Multicompartment Micelles by Aqueous Self-Assembly of μ-A(BC) n Miktobrush Terpolymers.

The aqueous self-assembly of μ-A(BC) n miktobrush terpolymers has been studied using dynamic light scattering and cryogenic transmission electron microscopy. In this system, the A block is hydrophilic poly(ethylene oxide), "O", the B block is hydrophobic poly(methylcaprolactone), "C", and the C block is hydrophobic and oleophobic poly(perfluoropropylene oxide), "F". Two terpolymers were examined: one with an average of about two C blocks and two F blocks and another with an average of about three C blocks and two F blocks. In both cases, the total molar mass is near 40 kg mol-1, and the volume fraction of the single O block is greater than 50% of the whole. Both samples form multicompartment micelle structures with subdivided solvophobic cores of C and F domains. The morphologies observed are generally analogous to those previously observed for the self-assembly of μ-ABC miktoarm star terpolymers, namely, "raspberry" and "hamburger" micelles; however, an intriguing multicompartment polymersome morphology with compartmentalized solvophobic bilayers is also observed. These results are interpreted in terms of the relative strengths of the competing interactions among the three blocks and the solvent and in terms of the constraints imposed by the miktobrush architecture.

Introduction

AB diblock copolymers self-assemble into a small set of micellar morphologies in A-selective solvents, and their synthesis and study have captivated research interest across various scientific disciplines.[1−11] More elaborate micelle structures have also been achieved by increasing the chemical and architectural complexity of the polymers; model examples include the solution self-assembly of ABC block terpolymers with two mutually incompatible and solvophobic blocks, B and C. In a solvent selective for the A block, ABC block terpolymers can form multicompartment micelles with solvophobic cores comprising distinct B and C domains.[12−25] The arrangement of B and C domains in the core will depend upon the block connectivity, and thus the architecture of the ABC triblock terpolymers, and their size and shape will be influenced by the relative interfacial tensions (and thus chemistry, χ, the Flory–Huggins interaction parameter, and N, the degree of polymerization) of chains A, B, and C and their respective interactions with the surrounding solvent.[19,26−29] For example, a mutually immiscible and linear ABC terpolymer in a solvent selective for A will typically form concentric B and C cores owing to the block connectivity.[30−34] However, an equivalent linear BAC terpolymer may form “raspberry” micelles with subdivided B and C domains in the core surrounded by looped A coronal chains.[35,36] Multicompartment micelles with nonconcentric morphologies and individually accessible hydrophobic cores enable applications that exploit the dual uptake and release properties of the separate domains.[37]

Multicompartment micelles with distinct subdivided cores were first demonstrated by the self-assembly of μ-ABC miktoarm star terpolymers. Because of the mandatory convergence of all three chains in the miktoarm architecture, the formation of concentric domains (“onion” micelles) is suppressed, thereby favoring multicompartment micelles with well-defined and subdivided solvophobic B and C cores. Furthermore, the ability to tune each block chemistry and size has enabled a suite of accessible morphologies from a single polymer architecture.[21,37−45] This flexibility is advantageous when tailoring the terpolymer and thus the solution morphology toward a particular application. We have recently built upon the architectural design of the previous μ-ABC miktoarm star approach by extending the number of repeat units of (BC) chains conjoined to A chains within the starlike architecture (the aforementioned μ-ABC structures would be those with n = 1), thus achieving μ-A(BC) “miktobrush” terpolymers.[46] We adopted a simple synthetic strategy to access a series of such terpolymers with increasing values of n and thus a potentially wide compositional range of μ-A(BC) polymers from three end-functionalized polymer building blocks. Because of the high incompatibility and unusual connectivity of the blocks within the μ-A(BC) terpolymers, we were interested in exploring their self-assembly in an aqueous solution and in uncovering the morphological consequences of this architectural variation. In this report, we describe the aqueous self-assembly of two members of this family of terpolymers, with n ≈ 2.5 but with differing composition, which can be considered analogous to the miktoarm stars with an average of six arms.

Methods

Materials

μ-A(BC)miktobrush terpolymers were synthesized by a reversible addition fragmentation chain transfer (RAFT) alternating, controlled radical copolymerization, as described previously.[46] Briefly, three end-functionalized polymers were synthesized and then were combined in the RAFT copolymerizations to give μ-O(CF)miktobrush terpolymers: poly(ethylene oxide) (PEO), “O” block (21 kDa) with a RAFT chain transfer agent (CTA) end group; poly(methylcaprolactone) (PMCL), “C” macromonomer block (5.3 kDa) with a maleimide end group; and poly(perfluoropropylene oxide) (PFPO), “F” macromonomer block (2.6 kDa) with a styrene end group. After the preparation in a mixed solvent system, the μ-O(CF) terpolymers were precipitated into a mixture of toluene/hexane (1:9, v/v) to remove any residual C or F macromonomers. All three polymer building blocks and μ-O(CF) terpolymers were characterized using size exclusion chromatography (SEC) equipped with triple detection (refractive index, ultraviolet, and multiangle laser light scattering) to determine molar masses (Mn and Mw), dispersity values (D̵), and purity, as determined by the presence of any residual macromonomers. All homopolymers used for the preparation of the miktobrush terpolymers were further characterized using 1H, 13C, and 19F NMR spectroscopies to determine their composition, molar mass (MnNMR), and end-group fidelity. The compositions of the homopolymers were confirmed using elemental analysis (Table S1). In the case of the miktobrush terpolymers, we relied on the elemental analysis of the two samples in this study to confirm their composition, given some modest internal inconsistencies in the 1H NMR spectra, which are likely associated with some solution aggregation in the nuclear magnetic resonance (NMR) solvent that limited the accuracy of the integrations (see Figures S1 and S2)—μ-O(C2F2): Calcd C = 53.80%, H = 7.95%, F = 9.19%, N = 0.08%, S = 0.17%. Found C = 53.61%, H = 7.92%, F = 8.85%, N = 0.14%, S = 0.33%. For the analysis of μ-O(C3F2): Calcd C = 55.38%, H = 8.14%, F = 8.03%, N = 0.10%, S = 0.23%. Found C = 54.70%, H = 8.09%, F = 7.64%, N = 0.17%, S = 0.19%. Integer values for the brush degree of polymerization are used in the article. For the calculated values based on the experimental elemental analysis, see Table S2.

Cryo-TEM

Cryo-transmission electron microscopy (cryo-TEM) samples were prepared by suspending approximately 100–200 nm thick menisci on lacey carbon TEM grids, which were then rapidly plunged into liquid ethane near its melting point (−183 °C) using an FEI Vitrobot to regulate the humidity and ethane immersion process. The resulting samples were stored at −196 °C before imaging using either a JEOL 1210 (120 keV) or an FEI Tecnai G2 Spirit BioTwin transmission electron microscope, at or near −178 °C.

Dynamic Light Scattering

Samples for dynamic light scattering (DLS) were prepared by filtering the solutions through 0.45 μm hydrophilic filters into dust-free 0.25 in. diameter glass tubes. DLS measurements were recorded using a Lexel model 75 Ar+ ion laser (wavelength, λ = 488 nm) and a home-built goniometer in a temperature-controlled silicon oil bath. The scattered light intensity was detected using a Brookhaven BI-DS photomultiplier and was processed using a digital correlator (BI-9000AT). Normalized intensity autocorrelation functions, g2(τ), were measured at six different scattering angles, at 15° intervals from 45° to 120° at 25 °C. The normalized electric field correlation functions, g1(τ), were obtained from the measured intensity correlation functions using the Siegert relation g2(τ) = 1 + β|g1(τ)|2, where 0 ≤ β ≤ 1 is the coherence factor. These functions were analyzed using both cumulant and double-exponential fitting procedures. Cumulant fitting gives information on the micelle size distribution (mean size and width) for a monomodal distribution.[47] The following cumulant expansion was used to fit the dataHere, Γ is the mean decay rate, and μ2 and μ3 are the second and third cumulants, respectively. The ratio μ2/Γ2 is a measure of the relative width of the size distribution. For double-exponential fits, which are utilized for a system exhibiting two distinct decay modes, the following function was usedwhere A1 and A2 represent the relative contributions to the scattered field of the decay rates, Γ1 and Γ2. Using either eq or 2, Γ was determined for different wavevector values (q = 4πnλ–1 sin(θ/2), where n is the refractive index of the solvent and θ is the scattering angle). The mean diffusion coefficient D was determined by a linear fit of the mean decay rate Γ versus q2 with an imposed zero intercept, where Γ = Dq2. The mean hydrodynamic radius, Rh, was determined using the Stokes–Einstein equation (Rh = kBT/6πηD, where kB, T, and η are the Boltzmann constant, absolute temperature, and solvent viscosity, respectively). Inverse Laplace transformations of the electric field correlation functions were also performed using the constrained regularization program REPES to obtain the distribution of relaxation times, which reflects the size distribution of the micelles.[48]

Results and Discussion

Here, we report our findings on two samples of μ-A(BC)miktobrush terpolymers: μ-O(C2F2) and μ-O(C3F2), where O represents the PEO block, C represents the PMCL block, and F represents the PFPO block.[46]Figure shows the chemical structure of the μ-O(CF) polymers; molecular characterization data are shown in Table .

Figure 1

Chemical structure of μ-O(CF) terpolymers, where the blue chain is the PEO, “O” block; the red chain is the PMCL, “C” block; and the green chain is the PFPO, “F” block.

Table 1

Composition and Self-Assembly Data of μ-O(CF) Terpolymers

terpolymer	PMCL, n	PFPO, n	fPEOa	fPMCLa	fPFPOa	Mn (kg mol –1)a	solution morphologyb
μ-O(C2F2)	1.9	1.9	0.63	0.28	0.09	36	HM and RM
μ-O(C3F2)	2.6	1.8	0.57	0.35	0.08	40	MLP

As determined using elemental analysis.

As determined using cryo-TEM. HM = hamburger-like micelles, RM = raspberry-like micelles, and MLP = multilamellar polymersomes.

The thin-film hydration technique was utilized to induce the self-assembly of the terpolymers in an aqueous solution. μ-O(C2F2) and μ-O(C3F2) terpolymers were dissolved in dichloromethane at a concentration of 10 mg/mL, and the solvent was removed under reduced pressure to form a thin film of the terpolymer on the walls of a glass vessel. Filtered (0.45 μm filter) deionized water was added to form a dilute (∼1 wt %) sample, and the mixtures were stirred at room temperature for extended time periods, as noted below. The solutions were then filtered (0.45 μm filter) once more to remove any large particles before DLS or electron microscopy measurements. The analysis of the μ-O(C2F2) solution using DLS after 1 week of gentle stirring at room temperature showed self-assembled aggregates with a mean hydrodynamic radius, Rh, of 40 nm with a relatively low dispersity (μ/Γ2 = 0.15) (see week-1 in Figure , top).

Figure 2

DLS size distributions for 1 wt % aqueous solutions of μ-O(C2F2) terpolymer (top) and μ-O(C3F2) terpolymer (bottom). Data shown were collected at a scattering angle of 90° and at a temperature of 25 °C.

The μ-O(C2F2) aggregates showed a broadening of their size distribution over 10 weeks and an increase in mean size, which is consistent with some evolution of the morphology.[49] As can be seen in Figure , from the cryo-TEM analysis of μ-O(C2F2) after 10 weeks, predominantly “hamburger” micelles were observed. In this geometry, the F blocks form the hamburger itself, the C blocks form the “buns”, and the O blocks stabilize the assembly. This structure minimizes the contact between the water and the fluoropolymer, the pair with the highest interfacial energy. The hamburger micelles typically showed clearly discernible central F core domains, which appear darker in cryo-TEM because of the higher electron density of PFPO relative to that of PMCL. The PFPO domains varied in both size and distribution, likely owing in some measure to the variability in the number of F and C chains in the miktobrush molecules and their respective chain length disparities. A density mapping procedure (Figure b) was applied across the core of an atypically large hamburger micelle (shown in Figure c, chosen for clarity) oriented roughly perpendicular to the plane of the grid. The intensity map shows a distribution of electron density consistent with a hamburger-like, alternating arrangement of C and F domains.

Figure 3

(a) Cryo-TEM image of a vitrified 1 wt % aqueous solution of the μ-O(C2F2) terpolymer, 10 weeks after preparation; (b) cryo-TEM image of the hamburger micelle selected for density mapping; and (c) density map across the center of an enlarged image of a micelle. Inset shows the structural cartoon of the hamburger micelle with a possible arrangement of the μ-O(C2F2) terpolymer.

For the hamburger micelles, the average PMCL and PFPO domain sizes were 30 ± 5 and 10 ± 5 nm, respectively (by measuring at least 50 micelles). The average C domain size is close to the fully stretched length (FSL) of the 5.3 kDa PMCL chain, which is estimated to be 32 nm. This result is consistent with the so-called superstrong segregation limit (SSSL),[50,51] whereby the entropic penalty of fully stretching a C chain from the F domain is overwhelmed by the enthalpic benefit of minimizing their interfacial interaction. The relatively flat nature of the F domain surfaces is also characteristic of this competition and has been observed previously for highly incompatible core domains in earlier studies of linear ABC and μ-ABC triblock terpolymers.[40]

The FSL of the 2.5 kDa F chain is 5 nm, and therefore the measured average domain size for the hamburger F domains of 10 nm using TEM strongly suggests a bilayer structure. As both C and F domain sizes resemble the FSL of the respective chains, we infer that the thermodynamic incompatibility of the F and C cores overrides any potential constraints imposed by the new miktobrush architecture for the μ-O(C2F2) terpolymer. Given the relatively few numbers of blocks (i.e., one O, two C, and two F chains), this is reasonable because the chains could behave comparably to the analogous miktoarm star polymers previously explored.

Upon imaging the same sample after 25 weeks, the majority of the structures resembled raspberry micelles, with distributions of discrete F domains (darker in cryo-TEM) interspersed within C domains (lighter in cryo-TEM), as shown in Figure . Some hamburger micelles were also observed, albeit with lower frequency. The overall size distribution using DLS did not change appreciably over this time interval, suggesting that the structure evolved more by intraparticle rearrangements than by the fission/fusion of distinct micelles.[49] The average size of the F domains is 8 ± 2 nm (by measuring the F domains of at least 50 micelles, including those in Figure ). This fits a similar model as that observed for the hamburger micelles, that is, fully stretched, back-to-back F chains within a PMCL matrix.

Figure 4

(a) Cryo-TEM image of a vitrified 1 wt % solution of the μ-O(C2F2) terpolymer, 25 weeks after self-assembly in water by thin-film hydration; (b) enlarged cryo-TEM image; and (c) structural model of a raspberry micelle and possible arrangement of terpolymer.

We next describe the aqueous self-assembly behavior of the μ-O(C3F2) terpolymer, with a slightly lower PEO volume fraction (fPEO = 0.57) and slightly less PFPO relative to PMCL (Table ). DLS analysis indicated that this polymer initially associated to form larger aggregates with an average hydrodynamic radius, Rh = 185 nm, and higher dispersity, μ/Γ2 = 0.28 (Figure , bottom). A double exponential fit of the autocorrelation function was consistent with two distinct populations of aggregates in solution, with one centered at Rh ≈ 75 nm and one larger population centered at Rh ≈ 250 nm. Cryo-TEM images were consistent with these DLS measurements, given the inherent differences between the two techniques (Figure ). Remarkably, the predominant morphology is that of a polymersome or polymeric vesicle. These images contain some multilamellar polymersomes (MLP), and in all cases, the bilayers appear to be compartmentalized (Figure ).

Figure 5

(a) Cryo-TEM image of a vitrified 1 wt % aqueous solution of the μ-O(C3F2) terpolymer, 10 weeks after self-assembly by thin-film hydration; (b and c) enlarged cryo-TEM images of multicompartment multilamellar polymersomes.

Figure 6

(a) Enlarged cryo-TEM image of a vitrified 1 wt % aqueous solution of the μ-O(C3F2) terpolymer, after 10 weeks; (b) density mapping across the perimeter of a polymersome reveals periodic density variations; (c) possible structural model of multicompartment polymersomes.

The FSL of the μ-O(C3F2) alternating styrene and maleimide backbone (including the RAFT CTA end group) can be estimated to be 3.3 nm (assuming a planar zigzag conformation, where the C–C single bond length is 0.154 nm, the C–C–C bond angle is 109.5°, the C–S single bond length is 0.180 nm, the C–C–S bond angle is 110°, and the C–S single bond length in S–C(S)–S is 0.152 nm, with a bond angle of 125°). These calculations, along with the thickness of 6 nm estimated from the cryo-TEM images, are consistent with a bilayer comprising two back-to-back μ-O(C3F2) terpolymers with hydrated O chains residing on either side, protecting the hydrophobic bilayer from the surrounding water.

To investigate the distribution of C and F domains in the bilayer, a density mapping procedure was performed on cryo-TEM images of a polymersome bilayer, as illustrated in Figure a,b. The resulting intensity plot highlights periodic increases in electron density separated center-to-center by 12 ± 2 nm. These periodic increases in intensity correlate with the increases in electron density within the structural domains, indicating the presence of F domains distributed periodically throughout the bilayer. Because the F domains are separated by 12 ± 2 nm and given that the FSL of the C chains is approximately 32 nm, the C domains are likely forming a continuous matrix around the isolated F domains, again shielding most of the F domains from the solvated O chains on the interior and exterior of the polymersome. Because of the connectivity of the blocks, some F chains must extend to the surface of the vesicle membrane, suggesting that the F blocks form channels that traverse the bilayer. However, it is important to note that it is not possible to infer the spatial arrangement of the F domains in the plane of the membrane because of the projection of the cryo-TEM image. In a previous report with miktoarm stars, a nearly hexagonal arrangement of F domains within the plane of the membrane was observed.[42]

Summary

We have examined the aqueous self-assembly of μ-O(CF)miktobrush terpolymers using DLS and cryo-TEM. The first terpolymer investigated, μ-O(C2F2) (fPEO = 0.63, fPMCL = 0.28, fPFPO = 0.09), at long incubation times forms hamburger and then raspberry-like micelles, likely owing to the larger volume fraction of PEO. Within the hamburger micelles, the PMCL chains form “buns” around an oblate PFPO disk core, thereby decreasing the interfacial penalty between the PFPO domain and the solvated PEO blocks. Over time, this morphology evolves to be more predominantly raspberry-like micelles with more dispersed F domains shielded from solvated PEO by PMCL chains. These findings are broadly consistent with previous research on μ-EOF and μ-EOC miktoarm star terpolymer systems, where raspberry-like multicompartment micelles were observed for μ-EOF (2-9-5) (fPEO = 0.59, fPEE = 0.17, fPFPO = 0.22). Although the PEO volume fractions of the respective blocks are similar, the molar masses and architectures differ. However, it appears that the miktobrush architecture does not significantly impact the observed morphology in this case, which can be attributed to the low number of chains that converge to form the terpolymer. Moreover, the volume fractions and interaction between the constituent polymers and the surrounding water exert the dominant influence on the solution structures. The propensity for the terpolymers studied to form multicompartment micelles with domains consisting of fully stretched blocks and flatter interfaces also implies that the superstrong segregation regime governs the microphase separation behavior of the B and C core domains, as in previous miktoarm star terpolymer systems containing F blocks.

The solution structures adopted by the μ-O(C3F2) terpolymer (fPEO = 0.57, fPMCL = 0.35, fPFPO = 0.08) are multilamellar vesicles or polymersomes with nanoscopic periodicity within the bilayer, attributed to dispersed domains of PFPO within a matrix of PMCL. Again looking at a comparison from a previous study on miktoarm star systems, μ-EOF (2-4-2.5) (fPEO = 0.48, fPEE = 0.32, fPFPO = 0.20) formed nanostructured bilayer sheets and vesicles with cylindrical F domains distributed approximately hexagonally in a continuous PEE matrix. The miktobrush architecture does influence the self-assembled structures in this case. Such multicompartmentalized nanostructures could have potential as drug delivery vehicles, nanoreactors, and protocells with tunable membrane permeability and double delivery capability.