performance measures were also transformed into Z‐scores to create composite scores as reported previously. The Z2 composite is the sum of the Z‐scores from T25FW and 9HPT. The Z3 composite is the sum of Z‐scores from T25FW, 9HPT, and overall vision tests.
To determine the effects of FRDA‐associated missense mutations on FXN import into the mitochondria, FXN variants containing a C‐terminal HA tag were cotransfected with mito‐GFP in HEK 293 cells. Levels of the FXN81–210 form of FXNI154F and FXNG130V are lower as determined by western blot compared to FXNWT, while no detectable exogenous FXN81–210 was detected following transfection of FXNG137V and FXNL106S constructs (Fig. 1). Confocal microscopy imaging was used to determine localization of the exogenous FXN proteins. FXNR165C and FXNW155R colocalize with mito‐GFP and have FXN immunoreactivity comparable to FXNWT (Fig. 2A). FXNI154F, FXNG130V, and FXNG137V colocalize with mito‐GFP but have lower FXN immunoreactivity compared to FXNWT (Fig. 2B). Finally, transfection of FXNL106S leads to no FXN immunoreactivity (Fig. 2C). All expressed mutant proteins colocalize with mito‐GFP with a Pearson's correlation coefficient greater than 0.98. While FXNR165C and FXNW155R have increased immunoreactivity compared to FXNI154F, FXNG130V, and FXNG137V, the FXNI154F, FXNG130V, and FXNG137V proteins retain their mitochondrial localization.
To investigate further the decrease in FXN81–210 levels of particular FXN‐mutant proteins, subcellular fractionation and separation of the soluble mitochondrial fraction and the insoluble mitochondrial pellet was performed. Consistent with immunocytochemistry results, transfection of FXNR165C or FXNW155R leads to FXN81–210 levels comparable to levels of FXNWT, while transfection of FXNI54F or FXNG130V produces lower levels of FXN81–210 (Fig. 3A and and3B).3B). While FXNI154F or FXNG130V lead to low FXN81–210 levels, expression of these mutant proteins leads to an increased level of FXN42–210 (Fig. 3A and C), suggesting these FXN variants are not processed readily from FXN42–210 to the FXN81–210 form. Furthermore, these variants also have increased ratios of insoluble to soluble FXN42–210 (Fig. 3E‐G), suggesting these proteins remain associated with the insoluble inner mitochondrial membrane rather than being released into the soluble portion of the mitochondrion.
To examine the mechanism by which FXNI154F and FXNG130V impair FXN processing, FXN‐mutant proteins were coimmunoprecipitated to study the strength of the interaction between FXN and MPP. The FXN42–210 forms of FXNI154F and FXNG130V are more readily coimmunoprecipitated by anti‐MPP than the FXN42–210 form of FXNWT, FXNR165C, and FXNW155R proteins (Fig. 4), suggesting stronger attachment between these variants and MPP.
Traditional therapies for FRDA include several strategies designed to increase FXN levels. To model this approach, transfected cells were treated with 10 μmol/L MG132, a proteasome inhibitor, to increase FXN1–210 levels in an effort to overcome impaired FXN processing. While FXNG130V and FXNI154F FXN1–210 levels increased, as did FXN42–210 levels, MG132 treatment did not increase FXN81–210 levels (Fig. 5). This suggests that amelioration of these missense mutations cannot be achieved with simple overexpression of precursor FXN, and that there is a true impediment to processing of these mutants to the FXN81–210 form.
To analyze the significance of these findings in patient‐derived cells and examine the processing of native FXN, western blots were performed on whole cell extracts prepared from control (CTRL) and FRDA G130V patient fibroblasts. Endogenous FXN42–210 and FXN81–210 FXN levels were detected and expressed as a ratio of FXN42–210/FXN81–210 (Fig. 6A). The ratio of FXN42–210 to FXN81–210 is increased in FRDA G130V patient fibroblasts compared to controls (P < 0.05). Patient fibroblasts were also immunostained with antibodies to FXN and mitofusin. FRDA G130V patient fibroblasts contain large globular structures (Fig. 6B) consistent with the increased insoluble FXN42–210 form detected by western blot and overexpression studies.
We then sought to establish whether patients carrying missense point mutations displayed distinct clinical abnormalities that could be related to the altered processing observed in vitro. Heterozygous (HTZ subjects) FRDA patients, with a missense mutation on one FXN allele and GAA expansion on the other, and typical homozygous (HMZ subjects) FRDA patients, with GAA expansions on both FXN alleles, have different clinical profiles when examined in a large natural history study (Table 1). In addition, patients carrying G130V mutations have significantly lower occurrence of cardiomyopathy, scoliosis, and diabetes, the most severe components of the disease, compared to other HTZ subjects (Table 1). Furthermore, clinical measures at baseline exam including ataxia stage, activities of daily living (ADL) scores, 9HPT‐1, T25FW‐1, Vision, 9HPT, Z2, and Z3 were significantly worse in other HTZ subjects, even though the groups were of similar disease duration, suggesting a less severe phenotype in G130V patients (Table 2). Patients with an I154F mutation fell between the phenotypic severity of G130V and other point mutations analyzed.
This study shows that the FRDA‐causing mutations FXNG130V and FXNI154F decrease FXN81–210 levels, but do not impair FXN localization to mitochondria. FXNG130V and FXNI154F appear to impair FXN processing from FXN42–210 to FXN81–210, as well as enhance binding of FXN42–210 to MPP. This impaired processing is also observed in primary fibroblasts from FRDA patients with a G130V mutation. Increasing FXNG130V and FXNI154F precursor levels does not lead to an increase in FXN81–210 levels, but does increase levels of FXN42–210. These are all consistent with a defect in peptide processing of these forms being a pathogenic mechanism in patients carrying these mutations. In addition, these two forms, especially G130V, are associated with milder features of FRDA than other point mutations or expanded GAA repeats, suggesting that these mild phenotypes may reflect the underlying FXN biochemistry.
In heterologous systems, disease‐associated mutations in FXN are abnormal in several mutation selective ways. The inability to detect FXNL106S and FXNG137V by immunostaining and western blot supports modeling studies suggesting that mutations residing within the protein core decrease protein stability.9 In vitro functional studies have also characterized FXNR165C and FXNW155R as dysfunctional mutations, causing decreased binding of FXN to Fe‐S cluster assembly complex.16, 17 Moreover, FXNR165C and FXNW155R had levels of FXN81–210 that were comparable to FXNWT, and there was no evidence for impaired processing of these two mutant forms from FXN42–210 to FXN81–210. Further functional studies in vivo may provide a correlation between the extent of dysfunction in these two FXN‐mutant proteins and severity of disease outcome.
Abnormalities in FXN processing have been explored mostly in yeast and bacteria expression systems.14, 19, 20, 21 Here, we show an increased level of the FXN42–210 form in disease‐associated missense mutations associated with milder phenotypes, not only in overexpression studies using mammalian systems, but also in primary fibroblasts from FRDA patients. FRDA patients who carry G130V express lower FXN81–210 levels than typical FRDA patients in fibroblasts, cheek swabs, and blood,22 yet have a milder clinical phenotype (Tables 1 and 2). In a large cohort of FRDA subjects, even though those with the G130V mutation have similar disease duration, they have significantly better FARS and ADL scores than individuals with other point mutations. FRDA G130V patients also have significantly lower occurrence of cardiomyopathy, scoliosis, and diabetes, and they surpass other point mutation carrying subjects on composite performance measures. As suggested in single cases previously, this demonstrates that FRDA patients with FXNG130V demonstrate greater neurological function and decreased disease severity at a similar length of disease duration.8, 10, 11, 15, 23, 24, 25, 26, 27, 28, 29 Patients with I154F mutations have clinical severities intermediate between other patients with point mutations and G130V patients, matching the data from cellular models of the molecular consequences of these mutants.
One explanation for the milder phenotype in patients carrying a G130V or I154F mutation is that the incompletely processed FXN42–210 carries some residual activity. In these mutants, this form is located in the mitochondria, and others suggest that FXN42–210 can perform Fe‐S cluster synthesis as well as participate in cysteine desulfurase activity as efficiently as FXN81–210.30, 31 Thus, the higher levels of FXN42–210 associated with FXNG130V could lead to the mild phenotype of patients with G130V if this intermediate form is functional. Alternatively, it is possible that a small but clinically significant amount of the FXN42–210 form is slowly converted to the mature form, leading to the milder phenotype in patients carrying G130V mutations compared with other mutations that yield absolutely no mature FXN. Further experiments examining the functional abilities of the FXN42–210 form of endogenous FXNG130V may help clarify these possibilities. Overall, the present study, in agreement with modeling studies and those in lower animal expression systems, identifies multiple mechanisms in mammalian heterologous systems by which FXN point mutations can lead to FRDA.
EC, JSB, IC, MN, and DRL conceived and designed the study. EC, JSB, and CI acquired and analyzed the data.
This work was supported in part by the NIH training program in neuropsychopharmacology at the University of Pennsylvania (R01MH109260), as well as by grants from the NIH (R21NS087343) and (R01NS081366 to MN) and the Friedreich Ataxia Research alliance to DRL, MN, and JSB. We also acknowledge the Collaborative Clinical Research Network in Friedreich Ataxia for their contribution to the clinical data.
This work was funded by University of Pennsylvania grant R01MH109260; NIH grants R21NS087343 and R01NS081366; Friedreich Ataxia Research alliance grant .
[Correction added on 8 August 2017 after first online publication: the last name of the fourth author was incorrectly spelled and was changed from Napierela to Napierala.]