Poligapolide, a PI3K/Akt Inhibitor in Immunodeficiency Virus Type 1 TAT-Transduced CHME5 Cells, Isolated from the Rhizome of Polygala tenuifolia

Abstract

The rhizome of Polygala tenuifolia WILLD (PT, family Polygalaceae) has been used in traditional Chinese medicine for inflammation, dementia, amnesia, neurasthenia and cancer. The phosphoinositide 3-kinase (PI3K)/Akt inhibitor(s) was isolated from PT by using the cytoprotective phenotype of human immunodeficiency virus type 1 (HIV-1) Tat-transduced CHME5 cells against lipopolysaccharide/cycloheximide. We isolated 9 constituents (1)–(9) from ethyl acetate fraction of PT, which potently showed anti-cytoprotective effect against HIV-1 TAT-transduced cells. Of them, (9R)-(−)-9-peptandecanolide (2), a new compound named poligapolide, most potently abolished the cytoprotective effect of HIV-1 Tat-transduced CHME5 cells. The compound (2) inhibited the phosphorylation of Akt and its downstream molecule, glycogen synthase kinase-3 beta (GSK3β) in PI3K/Akt cell survival signaling pathway, but did not suppress the phosphorylation of PI3K and pyruvate dehydrogenase lipoamide kinase isozyme 1. Based on these finding, poligapolide may abolish the cytoprotective phenotype of HIV-1 Tat-transduced CHME5 cells by inhibiting Akt phosphorylation in PI3K/Akt pathway.

The phosphoinositide 3-kinase (PI3K)/Akt cell survival pathway is particularly activated by several key human pathogenic viruses such as human immunodeficiency virus type 1 (HIV-1)¹⁾ and human papillomavirus (HPV).²⁾ This virus-induced activation of PI3K/Akt pathway involves specific viral proteins such as E6/E7 of HPV, NS5A of hepatitis C virus (HCV), Tax of human T-lymphotrophic virus type (HTLV) and Tat of HIV-1. Of them, HIV-1 Tat appears to inactivate phosphatase and tensin homology (PTEN), a negative regulator of the PI3K/Akt pathway.^3,4) For example, the transfection of HIV-1 Tat into CHME5 cells, a human microglial cell line, and human primary macrophages elevates their cytoprotective phenotypes against cellular stresses.^4,5) More specifically, the HIV-1 Tat-expressing CHME5 cells activate the PI3K/Akt pathway upon exposure to cellular stresses by reducing the level of PTEN.^3,5) These cells render strong resistance to extracellular stresses such as lipopolysaccharide (LPS) or nitric oxide. Tat-expressing human microglia and macrophages play important roles in the establishment of long-living HIV-1 reservoirs in the central nervous system,^6,7) which in turn induces neuronal death and HIV-1 associated neurodegenerative diseases.⁸⁾ Also the PI3K/Akt pathway is also highly activated in many cancer cells.⁹⁾ Therefore, PI3K/Akt inhibitors have been extensively searched as potential anti-cancer and anti-HIV-1 dementia agents. Nevertheless, the studies on the isolation of PI3K/Akt inhibitors from natural products have not been thoroughly.

The rhizome of Polygala tenuifolia WILLD (PT, Polygalaceae), which contains xanthones, and triterpenoid saponin, has been used in traditional Chinese medicine for inflammation, dementia, amnesia, neurasthenia and cancer.^10,11) As a part of our studies on anti-PI3K/Akt constituents from natural medicines, we isolated 4 anti-cytoprotective compounds, clionasterol, ethyl cholestan-22-en-3-ol, 3-O-β-D-glucosyl ethyl cholestan-22-en-3-ol, and 3-O-β-D-glucopyranosyl clionasterol from EtOAc fraction of PT against HIV-1 Tat-transduced CHME5 cells by inhibiting PDK1 phosphorylation, in the previous study.¹²⁾

In the present study, we continuously screened PI3K/Akt inhibitory constituents from PT, isolated a novel constituent named poligapolide and investigated its inhibitory effect against the activation of PI3K, pyruvate dehydrogenase lipoamide kinase isozyme 1 (PDK1), Akt and glycogen synthase kinase-3 beta (GSK3β) in HIV-1 Tat-transduced CHME5 cells.

Results and Discussion

To search PI3K/Akt inhibitors from PT, first we isolated 9 compounds in EtOAc fraction of PT (Fig. 1A) and measured their anti-cytoprotective effects by employing HIV-1 Tat-induced cytoprotective phenotype of CHME5 cells (Table 1). Among the isolates, compound 2 has never been reported previously. Compound 2 exhibited the most potent anti-cytoprotective activity, followed by compounds 5 and 9 (Fig. 1A).

Fig. 1. Structures of Constituents Isolated from the Rhizome of Polygala tenuifolia

(A) Structures of 9 isolated constituents. (B) Structure of PT2 analyzed with COZY experiments. (C) Structure of PT2 measured by HMBC experiment.

Table 1. The Anti-cytoprotective Effects of 9 Constituents Isolated from EtOAc-Soluble Fraction against HIV-1 Tat-Expressing CHME5 Cells

Compound	Dead cellsa) (%)
	Trypan blue staining		PI staining
	With LPS/CHX	Without LPS/CHX	With LPS/CHX	Without LPS/CHX
1	8.6±0.4	3.9±0.1	7.4±0.1	3.4±0.2
2	19.3±0.5	4.0±0.5	18.7±0.1	3.4±0.1
3	12.9±0.4	3.9±0.1	13.4±0.1	3.7±0.1
4	11.8±0.4	3.9±1.0	10.8±0.1	3.3±0.2
5	17.0±0.3	3.9±0.2	16.0±0.1	3.6±0.1
6	12.6±1.5	3.2±0.5	14.4±0.1	3.4±0.2
7	13.0±0.5	3.2±1.4	13.5±0.1	3.5±0.1
8	12.9±1.5	4.0±1.3	13.4±0.1	3.5±0.1
9	14.9±0.5	3.4±0.5	15.0±0.2	3.3±0.3
None	3.6±0.2	2.9±0.4	3.2±0.1	2.9±0.2
Miltefosine	20.6±1.1	8.8±0.2	21.5± 0.2	8.8±0.3

Compound 2 was obtained as white crystals. The ¹H-NMR spectrum showed the characteristic absorptions due to an oxygenated proton (δ 4.50, m, 1H, H-9), a methyl proton at δ 0.84 (t, J=5.0 Hz, 3H, H-16), and methylene protons at δ 2.49 (t, J=10.0 Hz, 1H, 2_a-H), and δ 1.76 (dddd, 3H, J=5.0, 10.0 5.0, 10.0 Hz, 3H, 8_a-, 3_a-, 11_a-H), the overlap proton signals δ 1.33–1.39 (3H, m, 4_a-, 7_a-, 12_a-H) and δ 1.23–1.29 (23H, m, 2_b-, 3_b-, 4_b-, 7_b-, 8_b-, 11_b-, 12_b-H and 5-, 6-, 7-, 10-, 13-, 14-, 15-, 16-, 17-H). In the ¹³C-NMR spectrum, there are 12 signals, including a carbonyl carbon δ 176.4 (C-1), an oxygenated carbon δ 76.79 (C-9) and a methyl carbon signal δ 14.0 (C-18).

The corresponding of carbons and protons were identified by heteronuclear single quantum coherence (HSQC) spectrum. A correlation spectroscopy (COSY) experiment showed that the coupling of the H-8_a (δ 1.76) to the H-9 oxygenated proton at (δ 4.44) was also observed. The H-2_a proton δ 2.49 was also coupled to the H-3_a proton δ 1.76 (Fig. 1B). An hetero-nuclear multiple bond connectivity (HMBC) experiment confirmed this structure. The strong coupling of the H-2a proton with carbonyl C-1 (δ 176.4), and C-3 (δ 25.7) and C-4 (δ 29.6) was obtained. The long correlation of H-9 to C-1 showed that compound 2 owned a lactone ring in structure.^13,14) The connection of H-9 to C-10 suggested that an alkyl attached to the lactone cycling at C-9 (Fig. 1C). The high resolution-electron spray ionization time-of-flight mass spectrum (HR-ESI-TOF-MS) spectrum of compound 2 showed a peak at m/z 267.1730, which was calculated to indicate as M₁=[M−CH₃]⁺ : [C₁₇H₃₁O₂]⁺. Additionally, those peaks at m/z 221.5189, 220.1516, 206.1405 were respectively computed for fragmented ions M₂=M-61=[M−C₃H₇−H₂O]⁺, M₃=M-62=[M−C₃H₈−H₂O]⁺, M₄=M-76=[M−C₄H₁₀−H₂O]⁺, which were matched to fragment of hydrophobic acid.¹⁵⁾ Thus, the molecular formula of compound 2 was determined as C₁₈H₃₄O₂. It possessed two saturation degrees which might guess a cycling and a carbonyl inside the structure of compound 2. When the spectra of compound 2 was compared with those of the previous literatures,^13–16) it was identified as 9-peptandecanolide. Moreover, when the optical rotation of compound 2 {[α]_D²⁰ −12.6 (c=0.04, CHCl₃)} was compared with that of (9R)-(−)-9-tetradecanolide {[α]_D²⁰ −27.6 (c=0.50, CHCl₃)},¹⁶⁾ the negative values or laevorotation of both compounds could be suggested to be similar in configuration at stereo center as (9S-) orientation. Therefore, compound 2 was determined as (9R)-(−)-9-peptandecanolide, named poligapolide.

Poligapolide exhibited the most potent abolishing effect against the cytoprotective CHME5 cells (Fig. 2). Poligapolide abolished Tat-expressing cytoprotective CHME5 cells dose-dependently by tryphan blue staining and PI/FCS assays, although treatment with LPS/CHX alone did not significantly induce cell death. Its anti-cytoprotective potency is comparable to that of miltefosine, which is a commercial Akt-inhibitory drug.³⁾ We also measured cytotoxic effect of poligapolide against control gene (pcDNA3.1-Hygro, without Tat) expressed CHME5 cells by a crystal violet method. No cytotoxic effect of poligapolide (20 µM for 48 h) were observed (<5%) under the condition used in this experiment (Fig. 2C). PI3K/AKT pathway is an important survival signal pathway in many cancer cells. It has been studied that HIV-1 infection also activates the PI3K/AKT pathways. Therefore, inhibitors of this pathway could serve as anticancer and anti-viral agents.^3,17) To confirm whether poligapolide could inhibit PI3K/Akt pathway, we measured its inhibitory effect against the activation of PI3K, PDK1, Akt and GSK3β in HIV-1 Tat-transduced CHME5 cells (Fig. 3). Poligapolide inhibited phosphorylation of Akt and its downstream molecule GSK3β, but did not suppress the activation of PI3K and PDK1, as compound K previously reported.¹⁸⁾ Therefore, it is suggested that poligapolide may abolish the cytoprotective function of HIV-1 Tat-transducing CHME5 cells by inhibiting Akt activation.

Fig. 2. Anti-cytoprotective Effect of Poligapolide Isolated from PT against Tat-Expressing CHME5 Cells

(A) Anti-cytoprotective effect by trypan blue staining assay. Poligapolide (0, 5, 10, 20 µM) was treated for 48 h. (B) Anti-cytoprotective effect by PI/FACS assay. Tat-expressing CHME5 cells were treated with and without 50 µg/mL LPS and 10 µg/mL CHX in the absence or presence of poligapolide (0, 5, 10, 20 µM) for 48 h. Normal control (Con) was treated with vehicle alone. Trypsinized cells stained with PI were measured by flow cytometry (C6 Flow Cytometer® System). (C) Cytotoxic effect by trypan blue staining assay. Tat-expressing CHME5 cells were treated with various concentrations of poligapolode for 48 h. The experiment was performed in tetraplicate. Values indicate the mean±standard deviation (n=4).

Fig. 3. Effect of Poligapolide on PI3K, p-PDK1, Akt, and GSK3β Phosphorylation in LPS/CHX-Stimulated Tat-Expressing CHME5 Cells

(A) Effect on PI3K, p-PDK1, Akt, and GSK3β phosphorylation by immunoblotting. Tat-expressing CHME5 cells were treated with and without LPS/CHX in the absence or presence of poligapolide (0, 5, 10, 20 µM) for 90 min. Immunoblotting for p-Akt, Akt, p-GSK3β, GSK3β, p-PDK1, PI3K, and GAPDH was performed for their lysates. (B) Quantification of the immunoblot data. Intensity of the immunoblotted bands is represented as the ratio of p-PI3K/PI3K, p-PDK1/PDK1, p-Akt/Akt and p-GSK3β/GSK3β. All values are the mean±standard deviation (n=3). ^# p<0.05, significantly different vs. normal control group treated with vehicle alone in the absence of LPS/CHX; * p<0.05, significantly different vs. control group treated with LPS/CHX alone.

In the present study, we isolated 3 sterols (1), (3) and (6), 3 xanthones (4), (7) and (8), 1 phenyl propanoid (5), 1 flavonoid (9), and 1 new macrolide compound (2) poligapolide from the rhizome of PT. Furthermore, isolation of a macrolide compound from PT is the first time. These constituents exhibited anti-cytoprotective effect against HIV-1 Tat-transduced CHME5 cells. Among them, compound (2) poligapolide isolated from PT abolished the cytoprotective HIV-1 Tat-transduced CHME5 cells most potently. Based on these findings, poligapolide, a PI3K/Akt inhibitor, may deliver anti-HIV effect by shortening the life span of HIV-1 infected macrophages.

Experimental

General Methods

1D- and 2D-NMR experiments were performed on a Varian NMR spectrometer operating at 500 MHz. HR-ESI-MS was measured on a JMS-700 MStation mass spectrometer. Optical rotations ([α]_D) were determined on a JASCO DIP-370 polarimeter using a 100 mm glass cell. IR spectra (KBr) were majored on a Bruker Equinox 55 FT-IR spectrometer. A Yamazen MPLC was used for purification and isolation with an ODS-80 MPLC column. Open column chromatography was performed using silica-gel (Kieselgel 60, 70–230 mesh and 230–400 mesh, Merck) or reverse phase silica-gel (LiChroprep® RP-18, 40–63 µm, MERCK). TLC tests were performed on Merck pre-coated Silica-gel 60 F₂₅₄ and/or RP-18 F_254s plates (0.25 mm), and compounds were observed by spraying the dried plates with 10% H₂SO₄, followed by heating.

Plant Materials

The rhizome of PT was collected at Jeongbong, Kangwondo, Korea, in June 2010 and identified by Adjunct professor Nam-Jae Kim, College of Pharmacy, Kyung Hee University (Seoul, Korea). A voucher specimen (KHUP1005021) was deposited at College of Pharmacy, Kyung Hee University, Korea.

Isolation of 9 Constituents from the Rhizome of PT

PT (10 kg) was extracted with 80% MeOH (×3) at room temperature for one month, filtered, and concentrated to give a MeOH extract (1.15 kg). The MeOH extract (1.15 kg) was suspended in hot distilled water, and then partitioned with EtOAc (4.0 L×3) (157 g). The EtOAc-soluble fraction (157 g) was chromatographed over a silica gel column (40×15 cm) eluted with a gradient of CHCl₃–MeOH (100 : 1 to 0 : 1, 5 L) to afford 12 sub-fractions (E1–E12). The sub-fractions (E1–E12) were continuously investigated chemical components. The fraction E3 (3.5 g) was passed through a silica gel column (35×5 cm) using an n-hexane–EtOAc mixture as solvent, with a stepwise gradient of (20 : 1 to 5 : 1, 3 L, with a stepwise gradient), then chloroform–acetone (10 : 1 to 0 : 1, 3 L, with a stepwise gradient) to give nine sub-fractions (E3.1–E3.9). The E3.3 (0.51 g) was chromatographed on silica gel column (30×5 cm), eluting by hexane–acetone (1 : 0 to 9 : 1, 3 L) and chloroform–acetone (20 : 1 to 5 : 1, 3 L) with a stepwise gradient then crystallizing in chloroform to obtain compound 1 (20.6 mg). E3.2 (1.3 g) was subjected on silica gel chromatography, eluting with n-hexane–EtOAc and hexane–acetone (10 : 1 to 0 : 1) to give a compound 2 (100.4 mg) (Fig. 1). The fraction E6 (8.5 g) was chromatographied on silica gel column (35×5 cm), eluting with chloroform–acetone (10 : 1 to 5 : 1, 3.5 L, with a stepwise gradient) to fractionate into 15 sub-fractions (E6.1–E6.15). The fraction E6.15 (0.53 g) was applied on silica gel column chromatography, eluting with chloroform–EtOAc (20 : 1 to 5 : 1, 2 L) and chloroform–methanol (9 : 1 to 0 : 1, 2 L) with a stepwise gradient, then it was purified on MPLC (LiChroprep® RP-18, 25×2 cm) and eluted with 5–100% MeOH, 2 L, with a linear gradient) to give out compounds 3 (12.5 mg). The purities of isolated compounds were examined over 95% by 2D-TLC and HPLC techniques.

Chondrillasterol (1): Colorless crystals; mp 155–159°C; ESI-MS (m/z) 411.68 [M−H]⁺; Calcd [C₂₉H₄₇O]⁺ ; IR ν_max (cm⁻¹) 3374.6, 2938.7, and 2867.9, 1641.6, 1457.3, 1385.6, 1036.7, and 881.2.

Poligapolide (2): White crystals; mp 110–115°C; [α]_D²⁰ −12.6 (c=0.04, CHCl₃); HR-ESI-MS (m/z) 267.17304 [M−CH₃]⁺; Calcd [C₁₇H₃₁O₂]⁺; IR ν_max (cm⁻¹) 1745.7 (C=O), 1248.9 (C–CO–O), and 1185.2 (C–O). ¹H-NMR (ppm, C₅D₅N, 500 MHz) δ: 4.50 (m, 1H, H-9), 2.49 (t, J=10.0 Hz, 1H, 2_a-H), and δ 1.76 (dddd, 3H, J=5.0, 10.0 5.0, 10.0 Hz, 3H, 8_a-, 3_a-, 11_a-H), 1.33–1.39 (3H, m, 4_a-, 7_a-, 12_a-H) and 1.23–1.29 (23H, m, 2_b-, 3_b-, 4_b-, 7_b-, 8_b-, 11_b-, 12_b-H and 5-, 6-, 7-, 10-, 13-, 14-, 15-, 16-, 17-H) and 0.84 (t, J=5.0 Hz, 3H, H-16). ¹³C-NMR (ppm, C₅D₅N, 125 MHz) δ: 176.4 (C-1), 76.79 (C-9), 34.9 (C-2), 32.2 (C-8, 16), 30.07 (C-7, 10, 12, 15), 30.00 (C-14), 29.9 (C-5), 29.8 (C-5), 29.7 (C-13), 29.6 (C-4), 25.7 (C-3, 11), 14.8 (C-17, CH₃).

3α-O-β-Pyranoglucosyl Spinasterol (3): Non-color crystals; mp 201–213°C; ESI-MS (m/z) 574.45 [M]⁺; Calcd [C₃₅H₅₈O₆]⁺.

1,2,3-Trihydroxy-6,7-dimethoxy Xanthone (4): Yellow amorphous powder; ESI-MS (m/z) 304.08 [M]⁺; Calcd [C₁₅H₁₂O₇]⁺.

3,4,5-Trimethoxy Methyl Cinamate (5): White amorphous crystals; ESI-MS (m/z) 237.09 [M−CH₃]⁺; Calcd [C₁₂H₁₃O₅]⁺.

3β-O-β-Pyranoglucosyl Chondrillasterol (6): White crystals; mp 204–210°C; ESI-MS (m/z) 556.43 [M−H₂O]⁺; Calcd [C₃₅H₅₆O₅]⁺.

1,2,3,6,7-Pentahydroxy Xanthone (7): Yellow amorphous powder; ESI-MS (m/z) 275.03 [M−H]⁺; Calcd [C₁₃H₇O₇]⁺.

6-Hydroxy-1,2,3,7-tetramethoxy Xanthone (8): Yellow amorphous powder; ESI-MS (m/z) 288.07 [M]⁺; Calcd [C₁₅H₁₂O₆]⁺.

3′,4′-Dimethoxy-7-diglucosyl-O-methylenoxy-5-hydroxyl Flavol (9): Yellow powder; mp 210–215°C; ESI-MS (m/z) 667.58 [M−H]⁺; Calcd [C₃₀H₃₅O₁₇]⁺.

Cells Live/Dead Assay

We used the human fetal microglia cell line CHME5 that expressed full-length Tat (pTat101) or a control gene (pcDNA3.1-Hygro, without Tat).¹⁵⁾ The cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM) with10% fetal bovine serum. For the live/dead assay of Tat expressing CHME5 cells, Tat-expressing CHME5 cells were treated with 50 µg/mL LPS and 10 µg/mL cycloheximide (Sigma Co., St. Louis, MO, U.S.A.). Stress in the presence or absence of test agents for 48 h.¹⁹⁾ Then the cells were trypsinized, mixed with trypan blue solution and the dead cells were calculated. Or the cells were trypsinized, stained with PI (1 µg/mL) for 15 min at room temperature and measured by flow cytometry (C6 Flow Cytometer® System, Ann Arbor, MI, U.S.A.).

Immunoblotting

The immunoblotting of PI3K, p-PI3K, Akt, p-Akt (Santa Cruz Biotechnology, Santa Cruz, CA, U.S.A.), p-PDK1, GSK3β and p-GSK3β (Cell Signaling Technology, Beverly, MA, U.S.A.). was performed according to the previous method of Le et al.¹²⁾

Acknowledgment

This study was supported by a Grant from World Class University Program through the National Research Foundation of Korea funded by the Ministry of Education, Science and Technology (R33-2008-000-10018-0). We also thank Dr. Dae Sik Jang, a professor of Kyung Hee University for structure assignment of PT2.

References

• 1) Deregibus M. C., Cantaluppi V., Doublier S., Brizzi M. F., Deambrosis I., Albini A., Camussi G., J. Biol. Chem., 277, 25195–25202 (2002).
• 2) Noh K. H., Kang T. H., Kim J. H., Pai S. I., Lin K. Y., Hung C. F., Wu T. C., Kim T. W., Mol. Ther., 17, 439–447 (2009).
• 3) Chugh P., Bradel-Tretheway B., Monteiro-Filho C. M., Planelles V., Maggirwar S. B., Dewhurst S., Kim B., Retrovirology, 5, 11 (2008).
• 4) Kim N., Kukkonen S., Gupta S., Aldovini A., PLoS Pathog., 6, e1001103 (2010).
• 5) Chugh P., Fan S., Planelles V., Maggirwar S. B., Dewhurst S., Kim B., J. Mol. Biol., 366, 67–81 (2007).
• 6) Koenig S., Gendelman H. E., Orenstein J. M., Dal Canto M. C., Pezeshkpour G. H., Yungbluth M., Janotta F., Aksamit A., Martin M. A., Fauci A. S., Science, 233, 1089–1093 (1986).
• 7) Cosenza M. A., Zhao M. L., Lee S. C., Neurobiol, 30, 478–490 (2004).
• 8) Irish B. P., Khan Z. K., Jain P., Nonnemacher M. R., Pirrone V., Rahman S., Rajagopalan N., Suchitra J. B., Mostoller K., Wigdahl B., Am. J. Infect. Dis., 5, 231–258 (2009).
• 9) Ferte C., Andre F., Soria J. C., Clin. Oncol., 7, 367–380 (2010).
• 10) Li C., Yang J., Yu S., Chen N., Xue W., Hu J., Zhang D., Planta Med., 74, 133–141 (2008).
• 11) Jiang Y., Zhang W., Tu P., Xu X., J. Nat. Prod., 68, 875–879 (2005).
• 12) Le T. K., Jeong J. J., Kim D. H., Biol. Pharm. Bull., 35, 1379–1383 (2012).
• 13) Nguyen N. T., Nguyen M. H. K., Nguyen H. X., Bui N. K. N., Nguyen M. T. T., J. Nat. Prod., 75, 1951–1955 (2012).
• 14) Evidente A., Lanzetta R., Capasso R., Andolfi A., Vurro M., Zonno M. C., Phytochemistry, 44, 1041–1045 (1996).
• 15) Pavia D. L., Lampman G. M., Kriz G. S., Vyvyan J. R., “Introduction to spectroscopy,” 4th edition, Publisher Books/Cole Cengage Learning, Boston, MA, pp. 445–482 (2009).
• 16) Bollbuck B., Kraft P., Tochtermann W., Tetrahedron, 52, 4581–4592 (1996).
• 17) Jeong J. J., Kim B., Kim D. H., Biol. Pharm. Bull., 36, 1088–1094 (2013).
• 18) Kim Y., Hollenbaugh J. A., Kim D. H., Kim B., PLoS ONE, 6, e21781 (2011).
• 19) Lucas A., Kim Y., Rivera-Pabon O., Chae S., Kim D. H., Kim B., PLoS ONE, 5, 10 (2010).