Friedreich Ataxia: Failure of GABA-ergic and Glycinergic Synaptic Transmission in the Dentate Nucleus

Abstract

Atrophy of large neurons in the dentate nucleus (DN) is an important pathological correlate of neurological disability in patients with Friedreich ataxia (FA). Thinning of the DN was quantified in 29 autopsy cases of FA and 2 carriers by measuring the thickness of the gray matter ribbon on stains with anti-glutamic acid decarboxylase (GAD), the rate-limiting enzyme in the biosynthesis of γ-amino-butyric acid (GABA). The DN was thinner than normal in all cases of FA, and atrophy correlated inversely with disease duration but not with age of onset or length of the homozygous guanine-adenine-adenine trinucleotide expansions. In 13 of the FA cases, frozen DN tissue was available for assay of frataxin. DN atrophy was more severe when frataxin was very low. Immunohistochemical staining for GAD revealed grumose reaction and preservation of small GABA-ergic neurons in the DN of FA patients. Residual small DN neurons and varicose axons also contained the glycine transporter 2, identifying them as glycinergic. Immunohistochemistry also confirmed severe loss of GABA-A and glycine receptors in the DN with comparable depletion of the receptor-anchoring protein gephyrin. Thus, loss of gephyrin and failure to position GABA-A and glycine receptors correctly may reduce trophic support of large DN neurons and contribute to their atrophy. By contrast, Purkinje cells may escape retrograde atrophy in FA by issuing new axonal sprouts to small surviving DN neurons where they form reparative grumose clusters.

INTRODUCTION

Friedreich ataxia (FA) results from the deficiency of a small mitochondrial protein, frataxin. The most common underlying mutation is a recessively inherited pathogenic homozygous expansion of guanine-adenine-adenine (GAA) trinucleotide repeat expansion in intron 1 of the frataxin gene (chromosome 9q21). Compound heterozygotes, such as those with a GAA trinucleotide repeat expansion on 1 allele and a point mutation or deletion on the other, are much rarer; however, such patients often have a similar neurological phenotype. Total deletion of the frataxin gene in a mouse model of FA and consequent complete absence of frataxin are lethal to the embryo (1). Human patients with FA have abnormally low levels of tissue frataxin but survive an average of 26 years after FA becomes clinically evident (2). "Onset" of the disease has been difficult to define because the neurological deficit remains hidden for years. After a child or young adult becomes overtly ataxic, parents or other family members may recall that the patient did not develop motor skills at the same pace as his or her siblings who remained well. Several hypotheses have been presented to explain delayed manifestations and progression to a state of profound disability. Declining frataxin levels due to somatic GAA trinucleotide repeat expansions is an attractive explanation (3–5), and mismatch repair enzymes may be involved (4–5). Beyond impairment of transcription by the formation of triplex deoxyribonucleic acid (6), frataxin gene regulation is also subject to epigenetic factors (7). The dentate nucleus (DN) is highly vulnerable to FA (8), and the profound ataxia in FA should also be attributed to atrophy of large glutamatergic DN neurons. Clinical disease progression can be measured by several available rating scales (9), and the paramagnetic effect of DN iron on magnetic resonance images is emerging as a biomarker of the disease (10, 11).

In contrast to clinical and imaging measures, however, the literature contains no information on the severity of the DN lesion as a correlate of age of onset, disease duration, length of GAA trinucleotide repeats, or frataxin levels. The DN receives the bulk of its afferent fibers from Purkinje cells, and the principal neurohumoral transmitter is γ-aminobutyric acid (GABA). Though the DN loses its large glutamatergic neurons, small GABA-ergic neurons survive (8). The DN also contains small glycinergic neurons, but the role of glycine as an inhibitory transmitter in the human cerebellar circuitry is not well characterized.

The work presented here advances our hypothesis that large DN neurons fail to express GABA-ergic and glycinergic receptors, or the receptor-anchoring protein gephyrin, and thereby lose trophic support from Purkinje cells in the cerebellar cortex and intrinsic glycinergic neurons in the DN. Selective atrophy of large DN neurons has major clinical implications because they are the main source of efferent cerebellar fibers. It is widely recognized that the cerebellar cortex in FA only rarely displays abnormalities (12). Survival of Purkinje cells, however, is functionally irrelevant because corticonuclear connections are interrupted by the loss of large DN neurons. In mice, Purkinje cells show little if any morphological evidence of retrograde degeneration after axonal interruption (13). The putative reason is recovery of trophic support by axonal sprouting from the divided main axon. Survival of Purkinje cells in FA may be analogous, but no prior information exists on how these neurons recover trophic support after most of the large DN neurons have disappeared.

MATERIALS AND METHODS

Specimens

Autopsy specimens were obtained through a formal tissue donation program supported by Friedreich's Ataxia Research Alliance and National Ataxia Foundation, or by the courtesy of other physicians. The program includes the collection of personal health information of the deceased subject and family history. The Institutional Review Board of the Veterans Affairs Medical Center has approved this research. The Table lists basic clinical information on 29 patients with FA from whom paraffin-embedded DN samples were available for immunohistochemistry and immunofluorescence. In 13 of these cases, frozen DN tissue was available for the assay of frataxin. Confirmation of the mutation was available in 28 patients, either by information gained from the clinical record or by amplification of the GAA trinucleotide repeat expansion in deoxyribonucleic acid extracted postmortem from the cerebellar cortex. One patient, FA 26 (Table), declined testing during life, and all tissue samples had been fixed at the time of autopsy and amplification of DNA extracted from formalin-fixed tissues was inconclusive. The neuropathological and cardiac phenotypes, however, were diagnostic of FA.

Table.

Basic Clinical Data of 29 Patients with Friedreich Ataxi

No	Sex	Age of onset (y)	Age of death (y)	Disease duration (y)	GAA1	GAA2	Cause of death
FA1	M	2	10	8	1016	1016	Cardiomyopathy
FA2	M	3	23	20	1200	1200	Cardiomyopathy
FA3	M	4	37	33	674	674	Cardiomyopathy
FA4	F	6	25	19	800	1110	Cachexia
FA5	F	6	27	21	668	864	Endocarditis
FA6	M	7	34	27	1114	1114	Cardiomyopathy
FA7	F	7	55	48	644	794	Cachexia
FA8	M	7	35	28	750	1000	Cachexia
FA9	F	7	28	21	681	837	Cardiac arrhythmia
FA10	F	8	15	7	840	840	Cardiomyopathy
FA11	F	8	36	28	715	840	Cardiomyopathy
FA12	M	8	27	19	700	1070	Cardiomyopathy
FA13	M	9	40	31	519	647	Cardiomyopathy
FA14	F	9	26	17	690	850	Cardiomyopathy
FA15	M	9	33	24	925	925	Cardiomyopathy
FA16	M	10	38	28	249	934	Cardiomyopathy
FA17	F	10	24	14	700	950	Cardiomyopathy
FA18	F	10	24	14	740	910	Myoglobinuria; renal failure
FA19	M	10	24	14	700	1050	Cardiomyopathy
FA20	F	10	25	15	750	850	Cardiomyopathy
FA21	F	11	42	31	761	990	Cardiomyopathy
FA22	M	14	19	5	841	1153	Cardiomyopathy
FA23	F	15	69	54	568	568	Cardiomyopathy; brain embolism
FA24	F	17	50	33	512	1122	Cardiomyopathy
FA25	F	20	61	41	593	694	Cardiomyopathy
FA26	F	20	48	28	NA	NA	Cardiomyopathy
FA27	F	34	77	43	466	841	Cachexia; cardiomyopathy
FA28	F	36	58	22	566	566	Cardiomyopathy; brain hemorrhage
FA29	F	54	83	29	106	235	Cachexia
Mean ± SD	F,18;M, 11	12.8 ± 11.2	37.7 ± 18.3	24.9 ± 11.8	696 ± 225	880 ± 213
Range		2–54	10–83	5–54	106–1114	235–1200

Patients were sorted according to age of onset. Patients FA2 and FA4 were siblings. F, female; M, male; y, years; GAA1, GAA2, guanine-adenine-adenine trinucleotide repeat expansion at allele 1 or allele 2; NA, not available.

Fourteen normal autopsy samples containing DN (9 males, 5 females) were secured from the academic autopsy service of Albany Medical College and Veterans Affairs Medical Center, Albany, NY. In 8 of these control samples, the amount of frozen tissue was sufficient for biochemical assay of frataxin. The age of death of the control subjects was 64.4 ± 8.4 years (mean ± SD). The FA tissue repository also included the DN of 2 subjects who were FA carriers. Both were men, aged 79 and 86 years, and their GAA trinucleotide repeats were 863/33 and 560/7, respectively. Autopsy delay for the FA cases was 20 ± 19 hours (mean ± SD) with a range of 3 to 96 hours. Autopsy delay for the control cases was 15 ± 8 hours (mean ± SD) with a range of 3 to 26 hours.

Enzyme-Linked Immunosorbent Assay of Frataxin

Portions of the gray matter ribbon of the DN were dissected with a No. 11 blade. Weighed samples of wet tissue (48–137 mg) were dispersed by 3 to 5 5-second-long bursts of ultrasonication in a lysis buffer containing 100 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% each of the nonionic detergents Triton X-100 and Nonidet P-40, 5 mM ethylenediamine tetraacetic acid, 5 mM ethylene glycol tetraacetic acid, and 1% protease inhibitor cocktail (vol/vol; Sigma, St. Louis, MO) (14). The mixtures were chilled on ice between bursts. The homogenates were centrifuged for 2 hours at 14,000 × g and 4°C. The supernatant was collected, and aliquots were diluted 1:10 in phosphate buffered saline (PBS) to reduce the detergent concentrations to 0.1%. The diluted extracts were then filtered through a centrifugal filter device with a molecular weight cut-off of 30 kDa (EMD Millipore, Billerica, MA) at 14,000 × g for 45 minutes. The filtrate was collected, and enzyme-linked immunosorbent assay (ELISA) of frataxin proceeded as follows: Polystyrene ELISA plates (Santa Cruz Biotechnology, Santa Cruz, CA) were coated with monoclonal anti-frataxin (0.33 µg protein/ml; Abcam, Cambridge, MA) in 0.05 M carbonate buffer (pH 9.6) by an overnight incubation at 4°C. The plates were washed 3 times with a 1% solution of non-fat dry milk in PBS, containing 0.1% Tween 80 (NFDM-PBS-Tween 80). Well surfaces were then covered for 4 hours at room temperature with NFDM-PBS-Tween 80 to block non-specific absorption of antibodies. The next step was the application of diluted tissue lysate or recombinant human frataxin in NFDM-PBS-Tween 80. After an overnight incubation, the wells were drained and washed with NFDM-PBS-Tween 80. The detecting antibody was rabbit polyclonal anti-frataxin (whole serum) that was diluted 1:2000 in NFDM-PBS-Tween 80. A second overnight incubation at 4°C followed. After washing the plates 3 times with NFDM-PBS-Tween 80, the wells were filled with biotinylated anti-rabbit IgG (0.75 µg protein/ml) in NFDM-PBS-Tween 80 and maintained at room temperature for 2 hours. Unbound biotinylated antibody was removed by washing once with NFDM-PBS-Tween 80 and 3 times with plain PBS. The next step was a 1-hour incubation at room temperature in a solution of horseradish peroxidase-labeled streptavidin (0.25 µg/ml) in PBS. After 3 washes with PBS, a chromogenic solution of ortho-phenylenediamine (2 mM) and H₂O₂ (0.01%) in 0.1 M citric acid-sodium phosphate buffer (pH 5.0) was added to each well. A distinct color gradient developed within 2 to 3 minutes, and the addition of 2.5 M sulfuric acid (50 µl) stopped the reaction. Absorbance at 492 nm was determined using an ELISA plate reader (SpectraMax Plus, Molecular Devices, Sunnyvale, CA). The amount of frataxin in tissue lysates was determined by reference to a calibration standard curve, and results were expressed as ng/g original wet weight.

Antibodies for Immunohistochemistry and Immunofluorescence

The following antibodies to the listed proteins were obtained from commercial sources (abbreviation, host and type, and supplier in parentheses): glutamic acid decarboxylase (GAD, mouse monoclonal, MBL International, Woburn, MA); class-III-β-tubulin (mouse monoclonal, R & D Systems, Minneapolis, MN); GABA-A-receptor γ2-subunit (GABA-A-Rγ2, goat polyclonal, Santa Cruz Biotechnology); glycine receptor α1/2-subunits (GlyRα1/2, rabbit polyclonal, Thermo Fisher Scientific, Waltham, MA); gephyrin (mouse monoclonal, Santa Cruz Biotechnology); glycine transporter 2 (GlyT2, rabbit polyclonal, Santa Cruz Biotechnology); synaptophysin (rabbit polyclonal, Millipore, Temecula, CA); and non-phosphorylated neurofilament protein (mouse monoclonal, Covance, Emeryville, CA).

Antigen Retrieval and Immunohistochemistry

Antibody concentrations were optimized by trial and error; protein concentrations ranged from 0.15 µg/ml (for polyclonal anti-synaptophysin) to 2 µg/ml (for monoclonal anti-non-phosphorylated neurofilament protein and polyclonal anti-GlyRα1/2). Three antigen retrieval methods were used on 6-µm-thick rehydrated paraffin sections (antigens in parentheses): Incubation at 95°C for 30 minutes in a 1× solution (vol/vol) of DIVA, a proprietary antigen retrieval mixture (Biocare Medical, Concord, CA) (GAD and GlyRα1/2); incubation at 98.5°C for 20 minutes in 0.1 M Tris hydrochloride buffer (pH 9.5) containing 5% urea, followed by a 10-min cool-down period (GABA-A-Rγ2, GlyT2, and non-phosphorylated neurofilament protein); incubation in 0.01 M citric acid-sodium citrate buffer (pH 6) at 95°C for 10 minutes, followed by a 10-minute cool-down period (gephyrin, class-III-β-tubulin, and synaptophysin).

The sequence of incubations for the visualization of immunohistochemical reaction products was as previously described (15) and included amplification by biotinylated anti-rabbit, mouse, or goat IgG (0.6 µg protein/ml for DIVA-treated sections, 3 µg protein/ml for all others), horseradish peroxidase-labeled streptavidin (0.4 µg protein/ml for DIVA-treated sections, 2 µg protein/ml for all others), and diaminobenzidine/urea/hydrogen peroxide as the chromogen (Sigma).

Immunofluorescence

Double-label immunofluorescence was used to detect localization and colocalization of synaptophysin and class-III-β-tubulin; and of GABA-A-Rγ2 and GlyRα1/2. The initial steps included section rehydration and suppression of non-specific signals. Antigen retrieval and antibody dilution were identical to those of immunohistochemistry, but oxidation in H₂O₂-containing methanol was omitted. For double-label detection of class-III-β-tubulin and synaptophysin, the sequence of incubations was as follows (washing steps omitted): overnight incubation at 4°C in the first primary antibody (mouse monoclonal anti-class-III-β-tubulin), application of Alexa 488-labeled donkey anti-mouse IgG (2 µg protein/ml) for 4 hours at room temperature, re-suppression in 10% normal donkey serum, an additional overnight cold incubation in the second primary antibody (rabbit polyclonal anti-synaptophysin), followed by incubation in Cy3-labeled donkey anti-rabbit IgG (2 µg protein/ml) for 4 hours at room temperature. After washing, the sections were covered by a solution of 50% glycerol in PBS. The protocol for double-label immunofluorescence of GABA-A-Rγ2 and GlyRα1/2 was similar. Incubation with primary antibodies (goat IgG and rabbit IgG, respectively) was followed by Alexa 488- and Cy3-labeled secondary antibodies (to goat and rabbit IgG, respectively). The fluorescent secondary antibodies were purchased from Jackson ImmunoResearch (West Grove, PA). The sections were viewed in a laser scanning confocal microscope (Zeiss LSM 510 Meta). Exciting wavelengths were 488 and 543 nm for Alexa 488 and Cy3, respectively. The band pass filters were set at 500–530 nm for Alexa 488 and 565–615 nm for Cy3. Images were collected using a 63× 1.4 NA oil objective lens. Z-stacks were acquired using Nyquist sampling criteria and volume-rendered by the transparency render mode in VisArtplus basic mode of ZEN2009 Black software (Zeiss Microimaging, Thornwood, NY).

Measurement of DN Thickness

Three non-overlapping linear or curvilinear portions of each GAD-stained DN were photographed at a magnification of 100× with a Zeiss Axiophot microscope. The stored images were displayed and analyzed with the aid of a Zeiss AxioVision computer program (version 4.7.1) (Carl Zeiss, Gottingen, Germany). Measurements were made by drawing 10 straight lines at intervals across the thickness of each DN, as shown in Figure 1. Measuring 3 DN portions in this manner provided a total of 30 values for each patient or normal control. The computer program also generated data tables that were used to calculate mean DN thicknesses and SDs.

Figure 1.

Measurement of dentate nucleus (DN) thickness. Glutamic acid decarboxylase (GAD) stain. (A, B) Ten lines were drawn at right angles to the estimated central axis of the DN. The computer program determined the precise length of lines in µm and stored each measurement. Thinning of the DN and depletion of GAD-reactive terminals in Friedreich ataxia (FA) are evident (A). Analysis of all 3 segments of DN yielded a thickness of 150 ± 26 µm (mean ± SD) for FA13 (A) and 261 ± 22 µm for the normal control (B). Bars, 100 µm.

Data Analysis

The thickness of the DN was correlated with age of onset, disease duration, GAA trinucleotide repeat expansion in allele 1 (GAA1) and allele 2 (GAA2), and frataxin levels by standard regression analysis, using MiniTab v17.1 (MiniTab Inc., State College, PA) and OriginPro v.9.1 (OriginLab Corp., Northampton, MA).

RESULTS

Clinical Data

All patients listed in the Table for whom genetic testing was possible (n = 28) had homozygous GAA trinucleotide repeat expansions. Onset (age 2–54 years) and duration (5–54 years) ranged over a wide spectrum. As expected, disease onset correlated significantly with the length of the shorter GAA trinucleotide repeat (R² = 0.43 in linear regression y = 13.12× + 860.52, wherein y = GAA1 and x = age of onset in years; p < 0.001). The shortest GAA1 trinucleotide expansion was 106, conveying onset at age 54 (FA29, Table). During the life of this patient, examiners diagnosed dominant ataxia with incomplete penetrance rather than recessive ataxia and therefore did not request genetic testing for FA. The diagnosis was established on DNA extracted from frozen postmortem tissue. In 24 patients, heart disease contributed to or was the main cause of death (82.8%) (Table).

Frataxin Levels and Correlation with DN Atrophy

Frataxin levels in the DN of eight control cases were 195.0 ± 55.3 ng/g wet weight (mean ± SD). In 13 FA cases with sufficient frozen DN tissue, the frataxin level was 6.8 ± 3.1 ng/g. In the 2 carrier cases, the levels were 181.8 and 102.2 ng/g, respectively. Frataxin levels in FA correlated significantly with the thickness of the DN (R²=0.56 in linear regression: y = 117 + 6.3×, wherein y = DN thickness and x = frataxin in ng/g; p = 0.003) but not with age of onset, disease duration, GAA1, or GAA2. The highest frataxin levels were present in the DN of patients FA14, FA19, and FA20, measuring, in ng/g wet weight, 12.3, 10.0, and 12.5, respectively. The corresponding DN thicknesses were, in µm, 186, 184, and 186. FA patients FA2 and FA3 showed frataxin levels at or below the lower detection limit of 4 ng/g wet weight, corresponding to DN thicknesses of 143 and 111 µm.

Thickness of the DN and Correlation with Clinical Data

DN thicknesses also negatively correlated with longer disease duration (Fig. 2). DN thickness in 14 normal controls was 238 ± 21 µm (mean ± SD) with a range of 207 to 272 µm. In 2 FA patients with short disease duration (7 and 5 years in FA10 and FA22, respectively), DN thickness fell into the normal range. In an exceptional case of FA (patient FA16), DN thickness was normal at 213 µm despite disease duration of 28 years. DN thickness did not correlate significantly with age of onset (R² = 0.3; p = 0.8), GAA expansions on either the long (R² = 0; p = 0.99) or the short allele (R² = 0.03; p = 0.48). DN thicknesses were 142 µm and 202 µm in the 79- and 86-year-old FA carriers, respectively. Both values were below the range of the normal controls (207–272; Fig. 2).

Figure 2.

Correlation of dentate nucleus (DN) thickness and disease duration in Friedreich ataxia (FA). Normal controls are shown by open red triangles, FA carriers by blue circles, and FA cases by open black squares. Progressive thinning of the DN occurs with longer disease durations. Normal DN thickness is 238 ± 21 µm (mean ± SD). The values for both carriers are below the normal range. The linear regression line shown includes only the 29 FA patients and yields a slope of −1.29 at R² = 0.24, p < 0.01.

Neurons and Axons in the DN of FA

Reactivities of the neuropil of the DN in 1 normal control and 3 cases of FA with antibodies to non-phosphorylated neurofilament protein are shown in Figure 3A–D and to class-III-β-tubulin in Figure 3E–H. The main observations in FA are loss of large neurons and retention of small nerve cells. Small DN nerve cells are probably not completely exempt from destruction in FA. An example is shown in Figure 3C (patient FA9), in which only a single neuron stains for anti-class-III-β-tubulin. Despite extensive neuronal loss, thin axons persist to a large degree in the depleted DN (Fig. 3F–H). The better preservation of the DN in the FA patient with late onset (FA29) is apparent on both stains (Fig. 3D, H). The insets in Figure 3B–D and 3F–H of the FA cases display the integrity of Purkinje cells in the cerebellar cortex that lies immediately adjacent to the DN.

Figure 3.

Immunohistochemistry of non-phosphorylated neurofilament protein and class-III-β-tubulin in the cerebellum of a normal control case and 3 cases of Friedreich ataxia (FA). (A, E) In the normal dentate nucleus (DN), neuron sizes are heterogeneous. Small neurons are generally located at the junction of the DN gray matter and adjacent white matter (arrows). (B–D, F–H) In the FA cases, both immunostains show loss of large neurons in the DN while small neurons remain intact (arrows). Insets show the contrast between morphologically intact Purkinje cells and DN neurons in FA on the same tissue block. A–D, non-phosphorylated neurofilament protein; E–H, class-III-β-tubulin. B, F, patient FA20; C, G, patient FA9; D, H, patient FA29. Bars: 50 µm.

Grumose Reaction in the DN of FA

Figure 4 displays the status of synaptic terminals in the DN of 1 case of FA (patient FA27) and a normal control. Double-label fluorescence staining for synaptophysin and class-III-β-tubulin shows grumose reaction in FA (Fig. 4A) and numerous thin axons within and around the clusters of synaptic terminals. Neuronal cell bodies are no longer distinct (Fig. 4A). In Figure 4B, a surviving small neuron is also surrounded by numerous thin axons. Short segments of proximal dendrites show a thick layer of compact synaptic terminals, representing grumose reaction (Fig. 4B). The normal large DN neurons in Figure 4C show a delicate layer of synaptic terminals about their somata and dendrites. The normal DN lacks the thin hair-like axons that are characteristic of FA (Fig. 4A). A small neuron of the normal DN (Fig. 4D) displays intense fluorescence of axosomatic and axodendritic synapses but does not show the grumose clusters surrounding cells of comparable size in FA (Fig. 4B).

Figure 4.

(A–D) Double-label immunofluorescence of class-III-β-tubulin and synaptophysin in Friedreich ataxia (FA) (patient FA27) (A, B) and a normal control (C, D). Class-III-β-tubulin is shown in Alexa 488 (green), synaptophysin in Cy3 (red). The images were acquired by laser scanning confocal microscopy and resolved as maximum intensity projections. (A) A focus of synaptophysin-positive grumose reaction is surrounded and penetrated by a dense network of delicate axons. A neuronal cell body is not distinct in the center of the grumose cluster. (B) Two small neurons ("N") display an abnormally thick layer of synaptic terminals about their cell bodies and proximal dendrites (arrows). This appearance is thought to represent grumose "regeneration" rather than "degeneration". Many thin axons traverse the neuropil. Some are in the immediate vicinity of the synaptic cluster and may be the source of the crowded terminals. (C) Two large normal dentate nucleus (DN) neurons and many nearby dendrites are lined by thin layers of synaptic terminals. Axons in the neuropil are generally thicker than in FA (A, B) and less abundant. (D) A small neuron located in the white matter adjacent to the main DN shows a thin layer of synapses. Axons around the neuron are thicker than in a comparable location of the DN in FA (B). Bars: 20 µm.

GABA-ergic and Glycinergic Terminals in the DN of FA

Immunohistochemical results with antibodies to GAD and GlyT2 as indicators of GABA-ergic and glycinergic transmission, respectively, are shown in Figure 5. In 2 cases of FA (Fig. 5B, C; patients F20 and F9, respectively), the GAD stain shows disorganization and grumose reaction of the DN. The bulk of GAD reaction product occurs in synaptic terminals (Fig. 5A), and prompt fixation also allows visualization of this protein in the cytoplasm of small nerve cells (Fig. 5A, D, insets). Figure 5D shows the DN of the FA case with late-onset and disease duration of 29 years (FA29). The GAD-reactive neuropil is more abundant although negative images of large neurons are absent. The stain of GlyT2 generates punctate reaction product and delicate varicose axons in the normal DN (Fig. 5E). Small neurons with GlyT2-reactive somata are infrequent (Fig. 5E, arrow; Fig. 5E, inset). In FA, the GlyT2 stain suggests reduced density of reactive axons in 2 of 3 cases (Fig. 5G, H). Glycinergic terminals do not appear to contribute to grumose reaction in FA (Fig. 5F–H).

Figure 5.

(A–H) Immunohistochemistry of glutamic acid decarboxylase (GAD) (A–D) and glycine transporter 2 (GlyT2) (E–H) in the dentate nucleus (DN) in 1 normal control case (A, E) and 3 cases of Friedreich ataxia (FA) (B–D, F–H). GAD reaction product in afferent terminals in the normal DN (A) is very dense and generates negative images of large neurons ("N"). Small GAD-positive neurons are also present. The inset shows a small neuron with GAD-reactive cytoplasm and a layer of GAD-positive synaptic terminals. The cytoplasm of an adjacent large nerve cell ("N") is GAD-negative, but displays a thin layer of axosomatic terminals. In FA patients, GAD immunohistochemistry shows a disorganized DN and several regions of grumose reaction (B, C, arrows). The DN in patient FA29 (D) appears less disorganized though it is thinner than normal and devoid of negative images of neurons and grumose reaction. The inset in (D) shows 2 surviving small γ-amino-butyric acid (GABA)-ergic neurons. The GlyT2 reaction shows punctate product and scattered varicose axons in the normal DN (E). When present, cytoplasmic reaction product occurs only in small nerve cells (E, arrow, and inset). In (F) (patient FA20), punctate GlyT2 reaction product does not appear reduced though the DN is thinner; GlyT2-positive neurons are absent. Reduced GlyT2 reaction product is more convincing in (G) and (H), corresponding to patients FA9 and FA29, respectively. A–D, GAD; E–H, GlyT2. A, E, normal control; B, F, patient FA20; C, G, patient FA9; D, H, patient FA 29. Bars, 50 µm; insets, 20 µm.

GABA-A-Rγ2, GlyRα1/2, and Gephyrin in the DN of FA

Immunohistochemical reaction products of the GABA and glycine receptors, and of the inhibitory receptor-anchoring protein gephyrin are shown in Figure 6. Large and small neurons of the normal DN show GABA-A-Rγ2 and GlyRα1/2 immunoreactivity close to or within the plasma membrane of somata and dendrites. This localization is best seen on images obtained under higher magnification and oil immersion (Fig. 6A–H, insets). In contrast to the receptor proteins, gephyrin reaction product is present throughout the cytoplasm of soma and dendrites (Figs. 6I–L). In FA, the receptor stains show a decline in number and size of neurons, and a general disorganization of the DN neuropil. The loss of large DN neurons is very evident on the anti-gephyrin-stained sections (Figs. 6J–L). Only a few small neurons remain in the FA patients. Some nerve cells have lost their receptor or gephyrin immunoreactivity (Figs. 6B, F, J–L, insets, arrows). Figures 6D, H, and L confirm the relative preservation of the receptor proteins and gephyrin in the FA case with late onset (FA29), matching the GAD (Fig. 5D) and nerve cell stains (Fig. 3D, H).

Figure 6.

(A–L) Immunohistochemistry of γ-amino-butyric acid receptor (GABA)-A-Rγ2, glycine receptor α1/2-subunits (GlyRα1/2), and gephyrin in the dentate nucleus (DN) of 1 normal control case and 3 cases of Friedreich ataxia (FA). In the normal case, GABA-A-Rγ2 (A) and GlyRα1/2 (E), reaction products are most abundant near the plasma membrane of neuronal somata and proximal dendrites, which is more distinctly visible at higher power under oil immersion (insets). In FA cases (B–D, F–H), immunohistochemistry of the receptor proteins generates a peculiar plaque-like distribution of reaction product. Some intact nerve cells also display receptor reaction product (insets). Reaction product is absent from some small neurons (B, F, insets, arrows). Receptor reaction products are more abundant in (D) and (H), respectively, from patient FA29, whose disease began late and continued for 29 years. (I–L) Gephyrin reaction product is present in the cytoplasm of nerve cell bodies and dendrites of the control (I). Immunohistochemistry with anti-gephyrin shows severe loss of large neurons and retention of small nerve cells in FA patients (J–L). In (L) (patient FA29), the neuropil is better preserved than in the cases of early onset (J, patient FA20; K, patient FA9). Some neurons of small or intermediate size are gephyrin-negative (J–L, insets, arrows). A–D, GABA-A-Rγ2; E–H, GlyRα1/2; I–L, gephyrin. A, E, I, normal control; B, F, J, patient FA 20; C, G, K, patient FA9; D, H, L, patient FA29. Bars, 50 µm; insets, 20 µm.

Laser confocal double-label immunofluorescence images of GABA-A-Rγ2 and GlyRα1/2 in the DN of 1 normal control and a case of FA (FA20) are shown in Figure 7. GABA- and glycine receptors occur in close apposition along the plasma membranes of neuronal somata and dendrites.

Figure 7.

(A–F) Double-label immunofluorescence of γ-amino-butyric acid receptor (GABA-A-Rγ2) and glycine receptor α1/2-subunits (GlyRα1/2) in the dentate nucleus (DN) of a control patient (A–C) and patient FA20 with Friedreich ataxia (FA) (D–F). Images were generated by laser scanning confocal microscopy; fluorescence was obtained by Alexa 488- and Cy3-labeled secondary antibodies. The 2 normal DN neurons (A) (shown as "N") and the single DN neuron in FA 20 (D) ("N") display densely packed GABA- and glycine-receptors in close apposition to each other. Abundance and density of the receptors in the FA neuron do not appreciably differ from the normal state. A, D, GABA-A-Rγ2; B, E, GlyRα1/2; C, F, merged images. Bars, 10 µm.

DISCUSSION

Frataxin Deficit and Progression of the DN Lesion in FA

Tissue frataxin concentrations reported here for the DN in FA show a greater deficit than reported for lymphoblasts (16), peripheral blood cells, whole blood, buccal cells (17, 18), or skeletal muscle (19), obtained from living FA patients. The mean level of 6.8 ng/g wet weight in DN represents 3.5% of normal, in contrast to the reported 20% to 30% in blood cells (16, 17). The average frataxin level of 2 carriers (142 ng/g wet weight) represents 72.8% of the normal mean, which is similar to the previously reported percentage in leukocytes (16, 17). Two reasons probably account for the very low frataxin concentrations in the DN in FA patients: 1) the DN is fully differentiated at the time of disease onset; 2) neurons, which account for the bulk of mitochondrial frataxin in the DN (2), do not regenerate. It cannot be assumed, however, that frataxin levels in the DN represent neurological "end stage" of FA because the most common cause of death is cardiomyopathy (Table). The aforementioned regression analysis of DN thickness as a function of frataxin level suggests that relatively high residual levels of the protein may convey some protection against neuronal atrophy. It is unknown, however, whether other gray matter regions of the CNS that are not affected in FA contain more frataxin than reported here for the DN.

Measuring FA Progression by GAD Immunohistochemistry of the DN

GAD is a strongly antigenic protein; and the compact organization of GABA-ergic terminals defines the measurable thickness of DN (Fig. 2). Loss of large neurons accounts for thinning of the DN, but regression analysis cannot determine the start of the disease progress (Fig. 2). Rapidly progressive fatal cardiomyopathy such as in patients FA1, FA10, and FA22 (Table) provided the best insight into the correlation between DN thickness and short duration of the neurological disease (FA1, 8 years; FA10, 7 years; FA22, 5 years).

Normal Purkinje Cells in FA and Grumose Reaction

The apparent morphological integrity of Purkinje cells in FA must be reexamined. Why do Purkinje cells survive loss of their synaptic target (Fig. 3), which is contrary to other neurons of the CNS? Figures 3F–H shows a peculiar abundance of axons that traverse the DN though depletion of afferent fibers would be expected. It is likely that grumose reaction in the DN in FA (Figs. 4A, 5B, C) is analogous to sprouting from transected Purkinje cell axons in experimental animals (13, 20, 21). The intense immunoreactivities with antibodies to synaptophysin (Fig. 4A) and GAD (Fig. 5B, C) establish derivation of these clusters from GABA-ergic Purkinje cell axons. The term "grumose degeneration" should therefore, be revised to "grumose reaction" because proliferation of terminals may be beneficial. From Figure 4B, it appears likely that axonal sprouts establish contact with small surviving neurons (Fig. 5B), and this new connection may provide the necessary retrograde trophic support to Purkinje cells. Purkinje cells also have collateral connections in the cerebellar cortex that may contribute to their survival when large DN neurons disappear. The exemption from FA, however, may only be temporary. In some FA patients, small and intermediate-size DN neurons are sparse (Fig. 3C, G), and some lack GABA-A-Rγ2 (Fig. 6B, inset), GlyRα1/2 (Fig. 6F, inset), or gephyrin immunoreactivity (Fig. 6J, K, insets). They may ultimately disappear, at which time Purkinje cells may become atrophic.

Glutamatergic, GABA-ergic, and Glycinergic Neurons of the DN in FA

Based on the absence of established markers of inhibitory transmission such as GAD or GlyT2, large neurons in the DN are thought to be glutamatergic and excitatory. Considering the role of large DN neurons as the main source of efferent cerebellar output, the morphological integrity of neurons, axons, and terminals in the molecular and granular layers of the cerebellum becomes functionally irrelevant for motor control (8).

In contrast to the relatively abundant small GAD-positive neurons in the DN, GlyT2-reactive nerve cells are sparse (Fig. 5). The paucity of glycinergic somata and axons may be misleading because normal large and small neurons of the DN express abundant GlyRα1/2 (Figs. 6, 7). Inhibitory glycinergic transmission in the DN may be more important than previously considered (22). It is also possible that co-transmission of GABA and glycine from the same terminal occurs in the human DN, in analogy to the cerebellar nuclei of rats (23). GlyT2 immunoreactivity also occurs elsewhere in the cerebellum (24), but from the immunohistochemical reaction product shown in Figure 5, synaptic connections in the human DN may be largely intrinsic.

The deficit of GABA-ergic and glycinergic synaptic transmission in the DN of FA patients is highlighted by loss of GAD, GABA-A-Rγ2, and GlyRα1/2 immunoreactivities (Figs. 5, 6). Selection of an antibody to the γ2 subunit of the GABA-receptor was based on the wide distribution and abundance of this subunit in the CNS, its essential function in benzodiazepine responsiveness (25), and its prominent interaction with gephyrin (26). The presence of GABA-A-Rγ2 on small neurons correlates with the potential of these cells to generate a beneficial grumose reaction in response to axonal sprouts arising from Purkinje cell axons (Fig. 4B).

Gephyrin, named after the Greek word for bridge (27), is of interest to the study of the DN in FA because it anchors both GABA- and glycine receptors to the cytoskeleton of neurons just beneath the plasma membrane (28). Many other proteins are involved in the correct positioning of the receptors at inhibitory synapses and the main function of gephyrin is thought to provide a dynamic linkage to the neuronal cytoskeleton (28). Any disturbance in this mechanism in FA would have serious consequences on inhibitory and likely, trophic, input to the neurons of the DN. The localization of gephyrin reaction product in the cytoplasm (Fig. 6I–L) is not necessarily contrary to its function at the plasma membrane. The observation of gephyrin staining in neuronal cytoplasm is not new (29), and may indicate rapid biosynthesis and degradation of the protein (28). We hypothesize that intact biosynthesis of gephyrin in small DN neurons (Fig. 6J–L). maintains the correct positioning of GABA-A-Rγ2 and GlyRα1/2 and, consequently, proper GABA-ergic and glycinergic transmission. Since these small nerve cells do not send efferent fibers to the thalamus, their survival seems of limited importance to the ataxic disturbance of FA patients. Gephyrin reaction product is generally robust (Fig. 6I–L), and the absence of an immunoreaction in nerve cells of small and intermediate size (Fig. 6J–L, insets, arrows) may imply failed protein biosynthesis.

Interrupted GABA-ergic and Glycinergic Synapses and Frataxin Deficiency

Frataxin deficiency causes a lack of iron-sulfur clusters that affects complexes I, II, and III of the mitochondrial electron transport chain and the citric acid cycle enzyme cis-aconitase. The severe lack of frataxin in the DN in FA reported here, however, does not reduce mitochondriogenesis, as determined by immunohistochemistry with an antibody to mitochondrial complex V (30, 31). Frataxin deficiency should affect all mitochondria-rich cells of the DN, including large and small neurons, with comparable severity. The remarkable survival of small GABA-ergic neurons and their dentato-olivary fibers is not consistent with a shared lack of a single protein, and it may be necessary to interpret failed GABA-ergic and glycinergic transmission as a primary event rather than secondary to neuronal damage. Under this hypothesis, trophic loss due to impaired alignment of GABA-A-Rγ2, GlyRα1/2, and associated proteins precedes atrophy of large glutamatergic neurons. Iron-mediated oxidative injury may not properly apply to the lesion of the DN though the high levels of iron and copper in this gray matter ribbon, and the admixture of these metals in the collapsing DN may be a source of toxic oxygen species (15). Several investigators have sought to connect frataxin deficiency with cytoskeletal abnormalities in FA (32–34). Recently, Bayot et al suggested that destabilization of the actin network in cultured FA fibroblasts occurred because FA also silenced a gene located upstream from the frataxin gene, namely, phosphatidylinositol-4-phosphate-5-kinase 1B (34). Poor growth of cultured FA fibroblasts has been known for many years (35), but it is unlikely that observations on mesenchymal cells are completely relevant to DN neurons. Immunohistochemistry alone will not provide sufficient insight into the role of receptor-associated proteins in FA. In addition to gephyrin, many other proteins of interest have established or putative roles in the linkage of receptors at inhibitory synapses to actin filaments and microtubules (36, 37), among which neuroligin 2 and collybistin have already received detailed attention (37). The methods of proteomics appear well suited to the study of all proteins that contribute to the function of inhibitory synapses in the normal DN and the DN in FA. The hypothesis that intact GABA-ergic input to the large glutamatergic neurons of the DN provides a critical trophic function gains support from analogous observations on the inferior olivary nuclei in FA (8). Intact GABA-ergic dentato-olivary fibers arising from small DN neurons in FA support the neurons of the inferior olivary nuclei and their efferent glutamatergic climbing fibers. Any neoplastic or non-neoplastic lesion that interrupts the dentato-olivary tract and thereby deprives the inferior olivary nucleus of GABA-ergic input, leads to neuronal hypertrophy, vacuolation, and trans-synaptic death of olivary nerve cells.