Segregation of Molecules in Lipid Bilayer Spreading through Metal

Dec 18, 2008 - Hideki Nabika,† Naozumi Iijima,† Baku Takimoto,† Kosei Ueno,‡,§ Hiroaki Misawa,‡ and. Kei Murakoshi*,†. Division of Chemis...
0 downloads 0 Views 2MB Size
Download PDF
Anal. Chem. 2009, 81, 699–704

Segregation of Molecules in Lipid Bilayer Spreading through Metal Nanogates Hideki Nabika,† Naozumi Iijima,† Baku Takimoto,† Kosei Ueno,‡,§ Hiroaki Misawa,‡ and Kei Murakoshi*,† Division of Chemistry, Graduate School of Science, Hokkaido University, Sapporo, 060-0810, Japan, Research Institute for Electronic Science, Hokkaido University, Sapporo 001-0021, Japan, and PRESTO, Japan Science and Technology Agency, Kawaguchi 332-0012, Japan A new methodology for nanoscopic molecular filtering was developed using a substrate with a periodic array of metallic nanogates with various widths between 75 and 500 nm. A self-spreading lipid bilayer was employed as the molecular transport and filtering medium. Dye-labeled molecules doped in the self-spreading lipid bilayer were filtered after the spreading less than a few tens of micrometers on the nanogate array. Quantitative analysis of the spreading dynamics suggests that the filtering effect originates from the formation of the chemical potential barrier at the nanogate region, which is believed to be due to structural change such as compression imposed on the spreading lipid bilayer at the gate. A highly localized chemical potential barrier affects the ability of the doped dye-labeled molecules to penetrate the gate. The use of the self-spreading lipid bilayer allows molecular transportation without the use of any external field such as an electric field as is used in electrophoresis. The present system could be applied micro- and nanoscopic device technologies as it provides a completely nonbiased filtering methodology. Molecular manipulation at the nanometer scale is one of the key technological goals for ultrasensitive purification and separation of a small number of molecules in ultrasmall devices.1-5 Particularly, the use of a lipid membrane as a manipulation medium has attracted diverse attention due to the potential for manipulating biomaterials in their native environments.6,7 Recent activity has focused on electrophoretic manipulation of materials on lipid bilayers, such as lipids,8 vesicles,9 and proteins.10 In addition to the electrophoretic molecular manipulation, electro* To whom correspondence should be addressed. Phone: +81-11-706-2704. Fax: +81-11-706-4810. E-mail: [email protected]. † Division of Chemistry, Hokkaido University. ‡ Research Institute for Electronic Science, Hokkaido University. § PRESTO. (1) Pamme, N. Lab Chip 2007, 7, 1644. (2) Fu, J.; Schoch, R. B.; Stevens, A. L.; Tannenbaum, S. R.; Han, J. Nat. Nanotechnol. 2007, 2, 121. (3) Schoch, R. B.; Bertsch, A.; Renaud, P. Nano Lett. 2006, 6, 543. (4) Fu, J.; Yoo, J.; Han, J. Phys. Rev. Lett. 2006, 97, 018103. (5) Han, J.; Craighead, H. G. Science 2000, 288, 1026. (6) Sackmann, E. Science 1996, 271, 43. (7) Castellana, E. T.; Cremer, P. S. Surf. Sci. Rep. 2006, 61, 429. (8) Kam, L.; Boxer, S. G. Langmuir 2003, 19, 1624. (9) Yoshina-Ishii, C.; Boxer, S. G. Langmuir 2006, 22, 2384. (10) Groves, J. T.; Wu ¨ lfing, C.; Boxer, S. G. Biophys. J. 1996, 71, 2716. 10.1021/ac802130e CCC: $40.75  2009 American Chemical Society Published on Web 12/18/2008

osmotic flow techniques have been investigated on bilayers that contain charged molecules.9 Minute differences between the electrophoretic and electro-osmotic flows control comprehensive molecular transportation and thus molecular separation ability. Furthermore, intermixing between molecules during the electrophoresis has been successfully reduced by the addition of cholesterol to the bilayer.11 Despite these growing developments, the electrophoresis method imposes severe restrictions in that only charged molecules can be manipulated. To overcome this limitation, a self-spreading lipid bilayer can be used as an alternative tool for molecular transportation.12-16 Lipid bilayer self-spreading is a thermodynamically driven collective molecular flow, in which there is no need to apply a bias to induce the molecular flow. The self-spreading bilayer can transport any molecule, even noncharged molecules, in any direction. The driving energy of the self-spreading originates from exothermic free energy change at the formation of the bilayer-substrate contact.12,13 The spreading dynamics are, therefore, controllable via the controlling of bilayer-substrate interactions.17 On the basis of this principle, the lipid bilayer can spread on a substrate with a grating structure or a microfluidic channel.18-20 Further development of the molecular separation or filtering ability on the self-spreading bilayer would open up a new and completely biasfree microfluidic molecular manipulation system. We have previously reported the molecular filtering ability of the self-spreading lipid bilayer using a substrate with a periodic array of metallic nanogates.21 It was found that during the selfspreading of the lipid bilayer, the dye-labeled molecules doped into the bilayer were filtered depending on the gate width. The two-dimensional elastic nature of the lipid bilayer suggests that (11) Daniel, S.; Diaz, A. J.; Martinez, K. M.; Bench, B. J.; Albertorio, F.; Cremer, P. S. J. Am. Chem. Soc. 2007, 129, 8072. (12) Ra¨dler, J.; Strey, H.; Sackmann, E. Langmuir 1995, 11, 4539. (13) Nissen, J.; Gritsch, S.; Wiegand, G.; Ra¨dler, J. O. Eur. Phys. J. B 1999, 10, 335. (14) Nissen, J.; Jacovs, K.; Radler, J. O. Phys. Rev. Lett. 2001, 86, 1904. (15) Suzuki, K.; Masuhara, H. Langmuir 2005, 21, 537. (16) Nabika, H.; Fukasawa, A.; Murakoshi, K. Langmuir 2006, 22, 10927. (17) Nabika, H.; Fukasawa, A.; Murakoshi, K. Phys. Chem. Chem. Phys. 2008, 10, 2243. (18) Furukawa, K.; Sumitomo, K.; Nakashima, H.; Kashimura, Y.; Torimitsu, K. Langmuir 2007, 23, 367. (19) Furukawa, K.; Nakashima, H.; Kashimura, Y.; Torimitsu, K. Lab Chip 2006, 6, 1001. (20) Suzuki, K.; Masuhara, H. Langmuir 2005, 21, 6487. (21) Nabika, H.; Sasaki, A.; Takimoto, B.; Sawai, H.; He, S.; Murakoshi, K. J. Am. Chem. Soc. 2005, 127, 16786.

Analytical Chemistry, Vol. 81, No. 2, January 15, 2009

699

Figure 1. (a) Bright field image of the nanogate channel substrate. (b) AFM images of each gate channel. The perpendicular spacing between two metallic architectures is defined as dgate. (c) Enlarged and three-dimensional AFM image of the squared area in (b). White square in (c) indicates the narrowest region in the nanogate with the depth of ca. 100 nm.

the filtering phenomena are a result of the formation of a compressed state at the gate during spreading. The compression induces a local chemical potential barrier at the gate, which leads to a reduction in the penetration ability of the doped molecules through the gate. This concept of the molecular filtering effect was recently confirmed by use of a single nanogate.22 The mechanism is somewhat similar to reported DNA filtering systems using narrow gates comparable to the size of DNA. In this case the energetic barrier at the gate controls the molecular transportation dynamics and molecular filtering ability.2,4,5 An advantage to using the self-spreading bilayer and nanogate system for filtration is that the filtering ability is based on the chemical potential barrier that is induced at the narrow gate. The system can be extended to recognize and separate molecules of varying physical properties such as size, configuration, charge, polarity, hydrophobicity, chirality, and so on. However, any quantitative discussion leading to an understanding of the mechanism for the filtering effect has not yet been investigated. In the present paper, we have fabricated microchannels with a periodic array of nanogates having a well-defined width at a resolution of a few nanometers. This width has been verified using electron beam lithography. Spreading a lipid bilayer on these substrates enables us to quantitatively discuss the gate width dependence of the spreading dynamics, molecular filtering efficiency, and inherent thermodynamic aspects. The observed gate width dependences were found to be closely correlated with each other. The formation of a locally compressed phase at the gate, which would be still in fluid phase, was suggested as the origin of the observed change in the spreading dynamics and molecular filtering phenomenon. MATERIALS AND METHODS A bright field image of the nanogate channel substrate is shown in Figure 1a. Periodic arrays of metallic nanogates were fabricated on cleaned coverslips (Matsunami Co., Japan) by highresolution electron beam lithography (ELS-7700H, Elionix Co., Ltd. Japan). After standard development (Zeon Co., Ltd., Japan), a 2 nm chromium and 30 nm gold bilayer were deposited using a (22) Kashimura, Y.; Durao, J.; Furukawa, K.; Torimitsu, K. Jpn. J. Appl. Phys. 2008, 47, 3248.

700

Analytical Chemistry, Vol. 81, No. 2, January 15, 2009

sputtering technique (ULVAC, MPS-4000, Japan). The substrate was composed of one control channel without gates and four channels having gates with different gate widths. Atomic force microscopy (AFM) images of each gate channel are shown in Figure 1b. Each gold architecture is a rhombus with 500 nm × 1000 nm dimension. Head-to-head distance between two architectures along the spreading direction is 1000 nm in each channel. The perpendicular distance is defined as the gate width (dgate) as indicated in Figure 1b. The dgate was tuned to be 500, 200, 100, and 75 nm. According to the width of each dgate, the channels were labeled CH500, CH200, CH100, and CH75, respectively. The three-dimensional image of a couple of the architectures are shown in Figure 1c. Judging from the AFM images, the distance of the narrowest gate region is assumed to be ca. 100 nm, which is depicted by the white squares in Figure 1c. L-R-Phosphatidylcholine from egg yolk (egg-PC, Sigma-Aldrich) was used as the lipid for the spreading bilayer. As the target molecules for the filtering, three dye-labeled lipids, Texas Red 1, 2-dihexadecanoyl-sn-glycero-3-phosphoethanolamine, triethylammonium salt (TR-PE, Molecular Probes), N-(7-nitrobenz-2-oxa-1,3diazol-4-yl)-1,2-dihexadecanoyl-sn-glycero-3-phosphoethanolamine, triethylammonium salt (NBD-PE, Molecular Probes), and 2-(12-(7-nitrobenz-2-oxa-1,3-diazol-4-yl)amino)dodecanoyl-1-hexadecanoyl-sn-glycero-3-phosphocholine (NBD-PC, Molecular Probes) were used. The structures of the three dye-labeled lipids are shown in Figure 2. These lipid molecules were used without further purification. Water used in all of the experiments was purified by a Milli-Q system. First, 1 mg/mL chloroform solution of egg-PC and 0.1 mM chloroform solutions of each dye-labeled lipid were prepared as stock solutions. TR-PE solution was added to egg-PC solution so that the final molecular fraction becomes 99.9 mol % egg-PC and 0.1 mol % TR-PE. Similarly, mixed solutions each containing 1 mol % NBD-PE or NBD-PE were prepared. A small amount of mixed solution was deposited at the entrance of the channels on the substrate. After evaporating the chloroform, the substrate was immersed into 100 mM Na2SO4 aqueous solution. Just after the immersion, the lipid bilayer spontaneously spread from the lipid aggregate into the guide of the channel. The spreading behavior was monitored at room temperature by use of an epi-fluorescence microscope (BX-51,

Figure 2. Chemical structures of three dye-labeled lipids used in the present study: (a) NBD-PC, (b) TR-PE, and (c) NBD-PE.

Figure 3. Snapshots of the fluorescence microscope image on the edge region of the self-spreading lipid bilayer on (a) control, (b) CH500, (c) CH200, (d) CH100, and (e) CH75 channels.

Olympus, Japan) equipped with a water immersion objective lens (×40, NA ) 0.80). A high-pressure mercury lamp was used as light source. To avoid significant photobleaching, the light intensity was attenuated to 1 mW using ND filters. The fluorescence images were acquired by an ORCA-ER CCD camera (Hamamatsu Photonics, Japan). RESULTS AND DISCUSSION Fluorescence microscopy revealed that a single bilayer of the lipid molecules spread into all of the gate channels, as shown in Figure 3. The bilayer spread from left to right in each channel. In the control channel, the spreading bilayer was confirmed by the image of a continuous bright layer, compared to the characteristic dark spots in the gate channels. Judging from the periodicity and configuration of the dark spots, the observed dark spots correspond to the Au architectures. Since the Au architecture acts as an effective barrier for lipid diffusion, the bilayer cannot spread over the architectures.23 The observed image using fluorescence microscopy proves that the bilayer spreads only through the metallic gate. Successive observation of the fluorescence microscope images revealed that the rate of the spreading became slower as the width of the gate became narrower (Figure 4). In addition to the effect of gate width on the rate, the molecular distribution of the dyelabeled lipid in the bilayer was also altered by the presence of (23) Groves, J. T.; Ulman, N.; Boxer, S. G. Science 1997, 275, 651.

the gates. In the case of the NBD-PC system, the fluorescence intensity was drastically reduced as the lipid spread through the gate. The decrease in the fluorescence intensity in the gate channel implies that the observed filtering effect of the dye-labeled molecules is due to the bilayer spreading through the nanogate. Although a decrease in the spreading rate for the NBD-PE and TR-PE systems was apparent, the change in the fluorescence intensity was not significant compared with that of NBD-PC system. These differences in the fluorescence changes of each dye-labeled molecule could be attributed to a molecular-dependent filtering effect of the spreading bilayer through the nanogate. The dye dependency becomes more evident by comparing intensity line profiles within a set distance after the spreading occurs (Figure 5, left).24 Line profiles in each system were normalized to the intensity of the control channel at the spreading edge (position ) 0). The filtering effect was not apparent in NBDPE, whereas the TR-PE and NBD-PC systems demonstrated a dgatedependent filtering effect. With decreasing dgate, the decrease in the intensity becomes more significant. The intensity change was also found to be dependent on the spreading distance in the gate region. Figure 5, right, shows the changes in the relative intensity at the spreading edge as a function of the spreading distance in the gate region. The NBD-PE system does not show any change in the fluorescence intensity, suggesting that no filtering effect appeared. Both of the TR-PE and NBD-PC systems, however, showed that the fluorescence intensity decreased as the spreading distance increased. For the TR-PE system, no obvious change was found for less than 50 µm spreading. A decrease in the intensity became apparent after the spreading for 100 µm. On the other hand, a significant decrease appeared at the initial spreading for 50 µm in the NBD-PC system. To evaluate the molecular dependency on filtering, the filtering efficiency per gate, ε, was estimated by dividing the percent decrease in the edge intensity by the number of gate as a function of dgate for respective dye molecules as shown in Figure 6. The value of ε was averaged over the spreading distance where the filtering effect was apparent, i.e., 50-200 µm for TR-PE and 0-50 µm for NBD-PC, respectively. For the NBD-PE system, the value of ε was averaged over whole spreading distance region (0-200 µm). The results show that ε was higher in order of NBD-PC > TR-PE . NBD-PE at all of the dgate channels. Also, it was found that ε increases with decreasing dgate. The distinct molecular dependence supports the conclusion that there is molecular selectivity in the present system. To gain insight into the observed filtering phenomena, a quantitative analysis on the change in the spreading dynamics was carried out. The spreading distance, which was obtained by (24) For quantitative intensity analysis, at least 50 line profiles were averaged on each channel. The smoothing process, by which three data points are averaged with their neighbors, was done for each averaged profile. After the averaging and smoothing procedures, three additional calibrations were performed. The first of these was a background calibration, in which the background intensity was simply subtracted from the profile. The second was an area calibration, in which an underestimation in the fluorescence intensity due to the presence of nonwettable gold architecture, which is involved during the averaging processes, was calibrated by dividing the profile by an area fraction of bare glass region. The area fractions were 0.833 (CH500), 0.792 (CH200), 0.773 (CH100), and 0.767 (CH75). The last was a calibration for photobleaching. The intensity lowering due to the photobleaching is time-dependent. The line profiles were calibrated by the calibration curve between fluorescence intensity and photoirradiation time that was acquired using of supported membrane.

Analytical Chemistry, Vol. 81, No. 2, January 15, 2009

701

Figure 4. Three-dimensional snapshots of fluorescence microscope images. Intensity of the fluorescence is shown as blue to red colors, indicating strength.

Figure 6. Gate width dependence of the filtering efficiency per gate: O, NBD-PC; 4, TR-PE; ×, NBD-PE. The dashed lines are guides for the eyes.

shown. The effect becomes more pronounced with decreasing dgate. Double-logarithmic plots of the spreading velocity (v) against t are shown in Figure 7b. The dynamics of the spreading lipid bilayer are expressed by kinetic spreading coefficient β.12,13 Figure 5. Left: fluorescence intensity line profiles of the bilayer after spreading for 150 µm on each channel. The spreading edge is defined as position ) 0. Right: change in the relative fluorescence intensity at the spreading edge against the spreading distance on the gate region. Dotted line, control (left panel only); solid black line, CH500; blue, CH200; red, CH100; light green, CH75.

the position of the spreading edge, was plotted against the spreading time, t, for each respective channel for NBD-PC in Figure 7a. Calibration to prevent overestimation of the spreading distance due to the presence of nonwettable gold architectures was performed by multiplying the distance by the area fraction of the bare glass region used in the line profile calibration. A retardation of the spreading based on the size of dgate was clearly 702

Analytical Chemistry, Vol. 81, No. 2, January 15, 2009

1 1 log v ) log β - log t 2 2 β)

Edh η

(1)

(2)

where E, dh, and η are the spreading driving energy, thickness, and viscosity of the water layer between the spreading bilayer and the substrate, respectively. Solid lines shown in Figure 7b are best fits to eq 1. Using dh ) 2 nm and η ) 10-3 kg/ms2,12,13 we obtained β ) 40-50 µm2/s for the bilayer spreading in the control channels. It should be noted here that the viscosity of pure water would be changed at nanochannels. However, it

Figure 7. (a) Self-spreading distance and (b) double-logarithmic plot of the spreading velocity of the egg-PC/NBD-PC bilayer: black, control; blue, CH500; light blue, CH200; green, CH100; orange, CH75. (c) The spreading coefficient (upper) and the energy dissipated at a single gate (lower) as a function of the gate width: O, NBD-PC; 4, TR-PE; ×, NBD-PE.

has been reported that the viscosity was almost unchanged in the presence of an electrolyte; thus, we assumed the value of bulk water in the above calculation.25 In all of the gate channels, the double-logarithmic plot shows the decrease in the intercept component while maintaining a constant slope of -0.5. This suggests that eq 1 is valid even when the bilayer spreads through the gates and allows for a quantitative analysis of the gate width dependence on the spreading dynamics. In turn, we can attribute changes in the spreading dynamics to a gate-induced modulation of β. As expected from the double-logarithmic plots, β shows a gradual decrease with decreasing dgate for all of the systems (upper panel in Figure 7c). The change in β is attributable to the change in E by assuming constant dh and η. The relationship between the decreases in β (∆β ) βcontrol - βgate) and E (∆E ) Econtrol - Egate) is then expressed as follows.

∆β )

∆Edh η

(3)

Because the decrease in E originates from the presence of the nanogate, dividing ∆E by the number of gates yields the dissipation energy per single gate, which is defined as Egate. The results are displayed in the lower panel in Figure 7c. The values for Egate show a gradual increase with the decrease in dgate. This result implies that a higher energetic dissipation is imposed at narrower gate sizes. Contrary to the filtering effect on the dye molecules (Figure 6), the retardation effect on the spreading rate does not show any apparent dependence on the types of molecules incorporated in the lipid bilayer. The present result suggests that the dominant component of the spreading bilayer, egg-PC in the present study, determines the energetic dissipation at the nanogate. If we assume that the length of the narrowest region at the gate is ca. 100 nm as depicted in Figure 1c, it is estimated that ca. 7.5 × 103 molecules are present at the narrowest region for dgate ) 75 nm gate, for example.26 Since Egate seems to arise from only this narrowest region, the energy dissipation by the (25) Tas, N. R.; Haneveld, J.; Jansen, H. V.; Elwenspoek, M.; van den Berg, A. Appl. Phys. Lett. 2004, 85, 3274.

presence of the nanogate may be due to a local increase in the free energy ∆Ggate of the bilayer at its respective single gate. Using values of Egate ) 2 × 10-17 J and the number of molecule (7.5 × 103 molecules ) 1.2 × 10-20 mol) for dgate ) 75 nm gate, the value of ∆Ggate is estimated to be 1.7 kJ/mol. This value is comparable to the reported ∆G at the compression for fluid phase lipid monolayer (1-4 kJ/mol) on an aqueous phase containing only monovalent cations.27 Agreement of ∆Ggate in the present system with the previously documented value supports our proposal that the bilayer is compressed at the gate region during the spreading process through the nanogates. Structural deformation or compression is frequently observed when macromolecules or molecular assemblies are used to penetrate into a narrow channel.2,4,5 This structural deformation imposes an energetic dissipation that alters the transportation dynamics, which is observed as the retardation of the spreading in the present system. The molecular filtering effect could be attributed to a mechanism similar to the partitioning phenomena of doped molecules in the lipid bilayer vesicles or supported membranes. Generally, doped molecules such as dye-labeled lipids exhibit inhomogeneous distribution in the lipid bilayer, when the bilayer has more than two coexisting phases.28 Segregation is caused by chemical potential difference between phases with different density and/ or the packing structure of lipid molecules. In the present system, compression of the bilayer at the nanogap results in the formation of a compressed phase which is highly localized at the gate region. As a general tendency, relatively bulky dye-labeled lipids are likely to be excluded from these dense phases. Thus, the formation of the compressed ultrasmall domain at the gate brings a smaller partitioning coefficient relative to that of a noncompressed phase. If the doped dye-labeled lipid molecules are to be excluded from the compressed phase, apparent penetration ability of these molecules through the gates becomes small. The decrease in the penetration ability is observed as the molecular filtering effect of the bilayer spreading in the gate channels. The partitioning characteristics are critically dependent on the nature of the doped molecule. A previous report has clearly demonstrated the difference in the partitioning characteristics between NBD-PE and TRPE.29 The more bulky TR-PE molecule was found to be excluded more effectively from a densely packed phase than NBD-PE. This tendency agrees well with our observation that TR-PE is filtered, whereas NBD-PE can pass through the gate readily. The higher filtering efficiency observed for NBD-PC is due to a formation of a complex looping structure so that the polar NBD group is situated near the headgroup.30-32 The looping structure makes the effective molecular volume larger, which results in a highly exclusive interaction with surrounding lipids in the compressed phase. By comparing between NBD-PE and NBD-PC, it was found (26) The area of the narrowest region is dgate nm × 100 nm. If the molecular area is assumed to be 1 nm2/molecule, this area can accommodate dgate × 100 molecules. (27) Vodyanoy, V.; Bluestone, G. L.; Longmuir, K. J. Biochim. Biophys. Acta 1990, 1047, 284. (28) Vaz, W. L. C.; Melo, E. J. Fluores. 2001, 11, 255. (29) Dietrich, C.; Bagatolli, L. A.; Volovyk, Z. N.; Thompson, N. L.; Levi, M.; Jacobson, K.; Gratton, E. Biophys. J. 2001, 80, 1417. (30) Loura, L. M. S.; Fernandes, F.; Fernandes, A. C.; Ramalho, J. P. P. Biochim. Biophys. Acta 2008, 1778, 491. (31) Loura, L. M. S.; Ramalho, J. P. P. Biochim. Biophys. Acta 2007, 1768, 467. (32) Tsukanova, V.; Grainger, D. W.; Salesse, C. Langmuir 2002, 18, 5539.

Analytical Chemistry, Vol. 81, No. 2, January 15, 2009

703

Figure 8. Filtering efficiency as a function of Egate: O, NBD-PC; 4, TR-PE; ×, NBD-PE. The dashed lines are guides for the eyes.

that the difference in the position of the NBD moiety results in a significant difference of the filtering effects. Since egg-PC was used in the present study, it is possible that specific lipid components in egg-PC would also be filtered at the nanogate. However, judging from that the structural difference among the components is relatively small, the filtering effect would be negligible compared to the doped dye-labeled lipid. Another distinguished feature in the present molecular filtering system is a dependency of the filtering ability on the self-spreading distance, as shown in Figure 5, right. There could be two important factors which contribute to the dependence of the spreading distance on the filtering efficiency. First is the distancedependent spreading rate. As previously described in Figure 7, the spreading rate decreases with time. It can be assumed that the compression effect will not be induced when a static lipid bilayer is situated around the gates. In other words, the compression is induced by the dynamic flow of the self-spreading lipid bilayer into a narrow gate. Thus, the compression effect, i.e., the chemical potential barrier that causes the filtering ability, is reduced with decreasing spreading rate and leads to a change in the filtering efficiency. The second possible contribution is a distance-dependent lipid density in the self-spreading bilayer.13 Our filtering model is based on the chemical potential difference between the noncompressed spreading bilayer and compressed lipid domain at the gate. The lipid density in the self-spreading bilayer is expected to have a direct influence on the filtering ability. Although the chemical potential difference is strongly dependent on the doped molecule, the distance that exhibits the maximum chemical potential difference, i.e., the maximum filtering efficiency, should also be dependent on the choice of the target molecule. The quantitative difference in the filtering effects on the molecules is shown in Figure 8. The respective plot of ε as a (33) Daniel, S.; Diaz, A. J.; Martinez, K. M.; Bench, B. J.; Albertorio, F.; Cremer, P. S. J. Am. Chem. Soc. 2007, 129, 8072. (34) Dietrich, C.; Volovyk, Z. N.; Levi, M.; Thompson, N. L.; Jacobson, K. Proc. Natl. Acad. Sci. U.S.A. 2001, 98, 10642. (35) Takamoto, D. Y.; Lipp, M. M.; von Nahmen, A.; Lee, K. Y. C.; Waring, A. J.; Zasadzinski, J. A. Biophys. J. 2001, 81, 153. (36) Kastl, K.; Menke, M.; Luthgens, E.; Faib, S.; Gerke, V.; Janshoff, A.; Steinem, C. ChemBioChem 2006, 7, 106. (37) Menke, M.; Gerke, V.; Steinem, C. Biochemistry 2005, 44, 15296. (38) Wang, T.-Y.; Leventis, R.; Silvius, J. R. Biochemistry 2001, 40, 13031.

704

Analytical Chemistry, Vol. 81, No. 2, January 15, 2009

function of Egate provides a measure of the selectivity of the molecular filtering. By changing Egate via the control of dgate, we can tune the filtering efficiency of respective molecules. Higher selectivity for NBD-PC was observed in this study. The selectivity for NBD-PC was more than 3 times that of TR-PE and an order of magnitude greater than NBD-PE within the studied values of Egate. The highly localized compressed region may cause a strong field gradient of the chemical potential at the boundary to the outer region under normal conditions. Such heterogeneous distribution in the ultrasmall size region results in successful partitioning of molecules. Pronounced differences between NBD-PC and NBD-PE demonstrate that isomerselective separation is achievable in the present filtering system. Although microfluidic separation of TR-lipid isomers has already been reported, the system relies solely on electrophoresis.33 By taking advantage of the present nonbiased system, any type of molecule, including noncharged molecules, can be manipulated. Furthermore, the chemical potential is determined by wide variety of molecular parameters such as size, configuration, hydrophilicity/hydrophobicity, polarity, charge, and chirality. This will allow diverse molecular selectivity and applicability. Of course, controlling the composition of the spreading bilayer is another important parameter to tune the chemical potential difference. The partitioning phenomenon in the lipid bilayer is not limited to dye-labeled lipids and is readily applicable for any other molecules such as proteins or peptides.34-38 All of these insights suggest possible applications of the present system for miniaturized molecular filtering systems that can be utilized without applying an external field. CONCLUSION We have successfully demonstrated a molecular-dependent filtering phenomenon using the self-spreading lipid bilayer as a molecular transporting and filtering medium. An arrayed structure of the nanogates enables a quantitative explanation of the compression effect. Our estimation shows that the formation of a densely packed phase at the gate region contributes to the appearance of both spreading dynamics attenuation and molecular filtering phenomenon. It is conceivable that our methodology could lead to the development of a completely nonbiased means for molecular filtering and separating systems. This novel methodology may have practical applications in microscopic and nanoscopic device technology. ACKNOWLEDGMENT This work was partially supported by the Grants-in-Aid for scientific research 18750001, 16205026, and 17034002, from the Ministry of Education, Science and Culture, Japan. Especially, that on Priority Area “Strong Photon-Molecule Coupling Fields (No. 470)” is acknowledged. Received for review October 8, 2008. Accepted November 29, 2008. AC802130E