Internal and External Catalysis in Boronic Ester Networks

Abstract

In dynamic materials, the reversible condensation between boronic acids and diols provides adaptability, self-healing ability, and responsiveness to small molecules and pH. The thermodynamics and kinetics of bond exchange determine the mechanical properties of dynamic polymer networks. Here, we investigate the effects of diol structure and salt additives on the rate of boronic acid–diol bond exchange, binding affinity, and the mechanical properties of the corresponding polymer networks. We find that proximal amides used to conjugate diols to polymers and buffering anions induce significant rate acceleration consistent with internal and external catalysis, respectively. This rate acceleration is reflected in the stress relaxation of the gels. These findings contribute to the fundamental understanding of the boronic ester dynamic bond and offer molecular strategies to tune the macromolecular properties of dynamic materials.

Graphical Abstract

The reversible condensation between boronic acids and diols has been applied to sensors,^1–6 drug delivery,^{3, 6–9} and dynamic hydrogels^{3, 6, 8, 10} due to its biocompatibility, selectivity, and reactivity under ambient-temperature aqueous conditions. Among dynamic covalent chemistries^11–13, boronic acid–diol exchange offers the fastest uncatalyzed rates, enabling rapid responses to physical and chemical stimuli.^{1, 14, 15} Within polymer networks, boronic esters endow self-healing behavior, stimuli responsiveness, and viscoelasticity. Experiment, simulations, and theory have revealed that the macroscopic properties of dynamic networks are highly dependent on the thermodynamics and kinetics of the reversible crosslink.^16–27 Specifically, the equilibrium constant (K_eq) of the dynamic reaction determines crosslink density in networks, and thus the stiffness of the polymer network (frequency-independent plateau modulus, G_p) (Figure 1a). On the other hand, important dynamic behaviors such as self-healing and viscoelasticity are governed by the kinetics of the reversible reaction. In dynamic networks that undergo exchange through dissociative mechanisms, the rate of dissociation (k_r) is thought to determine the viscoelasticity of the network, which is characterized by the crossover frequency (ω_c) at which material transition occurs from solid-like to liquid-like behavior (Figure 1a). Therefore, to better control and design boronic ester-based materials for specific applications, it is essential to understand the molecular factors that influence both kinetics and thermodynamics of this dynamic bond.

Figure 1.

(a) Translation of molecular parameters into macroscopic network properties. Solid lines and dashed lines represent storage modulus (G’) and loss modulus (G”) of each material with distinct molecular parameter, respectively. The binding constant (K_eq) relates to the stiffness (plateau modulus, G_p) while the rate of dissociation (k_r) determines the viscoelasticity in dissociative networks. (b) Summary of factors affecting the boronic acid–diol reversible condensation and focus of this study (in bold). (c) This work uncovers internal catalysis in boronic ester networks by the linker used for diol conjugation, and external catalysis by buffering anions.

The dynamic reaction between boronic acids and diols occurs via multi-step mechanisms and are sensitive to many factors: pKa of the boronic acid and diol, sterics and dihedral angle of the diol, pH of the solution, and buffer composition and concentration (Figure 1b).^{10, 28, 29} As a result, the reaction landscape varies significantly depending on the choice of boronic acid, diol partner, and reaction conditions. Efforts to understand how these factors affect the thermodynamics of the dynamic bond have focused on maximizing the binding affinity (K_eq) at physiological pH.^{10, 28, 30–34} Those studies have introduced new classes of boronic acids with high binding affinities, such as Wulff-type boronic acids and benzoxaboroles, and identified diols and polyols with high hydrolytic stability (e.g. nopoldiol and salicylhydroxamic acid).^35–40 In contrast, design principles to tune the kinetics of the reaction have not been systematically developed (see Supporting Information (SI), Table S1 for a summary of kinetic studies).^41–49 Furthermore, factors of particular relevance to polymer networks, such as the conjugation chemistry used to attach the dynamic bond to polymers or commonly used buffers for hydrogels, are not typically included in small-molecule model studies.

Here, we report the effect of diol structure and salt additives on both kinetics and thermodynamics of the reversible boronic acid–diol condensation and their impact on the mechanical properties of boronic ester networks. We measured the exchange kinetics and binding affinity between different diols/polyols and phenylboronic acid (PBA) by one-dimensional selective exchange spectroscopy (1D EXSY) and ¹H NMR. We then synthesized star polymers with analogous diol/polyol and boronic acid end groups and measured the mechanical properties of the resulting gels using shear rheology. Through these measurements, we uncovered two new factors that affect the mechanics of boronic ester networks: the presence of amide groups proximal to the diol, and the composition of the buffer (Figure 1c). We showed that amide groups used to conjugate diols to polymers act as internal catalysts via a solvent-insertion mechanism,⁵⁰ accelerating both esterification and hydrolysis. Additionally, we observed salt-dependent rate constants correlated to the pKa of the anion’s conjugate acid. These changes in molecular thermodynamics and kinetics are translated into gels with a wide range of mechanical properties, with shear moduli ranging from 1.8 kPa to 27 kPa and relaxation times from 0.1 s to 571 s, based only on changes in diol structure or salt additive. Our results inform new design parameters that can be used to tune boronic ester-based dynamic materials and complement existing studies that have focused on the boronic acid partner. Furthermore, these findings emphasize the importance of designing small-molecule model systems that capture essential components of the polymer network when relating molecular parameters to network properties.

To study the effect of diol structure on boronic ester formation and hydrolysis, we selected a representative 1,2-diol, 1,3-diol, triol, and polyol with varying degrees of flexibility and steric encumbrance (Figure 2a, “simplified alcohols”). The reaction kinetics of the reversible condensation between PBA and each alcohol were studied by using 1D EXSY. EXSY NMR detects magnetization transfer via chemical exchange, allowing quantitative measurement of the rate of dynamic processes occurring slower than the NMR time scale. In 1D EXSY, a peak of interest is selectively excited, and the intensities of the irradiated and exchanged peaks are monitored as a function of mixing time.

Figure 2.

(a) The reversible condensation between PBA and alcohols, and the structures of the alcohols used in this study. (b) Representative ¹H NMR (top, at equilibrium) and 1D EXSY with different mixing times (bottom) with PBA and NP at 25 °C. The peak at 7.97 ppm (ortho-proton of PBA) is selectively irradiated. (c) Analysis of 1D EXSY data. The concentration change of PBA over mixing time was fit with a first-order reversible rate law. (d) Arrhenius plot for pseudo-first-order rate constants at different temperatures; activation energies were calculated from the slope.

The dynamic reaction was studied under pseudo-first order conditions (see SI for details). Under these conditions, the ortho-protons of the boronic acid and ester can be selectively irradiated in the 1D EXSY experiment (Figure 2b). The decrease in acid concentration following irradiation with respect to mixing time was fit to a first-order reversible rate law to obtain pseudo-first-order rate constants for the forward (esterification, k_obs,f) and reverse (hydrolysis, k_obs,r) reactions (Figure 2c).⁵¹ These measurements were repeated at several temperatures to determine the activation energies for esterification and hydrolysis (Figure 2d).

The measured kinetic parameters and equilibrium constants for each alcohol are summarized in Table 1; activation energies (E_a,f for esterification and E_a,r for hydrolysis) are reported in Table S3. Consistent with previous work,²⁹ the flexible 1,3-diol neopentylglycol (NP) exhibited a significantly lower binding affinity to PBA compared to cis-1,2-cyclopentanediol (CP), which has a locked geometry that favors ester formation but slows kinetics dramatically. Increasing the number of OH groups also enhanced the binding affinity (NP < Tris < Sorb). As the functionality increases, the rate of esterification increases, and the rate of hydrolysis decreases, albeit to a lesser extent.

Table 1.

Equilibrium and rate constants for the reversible esterification/hydrolysis of PBA with various alcohols.

Alcohols	K eq	kobs,f (s−1)	kobs,r (s−1)	k rel,f c	krel,rc°
NP	0.577 (±0.009)	0.0473 (±0.0004)	0.082 (±0.001)
Tris	1.389 (±0.006)	0.1100 (±0.0003)	0.0792 (±0.0003)
Sorb	7.4 (±0.6)	0.23 (±0.01)	0.031 (±0.002)
CP	15a (±2)	0.0216 (±0.0004)	0.0014b (±0.0002)
Tris-amide	0.182 (±0.006)	2.20 (±0.04)	12.1 (±0.3)	20	150
NP-amide	0.877 (±0.005)	0.327 (±0.001)	0.373 (±0.002)	6.9	4.5
(syn)-CP-amide	1.45 (±0.02)	0.210 (±0.002)	0.145 (±0.002)	9.7	100
(anti)-CP-amide	5.4 (±0.4)	0.107 (±0.006)	0.020 (±0.001)	5.0	14
Glu-amide	6.4 (±0.5)	0.70 (±0.03)	0.110 (±0.008)	3.0	3.5

^aEstimated using K_eq = ([ester][H₂O]²)/([acid][diol]) because hydrolysis was too slow for the rate constant to be extracted by fitting; see SI for details.

^bEstimated from K_eq using k_obs,r = k_obs,f/K_eq.

^cCalculated by k_obs (amide derivatives)/k_obs (simplified alcohols).

We next synthesized star polymers with end groups that contain the structures of the simplified alcohols (Figure 3a; see SI for synthetic details). A mixture of anti and syn diastereomers was used to prepare PEG-CP; as a result, its end groups represent a 4:1 ratio of anti and syn diastereomers in all cases unless otherwise noted (see SI for assignment). Each alcohol-terminated polymer was mixed with PEG-PBA in a 1:1 ratio to form a 10 w/v% solution in propylene carbonate containing 1% (v/v) D₂O. This solvent mixture was required based on the low binding constants of some of the alcohols tested, which preclude gelation in fully aqueous environments. The gels were characterized by oscillatory shear rheology within the linear viscoelastic regime to determine the frequency-independent plateau modulus (G_p) (Figure 3b; see SI for amplitude sweeps).

Figure 3.

Mechanical and self-healing properties of the gels formed from 4-arm PEG-PBA and PEG-alcohol (1:1, 10 w/v% in propylene carbonate with 1% D₂O). (a) Chemical structures of modified PEG polymers. PEG-CP end groups represent a mixture of diastereomers (4:1 anti:syn; see SI for assignment). (b) Frequency sweeps at 5% strain obtained by oscillatory shear rheology. (c) Normalized stress relaxation profiles following 5% shear step strain. (d) Calculated τ for the different gels (bar graph, left axis) and their plateau moduli, G_p (black dot, right axis). Mechanical data are the average of 3 samples; error is the standard deviation of the mean. (e) Photographs of PEG-Tris, PEG-Glu and PEG-CP gels undergoing self-healing.

In a dynamic network, the number of elastically active polymer strands, which determines G_p, is dictated by K_eq of the crosslink. Tibbitt recently modified the phantom network model to directly relate G_p to K_eq in dynamic networks and demonstrated the validity of this model in a boronic ester hydrogel.²¹ Based on the K_eq values obtained for the simplified alcohol in Table 1, we expected the relative plateau moduli of the corresponding gels to follow the trend PEG-NP < PEG-Tris < PEG-Glu < PEG-CP. However, we observed that the moduli followed the trend PEG-Tris << PEG-NP < PEG-CP < PEG-Glu (Figure 3d).

To relate the kinetics of the dynamic bond and the viscoelasticity of the gels, we applied a step strain (5%) to the gels and measured the stress relaxation over time (Figure 3c). These data were fit to a Maxwell model to obtain the characteristic relaxation time τ, the time required for the material to relax stress to 1/e of its initial value. The theoretical foundation established by Semenov and Rubinstein revealed an inverse relationship between τ and the dissociation rate of the crosslink (τ ∝ 1/ k_obs,r),^{23, 24} and experimental studies have confirmed this relationship.^{17–21, 26, 52–54} Based on the measured k_obs,r values of the simplified alcohols, we expected to observe the trend PEG-NP ~ PEG-Tris < PEG-Glu << PEG-CP. However, the trend in relaxation times was PEG-Tris < PEG-Glu < PEG-NP << PEG-CP (Figure 3d).

Analogously, the rate of self-healing ability in the network should be determined by the rate of crosslink formation (k_obs,f).^{26, 27, 55} We qualitatively compared self-healing of PEG-Tris, PEG-Glu and PEG-CP gels by cutting and re-forming the gels, and observed the fastest self-healing in PEG-Tris (Figure 3e). Based on the measured k_obs,f trend, we expect self-healing to occur fastest for PEG-Glu. Taken together, the discrepancies between the small-molecule parameters and the mechanical properties of the networks led us to re-evaluate the use of simplified alcohols as model compounds.

The key structural difference between the simplified alcohols in Figure 2a and the crosslinkers in the gel is the presence of amide groups used for conjugation. Previously, our lab has taken advantage of photoswitchable Lewis basic groups proximal to the boronic ester to modulate the kinetics and thermodynamics of its dynamic covalent reactions.^56–58 Anslyn has shown that the basic aminomethyl group in Wulff-type boronic acids catalyzes the addition and elimination of alcohols and water.^{42, 59} However, analogous internal catalytic effects for the diol component of this dynamic covalent reaction are far less studied. Hall showed that the proximity of an ether linkage to nopoldiol had a 1.4–1.8-fold effect on esterification rate,³⁶ but the effect on hydrolysis rate was not quantified. We hypothesized that the amide groups introduced for polymer functionalization perturb the rate and equilibrium constants of the esterification and hydrolysis reactions, through inductive effects, direct coordination to the boron center,^{60, 61} or internal catalysis.^{62, 63}

Therefore, we synthesized amide derivatives of each alcohol (Figure 1a, “amide derivatives”) and measured the kinetics of their dynamic reactions with PBA using 1D EXSY (Table 1). Indeed, we observed dramatic increases in both esterification and hydrolysis rates for the amide derivatives compared to their simplified alcohol counterparts. These increases ranged from a factor of 3 (Sorb vs. Glu-amide) to greater than an order of magnitude (Tris vs. Tris-amide, 20-fold increase in k_obs,f and 150-fold increase in k_obs,r). With Glu-amide, we observed an additional peak that undergoes exchange with both acid and ester (Figure S18). We propose this peak is an intermediate species as multiple binding modes can exist with ring-opened gluconolactone.¹⁸

To obtain insight into the relative contributions of inductive and neighboring-group effects, we prepared the anti and syn diastereomers of CP-amide and separately measured their reaction kinetics with PBA. Relative to unfunctionalized CP, the minor isomer of CP-amide (syn) experiences more significant rate acceleration (9.7-fold for esterification and 100-fold for hydrolysis) than the major isomer (anti, 5.0-fold and 14-fold, respectively) (Table 1). We propose that (syn)-CP-amide experiences greater rate acceleration than the anti diastereomer because the amide carbonyl is on the same face of the cyclopentane ring as the diol and is thus well positioned to activate incoming nucleophiles (Figure S19). Furthermore, using ¹¹B NMR, we obtained evidence that the boronic ester of NP-amide forms a solvent-inserted tetrahedral species in methanol-d₄, which is not observed with NP (Figure S20).^{50, 64} These results suggest that internal catalysis occurs via a solvent-assisted mechanism, wherein the amide carbonyl facilitates proton transfer when water or alcohol attacks the boron center.

The amide derivatives exhibited more pronounced rate acceleration for hydrolysis than for esterification, except for NP-amide. As a result, the binding constants for Tris-amide, CP-amide, and Glu-amide with PBA are lower than those of their simplified alcohol counterparts. Therefore, when the amide effect is considered, the trend in binding affinities for the amide model systems (Tris-amide < NP-amide < CP-amide < Glu-amide) is consistent with the trend in plateau modulus observed for the gels. Furthermore, the self-healing trends for the networks are qualitatively consistent with the k_obs,f trends for the amide derivatives, with Tris-amide and PEG-Tris displaying the highest k_obs,f and self-healing rate, respectively.

The k_obs,r values predict a trend in bulk relaxation times (Tris-amide < NP-amide < Glu-amide < (anti)-CP-amide) that more closely resembles the gel data. The ratio of anti:syn isomers in PEG-CP affects the relaxation time, with a greater fraction of syn isomer (anti:syn = 1.5:1 vs. 4:1) corresponding to faster stress relaxation (τ = 355 vs. 571 s), again providing evidence for the relevance of internal catalysis in the mechanism by which the network relaxes stress (Figure S38). While the measured rate constant for Glu-amide hydrolysis is ~3-fold slower than NP-amide hydrolysis based on our EXSY experiments, PEG-Glu relaxes stress ~2-fold faster than PEG-NP. However, this discrepancy is ascribed in part to the buildup of an intermediate in the dynamic reaction of Glu-amide, which may affect the rate constants measured by EXSY.

Overall, our data agree with the paradigm first articulated by Craig for dynamic networks: “strong means slow”, meaning that the gels that resist flow have slower kinetics for the crosslink-breaking step.⁵³ Previous studies of supramolecular networks have shown that the frequency sweeps for chemically distinct networks can be superimposed when scaled to a molecular parameter such as k_r or E_a,r (“time-crosslinker superposition”).^{17, 65} However, those systems undergo flow through single-step, dissociative mechanisms. For our boronic ester networks, it was not possible to superimpose the frequency sweep data by scaling to 1/k_obs,r (Figure S37). We ascribe these differences to the complexity of boronic ester hydrolysis, a multi-step mechanism in which the dominant pathway and speciation can change depending on the structure of the boronic acid and diol, and environmental factors like solvent and pH.^{41, 48, 66} Consistent with changes in mechanism or rate-limiting step, the small-molecule activation energies are not consistently lowered by the presence of an internal catalyst (e.g. E_a,r is 20 kJ/mol for Tris and 36 kJ/mol for Tris-amide; see Table S3).

Many boronic ester-based hydrogels are designed for biomedical applications, so are typically formed and studied in the presence of buffers or media containing inorganic salts, amino acids, glucose, and vitamins. While previous studies have focused on the pH of the medium rather than the buffer itself, Springsteen and Wang showed that the binding affinity of a catechol dye to PBA is sensitive to buffer composition and concentration.^{30, 67} Furthermore, London and Gabel showed that the presence of phosphate anion increased the dissociation rate of 4-fluorophenylboronic esters.⁶⁸ In the course of our studies, we also observed significant changes in viscoelasticity of gels based on the identity and concentration of buffer salts. Therefore, we used 1D EXSY to systematically determine the effect of ions on the rate of boronic ester exchange. NP was used as the model diol for this study, and various salts were introduced to the solution. To rule out any changes in rate due to pH, the pH of each solution was adjusted to 7 using either DCl or NaOD.

The pseudo-first-order rate constants for esterification and hydrolysis in the presence of various salts are shown in Table 2. Using a common buffer, phosphate-buffered saline (PBS), we observed two orders of magnitude increase of the rate constants for both the forward and reverse reactions. The effect of phosphate anion was further confirmed with NBu₄H₂PO₄, which revealed similar rate acceleration. London and Gabel proposed that phosphate anions catalyze proton transfer from water during hydrolysis.⁶⁸ We further observed that the extent of acceleration was correlated with the pKa of the conjugate acid: KOAc provides ~7-fold acceleration, compared to D₂O alone, and NBu₄PF₆ shows no effect (see Table S4 for additional salts). K_eq was not affected by salt in acetone (Table 2), but a two-fold change was observed in propylene carbonate (Figure S43).

Table 2.

Rate constants for the reversible esterification/hydrolysis of PBA with NP in the presence of different salts at pH 7.

Salt	pKaa	kobs,f (s−1)	kobs,r (s−1)
No salt	-	0.0473 (±0.0004)	0.082 (±0.001)
PBS (NaH2PO4 + Na2HPO4)	H2PO4− = 7.21	5.59 (±0.07)	8.2 (±0.1)
NBU4H2PO4		4.90 (±0.06)	7.9 (±0.1)
KOAc	CH3COOH = 4.76	0.371 (±0.002)	0.574 (±0.005)
NBU4PF6	-	0.0587 (±0.0006)	0.095 (±0.001)

^aFrom pKa data compiled by R. Williams.⁶⁹

We anticipated that the dramatic effects on hydrolysis rate would also translate to the viscoelasticity of boronic ester gels. Indeed, compared to the PEG-NP gel without salts, relaxation times are 10 and 1000 times faster in the presence of 0.01 M KOAc and 0.01 M NBu₄H₂PO₄, respectively (Figure 4a). This 1000-fold range in relaxation times is visually apparent and is achieved solely through the addition of salts, without any change in pH (Figure 4b).

Figure 4.

Effects of anion coordination on the mechanical properties of the PEG-NP gels (10 w/v% in propylene carbonate with 1% 1M salt in D₂O) formed with solution containing different salts (final concentration of salt: 0.01M). pH of the 1M stock solution of the salt in D₂O was adjusted to 7 with either DCl or NaOD. (a) Frequency sweep with 5% strain. (b) Photographs of PEG-NP gels at pH 7 in the absence and the presence of 0.01M NBu₄H₂PO₄ salts.

Small-molecule model studies, including our own,⁵⁸ are often performed with simple diols like ethylene glycol or fructose in organic solvents such as DMSO, which simplifies analysis. In comparison, the attachment of diols to polymers requires the installation of Lewis basic functional groups like amides, triazoles,⁷⁰ and oximes.⁷¹ Furthermore, the resulting materials are often studied in complex aqueous environments like buffer or media. Our findings suggest that small-molecule model studies based on simple diols in organic solvents do not sufficiently capture the behavior of materials, particularly in the translation from molecular kinetics to dynamic mechanical properties.

By comparing the small-molecule kinetics of dynamic covalent reactions to the macroscopic properties of covalent adaptable networks, we have discovered two factors that affect gel mechanics by accelerating boronic acid esterification and hydrolysis: proximal amides commonly used to conjugate crosslinkers to polymers, and anions found in buffers and media. The extent of this rate acceleration can be precisely tuned through the position of the amide relative to the diol and the pKa of the anion’s conjugate acid, and therefore can be viewed as tunable design parameters for dynamic materials. The internal catalysis effect may extend to other Lewis basic groups used for polymer modification, with both distance and three-dimensional structure influencing the magnitude of this effect^{36, 70, 71}, or to other classes of boronic acids such as benzoxaboroles and Wulff-type boronic acids. These effects could be exploited in the design of dynamic materials that change physical properties in response to anions or diols. We envision this work will contribute to a growing understanding of the boronic ester dynamic bond and provide strategies to molecularly engineer dynamic materials.