Molecularly imprinted polymers via living radical polymerization: Relating increased structural homogeneity to improved template binding parameters

Molecularly imprinted polymers via living radical polymerization: Relating increased structural homogeneity to improved template binding parameters

Reactive & Functional Polymers 78 (2014) 38–46 Contents lists available at ScienceDirect Reactive & Functional Polymers journal homepage: www.elsevi...

646KB Sizes 1 Downloads 59 Views

Reactive & Functional Polymers 78 (2014) 38–46

Contents lists available at ScienceDirect

Reactive & Functional Polymers journal homepage: www.elsevier.com/locate/react

Molecularly imprinted polymers via living radical polymerization: Relating increased structural homogeneity to improved template binding parameters Vishal D. Salian, Charles J. White, Mark E. Byrne ⇑ Biomimetic & Biohybrid Materials, Biomedical Devices, and Drug Delivery Laboratories, Department of Chemical Engineering, Auburn University, Auburn, AL 36849, USA

a r t i c l e

i n f o

Article history: Received 14 June 2013 Received in revised form 1 February 2014 Accepted 13 February 2014 Available online 22 February 2014 Keywords: Molecular imprinting Living radical polymerization Controlled polymerization

a b s t r a c t This work examined imprinted polymer networks prepared via controlled/living radical polymerization (LRP) and conventional radical polymerization (CRP) on chain growth, network formation, and efficiency of producing molecularly imprinted, macromolecular memory sites for the template molecule, diclofenac sodium. LRP extended the reaction-controlled regime of the polymerization reaction and formed more homogeneous polymer chains and networks with smaller mesh sizes. In addition, LRP negated the effect of the template on polymer chain growth resulting in polymers with a more consistent PDI independent of template concentration in the pre-polymerization solution. Improved network homogeneity within imprinted poly(HEMA-co-DEAEM-co-PEG200DMA) networks prepared via LRP resulted in a 38% increase in template binding affinity and 43% increase in the template binding over imprinted networks prepared via CRP and a 97% increase in affinity and 130% increase in capacity over non-imprinted networks prepared by LRP. By varying certain parameters, it was possible to create imprinted networks with even higher template binding affinities (155% over non-imprinted) and capacities (261% over non-imprinted). This work is the first to conclusively demonstrate that the observed improvement in binding parameters in weakly crosslinked, imprinted polymer networks could be explained by the more uniform molecular weight evolution associated with the LRP mechanism and the longer lifetime of an active polymer chain relative to the total polymerization time, which allowed for the formation of a more homogenous imprinted polymer network. Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction Molecularly imprinted polymers are engineered to recognize and preferentially bind a specific template molecule due to the creation of high affinity memory sites within the polymer network. In addition to the selective binding of template molecules, increased stability and cost-effectiveness make molecularly imprinted polymers (MIP) highly attractive potential alternatives for many analytical applications, such as sensors, solid-phase extraction membranes or filters, chromatography, coatings, and binding assays [1–3]. In the last decade, the introduction of molecularly imprinted polymers into the pharmaceutical field as carriers for the controlled release of drugs, peptides, and proteins has resulted in increased interest [4–9]. Despite the many attractive attributes and potential high market value of molecularly imprinted polymers and hydrogels in many fields, such as ophthalmology and ⇑ Corresponding author. Tel.: +1 334 844 2862; fax: +1 334 844 2063. E-mail address: [email protected] (M.E. Byrne). http://dx.doi.org/10.1016/j.reactfunctpolym.2014.02.003 1381-5148/Ó 2014 Elsevier Ltd. All rights reserved.

drug delivery [10–20], wide use of these materials clinically or industrially has not yet been realized. This is partly because of deficiencies in their binding properties. MIPs, in general, have low average binding affinities and a high degree of binding site heterogeneity. Attempts to improve the binding site heterogeneity have been reported [21] but have resulted in decreased template binding affinity and/or capacity, which is especially detrimental to drug delivery applications. For drug delivery from imprinted materials, affinity and capacity are crucial design parameters. Rational methods to improve the template binding characteristics have been hindered by a lack of understanding of the synergy between the creation of the macromolecular memory sites and the mechanics of the polymerization reaction. This is especially true of conventional free radical polymerization (FRP), the most popular polymerization method due to its versatility and economic viability. Unfortunately, FRP gives rise to heterogeneity within the resulting polymer network, which further exacerbates the binding site heterogeneity and loss of template binding capacity and/or affinity observed in molecularly imprinted polymers. Novel

V.D. Salian et al. / Reactive & Functional Polymers 78 (2014) 38–46

techniques, like controlled or living radical polymerization (LRP), offer hope in terms of achieving improved template binding, decreased binding site heterogeneity, and greater understanding of the imprinting process. LRP results in the synthesis of improved, more efficient polymer networks with more homogeneous network structures [22,23] and can lead to better binding parameters in imprinted polymers [24–30]. In addition, LRP can lead to more control over network structures and a better understanding of their structure–property relationships; however, these reactions are relatively new introductions to the imprinting field and have not been used extensively in the synthesis of molecularly imprinted polymers. The first reports of LRP enhancing template binding properties with highly cross-linked imprinted polymer networks were published in 2006 [25,26,30]. Boonpangrak et al. [30] used covalent imprinting due to the use of a high temperature polymerization reaction, while Vaughan and Byrne [25,26] used UV polymerization, which allowed the use of more favorable, non-covalent imprinting techniques. In addition, Vaughan et al. [24] were the first to demonstrate that weakly cross-linked imprinted networks prepared using LRP techniques resulted in increased template binding capacity and affinity. Pan et al. [29] used reversible addition– fragmentation chain transfer (RAFT) polymerization to create MIP microspheres using precipitation polymerization, and they reported higher binding capacity per unit surface area in the RAFT polymerized microspheres over the microspheres prepared via FRP. The loss of template binding affinity and/or capacity observed in imprinted networks prepared by FRP is due to mismatch between the rapid chain growth during polymerization and slow chain relaxation of the polymer chain, which results in structural heterogeneity in polymer networks generated by FRP [22,23]. As the polymerization reaction proceeds (increased conversion/decreased monomer concentration), the bimolecular termination reaction between two propagating polymer chains is hampered due to diffusional limitations. Heterogeneity in these networks is largely attributed to spatial variations and localized polymerization in the vicinity of chain initiation. By contrast, the LRP rate of polymerization proceeds much more slowly, allowing increased relaxation times for the propagating chains, making the formation of homogeneous polymer networks more thermodynamically favorable. LRP replaces the bimolecular termination between polymer chains with a reversible termination reaction controlled by a macroradical iniferter, which prevents or delays auto-acceleration of monomer conversion. Combining molecular imprinting with LRP synthesis techniques has been shown to result in enhanced template binding affinity and capacity [25–30] in addition to slower template transport [24]. However, there has not been an effort to understand the effect of LRP on (a) polymer network structure in imprinted polymer networks and (b) the template binding cavities in the imprinted polymer network. A greater understanding of these relationships could lead to the rational design of imprinted polymer networks and gels with the ability to better control their template binding and template transport properties. The work described in this paper examines the variation between polymer network structures prepared through FRP and LRP techniques, and their respective binding parameters. In addition, the relationship between different parameters, the template–monomer interactions of the pre-polymerization solution, and the template binding properties of the formed polymer networks are described. The parameters varied were the template concentration, the functional monomer concentration, and the solvent content in the pre-polymerization solution. During polymerization, reaction analysis was used to determine the double bond conversion of the imprinted polymer. In addition, analysis of the reaction signature in the absence of light was used to calcu-

39

late the kinetic coefficients for the reaction. Further analysis of the network structure was carried out by analyzing the growing polymer chains of the imprinted network. Lastly, equilibrium binding studies were carried out to evaluate the template binding affinity and capacity of the networks. By exploring weakly crosslinked systems, the effect of the template concentration on polymer chain growth, in terms of dispersity and molecular weight, can be described. In addition, the effect of LRP on the formation of template binding cavities in imprinted networks while varying different parameters was analyzed. This work may allow much better control of the properties of molecularly imprinted polymers, which will result in their broader use in novel technologies. 2. Materials and methods Monomers, inhibitor removal packing sieves, initiator, chain transfer agent, and template were purchased from Sigma– Aldrich (Milwaukee, WI). Poly(ethyleneglycol(200))dimethacrylate (PEG200DMA) was used as received, while (diethylaminoethyl)methacrylate (DEAEM) and (hydroxyethyl)methacrylate (HEMA) had inhibitors removed via inhibitor removal packing sieves prior to polymerization. The initiator [azo-bis(isobutyronitrile) (AIBN)], template molecule (diclofenac sodium (DS)), and chain transfer agent (tetraethylthiuram disulfide (TED)) were used as received. De-ionized (DI) water was used as the template rebinding solvent, the wash solvent (to remove template and unreacted monomer), as well as the mobile phase in the high pressure liquid chromatography (HPLC) system. 2.1. Synthesis of poly(HEMA-co-DEAEM-co-PEG200DMA) networks Poly(HEMA-co-DEAEM-co-PEG200DMA) networks prepared via conventional free radical polymerization (FRP) were imprinted for DS with a 5% crosslinking percentage and made with 0.336 mL of DEAEM (1.673 mmol), 3.659 mL of HEMA (30.118 mmol), 0.538 mL of PEG200DMA (1.673 mmol), 20 mg of AIBN (0.121 mmol), and 150 mg of DS (0.472 mmol). Table 1 describes all prepared polymer formulations in greater detail. The solutions were mixed and sonicated until all solids were dissolved. Non-imprinted polymers were prepared similarly except the template was absent. The poly(HEMA-co-DEAEM-co-PEG200DMA) imprinted networks prepared via controlled or living radical polymerization (LRP) were synthesized with 4.20 mg of TED (0.014 mmol) and 40 mg of AIBN (0.242 mmol). Solutions were transferred to an inert (nitrogen) atmosphere for radical UV photopolymerization [MBraun Labmaster 130 1500/1000 Glovebox (Stratham, NH) (Temp = 25 °C, Pressure = 3.6 kPa)]. Monomer solutions were pipetted between two 600  600 glass plates coated with trichloromethylsilane (to prevent adhesion of the polymer matrix to the glass) and separated by 0.25 mm Teflon spacers. The solutions were left uncapped and open to the nitrogen atmosphere until the O2 levels inside reached negligible levels (<1 ppm) as determined by the attached solid state O2 analyzer. The polymerization reaction was carried out for 8 min for the poly(HEMA-co-DEAEMco-PEG200DMA) networks prepared via FRP, whereas the reaction time was 24 min for the polymers prepared via LRP. The intensity of light from a UV Flood Curing System (Torrington, CT) was 40 mW/cm2 at 325 V. Double bond conversion analysis was also performed on both the pre-polymer solutions and the polymerized films using a Nicolet 6700 Fourier Transform Infrared Spectrometer. A small amount of the pre-polymer solution was pipetted into a quartz well plate until the depth of the solution matched the thickness of the corresponding cured polymer films. FTIR analysis was performed between 1600–1700 cm1 on each of the pre-polymer solutions in triplicate. The carbon–carbon double bond was

40

V.D. Salian et al. / Reactive & Functional Polymers 78 (2014) 38–46

Table 1 Poly(HEMA-co-DEAEM-co-PEG200DMA) polymer network formulations. Polymer

DS (mg)

DEAEM (mL)

HEMA (mL)

PEG200DMA (mL)

Non-imprinted (FRP0, LRP0) (T/FM = 0, FM = 5%) Imprinted (FRP1, LRP1) (T/FM = 0.1, FM = 5%) Imprinted (FRP3, LRP3) (T/FM = 0.3, FM = 5%) Imprinted (FRP5, LRP5) (T/FM = 0.5, FM = 5%) Imprinted (FRP3-10FM, LRP3–10FM) (T/FM = 0.3, FM = 10%) Imprinted (FRP3-S, LRP3-S) (T/FM = 0.3, FM = 5%, 50% Solvent)

0

0.336

3.659

0.538

53.22

0.336

3.659

0.538

150

0.336

3.659

0.538

266.12

0.336

3.659

0.538

300

0.672

3.456

0.538

150

0.336

3.659

0.538

Notes: For polymers prepared via FRP, 20 mg of AIBN was added to the polymer formulation while 40 mg of AIBN, and 4.2 mg of TED was added for polymers prepared via LRP. For polymers prepared in the presence of solvent, 4.703 mL of water was added to the polymer formulation. Polymer sample names were assigned by using the type of reaction and the T/FM ratio. Any other deviation from the base formulation was indicated after a hyphen. For example, FRP3-S indicates the polymer was synthesized via FRP procedures, formulated to a T/FM ratio of 0.3, and that solvent was present in the pre-polymer solution.

identified, and the area under the curve calculated. The analysis was repeated in triplicate for the polymeric films. Fractional double bond conversion was determined by taking the ratio of the film to solution area under the curve for each sample formulation and subtracting the ratio from one. The fractional conversions were then converted to percentage and reported. It should be noted that the creation of polymer networks for analysis of kinetic parameters and kinetic chain length distributions was carried out via UV polymerization in a differential photocalorimeter (DPC). Separate DPC studies producing films of equivalent thicknesses revealed exact reaction times and kinetic parameters. The DPC employed an EXFOS high pressure 100 W mercury vapor short arc lamp, and the wavelengths of the light sources were matched utilizing band pass filters (320–500 nm) with intensities being matched by radiometry (International Light IL1400A radiometer). The temperature and ambient atmosphere were also matched between experiments. Optically clear, thin films were produced which limited variations or attenuation within the films, which could lead to non-uniformity within the film. After polymerization, the glass plates were soaked in DI water, and the polymers were quickly peeled off the plates and cut into circular disks using a size 10 cork borer (13.5 mm). The gels were washed in a well-mixed 2 L container of DI water for 7 days with a constant 5 mL per minute flowrate of DI water. Absence of detectable template released from the polymer was verified by removing random gels, placing them in fresh DI water with adequate mixing, and sampling the supernatant via spectroscopic monitoring. The disks were allowed to dry under laboratory conditions (temperature = 20 °C) for 24 h and then transferred to a vacuum oven (91.4 kPa, 33–34 °C) for 24 h until the disk weight change was less than 0.1wt%. 2.2. Analysis of kinetic chain length of polymers The molecular weight distributions of uncrosslinked polymer chains were characterized by a modified HPLC system (Shimadzu, Columbia, MD) used for size exclusion chromatography (SEC). The SEC setup consisted of two PL Aquagel size exclusion columns in series (Varian LLC, Santa Clara, CA) for separation of various molecular weight fractions of the polymer which were detected using a RID10A refractive index detector (Shimadzu, Columbia, MD). DI water (temperature = 20 °C) was used as the mobile phase for the system.

Prior to running the samples, the system was calibrated using narrow molecular weight distribution poly(methacrylic acid) [poly(MAA)] standards (Polymer Source Inc., Dorval, Quebec). The molecular weights used were 2 ± 1.25, 10 ± 1.1, 16 ± 1.05, 124 ± 1.25, 500 ± 1.08, and 1100 ± 1.1 KDa. The packing of both SEC columns were polystyrene-based and were recommended by both the column and the HPLC manufacturers as compatible with the mobile phase, the synthesized copolymer samples, and the polymeric calibration standards used in this work. Flow rates, injection volume, and injection concentration were within manufacturer recommendations. The polymer chains obtained after early termination of polymerization were dissolved in the mobile phase to achieve a concentration of 5 mg/mL. A 50 lL aliquot of the solution was then injected into the system using a Rheodyne (Oak Harbour, WA) 7725(i) manual injection unit. Using the subsequent peaks obtained on the chromatograph, the weight average molecular weight (Mw), the number average molecular weight (Mn), and the polydispersity index (PDI) were calculated. The molecular weight (Mi) for a particular weight fraction was described by the x-coordinate of the corresponding point on the chromatograph while the number of chains (Ni) was described by the y-coordinate. The SEC was calibrated according to manufacturer recommendations and in accordance with universal calibration. Universal calibration of the column has been well documented to be independent of the actual geometry for a vast majority of polymers, including linear, coiled, and branched (e.g. comb and star shaped). The Mark–Houwink equation was used to convert the hydrodynamic volume of synthesized copolymer samples to molecular weight, and the Mark–Houwink equation parameters (K = 8.9  105, a = 0.75, at the temperature of our column, 20 °C) were obtained from the literature [31,32]. The copolymers synthesized in this work predominately consisted of the single monomer (HEMA) (>90%) so branching/crosslinking and, as a result, terpolymer formation was minimized. The majority of radical chain transfer to polymer or solvent and the formation of branches and/or crosslinks typically occurs at high conversion or very low monomer concentration as the propagation step is more thermodynamically favorable than the chain transfer mechanism. The reactions were stopped before monomer conversion reached 20% to prevent excessive branching or crosslinking. Therefore, the extent of crosslinking is negligible at these low

V.D. Salian et al. / Reactive & Functional Polymers 78 (2014) 38–46

monomer conversions and does not significantly affect the hydrodynamic radius of the polymer in solution. In addition, low composition of comonomer species in the polymer have minimal influence on the Mark–Houwink parameters, making the Mark– Houwink equation an appropriate conversion for our system.

41

were vortexed for 10 s, and the equilibrium concentrations were measured. A mass balance was used to determine the bound amount of template within the polymer. All polymers were analyzed in triplicate, and all binding values are based upon the dry weight of the gel. The Langmuir–Freundlich isotherm was used to determine binding parameters.

2.3. Analysis of kinetic parameters 3. Results and discussion A dark reaction was used to determine the kinetic reaction profile for the poly(DEAEM-co-HEMA-co-PEG200DMA) networks [33,34]. For each polymerization reaction, the UV light was shut off at a specific time point during the reaction. The rate of polymerization was calculated via reaction analysis using the heat flow vs. time from the DPC, average molecular weight of the polymerization solution, and the theoretical heat of reaction for methacrylate monomers (13.1 kcal mole1 [34,35]). Fractional double bond conversion was determined by dividing the experimental heat of reaction by the theoretical heat of reaction. The experimental heat of reaction was determined by the area under the heat flow versus time curve from the DPC. The effective termination and propagation coefficients, kt and kp, were calculated from Eqs. (1) and (2), and the derivation of these equations can be found in [36–38].

kp 0:5 kt

¼

Rp ½MðfIo e½IÞ

0:5

ð1Þ

where [M] is the monomer concentration, f is the initiator efficiency, Io is the light intensity, e is the extinction coefficient, and [I] is the initiator concentration. The initiator efficiency (f) varies between no initiation (0) and ideality (1), but usually ranges between 0.3 and 0.8. The ideal value is assumed in our calculations. The actual value is dependent on diffusional effects and monomer affinity, specific to experimental formulation and temperature of the solution, and is difficult to measure exactly [36–38]. The calculated effective propagation coefficient will therefore scale proportionately according to 1/f0.5. The unsteady state equation used to decouple the propagation coefficient, and the termination coefficient is shown below. 0:5

kt

¼

 0:5  kp =kt ½Mt¼t1 ½Mt¼t0  2ðt 1  t 0 Þ Rpt¼t1 Rpt¼t0

ð2Þ

where t1 and t0 are the final and initial times, [M]t=t1 and [M]t=t0 are the corresponding monomer concentrations, respectively, and Rpt=t1 and Rpt=t0 are the rates of polymerization at final and initial times, 0:5 respectively. In order to solve for kp, Eq. (1) is first solved for kp =kt at t = t1. The calculated ratio found in Eq. (1) is used in Eq. (2) to 0:5 solve for kt . Once the polymerization termination coefficient is determined, the propagation coefficient can be determined as shown in Eq. (3).

kp ¼

kp 0:5

kt

Fig. 1A shows equilibrium binding isotherms for weakly crosslinked poly(HEMA-co-DEAEM-co-PEG200DMA) imprinted and non-imprinted networks prepared using conventional free radical polymerization (FRP) and controlled or living radical polymerization (LRP). Imprinted polymer networks prepared via FRP exhibited higher template binding compared to the corresponding non-imprinted polymer networks. Imprinted polymer networks prepared using LRP exhibited still higher template binding while their corresponding non-imprinted networks matched the template binding demonstrated by non-imprinted polymer networks prepared via FRP. Calculation of binding parameters (Table 2) revealed an increase in both the number of binding sites (Qmax) and binding affinity (Ka) in imprinted polymers prepared via FRP (Ka = 15.7 ± 0.12 mM1, Qmax = 16.7 ± 0.64 mg/g) over the corresponding non-imprinted polymer networks prepared via FRP (Ka = 10.1 ± 0.20 mM1, Qmax = 9.6 ± 0.38 mg/g) indicating the successful creation of molecular memory in the imprinted networks. The imprinted networks showed 74% higher binding capacities and 55% higher binding affinities. Similarly, imprinted polymers networks prepared via LRP (Ka = 21.7 ± 0.17 mM1, Qmax = 23.9 ± 0.60 mg/g) demonstrated 130% increase in the number of binding sites (Qmax) and 97% increase in binding affinity over the corresponding non-imprinted polymer networks prepared via LRP (Ka = 11.0 ± 0.34 mM1, Qmax = 10.4 ± 0.48 mg/g). Non-imprinted networks prepared via LRP did not demonstrate a statistically significant enhancement of binding parameters over the nonimprinted networks prepared via FRP indicating LRP enhanced the effect of molecular imprinting on the template binding parameters of the networks. Propagation in a radical initiated polymerization reaction has two stages: the reaction-controlled stage and the diffusion-controlled stage. During the reaction-controlled stage, the monomer addition to the growing macroradical is controlled almost exclusively by the reactivity of the radical. As the reaction proceeds, the concentration of free monomer around the macroradical is de-

! 0:5

 ðkt Þ

ð3Þ

2.4. Template binding experiments and analysis of binding parameters A stock solution of 1 mg/mL of DS was prepared and diluted to five concentrations (0.05, 0.10, 0.15, 0.20, and 0.25 mg/mL). Initial absorbance of each concentration were measured using a Synergy UV–VIS spectrophotometer (BioTek Instruments, Winooski, VT) at 276 nm, the wavelength of maximum absorption. After the initial absorbance was taken, a dry, washed poly(HEMA-co-DEAEM-coPEG200DMA) polymer disk was inserted in each vial, and the vials were gently mixed until equilibrium. Separate dynamic studies were performed to assure equilibrium conditions were reached. After equilibrium was reached over a 7-day period, the solutions

Fig. 1. Equilibrium mass loading of template molecules of poly(HEMA-co-DEAEMco-PEG200DMA) networks. Equilibrium binding isotherms for films of poly(HEMAco-DEAEM-co-PEG200DMA) imprinted polymers prepared via LRP (N), prepared via FRP (4) and non-imprinted polymer networks prepared via LRP (d) and FRP (s) are shown in Fig. 1A. Error bars represent the standard error (n = 3). LRP leads to significantly higher loading and affinity as compared to FRP.

42

V.D. Salian et al. / Reactive & Functional Polymers 78 (2014) 38–46

Table 2 Binding capacities, affinities and coefficients of diffusion for diclofenac sodium in poly(HEMA-co-DEAEM-co-PEG200DMA) polymer networks. Polymer

Ka (mM1)

Qmax (mg/g)

Characteristics

FRP0 LRP0 FRP1 LRP1 FRP3 LRP3 FRP5 LRP5 FRP3–10FM LRP3–10FM FRP3-S LRP3-S

10.1 ± 0.20 11.0 ± 0.34 12.8 ± 0.21 13.2 ± 0.41 15.7 ± 0.12 21.7 ± 0.17 15.4 ± 0.18 19.2 ± 0.25 21.6 ± 0.38 25.7 ± 0.40 13.3 ± 0.29 13.6 ± 0.30

9.6 ± 0.38 10.4 ± 0.48 11.8 ± 0.54 14.2 ± 0.62 16.7 ± 0.64 23.9 ± 0.60 17.1 ± 0.41 25.4 ± 0.54 31.3 ± 0.52 34.7 ± 0.57 20.7 ± 0.33 22.3 ± 0.49

Non-living, non-imprinted, T/FM = 0, FM = 5% Living, non-imprinted, T/FM = 0, FM = 5% Non-living, imprinted, T/FM = 0.1, FM = 5% Living, imprinted, T/FM = 0.1, FM = 5% Non-living, imprinted, T/FM = 0.3, FM = 5% Living, imprinted, T/FM = 0.3, FM = 5% Non-living, imprinted, T/FM = 0.5, FM = 5% Living, imprinted, T/FM = 0.5, FM = 5% Non-living, imprinted, T/FM = 0.3, FM = 10% Living, imprinted, T/FM = 0.3, FM = 10% Non-living, imprinted, T/FM = 0.3, FM = 5%, 50% Solvent Living, imprinted, T/FM = 0.3, FM = 5%, 50% Solvent

Note: Polymer sample names were assigned by using the type of reaction and the T/FM ratio. Any other deviation from the base formulation was indicated after a hyphen. For example, FRP3-S indicates the polymer was sythesized according to FRP procedures, formulated to a T/FM ratio of 0.3, and solvent was present in the pre-polymer solution.

pleted and, as a result, the reaction is controlled by the diffusion of the free monomer and the macroradical. As a result of the unavailability of monomers, many radicals undergo termination. The diffusion-controlled stage is now reached where the termination of the reaction begins to gain importance over propagation. More homogenous polymer networks result when the propagation phase is extended. Fig. 2 and Table 3 show the observed propagation coefficient versus the double bond conversion for the polymerization reaction to produce the weakly crosslinked networks. In the reaction-controlled stage, the propagation reaction dominates, and as a result, the apparent propagation coefficient stays constant. When the diffusion-controlled stage begins, the contribution of the propagation reaction decreases while that of the termination reaction increases. As a result, when the apparent propagation coefficient begins to drop, one can assume a transfer from the reaction-controlled to the diffusion-controlled stage. LRP involves the introduction of chain transfer molecules which combine with the macroradicals to form dormant species. This process allows for the formation of a more thermodynamically favorable network since the decrease in the rate of propagation allows the polymer radicals more time to reorganize to minimize the Gibb’s free energy of the system. With LRP, the macro-iniferter can decay back into a dithiocarbamyl radical and a propagating chain during polymerization. Therefore, LRP adds a reversible termination step and

Fig. 2. Effective propagation coefficient from kinetic analysis of weakly crosslinked poly(HEMA-co-DEAEM-co-PEG200DMA) network formation. The propagation coefficient for the poly(HEMA-co-DEAEM-co-PEG200DMA) imprinted network (d) and non-imprinted network (s) prepared via FRP show similar trends where the propagation coefficient decreases after initially remaining constant. kp,eff of imprinted gels drops at earlier times than the non-imprinted gels. The imprinted (N) and non-imprinted (4) networks prepared via LRP show a more constant rate of propagation indicating a longer reaction-controlled propagation mechanism.

Table 3 Monomer conversion (Xi) and polymer effective propagation coefficient (kp,eff) values for imprinted and non-imprinted conventional free radical polymerization and controlled living radical polymerization. Imprinted

Non-imprinted

Xi

kp,eff

0.04 0.14 0.23 0.36 0.5 0.64 0.06 0.16 0.22 0.38 0.48 0.54 0.6

Xi

kp,eff

273 ± 23.0 376 ± 42.5 230 ± 13.0 90 ± 2.0 66 ± 10.5 10 ± 1.5

0.06 0.18 0.285 0.42 0.51 0.63

297 ± 30 334 ± 35 273 ± 30 193 ± 20 121 ± 15 15 ± 5

10.6 ± 5 13.8 ± 5 12.5 ± 5 15.2 ± 5 11.3 ± 5 11.3 ± 5 9.6 ± 5

0.04 0.1 0.24 0.38 0.44 0.5 0.56

8.6 ± 5 12.5 ± 5 11.4 ± 5 13.9 ± 5 11.3 ± 5 10.0 ± 5 10.0 ± 5

(L/(mol s))

(L/(mol s))

FRP

LRP

diffusing species can be different from conventional FRP. Using LRP leads to a delayed transition from the reaction-controlled phase to the diffusion-controlled phase. In addition, we hypothesized that the extended propagation phase in LRP would result in polymer networks with fewer imperfections in the polymer network, such as unreacted pendant double bonds and/or primary and secondary cycles, allowing a more homogenous structure. The double bond conversion determined by FTIR for networks produced by FRP and LRP (FRP0,1,3,5 and LRP 0,1,3,5) was statistically similar at 90 ± 6% and 92 ± 5%, respectively. As a result, even though the final double bond conversion was similar for both reactions, polymer networks prepared using LRP are hypothesized to have a more homogenous network because a larger proportion of polymer was formed in the reaction-controlled phase. It is also of great interest to note the large difference in the order of magnitudes for the effective propagation coefficients for LRP and FRP. The calculated effective propagation coefficients for LRP are at least an order of magnitude smaller than those calculated for FRP, indicating a much slower reaction rate but greater polymer relaxation times, etc. and a more homogenous polymer network. Below 50% monomer conversion, the difference in order of magnitude between the two approaches two orders of magnitude. The calculated effective propagation coefficients had standard deviation (<10%) and demonstrated repeatability. All methods of experimental determination of polymerization effective propagation coefficients are subject to some deviation due to differences in experimental variability even in well-behaved homopolymerizations and can

V.D. Salian et al. / Reactive & Functional Polymers 78 (2014) 38–46

be misleading in the presence of high occurrences of branching and/or crosslinking and/or chain transfer. However, an approximation of the effective propagation coefficients for lightly crosslinked networks can be found with reasonable confidence and accuracy, particularly for use to compare between two systems as shown here. The application of LRP techniques greatly slowed the polymerization rate, further extending the time available for the macroradicals to reach minimal free energy states within the polymer network and allowing the formation of more homogeneous networks. The discussion above illustrates the effect of LRP on the kinetics of the polymerization reaction. This analysis can be useful in understanding the effect of LRP on polymer chain growth and network formation. Analyzing the polymer chains composing the polymer network would be even more useful in terms of explaining the role of LRP in enhancing molecular imprinting in polymer networks. However, the insolubility of crosslinked polymer networks makes it extremely difficult to analyze their structure with SEC methods directly. After the gelling point, the network cannot just be dissolved and analyzed via SEC without destroying the network or without loss of information available to uncrosslinked polymers. Thus, the polymer chains formed in the early stages of polymerization (before gel formation begins) were analyzed using SEC. The polymer chains formed in the initial stages of polymerization are the building blocks of the polymer networks. Any variations in the polymer chain size as well as molecular weight distribution of the chains would be incorporated into the polymer network. Thus, we hypothesized that by analyzing the polymer chains formed before gelation occurred, we could approximate the structure and architecture of the polymer network of the gels. Fig. 3A shows the average molecular weight of the polymer chains as a function of the template/functional monomer (T/FM) ratio. The T/FM ratio is the ratio of molar concentrations of the template, DS, and the functional monomer, DEAEM. It is used as

Fig. 3. Effect of template concentration on kinetic chain length and polydispersity index of poly(HEMA-co-DEAEM-co-PEG200DMA) chains. Weight average molecular weight (A) and polydispersity index (B) versus template/functional monomer (T/ FM) ratio for pre-gelation poly(HEMA-co-DEAEM-co-PEG200DMA) chains prepared via FRP (h) and LRP (j) are shown here. Error bars represent the standard error with n = 3. Presence of template in the reaction mixture leads to the formation of polymers with shorter chains and higher dispersity due to an early transition to the diffusion-controlled regime of propagation. LRP counterbalances it by extending the reaction-controlled regime.

43

a measure of template concentration in the pre-polymerization solution. The T/FM ratio in the pre-polymer mixture was varied by increasing concentration of template in the pre-polymerization mixture while keeping the functional monomer concentration constant. As the template concentration was increased (increasing T/ FM ratio) for networks prepared via FRP, the average molecular weight of the corresponding polymer chain decreased. The decrease in average molecular weight of the polymer chains occurred despite the conversion remaining the same. Therefore, the decrease in average molecular weight cannot be attributed to less monomer being incorporated into the growing chains. Rather at higher template concentrations, there is a tendency of a larger number of comparatively shorter polymer chains to be formed. This may be a result of the diffusional limitations introduced by the presence of the template molecule. Thus, we hypothesize that the presence of template causes the polymerization reaction to shift from a reaction-controlled stage to a diffusion-controlled stage earlier than expected. This is confirmed by Fig. 2 where the effective propagation coefficient for the imprinted polymer gel begins to drop earlier than the effective propagation coefficient for the non-imprinted gel. LRP has been demonstrated to extend the propagation period of a polymerization reaction [39,40]. The active chain transfer radical combines with the polymer radical to undergo a reversible termination reaction, forming a meta-stable species. This results in delayed auto-acceleration and compensates for any diffusional limitation to propagation introduced by the presence of the template molecule. Thus, the addition of a chain transfer agent to the pre-polymerization mixture counterbalances the retardation of kinetic chain length of the resultant polymer chains caused by the diffusional limitations presented by an increasing concentration of monomer–template complexes. This is demonstrated in Fig. 3A where the average molecular weight of polymer chains created via LRP does not decrease despite an increase in the template concentration. In addition, the longer propagation period combined with the decreased rate of propagation allows the polymer chains to grow uniformly and simultaneously resulting in more monodisperse polymers prepared via LRP as compared to polymers prepared via FRP. In addition, polymers prepared using LRP show smaller average molecular weights consistent with decreased mesh sizes [24]. Fig. 3B shows the polydispersity index of the polymer chains as a function of the T/FM ratio. An increase in the T/FM ratio in the pre-polymerization mixture led to higher dispersity in the resultant polymer chains created by FRP. The increased dispersity can be attributed to an early transition to the diffusion-controlled stage of propagation. LRP, however, extends the reactioncontrolled stage of propagation negating the effect of the template on the polymerization reaction, leading to a more consistent dispersity independent of template concentration in the pre-polymerization solution. The increased uniformity in the polymer chains manifests itself as increased homogeneity in the polymer network. An increase in the homogeneity in the polymer network implies better binding sites as well as greater availability of binding sites within the polymer network for the template which could explain the observed improvement in the binding affinity and capacity in imprinted polymers due to LRP. In Fig. 3, we demonstrate the effect on chain growth in polymer gels of varying T/FM ratio while keeping the monomer concentration constant. Conversely, in Fig. 4, the T/FM ratio was maintained constant while the functional monomer concentration was varied to determine the effect on polymer chain growth. Fig. 4 compares the average molecular weight and PDI of imprinted polymer gels with varying concentration of functional monomer prepared via FRP and LRP. From Fig. 4, it can be observed that an increase in the concentration of functional monomer results in the creation of polymer chains with lower average molecular weight as well

44

V.D. Salian et al. / Reactive & Functional Polymers 78 (2014) 38–46

Fig. 4. Effect of Functional Monomer Concentration on Kinetic Chain Length and Polydispersity Index of poly(HEMA-co-DEAEM-co-PEG200DMA) Chains. Weight average molecular weight (A) and polydispersity index (B) versus functional monomer concentration for pre-gelation poly(HEMA-co-DEAEM-co-PEG200DMA) chains prepared via FRP (h) and LRP (j) are shown here. Error bars represent the standard error with n = 3.

as higher PDI. The use of LRP decreased kinetic chain lengths as well as PDI which is consistent with results described above. Fig. 5 compares the average molecular weight and PDI of imprinted polymer networks prepared in the presence of a polar solvent (water) with imprinted polymer networks prepared with no solvent present. We can see that presence of solvent during the creation of imprinted polymers using FRP results in an increase in the kinetic chain length and a decrease in the polydispersity index of polymer chains formed. The increase in the kinetic chain length is consistent with the increase in the observed propagation coefficient for polymer gels prepared in the presence of solvent. The small, polar water molecule readily associates with the monomer molecules aiding molecular diffusion. The increased effective propagation coefficient observed here is consistent with results reported in the literature for non-imprinted polymers [41,42]. The improved molecular diffusion in the polymerization solution may also explain the lower PDI since better transport of the polymer radicals would allow for more uniform growth. The use of LRP resulted in decreases in both the average molecular weight and the PDI which is consistent with the lower effective propagation coefficient and extended polymerization time discussed earlier. It is important to note that LRP overrides the effect of solvent in this case and the presence of solvent during the creation of imprinted polymers via LRP does not have a significant effect on either the average molecular weight or the PDI. Equilibrium binding isotherms for imprinted polymer networks prepared via FRP and LRP with varying template/functional monomer (T/FM) ratios in the pre-polymerization solution are shown in Fig. 6. The binding capacity and affinity for all networks were calculated using the Langmuir–Freundlich isotherm and are listed in

Fig. 5. Effect of solvent on kinetic chain length and polydispersity index of poly(HEMA-co-DEAEM-co-PEG200DMA) Chains. Weight average molecular weight (A) and polydispersity index (B) versus solvent content during polymerization of poly(HEMA-co-DEAEM-co-PEG200DMA) chains prepared via FRP (h) and LRP (j) are shown here. Error bars represent the standard error with n = 3.

Fig. 6. Equilibrium binding isotherms for diclofenac sodium binding by weakly crosslinked imprinted poly(HEMA-co-DEAEM-co-PEG200DMA) networks prepared via (A) FRP and (B) LRP with varying T/FM ratios. Data points represent poly(HEMAco-DEAEM-co-PEG200DMA) networks prepared with varying T/FM ratios: T/FM = 0 (—); T/FM = 0.1 (d); T/FM = 0.3 (N); T/FM = 0.5 (j). Error bars represent the standard error with n = 3.

V.D. Salian et al. / Reactive & Functional Polymers 78 (2014) 38–46

45

Table 2. From Table 2, one can observe that both the binding capacity and affinity increase as the T/FM ratio increases to 0.5. The T/FM ratio is a very important parameter which affects the strength and extent of monomer–template complexation in the pre-polymerization solution. This is especially true for systems where multiple non-covalent interactions can be formed between the template and functional monomer. As the T/FM ratio is increased, both template binding capacity and the mean template affinity increase since an increase in template concentration results in an increase in the number of template binding sites, which have higher binding affinities as compared to non-specific binding sites. However, after a certain optimal T/FM ratio as the template concentration is increased further, the template affinity begins to decrease [43]. The formation of good high affinity binding sites in systems with multiple weak non-covalent template–monomer interaction requires the presence of an excess of the functional monomer concentration to ensure all the complexation points on the template are satisfied by functional monomers. As the template concentration is increased, the T/FM ratio approaches the stoichiometric ratio and the excess monomer around a template molecule decreases. As a result, all of the complexation points on the template may not be satisfied resulting in the formation of specific binding sites with lower affinities even while the overall binding capacity may increase. However, the poly(HEMA-co-DEAEM-coPEG200DMA) networks primarily undergo ionic bonding with DS. The hydrogen bonding sites on DS may not be available for specific interaction with the template molecule due to intramolecular hydrogen bonding [44]. As a result, multi-point complexation is not expected to play a significant role in the formation of the template binding site. Thus, the increased affinity observed with increasing T/FM ratio can be explained by a higher proportion of specific interaction during template rebinding. From Fig. 6B, one can see that the template loading in the imprinted gels prepared using LRP increase with increasing T/FM ratio. Further, all gels prepared via LRP show an improvement over their corresponding gels prepared via FRP, and the extent of the improvement increases as the template concentration increases. Lastly, the capacity of the FRP gels increases by 23% for FRP1, 74% for FRP3, and 78% for FRP5 over FRP0; whereas the LRP gels show much larger increases of 37% for LRP1, 130% for LRP3, and 144% for LRP5 over LRP0. Similar results were observed when comparing binding affinities. It is important to note that the use of LRP in non-imprinted gels does not result in a significant improvement in their binding properties.

Fig. 7 shows equilibrium binding isotherms for weakly crosslinked polymer networks at varying functional monomer concentrations prepared with and without LRP. It is important to note that the T/FM ratio in the pre-polymerization mixture was kept at a constant value of 0.3 for all gels. The binding capacity and affinity values are presented in Table 2. From Fig. 7, it can be observed that the template loading increases significantly as the concentration of functional monomer is increased for both gels prepared via FRP and LRP. It was also noted that gels prepared without the positively charged functional monomer, DEAEM, displayed negligible binding of the template molecule. This can be explained by the absence of the charged DEAEM, which provided the most significant interaction. As reported in literature [43], although there exist two hydrogen bonding sites on the template, they may not be available for specific interaction with the template molecule due to intramolecular hydrogen bonding. Fig. 8 compares the equilibrium binding isotherms for networks prepared in the presence of water as a solvent as well as those prepared in its absence. The purpose of this experiment was to analyze the effect of a small solvent molecule which could alleviate some of the diffusional limitations inherent to late stage bulk polymerization reactions. Thus, a large amount of water was used (50% by weight). All other parameters were identical. The gels prepared via FRP demonstrated higher template binding when solvent was present during polymerization. Conversely, for polymers prepared via LRP, the template binding is lower when solvent was present during polymerization. Water is a small, polar, protic solvent. As discussed before, the presence of small solvent molecules during polymerization may lead to the formation of more homogenous networks in polymers prepared via FRP explaining the increased template binding. The increased homogeneity in the network structure is supported by the decreased PDI shown in Fig. 5B. However, for polymers prepared via LRP, the improvement in network structure was not significant enough to result in greater template binding. In addition, polar, protic solvents are known to interfere with non-ionic, non-covalent interactions, such as hydrogen bonding. Although, the primary ionic interaction between the template and the polymer chains should not be affected by the presence of water, it may disrupt weaker, secondary non-covalent interactions necessary for stabilizing the template binding cavity, resulting in lower template binding. However, for polymers prepared via FRP, the improved network homogeneity has a more significant impact than the disruption of secondary interactions by the solvent result-

Fig. 7. Equilibrium binding isotherm for diclofenac sodium binding by weakly crosslinked imprinted poly(DEAEM- co-HEMA- co-PEG200DMA) networks prepared via FRP and LRP with varying functional monomer concentration. Data points represent poly(HEMA-co-DEAEM-co-PEG200DMA) networks prepared with varying functional monomer (FM) concentrations: 0% FM (N,4); 5% FM (d,s); and 10% FM (j,h). Closed symbols (j,d,N) represent networks prepared via LRP while open symbols (h,s,4) represent networks prepared via FRP. Error bars represent the standard error with n = 3.

Fig. 8. Equilibrium binding isotherms for diclofenac sodium binding by weakly crosslinked imprinted poly(HEMA-co-DEAEM-co-PEG200DMA) networks prepared via FRP and LRP. Data points represent poly(HEMA-co-DEAEM-co-PEG200DMA) networks prepared in the presence (50% by weight) (N,4); and absence of solvent (d,s). Closed symbols (d,N) represent networks prepared via LRP while open symbols (s,4) represent networks prepared via FRP. Error bars represent the standard error with n = 3.

46

V.D. Salian et al. / Reactive & Functional Polymers 78 (2014) 38–46

ing in higher template binding. If the solvent interfered with the primary interaction, we would not see a significant increase in template binding for polymers prepared via FRP despite any possible improvement in network homogeneity. Conversely, if a noninterfering solvent were used, the imprinted polymers prepared via LRP in the presence of solvent would demonstrate higher template binding. It is important to note that the use of LRP in non-solvated systems showed the best results, but the improved template binding due to LRP may diminish as the solvent content in the prepolymerization solution is increased. However, it should also be noted that LRP has been shown to result in significantly improved template binding in highly crosslinked polymer networks even when a significant amount of solvent is present during polymerization [25–27].

Acknowledgements This work was supported by a grant from the National Science Foundation (NSF-CBET-0730903, Grant G00003191). References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12]

4. Conclusions This work highlighted the improvement in network homogeneity and imprinting efficiency in weakly crosslinked polymer networks prepared via controlled radical polymerization. It was demonstrated that the observed improvement in binding parameters in weakly crosslinked, imprinted polymer networks could be explained by the extension of the reaction-controlled regime during propagation of the macroradical, which allows the system to achieve a minimal global free energy configuration. This resulted in shorter polymer chains with lower PDI and, subsequently, polymer networks with greater homogeneity and smaller mesh sizes. It is likely that the more homogeneous network is due to the larger number of chains, the more uniform MW growth associated with the LRP mechanism, and the longer lifetime of a polymer chain relative to the total polymerization time. In addition, controlled radical polymerization negated the effect of the template on polymer chain growth resulting in polymers with a more consistent PDI independent of template concentration in the pre-polymerization solution. All imprinted poly(HEMA-co-DEAEM-co-PEG200DMA) networks prepared via controlled radical polymerization demonstrated significantly higher template binding capacity as well as affinity. It was also found that increased functional monomer concentration had a more significant effect on template binding properties as compared to template concentration, indicating the importance of the primary ionic non-covalent interaction to the formation of good imprinting sites. Lastly, the presence of solvent during conventional radical polymerization was shown to result in higher template binding as well as greater network homogeneity. However, the use of controlled radical polymerization in non-solvated systems showed better results while its use for creating polymers in the presence of solvent resulted in a decrease in template binding owing possibly to solvent interference during the formation of the template binding cavity. Thus, controlled radical polymerization can be an excellent tool for creating more efficient imprinted polymer networks with lower heterogeneity but care must be taken to account for the effect of other parameters during polymerization on the growth of the polymer network.

[13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44]

K. Haupt, K. Mosbach, Chem. Rev. 100 (2000) 2495–2504. C. He, J. Sun, Q. Wang Patent, Patent Number CN101029136A, 2007. M.E. Davis, A. Katz, W.R. Ahmad, Chem. Mater. 8 (1996) 1820–1839. M. Byrne, K. Park, N. Peppas, Adv. Drug Delivery Rev. 54 (2002) 149–161. M. Byrne, V. Salian, Int. J. Pharm. 364 (2008) 188–212. M. Byrne, J. Hilt, N. Peppas, J. Biomed. Mater. Res., Part A 84A (2007) 137–147. J.Z. Hilt, M.E. Byrne, Adv. Drug Delivery Rev. 56 (2004) 1599–1620. N.M. Bergmann, N.A. Peppas, Prog. Polym. Sci. 33 (2008) 271–288. D. Cunliffe, A. Kirby, C. Alexander, Adv. Drug Delivery Rev. 57 (2005) 1836– 1853. M. Ali, M.E. Byrne, Expert Rev. Clin. Pharmacol. 1 (2008) 145–161. C.J. White, M. Byrne, Expert Opinion Drug Delivery 7 (2010) 765–780. C.J. White, M.K. McBride, K.M. Pate, A. Tieppo, M.E. Byrne, Biomaterials 32 (2011) 5698–5705. C.J. White, Chemical Engineering, Auburn University, Auburn, AL, 2011. C.J. White, A. Tieppo, M.E. Byrne, J. Drug Delivery Sci. Technol. 21 (2011) 369– 384. A. Tieppo, C.J. White, A.C. Paine, M.L. Voyles, M.K. McBride, M.E. Byrne, J. Controlled Release 157 (2012) 391–397. A. Tieppo, K.M. Pate, M.E. Byrne, Eur. J. Pharm. Biopharm. 81 (2012) 170–177. M. Ali, S. Horikawa, S. Venkatesh, J. Saha, J.W. Hong, M.E. Byrne, J. Controlled Release 124 (2007) 154–162. L. Xinming, C. Yingde, A.W. Lloyd, S.V. Mikhalovsky, S.R. Sandeman, C.A. Howel, L. Liewen, Cont. Lens Anterior Eye 31 (2008) 57–64. C. Alvarez-Lorenzo, A. Concheiro, MML Ser. 7 (2006) 279–336. C. Alvarez-Lorenzo, H. Hiratani, A. Concheiro, Am. J. Drug Delivery 4 (2006) 131–151. R.J. Umpleby II, G.T. Rushton, R.N. Shah, A.M. Rampey, J.C. Bradshaw, J.K. Berch Jr., K.D. Shimizu, Macromolecules 34 (2001) 8446–8452. A.R. Kannurpatti, K.J. Anderson, J.W. Anseth, C.N. Bowman, J. Polym. Sci., Part B: Polym. Phys. 35 (1997) 2297–2307. A.R. Kannurpatti, J.W. Anseth, C.N. Bowman, Polymer 39 (1998) 2507–2513. A.D. Vaughan, J.B. Zhang, M.E. Byrne, AICHE J. 56 (2010) 268–279. A.D. Vaughan, M.E. Byrne, in: American Institute of Chemical Engineers, 2006, pp. 324c/321-324c/325. A.D. Vaughan, M.E. Byrne, PMSE Prepr. 94 (2006) 762–763. A.D. Vaughan, S.P. Sizemore, M.E. Byrne, Polymer 48 (2007) 74–81. V.D. Salian, A.D. Vaughan, M.E. Byrne, J. Mol. Recognit. 25 (2012) 361–369. G. Pan, B. Zu, X. Guo, Y. Zhang, C. Li, H. Zhang, Polymer 50 (2009) 2819–2825. S. Boonpangrak, M.J. Whitcombe, V. Prachayasittikul, K. Mosbach, L. Ye, Biosens. Bioelectron. 22 (2006) 349–354. A. Katchalsky, H. Eisenberg, J. Polym. Sci. 6 (1951) 145–154. R.J. Fort, T.M. Polyzoidis, Eur. Polym. J. 12 (1976) 685–689. K.S. Anseth, L.M. Kline, T.A. Walker, K.J. Anderson, C.N. Bowman, Macromolecules 28 (1995) 2491–2499. K.S. Anseth, C.M. Wang, C.N. Bowman, Polymer 35 (1994) 3243–3250. J.H. Ward, A. Shahar, N.A. Peppas, Polymer 43 (2002) 1745–1752. P. Flory, Principles Polym. Sci. (1953). G.G. Odian, Principles of Polymerization, John Wiley & Sons, 2007. G. Moad, D.H. Solomon, The Chemistry Of Radical Polymerization, Elsevier Science & Tech, 2006. H. Gao, K. Min, K. Matyjaszewski, Macromolecules (Washington, DC, US) 40 (2007) 7763–7770. H. Gao, A. Miasnikova, K. Matyjaszewski, Macromolecules (Washington, DC, US) 41 (2008) 7843–7849. S. Beuermann, M. Buback, P. Hesse, I. Lacik, Macromolecules 39 (2006) 184– 193. M. Stach, I. Lacik, D. Chorvat, Jr., M. Buback, P. Hesse, R.A. Hutchinson, L. Tang, Macromolecules (Washington, DC, US) 41 (2008) 5174–5185. H. Hiratani, Y. Mizutani, C. Alvarez-Lorenzo, Macromol. Biosci. 5 (2005) 728– 733. M.T. Ledwidge, S.M. Draper, D.J. Wilcock, O.I. Corrigan, J. Pharm. Sci. 85 (1996) 16–21.