Ionically Cross-Linked Polymer Networks for the Multiple-Month Release of Small Molecules.

Long-term (multiple-week or -month) release of small, water-soluble molecules from hydrogels remains a significant pharmaceutical challenge, which is typically overcome at the expense of more-complicated drug carrier designs. Such approaches are payload-specific and include covalent conjugation of drugs to base materials or incorporation of micro- and nanoparticles. As a simpler alternative, here we report a mild and simple method for achieving multiple-month release of small molecules from gel-like polymer networks. Densely cross-linked matrices were prepared through ionotropic gelation of poly(allylamine hydrochloride) (PAH) with either pyrophosphate (PPi) or tripolyphosphate (TPP), all of which are commonly available commercial molecules. The loading of model small molecules (Fast Green FCF and Rhodamine B dyes) within these polymer networks increases with the payload/network binding strength and with the PAH and payload concentrations used during encapsulation. Once loaded into the PAH/PPi and PAH/TPP ionic networks, only a few percent of the payload is released over multiple months. This extended release is achieved regardless of the payload/network binding strength and likely reflects the small hydrodynamic mesh size within the gel-like matrices. Furthermore, the PAH/TPP networks show promising in vitro cytocompatibility with model cells (human dermal fibroblasts), though slight cytotoxic effects were exhibited by the PAH/PPi networks. Taken together, the above findings suggest that PAH/PPi and (especially) PAH/TPP networks might be attractive materials for the multiple-month delivery of drugs and other active molecules (e.g., fragrances or disinfectants).

Introduction

Hydrogels are widely utilized in controlled release applications, where they can offer numerous advantages such as tunable solute permeability, degradability, stimulus-sensitivity, and having water-swollen structures that mimic living tissues.[1−4] When used in drug delivery, they can enable sustained drug release, which maintains the drug concentrations within the therapeutic window for a prolonged duration and, in turn, can provide increased efficacy.[5] When applied to the delivery of small molecules, however, hydrogels provide relatively weak barriers to diffusion. Thus, release of small, water-soluble drug molecules from hydrogels is typically too fast to enable their sustained delivery over many days, weeks, or months.[1] This stems from their highly hydrated hydrophilic structures and can significantly limit the scope of their applications.

To overcome this problem, several approaches have been explored, where long-term payload release over multiple days, weeks or months has typically been achieved at the expense of increased design complexity. Such approaches include: (1) chemical conjugation of the drug molecules to the hydrogel, so that their release is controlled by the rates of their slow (enzymatic or hydrolytic) cleavage;[6−9] (2) modification of the scaffold to reversibly bind the drug, thereby inhibiting its diffusion from the gel;[10,11] and (3) dispersion of colloidal structures (e.g., liposomes or polymeric micro- and nanoparticles) in the hydrogel network, which either serve as additional diffusion barriers or as substrates for drug binding.[12−17] Though long-term release of small molecules from hydrogels can be achieved using these approaches, each of them adds significant complexity to the controlled release device and (especially in the cases of drug conjugation or incorporation of drug-binding substrates) is often drug-specific. Similarly, while there are drug carriers that achieve long-term release without the use of gels (e.g., liposomes,[18,19] matrices prepared from water-insoluble polymers[20,21] or osmotic pumps[22]), they rarely combine all the attractive physicochemical properties of hydrogels. Accordingly, there is a need for simpler and more-versatile hydrogel systems for achieving long-term small molecule release.

To this end, we have recently shown that when poly(allylamine hydrochloride) (PAH) is mixed with strongly binding multivalent anions, pyrophosphate (PPi) and tripolyphosphate (TPP) in water (see structures in Scheme ), it self-assembles into gel-like, ionically cross-linked coacervates with high storage moduli (G∞′ ∼ 4 × 105 Pa).[23,24] Such high storage moduli are unusual for self-assembled hydrogels (whose G∞′-values seldom exceed 105 Pa[25−27]) and are indicative of very high cross-link densities. Indeed, for G∞′-values of this magnitude the scaling relationship between the plateau storage modulus, G∞′, and point cross-link density, c (i.e., G∞′ ∼ kBTc, where kBT is the Boltzmann factor[28]), suggests a hydrodynamic mesh size (ξH ∼ c1/3) on the order of a nanometer. On the basis of this small mesh size, we hypothesized that PAH/PPi and PAH/TPP complexes (which form from readily available and inexpensive ingredients) can serve as strong diffusion barriers and be effective vehicles for long-term controlled release. Because these ionic complexes utilize their small pore size, we also postulated that their ability to release small hydrophilic molecules over extended timescales might be independent of the chemical functionality of their payloads (which might make them potentially more versatile than their affinity or covalent conjugation-based counterparts). Furthermore, since PAH is already used as an orally administered drug (marketed as Renogel and Renvela by Sanofi)[29] and both PPi and TPP are on the U.S. Food and Drug Administration’s generally recognized as safe (GRAS) list, PAH/PPi and PAH/TPP complexes could be attractive for biomedical and pharmaceutical applications.

Scheme 1

Molecular Structures of (a) PAH and the Sodium Salts of (b) PPi and (c) TPP

Adapted from ref (23). Copyright 2014 American Chemical Society.

To test the above hypotheses, here we used PAH/PPi and PAH/TPP ionic networks to encapsulate and release two dissimilar small molecule payloads, anionic Fast Green FCF (FG) and zwitterionic Rhodamine B (RhB), both of which were dyes with molecular weights below 1000 Da (see Scheme ). The stability of PAH/PPi and PAH/TPP complexes in phosphate buffered saline (PBS) was confirmed using gravimetry and dynamic rheology. Then, to study the effects of payload/network binding on the small molecule uptake and release properties, the binding of the model payloads to the PAH/PPi and PAH/TPP networks was analyzed via isothermal titration calorimetry (ITC). The uptake and release properties were then probed by UV–vis spectroscopy. Finally, to inquire into the possibility of using these networks for drug delivery applications (e.g., for transdermal or buccal drug delivery), their cytocompatibility with human fibroblast cells was also explored.

Scheme 2

Molecular Structures of (a) FG and (b) RhB Dyes

Materials and Methods

Materials

Millipore Direct-Q 3 deionized water (18.2 MΩ·cm) was used in all experiments. PAH (roughly 98% pure; nominal molecular weight ≈ 120–200 kDa) was purchased from Alpha Aesar (Ward Hill, MA, USA). PPi (≥95% pure), TPP (≥98% pure), RhB (≥95% pure), and lysozyme (≥90% pure) were purchased from Sigma-Aldrich (St. Louis, MO, USA). Certified 6 M hydrochloric acid (HCl) and FG (98% pure) were acquired from Fisher Scientific (Waltham, MA, USA), while ≥97% pure sodium hydroxide (NaOH) was purchased from VWR (West Chester, PA, USA). PBS solutions were prepared from premade powders purchased from Fisher Scientific. All materials were used as received.

Preparation of Dye-Loaded Ionic Networks

The gel-like PAH/PPi and PAH/TPP coacervates were prepared at room temperature (RT) by adding either 0.93 mL of 7.5 wt % (305 mM) PPi or 0.80 mL of 7.5 wt % (220 mM) TPP to 0.75 mL of 10.0 wt % (1190 mM) PAH solution inside a 2.0 mL microcentrifuge tube (whereupon the mixtures were gently swirled for 3–5 s). To incorporate model payloads (FG or RhB dyes) into these ionic networks, the dyes were added to the PAH, PPi and TPP solutions at matching concentrations of 0.5, 1.0, or 4.0 mg/mL (which corresponds to either 0.6, 1.2, or 4.9 mM FG, or 1.0, 2.1, or 8.4 mM RhB). All solutions were titrated to pH 7.0 using HCl and NaOH prior to mixing. To collect the coacervates, the mixtures were then centrifuged at 15 000 rpm for 90 min. This yielded gel-like plugs at the bottoms of the microcentrifuge tubes, each roughly 7–8 mm in thickness (see Scheme ).

Scheme 3

Preparation of the Dye-Loaded Ionic Networks

Gravimetric Analysis

The stability of PAH/PPi and PAH/TPP coacervates was probed via gravimetry. Here, the initial mass of each PAH/PPi and PAH/TPP complex described in Section was measured after removing the supernatant from the microcentrifuge tube. The complexes were then shaken in an Eppendorf Thermomixer (Hamburg, Germany) at 400 rpm in 1 mL of 1× PBS (at 37 °C) with the PBS solution replaced daily, and weighed weekly to determine changes in their weights. The measured weights were then normalized to the initial values as W(t)/W0, where W(t) was the weight at time t and W0 was the weight prior to PBS immersion. Because the ionic complexes essentially behaved like gels (i.e., they did not flow upon agitation or tube inversion), they remained firmly fixed to the bottoms of the centrifuge tubes. Thus, all solvent replacement steps were performed by simply pipetting out the old PBS, rinsing the PAH/PPi and PAH/TPP complexes with 1 mL of fresh PBS for 3–5 s (to ensure that the released dye was fully removed) and, lastly, replacing the buffer used for the rinse with another 1 mL of fresh PBS.

Dynamic Rheology

In addition to probing their changes in weight, the stability of PAH/PPi and PAH/TPP complexes was confirmed by monitoring changes in their dynamic rheology. The samples for these measurements (which required larger coacervate volumes) were prepared by slowly adding either 26 mL of 3.9 wt % (153 mM) PPi solution or 14 mL of 5.7 wt % (164 mM) TPP solution at pH 7.0 to beakers containing 1000 mL of a 0.1 wt % (10.7 mM) PAH solution (also at pH 7.0), which was stirred at 300 rpm with a cylindrical magnetic stir bars (5 cm × 1 cm). The complexes were then equilibrated for 3 d, whereupon they were collected from the bottoms of the beakers. The PAH/PPi and PAH/TPP samples (0.5 g each) were then stored in 5.0 mL of 1× PBS for up to 14 d with the PBS being replaced daily. After either 3 or 14 d of storage the samples were loaded into a Rheometric Scientific RDA III (Piscataway, NJ, USA) strain-controlled rheometer equipped with 25-mm parallel plates and compressed to a 0.5 mm gap thickness. Prior to each measurement the samples were allowed to relax between the plates until the normal force was below 100 g and all excess sample was removed from the sides of the plates. After performing strain amplitude sweeps to determine the viscoelastic region, frequency sweeps were performed at a strain amplitude of 1.0% and angular velocities (ω-values) of 0.1–500 rad/s. Each frequency sweep was performed at RT and repeated three times to ensure reproducibility.

ITC Analysis of Payload/Network Binding

The binding of model payloads (FG and RhB) to the PAH/PPi or PAH/TPP complexes was probed by ITC using a MicroCal VP-ITC instrument (GE Healthcare; Northampton, MA, USA). To load the gel-like complexes into the sample cell (which is not designed for macroscopic solids), colloidal PAH/PPi and PAH/TPP complexes were formed by adding either 29.0 μL of 3.93 wt % (154 mM) PPi solution or 19.0 μL of 5.84 wt % (169 mM) TPP solution to 10.0 mL of 0.016 wt % (1.7 mM) PAH solution. The FG and RhB solutions were then prepared at concentrations of 0.43 wt % (5.3 mM) and 0.80 wt % (16.7 mM), respectively. To minimize buffer mismatch, all solutions and PAH/ionic cross-linker dispersions were titrated to pH 7.0 using HCl and NaOH. ITC experiments were then performed at 25 °C by placing the PAH/PPi or PAH/TPP dispersions in the sample cell, and loading the dye solutions into the injector. The dye solutions were titrated using twenty 15-μL injections, with 20 min equilibration intervals between each injection. The contents of the sample cell were stirred at 307 rpm with the impeller-shaped injector tip. The instrument software was used to integrate the raw ITC data, whereupon the heat of dilution (which was obtained by titrating the dye solutions into pH-matched DI water) was subtracted from the integrated data to obtain the final thermograms. To ensure reproducibility, each ITC measurement was repeated twice.

Payload Uptake and Release Analysis

The loading capacity (LC) and loading efficiency (LE) for each dye loading procedure (in Section ) was calculated by subtracting the amount of dye in the supernatant from the total amount of dye inside the microcentrifuge tubes. The total amount of dye in the supernatant was determined by measuring the supernatant mass in each tube and quantifying the supernatant dye concentration via UV–vis spectroscopy, which was achieved using a Varian Cary 50 spectrophotometer (Cary, NC, USA), at a wavelength of 614 nm (ε = 1.26 × 102 mL mg–1 cm–1) for FG and 555 nm (ε = 2.05 × 102 mL mg–1 cm–1) for RhB. The LC and LE were then calculated using the following equations:where Ci is the dye concentration in the parent PAH and PPi/TPP solutions, Cs is the final dye concentration in the supernatant, Vt is the total solution volume initially added to the microcentrifuge tube, Vs is the supernatant volume recovered after centrifugation, and W0 is the coacervate weight. In addition to varying the dye concentrations (as described in Section ), the LC- and LE-values of the dye-loaded complexes were measured at variable PAH concentrations and a constant initial dye concentration of 1 mg/mL (either 1.2 mM FG or 2.1 mM RhB). Here, the dye-loaded complexes were prepared as outlined earlier; however, the PAH concentration was varied between 2.5 and 10.0 wt % (or 274 and 1190 mM). The concentrations of the PPi and TPP solutions were also varied between 1.9 and 7.5 wt % (i.e., between either 72.8 and 305 mM PPi or 52.7 and 220 mM TPP) to keep the ionic cross-linker:PAH molar ratio constant (at 0.33 PPi:PAH and 0.20 TPP:PAH). The LC- and LE-values were then determined using Eqs and 2.

The release rates from PAH/PPi and PAH/TPP complexes were characterized using samples prepared at either 1.0 or 4.0 mg/mL dye concentrations (see Section ). Upon preparing each sample, the supernatant was removed from the microcentrifuge tube and, to ensure that all unencapsulated dye was removed, the inside of the tube was rinsed with DI water. The release medium, 1 mL of 1× PBS, was then added to the microcentrifuge tube. During the release studies, the microcentrifuge tubes were shaken at 400 rpm in an Eppendorf Thermomixer at 37 °C. The old PBS solution was replaced with fresh PBS daily (as described in Section ) and UV–vis spectroscopy was used to quantify the dye release within each one-day interval (using the same extinction coefficients as those used to determine the loading). The release from each sample was monitored until the amount of dye released each day became too low to yield a discernible peak in the UV–vis spectrum. Furthermore, to probe the release from these materials in nonmedical (e.g., household) applications, an additional release experiment was performed (using the same complex/dye compositions) with tap water at RT as the release medium. Each release experiment was repeated thrice to ensure reproducibility.

Cytocompatibility Analysis

The cytocompatibility of PAH/PPi and PAH/TPP complexes was probed in vitro using human dermal fibroblasts harvested from neonatal foreskins. The fibroblasts were expanded to a fifth passage in Dulbecco’s modified eagle medium (DMEM) containing 10% fetal bovine serum (FBS) and 100 μg/mL penicillin–streptomycin (all Life Technologies; Carlsbad, CA, USA). They were then seeded at a density of 10 000 cells/cm2 in 24-well plates. After 14 h of incubation at 37 °C and 5% CO2, PAH/PPi or PAH/TPP complexes (300 μL each) were added to the cells, such that 9 samples of each (including a coacervate-free control) were available. The plates were then incubated for 8 days and analyzed for cell morphology, metabolism, total protein, and total DNA content as described below.

Cell Morphology

Cell morphology was assessed using phase contrast microscopy 1, 4, and 8 d after the PAH/PPi or PAH/TPP complex addition. The potential cytotoxicity of PAH/TPP and PAH/PPi complexes was evaluated using ISO 10993–5:2009 guidelines where (at each time point) the cells were scored on a scale of 0–4 based on the qualitative microscopic observations/metrics summarized in Table .[30]

Table 1

Morphological Grading of Cell Culture Conditionsa

grade	reactivity	conditions of cell culture
0	none	discrete intracytoplasmic granules (which become visible at 40× magnification); no cell lysis; no inhibition of cell growth
1	slight	no more than 20% of cells are rounded, loosely attached and without intracytoplasmic granules, or show changes in morphology; lysed cells are occasionally present; growth inhibition is only slight
2	mild	not more than 50% of cells are round, devoid of intracytoplasmic granules; no extensive cell lysis; growth inhibition is below 50%
3	moderate	not more than 70% of cells are round or are lysed; cell layers not completely destroyed; but growth inhibition exceeds 50%
4	severe	nearly complete or complete destruction of cell layers

Adapted from ISO 10993-5:2009.

Pico Green Assay for Total Double-Strand DNA Content

The total DNA content within each well was determined after 8 d of culture in order to quantify differences in their relative cell numbers. The cells were washed with PBS three times and 1 mL of 25 mg/mL (1.7 mM) lysozyme buffer was added to each well to digest the cells. The plate was then incubated for 18 h with pipet mixing every 6–7 h, after which the digests were stored at −20 °C freezer until assaying. To perform the assay, 150 μL of each digest sample was added in triplicate to a 96-black-well plate, whereupon 150 μL of Quant-iT Pico Green double stranded DNA (dsDNA) reagent solution (Life Technologies) was added to each well and incubated at RT for 5–10 min. Fluorescence was then measured using an Infinite M200 spectrofluorometer (Tecan, Maennedrof, Switzerland) at an excitation wavelength of 480 nm and an emission wavelength of 520 nm, and compared against a standard curve (obtained using Quant-iT Pico Green dsDNA standards) to reveal the total dsDNA mass in each well.

Presto Blue Assay for Cell Metabolism

The PAH/PPi and PAH/TPP network effects on the cell metabolism were then investigated using the Presto Blue assay. Here, media was removed and a 10% Presto Blue solution (Life Technologies, Temecula, CA), prepared in fibroblast growth medium, was added to wells following the manufacturer’s protocol. The Presto Blue solution contains resazurin (7-hydroxy-3H-phenoxazin-3-one-10-oxide), which is metabolically reduced to resorufin (7-hydroxy-3H-phenoxazin-3-one) resulting in an absorbance shift that reflects the metabolic rate.[31] The cells and standards were then incubated with 10% Presto Blue solution for 60 min to allow substrate conversion. The supernatant from each well was transferred to a 96-black-well plate in triplicate and fluorescence was measured at 555 nm excitation and 590 nm emission wavelengths with an Infinite M200 spectrofluorometer. The fluorescence readings were then compared against a standard curve, constructed using 0, 50, 100, and 1000 μM dithiothreitol (DTT) reducing solutions instead of cells, to obtain an equivalent DTT concentration (mM).

Statistics

All statistical analyses in the cytocompatibility experiments were performed using JMP 11 Pro (SAS Institute, Cary, NC, USA). ANOVA with Tukey’s post hoc analysis was performed to detect significant differences between groups. An α-level of 0.05 was used to determine significance between groups. Asterisks indicate where a significant difference exists between groups. Data are reported as mean ± standard deviation.

Results and Discussion

Stability of PAH/PPi and PAH/TPP Complexes

Changes in the wet weights of the gel-like PAH/PPi and PAH/TPP complexes in 1× PBS at 37 °C were tracked over multiple months (see Figure ). These weights increased to roughly 110% of their initial values after the first week (likely due to minor changes in swelling). After this initial increase, however, the normalized weights of PAH/PPi and PAH/TPP coacervates stayed roughly constant (at around 100–110% of their initial weights), exhibiting almost no change over 200 d. These stable weights suggested that PAH/PPi and PAH/TPP complexes degrade very slowly under physiological conditions. Similar trends occurred when FG and RhB were loaded into the network (Figures S1 and S2), though the initial swelling varied with the type and quantity of payload used (with the W(t)/W0-values after 1 week ranging between 100% and 135%).

Figure 1

Evolution in the normalized (blue squares) PAH/PPi and (red circles) PAH/TPP complex weight (equal to the percentage of their original weight) as a function of storage time in PBS (n = 3). All data are mean ± SD and the lines are guides to the eye.

The stability of these complexes was probed further by dynamic rheology, which was characterized after 3 and 14 d of storage in PBS. Despite the variability in storage times, the rheology of the samples remained unchanged. The frequency sweeps (see Figure ) revealed similar rheological properties to those measured in 150 mM NaCl,[24] where the ionic strength was comparable to that in PBS. The G∞′-values of the PAH/PPi and PAH/TPP complexes were near 4 × 105 Pa with G′ and G″ approaching one another at lower ω-values, thus suggesting that these gel-like coacervates would flow under a sustained stress (see Figure ). The high G∞′-values showed that these gel-like networks retained their high cross-link densities in PBS, while the invariance in their rheology over a 2-week period confirmed their stability.

Figure 2

G′ (closed symbols) and G″ (open symbols) for (a) PAH/PPi and (b) PAH/TPP complexes analyzed after 2 weeks of storage in PBS.

Binding of Model Payloads to PAH/PPi and PAH/TPP Networks

Because PAH is cationic and PPi and TPP are anionic, charged solutes may bind to the ionically cross-linked PAH networks. Though both model payloads used in this work are ampholytic (see Scheme ), FG has a net negative charge while RhB (at least at near-neutral pH[32]) has a net charge of zero. Given this dissimilar charge, we hypothesized that the two model payloads differed in their affinity for the PAH/PPi and PAH/TPP networks. These differences in payload/network binding were probed by ITC, where FG and RhB solutions were titrated into PAH/PPi and PAH/TPP dispersions. The resulting thermograms are shown in Figure , and indicate the binding enthalpy per mole of injected dye as a function of the dye:PAH molar ratio inside the sample cell. These thermograms roughly correspond to the first derivatives of the binding isotherms plotted as functions of the dye:PAH monomer molar ratios.[33,34] Accordingly, the plots obtained from the FG titrations (Figures a and b) are indicative of strong cooperative binding, with the exothermic binding signals (and consequently the binding strength) increasing with the dye:PAH molar ratio up to ratios of roughly 0.1 to 0.2:1. Conversely, the thermograms from the RhB titrations (Figures c and d) exhibited much weaker exothermic signals, which gradually decreased as the dye:PAH molar ratio increased, and indicated RhB binding to be very weak and noncooperative.[34,35] Thus, the choice of dye molecules in this work enabled the uptake and release of both binding and essentially nonbinding payloads to be characterized.

Figure 3

ITC data for (a, b) FG and (c, d) RhB titrated into (a, c) PAH/PPi and (b, d) PAH/TPP dispersions.

Payload Uptake

The formation of PAH/PPi and PAH/TPP networks in the presence of dissolved dye (see Section ) caused the dye molecules to become entrapped in the PAH/PPi and PAH/TPP networks. The LC- and LE-values, however, depended on the dye type and initial dye concentrations in the parent PAH, PPi and TPP solutions (see Table ). The LCs and LEs of FG were much higher than those of RhB. This reflected the strong FG binding to the PAH/PPi and PAH/TPP networks which, like in previous studies on ionic payload uptake by polyelectrolytes,[36,37] caused this dye to be entrapped much more efficiently than the virtually nonbinding RhB. Moreover, the LE-values for FG increased with the dye concentration, while the RhB LE-values remained roughly constant. This increase in the LEs for FG at higher dye concentrations was consistent with the cooperative binding between FG and the coacervates revealed by ITC (see Figures a and b). Conversely, the nearly constant LE-values obtained for RhB (where the LCs scaled almost linearly with the initial dye concentration; Table ), suggested that this weakly interacting payload gets encapsulated through simple physical entrapment, with little or no binding to the gel-like networks.

Table 2

LE and LC-Values Achieved at Different Initial Dye Concentrations (n = 3)a

	PAH/PPi		PAH/TPP
	LE (%)	LC (mg/g)	LE (%)	LC (mg/g)
FG (0.5 mg/mL)	56.48 ± 3.04	3.16 ± 0.17	64.01 ± 1.45	3.31 ± 0.07
FG (1.0 mg/mL)	63.03 ± 1.37	7.04 ± 0.15	72.94 ± 3.09	7.54 ± 0.32
FG (4.0 mg/mL)	68.54 ± 2.09	30.62 ± 0.93	74.74 ± 3.47	30.89 ± 1.43
RhB (0.5 mg/mL)	9.83 ± 2.79	0.55 ± 0.16	8.71 ± 2.10	0.45 ± 0.11
RhB (1.0 mg/mL)	6.38 ± 0.79	0.71 ± 0.09	6.27 ± 1.04	0.65 ± 0.11
RhB (4.0 mg/mL)	7.76 ± 2.77	3.46 ± 1.24	6.62 ± 1.41	2.74 ± 0.58

All data are mean ± SD.

To further analyze payload uptake, the LEs and LCs for each dye type were also measured at a constant initial dye concentration (1 mg/mL) while varying the PAH concentration used during encapsulation (Table ). This was done to vary the fraction of the sample volume where the physical entrapment of the payload molecules occurred. The LEs of the RhB-loaded complexes increased linearly with PAH concentration, while their LCs remained roughly constant (see Figure a). Thus, consistent with the physical entrapment hypothesis, as less PAH was used and less coacervate formed, fewer of the RhB molecules could be enmeshed by the ionic networks. The mass of dye entrapped per unit of network volume, however (which was reflected by its LC), remained nearly constant (Figure b).

Table 3

LE and LC-Values Achieved Using Different PAH Concentrations (n = 3)a

	PAH/PPi		PAH/TPP
	LE (%)	LC (mg/g)	LE (%)	LC (mg/g)
FG (2.5% PAH)	N/A	N/A	N/A	N/A
FG (5.0% PAH)	61.69 ± 1.13	13.78 ± 0.25	72.40 ± 0.53	14.96 ± 0.11
FG (10% PAH)	63.03 ± 1.37	7.04 ± 0.15	72.94 ± 3.09	7.54 ± 0.32
RhB (2.5% PAH)	1.49 ± 0.30	0.67 ± 0.13	1.59 ± 0.65	0.66 ± 0.27
RhB (5.0% PAH)	3.76 ± 0.45	0.84 ± 0.10	3.78 ± 0.32	0.78 ± 0.07
RhB (10% PAH)	6.38 ± 0.79	0.71 ± 0.09	6.27 ± 1.04	0.65 ± 0.11

All data are mean ± SD.

Figure 4

Effect of parent PAH solution concentration on the (a) LE and (b) LC-values achieved using (blue squares) PAH/PPi and (red circles) PAH/TPP complexes and a weakly binding model payload (RhB) (n = 3). All data are mean ± SD and the lines are guides to the eye.

In contrast, the LEs of the FG-loaded complexes remained fairly constant with the PAH concentration (Table ). This might be because the decreased PAH concentrations (at a constant FG concentration) increased the dye:PAH molar ratios which, due to the cooperativity of the FG/network binding, increased the fractional coverage of the PAH binding sites. Additionally, the extent of electrostatic FG/network binding may have been amplified by a reduction in the ionic strength during dye uptake (i.e., at lower PAH, PPi, and TPP concentrations fewer monovalent counterions were released upon their complexation). The LE therefore remained roughly constant, despite the reduction in PAH concentration. Furthermore, when 2.5 wt % PAH was used during the encapsulation procedure, the FG-loaded complexes did not form compact gel-like plugs at the bottoms of the microcentrifuge tubes. Instead, the PAH/ionic cross-linker/dye complexes became much more tacky and mechanically weak, and remained coated on the centrifuge tube walls even after centrifugation. This was likely because the FG:PAH ratio became so high that the FG bound to the PAH began displacing the PPi and TTP anions (which drastically changed the properties of the ionic complexes). Interestingly, while these weakened, tacky complexes formed when 2.5 wt % PAH and 1 mg/mL FG were used during encapsulation, robust gel-like coacervates were obtained when 10 wt % PAH and 4 mg/mL FG were used (despite the two formulations having the same FG:PAH ratios). This effect remains poorly understood and suggests that higher PAH and cross-linker concentrations might mitigate the adverse effects of payload binding on network integrity.

Overall, Tables and 3 show that the LE and LC-values for both binding and nonbinding small molecule payloads can be controlled by varying the payload and PAH/cross-linker concentrations used during the loading process. Consistent with other ionic networks,[36] they also indicate that encapsulation is more efficient for payloads that bind to the PAH/PPi and PAH/TPP complexes. The payload/network binding, however (as seen for the 2.5 wt % PAH samples), can affect the physicochemical properties of the ionic networks, leading to their degradation in the limit of high LC-values. Thus, the LC-values of PAH-binding payloads may sometimes be limited by their effects on the network properties rather than by the efficiency of their entrapment within the matrix. Another limitation of the loading method used in this work is the very low LE (below 10%) achieved with the nonbinding RhB. These LE-values, however, can likely be significantly improved by (1) dissolving all of the payload in the parent PAH solution (rather than in both the PAH and PPi/TPP solutions as done here), i.e., so that all of the dye would be initially enmeshed in PAH chains; and (2) using more concentrated parent PAH and PPi/TPP solutions, which would enlarge the ultimate volume of the PAH/PPi and PAH/TPP networks and limit dye release during their formation.

Payload Release Kinetics

The PAH/PPi and PAH/TPP complexes loaded with FG released only a very small fraction (<5%) of their payload over timescales ranging between 42 and 164 d (see Figures a and b), whereupon the release became too slow to obtain discernible UV–vis peaks at daily intervals. The complexes prepared at a FG concentration of 4.0 mg/mL exhibited a slower fractional release than those prepared using 1.0 mg/mL dye (Figures a and b). This might have stemmed from the cooperative binding of FG to the ionic network (see Section ). When the initial FG concentration increased from 1.0 mg/mL to 4.0 mg/mL, its binding to the network became stronger and (since binding lowers diffusion rates[38]) may have prolonged its release. Due to their higher LC-values (Table ), however, complexes prepared using 4.0 mg/mL FG released a higher total mass of dye than those prepared using 1.0 mg/mL FG (see Figures c and d). This increase in the total release rate suggests that the payload delivery rates can be easily tuned by varying their concentrations during the encapsulation procedure. Furthermore, the greater total release rate allowed the dye release from the samples prepared using 4.0 mg/mL FG to be tracked longer than those prepared using 1.0 mg/mL FG. Because less dye was released from the complexes prepared at a FG concentration of 1.0 mg/mL, discernible UV–vis peaks were not obtained beyond 42 d for the PAH/PPi complexes and 80 d for the PAH/TPP complexes (as opposed to 164 d for both complex types when 4.0 mg/mL FG was used).

Figure 5

Release profiles obtained from (a, c) PAH/PPi and (b, d) PAH/TPP complexes loaded using (blue squares) 1 mg/mL and (red circles) 4 mg/mL parent FG solutions plotted as (a, b) percent of dye released and (c, d) total mass of releases dye (n = 3). All data are mean ± SD and the lines are guides to the eye.

The slow payload release was also achieved when the complexes were loaded with RhB (see Figure ). Again, although further release still occurred, the times over which the experiments were performed were limited to the early portions of the release profiles, where discernible UV–vis peaks could be obtained (and where only 1–3% of the payload was released; Figures a and b). Because RhB had much lower LEs than FG (see Table ), there was significantly less RhB loaded into the ionic networks compared to the FG. Consequently, a lower total mass of RhB was released (see Figures c and d). This made it difficult to obtain discernible UV–vis peaks beyond 14–40 d (14 d when 1.0 mg/mL of RhB was used and 40 d when 4.0 mg/mol RhB was used). Furthermore, unlike the complexes loaded with FG, those loaded with RhB released approximately the same percentage of their payload over time regardless of the initial dye concentration (see Figures a and b). This was likely because the binding of RhB to the network was very weak. Without significant binding, the fractional release rate was simply determined by the mesh size (which controlled the diffusion coefficient[39−41]) and thickness of the gel-like network,[42] and was virtually independent of payload content. Thus, the coacervates loaded with RhB dye released approximately the same percentage of dye over time regardless of the initial dye concentration, and the mass of dye release was roughly proportional to LC.

Figure 6

Release profiles obtained from (a, c) PAH/PPi and (b, d) PAH/TPP complexes loaded using (blue squares) 1 and (red circles) 4 mg/mL parent RhB solutions plotted as (a, b) percent of dye released and (c, d) total mass of releases dye (n = 3). All data are mean ± SD and the lines are guides to the eye.

While the slow release of FG molecules may in part be attributed to their binding to the ionic networks, the slow release of nonbinding RhB indicates the low solute permeabilities of PAH/PPi and PAH/TPP networks as the reason for the extended release profiles. The low permeabilities of these networks likely reflect their small hydrodynamic mesh sizes and are qualitatively consistent with previous studies, where the diffusion coefficients and release rates diminished when the network pore size (and water content) were reduced.[39−41] Further, the slow RhB release shown in Figure suggests that PAH/PPi and PAH/TPP complexes can be used for the multiple-month delivery of small molecules, regardless of whether the molecules bind to the ionic networks. Because the slow release does not require payload/network affinity, the PAH/PPi and PAH/TPP complexes can likely release small, water-soluble molecules over many months regardless of the chemical functionality (or ionization states) of their payloads. Moreover, this long-term release is not only limited to a PBS medium. Release experiments were also performed using tap water at RT as the release medium, and yielded slow release on similar timescales (Figures S3 and S4).

Furthermore, consistent with the small percentage of the dye released even after multiple months (see Figures and 6), the color of the PAH/PPi and PAH/TPP complexes (blue in the case of FG and pink in the case of RhB) faded only near the surface of the gel-like plugs. This revealed that, over the timescales of these experiments, diffusion-mediated release of dye molecules was only occurring from near the surface of the ionic networks. Accordingly, to achieve faster and more-complete release, release experiments were also performed on 1–2 mm (rather than 7–8 mm) thick PAH/PPi and PAH/TPP ionic networks, and a much higher percent of the loaded dye was released (up to 15% within 14 d; see Figure S5). Thus, in addition to varying the LC within these ionic networks, their release performance can also not-surprisingly be tuned by varying their thickness.

Interestingly, though for diffusion-controlled release the mass of payload released was expected to scale with the square root of time (i.e., as M(t) ∝ t1/2),[42] some of the release profiles (especially those obtained using FG) exhibited sharp reductions in release rates after 10–20 d of release time (e.g., see Figures a and c, S3, and S5a). These reductions in release rates were much more abrupt that those predicted by the above scaling relationship (which assumes unchanging diffusion coefficients)[42] and were likely caused by a shrinkage in the coacervate pore size that occurred over time. Such structural rearrangements are supported by changes in the visual appearance of the samples, which gradually became less opaque (indicating the loss of larger pores[43]), and may have led to a reduction in payload mobility. Consistent with this interpretation, the sudden reductions in release rates were greater for the larger FG dye (which was easier than RhB to entrap) and for the least swollen dye-loaded networks, that is, PAH/PPi networks loaded with 1 mg/mL FG (Figure S1). Moreover, this view is consistent with experiments by Weinbreck et al., which revealed the diffusivity of polyelectrolyte molecules in protein/polyelectrolyte coacervates to undergo a similar reduction with time (which also coincided with decreased coacervate opacity).[44] Another possible interpretation of the sudden changes in FG release rates is that the binding sites available for FG might have been saturated, such that the nonbinding FG was released much faster than the remaining, PAH-bound FG. The FG:PAH monomer molar ratios used in the release experiments, however (which were between 1.5 × 10–3 and 6.7 × 10–3:1), were very low relative to those where binding site saturation occurred (FG:PAH monomer ratios near 0.6:1; see Figure a and b), which makes the changes in pore size a much more likely explanation for the sudden changes in release rates.

Another factor that had the potential to affect dye release was its accumulation in the release media, which could have slowed down the release process.[45] Because of the daily solvent replacement, however, this accumulation was (especially after the first few days) very low. Indeed, though the average FG concentrations in the release media after day 1 (at the highest FG loading) reached around 2.5 × 10–3 wt % (0.030 mM), the supernatant FG concentrations accumulated after day 7 did not exceed 2.6 × 10–4 wt % (3.2 × 10–3 mM). Combined with the ability of the ionic networks to continue releasing dye even when the release media was not replaced after the first day (data not shown), this suggests that–at least for the majority of the release process–the dye in the release media had little impact on the highly extended release profiles. These observations again confirm that the multiple-month release from PAH/PPi and PAH/TPP networks is primarily enabled by their low solute permeabilities, and not payload/network binding or dye accumulation in the release media.

Cytocompatibility Analysis

For their safe use in drug delivery applications, the PAH/PPi and PAH/TPP networks must be nontoxic. As a first step to evaluating safety, their cytocompatibility with human dermal fibroblasts (such as would be encountered in a dermal drug delivery device) was explored. Without the PAH/PPi and PAH/TPP complexes (i.e., the untreated controls), the cells readily proliferated and formed uniform layers with expected morphology. This confirmed that there was no cytotoxic reactivity (grade 0 reactivity) throughout the 8-day control experiment. The PAH/TPP complexes, however, caused a “slight” (grade 1) inhibition of fibroblast growth on days 1 and 4, where the presence of some rounded cells indicated minor cellular dysregulation, and no inhibition (grade 0) on day 8 (center column in Figure ). This mild inhibition at early time points might reflect the leaching of unassociated TPP from the PAH/TPP network, which might affect the cell culture. While the inhibitory effects were mild in the case of PAH/TPP complexes, they were stronger when PAH/PPi networks were used, showing a “moderate” (grade 3) cytotoxic reactivity on day 1 (with ≤ 70% of cells being round or lysed and a ≤ 50% growth inhibition) and a “slight” (grade 1) cytotoxic reactivity on days 4 and 8 (right column in Figure ).

Figure 7

Phase contrast images of human dermal fibroblasts cultured in the presence of PAH/TPP (left column) and PAH/PPi (center column) ionic networks and ionic network-free controls (right column) imaged 1, 4, and 8 d after ionic network addition. The green circles show examples of round cells while the red dashed lines highlight the inhibition zones around the PAH/PPi networks.

Microscopy confirmed growth inhibition (highlighted by the dashed lines in Figure ) around the PAH/PPi complexes. This stronger cytotoxicity of the PAH/PPi complexes might have reflected the higher free PPi concentration relative to that of TPP. While the pentavalent TPP was mixed with PAH at a 0.20:1 TPP:PAH monomer molar ratio (which yielded near-stoichiometric complexation), the tetravalent PPi was mixed with PAH at a 0.33:1 PPi:PAH monomer molar ratio. This higher ratio was used to accelerate the assembly of PAH/PPi complexes into macroscopic gel-like networks[23] and significantly exceeded the stoichiometric PPi:PAH monomer molar ratio of 0.25:1. The use of excess PPi in these mixtures likely gave rise to significant PPi leaching into the surrounding media and, consequently, may have led to the primary cytotoxic response. Indeed, despite being produced intracellularly through adenosine triphosphate (ATP) hydrolysis[46] and being on the U.S. Food and Drug Administration’s GRAS list, addition of PPi to cell cultures has been reported to slightly inhibit cell proliferation.[47] Furthermore, since the formation of PAH/PPi and PAH/TPP complexes can shift the initial pH and monovalent ion concentrations within the ionic networks, it is possible that the mild inhibitory effect may have stemmed from slight perturbations in the pH and osmolarity in the adjacent media.

Despite these inhibitive effects, after 8 d of culture the total cell numbers (inferred from the total DNA content revealed by the Pico Green assay) appeared to be unaffected by the presence of ionic networks, and showed no statistically significant difference between the wells containing PAH/PPi and PAH/TPP networks and the network-free controls (Figure a). To further analyze the effect of PAH/TPP and PAH/PPi networks on cell function, cell metabolism in their presence was measured on day 8 using the Presto Blue assay. The metabolism of the PAH/TPP group was similar to that of the control group (p > 0.05 by ANOVA) and the metabolism of the PAH/PPi group was only roughly 20% lower (p < 0.05 by ANOVA; Figure b). This suggests that, despite the morphological differences revealed in Figure , the total metabolic activity of cells was unaffected by the PAH/TPP complexes and was only slightly diminished by the PAH/PPi complexes.

Figure 8

Quantitative analysis of PAH/PPi and PAH/TPP network effects on (a) total DNA and (b) Presto Blue cell metabolism assay results after 8 day of culture (n = 3). The asterisk (*) denotes significance by ANOVA (p < 0.05). All data are mean ± SD.

In summary, despite the apparent cytotoxicity of PAH/PPi networks seen in Figure , the cytotoxic response to the near-stoichiometric PAH/TPP complexes was minimal. Similarly, the weak impact of PAH/PPi and PAH/TPP networks on the fibroblast cell numbers and metabolism (in Figures a and b), suggests that ionically cross-linked PAH networks can be essentially nontoxic. This opposes the cytotoxicity of PAH reported in previous studies[48,49] and likely reflects the cytotoxic PAH amine groups being almost-fully complexed with either PPi or TPP (which limits cell exposure to these cytotoxic moieties). Indeed, similar improvements in PAH cytocompatibility have been reported upon its complexation with dextran sulfate, evidently due to an analogous reduction in PAH amine group availability.[49] Because of the very strong binding and high stability of PAH/TPP complexes (Figure ), their ionic networks exhibit promising cytocompatibility. Furthermore, though the PAH/PPi and networks explored herein appear to be somewhat cytotoxic, it may be possible to ultimately improve their cytocompatibility by optimizing the PPi:PAH stoichiometry (i.e., to minimize PPi leaching) or adjusting their pH and osmolarity.

Technological Implications

The release of small molecules from PAH/PPi and PAH/TPP networks in Figures and 6 occurs over timescales that are comparable (or longer than) those typically achieved from other/more-complicated hydrogels for long-term drug delivery (i.e., drug depots where covalent drug conjugation, or incorporation of drug-entrapping or drug-binding colloids is required[14−17]). Moreover, it reveals that: (1) release over multiple-month timescales can be achieved regardless of whether the small molecule payload binds to the PAH/cross-linker network; and (2) that the rate at which a payload is delivered from these matrices can be easily tuned by varying its initial loading. Combined with their mild and facile preparation, these observations suggest that PAH/PPi and PAH/TPP networks could provide a simple and versatile platform for the long-term release of water-soluble small molecules. Furthermore, the established use of PAH, PPi and TPP in pharmaceutical[29,50,51] and food applications,[52−54] and the low in vitro toxicity of the PAH/TPP ionic networks (Figures and 8) could make these materials attractive candidates for long-term drug delivery. Though establishing their safety in their ultimate applications will require longer-term in vivo toxicity testing (over the multiple-month timescales intended for their use), these materials might provide simple and effective alternatives for delivering water-soluble small molecule drugs over multiple weeks or months.

In addition to their potential in long-term controlled release, we have recently shown PAH/PPi and PAH/TPP complexes to adhere to a variety of dissimilar substrates (ranging from glass, to Teflon, to human skin) while under water.[23,24] Thus, these multifunctional materials can simultaneously serve as both controlled release systems and wet adhesives, which could be beneficial for dermal, buccal or even ocular drug delivery. Similarly, these adhesive ionic networks might provide a sustained release platform for applications outside of drug delivery, such as release of fragrances or disinfectants in household applications (e.g., in kitchens or bathrooms), as demonstrated by their long-term release of FG and RhB into tap water (Figures S3 and S4).

More broadly, the properties of PAH/PPi and PAH/TPP complexes revealed herein highlight the excellent potential of coacervates in the long-term release of water-soluble small molecules. Coacervates from mixtures of oppositely charged polymers, such as gelatin and gum arabic, have been used for many years for encapsulation of hydrophobic molecules (e.g., essential oils)[55−58] and for the long-term release of proteins.[59−61] Over the past decade, colloidal coacervate complexes formed via complexation of PAH with other, weaker-binding multivalent anions (such as citrate and phosphate) have also been used for the uptake and transport of PAH-binding payloads, including small, water-soluble molecules.[36,62] These PAH-based coacervates, however, had fluid-like properties[23] and flowed when agitated in the presence of PBS (data not shown). Accordingly, they are unlikely to be suitable as macroscopic drug depots and, in the absence of strong drug/network interaction (when used as colloidal capsules), release most of their model small molecule payloads within minutes.[63] The strongly associated PAH/PPi and PAH/TPP networks, however, have much longer relaxation times and therefore maintain their shape upon agitation. Combined with their simple preparation, stability and low solute permeability, this mechanical stability could make PAH/PPi and PAH/TPP networks attractive for the long-term release of small molecules in drug delivery and other (e.g., household, environmental and industrial) applications.

As an outlook for future work, it is helpful to also discuss some current limitations of PAH/PPi and PAH/TPP networks (along with some possible approaches for overcoming them). The first is that the release rates achieved in this initial study (where just a few percent of the dye were released over months; see Figures and 6) were so slow that they would only be suitable for small, water-soluble drugs (or other active ingredients) that are effective at very low concentrations. These drugs might include analgesics, such as lidocaine and bupivacaine (which are effective even at μg/mL concentrations[64,65] and could be used to treat chronic pain), or potent anti-inflammatory drugs such as ibuprofen (e.g., to improve host response to biosensors, implants or indwelling catheters).[66] To make their release profiles suitable to a broader range of applications (including those requiring higher drug doses or less-extended release times), however, the PAH/PPi and PAH/TPP complexes should be tailored to: (1) enable all (rather than 1–5%) of their payloads to be released over the several-month timescales used in this work; and (2) make the release rates more uniform–i.e., so that the sharp reduction in the release rates seen at the lower FG loading in Figure would no longer occur. Both of these effects can likely be achieved by developing strategies for adjusting the ionic network mesh size. Furthermore, higher drug doses can likely be achieved by improving the LE and LC-values (especially those of nonbinding payloads such as RhB) using the approaches described in the end of Section .

Finally, PAH is not a degradable polymer and the highly stable PAH/PPi and PAH/TPP complexes (see Figure ) are unlikely to be cleared from their application site unless they are mechanically or manually removed. Though this might limit the use of these materials as implants, this limitation can likely be overcome by introducing hydrolytically or enzymatically cleavable linkages into the polyamine structure. Indeed, similar results have already been obtained for nanocarriers prepared by cross-linking poly(l-lysine) (PLL) with citrate ions, where the PLL chains were digested with trypsin to release the capsule-bound payload.[36] Tailoring these ionically cross-linked network structures to optimize their payload uptake, release properties and degradability will be subjects of our future work.

Conclusions

Ionically cross-linked gel-like coacervates prepared through the self-assembly of PAH with either PPi or TPP can encapsulate small molecules and release them over multiple-month timescales. This slow release apparently stems from their very high ionic cross-link densities, which make the PAH/PPi and PAH/TPP networks strong barriers to diffusion. Release over multiple months can be achieved regardless of whether the payload molecules bind to the ionic networks, and regardless of whether PBS at 37 °C or tap water at RT is used as the release medium. The rates at which a payload is delivered over these extended timescales can be tuned by varying its initial loading (i.e., by varying the compositions used in the encapsulation process). Furthermore, PAH/PPi and PAH/TPP networks are highly stable in PBS and (at least in the case of PAH/TPP complexes tested in vitro) appear to be nontoxic to human dermal fibroblast cells. Their long-term stability and controlled release properties, and apparent cytocompatibility, make these ionically cross-linked PAH networks intriguing as potential materials for long-term (multiple-month) release in biomedical, household and industrial applications.