Cerebrospinal fluid samples were obtained at the time of diagnosis from patients in three French ALS centers (Tours, Lille and Paris). The diagnosis of ALS was made according to diagnostic criteria for definite or probable ALS based on the El Escorial World Federation diagnostic criteria [12]. In routine practice, a CSF analysis is performed in all patients in whom a diagnosis of ALS is suspected. All patients involved in our study gave informed consent. Patients from Lille and Paris gave written consent to perform research with their CSF samples. In routine practice, a CSF analysis is performed in all patients in whom a diagnosis of ALS is suspected. For our study, we used a part of CSF sample to perform complementary analysis with NMR. In Tours, before starting this study, we asked our local ethics committee about the requirement of informed consent. We obtained verbal permission from the Persons Protection Committee's President of Tours to use CSF samples obtained by routine check up and without written consent of patients and controls. Patients from Tours gave verbal consent to use their CSF samples for medical research. One of the clinicians was responsible to give to patients the appropriate information about the use of their CSF samples in medical research. Ethics committees approved consent procedure.

For each patient, information on diagnosis, gender, age, site-of-onset and age-at-onset were obtained. The site-of-onset was defined as either bulbar or limb-onset. The age-at-onset was defined as the time at which motor weakness was first noted by the patient. We compared clinical, demographic data and biological parameters of ALS patients between each center. According to ethical considerations, it is not conceivable to collect CSF from healthy subjects. Then, the control group encompassed individuals with non-neurodegenerative diseases having a routine lumbar puncture at the time of diagnosis. Standard clinical laboratory tests including bacteriological, biochemical analyses of CSF and blood glucose levels were collected for each subject.

CSF samples were stored in polypropylene tubes at −80°C immediately after collection and until analysis [13]. Before the NMR experiments, samples were thawed, and centrifugated at 3000g for 5 minutes. Then 0.5 mL of CSF were mixed with 0.1 mL of deuterium oxyde solution to calibrate the NMR spectrometer. Final adjustment of pH to 9.5±0.15 was done with HCl or NaOH solutions using a pH-meter equipped with a glass combination electrode. Finally, 0.6 mL of the prepared solutions were transferred to a 5-mm NMR tube (CortecNet, Paris, France) for ¹H NMR analysis.

The ¹H NMR spectra were performed on a Bruker DRX-500 spectrometer (Bruker SADIS, Wissembourg, France), operating at 11.7 T, with a Broad Band Inverse (BBI) probehead equipped with Z gradient coil. NMR measurements were done at 298K, with non-spinning samples. CPMG ¹H NMR spectra were recorded with a spin-spin relaxation delay of 80ms (echo time 200 µs repeated 200 times), collected with 128 transients, eight dummy scans and a d1 of 10s into 32k data points with a spectral width of 7500 Hz and an acquisition time of 2.04s. The water peak was irradiated during d1 delay. Unambiguous assignments were performed using a two-dimensional correlation spectroscopy. Eight transients per increment and 256 increments were collected into 4K data points. COSY spectra were acquired with 3s relaxation delay, 6000Hz spectral width in both directions. The water signal was irradiated during the recycling time. Shimming of the sample was performed automatically on the water signal. For quantification of metabolite peaks, the electronic reference (ERETIC) signal was set during acquisition time [14]. The ERETIC pulse was generated by a second radio frequency channel phase-synchronized with the NMR signal of the sample. ERETIC intensity, frequency and line broadening were calibrated with a reference tube to determine metabolite concentrations.

Data were processed using XWinNMR version 3.5 software (Bruker Daltonik, Karlsruhe, Germany). Prior to Fourier transformation (FT), the 1D FIDs were zero-filled to 64K data points which provided sufficient data points for each resonance, and a line broadening factor of 0.3Hz was applied. All spectra were corrected for phase distortion and the baseline was manually corrected for each spectrum. Each spectrum was integrated using WinNMR software integral function.

Quantification of metabolite peaks was performed with the ERETIC peak as a quantitative reference. Calibration of the ERETIC peak, which had the same area in all spectra, was made using a mixture of 9.13mM of citrate, 9.20mM of lactate, and 9.10mM of glutamate as the reference tube. According to this reference tube, the measured area of the ERETIC peak (A_E) symbolized a calculated concentration (C_E) of 4,67mM. Metabolite concentration (C_X) was calculated according to the following formula:where Cx is the metabolite concentration in the CSF sample, Nx is the number of protons of the quantified peak metabolite, C_E is the calculated concentration for the ERETIC peak, A_E is the ERETIC area peak, and f_D is the dilution factor of the CSF sample.

Spectral ¹H assignments were achieved according to the literature values of chemical shifts in various media and biofluids [15]. Because beta-methyl-L-amino-alanine (BMAA) has been recently reported by several authors as potentially involved in the etiology of ALS, we searched for its presence [16].

Clinical data and CSF samples were obtained from 50 patients with ALS (22 from Tours, 19 from Lille and 9 from Paris) and 44 controls. No differences in clinical, demographic or biological data were found between patients from Tours, Lille and Paris, so we pooled these samples into a single group: the ALS group. 61% of ALS patients were male and had median age-at-onset of 62.5 years (22.9–90.6). 65% of ALS patients had limb-onset. The control group encompassed patients with the following diseases : sarcoidosis (one patient), psychiatric disorders (three patients), cerebellar syndrome (one patient), peripheral neuropathy (thirteen patients), chronic inflammatory demyelinating polyneuropathy (six patients), muscular cramps (one patient), facial paralysis (one patient), traumatic cerebral injury (one patient), ataxia with axonal neuropathy (one patient), normal pressure hydrocephalus (two patients), paresthesia (two patients), epileptic seizures (one patient), spinal cord infarction (one patient), cervical spondylotic myelopathy (three patients), adrenomyeloneuropathy (one patient), non-organic hypoesthesia (one patient), neuroborreliosis (one patient), dystonia (one patient), Hallervorden-Spatz syndrome (one patient), non neurologic deafness (one patient), and oculomotor paralysis (one patient). Controls were matched with ALS patients by age and sex. There were no differences in standard bacteriological and biochemical tests of CSF, or in blood glucose levels between groups (Table 1).

The ¹H-NMR signals of all common metabolites including amino-acids, organic acids, and carbohydrates were assigned according to previous publications. [17]. Qualitative analysis of CSF by superimposition of spectra from ALS patients and controls did not reveal any differences in metabolite composition.We found particularly no trace of BMAA in the CSF of our ALS patients. Accurate visual comparison of the spectra was difficult because of dilution factor. An example of CSF spectra from a typical ALS patient and control are shown on figure 1.

We analysed 17 CSF metabolites in ALS and non ALS patients, defined as follows: amino-acids (alanine, glutamine, tyrosine), organic acids (citrate, acetate, α-hydroxybutyrate (AHBT)), ketone bodies (β-hydroxybutyrate (BHBT), acetone, acetoacetate), glucose, fructose, metabolites involved in glucose metabolism (pyruvate, lactate, creatinine and creatine, recently identified as markers of mitochondrial dysfunction [18], the anti-oxidant molecule ascorbate, and formate as well as ethanol.

Results of quantitative analyses of metabolites in CSF from ALS patients and controls are presented in table 2. We noted 3 metabolites for which mean CSF concentrations were significantly different between groups (p<0.003). We found significantly lower concentrations of acetate in ALS patients than in controls (53 vs 104 µmol/L, p = 0.0002). There were higher concentrations of ascorbate and pyruvate in ALS patients compared to non ALS patients (p = 0.003 and p = 0.002, respectively). The lactate/pyruvate ratio, a marker of cytosolic oxido-reduction state was lower in ALS patients than in controls (56 vs 115, p = 0.0029). Acetone concentrations trended to be higher in ALS patients (p = 0.015). Box plots representing acetate, acetone, ascorbate, and pyruvate concentrations in the ALS and non ALS groups are shown in figure 2.

PCA analysis was applied to the NMR data using the 17 metabolites integrated in all patients. The resulting PCA score plot is shown in figure 3, with the two first PCs explaining 55.65% of the variation in the selected metabolites. The PCA scores plot for the quantitative NMR data showed separated clusters for ALS and non ALS patients. Each spectrum can be viewed as an observation in PCA space where the proximity of observations represents the similarity of the metabolic profiles of CSF samples. The PCA score plot shows the metabolites contributing to the separation of the two clusters and allows the identification of metabolites acting on this clustering. The ALS patients seemed to be characterized by high concentrations of pyruvate, ascorbate, and acetone whereas the non-ALS group seemed to be weighted by lower concentrations of these metabolites and by high concentrations of formate, acetate and β-hydroxybutyrate.

All metabolites were entered into the discriminant model. According to this panel of metabolites, we evaluated the clinical status of patients and identified that they were predicted in the correct group with a probability evaluated at 81.6%.