Neuroprotection of microglial conditioned medium on 6-hydroxydopamine-induced neuronal death: role of transforming growth factor beta-2


Address correspondence and reprint requests to Barbara Monti, Department of Biology, University of Bologna, Via Selmi 3, 40126 Bologna, Italy. E-mail:


Microglia, the immune cells of the CNS, play essential roles in both physiological and pathological brain states. Here we have used an in vitro model to demonstrate neuroprotection of a 48 h-microglial conditioned medium (MCM) towards cerebellar granule neurons (CGNs) challenged with the neurotoxin 6-hydroxydopamine, which induces a Parkinson-like neurodegeneration, and to identify the protective factor(s). MCM nearly completely protects CGNs from 6-hydroxydopamine neurotoxicity and at least some of the protective factor(s) are peptidic in nature. While the fraction of the medium containing molecules < 30 kDa completely protects CGNs, fractions containing molecules < 10 kDa or > 10 kDa are not neuroprotective. We further demonstrate that microglia release high amounts of transforming growth factor-β2 (TGF-β2) and that its exogenous addition to the fraction of the medium not containing it (< 10 kDa) fully restores the neuroprotective action. Moreover, MCM neuroprotection is significantly counteracted by an inhibitor of TGF-β2 transduction pathway. Our results identify TGF-β2 as an essential neuroprotective factor released by microglia in its culture medium that requires to be fully effective the concomitant presence of other factor(s) of low molecular weight.

Abbreviations used



activin receptor-like kinase


brain derived neurotrophic factor


Basal Medium Eagle


cerebellar granule neurons


insulin-like growth factor-I


microglia-conditioned medium


3-(4,5-dimethylthiazol-2-yl)2,5-diphenyl-tetrazolium bromide


phosphate-buffered saline


Parkinson’s disease


transforming growth factor-β

Microglia are the main immune effector cells in the brain and are particularly sensitive to changes in the surrounding environment, becoming readily activated in response to brain infection or injury (for review see Garden and Möller 2006; Hanisch and Kettenmann 2007). Insights into the physiological roles of microglia in healthy brain have been obtained slowly in the beginning, but our understanding has rapidly improved during the last few years. By now, a great deal of information has been obtained on essential roles of microglia in influencing brain function from development to aging, in promoting neuronal survival and in maintaining neural circuits (Streit et al. 1999; Streit 2002; Raivich 2005; Schwartz et al. 2006). Microglial activation is primarily intended to protect neurons and is a component of the regenerative process. However, in several neuropathologies, where chronic inflammation is present, the inflammatory products derived from activated microglia may secondarily promote neurodegeneration and contribute to neuronal loss (González-Scarano and Baltuch 1999; Polazzi and Contestabile 2002; Streit et al. 2008). In this setting, the possibility to shift the balance of microglial cells from pro-inflammatory to anti-inflammatory states offers a good strategy to develop therapies for neurodegenerative diseases. The knowledge of the neuroprotective mechanisms of microglial cells, emerging from studies of in vitro neuron/microglia cross-talk, may help in identifying and targeting specific microglial-produced molecules that could account for neuronal survival. In recent years, we and others have performed in vitro studies based on co-culture of neurons and microglia or on the use of conditioned media to elucidate the neuroprotective functions of microglial cells (Toku et al. 1998; Zietlow et al. 1999; Park et al. 2001; Polazzi et al. 2001; Figueiredo et al. 2008;Lai and Todd 2008). In particular, we have demonstrated that microglia conditioned media prevent apoptosis of cerebellar granule neurons (CGNs) induced by shift to low potassium and that diffusible signals from apoptotic neurons increase microglial neuroprotective action (Polazzi et al. 2001). By using this approach, we have also recently demonstrated that apoptotic death of CGNs caused by staurosporine, as well as by a mild excitotoxic stimulus delivered through sub-chronic glutamate treatment, was significantly reversed by microglia conditioned media, while a stronger and acute excitotoxic insult, resulting in prevailingly necrotic death, was only marginally counteracted by the same media (Eleuteri et al. 2008). This suggests that microglial cells, at least in vitro, release specific molecules that could preferentially rescue neurons from apoptotic death with respect to necrotic one. The peculiar state of the microglia detached from mixed glial cultures and maintained in in vitro conditions of very high enrichment (nearly 100% pure microglia) is not a simple functional condition. According to a now widely-accepted idea, cultured microglia is not the exact equivalent of the so-called ‘resting microglia’ present in healthy brain. Rather, cultured microglial cells are activated in a way similar to the one of microglia present in chronic neurodegenerative diseases, but different from the one of over-activated microglia characterizing acute brain insults (Streit et al. 1999). These chronically activated in vivo microglia are likely beneficial for maintenance of brain function, as also suggested by the fact that age-related neurodegenerative processes are accompanied by microglia senescence and by its consequent elimination (Streit 2006;Streit et al. 2008). This strongly suggests that their counterpart in culture should theoretically perform similar roles.

Starting from previous observations, in the present study we have extended our in vitro‘microglial neuroprotective model’ to an in vitro model of Parkinson-like neurodegeneration, i.e. 6-hydroxydopamine (6-OHDA)-induced death in CGNs, a widely used model to study mechanisms of neuronal survival/death (Contestabile 2002). The neurotoxin 6-OHDA has been shown to induce CGNs death through production of free radicals, alteration of protein degradation mechanisms and accumulation of ubiquitinated proteins (Dodel et al. 1999; Lin et al. 2003; Chen et al. 2004; Ma et al. 2006; Monti et al. 2007), all typical features of the Parkinson’s Disease (PD)-like death caused by the toxin on its specific target represented by dopaminergic neurons of the substantia nigra (reviewed by Blum et al. 2001; Bovéet al. 2005; Olanow 2007). Here we report that a medium conditioned by microglia for 48 h (MCM) is highly protective towards CGNs death induced by 6-OHDA. Through various experiments of inactivation and fractionation of the medium, as well as through protein identification and pharmacological manipulation, we also demonstrate that full neuroprotection is granted by the medium containing substances under 30 kDa of molecular weight, that more than one protective molecules must be present in MCM and that transforming growth factor-β2 (TGF-β2) released by microglia is one of these molecules.

Materials and methods

Microglial cell culture and microglial conditioned medium preparation

Microglial cells were prepared from cerebral cortex of newborn Wistar rats, as previously described (Levi et al. 1993). All animal experiments were authorized by a local bioethical committee and were performed in accordance with the Italian and European Community law on the use of animals for experimental purposes. Briefly, brain tissue was cleaned from meninges, trypsinized for 15 min at 37°C and, after mechanical dissociation, cell suspension was washed and plated on poly-l-lysine (Sigma-Aldrich, St. Louis, MO, USA, 10 μg/mL) coated flasks (75 cm2). Mixed glial cells were cultured for 10–13 days in Basal Medium Eagle (BME, Invitrogen, Paisley, UK) supplemented with 100 mL/L heat-inactivated fetal bovine serum (Invitrogen), 2 mmol/L glutamine (Sigma-Aldrich) and 100 μmol/L gentamicin sulphate (Sigma-Aldrich). Microglial cells were harvested from mixed glial cells cultures by mechanical shacking, resuspended in serum free BME and plated on uncoated 40 mm dishes at a density of 1.5 × 106 cells/1.5 mL medium/well. Cells were allowed to adhere for 30 min and, then, washed to remove non-adhering cells. After 48 h, MCM was collected, filtered through 0.22 μm filters, aliquoted and stored at −20°C until used. To get insight into the nature and the range of molecular weight of the neuroprotective factor(s) present in MCM, some aliquots were boiled for 30 min before use. Other aliquots of MCM were treated with 1 μg/mL of proteinase K or peptidase (Sigma-Aldrich) for 1 h at 37°C. Enzymatic activity was then eliminated by heating the media at 95°C for 10 min, a procedure that did not compromise the neuroprotective efficacy of MCM. Finally, other aliquots of MCM were filtered by using Microcon-YM-10 or Microcon-YM-30 filters (Millipore Co., Billerica, MA, USA). In this way, we obtained different fractions: one containing low molecular weight (< 10 kDa or < 30 kDa) substances and the others containing high molecular weight (> 10 kDa or > 30 kDa) substances. The fractions were separately reconstituted in the original volume of medium to recreate a similar concentration of active factors and then used to test protection of neuronal cultures. In some experiments, to evaluate TGF-β2 neuroprotective role, we treated MCM with a neutralizing antibody specific for TGF-β2 (T4442 from Sigma-Aldrich or sc-90 Santa Cruz Biotechnology Inc., Santa Cruz, CA, USA), as previously shown by (Elvers et al. 2005). Briefly, 20 μg/mL of the neutralizing antibody was added to MCM and incubated for 1 h at 24°C, before testing its neuroprotective action. In a set of experiments, to evaluate the neuroprotective effect of TGF-β2, we have added 10 ng/mL of human recombinant TGF-β2 (Sigma-Aldrich) to the fraction of MCM < 10 kDa that does not contain this cytokine. We have also evaluated whether the neuroprotective effect of MCM was inhibited when the signal transduction cascade of TGF-βs was blocked by the compounds SB 431542 or SB 525334 (Sigma-Aldrich) (Inman et al. 2002; Laping et al. 2002; Grygielko et al. 2005).

Cerebellar granule neurons cultures

Primary cultures of cerebellar granule neurons cultures (CGNs) were prepared from 7-day-old rats of Wistar strain, as previously described (Gallo et al. 1987). Briefly, cells were dissociated from cerebella and plated on 40 mm dishes or in 24-well plates, previously coated with 10 μg/mL poly-l-lysine, at a density of 2 × 105 cells/cm2 in BME supplemented with 100 mL/L heat-inactivated fetal bovine serum, 2 mmol/L glutamine, 100 μmol/L gentamicin sulphate and 25 mmol/L KCl (all from Sigma-Aldrich). 16 h later, 10 μM cytosine arabino-furanoside (Sigma-Aldrich) was added to avoid glial proliferation. After 7 days in vitro, differentiated neurons were shifted to serum free BME medium (Invitrogen) containing 25 mmol/L KCl and treated with different concentration of 6-OHDA (Sigma-Aldrich), in presence or absence of MCM. We have also tested the neuroprotective effect of the different fractions of MCM, MCM previously heat-inactivated or MCM previously treated with proteinase K or peptidase (see above). All microglia conditioned media or their fractions were added with KCl to reach a final 25 mmol/L K+ concentration. After 24 h neuronal survival was analyzed through 3-(4,5-dimethylthiazol-2-yl)2,5-diphenyl-tetrazolium bromide (MTT) test or nuclei count, after Hoechst staining.

MTT assay

The viability of CGNs cultures was evaluated by thiazolyl blue (MTT) assay (Hansen et al. 1989). This method is based on the conversion of the tetrazolium salt to a colored compound, a reaction that only occurs in viable cells since the chemical reaction is carried on by mitochondrial dehydrogenases. MTT (Sigma-Aldrich) was added to the culture-medium to reach a final concentration of 0.1 mg/mL. Following 15 min of incubation at 37°C in the dark, formed crystals were dissolved in 0.1 mol/L Tris-HCl buffer containing 50 mL/L Triton X-100 (all from Sigma-Aldrich) and the absorbance was read at 570 nm in a Multiplate Spectophotometric Reader (Bio-Rad Laboratories, Hercules, CA, USA).

Hoechst staining

To quantify neuronal cell death, normal and condensed nuclei were counted after Hoechst staining. CGNs were fixed for 20 min with 40 g/L paraformaldehyde in phosphate buffer, washed in phosphate-buffered saline (PBS) and incubated for 5 min at room temperature with 0.1 μg/mL Hoechst 33258 (all from Sigma-Aldrich). Cultures were observed and photographed with a fluorescence microscope, by using a 20× objective, and count was performed in three randomly selected fields of each dish. Quantitative evaluation of cell death was determined by calculating the percent ratio: condensed nuclei/condensed + normal nuclei (Monti et al. 2001).

Western blot analysis

To analyze protein expression of insulin growth factor-I (IGF-I), brain derived neurotrophic factor (BDNF) and transforming growth factor-βs family members, in their mature or precursor forms, on microglial cell cultures and on microglial conditioned medium, freshly detached microglial cells were plated in serum free BME at the same density used to obtain MCM. After 2 or 48 h, microglial cell were collected directly in Loading buffer 2× (0.5 mol/L Tris-HCl pH 6.8; 40 g/L sodium dodecyl sulfate; 20 mL/L Glicerol; 2 g/L Bromophenol Blue; 0.2 mol/L dithiothreitol; all chemicals were from Sigma-Aldrich). 1.5/106 cells from each well were collected in 50 μL of 2× Loading buffer and the corresponding media were also collected. 1 mL aliquots for each condition were lyophilized, by using Microcon-YM-3 (Millipore), and resuspended in 15 μL of Loading buffer 2×. Fifteen μL/lane of each of either microglial cells or concentrated MCM samples were loaded onto a 15% (wt/vol) sodium dodecyl sulfate–polyacrilamide gel (Bio-Rad). After electrophoresis and transfer into nitrocellulose membrane (GE Healthcare, Chalfont St. Giles, UK), the membranes were blocked for 1 h with a Blocking Solution made of 50 g/L non-fat dried milk (Bio-Rad)/1 mL/L Tween 20 in phosphate buffer solution (Sigma-Aldrich), pH 7.4, and incubated overnight at 4°C with primary antibodies (rabbit polyclonal anti-IGF-I, anti-BDNF, anti-TGF-β1, anti-TGF-β2, anti-TGF-β3, anti-inhibin β-A, anti-inhibin β-B, anti-Nodal and goat polyclonal anti-GDF (growth differentiation factor)-1, Santa Cruz Biotechnology, 1 : 1000) in 1 mL/L Tween 20/PBS. Subsequently, nitrocellulose membranes were incubated with a secondary antibody, a donkey anti-rabbit antibody or a donkey anti-goat antibody conjugated to horseradish peroxidase (respectively 1 : 2000 and 1 : 3000, Santa Cruz Biotechnology) for 90 min at 24°C in 1 mL/L Tween 20/PBS. The labeled bands were visualized by enhanced chemiluminescence method (ECL, GE Healthcare). To further check the equal loading of the cell samples, a staining for the β-actin content was performed on microglial samples, by using a primary monoclonal antibody developed against β-actin (Sigma-Aldrich).

Real-time PCR

For total RNA extraction, 3 × 106 cells were directly lysed in 1 mL of Tri-reagent (Sigma), according to the manufacturer’s protocol. RNA pellets were resuspended in DEPC (diethyl pyrocarbonate)-treated deionized water. RNA levels were quantified using a NANOdrop UV spectrophotometer (Bio-Rad) and stored at −80°C until used for cDNA synthesis. For each sample, 1 μg of Dnase-treated RNA (Fermentas, Burlington, Ontario, Canada) was retro-transcribed using the Superscript III First-Strand Synthesis SuperMix for qRT-PCR kit (Invitrogen) following the manufacturer’s protocol. cDNAs were stored at −20°C until used for qRT-PCR.

TGF-β1 (accession number NM_021578), TGF-β2 (accession number NM_031131), TGF-β3 (accession number NM_013174) and β-actin (accession number NM_031144) qRT-PCR primers were designed using the software Beacon Designer 7.2 (PREMIER Biosoft International, Palo Alto, CA, USA). The following primer pairs (from Sigma genosys) were used:

Product size: 173 bp
Product size: 115 bp
TGF-β3Forward: 5′ ATCCACTGTCCATGTCATACC 3′ Tm = 54.4°C
Reverse: 5′ TCATCTTCGTTGTCCACTCC 3′ Tm = 54.3°C
Product size: 101 bp
β-actinForward: 5′ AGCAGATGTGGATCAGCAAG 3′ Tm = 54.9°C
Reverse: 5′ AACAGTCCGCCTAGAAGCAT 3′ Tm = 56.0°C
Product size: 81 bp

cDNA samples were diluted to a concentration of 20 ng/μL. A master mix of IQ SYBR Green Supermix (Bio-Rad) was mixed with the adequate primers (final concentration 0.2 μmol/L), cDNA (40 ng per replicate) and deionized water. qRT-pCR was performed using an iCycler IQ (Bio-Rad) thermal cycler and real-time PCR measurements were carried out in duplicate with three biological replicates. mRNA relative expression levels were normalized using β-actin as the internal control. PCR products were electrophoresed on 2.5% (wt/vol) agarose gels containing gel-red and visualized under ultraviolet light.

Statistical analysis

All quantitative data are presented as means ± SE from independent experiments. Statistical significance between different treatments was calculated by using one way analysis of variance (anova) followed by post-hoc comparison through Bonferroni’s test. A value of p < 0.05 was considered statistically significant.


In agreement with previous results (Dodel et al. 1999; Monti et al. 2007), cultures of fully differentiated CGNs at 7 days in vitro are dose-dependently sensible to the neurotoxic insult caused by 24-h exposure to 6-OHDA, based on the MTT assay and on nuclei count following Hoechst staining (Fig. 1). Exposure of CGNs to 6-OHDA in presence of medium previously conditioned for 48 h by cultured microglial cells (microglia conditioned medium, MCM) resulted in nearly complete preservation of cell viability, evaluated through MTT assay, even at the most toxic 6-OHDA concentrations (Fig. 1a). Any in vivo toxin hitting neurons would likely also affect microglia. In order to gain better insight into the functional significance of the neuroprotection granted by MCM, we verified whether the MCM neuroprotective effect was maintained following exposure of microglia to 6-OHDA. While a toxic effect towards microglia was exerted by 6-OHDA at high concentration (data not shown), the medium conditioned by microglia exposed to 20 μmol/L 6-OHDA for 48 h protected CGNs, challenged with 6-OHDA themselves, similarly to the normal MCM (Fig. 1b). To obtain a more precise assessment of cell death caused by 6-OHDA and of the extent of the MCM neuroprotection, we counted the condensed and normal nuclei in randomly selected fields of cultures stained with Hoechst (Fig. 1c). The quantification of the condensed/total nuclei ratio confirmed that 20 μmol/L 6-OHDA was highly toxic, resulting in more than 80% CGNs death after 24 h of toxin exposure, and that MCM was able to completely revert this effect (Fig. 1d).

Figure 1.

 Microglial conditioned medium (MCM) enhances neuronal viability and survival in CGNs exposed to 6-OHDA. (a) MTT assay of CGNs cultures treated for 24 h with increasing concentrations of 6-OHDA in the presence or absence of 48 h MCM to test its neuroprotection. Each point is the mean ± SE of six different experiments run in triplicate; *p < 0.05, ***p < 0.001 with respect to the conditions of treatment with 6-OHDA in non-conditioned medium. (b) MTT assay of CGNs treated for 24 h with 6-OHDA 20 μmol/L in the presence or absence of medium obtained by microglial cells exposed themselves to 6-OHDA 20 μM for 48 h to test its neuroprotection. Each point is the mean ± SE of three different experiments run in triplicate; ≠p < 0.001 compared to controls; ***p < 0.001 with respect to the conditions of treatment with 6-OHDA in non-conditioned medium. (c) Representative Hoechst staining of apoptotic death. CGN cultures were treated for 24 h with 20 μmol/L 6-OHDA in the presence or absence of MCM. (d) Quantification based on count of condensed and normal nuclei following Hoechst staining. Bars are the mean ± SE of three different experiments run in duplicate; ≠p < 0.001 respect to the control condition, ***p < 0.001 with respect to the condition of treatment with 6-OHDA in non-conditioned medium. Calibration bar 40 μm.

In an effort to identify the nature of the putative neuroprotective factor(s) present in the conditioned medium, we exposed MCM to heath inactivation or to a pre-treatment with wide spectrum peptidases/proteases, before testing it for neuroprotection of CGNs exposed to 6-OHDA. As shown by Fig. 2(a) and (b), both types of treatments resulted in a significant, but not complete abrogation of MCM neuroprotection, thus suggesting that peptidic factor(s) certainly contributed to, but not completely accounted for this MCM neuroprotective action. To get more information on putative neuroprotective peptide(s), we performed a molecular weight-based fractionation of MCM, by passing it through filters provided with a 10 kDa cut-off, before reconstituting the medium to the initial volume and testing it for its neurotoxicity/neuroprotection on 6-OHDA-treated CGNs. By itself, none of the two fractions obtained was able to replicate the neuroprotection, as done by the complete MCM (Fig. 3a and b). Interestingly, even if the < 10 kDa fraction was neurotoxic towards CGNs per se, it significantly protected them from 6-OHDA neurotoxicity (Fig. 3).

Figure 2.

 MCM protects CGNs exposed to 6-OHDA through partial contribute of peptidic factors. CGN cultures were treated for 24 h with 30 μmol/L 6-OHDA in the presence or absence of MCM previously heat-inactivated (a) or previously enzimatically treated with proteinase K or peptidase (b). CGNs viability has been measured through MTT assay. (a) Bars are the mean ± SE of six different experiments run in triplicate; ≠p < 0.001 with respect to the conditions of treatment with 6-OHDA in non-conditioned medium; **p < 0.01 with respect to the conditions of treatment with 6-OHDA in non-heat-inactivated MCM. (b) Bars are the mean ± SE of six different experiments run in triplicate; ≠p < 0.001 with respect to the conditions of treatment with 6-OHDA in non-conditioned medium. **p < 0.01 with respect to the conditions of treatment with 6-OHDA in non-enzymatically-treated MCM.

Figure 3.

 Absence of neuroprotection by fractions of MCM obtained through filtration with cut-off at 10 kDa molecular weight. CGNs cultures were treated for 24 h with 20 μmol/L 6-OHDA in the presence or absence of MCM or its fractions containing high molecular weight (> 10 kDa) substances and low molecular weight (< 10 kDa) substances. Neuronal death has been measured by Hoechst staining (a), followed by nuclei count (b). Bars are the mean ± SE of three different experiments run in duplicate ≠p < 0.001 with respect to the control medium, #p < 0.001 with respect to control medium and MCM; **p < 0.01, ***p < 0.001 with respect to the conditions of treatment with 20 μmol/L 6-OHDA with non-conditioned medium. Calibration bar 40 μm.

To more precisely target putative neuroprotective peptide(s), we tested MCM for the presence of proteins known to be neuroprotective. In particular, we performed western blot analysis of both cell lysates and MCM to verify whether some of these potentially neuroprotective factors were synthesized by microglial cells and accumulated in the medium. We decided to analyze the expression of BDNF, IGF-I and TGF-β proteins, as they are known to be potentially synthesized by microglia and to be neuroprotective towards CGNs (D’Mello et al. 1993; Lindholm et al. 1993; Nakajima et al. 2001; Elvers et al. 2005). As shown in Fig. 4, BDNF appeared to be synthesized, but not released in the medium by microglia, while IGF-I, TGF-β1 and TGF-β3 were both present in the cells and released into the medium conditioned for 48 h in their precursor high molecular weight forms. TGF-β2 precursor was constitutively synthesized by microglia and accumulated into the medium in its active form of estimated molecular weight of 12.5 kDa, reaching an apparently high level after 48 h of culture. In order to exclude that non-specific TGF-β-like ligands (see below) could interfere with the revealed immunoreactivity, we verified through real time PCR whether microglial cells actually synthetized TGF-β mRNAs. All the three TGF- β mRNAs were synthesized by microglial cells in culture both shortly after plating and after 48 h of culture, as shown by the presence of bands of the expected size in the gel and by real time PCR quantification (Fig. 5).

Figure 4.

 Protein expression analysis of potential neuroprotective factors in microglial cells cultures and microglial conditioned medium. Microglial BDNF, IGF-I, TGF-β1, TGF-β2 and TGF-β3 expression and release into the conditioned medium have been detected by using western blot analysis on both microglial cell lysates (after 2 h and 48 h of culture) and concentrated aliquots of MCM (at 2 h and 48 h). β-Actin staining on microglial cell samples has been performed as a loading control.

Figure 5.

 PCR analysis and quantification of TGF-β1, TGF-β2 and TGF-β3 expression in microglial cells cultures. Microglial TGF-β1, TGF-β2 and TGF-β3 RNA expression have been detected by using specific primer pairs through real-time PCR analysis in microglial cell at 2 h and 48 h. (a) Table of TGF-β1, TGF-β2 and TGF-β3 expression in Δct relative to actin (Δct = Ct of each gene-Ct of β-actin) at 2 h and the fold-expression of 48 h compared to 2 h. For the 2 h stage, three experiments in duplicate were run; for the 48 h, four experiments in duplicate; b is the negative control, without cDNA. (b) Representative gel of PCR analysis of microglial TGF-β1, TGF-β2 and TGF-β3 RNA expression at 2 and 48 h (b is the negative control).

Focusing on TGF-β2, which was the only tested peptide present in MCM in its mature form, by densitometric quantification of its content in western blots from MCM compared with western blots made from known amounts of recombinant TGF-β2, we estimated a peptide concentration of about 60 ng/mL in media conditioned for 48 h (data not shown). Therefore, TGF-β2 appeared to be a promising candidate as a possible neuroprotective agent, being released in MCM at high concentration in its active form and having previously shown a neuroprotective action towards CGNs and in PD-models (Elvers et al. 2005; Andrews et al. 2006). Attempts to abrogate MCM neuroprotection by blocking TGF-β2 with neutralizing antibodies (T4442 from Sigma-Aldrich or sc-90 Santa Cruz Biotechnology Inc.) were hampered by the fact that concentrations of antibodies still unable to remove the high amount of the peptide present in the medium, as verified through immunoprecipitation experiments, resulted toxic by themselves to CGNs (data not shown). We, therefore, tried to demonstrate the involvement of TGF-β2 in MCM neuroprotection by using two different approaches. First, we tested whether addition of exogenous TGF-β2 to the MCM fraction, previously depleted of it through the fractionation procedure, was able to restore a neuroprotective action similar to the complete medium. This was actually the case, as supplementation of the < 10 kDa MCM fraction with 10 ng/mL, a concentration known to exert in vitro biological activity (Elvers et al. 2005) fully protected CGNs from exposure to 6-OHDA (Fig. 6a). Noticeably, the same addition of TGF-β2 to a control medium, not previously conditioned by exposure to microglia, was ineffective. This clearly demonstrated that the efficacy of TGF-β2 neuroprotection depended in our model on the cooperative action with other substance(s) present in the conditioned medium, in particular in its low molecular weight fraction. Furthermore, addition of exogenous TGF-β2 was also able to abrogate the intrinsic toxicity of < 10 kDa MCM fraction, thus recreating the effect of the complete conditioned medium (Fig. 6a). Second, we tested the role of TGF-β2 in the neuroprotection granted by MCM by adding to this medium inhibitors of the intracellular pathways mediating the receptorial effects of TGF-βs, i.e. the compounds SB 431542 and SB 525334 that act on ALK (Activin Receptor-like Kinase)-family kinases (Inman et al. 2002; Laping et al. 2002; Grygielko et al. 2005). The neuroprotective effect of MCM was significantly attenuated by the ALK antagonist SB 431542 (Fig. 6b), thus supporting the role of TGF-β-activated pathway in MCM protection of CGNs. A similar effect was also exerted by another ALK antagonist, SB 525334 (data not shown). As both these inhibitors are not selective for ALK5, which is the specific downstream mediator of TGF- β2, we performed experiments able to exclude that other proteins of the activin family acting through ALK signaling (Schmierer and Hill 2007) were involved in MCM-mediated neuroprotection. Western blots of four members of activin family showed that only precursor forms of these proteins, with estimated molecular weight > 30 kDa, were present in the medium conditioned by microglia (Fig. 7a). As mature forms of these proteins were not detected in MCM, their involvement in neuroprotection was unlike. To experimentally support this, we tested whether the neuroprotective effect of MCM was preserved after excluding all molecules with molecular weight > 30 kDa through selective filtration. Results showed that the conditioned medium deprived of > 30 kDa molecules retained its full neuroprotective activity (Fig. 7b).

Figure 6.

 Involvement of TGF-β2 in the neuroprotective action of MCM. (a) Human recombinant TGF-β2 (10 ng/mL) was added to the fraction of MCM containing low molecular weight (< 10 kDa) substances and its neuroprotective action was compared with that of complete medium on CGN cultures treated for 24 h with 20 μmol/L 6-OHDA. Nuclei count after Hoechst staining. Bars are the mean ± SE of three different experiments run in duplicate; **p < 0.01, ***p < 0.001 with respect to the conditions of treatment with 20 μmol/L 6-OHDA in non-conditioned medium; #p < 0.001 with respect to the condition of treatment with fraction of MCM containing low molecular weight (< 10 kDa) substances without the addiction of TGF-β2. (b) A specific inhibitor of TGF-βs signal transduction pathway, SB 431542 (10 μmol/L), was added to the cultures of CGNs treated for 24 h with 20 μmol/L 6-OHDA in the presence or absence of MCM. Nuclei count after Hoechst staining. Bars are the mean ± SE of two different experiments run in duplicate; *p < 0.05, ***p < 0.001 with respect to the conditions of treatment with 6-OHDA in non-conditioned medium; #p < 0.01 with respect to the condition in the absence SB 431542.

Figure 7.

 Analysis of activin family members in microglial conditioned medium. (a) Western blot analysis of activin family factors in MCM. Microglial-derived inhibin β-A (activin β-A), inhibin β-B (activin β-B), GDF-1 (vg-1) and nodal released into 1 mL of the conditioned medium have been detected by using western blot analysis on concentrated aliquots (at 2 h and 48 h). (b) The fraction of MCM deprived of molecules of molecular weight > 30 kDa retains full neuroprotective activity. CGNs cultures were treated for 24 h with 20 μmol/L 6-OHDA in the presence or absence of MCM or its fractions containing > 30 kDa or < 30 kDa substances. CGNs viability has been measured through MTT assay. Bars are the mean ± SE of three different experiments run in triplicate; ≠p < 0.001 with respect to the conditions of control in non-conditioned medium; **p < 0.01 with respect to the conditions of treatment with 6-OHDA in absence of MCM.

As a final experiment, we tested whether the putative neuroprotective factor(s) present in the < 10 kDa MCM fraction and required to obtain the full TGF-β2-mediated protection were also heat-labile molecules. As shown by Fig. 8, this was not the case, as the neuroprotective effect of TGF-β2 was exactly the same when added to the native or to the heat-inactivated < 10 kDa fraction, suggesting a non-peptidic nature of these low molecular weight factor(s).

Figure 8.

 Low molecular weight, heat-stable factors cooperate with TGF-β2 in fulfill its neuroprotection. Human recombinant TGF-β2 (10 ng/mL) was added to the fraction of MCM containing low molecular weight (< 10 kDa) substances, previously heat-inactivated, then its neuroprotective action was compared with that of not heat-exposed medium on CGNs cultures treated for 24 h with 20 μmol/L 6-OHDA. Bars represent the mean ± SE of two different experiment run in duplicate ≠p < 0.001 with respect to the conditions of treatment with 6-OHDA in non-conditioned medium; ***p < 0.001 with respect to the corresponding conditions in the absence of TGF-β2.


With the present report, we have demonstrated a strong and highly-specific neuroprotective action of MCM on an in vitro model of Parkinson-like neurodegeneration, the 6-OHDA-induced CGNs death. This neuroprotection is exerted through the release into the medium of peptidic molecules, which cooperate with low molecular weight, heat-resistant factor(s). Moreover, based on multiple evidence, we have identified TGF-β2 as one of these neuroprotective agents.

As outlined in the introductory section, microglia can secrete inflammatory cytokines and other neurotoxic molecules, but also neuroprotective factors. A robust inflammatory reaction is observed in PD and other neurodegenerative pathologies and reactive microglia are prominent in affected brain areas of these patients (recently reviewed by Klegeris et al. 2007; McGeer and McGeer 2008; Smith 2008; Villoslada et al. 2008;Whitton 2007). Therefore, blocking inflammation or shifting the balance from pro-inflammatory to anti-inflammatory states of microglia offers a promising strategy for therapy. In this setting, the knowledge of the neuroprotective mechanisms of microglia, coming out from in vitro studies of neuron/microglia cross-talk, may help to identify and target specific microglial molecules accounting for neuronal survival. With the present report, we demonstrate a novel neuroprotective action of MCM on 6-OHDA neurotoxicity in CGNs, having previously described similar effects in different models of CGNs induced apoptosis (Polazzi et al. 2001; Eleuteri et al. 2008). 6-OHDA is a neurotoxin with selectivity for dopaminergic neurons because of the presence of the dopamine transporter expressed in these cells, but CGNs are quite sensitive to 6-OHDA too (Dodel et al. 1999; Lin et al. 2003; Chen et al. 2004; Ma et al. 2006; Monti et al. 2007). CGNs have been used here as they are commonly considered a good model to study the cellular, biochemical and molecular mechanisms of neuronal survival/death and are among the most widely-used neuronal primary cultures (Contestabile 2002). In this model, 6-OHDA results in neurodegeneration presenting most of the characteristic features of Parkinson-like neuronal death, i.e. an apoptotic process with oxidative stress, proteasome impairment and protein aggregation (Dodel et al. 1999; Lin et al. 2003; Chen et al. 2004; Ma et al. 2006; Monti et al. 2007). Here, we have found that the pro-survival action exerted by MCM completely reversed 6-OHDA neurotoxic effect, notwithstanding its high toxicity, thus suggesting a high specificity of MCM-secreted survival factors for this model of cell death. Indeed, in different models of CGNs death, MCM only partially counteracted neurotoxic insults (Polazzi et al. 2001; Eleuteri et al. 2008). Moreover, we have demonstrated here that MCM exerts its neuroprotective effect even if microglia are exposed themselves to 6-OHDA, thus suggesting that this neurotoxin does not induce microglial pro-inflammatory activation.

We further provide evidence that TGF-β2 accumulated in the medium conditioned by microglial cells is an essential, even if not unique, factor mediating the MCM neuroprotective effect. Indeed, addition of the exogenous peptide to the conditioned medium fraction, previously depleted of endogenous TGF-β2 through molecular weight-based fractionation, was able to restore the neuroprotective action of the complete medium and also to efficiently counteract the intrinsic toxicity of this < 10 kDa MCM fraction towards CGNs. On the other hand, TGF-β2 requires the presence of at least one factor released by microglia in this low molecular weight range to ensure its neuroprotective action. This is demonstrated by the fact that the > 10 kDa MCM fraction, in which TGF-β2 is present in large amount, lacks neuroprotection by itself and that the addition of the exogenous peptide to an unconditioned medium is ineffective too. The finding that TGF-β2 needs to cooperate with other factors to explicate its biological activity on CGNs is in agreement with previous data (Kane et al. 1996), demonstrating that this cytokine differentially regulates cerebellar granule cell neurogenesis depending on the presence of serum factors. Regarding the possible cooperation of other factors with TGF-β2, it should be considered that the conditioned medium here used derives from unstimulated cultured microglia. While evidence exists for functional interactions modulating TGF-β activity in response to pro-inflammatory cytokines, such as interleukin 6 (Samanta et al. 2008), no production of these cytokines does occur from primary microglia not stimulated with LPS (lipopolysaccharide) or other activating factors (Horvath et al. 2008; Vollmar et al. 2008).We also provide pharmacological evidence for the neuroprotective role of TGF-β2 contained in MCM, by significantly decreasing its efficacy through an antagonist of its main receptor/transduction system, which acts through kinases of ALK family (Laping et al. 2002; Grygielko et al. 2005). Since this pharmacological tool could not unequivocally identify TGF-β2 as the neuroprotective factor, because of the possible interference with other proteins acting through ALK pathway, we verified that our cultured microglial cells produced the specific mRNA for TGF-β2, as well as for TGF-β1 and TGF-β3. Notwithstanding some authors could not detect TGF-β2 production in microglia in early studies (Morgan et al. 1993; Constam et al. 1992), more recently other groups demonstrated the microglial expression of TGF-β2 RNA and protein both in vitro and in vivo (De Groot et al. 1999; Ma and Streilein 1999; Yuan and Neufeld 2001). Our PCR analysis revealed a continuous production of TGF-β2 mRNA from microglial cells throughout the culture period that could well account for the accumulation of the mature form of the protein in the medium after 48 h. Furthermore, we demonstrated that other proteins of the TGF-βs family, which could be involved in microglial neuroprotective action, including activins, nodal and vg-1 (Schmierer and Hill 2007), were released in the conditioned medium at molecular weight > 30 kDa and thus could not contribute to neuroprotection as MCM deprived of its > 30 kDa fraction retained its full neuroprotective capacity. These experiments, together with our demonstration that addition of exogenous TGF-β2 to the fraction of MCM previously deprived of it through selective filtration efficiently granted neuroprotection, strongly support our identification of TGF-β2 as a neuroprotective factor constitutively released by microglia in its medium.

Microglia-conditioned medium inactivation shows that neuroprotection granted against 6-OHDA toxicity is in part related to heat-resistant molecule(s), as the effect does not completely disappear after heat inactivation or peptide enzymatic destruction of the medium. This is in agreement with previous results, showing that the neuroprotective action of MCM on neurons in different neurodegenerative conditions was not completely abolished by heat inactivation (Toku et al. 1998; Zietlow et al. 1999; Park et al. 2001; Polazzi et al. 2001). Present experiments of heat inactivation on the < 10 kDa MCM fraction demonstrate the presence in this fraction of heat-resistant neuroprotective molecule(s), which are required to ensure the full protection of the complete medium or of the exogenous TGF-β2 added to this same fraction. Hypotheses on the possible nature of heat-resistant neuroprotective molecule(s) are at present speculative, as they obviously require to be subjected to specific experimental analysis. On the basis of literature data, potential candidates may belong to prostaglandins and other lipids or sugars, as low molecular weight molecules of this kind are known to be released by microglia and to result neuroprotective under some conditions (Hicks et al. 1998;Bachis et al. 2002; Carrasco et al. 2008; Zhang et al. 2008).

The family of TGF-βs comprises multifunctional growth factors abundantly expressed in the nervous system and promoting neuronal survival in a variety of physiological and neuropathological conditions (Flanders et al. 1998; Krieglstein et al. 2002). TGF-β2, in particular, protects neurons from excitotoxic insults, promotes differentiation and survival of nigral dopaminergic neurons, modulates CREB (cAMP-response element binding protein) activity in hippocampal neurons, thus promoting synaptic plasticity, and negatively regulates the inflammatory response of the central nervous system (Bruno et al. 1998; Andrews et al. 2006; Roussa et al. 2006; Fukushima et al. 2007; Siglienti et al. 2007). Cerebellar granule neurons have been demonstrated to be a target of TGF-β2 during development. In particular, TGF-β2 has been shown to take part in the neurogenesis of this neuronal population through regulation of precursor proliferation, apoptotic elimination and maturation (Constam et al. 1994; Kane et al. 1996; de Luca et al. 1996; Kaltschmidt and Kaltschmidt 2001; Elvers et al. 2005). Our results indicate that TGF-β2, by acting together with other microglia-derived molecules, represents a survival factor for these neurons when they are exposed to a neurotoxic insult, such as the one represented by 6-OHDA. The interest of this observation is increased by the fact that 6-OHDA is a known neurotoxin for dopaminergic nigral neurons found to massively degenerate in Parkinson’s disease and by the fact that TGF-β2 protects these neurons both in vitro and in vivo (Macauley et al. 2004; Roussa et al. 2006). Furthermore, transgenic mice expressing low levels of TGF-β2 showed an age-related nigro-striatal dopaminergic deficit (Andrews et al. 2006) and a recent population study preliminarily demonstrated a trend towards association of the 5′ region of TGF-β2 gene and susceptibility to PD (Goris et al. 2007; Klegeris et al. 2007). Thus, our present study should stimulate novel researches on the role played by microglia in the beginning and progression of PD, with the aim of identifying new potential targets for therapy. Another member of the same protein family, TGF-β1, has been reported to be potentially neuroprotective under various experimental conditions (Flanders et al. 1998; Zhu et al. 2002, 2004; Boche et al. 2003). Our microglial cells expressed both mRNA and precursor of the protein, but no accumulation of mature TGF-β1 was detected in MCM even after 48 h of culture, thus suggesting that this factor does not play a role in MCM neuroprotection.

The relevance of our present findings should be discussed with respect to how and to what extent the artificial situation of in vitro isolation of microglial cells reflects the situation occurring in vivo. In vivo‘resting microglia’ is actually continuously active (Raivich 2005) and ‘activated microglia’ is present in the brain not only with harmful phenotypes, but also with beneficial phenotypes that actively contribute to limit neuronal damage and to promote neuron recovery (Schwartz et al. 2006). However, this ‘neuroprotective phenotype’ is still not fully characterized. Accordingly, in vivo immunosuppressant treatment inhibits the neurodegenerative role of microglia in a PD rat model, without completely blocking microglial activation itself (Wright et al. 2008). The various microglial states are not autonomously determined by microglial cells but are, instead, the result of a continuous and reciprocal exchange of molecular messages (Kreutzberg 1996; Neumann 2001; Polazzi and Contestabile 2002, 2006; Schwartz et al. 2006). The impossibility to re-create such complex interrelationships in in vitro systems is an obvious limitation of studies based on interactions in co-cultures or on the use of conditioned media to reciprocally test the responses of the different players. On the other hand, as considered in the introduction, this approach makes easier to isolate and identify specific factors involved in microglia-neuron cross-talk, as it is the case for TGF-β2 in the present study. In the case of experiments here presented, for instance, results suggest a continuous production of neuron survival factors from microglia and, therefore, they are consistent with the idea that microglia constitutively serve a functionally supportive and neuroprotective role (Streit et al. 2008). While obtained in separate cultures, our result, demonstrating that 6-OHDA toxicity towards microglia does not disrupt the neuroprotective action of the conditioned medium, suggests that the occurrence of neurotoxic insults affecting both neurons and microglia in vivo may not compromise microglial neuroprotection. Furthermore, in line with the idea that microglia in culture may reflect a functional state of activation of these cells, in some way similar to a state found in chronic neurodegenerative diseases (Streit et al. 1999), these experimental evidences in culture represent an important conceptual advancement in understanding microglial role in normal and pathological brain and in targeting modulation of microglia activation for therapeutic purposes in neurodegenerative diseases.


The present work was supported by a young investigator grant of the University of Bologna for strategic research to B.M. The authors are grateful to L. Minghetti and D. Kirik for critically reading the manuscript. The skilful technical assistance of Miss Monia Bentivogli is gratefully acknowledged.