9752
Langmuir 2007, 23, 9752-9759
Two-Dimensional Protein Crystals on a Solid Substrate: Effect of Surface Ligand Concentration Chengfei Lou,† Zifu Wang,‡ and Szu-Wen Wang*,† Department of Chemical Engineering and Materials Science, DeVelopmental Biology Center, UniVersity of California, IrVine, California 92697 ReceiVed May 14, 2007. In Final Form: June 26, 2007 Proteins imbedded in solid-supported lipid bilayers can serve as model systems for investigations of cellular membranes and protein behavior on surfaces. We have investigated the self-assembly of streptavidin on mica-supported bilayer membranes. Using fluorescence microscopy and atomic force microscopy, our studies reveal that the concentration of surface ligand influences the molecular packing of the resulting protein arrays, which in turn affects overall crystal morphology. Two-dimensional streptavidin crystals are obtained when the biotinylated lipid density on the substrate reaches 1.5% mole fraction, yielding high-aspect morphologies that comprise primarily of crystals with P1 symmetry. At 3% and above, crystals with C222 symmetry are formed and result in H-shaped and confluent structures. In intermediate densities between 2 and 3%, a coexistence of P1 and C222 crystal forms is observed. The relationship between macroscopic morphology and molecular configuration is similar to previously reported data obtained at the air/water interface. This suggests that, under our experimental conditions, protein interactions with the supporting substrate are less significant for defining self-assembly behavior than interactions between protein molecules. Ligandinhibition and fluorescence recovery after photobleaching were used to elucidate the concentration-dependent mechanism for the divergent crystal forms. We have measured the diffusion coefficient of molecules in P1-forming conditions to be approximately twice that of molecules in C222-forming concentrations, which is consistent with proteins bound to the surface through one and two ligands, respectively. The differential flexibility associated with the binding state is therefore likely to alter the subtle protein interactions involved in crystallization.
Introduction Synthetic systems with proteins attached to solid-supported phospholipid membranes have received wide attention for their potential applications in diverse areas. Since natural biological membranes are intricately complex, these platforms have served as artificial models for dissecting the biophysics of cell interactions with membrane proteins and extracellular matrix components.1 With the ability to pattern solid substrates and finely control the bilayer architecture of these structures, applications in biosensing have been pursued, including those for high-throughput screening in drug discovery efforts.1,2 In these and other sensor applications, the proteins’ interactions with neighboring macromolecules within the membrane and with their supporting substrates can be important to modulate because they can affect the quality of the detected signal.2-4 Such interactions have been utilized to fabricate patterned supramolecular assemblies of hybrid materials. For example, gold nanoparticles and quantum dots have been ordered using self-assembled chaperonin arrays.5 S-layer proteins from different microbial sources have been shown to crystallize into varying configurations,6 and these proteins can be fused with the protein * Corresponding author. Address: Department of Chemical Engineering and Materials Science, University of California, Irvine, 916 Engineering Tower, Irvine, CA 92697-2575. E-mail:
[email protected]. Phone: 949824-2383. Fax: 949-824-2541. † Department of Chemical Engineering and Materials Science. ‡ Developmental Biology Center. (1) Tanaka, M.; Sackmann, E. Nature 2005, 437, 656-663. (2) Cooper, M. A. Nat. ReV. Drug DiscoVery 2002, 1, 515-528. (3) Murphy, L. Curr. Opin. Chem. Biol. 2006, 10, 177-184. (4) Anrather, D.; Smetazko, M.; Saba, M.; Alguel, Y.; Schalkhammer, T. J. Nanosci. Nanotechnol. 2004, 4, 1-22. (5) McMillan, R. A.; Paavola, C. D.; Howard, J.; Chan, S. L.; Zaluzec, N. J.; Trent, J. D. Nat. Mater. 2002, 1, 247-252. (6) Sleytr, U. B.; Egelseer, E. M.; Ilk, N.; Pum, D.; Schuster, B. FEBS J. 2007, 274, 323-334.
streptavidin, which enables functionalization of the surface through a biotin linkage.7 These chimeric proteins enable the creation of ordered arrays for functional biosensor surfaces,6,8 with S-layers proteins being the template upon which a secondary streptavidin array is created. Our investigations show that the streptavidin molecule itself, without an intermediate templating fusion protein, can also self-assemble on a solid surface into defined configurations. This solid-supported protein system not only enables the presentation of biological ligands or nanoparticles in a controlled spatial assembly, but can also serve as a model system for probing the fundamentals of protein-protein interactions in a lipid bilayer. Streptavidin has been shown to form two-dimensional crystalline arrays under favorable conditions, both at the air/water9-11 and solid/liquid interfaces.12,13 The protein’s symmetry and unusually strong binding to biotin also make it useful for orienting ligands in a controlled direction.14 Previous work demonstrated that the molecular configuration of streptavidin arrays at the air/water interface can be manipulated by both environmental conditions and changes to specific amino acid interactions.15,16 (7) Moll, D.; Huber, C.; Schlegel, B.; Pum, D.; Sleytr, U. B.; Sara, M. Proc. Natl. Acad. Sci. U.S.A. 2002, 99, 14646-14651. (8) Huber, C.; Liu, J.; Egelseer, E. M.; Moll, D.; Knoll, W.; Sleytr, U. B.; Sara, M. Small 2006, 2, 142-150. (9) Blankenburg, R.; Meller, P.; Ringsdorf, H.; Salesse, C. Biochemistry 1989, 28, 8214-8221. (10) Darst, S. A.; Ahlers, M.; Meller, P. H.; Kubalek, E. W.; Blankenburg, R.; Ribi, H. O.; Ringsdorf, H.; Kornberg, R. D. Biophys. J. 1991, 59, 387-396. (11) Edwards, T. C.; Koppenol, S.; Frey, W.; Schief, W. R.; Vogel, V.; Stenkamp, R. E.; Stayton, P. S. Langmuir 1998, 14, 4683-4687. (12) Reviakine, I.; Brisson, A. Langmuir 2001, 17, 8293-8299. (13) Calvert, T. L.; Leckband, D. Langmuir 1997, 13, 6737-6745. (14) Wilchek, M.; Bayer, E. A. Methods Enzymol. 1990, 184, 467-469. (15) Wang, S. W.; Robertson, C. R.; Gast, A. P. Langmuir 1999, 15, 15411548. (16) Wang, S. W.; Robertson, C. R.; Gast, A. P. J. Phys. Chem. B 1999, 103, 7751-7761.
10.1021/la701399s CCC: $37.00 © 2007 American Chemical Society Published on Web 08/11/2007
Protein Crystals on a Solid Substrate
Langmuir, Vol. 23, No. 19, 2007 9753
Figure 1. Schematic of protein crystallization protocol. Langmuir-Blodgett deposition is used to form the bilayer membrane, and incubation of streptavidin with this supported membrane yields self-assembled monolayers of protein crystals.
These characteristics impart potentially useful properties for facilitating the creation of novel inorganic-biological hybrid materials. Although crystallization at the air/water interface has been a useful approach for investigating the fundamentals of protein array fabrication, applications toward biosensor devices must incorporate a solid substrate. Prior exploration of streptavidin crystals grown on solid substrates have shown that, like those grown at the air/water interface, dendritic-X-shaped morphologies can be obtained.13 Furthermore, two crystal forms (with P2 and C222 symmetry) have been observed.12 However, the correlation of morphological crystal development with lattice parameters has not been reported. In addition, air/water experiments indicate the existence of a crystal with P1 symmetry that yields needlelike domains; this form had not yet been identified on a solid substrate. In this work, we seek to relate large-scale crystalline morphologies with the molecular configuration of protein arrays grown on solid substrates and to control the spatial arrangement within these arrays. We have focused on the effects of modulating ligand concentration within the membrane bilayer. Using fluorescence microscopy and atomic force microscopy, our results demonstrate that the crystalline phase of the protein array is altered by the biotin concentration on the solid substrates. Although changes in protein-protein interactions induced by environmental conditions or genetic modifications are key in generating different crystal polymorphs,15,16 the mechanism for solid-solid phase change upon altering ligand density is not obvious. We have performed diffusion analysis using fluorescence recovery after photobleaching as a means to elucidate the mechanism of this concentration-induced change. Materials and Methods Materials. Phospholipids 1,2-dipalmitoyl-sn-glycero-3-phosphoethanolamine (DPPE) and 1,2-ditridecanoyl-sn-glycero-3-phosphocholine (DTPC) were purchased from Avanti Polar Lipids, and N-((6(biotinoyl)amino)hexanoyl)-1,2-dihexadecanoyl-sn-glycero-3phosphoehtanolamine (biotin-X-DHPE) was purchased from Invitrogen. Fluorescein-5-isothiocyanate (FITC), Sephadex G50, and glutaraldehyde were from Sigma-Aldrich. Recombinant streptavidin was from Roche Applied Science, and chloroform and methanol were obtained from EMD Chemicals. Succinimidobiotin (NHSbiotin) was from Pierce, and grade V1 mica was from Ted Pella.
The water used in this investigation was purified with a Milli-Q purification system (Millipore). Phospholipid Bilayer Preparation. Langmuir-Blodgett monolayers were deposited on mica surfaces using a Langmuir trough (KSV 5000) using previously reported procedures as a departure point.13 To generate the lipid monolayer in direct contact with the substrate, DPPE dissolved in 15:1 chloroform:methanol was spread onto an aqueous surface and compressed to a surface pressure of 38 mN/m, resulting in a mean molecular area of 40 Å2 per lipid molecule. The DPPE monolayer was then transferred onto a 1-cm × 1.5-cm freshly cleaved mica substrate at constant pressure by pulling the substrate out of the liquid. A schematic of the lipid bilayer deposition and protein crystallization is shown in Figure 1. To create the outer leaflet of the bilayer, lipid mixtures of biotinX-DHPE were mixed with DTPC in chloroform. Mixtures ranged from 0.75 to 5 mol % biotin-X-DHPE, with a constant total concentration of 0.5 mg/mL. This lipid mixture was spread on a clean aqueous surface and compressed to a surface pressure of 38 mN/m (mean molecular area of 64 Å2 per molecule). The monolayer was then transferred onto the mica-supported DPPE surface by dipping the substrate into the liquid phase. The resulting supported bilayer remained in an aqueous environment for the duration of the protein crystallization protocol. Streptavidin Crystallization on Mica-Supported Substrates. All streptavidin was labeled with FITC and characterized according to previous procedures.17 Briefly, protein and dye were mixed overnight in the dark in 0.1 M sodium bicarbonate buffer (pH 9). Unreacted dye was removed by a Sephadex size exclusion column. The protein concentration and labeling ratio were determined from absorbance readings at 280 and 496 nm (UV-2101PC, Shimadzu). Final labeling concentrations yielded ratios of 1.2-1.5 FITC per streptavidin tetramer. To grow streptavidin crystals on the mica substrate, we first transferred the mica-supported bilayer to an incubating dish containing approximately 6.5 mL of 50 mM NaNO3 and 5 mM sodium phosphate at pH 7. A solution of FITC-labeled streptavidin was injected into the buffer subphase to achieve a final protein concentration of 7-8 µg/mL, and the samples were then incubated for 4 h. We have determined that this length of time is sufficient to saturate crystal growth at these protein concentrations. Since 7-8 µg/mL gives an excess of streptavidin relative to the available biotin binding sites on the surface of the substrate, we exchanged the protein solution with streptavidin-free buffer to remove background fluorescence prior to imaging. The samples were then placed in a (17) Nargessi, R. D.; Smith, D. S. Methods Enzymol. 1986, 122, 67-72.
9754 Langmuir, Vol. 23, No. 19, 2007 viewing cell, and fluorescence images of streptavidin crystals on substrates were obtained on a fluorescence microscope (BX51, Olympus) with a cooled CCD camera (ORCA-285, Hamamatsu). Modulation of Available Biotin-Binding Sites on Streptavidin. Streptavidin can be bound to the substrate on one face (which includes two binding sites), and we investigated the dependence of the crystal form on the number of subunits bound to the lipid bilayer. Streptavidin in solution was first premixed with soluble NHS-biotin (stock of 0.1 mg/mL in 0.1 M sodium phosphate buffer at pH 7 containing 1% DMSO) to achieve final streptavidin-to-biotin molar ratios of 1:1, 1:2, and 1:4. This solution was incubated for 4 h to allow proteinligand binding. Since streptavidin bound to biotin in solution has a lower dissociation rate than surface-bound biotin,18 preincubation of free biotin with streptavidin will block the binding sites available for attaching to the surface monolayer. Crystallization was performed under the typical protocols described above, with the exception that the concentration of biotin-X-DHPE in the substrate bilayer was fixed at 5%, and the overall streptavidin concentration was increased to 15 µg/mL. Furthermore, due to the long crystallization times in 50 mM NaNO3 and 5 mM sodium phosphate buffer, the salt concentration was increased to 500 mM NaNO3. We have determined that ionic strength of the buffer, while affecting growth rate, does not affect final crystal form (manuscript in preparation). AFM Characterization. To obtain molecular configuration within the two-dimensional crystals, we used an atomic force microscope (AFM) (MultiMode Nanoscope IIIa, Digital Instruments) equipped with a 120-µm (“J”) scanner. High-resolution probes (MikroMasch) were used in tapping mode to reduce the potential damage to the sample surface. To preserve the spatial arrangement within the protein crystals, we injected glutaraldehyde to a final concentration of 0.5% into the solution surrounding the crystals and incubated these samples overnight. Samples were then rinsed with water and dried prior to AFM imaging. The AFM images were acquired and processed with Nanoscope v5.2 (Veeco) and Scion Image software packages. Raw images were flattened and subject to two-dimensional (2D) fast Fourier transformation (FFT) to obtain 2D diffraction patterns. Reciprocal lattice reflections that were relatively sharp and yielded high intensities were selected, and the periodic lattice parameters were calculated by the software based on these selected spots. Noise outside of these spatial frequency points was filtered, and the Nanoscope software then reconstructed the image using the resulting filtered FTT. Fluorescence Recovery after Photobleaching (FRAP). Streptavidin bound to one ligand in a bilayer membrane has been shown to exhibit diffusion coefficients larger than those in which streptavidin is bound to the bilayer through two ligands.19 To determine how biotin binding affects crystallinity, therefore, we used fluorescence recovery after photobleaching (FRAP) to measure the lateral diffusion rates of streptavidin as a function of surface ligand concentration. In this technique, the fluorescence signal in a region of the noncrystalline protein monolayer is destroyed using an intense pulse of light. Fluorescence recovery is recorded over time as nonbleached protein molecules re-enter this area by diffusion.20 Since there is a large excess of protein added to the incubation dish, and since avidin/ streptavidin-biotin has an unusually tight interaction (KD ∼ 10-15 M),21,22 we assume the protein has saturated all of the available biotin ligands within the bilayer up to the limit of confluency. Therefore, the two-dimensional protein density can be modulated by the concentration of biotin-X-DHPE in DTPC. We measured the diffusion coefficient in samples containing 0.75-3 mol % biotinylated lipid on the surface. Beyond 3%, diffusion coefficients in noncrys(18) Perez-Luna, V. H.; O’Brien, M. J.; Opperman, K. A.; Hampton, P. D.; Lopez, G. P.; Klumb, L. A.; Stayton, P. S. J. Am. Chem. Soc. 1999, 121, 64696478. (19) Gambin, Y.; Lopez-Esparza, R.; Reffay, M.; Sierecki, E.; Gov, N. S.; Genest, M.; Hodges, R. S.; Urbach, W. Proc. Natl. Acad. Sci. U.S.A. 2006, 103, 2098-2102. (20) Axelrod, D.; Koppel, D. E.; Schlessinger, J.; Elson, E. L.; Webb, W. W. Biophys. J. 1976, 16, A217-A217. (21) Green, N. M. Methods Enzymol. 1990, 184, 51-67. (22) Green, N. M. Biochem. J. 1963, 89, 585-&.
Lou et al. talline regions were difficult to reliably measure as crystal domains became confluent. Samples of streptavidin on mica-supported bilayers were prepared and incubated as described above. The proteins bound to the substrates were then transferred to buffer free of streptavidin in solution and observed with a multi-spectral laser-scanning microscope (LSM 510 META, Zeiss) at room temperature (∼22 °C). Regions determined to be noncrystalline were used in FRAP investigations; crystalline domains did not exhibit measurable diffusion within the time scale investigated. Noncrystalline regions were photobleached with a Ti: sapphire two-photon femtosecond laser (Mai Tai, Spectra-Physics) at a wavelength of 800 nm. Fluorescence intensity and recovery within the bleached area were recorded by LSM 510 software (Zeiss), and the data was processed and fit with Origin (OriginLab). Diffusion coefficients were extracted following the mathematical treatment of Axelrod,20 which is valid for two-dimensional diffusion in a diskshaped area. Since the streptavidin-biotin bond is very strong relative to typical protein-ligand interactions, and since unbound proteins in the viewing cell are removed prior to photobleaching, we assume that intensity recovery is due only to lateral diffusion and not to exchange with free protein in the buffer.
Results and Discussion a. Effect of Biotin Concentration on Crystal Morphologies. Two-dimensional streptavidin crystals were grown on micasupported phospholipid bilayers and imaged with fluorescence microscopy. As we varied the percentage of the biotinylated lipid in the outer monolayer leaflet, we found that crystal morphologies varied greatly, as described in Figure 2. A lipid layer with 1% or less biotin-X-DHPE (in a binary mixture of DTPC/biotin-X-DHPE) resulted in no apparent crystalline domains, although streptavidin binding to the substrate was observed. At 1.5% biotin-X-DHPE, crystalline domains displayed relatively large aspect ratios, with the majority being needles (Figure 2A) coexisting with a smaller amount of very narrow H-shaped morphologies (Figure 2B). Crystals at 2% and 2.5% biotin-X-DHPE were dominated by relatively large H- or X-shaped domains interspersed with small needle-like crystals (Figure 2C-F). These domains often exhibited needle-like or dendritic edges at these lipid concentrations. Prior streptavidin crystallization studies at the air/water interface have shown that dendrites are attributed to transport limitations23 and needle-like edges are due to the presence of crystals with P1 symmetry in coexistence with other crystal forms within the domains.16 As the biotinylated lipid concentration was increased further to 3% and above, the X-shaped crystals became less dendritic and yielded smoother edges (Figure 2G), and overall crystalline surface coverage became denser with increasing biotin concentration. By 5% biotin-X-DHPE, crystalline domains grew until confluent, giving shapes with no distinct morphology (Figure 2H). Crystallinity of all these domains was confirmed by viewing the two-dimensional morphologies with a polarizer analyzer. Figure 3 demonstrates that these domains polarize light as shown by the intensity variations upon rotatation of the analyzer, indicating the high molecular order within these domains. Crystallinity is also corroborated by AFM data, as described below. Under these growth conditions, we have found that there is a critical concentration between 1 and 1.5% biotin-X-DHPE, below which crystals do not form. Higher concentrations yield morphologies similar to those previously reported. For example, Calvert and Leckband reported dendritic-X and confluent domains at 2 and 5% biotinylated lipid, respectively,13 similar to our observed crystals (Figures 2C and 1H). Our investigations (23) Ku, A. C.; Darst, S. A.; Kornberg, R. D.; Robertson, C. R.; Gast, A. P. Langmuir 1992, 8, 2357-2360.
Protein Crystals on a Solid Substrate
Langmuir, Vol. 23, No. 19, 2007 9755
Figure 2. Crystal morphologies and corresponding crystal types grown at different concentrations of biotinylated lipid in DTPC. No crystals are observed at or below 1%. Domains with large aspect ratios exist at 1.5% biotin-x-DHPE, including needle-like morphologies (A) and narrow H-shaped domains (B). At 2% (C and D) and 2.5% (E and F) biotin-X-DHPE, large X and H-shaped domains coexist with needle-like domains. X-shaped crystals are observed at 3% biotin-X-DHPE (G). At 5% biotin-X-DHPE, domains grow to confluency and individual morphologies are amorphous (H). Image H is polarized to distinguish between different domains. AFM data shows that crystals at 1.5% biotin-X-DHPE primarily exhibit P1 symmetry and crystals at 3% biotin-X-DHPE and above exhibit C222 symmetry. Crystalline domains resulting from intermediate concentrations of biotinylated lipid comprise a coexistence of the different forms within the individual domains. (Scale bar in each picture is 50 µm.)
Figure 3. Domains are confirmed to be crystalline by exhibiting optical anisotropy. Shown here are identical views of crystals grown at 3% biotinylated lipid and observed with a polarizer analyzer upon 90° rotation. (Scale bar is 50 µm.)
additionally yield a coexistence of different morphologies between these concentration ranges. The lattice parameters within these different types of domains have not been previously determined, and we used AFM to elucidate molecular ordering. b. Correlation of Crystal Lattices with Morphologies. Previous investigations of two-dimensional streptavidin crystals grown at the air/water interface found that macroscopic crystal morphology is related to the molecular configuration within these domains.15,16 Three distinct two-dimensional streptavidin crystal
types have been identified, with unit cell parameters exhibiting P1,24 P2,25 and C22210 symmetry. The macroscopic domains in crystals at the air/water interface consist of these three crystalline forms, either individually or in coexisting combinations, and the different morphologies correlate to different coexistence ratios. Therefore, we can infer molecular-scale order by examining macroscopic structure of the crystal with fluorescence microscopy.15 A prior investigation of streptavidin crystals grown on solid substrates has shown that, like those grown at the air/water interface, both C222 and P2 crystal forms can be obtained.12 However, due to the nature of the experimental protocol in this previous study, these crystal forms were not related to largescale macroscopic morphologies. Furthermore, the crystal with P1 symmetry has not yet been reported to form on a solid substrate. In this investigation, we grew crystals on our mica substrate and used AFM analysis to probe molecular order within resulting macroscopic structures. A summary of the crystalline phase and its related macroscopic domains as a function of biotinylated lipid concentration is presented in Figure 2. Our data show that X-shaped morphologies and confluent regions grown at 3% (and above) biotin-X-DHPE comprised of molecules packed in the C222, square lattice arrangement (Figure 4). Average lattice parameters are a ) 6.3 ( 0.2 nm, b ) 6.1 ( 0.2 nm, and γ ) 90.4 ( 2.9°. These values are close to previously published values of a ) b ) 5.8 nm and γ ) 90° for C222 crystals generated at the air/water interface.10 In a previous study, streptavidin crystals which were grown on mica-supported lipid bilayers (formed by unilamellar vesicles) also yielded C222 crystals with lattice parameters reported to be a ) b ) 8 nm and γ ) 93.2°.12 These unit cell values, however, are a ) b ) 5.7 nm if an indexing scheme consistent with Darst (24) Hemming, S. A.; Bochkarev, A.; Darst, S. A.; Kornberg, R. D.; Ala, P.; Yang, D. S. C.; Edwards, A. M. J. Mol. Biol. 1995, 246, 308-316. (25) Wang, S. W.; Poglitsch, C. L.; Yatcilla, M. T.; Robertson, C. R.; Gast, A. P. Langmuir 1997, 13, 5794-5798.
9756 Langmuir, Vol. 23, No. 19, 2007
Lou et al.
Figure 4. AFM images of crystals with C222 symmetry. (A) Raw image of confluent C222 crystals grown at 5% biotin-X-DHPE, (B) its enhanced image reconstructed from filtered data, and (C) the corresponding C222 diffraction pattern. (Scan scale is 200 nm.)
Figure 5. AFM images of crystals with P1 symmetry. (A) Raw image of P1 crystals grown at 1.5% biotin-X-DHPE, (B) its enhanced image reconstructed from filtered data, and (C) the corresponding P1 diffraction pattern. (Scan scale is 200 nm.)
et al.10 is used. Our slightly larger unit cell lengths relative to prior solid-supported crystals possibly can be attributed to the degree of hydration. Reviakine and Brisson determined their lattice parameters under fully hydrated conditions while our samples are necessarily dried prior to AFM imaging.12 We also performed AFM on samples with needle-like domains to determine whether they correspond to crystalline lattices with P1 symmetry, as is the case with air/water interface crystals. In Figure 5, we present a typical raw image of a P1 crystal, its reconstruction from filtered data, and the corresponding diffraction pattern. Unit cell parameters of crystals with P1 symmetry were a ) 5.7 ( 0.1 nm, b ) 5.0 ( 0.2 nm, and γ ) 110 ( 0.1°. These values are close to a ) 5.8 nm, b ) 5.0 nm, and γ ) 113°, which are those previously reported for P1 crystals grown at the air/ water interface.24 Although C222 crystals are relatively straightforward to image with the AFM, we found that obtaining clear images showing the molecular order within the needle-like domains (which comprise mainly the P1 crystal phase) was much more difficult. P1 crystals are very fragile and, in the process of obtaining molecular-scale images, frequently fractured. In crystals formed at intermediate biotin-X-DHPE concentrations of 2-2.5%, P1 regions that were adjacent to (and coexisting with) the stronger C222 regions were more likely to remain intact. These observations are consistent with the characteristics of P1 crystals generated at the air/water interface.15 At biotin-X-DHPE concentrations of 2-2.5%, we also obtained “intermediate” crystal phases in which lattice parameters could not obviously be classified as P1, P2, or C222. Although these unit cell values were closest to P2 parameters, nevertheless they varied in range for a, b, and γ. The diffraction data for these intermediate crystals were reminiscent of amalgam crystals described in a previous air/water investigation.15 P2 crystals have characteristics of both C222 and P1, and in fact have been found in conditions that overlap with P1 and C222. The absence of a definite crystal with P2 symmetry, while P1 and C222 are
observed to coexist, could be indicative of P2’s fragile and transitory nature. c. Crystal Form Dependence on Number of Protein Subunits Bound to the Surface. We hypothesize that the preference for P1 at low biotin-X-DHPE concentrations and C222 at high concentrations is related to the number of protein subunits that are bound to the surface via the biotinylated lipid. A streptavidin molecule comprises four identical subunits arranged in 222 point group symmetry.26 As these tetrameric molecules attach onto the solid substrate, two types of streptavidin-biotin couplings will be generated: one biotin bound to a single subunit of a streptavidin tetramer (Figure 6, left) and two biotin molecules bound to the two adjacent subunits of a single streptavidin tetramer (Figure 6, right). The other two ligand-binding sites on the opposite face of the protein remain unbound and exposed to aqueous solution. At low biotin-X-DHPE concentrations, we expect that the large excess of streptavidin relative to biotinylated lipid results in a dominance of singly bound protein molecules. As the biotin concentration is increased, the number of protein molecules that are attached to the surface by two binding sites will correspondingly increase. Steric constraints will, at a critical point, prevent further increases in surface protein concentration. At the surface pressure of 36 mN/m, biotin-X-DHPE (at 1.55%) in DTPC has an approximate mean molecular area of 64 Å.2 With projection areas of 2669 Å2 per streptavidin molecule for the P1 crystal form and 3364 Å2 for C222 symmetry, and assuming enough protein is on the surface to generate a confluent crystalline monolayer, we estimate that crystals formed below 1.9% biotin-X-DHPE will be dominated by singly bound streptavidin tetramers, while crystals formed above 4.8% will be attached to the surface at the two biotin-binding sites (see the Supporting Information). Experimentally, these concentrations (26) Weber, P. C.; Ohlendorf, D. H.; Wendoloski, J. J.; Salemme, F. R. Science 1989, 243, 85-88.
Protein Crystals on a Solid Substrate
Langmuir, Vol. 23, No. 19, 2007 9757
Figure 6. Model for P1 and C222 crystal formation is related to the number of subunits bound to the solid substrate. (A) Side view and (B) top view. Each protein has four available sites, but geometry only enables one or two of the sites to be bound to the solid surface.
correspond to crystals generated which exhibit P1 and C222 symmetry, respectively. To test whether the crystal form can be altered by modulating the number of sites attached to the solid surface, we incubated streptavidin with different amounts of free biotin (NHS-biotin) prior to crystallization on the lipid bilayer. The unusual strength of the streptavidin-biotin interaction has been attributed to its slow dissociation kinetics.22 Since streptavidin bound to biotin in solution has a dissociation rate constant that is 2 orders of magnitude lower than the value for streptavidin conjugated to surface-attached biotin,18 we expect that prebound, soluble biotin will largely remain associated with the streptavidin tetramer and limit the available sites for attachment to the solid substrate. If the free biotin is bound to the subunits directed away from the surface, our model predicts that this binding should have little effect on crystallization. However, if the free biotin is bound to one of the two subunits which face the solid substrate, streptavidin will only be anchored to the lipid bilayer at one binding site. Therefore, we predict that an increase in the amount of free biotin added prior to crystallization should result in a greater amount of crystals with P1 symmetry. The results of this experiment are presented in Figure 7. At 5% biotin-X-DHPE with no free biotin, we obtain confluent crystals determined to exhibit C222 symmetry. At a 1:1 molar ratio of free biotin to streptavidin tetramer, we observe dendritic-X morphologies which are indicative of C222 packing. When the concentration of free biotin is increased to a 2:1 molar ratio, we find crystalline domains with high aspect ratios, which are characteristic of P1 crystals in coexistence with small amounts of P2 or C222 (Figure 2).15 We were not able to determine lattice parameters for crystals grown with streptavidin prebound to biotin at this ratio, which is also consistent with the fragile nature of P1 crystals. Our data therefore suggest a preference for P1 crystals in populations of streptavidin that are able to attach to the substrate via a single subunit, whereas proteins bound to the bilayer by two subunits generate C222 crystals. As expected, streptavidin with all binding sites blocked with biotin (4 biotin:1 streptavidin ratio) could not bind to the surface and resulted in no crystals. Additional evidence shows that biotin in solution does indeed specifically interact with the biotin-binding pockets of streptavidin rather than elsewhere on the protein. When soluble NHS-biotin is injected into the liquid phase containing preformed twodimensional streptavidin crystals at ratios of 3:1 and 4:1 NHSbiotin:streptavidin, the fluorescent intensity of the lipid surface
Figure 7. Fluorescence microscopy images of crystals grown after the addition of free biotin to the streptavidin solution, prior to incubation with lipid bilayer. (A) No free biotin added. (B) 1:1 ratio of free biotin to streptavidin tetramer. (C) 2:1 ratio of free biotin to streptavidin tetramer. (D) 4:1 ratio of free biotin to streptavidin tetramer. (Scale bar is 50 µm.)
becomes dark within an hour. This demonstrates that the soluble biotin is able to specifically access the biotin-binding sites and compete with the surface-bound biotin at these ligand:protein ratios, resulting in detachment of the fluorescently tagged protein from the surface. d. Fluorescence Recovery after Photobleaching (FRAP). Several studies have examined the relationship between lateral mobility and the size of an entity diffusing in a lipid bilayer.19,27,28 Streptavidin molecules which are bound to a bilayer membrane through one subunit have been shown to diffuse faster than those attached by both available subunits; this discrepancy has been related to the size, since the diameter of the diffusing system is larger in the latter case.19 Therefore, we used FRAP to further elucidate the binding state of the protein molecules at the different biotin concentrations. Proteins were incubated with the lipid bilayer substrate for at least 4 h. At biotin-X-DHPE concentrations in which crystals formed, this length of time ensured saturation of crystal growth. (27) Saxton, M. J. Biophys. J. 1993, 64, 1053-1062. (28) Saffman, P. G.; Delbruck, M. Proc. Natl. Acad. Sci. U.S.A. 1975, 72, 3111-3113.
9758 Langmuir, Vol. 23, No. 19, 2007
Figure 8. Diffusion coefficient vs biotinylated lipid concentration. The relationship between these two variables is observed to exhibit biphasic behavior, with one dependence in the non/nascent-crystalline regime (blue solid line), and another in the regime with crystal growth (green solid line). FRAP measurements in the crystal-forming regime were performed in noncrystalline areas of the sample. The expected dependence within the crystal-forming regime, if the proteins are attached to the surface by a constant average number of biotins at the different concentrations, is shown in red (dotted line). Each data point in the crystal-forming regime (green) is an average of four replicates, while those in the noncrystalline regime (blue) are based on three replicates.
FRAP analysis was then performed in the remaining noncrystalline regions. Results showing the relationship between diffusion coefficient and biotin-X-DHPE concentration are summarized in Figure 8. Biphasic activity is observed which corresponds to the presence or absence of crystals. In the region in which there are no crystals (undersaturated protein concentration) or small nascent crystals (protein concentration just above saturation), the diffusion coefficient of streptavidin is inversely proportional to the concentration of surface-bound biotin (see Figure 8). Our results are in agreement with previous studies which concluded that protein diffusivity in a freely diffusing lipid monolayer increases as protein concentration is decreased.29,30 This dependence on concentration is attributed to the probability of finding a vacant space adjacent to the protein.31 Our diffusion coefficients for streptavidin bound to biotin-X-DHPE in DTPC on a solid substrate are close to 3.1 × 10-8 cm2/s, which is the value previously reported for 1% fluorescently labeled DHPE in DTPC.13 Our lower value at 1% biotin-X-DHPE can be explained by the larger diameter and molecular weight of the diffusing streptavidin-DHPE complex relative to the size of FITC-DHPE. Diffusion coefficients determined for proteins in the crystalforming concentrations also results in a linear profile, although the slope of this relationship is different. Since crystals have been grown to maximum coverage, we expect that the protein concentrations remaining in the noncrystalline regions (in which diffusion measurements were obtained) are at saturation. If the average diffusing protein species are identical between 1.5 and 3% biotin, then the resulting diffusion coefficients should be constant over the concentration range. However, a linear profile with nonzero slope is observed. Furthermore, the concentrations which correspond to growth of P1 and C222 crystals (1.5 and 3%, respectively) give diffusion coefficients that exhibit a 2-fold disparity. This difference in diffusion coefficients is consistent with a recent investigation that found the relationship between diffusion (29) Subramaniam, S.; Seul, M.; Mcconnell, H. M. Proc. Natl. Acad. Sci. U.S.A. 1986, 83, 1169-1173. (30) Tank, D. W.; Wu, E. S.; Meers, P. R.; Webb, W. W. Biophys. J. 1982, 40, 129-135. (31) Minton, A. P. Biophys. J. 1989, 55, 805-808.
Lou et al.
coefficient (D) and protein dimension (R) to be D ∼ 1/R.19 Specifically in this cited study, the diffusion coefficient for streptavidin bound to the model membrane bilayer through one streptavidin subunit was twice that of streptavidin bound via two subunits. Similar trends were also obtained in studies using gold nanoparticles as model proteins; nanoparticle diffusion was shown to be slower in multivalent binding to the surface relative to single binding.32 Our diffusion data, therefore, supports the model that P1 crystals consist of proteins bound to the substrate by a single binding site and proteins in C222 crystals are attached at both sites. Intermediate diffusion coefficient values, then, correspond to average intermediate ratios of one vs two-bound states. An alternative model for the reduction in diffusion coefficients proposes that P1 and C222 crystals exhibit different protein saturation concentrations. To address this possibility, we used the fluorescence intensity in the noncrystalline regions to compare the relative protein concentration at saturation. At 1.5, 2.0, and 2.5% biotin-X-DHPE, the intensities were 1710 ( 240, 1590 ( 190, and 1450 ( 450 (arbitrary fluorescence units with the substrate background subtracted), respectively. (Data for 3% could not be directly compared due to use of a modified laser source.) These values reveal that the surface protein concentration is relatively constant over the range of biotinylated lipid concentrations. Although there may be a slight decrease in the average protein concentration as the amount of surface biotin is increased, this difference is not large relative to overall intensity. Furthermore, a decrease in protein concentration should yield increasing diffusion coefficient values, but we observe, in fact, that the coefficient decreases from 1.5 to 2.5% biotin-X-DHPE. This data therefore supports our model that the 2-fold change in diffusion coefficient is due to binding configuration rather than different protein saturation concentrations. We observed an intriguing result in the course of these FRAP experiments. During our initial attempts to examine diffusion at >1.5% biotin-X-DHPE concentrations, crystals were found to form in the photobleached area, thereby disrupting diffusion coefficient measurements. To circumvent this obstacle, we attempted to use avidin, a glycosylated protein that is homologous to streptavidin but is not expected to form 2-D crystals due to the surface oligosaccharide moieties.23 Curiously, the resulting avidin monolayers formed a gel-like monolayer in which the fluorescence of the bleached area could not recover up to the 2 h the sample was examined. (As a comparison, typical fluorescence recovery times for streptavidin were only a few minutes.) This unexpected rigidity has also been recently reported for avidin monolayers on unilamellar vesicles;33 unlike streptavidin-coated or bare vesicles, avidin vesicles do not deform under osmotic stress. These observations suggest that protein surface characteristics, such as charge, should also be considered in models of lipid membranes. Numerous studies have investigated the dependence of diffusion in lipid bilayers to factors including protein concentration,29,31 lipid composition of bilayer,29 size of diffusing system,19,27,32 and viscosity.28,34 Streptavidin and avidin are remarkably similar in structure and size, but differ greatly in surface glycosylation and isoelectric point.21,35 Our data suggest the oligosaccharide groups may create a network of hydrogen bonds between adjacent avidin molecules that is not possible in (32) Lee, G. M.; Ishihara, A.; Jacobson, K. A. Proc. Natl. Acad. Sci. U.S.A. 1991, 88, 6274-6278. (33) Horton, M. R.; Manley, S.; Arevalo, S. R.; Lobkovsky, A. E.; Gast, A. P. J. Phys. Chem. B 2007, 111, 880-885. (34) Peters, R.; Cherry, R. J. Proc. Natl. Acad. Sci.-Biol. 1982, 79, 43174321. (35) Livnah, O.; Bayer, E. A.; Wilchek, M.; Sussman, J. L. Proc. Natl. Acad. Sci. U.S.A. 1993, 90, 5076-5080.
Protein Crystals on a Solid Substrate
Langmuir, Vol. 23, No. 19, 2007 9759
streptavidin, resulting in a rigid gel-like protein monolayer. Intermolecular interactions between functional groups on protein surfaces, therefore, may significantly affect diffusion behavior. e. Model for P1 vs C222 Growth. Streptavidin tetramers which are bound to a single biotinylated lipid on the substrate only have one anchor; therefore, relative to proteins bound by both subunits, they will have a greater range of movement, both to diffuse laterally along the bilayer and to tilt out of the plane. In contrast, we expect that tetramers with two subunits anchored to the substrate will diffuse slower due to the increased effective radius19,28 and will exhibit less rotation out of the two-dimensional plane. The molecular structure of streptavidin molecules placed in the C222 unit cell configuration shows that the regions of streptavidin-streptavidin interaction are small and specific, with interactions isolated only to the “corners” of the tetramer.11 In contrast, P1 crystals are more compact and have interactions with adjacent molecules at several sites throughout its surface.16 Since the formation of C222 crystals depends on much smaller regions of interaction, diffusion within the two-dimensional plane could require a greater degree of rigidity in order to capture a molecule within its crystal. This is consistent with C222 crystal formation at biotin-X-DHPE concentrations which favor two subunits bound to the substrate. Since C222 crystals appear to be more stable than the fragile P1 form, they could be more thermodynamically favored, whereas the P1 form could be a kinetically preferred metastable configuration. The results of the experiment in which streptavidin was prebound with free biotin in solution also rule out an alternative model. In previous air/water investigations, the anisotropic nature of C222 crystals were shown to be related to interactions between biotin-bound subunits and biotin-unbound subunits, with preferred growth along the direction of biotin-bound subunits.15 Crystal structures revealed that binding to biotin alters the position of a flexible loop,36 and this difference between bound and unbound subunits results in intermediate interaction changes and anisotropic growth.15 Indeed, when streptavidin was bound to a lipid monolayer through surface histidines rather than biotin binding, resulting crystals were square, reflecting the identical nature of all four tetramers.37 In this investigation, our data (in Figure 7) does not support the origin of the concentration-dependent P1 vs C222 crystal forms to be this flexible loop. Such a model would predict that the aspect ratio of crystals grown with free biotin would not change since binding sites would all remain bound, either to free biotin or to surface-bound biotin. Our hypothesis that P1 is obtained when streptavidin can sample more flexible configurations is also consistent with previously reported phenomena.16 At the air/water interface, crystals with the P1 unit cell parameters were observed to form either a lower-quality, less-ordered paracrystal, or a crystal with longer-range order. Although the paracrystalline form was observed initially, the more thermodynamically stable higherordered P1 crystal formed over time. This observed transition demonstrates that the P1 crystal has a certain degree of flexibility associated with it which has not been observed for the C222 crystal.
Conclusions
(36) Freitag, S.; LeTrong, I.; Klumb, L.; Stayton, P. S.; Stenkamp, R. E. Protein Sci. 1997, 6, 1157-1166. (37) Frey, W.; Schief, W. R.; Pack, D. W.; Chen, C. T.; Chilkoti, A.; Stayton, P.; Vogel, V.; Arnold, F. H. Proc. Natl. Acad. Sci. U.S.A. 1996, 93, 4937-4941.
Our results have shown that the density of ligand in a twodimensional, solid-supported bilayer membrane can play a key role in controlling the molecular configuration and morphology of resulting protein crystals. Previous studies have reported two crystal forms (with C222 and P2 symmetry) on solid substrates, and in this work, we have confirmed the existence of the third predicted crystal type (with P1 symmetry). By modulating the density of biotinylated lipid in the outer bilayer leaflet, we obtain crystalline domains with various morphologies that correspond to different unit cell parameters. Below a critical biotin-X-DHPE concentration (1%), we observed no crystals. As the concentration is increased (1.5%), needle-like crystals with P1 symmetry appear. At intermediate ligand concentrations, coexistence of multiple crystal forms is observed, both within single crystalline domains and in independent domains. At higher concentrations (>3%), crystals with C222 symmetry comprise X-shaped and confluent domains. The correlation between morphology and unit cell parameters is consistent with previously reported results for protein arrays grown at the air/water interface.15 This similarity in morphologyto-lattice mapping suggests that the fine interactions between protein molecules may be more important than the proteins’ interactions with the solid substrate in determining molecular configuration. Our data gives evidence to the hypothesis that streptavidin molecules which are bound to the interface by one binding site have more flexibility and preferentially forms crystals with P1 symmetry. Proteins which attach to the surface by the pair of subunits on one face of the molecule exhibit less movement out of the plane and preferentially self-assemble into crystals with C222 symmetry. The actual formation of P1 vs C222 crystals matches well with the predicted concentration of biotinylated lipid for one and two-bound sites, respectively. FRAP measurements and blockage of available attachment sites on the protein further support this model. Our results demonstrate that the delicate balance of intermolecular interactions plays an important role in protein self-assembly and mobility in lipid bilayers, and the protein surface characteristics which govern such interactions should be considered in biological membrane models. Acknowledgment. We gratefully acknowledge Grace Lu and Joseph Fan for AFM use and assistance, Michael Dennin for donation of the Langmuir trough, Frank Shi for use of his optical table, and Deborah Leckband for helpful discussions. This work was funded by the ACS-Petroleum Research Fund and NSF (ECS-0609195). Supporting Information Available: Calculated concentrations of biotin-X-DHPE corresponding to singly-bound and doubly-bound streptavidin tetramers. This material is available free of charge via the Internet at http://pubs.acs.org. LA701399S
