Astrocytes require insulin-like growth factor I to protect neurons against oxidative injury

Oxidative stress is a proposed mechanism in brain aging, making the study of its regulatory processes an important aspect of current neurobiological research. In this regard, the role of the aging regulator insulin-like growth factor I (IGF-I) in brain responses to oxidative stress remains elusive as both beneficial and detrimental actions have been ascribed to this growth factor. Because astrocytes protect neurons against oxidative injury, we explored whether IGF-I participates in astrocyte neuroprotection and found that blockade of the IGF-I receptor in astrocytes abrogated their rescuing effect on neurons. We found that IGF-I directly protects astrocytes against oxidative stress (H 2O 2). Indeed, in astrocytes but not in neurons, IGF-I decreases the pro-oxidant protein thioredoxin-interacting protein 1 and normalizes the levels of reactive oxygen species. Furthermore, IGF-I cooperates with trophic signals produced by astrocytes in response to H 2O 2 such as stem cell factor (SCF) to protect neurons against oxidative insult. After stroke, a condition associated with brain aging where oxidative injury affects peri-infarcted regions, a simultaneous increase in SCF and IGF-I expression was found in the cortex, suggesting that a similar cooperative response takes place in vivo. Cell-specific modulation by IGF-I of brain responses to oxidative stress may contribute in clarifying the role of IGF-I in brain aging.


Introduction
Oxidative stress is usually considered a mechanism of brain aging 1 . However, contradictory data 2 and lack of firm evidence 3 makes it difficult to firmly establish its actual significance in this process (see  for a recent review). One important aspect that requires further clarification in this regard is the relationship between oxidative stress and insulin peptides, a well conserved family of hormones firmly linked to aging. Extensive work in vertebrates and invertebrates indicates that the insulin-like growth factor I (IGF-I)/insulin signalling (IIS) pathway has a negative impact on aging. It has been argued that this detrimental action is mediated by reducing cell defences to oxidative stress 5-7 which, in turn is harmful for neuronal survival 1 . However, IGF-I has been shown to be largely neuroprotective 8 , even in conditions such as ischemic injury or brain trauma where oxidative stress is most likely a major pathogenic mechanism 9 . Thus, it is unclear whether or not IGF-I protects the brain against oxidative stress as the current evidence is contradictory.
A possible explanation for these apparently contradictory observations may be that modulation of the cellular response to oxidative stress by IGF-I is cell-dependent 10 . Until now, only neurons have been studied in this regard. However, astrocytes, a major cellular element of the brain, are essential contributors to neuronal homeostasis and are coupled to neurons in the response to oxidative stress in order to help protect them 11 . It is thus possible that IGF-I participates in the response of astrocytes to oxidative stress as part of the overall brain response encompassing all types of brain cells, not only neurons. Contrary to what we previously observed in neurons 12 , we report here that IGF-I protects astrocytes against oxidative stress and, very significantly, also co-operates with astrocytes to protect neurons.

Animals
We used postnatal rats and mice for in vitro cultures (P0-3 days for astrocytes and P7 for neurons) and 3 month old mice for in vivo experiments. P2 Wistar rats (8 g ± 0.04 body weight, n=240, Harlan, Spain), P3 (2 g ± 0.03, n=36, Harlan) and 3 months old (27.6 g ± 0.812; n= 24) C57BL6 mice and P7 GFP transgenic mouse pups (4.25 g ± 0.22, n=126; in-house colony) were used. Pups used were of both sexes and no attempt to sex them was done. Adult mice were male. Rat tissue was used in all in vitro experiments except when using GFP cell derived from transgenic mice. In vivo experiments were done in mice for future comparisons with transgenic mice. All efforts were made to minimize suffering and reduce the number of animals. Animals were kept under light/dark, 12 h/12 h) conditions following EU guidelines (directive 86/609/EEC) and handled according to institutionally-approved procedures (CSIC bio-ethics subcommittee project code SAF2010-1703). Animals were fed ad libitum with laboratory rodent chow (Teklad Global 2018S) and kept in standard laboratory cage conditions with 4 animals/cage.

Reagents
Antibodies used in this study are detailed in Table 1. The different drug inhibitors used in the study are given in Table 2. Hydrogen peroxide (H 2 O 2 ) and the calcium chelator BAPTA-AM were purchased from Sigma (Steinheim, Germany). IGF-I and SCF were purchased from Prospec-Tany Technogene, (Israel).
Cell culture and transfections Cerebellar granule cultures were produced from either P7 rat or GFP transgenic mouse cerebella as previously described 13 . In brief, cells were plated onto 6 or 12-well dishes coated with poly-l-lysine (1 μg/ml) at a respective final density of 1.5×10 6 /well or 0.45×10 6 / well. Cells were incubated at 37°C/5% CO 2 in Neurobasal (Gibco, USA) medium supplemented with 10% B27 (Gibco), glutamine (5 mM) and KCl (25 mM). All experiments were carried out in 2-7 day old cultures, with neurons showing neurite extensions. Different times in vitro were used to analyze time-dependent parameters such as cell survival. Rat granule neurons were transfected 24 h after plating. The DNA: transfection agent ratio (Neurofect, Genlantis, San Diego, USA) was 1:7. The percentage of neurons transfected was 5-10%, as assessed with a GFP vector. Neurons were left untreated for at least 48 hours. On the day of the experiment, medium was replaced with Neurobasal + 25 mM KCl. Two

Amendments from Version 1
We very much appreciate the expert reviews that will help us improve this manuscript and our future work. As indicated by Dr Margeta we describe two sets of in vitro actions of IGF-I that may be roughly labeled as "astroprotection" and "neuroprotection through astrocytes". We assumed that enhanced resilience of astrocytes to oxidative injury provided by IGF-I would positively impact on their ability to protect neurons in vivo. However, we do not have any direct experimental evidence that this is the case. We are planning quantitative proteomics of astrocytes with and without IGF-I receptors to address this lack of evidence. We have now deleted any comment suggesting this connection and thank the reviewer for her insight as we were over-interpreting our findings. Astro-protective effects of IGF-I (anti-oxidant and Txniprelated, independent of neuroprotection) are now more specifically addressed as such. We now include the experiments required by the reviewer previously not shown (see new Figure 1D, and Figure 7A right panel) and have added the finding that astrocyte SCF release (and not only SCF mRNA, new Figure 6B right panel) is increased in response to oxidative stress. The latter reinforces our interpretation that IGF-I cooperates with astrocytes to protect neurons through SCF. The rest of the technical details requested and formal changes have been incorporated to the manuscript. hours later, IGF-I (10 -7 M) and/or hydrogen peroxide (H 2 O 2 ) at doses of 50-150 μM were added. Inhibitory drugs were given 45 min before treatments. We used H 2 O 2 as an oxidant stimulus because it is an endogenously produced reactive oxygen species (ROS) that serves as a precursor to hydroxyl radicals and possesses signalling capacities 14 . Astrocyte cultures were prepared from P3 rat or GFP mouse forebrain, as previously described 15 after animals were sacrificed by decapitation. Cells were grown on Dulbecco's modified Eagle's medium F12 (DMEM-F12) supplemented with 10% fetal calf serum. After 12 days astrocytes were seeded at 2.5×10 5 or 1.25×10 5 cells/well in 6-well and 12-well culture plates, respectively. On the day of the experiment cells were treated with IGF-I (10 -7 M), H 2 O 2 (50-200 μM) and/or inhibitors, as above. For transfection, astrocytes were seeded at 2.5×10 5 or 1.25×10 5 cells/well in 6-well and 12-well culture plates respectively, and after 16 h constructs were mixed with Fugene HD (Roche, Switzerland) in a 1:3 ratio, and added following the manufacturer's instructions. Alternatively, astrocytes were electroporated (2×10 6 astrocytes with 2 μg DNA or shRNA) before seeding using an astrocyte Nucleofector Kit (Lonza, Switzerland). After electroporation, cells were plated to obtain a final cell density on the day of the experiment similar to that obtained with the transfection method. All experiments were performed after 48 h. The transfection efficiency was 20-30% and 60-80% for electroporation, as assessed with a GFP vector. At least three independent experiments were done in duplicate wells.

Co-cultures
For co-cultures, 1.25×10 5 wild type mouse astrocytes/well were seeded on 12-well plates and grown with DMEM-F12 plus 10% FBS. After 48-72 hours, GFP neurons were isolated and plated onto astrocytes. We used forebrain astrocytes and cerebellar neurons because in our experience the forebrain and cerebellum yielded very high numbers of astrocytes and neurons, respectively (thus minimizing animal use). Furthermore, in this study we were interested in exploring general, rather than region-specific neuroprotective characteristics of astrocytes. Nevertheless, we also carried out co-cultures with neurons and astrocytes from the same region (forebrain) and the results obtained were identical than when using cells from differing regions (see Figure 2 in results). Culture medium was changed to DMEM-F12 plus B27, 4 mM glutamine and 25 mM KCl (the latter only in the case of neurons). Two days later, cocultures were treated with 100 nM IGF-I ± 50-100 μM H 2 O 2 . Pictures were taken every 24 hours up to 5-7 days as above. For protein silencing or overexpression, 2×10 6 astrocytes were electroporated in a Nucleofector ® II (Amaxa Biosystems Lonza, Switzerland) and seeded at 1.25×10 5 /well. Co-cultured neurons were seeded as described above. Viability of neurons was assessed by counting the number of cells expressing GFP using Incucyte software (2010A) with a set cell size threshold to avoid including GFP + cell debris and dying cells. This threshold ranged from 8-36 μm 2 to 70-200 μm 2 depending on the experiment. Viability is expressed as percentage of GFP + cells at the beginning of the experiment (time 0). At least three independent experiments were done.

Cell assays
Cell viability was determined by four different methods. The first assessed astrocyte death by quantification of the amount of lactate dehydrogenase (LDH) released from damaged astrocytes into the culture medium. LDH levels were measured after 16 h of treatment with different H 2 O 2 concentrations using a commercial kit (Roche Diagnostics, Germany). When using transfected astrocytes, a GFP-pCMV vector and the different constructs under evaluation were used in a 1:5 ratio. In this case, GFP + astrocytes were scored prior to treatment to determine baseline survival (time 0) and at different times as indicated in the results. Alternative viability assays for astrocytes included measuring cell metabolism with fluorescein diacetate (0.1 μg/ml FDA) or number of propidium iodide (PI) cells 12 as specified in the results section. For the latter, astrocytes or neurons were stained with 2 μg/ml PI as a marker of dead cells plus DAPI staining as a marker of total cell number. PI + and DAPI + cells were counted under a Leica CTR 6000 fluorescence microscope. Percentage of viable cells indicates the number of PI + cells related to total cell number. The experiments were done in triplicate and a total of three independent experiments were done. For neuronalspecific viability assays cerebellar neurons from GFP mice were seeded on 12-well plates (4.5×10 5 cells) coated with poly-L-Lysine and grown with Neurobasal medium plus B27, 4 mM glutamine and 25 mM KCl. After 4-5 days, cultures were treated with 100 nM IGF-I in the presence or absence of 50-100 μM H 2 O 2 . Pictures of GFP + cells (green fluorescence) were taken every 24 hours up to 3 days in an Incucyte TM 2010A Rev2 system (Essen BioScience, USA). Viability of neurons in co-culture experiments was measured as described above.

Immunoassays
Western blotting was performed as described 13 . Cells were washed once with ice-cold PBS and lysed with 1% NP-40, 150 mM NaCl, 20 mM Tris, pH 7.4, 10% glycerol, 1 mM CaCl 2 , 1 mM MgCl 2 , 400 μM sodium vanadate, 0.2 mM PMSF, 1 μg/ml leupeptin, 1 μg/ml aprotinin and 0.1% phosphatase inhibitor cocktails I and II (Sigma-Aldrich). To normalize for protein load, membranes were reblotted (Re-Blot, Chemicon, USA) and incubated with an Densitometric analysis was performed using Analysis Image Program (Bio-Rad, USA). A representative blot is shown from a total of at least three independent experiments. IGF-I levels in culture medium were measured using Quantikine ELISA for mouse/rat IGF-1 (R&D Systems, USA). In brief, cells were treated as described above and 1 ml of culture medium was collected after 24 hours, spun to eliminate cell debris, and stored at -80ºC. Samples were lyophilized overnight and resuspended in 150 μl of calibrator buffer. After vortexing, samples were centrifuged 10 min/14,000 rpm (Hettich, Germany) and assayed according to manufacturer's instructions. A total of three and four independent experiments were done for neurons and astrocytes, respectively. SCF levels in culture medium were measured by western blot after collecting the supernatants and processing them as described above for IGF-I. After lyophilisation, samples were resuspended in western blot lysis buffer and protein levels were measured by Bradford (Biorad, Germany) following the manufacturer's instructions to normalize for protein load in SDS-PAGE gels.

Luciferase assays
Luciferase assays were done as previously reported 12 . In brief, cells were transfected with a reporter construct bearing six canonical FOXO binding sites (6×DBE-luciferase) and co-transfected with different constructs, as indicated in each experiment. Transfections were performed in triplicate dishes. Luciferase counts were normalized using TK-Renilla luciferase. At given times, neurons were lysed in passive lysis buffer (PLB) and luciferase activity was analysed using a luminometer and dual luciferase assay kit according to the manufacturer (Promega, USA). Background luminescence was subtracted. Luciferase activity was expressed as fold of increase over control levels. At least three independent experiments were done.

Flow cytometry
After 18h of exposure to H 2 O 2 , cell death was assessed. Cells were detached using 0.25% Trypsin-1.3 mM EDTA (Invitrogen) during 5-10 minutes, centrifuged (200×g, 5 min/4ºC), and resuspended in cold PBS. Propidium iodide (PI 5 μg/ml; Sigma) in PBS was added prior to flow cytometry analysis using a FACSAria cytometer (BD Biosciences). Fluorescence intensity, forward scatter (FSC), and side scatter (SSC) were collected in logarithmic scale. The emission filter used was 600-620 nm band pass (FL3). A fluorescence blank was measured and subtracted from the fluorescence of the sample. Dead cells were identified as red fluorescence positive events with low FSC (small PI permeable cells). Debris was always excluded from the analysis. At least three independent experiments were conducted.

ROS measurement
Mitochondrial O 2 -production levels were measured by using the fluorescent probe MitoSOX TM Red (Life Technologies, USA). Briefly, astrocytes were pre-treated overnight with IGF-I and then 200 μM H 2 O 2 were added during 1 hour. Cells were incubated with 1.5 μM MitoSOX TM Red in DMEM-F12 for 10 min/37ºC and washed 3 times with PBS. Astrocytes were then trypsinized and fluorescence was measured by flow cytometry (510 nm excitation/580 nm emission) using the cytometer, as described 15 . A total of six independent experiments were done. Alternatively, ROS generation was assessed in astrocytes cultured on coverslips with the fluorogenic marker carboxy-H 2 DCFDA (Molecular Probes, USA) during 30 min/37ºC, protected from the light. When using this ROS marker it is not possible to distinguish endogenous ROS from exogenously applied H 2 O 2 . Nevertheless, we compared this method to the oxidation of luminol (which detects superoxide anions) that distinguishes H 2 O 2 from other ROS and we obtained identical results with either method (data not visualized). The reason we used carboxy-H 2 DCFDA is because we could obtain both qualitative (cell images) and quantitative (fluorimetry assay) measurements within the same assay. After incubation with carboxy-H 2 DCFDA, cells were gently washed 3 times with warm DMEN, and mounted, or, alternatively, lysed for fluorimetry. Pictures were taken at 40× magnification using a Leica fluorescence microscope (Germany). A representative picture is shown. Fluorescence intensity in lysed cells was measured using a FluoroStar fluorimeter.
Growth factor gene array An RT 2 Profiler TM PCR Array (SABiosciences, USA) was used to screen a battery of growth factors following the manufacturer's recommendations. After treatment, astrocytes were lysed and RNA extracted using Trizol (Life Technologies, USA). The resulting cDNA synthesis reaction was diluted in water, mixed with the qPCR master mix, and loaded in a 96 well PCR Array plate. PCR was performed following manufacturer's instructions.

Brain focal ischemia
Three-month old male mice (4-6 per group) were anesthetized with 3% isoflurane (in 70% N 2 O, 30% O 2 ) for induction and with 2% isoflurane for maintenance. Rectal temperature was maintained at 36.5 C with a heating pad. The frontal branch of the medial cerebral artery (MCA) was exposed and occluded permanently by suture ligation as previously reported, with modifications 16 . Briefly, an incision perpendicular to the line connecting the lateral canthus of the left eye and the external auditory canal was made to expose and retract the temporalis muscle. A burr hole was drilled, and frontal and parietal branches of the MCA were exposed by cutting and retracting the dura. The frontal branch of the MCA was elevated and ligated with a suture nylon monofilament 8/0. Following ligation, a sharp decrease of blood flow was evidenced with a laser Doppler flowmetry (Järfalla, Sweden). Following surgery, mice were returned to their cages, kept at room temperature and allowed free access to food and water. All physiological parameters measured: rectal temperature, mean arterial pressure and blood glucose levels were not different between groups. Sixteen hours after medial cerebral artery occlusion (MCAO), animals were killed by neck dislocation by an experienced researcher to assess infarct outcome. The brain was removed and the infarcted area isolated and processed for RNA and protein isolation.

Quantitative PCR
Total RNA isolation from cell lysates or brain tissue was carried out with Trizol. One μg of RNA was reverse transcribed using High Capacity cDNA Reverse Transcription Kit (Life Technologies) according to the manufacturer's instructions. For the quantification of specific genes, total RNA was isolated and transcribed as above and 62.5 ng of cDNA was amplified using TaqMan probes for Txnip1, IGF-I or SCF and 18S as endogenous control (Life Technologies). Each sample was run in triplicate in 20 μl of reaction volume using TaqMan Universal PCR Master Mix according to the manufacturer's instructions (Life Technologies). All reactions were performed in a 7500 Real Time PCR system (Life Technologies). Quantitative real time PCR analysis was carried out as previously described 17 . Results were expressed as relative expression ratios on the basis of group means for target transcripts versus reference 18S transcript. At least three independent experiments were done.

Statistical analysis
Data are expressed as mean ± SEM. Differences among groups were analyzed by one-or two-way ANOVA followed by a Newman-Keul's or Student's t-test using Graph Pad Prism 5 software. A p<0.05 was considered significant.

Results
Astrocyte neuroprotection against oxidative stress requires IGF-I signalling onto astrocytes Whereas neurons cultured without astrocytes are very sensitive to acute oxidative insult elicited by H 2 O 2 ( Figure 1A), when cultured with astrocytes, neurons become very resilient ( Figure 1A). To determine whether IGF-I participates in the neuroprotective effects of astrocytes against oxidative stress we first confirmed that it is endogenously produced by these cells. As shown in Figure 1B, not only astrocytes but also neurons (albeit at much lower levels) secrete IGF-I into the culture medium. In response to H 2 O 2 astrocytes secrete lower, but still substantial, amounts of IGF-I, and so IGF-I may still participate in neuroprotection by astrocytes. To directly test this possibility we blocked IGF-I signalling in astrocytes with a  dominant negative (DN) IGF-IR 18 ( Figure 1C) and determined their ability to protect neurons against oxidative challenge. As shown in Figure 1D, a significantly greater percentage of neurons co-cultured with mock-transfected astrocytes survived after H 2 O 2 challenge than when cultured with astrocytes transfected with DN IGF-IR.
We next used pharmacological blockade of the IGF-I receptor using picropodophyllin (PPP), an antagonist of IGF-IR ( Figure 2A). As in this case both the neuronal and astrocyte receptors are blocked, we first determined whether neurons are affected by PPP blockade of the IGF-I receptor. In the presence of H 2 O 2 , neurons cultured alone die regardless of the presence or absence of proper IGF-I signalling since PPP did not increase neuronal death ( Figure 2B). This agrees with our previous findings that IGF-I does not protect cultured neurons against oxidative stress 12 . Confirming the results seen with astrocytes transfected with dominant negative IGF-I receptor, a reduction in neuroprotection by astrocytes was seen when co-cultured neurons were exposed to PPP. In the presence of H 2 O 2 , significantly fewer co-cultured neurons survived with PPP (p<0.01 H 2 O 2 +PPP vs. H 2 O 2 alone; Figure 2C). To rule out region-specific actions of astrocytes on neuroprotection we then co-cultured neurons and astrocytes from the same brain region (forebrain) and treated them with PPP. As shown in Figure 2D, forebrain neurons were similarly sensitive to blockade of IGF-IR when co-cultured with forebrain astrocytes. The observation that even supra-physiological doses of IGF-I (100 nM) added to the co-cultures only produced a modest additional effect on neuronal survival after oxidative insult confirmed the idea that endogenous IGF-I is required by astrocytes for neuroprotection ( Figure 2E). Hence, endogenous production of IGF-I is necessary and sufficient to protect neurons.

Figure 2. Endogenously produced IGF-I protects neurons against oxidative injury. A)
The IGF-IR inhibitor PPP blocks IGF-I signalling in astrocytes. Astrocytes were treated with 120 nM PPP 1h before adding IGF-I while pAkt levels were measured 10 minutes after adding IGF-I. IGF-I protects astrocytes against oxidative stress IGF-I-dependent neuroprotection by astrocytes appears to also involve a direct action of IGF-I on astrocytes. Because it is known that astrocytes are more resistant to oxidative damage than neurons, we explored whether IGF-I was involved in this greater resilience. Contrary to neurons ( Figure 3A), IGF-I protected astrocytes against H 2 O 2 -induced death ( Figure 3B). The protective effect of IGF-I involved blockade of the activation of FOXO 3, a transcription factor involved in brain responses to oxidative stress 19 , by H 2 O 2 ( Figure 3C). Inhibition of FOXO 3 by IGF-I was mediated by Akt; i.e.: an Akt-insensitive mutant of FOXO (M-FOXO3) abrogated IGF-I effects while wild type FOXO3 did not interfere with its protective actions ( Figure 3D). Indeed, in astrocytes IGF-I activates Akt in the presence of H 2 O 2 ( Figure 3E), whereas in neurons H 2 O 2 blocks this canonical pathway 12 . Underlying the protective actions of IGF-I on astrocytes was its ability to block excess ROS after exposure to H 2 O 2 as determined by flow cytometry using MitoSOX ( Figure 4A) or fluorometry with carboxy-H 2 DCFDA ( Figure 4B).
We then determined possible mediators of the anti-oxidative actions of IGF-I on astrocytes. We examined whether modulation of SODs could be involved because these anti-oxidant enzymes constitute an important detoxifying mechanism in cases of excess ROS. We found that cytosolic Cu/ZnSOD was increased by IGF-I, H 2 O 2 , or both ( Figure 5A), while mitochondrial MnSOD was increased only by H 2 O 2 ( Figure 5B). Thus, increases in SOD levels form part of    as compared to mock-transfected astrocytes. Accordingly, IGF-I, which inhibits FOXO, also reduced TXNIP1 levels ( Figure 6A). Strikingly, H 2 O 2 , which stimulates FOXO activity in astrocytes ( Figure 3D), also inhibited TXNIP1 ( Figure 6A), suggesting alternative routes of TXNIP1 regulation in the presence of H 2 O 2 . When IGF-I and H 2 O 2 were simultaneously added to astrocytes, TXNIP1 levels were markedly decreased (p<0.05 vs. IGF-I or H 2 O 2 alone, Figure 6A). To determine the impact of downregulation of TXNIP1 the astrocyte response to H 2 O 2, and IGF-I does not appear to interfere with these enzymes. Because FOXO participates in cellular responses to ROS, we looked for signals downstream of FOXO inactivation by IGF-I such as thioredoxin inhibitor 1 (TXNIP1), a pro-apoptotic protein dependant on FOXO activity and related to anti-oxidant responses 20 . We first confirmed that in astrocytes TXNIP1 is also controlled by FOXO; i.e.: in astrocytes expressing dominant negative Foxo, TXNIP1 levels were 89% reduced  on astrocyte survival we inhibited its expression with shRNA (blot in Figure 6B left panel) and found that astrocytes became resistant to H 2 O 2 when TXNIP1 levels were low ( Figure 6B). Overexpression of TXNIP1 did not alter the response of astrocytes to H 2 O 2 whereas co-culture of neurons with astrocytes depleted of TXNIP1 did not result in enhanced neuronal survival ( Figure 6B right panel), indicating that this route is involved in the response of astrocytes to oxidative stress but not in neuroprotection. Interestingly, in neurons, TXNIP1 was downregulated only in the presence of H 2 O 2 , but not after IGF-I treatment ( Figure 6C). Thus, IGF-I down-regulates TXNIP1 only in astrocytes, not in neurons.
We then analyzed possible pathways involved in the inhibitory effect of H 2 O 2 and IGF-I on TXNIP1. Using kinase inhibitors we ruled out the idea that the main kinases downstream of the IGF-I receptor or H 2 O 2 were involved. In fact, inhibition of most of these kinases resulted in altered basal levels of TXNIP1 (not visualized), suggesting that basal levels of this protein are tightly regulated in astrocytes. Other inhibitory drugs of different pathways where IGF-I participates (PKC, PKA, CnA, PDK-1, NFκB, among others) gave similar negative results. However, inhibition of Ca ++ flux with 5 μM BAPTA abrogated TXNIP1 decreases in response to either H 2 O 2 or IGF-I while only slightly, but not significantly affecting basal levels ( Figure 6D).

IGF-I cooperates with SCF produced by astrocytes to protect neurons against oxidative stress
We next analyzed possible neuroprotective effects of IGF-I through astrocytes. Using a commercial gene array for growth factors we screened growth factor production by IGF-I-treated astrocytes in response to H 2 O 2 . Among the several growth factors that increased, stem cell factor (SCF) showed the highest elevation (Table 3). We confirmed by qPCR that SCF mRNA was increased after H 2 O 2 whereas IGF-I decreased it ( Figure 7A upper panel). Accordingly, levels of soluble SCF (sSCF) in culture medium from astrocytes treated with H 2 O 2 were also increased ( Figure 7A, lower panel). As SCF has been shown to be neuroprotective 21 , we determined whether it protects neurons against H 2 O 2 and found that while SCF alone did not exert any protection, co-treatment with IGF-I resulted in significantly greater neuronal survival (p<0.05; Figure 7B). We then examined pathways underlying this cooperative action of IGF-I and SCF. Under basal conditions, the activity of extracellular signal-regulated kinase (Erk; measured as pErk/Erk ratio), a canonical kinase in IGF-I signalling, was increased by IGF-I as expected, and to a lesser extent also by SCF ( Figure 7C). Basal Erk activity was also increased by H 2 O 2 . However, Erk was no longer activated by IGF-I or SCF in the presence of H 2 O 2 . Only when both were added together to H 2 O 2 -challenged cultures Erk activity was increased ( Figure 7C). No interactions were found with Akt, the other canonical kinase pathway activated by IGF-I.
To determine the in vivo relevance of these observations we submitted mice to brain ischemia as this brain insult is associated to oxidative stress 22 , and both IGF-I 9 and SCF 23 have been shown to be neuroprotective after ischemia. We found that IGF-I mRNA is increased after middle cerebral artery occlusion (MCAO) both in the ipsilateral and contralateral cortex, while only the contralateral side showed increased SCF mRNA levels compared to intact mice A battery of growth factors was screened with an RT 2 Profiler TM PCR Array. In brief, astrocytes were treated or not with IGF-I+H 2 O 2 for 16 h and total RNA was isolated. After performing the RT-PCR, total cDNA was assayed for PCR Array. PCR data was analyzed with the RT 2 Profiler PCR Array Data Analysis version 3.5 software provided by the manufacturer. Significantly up-or downregulated genes are shown.
( Figure 7D). However, levels of SCF protein were elevated after MCAO in both the damaged and contralateral sides compared to normal mice ( Figure 7E). This suggests that after brain ischemia the contralateral cortex produces higher amounts of SCF that eventually reach the ischemic side. Under this condition IGF-I may interact with SCF to promote neuronal survival in the ipsilateral cortex.

Discussion
The present results indicate that IGF-I exerts a protective action on astrocytes contributing to the resilience of these glial cells against oxidative stress. IGF-I also cooperates with astrocytes to protect neurons. These observations highlight the importance of cellspecific and cell-cooperative aspects of IGF-I protection against A representative blot is shown. Six, five and four animals were used per group, respectively. Levels of actin in each sample were measured to normalize for total protein levels.
oxidative challenge. Thus, a better understanding of the trophic role of IGF-I in the brain requires taking into account its effects on astrocytes (and other brain cells) and the functional links of these cells with neurons. While these observations do not help settle the role of oxidative stress in brain aging they put forward an important aspect of possible mechanisms involved in aging; regulatory signals such as IGF-I may not modulate the response of the different cells and even tissues to oxidative stress in the same way.
The protection provided to astrocytes by IGF-I against oxidative stress may contribute to the greater resilience of these cells to oxidative challenge. In addition, astrocytes are coupled to neurons in the response to oxidative stress and provide them with ample detoxification support 11 . Among different anti-oxidant defences provided by astrocytes to neurons, we now find that IGF-I, which cannot protect isolated neurons against excess ROS 12 cooperates with SCF secreted by astrocytes to support neurons (Figure 8). While in response to oxidative stress the production of IGF-I by cultured astrocytes and neurons is decreased, after brain ischemia IGF-I levels are actually higher due to increased synthesis and accumulation in microglia, vessels and astrocytes 24 . Therefore, in vivo, astrocytes and neurons will receive IGF-I input from various local sources, suggesting that the response of increased IGF-I after brain ischemia reflects an endogenous neuroprotective mechanism against oxidative injury. This conclusion apparently contradicts other evidence that IIS activity is pro-oxidant. Thus genetic ablation of ISS components in the nematode Caenorhabditis elegans 25 or in higher organisms such as the fruit fly 26 or mice 27 , increases organism resistance to oxidative stress. For example, mice with reduced IGF-I activity (hemizygous for the IGF-I receptor) have lower levels of ROS in the brain 28,29 . However, these mice developed greater cell damage after oxidative injury 29 . Conceivably, the effects of modulating IGF-I signalling prior to ROS insult (as when using genetic models) may not be the same compared to after insult. For example IGF-I protects nerve cells and/or the brain against diverse types of ROS-related insults [30][31][32][33][34] . In this regard, we recently reported that in a cellular model of Friedreich's ataxia (which elicits oxidative damage) neurons responded to IGF-I only when they became frataxin deficient, but not under normal conditions 15 . Collectively these observations emphasize the importance not only of cell type but also of context dependency of IGF-I neuroprotection in relation to oxidative stress.
A role for oxidative stress in many neurodegenerative diseases is gaining increasing acceptance 35 . Aberrant production of ROS in the central nervous system is linked to neurodegenerative diseases such as Alzheimer's dementia, Parkinson's disease or stroke, all of them associated to aging 36 . However, as already commented, the role of oxidative stress in brain aging is still unclear. An attempt to explain these apparently opposing observations is that moderate ROS levels may activate survival pathways 37 . The present findings agree with this proposal. Thus, doses of H 2 O 2 up to 100 μM do not elicit astrocyte death probably because IGF-I helps maintain their anti-oxidant capacity. In this regard our results show that astrocytes in response to IGF-I and/or H 2 O 2 activate antioxidant signalling including upregulation of Cu/ZnSOD and MnSOD coupled to downregulation of pro-oxidant proteins such as Txnip1. Txnip1 inhibits thioredoxin (Trx), a protein that reduces protein disulfides as well as H 2 O 2 . The Txnip-Trx axis plays an important role in different brain diseases in which oxidative stress is implicated 38 .
There is ample evidence that different trophic factors, including SCF 39 , contribute to reduce cell damage due to oxidative stress after brain stroke 40 . We have found that in vitro IGF-I and SCF exert a cooperative neuroprotective effect against oxidative stress, suggesting that they may exert a similar beneficial role in vivo as after brain stroke both factors are upregulated in the lesioned area. Indeed, a cooperative neuroprotective effect of SCF with insulin has been reported 41 . The intracellular mechanisms mediating cooperation between these two factors involve Erk, a kinase activated by IGF-I.  Figure 3A), and probably also through astrocytes. In the presence of H 2 O 2 (right side) the actions of IGF-I on neurons and astrocytes can be summarized in 5 points: 1) IGF-I loses its ability to directly protect neurons, 2) IGF-I secretion by astrocytes is diminished, 3) IGF-I reinforces astrocyte defences against oxidative stress by down-regulating pro-oxidant mechanisms such as TXNIP1. 4) IGF-I cooperates with SCF secreted by astrocytes to promote neuronal survival. 5) However, the precise mechanism(s) downstream of astrocyte IGF-I receptors underlying enhanced astrocyte neuroprotection remains to be determined. Cytotoxic effects are depicted in red while cytoprotective actions are indicated in blue trace. 2.
1. In the revised article, the authors have addressed most of my concerns. There still remain a few relatively minor issues, mostly involving data presentation / manuscript readability rather than scientific validity.

Open Peer Review
As in the prior version of the manuscript, the authors claim in Fig. 8 (model figure) that IGF-1 has a neuroprotective effect under basal conditions (i.e. in the absence of H O ) and now explicitly cite Fig. 3A in support of this claim. However, as previously, Fig. 3A includes no information on the statistical significance of IGF-1 effect on the neuronal death in the absence of H O (by eye, the IGF-1 effect seems very modest). In the legend for Fig. 3A, the authors state that "#" means " " but symbol "#" is not used in the figure -perhaps it has p<0.05 vs control been accidentally omitted? I appreciate that, as I suggested, the authors repeated statistical analyses for many experiments and are now using two-way rather than one-way ANOVA where appropriate. However, the results of this analysis are buried in the figure legends and are described in a fairly non-transparent, difficult to understand way; the figures themselves are largely unmodified. To make the paper more accessible to future readers, the graphs for experiments analyzed by 2-way ANOVA should show statistical significance for all post-hoc comparisons that were performed (including the effects that were not significant, to make this clear); for a 2x2 experimental condition grid this means all 4 post-hoc comparisons should be shown in terms of statistical significance. As an example, for the experiment in Fig. 2B, the authors should show statistical significance for (1)  In response to the minor points raised by the reviewer: The symbol # in Figure 3A is missing due to a typographical error. 2. 1.
The symbol # in Figure 3A is missing due to a typographical error.
We now incorporate the requested This manuscript by Laura Genis and co-workers is very interesting and the data presented are of significant scientific value. These studies provide further understanding of the cellular and molecular mechanisms involved in neuronal damage and rescue following oxidative stress. Of particular interest are the protective effects of IGF-I on neuronal cells mediated via the astrocytes. To these, the synergistic/additive effects of SCF on IGF-I are novel and interesting.
I have only few minor suggestions -mainly to improve graphic illustrations: In figure 1B, IGF-I ELISA, the conditioned media from astrocytes and neuronal cells is analysed for IGF-I levels but the values are express in ng/ug of protein, why?
In Figure 3E I asume that the graph shows the % phosphorylation for the IGF-I treatment? This should be properly label.
In Figure 4 a statistical significance is shown for the 'untreated', but it should shown for the IGF-I which actually prevents MitoO2. I have a similar comment for the graph below in figure 4B.

I have read this submission. I believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard.
No competing interests were disclosed. In this paper, the authors attempt to elucidate the role of IGF-I in the astrocyte-mediated protection of neurons against oxidative stress. While this is an important topic and the authors present a lot of interesting data, the paper is unfocused and the results do not fully support the conclusions. As such, I feel that the paper cannot be approved for indexing until substantively revised.

Major comments:
The paper presents three essentially independent sets of data and then tries to connect them into a single coherent story, without direct experimental evidence that it is appropriate to do so. Specifically, Figs. 1 and 2 present evidence that IGF-I plays an important role in the astrocyte-mediated neuroprotection, both at baseline conditions and under oxidative stress; this is the most interesting part of the paper (and the part that is most relevant to the paper's current title). Figs. 3-6 show data that elucidate some aspects of IGF-I effect on astrocytes, but do not establish the importance of these effects/mechanisms for IGF-I-and astrocyte-mediated neuroprotection.
(Notably, the authors have actually established that one of these mechanisms, IGF-I-induced decrease in the expression of astrocytic TXNIP1, play a role in the astrocyte-mediated does not neuroprotection. Surprisingly, these key results are not shown despite the fact that an entire figure [ Fig. 6] is devoted to the IGF-I modulation of TXNIP1).
Finally, Fig. 7 shows that IGF-I and SCF applied together (but not separately) have a neuroprotective effect in the absence of astrocytes and that their expression is increased in an in stroke model. However, the authors again fail to show that these observations are in any way vivo relevant for the astrocyte-mediated neuroprotection shown in Figs. 1 and 2. To connect these currently unconnected experimental threads, the authors need to use their neuron-astrocyte co-culture system to establish the link between astrocyte-mediated neuroprotection and (1) IGF-I-mediated increase in the expression of astrocyte antioxidant enzymes, (2) IGF-I-mediated decrease in the astrocyte ROS levels, and (3) astrocyte secretion of SCF; the experiments should be performed both at baseline and under oxidative stress conditions. Alternatively, the paper needs to be re-written in a way that makes it very clear (1) that the data presented in the paper represent a series of independent observations that do not add up to a coherent whole and (2) that the mechanism mediating IGF-I-induced neuroprotection in the mixed neuron-astrocyte environment currently remains unexplained. If choosing this route, the authors should also change the title of the paper to something more neutral and descriptive.
The authors do not show some important experimental data, ostensibly "for clarity"; these data need to be included in the revised paper. Specifically, as already stated in point #1, the authors mention (but do not show) that depleting astrocytes of TXNIP1 does not result in increased neuronal survival (page 16). This finding, if properly established, indicates that the IGF-I-mediated " when in fact, IGF-I has no clear neuroprotective effect when applied to directly onto neurons neurons cultured alone (Fig. 3A, 0 µM H O condition) -the IGF-I-induced decrease in neuronal cell death is small [~5% based on the graph] and does not appear statistically significant, although the authors do not comment on this one way or the other. Similarly, the figure legend mentions that IGF-1 down-regulates astrocytic TXNIP1 -a finding that is accurate but not relevant for the astrocyte-mediated neuroprotection illustrated by the figure. Thus, Fig. 8 should either be altered to more meaningfully represent the paper's findings or, if the authors decide to re-write the paper in a more descriptive fashion, could be eliminated altogether.

Minor comments:
Why are neurons cultured under depolarizing conditions (25 mM KCl)?
The co-culture experimental set-up should be described under a separate heading, not buried under " ".

Cell assays
For similar experiments, the authors should use a similar type of plot to make the paper more readable. For example, experiments in Figs. 3A and 3B have a very similar overall design -why are the results shown very differently? Neuronal viability is established by counting " "; the authors therefore need to all GFP-positive cells show that their neuronal cultures are at least 95% and preferably 99% pure (i.e., do not contain a significant population of GFP-positive glial cells).
Representative flow cytometry plots should be included in Fig. 4A.
To enhance readability, all figure panels should be labelled "astrocytes", "neurons", or "neurons + astrocytes", as appropriate. (The information is currently largely buried in the figure legends.) In Fig. 5, authors should specify whether they are measuring mRNA or protein level (I assume the latter, but it's difficult to be sure).
I have read this submission. I believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above.
No competing interests were disclosed. Competing Interests: 2 2

Ignacio Torres Aleman
We would like to respond to some of the points raised by the reviewers.
In response to points raised by the first reviewer : Marta Margeta

1: Why are neurons cultured under depolarizing conditions (25 mM KCl)?
As shown in cerebellar granule cells require http://cshprotocols.cshlp.org/content/2008/12/pdb.rec11550 25 mM KCl to survive in vitro. We checked ourselves a long time ago this specific requirement and confirmed that without it these neurons do not survive. 4: Neuronal viability is established by counting "all GFP-positive cells"; the authors therefore need to show that their neuronal cultures are at least 95% and preferably 99% pure (i.e., do not contain a significant population of GFP-positive glial cells).
We established neuronal-enriched culture methods long ago using neurofilament (Fig 1A in  Neurosci., Cell morphology also helps to avoid counting non-neuronal GFP cells.
In response to points raised by the third reviewer : Vince Russo 1: In figure 1B, IGF-I ELISA, the conditioned media from astrocytes and neuronal cells is analysed for IGF-I levels but the values are express in ng/ug of protein, why?
IGF-I levels are expressed as ng/µg protein because supernatants were concentrated by lyophilization and re-suspended in ELISA buffer. Protein in supernatants was measured by Bradford to normalize IGF-I levels.
No competing interests were disclosed.

Ignacio Torres Aleman
We appreciate the comments of the reviewer. We understand that we did not succeed in conveying the notion that IGF-I exerts specific actions on astrocytes related to Txnip1 ...etc that do not relate directly to neuroprotection by astrocytes; we may call it " ". We will carefully re-write the astroprotection by IGF-I manuscript to deal with this problem. The experiments with SCF were already done but not included in the manuscript so we now can incorporate them: SCF secretion by astrocytes increases in response to