Study Design and Subjects
This was a phase II pilot, prospective, open-label, single-center study, which was approved by the Spanish Agency of Medicines and Medical Devices and by the Clinical Research Ethics Committee of Bellvitge University Hospital and was registered at ClinicalTrials.gov (NCT01495260). The trial was conducted in accordance with the ethical standards of the Declaration of Helsinki. The subjects were followed up for 2 years (8 visits). Thirteen subjects, 12 men and 1 woman, were selected for the study. The age of the subjects ranged from 24 to 64 years. All the subjects had a confirmed diagnosis of AMN on the basis of elevated VLCFA levels and an identified pathogenic variant of the ABCD1 gene. All the subjects met all the inclusion criteria and none of the exclusion criteria in accordance with the protocol.
Symptomatic subjects over 18 years of age, with confirmed diagnosis of AMN by elevation in VLCFA and the presence of a mutation in the ABCD1 gene.
Cerebral inflammatory disease verified with gadolinium enhancement, cerebral disease with cognitive impairment, Expanded Disability Status Scale (EDSS) > 6.5, hypersensitivity to compounds related to cysteine, peptic ulcer, asthma, severe respiratory failure, impaired hepatic function, or other basic blood and urine tests, and pregnant, lactating, or childbearing-aged women.
Clinical and demographic data for the subjects at the initiation of the study are available in Table 1. VLCFA levels and pathogenic variants in ABCD1 of the subjects are available in Supplemental Data (Table 1).
For statistical purposes, plasma and peripheral blood mononuclear cells (PBMCs) from 25 healthy age- and sex-matched individuals and cerebrospinal fluid (CSF) from 9 subjects with nonrelated diseases were used as controls.
Because this combination of antioxidants has not, to our knowledge, been used previously, we used 2 different doses owing to safety considerations. Patients initially received a lower dose A: NAC (800 mg), LA (300 mg), and vit E (150 IU) orally daily for 2 months. After this treatment period, a 2-month washout period was introduced, and protein oxidative damage biomarkers in plasma were tested. In patients showing normalization of biomarkers, the treatment was restarted for 12 months at the same dose. In patients showing no normalization of oxidative damage biomarkers, the dosage was increased to dose B for 3 months: NAC (2400 mg), LA (600 mg), and vit E (300 IU), single high doses already reported in the literature. After this treatment period, the biomarkers were tested again. A further washout period of 2 months ensued. If normalization of the levels was attained with dose B, the treatment was restarted for 12 more months with the higher dose. In the eventuality that protein oxidative damage biomarkers were not reduced, the patient was considered a nonresponder, and the treatment was discontinued at that point. Plasma and PBMC extractions were assessed at all visits (Fig. 1).
Only 1 patient (patient 8) exhibited normalized oxidative damage biomarkers in plasma and PBMCs with the lower dose of antioxidants (dose A). Thus, and following the protocol, the patient was maintained on the same treatment regimen (dose A) for an additional year until the end of the study. The remaining subjects were treated with dose B for an additional year.
Oxidative Protein Damage Markers
Metal-catalyzed oxidation products: AASA (aminoadipic semialdehyde) and GSA (glutamic semialdehyde); mixed glycoxidation/lipoxidation products: CEL (Nepsilon-(carboxyethyl)-lysine) and CML (Nepsilon-(carboxymethyl)-lysine); and lipid peroxidation-derived products: MDAL (Nepsilon-malondialdehyde-lysine) were quantified in the plasma as previously reported . AASA, CML, CEL, and MDAL were expressed as micromoles per mole lysine.
Oxidative DNA Damage Marker
8-Oxo-dG (7,8-dihydro-8-oxo-2-deoxyguanosine) was tested in the urine by HPLC according to the methodology described by Haghdoost et al. . The 8-oxo-dG content was expressed as nanogram per milligram creatine.
The reference values for oxidative damage biomarkers were calculated according to the sample distribution of 25 healthy individuals. The point that defines a superior moderate outlier in the sample of controls was taken as the reference limit (Q75 + 1.5 × IQR, Q75 = 75% percentile, IQR = interquartile range).
We quantified the following molecules with previously reported methods : i) plasma eicosanoids and oxidized polyunsaturated fatty acids (AA, DHA, PGD2, PGE2, PGF2α, 6-keto-PGF1α, 9S-HODE, 13S-HODE, 12S-HETE, 15S-HETE, and TXB2); ii) plasma inflammatory cytokines, chemokines, and signalling receptors (HGF, IL-6, IL-8, MCP1, TNF, and adiponectin); iii) PBMC cytokines and signalling receptors (IFNA2, IL-37, IL-4, PPARα, IL-10, IL-36A, IL-36RN, IL-13, CCR3, CXCL5, IL-9R, and STAT1). Briefly, eicosanoids and oxidized polyunsaturated fatty acids were measured by a triple quadrupole mass spectrometry–based metabolite quantification assay (Biocrates Life Science AG, Innsbruck, Austria); HGF, IL-6, IL-8, MCP1, TNF, and adiponectin were determined by immunoassays using the Milliplex kit (Luminex xMAP Technology, EMD Millipore Corporation, Billerica, MA), and IFNA2, IL-37, IL-4, PPARα, IL-10, IL-36A, IL-36RN, IL-13, CCR3, CXCL5, IL-9R, and STAT1 gene expression was quantified by real-time PCR with Taqman® probes in total RNA from PBMCs. We also tested neopterin in the cerebrospinal fluid according to the methodology described by Molero-Luis et al. .
All the subjects were clinically examined using a standardized neurologic examination, including central motor, sensory, cerebellar, cerebral, and peripheral nerve motor functions. A standard neurologic history and examination were used to score the subjects on the Kurtzke EDSS . The 6-minute walk test (6MWT)  was performed at all follow-up visits. For statistical purposes, this test was considered at 0 versus 12 months.
Visual evoked potentials (VEP), brainstem auditory evoked potentials (BAEP), motor evoked potentials (MEP), somatosensory evoked potentials (SEP), and laser-evoked potentials (LEP) were obtained through standard techniques using Synergy electromyographs (Oxford Instruments, Surrey, UK). Normative data were obtained from 10 healthy adults examined under the same experimental conditions . Abnormality was defined as the presence of a latency, amplitude, or conduction velocity > 2.5 SD above the mean values in healthy adults.
Monocular pattern reversal VEP was recorded by means of a pattern reversal black-and-white checkerboard presented at 1.9 Hz on a cathode ray tube screen. The visual angle was 35′ and 70′. Monocular stimulation was performed by covering 1 eye with an eye patch, and full-field stimuli were given. The recording electrode was placed at Oz according to the international 10/20 system and referenced to Fz. The ground electrode was placed over the Cz. Signals were recorded using Synergy electromyographs (Oxford Instruments) with a bandpass filter set at 1 to 100 Hz, with a sweep time of 500 ms. At least 2 averages of 100 artifact-free trials were recorded. We measured P100 latency and amplitude.
BAEPs were obtained for clicks presented at 70 dB above the hearing threshold using a Synergy electromyograph (Oxford Instruments). A click sound was presented to the unilateral ear on the reference side at a rate of 21.1 Hz using headphones; each side was examined separately. The bandpass filter was set at 100 Hz to 1.5 kHz. Evoked responses of 1500 stimuli were averaged, and at least 2 trials for each ear were conducted. The recording electrode was placed over the vertex (Cz), the reference electrode was placed on the unilateral earlobe, and the ground electrode was placed over the Fz. The peak latencies of I, II, III, IV, and V waves were measured, and the interpeak latency between I, III, and V waves was calculated.
During the magnetic stimulation, the patients lay comfortably in a supine position. MEPs were recorded from the abductor digiti minimi of the hypothenar and abductor hallucis muscles of the foot using Ag/AgCl surface cup electrodes placed in a belly tendon montage. Signals were recorded using Synergy electromyographs (Oxford Instruments) with filters set at 3 Hz and 10 kHz. Magnetic stimulation was conducted using a monophasic stimulator (Magstim 200; The Magstim Co. Ltd., UK) and a round magnetic stimulating coil (10-cm diameter; The Magstim Co. Ltd.) to stimulate upper limbs and a double cone coil (11-cm diameter; The Magstim Co. Ltd.) to stimulate lower limbs. The stimulus was applied over the vertex of the scalp. To obtain preferential activation of each hemisphere, induced current flowed from the posterior to the anterior direction over the motor area. We measured the threshold for evoking a response in the target muscles. Responses were recorded at rest and with a minimum voluntary contraction (facilitation), because MEP size increases and its latency shortens when the muscle is voluntarily contracted. We recorded 4 consecutive responses at rest and with facilitation and measured the response with a minimum latency and maximum amplitude. The central motor conduction time (CMCT) was calculated with the following formula: CMCT = CM−((DL + F wave−1) / 2), in which CM is the latency between the cortex and the muscle explored, DL is the distal latency of the motor response, and F is the shortest F wave latency.
For SEP recording, patients lay in the supine position in a warm and semidarkened room to make the patient as comfortable as possible and help him to relax. SEPs were elicited after electrical stimulation (0.2-ms duration, constant current, 4.13-Hz stimulus rate) with skin electrodes from both median nerves at the wrist and posterior tibial nerves at the ankle. The intensity of electrical stimuli was set slightly above the motor threshold. The number of sweeps averaged was between 500 and 1000. Samples with excessive interference were automatically rejected from the average. The EP tracings were replicated and superimposed to demonstrate the reproducibility of the components measured. The filter bandpass was 3 to 3000 Hz. The analysis time was 50 ms for median nerve SEPs and 100 ms for tibial nerve SEPs. Recording electrodes for the median nerve SEPs, with an electrical impedance of less than 4 kΩ, were placed bilaterally in the supraclavicular fossa (Erb’s point), in the skin overlying the sixth cervical spinous process (Cv6) and in the parietal scalp region (C3′ and C4′). The Erb’s point electrode was referred to the contralateral Erb’s point, the Cv6 electrode was referred to an electrode located immediately above the thyroid cartilage and scalp electrodes (C3′ and C4′) were referred to Fz. For the posterior tibial nerve SEPs, the recording electrodes were placed over the tibial nerve in the popliteal fossa, over the low back on the skin overlying the spinous processes of vertebra L1, over the skin overlying the cervical spine and in Cz′ (2 cm posterior to the Cz) in the scalp. The popliteal fossa electrode was referred to an electrode placed 3 cm proximally, the first lumbar vertebra electrode was referred to an electrode located immediately above the umbilicus, the Fz electrode was referred to cervical electrode, and Cz′ was referred to Fz. For median nerve SEPs, we evaluated peaking latencies of Erb’s point N9, generated by the brachial plexus volley and recorded by the Erb’s electrode; the spinal N13 response, generated within the cervical gray matter; and contralateral scalp N20, recorded from the parietal electrode and generated in the primary somatosensory (SI) area. We also calculated N9-N13, N9-N20, and N13-N20 interpeak latencies and measured the peak-to-peak amplitude between N20 and the subsequent positive peak (P25). For tibial nerve SEPs, we evaluated the peaking latencies of the N8 recorded at the popliteal fossa, N22 recorded at L1 generated by the lumbosacral dorsal horn neurons, P30 recorded at the sixth cervical spinous process, and the scalp P40 response, probably generated in the SI area. We also calculated N22-P40 to assess the conduction time in central somatosensory pathways and measured the peak-to-peak amplitude between P40 and the subsequent negative peak (N50).
LEPs were obtained by delivering brief radiant heat pulses to the face, dorsum of the hand and leg by means of a NdYAP laser stimulator (wave length 1.34 μm; beam diameter 5 mm; energy 1.5–3.5 J; duration 5 ms). The stimulus intensity was approximately 1.5 to 2 times the mean pain threshold of healthy subjects. Signals were recorded using a Synergy electromyograph (Oxford Instruments). LEPs were recorded from Cz with reference to linked earlobes, using silver/silver chloride cup electrodes of 9-mm diameter filled with a conductive adhesive gel. The amplifier bandpass frequency filter was 0.2 to 100 Hz. The analysis time was 1 s. Impedance was maintained at less than 4 kΩ. Ten artifact-free LEPs were recorded for each side and averaged offline. We measured the latency of the first negative peak (N2) and of the subsequent positive peak (P2), as well as the peak-to-peak amplitude (N2/P2 amplitude).
Nerve Conduction Studies
Synergy electromyographs (Oxford Instruments) were used for these studies. The following parameters were measured: nerve conduction velocity, distal latency (with the distance kept constant), and amplitude of response. Recordings were performed using standard methods . Abnormality was defined as the presence of a latency, amplitude, or conduction velocity > 2 SD or 1 SD. In motor nerve conduction studies, at least 1 unilateral median, ulnar, and peroneal nerve was evaluated. Distal and proximal compound muscle action potential (CMAP) amplitude, distal motor latency (DML), duration of proximal and distal CMAP (dCMAP), and motor nerve conduction velocity (MNCV) were also assessed. The sensory tests were performed in the sural nerve, except in 1 patient (patient 7), who had suffered a previous bilateral sural biopsy. In this patient, the studies were performed in the superficial peroneal nerve. The motor nerve conduction studies were performed in the peroneal nerve except for patient 1, for whom the studies were performed in the tibial nerve. With regard to these studies, 7/13 subjects presented a mild peripheral neuropathy at the beginning of the trial. A pattern suggestive of neuropathy with demyelinating features was observed in 2/13 subjects, and axonal neuropathy in 5/13 subjects.
MRI with gadolinium contrast was performed in a 1.5-T apparatus (Philips Healthcare, Best, The Netherlands). Diffusion tensor imaging (fractional anisotropy maps) was obtained for each patient. Sagittal T1-weighted, coronal-enhanced IRT1-weighted imaging, and FLAIR sequences were obtained for the axial plane and coronal sequences with pulse inversion recovery providing high contrast between white matter lesions and other adjacent structures. The subjects were classified into 3 groups according to severity of the lesions in FLAIR/T1 images. Three subjects (3, 8, and 9) showed a normal pattern, 7 exhibited moderate lesions in the corticospinal tract (CST) and peritrigonal white matter (mild hyperintensity on FLAIR sequence, with a bilateral and symmetrical distribution), and 3 (subjects 1, 4, and 6) exhibited clearly pathological lesions in the CST. In those 3 subjects, the lesions were more severe in the internal capsule and midbrain, peritrigonal white matter, optic radiation, and corpus callosum; in addition, the pattern of lesions was more focal, with prominent altered signals.
DTI was acquired using a single-shot sequence echo planar imaging. The image matrix was 144 and a field of vision (FOV) of 234 × 170 mm was the direction of phase AP. The entire skull was studied, including stem, with 60 axial sections 2 mm thick, which were acquired parallel to the anterior and posterior white matter commeasured with isometric voxel 2 × 2 × 2 mm. The value of b was 400 to 800 s/mm2 with an average resolution, and 16 points were calculated. From this sequence, the isotropic image and fractional anisotropy (FA) maps for each of the patients were obtained. Regions of interest (ROIs) were defined coinciding with MRS voxel, and the FA values were obtained at the bilateral corticospinal tract. A third ROI was performed in 3 patients in whom impaired signal was observed in the white matter on conventional sequences. Visual inspection of the diffusion images was performed to exclude moving images or artifacts. The single-voxel 1H-MRS technique was used to acquire data from MRS. A VOI from 1.5 to 2 cm3 was placed in the left parieto–occipital white matter. Two spectra were acquired in the same VOI: 1) SE STE (TR/TE/stockings, 2000/30/96 -192) and 2) (TR/TE/averages, 2000/136/128-256) SE LTE. Myo/Cr, Cho/NAA, and NAA/Cr ratios were calculated for each patient. The same radiologist performed the evaluation of the images and calculations consecutively.
Two patients were removed from the statistical analysis. Subject 1, who showed a more severe lesion pattern on MRI among the patients, presented a behavioral change after 9 months in the trial, with a jocular attitude, puerility, and impulsiveness, signs of frontal release (glabellar and palmomental reflexes), and developed inflammatory lesions 6 months after the end of the treatment. Subject 6 presented hypomania, perseveration, and disinhibition, as previously noted in the pre-existing medical history, although the principal investigator considered during the inclusion period that he would be able to follow through with treatment. Although the development of psychiatric disease and cerebral manifestations is common in AMN, because we did not include an untreated comparison group, we cannot formally exclude an unlikely treatment-related event. Furthermore, the low number of patients and short observation period in our cohort preclude from extrapolating how many would be expected to develop the cerebral AMN phenotype in our cohort.
Protein oxidative damage, inflammation biomarkers, and Q-PCR array expression data were examined for normality by the Shapiro–Wilk test. Significant differences were determined using a 1-tailed Student’s T test if the data were normally distributed or a 1-tailed Wilcoxon rank sum test otherwise. When analyzing the treatment effect by comparing patient scores at different time points, we used the 1-tailed paired Student’s T test if the data were normally distributed or the 1-tailed Wilcoxon signed rank test otherwise. We chose the 1-tailed T test instead of the 2-tailed because we aimed to elucidate whether or not the analyzed parameter deviated from the controls in the same sense for all subjects. We further analyzed whether these values were affected by the treatment and changed in concert as a block in a given sense. The patient on dose A was excluded from the statistical study, as the dose used was much lower than the rest of patients.
Principal component analysis (PCA) was applied as previously reported . We performed 3 different analyses grouping the variables under the following categories: 1) clinical variables and 2) expression levels of inflammatory markers and mediators. We considered those factors that changed significantly in response to treatment, and we analyzed their levels prior to treatment as variables, together with the age of the patients. To extract the eigenvalues and calculate the factor loadings, PCA was performed with the S-Plus princomp function in the R programming environment.
Penalty regression methods were conducted using the glmnet package  in R. Briefly, we applied a likelihood ratio test using the lmtest package  to differentiate the influence of the variables selected by the penalty regression methods from confounding variables such as age and distance walked in the 6MWT before treatment . The statistical analyses were performed using the Bioconductor packages in the R programming environment . To investigate whether the distance walked at the end of the assay in the 6MWT could be predicted by the levels of oxidative damage and inflammatory markers, we distributed the data into 3 groups of variables: i) inflammatory cytokines, chemokines, and receptors; ii) inflammation-associated lipids; and iii) the expression of inflammatory cytokines, chemokines, and receptors. We also included the age and 6MWT distance for each patient before treatment as independent variables. We applied Y = b0 + b1X1 + b2X2 … bnXn to the independent variables (X1… Xn), which were defined as the 6MWT distance before treatment, age, MCP1, and 15S-HETE levels, to better predict the dependent variable (Y), i.e., the 6MWT distance at the end of the study. To validate the penalized regression model, the analysis included a leave-one-out cross-validation (LOOCV) procedure characterized by using a single sample as validation data for testing the model, whereas the remaining samples were utilized as training data. The cross-validation process was then repeated as many times as the number of samples used in the model .