Improved Oxidative Biostability of Porous Shape Memory Polymers by Substituting Triethanolamine for Glycerol

Abstract

While many aromatic polyurethane systems suffer from poor hydrolytic stability, more recently proposed aliphatic systems are oxidatively-labile. The use of the renewable monomer glycerol as a more oxidatively-resistant moiety for inclusion in shape memory polymers (SMPs) is demonstrated here. Glycerol-containing SMPs and the amino alcohol control compositions are compared, with accelerated degradation testing displaying increased stability (time to complete mass loss) as a result of the inclusion of glycerol without sacrificing the shape memory, thermal transitions, or the ultralow density achieved with the control compositions. Gravimetric analysis in accelerated oxidative solution indicates that the control will undergo complete mass loss by approximately 18 days, while lower concentrations of glycerol will degrade fully by 30 days and higher concentrations will possess approximately 40% mass at the same time. In real time degradation analysis, high concentrations of glycerol SMPs have 96% mass remaining at 8 months with 88% gel fraction remaining that that time, compared to less than 50% mass for the control samples with 5% gelation. Mechanically, low glycerol-containing SMPs were not robust enough for testing at three months, while high glycerol concentrations displayed increased elastic moduli (133% of virgin materials) and 18% decreased strain to failure. The role of the secondary alcohol, as well as isocyanates, is presented as being a crucial component in controlling degradation; a free secondary alcohol can more rapidly undergo oxidation or dehydration to ultimately yield carboxylic acids, aldehydes, carbon dioxide, and alkenes. Understanding these pathways will improve the utility of medical devices through more precise control of property loss and patient risk management through reduced degradation.

Graphical Abstract

INTRODUCTION

The use of polyurethanes in medical devices spans decades, with pacemaker wires and breast implants being some of the best known examples of long-term implantable devices.^1–9 In most cases, aromatic diisocyanates and either polyester or polyether chain extenders were used, which upon degradation and as synthesis byproducts will form aromatic amines, which are carcinogenic, and other degradation products.^5–8 Carbamate scission primarily is a result of hydrolysis, catalyzed by pH change, which will form the amines.^5–9 The amines may also form during the synthesis process, with the end result still being the risk of patient exposure to toxic substances.⁷ In an effort to improve upon these limitations, Wilson et al developed a series of highly crosslinked, thermoset polyurethanes from aliphatic, symmetric monomers.⁴

Wilson’s polymers possessed a shape changing functionality, rendering them shape memory polymers (SMPs), which were previously proposed for a variety of medical applications, such as aneurysm and vascular occlusion devices.^1–3 The driving force of shape memory in thermosets are the crosslinks, which determine the original geometry of the bulk material; deformation of the chains and the relative entropic maxima are what allow for the strain fixation (strain fixity) into a temporary, secondary shape.^3–4,10 The external stimulus that drives the shape change originally is also used to return the SMP to its original configuration; for several polymer species, including polyurethanes, water and other hydrogen bonding molecules may induce a solvent shape memory effect through interruption of the fixed shape by plasticizing the polymer at lower temperatures.^11–14 Thermoplastic systems typically display higher recovery forces with lower recoverable strains, while thermoset SMPs have the opposite trend, and with sufficient strain recovery are ideal for void space occluding applications, such as in aneurysm treatment devices in clinical settings.^4,15–19

Highly porous materials are of great interest across a variety of disciplines, particularly in medicine.^20–21 Wilson’s highly porous SMPs have demonstrated biocompatibility and rapid thrombus formation as aneurysm occlusion devices, although recent studies have demonstrated mass loss via oxidative chain scission of the SMP backbone.²² The amino-alcohols used to achieve the highly crosslinked structure can be readily oxidized, which may be of concern in specific medical device applications where reduced gravimetric loss, and therefore decreased degradation product accumulation, are desired as a method of reducing regulatory or patient risk.^22–27 While it is assumed that even with rapid material loss there is minimal patient risk associated with the implantation of these materials, there is still a need for more biostable (slower degrading) materials as alternatives for other applications where higher macrophage loads are possible. The tertiary amine crosslinking monomers will rapidly oxidize and fragment to form lower amines, aldehydes and acids, and eventually ammonia, during oxidation of the thermoset network; the replacement of these amino-alcohols with a less susceptible monomer will result in increased oxidative stability and further reduce patient risk.^22–27

Glycerol was selected to replace triethanolamine (TEA), one of the amino-alcohols used in urethane synthesis, as it is the most susceptible of those examined.²² In addition to the tri-functionality of glycerol, which will theoretically not reduce the crosslink density and strain recovery of the bulk material, it is also inexpensive and a renewable resource that has gained attention for a variety of applications.^28–40 While glycerol has increased in popularity for polymers in recent years, there are still several factors that must be studied to access the monomer’s utility in more biostable SMP formulations. In literature, dendritic glyercol and polyglycerols are known to yield networks containing free secondary alcohols on the glycerol unit after synthesis, which may contribute to oxidation.^36,40 Degradable polymers have been synthesized utilizing glycerol, but the degradable linkages have been determined to be from the copolymer component, such as sulfide or ether linkages.³⁷ While a number of studies have examined molecular glycerol for oxidation behavior, only with catalysts are the reported yields significant.^28–35 To our knowledge, no work has been done to characterize what, if any, degradation occurs for glycerol-based urethanes as a function of the secondary alcohol being free or incorporated into a carbamate.

Here we present the substitution of glycerol in place of the oxidatively labile amino alcohol monomers are a method of improving biostability over the SMP lifespan. The presented SMPs maintain the high porosity, low density, and prized urethane chemistry so highly desired in commercial spaces. Stoichiometric amounts of the amino-alcohol hydroxypropyl ethylene diamine (HPED) and glycerol were varied for characterization of degradation behavior, as well as basic material properties. The role of glycerol crosslinker in the oxidation process was determined through gravimetric analysis, mechanical testing, and model compound studies. Free secondary alcohols in the glycerol molecule demonstrated the most rapid fragmentation through a series of oxidation and/or acid catalyzed reactions; glycerol with three carbamate linkages displayed slow oxidation to ketones and 1,3 diketones with little fragmentation occurring. While low concentrations of glycerol were not found to effectively alter the degradation properties, higher concentrations demonstrated substantially improved oxidative resistance. The role of the secondary alcohol in oxidation was examined, comparing free alcohol oxidation with a secondary carbamate linkage using model compound studies. Mass spectrometry was used to characterize the degradation products and their relative abundances, and mechanisms of degradation in hydrogen peroxide are proposed.

EXPERIMENTAL

Materials:

N,N,N’,N’-tetrakis(2-hydroxypropyl)ethylenediamine (HPED, 99%), triethanolamine (TEA, 98%, Alfa Aesar), glycerol (Gly, Sigma Aldrich, 99%) and 2,2,4-trimethyl hexamethylene diisocyanate (TMHDI, 98%, TCI America, a mixture of 2,2,4 and 2,4,4 monomers) were used as monomers. Hexamethylene diisocyanate (HDI, 98%) and isophorone diisocyanate (IPDI, 98%) were also purchased from Sigma Aldrich and were used without modification. Deuterated chloroform (CDCl₃), ethyl isocyanate (EtNCO, 98%), ethanol (190 proof), cobalt chloride anhydride (CoCl2), hydrogen peroxide (50% v/v), methanol (LCMS grade), and acetonitrile (LCMS grade) were obtained from Sigma Aldrich and used without further modification.

Spectroscopic and X-ray Crystallography Characterization:

Solution state nuclear magnetic resonance (¹³C 125 MHz and ¹H 300 MHz) was performed on a Mercury 300 MHz spectrometer operating in the Fourier transform mode with CDCl₃ as the standard. Solid state CP (cross-polarization) MAS NMR (¹³C 100 MHz) was performed using a 4.0 mm ¹³C probe on an Avance-400 Solids spectrometer (Bruker). A frequency of 123.63 MHz was used, with a spin speed of 3.5 kHz and 4.5 kHz and 90˚ pulse lengths of 4.5 μs. A recycle delay of 5 s and a contact time of 1.6 ms were the used measurement parameters. Edited CP MAS NMR experiments were performed as well, to observe positively-phased signals for C and CH carbons, negatively-phased CH_2, and zero-intensity signals for CH carbons to assist in peak assignments, with an external TMS.

Attenuated total reflectance Fourier transform infrared spectroscopy (ATR-FTIR) spectroscopy was used to determine any spectroscopic changes to the bulk material. ATR-FTIR spectra were taken using a Bruker ALPHA infrared spectrometer (Bruker, Billerica, MA) using 32 scans per spectra for both background and samples. Spectra data was collected in absorption mode with a resolution of 4 cm^-1. OPUS software was used to examine spectra, identify peaks, and perform baseline and atmospheric corrections. Examinations were performed in triplicate to confirm results, and comparisons were made to previously published SMPs.

Liquid chromatography mass spectrometry (LCMS) was performed using a single quad OrbiTrap (ThermoFisher) with Exactive software and a C18 normal phase silica column. The capillary and heater temperatures were set for 50C to prevent thermal-induced fragmentation.

X-ray diffraction (XRD) characterization was performed using a Bruker D8 Advance, with a Cu x-ray tube of 1 kW maintained at 40 kV and 25 mA. The Bragg-Brentano para-focusing standard mode was used with a 1 mm DS slit and a two-circle 218 mm diameter computer controlled goniometer.

Synthesis:

A traditional two-step polyurethane synthesis method was used to produce the porous SMPs.^1–4 An isocyanate premix with approximately 40% of required alcohol groups was homogenously mixed and cured at 50*C for 36 hours. The isocyanate premix was added to surfactants (DC 5943, 7.5% wt) and mixed using a high speed shear mixer. An alcohol premix, containing the remaining alcohols, catalysts, and water was mixed and added to the isocyanate premix. After mixing both components with a physical blowing agent, the mixture was cured in the oven at 90˚C for 30 minutes. Foams were cold cured overnight at room temperature, and were washed using RO water and IPA with sonication, followed by overnight drying under vacuum.

Model Compounds:

Degradation products were determined using model compound studies. The behavior of TEA and HPED in the urethane network was examined by Weems et al, and is not repeated here.²² Model compound synthesis of an idealized urethane and urea were used to determine the spectroscopic signal. Glycerol and ethyl isocyanates were added to produce 2G (OH:NCO 3:2) and 3G models (1:1 OH:NCO) by reacting the two reagents in a round bottom flask overnight at 40˚C, followed by vacuum at 50˚C for 12 hours. Diisocyanates were reacted with excess amounts of ethanol to produce small molecule dicarbamates. TMHDI, HDI, and IPDI were examined in this manner. Synthetic schemes and specific procedures for each compound are presented in Supplemental Materials.

Thermomechanical Characterization:

Differential scanning calorimetry (DSC) was also used to measure both wet and dry T_g using a Q-200 DSC (TA Instruments, Inc., New Castle, DE).¹⁶ Samples of approximately 5.0 mg ± 1.0 mg were sealed in TZero aluminum pans and placed in the heating cell. The test profile was as follows: equilibration at −40°C, heating to 120°C at 10°C/min, cooling 10°C/min to −40°C and holding for 5 minutes, and a final heating to 120°C at 10°C/min. The half-height transition of the final heating cycle was the reported T_g. Wet samples were weighed and sealed in the same manner, and were then heated from −40°C to 80°C.

Shape Memory Characterization:

Shape recovery of cylindrical foam samples (6 mm diameter, 10 mm length, six samples per series) crimped over a wire was examined at 50°C in RO water to determine the volume recovery behavior (strain recovery).¹⁶ Samples were crimped to a minimal diameter (approximately 1.0 mm using a SC150–42 Stent Crimper (Machine Solutions, Flagstaff, AZ). To crimp, samples were equilibrated at 100°C for ten minutes and then radially compressed and cooled to room temperature. Samples relaxed for 12 hours and were tested over the course of 30 minutes. Image J was used for analysis of the change in diameters over time.

Microscopy Analysis:

Foam cell structure was determined by cutting axial and transverse samples that were examined using scanning electron microscopy (SEM). Samples were mounted onto a stage and sputter coated with gold using a Cressington Sputter Coater (Ted Pella, Inc., Redding, CA) for 60 seconds at 20 mA. Samples were then examined using a Joel NeoScope JCM-5000 SEM (Nikon Instruments Inc., Melville, NY) at 11X magnification and 15 kV under high vacuum.

Degradation Testing:

For analysis of degradation rates, cleaned samples were completely immersed in respective solutions of 2% H₂O₂ (real time oxidation) and 20% H₂O₂ with 0.1 M CoCl₂ (accelerated oxidation), both stored at 37˚C.²² Sample solutions were changed every 3 days to ensure a relatively stable ion concentration, and once per week samples were removed, cleaned in EtOH, and dried in at 50˚C under vacuum (30 in Hg), after which sample mass was recorded and samples returned to fresh solution.

Gel fraction was determined for five samples of selected compositions. Samples (approximately 0.10 mg) were immersed in 30 mL of ethanol and heated to 50˚C for five days. The samples were then removed, blotted dry, and placed in a vacuum oven at 50˚C (30 in Hg) for 2 days to remove any solvent before being weighed.

Model compounds were dissolved in 50% H₂O₂ and placed in a 37C oven for four weeks, with samples withdrawn for analysis on a weekly basis; peroxide was refreshed at the same time. Samples were diluted 100 fold in a 50% methanol (49.95% water, 0.05% formic acid) solution, and injected into the LCMS column. The injection volume was 10μL, and the flow gradient was varied from 0% acetonitrile to 95% acetonitrile over a 3-minute gradient. Exactive software was used for analysis of the chromatography scan and the peaks.

RESULTS AND DISCUSSION

Material Characterization:

Stoichiometric amounts of HPED and glycerol were varied for characterization of degradation behavior, as well as basic material properties to produce porous SMPs with 20% glycerol (20 Gly, containing 10% by endgroups glycerol, 40% HPED, and 50% TMHDI), 40% glycerol (40 Gly), and 80% glycerol (80 Gly). Initial gel fraction for all materials was found to be greater than 95%, regardless of the concentration of glycerol or amino alcohols. Aliphatic diisocyanate species (not shown) was also found to not significantly impact the bulk crosslink density, agreeing with previous studies using amino alcohols with varied diisocyanates.²²

It is expected that with high concentrations of free secondary alcohols that shape recovery may be inhibited, due to increased hydrogen bonding as well as adhesive forces within the pores and along the struts/membranes of the SMPs.^41–42.13C NMR (SI Figure 1) confirms that despite the reactivity differences between primary and secondary alcohols, there will be low concentrations of free secondary alcohols remaining when stoichiometry is balanced; preliminary experiments (not examined further) with excess alcohols greater than 10% displayed this behavior.

Figure 1.

Shape memory testing of the SMPs comparing the control with the first and tenth cycle of shape recovery for the 20 Gly SMP with 80 Gly and control (0 Gly), with the cyclic trend over 10 cycles displaying no change in strain recovery (insert).

Shape memory behavior of the materials were tested using in vitro conditions (37°C and immersed in RO water), demonstrating that glycerol content would decrease the rate of shape recovery without altering the ultimate recoverable strain (all samples recovered more than 99% strain) (Figure 1). Cyclic testing of 20 Gly demonstrated that repeated compression and expansion did not impact the recovery rate or final strain over 10 cycles. It is expected that the decrease in the rate of recovery comparing the glycerol-containing SMPs and control compositions is most likely due to the hydrophobicity of the remaining alcohol monomer (HPED) which contains four methyl groups and has a higher degree of crosslinking, thereby increasing SMP backbone rigidity and T_g.²²

The T_gs of the SMPs were found to be approximately 74˚C when dry (Table 1), although the plasticized T_g, which is more representative of the environment experienced by implanted SMPs, ranges from 42 to 48˚C, agreeing with previous studies that have utilized TMHDI with HPED and TEA.^4,22 This behavior may also contribute to the shape recovery behavior, as the higher T_g (relative to room temperature) allows for substantial strain fixation at ambient conditions without being so high as to make the polymer matrix glassy and therefore fragile.

Table 1.

Basic characteristics of SMPs before and after oxidation.

Composition	Wet  DSC	Dry  DSC	Gel Fraction (%)	% Crystallinity
	(˚C)	(˚C)
Control	47.8	74.0	99.8	0
Control 8 months	28.2	65.1	~5	0*
20 Gly 0 months	47.2	73.9	97.9 ± 0.4	0
20 Gly 8 months	22.6	54.6	15.3 ± 9.8	0*
80 Gly 0 months	42.5	74.9	97.8 ± 0.8	0
80 Gly 8 months	35.2	67.9	87.8 ± 1.6	0*

^*No crystalline sections were found by DSC or XRD, but thermal examination indicates that crystallization is possible

Thermal sweeps using DSC also allowed for examinations of possible crystalline domain formation, which was supported by XRD analysis (Figure 2), with virgin samples displaying no crystallinity. The lack of crystallinity in virgin materials is expected, as these are highly crosslinked polymers with little space between crosslinks, and agree with results presented by Wilson.⁴ This characterization also indicates that long segments of polyureas are not forming during the foaming process, and that the urethanes are well distributed about the SMP. If long polyurea segments were forming as a result of poor isocyanate premix synthesis or improper mixing during the foaming process, crystalline domains would form in the SMP matrix as demonstrated with polyurea chains produced from isophorone diisocyanate.¹⁹ Further characterization of the degraded samples is discussed below.

Figure 2.

XRD plots of degraded glycerol SMPs after incubation in 2% H₂O₂ at 37°C, demonstrating no significant crystallization has occurred (greater than the 2% uncertainty threshold of the analysis) (A), and DSC heating profiles (B), displaying the ability to crystallize for the 20 Gly 8 months degraded samples, which may also be present in the 80 Gly 8 months degraded samples at a substantially reduced concentration.

Bulk Degradation Characterization:

Scanning electron microscopy (SEM) images confirm the gravimetric results discussed later, with the control and 20 Gly losing all morphological features by the end of the degradation studies (8 months real time degradation), agreeing with previous studies examining in vitro and in vivo driven changes in control materials containing TEA as well as the behavior of other degradable polyurethane foams.^22,43–44 The 80 Gly was found to have minimal changes in the pore morphology and membranes (Figure 3), indicating a greater oxidative stability. Additionally, the T_g was found to decrease by approximately 13˚C, which may be due to the reduction in crosslinking and increase in chain end mobility, discussed in greater detail with the degradation assessment.

Figure 3.

SEM images of SMPs over the course of degradation in 2% H₂O₂ displaying control (top, A: 0 days, B: 3 months, C: 8 months), 20 Gly (middle, D: 0 days, E: 3 months, F: 8 months), and 80 Gly (bottom, G: 0 days, H: 3 months, I: 8 months) SMPs at 0 (left), 3 (right), and 8 months (centers).

While none of the SMP samples were shown to have formed crystalline segments as a result of oxidation, DSC did reveal a cold crystallization peak to be present in the degraded samples without any obvious melting transitions. Further examination of degraded samples also revealed minimal crystallinity formation with higher concentrations of glycerol. However, with lower concentrations, such as the presented 20 Gly materials at 8 months, crystallinity may be induced at higher temperatures, as demonstrated through DSC. This indicates that the tertiary amine scission to secondary and primary amines as well as aldehydes and carboxylic acids are producing linear polymer chains, crosslinked by the urethanes linkages off of the glycerol moiety. The higher chain integrity of the glycerol allows the amines to fragment into smaller molecules while the glycerol will oxidatively fragment at a reduced rate (Figure 4).

Figure 4.

(A)Proposed oxidative degradation pathway of the 2G molecule, examined at 37 °C in 50% H₂O₂ based on results using LCMS. (B) Accelerated mass loss (20% H₂O₂ with 0.1 M CoCl₂) examined over 35 days, and (C) real time mass loss (2% H₂O₂) examined over the course of real time oxidation out to eight months (samples were maintained at 37˚C). (n=7) Representative stress-strain curves for 20 Gly SMPs (D) and 80 Gly SMPs (E) at each examined degradation time point.

Gravimetric analysis of the samples in accelerated oxidative solution (20% H₂O₂ with 0.1 M CoCl₂, found in previous studies to be the equivalent of 50% H₂O₂ without catalyst) indicates that the control will undergo complete mass loss by approximately 18 days, while lower concentrations of glycerol will degrade fully by 25 to 35 days (for 20 and 40 Gly, respectively).²² At 35 days, 80 Gly had approximately 40% of mass remaining, and while the rate of mass loss is similar to 40 Gly, the time to initial mass loss is substantially improved from ca. 2 days to 12 days. In real time testing performed in conjunction with SEM analysis (2% H₂O₂, found in previous studies to be the most accurate for real time degradation modeling), 80 Gly had ca. 96% mass remaining at 8 months, compared to less than 50% remaining for the control samples, expected due to the decreased tertiary amine concentration.²²

Gel fraction analysis provides insight into the less mobile degradation products and oligomers within the bulk material. Such products may form after chain scission but are not highly water soluble and are therefore introducing error into the gravimetric analysis as well as reducing the understanding of polymer degradation rates. Gel fraction of 80 Gly for real time oxidation conditions reduces from 98% to 88% over the course of testing, but the control and 20% Gly reduce to approximately 5% and 15%, respectively (Figure 5). With approximately identical T_gs for control and Gly-based SMPs, the mass loss from oxidative degradation is dependent on the amines. While the crosslink density is theoretically the same, the network should be less fragmented at the glycerol units compared with the control TEA units at early time points.

Figure 5.

Gel fractions of SMP samples at time 0, 3 months and 8 months, comparing varied glycerol concentration effect on oxidative stability (with the remaining alcohol fraction of 100 consisting of the amino alcohol HPED). Samples were incubated in 2% H₂O₂ at 37°C. (n=7)

To support the gravimetric and microscopic analysis, tensile testing (ASTM Type IV D dogbones, 5 mm/min strain rate) was used to provide a dynamic aspect to the degradation studies, and revealed an increase in modulus over time (Table 2), along with a decrease in strain to failure and tensile strength. The 20 Gly samples were not robust enough to test after three months; 80 Gly SMPs had elastic moduli 133% of the original at eight months (80 Gly 8 months). The strain to failure decreased by approximately 18%, and tensile strength decreased by 33%. For material evaluation in an environment that most closely simulates in vivo conditions, the mechanical samples provide the shortest time for performance testing.

Table 2.

Mechanical property changes in 20 Gly and 80 Gly SMP samples over the course of real time oxidative testing (2% H₂O₂ solution) (n=7).

	Young’s   Modulus (MPa)	Strain to Failure (%)	Tensile Strength (MPa)
20 Gly 0	1.77 ± 0.20	29.6 ± 4.8	0.23 ± 0.02
20 Gly 1 month	1.74 ± 0.20	18.9 ± 5.0	0.17 ± 0.03
20 Gly 1.5 month	2.29 ± 0.43	3.8 ± 0.8	0.05 ± 0.03
20 Gly 2 monthϯ	2.25ϯ	3.1ϯ	0.06ϯ
20 Gly 2.5 month*	38.94 ± 0.10	0.5 ± 0.1	0.08 ± 0.00
80 Gly 0	1.29 ± 0.13	34.4 ± 0.5	0.24 ± 0.04
80 Gly 1 month	0.87 ± 0.31	32.3 ± 5.2	0.18 ± 0.03
80 Gly 1.5 month	1.68 ± 0.21	31.5 ± 10.0	0.27 ± 0.10
80 Gly 2 month	1.57 ± 0.71	41.8 ± 8.3	0.25 ± 0.06
80 Gly 2.5 month*	1.38 ± 0.20	31.9 ± 0.06	0.24 ± 0.05
80 Gly 3 month	1.68 ± 0.41	32.6 ± 13.3	0.15 ± 0.04
80 Gly 8 month	1.88 ± 0.70	28.5 ± 6.1	0.16 ± 0.01

^ϯdenotes only 1 sample was able to be tested, the others had undergone failure prior to testing

^*denotes only 3 samples were able to be tested, the others had undergone failure prior to testing

Spectroscopic analysis of the bulk SMPs revealed a carbamate carbonyl peak shift (FTIR) after degradation due to the formation of adjacent carboxylic acids, as has been found previously with the degradation of the amino alcohols (Figure 6).²² This shift was found in the 20 and 40 Gly, but not the 80 Gly, which further agrees with the gravimetric results and indicates a reduced oxidation and lack of -COOH formation; these conclusions are also supported by ¹³C NMR (SI Figure 2–6). The lack of carbonyl peak shift indicates a lack of (1)-like moieties in the SMPs, supported by the increase in oxidative stability with sufficiently high concentrations of glycerol used to synthesize the SMP.

Figure 6.

Fourier Transform Infrared (FTIR) spectroscopic signals (representative curves) of the original and degraded glycerol-containing SMPs displaying 20 Gly samples (A) at 0, 3, and 8 months and (B) 80 Gly samples at 0 and 8 months after incubation in 2% H₂O₂ (real time oxidation solution).

Unfortunately, moving to 100% glycerol led to alterations in the synthetic procedure related to miscibility problems and was not further studied. Optimization of the SMP synthesis process to reduce the secondary alcohol concentration and incorporation of 100% glycerol as a source of alcohol monomers is expected to ultimately yield SMPs with greater oxidative stability compared with those presented here. However, spectroscopic analysis of the current SMPs does not reveal an abundance of secondary alcohols in the compositions with the current procedure and is supported by the remaining 95.7% mass of the 80 Gly sample (87.1% gel fraction) in real time oxidative solutions at 8 months, as opposed to the 20 Gly sample (51.6% mass remaining and 15.3 ± 9.8 % gel fraction) and control sample (49.1% mass remaining and approximately 5% gel fraction). This testing does indicate a threshold for this biostability at approximately 6 to 7 years for complete mass loss, based upon previous characterization of similar formulations.²²

Model Compounds and Polymer-Chain Scission Analysis:

In order to determine the contribution of each component to the oxidative degradation, liquid chromatography mass spectrometry (LCMS) was used. Relative abundances and species of model compounds (1) (2G (containing 2 urethane linkages and a free secondary alcohol) and (11) (3 urethane linkages) produced from oxidation in 50% H₂O₂ solution (at 37 ˚C over 4 weeks) were found using LCMS, supported by ¹³C NMR (SI Figure 1, 7–8). The progression of the 2G oxidation is demonstrated: (1) is consumed by formation of the ketone (2), then a 1,2 diketone (3) which undergoes fragmentation. While there is rapid oxidation apparent, as noted by the formation of carboxylic acids and aldehydes at week 1, it is still apparent that the majority of (1) does not undergo fragmentation until ca. weeks 2–3 (Figure 7). It appears that by week 4 most of the fragmentation of the starting material has occurred. However, 3G (11) will oxidize to the ketone (12) and diketone (13) without fragmentation over the same time. This supports the results indicated by increased oxidative stability for the 80 Gly samples compared with the control and 20 Gly samples. Though the glycerol-urethane linkages will imbue increased stability for the 20 Gly SMPs, the remaining amino alcohol linkages are still susceptible and will fragment at the same rate as the control, leading to a fragmented network in a similar manner as before, and explains the same gravimetric change behavior for the 20 Gly and control in accelerated testing. Additionally, the HPED has been demonstrated to be more stable compared to the TEA, which will also impact the rate of mass loss with the 80 Gly when accounting for the initial amino alcohol degradation. [22]

Figure 7.

Degradation products and starting model compounds relative abundances (50% H₂O₂ at 37°C) 2G (A) and 3G (B) with the proposed oxidation scheme (C).

Figure 8.

Relative abundances of degradation products due to oxidation of the diisocyanate-backbone of HDI (A), TMHDI (B), and IPDI (C) after incubation in 50% H₂O₂ at 37°C, determined using LCMS.

As can be determined from the model degradation studies, by 4 weeks, 2G will have formed more than 4 times the number of fragments than 3G, and will be completely consumed in approximately 2–3 weeks, while non-oxidized 3G starting material is still present in ca. 50% by mass. More importantly, our results indicate that the SMP network which contain few remaining alcohol moieties from glycerol monomer will remain intact, as the model compounds may non-specifically oxidize but do not undergo substantial further fragmentation or degradation under these highly oxidative conditions, which will thereby reduce the formation of risky degradation byproducts, such as oxalic acid in the control formulations.⁴⁵

While it was not expected that the urethane linkage was contributing to the gravimetric behavior, model compound studies were also pursued for varied aliphatic; aliphatic diisocyanate-based model compounds were found to possess high oxidative stability (Figure 8).

As a representative case, HDI may produce oxidation products including ketones and carbamic acids (Figure 9), but the abundances of these products were below ca. 5% after 5 weeks in accelerated oxidative conditions. TMHDI and IPDI presented similar results, although product abundances may be as high as 10% with these species. Additionally, IPDI may further fragment, producing amines, but this is expected to again occur in low concentrations in this environment, and more importantly, these are aliphatic diamines and will be less cytotoxic compared with aromatic hydrolysis products. It is not known whether the main contribution to this minimal fragmentation is oxidation or acid-catalyzed hydrolysis, but the rates of mass loss indicate that the diisocyanate component and the urethane linkages play a minimal role in the oxidative degradation process.^46–47

Figure 9.

Major degradation products for the fragmentation of diisocyanate monomer analogs after incubation in 50% H₂O₂ at 37°C; (A) hexamethylene diisocyanate (HDI), (B) trimethyl hexamethylene diisocyanate (TMHDI), isophorone diisocyanate (IPDI).

During this time, the maximum loss was approximately 5% of the initial IPDI model, which formed a carbamic acid (carbamate), due to hydrolysis that may be favored with the ethanol end groups, or an epoxide or ketone along with a carbamoyl moiety. The latter two are fragments most likely associated with oxidation that would occur in the bulk SMP and correspond to approximately 2% of the composition at 5 weeks. For TMHDI model, ketones are the most common product; the carbamic acid (carbamate) is again a product most likely associated with the model compound used rather than what would be seen during degradation, although acid hydrolysis of the urethane is possible due to the peroxide. The ketone group can be associated with carbon 3 or 5 (2 or 4 for the isomer), with carbon 3 oxidation being thermodynamically favorable due to electron donating effects of the surrounding methyl and dimethyl substituted carbons involved in radical stabilization. The tertiary alcohol formed (at carbon 2 or 5) will not undergo further changes, and is stable, as are the ketone products; the alcohol appears to be the most thermodynamically favored product. The same behavior is found for HDI, with ketone and carbamate products formed. The HDI analog appeared to be the most stable, and prone to crystallization. This indicates that HDI-based SMPs could have a lower mass loss rate compared with TMHDI-based SMPs despite the lower T_g, as a result of superior chain packing.^48–49 It also provides a measure of understanding of the behavior relative to thermoplastic, semi-crystalline polyurethanes, which undergo significantly slower rates of degradation.^48–49 Additionally, previous studies using aromatic diisocyanates have predicted the formation of aromatic amines during oxidative degradation; the corresponding aliphatic amines resulting from urethane-bond scission were not found, due to the slow hydrolysis rate of alkyl carbamates relative to the carbamate hydrolysis here.²²

CONCLUSIONS

Overall, we believe that the tertiary amines present in the SMP (from HPED) will oxidize preferentially, and after the amines are consumed, the glycerol will be the main molecular component to oxidize and fragment. At this point, the glycerol oxidation will be the primary cause of mass loss from the bulk SMP, albeit at a much slower rate compared with the amines. However, as we demonstrate, unreacted secondary alcohols will be the site of more rapid oxidation and subsequent polymer fragmentation. As the concentration of glycerol increases in the SMP, the fragmentation of amines will have less impact on the initial mass loss, which is demonstrated by the 80 Gly gravimetric results.

For rapidly testing the biostability of multiple compositions prior to devoting resources to long-term studies, a model compound study combined with mechanical testing seems to provide rapid insight into the relative biostability of aliphatic-based SMPs. Gravimetric studies and microscopic analysis provide excellent comparisons for determining the role of composition in oxidative stability, although at a slower rate when utilizing environmental conditions for real time approximations. For the glycerol-containing SMPs, the high degree of crosslinking without containing residual alcohols seems to produce a more biostable product. We further demonstrate that the oxidative degradation that occurs will be most likely due to the fragmentation of these free alcohols and any amino alcohols present, and that the completely reacted glycerol will be biostable once the diketone has formed. Based upon the known biocompatibility of glycerol-based polymers, and the presented attributes with regard to degradation, these SMPs do have excellent potential for improving the stability from the current iterations of SMPs, enhanced more than 3 times compared with the formulations containing TEA.