A new combination MALDI matrix for small molecule analysis: application to imaging mass spectrometry for drugs and metabolites

Abstract

Since the development of matrix-assisted laser desorption/ionization (MALDI) mass spectrometry, this procedure has been specifically used for analyzing proteins or high molecular weight compounds because of the interference of matrix signals in the regions of the low mass range. Recently, scientists have been using a wide range of chemical compounds as matrices that ionize small molecules in a mass spectrometer and overcome the limitations of MALDI mass spectrometry. In this study, we developed a new combination matrix of 3-hydroxycoumarin (3-HC) and 6-aza-2-thiothymine (ATT), which is capable of ionizing small molecules, including drugs and single amino acids. In addition to ionization of small molecules, the combination matrix by itself gives less signals in the low mass region and can be used for performing imaging mass spectrometry (IMS) experiments on tissues, which confirms the vacuum stability of the matrix inside a MALDI chamber. The drug donepezil was mapped in the intact tissue slices of mice simultaneously with a spatial resolution of 150 μm during IMS. IMS analysis clearly showed that intact donepezil was concentrated in the cortical region of the brain at 60 min after oral administration. Our observations and results indicate that the new combination matrix can be used for analyzing small molecules in complex samples using MALDI mass spectrometry.

---

Introduction

Matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS) is a powerful tool for analyzing various types of biomolecules in omics research. This technique has been successfully used throughout different fields of science since its development in the 1980s.¹ Singly charged ions can be detected from the mixture using the mass spectrum obtained from MALDI-MS; this advantage of MALDI-MS has led to a new era of analysis of biological and chemical molecules. The unique ionization platform of MALDI, 2D sample array, and the advances in laser technology have enabled the application of this technique in the field of imaging, which is a breakthrough in the field of MALDI-MS. MALDI-imaging mass spectrometry (IMS) has become a revolutionary method for identifying and mapping spatial arrangement and concentration of biological molecules and molecular complexes directly from tissue sections. IMS plays a crucial role in characterizing biomolecules in cellular biology studies and is simultaneously used for drug discovery because it is a unique method currently used for screening drug targets and mapping drug distribution throughout tissues and organs.^2–4

Selection of the correct matrix, especially for small molecule analysis, is a challenging process. As the low m/z regions of MALDI-IMS gives large interference signals from a large population of matrix-related adduct clusters and endogenous biomolecules from the tissue sample, it is expected to have overlapped signals with the same nominal mass of the target molecule.³ This problem is prominent in low spatial resolution mass spectrometers such as the time-of-flight system. To overcome the problem, several strategies can be attempted including high resolution MALDI-IMS with Fourier transform ion cyclotron (FT-ICR) mass spectrometry,³ the alternative use of a deuterated matrix⁵ or tandem mass spectrometric analysis (MS/MS) of targeted drug molecules prior to IMS analysis or on tissue MS/MS analysis after IMS.^6,7

The use of high resolution mass spectrometers for example FT-ICR or Orbitrap can solve the problem of false-positive identification associated with overlapping peaks.³ Because of their advanced resolving power these MS instruments have the capability of separating target molecules or their metabolites from matrix clusters and endogenous compounds that have similar molecular weight. However as the acquisition time is much slower, it increases the total image acquisition time.⁸ MS/MS analysis can provide structural information of targeted compounds and can identify the target molecule from other biomolecules with the similar nominal mass. For more accurate and specific MALDI-IMS analysis of the target molecule and its metabolites in tissue, a number of strategies based on MS/MS imaging have been described including selected reaction monitoring (SRM) mode⁶ and multiple reactions monitoring (MRM) mode.⁷ In addition, there is a very recent development of MALDI-IMS for the target molecule by reducing interference of the matrix clusters.⁵ The use of deuterated matrix uncovers the masked signals of the target molecules analyzed with the conventional matrix. Change to the deuterated matrix of the corresponding matrix generates a different set of matrix cluster ions which previously masked the target molecules.

However, selection of an appropriate matrix is critical to ensure successful MALDI-IMS. Using commonly known matrices such as 2,5-dihydroxybenzoic acid (DHB), α-cyano-4-hydroxycinnamic acid (CHCA), 9-aminoacridin (9-AA), and 2-mercaptobenzothiazole (2-MBT) for imaging analysis of small molecules have their own particular difficulties. An appropriate matrix has at least 2 major requirements: (1) ensuring a reduced background signal to facilitate identification of analyte peaks from the target samples and (2) sustaining the stability of analyte peaks under vacuum conditions during data acquisition. Therefore, it is essential to develop a matrix that has little noise in the low molecular weight region and is stable in vacuum for successful analysis of small molecules using MALDI-MS and MALDI-IMS.

Thus, binary or so-called combination matrices have recently been suggested to improve matrix efficiency and obtain a good mass spectrum of diverse molecules. For example, a binary matrix of 2,5-DHB and 2-hydroxy-5-methoxybenzoic acid (super DHB) improves the resolution of mass spectra of carbohydrates and glycoproteins.⁹ Another binary matrix of CHCA and 2,5-DHB is used in MALDI-MS for peptide mass mapping and analysis of intact glycoproteins and phospholipids because this matrix provides improved mass spectra.^10,11 In these cases, the efforts were focused on improving the quality of mass spectra of proteins, protein–carbohydrate complexes, or lipids, which have molecular weight >500 Da. The binary matrix of CHCA–9-AA effectively reduced noisy peaks in the low molecular weight region.¹² However, the binary matrix of CHCA–9-AA was only applicable to analytes with pK_a values substantially different from those of matrices. Moreover, the abovementioned binary matrices have not been used for IMS of small molecules directly on real tissue samples. One recent successful report on small molecule identification was stated by Chen et al., where a new matrix N-(1-naphthyl) ethylenediamine dihydrochloride (NEDC) has been applied on rat brain tissue to detect glucose level.¹³

Analysis of small molecules especially different types of drugs using MALDI-IMS has been a major focus of the pharmaceutical industries because MALDI-IMS provides a fast and authentic proof of drug distribution in target organs compared to traditional imaging methods such as magnetic resonance imaging, autoradiography, or fluorescence microscopy. MALDI-IMS analyses provide drug distribution data in the target organ without extra labeling processes and identify unexpected toxicity of the parent drug or its metabolites to non-target organs.⁸ MALDI-IMS can simultaneously distinguish distribution of targeted drugs and their metabolites even in more complex tissue structures for example brain, which is not possible with other traditional imaging techniques. For the successful analysis of MALDI-IMS, at present the major concern of researchers is the choice of suitable matrices. To achieve this goal, this study aimed to develop a new matrix that can overcome the problems in analyzing small molecules mainly drugs directly on tissue.

In this study, we present a new combination matrix of 3-hydroxycoumarin (3-HC)¹⁴ and 6-aza-2-thiothymine (ATT)¹⁵ that is effective in identifying and imaging small biomolecules, including drugs (gefitinib, m/z 447 and donepezil, m/z 380) and amino acids (alanine, m/z 89, threonine, m/z 119, methionine, m/z 149, and tryptophan, m/z 204). All these amino acids and drugs have low molecular weights and are difficult to detect because of matrix ion interference. To determine the optimal matrix conditions, we tested different combinations of 3-HC and ATT, while adding trifluoroacetic acid (TFA) and/or piperidine as additives. The new combination matrix was applied on the brain, kidney, and liver tissue sections of mice treated with donepezil.¹⁶ These results support the potential of this combination matrix as an imaging tool for distribution of drugs and their metabolites through tissues.

Experimental

Materials

The MALDI matrices ATT, 3-HC, 2,5-DHB, CHCA, 2-amino-6-nirobenzothiazole (ANBT), 2-amino-5-nitrothiazole (ANT), 2-marcaptobenzothiazole (MBT), and 9-nitroanthracene (9 NA) were purchased from Bruker Daltonics, Germany. TFA, piperidine and amino acids (Ala, Thr, Met, and Trp) were purchased from Sigma-Aldrich, Steinheim, Germany.

Donepezil and gefitinib were kind gifts from Eisai Korea and AstraZeneca Korea, respectively. All materials were of laboratory grade and were used without purification.

Matrix preparation

Different concentrations of ATT and 3-HC with or without TFA and piperidine were used to detect amino acids and drug molecules in MALDI profiling mode. For example, in Fig. S1,† C-1 represents the condition where the combination matrix was dissolved in 50% acetonitrile with 0.2% TFA and 1% piperidine, C-2 was the condition where 0.1% TFA was added with 50% acetonitrile to dissolve the matrix and in condition C-3, the combination matrix was dissolved in 50% acetonitrile with 0.2% TFA. After several trials, a suitable matrix condition, which provides the highest sample signals with lower background peaks, was selected. Finally, 10 mg mL⁻¹ of both ATT and 3-HC matrices in 50% acetonitrile (ACN) with 0.2% TFA (C-3) was selected as the most preferable matrix condition for both profiling and imaging of small molecules (Fig. S1†).

Animal handling and tissue preparation

Male ICR mice (Orient Co., Ltd, a branch of Charles River Laboratories, Seoul, Korea; 30 g body weight at the beginning of the study) were maintained under a 12 h light–dark cycle and housed under controlled temperature (23 ± 1 °C) and humidity (55% ± 2%). All animal experimental procedures were performed in accordance with the approved guidelines of the institutional animal care and committee of Konkuk University. Animals had access to rat chow and water ad libitum and were used for experiments at 10 weeks of age. Donepezil (50 mg kg⁻¹) dissolved in saline was administered orally. After 1 h, animals (n = 3) were killed and perfused with ice-cold saline for 20 min. The brain, liver, and kidney were rapidly removed, immersed into a tube containing liquid nitrogen, and transferred to a −80 °C refrigerator. Frozen brain, kidney, and liver tissues were sectioned using a cryo-cut microtome (Leica, Nussloch, Germany) at −20 °C. Typically, 12 μm thick sections were made for IMS experiments. Tissue sections were thaw-mounted onto MALDI indium tin oxide (ITO)-coated slides (Bruker Daltonics, Germany). Tissue sections were dehydrated using desiccators for 20 min and stored at −80 °C until use.

MALDI-MS and IMS conditions

To prepare the combination matrix, both matrices were dissolved separately at a concentration of 10 mg mL⁻¹ in 50% ACN plus 0.2% TFA. The final matrix solution of ATT–3-HC was prepared by thoroughly mixing the 3-HC and ATT solutions at volume ratios of 1 : 1 by pipetting and using a vortex meter. The amino acids alanine, methionine, threonine, and tryptophan were diluted to 1 mg mL⁻¹ using HPLC-grade water. 1 mg mL⁻¹ solutions of gefitinib and donepezil were prepared and diluted to 1 ng mL⁻¹ using HPLC-grade water. For profiling, 1 μL of the prepared analyte solution was thoroughly mixed with 1 μL of the matrix solution by pipetting in a 0.5 mL Eppendorf tube. We applied 2 μL of the matrix–sample mixture on the MALDI stainless-steel sample plate. The droplet was allowed to dry at room temperature before analysis.

For imaging, an ImagePrep™ instrument (Bruker Daltonics) was used to spray a total of 2 mL matrix solution on a tissue section. After matrix application, the MTP slide adapter (Bruker Daltonics) mounted with the tissue section ITO slide was directly transferred to the Ultraflex MALDI mass spectrometer (Bruker Daltonics). MS data acquisition was performed by averaging signals from 1000 consecutive laser shots. The MS data were acquired in the m/z range between 0 and 500. Before acquisition, the spectra were calibrated using a calibration standard. The spatial resolution of all the mass images shown in this work was 150 μm. The total spectra obtained were baseline subtracted. MS/MS analysis was performed directly on tissue. All acquired profiling, imaging and MS/MS data were analyzed by Flexanalysis and Fleximaging software (Bruker Daltonics).

Results and discussion

Optimum matrix development and application for analyzing single amino acids

To determine the matrix elements and their compositions, we evaluated the usefulness of each of the chemical candidates as a matrix. Although matrix selection and optimization are still often a trial and error process, it is critical to match matrix and analyte polarity.¹⁷ Mixing of two matrices gives a new matrix with different polarity. Few chemical compounds have been used as a matrix for analysis of small molecules due to interference of matrix peaks in the corresponding m/z regions. To be a successful candidate for small molecule analysis, a matrix needs to have less matrix peaks in the region of low molecular weight and high ionization efficiency for target molecules. To accomplish this, we compared different kinds of single matrices to analyze the matrix peaks in a low molecular weight region up to m/z 500. The candidates for the matrix were 3-HC, 9-NA, ATT, ANBT, ANT, and MBT, which are widely used matrices for small molecule analysis (Fig. 1A and S2†). All these matrices were dissolved in 50% ACN with 0.2% TFA and analyzed using the same MALDI mass spectrometer. 9-NA, ANBT, ANT, and MBT showed numerous matrix peaks in low molecular weight regions, which negatively affect the small molecule analysis under the same MALDI conditions. This result showed that most of the single matrices, although commonly used in MALDI analysis, are not suitable for analytes below m/z 500. However, ATT and 3-HC showed fewer matrix peaks and reproducibility during sample analysis compared to the other candidates (Fig. 1A). Previous studies indicate that ATT is a well-known agent for making homogeneous crystals with good ionization capacity. However, ATT gives a relatively noisier background than 3-HC. Although ATT and 3-HC matrices fulfill the first requirement of giving fewer peaks in the low molecular weight region, however, the ionization efficiency of these agents was not as effective as we expected when we analyzed targeted samples separately using ATT or 3-HC as the matrix (Fig. 1).

Fig. 1 Comparison of MALDI-TOF spectra of matrices (A), donepezil (1 ng per plate) (B) and gefitinib (1 ng per plate) (C) analyzed with ATT, 3-HC and the combination matrix in the low molecular weight region (arrow indicates peaks for [M + H]⁺ of the target molecules).

Because both 3-HC and ATT showed a high potential for small molecule analysis, we examined the effectiveness of the binary matrix of 3-HC and ATT. We combined these 2 elements in a 1 : 1 ratio, each dissolved in 50% ACN with 0.2% TFA at a concentration of 10 mg mL⁻¹ and obtained a noteworthy result with the condition (Fig. S1†). To evaluate the combination matrix, we used four different amino acids (Ala, Met, Thr and Trp) and two chemical drugs (gefitinib and donepezil). Solutions of chemical drugs and amino acids were prepared (ranging from 1 ng mL⁻¹ to 1 mg mL⁻¹) and mixed with the same volume of the matrix solution. The mixture solutions (2 μL each, corresponding to 1 pg to 1 μg on the target plate) were dropped on the MALDI target plate for the profiling experiment. After drying the spots, each spot was tested under the same MALDI condition under positive ionization mode.

Fig. 1A shows the background peaks of single ATT, 3-HC and the combination matrix in the region below m/z 500 from the MALDI spectrum. The combination matrix shows few peaks in the neighboring regions of the mentioned m/z region (Fig. 1A). Small molecules were analyzed using ATT and 3-HC alone and their combination, and the spectrum obtained with the different matrices was compared in terms of intensity and the S/N ratio (Table 1 and Fig. 1). As shown in Fig. S3,† relative intensities of amino acids and drugs were higher than the neighboring background peaks of the combination matrix itself in the case of all samples. Small molecules with the combination matrix were ionized efficiently and showed clear peaks of small molecules at their expected m/z values with the protonated forms, [M + H]⁺. The representative MALDI spectra presented were obtained from 1 ng of analytes with different matrix conditions (Fig. 1B and C). The peak corresponding to 1 ng of donepezil appeared at a higher S/N ratio of 1474 using the combination matrix then using ATT (S/N 330), and 3-HC (S/N 17) (Fig. 1B). For gefitinib, S/N ratios were 7.1 with 3-HC, 77.5 with ATT and 142 with the combination matrix (Fig. 1C). These results showed that the combination of ATT and 3-HC, which we named “combination matrix,” provides optimal efficiency in target ionization and has minimal interference for small molecule analysis.

Table 1 Comparison of intensities and signal-to noise (S/N) ratios of drugs obtained with different matrices

	Amount on the plate (g)	3-HC		ATT		Combination
		Intensity	S/N	Intensity	S/N	Intensity	S/N
Donepezil	1 × 10^−6	656 883.0	1582.7	640 188.6	2645.3	813 412.2	1340.9
	1 × 10^−9	140 648.9	12 415	313 284.2	1369	680 554.2	744.3
	1 × 10^−12	2452.61	194.5	18 970.99	23.1	12 125.25	45.4
Gefitinib	1 × 10^−6	356 631.6	604.5	436 110.7	181.2	71 8492.4	262.2
	1 × 10^−9	143.0	7.1	42 860.5	77.5	60 735.6	142.4
	1 × 10^−12	84.0	33.3	12 551.5	18.9	530.0	29.9

---

Application of the combination matrix for IMS of donepezil

Donepezil has been widely used for the treatment of Alzheimer's disease. Previous studies indicate that intact donepezil easily crosses the blood brain barrier and quickly reaches the neuronal cells after administration. However, metabolites of donepezil were mainly found in other organs like the kidney and liver. On the basis of the distinct signal of donepezil obtained using the combination matrix, we performed IMS on donepezil-treated tissue sections of the brain, kidney, and liver. As a blank control, a tissue section of brain without donepezil treatment was included (Fig. 2B). Our aim was to verify whether our newly developed matrix was capable of identifying the intact drug and its metabolites from a complex sample like the actual tissue. At 60 min after oral administration to healthy male ICR mice, the mice were sacrificed and organs were collected for IMS. The combination matrix sprayed on each tissue section showed enhanced homogenous crystallization, which is an important requirement of the matrix for imaging (Fig. 2A and B).

Fig. 2 Typical IMS images and MALDI spectra from the 200–400 m/z region and images of donepezil on tissue (n = 3). (A) Donepezil distribution in the brain tissue, where the intensity of the drug is higher in the cortex region. (B) Brain images from drug-free mouse as a blank control. (C) Donepezil distribution in the kidney tissue. The drug showed a very scattered distribution in the kidney. (D) Drug distribution in the liver tissue, where the drug is located only on the outer surface of the tissue. IMS was performed with the combination matrix. Optical images of brain sections after matrix crystallization were also shown (B and D).

After oral dosing of donepezil it has been reported that the activity of brain AchE is decreased significantly, albeit to a small extent compared with the AchE activity in plasma.^18,19 The ratio of donepezil in the brain and plasma is reported to be 4.4 : 1 ± 0.3 in young animals and 5.3 : 1 ± 1.8 in old animals, respectively.¹⁹ Although these studies clearly suggest that the infiltration and accumulation of donepezil in the brain are responsible for the decreased AchE activity in the brain, they do not provide the detailed anatomical correlationship of pharmacokinetic behaviors of donepezil in the brain with the observed pharmacodynamic changes in the brain including the inhibition of AchE activity or differential expression of muscarinic acetylcholine receptors.²⁰ Combining the pharmacokinetic methods presented here with the imaging mass spectrophotometric determination of changes in cholinergic components in the brain will provide in depth information on the brain region specific PK/PD parameters, which might be also applicable to other chemicals and therapeutic reagents as well.

The total MS spectra in the m/z range 200–400 obtained from the brain, kidney, and liver are shown in Fig. 2. Because MALDI imaging was performed directly on tissue slices without extra processing, several peaks from different biomolecules appeared in addition to the total MS spectra. However, a distinct signal of donepezil was obtained. Previous profiling studies indicated that the donepezil peak that appeared at m/z 380 was also found on an average spectrum after imaging the entire tissue. The donepezil peak shows a clearer distribution in the brain and kidney tissue sections than on the liver tissue section (Fig. 2). However, on the kidney tissue section, the donepezil peak did not show any significant localization. The donepezil signal was more intense in the cortex region of the brain tissue. This conforms to the fact that most of the beta-amyloid plaques are located in the cortex region for Alzheimer's disease^21,22 and that the drug can effectively reach its target site. Tissue imaging and drug mapping confirmed that donepezil can be successfully delivered to its target organ and is active at 60 min after administration.

The donepezil administrated mouse brain section, sprayed with the conventional 2,5-DHB matrix was also analyzed by IMS (Fig. S4†). However, after the imaging of the whole drug administrated brain section, the donepezil peak did not appear (Fig. S4†). On the other hand, the newly developed combination matrix successfully identified donepezil from three different organs of mouse by IMS.

MS/MS analysis of donepezil on brain tissue

Each tissue section has several different complex biomolecules with similar molecular weights. Thus, detection of specific compounds directly from tissue sections without further processing is difficult. However, the MS technique successfully overcomes such problems. Using tandem MS, it is possible to fragment particular molecules and according to their fragmentation pattern their chemical structure can be confirmed. In the present experiment, we identified donepezil at m/z 380 on the brain tissue section by profiling. To confirm that the signal is only from donepezil and not from other biomolecules present in the drug-treated tissue section, we performed the MS/MS experiment at m/z 380 directly on the brain tissue. Apparent signals were detected at m/z 91, 176, 186, 193, 206, 289, and 366 (Fig. 3). Peaks at m/z 91 and m/z 289 correspond to 2 major fragments of donepezil (Fig. 3, inset), and m/z 176, 186, 193, 206, and 366 were also detected as fragment peaks appeared from donepezil from the MS/MS experiment.¹⁶ The MS/MS spectra of m/z 380 showing a similar peak pattern as reported previously confirmed the spatial distribution of m/z 380 representing donepezil delivery to the brain.

Fig. 3 MS/MS analysis of donepezil directly on the drug-treated tissue section.

Conclusion

Here, we describe a new “combination matrix” that is capable of efficiently ionizing small molecules such as drugs and/or amino acids with minimal interference peaks from the matrix. In addition, our newly developed matrix can sustain for a long period under high vacuum inside a MALDI chamber to perform whole tissue imaging and show the spatial distribution of candidate drugs throughout the target organs. Finally by tandem MS analysis on the tissue, the signal of drug could be clearly identified and distinguished from complex interfering biomolecules in the tissue samples.

Acknowledgements

The research was supported by the Converging Research Center Program (2012K001533) and WCU (World Class University) Program (Project No. R33-10128) through the National Research Foundation funded by the Ministry of Education, Science and Technology, Korea.

References

1. M. Karas and F. Hillenkamp, Anal. Chem., 1988, 60, 2299.
2. G. Marko-Varga, T. E. Fehniger, M. Rezeli, B. Dome, T. Laurell and A. Vegvari, J. Proteomics, 2011, 74, 982.
3. D. S. Cornett, S. L. Frappier and R. M. Caprioli, Anal. Chem., 2008, 80, 5648.
4. S. Khatib-Shahidi, M. Andersson, J. L. Herman, T. A. Gillespie and R. M. Caprioli, Anal. Chem., 2006, 78, 6448.
5. M. Shariatgorji, A. Nilsson, R. J. A. Goodwin, P. Svenningsson, N. Schntu, Z. Banka, L. Kladni, T. Hasko, A. Szabo and P. E. Andren, Anal. Chem., 2012, 84, 7152.
6. G. Hopfgartner, E. Varesio and M. Stoeckli, Rapid Commun. Mass Spectrom., 2009, 23, 733.
7. B. Prideaux, V. Dartois, D. Staab, D. M. Weiner, A. Goh, L. E. Via, C. E. Barry III and M. Stoeckli, Anal. Chem., 2011, 83, 2112.
8. S. Castellino, M. R. Groseclose and D. Wagner, Bioanalysis, 2011, 3, 2427.
9. N. Y. Hsu, W. B. Yang, C. H. Wong, Y. C. Lee, R. T. Lee, Y. S. Wang and C. H. Chen, Rapid Commun. Mass Spectrom., 2007, 21, 2137.
10. S. Laugesen and P. Roepstorff, J. Am. Soc. Mass Spectrom., 2003, 14, 992.
11. S. R. Shanta, L. H. Zhou, Y. S. Park, Y. H. Kim, Y. Kim and K. P. Kim, Anal. Chem., 2011, 83, 1252.
12. Z. Guo and L. He, Anal. Bioanal. Chem., 2007, 387, 1939.
13. R. Chen, W. Xu, C. Xiong, X. Zhou, X. Xiong, Z. Nie, L. Mao, Y. Chen and H. Chang, Anal. Chem., 2012, 84, 465.
14. Z. Zhang, L. Zhou, S. Zhao, H. Deng and Q. Deng, J. Am. Soc. Mass Spectrom., 2006, 17, 1665.
15. P. Lecchi, H. M. Le and L. K. Pannell, Nucleic Acids Res., 1995, 23, 1276.
16. K. Matsui, M. Mishima, Y. Nagai, T. Yuzuriha and T. Yoshimura, Drug Metab. Dispos., 1999, 27, 1406.
17. M. W. F. Nielen, Mass Spectrom. Rev., 1999, 18, 309.
18. T. Kosasa, Y. Kuriya, K. Matsui and Y. Yamanishi, Eur. J. Pharmacol., 1999, 386, 7.
19. C. W. Goh, C. C. Aw, J. H. Lee, C. P. Chen and E. R. Browne, Drug Metab. Dispos., 2011, 39, 402.
20. M. Mesulam, Learn. Mem., 2004, 11, 43.
21. C. Priller, T. Bauer, G. Mitteregger, B. Krebs, H. A. Kretzschmar and J. Herms, J. Neurosci., 2006, 26, 7212.
22. P. R. Turner, K. O'Connor, W. P. Tate and W. C. Abraham, Prog. Neurobiol., 2003, 70, 1.

---

Footnotes

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c2an35782h

‡ These authors contributed equally.

---