Role of highly branched, high molecular weight polymer structures in directing uniform polymer particle formation during nanoprecipitation

The new macromolecular architecture, hyperbranched polydendrons, are composed of a broad distribution of molecular weights and architectural variation; however, nanoprecipitation of these materials yields highly uniform, dendron-functional nanoparticles. By isolating different fractions of the diverse samples, the key role of the most highly branched structures in directing nucleation and growth has been explored and determined.

Polymeric nanostructures have been formed using various techniques including direct synthesis, such as micro/mini-emulsion polymerisation 1,2 and polymerisation induced self-assembly, [3][4][5][6] or the post-synthesis self-assembly of polymers of varying architectures, hydrophilicity or hydrophobicity. [7][8][9] Alternatively, physical fabrication methods such as the ''Particle Replication In Non-wetting Templates'' (PRINTt) approach 10 may be employed. In many self-assembly routes, a ''solvent switch'' is used to modify the media environment and drive inter-chain interactions. This can be achieved by changes in pH, 11 addition of salt, 12 removal of a co-solvent or the slow exchange of good solvent with an antisolvent 13,14 (often water); solution and emulsion approaches have been reported.
A rapid and scale-able solvent switch technique that has gained considerable interest is nanoprecipitation, 15 especially in the formation of degradable nanoparticles for biomedical and nanomedicine applications. In general, polymers are dissolved in a good water-miscible solvent and rapidly added to an aqueous anti-solvent. Descriptions of both spontaneous, energybarrier-free nucleation (spinodal decomposition) 16 and nucleationand-growth mechanisms have been reported. In the latter case, providing the polymer is above its thermodynamic solubility limit within the resulting mixed solution, homogeneous nucleation of particles will occur. 16 The particle radii will initially be small, with a free energy of formation that is governed by parameters such as surface tension, particle radius and the free energy difference between media and particle. The free energy will pass through a maximum at a critical nucleus radius leading to rapid growth of larger particles. In either mechanism, growth will continue by depletion of polymer from the surrounding environment or through a process known as diffusion-limited cluster-cluster aggregation, assuming the balance of attractive and repulsive forces will allow interaction and steric/charge stabilisation does not prevent particles from achieving sufficient proximity for effective collision. 15 To ensure rapid polymer nucleation, several groups utilise microfluidic or rapid-mixing techniques that allow distinct separation of the nucleation and growth/aggregation stages. 17,18 Studies of the impact of polymer architecture and molecular weight on nanoprecipitation outcomes are very rare. 15 We have reported a new complex polymer architecture -''hyperbranched polydendrons'' [19][20][21] -that overcomes the complexity of dendrimer synthesis by branching between linear-dendritic hybrid polymer chains during propagation (Scheme 1); weight average molecular weights 410 6 g mol À1 , representing 490 conjoined primary linear-dendritic chains, have been achieved. Hyperbranched polydendrons contain considerable variation in: (1) the number of branched chains ( Fig. 1), (2) the position of branched points ( Fig. 1) and, (3) molecular weight; a significant concentration of un-branched primary chains (linear-dendritic hybrid polymers) is also present. Despite this diversity, the complex mixtures nanoprecipitate to form stable, highly uniform nanoparticle dispersions. Each particle is surrounded by dendron functionality and outcomes are tunable to allow the multi-valency benefits of dendrimers to be achieved at sizes unobtainable by dendrimer synthesis alone (50-200 nm). Hyperbranched polydendrons mimic our studies of branched polymers without dendron chainends, 22,23 forming more uniform nanoprecipitates with increased stability and reduced size compared to particles generated from their chemically-identical linear primary chain building blocks. As such, they are ideal candidates for detailed examination of the role of the very highly branched, very high molecular weight fraction that is present within the molecular weight distribution.
Here we isolate the very high molecular weight fraction of hydrophobic hyperbranched polydendrons and study its role in directing successful nanoprecipitation to allow a greater mechanistic understanding and offer more general opportunities for enhanced nanoprecipitation processes.
The polymers were synthesised using previously reported approaches, 19-21 namely copper-catalysed methanolic atom transfer radical polymerisation at 30 1C, initiated using a new generation 2 (G 2 ) dendron initiator, 1, and a monomer mixture of 2-hydroxypropyl methacrylate, 2 (HPMA), and ethylene glycol dimethacrylate, 3 (EGDMA) in a 1 : 2 : 3 molar ratio of 1 : 50 : 0.8 (target number average degree of polymerisation for each primary chain of 50 monomer units). This ratio also ensures o1 brancher per primary chain and avoids gelation. The number average molecular weight (M n ) and weight average molecular weight (M w ) of the resulting branched polymer (G 2 -p(HPMA 50 -co-EGDMA 0.8 )) were determined by triple detection size exclusion chromatography (DMF/0.01 M LiBr eluent; SEC) to be 115 700 g mol À1 and 1 538 000 g mol À1 respectively (dispersity (Ð) = 13.3; Fig. 1). The equivalent linear polymerisation in the absence of 3 yielded a polymer (G 2 -p(HPMA 50 )) of M n = 12 300 g mol À1 and M w = 15 500 g mol À1 (Ð = 1.26). Branching with 3 does not impact the formation of primary chains; 24 therefore, it is correct to consider the weight average structures within the G 2 -p(HPMA 50 -co-EGDMA 0.8 ) as comprising approximately 100 conjoined G 2 -p(HPMA 50 ) chains. This is highly important as M w represents the molecular weight of macromolecules at the mean of the weight fraction and, therefore, approximately half of the physical mass of the sample comprises branched structures with a minimum of 100 conjoined chains; the M n value indicates a large number of linear and lesser branched structures, as expected. G 2 -p(HPMA 50 -co-EGDMA 0.8 ) was fractionated using alcoholic dialysis (3 days) with either methanol or 2-propanol, and tubing with two nominal molecular weight cut-off values (MWCO = 100 kg mol À1 and 300 kg mol À1 ). Two good solvents for p(HPMA) were studied due to the potential for variable wetting and pore swelling within the tubing; outer solutions were collected and fully replenished each day. Solutions within the dialysis tubing, and the combined solvent fractions collected from the reservoir, were dried and the recovered polymer studied by triple detection SEC ( Fig. 1 and Table 1).
The different MWCO dialysis tubing had a minimal impact on the separation, presumably due to non-aqueous solvent use; Scheme 1 Cu-catalysed ATRP synthesis of hyp-polydendrons using HPMA, 2 (both isomers shown), EGDMA, 3, and a G 2 dendritic initiator, 1.  (Table 1). Methanolic dialysis allowed enhanced isolation of the high molecular weight fraction with the lowest dispersity values.
Nanoprecipitation of the eight fractions (Table 1) and the linear-dendritic hybrid G 2 -p(HPMA 50 ) was conducted from THF solutions (5 mg mL À1 ) using deionised water as the anti-solvent and targeting final nanoparticle concentrations of 1 mg mL À1 after THF removal. Dynamic light scattering analysis (DLS; Table 2) of the unfiltered samples yielded z-average diameter (D z ) values of 620-770 nm for nanoprecipitates of G 2 -p(HPMA 50 ) and the least branched materials fractionated using 2-propanol; similar methanol dialysis fractions generated significant macrophase separated material. When the high molecular weight fractions isolated using 2-propanol or methanol were nanoprecipitated under identical conditions, considerably more uniform size distributions were observed with D z values ranging from 135-205 nm and polydispersity (PdI) values ranging from 0.068-0.095 (Table 2).
To investigate the role of the very high molecular weight species in directing the nanoprecipitation process of highly disperse samples, the high molecular weight fraction of G 2 -p(HPMA 50co-EGDMA 0.8 ) collected using the 100 kg mol À1 MWCO tubing and methanolic dialysis (G 2 -p(HPMA 50 -co-EGDMA 0.8 ) MeOH ) was systematically added at varying weight ratios to the G 2 -p(HPMA 50 ) sample prior to nanoprecipitation. This fraction was selected as it possessed the highest M n and M w values and the lowest Ð value of the high molecular weight fractions, thereby allowing the most extreme case of molecular weight and architectural diversity after mixing with G 2 -p(HPMA 50 ). Nanoprecipitation of a G 2 -p(HPMA 50 )/ G 2 -p(HPMA 50 -co-EGDMA 0.8 ) MeOH mixture containing just 1 wt% branched material generated significantly lower D z and PdI values than those seen for G 2 -p(HPMA 50 ) alone; 198 nm (PdI = 0.034) versus 670 nm (PdI = 0.215) respectively ( Fig. 2A).
Further addition of G 2 -p(HPMA 50 -co-EGDMA 0.8 ) MeOH led to a range of nanoparticles with steadily decreasing D z values (PdI = 0.034-0.110), with the lowest value (145 nm) achieved from nanoprecipitation in the absence of linear G 2 -p(HPMA 50 ). The impact of highly branched, high molecular weight material within the nanoprecipitation of linear G 2 -p(HPMA 50 ) can be rationalised by consideration of the nucleation process in the mixed solvent/anti-solvent environment. In the absence of branched polymer, the diffusion of good solvent will lead to the media becoming increasingly less able to solvate the linear-dendritic hybrid polymer chains; these will collapse to form small particles comprising single or small numbers of collapsed macromolecules. It is unlikely that these small nuclei will exceed the critical radius (r*), and slow assembly of chains will lead to poorly defined nucleation and growth stages (Fig. 2B); clearly distinct nucleation and growth are key to the formation of narrow size distributions. 15 Conversely, the population of unimolecular structures comprising 4100 conjoined chains will lead to the rapid formation of considerably larger nuclei, many of which would be expected to exceed r* and allow a distinct nucleation stage; the lineardendritic polymer would require a considerable number of single chains to coalesce in a rapid and concerted multi-body collision to achieve the same outcome. We recently reported a lack of discernible light scattering immediately after addition of linear polymer solutions to anti-solvents, in contrast to the observed strong scattering from branched polymers under identical conditions; this would support the absence of rapid nucleation in lower molecular weight linear materials. 23 Fractionation to generate G 2 -p(HPMA 50 -co-EGDMA 0.8 ) MeOH also increased the number of conjoined chains within the weight average branched structure to 4150 (number average structures contain 460 chains) ( Table 1).
Decreasing D z values with increasing content of G 2 -p(HPMA 50co-EGDMA 0.8 ) MeOH (Fig. 2A) supports a nucleation-driven mechanism as increased nucleation density from additional highly branched polymer will lead to a larger number of smaller nanoparticles at constant polymer mass. 25 The number of nanoparticles formed can be approximated by assuming the polymer density and utilising the number average diameter (D n ) values (Table S1, ESI †), and it is important to note that the number of chains and chain ends does not vary throughout this study. Previous studies have shown that increasing polymer molecular weight alone will lead to little difference in D z values within nanoprecipitations of linear polymer systems, despite considerable variation in solution viscosity,  with increased particle sizes resulting for some polymers. [26][27][28] The effects seen here are, therefore, clearly driven by the number of conjoined chains and the architecture of the high molecular weight fraction. The ability to direct nanoprecipitation outcomes by introducing small fractions of highly branched polymers may have considerable value in ensuring consistency in medical products, decreasing complexity of mixing processes and in controlling nanoparticle sizes and polydispersities. To achieve such results with as little as 1 wt% of branched polymer opens significant opportunities, offers simplification within large scale processes and provides mechanistic insight.
The authors gratefully acknowledge financial support from the University of Liverpool and the Engineering and Physical Sciences Research Council for a PhD studentship (FL) and grant funding (EP/I038721/1) that underpinned the research.