System ${\text{x}}_{\text{C}}^{-}$ is a mediator of microglial function and its deletion slows symptoms in amyotrophic lateral sclerosis mice

Animals

Mice were hSOD1^G85R, hSOD1^G37R (Boillee et al., 2006b), control hSOD1^WT [C57Bl/6-Tg(SOD1)3Cje/J, Jackson Laboratory], and mice deleted for the Slc7a11 gene (xCT^−/− mice) (Sato et al., 2005), all on a C57BL/6J background (Janvier). All procedures were performed in accordance with the guidelines for care and use of experimental animals of the European Union and approved by the ethics committee for animal experimentation n°5 in Ile-de-France.

Survival analysis

A two-step mating strategy gave hSOD1^G37R:xCT^−/− (n = 21), hSOD1^G37R:xCT^+/− (n = 35) and hSOD1^G37R:xCT^+/+ (n = 24) mice, which were weighted weekly as an objective and unbiased measure of disease course (Boillee et al., 2006b; Lobsiger et al., 2009) and their grip strength measured (Bioseb, grip test). The three groups of mice started with comparable weight and grip-strength means.

Primary microglial cell culture

Cells were derived from postnatal Day 0 or 1 mouse pup cerebral cortex according to Thery et al. (1991). Highly pure microglia (>99%, quantified after immunostaining with microglial specific antibodies CD11b/Itgam, F4/80/Emr1 and Hoechst 33342 nuclear staining dye) were plated for the different assays and their survival measured. For immunostaining anti-xCT (Novus Bio; 1:5000), anti-CD11b (Serotec, 1:400) and F4/80 (Serotec, 1:100) antibodies were used. Fluorescence signal (integrated density) was measured per cell and compared to the control condition (xCT^+/+ microglia without lipopolysaccharide treatment), using ImageJ software (n = 3 experiments per genotype).

Glutamate assay

Glutamate released (for 30 h after adding lipopolysaccharide, 20 000 cells/well, n = 3–4 experiments per genotype and treatment) was measured with the glutamate dehydrogenase-based colorimetric assay of Beutler (1985).

Nitric oxide assay

Nitric oxide production was assessed by measuring nitrite levels (a stable by-product of nitric oxide) with the colorimetric Griess method for 50 000 cells/well (n = 3 experiments per genotype and treatment).

Luminex assay

Microglia were plated at a density of 5 × 10⁴ cells/96-well-plates (5–6 wells/condition, n = 3–4 experiments per genotype and treatment). All samples (25 µl of medium) were run in duplicates with Milliplex Map kits and analysed with the Magpix system (Life Technologies).

Glutathione assay

Total glutathione levels were measured in the spinal cord of 1-year-old mice using the QuantiChrom Glutathione Assay Kit (BioAssay Systems).

RNA extraction and real-time PCR

RNA extraction for tissues and cells was performed with Qiagen RNeasy® Kits (Qiagen). Reverse transcription was performed with SuperScript® III (Life Technologies) using 500 ng of RNA for spinal cord tissues, 250 ng for cultures and human spinal cord tissues, and 20 ng or 2 ng for laser-microdissected motor neurons and adult mouse spinal cord microglia, respectively. Quantitative PCRs were performed with SYBR® Green Master Mix (Applied Biosystems).

Laser microdissection of motor neurons

Motor neuron laser microdissection was performed as previously described (Lobsiger et al., 2007). From one lumbar (L4–L6) spinal cord, 350 motor neurons were collected yielding ∼20 ng of RNA.

Isolation of microglia from adult mouse spinal cord

Microglial cells were isolated as described by De Groot et al. (2000), followed by purification with anti-CD11b microbead-coupled antibodies and Miltenyi MS columns. Approximately 7–10 × 10⁴ CD11b⁺ cells were recovered from one control spinal cord.

Immunohistochemistry on spinal cord tissues

Thirty-micrometre paraformaldehyde fixed cryosections were stained with: anti-Iba1 (Wako Chemicals, 1:500), anti-Mac-2 (Cederlane, 1:500), anti-GFAP (Dako, 1:4000) and anti-ChAT (Millipore, 1:200) for n = 3–4 animals per genotype and per time point.

Motor neuron counts

Motor neuron numbers were determined from 30-μm serial sections across the entire lumbar spinal cord and counted in every 12th ChAT stained section corresponding to a total of 23–26 sections per animal for n = 3 animals per genotype and time point.

Human samples

Patients with ALS were enrolled in a brain donation programme declared to the Ministry of Research and Universities, as requested by French Law. An explicit consent was signed by the patient, or by the next of kin, in the name of the patient, in accordance with the French Bioethical Laws.

Immunoblots

Proteins (16 µg) were separated on NuPAGE™ 4–12% Bis–Tris Gel (Invitrogen) and electrophoretically transferred to nitrocellulose membranes stained with anti-xCT (Abcam; 1:1500) and anti-beta-actin (Sigma; 1:6000) antibodies.

Statistical analysis

Two-way ANOVA tests followed by a Bonferroni multiple comparison test were used to compare groups with different genotypes and different treatments (with or without lipopolysaccharide). One-way ANOVA tests followed by a Bonferroni multiple comparison test were used to compare multiple groups, Student’s t-test to compare means of two groups and Kaplan-Meier Log Rank test for survival curves.

Primary microglial cells express system ${\text{x}}_{\text{C}}^{-}$ and xCT levels are increased on microglial activation

To study the implication of system ${\text{x}}_{\text{C}}^{-}$ and microglial glutamate release in ALS, we first studied the expression of system ${\text{x}}_{\text{C}}^{-}$ in primary mouse microglial cells. The mRNA for xCT, the specific subunit of system ${\text{x}}_{\text{C}}^{-}$, was expressed by control (xCT^+/+) microglia and strongly increased (18-fold) when cells were activated with lipopolysaccharide, the most commonly used factor to activate microglia (Fig. 1A). At a protein level, xCT was present in the cytoplasm and on the plasma membrane of xCT^+/+ microglia (with the signal absent in xCT^−/− cells) (Fig. 1B) and likewise strongly increased (3-fold) upon activation with lipopolysaccharide (Fig. 1B and C). Next, we confirmed that primary microglial cells from both mutant (hSOD1^G37R, hSOD1^G85R) and wild-type (hSOD1^WT) human SOD1 expressing mice were also able to induce xCT levels after lipopolysaccharide activation (Fig. 1A and C).

Importantly, as no differences between wild-type hSOD1, mutant hSOD1 or non-transgenic microglia were detected, these results suggest that induction of system ${\text{x}}_{\text{C}}^{-}$ was solely dependent on the microglial activation state and not the presence of a specific ALS-linked gene mutation.

Deletion of xCT reduces the pro-inflammatory phenotype of activated primary microglia

To study system ${\text{x}}_{\text{C}}^{-}$ involvement in glutamate release by primary microglial cells (and linked to it, its potential contribution in excitotoxicity), we measured glutamate concentration in culture medium. Unstimulated microglia released glutamate at 10–50 μM after 30 h in culture and this increased by 70% upon lipopolysaccharide activation (Fig. 2A). Importantly, xCT deletion in microglial cells from xCT^−/− mice, led to a 70% decrease in glutamate release compared to control (xCT^+/+) microglia (Fig. 2A). In addition, treating xCT^−/− microglial cells with lipopolysaccharide did not raise glutamate production (Fig. 2A). Therefore, our findings show that microglial cells release glutamate mainly through their system ${\text{x}}_{\text{C}}^{-}$. Of note, mutant or wild-type hSOD1 expressing microglia released similar amounts of glutamate as control microglia, both with and without lipopolysaccharide activation (Fig. 2B). These data show (similar to the xCT expression, above) that glutamate release did not depend on the presence of mutant or wild-type hSOD1 but solely on the microglial activation state.

Although system ${\text{x}}_{\text{C}}^{-}$ blockers are known to lack specificity (Lewerenz et al., 2013), we tested one, sulfasalazine, and found that it blocked microglial glutamate release both in control conditions and after lipopolysaccharide treatment (Fig. 2C), consistent with system ${\text{x}}_{\text{C}}^{-}$ being responsible for the measured glutamate release.

To test our hypothesis that xCT deletion would not just reduce excessive glutamate release, but could also directly influence the global M1/M2 activation state, we assessed microglial production of known pro-inflammatory/neurotoxic (M1) and anti-inflammatory/neurotrophic (M2) factors. First, the classic microglial neurotoxic (M1) factor nitric oxide was assessed. Lipopolysaccharide strongly induced (by 10-fold) microglial nitric oxide production, whereas xCT deletion led to 65% less nitric oxide production upon lipopolysaccharide activation (Fig. 3A). Next, using a multiplex enzyme-linked immunosorbent assay (Luminex), different growth factors (G-CSF/Csf3, GM-CSF/Csf2, VEGFa), cytokines (TNFa, IFNg, IL1a, IL1b, IL3, IL4, IL6, IL10, IL12b, IL13, IL17) and chemokines (CCL5, CCL2, CCL3, CCL4) were measured in media conditioned by control (xCT^+/+) or xCT^−/− microglia, activated or not by lipopolysaccharide (Fig. 3B–E and Supplementary Fig. 1). Except for IL13 and IFNg, which were not detectable, all the factors assessed were released by microglia. Interestingly, compared to controls, in xCT-deleted microglia, we found a significant downregulation of the lipopolysaccharide-induced production of the growth factor G-CSF/Csf3 (−80%, Fig. 3D), but also of the pro-inflammatory factors TNFa (−60%, Fig. 3B), IL6 (−70%, Fig. 3C), and the chemokine CCL3 (−25%, Fig. 3E). Of note, not all measured factors were dependent on xCT expression; CCL2 production was stable, underlining the specificity of the changes found (Supplementary Fig. 1).

Thus, as upon activation, xCT-deleted microglial cells released not just less glutamate, but also less nitric oxide and other pro-inflammatory/neurotoxic (M1) factors, system ${\text{x}}_{\text{C}}^{-}$ can therefore be considered a regulator of the microglial activation state.

xCT expression is increased in microglia of ALS mice and correlates with inflammation in human ALS spinal cord

Although system ${\text{x}}_{\text{C}}^{-}$ is considered a rather glial transporter, its lack of expression by adult motor neurons was not known, but was an essential requirement for our non-cell autonomous hypothesis. Thus, we established, by comparing xCT mRNA expression from isolated, laser-microdissected, adult motor neurons to whole lumbar spinal cord, that motor neurons did not express significant levels of system ${\text{x}}_{\text{C}}^{-}$ in control and mutant SOD1 ALS mice (Fig. 4A and Supplementary Fig. 2B), indicating that potential changes of xCT expression and effects of xCT deletion during disease, cannot be linked to a contribution of motor neurons.

Because primary microglial cells, when activated, increased xCT expression and as microglia are activated during ALS, we analysed lumbar spinal cord xCT mRNA expression at four different disease stages in slowly progressing mutant hSOD1^G37R ALS mice, which reach end stage on average at 13.5 months (Fig. 4B). In non-transgenic control mice, xCT levels were not modified during ageing, and wild-type hSOD1^WT did not influence xCT expression (Supplementary Fig. 2A). Importantly, however, in ALS mice, xCT expression, while unchanged presymptomatically, was strongly increased at both disease onset and symptomatic stages, and then decreased at end stage (Fig. 4B).

To analyse whether microglial cells were a major source of disease-associated increased xCT expression, we isolated adult microglia from spinal cords of ALS mice during disease. First, compared to whole lumbar spinal cord, xCT mRNA was clearly enriched in microglia (Fig. 4A). Isolated microglia expressed mRNA for the microglial/macrophage marker CD68 and to higher levels when isolated from symptomatic compared to presymptomatic ALS mouse spinal cords (Supplementary Fig. 2C). Consistent with a contribution from activated microglia to disease-associated xCT induction, xCT mRNA levels were strongly increased in isolated microglia during disease progression (by 45-fold between symptomatic and presymptomatic stages) (Fig. 4C).

Importantly, xCT mRNA and protein were also expressed in the spinal cord of human ALS cases and xCT mRNA levels correlated with levels of CD68 (Fig. 4D and Supplementary Table 1) in agreement with our results in mice where xCT expression was increased when microglial cells were activated.

Deletion of xCT in ALS mice leads to increased expression of M2 markers, suggesting a switch towards a more neuroprotective microglial phenotype

As deleting xCT reduced neurotoxic factor production of cultured microglia, and microglial xCT expression was increased during disease in ALS mice, we next investigated the impact of xCT deletion on microglial phenotype, in vivo, over the course of the disease. Therefore, we crossed hSOD1^G37R mice with xCT^−/− mice.

First, we analysed general microgliosis in lumbar spinal cords of ALS mice, deleted or not for xCT, over the course of the disease, by using the classic microglial markers Iba1/Aif1 and Mac-2/Lgals3 (Fig. 5 and Supplementary Fig. 3). Although microglial activation increased during the course of the disease, overall microgliosis (as assessed by Iba1/Aif1 and Mac-2/Lgals3 immunostainings) was comparable between hSOD1^G37R ALS mice with or without xCT, at every tested disease stage (Fig. 5 and Supplementary Fig. 3).

As classic microglial activation markers (Iba1/Aif1, Mac-2/Lgals3) do not necessarily predict more specific M1/M2 microglial activation states, we compared mRNA expression levels of specific microglial-associated M1 (iNOS/Nos2, NOX2/Cybb, Il1b, IL12p40/Il12b) and M2 (Ym1/Chil3, Arg1, Igf1) markers, during disease progression in the lumbar spinal cord of hSOD1^G37R:xCT^+/+ and hSOD1^G37R:xCT^−/− ALS mice (Fig. 6 and Supplementary Fig. 4).

In presymptomatic hSOD1^G37R ALS mice, the expression of both M1 and M2 markers was comparable to control (xCT^+/+) mice (Supplementary Fig. 4A). For the M1 markers, Il1b and NOX2/Cybb were increased over the course of the disease (Fig. 6B and C). For the M2 markers, Igf1 was increased over the course of the disease (Fig. 6F), whereas the patterns of expression of Arg1 and Ym1/Chil3 were more complex. Indeed, Arg1 expression was stable during the disease and only increased at end stage (Fig. 6E). Ym1/Chil3 levels were increased at the symptomatic stage, but transiently decreased to almost control levels at the late symptomatic stage, to strongly increase again at disease end stage (Fig. 6D).

Upon deletion of xCT, surprisingly, already at baseline in non-ALS mice as well as in presymptomatic ALS mice, we detected a striking difference between xCT^−/− and xCT^+/+ mice for the M2 marker Ym1/Chil3, with mRNA levels more than 10-fold higher in xCT^−/− mice (Fig. 6A and D). This difference (increase in Ym1/Chil3 expression) was even more striking when directly comparing adult spinal cord microglial cells isolated from xCT^+/+ and xCT^−/− mice (Supplementary Fig. 5).

Over the disease course, comparing ALS mice with and without xCT, we could distinguish two different disease phases, with an early disease phase (onset and symptomatic stage), where xCT deletion led to an increased expression of Il1b (M1) and a suppression of Ym1/Chil3 (M2) (Fig. 6B and D) and a late disease phase (late symptomatic stage), where xCT deletion led to an increased expression of the neuroprotective M2 markers Ym1/Chil3 and Arg1 (Fig. 6D and E).

Consistent with what we already detected in vitro, deletion of system ${\text{x}}_{\text{C}}^{-}$ in ALS mice, conferred a more anti-inflammatory/neuroprotective microglial-associated phenotype and this was especially evident in the late disease phase. This suggested a switch towards a more neuroprotective microglial phenotype.

A dual action from xCT deletion leads to an initial speed-up followed by a beneficial prolongation of the symptomatic disease phase

As xCT deletion led to a more anti-inflammatory/neuroprotective environment in ALS mouse spinal cords, we followed hSOD1^G37R mice, deleted or not for xCT to analyse the consequence of removing xCT on the ALS disease course.

First, baseline xCT-deleted mice showed no phenotype and had a normal lifespan (958 ± 27.7 days, n = 20 and 978 ± 25.3 days, n = 24; for xCT^+/+ and xCT^−/− mice, respectively; equally gender-matched). Nevertheless, to ensure absence of subtle pathological abnormalities related to ALS, we confirmed lack of motor neuron loss and general glial activation in 1-year-old xCT^−/− mice (Supplementary Fig. 6A and C). Likewise, although system ${\text{x}}_{\text{C}}^{-}$ can contribute to glutathione synthesis, previous studies already showed that global brain glutathione content (measured in the striatum and hippocampus) was not reduced in xCT-deleted mice (De Bundel et al., 2011; Massie et al., 2011), which we now also confirmed for spinal cord (Supplementary Fig. 6B).

Next, analysis of disease course showed that xCT deletion in mutant SOD1 ALS mice had no impact on disease onset, defined by two methods (Fig. 7A and A’), (i) the age when reaching weight peak (hSOD1^G37R:xCT^+/+: 247 ± 6.5 days, n = 28; hSOD1^G37R:xCT^+/−: 254 ± 8.8 days, n = 40; hSOD1^G37R:xCT^−/−: 242.8 ± 7.8 days, n = 24); and (ii) the age when reaching grip strength peak (hSOD1^G37R:xCT^+/+: 197.6 ± 1.9 days, n = 28; hSOD1^G37R:xCT^+/−: 191.3 ± 1.2 days, n = 40; hSOD1^G37R:xCT^−/−: 183.9 ± 1.9 days, n = 24), consistent with the idea that effects linked to modulation of microglia should rather contribute to the symptomatic phase of the disease. Surprisingly, hSOD1^G37R :xCT^−/− mice reached the specific early disease stage (the beginning of the symptomatic phase, defined by appearance of motoric symptoms and 10% weight loss) 1 month earlier than hSOD1^G37R :xCT^+/+ mice (−32.5 days; P = 0.0233) (328.4 ± 9.2 days, n = 24 and 360.5 ± 7.7 days, n = 26, respectively) (Fig. 7B), with a similar speed-up, when defining early disease by 35% grip strength loss (−30.8 days; P = 0.0329) (Fig. 7B’).

However, this earlier reaching of the symptomatic stage (10% weight loss) did not decrease survival, with hSOD1^G37R:xCT^−/− mice reaching end stage (complete hind limb paralysis) at the same time as hSOD1^G37R:xCT^+/+ mice (Fig. 7C) (410 ± 6 days, n = 21 and 405 ± 7.5 days, n = 24, respectively). This dual action of xCT deletion led to: (i) a speed-up to reach early disease (Fig. 7B, B’ and D) but in contrast, and importantly; (ii) an increase of the duration of the late disease phase after appearance of symptoms (the time between 10% weight loss and end-stage or 35% grip strength loss and end-stage; +28 days, P = 0.0278) in hSOD1^G37R:xCT^−/− mice (85.5 ± 8.2 days, n = 21) as compared to hSOD1^G37R:xCT^+/+ mice (57.4 ± 5.9 days, n = 24) (Fig. 7D and Supplementary Fig. 7).

Our tracking of the mouse cohort consistently indicated that, after appearance of symptoms, hSOD1^G37R :xCT^−/− mice seemed to live longer and behave better than hSOD1^G37R :xCT^+/+ mice during the symptomatic phase. To quantify, we measured the survival time to reach a defined advanced symptomatic disease stage (20% of weight loss). The time elapsed, after appearance of symptoms, to reach this advanced symptomatic stage was indeed increased (Supplementary Fig. 7A). Interestingly, and consistent with our observation, while 40% of hSOD1^G37R :xCT^+/+ mice had already died at this disease stage, this was the case for only 20% of the hSOD1^G37R :xCT^−/− mice (Fig. 7E). Even more so, compared to hSOD1^G37R :xCT^+/+ mice, hSOD1^G37R :xCT^−/− mice lived longer after reaching this disease stage (Fig. 7E and Supplementary Fig. 7B, P = 0.01, which is underestimated as n = 17/21 for the hSOD1^G37R:xCT^−/− group and only n = 15/24 for the hSOD1^G37R :xCT^+/+ group). In line with this finding, during the late symptomatic disease phase, hSOD1^G37R:xCT^−/− mice had ∼20% (P < 0.05) more remaining grip strength than hSOD1^G37R:xCT^+/+ mice (57 ± 1.6% and 39 ± 1.7%, respectively) (Fig. 7F).

Deletion of xCT in ALS mice results in more surviving motor neurons at disease end stage

The three groups of ALS mice, hSOD1^G37R:xCT^+/+, hSOD1^G37R:xCT^+/− and hSOD1^G37R:xCT^−/−, started out with the same number of motor neurons (30.9 ± 0.4, 35.5 ± 4.6 and 33.7 ± 1.0 per spinal cord section, respectively) (Fig. 8) and, as expected, during disease, motor neurons degenerated in hSOD1^G37R:xCT^+/+ mice with only 35% remaining at end stage (Fig. 8). Interestingly, however, hSOD1^G37R:xCT^−/− mice had, with 45% remaining, a significant increase in survival of motor neurons (Fig. 8). This suggested an overall less neurotoxic environment and slowing of motor neuron degeneration due to xCT deletion, consistent with our results, in vitro, showing decreased secretion of pro-inflammatory/neurotoxic (M1) factors from primary microglial cells (Fig. 3) and, in vivo, with the switch to a spinal cord environment associated with a more anti-inflammatory/neuroprotective (M2) microglial phenotype in late symptomatic xCT-deleted ALS mice (Fig. 6).