pubs.acs.

org/Langmuir 2009 American Chemical Society

Fmoc-Diphenylalanine Self-Assembly Mechanism Induces Apparent pKa Shifts

Claire Tang,, Andrew M. Smith,, Richard F. Collins, Rein V. Ulijn,,, and Alberto Saiani*,

School of Materials, The University of Manchester, Grosvenor Street, Manchester M1 7HS, United Kingdom, Manchester Interdisciplinary Biocentre (MIB), The University of Manchester, 131 Princess Street, Manchester M1 7DN, United Kingdom, and WestCHEM, University of Strathclyde, Thomas Graham Building, 295 Cathedral Street, Glasgow G1 1XL, United Kingdom Received February 23, 2009. Revised Manuscript Received May 20, 2009

We report the effect of pH on the self-assembly process of Fmoc-diphenylalanine (Fmoc-FF) into fibrils consisting of antiparallel -sheets, and show that it results in two apparent pKa shifts of 6.4 and 2.2 pH units above the theoretical pKa (3.5). Using Fourier transform infrared (FTIR) spectroscopy, transmission electron microscopy (TEM), wide angle X-ray scattering (WAXS), and oscillatory rheology, these two transitions were shown to coincide with significant structural changes. An entangled network of flexible fibrils forming a weak hydrogel dominates at high pH, while nongelling flat rigid ribbons form at intermediate pH values. Overall, this study provides further understanding of the self-assembly mechanism of aromatic short peptide derivatives.

Introduction
Over the past decade, significant efforts have been made to develop a new generation of biomaterials based on the selfassembly of peptides and their derivatives.1,2 By exploiting the spontaneous or induced molecular arrangement of peptides and their derivatives in aqueous solutions, nanostructured hydrogels can be formed under specific conditions.3-7 These highly hydrated scaffolds have potential applications in tissue engineering,5,8 3D cell culture,9-11 and templating.12,13 Such peptide based biomaterials exploit known biological architectures such as the -sheet and are usually composed of peptide components exceeding 10 amino acids in length. A relatively new class of hydrogel
*Corresponding author. E-mail: a.saiani@manchester.ac.uk.
(1) Lehn, J.-M. Supramolecular Chemistry: Concepts and Perspectives; VCH Press: New York, 1995. (2) Whitesides, G. M.; Grzybowski, B. Science 2002, 295, 24182421. (3) Zhang, S. Nat. Biotechnol. 2003, 21, 11711178. (4) Mart, R. J.; Osborne, R. D.; Stevens, M. M.; Ulijn, R. V. Soft Matter 2006, 2, 822835. (5) Haines-Butterick, L.; Rajagopal, K.; Branco, M.; Salick, D.; Rughani, R.; Pilarz, M.; Lamm, M. S.; Pochan, D. J.; Schneider, J. P. Proc. Natl. Acad. Sci. U.S. A. 2007, 104, 77917796. (6) Tsonchev, S.; Niece, K. L.; Schatz, G. C.; Ratner, M. A.; Stupp, S. I. J. Phys. Chem. B 2008, 112, 441447. (7) Saiani, A.; Mohammed, A.; Frielinghaus, H.; Collins, R.; Hodson, N.; Kielty, C. M.; Sherratt, M. J.; Miller, A. F. Soft Matter 2009, 5, 193202. (8) Lee, K. Y.; Mooney, D. J. Chem. Rev. 2001, 101, 18691880. (9) Beniash, E.; Hartgerink, J. D.; Storrie, H.; Stendahl, J. C.; Stupp, S. I. Acta Biomater. 2005, 1, 387397. (10) Jayawarna, V.; Richardson, S. M.; Hirst, A. R.; Hodson, N. W.; Saiani, A.; Gough, J. E.; Ulijn, R. V. Acta Biomater. 2009, 5, 934943. (11) Zhou, M.; Smith, A. M.; Das, A. K.; Hodson, N. W.; Collins, R. F.; Ulijn, R. V.; Gough, J. E. Biomaterials 2009, 30, 25232530. (12) Hartgerink, J. D.; Beniash, E.; Stupp, S. I. Science 2001, 294, 16841688. (13) Reches, M.; Gazit, E. Science 2003, 300, 625627. (14) Vegners, R.; Shestakova, I.; Kalvinsh, I.; Ezzell, R. M.; Janmey, P. A. J. Pept. Sci. 1995, 1, 371378. (15) Jayawarna, V.; Ali, M.; Jowitt, T. A.; Miller, A. F.; Saiani, A.; Gough, J. E.; Ulijn, R. V. Adv. Mater. 2006, 18, 611614. (16) Mahler, A.; Reches, M.; Rechter, M.; Cohen, S.; Gazit, E. Adv. Mater. 2006, 18, 13651370. (17) Schnepp, Z. A. C.; Gonzalez-McQuire, R.; Mann, S. Adv. Mater. 2006, 18, 18691872. (18) Toledano, S.; Williams, R. J.; Jayawarna, V.; Ulijn, R. V. J. Am. Chem. Soc. 2006, 128, 10701071.

scaffolds formed from the self-assembly of much shorter peptides modified with aromatic ligands has recently been developed.14-20 The aromatic moieties play a key role in the self-assembly process through - interactions, while the peptide components are stabilized via hydrogen bonding. The main focus of our work to date has been to characterize the self-assembling behavior of dipeptides possessing an N-terminal fluorenyl-9-methoxycarbonyl (Fmoc) group. The overall aim is to establish design rules allowing the preparation of biomaterials, in particular hydrogels, with tailored properties. We have recently shown that Fmoc-diphenylalanine (Fmoc-FF) can form hydrogels at physiological pH on which cells have been shown to grow and proliferate.10,11,15 We subsequently proposed a molecular model for the self-assembled structure formed by this system.19 Fmoc-FF peptides were shown to form antiparallel -sheets that self-assemble laterally through - stacking of the fluorenyl groups and phenyl rings (phenylalanine side chain) present on the edge of the -sheets. Due to the natural twist of the -sheets, four sheets come together to form a cylindrical fibril with an external diameter of 3.0 nm. These fibrils were then shown to further self-assemble laterally, forming large flat ribbons under specific pH conditions. For a detailed discussion of the model, we refer the reader to ref 19. The R-carboxylic acid groups of amino acids are in their anionic form at neutral pH. N-protected nonpolar peptides have pKa values of about 3.5.21 (SPARC web calculator to be found at http://ibmlc2.chem.uga.edu/sparc.) In dilute solutions, one would therefore expect terminal carboxylic acid groups to be ionized at pH values higher than 3.5 and the molecules to be negatively charged. Self-assembly would presumably be unfavored at pH higher than 3.5 due to electrostatic repulsion between the negatively charged molecules. Indeed, a
(19) Smith, A. M.; Williams, R. J.; Tang, C.; Coppo, P.; Collins, R. F.; Turner, M. L.; Saiani, A.; Ulijn, R. V. Adv. Mater. 2008, 20, 3741. (20) Zhang, Y.; Gu, H.; Yang, Z.; Xu, B. J. Am. Chem. Soc. 2003, 125, 13680 13681. (21) Ulijn, R. V.; Moore, B. D.; Janssen, A. E. M.; Halling, P. J. J. Chem. Soc., Perkin Trans. 2 2002, 10241028.

Langmuir 2009, 25(16), 94479453

Published on Web 06/19/2009

DOI: 10.1021/la900653q

9447

Article

Tang et al.

Figure 1. Equilibrium between the dipeptide neutral acidic form (Fmoc-FF) and its basic ionized form (Fmoc-FF-).

number of Fmoc-peptides only showed self-assembly at low pH.15 However, Fmoc-FF (Figure 1), a highly hydrophobic peptide derivative, was found to self-assemble at neutral pH.19 It is known that pKa can shift dramatically in protein and peptide self-assembly, especially in hydrophobic environments.22 For example, a shift of 6.1 has been observed for aspartic acid side chain carboxylic acid in a family of protein based polymers composed of the repeat of polypentapeptides.23 pKa shifts have also been encountered in fatty acids. For instance, palmitic acid (C16) revealed a pKa shift of up to 3.9 which is related to the formation of foam by these molecules.24 Here we show that Fmoc-FF self-assembly results in a suppressed ionization leading to dramatic pKa shifts related to significant structural transitions.

Materials and Methods

Materials. Fmoc-FF peptides were purchased from Bachem
(Bubendorf, Switzerland) and used without any further purification. The purity of the compound was checked by HPLC (>98%) and mass spectrometry. HPLC grade water was purchased from Merck, and deuterated water (99.9 atom % D) from Sigma-Aldrich. Potentiometry. pH measurements were performed using a Hanna Instruments pH210 pH-meter equipped with a Hamilton Spintrode pH-probe (Ag/AgCl reference system, 3 M KCl electrolyte, ceramic diaphragm, sensitivity 58 mV/pH unit at 25 C). The pH-meter was calibrated before each experiment to check the response of the electrode with two buffer solutions purchased from Fisher Scientific: phthalate pH 4.01 and phosphate pH 7.01 buffer solutions. Titration Experiments. Depending on the desired concentration, the required amount of Fmoc-FF was suspended into 2 mL of HPLC grade water. Sodium hydroxide (0.5 M, e.g., 55 L for 10 mmol L-1 sample) was added to the aqueous suspensions of Fmoc-FF until pH 10.5 was reached. The samples were vortexed and sonicated (VWR ultrasonicator bath, 30 W) for 1 min to fully dissolve the modified peptide. To ensure that these conditions did not lead to significant hydrolysis of the carbamate and loss of the fluorenyl moiety, HPLC was used to estimate the cleavage percentage. After 10 min at pH 10.5, less than 1% cleavage was observed. The titration experiments were performed by stepwise addition of small volumes of diluted HCl (0.085 M). After each addition the samples were heated to 75-80 C, vortexed, sonicated, and subsequently cooled back to room temperature using a water bath. pH values were recorded before and after heating the samples. The samples were deemed fully mixed when no significant difference ((0.3 pH unit) was observed between the two pH measurements. Due to the strong and constant agitation applied, samples were liquid at all times during the titration experiments. As a control, water was also titrated using the same methodology described above; that is, NaOH was added to the
(22) Stryer, L. Biochemistry, 4th ed.; Freeman: New York, 1995. (23) Urry, D. W.; Peng, S. Q.; Parker, T. M.; Gowda, D. C.; Harris, R. D. Angew. Chem., Int. Ed. Engl. 1993, 32, 14401442. (24) Kanicky, J. R.; Poniatowski, A. F.; Mehta, N. R.; Shah, D. O. Langmuir 2000, 16, 172177.

water in order to bring the pH of the solution to 10.5, and then HCl was added stepwise. Sample Preparation. Depending on the desired concentration, the required amount of Fmoc-FF was suspended into 2 mL of HPLC grade water. Sodium hydroxide (0.5 M) was added to the aqueous suspensions of Fmoc-FF until pH 10.5 was reached (55 L for 10 mmol L-1 sample). The samples were vortexed and sonicated for 1 min to fully dissolve the modified peptide. Depending on the concentration and on the target pH, a required volume of dilute hydrochloric acid (0.085 M) was then added dropwise while the solution was vortexed and sonicated until the target pH was obtained. Next, the samples were heated to 75-80 C for 1 min and homogenized. The samples were subsequently cooled and maintained at 4 C for 12 h (overnight) to promote gelation. Reported pH values were those measured after storage. They were found to be identical to the pH values measured before heating within (0.3 units. Fourier Transform Infrared (FTIR) Spectroscopy. Multiple bounce attenuated total reflectance FTIR experiments were undertaken using samples prepared in deuterated water. Spectra were recorded on a Thermo Nicolet 5700 spectrometer equipped with a smart ark trough plate comprising a zinc selenide crystal. The samples were spread directly on the surface of the trough plate. Spectra were acquired in the 4000-400 cm-1 range with a resolution of 4 cm-1 over 128 scans. The deuterated water spectrum was used as background and subtracted from all spectra. Agar Scientific) was placed on 10 L of sample for 30 s. After blotting on Whatman 50 filter paper, the loaded grid was washed in double distilled water for 30 s and blotted. The sample was then stained with 10 L of 2% (w/v) uranyl acetate (centrifugated for 5 min beforehand) for 1 min and blotted for 10 s. Data were collected on a Tecnai 10 transmission electron microscope operating at 100 keV onto Kodak SO-163 films at a calibrated magnification of 43 200. Micrographs were scanned at 1600 dpi using a UMAX 2000 transmission scanner, giving a specimen pixel-1. Data were converted to LINUX level increment of 3.66 A format and were then analyzed using EMAN.25 Using BOXER, a total of 550 straight and nonoverlapping fiber areas were interactively selected and a projection average was created using reference free alignment with a maximum shift value of (4 pixels. Wide Angle X-ray Scattering (WAXS). Wide angle X-ray scattering experiments were conducted using 10 mmol L-1 samples. Wet samples were spread onto glass slides as thin films and allowed to air-dry for 48 h prior to data collection. Experiments were performed on a Philips XPert diffractometer equipped with ). a copper source (wavelength of 1.54 A Mechanical Properties. Mechanical properties were assessed using a strain-controlled rheometer (Bohlin C-CVO) equipped with a Peltier device to control temperature. A parallel-plate geometry was used with a diameter of 40 mm. To ensure the measurements were made in the linear regime, amplitude sweeps were performed and showed no variation in G0 and G00 up to a strain of 1%. The dynamic moduli of the hydrogel were measured
(25) Ludtke, S. J.; Baldwin, P. R.; Chiu, W. J. Struct. Biol. 1999, 128, 8297.

Transmission Electron Microscopy (TEM) and Image Analysis. A glow discharged carbon coated copper grid (400 mesh,

9448 DOI: 10.1021/la900653q

Langmuir 2009, 25(16), 94479453

Tang et al.

Article

Figure 2. Titration curves (pH versus moles of added HCl) of water and Fmoc-FF samples at 0.01, 0.1, 1, 5, and 10 mmol L-1. as a function of frequency in the range 10-2-102 rad s-1 with a strain of 1%. To keep the sample hydrated, a solvent trap was used and the atmosphere within the sample chamber was saturated with water. The experiments were performed at 25 C and repeated at least three times (each time on new samples) to ensure reproducibility.

Figure 3. nHCl (moles of HCl added corrected for moles of HCl needed to titrate water) versus nFmoc-FF (moles of peptide present in the sample) for the 1, 5, and 10 mmol L-1 samples at pH values of 8.9, 7.0, and 4.3. Inset: Degree of ionization, R, derived from the slope of the fitted linear curves (for more details, see the text).

Results and Discussion Hydrogels of aromatic short peptide derivatives were previously formed by sequential pH change,15 dilution from fluorinated solvents,16 as well as enzymatic hydrolysis of corresponding esters.26 Here we focus on the pH dependence of the Fmoc-FF self-assembly process using a modified sample preparation method, described above, in order to accurately control the pH of the samples. This new method enabled us to improve the homogeneity of the samples and to reduce the formation of kinetically trapped aggregates. To investigate the ionization behavior of Fmoc-FF, we performed titration experiments by measuring the variation of the samples pH as a function of added HCl. The experiments were carried out on samples at 0.01, 0.1, 1, 5, and 10 mmol L-1 concentrations. Water was also titrated in the same way as a control (Figure 2). At 0.01 mmol L-1, Fmoc-FF was found to be soluble at all pH values. The titration curve obtained at this concentration was similar to that of water. At 0.1 mmol L-1, Fmoc-FF was found to be soluble at pH values above 7, while at lower pH values the peptide was found to precipitate. The overall pH variation as a function of added HCl was in this case too similar to that of water. For samples prepared at 1 mmol L-1 and above, Fmoc-FF was found to be fully soluble at high pH (g10.5) only. At all concentrations tested, the peptide derivative was fully dissolved at pH 10.5. At this pH, most of the Fmoc-FF molecules are expected to be ionized. When HCl was added to the solutions, the pH gradually dropped for all the samples (Figure 2). Once a pH of 9.2 was reached, the pH of the 5 and 10 mmol L-1 samples was found to increase slightly to 9.5-10.2 and in the case of the 10 mmol L-1 sample to become constant. For the 1 mmol L-1 sample, no increase in pH was observed and a transition was observed at a slightly lower pH of 8.6. As the samples went through this first transition, they became slightly cloudy. Once the first transition was complete, the pH of all samples was found to decrease again with addition of HCl. The samples became more
(26) Das, A. K.; Collins, R. F.; Ulijn, R. V. Small 2008, 4, 279287.

Figure 4. FTIR spectra of Fmoc-FF samples at 10 mmol L-1

prepared in D2O at pH 10.5 starting point of the titration experiment, pH 9.1 below apparent pKa 1, and pH 6.8 and 4.2 above and below apparent pKa 2, respectively.

turbid as the pH decreased. At a pH of 6.2-5.2, a second transition was observed and the pH of the 5 and 10 mmol L-1 samples was found to become constant. As the samples went through this second transition, a white precipitate appeared. For samples with pH values below 5, phase separation occurred with the emergence of a clear liquid phase at the top and a white precipitate at the bottom of the test tube when left at room temperature. As the pH was decreased further, the precipitation and phase separation were found to become more pronounced and rapid. At low pH (<3), all the titration curves were found to merge with the water titration curve. The two transitions observed on the titration curves of the 1, 5, and 10 mmol L-1 samples were more marked as the peptide concentration was increased. They are reminiscent of pKa-type transitions. These two shifted apparent pKa transitions (pKa 1 and pKa 2, Figure 2) are related as we will see below to two marked structural transitions resulting from the peptide self-assembly. As FmocFF is a weak acid, at high pH, we can assume in a first approximation that addition of 1 mol of HCl results in the neutralization of 1 mol of Fmoc-FF-, its conjugated weak base (Figure 1). Assuming that at pH 10.5 (before addition of HCl) all
DOI: 10.1021/la900653q

Langmuir 2009, 25(16), 94479453

9449

Article

Tang et al.

Figure 5. TEM micrographs of Fmoc-FF samples at 10 mmol L-1 (A) at pH 10.5 starting point of the titration experiment; (B) at pH 9.0
below apparent pKa 1 (right-hand side inset: projection average of the fibers, scale bar represents 6 nm), (C) at pH 7.4 above apparent pKa 2, and (D) at pH 4.7 below apparent pKa 2. Scale bars represent 100 nm. Left hand side insets correspond to photographs of the samples showing their macroscopic appearance.

Fmoc-FF molecules are ionized, the peptide degree of ionization, R, can simply be written as R 1nHCl nFmoc-FF 1

were nHCL is the number of moles of HCl added corrected for the number of moles of HCl needed to titrate the water and nFmoc-FF is the number of moles of peptide present in the sample. In Figure 3, nHCl is plotted as a function of nFmoc-FF for the 1, 5, and 10 mmol L-1 samples at three pH values: 8.9, just below the apparent pKa 1 transition, 7 and 4.3, just above and below the apparent pKa 2 transition, respectively. The degrees of ionization derived from the slope of the fitted linear curve are given in the inset of Figure 3. As can be seen for the three samples, R is the same at a fixed pH, suggesting that the same processes are occurring in the concentration range investigated. In order to relate the ionization behavior of Fmoc-FF to its self-assembly properties, we have performed a structural investigation of the samples around four different pHs: 10.5, the starting point of the titration; 9.0, below the apparent pKa 1 transition; 7.4, above and 4.7, below the apparent pKa 2 transition. The samples were prepared as described above. The main difference compared to the sample used for the titration experiments is that, in this case after homogeneization, the samples were stored a 4 C overnight to promote gelation. First, we discuss the 10 mmol L-1 samples. At pH 10.5, above the shifted apparent pKa 1, Fmoc-FF was solubilized by addition of NaOH. As can be seen from Figure 4, the FTIR spectrum of the sample at pH 10.5 showed a barely detectable peak at 1625 cm-1, which is usually characteristic of the presence of -sheet structures.27 At this same pH, TEM micrographs showed a small population of fibers (Figure 5A) compared to
(27) Barth, A.; Zscherp, C. Q. Rev. Biophys. 2002, 35, 369430.

Figure 6. WAXS spectra of Fmoc-FF samples at 10 mmol L-1

dried at pH 8.8 below apparent pKa 1 and pH 7.6 and 3.6 above and below apparent pKa 2, respectively.

what was observed at lower pH values as will be discussed later. At this high pH, most of the Fmoc-FF molecules are expected to be ionized, and therefore, due to the electrostatic repulsion between peptides, their self-assembly is thought not to be favored. When the pH was reduced to 9.0 (just below the apparent pKa 1 transition), a weak, translucent, self-supporting hydrogel was obtained. The infrared spectrum of the sample displayed a strong peak at 1625 cm-1 consistent with the presence of -sheet structures and a second one at 1687 cm-1 characteristic of an antiparallel arrangement of the -sheets (Figure 4).27 It should be noted that although the positions of these two peaks are typical to an antiparallel arrangement of the -sheets, their relative intensity is not. For ideal antiparallel -sheet arrangements, usually encountered in longer peptide sequences and proteins, the band
Langmuir 2009, 25(16), 94479453

9450 DOI: 10.1021/la900653q

Tang et al.

Article

Figure 7. (A) Cross sections of the fibril previously described in ref 19 corresponding to four -sheets forming a square cross section that rotates around a central axis. As a consequence, when viewed down the long section, it appears to give a cylinder. Slices 1.5 nm apart are rotated 45 to each other. (B) Schematic representation of a view down the length of a fiber (top) with 1 and 2 (left) representing two slices 1.5 nm apart rotated by 45 through the depicted structure showing flat and ridged surfaces, and fiber projection map (right) obtained from the TEM micrographs showing the parallel arrangement of two fibrils. The light and darker regions correspond to the alternation of flat (light) and ridged (dark) surfaces. Scale bar represents 6 nm. (For more details, see the text.)

at 1687 cm-1 is expected to be much weaker.28,29 The exact origin of this effect is not yet fully understood. It is thought that our peptide sequence being short (two amino acids), steric, and/or other constraints results in the antiparallel -sheet arrangement not being ideal. It should be noted that IR band intensities are very sensitive of small changing in bonds environment. The presence of these two peaks clearly indicates that at this stage self-assembly of Fmoc-FF has occurred. This observation was confirmed by TEM (Figure 5B). Analysis of the micrograph revealed the presence of an extended entangled network composed of mainly (80%) thin flexible fibers with a width of 6.0 ( 0.4 nm. Image analysis of the averaged fiber projection map (inset of Figure 5B) revealed a thin high-contrast line running along the middle of the fibers parallel to the long axis. This feature suggests that the fibers either correspond to a hollow tube of 6.0 nm diameter or are formed by the lateral association of two smaller fibrils, each of 3.0 nm in diameter. The latter interpretation is favored; indeed, the accumulation of stain down the center of a hollow fiber of this size would be difficult with the particular staining technique used here. From the TEM micrograph, the diameter of the central hole would be less than 0.5 nm. Keeping in mind that the fibers are first deposited on the grid and then stained, the even staining of the central part of the tube over several micrometers, as observed here, would require considerable capillary action and would likely result in incomplete staining. The fact that the stain is even along the full length of the fibers is more consistent with a surface coating staining action caused by the accumulation of stain in the cavity at the interface between two fibrils. Each fibril would then correspond to the cylindrical structure described previously.19 The fibril diameter was confirmed by WAXS. As can be seen in Figure 6, a strong reflection at 2.5 nm-1 is observed corresponding to a repeat distance of 2.6 nm in good agreement with the fibril size estimated by TEM. The exact reason and origin of the association of these fibrils in pairs has not yet been elucidated. It is also of interest to note that along the length of the paired fibrils a clear 3.0 nm periodic repeat can be observed (inset of Figure 5B). The cylindrical nature of the model proposed by Smith et al.19 is due to four -sheets forming a square
(28) Schneider, J. P.; Pochan, D. J.; Ozbas, B.; Rajagopal, K.; Pakstis, L.; Kretsinger, J. J. Am. Chem. Soc. 2002, 124, 1503015037. (29) Rapaport, H.; Grisaru, H.; Silberstein, T. Adv. Funct. Mater. 2008, 18, 28892896.

Figure 8. Mechanical spectra of Fmoc-FF samples at 10 mmol L-1 at pH 9.0 below apparent pKa 1 and pH 7.6 above apparent pKa 2.

cross section, with each face corresponding to one -sheet, rotating around a central axis (Figure 7A). Therefore, the pitch of the helical structure would be 12 nm, corresponding to four times the periodic repeat unit observed by TEM. As a consequence, an alternation of flat to ridged to flat surfaces from the square cross section (Figure 7B) with a periodicity of 3.0 nm should be observed. As revealed in the TEM micrograph, the alternation of bright and dark patches along the long axis of the fibers would then result from the difference in stain densities between the flat and ridged surfaces, with the ridged surfaces appearing darker as they collect more stain. The gel-like nature of the sample at this pH was confirmed by dynamic rheology. As can be seen from Figure 8 at pH 9.0, the elastic modulus (G0 ) was higher than the viscous modulus (G00 ), indicating the material was elastic rather than viscous. Both G0 and G00 were weakly dependent on frequency between 10-2 and 102 rad s-1. This behavior is characteristic of entangled polymer networks and is in good agreement with the type of network topology revealed by TEM. G0 values were in the range of 1-10 Pa, which confirms the hydrogels were weak compared to the Fmoc-FF hydrogels previously reported (104 Pa).16,19 We hypothesize that the high modulus observed in our previous
DOI: 10.1021/la900653q

Langmuir 2009, 25(16), 94479453

9451

Article

Tang et al.

Figure 9. Proposed self-assembly mechanism of Fmoc-FF from high to low pH above the critical gelation concentration as a function of the peptide degree of ionization, R.

work results from the presence of kinetically trapped aggregates which reinforce the mechanical properties of the material. G0 values of different orders of magnitude have been reported for peptide based hydrogels depending on the kinetics of the selfassembly process.30,31 By increasing the ionic strength of solutions used, Pochan et al.30 were able to accelerate the self-assembly kinetics of their peptides. This resulted in the formation of networks with higher cross-link density leading to stiffer hydrogels. G0 values were found to increase from 100 Pa to 3000 Pa when 20 mM and 400 mM of salt solutions, respectively, were used. Investigating enzyme-triggered -peptide hydrogels, Xu et al.31 have also shown that the morphology of the network generated depends strongly on the kinetics of hydrogelation. Using low concentrations of enzyme, thin, uniform fibrils were formed slowly resulting in relatively weak hydrogels with G0 of 300 Pa. On the other hand, when higher concentrations of enzyme were used, the formation of nanofibrils was found to be almost instantaneous, leading to the formation of heterogeneous bundles of fibers resulting in hydrogels with higher G0 of 4000 Pa. In our case, heating the samples at 75-80 C increases their homogeneity and storing of the samples at 4 C results in a very slow hydrogelation dynamic producing more homogeneous hydrogels with lower moduli. As clearly shown, the self-assembly of the peptide occurs during the apparent pKa 1 transition. The neutralization of Fmoc-FF- is thought to favor self-assembly as the electrostatic repulsion between peptides is decreased. As the Fmoc-FF- molecules are neutralized, they self-assemble to form the cylindrical structures described above, which maintains the samples pH constant (Figure 9). The samples pH remains constant until the self-assembly process is completed, at which point the pH starts to decrease again. At the end of apparent pKa 1 transition, only 34% (i.e., R = 0.66) of the Fmoc-FF- molecules are thought to be neutralized, suggesting that the self-assembled fibers incorporate a significant amount of still ionized peptide. In our model,19 most of the terminal acid groups are located on the surface of the fibril. We therefore expect the surface of the fibrils to be negatively charged. At pH 6.8, just above the apparent pKa 2 transition, the sample was turbid and a thick viscous solution rather than a weak gel was
(30) Ozbas, B.; Kretsinger, J.; Rajagopal, K.; Schneider, J. P.; Pochan, D. J. Macromolecules 2004, 37, 73317337. (31) Yang, Z.; Liang, G.; Ma, M.; Gao, Y.; Xu, B. Small 2007, 3, 558562.

obtained. As can be seen from Figure 8 at pH 7.6, G0 was found to be lower than G00 until a crossover of the moduli was observed beyond 10 rad s-1, confirming that when the pH is decreased the materials become viscous. Both G0 and G00 showed a strong frequency dependence between 10-2 and 102 rad s-1. Such rheological behavior is characteristic of liquid-like materials. From the FTIR data, it is apparent that -sheets are still present at pH 6.8 (Figure 4). The area of the peaks at 1625 and 1687 cm-1 at pH 9.1 and 6.8 are comparable, suggesting the presence of a similar overall amount of -sheets at both pHs. TEM micrographs at pH 7.4 revealed the presence of large, flat ribbons resulting from the lateral association of many fibrils (Figure 5C) similar to the ones observed by Smith et al.19 This was confirmed by WAXS. The same pattern as reported in our previous work was obtained at this pH (Figure 6), showing a series of scattering peaks corresponding to the lateral assembly of the fibrils in large flat ribbons. A peak at 0.43 nm can also be observed corresponding to the spacing between -strands. This spacing is lower than the one usually observed for ideal -sheet arrangements: 0.47 to 0.48 nm; although such a low spacing was reported for silk hydrogels.32 This low spacing could suggest, in agreement with the infrared results, that the -sheet arrangement in our system is probably not ideal. For further details on the interpretation of the scattering peaks observed, we refer the reader to ref 19. As pH decreases from 9.1 to 6.8, the ionization degree of Fmoc-FF decreases from 0.66 to 0.55 (Figure 3). As a consequence, it is thought that the overall surface charge of the fibers decreases, allowing them to self-assemble further and form large flat ribbons through lateral hydrophobic interactions (Figure 9). The reason of this apparent one-dimensional aggregation of the fibrils is not clear but suggests the presence of specific directional interfiber interactions or energetic limitations on the reordering of the fibers into the ribbons. These structures are expected to be highly rigid and would therefore not entangle easily and form hydrogels but form liquidlike nematic structures. At pH 4.7, below the shifted apparent pKa 2, precipitate formation was observed. The TEM micrograph revealed the presence at this pH of rigid rodlike elongated objects. The lateral dimension of these objects is similar to the one of the flat ribbon observed at higher pH, suggesting that they probably derive from
(32) Kim, U.-J.; Park, J.; Li, C.; Jin, H.-J.; Valluzzi, R.; Kaplan, D. L. Biomacromolecules, 2004, 5, 786792.

9452 DOI: 10.1021/la900653q

Langmuir 2009, 25(16), 94479453

Tang et al.

Article

them. It is thought that during the apparent pKa 2 transition further assembly of the ribbons occurs as the ionization degree of the peptide decreases further from 0.55 to 0.24. As a result, these larger rigid rodlike structures, probably corresponding to aggregated ribbons, are formed and precipitate out of solution. As HCl is added, further aggregation and precipitation of the ribbons occurs which maintains the samples pH constant (Figure 9). Once the process is complete, the pH of the sample decreases again. Below the apparent pKa 2 transition, a fully separated two-phase system is obtained with a clear liquid phase, corresponding to water, and a solid phase, corresponding to the precipitated rodlike objects containing the peptides, coexisting. WAXS and FTIR experiments were also performed at pH 4.8. The WAXS pattern showed the same scattering peaks as observed at pH 7.6, but the peak intensities were much lower, suggesting a loss of structure. This was confirmed by FTIR. The peaks corresponding to antiparallel -sheet are still observed, but their intensity is much weaker. Additional peaks are on the other hand observed in the amide II region at 1694 and 1649 cm-1. The peak at 1649 cm-1 has been assigned in the literature to peptides adopting a random coil conformation.27 This would suggest that as the ribbons aggregate and precipitate, the peptides lose their antiparallel -sheet organization. The exact nature of these precipitated aggregated ribbons and their internal structure has not been yet elucidated. As mentioned previously, the system phase separates at low pH (<3) with a liquid phase corresponding mainly to water (Figure 9). As a result, at these low pH values, all the titration curves are found to merge with the water titration curve. Similar results as the one described above for the 10 mmol L-1 sample were obtained for the 1 and 5 mmol L-1 samples. The only major difference was that while for the 5 mmol L-1 sample a weak gel was also obtained just below the apparent pKa 1 transition, for the 1 mmol L-1 sample a viscous solution was obtained, suggesting that for this system the critical gelation concentration lies between 1 and 5 mmol L-1.

2.2 pH units (Figure 2). Although the modified peptides selforganize into antiparallel -sheets within the pH range investigated, they show different structural behaviors depending on the pH. At high pH, most of the molecules are ionized and in solution; therefore, almost no self-assembly is observed. The first shifted apparent pKa transition, pKa 1, at pH 10.2-9.5 corresponds to the self-assembly of both protonated and nonprotonated molecules into paired fibrils consisting of antiparallel -sheets. The charges associated with the ionization of the terminal carboxylic groups of the peptides being located on the surface of the fibrils, the paired fibers are negatively charged. Above the critical gelation concentration, the fibers entangle, resulting in the formation of a three-dimensional network and a hydrogel. The gradual neutralization of the peptide molecules between pH 9.5 and 6.2 results in the decrease of the fiber surface charge, allowing them to self-assemble laterally through hydrophobic interactions and form large rigid ribbons. The parameters that control the ribbons width limit have not been yet elucidated. A second shifted apparent pKa transition, pKa 2, is observed at pH 6.2-5.2. This transition was shown to be related to further aggregation of the ribbons, triggered by further neutralization of the peptide, and their precipitation out of solution. Our results suggest that during precipitation there is a loss of the antiparallel -sheet structure adopted by the peptides. Finally, below the apparent pKa 2 transition, almost all Fmoc-FF- molecules are neutralized and phase separation occurs. Overall, this study provides further understanding of the self-assembly mechanism of aromatic short peptide derivatives. The methodology used for this study may be exploited to understand the self-assembly behavior of similar systems and to design rules for the preparation of hydrogel scaffolds for biomedical and nanotechnology applications. Acknowledgment. The authors gratefully acknowledge the EPSRC and Leverhulme Trust for their financial support as well as Judith Shackleton and Paula Crook from the University of Manchester for their assistance with the WAXS and FTIR measurements.

Conclusion
We have shown that Fmoc-FF molecules self-assemble upon lowering pH (Figure 9) due to dramatic pKa shifts of 6.4 and

Langmuir 2009, 25(16), 94479453

DOI: 10.1021/la900653q

9453