C-Terminal Residue of Ultrashort Peptides Impacts on Molecular Self-Assembly, Hydrogelation, and Interaction with Small-Molecule Drugs

Single molecular changes on a tripeptide can have dramatic effects on their self-assembly and hydrogelation. Herein, we explore C-terminal residue variation on two consistent ultrashort peptide backbones, i.e. acetylated-Leu-Ile-Val-Ala-Gly-Xaa and acetylated-Ile-Val-Xaa (Xaa = His, Arg, Asn). The objective of this study is to identify candidates that can form hydrogels for small-molecule drug (SMD) delivery. Haemolysis and cytotoxicity (with human adipose-derived mesenchymal stem cells) assays showed that the new soluble peptides (Xaa = His, Arg) are cytocompatible. Gelation studies showed that all but acetylated-Ile-Val-Arg could gel under physiological conditions. Longer peptidic backbones drive self-assembly more effectively as reflected in field emission scanning electron microscopy (FESEM) and circular dichroism spectroscopy studies. Rheological studies revealed that the resultant hydrogels have varying stiffness and yield stress, depending on the backbone and C-terminal residue. Visible spectroscopy-based elution studies with SMDs (naltrexone, methotrexate, doxorubicin) showed that besides the C-terminal residue, the shape of the SMD also determines the rate and extent of SMD elution. Based on the elution assays, infrared spectroscopy, and FESEM, we propose models for the peptide fibril-SMD interaction. Our findings highlight the importance of matching the molecular properties of the self-assembling peptide and SMD in order to achieve the desired SMD release profile.


Results and Discussion
Peptide Preparation and Assessment of Cytotoxicity. The acetylated Leu-Ile-Val-Ala-Gly (Ac-LIVAG) and acetylated Ile-Val (Ac-IV) backbones have been previously reported to be nonpolar oligopeptidic backbones that promote peptide self-assembly in an antiparallel manner ( Fig. 1) 37 . These studies have primarily focussed on peptides that have serine, threonine, cysteine, aspartic acid, glutamic acid, and lysine as the C-terminal amino acid residue. In this report, we expand the study to include histidine, arginine, and asparagine. Both histidine and arginine are basic amino acids, with histidine and arginine being less and more basic than lysine respectively. Asparagine possesses a neutral (carboxamide) side chain (under physiological conditions) and can be considered to be more basic than aspartic acid (carboxylic acid side chain). These new peptides ( Fig. 1) will allow us to evaluate the effect of side chain basicity on self-assembly, hydrogelation, and peptide fibril interaction with small-molecule drugs (SMDs).
The peptides were prepared via standard solid-phase peptide synthesis. The purification of all peptides was straightforward, except for peptide 1D. Peptide 1D was found to be essentially insoluble in acetonitrile/water. As a result, it frequently precipitated on the reversed-phase column during HPLC (high performance liquid chromatography) purification with a gradient of acetonitrile/water, leading to poor yields (<10%). The solubilities of peptides 1 and 2 in phosphate-buffered saline (PBS; pH 7.4) to form clear solutions were then tested; peptides 1A and 2A were also assessed as references. The range of solubility for the peptides is shown in Table 1. Peptides 1D and 2D were only sparingly soluble in PBS. As Table 1 indicates, the hexapeptides dissolve to a lower extent than the tripeptides. This is reflective of the greater propensity of the hexapeptides to self-assemble and aggregate due to their longer hydrophobic backbone. In addition, peptides with lysine and arginine dissolve to a greater extent due to their charged side chains at pH 7.4.
The cytocompatibility of the new peptides was assessed in order to ascertain their suitability for eventual drug delivery applications. The hemolytic ability was evaluated using rabbit red blood cells. This is a standard assessment of materials for biomedical purposes (protocol ISO 10993-4/NIH 77-1294) 42 , particularly for implants, hemostatics and wound dressings. Following red blood cell incubation with varying concentrations of peptide, the level of hemoglobin detected in the supernatant was comparable to the baseline values obtained in buffered solution. This indicates that the peptides do not disrupt the membranes of red blood cells, even at their respective maximal soluble concentrations. Since the peptides will not cause hemolysis, they can potentially be used as implantable scaffolds, drug delivery matrices and topical wound dressings.
Human mesenchymal stem cells are multipotent stem cells that are being actively investigated for tissue engineering applications. They can be harvested from adipose tissue (hASCs) and are capable of differentiating into Scientific REPoRTS | (2018) 8:17127 | DOI: 10.1038/s41598-018-35431-2 many tissue types, representing a very important source of stem cells for tissue regeneration. It is thus of interest to assess if the new peptide hydrogels are suitable as 3D matrices for the expansion and differentiation of hASCs [43][44][45] . A positive indication of cytocompatibility would be the survival of the cells following incubation with the peptides. For comparison, peptides 1A and 2A were included as part of the study. The cytocompatibilty of the peptides was evaluated by quantifying the metabolic activity using the WST-1 reagent and visualising the live/dead cells using calcein AM/ethidium homodimer staining. All the peptides demonstrated good viability, even at the solubility limit of the peptides (Fig. 2b). The metabolic activity was consistently comparable to control cultures (absence of peptides); the lone exception was peptide 2C with slightly lower (80%) metabolic activity. Fluorescence imaging of the hASCs stained with calcein AM and ethidium homodimer showed that the cells are metabolically active (presence of calcein fluorescence in the cells) and that the cellular membranes were intact (absence of ethidium bromide fluorescence in the cells); only peptide 1C demonstrated some ethidium bromide signal along with strong calcein fluorescence ( Supplementary Fig. S1). This bodes well for the development of these peptides as implantable scaffolds.
We assessed the cytocompatibility with these two techniques as they offer differing insights to how amphiphilic peptides can interact with cells. As Lum et al. have found, although their cationic hydrophobic peptides Figure 1. Depiction of how the hydrophobic amino acid backbones may self-assemble in an anti-parallel manner to accommodate the different sizes of the amino acid side chains. The darker the tone, the larger the side chain. Indigo refers to the hydrophobic amino acid and yellow refers to the hydrophilic C-terminal amino acid (Xaa). For X = (A-C) the peptides are amidated at the C-terminal (viz. a carboxamide); for X = (D, E) the C-terminal is a carboxylic acid.  Table 1. Soluble peptide concentrations and minimal peptide gelation concentrations in phosphate-buffered saline solution (pH 7.4). The pK a values are approximations based on N-acetyl-L-lysinamide for 1A/2A, N-acetyl-L-histidinamide for 1B/2B, and N-acetyl-L-argininamide for 1C/2C, which in turn were calculated values based on the Advanced Chemistry Development (ACD/Labs) software v11.02. As the hydrophobic backbones are non-ionizable, they will only negligibly affect the pK a of the C-terminal residue side chain. This will not affect the relative magnitudes of pK a between peptides with different C-terminal residues. The zeta potential measurements were made in water at a peptide concentration of 10 mM at 25 °C. The values (mean of triplicate measurements) indicate that the fibrils formed by the hexapeptides are more positively charged than those of the tripeptides. were non-haemolytic, they exhibited substantial toxicity towards the two human cell lines that were used for assessment 46 . The balance between hydrophilicity and hydrophobicity is certainly very important in the design of self-assembling peptides, and this has certainly been achieved with the RADA (Arg-Ala-Asp-Ala) peptide 47 . Strikingly, this is a motif that is amenable to adjustment of the hydrophilic-hydrophobic components -when the hydrophilicity and hydrophobicity are respectively dialled down (changing Arg to Lys) and up (changing Ala to Leu), the resultant KLDL peptide continues to remain cytocompatible and biologically active 48 ; similarly, when RADA was converted to EAKA (Glu-Ala-Lys-Ala), no difference in toxicity was measured 49 . This is a flexibility that we also observe with the Ac-LIVAG-and Ac-IV-consistent backbones: interconversion of the C-terminal residue among Lys, His, or Arg preserves the cytocompatibility of the peptides while introducing variability to the side chains.

Rheological Studies of Peptide Hydrogels, Field Emission Scanning Electron Microscopy Studies of Peptide Nanostructures, and Circular Dichroism Spectroscopy Studies of Peptide Solutions.
Compared to peptide 1A, which was previously reported to gel at a low concentration of 10 mg/mL (6 mM) in phosphate-buffered saline (PBS) 50,51 , peptide 2A was found to gel at 18 mM. Although peptide 1C could form a gel at 40 mg/mL (30 mM), the hydrogel formed was opaque and inhomogeneous; this is a big contrast to peptide 1A that forms a clear and transparent hydrogel (Fig. 3a). Peptides 1D and 2D are only sparingly soluble and do not form self-supported hydrogels; only dense clumps of precipitate were observed. This is a big contrast to peptides 1E and 2E, which were previously reported to form a self-supported hydrogel even at low peptide concentrations. On the other hand, perhaps unsurprisingly, the tripeptides generally require higher concentrations to gel. Peptide 2A was previously reported to require a higher peptide concentration to gel than peptide 1A 50 . This has also been observed for peptide 2B, in which 30 mM is required (c.f. 15 mM for peptide 2A). However, 2C did not gel even at the highest peptide concentration (90 mM) assessed, remaining as a solution in PBS. As with peptide 1D, peptide 2D could not gel and only precipitated. These observations suggest that self-assembly and hydrogelation require an intricate balance of the hydrophilic and hydrophobic elements in the peptide motif. When the hydrophobic characteristics predominate, the resulting peptides demonstrate low solubility and are more likely to form colloids or aggregate as precipitates in solution. When the hydrophilic components are dominant, the enhanced solubility reduces the driving forces for self-assembly. Consequently, hydrogelation is observed at markedly higher concentrations. Such a balance of hydrophilicity/hydrophobicity has been exploited (via enzymatic control) for the formulation of hydrogels 52,53 . The stiffness of the peptide hydrogels was assessed via rheology. The peptides (peptides nX; n = 1-2, X =A-C) were dispersed in phosphate-buffered saline to final concentrations of 40 mg/mL. As Fig. 3a illustrates, only peptides 1A and 1B formed transparent hydrogels; the rest of the peptides (except for peptide 2C that did not gel) formed opaque hydrogels. This implies that the fibrils of peptides 1A and 1B had sufficient time to disperse in solution before entrapping water to gel whereas peptides 1C, 2A, and 2B gelled quickly upon fibril formation, leading to opaque hydrogels. These observations indicate that the C-terminal residue affected the kinetics of hydrogelation, viz. fibril-water interactions, in which the Arg of peptide 1C is capable of mediating stronger hydrogen bonding interactions between the fibrils and the surrounding water (compared to Lys of peptide 1A). Thus, this suggests that the greater number of hydrogen bond donors/acceptors on Arg (seven) compared to Lys (three) or His (two) is responsible for mediating greater fibril-water interaction for the hexapeptides. In the case of tripeptide 2C, however, the strong peptide 2C-water interaction overwhelms self-assembly conferred by the 2-residue backbone (as for peptides 2A and 2B). These observations point to the importance of the balance between inter-peptide interaction and peptide fibril-water interaction in determining whether hydrogelation occurs or not. This has been found to be important in starch gels 54 , and this was precisely the consideration made in the design of various peptide nanostructures 55 and a vancomycin hydrogel 56 . The rheological studies show that the hydrogels (except peptide 2C that did not gel) exhibit a range of stiffness from 5 kPa to 100 kPa. One might expect that a backbone with five amino acid residues (c.f. two amino acid residues) would be able to serve as a more effective anchor for the self-assembly process, leading to similar fibrillar nanostructures and consequently similar hydrogelation properties. However, this is not the case: the stiffness of the peptide hydrogels 1(A-C) varied between 7-100 kPa whereas that of peptides 2(A,B) varied between 5-20 kPa ( Supplementary Fig. S2). The oscillatory strain sweep also bears out the same point. As Fig. 3b illustrates, despite their different consistent backbones, peptides 2A, 3A, 1B, and 2B break down at around the same strain (0.3%); the exception is that of peptide 1A, which broke down only at 3% strain, i.e. 10× that of the other hydrogels. Evidently, the C-terminal residue exerts a sufficiently significant effect that the stiffness and the yield stress of the hydrogels varied widely. This is intriguing as (1) peptide 1A would exhibit a different stress yield from the other two hexapeptides (similar to each other) and (2) the two hexapeptides (1B, 1C) and two tripeptides (1A, 1B) would exhibit the same yield stress. While various external macroscopic factors (e.g. temperature, pH) and peptide sequence have been reported to influence the rheological properties of hydrogels 57 , this is the first instance, to the best of our knowledge, that a molecular change in a single residue of the peptide or length of the hydrophobic backbone are found to impact upon these rheological properties so profoundly.
Field emission scanning electron microscopy (FESEM) was utilised to analyse and contrast the nanostructures within the peptide hydrogels. As Fig. 4a illustrates, the fibrils of peptide 1A are more condensed together compared to the distinct fibrils observable for peptides 1(B,C) and 2(A,B). While some fibrillar feature is observed for peptide 2C, the fibrils appear to be very closely stuck to each other. Analysis of the electron micrographs reveals that mean peptide widths of peptides 1(A-C) and 2B are similar; only the mean fibril width of peptide 2A differs significantly from the other peptides (Fig. 4b). The observation that the fibrillar morphologies and mean peptide widths of peptides 1(A-C) are similar, but not among peptides 2(A-C), suggests that the five-amino acid consistent backbone (Ac-LIVAG) is long enough to facilitate similar self-assembly of homologous peptides (Ac-LIVAG-Xaa) with a lesser impact by the C-terminal residue (Xaa). On the other hand, the two-amino acid consistent backbone (Ac-IV) appears to be too short to direct similar self-assembly of homologous Ac-IV-Xaa peptides, with their C-terminal residue playing a much more significant role.
It is apparent that the identity of the C-terminal residue affects peptide self-assembly to differing extents depending on the length of the hydrophobic backbone. As the electron micrographs illustrate, the morphology of the hexapeptide fibrils are quite similar whereas those of the tripeptide fibrils are quite different from each other (Fig. 4). While the impact of the side chain of the C-terminal residue on peptide self-assembly has not been reported before, the impact of the chirality of the C-terminal residue is known. In the elegant work of Wang et al. 58 , it was shown that a change in the chirality of only the C-terminal residue of Ile-Ile-Ile-Lys did not impede self-assembly, but affected the "handedness" of the peptide fibrils. The impact of this single chirality change on hydrogelation, though, is not known from this work. However, the work of Marchesan et al. reveals that for Phe-Phe-Val, the chirality of the C-terminal residue (Val) has to be the same as the second residue (Phe) in order for hydrogelation to proceed 59 . In addition, they found that the two enantiomers, i.e. D Phe-Phe-Val and Phe-D Phe-D Val, turned out to gel at different rates and resulted in hydrogels with different storage moduli, just as we have observed with our C-terminal changes. All these results point to the influential role that the C-terminal residue plays in peptide fibril-water interaction in hydrogelation.
Circular dichroism (CD) spectroscopy was also utilised to study the secondary structures that arise from the self-assembly of the peptides in solution. As Fig. 4c illustrates, all the peptides form polyproline II-type helical structures in solution, which is suggested by the presence of two negative peaks close to 197 nm and 225 nm as reported by Rucker and Creamer 60 and Shi et al. 61 . Among the hexapeptides, the different intensities of the negative peaks indicate that the extent to which each hexapeptide forms the polyproline II helical structure in solution is different. This suggests that the various C-terminal residues of the hexapeptides have led to differences in peptide self-assembly. On the other hand, the shorter backbone of the tripeptides has led to even bigger effects on self-assembly: not only are the intensities of the negative peaks different, the peak wavelength is different for peptides 2A and 2C, which exhibit a red shift in one of the negative peaks from 197 nm (for peptide 2B) to ~200 nm, which suggests a significant difference in their polyproline II helical structure compared to the rest of the peptides. Thus, these CD studies also illustrate the profound impact both the backbone length and C-terminal residue exert on peptide self-assembly.

Studying the Interaction of Peptide Fibrils and Small-Molecule Drugs via Elution Assay.
Three medically relevant small-molecule drugs (SMDs), i.e. naltrexone, methotrexate, and doxorubicin, were used to study how the C-terminal residue affects the interaction between the peptide fibrils and SMDs. Naltrexone can be used to treat pruritus 62 , methotrexate can be used to treat psoriasis, arthritis, and even certain cancers 63 , and doxorubicin has been a prominent anticancer drug for decades 64 . Naltrexone, methotrexate, and doxorubicin were selected to study the effect of charge and molecular shape in their interaction with the peptide fibrils. At physiological pH 7.4, naltrexone (pK a~8 .1) 65 and doxorubicin (pK a~8 .3) 66 are positively charged while methotrexate is negatively charged (pK a~1 1) 67 . Molecular modelling shows that naltrexone is a relatively spherical molecule, methotrexate is a semi-planar molecule, and doxorubicin is largely a planar molecule (Fig. 5).
In order to probe how well the SMDs interact with the positively-charged peptide fibrils (reflected by the zeta potential measurements in Table 1), we utilised a simple elution assay analogous to that employed by Branco et al. 68 : the SMD-encapsulated hydrogel, with a drug encapsulation efficiency of 100%, was layered with PBS (pH 7.4) for two hours, after which the PBS was removed and analysed for the amount of SMD eluted (Supplementary Fig. S3). Thus, the stronger the peptide fibril-SMD interaction, the lesser the amount of SMD that would be eluted. We compared the SMD elution profiles of the peptide hydrogels with the theoretical maximal amount of SMD that would diffuse from the hydrogel into the extra volume of PBS if there were no obstruction to diffusion. Figure 6a compares and contrasts the naltrexone elution profiles of different concentrations of peptide 1A hydrogels . It shows that the elution of naltrexone is essentially independent of peptide 1A concentration or peptide identity. One cause that could lead to such an effect is that there is minimal interaction of naltrexone with the peptide fibrils, viz. naltrexone is diffused within the water medium. This reason is also supported by the near-complete elution of naltrexone from the peptide hydrogels with sufficient elution counts (Fig. 6a). While peptide hydrogels 1A, 1B, and 1C formed stable hydrogels with naltrexone, only peptide hydrogel 1A could do so with methotrexate (with some coagulation within the hydrogel); both 1B and 1C formed only hydrogel clumps with methotrexate and consequently could not be assessed with the elution assay. For doxorubicin, only peptides 1A and 1B formed stable hydrogels with it. Figure 6b illustrates the difference that the C-terminal residue and the SMD makes to the peptide fibril-SMD interaction. While the naltrexone elution profiles of peptide 1A and 1B are similar (blue and grey), their doxorubicin elution profiles are very different (orange and green): 40% of the encapsulated doxorubicin could be eluted from peptide 1A hydrogel whereas only 10% could be eluted from peptide 1B hydrogel. The time-dependent release of naltrexone and doxorubicin from both peptides 1A and 1B was also determined and compared with each other. As Fig. 6c illustrates, burst release of naltrexone and doxorubicin from both peptide hydrogels was observed. However, naltrexone was released to a greater extent than doxorubicin from both peptide hydrogels, which indicates the greater extent of interaction of doxorubicin (c.f. naltrexone) with the peptide fibrils. In addition, while naltrexone was released to a similar extent from either peptide 1A or 1B hydrogels, doxorubicin was released to a greater extent from peptide 1A (c.f. 1B), which reflects the greater interaction of doxorubicin with peptide 1B fibrils. These observations imply that the C-terminal residue (Xaa) plays an influential role in the interaction of the peptide fibril with different SMDs.
Given that both naltrexone and doxorubicin are positively charged at pH 7.4, their electrostatic interactions with peptide 1A (or peptide 1B) fibrils ought to be similar. However, since their elution profiles with each peptide hydrogel are different (1A: blue vs orange; 1B: grey vs green), this implies that another factor is at work. This factor is also evident by comparing the elution of different SMDs from peptide 1A hydrogels. The elution of methotrexate from peptide 1A hydrogel was lower (and slower) than that of naltrexone, and this might be attributable in part to the electrostatic attraction between the positively charged peptide 1A fibril and negatively charged methotrexate. However, despite being positively charged, doxorubicin was eluted to an even lower extent than the negatively charged methotrexate. Thus, there is a factor that helps the positively charged doxorubicin to interact with the positively charged peptide fibrils so as to reduce its elution.
As alluded to earlier, this factor could be the molecular shape of the SMD. Doxorubicin is a planar and (largely) aromatic molecule that could bind to the peptide fibrils in a manner analogous to thioflavin T, a planar aromatic molecular probe used in binding studies with peptide amyloid fibrils 69 . Due to the inherent insolubility and variability in binding of thioflavin T to peptide amyloid fibrils, the molecular details of the binding are currently still elusive. However, model studies utilising a peptide self-assembly mimic (PSAM) have shed light on this important interaction. Based on the molecular docking on an X-ray crystal structure of the PSAM 70 and molecular dynamics simulation 71 , four possible binding sites were identified. The most prominent binding site comprises a hydrophobic groove that is bordered by two ladders (separated by one amino acid residue) along the long axis of the fibril: in one ladder, there are five Tyr side chains on five adjacent cross-beta strands; in the other ladder, there are five Leu side chains (Fig. 7a). Analogous to how acetylcholine esterase binds thioflavin T 72 , the PSAM is able to bind thioflavin T via a combination of π-π stacking (via Tyr side chains) and hydrophobic (via Leu side chains) interactions.
Since the Ac-LIVAG-Xaa class of peptides have been shown to assume cross-β peptide structures 37,50 , we propose that doxorubicin might bind analogously to the hexapeptides as thioflavin T binds to the PSAM. For peptide 1B, whose C-terminal residue is histidine, it has an aromatic imidazole side chain that could participate in π-π interaction 73 with doxorubicin, just as how the aromatic phenolic side chain (of Tyr) in PSAM formed the "first ladder" that interacts with thioflavin T. (Fig. 7a) Although the imidazole side chains of the His residues on neighbouring antiparallel cross-β strands are five residues away on the same protofibril and are too far away from each other to form a binding site for doxorubicin via π-π interaction, the imidazole side chains can potentially form a binding groove with the side chains of the neighbouring protofibril instead (Fig. 7b). Intra-protofibril binding doxorubicin is possible, but this would have been mediated via imidazole NH-π interaction with doxorubicin ( Supplementary Fig. S4). In fact, as Liao et al. noted 74 , the His NH-π interaction is stronger than π-π interaction, so it may be that both intra-and inter-protofibril binding of doxorubicin are relevant. Either mode of binding of doxorubicin should not affect the secondary structure of the peptide fibrils and this can be observed from the lack in change of the amide I absorption band (ν = 1635 cm −1 ), which is reflective of β structures 75,76 , in the presence of doxorubicin (Fig. 7d); electron micrographs also reveal little change in peptide morphology in the presence of doxorubicin (Fig. S5). As Lindberg et al. noted, the binding of thioflavin T by peptide fibrils of insulin also did not alter its structure 77 . Naturally, such binding interaction of doxorubicin with the peptide fibril would decrease its rate of elution from the hydrogel. However, how would peptide 1A fibrils, which do not contain aromatic residues, interact with doxorubicin? As Wu et al. 71 and Sabaté et al. 78 noted, the positively charged thioflavin T binds poorly to PSAM and insulin fibrils respectively when there are positively charged residues close to the binding sites. Analogously, doxorubicin would bind poorly to peptide 1A fibrils which possess positively charged side chains. One possibility is that the hexapeptide may interact with doxorubicin via amide-H mediated cation-π interaction. It is known that cation-π interactions feature prominently in the stabilization of the tertiary structure of proteins 79 . In proteins, the cation-π interaction is commonly found to be mediated via the ammonium or guanidinium side chains of Lys and Arg with aromatic residues (Phe, Tyr, Trp) 80 . While the C-terminal residue of peptide 1A is Lys and could possibly interact with the aromatic surface of doxorubicin, the side chain of Lys is long (with four methylene units between the alpha carbon and the ammonium group) and projects the ammonium group into the aqueous medium of the hydrogel where the salt effect is strong and the corresponding cation-π interaction would be weak 81 . However, the amide proton can also participate in cation-π interactions, in which the partial positive charge on the amide proton is directed into π electron cloud of the aromatic ring [82][83][84][85] . As the crystal structure of Ac-LLE shows, the tripeptide packs in an antiparallel manner and this leads to the amide protons pointing in the same direction on one layer of Ac-LLE (Fig. 7c) 41 . Since the hexapeptide also self-assembles in an antiparallel manner, it is conceivable that the amide protons could be aligned in the same direction. Thus, the hexapeptides could interact with both faces of doxorubicin via amide-H mediated cation-π interactions, leading to the intercalation of doxorubicin into the peptide fibril at certain junctures (Fig. 8). Given that the doxorubicin:peptide ratio is 1:10, such intercalation should not affect the fibril structure significantly -as the infrared spectra and electron micrographs indicate, there is little change in the presence of doxorubicin (Fig. 7d, S5). Naturally, if such cation-π interaction were to contribute to binding of doxorubicin to peptide 1A, then this binding mechanism would be operational for peptide 1B too. This could account for why doxorubicin was eluted to a lesser extent (and slower) from peptide 1B than from peptide 1A: there are two separate mechanisms binding doxorubicin to peptide 1B whereas there is only one mechanism for peptide 1A. Such cation-π interaction could also account for the greater (and faster elution) of methotrexate from peptide 1A hydrogel compared to doxorubicin. Methotrexate has a small planar component that could potentially intercalate into the fibril (Fig. 8) and it would understandably be less effective at doing so compared to doxorubicin given its smaller aromatic surface, leading to poorer binding. Given that the potentially small extent of intercalation and electrostatic attraction between the positively charged peptide fibril and negatively charged methotrexate still led to the greater (and faster) elution of methotrexate compared to doxorubicin, it implies that intercalation has a greater impact upon fibril-SMD interaction compared to electrostatic attraction. This is significant in view of the previous observation that negatively charged proteins were impeded in their release from hydrogels with positively charged peptide fibrils 36 . This would account for the lower (and slower) elution of doxorubicin relative to naltrexone even though both SMDs are positively charged at pH 7.4; presumably, the spherical shape of naltrexone would not enable it to intercalate within the peptide fibril (Fig. 8).

Conclusion
In this study, we have found that a longer consistent hydrophobic backbone of acetylated Leu-Ile-Val-Ala-Gly can indeed serve as an anchor for self-assembly even as the C-terminal residue is varied. While the C-terminal does not affect the peptides' cytocompatibility (non-haemolytic and non-cytotoxic), the stiffness and yield stress differ depending on the identity of the C-terminal residue. Thus, it might be possible to co-assemble different peptides (same backbone, different C-terminal residue) so as to produce hydrogels with a greater spectrum of tunability of various properties 86 . In addition, we have also found that the C-terminal residue plays an important role in affecting the rate and extent to which the peptide fibril interacts with an SMD. Interestingly, it might be that a planar aromatic SMD can interact with the peptide fibril via a novel intercalation mechanism. The details of this possible intercalation mechanism and the co-assembly of peptides are currently under investigation and will be reported in due course.

Materials and Methods
All peptides were prepared via fluoren-9-ylmethoxycarbonyl (Fmoc)-based solid phase peptide synthesis and purified via reversed-phase high performance liquid chromatography-mass spectrometry (HPLC-MS) on an Agilent 6130 Quadrupole LC/MS system. Peptide hydrogels were prepared by weighing out the appropriate peptide mass in 2 mL-vials. After the requisite volume of phosphate-buffered saline (PBS) was added, the peptide-water mixture was shaken and sonicated for 10 s. The hydrogels were allowed to anneal for 24 h before they were used for further experiments. These procedures, along with the acquisition of precursors for peptide synthesis, were previously described in detail by us 41 . Naltrexone, methotrexate, and doxorubicin were purchased from Sigma-Aldrich. For the drug elution studies, a stock solution of 20 mM in Millipore water of each of these drugs was prepared. Methotrexate was neutralised with two equivalents of NaOH to produce the anionic form in solution. The stock solution was diluted to a final concentration of 2 mM in PBS prior to the preparation of the hydrogel for small-molecule drug elution studies (see below).
Haemolysis assay of peptides with rabbit red blood cells. The procedure of Jiang et al. was followed 87 . Briefly, rabbit blood purchased from the Biological Resource Centre, Singapore, was centrifuged and the serum was removed. The red blood cells were rinsed with PBS, centrifuged, and the supernatant was removed; this process was repeated twice. The red blood cells were then incubated with the respective peptides in PBS in a 96-well plate (1 × 10 4 cells/well) for 24 h at 37 °C. After that, the supernatant in each well was transferred with a multichannel pipettor into another 96-well plate and analyzed by a TECAN plate reader at 576 nm. 1% Triton-X100 was used as a positive (lytic) control. Each peptide was assessed for a total of eight times. Field emission scanning electron microscopy (FESEM) study of peptide morphology. Peptide morphology was analysed with a JEOL JSM-7400F field emission scanning electron microscope. The freeze-dried sample was loaded onto the sample stage of the electron microscope and analysed under a vacuum of 10 Pa and a working current of 10A (5 kV). A detailed procedure was previously reported by us 41 . The diameter of the peptide nanofibers was determined by analysis of SEM images using ImageJ software (http://imagej.nih.gov/ij/docs/). Briefly, pictures were converted to 8-bit images and the threshold was adjusted in order to distinguish each nanofibril. Subsequently, the diameter of peptide nanofibrils was measured manually at 10 random locations in each micrograph.
Circular dichroism spectroscopy studies of peptides. Circular dichroism spectroscopy studies were carried out with a Jasco J-1500 circular dichroism (CD) spectrometer, which was calibrated with (+)-camphor-10-sulfonic acid. A 0.1 mm quartz cuvette was utilised for the measurements, which was kept at a constant temperature of 20 °C with the Jasco CTU-100 circulating thermostat unit. The spectra were recorded at soluble peptide concentrations in PBS as reflected in Table 1 at a resolution of 1 nm.
Zeta potential measurements of peptides. The zeta potential of the peptide fibrils was measured with an Anton Paar Litesizer 500 Particle Sizer. An omega zeta potential cuvette was utilised for the measurements, which was kept at a constant temperature of 25 °C. Triplicate measurements of each peptide sample were made at concentrations of 10 mM in water.
Small-molecule drug elution studies with naltrexone, methotrexate, and doxorubicin. The peptides were weighed out in 4 mL-vials, after which phosphate-buffered saline (PBS) containing the small-molecule drug (SMD final concentration = 2 mM) was added to a final volume of 0.30 mL each. The peptide-SMD-PBS mixtures were then shaken and sonicated for 10 s, after which the mixtures were allowed to anneal for 24 h. After that, 0.75 mL of fresh PBS was slowly added with a pipettor to the surface of the hydrogel so as not to disturb the surface of the gel. After 24 h at 25 °C, the PBS was carefully extracted with a pipettor for UV/vis analysis; a fresh volume (0.75 mL) of PBS was then replaced. This process was repeated as many times as the number of pre-determined elution samples. The extracted PBS was transferred into a cuvette (pathlength = 10 mm) and analysed with an Agilent 8543 UV/visible spectrometer for the presence of the SMD. Naltrexone, methotrexate, and doxorubicin were respectively analysed at 280 nm, 390 nm, and 480 nm in PBS. A control experiment was set up for each peptide hydrogel, for which no drug was added. This is to account for absorption at the respective wavelengths due to eluted peptide. The elution profile of each peptide is compared with a theoretical maximal elution profile, viz. the theoretical maximal amount of SMD that would diffuse from the hydrogel into the extra volume of PBS if there were no obstruction to diffusion. The cumulative eluted percentage of SMD at each point is given by: [1 − (1/3.5) n ] 100%, where n is the number of elution. The time-dependent release of naltrexone and doxorubicin was carried out in a similar manner to the elution studies, except that the supernatant was replaced on top of the hydrogel after each UV/visible spectroscopic measurement at each time point.
Attenuated total reflection infrared spectroscopy studies of peptides. ATR-FTIR spectroscopy studies were carried out with a Perkin-Elmer Spectrum 100 IR spectrometer, which was fitted with a germanium-based PIKE MIRacle attenuated total reflectance (ATR) sampling accessory. Briefly, a freeze-dried sample of the hydrogel was loaded on the sample stage of the spectrometer. At least 16 scans were recorded. The spectra were recorded by subtracting a reference spectrum of air. A detailed procedure was previously described by us 88 .