RESEARCH ARTICLE

Early VGLUT1-specific parallel fiber synaptic deficits anddysregulated cerebellar circuit in the KIKO mouse model ofFriedreich ataxiaHong Lin1, Jordi Magrane2, Elisia M. Clark1,3, Sarah M. Halawani1, Nathan Warren1, Amy Rattelle1 andDavid R. Lynch1,3,*

ABSTRACTFriedreich ataxia (FRDA) is an autosomal recessiveneurodegenerative disorder with progressive ataxia that affects boththe peripheral and central nervous system (CNS). While later CNSneuropathology involves loss of large principal neurons andglutamatergic and GABAergic synaptic terminals in the cerebellardentate nucleus, early pathological changes in FRDA cerebellumremain largely uncharacterized. Here, we report early cerebellarVGLUT1 (SLC17A7)-specific parallel fiber (PF) synaptic deficits anddysregulated cerebellar circuit in the frataxin knock-in/knockout(KIKO) FRDA mouse model. At asymptomatic ages, VGLUT1levels in cerebellar homogenates are significantly decreased,whereas VGLUT2 (SLC17A6) levels are significantly increased, inKIKO mice compared with age-matched controls. Additionally,GAD65 (GAD2) levels are significantly increased, while GAD67(GAD1) levels remain unaltered. This suggests early VGLUT1-specific synaptic input deficits, and dysregulation of VGLUT2 andGAD65 synaptic inputs, in the cerebellum of asymptomatic KIKOmice. Immunohistochemistry and electron microscopy further showspecific reductions of VGLUT1-containing PF presynaptic terminalsin the cerebellar molecular layer, demonstrating PF synaptic inputdeficiency in asymptomatic and symptomatic KIKO mice. Moreover,the parvalbumin levels in cerebellar homogenates and Purkinjeneurons are significantly reduced, but preserved in other interneuronsof the cerebellar molecular layer, suggesting specific parvalbumindysregulation in Purkinje neurons of these mice. Furthermore, amoderate loss of large principal neurons is observed in the dentatenucleus of asymptomatic KIKO mice, mimicking that of FRDApatients. Our findings thus identify early VGLUT1-specific PFsynaptic input deficits and dysregulated cerebellar circuit aspotential mediators of cerebellar dysfunction in KIKO mice,reflecting developmental features of FRDA in this mouse model.

KEY WORDS: VGLUT1, Parallel fiber synapse, Cerebellar circuit,KIKO, Friedreich ataxia

INTRODUCTIONFriedreich ataxia (FRDA) is the most common form of hereditaryataxia, affecting ∼5000 people in the USA. Most affectedindividuals inherit two alleles containing expanded GAA repeatsin intron 1 of the frataxin (FXN) gene, resulting in chromatincondensation and reduced expression of the mitochondrial proteinfrataxin (Bidichandani et al., 1998; Campuzano et al., 1996;Chutake et al., 2014; Grabczyk and Usdin, 2000). This protein isassociated with iron-sulfur cluster formation and ATP production(Chiang et al., 2016; Lu and Cortopassi, 2007; Martelli and Puccio,2014; Parent et al., 2015; Rötig et al., 1997; Vaubel and Isaya,2013). Symptoms begin as early as 5 years old and worsen overtime, with an initial presentation of ataxia, absent lower limbreflexes, and loss of position and vibration sense. This pattern ofneurodegeneration results in progressive ataxia, dysmetria anddysarthria (Lynch et al., 2012; Lynch and Seyer, 2014; Pandolfo,2009). Postmortem studies show degeneration of the large sensoryneurons and their axons in the spinal cord, while the cerebellumovertly degenerates to a lesser extent, largely within the dentatenucleus (DN) (Koeppen et al., 2011a,b, 2009).

While the cerebellar cortex is viewed as being spared and the DNis viewed as a key site of FRDA pathology (Koeppen et al., 2011a,2015), Purkinje cell injury and remodeling has also been reported inpostmortem brain samples of FRDA patients (Kemp et al., 2016).However, in most cases, brain samples are available from late-stageindividuals who presented with the disease for >30 years. Anyevidence of early pathological changes might be lost in the ensuingyears of disease progression. FRDA mouse models can be usefultools to search for the earliest pathological changes during diseaseprogression. Complete frataxin knockout is lethal prenatally, andneuron-specific knockouts have an early onset phenotype thatresembles the fully developed changes of FRDA (Cossee et al.,2000; Simon et al., 2004); however, this might be too severe tosearch for the earliest features of the disorder, owing to completeknockout of frataxin, whereas GAA expansion in FRDA decreasesfrataxin levels to 2-20% of those in controls (Lazaropoulos et al.,2015; Sahdeo et al., 2014). Identification of early changes requires amodel in which the phenotype is present, but slowly evolving, in thesame manner as FRDA patients progress.

The frataxin knock-in/knockout (KIKO) mouse model of FRDAmeets these features. It has a knock-in-expanded GAA repeat on oneallele (230 GAAs) and a knockout of FXN on the other allele,leading to mice with moderate overall deficiency of frataxin early inlife (20-30% of control levels), comparable to the levels in mildlyaffected patients. No overt neuronal loss appears in initial studiesbut mRNA panels from tissue share many features with those frompatients (Miranda et al., 2002). Recent neurobehavioral studies inKIKO mice show cerebellar ataxia, decreased peripheral sensitivity,Received 16 March 2017; Accepted 30 October 2017

1Departments of Pediatrics and Neurology, The Children’s Hospital ofPhiladelphia, Philadelphia, PA 19104, USA. 2Feil Family Brain and Mind ResearchInstitute, Weill Cornell Medical College, New York, NY 10065, USA.3Perelman School of Medicine, University of Pennsylvania, Philadelphia,PA 19104, USA.

*Author for correspondence (lynchd@mail.med.upenn.edu)

H.L., 0000-0003-2884-7791; S.M.H., 0000-0002-3059-1089; D.R.L., 0000-0001-7168-214X

This is an Open Access article distributed under the terms of the Creative Commons AttributionLicense (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use,distribution and reproduction in any medium provided that the original work is properly attributed.

1529

© 2017. Published by The Company of Biologists Ltd | Disease Models & Mechanisms (2017) 10, 1529-1538 doi:10.1242/dmm.030049

Disea

seModels&Mechan

isms

and decreased motor strength and endurance at 9 months of age,resembling clinical manifestations observed in late-onset FRDApatients (McMackin et al., 2016). This phenotypically abnormalmouse with solid biochemical deficits, but no overt cell loss, thusprovides a model to search for CNS abnormalities that mediate theearliest cerebellar features of FRDA. In the current study, weutilized KIKO mice to search for the early pathological changes inmouse cerebellum at asymptomatic ages.

RESULTSFrataxin is highly expressed in cerebellar Purkinje neuronsand DN large principal neurons of C57BL/6 wild-type miceWe first examined the expression and distribution of frataxinin C57BL/6 wild-type mouse cerebellum using doubleimmunohistochemical staining with anti-PV and anti-frataxinantibodies. In mouse cerebellar cortex, PV immunoreactivity isabundant in Purkinje neurons and other interneurons in themolecular layer (ML) (Fig. 1B,B′), while frataxin immunoreactivity(Fig. 1A,A′) is widely distributed in the ML, Purkinje layer (PL)and granular layer (GL), with a high level of expression offrataxin in PV-positive Purkinje neurons (Fig. 1A-C,A′-C′).In the cerebellar DN, frataxin is highly expressed in thelarge principal neurons surrounded by PV-positive synapses(Fig. 1D-F,D′-F′).

We then examined the expression and distribution of frataxin inrelation to mitochondria in mouse cerebellum using doubleimmunohistochemical staining with antibodies against ATP5A(ATP5A1), as a mitochondrial marker, and frataxin. In mousecerebellar cortex, ATP5A immunoreactivity is widely distributedin the ML, PL and GL, with a high level of expression in thesoma and dendrites of Purkinje neurons (Fig. 1H,H′). Frataxinimmunoreactivity (Fig. 1G,G′) colocalizes with ATP5Aimmunoreactivity (Fig. 1H,H′) in mouse cerebellar cortex (Fig. 1I,I′).In cerebellar DN, ATP5A immunoreactivity (Fig. 1K,K′) is highlyexpressed in large principal neurons and colocalizes with frataxinimmunoreactivity (Fig. 1J,J′) in the principal neurons (Fig. 1L,L′).This suggests crucial roles of frataxin in maintaining cerebellarmitochondrial function under normal physiological conditions.

Frataxin and the mitochondrial marker ATP5A are moreabundant in VGLUT1-positive glutamatergic synapticterminals than in GAD65-positive GABAergic synapticterminals in wild-type mouse cerebellar cortexWe further examined the distribution of frataxin in glutamatergicand GABAergic synaptic terminals of the cerebellar cortex in wild-type C57BL/6 mice using double immunohistochemical stainingwith anti-frataxin and anti-VGLUT1 (SLC17A7) or anti-GAD65(GAD2) antibodies. VGLUT1 immunoreactivity is most prominent

Fig. 1. Frataxin expression and distribution in mouse cerebellum. (A-F′) Confocal images of frataxin (FXN, red) and PV (green) fluorescence, andmerged images with DAPI-stained nuclei (blue), showing wide distribution of FXN in the cerebellar cortex (A-C′) and DN (D-F′) of C57BL/6 normal mice.(G-L′) FXN (red) andmitochondrial marker ATP5A (green) fluorescence, and merged images with DAPI-stained nuclei (blue), showing colocalization of FXNwithATP5A in the cerebellar cortex (G-I′) and DN (J-L′) of C57BL/6 normal mice. Scale bars as indicated.

1530

RESEARCH ARTICLE Disease Models & Mechanisms (2017) 10, 1529-1538 doi:10.1242/dmm.030049

Disea

seModels&Mechan

isms

in the ML, with lower intensity in the GL (Fig. 2A,A′). Frataxinimmunoreactivity (Fig. 2B,B′) colocalizes with VGLUT1-positiveglutamatergic synaptic terminals in theML and GL (Fig. 2C,C′). Bycontrast, frataxin immunoreactivity (Fig. 2E,E′) does not colocalizewith GAD65-positive GABAergic synaptic terminals (Fig. 2D,D′)in the ML and GL (Fig. 2F,F′).We then examined the expression and distribution of

mitochondria in relation to frataxin, glutamatergic andGABAergic synaptic terminals in mouse cerebellar cortex usingdouble immunohistochemical staining with anti-ATP5A and anti-VGLUT1 or anti-GAD65 antibodies. While there is morecolocalization of ATP5A immunoreactivity (Fig. 2H,H′) withVGLUT1-positive glutamatergic synaptic terminals (Fig. 2G,G′) inthe ML and GL (Fig. 2I,I′), there is much less colocalization ofATP5A immunoreactivity (Fig. 2K,K′) with GAD65-positiveGABAergic synaptic terminals (Fig. 2J,J′) in the ML and GL(Fig. 2L,L′). Taken together, our findings indicate that frataxin andthe mitochondrial marker ATP5A are more abundant in VGLUT1-positive glutamatergic synaptic terminals than in GAD65-positiveGABAergic synaptic terminals of wild-type mouse cerebellarcortex, suggesting that frataxin and mitochondria have moreimportant roles in VGLUT1-containing glutamatergic synapticterminals than in GAD65-containing GABAergic terminals.

Early specific reduction of VGLUT1 levels in cerebellum offrataxin KIKO mice at asymptomatic agesWe examined whether frataxin levels are decreased in cerebellarhomogenates of KIKO mice at asymptomatic (1, 3 and 6 monthsold) and potentially symptomatic (9 months) ages (McMackin et al.,2016). As expected, frataxin levels are significantly reduced incerebellar homogenates of KIKO mice compared with age-matchedcontrols at all ages (16-29% residual frataxin) (Lin et al., 2017). Wethen searched for early cerebellar synaptic abnormalities in KIKOmice by examining levels of glutamatergic presynaptic markers[VGLUT1 and VGLUT2 (SLC17A6)] and GABAergic synapticmarkers [GAD65 and GAD67 (GAD1)] in cerebellar homogenatesof KIKO mice at asymptomatic (1, 3 and 6 months old) andpotentially symptomatic (9 months) ages (McMackin et al., 2016).In wild-type control mice, VGLUT1 levels in cerebellarhomogenates are decreased at P180 and P270 compared with P30and P90 (Fig. S1A), whereas VGLUT2 (Fig. S1B) and GAD67(Fig. S1C) remain unaltered, and GAD65 levels are modestlyincreased at P90, P180 and P270 compared with P30 (Fig. S1C). InKIKO mice, VGLUT1 levels in cerebellar homogenates aresignificantly decreased at asymptomatic (1, 3 and 6 months old)(28%, 43%, 42% reduction; P<0.01, P<0.001, P<0.05 versuscontrols, respectively) and potentially symptomatic (9 months old,

Fig. 2. Frataxin and the mitochondrial marker ATP5A are more abundant in VGLUT1-positive glutamatergic synaptic terminals than in GAD65-positiveGABAergic synaptic terminals in C57BL/6 mouse cerebellar cortex. (A-C′) Confocal images of VGLUT1 (green) and FXN (red) fluorescence, andmerged images with DAPI-stained nuclei (blue), showing FXN in VGLUT1-positive glutamatergic synaptic terminals. (D-F′) GAD65 (green) and FXN (red)fluorescence, and merged images with DAPI-stained nuclei (blue), showing GAD65-positive GABAergic synaptic terminals. (G-I′) VGLUT1 (red) and ATP5A(green) fluorescence, and merged images with DAPI-stained nuclei (blue), showing ATP5A in VGLUT1-positive glutamatergic synaptic terminals. (J-L′) GAD65(red) and ATP5A (green) immunofluorescence, and merged images with DAPI-stained nuclei (blue), in GAD65-positive GABAergic synaptic terminals. Scalebars as indicated.

1531

RESEARCH ARTICLE Disease Models & Mechanisms (2017) 10, 1529-1538 doi:10.1242/dmm.030049

Disea

seModels&Mechan

isms

21% reduction, P<0.05 versus controls) ages (Fig. 3A,B). Bycontrast, the levels of VGLUT2 are significantly increased at allages (P<0.05, P<0.01 and P<0.001 versus controls) (Fig. 3C,D).Moreover, similar to VGLUT2, the levels of the GABAergicpresynaptic marker GAD65 are significantly increased (P<0.05,P<0.001 versus controls) (Fig. 3E,F), whereas GAD67 levelsremain unaltered (Fig. 3E,G), in cerebellar homogenates of KIKOmice compared with age-matched controls. These findingsdemonstrate early VGLUT1-specific synaptic input deficits anddysregulated VGLUT2 and GAD65 synaptic inputs in thecerebellum of asymptomatic KIKO mice.

Early VGLUT1-specific parallel fiber synaptic input deficitsin cerebellar ML of asymptomatic KIKO miceTo further pursue the VGLUT1-specific synaptic deficits in KIKOmouse cerebellum, we utilized the KIKO mouse and a transgenicmouse expressing fluorescent Dendra-labeled mitochondria in thenervous system (mitoDendra-KIKO mouse) (Magrané et al., 2014).The mitoDendra transgene allows us to study the exact relationshipof mitochondria to the synaptic changes of KIKO mice, andprovides a marker of the location of mitochondria relative to detailedcerebellar anatomy. Dendra-labeled mitochondria are widelyexpressed in the cerebellar cortex of mitoDendra mice

(Fig. 4A,A′,D,D′). We observed that frataxin immunoreactivity(Fig. 4B,B′,E,E′) colocalizes with Dendra-labeled mitochondria asexpected (Fig. 4C,C′,F,F′). The overall levels of frataxin andmitoDendra are markedly reduced in the cerebellar cortex ofmitoDendra-KIKO mice (Fig. 4E,E′) compared with controls(Fig. 4B,B′) at 3 months of age, consistent with frataxin andmitochondrial deficiency in mitoDendra-KIKO mice.

We then compared VGLUT1 immunoreactivity in the cerebellumof mitoDendra-KIKO mice with that of age-matched controls at3 months of age. In the cerebellar cortex of controls, similar to wild-type C57BL/6 mice (Fig. 2), VGLUT1 immunoreactivity is mostprominent in the ML (Fig. 5A), in which parallel fibers (PFs) arisefrom granular cells and make numerous VGLUT1-positive synapseson dendrites of Purkinje neurons. VGLUT1 immunoreactivity issparser in the GL (Fig. 5A). Interestingly, the overall levels ofVGLUT1 immunoreactivity are markedly decreased in the ML, butnot the GL, of KIKO mice (Fig. 5D,E) compared with age-matchedcontrols (Fig. 5A-C), suggesting specific deficiency ofVGLUT1-containing PF synaptic inputs onto Purkinje neurons.VGLUT1-positive glutamatergic presynaptic terminals aredramatically decreased in the ML of KIKO cerebellum (Fig. 5D′-F′) compared with controls (Fig. 5A′-C′). Quantification of VGLUT1-positive puncta confirms the reduction of VGLUT1-containing

Fig. 3. Specific reduction of VGLUT1 levels and increased VGLUT2 and GAD65 levels in cerebellar homogenates of frataxin-deficient KIKO mice atasymptomatic and potentially symptomatic ages. Western blot and quantification analysis of cerebellar homogenates (30 μg per lane) showing (A,B)VGLUT1, (C,D) VGLUT2 or (E-G) GAD65/67 levels, with actin or GAPDH as internal control, in KIKOmice and age-matched controls at postnatal days P30, P90,P180, and P270 (n=3-8 for KIKO and control mice per time point). Data are mean±s.e.m. *P<0.05, **P<0.01, ***P<0.001, two-tailed, unpaired Student’s t-test.Blots in Fig. 3 were stripped and reprobed with multiple antibodies in Fig. 7 and in figures in Lin et al. (2017); anti-actin and anti-GAPDH served as the loadingcontrol for each.

1532

RESEARCH ARTICLE Disease Models & Mechanisms (2017) 10, 1529-1538 doi:10.1242/dmm.030049

Disea

seModels&Mechan

isms

presynaptic terminals in the ML of KIKO cerebellum compared withcontrols (Fig. 5G-I, P<0.001). Our findings thus demonstrate earlyVGLUT1-specific PF synaptic input deficits onto Purkinje neurons inFRDA mouse cerebellum.

To further pursue the association of VGLUT1-specific PFsynaptic inputs with cerebellar dysfunction in KIKO mice, weutilized transmission electron microscopy to examine theultrastructure of PF synaptic terminals on Purkinje neurons in the

Fig. 4. Reduced frataxin levels in the cerebellumof asymptomatic KIKO mice carrying amitoDendra marker. (A-F′) Confocal images ofmitoDendra (green) and FXN (red)immunofluorescence, and merged images withDAPI-stained nuclei (blue), showing a reduction inthe overall levels of FXN and mitoDendra in thecerebellar cortex of mitoDendra-KIKO mice(D-F,D′-F′) compared with control mice (A-C,A′-C′)at 3 months of age (P90). Scale bars as indicated.

Fig. 5. VGLUT1-specific PF synaptic inputs are impaired in the cerebellar ML of asymptomatic KIKOmice carrying amitoDendramarker. (A-F) Confocalimages of VGLUT1 (red) and mitoDendra (green) fluorescence, and merged images with DAPI-stained nuclei (blue), showing specific reduction of VGLUT1immunoreactivity in the cerebellar ML, but not the GL, of P90 mitoDendra-KIKO mice (D-F) compared with age-matched control mice (A-C). (A′-F′) Highermagnification confocal images showing marked reduction of VGLUT1-positive presynaptic terminals in the cerebellar ML of P90 mitoDendra-KIKO mice(D′-F′) compared with control mice (A′-C′). (G,H) Confocal images of VGLUT1-positive presynaptic terminals (red immunofluorescence) in the cerebellar ML ofP90 mitoDendra-KIKO mice (H) compared with control mice (G). (I) Quantification of VGLUT1-positive puncta showing significant reduction of VGLUT1-positivepresynaptic terminals (n=6 sections for two KIKO mice and n=9 sections for three controls). Data are mean±s.e.m. ***P<0.001, two-tailed, unpaired Student’st-test. Scale bars as indicated.

1533

RESEARCH ARTICLE Disease Models & Mechanisms (2017) 10, 1529-1538 doi:10.1242/dmm.030049

Disea

seModels&Mechan

isms

cerebellar ML at asymptomatic (P150) and potentially symptomatic(P360) ages. In wild-type control mice, the numerous mitochondriaare fully packed in the small dendritic branches and spiny branchletsof Purkinje neurons (Fig. 6A,C; Fig. S2A,C) (d, dendritic branchesand branchlets), suggesting important physiological roles ofmitochondria in Purkinje neuron dendritic function. The abundantPF presynaptic terminals contain round-shaped mitochondria andform synapses on the thorns of Purkinje neuron dendritic branches(Fig. 6A,C,E; Fig. S2A,C) (pf, parallel fiber; t, thorns). In P360KIKO mice, we observed an overall impaired ultrastructure with anenlarged empty space in the cerebellar ML (Fig. 6D; Fig. S2D)compared with age-matched controls (Fig. 6C; Fig. S2C). The smalldendritic branches and spiny branchlets of Purkinje neurons havefewer and smaller mitochondria with disrupted cristae (Fig. 6D;Fig. S2D) compared with those of controls (Fig. 6C; Fig. S2C).Furthermore, the number of PF synapses (pf ) in the cerebellarmolecular layer is markedly reduced in KIKO mice (Fig. 6D,F;Fig. S2D) compared with controls (Fig. 6C,E; Fig. S2C),demonstrating severe PF synaptic deficits in older KIKO mice.Interestingly, the overall impairment of ultrastructure in thecerebellar ML is also observed in P150 KIKO mice (Fig. 6B;Fig. S2B) compared with age-matched controls (Fig. 6A; Fig. S2A),but to a lesser extent compared with P360 KIKO mice (Fig. 6D;Fig. S2D). Similar reduction of PF synapses occurs in the cerebellarML compared with controls (Fig. 6A,B; Fig. S2A,B). Theimpairment of mitochondria is also observed in the smalldendritic branches and spiny branchlets of Purkinje neurons inP150 KIKO mice compared with controls (Fig. S2A,B). Thissuggests that PF synaptic deficits might contribute to cerebellardysfunction and possibly progressive ataxia in this FRDA model.

Parvalbumin levels are specifically reduced in cerebellarPurkinje neurons of asymptomatic KIKO miceParvalbumin (PV) levels frequently reflect neuronal excitatoryactivity, and PV is highly expressed in cerebellar Purkinje neurons(Patz et al., 2004; Vig et al., 1996). We examined PV levels incerebellar homogenates of KIKO mice compared with control miceat P30, P90, P180 and P270. In wild-type control mice, PV levels incerebellar homogenates remain unaltered or slightly decreased at

P180 and P270 compared with P30 and P90 (Fig. S1D). In KIKOmice, PV levels are significantly reduced in cerebellar homogenatesfrom P30 to P270 (39-69% reduction, P<0.05 and P<0.01) (Fig. 7A,B), suggesting PV dysregulation in KIKO cerebellum. In order toidentify where changes in PV levels occur, we examined PVimmunoreactivity by immunohistochemistry in the cerebellum ofmitoDendra-KIKO mice compared with age-matched controls atP90. In the cerebellar cortex of control mice, similar to wild-typeC57BL/6 mice (Fig. 1), PV is highly expressed in Purkinje neuronsand other interneurons in the ML (Fig. 7C,C′). Remarkably, PVlevels are specifically reduced in Purkinje neurons, but preserved inother interneurons of the ML (Fig. 7C′,E′,F′,H′, arrows), inasymptomatic KIKO mice (Fig. 7F-H,F′-H′) compared with age-matched control mice (Fig. 7C-E,C′-E′), demonstrating specific PVdeficits in the Purkinje neurons of asymptomatic KIKO mice. Thenumber of Purkinje neurons appears to be unaltered, consistent withPurkinje cell dysfunction rather than neuronal loss in this FRDAmouse model (Kemp et al., 2016). This suggests that PV deficits inPurkinje neurons are downstream from synaptic abnormalities inKIKO cerebellum as PV levels frequently reflect excitatory neuronalactivity, implicating early Purkinje cell dysfunction, instead ofneuronal loss, in the evolving phenotype of this FRDAmouse model.

Moderate loss of principal neurons in cerebellar DN of KIKOmice at asymptomatic ageCerebellar Purkinje neurons project to the large principal neurons inthe DN. We observed reduction of PV-positive synaptic terminalssurrounding the large principal neurons in the DN of mitoDendra-KIKO mice (Fig. 8D,F, arrows) compared with control mice(Fig. 8A,C, arrows), consistent with PV reduction in the Purkinjeneurons of KIKO mice (Fig. 7). Interestingly, the number of largeprincipal neurons appears to be reduced in the DN of mitoDendra-KIKO mice at P90 (Fig. 8D-F) compared with controls (Fig. 8A-C).We further examined frataxin immunoreactivity in the DN ofcerebellum sections. The overall levels of frataxin are decreased inthe DN of KIKO mice. In addition, the number of frataxin-positivelarge principal neurons is moderately but significantly reduced inthe DN of mitoDendra-KIKO mice (Fig. 8J-L) compared withcontrol mice at P90 (Fig. 8G-I,M, P<0.05), matching the atrophy of

Fig. 6. Impaired ultrastructures of cerebellarMLand PF synaptic deficits on Purkinje neurons inasymptomatic and potentially symptomaticKIKO mice. (A-F) Electron microscopy images ofcerebellar ML in asymptomatic (P150) andpotentially symptomatic (P360) KIKO mice (B,D,F)compared with controls (A,C,E). KIKO mice showimpaired ultrastructures of cerebellar ML withenlarged empty space, fewer and smallermitochondria in the dendritic branches (d) ofPurkinje neurons, and decreased PF presynapticterminals (pf ) on the thorns (t) of Purkinje neurondendritic branches in P150 (B) and P360 (D) KIKOmice compared with age-matched control mice(A,C). A high abundance of PF synaptic terminalscontain round-shaped mitochondria and formsynapses on the thorns of Purkinje neurondendritic branches in the cerebellar ML of controlmice (E) and are markedly reduced in P360 KIKOmice (F). Scale bars as indicated.

1534

RESEARCH ARTICLE Disease Models & Mechanisms (2017) 10, 1529-1538 doi:10.1242/dmm.030049

Disea

seModels&Mechan

isms

large principal neurons in the DN observed in FRDA patients(Koeppen et al., 2011a, 2015). This has not been identified before inthis FRDA model, especially at asymptomatic ages, showing therelevance of this model to study neuropathological changes inFRDA.

DISCUSSIONThe present study uses a frataxin-deficient FRDA mouse model(KIKO mouse) to show early VGLUT1-specific PF synaptic inputdeficits and dysregulated cerebellar circuit at asymptomatic ages.The levels of VGLUT1 and PV are significantly decreased, whereasVGLUT2 and GAD65 levels are significantly increased, incerebellar homogenates of asymptomatic KIKO mice, suggestingearly VGLUT1-specific synaptic deficits and dysregulatedVGLUT2 and GAD65 synaptic inputs in KIKO cerebellum.Significant reduction of VGLUT1-positive presynaptic terminalsand PF synaptic deficits in the cerebellar ML further demonstrateearly VGLUT1-specific PF synaptic deficits in asymptomatic andsymptomatic KIKO mice. PV levels are specifically reduced inPurkinje neurons, but preserved in other interneurons of the ML,suggesting specific PV dysregulation in Purkinje neurons.Consistent with DN atrophy in FRDA patients, a moderate loss oflarge principal neurons is also observed in the DN of asymptomaticKIKOmice (Koeppen et al., 2011a). Our findings thus identify earlyVGLUT1-specific PF synaptic deficits and dysregulated cerebellarcircuit as potential mediators of cerebellar dysfunction, whichreflects developmental features of FRDA in this mouse model.

Early synaptic dysfunction and degeneration have beenimplicated in the pathogenesis of other neurodegenerativediseases including Alzheimer’s, Huntington’s and Parkinson’sdiseases (Cai and Tammineni, 2016; Calkins et al., 2011; Onyangoet al., 2017; Shirendeb et al., 2012). Glutamatergic and GABAergicsynaptic degeneration are found in the cerebellar DN of FRDApostmortem brain samples (Koeppen et al., 2011a, 2015). Ourfindings demonstrate significant reduction of VGLUT1 levels inKIKO cerebellum at asymptomatic and symptomatic ages.Conversely, VGLUT2 levels are increased in KIKO cerebellum.VGLUT1 is a selective marker for PFs that arise from granule cellsand form numerous synapses onto dendrites of Purkinje neurons inthe ML (Fremeau et al., 2001). VGLUT2 is a selective marker forclimbing fiber (CF) synaptic terminals that project onto Purkinjeneurons and undergo supernumerary elimination from multiple- tomono-innervation during development (Fremeau et al., 2001;Hashimoto and Kano, 2013). Proper formation of PF synapses isnecessary for eliminating CF innervation on Purkinje cells duringdevelopment (Hashimoto et al., 2009; Vanni et al., 2015; Watanabeand Kano, 2011). Our findings thus imply that reduction ofVGLUT1-containing PFs might lead to abnormal CF synapseelimination and increases in VGLUT2 levels in KIKO cerebellum atearly ages. Furthermore, our findings show specific reduction ofVGLUT1-positive presynaptic terminals and PF synapses in thecerebellar ML of KIKO mice at asymptomatic and potentiallysymptomatic ages, demonstrating PF synaptic deficits in KIKOcerebellar cortex at asymptomatic ages. This suggests that early

Fig. 7. PV levels are specifically reduced inPurkinje neurons, but preserved in otherinterneurons of the ML, in asymptomatic KIKOmice. (A,B) Western blot and quantificationanalysis of cerebellar homogenates (30 μg perlane) showing PV levels, as well as actin orGAPDH as an internal control, in the cerebellum ofKIKO mice and age-matched control mice at P30,P90, P180 and P270 (n=3-8 for KIKO and controlmice per time point). Data are mean±s.e.m.*P<0.05, **P<0.01, two-tailed, unpaired Student’st-test. Blots in Fig. 7 were stripped and reprobedwith multiple antibodies in Fig. 3 and in figures inLin et al. (2017); anti-actin and anti-GAPDH servedas the loading control. (C-H) Confocal images ofPV (red) and mitoDendra (green) fluorescence,and merged images with DAPI-stained nuclei(blue), showing marked reduction of PV andmitoDendra in the cerebellar cortex of P90mitoDendra-KIKO mice (F-H) compared with age-matched control mice (C-E). (C′-H′) Highermagnification confocal images showing markedreduction of PV in Purkinje neurons, but not in otherinterneurons of the ML (arrows), of P90mitoDendra-KIKO mice (F′-H′) compared withcontrol mice (C′-E′). Scale bars as indicated.

1535

RESEARCH ARTICLE Disease Models & Mechanisms (2017) 10, 1529-1538 doi:10.1242/dmm.030049

Disea

seModels&Mechan

isms

VGLUT1-specific PF synaptic deficits might contribute toabnormal cerebellar circuit development in this FRDA model.Interestingly, our findings not only show imbalanced VGLUT1- andVGLUT2-containing glutamatergic synaptic inputs, but also alteredGABAergic synaptic inputs with increased GAD65 levels in KIKOcerebellar homogenates, whereas GAD67 levels remain unaltered.GAD65 is primarily located at GABAergic synaptic terminals andserves as a marker for other interneurons, while GAD67 is primarilylocated in the soma of GABAergic interneurons and most abundantin Purkinje neurons. Heterologous competition among GABAergicand glutamatergic inputs is important for normal cerebellar circuitdevelopment, and GABAergic transmission onto Purkinje cells is acrucial factor for CF synapse elimination (Hashimoto et al., 2009;Nakayama et al., 2012). Our findings thus suggest that increasedGAD65 levels could result from early VGLUT1-specfic PF synapticdeficits, which also influence CF synapse elimination duringdevelopment, leading to an increase of VGLUT2 levels. Ourfindings thus strongly suggest that abnormal cerebellar synaptic andcircuit development in KIKO cerebellum might be potentialmediators of cerebellar dysfunction and ataxia in this FRDA model.Purkinje cells are the most distinctive neurons in the brain as they

receive more synaptic inputs than any other type of neuron in thebrain. Estimates of the number of spines on a single human Purkinjeneuron run as high as 200,000; therefore, synaptic inputs onPurkinje neurons are crucial for cerebellar neuronal functions.Imbalanced glutamatergic and GABAergic synaptic inputs onPurkinje neurons might lead to Purkinje neuronal dysfunction inthis FRDA model. Our findings identify early and specific PVreduction in Purkinje neurons at asymptomatic ages. PV is highlyexpressed in Purkinje neurons and its levels depend on neuronal

activity (Patz et al., 2004). PV reduction thus reflects reducedneuronal activity in Purkinje neurons resulting from imbalancedexcitatory and inhibitory synaptic inputs in KIKO cerebellum.Moreover, our findings identify moderate loss of large principalneurons in the DN of KIKO cerebellum at asymptomatic ages,consistent with that of FRDA patients (Koeppen et al., 2011a).Taken together, our findings thus suggest that a dysregulatedcerebellar circuit including imbalanced glutamatergic andGABAergic synaptic inputs, PV dysregulation in Purkinjeneurons and moderate loss of large principal neurons couldunderlie cerebellar dysfunction and ataxia in this FRDA model,providing novel pathogenic mechanisms and potential therapeutictargets for treatment of FRDA patients.

MATERIALS AND METHODSMaterialsC57BL/6 mice were from Charles River Laboratories and frataxin KIKOmice were from the Jackson Laboratory (B6. Cg-Fxntm1.1Pand Fxntm1Mkn/J;stock number 012329). KIKO mice were twice crossbred with Thy1-mitoDendra mice [B6SJL-Tg (Thy1-COX8A/Dendra)57Gmnf/J; stocknumber 025401] to generate control-mitoDendra and KIKO-mitoDendramice (Magrané et al., 2014). All animals were treated according to theprotocols approved by The Children’s Hospital of Philadelphia InstitutionalAnimal Care and Use Committee and Weill Cornell Medical CollegeInstitutional Animal Care and Use Committee. Antibodies included anti-GAD65 [Developmental Studies Hybridoma Bank, GAD-6, mousemonoclonal, 1:50 for immunohistochemistry (IHC)], anti-GAD65/67[Millipore, AB1511, 1:1000 for western blotting (WB)], anti-VGLUT1(Synaptic Systems, 135302, 1:1000 for IHC and 1:2000 for WB; SynapticSystems, 135311, 1:500 for IHC), anti-VGLUT2 (Synaptic Systems,135402, 1:1000 for WB), anti-frataxin (Abcam, ab175402, 1:250 for IHC

Fig. 8. Moderate loss of large principal neurons in the DN of asymptomatic KIKO mice. (A-F) Confocal images of PV (red) and mitoDendra (green)fluorescence, andmerged images with DAPI-stained nuclei (blue), showing reduction of PV-positive synapses and the surrounding principal neurons in the DN ofP90 mitoDendra-KIKO mice (D-F) compared with age-matched controls (A-C). (G-L) Confocal images of FXN (red) and mitoDendra (green) fluorescence, andmerged images with DAPI-stained nuclei (blue), showing reduction of large principal neurons in the DN of P90 mitoDendra-KIKO mice (J-L) compared withcontrols (G-I). (M) Quantification analysis showing a moderate, but significant, decrease in the number of FXN-positive large principal neurons in the DN ofmitoDendra-KIKOmice (J) compared with controls (G) (n=9 sections for three KIKOmice and three controls). Data aremean±s.e.m. *P<0.05, two-tailed, unpairedStudent’s t-test. Scale bars as indicated.

1536

RESEARCH ARTICLE Disease Models & Mechanisms (2017) 10, 1529-1538 doi:10.1242/dmm.030049

Disea

seModels&Mechan

isms

and 1:1000 for WB), anti-PV (Millipore, AB1572, 1:400 for IHC and1:1000 for WB; Swant, PV27, rabbit polyclonal, 1:1000 for IHC and WB),anti-ATP5A (MitoSciences, MS507, 1:500 for IHC), anti-GAPDH (NovusBiologicals, 1D4, NB300-221, 1:1000 for WB) and anti-actin [Abcam,ACTN5(04), ab3280, 1:5000 for WB] (Lin et al., 2014a,b, 2010).

Tissue preparation and immunohistochemistryFor tissue homogenate preparation, cerebella of KIKO mice and age-matched heterogeneous controls [both wild-type/wild-type (WTWT) andknock-in/wild-type (KIWT) mice] at 1, 3, 6, and 9 months, which representapproximately postnatal days P30, P90, P180 and P270±10 days, of eithersex were harvested. KIWT mice are the equivalent to human heterozygouscarriers. The tissues were mechanically homogenized in 20 ml lysis bufferper 1 g weight, and lysed for 1 h at 4°C. Lysis buffer contained 150 mMNaCl, 1 mM EDTA, 100 mM Tris-HCl, 1% Triton X-100, and 1% sodiumdeoxycholate, pH 7.4, supplemented the day of use with 1:500 EDTA-freeprotease inhibitor cocktail III (Calbiochem, 53914). Debris was cleared bycentrifugation at 39,000 g for 1 h at 4°C. Supernatants were stored at −80°Cuntil use.

For immunohistochemical studies, KIKO-mitoDendra mice and age-matched KOWT-mitoDendra controls of either sex at P90 were perfusedwith 4% paraformaldehyde and their cerebella were harvested. KOWT micewere used as controls and are equivalent to the hemizygous carriers. A seriesof brain coronal floating sections (50 µm thick) was obtained using avibratome (VT1200S; Leica, Deerfield, IL) in PBS; the sections were storedin PBS with 30% glycerol (vol/vol) and 30% ethylene glycerol (vol/vol) at−20°C. Floating sections were blocked with 5% normal goat serum and 1%bovine serum albumin in combination with 0.3% (vol/vol) Triton X-100 inPBS at room temperature for 1 h, then incubated with primary antibodies at4°C overnight and then secondary antibodies conjugated to Alexa Fluor 488(Invitrogen, A11029) or Alexa Fluor 568 (Invitrogen, A11036) at roomtemperature for 60-90 min. Following several washes with PBS, the stainedsections were mounted on slides with Vectashield with DAPI (VectorLaboratories, H-1200).

Transmission electron microscopyFor the ultrastructural studies, KIKOmice and age-matched KIWT controls ofeither sex at P150 and P360 (n=2 per group) were perfused with 2.5%glutaraldehyde and 2% paraformaldehyde and their cerebella were harvested.Tissues for electron microscopic examination were fixed with 2.5%glutaraldehyde, 2.0% paraformaldehyde in 0.1 M sodium cacodylate buffer,pH 7.4, overnight at 4°C. After subsequent buffer washes, the samples werepostfixed in 2.0% osmium tetroxide for 1 h at room temperature, and thenwashed again in buffer followed by distilledH2O. After dehydration through agraded ethanol series, the tissue was infiltrated and embedded in EMbed-812(Electron Microscopy Sciences, Fort Washington, PA). Thin sections werestained with uranyl acetate and lead citrate and examined with a JEOL 1010electron microscope (JEOL USA, Peabody, MA) fitted with a Hamamatsudigital camera and AMT Advantage image capture software. Sampleprocessing and staining were performed at the Electron MicroscopyResource Laboratory at University of Pennsylvania Perelman School ofMedicine Biomedical Research Core Facilities.

Western blot analysisWestern blotting was performed as described previously (Lin et al., 2014b).Protein content was determined using the BCA Protein Assay kit (ThermoFisher Scientific). Equal amounts of total protein (30 µg tissue homogenateper lane) were subjected to 4-12% NuPAGE Gel for electrophoresis andtransferred to nitrocellulose membranes. Membranes were blocked with 3%nonfat milk and incubated with primary antibody overnight at 4°C. Blotswere then incubated with appropriate horseradish peroxidase (HRP)-conjugated secondary antibodies (Cell Signaling Technology) for 2 h atroom temperature and then washed; reaction bands were visualized using aluminol-enhanced chemiluminescence (ECL) HRP substrate (ThermoFisher Scientific). Each blot was then incubated with stripping buffer (2%SDS, 50 mM Tris, pH 6.8, and 100 mM β-mercaptoethanol) for 45 min atroom temperature to remove the signals and reprobed for other proteins,

including actin or GAPDH as an internal control. Reaction product levelswere quantified by scanning densitometry using NIH ImageJ software(https://imagej.nih.gov/ij/) and normalized to the levels of actin andGAPDH.

ImmunohistochemistryFor immunohistochemistry, the floating sections were blocked with 5%normal goat serum and 1% bovine serum albumin in combination with 0.3%(vol/vol) Triton X-100 in PBS at room temperature for 1 h, then incubatedwith primary antibodies at 4°C overnight and then secondary antibodiesconjugated to Alexa Fluor 488 or 594 (Invitrogen) at room temperature for60-90 min. Following several washes with PBS, the stained sections weremounted on slides using Vectashield with DAPI.

Fluorescence imaging and quantificationFluorescence images were obtained with an Olympus FluoView or Leica SP8laser scanning confocal microscope. Confocal scans were performed inmouse cerebellum, and the imaging parameters were identical for KIKOmiceand knockout/wild-type (KOWT) controls. KOWTmice are the equivalent ofhemizygous carriers. Control sections were included in all experiments tonormalize for expected variations in antibody staining intensity performed ondifferent days. The confocal images were acquired at the focal plane with amaximal number of VGLUT1-positive puncta or large principal neurons fromthree sections per animal, and from two to three animals per group. ImageJsoftware was used to quantify the number of VGLUT1-positive puncta in thecerebellar cortex or large principal neurons in the DN in the acquired confocalimages. Thresholds were set at three standard deviations above the meanstaining intensity of six nearby regions in the same visual field. Thresholdedimages present a fixed intensity for all pixels above threshold after havingremoved all of those below, and VGLUT1-positive puncta in the thresholdedimages were quantified (Lin et al., 2014b; Lozada et al., 2012).

Statistical analysisData are shown as mean±s.e.m. Experimental results were analyzed usingtwo-tailed, unpaired Student’s t-test to compare two conditions. Significancewas set at P<0.05.

AcknowledgementsWe thank Dr Hajime Takano, Dr Guoxiang Xiong and Biao Zuo for assistance withconfocal imaging, immunohistochemistry and electron microscopy, respectively.

Competing interestsThe authors declare no competing or financial interests.

Author contributionsConceptualization: H.L., J.M., D.R.L.; Methodology: H.L., J.M., E.M.C., S.M.H.,N.W., A.R., D.R.L.; Validation: H.L., J.M., E.M.C.; Formal analysis: H.L.;Investigation: H.L., J.M., E.M.C., S.M.H., N.W.; Resources: H.L., J.M., D.R.L.;Writing - original draft: H.L., D.R.L.; Writing - review & editing: H.L., J.M., E.M.C.,D.R.L.; Supervision: H.L., J.M., D.R.L.; Project administration: H.L., D.R.L.; Fundingacquisition: H.L., J.M., D.R.L.

FundingThis work was supported by Friedreich’s Ataxia Research Alliance (Center ofExcellence Grant to D.R.L. and New Investigator Grant to J.M.) and Children’sHospital of Philadelphia (Foerderer Grant for Excellence to H.L.).

Supplementary informationSupplementary information available online athttp://dmm.biologists.org/lookup/doi/10.1242/dmm.030049.supplemental

ReferencesBidichandani, S. I., Ashizawa, T. and Patel, P. I. (1998). The GAA triplet-repeat

expansion in Friedreich ataxia interferes with transcription and may be associated

with an unusual DNA structure. Am. J. Hum. Genet. 62, 111-121.Cai, Q. and Tammineni, P. (2016). Mitochondrial aspects of synaptic dysfunction in

Alzheimer’s disease. J. Alzheimers Dis. 57, 1087-1103.Calkins, M. J., Manczak, M., Mao, P., Shirendeb, U. and Reddy, P. H. (2011).

Impaired mitochondrial biogenesis, defective axonal transport of mitochondria,

1537

RESEARCH ARTICLE Disease Models & Mechanisms (2017) 10, 1529-1538 doi:10.1242/dmm.030049

Disea

seModels&Mechan

isms

abnormal mitochondrial dynamics and synaptic degeneration in amousemodel ofAlzheimer’s disease. Hum. Mol. Genet. 20, 4515-4529.

Campuzano, V., Montermini, L., Molto, M. D., Pianese, L., Cossee, M.,Cavalcanti, F., Monros, E., Rodius, F., Duclos, F., Monticelli, A. et al. (1996).Friedreich’s ataxia: autosomal recessive disease caused by an intronic GAAtriplet repeat expansion. Science 271, 1423-1427.

Chiang, S., Kovacevic, Z., Sahni, S., Lane, D. J. R., Merlot, A. M., Kalinowski,D. S., Huang, M. L.-H. and Richardson, D. R. (2016). Frataxin and the molecularmechanism of mitochondrial iron-loading in Friedreich’s ataxia. Clin. Sci. (Lond.)130, 853-870.

Chutake, Y. K., Costello, W. N., Lam, C. and Bidichandani, S. I. (2014). Alterednucleosome positioning at the transcription start site and deficient transcriptionalinitiation in Friedreich ataxia. J. Biol. Chem. 289, 15194-15202.

Cossee, M., Puccio, H., Gansmuller, A., Koutnikova, H., Dierich, A., LeMeur, M.,Fischbeck, K., Dolle, P. and Koenig, M. (2000). Inactivation of the Friedreichataxia mouse gene leads to early embryonic lethality without iron accumulation.Hum. Mol. Genet. 9, 1219-1226.

Fremeau, R. T., Jr, Troyer, M. D., Pahner, I., Nygaard, G. O., Tran, C. H., Reimer,R. J., Bellocchio, E. E., Fortin, D., Storm-Mathisen, J. and Edwards, R. H.(2001). The expression of vesicular glutamate transporters defines two classes ofexcitatory synapse. Neuron 31, 247-260.

Grabczyk, E. and Usdin, K. (2000). The GAA*TTC triplet repeat expanded inFriedreich’s ataxia impedes transcription elongation by T7 RNA polymerase in alength and supercoil dependent manner. Nucleic Acids Res. 28, 2815-2822.

Hashimoto, K. and Kano, M. (2013). Synapse elimination in the developingcerebellum. Cell. Mol. Life Sci. 70, 4667-4680.

Hashimoto, K., Yoshida, T., Sakimura, K., Mishina, M., Watanabe, M. and Kano,M. (2009). Influence of parallel fiber-Purkinje cell synapse formation on postnataldevelopment of climbing fiber-Purkinje cell synapses in the cerebellum.Neuroscience 162, 601-611.

Kemp, K. C., Cook, A. J., Redondo, J., Kurian, K. M., Scolding, N. J. andWilkins,A. (2016). Purkinje cell injury, structural plasticity and fusion in patients withFriedreich’s ataxia. Acta Neuropathol. Commun. 4, 53.

Koeppen, A. H., Morral, J. A., Davis, A. N., Qian, J., Petrocine, S. V., Knutson,M. D., Gibson,W.M., Cusack,M. J. and Li, D. (2009). The dorsal root ganglion inFriedreich’s ataxia. Acta Neuropathol. 118, 763-776.

Koeppen, A. H., Davis, A. N. and Morral, J. A. (2011a). The cerebellar componentof Friedreich’s ataxia. Acta Neuropathol. 122, 323-330.

Koeppen, A. H., Morral, J. A., McComb, R. D. and Feustel, P. J. (2011b). Theneuropathology of late-onset Friedreich’s ataxia. Cerebellum 10, 96-103.

Koeppen, A. H., Ramirez, R. L., Becker, A. B., Feustel, P. J. and Mazurkiewicz,J. E. (2015). Friedreich ataxia: failure of GABA-ergic and glycinergic synaptictransmission in the dentate nucleus. J. Neuropathol. Exp. Neurol. 74, 166-176.

Lazaropoulos, M., Dong, Y., Clark, E., Greeley, N. R., Seyer, L. A., Brigatti, K.W.,Christie, C., Perlman, S. L., Wilmot, G. R., Gomez, C. M. et al. (2015). Frataxinlevels in peripheral tissue in Friedreich ataxia. Ann. Clin. Transl. Neurol. 2,831-842.

Lin, H., Vicini, S., Hsu, F.-C., Doshi, S., Takano, H., Coulter, D. A. and Lynch,D. R. (2010). Axonal alpha7 nicotinic ACh receptors modulate presynaptic NMDAreceptor expression and structural plasticity of glutamatergic presynaptic boutons.Proc. Natl. Acad. Sci. USA 107, 16661-16666.

Lin, H., Hsu, F.-C., Baumann, B. H., Coulter, D. A., Anderson, S. A. and Lynch,D. R. (2014a). Cortical parvalbumin GABAergic deficits with alpha7 nicotinicacetylcholine receptor deletion: implications for schizophrenia. Mol. Cell.Neurosci. 61, 163-175.

Lin, H., Hsu, F.-C., Baumann, B. H., Coulter, D. A. and Lynch, D. R. (2014b).Cortical synaptic NMDA receptor deficits in alpha7 nicotinic acetylcholine receptorgene deletion models: implications for neuropsychiatric diseases. Neurobiol. Dis.63, 129-140.

Lin, H., Magrane, J., Rattelle, A. J., Stepanova, A., Galkin, A., Clark, E. M., Dong,Y.-N., Halawani, S. W. and Lynch, D. R. (2017). Early cerebellar deficits inmitochondrial biogenesis and respiratory chain complexes in the KIKO mousemodel of Friedreich ataxia. Dis. Model. Mech. 10, 1343-1352.

Lozada, A. F., Wang, X., Gounko, N. V., Massey, K. A., Duan, J., Liu, Z. and Berg,D. K. (2012). Glutamatergic synapse formation is promoted by alpha7-containingnicotinic acetylcholine receptors. J. Neurosci. 32, 7651-7661.

Lu, C. and Cortopassi, G. (2007). Frataxin knockdown causes loss of cytoplasmiciron-sulfur cluster functions, redox alterations and induction of heme transcripts.Arch. Biochem. Biophys. 457, 111-122.

Lynch, D. R. and Seyer, L. (2014). Friedreich ataxia: new findings, new challenges.Ann. Neurol. 76, 487-488.

Lynch, D. R., Deutsch, E. C., Wilson, R. B. and Tennekoon, G. (2012).Unanswered questions in Friedreich ataxia. J. Child Neurol. 27, 1223-1229.

Magrane, J., Cortez, C., Gan, W.-B. and Manfredi, G. (2014). Abnormalmitochondrial transport and morphology are common pathologicaldenominators in SOD1 and TDP43 ALS mouse models. Hum. Mol. Genet. 23,1413-1424.

Martelli, A. and Puccio, H. (2014). Dysregulation of cellular iron metabolism inFriedreich ataxia: from primary iron-sulfur cluster deficit to mitochondrial ironaccumulation. Front. Pharmacol. 5, 130.

McMackin, M. Z., Henderson, C. K. and Cortopassi, G. A. (2016).Neurobehavioral deficits in the KIKO mouse model of Friedreich’s ataxia.Behav. Brain Res. 316, 183-188.

Miranda, C. J., Santos, M. M., Ohshima, K., Smith, J., Li, L., Bunting, M.,Cossee, M., Koenig, M., Sequeiros, J., Kaplan, J. et al. (2002). Frataxin knockinmouse. FEBS Lett. 512, 291-297.

Nakayama, H., Miyazaki, T., Kitamura, K., Hashimoto, K., Yanagawa, Y., Obata,K., Sakimura, K., Watanabe, M. and Kano, M. (2012). GABAergic inhibitionregulates developmental synapse elimination in the cerebellum. Neuron 74,384-396.

Onyango, I. G., Khan, S. M. and Bennett, J. P., Jr. (2017). Mitochondria in thepathophysiology of Alzheimer’s and Parkinson’s diseases. Front. Biosci. 22,854-872.

Pandolfo, M. (2009). Friedreich ataxia: the clinical picture. J. Neurol. 256 Suppl. 1,3-8.

Parent, A., Elduque, X., Cornu, D., Belot, L., Le Caer, J.-P., Grandas, A.,Toledano, M. B. and D’Autreaux, B. (2015). Mammalian frataxin directlyenhances sulfur transfer of NFS1 persulfide to both ISCU and free thiols. Nat.Commun. 6, 5686.

Patz, S., Grabert, J., Gorba, T., Wirth, M. J. and Wahle, P. (2004). Parvalbuminexpression in visual cortical interneurons depends on neuronal activity and TrkBligands during an Early period of postnatal development. Cereb. Cortex 14,342-351.

Rotig, A., de Lonlay, P., Chretien, D., Foury, F., Koenig, M., Sidi, D., Munnich, A.and Rustin, P. (1997). Aconitase and mitochondrial iron-sulphur proteindeficiency in Friedreich ataxia. Nat. Genet. 17, 215-217.

Sahdeo, S., Scott, B. D., McMackin, M. Z., Jasoliya, M., Brown, B., Wulff, H.,Perlman, S. L., Pook, M. A. and Cortopassi, G. A. (2014). Dyclonine rescuesfrataxin deficiency in animal models and buccal cells of patients with Friedreich’sataxia. Hum. Mol. Genet. 23, 6848-6862.

Shirendeb, U. P., Calkins, M. J., Manczak, M., Anekonda, V., Dufour, B.,McBride, J. L., Mao, P. and Reddy, P. H. (2012). Mutant huntingtin’s interactionwith mitochondrial protein Drp1 impairs mitochondrial biogenesis and causesdefective axonal transport and synaptic degeneration in Huntington’s disease.Hum. Mol. Genet. 21, 406-420.

Simon, D., Seznec, H., Gansmuller, A., Carelle, N., Weber, P., Metzger, D.,Rustin, P., Koenig, M. and Puccio, H. (2004). Friedreich ataxia mouse modelswith progressive cerebellar and sensory ataxia reveal autophagicneurodegeneration in dorsal root ganglia. J. Neurosci. 24, 1987-1995.

Vanni, V., Puglisi, F., Bonsi, P., Ponterio, G., Maltese, M., Pisani, A. andMandolesi, G. (2015). Cerebellar synaptogenesis is compromised in mousemodels of DYT1 dystonia. Exp. Neurol. 271, 457-467.

Vaubel, R. A. and Isaya, G. (2013). Iron-sulfur cluster synthesis, iron homeostasisand oxidative stress in Friedreich ataxia. Mol. Cell. Neurosci. 55, 50-61.

Vig, P. J. S., Fratkin, J. D., Desaiah, D., Currier, R. D. and Subramony, S. H.(1996). Decreased parvalbumin immunoreactivity in surviving Purkinje cells ofpatients with spinocerebellar ataxia-1. Neurology 47, 249-253.

Watanabe, M. and Kano, M. (2011). Climbing fiber synapse elimination incerebellar Purkinje cells. Eur. J. Neurosci. 34, 1697-1710.

1538

RESEARCH ARTICLE Disease Models & Mechanisms (2017) 10, 1529-1538 doi:10.1242/dmm.030049

Disea

seModels&Mechan

isms