Biomaterials 35 (2014) 3427e3434
Contents lists available at ScienceDirect
Biomaterials journal homepage: www.elsevier.com/locate/biomaterials
Deciphering the mechanism of protein interaction with silk fibroin for drug delivery systems Oliver Germershaus 1, Vera Werner 1, Marika Kutscher, Lorenz Meinel* Institute for Pharmacy, University of Würzburg, Am Hubland, DE-97074 Würzburg, Germany
a r t i c l e i n f o
a b s t r a c t
Article history: Received 7 December 2013 Accepted 22 December 2013 Available online 22 January 2014
Silk fibroin (SF) is an exceptional drug delivery carrier with respect to stabilizing, protecting, and delivering sensitive biologics. A synopsis of thermodynamic, static light scattering, hydrophobicity probing, and nanoparticle tracking analyses served as a basis to decipher the mechanism of interaction between SF and two model proteins, protamine and polylysine. The impact of salts aiding (chaotropic), not affecting (neutral), or opposing (cosmotropic) SF unfolding was a major determinant, ranging from complete abolishment to maximal interaction efficacy. Evidence is provided, that the underlying mechanism of the remarkable ability to tailor drug/SF interaction throughout such large ranges and by appropriate salt selection is the control of structural breakdown of SF micelles as present in pure SF ad initium. This study provides a mechanistically justified and hypothesis driven blueprint for future experimental designs addressing the controlled interaction of biologics and SF. Ó 2014 Elsevier Ltd. All rights reserved.
Keywords: Drug delivery of biologics Silk fibroin Mechanism of interaction Thermodynamics
1. Introduction The drug delivery of biologicals is a high growth segment of the pharmaceutical industry, with key drivers of innovation being the stabilization of these sensitive molecules opening a potential for storage at room temperature and deployment of challenging administration routes [1]. Another frequent challenge in the production of current drug delivery systems is the need for processing in organic solvents, at extreme pH values, and mechanical stress all of which potentially challenging the protein’s integrity [2,3]. Silk fibroin (SF), a biopolymer isolated from whole silk cocoons of the silkworm Bombyx mori (L., 1758), has demonstrated superb stabilization for and controlled release of biologicals [1], thereby providing exciting new avenues for drug delivery [4e10] while further opening the interface to biomaterials and/or tissue (re-) generation [11]. Protocols have been presented, allowing manufacturing of protein loaded SF drug delivery systems under mild condition [8] and by translating insights of the natural (allwater based) silk spinning processes into advanced biomaterial and drug delivery system production, respectively [12,13]. Successful development of delivery systems was reported for therapeutically relevant proteins and peptides, including nerve growth factor (NGF) [7], insulin-like growth factor I (IGF-I) [8], fibroblast growth * Corresponding author. Tel.: þ49 931 318 54 71; fax: þ49 931 318 46 08. E-mail address: [email protected] (L. Meinel). 1 These authors contributed equally to the manuscript. 0142-9612/$ e see front matter Ó 2014 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.biomaterials.2013.12.083
factor 2 (FGF-2) [14], bone morphogenetic protein 2 (BMP2) [4,5,15], monoclonal antibodies (mAb) [11], enzymes [16], and others [1,6]. Most of these studies deployed the excellent SF material properties [17e19] and biocompatibility [20,21] for cell growth and differentiation [22,23] and combined these advantageous properties with bio-functionalities through co-processing the biopolymer and the therapeutic or by means of its covalent decoration before or after scaffolding into three dimensional carriers [1,5,24]. More recently, these solid SF drug delivery carriers were complemented by semisolid drug delivery systems and freeze-dried gels for mAb delivery, detailing the design space and mechanism of release by which SF may be used as a drug delivery biopolymer [11,25]. However, the precise mechanism of interaction of SF and biologicals in aqueous solution is still poorly understood and, therefore, scientists were largely forced to follow empirical approaches. Nevertheless, these studies linked better SF loading to positively rather than negatively charged drug molecules, hypothetically assuming Coulomb forces establishing between the negatively charged SF at neutral pH (pI w 4.5) [26] and positively charged, alkaline drugs [9,27]. Other hypotheses explained sustained delivery profiles with hydrophobic attraction, linking the rather hydrophobic patches of SF to the hydrophobicity of small molecule drugs, such as propranolol, or hydrophobic patches of antibodies [1,11]. This study aims at expanding this mostly empirically built knowledge base by elucidating the mechanism of SF interaction with biologicals, using the basic model proteins polylysine and
3428
O. Germershaus et al. / Biomaterials 35 (2014) 3427e3434
protamine, respectively. These proteins were selected based on their net positive charge at neutral pH (facilitating electrostatic forces to the net negatively charged SF). Furthermore, these model proteins were chosen to include some structural variability, with polylysine representing a strongly and homogenously charged polyelectrolyte characterized by significant intramolecular electrostatic repulsion resulting in random coil conformation [28], while protamine provides more structural complexity, including hydrophobic sections/b-sheet [29]. In addition to the variation of the model drug, the effect of specific ions resulting in cosmotropic or chaotropic environment, respectively, on proteineSF interaction was studied.
and 2.5 nm, respectively. Scattered light was observed for 60 s with an integration time of 0.1 s and a data interval of 1.2 s. The samples were produced as described before but volumes were increased by a factor of 10 to allow reliable measurement in a 3.5 ml quartz cuvette (cuvette type 101-QS, Hellma, Müllheim, Germany). 2.5. Zeta potential measurements The surface charge was determined using a Delsa Nano HC (Beckmann Coulter, Brea, CA) using a 658 nm laser and a scattering angle of 15 . The samples were produced as described before but were diluted 1:12.5 for protamine and 1:7.5 for polylysine with ultrapure water immediately before each measurement.
2. Experimental details 2.6. Nanoparticle tracking analysis (NTA) 2.1. Materials Cocoons of B. mori (L., 1758) were obtained from Trudel AG (Zurich, Switzerland). Protamine sulfate salt from salmon grade x, amorphous powder and Poly-L-lysine hydrobromide (polylysine, 15e30 kDa by viscosity) was from SigmaeAldrich (Munich, Germany). All other chemicals used were at least of analytical grade and from SigmaeAldrich if not stated otherwise.
The colloid kinetics of SF and protamine alone, as well as of SF/ protamine mixtures at different mixing ratios, were determined with a NanoSight NS 500 instrument (NanoSight, Wiltshire, UK) using a 40 mW laser at 638 nm and the NTA Software Version 2.3. All samples were prepared as described before, and were diluted 1:10 using the respective buffer/salt solution immediately before each experiment. Movies were cut using Movie Maker (Microsoft, Redmont, WA).
2.2. Preparation of SF solution 2.7. Fluorescence spectroscopy SF solutions were prepared from silk cocoons as previously described with modification [30]. Cocoons were boiled in a stirred aqueous solution of 0.02 M Na2CO3 twice for 1 h each, and then washed with water 10 times. The dried fibers were dissolved in 9 M aqueous LiBr solution at 55 C until completely dissolved, yielding a concentration of 20% (w/v) SF which was filtered through a 5 mmsyringe filter (Versapor, Pall Life Sciences, Washington, NY) and dialyzed against ultrapure water for 2 days, changing the water 5 times and using SpectraPor dialysis membranes (MWCO 6e 8000 Da, Spectrum Labs, Rancho Dominguez, CA). The SF solution had a final concentration of 1.5e2% (w/v) and was stored at 2e8 C. Prior to use, the SF solution was filtered through a 5 mm-syringe filter (Versapor, Pall Life Sciences, Washington, NY).
Fluorescence intensities were recorded using an LS 50B luminescence spectrometer (Perkin Elmer, Waltham, MA) setting excitation at 490 nm and emission at 570 nm. Samples were prepared as follows: 2036 ml of an SF solution at 3.2 mM in the respective buffer/salt solution were mixed with 4 ml of a Sypro Orange (SO) solution, diluted 1:10 in DMSO from a stock solution (total dilution of working solution 50,000). The protamine solution was added in steps of 20 ml each, measuring the fluorescence spectrum at each titration point (vide supra). Relative fluorescence was calculated by setting the fluorescence of the SF solution without protamine as reference (corresponding to a relative fluorescence of 1). 2.8. Fluorescent labeling of protamine
2.3. Isothermal titration calorimetry (ITC) The ITC titrations were performed using a MicroCal iTC200 (GE Healthcare, Buckinghamshire, GB) to measure the heat of reaction when a protamine solution (1.622 mM) was titrated into an SF solution (32 mM) or a polylysine solution (98 mM) was titrated into an SF solution (19 mM) at 25 C and stirred at 400 rpm. Both, SF and titrant (protamine or polylysine) were dissolved in the same medium, composed of a 25 mM histidine buffer, pH 7.6 supplemented with sodium chloride, sodium thiocyanate, or sodium sulfate yielding an ionic strength of 192 mM, respectively. Each injection consisted of 2 ml of titrant injected during 4 s each (first titration at 0.2 ml of titrant injected during 0.4 s), with a spacing of 150 s between each of a total of 20 injections. The cell volume was 200 ml. The first injection was always omitted to minimize the impact of equilibration artifacts, following the manufacturer’s recommendations. All curves were corrected by a blank titration, performed by titrating the respective model protein (protamine or polylysine) in buffer/salt solution without SF.
8 ml of a 1.0 mg/ml solution of Fluorescein isothiocyanate was added drop-wise to 100 ml of a stirred 2.0 mg/ml solution of protamine in 0.1 M sodium carbonate buffer, pH 9.0. The mixture was stirred for 8 h at 4 C protected from light. Afterward the solution was dialyzed against ultrapure water for 36 h, using a SpectraPor dialysis membrane (MWCO 6-8000 Da, Spectrum Labs, Rancho Dominguez, CA) and the solution was freeze-dried for 24 h at 5 C and 0.16 mbar (Alpha 1-4, Christ, Osterrode, Germany). 2.9. Microscopic particle analysis A 32 mM SF solution was mixed with a 1.622 mM solution of fluorescently labeled protamine at a molar ratio of 10.4. An Axio Observer.Z1 microscope equipped with a Plan-Apochromat 40/ 0,95 objective and an AxioCam MRm3 camera (Carl Zeiss Microscopy, Jena, Germany) was used. Excitation wavelength was set to 450e490 nm and emission wavelength to 500e550 nm. 3. Results
2.4. Static light scattering (SLS) Light scattering was performed using an LS 50B luminescence spectrometer (Perkin Elmer, Waltham, MA) at a fixed angle of 90 . Excitation- and emission wavelength and -slits were set to 638 nm
The protamine/SF interaction was studied by titrating SF into a solution of protamine in low ionic strength buffer of 0.31 mM, respectively. Endothermic signals with progressively declining intensity were recorded up to a ratio of 1.5 at which the reaction
O. Germershaus et al. / Biomaterials 35 (2014) 3427e3434
1.6
5 Hm [kcal/mol]
1.4 0
3429
200
10
150
0
100
–10
50
–20
1.2
–5
1 0.8
–10
0.6 –15 0
2
4 6 8 Molar ratio
0 0
10
2
4 6 8 Molar ratio
10
–30
Fig. 1. Interaction of SF and protamine in histidine buffer. (A) Development of isothermal enthalpy (open circles) and SO fluorescence (filled squares) during addition of increasing amount of protamine to SF. (B) Development of right angle light scattering at 638 nm (open circles) and zeta potential (filled squares) during titration of SF with protamine.
enthalpy became exothermic with a maximum of heat liberation observed at a ratio of 3 and a subsequent decline of heat per additional titration, thereafter (Fig. 1A). The change of hydrophobicity of the colloidal solution during the titration as probed by Sypro Orange (SO) correlated to the heat signal, with a continuous increase in fluorescence intensity (hence hydrophobicity of the colloidal solution) at low ratio, rapidly reaching maximum fluorescence at a ratio of 2.2 and continuously declining, thereafter (Fig. 1A). Similarly, light scattering profiles were characterized by a constant increase of scattering intensity up to a ratio of 2.6, followed by a phase of a steep increase in scattering intensity reaching its maximum at a ratio of 3 and rapidly declining, thereafter (Fig. 1B). Zeta potential of pure SF solution was at 25 mV and rapidly increasing until reaching net charge neutrality at a ratio of 3 coinciding with maximal scattering intensity. The zeta potential remained stable and plateaued at about 6 mV (Fig. 1B). The effect of increasing the ionic strength was studied using sodium chloride, a salt having only negligible effects on protein solvation [31e34]. Interestingly, in the presence of sodium chloride exothermic signals were obtained already at early titration points, with maximum exothermicity at a ratio of 1.6, after which the exothermic signals declined with increasing ratio (Fig. 2A). The change of hydrophobicity of the colloid solution (as indicated by relative SO fluorescence) slightly increased at low ratio, peaking at a ratio of 1.1 and gradually declining, thereafter (Fig. 2A). Overall the scattering
1.6
5
1.4 0
200
10
150
0
100
–10
50
–20
1.2
–5
1
0.6 2
4
6
8
10
R
0.8
–10 –15 0
intensity was low as compared to titrations performed in pure buffer (vide supra) and increasing at a ratio of 6.3. Likewise, the zeta potential was substantially reduced with surface charge neutrality being shifted to a higher ratio as compared to the titration in plain buffer (Fig. 2B). The impact of the chaotropic salt sodium thiocyanate replacing sodium chloride was studied at otherwise identical ionic strength (Fig. 3) [31e34]. Analogous to the observation with sodium chloride low ratios had exothermic effects (Fig. 3A). In contrast to sodium chloride, the exothermicity constantly increased up to ratios of 10, whereas the hydrophobicity pattern (relative SO fluorescence) throughout the titration was qualitatively comparable to the results obtained with sodium chloride with a more pronounced decrease at low ratios and maximum hydrophobicity being measured at a ratio of 2.6. The scattering intensity increased constantly with a low slope up to a ratio of 5.9, followed by a substantial increase (Fig. 3B). Interestingly, the scattering intensity did not decrease at higher ratios and in contrast to the other profiles obtained, but leveled off at high scattering intensity. Similar to the data obtained for sodium chloride, the zeta potentials were overall lower than those observed in buffer alone (Fig. 3B). The impact of cosmotropic salts was studied using sodium sulfate [31e 34]. Sodium sulfate radically reduced the interaction of protamine with SF and as detailed by minimal exothermic effects as recorded by ITC, with the hydrophobicity of the colloidal solution being linearly reduced with increasing ratio (Fig. 4A). An increase in light
0 0
2
4
6
8
10
–30
Fig. 2. Interaction of SF and protamine in histidine buffer containing 192 mM sodium chloride. (A) Development of isothermal enthalpy (open circles) and SO fluorescence (filled squares) during addition of increasing amount of protamine to SF. (B) Development of right angle light scattering at 638 nm (open circles) and zeta potential (filled squares) during titration of SF with protamine.
3430
O. Germershaus et al. / Biomaterials 35 (2014) 3427e3434
A
B
1.6
5
1.4 0
10
600
0
400
–10
200
–20
1.2
–5
1
0.6 2
4
6
8
R
0.8
–10 –15 0
800
0 0
10
2
4
6
8
10
–30
Fig. 3. Interaction of SF and protamine in histidine buffer containing 192 mM sodium thiocyanate. (A) Development of isothermal enthalpy (open circles) and SO fluorescence (filled squares) during addition of increasing amount of protamine to SF. (B) Development of right angle light scattering at 638 nm (open circles) and zeta potential (filled squares) during titration of SF with protamine.
scatter was observed up to a ratio of 6.5, remaining stable, thereafter (Fig. 4B). Similarly, zeta potentials increased up to a ratio of 5 and remained constant thereafter, failing to reach surface charge neutrality (Fig. 4B). Particle formation and interaction was followed by NTA on the nanometer scale (Fig. 5; Supplementary video 1). In the case of SF/protamine interaction, the resulting particle size distributions were rather broad, representing a mixture of particles with diameters from hundreds of nanometers up to several micrometers and complicating reliable analysis of single particle tracks (data not shown). Consequently, we focused on particle appearance and particle interaction at the different titration points. NTA of pure SF colloidal solutions generally exhibited broad particle size distributions, ranging from approximately 100 nm to well above 400 nm, with mean diameter in the range of 250e300 nm (data not shown). Particle morphology was dramatically impacted by the SF/protamine ratio and the buffer composition. In pure histidine buffer, a progressive increase in the number of fast moving, i.e. smaller particles was observed up to a mixing ratio of 2 (Fig. 5, A2). At subsequent titration points, formation of progressively larger, slower moving particles commenced and particle interaction increased at mixing ratios above 2, resulting in the formation of duplet and triplet clusters. After addition of sodium chloride formation of small particles was observed at mixing ratios from 1.1 to 3.2 with increasing aggregation of particles, thereafter (Fig. 5, B1e4). The experiments using sodium thiocyanate generally
A
1.6
5
1.4 0
yielded a similar outcome (Fig. 5, C1e4) but in contrast to sodium chloride the number of larger particles was reduced. Essentially the same observation was made for sodium chloride or sodium thiocyanate as compared to histidine buffer and as assessed qualitatively. In contrast, in the presence of sodium sulfate no major changes of particle size or morphology were observed throughout the entire ratio tested (Fig. 5, D1e5). The characterization of particle formation was corroborated by fluorescent microscopy using fluorescently labeled protamine at a protamine/SF ratio of 10.4 (Fig. 6). Particle formation in the micrometer range, as observable by fluorescence microscopy, was found for the histidine buffer and these observations were similar in the experiments conducted in presence of either sodium chloride or sodium thiocyanate, respectively. No particles in the micrometer range were observed in the case of sodium sulfate containing buffer. Particle counting and particle size analysis detailed numerous smaller particles in the sodium thiocyanate and larger but fewer particles in the case of sodium chloride containing buffers, respectively (Fig. 6B). The interaction of SF with basic proteins was further studied using polylysine (Fig. 7). The enthalpy change per titration was higher for polylysine as compared to protamine and complexation completed at substantially lower molar ratios. In contrast to protamine, polylysine interaction with SF was endothermic throughout the main part of the titration. In spite of these differences, the general interaction pattern between SF and polylysine decreased in the order of buffer
B
200
10
150
0
100
–10
50
–20
1.2 –5
1 0.8
–10
0.6 –15 0
2
4
6
8
10
0 0
2
4
6
8
10
–30
Fig. 4. Interaction of SF and protamine in histidine buffer containing 64 mM sodium sulfate (identical ionic strength as 192 mM sodium chloride). (A) Development of isothermal enthalpy (open circles) and SO fluorescence (filled squares) during addition of increasing amount of protamine to SF. (B) Development of right angle light scattering at 638 nm (open circles) and zeta potential (filled squares) during titration of SF with protamine.
O. Germershaus et al. / Biomaterials 35 (2014) 3427e3434
3431
Fig. 5. Evaluation of nanoparticle characteristics during the titration in different buffer and salt solutions using nanoparticle tracking analysis. Titrations were performed in histidine buffer, sodium chloride, sodium thiocyanate and sodium sulfate, respectively. Representative video stills are shown for the SF to protamine ratio of 1.1, 2.1, 2.6 and 3.7 for histidine buffer and 1.3, 2.6, 3.2, and 4.5 for all other salts deployed, respectively. Arrows highlight duplex particles. Original videos are uploaded as supporting information.
larger than sodium chloride larger than sodium thiocyanate larger than sodium sulfate (vide supra; Fig. 7A). Static light scattering revealed an increasing scattering intensity up to a ratio of approximately 0.37 for pure buffer, 0.16 for sodium chloride, 0.37 for sodium thiocyanate and 0.32 for sodium sulfate addition, respectively. In all experiments, the scattering intensity plateaued, thereafter (Fig. 7B). Supplementary Video related to this article can be found at http://dx.doi.org/10.1016/j.biomaterials.2013.12.083 4. Discussion These results detailed the interaction between SF and model protein drug, paving the way for a mechanistically driven and optimized drug loading on this biopolymer. The exceptional progresses seen for SF in the past provided the mechanistic foundation leading to predicted mechanical properties [19,35], a perplexing versatility in form, structure and function [36,37], and demonstrated in vitro and in vivo biocompatibility [8,20,38]. These successes among others sparked the interest in this biopolymer within a rapidly growing scientific community. Along with the insights outlined here and building off previous reports [1,10], the interpretation on the mechanism of drug loading will support the important research at the interface to drug delivery and might
expand the deployment of SF for the purpose of drug delivery through the rational hypothesis presented here within. Our results suggested, that protein drugeSF interaction at low molar ratio was by (i) association of protein molecules to free SF molecules resulting in nanocomplex formation, and to (ii) preexisting micelles as present in pure SF solution. The nanocomplex formation (i) is assumed to occur continuously throughout adsorption of the basic model protein (Fig. 5) and progressed towards charge neutralization of the ad initium negatively charged SF (Fig. 1A). At neutrality, aggregation commenced leading to abrupt phase separation (Fig. 1B). Thermodynamically, nanocomplex formation/coacervation is typically entropically driven, dominated by the release of counterions [39], a finding which had been corroborated here within by observing reduced (Fig. 1B and 2B) or completely abolished interaction with increasing ionic strength (data not shown) and as described before for similar systems [40e 42]. Mixing of polycations and polyanions typically sparks two events, (i) the formation of soluble or insoluble interpolymeric complexes paralleled or followed by (ii) coacervation and phase separation [43,44]. Therefore, the endothermic signals at low mixing ratio (Fig. 1A) may at least in part reflect ion pairing, an interpretation supported by the increasing formation of nanocomplexes and ultimately phase separation (Fig. 1B). This
3432
O. Germershaus et al. / Biomaterials 35 (2014) 3427e3434
(Fig. 5) (iii) flipping into an exothermic phase at a ratio of 2e3, and (iv) ultimately fading into a terminal phase of diminishing exothermic signals and decreasing hydrophobicity, likely a result of shielding effects of adsorbed protamine on the surfaces. Therefore, a mechanism of drug loading was detailed, within which structural dynamics of SF in presence of protamine are controlled by proper salt choice, providing a convenient tool for future manufacturing processes while deploying compendial or largely uncritical salts [48]. We hypothesized, that these differences in protamineeSF interaction result from stabilization or destabilization of SF micelles under cosmotropic or chaotropic conditions, respectively [18,27,31,33,34,39,49e51]. The relative importance of specific ion effects can be deduced from conditions of the natural silk spinning process [52] as well as from effects of different cosmotropic or chaotropic salts on stability of amphiphilic block copolymer micelles [53,54]. The increase in ionic strength by sodium chloride reduced the zeta potential and hence electrostatic interaction (Fig. 2B) while the impact on SF micelle stability was minimal (Fig. 2B; 5B). In contrast, chaotropic environments substantially impacted hydrophobic collapse of SF leading to micelle destabilization (Fig. 5C), efficient formation of coacervates (Fig. 6) and a continuous increase of exothermic signals with increasing ratio with no detectable saturation (Fig. 3A). The opposite was observed for cosmotropic conditions under which heat turnover
A
B
300 Buffer NaCl NaSCN
Particle count
250 200 150
A
100 50 0 0
5
10 15 20 25 Ferret diameter [μm]
30
Fig. 6. Microscopic analysis of complex coacervate at a protamine/SF mixing ratio of 10.4. (A) Representative micrographs of particles formed in histidine buffer (top left), sodium chloride (top right), sodium thiocyanate (bottom left) and sodium sulfate (bottom right). Scale bar represents 25 mm. (B) Analysis of Ferret diameter of particles in histidine buffer (open circles), sodium chloride (filled squares) and sodium thiocyanate (filled diamonds).
endothermic interaction at low ratios was principally corroborated for polylysine (although strikingly different in absolute terms as compared to protamine; Fig. 7). Previous studies reported similar interaction between polylysine and the (negatively charged) polysaccharide pectin, within which in spite of significant heat consumption, spontaneous interaction occurred [45]. The entropic dependence of the polylysine interaction with SF was impressively documented by the strong signal decrease observed for all high ionic strength conditions and less affected by the selected salt to reach this ionic strength (Fig. 7A). In contrast, a substantial impact on the heat signals as a function of the selected salt was observed for the interaction with protamine (Figs. 1e4). SF primary sequence, characterized by repetitive hydrophobic and hydrophilic elements is structurally comparable with synthetic amphiphilic triblock copolymers and results in formation of micellar structures in SF solutions [46]. The heat signal (Fig. 1A) and the hydrophobic probing (Fig. 1A) of protamine and SF shared a striking similarity to analogous studies using cutin hydrolase as model protein and dodecyl sulfate as a micelle forming surfactant [47]. Therefore, the presence of protamine (but not of polylysine) at a ratio of
Contents lists available at ScienceDirect
Biomaterials journal homepage: www.elsevier.com/locate/biomaterials
Deciphering the mechanism of protein interaction with silk fibroin for drug delivery systems Oliver Germershaus 1, Vera Werner 1, Marika Kutscher, Lorenz Meinel* Institute for Pharmacy, University of Würzburg, Am Hubland, DE-97074 Würzburg, Germany
a r t i c l e i n f o
a b s t r a c t
Article history: Received 7 December 2013 Accepted 22 December 2013 Available online 22 January 2014
Silk fibroin (SF) is an exceptional drug delivery carrier with respect to stabilizing, protecting, and delivering sensitive biologics. A synopsis of thermodynamic, static light scattering, hydrophobicity probing, and nanoparticle tracking analyses served as a basis to decipher the mechanism of interaction between SF and two model proteins, protamine and polylysine. The impact of salts aiding (chaotropic), not affecting (neutral), or opposing (cosmotropic) SF unfolding was a major determinant, ranging from complete abolishment to maximal interaction efficacy. Evidence is provided, that the underlying mechanism of the remarkable ability to tailor drug/SF interaction throughout such large ranges and by appropriate salt selection is the control of structural breakdown of SF micelles as present in pure SF ad initium. This study provides a mechanistically justified and hypothesis driven blueprint for future experimental designs addressing the controlled interaction of biologics and SF. Ó 2014 Elsevier Ltd. All rights reserved.
Keywords: Drug delivery of biologics Silk fibroin Mechanism of interaction Thermodynamics
1. Introduction The drug delivery of biologicals is a high growth segment of the pharmaceutical industry, with key drivers of innovation being the stabilization of these sensitive molecules opening a potential for storage at room temperature and deployment of challenging administration routes [1]. Another frequent challenge in the production of current drug delivery systems is the need for processing in organic solvents, at extreme pH values, and mechanical stress all of which potentially challenging the protein’s integrity [2,3]. Silk fibroin (SF), a biopolymer isolated from whole silk cocoons of the silkworm Bombyx mori (L., 1758), has demonstrated superb stabilization for and controlled release of biologicals [1], thereby providing exciting new avenues for drug delivery [4e10] while further opening the interface to biomaterials and/or tissue (re-) generation [11]. Protocols have been presented, allowing manufacturing of protein loaded SF drug delivery systems under mild condition [8] and by translating insights of the natural (allwater based) silk spinning processes into advanced biomaterial and drug delivery system production, respectively [12,13]. Successful development of delivery systems was reported for therapeutically relevant proteins and peptides, including nerve growth factor (NGF) [7], insulin-like growth factor I (IGF-I) [8], fibroblast growth * Corresponding author. Tel.: þ49 931 318 54 71; fax: þ49 931 318 46 08. E-mail address: [email protected] (L. Meinel). 1 These authors contributed equally to the manuscript. 0142-9612/$ e see front matter Ó 2014 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.biomaterials.2013.12.083
factor 2 (FGF-2) [14], bone morphogenetic protein 2 (BMP2) [4,5,15], monoclonal antibodies (mAb) [11], enzymes [16], and others [1,6]. Most of these studies deployed the excellent SF material properties [17e19] and biocompatibility [20,21] for cell growth and differentiation [22,23] and combined these advantageous properties with bio-functionalities through co-processing the biopolymer and the therapeutic or by means of its covalent decoration before or after scaffolding into three dimensional carriers [1,5,24]. More recently, these solid SF drug delivery carriers were complemented by semisolid drug delivery systems and freeze-dried gels for mAb delivery, detailing the design space and mechanism of release by which SF may be used as a drug delivery biopolymer [11,25]. However, the precise mechanism of interaction of SF and biologicals in aqueous solution is still poorly understood and, therefore, scientists were largely forced to follow empirical approaches. Nevertheless, these studies linked better SF loading to positively rather than negatively charged drug molecules, hypothetically assuming Coulomb forces establishing between the negatively charged SF at neutral pH (pI w 4.5) [26] and positively charged, alkaline drugs [9,27]. Other hypotheses explained sustained delivery profiles with hydrophobic attraction, linking the rather hydrophobic patches of SF to the hydrophobicity of small molecule drugs, such as propranolol, or hydrophobic patches of antibodies [1,11]. This study aims at expanding this mostly empirically built knowledge base by elucidating the mechanism of SF interaction with biologicals, using the basic model proteins polylysine and
3428
O. Germershaus et al. / Biomaterials 35 (2014) 3427e3434
protamine, respectively. These proteins were selected based on their net positive charge at neutral pH (facilitating electrostatic forces to the net negatively charged SF). Furthermore, these model proteins were chosen to include some structural variability, with polylysine representing a strongly and homogenously charged polyelectrolyte characterized by significant intramolecular electrostatic repulsion resulting in random coil conformation [28], while protamine provides more structural complexity, including hydrophobic sections/b-sheet [29]. In addition to the variation of the model drug, the effect of specific ions resulting in cosmotropic or chaotropic environment, respectively, on proteineSF interaction was studied.
and 2.5 nm, respectively. Scattered light was observed for 60 s with an integration time of 0.1 s and a data interval of 1.2 s. The samples were produced as described before but volumes were increased by a factor of 10 to allow reliable measurement in a 3.5 ml quartz cuvette (cuvette type 101-QS, Hellma, Müllheim, Germany). 2.5. Zeta potential measurements The surface charge was determined using a Delsa Nano HC (Beckmann Coulter, Brea, CA) using a 658 nm laser and a scattering angle of 15 . The samples were produced as described before but were diluted 1:12.5 for protamine and 1:7.5 for polylysine with ultrapure water immediately before each measurement.
2. Experimental details 2.6. Nanoparticle tracking analysis (NTA) 2.1. Materials Cocoons of B. mori (L., 1758) were obtained from Trudel AG (Zurich, Switzerland). Protamine sulfate salt from salmon grade x, amorphous powder and Poly-L-lysine hydrobromide (polylysine, 15e30 kDa by viscosity) was from SigmaeAldrich (Munich, Germany). All other chemicals used were at least of analytical grade and from SigmaeAldrich if not stated otherwise.
The colloid kinetics of SF and protamine alone, as well as of SF/ protamine mixtures at different mixing ratios, were determined with a NanoSight NS 500 instrument (NanoSight, Wiltshire, UK) using a 40 mW laser at 638 nm and the NTA Software Version 2.3. All samples were prepared as described before, and were diluted 1:10 using the respective buffer/salt solution immediately before each experiment. Movies were cut using Movie Maker (Microsoft, Redmont, WA).
2.2. Preparation of SF solution 2.7. Fluorescence spectroscopy SF solutions were prepared from silk cocoons as previously described with modification [30]. Cocoons were boiled in a stirred aqueous solution of 0.02 M Na2CO3 twice for 1 h each, and then washed with water 10 times. The dried fibers were dissolved in 9 M aqueous LiBr solution at 55 C until completely dissolved, yielding a concentration of 20% (w/v) SF which was filtered through a 5 mmsyringe filter (Versapor, Pall Life Sciences, Washington, NY) and dialyzed against ultrapure water for 2 days, changing the water 5 times and using SpectraPor dialysis membranes (MWCO 6e 8000 Da, Spectrum Labs, Rancho Dominguez, CA). The SF solution had a final concentration of 1.5e2% (w/v) and was stored at 2e8 C. Prior to use, the SF solution was filtered through a 5 mm-syringe filter (Versapor, Pall Life Sciences, Washington, NY).
Fluorescence intensities were recorded using an LS 50B luminescence spectrometer (Perkin Elmer, Waltham, MA) setting excitation at 490 nm and emission at 570 nm. Samples were prepared as follows: 2036 ml of an SF solution at 3.2 mM in the respective buffer/salt solution were mixed with 4 ml of a Sypro Orange (SO) solution, diluted 1:10 in DMSO from a stock solution (total dilution of working solution 50,000). The protamine solution was added in steps of 20 ml each, measuring the fluorescence spectrum at each titration point (vide supra). Relative fluorescence was calculated by setting the fluorescence of the SF solution without protamine as reference (corresponding to a relative fluorescence of 1). 2.8. Fluorescent labeling of protamine
2.3. Isothermal titration calorimetry (ITC) The ITC titrations were performed using a MicroCal iTC200 (GE Healthcare, Buckinghamshire, GB) to measure the heat of reaction when a protamine solution (1.622 mM) was titrated into an SF solution (32 mM) or a polylysine solution (98 mM) was titrated into an SF solution (19 mM) at 25 C and stirred at 400 rpm. Both, SF and titrant (protamine or polylysine) were dissolved in the same medium, composed of a 25 mM histidine buffer, pH 7.6 supplemented with sodium chloride, sodium thiocyanate, or sodium sulfate yielding an ionic strength of 192 mM, respectively. Each injection consisted of 2 ml of titrant injected during 4 s each (first titration at 0.2 ml of titrant injected during 0.4 s), with a spacing of 150 s between each of a total of 20 injections. The cell volume was 200 ml. The first injection was always omitted to minimize the impact of equilibration artifacts, following the manufacturer’s recommendations. All curves were corrected by a blank titration, performed by titrating the respective model protein (protamine or polylysine) in buffer/salt solution without SF.
8 ml of a 1.0 mg/ml solution of Fluorescein isothiocyanate was added drop-wise to 100 ml of a stirred 2.0 mg/ml solution of protamine in 0.1 M sodium carbonate buffer, pH 9.0. The mixture was stirred for 8 h at 4 C protected from light. Afterward the solution was dialyzed against ultrapure water for 36 h, using a SpectraPor dialysis membrane (MWCO 6-8000 Da, Spectrum Labs, Rancho Dominguez, CA) and the solution was freeze-dried for 24 h at 5 C and 0.16 mbar (Alpha 1-4, Christ, Osterrode, Germany). 2.9. Microscopic particle analysis A 32 mM SF solution was mixed with a 1.622 mM solution of fluorescently labeled protamine at a molar ratio of 10.4. An Axio Observer.Z1 microscope equipped with a Plan-Apochromat 40/ 0,95 objective and an AxioCam MRm3 camera (Carl Zeiss Microscopy, Jena, Germany) was used. Excitation wavelength was set to 450e490 nm and emission wavelength to 500e550 nm. 3. Results
2.4. Static light scattering (SLS) Light scattering was performed using an LS 50B luminescence spectrometer (Perkin Elmer, Waltham, MA) at a fixed angle of 90 . Excitation- and emission wavelength and -slits were set to 638 nm
The protamine/SF interaction was studied by titrating SF into a solution of protamine in low ionic strength buffer of 0.31 mM, respectively. Endothermic signals with progressively declining intensity were recorded up to a ratio of 1.5 at which the reaction
O. Germershaus et al. / Biomaterials 35 (2014) 3427e3434
1.6
5 Hm [kcal/mol]
1.4 0
3429
200
10
150
0
100
–10
50
–20
1.2
–5
1 0.8
–10
0.6 –15 0
2
4 6 8 Molar ratio
0 0
10
2
4 6 8 Molar ratio
10
–30
Fig. 1. Interaction of SF and protamine in histidine buffer. (A) Development of isothermal enthalpy (open circles) and SO fluorescence (filled squares) during addition of increasing amount of protamine to SF. (B) Development of right angle light scattering at 638 nm (open circles) and zeta potential (filled squares) during titration of SF with protamine.
enthalpy became exothermic with a maximum of heat liberation observed at a ratio of 3 and a subsequent decline of heat per additional titration, thereafter (Fig. 1A). The change of hydrophobicity of the colloidal solution during the titration as probed by Sypro Orange (SO) correlated to the heat signal, with a continuous increase in fluorescence intensity (hence hydrophobicity of the colloidal solution) at low ratio, rapidly reaching maximum fluorescence at a ratio of 2.2 and continuously declining, thereafter (Fig. 1A). Similarly, light scattering profiles were characterized by a constant increase of scattering intensity up to a ratio of 2.6, followed by a phase of a steep increase in scattering intensity reaching its maximum at a ratio of 3 and rapidly declining, thereafter (Fig. 1B). Zeta potential of pure SF solution was at 25 mV and rapidly increasing until reaching net charge neutrality at a ratio of 3 coinciding with maximal scattering intensity. The zeta potential remained stable and plateaued at about 6 mV (Fig. 1B). The effect of increasing the ionic strength was studied using sodium chloride, a salt having only negligible effects on protein solvation [31e34]. Interestingly, in the presence of sodium chloride exothermic signals were obtained already at early titration points, with maximum exothermicity at a ratio of 1.6, after which the exothermic signals declined with increasing ratio (Fig. 2A). The change of hydrophobicity of the colloid solution (as indicated by relative SO fluorescence) slightly increased at low ratio, peaking at a ratio of 1.1 and gradually declining, thereafter (Fig. 2A). Overall the scattering
1.6
5
1.4 0
200
10
150
0
100
–10
50
–20
1.2
–5
1
0.6 2
4
6
8
10
R
0.8
–10 –15 0
intensity was low as compared to titrations performed in pure buffer (vide supra) and increasing at a ratio of 6.3. Likewise, the zeta potential was substantially reduced with surface charge neutrality being shifted to a higher ratio as compared to the titration in plain buffer (Fig. 2B). The impact of the chaotropic salt sodium thiocyanate replacing sodium chloride was studied at otherwise identical ionic strength (Fig. 3) [31e34]. Analogous to the observation with sodium chloride low ratios had exothermic effects (Fig. 3A). In contrast to sodium chloride, the exothermicity constantly increased up to ratios of 10, whereas the hydrophobicity pattern (relative SO fluorescence) throughout the titration was qualitatively comparable to the results obtained with sodium chloride with a more pronounced decrease at low ratios and maximum hydrophobicity being measured at a ratio of 2.6. The scattering intensity increased constantly with a low slope up to a ratio of 5.9, followed by a substantial increase (Fig. 3B). Interestingly, the scattering intensity did not decrease at higher ratios and in contrast to the other profiles obtained, but leveled off at high scattering intensity. Similar to the data obtained for sodium chloride, the zeta potentials were overall lower than those observed in buffer alone (Fig. 3B). The impact of cosmotropic salts was studied using sodium sulfate [31e 34]. Sodium sulfate radically reduced the interaction of protamine with SF and as detailed by minimal exothermic effects as recorded by ITC, with the hydrophobicity of the colloidal solution being linearly reduced with increasing ratio (Fig. 4A). An increase in light
0 0
2
4
6
8
10
–30
Fig. 2. Interaction of SF and protamine in histidine buffer containing 192 mM sodium chloride. (A) Development of isothermal enthalpy (open circles) and SO fluorescence (filled squares) during addition of increasing amount of protamine to SF. (B) Development of right angle light scattering at 638 nm (open circles) and zeta potential (filled squares) during titration of SF with protamine.
3430
O. Germershaus et al. / Biomaterials 35 (2014) 3427e3434
A
B
1.6
5
1.4 0
10
600
0
400
–10
200
–20
1.2
–5
1
0.6 2
4
6
8
R
0.8
–10 –15 0
800
0 0
10
2
4
6
8
10
–30
Fig. 3. Interaction of SF and protamine in histidine buffer containing 192 mM sodium thiocyanate. (A) Development of isothermal enthalpy (open circles) and SO fluorescence (filled squares) during addition of increasing amount of protamine to SF. (B) Development of right angle light scattering at 638 nm (open circles) and zeta potential (filled squares) during titration of SF with protamine.
scatter was observed up to a ratio of 6.5, remaining stable, thereafter (Fig. 4B). Similarly, zeta potentials increased up to a ratio of 5 and remained constant thereafter, failing to reach surface charge neutrality (Fig. 4B). Particle formation and interaction was followed by NTA on the nanometer scale (Fig. 5; Supplementary video 1). In the case of SF/protamine interaction, the resulting particle size distributions were rather broad, representing a mixture of particles with diameters from hundreds of nanometers up to several micrometers and complicating reliable analysis of single particle tracks (data not shown). Consequently, we focused on particle appearance and particle interaction at the different titration points. NTA of pure SF colloidal solutions generally exhibited broad particle size distributions, ranging from approximately 100 nm to well above 400 nm, with mean diameter in the range of 250e300 nm (data not shown). Particle morphology was dramatically impacted by the SF/protamine ratio and the buffer composition. In pure histidine buffer, a progressive increase in the number of fast moving, i.e. smaller particles was observed up to a mixing ratio of 2 (Fig. 5, A2). At subsequent titration points, formation of progressively larger, slower moving particles commenced and particle interaction increased at mixing ratios above 2, resulting in the formation of duplet and triplet clusters. After addition of sodium chloride formation of small particles was observed at mixing ratios from 1.1 to 3.2 with increasing aggregation of particles, thereafter (Fig. 5, B1e4). The experiments using sodium thiocyanate generally
A
1.6
5
1.4 0
yielded a similar outcome (Fig. 5, C1e4) but in contrast to sodium chloride the number of larger particles was reduced. Essentially the same observation was made for sodium chloride or sodium thiocyanate as compared to histidine buffer and as assessed qualitatively. In contrast, in the presence of sodium sulfate no major changes of particle size or morphology were observed throughout the entire ratio tested (Fig. 5, D1e5). The characterization of particle formation was corroborated by fluorescent microscopy using fluorescently labeled protamine at a protamine/SF ratio of 10.4 (Fig. 6). Particle formation in the micrometer range, as observable by fluorescence microscopy, was found for the histidine buffer and these observations were similar in the experiments conducted in presence of either sodium chloride or sodium thiocyanate, respectively. No particles in the micrometer range were observed in the case of sodium sulfate containing buffer. Particle counting and particle size analysis detailed numerous smaller particles in the sodium thiocyanate and larger but fewer particles in the case of sodium chloride containing buffers, respectively (Fig. 6B). The interaction of SF with basic proteins was further studied using polylysine (Fig. 7). The enthalpy change per titration was higher for polylysine as compared to protamine and complexation completed at substantially lower molar ratios. In contrast to protamine, polylysine interaction with SF was endothermic throughout the main part of the titration. In spite of these differences, the general interaction pattern between SF and polylysine decreased in the order of buffer
B
200
10
150
0
100
–10
50
–20
1.2 –5
1 0.8
–10
0.6 –15 0
2
4
6
8
10
0 0
2
4
6
8
10
–30
Fig. 4. Interaction of SF and protamine in histidine buffer containing 64 mM sodium sulfate (identical ionic strength as 192 mM sodium chloride). (A) Development of isothermal enthalpy (open circles) and SO fluorescence (filled squares) during addition of increasing amount of protamine to SF. (B) Development of right angle light scattering at 638 nm (open circles) and zeta potential (filled squares) during titration of SF with protamine.
O. Germershaus et al. / Biomaterials 35 (2014) 3427e3434
3431
Fig. 5. Evaluation of nanoparticle characteristics during the titration in different buffer and salt solutions using nanoparticle tracking analysis. Titrations were performed in histidine buffer, sodium chloride, sodium thiocyanate and sodium sulfate, respectively. Representative video stills are shown for the SF to protamine ratio of 1.1, 2.1, 2.6 and 3.7 for histidine buffer and 1.3, 2.6, 3.2, and 4.5 for all other salts deployed, respectively. Arrows highlight duplex particles. Original videos are uploaded as supporting information.
larger than sodium chloride larger than sodium thiocyanate larger than sodium sulfate (vide supra; Fig. 7A). Static light scattering revealed an increasing scattering intensity up to a ratio of approximately 0.37 for pure buffer, 0.16 for sodium chloride, 0.37 for sodium thiocyanate and 0.32 for sodium sulfate addition, respectively. In all experiments, the scattering intensity plateaued, thereafter (Fig. 7B). Supplementary Video related to this article can be found at http://dx.doi.org/10.1016/j.biomaterials.2013.12.083 4. Discussion These results detailed the interaction between SF and model protein drug, paving the way for a mechanistically driven and optimized drug loading on this biopolymer. The exceptional progresses seen for SF in the past provided the mechanistic foundation leading to predicted mechanical properties [19,35], a perplexing versatility in form, structure and function [36,37], and demonstrated in vitro and in vivo biocompatibility [8,20,38]. These successes among others sparked the interest in this biopolymer within a rapidly growing scientific community. Along with the insights outlined here and building off previous reports [1,10], the interpretation on the mechanism of drug loading will support the important research at the interface to drug delivery and might
expand the deployment of SF for the purpose of drug delivery through the rational hypothesis presented here within. Our results suggested, that protein drugeSF interaction at low molar ratio was by (i) association of protein molecules to free SF molecules resulting in nanocomplex formation, and to (ii) preexisting micelles as present in pure SF solution. The nanocomplex formation (i) is assumed to occur continuously throughout adsorption of the basic model protein (Fig. 5) and progressed towards charge neutralization of the ad initium negatively charged SF (Fig. 1A). At neutrality, aggregation commenced leading to abrupt phase separation (Fig. 1B). Thermodynamically, nanocomplex formation/coacervation is typically entropically driven, dominated by the release of counterions [39], a finding which had been corroborated here within by observing reduced (Fig. 1B and 2B) or completely abolished interaction with increasing ionic strength (data not shown) and as described before for similar systems [40e 42]. Mixing of polycations and polyanions typically sparks two events, (i) the formation of soluble or insoluble interpolymeric complexes paralleled or followed by (ii) coacervation and phase separation [43,44]. Therefore, the endothermic signals at low mixing ratio (Fig. 1A) may at least in part reflect ion pairing, an interpretation supported by the increasing formation of nanocomplexes and ultimately phase separation (Fig. 1B). This
3432
O. Germershaus et al. / Biomaterials 35 (2014) 3427e3434
(Fig. 5) (iii) flipping into an exothermic phase at a ratio of 2e3, and (iv) ultimately fading into a terminal phase of diminishing exothermic signals and decreasing hydrophobicity, likely a result of shielding effects of adsorbed protamine on the surfaces. Therefore, a mechanism of drug loading was detailed, within which structural dynamics of SF in presence of protamine are controlled by proper salt choice, providing a convenient tool for future manufacturing processes while deploying compendial or largely uncritical salts [48]. We hypothesized, that these differences in protamineeSF interaction result from stabilization or destabilization of SF micelles under cosmotropic or chaotropic conditions, respectively [18,27,31,33,34,39,49e51]. The relative importance of specific ion effects can be deduced from conditions of the natural silk spinning process [52] as well as from effects of different cosmotropic or chaotropic salts on stability of amphiphilic block copolymer micelles [53,54]. The increase in ionic strength by sodium chloride reduced the zeta potential and hence electrostatic interaction (Fig. 2B) while the impact on SF micelle stability was minimal (Fig. 2B; 5B). In contrast, chaotropic environments substantially impacted hydrophobic collapse of SF leading to micelle destabilization (Fig. 5C), efficient formation of coacervates (Fig. 6) and a continuous increase of exothermic signals with increasing ratio with no detectable saturation (Fig. 3A). The opposite was observed for cosmotropic conditions under which heat turnover
A
B
300 Buffer NaCl NaSCN
Particle count
250 200 150
A
100 50 0 0
5
10 15 20 25 Ferret diameter [μm]
30
Fig. 6. Microscopic analysis of complex coacervate at a protamine/SF mixing ratio of 10.4. (A) Representative micrographs of particles formed in histidine buffer (top left), sodium chloride (top right), sodium thiocyanate (bottom left) and sodium sulfate (bottom right). Scale bar represents 25 mm. (B) Analysis of Ferret diameter of particles in histidine buffer (open circles), sodium chloride (filled squares) and sodium thiocyanate (filled diamonds).
endothermic interaction at low ratios was principally corroborated for polylysine (although strikingly different in absolute terms as compared to protamine; Fig. 7). Previous studies reported similar interaction between polylysine and the (negatively charged) polysaccharide pectin, within which in spite of significant heat consumption, spontaneous interaction occurred [45]. The entropic dependence of the polylysine interaction with SF was impressively documented by the strong signal decrease observed for all high ionic strength conditions and less affected by the selected salt to reach this ionic strength (Fig. 7A). In contrast, a substantial impact on the heat signals as a function of the selected salt was observed for the interaction with protamine (Figs. 1e4). SF primary sequence, characterized by repetitive hydrophobic and hydrophilic elements is structurally comparable with synthetic amphiphilic triblock copolymers and results in formation of micellar structures in SF solutions [46]. The heat signal (Fig. 1A) and the hydrophobic probing (Fig. 1A) of protamine and SF shared a striking similarity to analogous studies using cutin hydrolase as model protein and dodecyl sulfate as a micelle forming surfactant [47]. Therefore, the presence of protamine (but not of polylysine) at a ratio of
