article published online: 12 October 2014 | doi: 10.1038/nchembio.1669

A small molecule binding HMGB1 and HMGB2 inhibits microglia-mediated neuroinflammation

© 2014 Nature America, Inc. All rights reserved.

Sanghee Lee1, Youngpyo Nam2, Ja Young Koo1, Donghyun Lim3, Jongmin Park1, Jiyeon Ock2, Jaehong Kim2, Kyoungho Suk2* & Seung Bum Park1,3* Because of the critical role of neuroinflammation in various neurological diseases, there are continuous efforts to identify new therapeutic targets as well as new therapeutic agents to treat neuroinflammatory diseases. Here we report the discovery of inflachromene (ICM), a microglial inhibitor with anti-inflammatory effects. Using the convergent strategy of phenotypic screening with early stage target identification, we show that the direct binding target of ICM is the high mobility group box (HMGB) proteins. Mode-of-action studies demonstrate that ICM blocks the sequential processes of cytoplasmic localization and extracellular release of HMGBs by perturbing its post-translational modification. In addition, ICM effectively downregulates proinflammatory functions of HMGB and reduces neuronal damage in vivo. Our study reveals that ICM suppresses microglia-mediated inflammation and exerts a neuroprotective effect, demonstrating the therapeutic potential of ICM in neuroinflammatory diseases.

N

euroinflammation, considered mainly as innate immune responses in the central nervous system (CNS), is triggered in response to diverse inflammatory signals such as pathogen infection, injury or trauma, which may result in neurotoxicity1,2. An inflammatory response in the brain is strongly associated with the activation of microglia and is characterized by various indications, including the synthesis and secretion of proinflammatory cytokines and chemokines and the breakdown of the blood-brain barrier1,2. Although there is some debate about the cause-and-­effect relationship, it is generally believed that chronic and excessive microglial activation results in neuronal damage and death, which contribute to neurodegenerative diseases, including Alzheimer’s disease, Parkinson’s disease, multiple sclerosis and amyotrophic lateral sclerosis1–3. Therefore, there is a growing need for the discovery of new therapeutic agents targeting microglial activation and neuroinflammation. To find a better treatment for devastating neuroinflammatory diseases of the CNS, the drug discovery processes should be accompanied by the identification of potential therapeutic targets. In the post-genomic era, cell-based phenotypic screening has emerged as a promising approach for the discovery of new chemical entities (NCEs) as candidates for first-in-class drugs4. This paradigm shift has enabled innovation in the functional network of bioactive small molecules through the identification of new mechanisms of action5. However, this attractive approach is quite limited in the actual drug discovery process because of the lengthy process of target identification or deconvolution6,7. In addition, further development and clinical applications of NCEs can be extremely difficult without elucidation of their molecular targets, even though the NCEs show prominent therapeutic effects in vivo8. Therefore, efficient target identification has become an indispensable element in phenotype-based drug discovery. Here we report a new anti-neuroinflammatory agent, ICM (1), discovered by cell-based phenotypic screening and the subsequent identification of its molecular target using fluorescence difference in two-dimensional (2D) gel electrophoresis (FITGE) technology9.

Furthermore, we demonstrate the therapeutic potentials of ICM in CNS inflammatory disease.

RESULTS Discovery of the anti-neuroinflammatory agent ICM

Microglia, a type of immune cell in the brain, serve as the first line of defense in the CNS by recognizing external pathogens, and they contribute to innate and adaptive immunity by continuously surveying the microenvironment and protecting the brain. Once microglia are stimulated by immunogens or bacterial endotoxins, such as lipopolysaccharide (LPS), activated microglia have a central role in neuroinflammation and secrete various neurotoxic factors, such as IL-1β, TNF-α, PGE2, nitric oxide (NO) and superoxide anions (O2−)2. To identify anti-neuroinflammatory agents, we used cellbased phenotypic screening with LPS-induced nitrite release in the BV-2 mouse microglial cell line as a readout. Using high-throughput screening of an approximately 3,500-member in-house library constructed by the pDOS strategy10, we identified a new benzopyranembedded tetracyclic compound named ICM (Fig. 1a). ICM ­efficiently blocked LPS-induced nitrite release in a dose-dependent manner without any toxicity in BV-2 microglial cells (Fig. 1b and Supplementary Results, Supplementary Fig. 1a). We also confirmed that ICM inhibited nitrite release in a broad range of cell lines, including rat microglia (HAPI), mouse macrophages (RAW 264.7) and mouse primary microglial cultures (Supplementary Fig. 1b). However, ICM exhibited less potency in culture of primary astrocytes, another major glial cell type in the CNS, than in microglial cells, indicating an excellent selectivity of ICM toward microglia (Supplementary Fig. 1b). To confirm the anti-inflammatory effect of ICM, we examined the expression or production of other pro­ inflammatory mediators in murine microglial cell lines and primary microglial cells. The increased levels of inflammation-related genes, such as Il6, Il1b, Nos2 and Tnf, after LPS stimulation were markedly suppressed by treatment with ICM (Fig. 1c and Supplementary Figs. 3a and 4a), and LPS-induced secretion of the proinflammatory cytokine TNF-α was also reduced by ICM treatment (Fig. 1d

Department of Chemistry, Seoul National University, Seoul, Korea. 2Department of Pharmacology, Brain Science and Engineering Institute, BK21 Plus KNU Biomedical Convergence Program, Kyungpook National University School of Medicine, Daegu, Korea. 3Department of Biophysics and Chemical Biology/N-Bio Institute, Seoul National University, Seoul, Korea *e-mail: [email protected] or [email protected] 1

nature CHEMICAL BIOLOGY | Advance online publication | www.nature.com/naturechemicalbiology

1

article b O

N O

N

N H

25

LPS (–) LPS (+)

20 Nitrite [µM]

a

Nature chemical biology doi: 10.1038/nchembio.1669 – –

+ –

Il6

15

Il1b

10

Nos2

5

– 10

+ 1

+ 5

+ 10

On the basis of these results, we confirmed that ICM as a newly identified microglial inhibitor has both an anti-inflammatory and a neuroprotective effect.

Potential ICM targets

We recently reported a cell-based target identification method, FITGE9. To overcome the 0.1 1 10 100 Inflachromene, ICM (1) limitation of weak binding affinity or low [ICM] (µM) abundance of target proteins, we introduced DMSO DMSO DMSO DMSO d e f g covalent anchoring of a bioactive small molICM ICM ICM ICM NS NS 6,000 100 ecule to cellular target proteins in live cells 150 150 ** *** using a photocrosslinker. The resolution of the 80 4,000 100 100 in-gel protein analysis was clearly enhanced by 60 *** 2D gel electrophoresis with dual-color labeling 40 2,000 50 50 to differentiate specific target proteins from ** 20 nonspecific protein binders9. On the basis of 0 0 0 0 a structure-activity relationship study for ICM Control LPS Control LPS Control LPS Control LPS analogs (Supplementary Fig. 7), we designed and synthesized the chemical probe ICM-BP (2) Figure 1 | Discovery of ICM as an anti-inflammatory agent in LPS-induced microglial activation. (Fig. 2a). Benzophenone was embedded in (a) Chemical structure of ICM. (b) Dose-dependent effect of ICM on BV-2 cells in the absence or the phenyl substituent of triazole as a photo­ presence of LPS stimulation. The level of nitrite in the extracellular milieu was measured by Griess activatable crosslinking moiety, and an assay. (c) RT-PCR analysis revealing the inhibitory effect of ICM on the expression levels of alkyne group was incorporated at the end of LPS-induced inflammation-related genes in HAPI cells (full-length blots are shown in benzophenone as a functional tag for a bioSupplementary Fig. 29). Data are representative of triplicate experiments. (d) LPS-induced orthogonal click reaction with azide-linked TNF-α secretion in BV-2 cells was measured by ELISA with ICM (5 μM) treatment. (e) The fluorescent dyes or biotin. Before the applinuclear translocation of p65 (an NF-κB subunit) was investigated by immunofluorescence cation of ICM-BP for target identification, analysis with ICM (5 μM) treatment in HAPI cells. (f) The effect of ICM (10 μM) on microglial we evaluated its inhibitory activity toward neurotoxicity as measured by coculture of LPS-stimulated HAPI microglial cells and B35-eGFP LPS-induced nitrite release and confirmed that neuroblastoma cells. (g) The effect of ICM (10 μM) on coculture of LPS-stimulated primary ICM-BP exhibited inhibitory activity at a level microglial cells and CMFDA-labeled primary cortical neurons. The eGFP-positive cells and the comparable to that of the original compound, CMFDA-positive cells for viability assessment are presented as the percentage of nontreated ICM (Fig. 2b). cells in f and g. Cells were treated with ICM for 30 min before LPS treatment (100 ng/ml). With the optimized probe, we sought to All data represent the mean and s.d. in triplicate. *P < 0.05, **P < 0.001, ***P < 0.0001 compared identify the target protein of ICM in live cells to DMSO, as calculated by Student’s t-test. NS, not significant compared to DMSO. using competitive labeling, which facilitated an effective exclusion of nonspecific binding and Supplementary Fig. 3b). We previously reported that a series proteins. Once the cells were treated with ICM-BP in the absence of ICM analogs inhibited RANKL-induced osteoclastogenesis by or presence of ICM as a soluble competitor, they were irradiated perturbing the NF-κB and MAPK signaling pathways in bone mar- with UV light to photocrosslink ICM-BP to cellular target proteins row monocytes and macrophages11. To confirm that these signaling and subjected to the click reaction with an azide-linked fluorescent pathways were influenced by the anti-neuroinflammatory effect of dye. The resulting proteomes were analyzed by gel electrophoresis ICM, NF-κB and MAPK signaling were investigated in microglia. and gel-fluorescence scanning. As determined by one-dimensional ICM substantially suppressed the nuclear translocation of NF-κB (1D) gel analysis, a direct comparison of the labeling patterns of (Fig. 1e and Supplementary Fig. 3c) and the degradation of IκB ICM-BP with and without ICM competition showed two distinct (Supplementary Fig. 5a) in LPS-stimulated microglia. In addition, bands at 29 kDa and 45 kDa (Fig. 2c) regardless of LPS stimulation. ICM treatment inhibited LPS-induced phosphorylation of ERK, We selected the protein in the 29-kDa band, as it competed out at a JNK and p38 MAPK in microglia (Supplementary Fig. 5b). lower concentration of ICM more effectively than the 45-kDa band Microglia release various neurotoxic factors after activation. in a dose-dependent competitive assay (Supplementary Fig. 9b). Microglia-mediated neurotoxicity is thought to be responsible for Even though the probe labeling pattern in cell lysates was substancellular damage to neighboring neurons; the damaged neurons lead tially different from that in live cells, we failed to observe any speto subsequent reactivation of microglia, a response known as micro- cific binding events of ICM-BP or competitive binding with ICM gliosis2. The perpetual accumulation of neuronal damage and micro- in the cell lysates (Supplementary Fig. 9c). On the basis of these glial activation eventually lead to the death of adjacent neurons in results, the bands were excised from 1D gels of live cell labeling, and the brain and to various neurodegenerative diseases. Therefore, the the proteins of the 29-kDa band were explored by LC/MS/MS analinhibition of microglial activation might confer a protective effect ysis. According to MS analysis, however, numerous proteins were against microglia-mediated neurotoxicity. To evaluate the neuro- implicated as potential targets by 1D gel analysis (Supplementary protective effect of ICM in vitro, coculture assays of LPS-treated Fig. 10a). To narrow down the number of candidates, we performed microglia were performed with either enhanced GFP (eGFP)- a 2D gel analysis to improve the resolution of the in-gel analysis. labeled neuroblastoma cells or 5-chloromethylfluorescein diacetate ICM-BP–labeled proteomes with and without ICM competition (CMFDA)-labeled primary cortical neurons. ICM completely pre- were treated with Cy5-azide (red for specific binding) and Cy3-azide vented the death of cocultured neuroblastoma and primary neuronal (green for nonspecific binding), respectively. Two samples labeled cells by inhibiting microglia-mediated neurotoxicity (Fig. 1f,g and with different fluorescent dyes were mixed and analyzed together Supplementary Fig. 4c). Furthermore, ICM itself had no significant by 2D gel electrophoresis, which reduced the gel-to-gel variation. effect on the viability of neurons (Supplementary Fig. 6; P values of In-gel fluorescence scanning of 2D gels effectively revealed char0.3225, 0.9908 and 0.6763 for 1 μM, 5 μM and 10 μM ICM treatment). acteristic spots around 29 kDa and 45 kDa, corresponding to the 2

0 0.01

CMFDA-positive cell (%)

Gapdh

eGFP-positive cell (%)

O

Tnf

Nuclear translocation of p65 (%)

TNF-α (pg/ml)

HO

© 2014 Nature America, Inc. All rights reserved.

c LPS ICM (µM)

nature chemical biology | Advance online publication | www.nature.com/naturechemicalbiology

article

Nature chemical biology doi: 10.1038/nchembio.1669 a

b 12

O

O O

N O

HO

N

N H

O

ICM-BP (2)

d

pH 3

Nitrite [µM]

HN

c

DMSO ICM 0.5 µM ICM 5 µM ICM-BP 0.5 µM ICM-BP 5 µM

LPS – ICM-BP + ICM –

8

***

+ + –

+ + +

– + –

– + +

+ + –

+ + +

*

*** *** ***

4 0

– + +

Control

LPS

10

In-gel fluorescence

Silver

e

*

ICM-BP ICM Input

+ –

+ +

PD

© 2014 Nature America, Inc. All rights reserved.

IB: HMGB2

Figure 2 | Target identification of ICM by 1D and 2D gel analysis in live cells. (a) The chemical structure of ICM-BP. Blue represents a photoaffinity group. Red represents a tag moiety for the click reaction. (b) The inhibitory activity of ICM-BP on nitrite release compared to that of ICM in BV-2 cells. The data represent the mean and s.d. in triplicate. ***P < 0.0001 compared to DMSO, as calculated by Student’s t-test. (c) After light-induced labeling of ICM-BP (5 μM) to target the proteome in live cells, ICM-BP–labeled proteomes with or without ICM competition (80 μM) were analyzed by 1D gel electrophoresis in the absence or presence of LPS (100 ng/ml) stimulation. The arrow and asterisk designate the major bands with significant differences after ICM competition. Blots in c and e were carried out as individual experiments, and several dose and time ranges were tested to find the optimal conditions. (d) Target identification with ICM-BP (20 μM) using 2D gel analysis. ICM-BP–labeled proteomes with and without ICM competition (20 μM) were treated with Cy3-azide (green) and Cy5-azide (red), respectively. The merged fluorescence images of the whole gel (left) and the expanded image of the dotted box (right). The arrow and asterisk designate the major red spots in the 2D gel showing specific binding. Scale bars, 2 cm. (e) Pulldown (PD) assay performed with biotin-labeled proteome and streptavidin beads. ICM-BP (5 μM) with or without ICM competition (20 μM) was subjected to PD assay and immunoblotting (IB) for HMGB2 (full-length blots are shown in Supplementary Fig. 30).

bands in the 1D gel analysis (Fig. 2d). After MS/MS analysis, we obtained a shorter list of proteins detected from the 29-kDa spot than in 1D gel analysis, and we selected a number of potential candidates for target proteins that were identified in both the 1D and 2D gel analyses (Supplementary Fig. 10). Among those candidate proteins, we focused on the HMGB2 protein as a potential target, as there are a number of reports in the literature indicating that the HMGBs have crucial roles in inflammation12–14. To assess the specific binding event of ICM-BP with HMGB2, we performed an affinity pulldown assay using avidinbased enrichment of an ICM-BP–treated proteome after the click reaction for labeling with biotin-azide. HMGB2 efficiently bound with ICM-BP, whereas the excess ICM as a soluble competitor markedly lowered the level of binding for HMGB2 (Fig. 2e), which confirmed the specific interaction between ICM-BP and HMGB2.

Functional validation of HMGB as a target of ICM

Next we elucidated the HMGB-dependent anti-inflammatory activity of ICM with a loss-of-function study to determine whether the interaction between ICM and HMGB functionally influenced the desired inhibitory effect of ICM. There are four isoforms of HMGB, HMGB1–HMGB4, with a shared structural motif. Among them, HMGB1 and HMGB2 share 80% sequence homology, containing two tandem HMG box domains (box A and box B) and a long acidic tail at the C terminus14. Because of the structural and functional similarity of the two proteins, we speculated that ICM might also interact with HMGB1, and so we examined the inhibitory

effect of ICM for both HMGB1 and HMGB2. Transfection of short interfering RNA (siRNA) for HMGB1 and HMGB2 individually induced a reduction in the protein expression levels of HMGB1 and HMGB2, respectively. In addition, the dual knockdown of both HMGB1 and HMGB2 decreased in their expression levels simultaneously (Fig. 3a). The loss of function resulting from individual knockdown of HMGB1 and HMGB2 markedly inhibited LPS-induced nitrite release in BV-2 microglial cells, and dual knockdown synergistically enhanced the inhibitory effect (Fig. 3b). These results indicate a crucial role of both HMGB1 and HMGB2 in microglial activation. We then tested whether HMGB knockdown affected the anti-inflammatory effect of ICM. Transfection with siRNA for the two HMGBs negated the inhibitory effect of ICM, as it elevated LPS-induced nitrite release compared to mock and scramble siRNA transfections. In simultaneous knockdown of both HMGB1 and HMGB2, ICM treatment resulted in even higher nitrite release than after individual knockdown of either after LPS stimulation (Fig. 3c). The attenuated inhibitory effect of ICM after knockdown of the HMGBs might be due to the elimination of its target protein, which confirmed an HMGB-dependent antiinflammatory effect of ICM. Overall these results support that HMGBs as the target of ICM have a critical role in the anti-inflammatory effect of ICM.

ICM as a PTM modulator

In mammals, HMGBs are known as multifunctional proteins14. Originally, the HMGBs were known as DNA-binding nuclear proteins12–15. Since the unexpected role of HMGB1 as a late mediator of inflammation was reported16, numerous studies have elucidated interesting features of the HMGBs, especially HMGB1, in the regulation of inflammatory signaling12–14,17. HMGB1 shuttles between the nucleus and the cytosol, and cytoplasmic HMGB1 can be released into the extracellular space through the secretory lysosomal pathway or by passive diffusion in necrotic cells18–20. Once released, extracellular HMGB1, similar to proinflammatory cyto­ kines, interacts consecutively with various receptors to activate the inflammatory signaling pathway12,13. Other HMGBs are expected to have similar roles based on their structural similarity to HMGB1 (ref. 14). Therefore, we hypothesized that ICM regulates HMGB2mediated inflammatory signaling. To test this hypothesis, we first monitored the intracellular or extracellular translocation of HMGB2 after ICM treatment. LPS stimulation increased the cytoplasmic accumulation of HMGB2, as well as the secretion of HMGB2 into the extracellular milieu, but ICM completely suppressed LPS-induced cytoplasmic translocation or extracellular release of HMGB2 (Fig. 4a). Because of these observations, we then focused on the post-translational modification (PTM) of HMGB2. HMGBs have diverse PTM sites and are modified extensively by acetylation, phosphorylation, methylation, glycosylation and poly(ADP) ribosylation21. In particular, several PTMs might be closely related to the cellular translocation of HMGB1 in monocytes or macrophages22,23, as acetylation and phosphorylation were observed at lysine and serine residues, respectively, near two nuclear localization sequences (NLSs) in the HMGBs. Therefore, we sought to determine the potential effect of

nature CHEMICAL BIOLOGY | Advance online publication | www.nature.com/naturechemicalbiology

3

article a

Nature chemical biology doi: 10.1038/nchembio.1669 b

c

Considering the impact of microglia-mediated neurotoxicity in neuroinflammatory disease, we envisioned the therapeutic potential of ICM and tested the effect of ICM on microglial activation in vivo using an LPS-induced mouse neuroinflammation model. Briefly, ICM was injected intraperitoneally daily for 4 days, then the mice were euthanized, and the levels of Iba-1 as a marker of microglial activation were analyzed (Supplementary Fig. 15a). Quantitative analysis of different regions of the mouse brain, including the cortex, hippocampus and substantia nigra, revealed that ICM effectively blocked LPS-mediated microglial activation, even at a low dose (2 mg per kg body weight of the mouse) (Fig. 5a and Supplementary Fig. 16). To evaluate the protective role of ICM in the pathogenesis of neuroinflammatory disease, we used a mouse experimental autoimmune encephalitis (EAE) model, which is an animal model of multiple sclerosis, a prototype organ-specific autoimmune and inflammatory disease in the CNS. After myelin oligodendrocyte glycoprotein (MOG) peptide immunization, EAE developed by day 12 and reached a maximal score on day 17. To evaluate the therapeutic effect of ICM, we administered 10 mg/kg of ICM daily after MOG immunization (Supplementary Fig. 15b). We found that ICM administration significantly reduced the progression of disease, as determined by EAE clinical score (Fig. 5b). When the EAE phenotype reached its peak, most of the mice treated with vehicle showed hindlimb paralysis (grade 3), but most of the ICM-treated 4

0.4

M oc k siHM sc si- G H B + si- MG 1 si- H B HM M 2 G GB B1 2

Effect of ICM on LPS-induced nitrite release (normalized to DMSO)

GB 2 + sisi- H HM M GB GB 1 2

GB 1

siHM

siHM

oc k

M

© 2014 Nature America, Inc. All rights reserved.

The therapeutic effect of ICM

sc

LPS (–) LPS (+) NS

Nitrite [µM]

si-HMGB1 + si-HMGB2

si-HMGB2

si-HMGB1

sc

Mock

animals showed limp tails (grade 1) (Fig. 5b). ICM markedly delayed the onset of disease *** ** and decreased disease severity by reducing *** siRNA 20 0.3 the maximal clinical score and degree of ** HMGB1 ** hindlimb paralysis compared to vehicle treat* 10 0.2 ment (Supplementary Fig. 17). To examine * HMGB2 *** NS the activation of microglia and the inflamma0 Actin siRNA siRNA tory response, we euthanized mice at the peak of disease and performed histological analysis. Microglial activation in the lumbar spinal cords of EAE mice was clearly suppressed, and inflammatory lesions in EAE spinal cords were Figure 3 | Functional validation of HMGB2 as a direct binding target of ICM. (a) Representative substantially attenuated by ICM treatment western blot results in triplicate confirming siRNA-mediated knockdown of HMGB1 and HMGB2 (Fig. 5c,d). In addition, microglial cells isoin BV-2 cells (mock, control transfection; sc, scrambled siRNA; si-HMGB1, siRNA for HMGB1; lated from the spinal cords of EAE mice were si-HMGB2, siRNA for HMGB2) (full-length blots are shown in Supplementary Fig. 30). (b) The subjected to analysis of the mRNA levels of loss-of-function effect by siRNA knockdown with either si-HMGB1 or si-HMGB2 and double proinflammatory cytokines and chemokines, knockdown with both si-HMGB1 and si-HMGB2 confirmed by LPS-induced nitrite release in which confirmed that ICM treatment supBV-2 cells. (c) The HMGB-dependent effect of ICM. LPS-induced nitrite release after siRNA pressed the production of inflammatory transfection. ICM (5 μM) treatment was normalized to the individual DMSO treatments. All data markers, such as IL-1β, TNF-α, CXCL10, CCL2 represent the mean and s.d. in triplicate. *P < 0.05, **P < 0.001, ***P < 0.0001 compared to mock and IL-6, in microglia in vivo (Fig. 5e). The transfection, as calculated by Student’s t-test NS, not significant compared to mock transfection. colocalization of HMGBs and microglia in EAE Asterisks above the lines indicate significance between single and dual knockdown. mice supported a microglia-specific inhibitory effect of ICM (Supplementary Fig.  18). The ICM on LPS-induced phosphorylation and acetylation of HMGBs in effects of ICM on the proliferation of neuronal progenitor cells in the microglia. LPS treatment increased the phosphorylation and acetyla- subventricular zone (SVZ) and the inflammatory activation of develtion of HMGB2 in microglia, and these PTMs were effectively inhib- oping glial cells and neuronal progenitor cells were next assessed. ited by treatment with ICM (Fig. 4b). Identical inhibitory effects Injection of ICM did not significantly influence the proliferation of on the PTMs for HMGB1 were also observed after ICM treatment neuronal progenitor cells, as determined by counting BrdU-positive (Supplementary Fig. 14). Considering that PTMs of HMGB1 regu- proliferating cells in SVZ (Supplementary Fig. 19; P values of late the nuclear export of HMGB1, this ICM-induced PTM inhibition 0.0736 and 0.0717 for 10 nM and 50 nM ICM treatment). Similarly, can explain the ICM-induced suppression of HMGB2 transloca- ICM treatment did not have significant effects on NO or TNF-α tion. In other words, LPS-induced phosphorylation and acetylation production in glial cells or neuronal progenitor cells derived from lead to relocalization of HMGB2 to the cytoplasm, but ICM treat- striatum at embryonic day (E) 15 (Supplementary Fig. 20), further ment inhibits the PTMs and suppresses the cytoplasmic accumula- supporting the microglia selectivity of ICM. Moreover, we clearly tion of HMGB2. Therefore, we confirmed that ICM disrupts PTMs observed that ICM ameliorated neuroinflammation, consequently and sequentially blocks intracellular translocation and extracellular attenuating neurite damages and spinal cord demyelination in EAE secretion of HMGB2, thereby inhibiting the role of HMGB2 as a spinal cords, as determined by immunostaining of microtubuleproinflammatory cytokine. associated protein-2 (MAP-2) and myelin basic protein (MBP) 30

a

LPS ICM

– –

+ –

b

+ +

LPS ICM

– –

+ –

+ +

Input

hnRNP Nu HMGB2

IP: pSer IB: HMGB2

Actin Cyt HMGB2

Input IP: AcLys

HMGB2

CM

IB: HMGB2

Figure 4 | Mode-of-action study of ICM for perturbing the inflammatory function of HMGB2. (a) The effect of ICM on the translocation of HMGB2 from the nucleus (Nu) to the cytoplasm (Cyt) and extracellular milieu in conditioned medium (CM). (b) Anti-phosphoserine (pSer) and acetylated lysine (AcLys) immunoprecipitation (IP) followed by immunoblotting (IB) for HMGB2 indicates the effect of ICM on the post-translational modification of HMGB2. BV-2 cells were treated with ICM (10 μM) before LPS (200 ng/ml) and analyzed by subcellular fractionation or IP assay. Heterogeneous nuclear ribonucleoprotein (hnRNP) and actin were detected as housekeeping proteins in the nucleus and cytoplasm, respectively. Input represents 5% of the total lysate used in IP (full-length blots are shown in Supplementary Fig. 30). Blots were carried out as individual experiments, and several dose and time ranges were tested to find the optimal conditions.

nature chemical biology | Advance online publication | www.nature.com/naturechemicalbiology

Nature chemical biology doi: 10.1038/nchembio.1669

have been focused on HMGB1 (refs. 12,13,17), and the role of other HMGB isoforms remains unclear, especially in neuroinflammation. In * 2 400 this study, we found that HMGB2 has a criti*** cal role in microglia-mediated neuroinflam** *** *** ** * * * ** *** ** *** 1 mation through the discovery of ICM and the 200 ** * * * ** *** *** subsequent identification of HMGB2 as the ** target of ICM. ** 0 0 Because of their important roles as late mediLPS – – – + + + – – – + + + – – – + + + 0 5 10 15 20 25 30 ICM (mg/kg) – 2 10 – 2 10 – 2 10 – 2 10 – 2 10 – 2 10 ators of inflammation, HMGBs have received Time after immunization (d) increasing attention as therapeutic targets for Naive EAE Naive EAE c d sepsis, cancer, arthritis and other inflammaVehicle Vehicle ICM Vehicle Vehicle ICM tory conditions, including hepatitis and pancreatitis, as well as neuroinflammation14,25. In fact, the dual inhibitory activity of ICM toward both HMGB1 and HMGB2 ensures the therapeutic potential of ICM because of the significant roles of both proteins in neuro­ inflammation (Fig.  3b). The excellent therapeutic efficacy of ICM in vivo, even at low 50 25 e f Naive Naive doses (Fig. 5a), compared to its efficacy in the EAE + vehicle EAE + vehicle 40 20 in vitro system might be due to the synergistic EAE + ICM EAE + ICM 30 15 inhibitory activity of ICM for both HMGB2 20 10 and HMGB1. ICM effectively suppressed *** ** * 10 5 microglial activation, inhibited neuroinflam** ** ** mation and mediated a neuroprotective effect 0 0 Tnf Cxcl10 Ccl2 Il6 Il1b CSF Serum in the EAE spinal cord. According to the timecourse experiment, ICM exhibited a lasting Figure 5 | The therapeutic effect of ICM in vivo. (a) Suppression of glial activation by ICM anti-inflammatory effect and not just a simadministration was analyzed in the cortex, substantia nigra (SN) and hippocampus regions of ple delay of inflammation (Supplementary LPS-injected mouse brains. Quantification for immunofluorescence staining of Iba-1. (b) EAE Fig. 23). Interestingly, ICM retained its inhibiwas induced in C57BL/6 mice by immunization of the MOG35–55 peptide and pertussis toxin. tory activity when administered after LPS The animals were administered daily with either vehicle or ICM for 30 d. Data are presented stimulation as well as when given before LPS as a clinical score and represent the mean and s.e.m. (n = 7). (c) Histological analysis of ICM’s treatment (Supplementary Fig. 24), suggesteffects on the EAE model compared with preimmunized mice (naive). Frozen sections of lumbar ing the therapeutic potential of ICM in the spinal cords were stained with FluoroMyelin for myelin (top) and Iba-1–specific antibody for clinical setting. Therefore, ICM might provide microglial activation (bottom). (d) The frozen sections in c were subjected to hematoxylin and an important breakthrough for the treatment of eosin staining to show inflammatory lesions. Scale bars (c,d), 200 μm. (e) The effect of ICM on HMGB-mediated neuroinflammation as well as the mRNA expression of proinflammatory cytokines and chemokines in microglia isolated from neurodegenerative diseases. the spinal cords of naive or EAE mice. (f) The levels of HMGB2 in cerebrospinal fluid and serum of HMGB-targeted anti-inflammatory thernaive or EAE mice were measured by ELISA. ICM (10 mg per kg body weight) was administrated apy was previously focused on a scavengin the EAE experiment, and samples were prepared at the disease peak time (Supplementary ing strategy using HMGB1-specific antibody Fig. 15). The data in a, e and f represent the mean and s.d. in triplicate. *P < 0.05, **P < 0.001, and A-box mimic peptides or on a receptor***P < 0.0001 compared to vehicle, as calculated by Student’s t-test. blocking strategy with RAGE-specific antibody. However, both strategies suffer from (Supplementary Figs. 21 and 22). Because ICM blocked the extra- limited cellular uptake or metabolic instability14,16,26,27. A small cellular release of HMGBs as an inflammatory cytokine, we mea- molecule–based strategy can effectively overcome these limitations. sured HMGB2 levels in the serum and cerebrospinal fluid of EAE Recently, glycyrrhizin, a natural product isolated from the root of mice. MOG immunization caused an increase in HMGB2 plasma Glycyrrhiza glabra, was reported to be the first HMGB1-binding levels, whereas ICM treatment effectively suppressed the release molecule28. However, the anti-inflammatory efficacy of glycyrof HMGB2 into body fluids after EAE induction (Fig. 5f). These rhizin is very low in LPS-activated microglia, with an approximately results strongly support the therapeutic potential of ICM as an anti- 600-fold lower potency than that of ICM (Supplementary Fig. 25), neuroinflammatory and neuroprotective agent in vivo. and its interaction with other HMGBs has not been explored. In contrast, using functional studies and additional mechanistic studDISCUSSION ies, we demonstrated that ICM regulated an inflammatory function Despite the deteriorating role of neuroinflammation in many of the HMGBs through direct binding. neuro­logical diseases, the number of existing anti-inflammatory ICM exhibits a unique mode of action as a PTM modulator, drugs is quite limited because of insufficient efficacy or undesired regulating the secretion of HMGBs (Fig. 4). A computational study side effects24. Therefore, there is a growing need for the develop- of the binding site of ICM to HMGB1 led to a structural insight for ment of NCEs with new modes of action for the treatment of neuro­ the ICM-induced nuclear translocation of HMGBs. On the basis of inflammatory disease. To address these unmet needs, we used the docking simulation, the binding site of ICM was predicted to be high-throughput phenotypic screening and discovered ICM as a the DNA-binding domain in box A of HMGB, which is located adjanew anti-microglial and anti-inflammatory agent. Furthermore, we cent to the NLS (Supplementary Figs. 26 and 27). This result supidentified HMGBs as a direct protein target of ICM using FITGE ports the idea that ICM binding perturbed the PTM near the NLS technology. Although there are reports in the literature regarding site and then inhibited the nuclear translocation of HMGBs. In fact, the role of HMGBs as inflammatory cytokines, most of the studies the HMGBs were originally identified as DNA-binding proteins that b

3

Vehicle ICM 10 mg/kg

HMGB2 (ng/ml)

mRNA expression (fold change)

Ventral horn

Hippocampus

Ventral column

SN

Fluoromyelin

Cortex

Iba-1

Iba-1+ cell number

600

Clinical score

a

© 2014 Nature America, Inc. All rights reserved.

article

nature CHEMICAL BIOLOGY | Advance online publication | www.nature.com/naturechemicalbiology

5

© 2014 Nature America, Inc. All rights reserved.

article

Nature chemical biology doi: 10.1038/nchembio.1669

regulated transcription, replication or DNA repair15. If ICM universally blocked the functions of HMGBs, it would clearly cause severe side effects. In spite of the various properties of the HMGBs, however, ICM inhibited the inflammatory function of HMGBs without affecting the interaction of HMGBs with DNA (Supplementary Fig. 28), thereby exhibiting no severe adverse effects or cytotoxicity (Supplementary Figs. 1c and 6). In addition, further evaluation with neuronal progenitor cells in SVZ using BrdU-injected animals confirmed no statistically significant effect of ICM on neurogenesis in vivo (Supplementary Fig. 19), supporting the specific regulation of ICM on microglia-mediated neuroinflammation. In our previous study, we observed that ICM analogs suppressed osteoclastogenesis by inhibiting the NF-κB and MAPK signaling pathways in mouse macrophages and bone marrow monocytes10. The NF-κB and MAPK signaling pathways are the key regulators of inflammation induced by HMGBs17, and HMGB1 signaling has a crucial role in arthritis as well as in the activation of macrophages29. In this study, we found that ICM also inhibited the NF-κB and MAPK signaling pathways in microglia, resulting in an antiinflammatory effect in macrophages (Fig. 1e and Supplementary Figs. 2a and 5). The inhibitory activity of ICM in these signaling cascades provides strong support for the molecular target of ICM and explains the observed phenotype and mechanism of action of ICM in activated microglia. The identification of the target of ICM allowed the discovery of the ‘missing link’ for the integration of previous observations into a coherent explanation. In summary, the integration of phenotype-based screening with FITGE-based target identification led to the discovery of a new chemical entity, ICM. This convergent strategy and subsequent mechanistic studies revealed that a new small molecule, a PTM modulator of HMGBs, is an inhibitor of neuroinflammation with a broad window of therapeutic possibilities for various neuroinflammatory diseases. Received 19 March 2014; accepted 18 September 2014; published online 12 October 2014

Methods

Methods and any associated references are available in the online version of the paper.

References

1. Streit, W.J., Mrak, R.E. & Griffin, W.S.T. Microglia and neuroinflammation: a pathological perspective. J. Neuroinflammation 1, 14 (2004). 2. Block, M.L., Zecca, L. & Hong, J.-S. Microglia-mediated neurotoxicity: uncovering the molecular mechanisms. Nat. Rev. Neurosci. 8, 57–69 (2007). 3. Block, M.L. & Hong, J.-S. Microglia and inflammation-mediated neurodegeneration: multiple triggers with a common mechanism. Prog. Neurobiol. 76, 77–98 (2005). 4. Swinney, D.C. & Anthony, J. How were new medicines discovered? Nat. Rev. Drug Discov. 10, 507–519 (2011). 5. Kotz, J. Phenotypic screening, take two. SciBX. doi:10.1038/scibx.2012.380 (2012). 6. Terstappen, G.C., Schlüpen, C., Raggiaschi, R. & Gaviraghi, G. Target deconvolution strategies in drug discovery. Nat. Rev. Drug Discov. 6, 891–903 (2007). 7. Schenone, M., Dančík, V., Wagner, B.K. & Clemons, P.A. Target identification and mechanism of action in chemical biology and drug discovery. Nat. Chem. Biol. 9, 232–240 (2013). 8. Andreux, P.A., Houtkooper, R.H. & Auwerx, J. Pharmacological approaches to restore mitochondrial function. Nat. Rev. Drug Discov. 12, 465–483 (2013). 9. Park, J., Oh, S. & Park, S.B. Discovery and target identification of an antiproliferative agent in live cells using fluorescence difference in twodimensional gel electrophoresis. Angew. Chem. Int. Ed. Engl. 51, 5447–5451 (2012).

6

10. Oh, S. & Park, S.B. A design strategy for drug-like polyheterocycles with privileged substructures for discovery of specific small-molecule modulators. Chem. Commun. (Camb.) 47, 12754–12761 (2011). 11. Zhu, M. et al. Discovery of novel benzopyranyl tetracycles that act as inhibitors of osteoclastogenesis induced by receptor activator of NF-κB ligand. J. Med. Chem. 53, 8760–8764 (2010). 12. Lotze, M.T. & Tracey, K.J. High-mobility group box 1 protein (HMGB1): nuclear weapon in the immune arsenal. Nat. Rev. Immunol. 5, 331–342 (2005). 13. Sims, G.P., Rowe, D.C., Rietdijk, S.T., Herbst, R. & Coyle, A.J. HMGB1 and RAGE in inflammation and cancer. Annu. Rev. Immunol. 28, 367–388 (2010). 14. Yang, H. & Tracey, K.J. Targeting HMGB1 in inflammation. Biochim. Biophys. Acta 1799, 149–156 (2010). 15. Štros, M. HMGB proteins: interactions with DNA and chromatin. Biochim. Biophys. Acta 1799, 101–113 (2010). 16. Wang, H. et al. HMG-1 as a late mediator of endotoxin lethality in mice. Science 285, 248–251 (1999). 17. Ulloa, L. & Messmer, D. High-mobility group box 1 (HMGB1) protein: friend and foe. Cytokine Growth Factor Rev. 17, 189–201 (2006). 18. Gardella, S The nuclear protein HMGB1 is secreted by monocytes via a non-classical, vesicle-mediated secretory pathway. EMBO Rep. 3, 995–1001 (2002). 19. Rendon-Mitchell, B. et al. IFN-γ induces high mobility group box 1 protein release partly through a TNF-dependent mechanism. J. Immunol. 170, 3890–3897 (2003). 20. Scaffidi, P., Misteli, T. & Bianchi, M.E. Release of chromatin protein HMGB1 by necrotic cells triggers inflammation. Nature 418, 191–195 (2002). 21. Zhang, Q. & Wang, Y. High mobility group proteins and their posttranslational modifications. Biochim. Biophys. Acta 1784, 1159–1166 (2008). 22. Bonaldi, T. et al. Monocytic cells hyperacetylate chromatin protein HMGB1 to redirect it towards secretion. EMBO J. 22, 5551–5560 (2003). 23. Youn, J.H. & Shin, J.-S. Nucleocytoplasmic shuttling of HMGB1 is regulated by phosphorylation that redirects it toward secretion. J. Immunol. 177, 7889–7897 (2006). 24. Craft, J.M., Watterson, D.M. & Eldik, L.J.V. Neuroinflammation: a potential therapeutic target. Expert Opin. Ther. Targets 9, 887–900 (2005). 25. Ellerman, J.E., Brown, C.K. & Vera, M. Masquerader: high mobility group box-1 and cancer. Clin. Cancer Res. 13, 2836–2848 (2007). 26. Yang, H. et al. Reversing established sepsis with antagonists of endogenous high-mobility group box 1. Proc. Natl. Acad. Sci. USA 101, 296–301 (2004). 27. Fiuza, C. et al. Inflammation-promoting activity of HMGB1 on human microvascular endothelial cells. Blood 101, 2652–2660 (2003). 28. Mollica, L. et al. Glycyrrhizin binds to high-mobility group box 1 protein and inhibits its cytokine activities. Chem. Biol. 14, 431–441 (2007). 29. Kokkola, R. et al. Successful treatment of collagen-induced arthritis in mice and rats by targeting extracellular high mobility group box chromosomal protein 1 activity. Arthritis Rheum. 48, 2052–2058 (2003).

Acknowledgments

This work was supported by a Creative Research Initiative grant (2014R1A3A2030423), the Bio & Medical Technology Development Program (2012M3A9C4048780) and the Basic Research Laboratory (2010-0019766) funded by the National Research Foundation of Korea (NRF). K.S. was supported by a NRF grant funded by the Korean government (MSIP) (2008-0062282) and a grant of the Korea Healthcare Technology R&D Project, Korean Ministry of Health and Welfare (A111345). S.L., Y.N., J.Y.K., D.L., J.P. and J.K. are grateful for a BK21 Scholarship.

Author contributions

S.L. performed the biochemical assays, analyzed the data and prepared the manuscript. Y.N. and J.K. performed the biochemical assays and in vivo experiments and analyzed the data. J.Y.K. synthesized the compounds and conducted the computational study. D.L. and J.O. performed experiments for target validation. J.P. contributed to the target identification. K.S. and S.B.P. directed the study and were involved in all aspects of the experimental design, data analysis and manuscript preparation. All authors critically reviewed the text and figures.

Competing financial interests

The authors declare no competing financial interests.

Additional information

Supplementary information, chemical compound and chemical probe information is available in the online version of the paper. Reprints and permissions information is available online at http://www.nature.com/reprints/index.html. Correspondence and requests for materials should be addressed to S.B.P. or K.S.

nature chemical biology | Advance online publication | www.nature.com/naturechemicalbiology

ONLINE METHODS

© 2014 Nature America, Inc. All rights reserved.

Reagents. Chemicals, including LPS (L2880, Sigma) and glycyrrhizin (G0150, TCI), were purchased from commercial venders. Cell culture reagents, including FBS, media and antibiotic-antimycotic solution, were purchased from Gibco, Invitrogen. All antibodies for immunoblotting analysis were purchased from R&D systems, Santa Cruz, Abcam, Cell Signaling and Wako. The purified HMGB1 (ab73658) and HMGB2 (ab91926) proteins were purchased from Abcam. Fluorescent-labeled ODN and native ODN were purchased from InvivoGen. Cell culture. The BV-2 and HAPI microglial cell lines were obtained from American Type Culture Collection and cultured in DMEM supplemented with 1% (v/v) antibiotic-antimycotic solution and heat-inactivated 5% or 10% (v/v) FBS, respectively. RAW 264.7 cells were obtained from American Type Culture Collection and cultured in RPMI 1640 medium supplemented with heat-inactivated 10% (v/v) FBS and 1% (v/v) antibiotic-antimycotic solution. B35-eGFP neuroblastoma cells were obtained by stably transfecting an eGFP construct into B35 rat neuroblastoma cells. Cells were maintained in a humidified atmosphere of a 5% CO2 incubator at 37 °C and cultured in 100-mm cell culture dishes. For primary astrocytes and microglial cultures, the whole brains of 3-day-old mice were chopped and mechanically disrupted using a nylon mesh. The mixed glial cells obtained were seeded in culture flasks and grown at 37 °C in a 5% CO2 atmosphere in DMEM supplemented with 10% FBS, 100 U/ml penicillin and 100 μg/ml streptomycin. Culture media were changed initially after 5 d and then changed every 3 d. After 14 d in culture, primary astrocytes and microglia were obtained using shaking or a mild trypsinization method from mixed glial cells and maintained in DMEM supplemented with 10% FBS and penicillin-streptomycin. Primary cortical neuron cultures were prepared from E15 C57BL/6 mice. Briefly, mouse embryos were decapitated, and the brains were removed rapidly and placed in a culture dish with cold PBS. The cortices were isolated and then transferred to a culture dish containing 0.25% trypsin-EDTA in PBS for 30 min at 37 °C. After two washes in serum-free neuro­basal medium (Gibco-BRL), dissociated cortical cells were seeded onto poly D-lysine–coated plates using neurobasal medium containing 10% FBS, 0.5 mM glutamine, 100 U/ml penicillin, 100 mg/ml streptomycin, N2 supplement and B27 supplement. Cells were maintained by changing the medium every 2–3 d and grown at 37 °C in a 5% CO2 humidified atmosphere. For E15 striatum cell culture, striatal tissue was isolated from E15 C57BL/6 mice and placed in a culture dish with cold Hank’s balanced salt solution (HBSS) buffer. The tissue was transferred to a culture dish containing 0.025% trypsin-EDTA in HBSS for 15 min at 37 °C and then washed with FBS-containing medium. Striatal tissues were mechanically dissociated with a Pasteur pipette and suspended in 10 ml of neurobasal medium containing 1% FBS, 100 U/ml penicillin, 100 mg/ml streptomycin, 2 mM glutamine, 10 ng/ml nerve growth factor (NGF) and B27 supplement (Gibco-BRL). Cells were counted with a hemocytometer and seeded onto poly D-lysine–coated 96-well plates. After 1 d, the cultures were treated with 2 μM Ara-C (C1768, Sigma) to remove glial cells for some experiments. Animals. C57BL/6 mice (25–30 g) were supplied by Samtaco. The animals were maintained in temperature- and humidity-controlled conditions with a 12-h light, 12-h dark cycle. All animal experiments were approved by the Institutional Review Board of Kyungpook National University School of Medicine and were carried out in accordance with the guidelines in the US National Institutes of Health Guide for the Care and Use of Laboratory Animals. Griess assay. The Griess assay was used for quantification of the secretion of nitrite. Cells were treated with compounds in the absence or presence of LPS (100 ng/ml). After 24 h of incubation, the cell culture media were reacted with Griess reagent (0.1% naphthylethylenediamine dihydrochloride and 1% sulfanilamide in 2% phosphoric acid). Absorbance was measured at 550 nm using a microplate reader, and the level of nitrite was estimated by comparison with a sodium nitrite standard curve. Cell viability assay. Cell viability was measured with MTT (3-(4,5­dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide; M-6494, Life Techno­ logies) or WST (water-soluble tetrazolium salt; W201-12, Dojindo) assay kits, and the experimental procedure was based on the manufacturer’s manual.

doi:10.1038/nchembio.1669

RT-PCR. For the analysis of gene expression, cells were incubated with ICM for 6 h in the absence or presence of LPS. Total RNA was extracted from cells using TRIzol reagent (Invitrogen) according to the manufacturer’s instructions. Reverse transcription was conducted using Superscript II (Invitrogen) and oligo(dT) primers. PCR amplification was conducted using a DNA Engine Tetrad Peltier Thermal Cycler and a C1000 Touch Thermal cycler (Bio-Rad) at an annealing temperature of 55–60 °C for 20–30 cycles using specific primer sets (Supplementary Table 4). To analyze PCR products, each sample was electro­phoresed on a 1% agarose gel and detected under UV light. Gapdh was used as an internal control. ELISA TNF-α secretion was measured using a TNF-α ELISA kit (DY410, R&D Systems). BV-2 cells were treated with LPS in the absence or presence of ICM. After a 24-h incubation, the levels of TNF-α in the culture medium were measured with rat monoclonal anti-mouse TNF-α antibody (1:180 dilution, 840143, R&D Systems) as the capture antibody and goat biotinylated poly­ clonal anti-mouse TNF-α antibody (1:180 dilution, 840144, R&D Systems) as the detection antibody. The biotinylated anti–TNF-α antibody was detected by sequential incubation with streptavidin–horseradish peroxidase (HRP) conjugate (1:120 dilution, 890803, R&D Systems) and TMB substrates (DY999, R&D Systems). After incubation for 20 min, the color development was stopped by adding 2 N H2SO4. The absorbance was then read at 450 nm and 540 nm using a microplate reader (Molecular devices). The level of HMGB2 in body fluids was measured by a HMGB2-ELISA kit (CSB-EL010560MO, CUSABIO Life Science) according to the manufacturing guidelines. Immunofluorescence staining. Cells were pretreated with compounds, exposed to LPS for 1 h and then fixed with 4% paraformaldehyde for 30 min at room temperature and with cold methanol for 10 min at −20 °C. After permeabilization with 0.3% Triton X-100 and PBS for 10 min, the fixed cells were blocked with 1% normal horse serum for 1 h and incubated with mouse anti-p65 antibody (1:100 dilution, sc-8008, Santa Cruz) at 4 °C overnight. After washing with PBS containing 0.05% Tween-20 (PBST), Alexa Fluor-488–labeled goat anti-mouse IgG antibody was added to the sample, incubated for 1 h at room temperature and washed with PBST. Nuclei were visualized by 4′,6-diamidino2-phenylindole (DAPI) staining (Vector Laboratories). Samples were analyzed by fluorescence microscopy. Microglia and neuron coculture. For the coculture of microglial cells and neuro­ blastoma cells, HAPI microglial cells were exposed to ICM and LPS (100 ng/ml) for 8 h. The medium was replaced with fresh medium containing B35-eGFP neuroblastoma cells, and the cocultures of microglia and neuro­blastoma were incubated for 24 h. At the end of the cocultures, the B35-eGFP neuroblastoma cells were counted to assess cell viability. The coculture of primary microglial cells and cortical neurons was done in a similar manner. Cortical neurons were labeled with CMFDA (C2925, Molecular Probe) before initiating the coculture. The viability of cortical neurons was determined after the coculture for 48 h. Assessment of neurogenesis in SVZ. C57BL/6 mice (10 weeks) were given an intraperitoneal (i.p.) injection of 200 mg BrdU per kg body weight (B5002, Sigma) every 12 h for 2.5 days (total of five injections) and perfused 12 h after the final injection. Mice received a single intracerebroventricular (posterior, 0.5 mm; lateral, 1 mm; vertical, 1.75 mm) injection of vehicle or ICM (10 or 50 nM) using a stereotaxic apparatus before the first injection of BrdU. Mice were anesthetized with diethyl ether, transcardially perfused with cold saline and perfused with 4% paraformaldehyde (PFA) diluted in 0.1 M PBS. The brains were fixed using 4% PFA for 3 d and then cryoprotected with 30% sucrose solution for 3 d. Fixed brains were embedded in optimal cutting temperature (OCT) compound for frozen sectioning and then sectioned sagittally at 20 μm. To detect cell proliferation in vehicle- or ICM-injected brains, sections were washed with PBS, and BrdU antigen retrieval was performed using 2 N HCl treatment for 30 min at 37 °C followed by three washes in 0.1 M borate buffer (pH 8.0) and three washes in PBS. All sections were blocked for 1 h at room temperature with 5% donkey serum, 0.3% BSA and 0.3% Triton X-100 in PBS solution and then stained with rat anti-BrdU antibody (1:500 dilution, MCA2060, Serotec). Synthesis. The synthetic approach for ICM and its derivatives, including ICM-BP, is described in detail in the Supplementary Note.

nature CHEMICAL BIOLOGY

© 2014 Nature America, Inc. All rights reserved.

In-gel analysis for identification of the protein target. BV-2 cells were treated with the compounds for 30 min and then with LPS for 2 h. Then the cells were irradiated with 365-nm UV light for 30 min on ice. The cells were washed with PBS and stored at −80 °C. Cells were lysed in RIPA buffer containing a protease inhibitor cocktail, and the protein concentration was adjusted to 1 mg/ml. The proteome was labeled with Cy5-azide (40 μM), Tris[(1-benzyl1H-1,2,3-triazol-4-yl)methyl]amine (TBTA) (100 μM), CuSO4 (1 mM), Tris(2carboxyethyl)phosphine (TCEP) (1 mM) and t-BuOH (5%) for 1 h. Acetone was added to the mixture to precipitate the proteins, and the mixture was kept at −20 °C for 20 min. After centrifugation at 4 °C, 14,000 r.p.m. for 10 min, the pellet was washed twice with cold acetone. For 1D gel analysis, the pellet was dissolved in Laemmli sample buffer and analyzed by SDS-PAGE. For 2D gel analysis, the pellet was dissolved in rehydration buffer. The proteome labeled with ICM-BP (labeled with Cy5-azide) and the control (labeled with Cy3-azide) were mixed (1:1 ratio). The mixed proteomes were analyzed by 2D gel electrophoresis. In-gel fluorescence was scanned with a Typhoon Trio. Mass analysis. The residual peptides in desired bands and spots were extracted by trypsin in-gel digestion and then subjected to LC/MS/MS analysis. In detail, selected protein spots and bands were excised and dehydrated in acetonitrile for 10 min. The acetonitrile was removed and dried under vacuum. For mass analysis, the resulting gel pieces were re-swelled at 4 °C for 45 min in buffer containing trypsin and 50 mM (NH4)2CO3 and incubated overnight at 37 °C for trypsin digestion. The samples were centrifuged, and the supernatants were collected for mass analysis. The residual peptides in gel pieces were further extracted with 50% acetonitrile containing 20 mM (NH4)2CO3 and 5% formic acid three times at room temperature. The combined peptide samples were condensed down in a SpeedVac until the desired sample concentration was reached. LC/MS/MS experiments were performed at National Instrumentation Center for Environmental Management (NICEM) at Seoul National University. A hybrid quadrupole-TOF LC/MS/MS spectrometer (Q-Star Elite) had a nanoelectrospray ionization source and was fitted with a fused silica emitter tip. For each LC/MS/MS run, 1–2 μg of the fractionated peptides was injected into the LC/MS/MS system, and the peptides were trapped and concentrated on an Agilent Zorbax 300SB-C18 column. The peptide mixture was separated on an Agilent Zorbax 300SB nanoflow C18 column at a flow rate of 300 nl per min, and eluted peptides were electrosprayed through a coated silica tip (ion spray voltage at 2,300 eV). ProteinPilot Software 2.0.1 (Software Revision Number: 67476) was used to identify peptides and proteins and quantify differentially expressed proteins. Pulldown assay. Cell lysates were prepared according to the same protocol used for the in-gel analysis before click chemistry. The proteome was reacted with biotin-azide (50 μM) instead of Cy5-azide. The mixture was precipitated according to the same protocol used for the in-gel analysis. The pellets were dissolved in PBS containing 1.2% SDS using sonication and then diluted with PBS containing 0.2% SDS. The samples were incubated with avidin beads for 2 h at room temperature and washed with PBS several times. Samples were denatured in Laemmli sample buffer with heating and analyzed by SDS-PAGE and immunoblotting for HMGB2. Western blot analysis. The proteomes were analyzed by SDS-PAGE and transferred to PVDF membranes. The membranes were blocked with 2% BSA in TBST for 1 h. The membranes were incubated overnight at 4 °C with the specific primary antibody (MAPK signaling: ERK, JNK, p38 (1:1,000 dilution, 4695, 9252, 9212, Cell Signaling) and their phosphorylated forms (9101, 9251, 9211, Cell Signaling); HMGB1 and HMGB2 (1:2,000 dilution, ab67282, ab18256, Abcam)) to detect the desired proteins and then washed with TBST. The resulting membranes were exposed to HRP-conjugated secondary antibody (1:5,000 dilution, 7074, Cell Signaling) for 1 h at room temperature. After washing, the membranes were developed using an enhanced chemiluminescence (ECL) detection kit, and the chemiluminescent signal was detected by an imaging system. siRNA-mediated knockdown assay. The siRNA transfection of BV-2 microglial cells was performed using Lipofectamine 2000 (Invitrogen) according to the manufacturer’s instructions. At 48 h after transfection, the cells were used for experiments. Subcellular fractionation and secretion analysis. After treatment with a compound or LPS for 18 h, cells were lysed in subcellular fraction buffer (250 mM nature chemical biology

sucrose, 20 mM HEPES, 10 mM KCl, 1 mM EDTA, 1 mM EGTA, 1 mM DTT and protease inhibitor cocktail (Roche)), passed through a 25-G needle several times and incubated on ice for 20 min. After centrifugation at 3,000 r.p.m. for 5 min, the pellet and supernatant were separated. The nuclear pellet was dispersed in buffer, centrifuged to remove the washing buffer and then resuspended in standard lysis buffer with 10% glycerol and 0.1% SDS. The cytosolic fraction was obtained by centrifugation of the separated supernatant at 8,000 r.p.m. for 15 min. The conditioned medium was collected for the analysis of the extracellular level of HMGBs and concentrated with Amicon Ultra 10K filters (Millipore). The samples were analyzed by SDS-PAGE and western blotting. Immunoprecipitation analysis. For the analysis of the PTM levels of phosphorylation and acetylation on HMGB, immunoprecipitation and western blotting were performed. After a 4-h incubation with a compound in the absence or presence of LPS, cells were harvested and lysed in IP buffer containing a protease inhibitor cocktail and the appropriate inhibitor, i.e., PMSF, NaF and Na3VO4 for phosphorylation or nicotinamide and trichostatin A for acetylation. The concentration of protein was measured by bicinchoninic acid assays, and the lysates were incubated with anti-phosphoserine antibody (1:1,000 dilution, ab9332, Abcam) or anti–acetylated lysine antibody (1:1,000 dilution, 9441, Cell Signaling) at 4 °C overnight and subsequently precipitated using a protein G immunoprecipitation kit (Sigma) following the manufacturer’s protocol. Samples were analyzed with the identical procedure used for the western blot assays with anti-HMGB2 as the primary antibody. LPS neuroinflammation model. All experiments were carried out on 11-week-old male C57BL/6 mice (25–30 g). The animals were divided into four experimental groups in each experiment: group 1, treated with vehicle; group 2, treated with ICM; group 3, treated with LPS and vehicle; and group 4, treated with LPS and ICM. The ICM (2 or 10 mg per kg body weight) or vehicle (distilled water containing 5% DMSO and 40% polyethylene glycol) was administered i.p. daily for 4 d. LPS (from Escherichia coli 055:B5; Sigma) was administered i.p. at a dose of 5 mg per kg body weight on day 2 for a single challenge. EAE induction. C57BL/6 mice (7–8 weeks old, female) were immunized subcutaneously with 200 μg oMOG35–55 (MEVGWYRSPFSRVVHLYRNGK) (GLBiochem) in 100 μl of a solution containing 50% complete Freund’s adjuvant with 10 mg/ml of the heat-killed H37Ra strain of Mycobacterium ­tuberculosis (Difco) in areas draining into the axillary and inguinal lymph nodes. Pertussis toxin (200 ng per mouse; List Biological Laboratories) in PBS was administered i.p. on the day of immunization and again 48 h later. Animals were weighed and examined for disease symptoms daily. Clinical signs of disease were scored using a 0–5 scale, as follows: 0, no clinical signs; 1, limp tail; 2, weakness and incomplete paralysis of one or two hindlimbs; 3, complete hindlimb paralysis; 4, forelimb weakness or paralysis; and 5, moribund state or death. Vehicle or ICM was injected i.p. daily for 15 d after MOG immunization. Histological analysis. Mice were anesthetized with diethyl ether, transcardially perfused with cold saline and then perfused with 4% PFA diluted in 0.1 M PBS. Brains or lumbar spinal cords were fixed using 4% PFA for 3 d and then cryoprotected with a 30% sucrose solution for 3 d. Three animals were used per experimental group. Fixed brains and spinal cords were embedded in OCT compound (Tissue-Tek, Sakura Finetek) for frozen sectioning and then sectioned coronally at 20 μm. To detect microglial activation in LPS-injected brains or microglial activation, neurites and demyelination in EAE spinal cords, sections were incubated with rabbit anti–Iba-1 antibody (1:500 dilution, MB100-1028, Novus), FluoroMyelin (1:300 dilution, F34651, Invitrogen), anti–MAP-2 (1:200 dilution, M9942, Sigma) or anti-MBP antibodies (1:200 dilution, ab24567, Abcam). Sections were visualized directly or incubated with Cy3-conjugated anti-rabbit IgG antibody (711-165-152, Jackson Laboratory). Sections were stained with hematoxylin and eosin to assess inflammatory lesions. Spinal cord microglia isolation. Spinal cords were homogenized in HBSS with collagenase and DNase. The resulting homogenates were passed through a nylon cell strainer and centrifuged at 500g for 6 min. Supernatants were removed, and cell pellets were resuspended in 37% isotonic Percoll (Amersham Biosciences) at room temperature. A discontinuous Percoll density gradient doi:10.1038/nchembio.1669

was set up as follows: 70%, 37%, 30% and 0% isotonic Percoll. The gradient was centrifuged for 20 min at 2,000g, and microglia cells were collected from the interphase between the 37% and 30% Percoll layers. Microglial cells were washed and then resuspended in sterile HBSS.

Databases. MS results in the target identification were searched against a National Center for Biotechnology Information (NCBI) or International Protein Index (IPI) database using the Paragon and Pro Group algorithms. The X-ray crystal used in the docking simulation is from Protein Data Bank (PDB).

© 2014 Nature America, Inc. All rights reserved.

Statistics. Analyses of in vitro experiments and LPS-induced in vivo experiments were performed with Student’s t test using GraphPad Prism. Analysis of the EAE experiment was performed with the Mann-Whitney U test using

SPSS software. Data represent the mean and s.d or s.e.m., as indicated in the individual figure legends. The P value summary related to DMSO, mock transfection or vehicle is indicated in the individual figure legends.

doi:10.1038/nchembio.1669

nature CHEMICAL BIOLOGY