Necrotic neurons enhance microglial neurotoxicity through induction of glutaminase by a MyD88-dependent pathway
- 9.3k Downloads
Microglia are macrophage-like cells that constantly sense the microenvironment within the central nervous system (CNS). In the event of neuronal stress or injury, microglial cells rapidly react and change their phenotype. This response may lead to a deleterious type of microglial activation, which is often associated with neuroinflammation and neurotoxicity in several neuropathological conditions. We investigated the molecular mechanisms underlying triggering of microglial activation by necrotic neuronal damage.
Primary cultures of microglia were used to study the effect of necrotic neurons on microglial inflammatory responses and toxicity towards cerebellar granule neurons (CGN). The mouse hippocampal cell line, HT22, was used in this study as the main source of necrotic neurons to stimulate microglia. To identify the signal transduction pathways activated in microglia, primary microglial cultures were obtained from mice deficient in Toll-like receptor (TLR) -2, -4, or in the TLR adapter protein MyD88.
Necrotic neurons, but not other necrotic cell types, induced microglial activation which was characterized by up-regulation of: i) MHC class II; ii) co-stimulatory molecules, i.e. CD40 and CD24; iii) β2 integrin CD11b; iii) pro-inflammatory cytokines, i.e. interleukin 6 (IL-6), IL-12p40 and tumor-necrosis factor (TNF); iv) pro-inflammatory enzymes such as nitric oxide synthase (iNOS, type II NOS), indoleamine 2,3-dioxygenase (IDO) and cyclooxygenase-2 (COX-2) and increased microglial motility. Moreover, microglia-conditioned medium (MCM) obtained from cultures of activated microglia showed increased neurotoxicity mediated through the N-methyl-D-aspartate receptor (NMDAR). The activation of microglia by necrotic neurons was shown to be dependent on the TLR-associated adapter molecule myeloid differentiation primary response gene (MyD88). Furthermore, MyD88 mediated enhanced neurotoxicity by activated microglia through up-regulation of the expression and activity of glutaminase, an enzyme that produces glutamate, which is an NMDAR agonist.
These results show that necrotic neurons activate in microglia a MyD88-dependent pathway responsible for a pro-inflammatory response that also leads to increased neurotoxic activity through induction of glutaminase. This finding contributes to better understanding the mechanisms causing increased neuroinflammation and microglial neurotoxicity in a neurodegenerative environment.
KeywordsMicroglial Cell Microglial Activation Cerebellar Granule Neuron Necrotic Neuron Microglial Neurotoxicity
Necrotic-like neuronal death is associated with the development of several pathologic conditions, including brain ischemia subsequent to stroke or trauma , infectious-driven neuroinflammation, e.g. meningitis , HIV-associated dementia  and cerebral malaria , as well as noninfectious-driven neuroinflammation, e.g. multiple sclerosis (MS) . In such pathological conditions microglia, the resident CNS macrophage equivalents, may contribute to neurotoxicity . Upon neuronal injury, microglial cells are activated, acquiring increased motility and phagocytic activity. Moreover, expression of molecules that regulate immune responses, including MHC class I and II, FcγR, and the co-stimulatory molecules B7.1 and B7.2 are also increased during the process of microglial activation. This can be beneficial by removing cellular debris and can also support neuronal survival through the secretion of neurotrophic factors such as bFGF and NGF (reviewed in ). On the other hand, this may also lead to unfettered microglial activation with deleterious consequences  such as increased production of the pro-inflammatory cytokines IL-1β, TNF, IL-6 and IL-12 , neurotoxins including the pro-oxidant molecules nitrogen oxide (NO) and reactive oxygen species (ROS) , and metabolites such as quinolinic acid and glutamate .
Specific factors released by stressed or damaged neurons, including matrix metalloproteinase-3, (MMP-3), α-synuclein, neuromelanin , and ATP  have been shown in vitro to induce the production of inflammatory factors by microglia. On the other hand, during the last decade several other putative endogenous "danger signals" or "alarmins" , released by necrotic but not by apoptotic cells have also been shown to stimulate pro-inflammatory responses in dendritic cells and in monocytes/macrophages. These "danger signals" include the heat shock proteins (hsp) hsp60, hsp70, hsp90, and gp96 ; the calcium-binding S100 proteins , DNA, proteases, uric acid and the chromosomal protein high-mobility group B1 (HMGB1) (reviewed in ). These molecules, referred to as "damage-associated molecular patterns" , can be recognized by a series of pattern recognition receptors (PRR) that include TLRs , the receptor for advanced glycosylation end products (RAGE)  and scavenger receptors , but also by other cell surface proteins such as CD40  and CD91 , and the intracellular NOD-like receptors (NLRs) family as well, namely NALP-3 .
Phenotypic alterations and migration of microglial cells following neuronal cell damage have been extensively reported. However, the molecular mechanisms mediating this activation and microglial cell neurotoxicity require further elucidation. Here we address the effects of neuronal necrosis on microglial activation and neurotoxicity. In this paper we report that necrotic neurons, but not other necrotic cell types, can induce in microglia a pro-inflammatory phenotype that in turns causes NMDAR-mediated toxicity of neurons. This activation of and neurotoxic activity by microglia resulting from stimulation by necrotic neurons is shown to be dependent on MyD88, which mediates the up-regulation of glutaminase.
C57BL/6, B10scN (which lack the genomic region containing the tlr-4 gene), TLR-2 (C57BL/6) and MyD88 (C57BL/6) deficient mice were bred and maintained under specific pathogen-free conditions in our animal facilities in accordance with guidelines from the Animal User and Institutional Ethical Comities of the Instituto Gulbenkian de Ciência.
Primary cell cultures
Neonatal microglia cultures were prepared from newborn mice as described previously . Briefly, after removal of the meninges, brains were mechanically disrupted. Cells were cultured in high-glucose Dulbeco's Modified Eagle Medium (DMEM) medium containing Glutamax (Invitrogen, Carlsbad, CA) and supplemented with 10% FBS (Endotoxin less than 10 EU/ml, Invitrogen), 5 μg/ml insulin (Sigma-Aldrich Química, S.A, Portugal), 2.0 mg/ml L-glucose (Sigma) and 1% Penicillin-Streptomycin 50 mM (Invitrogen) for 14–21 days (naïve microglia). Granulocyte-macrophage colony-stimulating factor (GM-CSF) (Peprotech, Rocky Hill, NJ) (GM-CSF-microglia) was added at the beginning of the culture and subsequently every 3 days at a final concentration of 0.25 ng/ml. At day 14, confluent mixed glial cell cultures were shaken for 3 h at 250 rpm. Microglial cells were obtained from the supernatant after filtering through 100-μm cell strainers (BD, Becton, Dickinson and Company, Franklin Lakes, NJ). Cells were resuspended in RPMI 1640 with Glutamax (Invitrogen) supplemented with 5% FBS, 50 μM β-mercaptoethanol and 1% Penicillin-Streptomycin (RMPI-FBS). Cells were plated in 6 well plates (1 × 106) or 24 well plates (2 × 105) for 1 hour. Adherent cells were washed twice with warm Hanks balanced salt solution (HBSS) to remove cell debris and astrocytes. The purity of the primary cultures assessed by CD11b staining and FACS analysis was >90%.
Astrocytes and cerebellar granule neurons
An enriched population of primary astrocytes was obtained from the adherent cell population after shaking the mixed cell brain cultures. Primary cerebellar granule neurons (CGN) were prepared from 6-day-old mice as described previously . Briefly, cerebella were dissected and the cells were dissociated and plated at a density of 1 × 105 neurons/cm2 on poly-L-lysine (0.5 mg/ml, Sigma) coated plates. Cells were maintained in Basal Medium (BME) supplemented with 10% (v/v) FBS, 25 mM KCl (Sigma), 1% penicillin-streptomycin, 1% glutamax and 6 mg/ml glucose. After 24 hr, the medium was replaced to remove dead cells by the above culture medium supplemented with 1% (v/v) N-2 supplement and 50 μM cytosine arabinoside (Sigma) to prevent proliferation of non neuronal cells. Cultures were kept in an incubator at 37°C in 5% CO2 and were used after 8 days.
Induction of cell death
The mouse hippocampal cell line HT22 (kindly provided by J. Pocock) and WHEI 164 cells were cultured in DMEM containing 10% FBS. EL-4 cells were cultured in complete RPMI medium containing 10% FBS. All cell lines tested negative for Mycoplasma by PCR. Cells were trypsinized with Trypsin/EDTA (Invitrogen) and washed twice in endotoxin-free PBS. Cells were resuspended in PBS (107 cells/ml) and a necrotic-like cell death type was induced by 3 cycles of freeze and thaw or mechanically disrupted using a syringe and needle (0.5 × 16 mm). Cells were immediately added to microglial cells or kept on ice until they were used to minimize protease activity. Loss of cell viability was confirmed using trypan-blue (Sigma) (>95% cell death). To induce apoptosis, cells were UV-irradiated in a UV Stratalinker 2400 (Stratagene, La Jolla, CA) for 1 h. The percentage of apoptotic cells was assayed by FACS analysis of annexin staining according to manufactures instructions (BD Biosciences, Pharmingen™).
Primary cultures and cell lines were stimulated with necrotic cells (1:1 or 50–100 μg/ml), IFN-γ (Peprotech) (10–100 U/ml), LPS (Sigma) (100 ng/ml) or Pam3CSK4 (InvivoGen, San Diego, CA) (300 ng/ml) for 24 hours in RPMI 1640 with Glutamax (Invitrogen) supplemented with 5% FBS, 50 μM β-mercaptoethanol and 1% Penicillin-Streptomycin (RMPI-FBS).
Cell migration in studies of chemotaxis was assayed as described previously . Briefly, 500 μl of medium containing ATP (100 μM), necrotic cells (2 × 105 or 50 μg/ml), were applied to 24-well culture plates. Microglial cells (2 × 105) were plated in a 10 mm diameter tissue culture insert with an 8 μm pore size polycarbonate membrane (Nalgene Nunc, Roskilde, Denmark) placed into each well. Cells were incubated for 3 hours at 37°C in 5% CO2 atmosphere and migrated cells were collected from the lower chamber. After centrifugation cells were stained with trypan-blue and counted using a hemocytometer (Neubauer chamber) or mixed with 10 μm latex beads (Bechman Coulter, Fullerton, CA) and counted by flow cytometry (BD, FACSCalibur™)
Antibodies against CD11b (M1/70), CD11c (HL3), H-2Kb (AF-88.5), CD86 (GL1), CD40 (3/23), CD80 (16-10A1), CD14 (rmC5-3), CD45 (30-F11) and CD24 (M1/69) were purchased from BD Biosciences (BD, Pharmingen™). The antibody against FcγR (2.4G2) was prepared in house from hybridoma culture supernatants. Microglial cells were detached with cold PBS containing 0.5 mM EDTA and incubated with anti-Fc receptor in FACS buffer (PBS containing 2% FCS and 0.01% NaN3) for 30 min at 4°C before staining. After washing cells were incubated with streptavidinallophycocyanin or phycoerythrin (BD, Pharmingen™) for 20 min at 4°C to detect biotinylated antibodies. Cells were then washed and acquired using FACSCalibur™ and CELLQuest™ software (BD Becton, Dickinson and Company). Post acquisition analysis was done with CELLQuest™ or FlowJo™ (Tree Star, Ashland, OR) software.
CGN seeded in glass coverslips were incubated with MCM from non-stimulated cultures or from microglia activated with necrotic neurons at a final concentration of 20%. Where indicated, the MCM was obtained from microglial cells cultured in glutamine-free medium or in medium containing 20 μM of iNOS inhibitor, L-N6-(1-Iminoethyl) lysine (L-Nil, Tocris, UK); 1 mM of IDO inhibitor (Sigma), 1-Mehyl-DL-tryptophan (1-MT, Sigma), or 1,5 mM of glutaminase inhibitor, 6-diazo-5-oxo-L-norleucine (DON, Sigma). NMDA receptor antagonist, 10 μM MK-801 (Tocris), was added to neurons at same time as MCM. To quantify and assess neuronal cell survival, cells were incubated with 10 μg/ml 4'-6'-Diamidino-2-phenylindole (DAPI, Sigma) for 10 min at 37°C and 4 μg/ml propidium iodide (PI, Sigma) for 2 min also at 37°C. PI positive cells (~500) were counted in more than five fields per coverslip and cell death was expressed as the percentage (%) of PI positive cells from the total cells.
Analysis of cytokines, nitrites (NO2-) and glutamate
Cytokines were quantified in the supernatant of microglia cultured for 24 hours by ELISA according to manufacturer's instructions: TNF, IL-1β, IL-6 (R&D Systems, Minneapolis, MN) and IL-12 (p40 subunit) (BD OptEIA™). The production of NO by iNOS was measured indirectly by assaying nitrites in the culture supernatant using the Griess reaction. Supernatants were mixed with an equal volume of Griess reagent . Optical density measurements were averaged and converted to micromoles of nitrites per well using a standard curve of sodium nitrite. L-Glutamic acid concentration in the supernatants was determined using the Amplex Red reagent-based assay (Invitrogen) accordingly to the manufacturer's instructions.
Reverse transcription-PCR Analysis
Microglial cells were treated with IFN-γ (10 U/ml), necrotic HT22 cells (1:1) or left untreated for 24 hours. Total RNA was isolated from microglial cells using Trizol (Invitrogen, USA). RNA was reverse-transcribed from 2 μg of total RNA by 200 U of M-MuLV reverse transcriptase (Fermentas, UK) used accordingly to manufacturer's instructions. PCRs were performed with diluted cDNA using primers for iNOS (sense: 5'-AAG CTG CAT GTG ACA TCG ACC CGT-3'; antisense: 5'-GCA TCT GGT AGC CAG CGT ACC GG-3'), COX-2 (sense: 5'-CAG CAC TTC ACC CAT CAG TT-3'; antisense: 5'-CTG GTC AAT GGA GGC CTT TG-3'), IDO (sense: 5'-GTA CAT CAC CAT GCT GTA TG-3'; antisense: 5'-GCT TTC GTC AAG TCT TCA TTG-3'), Glutaminase (sense: 5'-GTA ATG ATC GTC AAC GGG GGA GGA C-3'; antisense: 5'-CCA GCA AGC CTT GCA ACC TTA ACC TTA ACC A-3') and HPRT (sense: 5'-GTA ATG ATC GTC AAC GGG GGA GGA C-3'; antisense: 5'-CCA GCA AGC TTG CAA CCT TA A CCT TAA CCA-3'). PCR products were resolved on 1% ethidium bromide-stained agarose gel.
Data is expressed as mean ± SD and was analyzed for significance using Student's t test. P-values are as follows: *** p < 0.005, ** p < 0.01 and * p < 0.05.
Activation of microglia by necrotic neurons
Parameters of activation in microglia stimulated with necrotic HT22 neurons.
Cytokine secretion (pg/ml)
36.6 ± 16.4
37.5 ± 10.3
15.2 ± 5.4*
67.7 ± 5.8
745.7 ± 15.5***
188.3 ± 4.3
1735.8 ± 85.2***
NO 2 - (μM)
12.5 ± 1.6***
Microglial activation by necrotic neurons is MyD88-dependent
TLR-2 and TLR-4 have been identified in macrophages and dendritic cells as putative receptors for endogenous signals such as heat shock proteins hsp70 and gp96  and the chromatin binding HMGB1 protein . We evaluated the putative role of these two TLRs in the microglial response to necrotic neurons using TLR-2-/- mice and B10ScN mice, which lack the genomic region containing the tlr-4 gene (Figure 2C). Microglia derived from these mouse lines maintained increased CD40 expression when incubated with necrotic neurons (Figure 2C). These data suggest that necrotic neurons contain ligands for My88-dependent TLR signaling other than TLR-2 or TLR-4 or, alternatively, suggest that several TLRs are activated and interfering with TLR-2 or TLR-4 signaling alone does not impair the microglial response.
Necrotic neurons enhance microglial glutamate-dependent neurotoxicity
Deletion of MyD88 in microglia blocks up-regulation of glutaminase and neurotoxicity by necrotic neurons
Taken together our results demonstrate that neurotoxicity induced by necrotic neurons in microglia is mediated through MyD88-dependent signaling cascades, leading to increased glutaminase expression and consequent increase in glutamate-mediated neurotoxicity.
Microglia and perivascular macrophages/dendritic cells in the CNS become activated upon neuronal injury. The goal of the present study was to understand the pathways of microglial activation induced by neuronal necrosis, and the effects of such activation on microglia-mediated neurotoxicity. We demonstrate that necrotic neurons induce several markers of microglial activation such as co-stimulatory molecules, MHC class II, pro-inflammatory cytokines and enzymes and enhanced motility. This activation, which involves a MyD88-dependent signalling pathway, also increases NMDAR-dependent microglial neurotoxicity by up-regulation of glutaminase expression and activity.
Bacteria-derived products are known to induce neuronal cell death through the activation of the MyD88 pathway in microglia . Here we demonstrate that "molecular danger patterns" exposed by necrotic neurons also activate microglia and enhance microglial neurotoxicity by a MyD88-dependent pathway. However, in the absence of TLR-2, TLR-4 and TLR-9 (data not shown) which signal through MyD88, the response of microglia to necrotic neurons is not affected. Either necrotic neurons activate other than those tested, or more than a single TLR are redundantly activated in response to the necrotic stimuli. In the later case, only by blocking a common signaling point, as in MyD88-deficient microglia, would activation of microglia by necrotic neurons be highly compromised as we have observed using assays of CD40 and MHC II expression and IL-6 and IL-12 production (Figure 3). A mechanism involving activation of NLRs by ligands exposed by necrotic neurons cannot be excluded since these PRRs induce IL-1β and IL-18 which act through MyD88. . However, we did not detect an increase in IL-1β secretion by microglia activated with necrotic neurons.
Microglia activated with necrotic neurons clearly show a pro-inflammatory phenotype dependent on MyD88. This type of response by microglia may contribute to enhanced neuropathogenesis after necrotic neuronal cell death. Interestingly, it has been shown recently that MyD88-deficient mice, but not TLR4/TLR2-deficient mice, show reduced brain inflammatory responses and decreased lesions after cold-induced cortical injury . However, pro-inflammatory mediators may have diverse effects after neuronal injury caused by brain ischemia. Whereas inhibition of COX-2 is neuroprotective , the role of pro-inflammatory cytokines such IL-6 and TNF is ambiguous in this context, with both deleterious and beneficial effects having been reported .
We have demonstrated that glutaminase expression is regulated by a MyD88-dependent pathway. By bioinformatics analysis using PROMO software [40, 41], we have identified three putative DNA-binding sites for the transcription factor Activation Protein-1 (AP-1) at 20 bp, 240 bp and 300 bp upstream of the start codon. Since a MyD88-dependent pathway was shown to activate AP-1, which together with NF-kB controls the induction of pro-inflammatory cytokines , AP-1 may be one of the putative transcription factors involved in the induction of glutaminase expression.
It has been reported that HSP60, present in the supernatant of total necrotic CNS cells, activates TLR4 on microglia leading to increased microglial neurotoxicity towards cortical neurons. This effect could be partially but significantly reduced by inhibiting iNOS activity, suggesting that other neurotoxic mechanisms may be involved . In our study, iNOS was induced in microglia activated by necrotic neurons but NO did not play a significant role in mediating the killing of CGN. This probably reflects different susceptibilities to neurotoxic molecules on the part of different neuronal cell populations. Of the variety of neurotoxic molecules that activated microglia can secrete , we clearly show that glutamate is the sole candidate as the agent that enhances neurotoxicity of CGN through NMDAR in microglia activated with necrotic neurons. These results are in accordance with previously described mechanisms of macrophage/microglial neurotoxicity [44, 45, 46]. Moreover, it has been shown that excessive activation of this receptor contributes to a large number of neurological disorders associated with microglial activation [47, 48].
We demonstrate that necrotic neurons induce a pro-inflammatory and glutaminase-mediated neurotoxic response by microglia. Both mechanisms are dependent on the adapter signalling protein MyD88 and may be highly relevant to increased neuropathology in the context of brain injury. Therefore, the MyD88-dependent pathway and glutaminase activity may constitute potential targets for new pharmacological strategies to control massive neuroinflammation and neurodegeneration.
We are very grateful to Dr. S. Akira (Osaka University) for providing TLR-2-/-, TLR-9-/- and MyD88-/- mice.
We thank Drs. Jocelyne Demengeot, Miguel Soares and Íris Caramalho for useful suggestions and for critical reading of the manuscript. We thank Carlos Duarte and Elsa Seixas for critical reading of the manuscript. We thank Hugo Soares for helping with the bioinformatics analysis.
This work was supported by the grant POCI/SAU-NEU/56986/04 from Fundação para Ciência e Tecnologia (Portugal). Teresa F. Pais, Maria H. Braz and Catarina Figueiredo were supported by Fundação para a Ciência e Tecnologia fellowships SFRH/BPD/5554/2001, SFRH/BD/6236/2001, SFRH/BD/9093/2002 respectively.
- 2.Leib SL, Kim YS, Chow LL, Sheldon RA, Tauber MG: Reactive oxygen intermediates contribute to necrotic and apoptotic neuronal injury in an infant rat model of bacterial meningitis due to group B streptococci. J Clin Invest. 1996, 98 (11): 2632-2639. 10.1172/JCI119084.PubMedCentralCrossRefPubMedGoogle Scholar
- 6.Minagar A, Shapshak P, Fujimura R, Ownby R, Heyes M, Eisdorfer C: The role of macrophage/microglia and astrocytes in the pathogenesis of three neurologic disorders: HIV-associated dementia, Alzheimer disease, and multiple sclerosis. J Neurol Sci. 2002, 202 (1–2): 13-23. 10.1016/S0022-510X(02)00207-1.CrossRefPubMedGoogle Scholar
- 7.Raivich G, Bohatschek M, Kloss CU, Werner A, Jones LL, Kreutzberg GW: Neuroglial activation repertoire in the injured brain: graded response, molecular mechanisms and cues to physiological function. Brain Res Brain Res Rev. 1999, 30 (1): 77-105. 10.1016/S0165-0173(99)00007-7.CrossRefPubMedGoogle Scholar
- 24.Cohen JaW GP: Neural Cell Culture: A Practical Approach. 1995, Oxford: IRL Press at Oxford University PressGoogle Scholar
- 29.Heyes MP, Saito K, Major EO, Milstien S, Markey SP, Vickers JH: A mechanism of quinolinic acid formation by brain in inflammatory neurological disease. Attenuation of synthesis from L-tryptophan by 6-chlorotryptophan and 4-chloro-3-hydroxyanthranilate. Brain. 1993, 116 (Pt 6): 1425-1450.CrossRefPubMedGoogle Scholar
- 33.Basu S, Binder RJ, Suto R, Anderson KM, Srivastava PK: Necrotic but not apoptotic cell death releases heat shock proteins, which deliver a partial maturation signal to dendritic cells and activate the NF-kappa B pathway. Int Immunol. 2000, 12 (11): 1539-1546. 10.1093/intimm/12.11.1539.CrossRefPubMedGoogle Scholar
- 38.Candelario-Jalil E, Gonzalez-Falcon A, Garcia-Cabrera M, Leon OS, Fiebich BL: Post-ischaemic treatment with the cyclooxygenase-2 inhibitor nimesulide reduces blood-brain barrier disruption and leukocyte infiltration following transient focal cerebral ischaemia in rats. J Neurochem. 2007, 100 (4): 1108-1120. 10.1111/j.1471-4159.2006.04280.x.CrossRefPubMedGoogle Scholar
- 43.Lehnardt S, Schott E, Trimbuch T, Laubisch D, Krueger C, Wulczyn G, Nitsch R, Weber JR: A vicious cycle involving release of heat shock protein 60 from injured cells and activation of toll-like receptor 4 mediates neurodegeneration in the CNS. J Neurosci. 2008, 28 (10): 2320-2331. 10.1523/JNEUROSCI.4760-07.2008.CrossRefPubMedGoogle Scholar
- 45.Takeuchi H, Jin S, Wang J, Zhang G, Kawanokuchi J, Kuno R, Sonobe Y, Mizuno T, Suzumura A: Tumor necrosis factor-alpha induces neurotoxicity via glutamate release from hemichannels of activated microglia in an autocrine manner. J Biol Chem. 2006, 281 (30): 21362-21368. 10.1074/jbc.M600504200.CrossRefPubMedGoogle Scholar
- 46.Takeuchi H, Mizuno T, Zhang G, Wang J, Kawanokuchi J, Kuno R, Suzumura A: Neuritic beading induced by activated microglia is an early feature of neuronal dysfunction toward neuronal death by inhibition of mitochondrial respiration and axonal transport. J Biol Chem. 2005, 280 (11): 10444-10454. 10.1074/jbc.M413863200.CrossRefPubMedGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.