Tissue Phantoms Constructed with Hydrophobic ... - ACS Publications

Jul 7, 2010 - We describe a protocol to create tissue phantoms with hydrophobic nanoporous particles. The nanopores of the particles are loaded with ...
2 downloads 0 Views 3MB Size
Anal. Chem. 2010, 82, 6712–6716

Tissue Phantoms Constructed with Hydrophobic Nanoporous Silica Particles Yulia A. Skvortsova,† Reygan M. Freeney, Zhenming Zhong,‡ and Maxwell L. Geng* Department of Chemistry and the Nanoscience and Nanotechnology Institute, The University of Iowa, Iowa City, Iowa 52242 We describe a protocol to create tissue phantoms with hydrophobic nanoporous particles. The nanopores of the particles are loaded with biological molecules at the desired compositions. Tissue phantoms are prepared by immersing dried particles into aqueous biological matrixes. The hydrophobicity of the pore surface prevents the solution from penetrating into the nanopores, thus preserving the designed molecular composition inside the particles. This protocol provides a unique approach to preparing biological systems in small domains, at micrometer and nanometer dimensions, with well-defined boundaries and tailored biological and optical properties. The nanoporous particle approach is easy when compared to the common preparation methods such as with polymers and vesicles as it involves direct loading of the biological molecules into the pores and does not require complex synthetic steps. The method is adaptable, with tunable pore and particle sizes, and robust, with a rigid boundary to protect the designed biological domain. In addition to tissue phantom preparation, this approach is applicable in systems where a well-defined biological domain is desired. There is an increasing interest in developing micrometer and nanometer domains with well-defined biochemical compositions and spectroscopic properties.1-8 These small domains can be used to mimic cell functions in artificial cells and in nanofactories2-6 or to simulate the initial formation of disease in tissue phantoms.1,7,8 They are valuable in the studies of cell functions and in the * Corresponding author. E-mail: [email protected] (M.L.G.). Phone: (319)3353167 (M.L.G.). Fax: (319)335-1270 (M.L.G.). † Current address: Health and Beauty R&D, Access Business Group, 7575 Fulton Street East, 50-1A, Ada, MI 49355. ‡ Current address: Novartis Pharmaceuticals Corporation, One Health Plaza, East Hanover, NJ 07936-1080. (1) Pogue, B. W.; Patterson, M. S. J. Biomed. Opt. 2006, 11, 041102. (2) Leluc, P. R.; et al. Nat. Nanotechnol. 2007, 2, 3–7. (3) Long, M. S.; Jones, C. D.; Helfrich, M. R.; Mangeney-Slavin, L. K.; Keating, C. D Proc. Natl. Acad. Sci. U.S.A. 2005, 102, 5920–5925. (4) Li, Y.; Lipowsky, R.; Dimova, R. J. Am. Chem. Soc. 2008, 130, 12252– 12253. (5) Cans, A.-S.; Andes-Koback, M.; Keating, C. D. J. Am. Chem. Soc. 2008, 130, 7400–7406. (6) Ge, X.; Conley, A. J.; Brandle, J. E.; Truant, R.; Filipe, C. D. M. J. Am. Chem. Soc. 2009, 131, 9094–9099. (7) Gioux, S.; Mazhar, A.; Cuccia, D. J.; Durkin, A. I.; Tromberg, B. J.; Frangioni, J. V. J. Biomed. Opt. 2009, 14, 034045. (8) Liang, X.; Oldenburg, A. L.; Crecca, V.; Chaney, E. J.; Boppart, S. A. Opt. Express 2008, 16, 11052–11065.

6712

Analytical Chemistry, Vol. 82, No. 15, August 1, 2010

development of imaging modalities for noninvasive disease diagnosis as an integrative component of tissue phantoms.9-11 The creation of microcompartments inside artificial cells is used to mimic phase transitions in crowded cellular environments;3-6 microencapsulation of enzymes is envisioned to be the core of creating nanofactories for combating diseases.2 In the development of imaging modalities for diagnostics, microscopic domains are crucial in simulating the earliest formation of diseased tissue in tissue phantoms for testing the spatial resolution of the imaging systems.1,7,8 The biomedical applications impose a few requirements on these small domains: they can be conveniently formed, without the need of extensive chemical synthesis; the biochemical composition and spectroscopic properties can be tuned with ease; a clear boundary can be maintained when the domains are imbedded into an aqueous biological matrix. The last requirement is especially critical in the preparation of tissue phantoms in the field of optical biopsy. The validation of diagnostic imaging methods depends on the demonstration that domains of distinctive spectral properties can be clearly differentiated in three-dimensional imaging. A clear boundary of the domains needs to be maintained in the phantom preparation and during optical imaging. At macroscopic dimensions, effective methods have been developed for the preparation of phantom domains as stacked layers or suspended optical cells.11-14 At microscopic and nanoscopic dimensions, these approaches are not easily implemented. New designs are desired for the facile and flexible preparation of microscopic and nanoscopic domains in a biological matrix. In this work, we create microscopic domains with hydrophobic nanoporous particles. The trapping of molecules inside nanopores and nanotubes has been intensely explored in biomedical applications such as controlled drug delivery and biocatalysis.15-19 These trapping (9) Durkin, A. J.; Jaikumar, S.; Richards-Kortum, R. Appl. Spectrosc. 1993, 47, 2114–2121. (10) Sokolov, K.; Galvan, J.; Myakov, A.; Lacy, A.; Lotan, R.; Richards-Kortum, R. J. Biomed. Opt. 2002, 7, 148–156. (11) Nieman, L.; Myakov, A.; Aaron, J.; Sokolov, K. Appl. Opt. 2004, 43, 1308– 1319. (12) Wagnieres, G.; Cheng, S. G.; Zellweger, M.; Utke, N.; Braichotte, D.; Ballini, J. P.; vandenBergh, H. Phys. Med. Biol. 1997, 42, 1415–1426. (13) Cubeddu, R.; Pifferi, A.; Taroni, P.; Torricelli, A.; Valentini, G. Phys. Med. Biol. 1997, 42, 1971–1979. (14) Manohar, S.; Kharine, A.; van Hespen, J. C. G.; Steenbergen, W.; van Leeuwen, T. G. J. Biomed. Opt. 2004, 9, 1172–1181. (15) Hillebrenner, H.; Buyukserin, F.; Kang, M.; Mota, M. O.; Stewart, J. D.; Martin, C. R. J. Am. Chem. Soc. 2006, 128, 4236. (16) Son, S. J.; Lee, S. B. J. Am. Chem. Soc. 2006, 128, 15974. (17) Yu, J.; Bai, X.; Suh, J.; Lee, S. B.; Sang Jun Son, S. J. J. Am. Chem. Soc. 2009, 131, 15574. 10.1021/ac902442g  2010 American Chemical Society Published on Web 07/07/2010

methods can potentially be adopted for preparation of tissue phantoms. A number of designs enabled the encapsulation of molecules in nanotubes for transportation in drug delivery, capping the opening of nanotubes with covalently linked polymer nanoparticles,15 synthesized gold plugs16 or gold, silver, and polymer caps mechanically hammered into the pore.17 When optimized, these capping approaches can fabricate plugs in a large fraction of the nanotubes and block molecular migration from them, while a small fraction of nanotubes remain open. Alternatively, direct chemical bonding of molecules inside the nanopores, an approach proven very successful in enzyme catalysis,18-20 can be adopted to form nonleaking biological domains with clear boundaries. In this paper, we describe a novel method of molecular encapsulation in nanopores that is simple and does not require any chemical synthesis or mechanical manipulation. We demonstrate that the method achieves encapsulation of all nanopores. EXPERIMENTAL DESIGN The nanoporous silica particles have an average diameter of 10 µm, a nominal pore size of 100 Å, and a surface area of 400 m2/g. The silica surface is covered with an organic monolayer of C18, rendering the pores hydrophobic (LUNA C18, Phenomenex, Torrance, CA). In the wetting studies, three additional types of nanoporous silica particles are utilized (Kromasil C4, C8, and C18, Akzo Nobel/Kromasil, Brewster, NY). Flavin adenine dinucleotide (FAD) was obtained at the highest purity from Sigma (St. Louis, MO). All solutions were prepared using deionized water purified with a Milli-Q system (Millipore, Bedford, MA). HPLC-quality acetonitrile was purchased from Fisher Scientific (Fair Lawn, NJ). Phosphate buffer saline (PBS) was prepared from sodium phosphate, sodium biphosphate, and sodium chloride; pH was adjusted to 7.4 with a sodium hydroxide solution. The images of the beads at each step of the phantom preparation were collected with the stage scanning confocal fluorescence microscope system built in-house and modified from those previously described.21-23 In brief, the excitation light at 454 nm, provided by an air-cooled argon laser (Model: 35-LAP431-220, Melles Griot, Carlsbad, CA), was focused into a diffraction-limited spot within the sample using a 100× oil-immersion objective. This wavelength provides optimum excitation of FAD fluorescence (Figure 1). The three-dimensional movement of the focal point across the C18 silica beads was achieved using a piezoelectric objective stepper and an xy-piezo flexure stage (Physik Instrumente, Germany) with subnanometer resolution. The movements of the stage were controlled with LabView programs written in house. Emitted fluorescence was collected by the same objective and passed through a 455 nm dichroic mirror, while scattered excitation light was reflected by the mirror. Out-of-focus signals were efficiently rejected by a 50 µm confocal pinhole, and the fluorescence signal from the focal plane was (18) El-Zahab, B.; Jia, H. F.; Wang, P. Biotechnol. Bioeng. 2004, 87, 178. (19) Alptekin, O.; Tukel, S. S.; Yildirim, D; Alagoz, D. J. Mol. Catal. B 2009, 58, 124. (20) Obert, R.; Dave, B. C. J. Am. Chem. Soc. 1999, 121, 12192.. (21) Lowry, M. Ph.D. Dissertation, University of Iowa, 2006; Freeney, R.; Lowry, M.; Geng, M. L., unpublished results. (22) Zhong, Z. M.; Lowry, M.; Wang, G. F.; Geng, L. Anal. Chem. 2005, 77, 2303–2310. (23) Zhong, Z. M.; Geng, M. L. Anal. Chem. 2007, 79, 6709–6717.

Figure 1. Fluorescence excitation and emission spectra of flavin adenine dinucleotide (FAD).

directed to an avalanche photodiode detector. The confocal volume achieved by this optical configuration is ∼250 nm in lateral resolution and ∼1 µm in axial resolution.22,24 To further improve the signal-to-noise ratio and effectively reject the excitation light, a 475 nm long-pass filter and a 530 nm band-pass filter with 55 nm bandwidth to isolate the fluorescence emission of FAD (Figure 1) were used in the emission channel. Intensity traces as a function of the stage movements experimentally collected were processed in MatLab (Mathworks Inc., Natick, MA) to construct the images. RESULTS AND DISCUSSION In the design of tissue phantoms, we pursued the goal of creating distinct spectral domains by imbedding hydrophobically modified nanoporous silica particles within a water-based agar matrix. The key principle of this design is the nonwettability of the hydrophobic nanopores by an aqueous solution.21 To create microscopic or nanoscopic biological domains, the solvent used to prepare the biological matrix, typically an aqueous solution, should not be able to diffuse across the boundary in order to maintain a clear, well-defined edge. Such diffusion would extract the biological molecules loaded in the domains out into the matrix, altering the molecular composition inside and, thus, defeating the objective of tissue phantoms. An effective design of tissue phantom can be accomplished through synthesizing an enclosure around the domain that does not allow the solutions to diffuse through. However, such preparation requires time-consuming synthetic steps well designed such that the synthesis conditions will not degrade the biological components. In our design, the particles are preloaded with the desired composition of biological molecules by simply soaking the particles in a solution. The molecules diffuse into the nanopores and are retained in these picoliter (particles of 10 µm diameter) to attoliter (particles of 100 nm diameter) containers. Importantly, the nanopore surface is tailored with an organic monolayer. A clear biological/optical boundary is achieved when the particles are dispersed into the aqueous matrix to make tissue phantoms, because there is no pore wetting,21 thus no molecular diffusion. The recognition that the hydrophobic pores cannot be wetted by an aqueous solution at the nanometer dimension is the driving principle for this design. Uniform silica beads 10 µm in diameter serve as perfect imaging objects for (24) Gao, Y.; Zhong, Z. M.; Geng, L. Appl. Spectrosc. 2007, 61, 956–962.

Analytical Chemistry, Vol. 82, No. 15, August 1, 2010

6713

Figure 3. Representative cross-section profiles of a bead wetted with FAD solution (A) and a bead loaded with FAD and dried (B). These intensity profiles are taken from particles in Figure 2, along the cross sections indicated by the dotted lines.

Figure 2. Confocal fluorescence images of blank silica beads (A), a bead loaded with FAD (B), and beads loaded with FAD and dried (C). The length scale of the images is in µm. Fluorescence intensity is expressed in counts/s.

testing optical methods for cancer imaging. In addition, they provide features of known shapes and dimensions comparable to eukaryotic cells. The hydrophobic silica beads were loaded with an endogenous tissue fluorophore, flavin adenine dinucleotide (FAD), which is one of the major fluorophores responsible for the laser-induced fluorescence of the colon tissue used for cancer detection.25-28 First, the particles were wetted with acetonitrile; this wetting step allows the subsequent loading of biological fluorophores. Then, the beads were separated from the solution by centrifugation and mixed with a 0.5 mM stock solution of FAD in PBS buffer. The particles, now loaded with FAD, were separated from the solution in the second round of centrifugation and dried in the oven at 40 °C overnight to completely remove the solvent from the nanopores. C18 beads dried in this fashion were added into a 1% by mass aqueous solution of agar at 50 °C. Immediate sonication for 2 min followed to disperse the particles in the solution. The resulting suspension was then poured into a mold and allowed to gel at 5 °C for several hours to form the tissue phantom. The presence of fluorescent impurities in nanoporous silica was detected in prior single molecule experiments.22 Repeated prewashing of the beads was necessary for such ultrasensitive experiments. The images of blank C18 beads (Figure 2A) show minimal background signal, about 400 times lower than the intensity of the fluorescence from FAD-loaded particles (Figure 2B). It was concluded that for the tissue phantom preparation the background fluorescence of the particles is negligible. FAD partitions into the nanopores of the beads when the solvent composition is optimal for the wetting of the pores. FAD is readily soluble in phosphate buffer, but the beads are not wetted (25) Schomacker, K. T.; Frisoli, J. K.; Compton, C. C.; Flotte, T. J.; Richter, J. M.; Nishioka, N. S.; Deutsch, T. F. Lasers Surg. Med. 1992, 12, 63–78. (26) Crowell, E.; Wang, G. F.; Cox, J.; Platz, C. P.; Geng, L. Anal. Chem. 2005, 77, 1368–1375. (27) Wang, G. F.; Platz, C. P.; Geng, M. L. Appl. Spectrosc. 2006, 60, 545–550. (28) Li, Y.; de Silva, P.; Xi, L.; van Winkle, A.; Lin, J. J. C.; Ahmed, S.; Geng, M. L. Biomed. Chromatogr. 2008, 22, 1374–1384.

6714

Analytical Chemistry, Vol. 82, No. 15, August 1, 2010

with aqueous solutions due to the nonpolar C18 coating on the silica surface of the nanopores. The loading of FAD into the pores is achieved by prior wetting of the beads with a less-polar solvent, such as acetonitrile or ethanol, and further addition of the fluorophore solution in aqueous buffer. Fluorescence images of the beads soaked in stock FAD solution show a relatively homogeneous distribution of FAD throughout individual beads (Figure 2B). The intensity profiles across silica beads show a relatively even concentration distribution of FAD throughout the particles (Figure 3A). The next step in the phantom preparation is the recovery of the hydrophobic environment of the pores, which is achieved by drying the beads. The drying step causes the dewetting of the pores and redistribution of the fluorophore molecules inside the beads (Figure 2C). Apparently, the solvent evaporation starts at the openings and gradually proceeds into the pores. Naturally water-soluble FAD relocates deeper into the core of the silica particles following the receding solvent front, resulting in a molecular distribution with a higher concentration toward the center of the beads (Figure 3B). It is noteworthy to mention that the fluorescence intensity of FAD in dry particles is about two times lower than the fluorescence signal from wetted FAD-filled beads (Figure 3). There are three possible causes for this intensity drop. First, a larger difference in the refractive indices between the silica and air in the dry nanopores causes more scattering of the laser light; thus, the excitation of the fluorophore inside dry particles is less efficient. Second, FAD, as many other fluorophores, is sensitive to the polarity of its microenvironment, which changes significantly in the process of particle drying. Third, the concentration of FAD in the nanopores after drying could be so high that fluorescence intensity decreases through concentration quenching. The drying conditions are being optimized in our laboratory to maintain a relative even distribution of the fluorophore concentration throughout the particle. To test the long-term stability of these spectral domains, we examined the wetting properties of the hydrophobic nanoporous particles. Particles whose nanopore surface is modified with C18 (Luna C18 and Kromasil C18), C8 (Kromasil C8), and C4 (Kromasil C4) were first loaded with FAD and dried. They were then immersed in an aqueous solution, and the absorbance of the solution was monitored continuously with UV-vis absorption spectroscopy. The entire absorption spectrum of the solution was acquired at each time point; the absorbance at the 450 nm peak of FAD follows the potential releasing kinetics by monitoring FAD

Figure 4. Wetting properties of hydrophobic nanoporous silica. (A) Luna C18 and Kromasil C18, C8, and C4 particles immersed in water. (B) Luna C18 particles: absorption spectrum after releasing in ethanol (top) and immersion in water (bottom).

concentration in the solution (Figure 4). The time series for all four types of particles (Figure 4A) clearly demonstrate that the absorbance at the FAD peak remained negligible in 48 h of immersion. We treated the time series with linear regression to test if there was a definitive increase of the solution absorbance. The fitting had R2 values of 0.540, 0.064, 0.003, and 0.114; two series showed slight positive slopes (2 × 10-7, 8 × 10-9) while the other two had small negative slopes (-5 × 10-8, -3 × 10-8), suggesting that the solution absorbance indeed remained negligible. We questioned if it is possible that FAD was released into the solution but was undetected because the amount of FAD loaded in the particles was beyond the detection sensitivity of the absorption spectrometer. To test this possibility, the particles were vortexed in a 35% ethanol/65% water solvent. Although ethanol is efficient in pore wetting, FAD has poor solubility in ethanol and does not partition into the solvent to be carried out of the nanopores. The mixed solvent effectively wetted the nanopores with ethanol and induced the release of FAD out of the particles by dissolving it in water. The solution was centrifuged to remove the particles, and the absorption spectrum of the solution was acquired. Figure 4B shows the example of Luna C18 particles. Clearly the absorbance of the loaded FAD (top curve) far exceeds the level of absorbance that we observed for the aqueous solution (bottom curve in Figure 4B and all time series in Figure 4A). These experiments further prove that the hydrophobic nanopores are nonwettable by an

aqueous solution, consistent with our earlier observations by confocal fluorescence imaging.20 The nonwettability is the consequence of the strong hydrophobic interactions at the nanometer dimension that renders the existence of a bulk aqueous phase thermodynamically unfavorable.29,30 The fact that all surface modifications examined in our experiments (C4, C8, and C18) prevented the water penetration suggests that even the coverage by the short hydrocarbon chain containing four carbons is effective in creating the biological domains with clear boundaries. These observations are consistent with our earlier confocal imaging results.21 This provides flexibility in phantom preparation in tuning the relative affinity of the biological molecules to the nanopore surface and, thus, affecting the loading. The pore loading was estimated from the absorbance of the releasing solution after pore wetting. The absorbance of FAD released from the Luna C18 particles was 0.051 at 450 nm (Figure 4B). With the measured molar absorptivity of 1.04 × 104 L mol-1 cm-1 and the pore surface area of 426 m2/g, the surface loading of FAD in the nanopores is estimated to be 0.57 nmol/ m2. This represents a low surface concentration in the C18 monolayer and an average intermolecular distance of 52 nm in the essentially two-dimensional space of the C18 layer. Compared to the surface coverage of 3.04 µmol/m2 of C18 ligands on the nanopore surface, there is one FAD molecule among ∼5300 C18 ligands. A 1 nm thickness of the C18 layer31 leads to an FAD concentration of 0.57 mM and, thus, a partition coefficient of 1.1 for FAD between Luna C18 and the PBS loading solution. The analysis of the other three types of particles yielded similar loading characteristics. This loading level also confirms the possibility that during the drying process, FAD can reach beyond millimolar concentrations where quenching of fluorescence occurs, leading to lowered image intensity (Figure 2C). Intensity of the FAD fluorescence can be restored by rewetting the pores of the particles, which proves that FAD was retained in the pores during the drying process. Addition of pure acetonitrile to the hydrophobic FAD-loaded beads effectively solubilizes the particles, but does not change the fluorescence intensity and distribution of FAD, because it is poorly soluble in nonpolar solvents. The subsequent addition of PBS solution to the sample immediately dissolves FAD, restoring its even distribution throughout the beads and recovering the intensity of the fluorescence to the original value. To mimic a real tissue matrix, the beads are embedded in 1% agar gel. Agar gel is one of the preferred matrixes for tissue phantoms due to its compatibility to biological molecules, allowing easy incorporation of cellular constituents.1 Restored hydrophobicity of the nanopores ensures the embedding of the beads in aqueous agar gel without “leakage” of the fluorophore into the surrounding matrix. Cross sections of the tissue phantom were imaged at different depths at 2 µm spacing. These confocal images (Figure 5) show that clear boundaries are maintained for these 10 µm optical domains inside the tissue phantom matrix. Optical slices of the tissue phantom demonstrate spherical boundaries of (29) Lum, K.; Chandler, D.; Weeks, J. D. J. Phys. Chem. B 1999, 103, 4570– 4577. (30) Meyer, E. E.; Rosenberg, K. J.; Israelachvili, J. Proc. Natl. Acad. Sci. U.S.A. 2006, 103, 15739–15746. (31) Brumaru, C.; Geng, M. L.,submitted to Langmuir for publication.

Analytical Chemistry, Vol. 82, No. 15, August 1, 2010

6715

Figure 5. z-sections of the three-dimensional tissue phantom composed of FAD-loaded silica particles dispersed in 1% agar gel. z-slices were taken at 2 µm increments in the axial direction. Relative z positions: (A) 0 µm; (B) 2 µm; (C) 4 µm; (D) 6 µm; (E) 8 µm; (F) 10 µm; (G) 12 µm; (H) 14 µm. (Excitation: 454 nm, step size in x-y directions: 0.2 µm.)

the FAD composed silica particles. The fluorophore molecules adsorbed on the outer surface of the beads can be removed by washing the beads with water. Since the exterior surface of the porous particles is negligible compared to the total surface area, the amount of FAD deposited on the outer surface is minor and this extra washing step was determined to be unessential in the phantom preparation. The hydrophobic silica particles have a natural tendency to aggregate in the aqueous matrix to minimize the total energy of the system. Sonication of the warm agar upon the addition of the beads promoted uniform distribution of the particles in all three dimensions. The particles loaded with endogenous fluorophore, FAD, and imbedded in the agar matrix form a three-dimensional nontoxic tissue phantom. The use of spherical beads to localize the endogenous fluorescent tissue components allows the simulation of the sample inhomogeneities. To prepare complete tissue phantoms, other endogenous fluorophores, such as NADH, will be loaded into the nanopores. Uniform 10 µm silica beads serve as perfect imaging objects for testing optical biopsy methods, providing features of known dimensions and shapes. The designed phantoms are model systems for the analysis of the effects of tissue optical properties, physiological fluorophore compositions, and the optical geometry on diagnostic imaging. Understanding of the microscopic origins of the observed spectral signatures could promote the development of new methodologies for spectral acquisition and analysis and enhance the effectiveness of the diagnostic protocols for cancer detection. Although the emphasis of this work is to create

6716

Analytical Chemistry, Vol. 82, No. 15, August 1, 2010

a microscopic domain with a boundary that the solvent cannot cross, it is possible to allow wetting by tuning the surface density of the hydrophobic chains in the nanopores. The distribution of biological molecules will then be determined by the partition between the two phases, resulting in a microcompartment with desired biological composition. We have developed a new method for creating tissue phantoms containing microscopic or nanoscopic domains with well-defined optical and biological compositions. The nonwettability of the hydrophobic nanopores by aqueous solutions maintains a clear boundary between these domains and the biological matrix that molecules cannot diffuse through. Compared to other methodologies of nanopore encapsulation, this new approach is simple, does not require any synthetic steps, and achieves encapsulation of 100% of the nanopores. ACKNOWLEDGMENT The support by the National Institutes of Health, the National Cancer Institute (Award No. CA100741), and by the National Science Foundation (Award No. 0911691) is gratefully acknowledged. Z.Z. was a graduate fellow of the Analytical Division of the American Chemical Society. We thank Kromasil for providing the silica particles as a gift. Received for review October 27, 2009. Accepted February 26, 2010. AC902442G