Synthesis and Micellization of Bottlebrush Poloxamers

Abstract

Bottlebrush polymers are characterized by an expansive parameter space, including graft length and spacing along the backbone, and these features impact various structural and physical properties such as molecular diffusion and bulk viscosity. In this work, we report a synthetic strategy for making grafted block polymers with poly(propylene oxide) and poly(ethylene oxide) side chains, bottlebrush analogues of poloxamers. Combined anionic and sequential ring-opening metathesis polymerization yielded low dispersity polymers, at full conversion of the macromonomers, with control over graft length, graft end-groups, and overall molecular weight. A set of bottlebrush poloxamers (BBPs), with identical graft lengths and composition, was synthesized over a range of molecular weights. Dynamic light scattering and transmission electron microscopy were used to characterize micelle formation in aqueous buffer. The critical micelle concentration scales exponentially with overall molecular weight for both linear and bottlebrush poloxamers; however, the bottlebrush architecture shifts micelle formation to a much higher concentration at comparable molecular weight. Consequently, BBPs can exist in solution as unimers at significantly higher molecular weights and concentrations than the linear analogues.

Graphical Abstract

Block polymer amphiphiles in aqueous environments assemble into diverse micelle and vesicle morphologies,^1,2 with applications in numerous fields including pharmaceutical delivery^3,4 and as nanoreactors.⁵ Additionally, single chains or unimers interact with phospholipid bilayers in therapeutically relevant ways, ranging from membrane solubilization to stabilization.^6–12

Poloxamers are a specific class of linear block polymer amphiphile of poly(ethylene oxide) (PEO) and poly(propylene oxide) (PPO) that are commercially available (e.g., Pluronics^®), biocompatible, and have demonstrated efficacy as cell membrane stabilizing agents against many types of stresses.^{11,13–15,16} A range of prior studies have focused on the mechanisms of poloxamer-lipid interactions; however, a complete picture of the stabilization mechanism is still lacking.^{17,18,27–30,19–26} Amphiphilic bottlebrush polymers have potential as a tool to yield additional mechanistic insights, through architectural variation. Bottlebrush polymers have multiple parameters to tune (e.g., backbone length, graft length, graft spacing) and aspects of the relationships among these parameters and molecular properties such as segmental dynamics,^31–33 molecular shape,³⁴ and ability to share space,^35–38 are understood. Thus, a synthetic platform that enables bottlebrush block polymers with PEO and PPO side chains, “bottlebrush poloxamers (BBPs),” with control over molecular parameters could enable pursuit of many structure-function hypotheses to elucidate mechanisms of polymer-lipid membrane interactions.

Synthesis of BBPs is challenging. Controlled radical polymerization can produce grafted PPO polymers; however, these methods require low monomer conversion or low grafting density to maintain modest dispersities.^4,39–43 In this work, we disclose a synthetic route to BBPs that circumvents these challenges and gives more control over parameters such as graft length and graft density. We then examine the micellization behavior of a set of BBPs to understand the relationship between the apparent critical micellization concentration (CMC_a) and the total molecular weight. The resulting scaling relationship provides a comparison of the thermodynamic drive towards micellization between the bottlebrush and the chemically analogous linear architectures.

Our synthetic route is shown in Scheme 1. Living anionic polymerization followed by chain end modification was used to make norbornene (NB)-functionalized macromonomers (MMs), which were then polymerized via ring opening metathesis polymerization (ROMP). ROMP has successfully been combined with living polymerization strategies such as anionic, RAFT, and ATRP and is compatible with PEO.^44–47 This strategy has several advantages. First, anionic polymerization can yield low dispersity PEO and PPO polymers of a targeted molecular weight with well-defined α-chain ends.^11,48,49 Second, ROMP of NB macromonomers reaches >99% conversion within minutes. Third, ROMP avoids chain transfer/coupling side reactions that are often observed during radical polymerization, to which PPO macromonomers are vulnerable due to the abstractable proton on the tertiary carbon site of every repeat unit.⁵⁰ Together, these advantages enable high-throughput synthesis of high molecular weight, high grafting density, low dispersity BBPs with control over graft length and graft end group. The nomenclature adopted throughout is BB-E_q,p-b-BB-P_n,m where “BB” indicates bottlebrush, “E” and “P” indicate PEO and PPO respectively, and the subscripts are number average degrees of polymerization in reference to Scheme 1.

Scheme 1:

Reaction scheme to synthesize bottlebrush poloxamers

(a) Anionic polymerization was initiated with potassium tert-butoxide and ran to full conversion at 40 °C in the presence of 18-crown-6 ether. Hydrogenation was achieved over palladium on carbon. Esterification used a 50% molar excess of norbornene-carboxylic acid and was facilitated by DIC and DMAP. (b) Sequential ROMP was performed using Grubbs 3^rd generation catalyst.

The black trace in Fig. 1 shows matrix-assisted laser desorption/ ionization mass spectrometry (MALDI) data of the product PPO from anionic polymerization. Notably, there are two populations with peak spacings of 58 g/mol, indicating both are PPO. The more intense family of peaks is the desired product with a tert-butyl α-chain end and an alcohol ω-chain end (t-PPO-OH). The family labelled with blue squares is an impurity with an alkene α-chain end, arising from deprotonation of PPO methyl groups during polymerization.⁵¹ This assessment is confirmed by ¹H nuclear magnetic resonance (NMR) spectroscopy, as shown in Fig. S1. We employed established methods to limit this side reaction using 18-crown-6 ether,^49,51 but roughly 7 mol% (from NMR) of the chains have this impurity. We expected this alkene impurity to be susceptible to cross-metathesis during ROMP, leading to high dispersity polymers with limited control over molecular weight, Fig. S2.⁵² Thus, we performed a post-polymerization hydrogenation to eliminate this alkene impurity. The MALDI data in Fig. 1 show a 2 g/mol increase in molecular weight for each of the peaks identified in the alkene chain end population, blue squares to green circles, which indicates that this chain end modification was successful without changing the molecular weight distribution.

Figure 1.

MALDI-ToF data of PPO polymer as synthesized via anionic polymerization (black) and post hydrogenation (orange). M_w = 1220 g/mol and Đ = 1.12 for both polymers. The most intense family of peaks in both samples is tert-butyl-PPO-OH (t-PPO-OH) and is unchanged during hydrogenation. Alkene and alkane α-chain end impurities are labelled accordingly.

The final step of MM preparation was to convert the ω-alcohol to a NB-ester through a modified protocol for the PPO chemistry.^3,53 Fig. 2a shows a close-up of the MALDI spectrum to highlight the peak shifting after chain end modification; the full spectrum can be found in Fig. S3a. The blue triangle population, present as the majority in the starting material (black spectrum) and a minority in the product (blue spectrum), is t-PPO-OH. After esterification, this population shifts by +4 g/mol in m/z (yellow stars), indicating successful addition of a NB unit. Note that adjacent peaks do not have the same degree of polymerization, thus the expected peak shift after an end group modification is the difference in molecular weight of the end groups minus an integer multiple of the repeat unit molecular weight. A similar shift, +6 g/mol, occurs between the blue square and yellow circle populations, which correspond to the alkene-PPO-OH and alkane-PPO-NB, respectively. Note the black trace is from the material prior to hydrogenation while the blue trace is post-hydrogenation and esterification. Fig. S3c shows size exclusion chromatography (SEC) traces of the PPO post anionic polymerization and the functionalized MM, and the molecular weight distributions are identical. ¹H NMR spectroscopy was used to assess the purity of the resulting polymer and to corroborate successful chain end modification. The peak shape and integration of the alkene multiplet at 6.09–6.24 ppm (Fig. 2B) suggests that 95 mol% of the NB is on a polymer chain end.⁵⁴ Excess NB-carboxylic acid was needed to achieve high conversion and was difficult to remove due to its similar hydrophobicity to PPO and the small molecular weight of the PPO. We deemed this level of purity to be acceptable, but there will be some incorporation of NB-carboxylic acid in the polymer backbone. When accounting for the small molecule (~5 mol%) and the alkane ω-end group (~7 mol%) impurities in the analysis of the integral values of the NB peaks (6, 7, 8), which were referenced to the t-butyl peak (1), we conclude that quantitative conversion of alcohol to NB at the ω-end groups was achieved. A similar protocol was used to synthesize a NB-functionalized MM from a commercial methyl ether PEO (M_n = 1970 g/mol, Đ = 1.04). MALDI and ¹H NMR characterization are shown in Fig. S4.

Figure 2.

(a) MALDI of t-PPO-OH and the NB functionalized macromonomer, t-PPO-NB, following the hydrogenation and esterification reactions. M_w = 1210 g/mol and Đ = 1.12 for the SM and M_w = 1330 g/mol and Đ = 1.14 for the product. (b) ¹H NMR spectrum of the t-PPO-NB macromonomer.

After synthesis of MMs, sequential ROMP afforded a grafted polynorbornene diblock polymer, where one block has PEO side chains and the other PPO sidechains. To our knowledge, this is the first report of a bottlebrush PEO-b-PPO polymer. The representative SEC traces in Fig. 3 show that both the PPO block and the diblock have narrow dispersities, achieve the target molecular weight, and shift systematically to shorter retention times for higher M_n. The ¹H NMR spectrum in Figure S6b shows complete conversion of the monomeric NB alkene peaks to backbone alkene peaks, confirming that the reaction achieved full conversion of monomer. To our knowledge, this is the first report of a grafted PPO polymer where the macromonomer was fully consumed, making synthesis of multi-blocks possible. SEC traces and ¹H NMR spectra for all polymers discussed in this report are shown in Figs. S5–S7.

Figure 3.

SEC traces with differential refractive index detection of a representative sequential ROMP polymer (BB-E_{45, 20}-b-BB-P_{15, 10}). The macromonomer traces are included for reference. From the multi-angle light scattering detector and a Zimm analysis, the BB-PPO aliquot has a M_w = 11,600 g/mol and Đ = 1.20 and the BB-diblock has a M_w = 55,300 g/mol and Đ = 1.10, both were within error of the targeted molecular weight. The small peak at ~17 minutes in the BB-PPO chromatogram is likely a small molecule impurity in the PPO macromonomer, possibly 18-crown-6 ether, that does not impact the subsequent polymerization

To understand the micellization behavior of these BBPs, we synthesized three polymers with constant side chain lengths and similar compositions (70–80 wt% PEO), with total molecular weights ranging from 26 kDa to 394 kDa. All materials had narrow molecular weight distributions (Đ < 1.15), and the characterization data are summarized in Table 1. Micelle solutions were prepared via direct dissolution in aqueous buffer and were equilibrated at 37 °C for at least 24 hours. The CMC_a and hydrodynamic radii (R_h) were determined via dynamic light scattering (DLS). DLS measures the diffusive relaxation of concentration fluctuations in dilute solution, which yields the particle size distribution by analyzing the correlation function via cumulant analysis or inverse Laplace techniques such as regularized positive exponential sum.^55–58

Table 1:

Materials and micellization data summary.

Polymer	Mn [g/mol]a	Đ a	wt% PEOa	wt% NBa	CMCa [mg/mL]b	Rh [nm]c	Rcore [nm]d (calc)	Rcore [nm]e (expt)
E450 (PEO-20k)	20000	1.10	100	0	N/A
E76-b-P29-b-E76 (F68)f	8400f		80f	0	70 ± 10f
E103-b-P39-b-E103 (F88)f	11400f		80f	0	6 ± 0.9f
E132-b-P50-b-E132 (F108)f	14600f		80f	0	0.4 ± 0.1f
BB-E45,10-b-BB-P15,5	26200	1.07	72	8	8 ± 3	17 ± 4	5	6 ± 1
BB-E45,20-b-BB-P15,10	55000	1.10	74	8	2 ± 0.5	25	8
BB-E45,160-b-BB-P15,43	394000	1.15	80	7	0.4 ± 0.1	150	30	15 ± 8

^aSEC with multi-angle light scattering detection and Zimm analysis

^bDetermined by DLS

^cDLS data reported for 10 mg/mL solution using the second cumulant model. Error is the standard deviation of three independent replicates.

^dModel described in Fig. S16: sum of core forming block’s backbone contour length and twice the radius of gyration of its side chains

^eDetermined by cryo-TEM

^fData from Alexandridis et al. Macromolecules. 1994, 27, 9, 2414–2425 measured via fluorescent dye method.⁵⁹

To estimate the CMC_a, we analyzed the excess scattering intensity (I_ex = I − I_solvent) as a function of polymer concentration.^60,61 Because I_ex scales with the product of the number of scatterers and the size of the scatters, the concentration at which I_ex increases relative to the control (linear PEO homopolymer with M_n = 20 kDa) indicates aggregation (Fig. 4a). To ensure that the aggregates are micelles, we performed cryo-transmission electron microscopy (TEM) on the smallest and largest BBPs. A representative micrograph of the largest polymer is shown in Fig. 4b, which clearly demonstrates a core-corona structure. Micrographs of the smallest polymer clearly show a core, although the corona block is too short to be observed. Additional micrographs are shown in Figs. S8–9.

Figure 4.

(a) Excess scattering intensity at 37 °C, as a function of polymer concentration. Linear PEO with an M_n = 20 kDa is a negative control for micellization. (b) cryo-TEM micrograph of BB-E(45,160)-b-BB-P(15,43) at a concentration of 10 mg/mL in aqueous buffer with 40,000x magnification. (c) Trend between CMC_a and M_n for linear and bottlebrush poloxamers of similar wt% PEO. Data for linear poloxamers (40 °C) is taken from Alexandridis et al.⁵⁹ Data are fit to the exponential function shown on the graph and the correlation coefficients are 0.99 and 0.58 for linear and brush, respectively.

On qualitative inspection of Fig. 4a, as M_n increases, the CMC_a decreases. Quantitative analysis is challenging because micellization is not a phase transition; therefore, a precise CMC_a is difficult to extract and is sensitive to the method employed.^{62, 63} A representative CMC_a was taken as the intersection of two logarithmic fits: one at low polymer concentrations, and one at high polymer concentrations, representing the free chain and micelle regimes, respectively. The error is estimated as the difference in the CMC_a when the number of data points in the micelle regime is increased by 1. To ensure these CMC_a estimates are representative, we also fit the DLS data to a closed association model,^62,64 which includes the CMC_a as a fitting parameter (Fig. S10). Consistent trends were observed with both methods.

The relationship between M_n and CMC_a for 80 wt% PEO linear poloxamers⁵⁹ and 72–80 wt% PEO BBPs are shown in Fig. 4c. The free energy of a micelle is the sum of the core-corona interfacial energy, the energy of chain deformation, and the entropy of mixing between solvent and corona units. The closed association model asserts that above the CMC_a there is an equilibrium between unimers ([unimers] = CMC_a) and micelles of a constant size. Minimization of the total free energy of the system (micelles, unimers, and solvent) yields an exponential scaling between the CMC_a and M_n.^60,62,65,66 Therefore, the data are fit to an exponential form as shown in Eq. 1.

$$CM{C}_{\text{a}}=\text{A}{\times \text{e}}^{-\text{α}\times M_{\text{n}}}$$ (1)

Strikingly, the coefficient in the exponent of the fit (slope in Fig. 4c) is two orders of magnitude smaller in the bottlebrush case than in the linear case. Because α ~ χ, the Flory-Huggins interaction parameter, the reduced scaling coefficient suggests a decreased degree of solvent-core incompatibility, and therefore reduced thermodynamic driving force for micellization, in BBPs than in linear poloxamers. One explanation for this is that the PPO units near the hydrophobic NB backbone are relatively dehydrated in the unimer state. Thus, some of the entropic gain in the solvent upon dehydrating PPO in going from unimer to micelle is lost. Additionally, a BBP with M_n ~ 400 kDa has a comparable CMC_a to a linear poloxamer with M_n ~ 14 kDa, which is consistent with a recent molecular dynamics result.⁶⁷ This finding is of practical significance because in BBPs the CMC_a is less sensitive to molecular weight, enabling synthesis of vastly higher M_n materials that will exist as unimers in aqueous solutions.

This reduction of α for BBPs compared to linear poloxamers cannot be explained by the hydrophobic backbone. If the added hydrophobicity of the backbone were a significant contributor to micellization, we would observe the opposite effect, namely the BBPs would form micelles at lower concentrations compared to linear poloxamers due to the added hydrophobicity. Rather, we speculate that the backbone is significantly shielded from the aqueous environment by the side chains to reduce the enthalpic contribution of water-backbone interactions. Similarly, the 5 mol% carboxylic acid impurity along the backbone is presumably not fully exposed to the aqueous environment and likely does not impact the CMC_a. Also, the effect of the alkane α-chain end impurity on micellization is assumed to be negligible because it represents less than 0.3wt% of the hydrophobic block and has a minimal difference in hydrophobicity compared to either the tert-butyl end group or the PPO repeat unit.

Analysis of the correlation functions via the second cumulant model (Eq. S1) shows that R_h increases approximately linearly with molecular weight, suggesting that the chains adopt an extended conformation in the micelles. Additionally, for every micellar BBP solution a biexponential model (Eq. S2), accounting for two relaxation rates, yields a superior fit than a single exponential model (Figs. S11–S13). Multi-angle DLS on a 3 mg/mL solution of the highest M_n polymer confirmed that both relaxation modes are diffusive and have R_h values larger than expected for a unimer, suggesting existence of a bimodal micelle size distribution (Fig. S14). The TEM micrograph in Fig. 4b also hints at a bimodal population of spherical micelles. This could be due to the direct dissolution method not allowing the micelle size distribution to fully equilibrate; however, the micelle size distribution does not change when annealed at 37 °C for 35 days (Fig. S15). This suggests the micelles have reached a steady state. Recent theoretical and experimental work has shown that the side chain length asymmetry of the two blocks can impact micelle morphology, aggregation number, and surface roughness due to packing frustration at the core-corona interface.^68–70,71 We hypothesize that the additional free energy contribution of side chain crowding near the core-corona interface creates at least two local minima in free energy-micelle size space that are separated by a significant energy barrier; thus leading to a bimodal micelle size distribution.

To estimate the micelle core radius (R_core), we modelled the core-forming block as a cylinder plus a hemisphere, where the cylinder length equals the contour length of the backbone and the hemisphere radius is twice the radius of gyration of a PPO side chain, Fig. S16. This simplified model assumes that the backbone is a rigid rod, the side chains are Gaussian coils, and that there is no chain overlap. Comparison of the calculated and experimental R_core values in Table 1 shows that this model is reasonable at short core block lengths (N_c,bb ~ 5), but is an overestimate for longer chains (N_c,bb ~ 40). This behavior resembles that of the Kratky-Porod wormlike chain model, and it suggests that in the micelle core a sufficiently long bottlebrush PPO can adopt a flexible conformation despite the high grafting density.^72,73 Finally, for polynorbornene grafted with 1 kDa PPO side chains a persistence length estimate of ~10 nm is consistent with a transition from rod-like to coil-like behavior occurring within the interval 5 < N_c,bb < 40. This estimate is markedly higher than the 0.7 nm persistence length of a linear polynorbornene polymer,⁷⁴ as expected for a densely grafted bottlebrush.^31,75

In conclusion, we report an efficient, high-throughput synthetic strategy to create bottlebrush poloxamers for the first time. The combination of living anionic and ROMP polymerization affords control over graft length and graft end-group and enables quantitative conversion of macromonomer. A series of hydrophilic bottlebrush poloxamers was synthesized over a range of molecular weights and the micellization behavior was compared to linear poloxamers. Bottlebrush poloxamers exhibit a remarkably reduced driving force for micellization compared to linear poloxamers as evidenced by a two order of magnitude smaller scaling exponent between M_n and CMC_a.