Microporous drug-eluting large silk particles through cryo-granulation

Abstract

A facile method for the preparation of large, microporous, drug-loaded particles is presented. High shear bollus injections of silk with cross-linker and drug colloids into super-cooled hexane were utilized to trigger phase separation of silk droplets, followed by immediate freezing at −60°C. A subsequent −20°C freeze-thaw of the frozen droplets resulted in self-assembly (crystallization) of the silk. The silk particles developed an internal interconnected microporous morphology with 0.1-10 µm in diameter pores. The silk particles ranged in diameter from 100 to 1,300 µm, with particle mean diameter and polydispersity controlled by the starting concentration of the cross-linking agent and silk, the rheology of the reaction mixture, and the injection pressure (80 - 300kPa). Cryogranulation provided a one-step process to produce microporous meso-scale silk particles with encapsulated drugs, such as doxorubicin chloride (DoxR), tobramycin sulfate (TS), kanamycin sulfate (KS) or gentamicin sulfate (GS). Almost 100% drug encapsulation efficiency was achieved in the process, and subsequent release profiles depended on the starting concentration of both the drug, silk, and pH of the elution medium. Kirby-Bauer tests and bioluminescent imaging confirmed the retention of anti-bacterial potency of the antibiotics pre-encapsulated in the cryo-particles, and macroparticles cytocompatibility towards human fibroblast and kidney cells.

Graphical Abstract

Meso-scale microporous drug-loaded silk fibroin particles were generated via high-shear dispersion of liquefied silk proteins in super-cooled non-aqueous medium. The silk spheres ranged in size from 300 to 1,000 µm and were chemically or physically cross-linked and retained therapeutic potency of the encapsulated drug.

Introduction

Regenerated silk fibroin (RSF) is a protein biopolymer with biocompatibility, aqueous processability, and useful mechanical characteristics for biomaterials in the forms of films, hydrogels, sponges, fibers, and particles [1–8]. RSF-based particles (RSFPs) with tunable morphologies and physical-chemical properties are useful as drug delivery systems (DDSs) [9, 10, 49]. For example, RSFPs less than 200 nm in diameter can be used as intravenous DDSs with extended half-lives in circulation [11], for controlled pharmacokinetics [12]. Particles several microns in diameter can be used subcutaneously and intravenously to localize at injection sites [13, 14, 15]. Large biopolymer-based particles with diameters i 50 to 1,000 µm have a broad range of potential applications in photonic materials, as chromatography column solid supports for separations, in field-responsive rheological fluids, and as injectable DDSs for chemotherapeutic treatment of tumors and bacterial infections [54]. Spherical and cubical meso-particles for drug release are conventionally produced from gelatin, polyvinyl alcohol, polyacrylamide and silicone [29]. Polyacrylic and polyvinylalcohol-co-acrylamide spheres, 100 to 800 µm in dimeter, have been produced by double emulsion polymerization [60]. Microfluidic coagulation and mechanical grinding of PVA-sheets [29] produce drug eluting polyvinyl alcohol-based particles of about 180–900µm in diameter. The coagulation approach has been successfully applied to fabricate silk sericin/lignin blended beads up to 2,000 µm in diameter for the removal of Cr(IV) from waste water [55].

Silk-based materials bind doxorubicin (DoxR), with beta-sheet content of the SF controlling release kinetics [12]. Silk-based micro- and meso-spheres has been explored, including those generated by salting-out [17], self-assembly [18], spray-drying [37–40], and mechanical grinding [19]. Microfluidic techniques using laminar flow encapsulation [1, 10] and electrospinning [20] offered effective control over the particle size over the range of 100–2,000 μm [12]. The bulk emulsification-based assembly of silk particles by a drop dissolution technique produced spheres with monolith internal morphology [18]. Continuous production of polymeric meso-particles by microfluidic approaches using a double emulsion processes required precise tuning of the injector nozzle diameter and control over the rheology and fluid dynamics of the injected polymer solutions [1, 3, 23]. In addition, careful selection of the starting monomer-to-initiator ratio, stirring rate, temperature and pH of the emulsification bath was required [48]. A faster granulation technique based on high shear phase separation of liquefied RSF in supercritical CO₂ produced monodisperse silk nano-particles [41–43], suggestive of an alternative for the assembly of silk. Hexanes have been used as dispersants in microfluidic fabrication of 100 to 600 µm meso-spheres of polylactic acid, starches, polyamines, and various proteins [32, 33, 50], and can be substituted for the expense of scale-up using CO₂-assisted setups for high shear emulsification of silk solutions. Here, we simplified the injection-based droplet approach and added high shear self-assembly of the liquefied silk droplets during forced continuous injections into hexane dispersant to improve the granulation of silk particle formation.

Microporosity in large polymer particles [52, 53] has been exploited to gain additional control over the release of therapeutics and such morphological control is often achieved by gas bubbling [25], solvent evaporation [53], lyophilization [2, 10] or leaching [31]. Macroporous silk hydrogels could also be achieved by cryotropic gelation of reactive silk/cross-linker colloids within an unfrozen liquid microphase (UFLMP) [22, 32, 23]. The ice-templating effect of cryogelation was realized in micro-porous PVA-based beads sized 200–600 μm [32] but was not optimized for the formation of silk-based particles. Various chemical or physical triggers for silk fibroin cryogelation in UFLMP, including chemical cross-linking of RSF with diglycidyl ethers [33, 34], or the β-sheet crystallization of RSF initiated by sonication, vortexing or additions of organic co-solvents [35] can be used for cryogenic stabilization of large particles of silk. For example, mixtures of cetone or ethanol, as well as oligomeric polyethylene glycols [3, 21, 36] often trigger physical curing of silk and could be employed at sub-zero conditions to obtain physically stabilized RSFPs in situ, pre-loaded with drug substances.

The goal of the present study was to develop a facile technological approach for the fabrication of uniformly and large sized silk particles to add to the options above, but with a focus on meso- and macro-scale particles. Such systems would be useful for many applications, including drug delivery, demonstrated here with DoxR-encapsulated large (meso-scale) RSFPs. High shear coagulation of the dispersed RSF, cross-linker and drug in supercooled hexanes achieved the self-assembly of large silk droplets. The size and shape of the particles was impacted by −80°C-cooled coagulation bath and a-20°C freeze-thawing process. The size distribution, morphology and cross-linking efficiency of the resulting RSFPs depended upon the concentration of the covalent cross-linking agents (e.g., EGDE, pPGDE), or the inducer of silk physical gelation, such as acetone or polyethylene glycol (PEG-400), as well as the concentration of the silk.

Experimental

Materials:

Hexanes, for analysis (Thermo Fisher Scientific Co., US), ethylene glycol diglycidyl ether (EGDE, Tokyo Chemical Industry Co., Ltd.), poly(propyleneglycol)dyglycidyl ether (pPGDE, average Mn 500), gentamicin sulfate; acetone for HPLC, GC and residue analysis, ≥99.9%, tobramycin sulfate (all Sigma-Aldrich, 650501); kanamycin sulfate (Thermo Fisher Scientific Co., US); borate buffer was prepared using BupH™ Borate Buffer Packs (Thermo Fisher Scientific Co., US).

Silk fibroin purification:

To obtain regenerated silk fibroin (RSF), Bombyx mori cocoons were extracted for 45 min in boiling aqueous solutions of sodium carbonate (0.02 M), then rinsed quickly with distilled water three times to remove the excess salt and sericins. The extracted silk was loosened and dehydrated overnight at 25°C under airflow. The dried silk mass was then dissolved in lithium bromide solution (9.3 M) for 4h at 60°C with occasional gentle agitation to yield a 20% w/v solution which was then cooled to 25°C and dialyzed against distilled water for 48 h using cellulose dialysis tubing (molecular weight cutoff, MWCO, 3,500, Fisherbrand by Fisher Scientific Co., Pittsburgh, PA) to remove the salt. Thus obtained aq. solution of fibroin (ASF) was then subjected to the dialysis against borax buffer (pH 8.5) for 24 hours. The resulting basic solution of fibroin was checked visually for optical clarity and centrifuged and decanted 3 times at 9,000 rpm at 4°C to remove residual solids or debris. The final concentration of aqueous silk solution was 9.5 % wt., determined gravimetrically based on the residual dry solids weight.

Cryogranulation:

Each batch of RSFPs was produced using the custom designed silk cryo-granulation assembly comprised of the coagulation bath, cooling jacket, 5mL syringe, pressure switch unit, and the air compressor (Supplementary Figure 1). The coagulation bath was comprised of a glass cylinder (d×h: 90×230 mm) filled with 1200 mL of hexane? and positioned coaxially inside a plastic beaker (d×h.: 150×180 mm) packed with crushed dry ice. The granule collection unit was comprised of a brass basket (d×h: 86×45 mm) draped with cellulosic liquid-permeable membrane (LensX™90 non-woven rayon blend tissues, Berkshire Corporation) deposited in the bottom section of the coagulation column. The setup was seated on top of an IKA 3810001 (Cole-Parmer Instrument Company, LLC) magnetic stirrer to maintain constant agitation of the freeze-hardening hexane bath at 100 rpm. The temperature of the pre-cooled coagulation bath was maintained at −60 to −65°C. The high shear injection unit was comprised of a 5-mL Cellink pressure tight syringe fitted with a 27G needle; the inlet of the syringe was connected by polyethylene tubing (inner d=3 mm) to the pressure unit from the air compressor and a digital pressure controller built into the Inkredible 3D-Bioprinter system (Cellink Inc.). Gelation of silk was initiated by blending the desired amounts of ASF in BupH (pH 8.5) with either EGDE, pPGDE, acetone or PEG-400 by brief vortexing at 500 rpm for 10 seconds. Final concentrations of SF ranged 4–5.5%wt., concentrations of EGDE or pPGDE ranged 5–30 mmol of epoxy groups per 1 g of dry SF, concentrations of acetone and PEG-400 ranged 5–20%vol. The reactive suspensions were then transferred to the injection syringe. Syringe was then connected to the air compressor pre-tuned to the desired pressure from 80 to 300 kPa and the needle tip was fixed 10 mm above the surface of the hexane bath. The reactive systems were then immediately injected into the supercooled coagulation bath at a desired pressure, and the frozen silk droplets formed from the contact of the pressurized silk/cross-linker masterbatch with the supercooled hardening bath were collected at the bottom of the column. The frozen droplets were incubated in hexanes for 5 minutes then slowly withdrawn, quickly wrapped in the rayon cloth and incubated for 20 hours at −20°C. Traces of hexanes were stripped off from the RSFPs by air convection in the cooling chamber and later by lyophilization for 24 hours at 0.06 mbar. The RSFPs were then washed with 200-x excess of DI water to remove any unreacted protein and the cross-linkers. The cryo-RSFPs obtained could further be fractioned by particle size via dry sieving using nested sieve columns, each sieving plate 80 mm in diameter, with the following screen openings (in µm): 1,400; 425; 355; 212; 180.0; 150.

Encapsulation of DoxR and antibiotics in RSFPs:

Doxorubicin chloride (DoxR), gentamicin sulfate (GS), tobramycin sulfate (TS), or kanamycin sulfate (KS) were dissolved in BupH buffer (pH 8.5) and blended with the stock ASF and 10%vol. acetone to obtain final drug concentrations in range from 1 to 20 mg/mL. The resulting drug/RSF/cross-linker was immediately injected at 150 kPa into the supercooled hexane bath. The frozen RSF/drug droplets were harvested from hexane, incubated for 20 hours at −20°C and freeze-dried to remove traces of hexane, water, and acetone. The total antibiotic content in the RSF beads was determined gravimetrically by subtracting the average dry weight of the blank RSFPs from the average dry weight of drug-loaded RSFPs.

Preparation of RSFPs modified with nRSFs:

To synthesize silk nano-particles, 45-minute extracted RSF solution in BuPh buffer (pH 8.5) was used [21]. In brief, the ASF was diluted to 5% wt. of RSF and mixed dropwise (0.2 drop/s; ca. 100 µL/drop) to pure acetone, with the final acetone fraction accounting for 80% (v/v). Silk particulated were left to precipitate overnight, then centrifuged for 30 min at 9,000 rpm, followed by aspiration of the supernatant and re-suspension/vortexing of the pellet in 35 mL of DI water. Crude SF suspensions were then sonicated for 30 s at 30% amplitude with a Branson Digital Sonifier 450 (Branson Ultrasonics, Danbury, CT, USA). The centrifugation-washing-suspension cycle was repeated twice to obtain silk nano-suspensions (nRSF). A 0.5-mL aliquot of the n-RSF was freeze-dried and the dry solids content was determined gravimetrically. Prior to cryo-granulation, the n-RSF suspension was subjected to a 15s sonication at 30% amplitude followed by size and zeta potential assessments with a Microbrook 2000L laser particle-size analyzer (BrookHaven Instruments Corporation, NY). A calculated volume of the n-RSF suspension was mixed with the DoxR for 2 hours at 25°C with constant agitation (200rpm). The resulting system, containing free DoxR and DoxR-loaded nRSFs was added to the ASF stock to reach concentrations of n-RSF in the system in range from 3 to 6 mg/mL. These mixtures were immediately mixed with 10% vol. acetone and subjected to the above cryo-granulation process at 150kPa, to produce nano-modified DoxR-loaded RSFPs.

Fourier transform infrared (FTIR) spectroscopy:

The IR-Spectra of the lyophilized RSFPs as well as pure RSF and cross-linking agents (in liquid from) were acquired using a single bounce diamond attenuated total refractance (ATR) module on a Fourier-transform infrared (FTIR) spectrometer (JASCO FTIR 6200 Spectrometer, JASCO, Tokyo, Japan) equipped with a MIRacle Ge crystal cell in the reflection mode. Air was used as the reference and subtracted automatically from sample readings. Each sample was analyzed in the frequency range from 400 to 4,000 cm⁻¹ with each measurement adding 32 interferograms with a resolution of 4 cm⁻¹. The fractions of secondary structures, including random coil, alpha-helices, beta-pleated sheets, and turns, were ascribed to the amide I region from 1,595 to 1,705 cm⁻¹. Random coil, alpha-helices and β-turns were ascribed to the absorption bands in the frequency ranges of 1,638–1,655 cm⁻¹, 1,656–1,663 cm⁻¹ and 1,663–1,695 cm⁻¹, respectively [44].

Differential scanning calorimetry (DSC):

An estimate of 3 mg of lyophilized RSFPs was packed in aluminum pans and measured in a Q600 TGA/DSC instrument (TA Company, New Castle, DE). The chamber of the instrument was purged with dry nitrogen at 50 ml/min. The instrument was calibrated with indium for heat flow and temperature. Aluminum and sapphire reference standards were used for calibration of the heat capacity. The measurements were performed using a standard mode at a heating rate of 8°C/min.

Rheology:

The shear viscosity of each freshly prepared injectable RSF/cross-linker system was measured by rotational rheometry (AR 2000, TA Instruments) using two 50mm circular titanium plates at 25°C. The normal force applied on the sample during lowering of the top plate was limited to 0.1 N. The shear rate was linearly increased from 0.1 to 5,000 1/s at the steady state deformation of 15%.

Size distribution of RSFPs:

Direct estimation of mean size distribution of RSFPs was performed using a combination of fluorescence microscopy and the following automated image processing with the ImageJ. The micrographs of cryo-RSFPs were recorded using a BZ-X710, All-in-One Fluorescence Microscope (Keyence, USA) with a DAPI filter and at 42x magnification. The recorded scans were further stitched and processed by contrasting and grain analysis using the ImageJ open access software.

Loading Efficiency and Release of DoxR:

A 100mg load of DoxR/RSF particles with sizes between 200–800 µm was sealed in water-permeable rayon membranes and incubated in 5 mL of PBS buffer at 37°C and constant agitation of the elution medium at 50 rpm. At determined time points, 100 μl aliquots of the elution medium were collected and the absorbance at 477nm (DoxR) and 480nm (DNR) was measured. The concentration of DoxR at each time point (20 mins to 14 days) was determined by extrapolation of the absorption to standard curves obtained for known concentrations of drug from 0.01 to 1 mg/mL. Non-drug cryo-RSFPs were used as negative controls.

Residual antibacterial activity:

Antimicrobial activity of tobramycin sulfate (TS), gentamicin sulfate (GS) and kanamycin sulfate (KS) encapsulated in RSFPs sized 100–500mcm was assessed by Kirby-Bauer disc diffusion tests, performed according to methods outlined in the NCCLS Approved Standard M2-A7. A stock suspension of bioluminescent S. aureus Xen29 was prepared in sterile physiological solution and turbidity (expressed as optical density; OD) was adjusted against the McFahrland standard to match the 0.5 level. A sterilized cotton swab was immersed in the bacterial suspension, and a lawn of bacteria was applied on the Luria Bertani (LB) plates. A 10-mm biopsy punch was used to cut two circular fragments from each of the freshly infected plates, and both fragments were coated with antibiotic-loaded RSFPs by quick blotting against the pool of particles. The freshly mounted samples were inserted back into the circular slots of the bacteria-seeded plates and further cultured for 48 hours at 37°C. In vitro imaging of the plates was performed after 24 hours of incubation using a Caliper IVIS Lumina II system (Caliper life Science, America). The exposure time (1s), excitation filter (430nm), and emission filter with emission wavelength from 575 to 650 nm (DsRed) were set prior to detection.

Scanning electron microscopy (SEM):

The morphology of the particle’s surface and inner phase of was visualized by field emission scanning electron microscopy (FE-SEM, JEOL, JSM- 6490, Japan) at a voltage of 5 kV. Before analysis, the freshly lyophilized particles were positioned onto a carbon paint adherent to the aluminum sample holder. Cryo-sectioning of large (500–1,000 µm) meso-particles was used to prepare samples for the in-depth study of microporous morphology: RSF-particles were quenched in liquid nitrogen and manually crushed using a pre-cooled porcelain mortar and pestle. The resulting crushed samples were freeze-dried and fixed on the sample holder. To assess the sponge morphology of the macroscopic cryogel scaffolds, the fully swollen and then vitrified cryogel samples were fractured and the fragments freeze-dried. All samples were made conductive by sputtering a thin layer of gold onto their surface.

Preparation of the RSF-scaffolds:

The silk-based scaffolds were fabricated using the cryogelation protocol to closely mimic the conditions applied to RSF during the cryo-granulation procedures. Solutions of RSF were mixed with either EGDE or acetone, to give a final 4.5%wt. of RSF, 30mmol/g of EGDE or 20%vol of acetone. These mixtures were molded in a 48-well plate and frozen using dry ice as coolant. The vitrified samples were incubated at −20°C overnight in hexane, then thawed at 4°C and washed extensively with DI water to obtain the chemically (EGDE) or physically (acetone) cross-linked RSF-sponges, sized 8mm in diameter and 5mm thick, with the pore diameter in range from about 30 to 120µm based on SEM images.

Cell Culture:

Immortalized human renal cortical epithelial cells (RPTEC/TERT1, ATCC, Manassas, VA, USA) and human fibroblasts (HNF) were cultured in DMEM:F12 (ATCC), 5pM triiodo-L-tyronine sodium salt (Sigma-Aldrich, St. Louis, MO, USA), 10 ng/ml recombinant human epidermal growth factor (Life Technologies, Grand Island, NY), 1 % ITS (Life Technologies), 25 ng/ml prostaglandin E1 (Millipore, Billerica, Ma), 25 ng/ml hydrocortisone (Sigma-Aldrich), 0.1 mg/ml G418 (Life Technologies), and penicillin-streptomycin (Life Technologies) [45]. The RSF sponges were sterilized by autoclaving and then seeded with 2.5×10⁵ cells/scaffold (P=4). The cultures were maintained for 21 days with the RPTEC media.

Cell Viability:

Cell viability was assessed on Days 10 and 21 using a LIVE/DEAD™ Viability/Cytotoxicity Kit for mammalian cells (L3224, Thermo Fischer Scientific). PBS (300 µl) supplemented with 10 mM calcein AM green and 1 mM ethidium homodimer-1 was added to the cells and incubated for 45 min. The samples were visualized using a BZ-X710, All-in-One Fluorescence Microscope (Keyence, USA) and confocal Laser Scanning Microscopy (Leica TCS SP5, Germany)

Statistics:

Particle size measurements were obtained in triplicate for two parallel batches of RSFPs, fabricated with the given set of parameters, with and expressed as mean ± standard deviation (SD). Rheological properties of the silk/cross-linker colloids and the drug release kinetics were acquired as three parallel measurements.

Results and Discussion

The main steps of the batch-based silk cryo-granulation process are presented in Supplementary Figure 1. The mixture of precursors was blended, consisting of RSF, a cross-linking agent chosen from EGDE, pPGDE, acetone or PEG-400, and one of the target therapeutic compounds chosen from doxorubicin chloride (DoxR), gentamicin sulfate (GS), kanamycin sulfate (KS) or tobramycin sulfate (TS). The mixture was then injected through a 27G-needle at pressures of 50–300 kPa into the hexane bath refrigerated at ca. −60 to −65°C. The uniqueness of the current granulation technique relied on the instantaneous splintering of the reactive RSF jet into discrete silk droplets upon contact with the water immiscible organic phase. This approach contrasts with the microfluidic processes where droplet separation is s via a laminar take-off of the liquefied RSF within the co-flow granulation devices [12, 32, 48]. The prior continuous microfluidic methods exploit the Rayleigh-Plateau instability within the stream of RSF to trigger droplet formation. Here the size and morphology of the resulting particles are controlled by the viscosity of the contact liquids [1, 10, 50], and the discrete droplets are achieved by RSF solutions with viscosities above 10P [12, 25]. Bollus injections of the plain 3–6%wt. RSF solutions produced continuous fibers 400–1,000µm thick. Dynamic rheometry showed that the pure RSF solutions had shear viscosities in range 8–9 P, several times higher than 0.2–3.5 P for the RSF/cross-linker mixtures (Fig. 1, Supplementary Table 1). Formation of discrete droplets was enabled in the reactive mixtures of RSF with either 10–20%vol. of acetone, 5–30 mmol/g of EGDE or PPGDE, or 5–15%vol. PEG-400, reduced shear viscosity of 0.2–4P and densities of 1.0–1.15 g/cm³ (Fig. 1). This observation highlights the difference between cryo-granulation and microfluidic techniques in terms of rheological properties.

Figure 1.

Rheological properties of the RSF/cross-linker mixtures for cryo-granulation; density of each batch was assessed gravimetrically; shear viscosity estimated based on the readings obtained on the rotational viscometer at a fixed 15% deformation and at a frequency range between 0.1-100 rad/s.

Hexane has a melting point between −94 and −96°C [56] and remains liquid upon cooling with dry ice to −60°C and −65°C, enabling rapid separation of the RSF 50–2,000 µm droplets and stabilization of external shape and dimensions through freezing. Unlike the −20°C-cooled petroleum ether bath implemented for the continuous production of PVA-based meso-spheres [32], the process described in the present work supports processing of up to 10 mL/min of the starting RSF mixture, significantly faster compared to the other co-flow assisted jetting devices [12, 32, 48]. The residual hexane was successfully removed from the frozen RSFPs by air convection in the freezing chamber, followed by 24-hour lyophilization.

Concentrations of RSF above 6% wt. resulted in premature gelation and clogging of the injector’s needle, and below 2.5%wt. RSF poorly shaped granules were produced with a tendency to shrink and disintegrate upon drying (Fig. 1D). The molecular weight of silk, as well as the rheological behavior can be controlled to some extent based on molecular weight of the silk [12]. Thus, we used reactive mixtures with 30mbs (higher molecular weight than the 60mbs) RSF with shear viscosities between 10–15 P produced continuous fibers, similar to those obtained from the 6%wt 45mbs RSF. Injections below 150kPa resulted in dripping of the reaction mixture, yielding large silk droplets which did not demonstrate further break up upon contact with the hexane bath and quickly froze, yielding spherical beads of around 2,000 µm. In this work we assessed the morphology and physico-chemical properties of the particulate materials produced only in course of the jetting regime Dripping-mode injections were observed in case of the 45mbs RSF-containing systems despite viscosities as low as 4–0.1P, mixtures of 30 mmol/g of EGDE or PPGDE or 15%vol of PEG-400 to RSF (especially at concentrations above 5%w), and lead to rapid precipitation of the protein which u tended to clog the needle (Supplementary Table 2). Similar issues were found with the acetone/RSF mixtures, which typically remained cryo-injectable at pressures above 100kPa, while dripping at pressures below 80kPa.

Apart from the emulsification of the starting RSF solutions, the additions of either diepoxides (EGDE, pPGDE) or desolvating agents (acetone, PEG-400) facilitated further solidification of silk within the frozen droplets. Structural stabilization of silk was achieved via internal cryogelation by a 20-hour freeze-thaw aging of the quenched droplets at −20°C. Cryogelation of RSF was triggered by covalent cross-linking with 5 or 30mmol/g of EGDE [33, 34]. Identical concentrations of the oligomeric pPGDE were used to differentiate between cross-linker type and the morphology of the resulting epoxide-cross-linked RSFPs. Physical cryogelation of RSF within the frozen droplets was achieved through addition of acetone or PEG-400, which initiate the transition of the water-soluble Silk I conformation (characterized by a mixture of random coil, α-helix and β-turn structures), into stable and water insoluble Silk II, predominantly composed of the β-sheets [7]. The lyophilized RSFPs, cross-linked both physically and chemically, were water-insoluble and had a powdery consistency. According to the post-granulation gravimetry, internal cryogelation with acetone resulted in gel fraction yields (mass ratio between the starting dry weight of the RSF being cross-linked and the final RSFPs weight) in range from 85 to 95%, whereas cryogelation in the presence of PEG-400 cross-linked between 68 and 75% RSF, while the EGDE- and pPGDE-assisted cryogelation cross-linked between 45 and 51% and between 59 and 74% of the starting silk, respectively. Thus, acetone/RSF mixtures were used as the matrix media to pre-disperse drug compounds to produce drug eluting beads.

The FTIR-spectra of lyophilized water soluble RSF (Fig. 2, curve 1) showed characteristic vibration bands between 1,630 and 1,650cm⁻¹ for amide I (C-O stretching) indicating the presence of primarily random coil and/or α-helix conformations, 1,540–1,520cm⁻¹ for amide II (secondary N–H bending) and 1270–1230cm⁻¹ for amide III (C–N and N–H functionalities) [4]. The RSFPs cross-linked with acetone were similar to that of the lyophilized RSF and contained chemical shifts of the above absorption frequencies as shoulders to the peaks at 1,685 and 1,622 cm⁻¹, 1,580.0 and 1,238.4 cm⁻¹, respectively, reflecting the formation of β-sheet structures (Fig. 2, curve 2). The FTIR-spectra of the EGDE- and PPGDE-stabilized RSFPs (Fig. 2, curves 3 and 5, respectively) with reference to the spectra of the pure cross-linkers (Fig. 2, curves 4 and 6, respectively) displayed main peaks at 1,623 cm⁻¹ assigned to a β-sheet conformation, as well as shoulders at 1,660 and 1685 cm⁻¹, which could be assigned to α-helix and β-turn conformations, respectively. This indicated a conformational transition from random coil (Silk I) to a β-sheet (Silk II) structure in the frozen droplets of RSFP precursors. The build-up in β-sheet content in silk cryogels had been shown to arise from the cryo-concentration effect causing a nearly 10-fold increase of RSF concentration within the unfrozen liquid microphase [33]. This process was accompanied by the loss of molecular mobility of silk macromolecules both due to the physical entanglements and covalent etheric cross-links between the protein chains [34]. Diepoxide cross-linking of RSF can be traced by new bands at 1,040−1,100 cm⁻¹ upon gelation, which were assigned to the ether stretching bands of EGDE and PPGDE (Fig. 2, curves 3&4, respectively) intermolecular cross-links.

Figure 2.

(A) ATR-FTIR spectra of: (1) freeze-dried silk fibroin before and after cryogelation triggered by (2) acetone, (3) EGDE or (5) PPGDE. FTIR spectra of chemically cross-linked RSF for (4) EGDE and (6) PPGDE. DSC characterization of silk microspheres at (D) −10 to 350°C, and (C) 160 to 245°C

Differential scanning calorimetry was used to characterize the thermal properties of the RSFPs. All particles released water, as indicated by a broad endothermal peak around 60–80°C. Further, heating the silk samples to 350°C resulted in thermal degradation of the cross-linked protein matrix as indicated by the sharp exotherms at 285–295°C (Fig. 2,b). Cryogelation of RSF caused an increase of the RSFP glass transition temperature by ca. 15–20°C, compared to the non-treated silk (Fig. 2,c). This result indicated that the intermolecular interactions between the amorphous domains of RSF were restricted by the cross-links induced by exposure to either diepoxides or acetone. Cross-linking of the RSFPs with acetone yielded the highest Tg values, up to 205°C, compared to the glass transition at ca. 198–200°C for the pPGDE, PEG-400EGDE-cured RSFPs, which could indicate β-sheet rich crystalline fractions of silk fibroin in these samples [51, 56].

The dynamics of droplet break up, as well as droplet size and shape, are governed by the competition between the interfacial tension holding the jet of the silk colloid intact, and the Rayleigh-Plateau instability within the focused jet, which increases at higher flow rates [25, 3]. The cryo-granulation used here controls the size and polydispersity of the final RSFPs by both flow rate of the reactive colloid (injected under the pressures tuned within the range of 80 to 300kPa) and the rheological properties of the colloids, in turn, defined by the cross-linker type and the cross-linker/RSF mass ratio. Figure 3 shows representative RSFP fluorescent micrographs converted into binary offsets and analyzed for mean particle size, expressed as the Feret diameters. The influence of the fabrication parameters on the resulting size distribution of RSFPs was fingerprinted by the relative frequency profiles (Fig. 3&4). The size distributions of water-swollen RSFPs could be varied over a wide range, between nominal D10/D90 cutoffs of ca. 200 and 1,000 µm, depending on the injection pressure and the composition of the injectable system (Supplementary Table 2).

Figure 3.

Estimation of mean RSFPs size by fluorescent microscopy of the water-swollen cryo-RSFPs; images taken under GFP filter, 10x magnification. The images were contrasted and converted to binary offsets, counted and the mean particle Feret diameter frequencies estimated with ImageJ software (ellipticity preset to 0.0-1.0).

Figure 4.

Number average distribution profiles of RSFPs by Feret diameter. Each plotted value was the mean result of three microscopic readings and calculations. Each particulate sample was fabricated in duplicate.

The numerical data and size distribution curves (Fig. 4) show that an increase of injection pressure decreased the mean Feret diameter of particles, as well as narrowed the size distribution expressed as the span of recorded diameters within the cutoff range between the D10 and D90 cumulative values. For example, an increase of the injection pressure from 80 to 150 and then to 300 kPa resulted in decreased mean diameter of RSF (4%wt)/PPGDE(5 mmol/g)-derived RSFPs (Mixture #5, Supplementary Table 1) from ca. 700 to 600 and 400µm, respectively; at the same time, the D10/D90 diameter span narrowed from ca. 550 to 500 and 350µm, respectively. These trends were also observed for the RSF/EGDE (5 mmol/g, Formulation #1, Supplementary Table 1), RSF/acetone (10 & 20%vol., Formulations #9&10), and the RSF/PEG-400 (5&15%vol., Formulations #13&14) systems. However, concentrations of RSF above 5%wt had poor control over RSFPs size distribution. Reactive systems with 5 and 30mmol/g of EGDE (Formulations # 3&4) demonstrated a broadening of the D10/D90 particle size distributions by ca. 5–35% with increased injection pressure from 150 to 300kPa, respectively. A similar tendency was observed for the RSF/pPGDE (Formulation #7,) and the RSF/acetone (Formulations #11&12), despite a decrease in mean particle size. Typically, the shear viscosities of the 5.5%wt. RSF systems were 2–8 times lower than that of the 4%wt. systems (Fig.1), which for the cryo-granulation approach may reflect the preferred viscosity range from 0.5 to 4P to enable efficient control over the mean and minimum particle diameter, and the RSFPs size distribution. The most uniform sized distribution profiles were for the RSF (4%wt)/acetone (20%vol) Formulation #10; cryo-injections at either 150 or 300kPa produced consistent D10/D90 sizes as narrow as ca. 170–250 µm, and the mean particle sizes of ca. 500 µm. The smallest particles with Feret diameter of ca. 150–200 µm were fabricated from the RSF/EGDE (Formulations #2, 3, 4) or RSF/pPGDE (Formulation #5) systems, while RSF/PEG-400 systems (Formulation #14) yielded the largest particles, with diameters of ca. 900–1,000 µm. The efficiency of the current cryo-granulation approach was confirmed gravimetrically by sieving; up to 90%wt. of the starting aqueous silk could be converted into the RSFPs with mean particle sizes between 200–900µm. The additional sieve screening of the RSFPs fabricated from the RSF/EGDE or RSF/pPGDE mixtures estimated between 70 and 90%wt. of the obtained beads sized in range from 800 to 160µm, which corresponded to ca. 0.6–3.6 grams of RSFPs per minute of cryogranulation with a liquid feed consumption rate around 20–30mL/min depending on the injection pressure. Cryogranulation from the RSF/acetone mixtures resulted in higher yields, 80–95%wt. sized ca. 200–800µm. These diameters can potentially be applied for the production of large batches of polymeric meso-particles, followed by more precise size refinement (i.e. via sieve screening) [26–28].The cryogranulation setup described above allowed processing of nearly 20–30mL/min of the starting liquefied silk and produced between ca. 1 to 4.0 g/min of silk particles, surpassing microfluidic routes which process from 0.1 to 3.0mL/min of liquefied silk [58–60]. Thus, the cryogranulation method was an efficient alternative to microfluidic and emulsification technologies [12, 32, 42], with large monodisperse silk spheres.

The injection pressure and the composition of the starting RSF/cross-linker systems controlled the morphologies of the meso-RSFPs formed through the high-shear self-assembly. The non-solvent dispersion coupled with internal cryogelation generated the microporous inner morphology and wrinkled surface topology with the obtained meso-RSFPs. The textured surfaces covered the spherical RSFPs, which for most combinations of RSF/cross-linker revealed an interconnected microporous morphology of the cores (Fig. 5). The pore sizes ranged from 0.1 to 10µm. This sponge-like morphology is a characteristic feature of the polymeric cryogels cross-linked either covalently or physically at sub-zero temperatures [23]. The acetone- and PEG-400-treated RSFPs revealed similar microporous morphologies, with micropores in the 0.1–10µm size, similar to silk/chitosan blend microparticles [52] or the PLGA microspheres prepared by evaporation of dichloromethane [53]. These pores in the RSFPs were significantly smaller compared to the large pores of ca. 50–150µm generated in the macro-scale bulk cryogel samples, cross-linked through continuous −20°C freeze-thaw aging following the addition of 5–30 mmol/g of EGDE [33, 34] or sonication [35]. This observation may underline the direct dependence between the dimensions of the RSF-based cryo-construct and the sizes of the inner pores. The cryogenic treatment of reactive RSF-mixtures contained in macro-volumes produces larger pores, compared to the discrete micro- or meso-gels. Discrete RSFPs with collapsed structures were produced from the RSF-based systems containing 20%vol. acetone; the particles had a non-spherical “scrambled egg-shell” morphology with thin pore walls and rough edges and eroded surfaces (Fig. 5C).

Figure 5.

Compilation of the SEM images for the RSFPs fabricated at different combinations of parameters including cross-linker type/starting mass balance/injection pressure. Each micrograph shows an 80x magnification of the particles general view and contains a 300x inserts of each batch typical surface topology. The leftmost column of each section contains 1500-200x magnification of the RSFPs microporous inner morphology.

No direct correlation between injection pressure and the surface topology of the cryo-RSFPs was observed. However, an increase of starting RSF concentration from 4 to 5.5%wt. yielded particles with more textured surfaces, including cracks, dimples or fractures. In particular, 5.5%wt. RSF with 10%vol. acetone resulted in coalesced beads, which slowed the dispersion step, quenched at the intermediate stage by the −50°C-cooled coagulant. SEM micrographs did not reveal a decrease in RSFP size with increased RSF content, unlike that which was observed for co-flow assisted capillary protocols [1]. RSFPs stabilized with oligomeric PEG-400 and PPGDE displayed less static repulsion and thus more facile processability in the dry state, compared to the EGDE- and acetone-cured particles. RSFPs pre-encapsulated with DNR and DoxR displayed smooth surfaces with occasional micropores.

To evaluate the cytocompatibility of the RSFPs, we studied 21-day proliferation of RPTEC/TERT1 kidney and human normal fibroblast cell lines. Cell proliferation on the particles was characterized by confocal imaging. All diepoxide-stabilized constructs supported cell growth over the 3-week period (Fig.6) and confirmed the cytocompatibility of hyaluronic acid-based [46] and collagen-based [47] hydrogel scaffolds cross-linked with EGDE.

Figure 6.

Z-stacked confocal imaging of the morphology and localization of (i-iv) RPTEC/TERT1 and (v,vi) HNF cells within the macro-scale silk-based cryogels labeled with calcein AM green (green) and distributed throughout the mesopores of the silk scaffolds (red); the chemically cross-linked scaffolds were synthesized using (a, e) 30mmol/g EGDE or (b, g) 30mmol/g PPGDE, and the physically cross-linked cryogels were fabricated using either (c) 20%vol. acetone, or (d) 15%wt. PEG-400. All images were taken after 21 days of cell culture. Scale bars are 50µm.

The affinity of silk towards anthracyclines provided prolonged release of DoxR, with faster release achieved at low pH levels [21, 12]. We obtained DoxR-encapsulated RSFPs with varying starting concentrations of RSF (4 and 5.5%wt.) and DoxR (2.25 and 4.5mg/mL) to trace the impact of the drug/biopolymer ratio on pH-dependent pharmacokinetics. We assumed that the entire dosage of the drug was encapsulated within the resulting RSFPs as DoxR and the model antibiotics were poorly soluble in hexanes and had limited elution during the DEB cryo-fabrication step. This was in contrast to externally gelled silk microspheres obtained by exposure to ethanol [10]. The DoxR release from RSFPs was pH-dependent; 10–20% higher at pH 5.2, compared to the release profiles obtained at the pH 7.2. After 14 days, the total amount of DoxR released at pH 7.2 was 28–46%, depending on the starting formulation, whereas the cumulative release at pH 5.2 ranged from 38–62% (Fig. 7A). The first 3–4 days of DoxR elution were characterized by a burst release, reaching ca.45% at pH 5.2, and ca. 38% at pH 7.2. Following the burst-release, consistent elution of DoxR at ca. 1–4%/day was established, which corresponded to a daily release of 0.1–0.21 mg/mL (from the 100mg batch of particles). Interestingly, the lowest level of release at pH 7.2 (ca. 26%) was detected for RSFPs encapsulated with 4.5mg/mL of DoxR (Fig. 7A, curve 2), compared to the RSFPs loaded with half the amount (2.25mg/mL, Fig. 7A, curve 1) of drug. Solubility of DoxR increases with decreased pH [10–12] and the 4.5mg/mL-loaded particles demonstrated a nearly doubled cumulative release at pH 5.2. A reverse effect of the acid conditions was demonstrated for the release form RSFPs obtained from the concentrated (5.5%wt.) RSF systems; nearly 45% release of DoxR at pH 7.2 (Fig. 7A) was observed, and this decreased by ca. 10% (to 35%) at pH 5.2 (Fig. 7A). This interplay between concentration of dry RSF and eluting medium pH was observed for the silk DEBs modified with nano-RSF; providing the same amount of RSF in the starting mixture (4.5%wt.) RSFPs obtained with 6mg/mL n-RSF (Fig. 7A) had the cumulative release of nearly 42% at pH 7.2, comparable to the release form RSFPs containing 5.5%wt. RSF (Fig. 7A).This was10% higher compared to the release rates achieved by RSFPs fabricated using only 3mg/mL of n-RSF (Fig. 7A).

Figure 7.

Release properties of the acetone-cross-linked RSFPs pre-laden with the doxorubicin chloride and antibiotics were evaluated photometrically and via the disc-diffusion method, respectively. S. Aureus XEN29 disc-diffusion inhibition zone was assessed by bioluminescent imaging at 48 hours post application. Numerical captions above the samples relate to the concentration of drug (mg/mL) in the starting RSF/antibiotic mixtures.

Susceptibility tests were used to confirm post-processing antimicrobial activity of cryoencapsulated antibiotics, including tobramycin sulfate (TS), gentamicin sulfate (GS) and kanamycin sulfate (KS). The antibiotic function was preserved, evaluated by bacterial zones of inhibition (Fig. 7B) established on XEN29 cultures after 12 hours, as opposed to the negative controls of blank RSFPs (Fig. 7B,i). The diameter of the inhibition zones around the GS-loaded particles was 33.2±2.1 and around 38.4±2.6 mm in case of the 10 and 20mg/mL pre-encapsulation concentrations (Fig. 7B,ii). Similarly sized inhibition zones were characteristic for TS-loaded RSFPs, measured at nearly 37mm for both 10 and 20 mg/mL pre-loaded particles (Fig. 7B,iv). Lower release efficiency was shown for the RSFPs loaded with either 10 or 20 mg/mL KS- (Fig. 7B,iii), which generated inhibition zones, respectively of around 21.3±1.8 and 10.2±0.1mm after 48 hours. Overall, the results of the Kirby-Bauer assay demonstrated good retention of function of the antibiotics during the cry-granulation procedures.

Conclusions

Forceful injections of RSF/cross-linker/drug compositions into water immiscible hexanes were used for production of silk beads with diameters ranging from 300 to 1,000µm. Cryo-conditioning of the injection step by cooling the coagulation bath to ca −65°C arrested the shape and surface morphology of the droplets, originated from the spontaneous self-assembly after contact of emulsified RSF with the non-solvent phase. Additions of either diepoxide ethers, acetone, or polyethylene glycol played a dual function by emulsifying the reactive system enabling droplet formation and facilitating the process of either chemical (in the case of diepoxide ethers) or non-covalent internal cryogelation of silk. The freeze-thawed RSFPs had a microporous morphology resulting from ice templating which accompanied cryogelation, with pore sizes from ca. 0.1 to 10µm in diameter. Mean diameter and width of the RSFPs distribution could be controlled by adjusting the bollus injection pressure from 80–300 kPa and by varying the viscosities of the starting reactive colloids. Injection pressures higher than 200kPa favored the self-assembly of smaller particles of ca. 300–500 µm, while injections at lower than 100kPa yielded the silk beads as large as 1,000 µm. The RSFPs obtained from the RSF/acetone and RSF/PPGDE mixtures had spherical morphologies with solid outer shells and spongy cores, while particles obtained with additions of EGDE- and PEG-400 had rough surfaces and prolonged or angular shapes. Drug-loaded particles demonstrated retention of bioactivity.