Jaclyn
Raeburn
a,
Cristina
Mendoza-Cuenca
b,
Beatrice N.
Cattoz
c,
Marc A.
Little
a,
Ann E.
Terry
d,
Andre
Zamith Cardoso
a,
Peter C.
Griffiths
c and
Dave J.
Adams
*a
aDepartment of Chemistry, University of Liverpool, Crown Street, Liverpool, L69 7ZD, UK. E-mail: d.j.adams@liverpool.ac.uk
bDepartment of Chemical Engineering, Faculty of Sciences, 18071 Granada, Spain
cSchool of Life Sciences, University of Sussex, Chichester II Building, Falmer, BN1 9QG, UK
dRutherford Appleton Laboratory, Science and Technology Facilities Council, Didcot, Oxfordshire OX11 0QX, UK
First published on 8th December 2014
Gels can be formed by dissolving Fmoc–diphenylalanine (Fmoc–PhePhe or FmocFF) in an organic solvent and adding water. We show here that the choice and amount of organic solvent allows the rheological properties of the gel to be tuned. The differences in properties arise from the microstructure of the fibre network formed. The organic solvent can then be removed post-gelation, without significant changes in the rheological properties. Gels formed using acetone are meta-stable and crystals of FmocFF suitable for X-ray diffraction can be collected from this gel.
Mahler et al. were the first to show that gels can also be formed using FmocFF by dissolving FmocFF in an appropriate water-miscible solvent,4 before diluting with water to drive the hydrophobic FmocFF to self-assemble into one-dimensional structures, leading to hydrogelation.2,13,14 Using this method, we recently showed that the final ratio of dimethylsulfoxide (DMSO) to water (expressed as the volume fraction of DMSO, ϕDMSO) strongly affects the mechanical properties of gels formed using FmocFF5 and another related gelator, FmocLG.15 The rheological properties were affected by the ϕDMSO, as was the ability of the gels to recover after shear. We also showed through the use of various buffers that the pH plays a pivotal role in controlling the mechanical properties of these FmocFF gels. Elsewhere, Gazit's group have formed gels from FmocFF using different compositions of DMSO, acetone, or hexafluoroisopropanol (HFIP) with water and reported mechanical properties, with the storage modulus (G′) being typically in the kPa region.4,10
For this solvent-triggered approach, some solvents may be preferred over others for use in biological systems due to any associated hazards with the solvent in the gel. Whilst pH-triggered gels consist of water and no solvent, which might be more attractive for some applications, the stability of “solvent-triggered” FmocFF gels over a wider range of pH conditions5 plus the fact that no mixing is required5 (thus increasing the homogeneity and reproducibility of these gels) makes the latter class of gels extremely interesting. Having the ability to potentially choose a desired solvent to tailor both the gel properties and suitability for different applications could be an extremely powerful approach. In this paper, we investigate the effects of specific solvents on both the gelation ability of FmocFF with different volume fractions of the solvent (ϕsolvent). We show how the choice of solvent influences the final mechanical properties of the gels formed and show that the solvent can be removed post-gelation without a significant impact on the rheological properties.
There is clearly therefore a phase separation event that leads to gelation. Using the other solvents, the duration of this sphere to fibre transition varies (Fig. S1, ESI†). However, it is clear by comparison of the microscopy in ethanol–water and from data collected for DMSO–water by Dudukovic et al. that the self-assembly process is similar and solvent independent.17 Slight differences in turbidity for the gels formed using different solvents (Fig. 3e) is more likely a result of variations in the sizes of spherulitic structures/fibrous structures rather than the process itself. We note that turbidity changes and the formation of spherulitic structures preceding gelation has never been reported for FmocFF gels prepared using pH triggered methods, indicating a fundamentally different self-assembly process occurs with these triggers.
Gelation occurred across slightly different ϕsolvent ranges, depending on the solvent used. At a final FmocFF concentration of 5 mg mL−1, gelation occurred from a ϕsolvent of 0.05 to 0.6 for DMSO and HFIP, but there were solubility issues at a ϕsolvent of 0.05 when ethanol and acetone were used. The higher volatility of these solvents compared to DMSO and HFIP may play a role in this due to the low volume of solvent required here. Otherwise, there appears to be only small deviations in the gelation ability of FmocFF using the four solvents despite their different hydrogen bonding ability, dielectric constant, etc., although no gel formed at a ϕacetone of 0.6 unlike for the other solvent systems.
We used small angle neutron scattering (SANS) to probe the fibrous structures formed. The scattering from a series of gels prepared from the different solvents at a ϕsolvent of 0.3 is presented in Fig. 2.
The data have been normalised for subtle differences in incoherent background in Fig. 2a, and separated by factors of 3 in Fig. 2b for clarity. The data fall into two pairs, though there is no major difference between the four datasets. There are also few distinguishing features in the data e.g. discontinuities at high Q. This implies that the scattering objects are not core–shell structures. Over the Q-range examined here, the data are most sensitive to the cross-section of the scattering objects. The decay rates follow the order acetone ∼ ethanol > DMSO ∼ HFIP, indicating a marginally larger cross-section of the structures formed in ethanol and acetone. The absolute intensities of all four cases are also similar suggesting a comparable number of scatterers per unit volume, and molecules per unit length along the fibril. The data were fitted in terms of the Kholodenko–Dirac worm as previously reported,15 which models the scattering structure as a series of semi-flexible rods of a given cross-section. The model smoothly interpolates between a rigid-rod Q−1 decay at low Q, into a decay consistent with a more flexible structure with characteristic Q−2 decay, and ultimately, at high Q, into a Q−4 dependence associated with the highly curved surface of the cross-section of the rod (see inset to Fig. 3b). In essence, Rcross-section = 34 (±3) Å for DMSO, 38 (±3) Å for ethanol, 44 (±3) Å for acetone, and 32 (±3) Å for HFIP. Hence, the primary fibres formed from each solvent are very similar. We note that the values obtained are in agreement with recent data for FmocFF fibres formed from DMSO.17
Despite the similar phase separation during the self-assembly process, comparison of the final gel microstructures shows that there are differences in the fibrous network depending on the solvent used. This is exemplified for FmocFF gels prepared at a final ϕsolvent of 0.3 in Fig. 3. Using DMSO (Fig. 3a), a largely homogeneous network of fibres is formed, where the fibres appear to be densely packed into larger domains of fibres, interconnected by slightly less dense regions of fibres (similar to that previously reported5). This microstructure is much more pronounced for the corresponding gels containing HFIP (Fig. 3d), where defined fibrous clusters are visible. FmocFF gels prepared using ethanol (Fig. 3b) appeared to the have the most homogeneous fibril network at the micrometer length scale. A dense network of thin entangled fibres was present throughout the gels. There are indications of similar fibrous domains visible in gels prepared from acetone (Fig. 3c), but there are also spherulitic structures. The residual spherulites in the acetone–water gels were crystalline, as can be determined by both polarised microscopy (Fig. S2, ESI†) and X-ray crystallography. We note that the gels formed from other solvents showed no such structures under polarised microscopy or X-ray diffraction. The crystals developed from acetone after gelation, and are not simply due to undissolved FmocFF. No crystals were observed in the ethanol gels. Since ethanol is also volatile like acetone, the lack of crystallisation in the ethanol gels suggests that crystallisation does not simply occur as a result of the solvent leaving the system.
A crystal of FmocFF, selected from gel phase, was suitable for single crystal X-ray diffraction studies using synchrotron radiation despite the extremely small size of the needle shaped fragment. In the single crystal structure FmocFF molecules are stacked parallel along the crystallographic b axis, Fig. 4a. The distance between FmocFF molecules stacked along b correlates to one unit cell length, or 5.0 Å. Hydrogen bonding interactions, Fig. 4b, and weak offset π–π interactions, at a distance of ∼5 Å, are evident between these FmocFF molecules. The single crystal structure of FmocFF was refined in the chiral monoclinic space group C2, with the asymmetric unit comprising two FmocFF molecules which interact through edge-to-face π–π bonding interactions. These two FmocFF molecules have the opposite relative orientation along b. Additional FmocFF molecules in the crystal lattice are related by twofold rotation axes; these axes are located centrally between four stacks of FmocFF molecules, Fig. 4c. Diffuse solvent, found in small lattice voids, was modelled as H2O.
Powder X-ray diffraction (pXRD) data was recorded on dried xerogel material formed using acetone (Fig. S5, ESI†). From the pXRD pattern reflections at 15, 12.6, 4.9, 4.2, 3.8 and 3.4 Å were measured. Comparison between the experimental pXRD pattern and that simulated from the single crystal structure reveal a close comparison between peak positions. From the experimental pXRD pattern, there is also a clear contribution from amorphous material. WAXS data has been reported for a pH-triggered gel, where reflections at 26, 16 and 4.3 Å were found.6 Reflections at approximately 10 and 4.7 Å have been interpreted as being due to the formation of β-sheet structures, which are thought to be the basis of the assembly of FmocFF,6 with a reflection at 16 Å thought to arise from stacking between Fmoc groups. SAED data on single fibres from a film of FmocFF, also revealed reflections at 4.8 and 23 Å, which were assigned to the fibre axis and a perpendicular axis respectively, with the aid of TEM imaging.18 Interestingly, these two reflections are very similar to the a- and b-axis lattice parameters obtained here for the single crystal data, 5.0 and 22.8 Å. Here, in line with our previous data, we ascribe the amorphous material observed by pXRD to the gel phase.19 It is not clear whether the crystalline material of FmocFF isolated during this study is representative of the proposed fibrous structure. We include the data here for completeness, but believe interpreting the single crystal structure in this instance is not informative for determining the packing in the fibrous structure in line with our previous data.19 Nonetheless, this data represents the first single crystal X-ray diffraction study for this important gelator.
We have reported previously on the importance of ϕsolvent for FmocFF5 and FmocLG15 gels prepared at different ϕDMSO. Here, we find that the rheological properties can be tuned by judicious choice of both solvent and solvent composition (strain and frequency sweep data for ϕsolvent of 0.3 are shown in Fig. S6 and S7, ESI†). For gels prepared using DMSO, ethanol, or acetone, somewhat similar mechanical properties were recorded between ∼ϕsolvent of 0.2 and 0.4, with the mechanical properties slightly lower for gels prepared at a ϕsolvent on either side of this range (Fig. 5). However, for HFIP, all ϕHFIP resulted in storage moduli (G′) < 10 kPa; except for a ϕHFIP of 0.05 which resulted in significantly mechanically stronger gels with a G′ of ∼26 kPa. In almost all the gels studied, G′ exceeded G′′ by a factor of >7, however in some cases this factor was only ∼4/5 (Fig. 5, showing tanδ (G′′/G′)). The critical strain at which the gelled state yields did not directly correlate with either the solvent used or the ϕsolvent. The critical strains varied between ∼2 and 20% for 5 mg mL−1 gels prepared from all solvents. This could be due to differences in the number of crosslink points or degree of association of the fibres, or even in the brittleness of the fibres themselves. The rheology for all gels was measured 24 hours after water addition to trigger gelation. These gels form quickly (see above), but the rheology does develop over time.16 However, by 24 hours, the rheology has essentially come to a plateau value (example data is shown in Fig. S8, ESI†).
The ability of the gels to recover after shear can also be controlled by solvent choice (Fig. 6). This is an important property for injectable gels, since ideally here the gel would recover quickly and completely back to its original state after flowing through a needle.20 For DMSO gels (ϕDMSO = 0.3), 100% of the mechanical strength is recovered upon cessation of high shear (Fig. 6a), even after several cycles. We note here that we intended to probe the ability of the gels to recover from a short period of high shear that breaks down the gel rather than the absolute value of G′. As a result, in this experiment, set up shear cycles of constant duration and as a result the modulus does not return to a constant value.
Previously, we drew a correlation between the gel microstructure and the ability of the gel FmocLG to recovery from shear.15 We showed that gels formed by a nucleation and growth process led to domains of fibres such as those found here (Fig. 3), which could percolate back to a gel after shear cessation and hence recover more successfully than those gels that were found to be consisting of a more densely packed network such as that found for ethanol gels. Yan et al. have also drawn similar conclusions for their injectable oligopeptide (MAX1 and MAX8) gels.21 From the microscopy, we would therefore expect that gels formed from HFIP and ethanol would recover well, whilst those formed from acetone would recover poorly. For gels formed from HFIP and acetone, this expectation holds. Gels formed using acetone only recovered 30% of their original G′. Gels formed using HFIP, however, could recover 95% of the original G′ at a ϕHFIP of 0.3. However, despite the similar microstructure observed by confocal microscopy (Fig. 3), gels formed using ethanol (ϕethanol 0.3) exhibit very poor recovery. These gels only regain 26% of their original mechanical properties after being subjected to high shear. It seems therefore that a simple comparison of the microscopy is not always a good predictor of mechanical properties.
It is clear from the above that there are distinct differences in gel properties depending on the choice of solvent. It is not entirely clear what influences these fluctuations at both the microstructural and bulk scales of FmocFF gels (and other systems). Considering the initial phase separation process seems very similar, presumably these differences appear to be at the latter stages of the gelation process i.e. fibril growth, association and beyond. To investigate this further, we investigated the scaling relationships between the concentration of FmocFF and the final mechanical properties i.e. G′ of the gel such that G′ ∝ Cx. A power law behaviour has been reported for many gelling systems, with x being indicative of the network type. Typically for colloidal gels, x is between 3 and 6.22 For cross-linked polymer networks, x is around 2.5.23–25 Values of x of around 1.4 have been reported for entangled semiflexible networks.24,26,27
Fig. 7 shows the scaling relationships drawn for FmocFF gels prepared with the four solvents at a ϕsolvent 0.3. The exponents depend on the solvent, which again shows that the network is tuneable by solvent. In all cases, the exponents are much lower than predicted by Mackintosh theory. This suggests that, over the concentration range studied here, the networks observed are more reminiscent of relatively dilute F-actin systems, which formed entangled chain networks rather than densely crosslinked networks.24,26 We previously found similar values for Fmoc–dipeptide gels formed using a pH switch.28
Concentrating on gels formed using DMSO, we find that the exponent derived was dependent on the ϕDMSO used to prepare the gels (Fig. 7e and S9, ESI†). Exponents ranging from 3.0 to 1.8 are derived for ϕDMSO between 0.05 and 0.6, inherently indicating a change in the network formed when the ϕDMSO is altered. This suggests that different networks are formed at different ϕDMSO. This could arise from differences in the kinetic process of gelation in these systems when the ϕDMSO is changed. Such differences are suggested by the turbidity data upon addition of water to the FmocFF–DMSO solution (Fig. S1b, ESI†). There are differences in the final turbidity change and in the duration of the turbidity events, depending on the ϕDMSO. We interpret this as the kinetics of the assembly process differing, leading to different types of fibre networks. Recently, Dudukovic and Zukoski reported a scaling relationship of 2.5 for FmocFF gels prepared using DMSO.29 This is much higher than the values we report here and more fitting with the model proposed by Mackintosh.29 The reasons for the difference are not clear, but may be due to slight changes in the sample preparation protocol, final pH etc. The pH of the gels was not reported by Dudukovic and Zukoski. Here, we find that the pH of the gels is between 4.3 and 4.7 for all solvents and ratios apart from for HFIP, where the pH is slightly lower (∼4).
For some applications, the presence of the solvent could be undesirable. For example, large solvent volumes could be deleterious for cell viability and biocompatibility. However, the tunability afforded using this solvent switch approach is useful. Hence, we considered the possibility of removing the solvent from the gel network post-gelation. FmocFF gels were prepared as described above, utilising a final gelator concentration of 5 mg mL−1 and a ϕDMSO of 0.3. We used D2O here as opposed to H2O to facilitate analysis of DMSO removal. Post-gelation, D2O was added to the top of the gels. This D2O was replaced regularly. The concentration of DMSO in the washing D2O was quantified using FT-IR by observation of the sulfoxide peak in these samples at 1020 cm−1 (Fig. 8a). This data shows that the DMSO can be removed via washing. ϕDMSO decreased to <0.1 from the initial value of 0.3 after just four washings (Fig. 8b). The ϕDMSO of the gels continued to decrease; after ten washings, the DMSO was almost completely removed (ϕDMSO of approximately 0.01).
A key question is whether the removal of the solvent affects the mechanical properties. Fig. 8c shows that the G′ slightly decreased from ∼22 kPa to ∼17 kPa after 10 washings. Hence, the solvent can be exchanged for D2O with little perturbation of the hydrogel network. The critical strains were also unaffected by the removal of DMSO from the system (Fig. S10, ESI†), with a critical strain of ∼8% being recorded after each washing.
The ability of the gels to recover from shear after the DMSO has been washed out slightly deteriorates, with the recovery going from 100% before washing to 70% after one washing step (Fig. 8d). However, there is no further decrease after this first washing step. This decreased ability to recover remains constant after successive washings. This diminished ability to recover at lower ϕDMSO is consistent with gels of similar ϕDMSO which have not undergone any washing process.5 Gels with almost no DMSO present at all still possess mechanical properties similar to that of the originally prepared FmocFF gels with ϕDMSO of ∼0.3. Hence, desired mechanical properties can potentially be targeted and the hydrogel be altered to contain almost 100% water with the mechanical properties being almost unperturbed in the process. This could be advantageous for systems where toxic and undesirable solvents may be required to solubilise the gelator. Potentially, any unwanted solvent can be removed and could, thus, increase the biocompatibility of the hydrogel material without compromising the material properties.
Tuning the properties of FmocFF gels by changing the solvent used is a very attractive feature. However, for many applications the presence of the organic solvent may be an issue. Removing the solvent by washing is possible as we have shown here, which expands the potential utility and opens up the possibility to replace the solvent portion of the hydrogel with potentially any solvent that may be required for an application. Even considerably harsh solvents – which may be necessary to solubilise a particular gelator – could be used due to the ability to remove the solvent from the hydrogel network.
We note that an exotherm was observed when water was added to the gelator solution; this increased as the ϕsolvent increased. This exotherm was more pronounced for DMSO–water systems30,31 – an exotherm of 19 °C was noted for a gel containing a ϕDMSO of 0.6. For ethanol, acetone and HFIP systems, exotherms were all ∼6 °C or less, regardless of the presence of gelator or not. Homogeneous hydrogels were formed at each ϕsolvent for all solvents. Gelation occurred within minutes. All gels were left to stand overnight in sealed sample tubes before any characterisation of their final properties and were all therefore at room temperature when analysed.
Crystal data for FmocFF·(H2O)0.875 crystallised from a H2O–acetone gel phase; CCDC entry: 1027570.† Formula C33H31.75N2O5.875; M = 550.35 g mol−1; monoclinic space group C2, colourless needle shaped crystal; a = 52.27(6), b = 5.010(6), c = 22.83(3) Å; β = 105.49(3)°; V = 5761(11) Å3; ρ = 1.269 g cm−3; μ = 0.082 mm−3; F(000) = 2326; crystal size = 0.070 × 0.009 × 0.008 mm3; T = 100(2) K; 11789 reflections measured (0.78 < Θ < 20.00°), 5782 unique (Rint = 0.1630), 2644 (I > 2σ(I)); R1 = 0.1027 for observed and R1 = 0.2274 for all reflections; wR2 = 0.3134 for all reflections; max/min residual electron density = 0.484 and −0.360 e Å−3; data/restraints/parameters = 5782/396/503; GOF = 1.008. The asymmetric unit for FmocFF·(H2O)0.875 comprises two complete FmocFF molecules, one fully occupied H2O molecule and one partially occupied H2O molecule. Single crystal samples of FmocFF were extremely small and very weakly diffracting, synchrotron radiation was therefore essential for structure solution. A resolution limit of 1 Å was applied during integration. Due to the limited number of reflections only the main residue atoms were refined anisotropically. Atomic displacement parameters were restrained during refinement (SIMU and DELU in SHELX). In addition the geometry of one aromatic ring was constrained during refinement (AFIX 66 in SHELX). Diffuse electron density was modeled as H2O. These H2O molecules were refined without riding proton atoms however the appropriate number of protons were included in the refined formula unit. For FmocFF, data quality was not of sufficient quality to determine proton positions these were therefore place is geometrically estimated positions using the riding model. For a displacement ellipsoid plot of the asymmetric unit, see Fig. S3.†
Frequency scans were performed from 1 rad s−1 to 100 rad s−1 under a strain of 0.5%. The shear moduli (storage modulus G′ and loss modulus G′′) were measured at a frequency of 10 rad s−1. All shear moduli measured were within the linear viscoelastic (LVE) region for the gels measured.
For the recovery test experiments, time sweep with a constant frequency of 10 rad s−1 and strain of 0.5% was first performed for 200 seconds, followed by higher strain of 300% for 60 seconds to totally destroy the gel to a liquid state. Restoration of the gel was monitored in the subsequent time sweep (with the frequency of 10 rad s−1 and strain of 0.5%) for 200 seconds again. The shear-recovery cycles were performed for 5 times for the same sample to check the reproducibility. The recovery ratios were calculated by the ratios of the average storage modulus (G′) after restoration with the original storage modulus obtained in the first step of time sweep.
All scattering data were (a) normalized for the sample transmission, (b) background corrected using an empty quartz cell or one filled with the appropriate solvent (this also removes the inherent instrumental background arising from vacuum windows etc.) and (c) corrected for the linearity and efficiency of the detector response using the instrument-specific software package. The data were put onto an absolute scale by reference to the scattering from a partially deuterated polystyrene blend. Data were fitted using the Kholodenko–Dirac worm-like chain model, which analyses the data as a Gaussian distribution of m connected cylindrical elements of statistical length l and radius Rax, such that the contour length, L, is L = ml. This model therefore interpolates between the expected Q−1 dependence for the rod-like character of the cylindrical elements, the Q−2 associated with the cross-section of the cylinder and a limiting Q−4 associated with the globular nature over large distances.
Footnote |
† Electronic supplementary information (ESI) available: Full experimental details, full rheological data, single crystal (monoclinic C2, a = 52.27(6), b = 5.010(6), c = 22.83(3) Å; β 105.49(3)°; V = 5761(11) Å3) and pXRD data. CCDC 1027570. For ESI and crystallographic data in CIF or other electronic format see DOI: 10.1039/c4sm02256d |
This journal is © The Royal Society of Chemistry 2015 |