Lithium delays progression of amyotrophic lateral sclerosis

Fornai et al. 10.1073/pnas.0708022105.

Supporting Information

Files in this Data Supplement:

SI Figure 6
SI Figure 7
SI Figure 8
SI Figure 9
SI Figure 10
SI Figure 11
SI Figure 12
SI Figure 13
SI Figure 14
SI Figure 15
SI Figure 16
SI Figure 17
SI Figure 18
SI Figure 19
SI Figure 20
SI Figure 21
SI Figure 22
SI Figure 23
SI Figure 24
SI Figure 25
SI Movie 1
SI Table 1
SI Text

SI Figure 6

Fig. 6. The effects of lithium on additional behavioral tests in G93A mice. Lithium improves the motor function of G93A mice as shown by additional tests. This improvement occurs in different behavioral tests aimed at exploring various features of motor activity. Graphs indicate various motor activities: (a) Rotarod test: indicates the time the mouse is able to stay on a rotating rod (rodspeed = 15 rpm); (b) Grip test: indicates the time the mouse is able to hold the grid with its hind limbs and, mostly, forelimbs before falling; (c) stride test: indicates the length of the step (stride, in cm) measured by the movement of the mouse's painted hind limb. Data are the means of the all strides lengths recorded in each group of mice (n = 20) during 2 min of observation. In each behavioral study the mice were tested twice a week starting at 90 days of age and during disease progression extending two weeks beyond the death of the last G93A saline-treated mouse. This meant a longer testing time for lithium-treated mice. The age of the mice is reported on the x axis. Results are shown as the mean ± SEM. *, P < 0.05 G93A+Lithium vs. G93A+saline.

SI Figure 7

Fig. 7. Lithium effect on lumbar spinal cord Lamina IX at 90 days. (a) Nissl and H&E staining of lumbar spinal cord sections from WT and G93A mice after 15 days of treatment. The images show the qualitative differences in cell density within Lamina IX following lithium in the G93A mice. (b) Reports cell counting by referring to the number of alpha motor neurons in Lamina IX, as described in SI, and using a size exclusion method, as reported by Martin (9). It is evident that at this age (90 days), which precedes the behavioural onset of the disease, the neuropathology is already present in the lumbar spinal cord of G93A mice. At this presymptomatic stage, the disease-delaying effects of lithium are significant even in this tract of the cord. *, P < 0.05 compared with wild type. #P < 0.05 compared with saline-treated G93A. (Scale bars, 70 mm.)

SI Figure 8

Fig. 8. Number and size of alpha motor neurons in the lumbar spinal cord. Motor neurons were selected within Lamina IX of the spinal cord according to their size [see Methods and see also Martin et al., 2007 (9)]. (a) Low magnification of H&E-stained spinal cord at lumbar level of WT and G93A mice treated either with saline or lithium. Note the representative difference in the cell density. (b) Higher magnification within Lamina IX showing the scarce density of motor neurons in G93A mice. (c) The count of alpha motor neurons, in 125 sections from each mouse, in 20 mice (total of 2,500 sections) according to a stereological quantification (see stereological analysis in SI) allows quantifying both the mean number of alpha motor neurons per section (left) and the total number per mouse (right) in the spinal cord at lumbar level (as delimited by the stereological analysis and defined by the size exclusion criteria to avoid a bias due to heterogeneous neuron size). It is evident that G93A mice in the last disease stage (mean age = 110.8 ± 5 days), have lost a significant number of these neurons according to both counting techniques. (d) Shows the significant increase both in diameter and area of the motor neurons belonging to G93A mice (as calculated by a computer-dedicated software, see Methods). These pathological changes are absent in G93A mice treated with lithium (mean age = 148 ± 4.3). The quantification of this effect is reported in (e) showing both the diameter (left graph) and the area (right graph) of Lamina IX alpha motor neurons of the different groups. Values are the mean ± SEM of a total number of 2,500 sections per group. The data on neuron size are the mean ± SEM of a total of 61,000 neurons for WT (consider that according to the stereological procedure not all sections were considered to build a single motor neuron). Comparison between groups both regarding cell number and cell size was made by using one-way ANOVA. *, P < 0.05 compared to WT saline-treated group; #, P < 0.001 compared to G93A saline-treated group. (Scale bars: a, 400 mm; b, 70 mm; d, 21 mm.)

SI Figure 9

Fig. 9. Effects of lithium on cervical spinal cord at the end of the disease. (a) Low magnification of H&E-stained cervical spinal cord of WT and G93A mice treated either with saline or lithium. Note the representative difference in the cell density within Lamina IX. (b) Higher magnification, within Lamina IX, showing the scarce density of motor neurons in G93A saline-treated mice and the improvement after lithium treatment. This effect was confirmed by the quantitative stereological analysis (c) performed as described in SI Methods. Note that the effect of lithium on motor neuron survival we found in the cervical spinal cord is comparable with that described in SI Fig. 7 for the lumbar spinal cord at a presymptomatic stage (90 days). This is in line with the spontaneous delay in cervical compared with lumbar motor neuron death in G93A mice. *, P < 0.05 compared with WT saline-treated group; #, P < 0.05 compared with WT and G93A saline-treated groups. (Scale bars: a, 400 mm; b, 70 mm.

SI Figure 10

Fig. 10. Lithium treatment protects the neurons of the nucleus ambiguus (NA). These data were obtained at the end of the disease (mean age = 110.8 ±5 for saline-treated G93A and 148 ± 4.3 for lithium-treated G93A mice) and age-matched controls. Histogram in a presents the mean number of alpha motor neurons in the NA selected according to the same size-exclusion criteria reported for the spinal cord to avoid a size heterogeneity bias, as elegantly reported by Martin et al. (9). Motor neurons were counted within serial, nonconsecutive sections obtained as described in stereological analysis (SI Methods). G93A Saline-treated mice show a significant reduction in the number of alpha motor neurons. Lithium treatment attenuates cell loss significantly (from 67 ± 5.1 to 91.3 ± 4.7). (b) Schematic map from the mouse atlas of Paxinos and Franklin (45) showing the localization of the NA. (c) Representative Nissl-stained slices of the NA. The Nissl staining allows appreciating the placement of the nucleus in the medulla oblongata (arrow) and the highly selective neuronal loss occurring in the G93A mice, which resembles a stereotaxic lesion. (d) The H&E staining at high magnification allows appreciating the severe neuronal loss in the mutant mice (G93A+saline) and the protective effects of lithium treatment (G93A+lithium). Motor neurons of the NA of G93A saline-treated mice appear rare and suffering, surrounded by a vacuolated tissue; by contrast, G93A mice treated with lithium have higher neuronal density and improved morphological features. Values shown in a represent means ± SEM., and they were analyzed by one-way ANOVA. *, P < 0.05 compared with WT saline-treated group; #, P < 0.05 compared with WT and G93A saline-treated groups. (Scale bars: c, 700 mm; d, 50 mm.)

SI Figure 11

Fig. 11. Lithium treatment decreases GFAP immunostaining. (a) Immunofluorescence of cervical and lumbar spinal cords of WT and G93A lithium-treated mice compared with the saline-treated mice showing GFAP staining. At the end of the disease, GFAP immunofluorescence appears highly intense in G93A saline-treated mice, whereas lithium treatment decreases GFAP-positive cells in both tracts of the spinal cord compared with saline-treated G93A mice. No GFAP staining was observed in age-matched, saline- or lithium-treated WT mice. (b) Representative immunoblots and relative densitometric analysis of GFAP expression in the spinal cord of saline- and lithium-treated WT and G93A mice. The significance of these findings is discussed in SI Discussion, -2- The Potential Role of Interference with Glial Activation). Values represent the mean ± SEM. of 5 repeated analyses compared by one-way ANOVA. *, P < 0.05 compared with WT saline-treated group; **, P < 0.01 compared with WT saline-treated group; #, P < 0.05 compared with WT and G93A saline-treated group. (Scale bars: 28 mm.)

SI Figure 12

Fig. 12. Lithium treatment increases the number of NeuN-positive neurons within Lamina VII. (a) Representative pictures showing NeuN-positive neurons in the Lamina VII of WT and G93A saline- and lithium-treated mice. As reported in the histogram (b), the number of NeuN-positive neurons within the Lamina VII of the lumbar spinal cord increased in the G93A mice after lithium treatment. Remarkably, this effect not only consisted of protection from the loss that occurred in the G93A saline-treated mice, but extended to a robust increase in neuron number compared with saline-treated controls. The number of NeuN positive neurons overlaps that reported within Lamina VII after H&E staining, calbindin 28K and gephyrin immunostaining. The counting criteria were the same under comparable stereological conditions. (c) Representative Western blot and densitometric analysis of NeuN expression in the lumbar spinal cord of WT and G93A mice, which confirms the increase in NeuN-staining after lithium treatment. Values represent means ± SEM., analyzed by one-way ANOVA. *, P < 0.05 compared with WT saline-treated group; #, P < 0.05 compared with WT and G93A saline-treated groups. (Scale bars: 13 mm.)

SI Figure 13

Fig. 13. Lithium treatment increases the number of calbindin 28K-positive neurons in Lamina VII. (a) Lumbar spinal cord sections from saline- and lithium-treated WT and G93A mice immunostained for calbindin 28K. (b) Histogram shows the significant loss of Lamina VII calbindin 28K-positive neurons in the G93A mutant mice compared with WT. Even in this case, the increase in the number of neurons in G93A mice after lithium treatment significantly surpasses that counted in controls. As observed after the other staining procedures, lithium induced no effect on neuron number in WT mice. (c) Representative Western blot and densitometric analysis of calbindin 28K expression in lumbar spinal cord in WT and G93A mice confirm the loss of the calbindin 28K-positivity in G93A saline-treated mice and the increase in the calbindin 28K-positivity after lithium treatment. Values represent means ± SEM, analyzed by one-way ANOVA. *, P < 0.05 compared with WT saline-treated group; #, P < 0.05 compared with WT and G93A saline-treated groups. (Scale bars: 16 mm.)

SI Figure 14

Fig. 14. Effects of lithium on BrdU incorporation (immunofluorescence). Representative pictures at immunofluorescence (a-d) show incorporation of BrdU (green fluorescence) within Lamina VII, where calbindin 28K-positive neurons are also shown (red fluorescence). Immunopositive cells for BrdU are counted in graph (e). There is a negligible signal in saline-treated controls, whereas this is substantially increased after lithium or in saline-treated G93A mice (e). A significant effect was obtained in G93A mice treated with saline. In graph (f) calbindin 28K is counted and provides data similar to those obtained with immunoperoxidase (SI Fig. 13). These results are similar to those obtained with all of the various staining procedures for Renshaw-like cells and correspond to the number of neurons counted by using NeuN or H&E after the size-exclusion criteria for Renshaw cells (see the summarizing SI Fig. 16). When BrdU was combined with calbindin 28K immunostaining, a positive merge occurred only in those G93A mice treated with lithium. BrdU was administered to wild-type and G93A 90-days-old mice at a dose of 50 mg/kg i.p. every other day for 10 days. This confirms the property of lithium to induce a preferential differentiation of NPC toward calbindin 28K positive neurons (See SI Discussion and refs. 9 and 10). *, P < 0.05 compared with WT saline-treated group; #, P < 0.05 compared with WT and G93A saline-treated groups. (Scale bars: a-d, 20 mm).

SI Figure 15

Fig. 15. Effects of lithium on BrdU incorporation (immunoelectronmicroscopy). At the ultrastructural level (a-e) we found the presence of nuclear ImmunoGold particles bound to BrdU (black arrows) from selected neurons of Lamina VII. According to morphological criteria, only Renshaw-like neurons were double-stained for BrdU and calbindin 28K (black and red arrows, respectively, in d and e). Colocalization of BrdU and calbindin 28K occurred only in the G93A mice treated with lithium (f, lower graph). BrdU ImmunoGold = 10 nm; calbindin 28K ImmunoGold = 20 nm. *, P < 0.05 compared with WT saline-treated group; #, P < 0.05 compared with WT and G93A saline-treated groups. (Scale bar: a-e, 0.2 mm.)

SI Figure 16

Fig. 16. Overview of the consistency of the increased neuron number produced by lithium within the Lamina VII of G93A mice. Lithium treatment increased the number of neurons within the Lamina VII of the lumbar spinal cord, as confirmed by H&E staining and immunohistochemical analyses with gephyrin, NeuN and calbindin 28K antibodies. The amount of the increase, which significantly exceeded the number of neurons counted in controls, is in sharp contrast with the severe cell loss that occurs within the Lamina VII of G93A mice in the absence of a pharmacological treatment (saline). Absolute counts of neurons show remarkable consistency regardless of the staining procedure. Values represent means ± SEM., analyzed by one-way ANOVA. *, P < 0.05 compared with WT saline-treated group; #, P < 0.05 compared with WT and G93A saline-treated groups.

SI Figure 17

Fig. 17. Effects of lithium on alpha-synuclein immunostaining. Representative pictures of Lamina IX (a) and Lamina VII (b) of alpha-synuclein immunostaining of WT and G93A mice after saline and lithium treatment. Pictures suggest an increase in alpha-synuclein staining in G93A mice, which is reverted by lithium administration. (c) Representative Western blot and densitometric analyses of alpha-synuclein expression confirming the lithium-induced occlusion in alpha-synuclein increased positivity of G93A mice. Values represent means ± SEM., analyzed by one-way ANOVA. *, P < 0.05 compared with WT saline-treated group; #, P < 0.05 compared with WT and G93A saline-treated groups. (Scale bars: a and b, 18 mm.)

SI Figure 18

Fig. 18. Effects of lithium on ubiquitin immunostaining. Ubiquitin immunohistochemistry within Lamina IX of the lumbar spinal cord shows a mild immunopositivity in both saline- (a) and lithium-treated (b) wild-type mice. In contrast, a marked ubiquitin immunopositivity appears in G93A mice (c), where ubiquitin aggregates occur in neuronal cell bodies and axons (arrows). (d) Lithium treatment reduces ubiquitin immunostaining and eliminates ubiquitin from axons. (Scale bars: a-d, 25 mm.)

SI Figure 19

Fig. 19. Lithium reduces both alpha-synuclein and ubiquitin in the spinal cord of G93A mice. Representative Western blots and densitometric analysis of alpha-synuclein (a) and ubiquitin (b) expression in the cervical and lumbar tracts of WT and G93A saline- and lithium-treated mice. Lithium significantly decreased the expression of alpha-synuclein in the lumbar tract, whereas the expression of ubiquitin decreased in both the cervical and lumbar tracts. Values represent means ± SEM., analyzed by one-way ANOVA. *, P < 0.05 compared with WT saline-treated group; #, P < 0.05 compared with WT saline-treated mice and G93A lithium-treated mice; ##, P < 0.01 compared with WT saline-treated mice and G93A lithium-treated mice.

SI Figure 20

Fig. 20. Effects of lithium on SOD1 immunostaining. Immunostaining for SOD1 is lighter in wild-type (a and b) compared with G93A mice (c and d). In particular, G93A mice treated with saline show intense and aggregated SOD1 immunostaining (c), which, similarly to ubiquitin, was also found in the axon of motor neurons (arrow). This SOD1 accumulation was markedly cleared after lithium administration (d). (e) Representative Western blot and densitometric analysis of a triplicate of SOD1 expression shows an accumulation of the protein in G93A saline-treated mice, whereas this is suppressed after lithium administration. Values represent means ± SEM., analyzed by one-way ANOVA. *, P < 0.05 compared with WT saline-treated group; #, P < 0.05 compared with WT and G93A saline-treated groups. (Scale bars: a-d, 33 mm.)

SI Figure 21

Fig. 21. Effects of lithium on the structure of mitochondria from G93A mice. (a-c) Representative images of markedly altered G93A mitochondria showing the enlarged cristae typical of these organelles in ALS. (d-f) Vacuoles containing deranged mitochondria are shown from saline-treated G93A mouse (arrows); at higher magnification, in e and f, detachment of the inner and the outer mitochondrial membranes from the mitochondrial matrix is clearly visible (asterisks). (g-i) Show mitochondria from G93A mouse treated with lithium in which no vacuolization process is detected and both cristae and matrix are well arranged. The size of these mitochondria is approximately half that of the pathological organelles described in saline-treated G93A mice. Note also that myelinization of the axons is the well preserved (arrows, g and h). (Scale bars: a-c, 0.12 mm; d, 0,55 mm; e, 0.15 mm; f, 0.13 mm; g-i, 0.12 mm.)

SI Figure 22

Fig. 22. Lithium increases the mRNA levels of oxidative phosphorylation component in G93A spinal cord. mRNA Levels of citochrome-b and ATP synthase-6 genes (a) were normalized to the mRNA levels of the internal control GAPDH for each target gene for which the ratio of the level in lithium treatment versus saline treatment was determined. Mean ± SEM. values (n = 3 for lithium- and saline-treated G93A mice) *, P < 0.05 (Student's t test). Similar results were obtained with the 18S internal control (data not shown). (b) Mitochondrial mass was determined from the mitochondrial DNA/ nuclear DNA ratio using RT-PCR as described in Materials and Methods. Mean ± SEM values (n = 3 for each experiment.)

SI Figure 23

Fig. 23. Effects of lithium administration on mitochondria in motor neuron cell cultures and cell lines. (a) SH-SY5Y cells plated on sterile cover-slips were either not treated (saline) or were treated for 72 h with 1 mM lithium (changing the medium and readding the substance every 24 h), stained for 60 min with 300 nM MitoTracker Red or MitoTracker Green (InVitrogen) and observed under the confocal fluorescence microscope. Arrows point to cytoplasmic regions of high fluorescence intensity that indicate the accumulation of mitochondria. (b) Immunoblotting of cytochrome C (monoclonal antibody purchased from Alexis) in homogenates of SH-SY5Y cells treated or not (saline) for 72 h with 1 mM lithium. One blot, representative of three, is shown. Using densitometry, the amount of cytochrome C increased in treated versus nontreated cells by a factor of 2.5. (c) Flow cytometry analysis of control and lithium-treated (1 mM for 72 h) SH-SY5Y cells labeled with rhodamine-123 (100 nM for 20 min), a fluorescent dye that accumulates specifically in healthy mitochondria. Cytofluorimetry profiles show consistent differences between the two populations. In particular, the proportion of cells showing a higher rhodamine labeling (fluorescence >10² arbitrary units; M2) is increased by threefold upon treatment with lithium). Data shown are representative of five independent determinations. (d) Neuronal cultures were treated with 0.5 and 1 mM lithium for 18 h (the time interval used in all of the cell culture experiments), then MitoTracker Red was added to the culture for one hour (lithium and non treated (NT)). Labeled mitochondria were quantified on the bases of MitoTracker Red fluorescence in treated and NT cultures. (e) Data are expressed in arbitrary units as background subtracted fluorescence intensity, assessed by selecting three small regions without cells on the coverslip. Images were collected by using a confocal laser microscope (CLSM510 Zeiss) and were analyzed by using the ImageJ analysis program. The values represent means ± SEM., analyzed by one-way ANOVA after Newman-Keuls post hoc test. *, P < 0.05 compared with WT. (Scale bars: a, 17 mm; d, 40 mm.)

SI Figure 24

Fig. 24. Autophagy in spinal cord cell cultures. Representative pictures of SMI32- and GFAP-positive cells in WT and G93A mixed spinal cord cultures before (a and b) and after the autophagy blocker 3-methyladenine (3-MA) (5 mM; c and d) and lithium (1 mM; e and f) exposure. (g) Quantification of the SMI32-positive neurons in WT and G93A culture after drug treatment. 3-MA Treatment decreases the number of SMI32-positive neurons, and neither lithium nor rapamycine (which induces autophagy only upstream to the blockade induced by 3-MA) are able to counteract its toxicity. Noteworthy, the toxicity induced by the autophagy blocker 3-MA is significantly higher in G93A MN than WT MN. Similar to the effects obtained in vivo during the disease (see Fig. 3 and the summary at SI Fig. 16), gephyrin-positive cells increase after (1 mM) lithium exposure as shown by representative pictures (h and i) and measured in histogram (j). Gephyrin-positive neuron number is normalized to control (untreated cells, NT). Values represent means ± SEM. from at least four independent experiments, analyzed by one-way ANOVA after Spjotvoll/Stoline posthoc test. *, P < 0.05 compared with NT and G93A treated with lithium; #, P < 0.05 compared with G93A. (Scale bars: a-f, 200 mm; h and i, 75 mm).

SI Figure 25

Fig. 25. Lithium prevents kainate toxicity in G93A motor neurons. (a) SMI32- and GFAP-positive cells in G93A mixed spinal cord culture. Lithium (1 mM) protects SMI32-positive cells against kainate (100 mM) toxicity. The histogram (b) depicts the number of surviving cells assessed by direct counting of SMI32-positive cells, and normalized to its own control (untreated cells). Note that the extent of the protective effect of lithium was related to the presence of the mutation. Values represent means ± SEM. from at least four independent experiments, analyzed by one-way ANOVA after Spjotvoll/Stoline posthoc test. *, P < 0.05 compared with G93A treated with kainate+saline. (Scale bars: a, 200 mm.)

SI Movie 1

Movie 1. Movie shows a saline-treated G93A mouse (the first appear in the movie, a white mouse) with a severe alteration of limb reflex and severe gait impairment close to paralysis. The second G93A mouse (brown mouse) was treated with lithium and it shows only a slight gait alteration and a normal extension reflex. Finally, wild type mouse (black mouse) shows a normal limb extension reflex, which is compared with the G93A mouse (white), which was treated with saline.

Table 1. Sequences of primers used for PCR

Gene	Genome	Primer sequences (5'-to3')
GAPDH	Nuclear	Forward:ACCACAGTCCATGCCATCACT
		Reverse:TCCACCACCCTGTTGCTGTAG
18S	Nuclear	Forward:GCAATTATTCCCCATGAACG
		Reverse:CACCTACGGAAACCTTGTTAC
ATP synthase 	Nuclear	Forward:ACAGGACCCTATGTGCTTGG
		Reverse:TCTCCATGTCGATTGCATCC
ATP synthase 6	Mitochondria	Forward:CCTTCCACAAGGAACTCCAA
		Reverse:GGTAGCTGTTGGTGGGCTAA
Cytochrome b	Mitochondria	Forward:ATTCCTTCATGTCGGGACGAG
		Reverse:GGGATGGCTGATAGGAGGTT

The table provides the sequences of the various primers used for the comparative RT-PCR. Two different housekeeping genes (GADPH and 18S) have been used to validate the quantification. We used specific primers pairs for ATP synthase6 and cytochrome b as markers of the mitochondrial genome and ATP synthaseas marker of the nuclear genome and the housekeeping genes as well.

SI Materials and Methods I

Genetic Background and Breeding Protocol of G93A Mice. All of the experiments were carried out in compliance with the European Council Directive (86/609/EEC) for the use and care of laboratory animals. B6SJL-TgN(SOD1-G93A)1Gur mice expressing the human G93A Cu/Zn superoxide dysmutase (SOD1) mutation were obtained from the Jackson Laboratories. For details see SI Materials and Methods II.

Behavior. Behavioural observations were made by blind observers once a day for all animal groups (n = 20 per group). For details see SI Materials and Methods II.

Tissue preparation staining procedures and histological analyses. See SI Materials and Methods II.

Electron microscopy. Mice (n = 10 from each group, WT, WT plus lithium, G93A, G93A plus lithium) were perfused and spinal cords were maintained in situ immersed in fixative solution [2% paraformaldehyde/0.1% glutaraldehyde in 0.1 M PBS (PBS), pH 7.4] overnight at 4°C and then removed from the spine. For details see SI Materials and Methods II.

Primary neuronal cultures, immunocytochemistry, cell labeling and toxicity. Mixed spinal cord cultures were prepared from 13-day-old embryos of a control female mated with a G93A male as described in ref. 37. Three days after plating AraC (10 mM) was added. For details see SI Materials and Methods II.

SH-SY5Y cell lines and treatments. Human neuroblastoma SH-SY5Y cell line was obtained from the American Type Culture Collection (ATCC) and cultured under standard culture conditions. Cells were seeded and cultivated for 24 h before starting the treatments with 1 mM lithium carbonate (Sigma), 50 mM asparagine (Sigma), and 400 nM rapamycin. Treatments lasted 72 h, by changing the medium and readding the substances every 24 h. Detailed protocols of the mitochondria labeling, fluorescence assessment of autophagy, and cytochrome C Western blotting are described in the SI Materials and Methods II.

Statistical analysis. Data are given as mean ± SEM. Group mean values were compared by ANOVA, followed by posthoc testing.

Clinical Trial. Study design and patients. We conducted a 15-month, parallel-group, randomized study of adults with ALS, diagnosed according to the El Escorial revised diagnostic criteria (47), with a disease duration of less than 5 years.

The study protocol was approved by the Neuromed IRCCS Ethical Committee, and all subjects provided written informed consent. Initial statistical analysis determined that at least 40 subjects were needed to determine, with 95% confidence interval, a survival increase greater than 6 months.

The present study was performed on 44 patients, 20 male and 24 female. No familial case was present. Eleven patients presented the bulbar form of the disease, and the remaining the classic onset. Sixteen patients (8 male and 8 female, 4 of whom had the bulbar form) were randomly selected to receive riluzole (Rilutek 50 mg, 1 cp x 2/die) plus lithium (Carbolithium: two daily doses of lithium carbonate each of 150 mg), and the remaining (12 male and 16 female, 7 of whom had the bulbar form) riluzole only (48). In this way, we carefully matched lithium-treated and control patients for bulbar forms and FVC at the time of their inclusion in the study. In particular, the FVC values were 89 ±10 and 91 ± 10 for lithium treated and control patients, respectively. Again, the bulbar forms were distributed similarly between the groups (4/16 = 25%) in the treated group and (7/22 = 32%) for controls. One physician was not blind to group assignment; however, clinical evaluation, measurement of FVC and data analysis was conducted by other physicians who were blind to group identities (single-blind study). In this way, the first physician was able to monitor lithium concentration and to adjust the daily dose from 300 mg up to 450 mg daily when lithium plasma levels were below 0.4 mEq/liter. In fact, the daily dose was selected so as to reach a plasma range of 0.4 - 0.8 mEq/liter.

Compliance and adverse effects were monitored throughout the study period.

Subjects were assessed six times (at baseline, and every 3 months, for 15 months). The primary endpoint of the present study was the survival rate. The secondary outcomes measured changes in global function, as scored by the ALSFRS-R (49), a widely used and extensively validated functional scale for ALS (normal score: 48); and by the Norris ALS scale. This disability score includes evaluation of the functioning of upper and lower limbs, also taking into account bulbar function. This score uses 34 items rated with a value from 0 to 3; and the normal score is 100. Quality of life (Short Form Health Survey-36: SF-36; 50) was also evaluated. In parallel, we assessed the disease progression with more objective measures, such as quantitative segmental muscle strength (by the MRC scale) and the pulmonary function (forced vital capacity, FVC). The use of these combined approaches is very useful in small brief clinical trials (see SI Discussion for a comparison of reliability between different scales).

Data analysis.The case analysis used all subjects who entered in the protocol study. Statistical analysis included a descriptive analysis of each group's data. Kolmogorov-Smirnov test of normality was non significant for all variables. Measures were compared every 3 months with respect to the baseline condition using an ANOVA for repeated measures. Comparisons between the two groups' mean scores were made by using the unpaired t test and, if necessary, ANOVA, with Bonferroni as posthoc test. Survival analysis was performed by the Kaplan-Meier curve.

The null hypothesis was rejected when P £ 0.05 for all tests.

Statistical analyses were performed by using MedCalc software (version 9.3.6.0).

SI Materials and Methods II

Genetic Background and Breeding Protocol of G93A Mice. Selective breeding maintained the transgene in the hemizygous state in an F1 hybrid C57BL6xSJL genetic background (3). Screening for the presence of the human transgene was performed on tail tips. Each experimental group contained both lithium-treated and untreated littermates, results of the transgenic mice were compared to those of their wild-type littermates. All of the experiments were performed when the mice were at the tetraplegic stage, unless otherwise specified.

Behavior. Behavioural observations were carried out by a blind observer once a day for all animal groups (n = 20 per group). In particular, starting at 10:00 a.m. each mouse was monitored to document the impairment, the progression and the lost of the hind limb extension reflex; the scoring system was the following 0 = normal reflex; 1 = one limb impairment; 2 = both limb impairment. Mice were further analyzed for gait abnormalities and scored according to a severity scale ranging from 0 = normal gait; 1 = mild gait impairment; 2 = medium gait impairment; 3 = severe gait impairment; 4 = absence of gait (paralysis). The extension reflex was evaluated holding the mice by their tails suspended, gait impairment was revealed by spontaneous locomotion in the open field cage. Paralysis was defined as the completed loss of gait associated with the loss of the physiological muscle tone. For an example of paralysis and the extension reflex see supplementary movie.

Motor impairment, motor strength and coordination were evaluated with the following tests.

Rotarod. Motor impairment was evaluated with a rotarod (Columbus Instruments, Columbus OH). The test begun at the age of 90 days, was performed twice a week. Each mouse was placed on a rod rotating at 15 rpm, and the time the mouse stayed on the rod until it fell off (during a 5 min interval) was recorded.

Grip strength. To measure the limb strength the mouse was placed, twice per week, on a wire lid gently shaken to induce the mouse to grip the grid (19). Then the lid was turned upside down and the time the mouse remained hold on (generally with the forelimbs) was recorded. Animals unable to grip the grid received a score of zero.

Stride length. The stride test was performed with slight modifications accordingly to the method reported by Gong (20). The mice were allowed to explore the cage for 1 min and then they were left to move freely for 2 min. The mouse's hind limb was painted with the ink and the track it left was recorded. Stride length was measured in centimeters. Mouse unable to walk was scored as zero.

Tissue Preparation Staining Procedures and Histological Analysis. Mice were anaesthetized by using chloral hydrate and then perfused transcardially with saline solution, followed by 4% paraformaldehyde in 0.1 M phosphate buffer, pH 7.3. The brain and the spinal cord, at levels corresponding to both the cervical and lumbar tract (identified by the presence of an increased diameter), were dissected out and postfixed in the same fixative solution for 24 h and then placed in 70% ethylic alcohol overnight at 4°C. Samples were dehydrated in increasing alcohol solutions, immersed in xylene for several hours and finally embedded in paraffin. Tissue blocks were sectioned coronally using a microtome and 7-8-mm-thick slices were obtained.

For histological analysis, the sections were collected in strict anatomical order. In particular, every tenth section of the spinal cord and brainstem a slice was collected and mounted on polylysinated slides. The sections were dried at 37°C, immersed in xylene to remove the paraffin, rehydrated by immersion in decreasing alcohol solutions and then stained by Nissl method. After dehydration, slides were immersed in xylene, cover-slipped with DPX plastic mounting media (Sigma) and observed at light microscopy (Nikon Eclipse 80i). Proper identification of the brainstem nuclei was confirmed by using comparable tables based on stereotaxic coordinates of the Paxinos and Franklin mouse atlas (45).

To confirm the cytoarchitecture and neuronal morphometry, from each animal a similar set of slides from spinal cord and brainstem was stained by using Hematoxylin & Eosin (H&E). Moreover, the same procedure was used to collect and stain sections for immunohistochemistry against calbindin-28K, gephyrin, NeuN and alpha-synuclein (see below).

The neuronal loss in the ventral horn of the cervical and lumbar spinal cord, in the ambiguus, facial, hypoglossal nuclei and dorsal motor nucleus of the vagus was estimated by having two different observers, blind to the treatment count the number of neurons at 20x magnification. Within the ventral horn, motorneurons in the Lamina IX were identified by using strict criteria based on morphological features (multipolar cells with non condensed nucleus and a clearly observable nucleolus) and size exclusion, i.e., counting only those neurons exhibiting a diameter of at least 30 mm. This latter criterion allowed us to consider with high specificity the population of alpha-MN. Size exclusion was also used to identify interneurons within the Lamina VII. For this purpose, only neurons exhibiting a diameter between 10 and 20 mm were considered for counting, whereas beta-MN and glial cells were automatically excluded.

Measurement of the neuronal diameter was determined by an image analysis software (Molecular Machine & Industries AG), which allowed drawing the diameter of the cell under observation and automatically reading the related numerical value. Likewise, the area of the alpha-MN was measured by delineating the cell contour and reading the related numerical value.

Each observer's measurements of neuronal number, diameter and area from each mouse were used to obtain the mean value of each group. Comparisons among groups were made by using a one-way ANOVA combined with Scheffè's posthoc test. Null hypothesis was rejected for P < 0.05.

Immunohistochemistry. We used antibodies against the glial fibrillary acidic protein (GFAP) (1:100, Sigma), ubiquitin (1:100, Sigma), alpha-synuclein (1:50, gift from Dr. Thomas Sudhof, University of Texas, Dallas, TX), calbindin-D28K (1:500, Sigma), gephyrin (1:100, Santa Cruz), neuronal nuclei (1:10, NeuN, Chemicon), SOD1 (1:500, Stressgen Bioreagents), BrdU (1:500, Sigma).

For the immunoperoxidase procedure, the secondary biotinylated antibodies (Vector Laboratories) were used at a dilution of 1:200, followed by incubation with ABC kit and diaminobenzidine (Vector Laboratories). Fluorescein-conjugated secondary antibodies (Vector Laboratories) were used at a dilution of 1:100, whereas the Cy3-conjugated secondary antibody (Chemicon) was used at a dilution of 1:400. The sections were observed by using a light microscope (Nikon Eclipse 80i).

For BrdU experiments, 90-days-old mice were injected with BrdU (50 mg/kg i.p.) every other day for 10 days. Mice were then anaesthetized, perfused and their spinal cord were dissected for immunohistochemistry. To denaturate DNA, sections were immersed in HCl 2 N for 20 min; then they were washed out in borate buffer (pH 8.5) and processed as described above with a fluorescein-conjugated secondary antibody.

Spinal Cord Tissue Immunoblotting and Densitometric Analysis. Cervical and lumbar spinal cords from each experimental group were homogenized. The membranes were incubated with primary antibody anti-alpha-synuclein (1:400, gift from Dr. Thomas Sudhof, University of Texas, Dallas, TX), -ubiquitin (1:500, Sigma), -gephyrine (1:500, Santa Cruz), -NeuN (1:500, Chemicon), -GFAP (1:1000, Sigma), -calbindine-D28K (1:1000, Sigma), or SOD1 (1:2000, Stressgen Bioreagents) at 4°C overnight. Quantification of immunoreactive bands was performed by using densitometry as described in ref. 46. The results were confirmed by duplication or triplication.

Electron Microscopy. Cervical tract and lumbar tract were dissected and postfixed in 1% OsO₄ in buffered solution, dehydrated in ethanol and embedded in Epon-araldite. Ultrathin sections obtained from anterior horns of both cervical and lumbar tract, were stained with uranyl acetate and lead citrate, and finally examined by Jeol Jem 100SX transmission electron microscope (Jeol). For each samples from cervical and lumbar tract we randomly chose two tissue blocks that we cut to obtain 20 grids for each blocks; each grid contained several nonserial sections to observe different neurons and count a final number of 5 cells per grids. The selected number of grids allowed us to analyze about 200 neurons for both cervical and lumbar tract. We selected alpha-motor neurons on the base by strict criteria described above (multipolar cell with not condensed nucleus and well evident nucleolus) and size exclusion, consisting of counting only those neurons exhibiting a diameter of at least 30 mm. Micrographs at 8,000 x from each neuron were taken to count the total number of mitochondria and determine the mitochondrial minimum and maximum diameter. Counts of mitochondrial minimum and maximum diameter were analyzed by using ANOVA with Scheffé's test for posthoc analysis.

DNA/RNA Extraction and Quantification of Mitochondrial DNA by Comparative PCR. WT and G93A male mice were treated with lithium carbonate (1mEq/kg) or saline starting at 75 d of age for 20 days, at the end of the treatment spinal cords were dissected and homogenized. The isolation of the DNA/RNA was performed by using Trizol reagent, following the manufacturer instructions (Invitrogen). The amounts of mitochondrial and nuclear DNA were determined by PCR using specific primer pairs for cytochrome b, as marker of the mitochondrial genome and ATP synthase subunit-beta as marker of nuclear genome. For the semiquantitative mRNA assay GAPDH and the S18 were used to determine the relative levels of the target mitochondrial mRNAs. The sequences of the primers are provided in the Table 1. Sequencing of the PCR amplicons confirmed their correctness.

Primary Neuronal Cultures, Immunocytochemistry, Cell Labeling and Toxicity. Immunocytochemistry on neuronal cultures was performed as described in ref. 37. Primary antibodies were mouse monoclonal SMI32 (1:6000; Sternberger Monoclonals), rabbit polyclonal GFAP (1:1000; Chemicon) and gephyrin (1:250; Santa Cruz). For spinal cord, 12-14 DIV cultures were used for toxicity experiments. 3-methyladenine (3MA, Sigma) was added to cultures to final concentration 5 mM for two hours at 37°C 5% CO₂. After lithium (1 mM, Sigma) or rapamycin (0.5 mM) were added for 18 h, than cell survival was assessed by counting SMI32 or gephyrin neurons. Kainate exposure (100 mM, 15 min) was performed as described in ref. 37. Lithium (1 mM) was added 20 min before kainate exposure. After lithium incubation, each dish was loaded with 20 nM the MitoTracker Red 580 (Invitrogen) for 1 h at 37°C. Then the cells were fixed in 4% paraformaldehyde for immunofluorescence. Digital images were taken in Leica SP5.

Experiments on Human Neuroblastoma SH-SY5Y Cell Line. Human neuroblastoma SH-SY5Y cell line was obtained from the American Type Culture Collection (ATCC) and cultured under standard culture conditions (37°C; 95% air:5% CO₂) in 50% MEM and 50% F12 nutrient medium supplemented with 10% heat-inactivated FBS (Invitrogen), 2 mM L-glutamine and 1% of a penicillin-streptomycin solution. Cells were seeded and cultivated for 24 h before starting the treatments with 1 mM lithium (Sigma), 50 mM asparagine (Sigma), and 400 nM rapamycin. Treatments lasted 72 h changing the medium and readding the substances every 24 h.

Mitochondria Staining. Mitochondria were labeled by using MitoTracker Red CMXRos and MitoTracker Green FM (Invitrogen). Cells plated on coverslips were incubated with 300 nM mitotracker solution for 1 h at 37°C, washed and then fixed in 37% paraformaldehyde 4% for 30 min. Stained cells were observed with the Leica DMIRE2 confocal fluorescence microscope (Leica Microsystems AG) equipped with Leica Confocal Software v. 2.61. Quantification of mitochondria was carried out by cytofluorimetric analysis on the basis of the fluorescence of rhodamine, a dye specifically retained within mitochondria. At the end of incubation, cells were collected, washed in PBS and incubated for 20 min at room temperature with 100 nM rhodamine-123 (Sigma). At least 10,000 cells were analyzed in a FacScan flow cytometer (Becton Dickinson) equipped with a 488 nm argon laser. Data were interpreted with the winMDI software.

Fluorescence Assessment of Autophagy. Formation of autophagosomes was directly monitored in living cells stably transfected with the expression vector pEGFPC2 (CLONTECH laboratories) encoding the fluorescence chimeric protein GFP-LC3, in which GFP was placed at the N terminus of LC3.

Western Blotting. Cells were homogenized in a buffer containing detergents and protease inhibitors. Fifty mg of cell proteins were denatured with Laemmli sample buffer, separated by electrophoresis on a 12.5% polyacrylamide gel, and then electroblotted onto nitrocellulose membrane (Bio-Rad). The same filter was subsequently probed for Cytochrome C and b-Actin, which were detected by specific monoclonal antibodies purchased, respectively, from Alexis Laboratories and Sigma. Immunocomplexes were revealed by using a peroxidase-conjugated secondary antibody and subsequent peroxidase-induced chemiluminescence reaction (Perkin-Elmer). Intensity of the bands was estimated by densitometry (Quantity One software).

Stereological Analysis. For each mouse, counts of the Lamina IX motor neurons and Lamina VII interneurons were made on a total of 125 serial, nonconsecutive, Nissl-stained slices, spaced 80 mm apart, at the level of the cervical or lumbar spinal cord.

This procedure allowed us to examine different nonoverlapping neurons present within up to 1 mm-length segment of spinal cord tissue.

The exact identification of the cervical and lumbar tract of the spinal cord was based on the presence of the corresponding enlargements of the spinal cord diameter.

The neuronal counts in the ambiguus, facial, hypoglossal nuclei and dorsal motor nucleus of vagus were carried out in serial, nonconsecutive slices, spaced about 50 mm. For the exact identification of each nucleus we used the Paxinos & Franklin Atlas (2004). In particular, for each nucleus we analyzed the sections within the following AP stereotaxic coordinates (expressed in mm posterior to the bregma):

a) 6.70-8.00 for the ambiguus nucleus;

b) 6.50-7.90 for the dorsal motor nucleus of the vagus nerve;

c) 7.00-8.12 for the hypoglossal nucleus;

d) 5.70-6.50 for the facial nucleus

Immunoelectron Microscopy. After plain electron microscopy two tissue blocks of cervical and lumbar tract from mice belonging to each group (WT, WT plus lithium, G93A, G93A plus lithium) were randomly chosen. We cut ultrathin sections on nickel grids to carry out immunocytochemistry using the postembedding method.

Ultrathin sections were deosmicated with saturated aqueous solution of Na-metaperiodate, after washing in phosphate buffer they were incubated for 24 h at 4°C with primary antibodies in buffer solution A (PBS, 1% goat-serum, and 0.2% saponin). All primary antibodies (beclin, calbindin 28K LC3 and BrdU) were diluted 1:10. After washing in phosphate buffer, sections were incubated with gold-conjugated secondary antibodies (Chemicon, gold particles, 10-15 nm) diluted 1:50 in buffer solution A for 1 h, at room temperature. Finally, sections were fixed with 1% glutaraldehyde, stained with uranyl acetate and lead citrate and examined under transmission electron microscope (TEM).

We processed 20 grids for each block of cervical and lumbar tract from mice belonging to each group. Ten grids from each block were processed for beclin and ten for LC3 immunoelectronmicroscopy. Every grid contained several nonserial sections to observe different neurons and count a final number of 5 cells per grid. The selected number of grids allowed us to observe about 100 detectable neurons for both cervical and lumbar tract and for the two different antigens. We selected alpha-motor neurons that had the following characteristics: multipolar cell with non condensed nucleus and clearly observable nucleolus, exhibiting a diameter of at least 30 mm. Several micrographs at 15,000´ from each neuron were taken to allow us to cover the total cytoplasmic area and count the total number of structures positive for beclin or LC3 present in the cytoplasm. Additional experiments were carried out to count the percentage of ImmunoGold particles corresponding to BrdU incorporation into the nucleus of cells within the Lamina VII. Lamina VII was identified by means of toluidine-stained ultrathin (1 mm) sections. Several micrographs were taken at 53,000 x from each neuron within Lamina VII to allow us to count the number of neurons containing BrdU and calbindin 28K ImmunoGold particles. Morphological data were analyzed by using ANOVA with Scheffé's test for post hoc analysis.

SI Discussion

-1- A Succint Comparison of the G93A Phenotype and ALS in Human Patients. The G93A phenotype has a rather stereotyped clinical course, which involves the hind limbs first and most severely and leads to gait alterations and paralysis. There is slighter involvement of the forelimbs, which leads to a defect in grasping. Recently, a defect of pulmonary function was measured but this is not leading to death. In fact, when the mice are left without artificial feeding they die because of the loss of locomotion (paralysis), which in turn, depends on the absence of hind limbs motor activity. This markedly contrasts with the variety of ALS in human patients, which may have different clinical onset and severity, starting either as a deficiency in the lower limbs, the upper limbs, the brainstem (bulbar form), or the motor cortex. In human patients, mainly the respiratory failure causes death.

This clinical difference is reflected by neuropathology. At the end of the disease, the pathology of the G93A mouse model is most severe in the lumbar spinal cord, whereas the cervical spinal cord and the motor nuclei of cranial nerves, although involved, are less affected. In fact, as mentioned above, only recently have findings been reported on the breathing dysfunction in G93A mice (5), it mimics what occurs in the bulbar involvement of ALS patients, but is not as severe and does not lead to death.

The neuropathology of the motor cortex in humans depends on the specific subtype of ALS, although there is a consistent and well established neuronal loss in various motor cortical areas. In contrast, data on the involvement of the motor cortex in G93A mice is under debate. An elegant study by Beal's group found an early energetic impairment in the cortex of G93A mice, but did not address the occurrence of cell loss and suggested a functional impairment of the cortico-spinal tract (6). A recent article by Chung et al. (4) reported careful morphometric measurement of all cortical neurons in G93A mice compared with their controls and found no difference in neuron number; however, a different expression of the protein calretinin was present. This article does not confirm what was reported five years ago in an article investigating descending fibers (3). Although evidence of a certain amount of cortico-spinal degeneration is reported in this work, it is much less severe compared with the involvement of the lumbar spinal cord (3). Even considering the susceptibility of motor neurons in the spinal cord to glutamate in the G93A mice, this phenomenon does not seem to occur in the motor cortex (1). Another fact, which adds evidence to the differential vulnerability of spinal compared to cortical motor neurons concerns the toxic role of glial factors. Although the latter have a detrimental effect on the survival of spinal motor neurons of G93A mice, they do not produce such an effect in their motor cortex (2).

As in the human disease, the G93A phenotype seems to spare the oculomotor nuclei to a large extent, but this has been poorly investigated. Although in the present work these motor nuclei appeared spared, neuropathology was evident in nucleus ambiguus, nucleus facialis, and hypoglossal nucleus. In particular, in SI Fig. 10, we provide evidence of the neuropathology affecting nucleus ambiguus.

At the intracellular level, the human disease presents typical neuronal inclusions that vary from Lewy-like bodies to Bunina bodies and ubiquitin aggregates. The mouse model shows many similarities with human findings (7); in the present work, we report the accumulation of ubiquitin, SOD 1 and alpha synuclein, which are cleared by lithium administration. As expected, the similarity is not complete; for instance, a recent article by Robertson et al. (8) failed to replicate in the mouse model the typical inclusions, featuring the simultaneous presence of ubiquitin and TDP-43 that characterize the human ALS.

However, the mitochondrial defect and the vacuolization of motor neurons, which are typical and consistent in G93A mice, are less documented in the human pathology; but, this seems to be an effect of the paucity of studies addressing these targets. In summary, the most robust difference between the mouse model and the human disorder both at clinical and pathological level consists of the stereotypical impairment of the lumbar spinal cord/hindlimb strength in mice compared with the variability in the temporal sequence and severity among the different motor areas in humans ALS.

Therefore, it is not surprising that, in the mouse model, when a neuroprotective agent is analyzed at the end of the disease (when the mice are dying), it does not show preservation of the number of lumbar motor neurons. However, when the motor neurons are counted at an earlier time interval (during the disease course), the number of lumbar motor neurons is higher in those mice under neuroprotective treatment. This confirms that the lifespan and, most importantly, the disease duration in treated mice is markedly increased by the neuroprotective agent. In fact, this is what we found after lithium administration. In the terminal stage (when the critical loss of lumbar motor neurons takes place), which occurs later in lithium treated mice, the number of cervical motor neurons or motor neurons of the cranial nerves is still significantly higher in lithium treated mice, as expected from the delayed degeneration occurring in these areas (which corresponds to the comparison made at earlier time points for the lumbar spinal cord).

-2- The Potential Role of Interference with Glial Activation. It was recently reported that the activation of glial cells is a key factor in sustaining motor neuron death (2). In fact, astrocytes carrying the SOD1 G93A mutation are able to release neurotoxins for motor neurons. One key finding of the present article is the significant attenuation induced by lithium administration of GFAP immunoreactivity in Lamina IX of the spinal cord (SI Fig. 11). This phenomenon persists even at the end of the disease when the number of lumbar motor neurons is maximally reduced. Thus, in the presence of a comparable motor neuron loss in the lumbar spinal cord (occurring at a more prolonged lifespan in lithium-treated mice), the activation of astrocytes appears to be attenuated in G93A mice receiving lithium. This effect suggests that a specific interference during the disease course might underlie the therapeutic effects. In fact a recent article by Gilad and Gilad (9) demonstrates that lithium at doses of 1 mM (comparable to those used in the present study) enhances neuronal survival, whereas it inhibits astroglial growth. In particular, these authors found that lithium treatment results in growth retardation and altered cell morphology of cultured astroglia, suggesting that direct effects on astrocytes and microglia may contribute to the neuroprotective effects of lithium on neurons. Again, these effects may contribute to the increase in Lamina VII interneurons, which contain calbindin 28K; in fact lithium was shown to promote selectively the neurogenesis of this subclass of neurons in the hippocampus (10). This effect, together with the inhibition of the growth of astrocytes (9) may contribute in leading neuronogenesis toward calbindin positive neurons while suppressing astrocytes formation. Further, while we were preparing this manuscript a report was published that lithium was able to promote neuronogenesis in vitro at a dose (1 mM) similar to the one we used in the present study (11). Interestingly, in vivo, chronic (4 weeks) administration of lithium produced a depression in the amount of microglia and macrophage activation, thus improving the survival of neuronal progenitor cells, as measured by BrdU immunostaining in the present study and by Su et al. (11). These effects of lithium are in line with other recent evidence (12) showing that improvement of autophagy via the autophagy inducer rapamycin, suppresses the microglial response, thus providing neuroprotection in traumatic brain injury. In fact, autophagy induction suppresses the immune response, which plays a detrimental effect in the spinal cord both via microglial cells and astrocytes. This effect on immunosuppression and inhibition of microglial activity (as expected from autophagy activation) could play a pivotal role in the neuroprotection provided by lithium in the spinal cord in our experimental setting, and in spinal cord injuries. In line with this, Brunet et al. (13) found that microglia surrounds dying motor neurons in the presence of large autophagic vacuoles, suggesting that inefficient autophagy triggers microglia activation. Thus, the detrimental role of glial cells in ALS is likely counteracted by these effects of lithium as autophagy inducer. Likewise, the survival of newly formed neurons and their guidance toward the calbindin 28K phenotype in the presence of suppressed astrocyte proliferation is well documented and supports what we observed in the spinal cord of G93A mice. These observations are fascinating and suggest the importance of investigating the effects of lithium in other degenerative conditions of the spinal cord, including the acute neuronal loss, which follows spinal trauma.

-3- The Significance of Increased Neuron Number After Lithium in G93A Mice. The present study was not aimed to analyze the issue of stem cells proliferation and differentiation in the spinal cord. However, when we examined the neuronal population mostly affected in the G93A mice, we found that it corresponded to those neurons in the Lamina VII of the spinal cord that stain for gephyrin and calbindin 28K and NeuN. When we administered lithium to G93A mice, we found a striking effect on these neurons, which were not simply preserved but increased above the number counted in control mice (wild-type treated with saline). Interestingly, lithium did not produce such an effect when it was administered to wild type mice. Thus, the net increase was due to both the disease state and the effects of lithium. Recent reports demonstrate that a chronic disease state like ALS is able to increase the number of neuronal progenitor cells (NPC) in different parts of the spinal cord (see for instance ref. 15). This is not specific to ALS but also occurs after a spinal trauma (18). All these conditions lead to increased incorporation of BrdU in the spinal cord. These new cells, which appear in the spinal cord after a spinal trauma (18) or during the course of ALS (15) follow established pathways of differentiation toward the glial phenotype. Thus, although the occurrence of newly formed neurons is also described (16, 17), the fate of the newly dividing NPC toward the glial phenotype was clearly established by Yang et al. (18) after spinal cord trauma and by Guan et al. (15), in a study published few weeks ago, during the course of ALS. These recent reports do not provide conclusive evidence on the significance and potential neurorescue of increased NPC during the course of ALS (compare the work in refs. 16 and 17 with the work of ref. 15). Thus, compared with controls, in these conditions there is no net increase in neuron number in the affected spinal cord. Within this scenario, our administration of lithium during the course of ALS carried out in the present study is critical. In fact, we measured an increase of BrdU incorporation in ALS mice treated with saline or lithium (SI Fig. 14 and 15). However, only ALS mice treated with lithium expressed a net increase in neuron number even when compared with saline-administered wild type mice (for a summary see SI Figs. 14 and 15). Interestingly, most of the BrdU incorporation we measured in lithium-treated G93A mice, occurred within calbindin 28K positive cells (SI Figs. 14 and 15), which correspond to those neurons we found augmented in the spinal cord of ALS mice treated with lithium. Providing support for this effect, recent reports demonstrate that lithium administration suppresses differentiation of NPC toward glial cells, while it promotes their differentiation into calbindin 28K-containing neurons (9, 10). This is in line with the other effects we described after lithium administration. In fact, as an activator of autophagy lithium reduces the glial response [as reported by Su et al. (11) and described in this manuscript]. It is intriguing that both in our ALS lithium-treated mice and in rats undergoing lithium administration after a spinal cord injury (11) the increased amount of NPC was directed toward the formation of calbindin 28K-containing neurons, whereas the glial differentiation was suppressed. This may account for the net increase in the number of calbindin 28K-containing neurons and for the neuroprotective effects of lithium due to the suppression of the glial response (see previous paragraph of the SI Discussion). The occurrence of increased BrdU incorporation during the course of the disease supports the concept that the net increase in neuron number we described in the Lamina VII of lithium-treated G93A mice is due to the preferential differentiation of NPC into Renshaw-like cells (see the similar neuron number counted by using a size-exclusion criteria, the calbindin and gephyrin positivity, NeuN and H&E staining and the colocalization of BrdU within calbindin positive neurons, as shown by light and electron microscopy). However, we think that further experiments, specifically aimed at solving the issue of the effects of lithium on the various differentiation steps of NPC in the spinal cord are needed. In fact, this topic is beyond the scope of the present article.

-4-Integrated Perspective of Concomitant Mechanisms. It is our hypothesis that the net increase in neuron number we observed in the present research together with the neurogenesis promoted by lithium in other brain areas (i.e., hippocampus) and the occurrence of neuronogenesis without increase in the net number of neurons in the G93A model reported by Chi et al., in Stem Cell Dev. (16) and Guan et al. (15), leads to hypothesize that the concomitant administration of lithium in the pathological condition of G93A mice ongoing neurodegeneration become a synergistic stimulation in which neurogenesis occurs at a level that surpasses the number of neurons we counted in the Lamina VII of saline-treated wild type mice. This effect, is typical of Lamina VII and occurs in Renshaw-like cells. Interestingly, these same neurons are more affected in the G93A mice than motor neurons; pioneer studies in humans seem to confirm such data. The colocalization of BrdU particles within these calbindin positive neurons confirms this hypothesis and supports the idea that lithium promotes a differentiation of NPC toward calbindin positive neurons while suppressing their glial differentiation. This leads to the hypothesis that the tonic inhibition of motor neurons and the counteracting effect on glutamatergic stimulation by the inhibitory effect of Renshaw cells is lost in ALS. This may be the event that makes the motor neurons in ALS more susceptible to glutamatergic toxicity. In fact the present article demonstrates that lithium abolishes the kainate-induced toxicity in a sort of disease-related manner (SI Fig. 24). Again, the clinical interpretation of fasciculation (i.e., the spontaneous contraction of single muscle motor units), which was traditionally interpreted as a dysfunctional message arising in suffering motor neurons, may well be due to the loss of the recurrent inhibitory circuit that is represented by Renshaw cells which disappear early. In fact, fasciculation is an early sign of the onset of ALS. We added such an alternative explanation enriched with the discussion of the role of glia as a fascinating hypothesis on multiple mechanisms, which remain to be investigated.

-5- Outcome Measures for Early Phase Clinical Trials. The significance of the effects of lithium in delaying the progression of ALS relies on the use of the common measures for early phase clinical trials, as reported very recently (14). In this study the authors compared ALSFRS-R, FVC measurement of muscle strength and the scale for the quality of life. In their validation study, these authors used linear mixed effects models to assess the associations among variables and Cox proportional-hazards models to examine the ability to predict survival. They found that all of the measurement correlate with each other, although the ALSFRS-R was the most effective in predicting survival (P = 0.002). In the present study, data obtained from the ALSFRS-R were complemented by FVC, segmental muscle strength measurement, and quality of life, and consistency was found along the different measurements. The classic, less update Norris scale was also used, and the results were in line with these measurements.

We believe that, despite concordance, this multiple approach is mandatory when brief small trials are performed. The one reported by Gordon et al. (14) referred to a 6 months trial, whereas the present study refers to 15 months. However, it does not cover the total disease duration and remains a small clinical trial that benefits from the use of combined approaches to assess disease progression.

This need is more and more urgent to be able to address those treatments, which can be useful to be continued on a large number of patients to assess the efficacy of potential neuroprotective therapies.

1. Tovar-y-Romo LB, Tapia R (2006) Cerebral neurons of transgenic ALS mice are vulnerable to glutamate release stimulation but not to increased extracellular glutamate due to transport blockade. Exp Neurol 199(2):281-290.

2. Nagai M, Re DB, Nagata T, Chalazonitis A, Jessell TM, Wichterle H, Przedborski S (2007) Astrocytes expressing ALS-linked mutated SOD1 release factors selectively toxic to motor neurons. Nat Neurosci 10(5):615-622.

3. Zang DW, Cheema SS (2002) Degeneration of corticospinal and bulbospinal systems in the superoxide dismutase 1(G93A G1H) transgenic mouse model of familial amyotrophic lateral sclerosis. Neurosci Lett 332(2):99-102.

4. Chung YH, Joo KM, Nam RH, Cho MH, Kim DJ, Lee WB, Cha CI (2005) Decreased expression of calretinin in the cerebral cortex and hippocampus of SOD1G93A transgenic mice. Brain Res 1035(1):105-109.

5. Tankersley CG, Haenggeli C, Rothstein JD (2007) Respiratory impairment in a mouse model of amyotrophic lateral sclerosis. J Appl Physiol 102(3):926-932.

6. Browne SE, Yang L, DiMauro JP, Fuller SW, Licata SC, Beal MF (2006) Bioenergetic abnormalities in discrete cerebral motor pathways presage spinal cord pathology in the G93A SOD1 mouse model of ALS Neurobiol Dis 22(3):599-610.

8. Sasaki S, Warita H, Murakami T, Shibata N, Komori T, Abe K, Kobayashi M, Iwata M (2005) Ultrastructural study of aggregates in the spinal cord of transgenic mice with a G93A mutant SOD1 gene. Acta Neuropathol (Berl) 109(3):247-255.

7. Robertson J, Sanelli T, Xiao S, Yang W, Horne P, Hammond R, Pioro EP, Strong MJ (2007) Lack of TDP-43 abnormalities in mutant SOD1 transgenic mice shows disparity with ALS Neurosci Lett 420(2):128-132.

9. Gilad GM, Gilad VH (2007) Astroglia growth retardation and increased microglia proliferation by lithium and ornithine decarboxylase inhibitor in rat cerebellar cultures: Cytotoxicity by combined lithium and polyamine inhibition. J Neurosci Res 85(3):594-601.

10. Kim JS, Chang MY, Yu IT, Kim JH, Lee SH, Lee YS, Son H (2004) Lithium selectively increases neuronal differentiation of hippocampal neural progenitor cells both in vitro and in vivo. J Neurochem 89(2):324-336.

11. Su H, Chu T-K, Wu W (2007) Lithium enhances proliferation and neuronal differentiation of neuronal progenitor cells in vitro and after transplantation into the adult rat spinal cord. Exp Neurol 206:296-307.

12. Erlich S, Alexandrovich A, Shohami E, Pinkas-Kramarski R (2007) Rapamycin is a neuroprotective treatment for traumatic brain injury. Neurobiol Dis 26(1):86-93.

13. Brunet N, Tarabal O, Portero-Otin M, Oppenheim RW, Esquerda JE, Caldero J (2007) Survival and death of mature avian motoneurons in organotypic slice culture: trophic requirements for survival and different types of degeneration. J Comp Neurol 501(5):669-690.

14. Gordon PH, Cheng B, Montes J, Doorish C, Albert SM, Mitsumoto H (2007 Jul) Outcome measures for early phase clinical trials. Amyotroph Lateral Scler 27:1-4.

15. Guan YJ, Wang X, Wang HY, Kawagishi K, Ryu H, Huo CF, Shimony EM, Kristal BS, Kuhn HG, Friedlander RM (2007) Increased stem cell proliferation in the spinal cord of adult amyotrophic lateral sclerosis transgenic mice. J Neurochem 102(4):1125-1138.

16. Chi L, Gan L, Luo C, Lien L, Liu R (2007) Temporal response of neural progenitor cells to disease onset and progression in amyotrophic lateral sclerosis-like transgenic mice. Stem Cells Dev 16(4):579-588.

17. Chi L, Ke Y, Luo C, Li B, Gozal D, Kalyanaraman B, Liu R (2006) Motor neuron degeneration promotes neural progenitor cell proliferation, migration, and neurogenesis in the spinal cords of amyotrophic lateral sclerosis mice. Stem Cells 24(1):34-43.

18. Yang H, Lu P, McKay HM, Bernot T, Keirstead H, Steward O, Gage FH, Edgerton VR, Tuszynski MH (2006) Endogenous neurogenesis replaces oligodendrocytes and astrocytes after primate spinal cord injury. J Neurosci 26(8):2157-2166.

19. Shin HJ, Cho SI, Lim RH, Lee JK, Lee AY, Noh SJ, Joo SI, Kim KW, Gwag JB (2007) Concurrent administration of Neu2000 and lithium produces marked improvement of motor neuron survival, motor function, and mortality in a mouse model of amyotrophic lateral sclerosis. Mol Pharmacol 71:965-975.

20. Gong YH, Parsadanian AS, Andreeva A, Snider WD, Elliott JL (2000) Restricted expression of G86R Cu/Zn superoxide dismutase in astrocytes results in astrocytosis but does not cause motoneuron degeneration. J Neurosci 20:660-665.