Polymers on drug-rich nanodroplets: why more polymer on the droplets did not mean better crystallization inhibition for enzalutamide

Background and motivation

Amorphous solid dispersions (ASDs) raise the oral exposure of poorly water-soluble drugs by generating supersaturated solutions on dissolution. In an ASD the polymer has several jobs: it keeps the amorphous drug from crystallizing in the solid, it helps release, and for many drugs it delays crystallization from the supersaturated solution once the drug is released, which matters for drugs that crystallize within the time they spend in the gut C1. If an ASD releases drug beyond the amorphous solubility, the maximum concentration of molecularly dissolved drug, the excess separates as amorphous drug-rich nanodroplets, by liquid–liquid phase separation (LLPS) when the droplets are above their glass transition temperature ($T_g$) or by glass–liquid phase separation (GLPS) when they are glassy C2. Earlier work from the same group and others showed that such droplets act as a depot: as drug crosses the membrane, the droplets dissolve and keep the solution at its maximum supersaturation, which sustains flux; droplets have been seen after ASD dissolution, after a pH shift, after dissolution of a salt, on dilution of an organic drug solution, and in aspirated human intestinal fluid (Indulkar 2016; Wilson 2018; refs 4–8, 47–50, 53–56) C2.

The depot only helps if the drug does not crystallize. In a solution that contains droplets, the drug has the same chemical potential in the droplets and in the surrounding water, and this chemical potential is higher than that of the crystal (Taylor and Zhang 2016) C3. Crystallization can therefore start in the water, inside the droplets, or at their interface, and adding more drug beyond the amorphous solubility does not raise the supersaturation further; it only adds droplets C3. The polymer has to suppress crystallization in all of these places, which raises the question of where the polymer goes.

Three earlier studies framed that question. Raina et al. (2015) found that chemically diverse polymers distribute between the aqueous phase and drug-rich droplets according to their hydrophobicity, hydrophilic polymers staying in the water, amphiphilic ones distributing between both phases and hydrophobic ones moving into the droplets; the good inhibitors were the ones found in both phases, which led to the suggestion that distributing between both phases is a requirement C4. Ueda et al. (2013), with a set of related HPMC derivatives, linked more polymer on the droplets to stronger inhibition, and Wang et al. (2018) found that the amount of HPMCAS coprecipitating with posaconazole correlated with how long supersaturation lasted C5. The working idea in the field was thus that more polymer associated with the droplets should mean better inhibition. For enzalutamide, a poorly soluble anti-androgen, HPMCAS was already known to be an effective crystallization inhibitor (Wilson 2020) C6.

This paper tests that idea with polymers of opposite charge. The authors measure nucleation induction times of enzalutamide at concentrations below and above its amorphous solubility, so that the amount of droplets changes, with three polymers at several concentrations. For two of them, the anionic HPMCAS and the cationic Eudragit E PO, they then measure how much polymer goes into the droplets, how the droplets look, how they are charged and how large they are, and they use molecular dynamics to look at how each polymer meets a small cluster of drug molecules A1·A2·A3·A5.

Background concepts

Amorphous solubility and droplet amount. The amorphous solubility is the highest concentration of drug that can stay molecularly dissolved; beyond it the excess forms a separate drug-rich phase. For enzalutamide under the conditions of this paper (pH 6.5, 50 mM phosphate buffer) the authors state a value of 42 μg/mL, which they also call the GLPS concentration, without giving how it was determined or citing a source M1·X11. They use it to convert added concentrations into droplet amounts: 45, 70 and 120 μg/mL of added drug correspond to 42 μg/mL dissolved plus 3, 28 and 78 μg/mL of drug in droplets, while 30, 35 and 40 μg/mL contain no droplets A1. Because the droplets are described as glassy, the paper speaks of GLPS rather than LLPS C2.

Nucleation induction time. The induction time is the time from creating the supersaturated solution until crystals are detected; it includes the nucleation time and the time a nucleus needs to grow to a detectable size P3. Here it is read from UV spectra as the moment the absorbance at 237 nm, where enzalutamide absorbs, starts to fall while the signal at 446 nm, where the drug does not absorb, rises from light scattering by crystals; Figure S2-1 shows the 237 nm band of a 70 μg/mL solution without polymer dropping from about 0.94 to 0.27 within 12 min P3·R2. Experiments were stopped at 12 h, so "no crystallization" means an induction time above 720 min R3.

Where crystals can form. Nucleation on a surface or interface, heterogeneous nucleation, usually needs to overcome a lower energy barrier than nucleation in bulk, and molecules adsorbed at an interface can either speed up or slow down nucleation there C44. This is the background to the authors' later argument that the droplet–water interface is a likely nucleation site and that what a polymer does at that interface matters C45.

Zeta potential and colloidal stability. The zeta potential is the electrical potential at the slipping plane of a particle in liquid. The paper cites the common rule that colloids are stabilized by electrostatic repulsion when the zeta potential is below −30 mV or above +30 mV, and that between these limits a dispersion tends to ripen; colloids can also be stabilized sterically by adsorbed polymer C27.

A polarity-sensitive label. The fluorescent dansyl group emits more strongly and at a shorter wavelength when its surroundings are less polar. Attached to HPMCAS, it reports whether part of the polymer sits in a less polar environment such as a drug-rich droplet C31·P1. The degree of labelling was obtained from 1H NMR integrals of the labelled polymer. The first relation gives the molar substitution of hydroxypropyl units, $n$, from the ratio of the backbone integral to the hydroxypropyl methyl integral, using the manufacturer's average degrees of substitution (DS) E1·M2:

In plain terms: the backbone region of the spectrum collects protons from the glucose ring and from the methoxy, hydroxypropyl, acetyl and succinyl groups, while the hydroxypropyl methyl peak counts only the hydroxypropyl chains; their ratio, here 4254.09/160.50 = 26.5, fixes how many propylene oxide units hang on each hydroxypropyl site R23. With the manufacturer's DS values (methoxy 0.69, hydroxypropyl 0.21, acetyl 0.27, succinyl 0.34) the paper obtains $n$ = 0.69 R23·M2. The paper writes the hydroxypropyl coefficient as 0.64$n$, whereas 3 × 0.21 gives 0.63$n$; with 0.63 the result would be 0.70, a small slip that does not change the label content X18. The second relation then gives the degree of substitution $x$ of the dansyl-ethylene label E2:

In plain terms: $a$ = 1 is one aromatic dansyl proton, outside the backbone region; $b$ = 11.68 is the number of backbone-region protons per glucose unit without the label; $cx$ = 10$x$ adds the label's own protons that fall in the backbone region; and $d$ = 4254.09 is the backbone integral when the aromatic proton is set to 1 E2·R23. Solving $x/(11.68+10x)=1/4254.09$ gives $x$ ≅ 0.003, a sparse label R23. The labelled polymer inhibited crystallization like the unlabelled one and partitioned between the phases in the same way (data not shown) R18.

Simulation measures. The radius of gyration $R_g$ describes how extended a polymer chain is; the solvent-accessible surface area (SASA) per atom describes how much of the chain is exposed to water; and the radial distribution function (RDF) between polymer and drug atoms gives how likely polymer atoms are found at a given distance from drug atoms P11·R20·R22. The paper also defines a drug–polymer interfacial area as the surface each partner hides from water when they touch, averaged over the trajectory E4:

In plain terms: at each time $t$ the solvent-accessible areas of the polymer alone and of the drug cluster alone are added, the area of the complex is subtracted, and the difference, which is the contact area, is averaged over the length $\Delta t$ of the production run with time step $\delta t$ E4·R22.

Methods in detail

The study has four parts: induction times that show how much harder crystallization is to stop once droplets exist; a measurement of how much HPMCAS and Eudragit E PO go into the droplets; three looks at the droplet surface (electron microscopy, charge and size, a fluorescent label); and a simulation of how each polymer chain meets a cluster of drug molecules. The final step sets the inhibition ranking against the association ranking.

Contribution 1: induction times on both sides of the amorphous solubility

① Problem. Most comparisons of polymeric crystallization inhibitors use supersaturated solutions, but ASDs often dissolve beyond the amorphous solubility. The question is how much the presence of droplets changes a polymer's apparent performance, and whether the ranking of polymers stays the same C13·C14.

② How. A 10 mg/mL methanol stock of enzalutamide was added to 35 mL of 50 mM phosphate buffer at pH 6.5 that already contained the polymer, to give 30, 35 or 40 μg/mL (no droplets) or 45, 70 or 120 μg/mL (droplets) P3·A1. HPMCAS (MF grade, Shin-Etsu) and Eudragit E PO (Evonik) were dissolved directly in the buffer; P177, a cellulose ester with a carboxylic-acid side chain made by the Edgar group, was first dissolved in DMSO, leaving 1 ppm of organic solvent M2·M3·M4·P3. Figure 3 shows polymer concentrations of 2.5, 5, 50 and 100 μg/mL; the methods text prints a different list (30, 35, 40, 45, 70 and 120 μg/mL), which repeats the drug concentrations and does not match the figure X1. Each condition was run in triplicate and followed for 12 h; temperature and stirring of these runs are not stated, as the method is taken from earlier work P3·X13. Without polymer, induction times at the six concentrations were compared by a Kruskal–Wallis test, and all data were fitted by a regression that estimates the probability that crystallization is inhibited, counting "no crystals after 720 min" as inhibited, with drug concentration, above/below GLPS, polymer type and polymer concentration as factors P4.

③ What it can show. Pairs of conditions just below and just above 42 μg/mL (40 and 45 μg/mL) isolate the effect of a small amount of droplets (about 3 μg/mL) at almost the same total drug, while 70 and 120 μg/mL test larger droplet amounts A1·R3. The authors use 180 min as a benchmark for a biologically relevant time frame, based on mean gastric residence of 1 h and small intestinal transit of 3 h (ref 61); for enteric polymers such as HPMCAS only intestinal transit is "likely" relevant, while for a polymer soluble at low pH such as Eudragit E PO gastric residence also counts R24·C47.

④ Difference from earlier work. Earlier inhibitor screens mostly measured induction times without droplets; here the same polymers are compared on both sides of the amorphous solubility, and the authors recommend that drug and polymer concentrations "should be carefully" chosen to resemble those of a dissolving ASD C14.

Contribution 2: how much polymer goes into the droplets

① Problem. Following Ueda et al. and Wang et al., the authors expected the weak performance of Eudragit E PO with droplets to come from too little polymer on the droplets ("might arise") C15. Testing this needs the amount of each polymer in the droplets.

② How (NMR). The method follows Ueda et al. (2020) P6. Buffer with 10% D₂O and a trimethylsilylpropanoic acid reference was pre-warmed to 37 °C and stirred at 1000 rpm; a 10 mg/mL enzalutamide stock in DMSO-d₆ was spiked in, 36 μL into 3 mL for 120 μg/mL (HPMCAS systems) or 21 μL into 3 mL for 70 μg/mL (Eudragit E PO systems) P6. After 15 min or 1 h 15 min of stirring, 600 μL samples were measured on an 800 MHz spectrometer with water suppression; polymer in solution was quantified against standard curves from 10 to 200 μg/mL after deconvolution of overlapping DMSO and drug peaks, in triplicate P6. Solution NMR sees only dissolved species, because molecules in the glassy droplets are too immobile to give resolved signals, so the polymer missing from the spectrum is assigned to the droplets C22. The paper states this relation in words E3:

In plain terms: whatever polymer is added but no longer found dissolved is taken to sit in or on the droplets E3. The same spectra give the enzalutamide still dissolved, from the loss of the drug signal relative to a polymer-free reference P6·C22. Table 2 reports the result as "w/w % polymer/ENZ nanodroplets"; the formula for this percentage is not printed, but the tabulated values are reproduced by dividing the polymer in the droplets by the sum of polymer and drug in the droplets, with the drug in the droplets taken as the total drug minus the drug in solution of Figure 5 (for example 32.6/(32.6 + 120 − 20.5) = 24.7% for 100 μg/mL HPMCAS) E3·X4. The NMR spectra report the enzalutamide peak used for quantification at 2.69 ppm, the same position the paragraph gives for residual DMSO, and elsewhere place enzalutamide at 2.98 ppm, so the exact peak is not clear from the text X10.

② How (colorimetry). As a check, droplets were pelleted by centrifugation (printed as "21.1 g", 40 min, 37 °C) and polymer was assayed in pellet and supernatant: HPMCAS with phenol and sulfuric acid at 490 nm, Eudragit E PO as a complex with acid orange V extracted into chloroform at 480 nm P7.

③ What it can show and ④ difference. Unlike a correlation across polymers that differ in many ways, this measurement compares two polymers that both carry hydrophobic groups but opposite charges C52. Two features of the design limit the comparison: the Eudragit E PO systems contain 70 μg/mL drug and the HPMCAS systems 120 μg/mL, so the droplet mass differs, and the Eudragit E PO samples were read after 15 min while most HPMCAS samples were read after 1.25 h R7·C20.

Contribution 3: looking at the droplet surface

Electron microscopy. Samples taken before and after the induction time were allowed to run down a tilted carbon-coated grid, a method from the group's earlier work, and imaged in bright field at 200 keV; EDX located fluorine, which enzalutamide carries (four atoms) and the polymers do not, and fast Fourier transforms (FFT) of image regions distinguished amorphous droplets (diffuse halo) from crystals (spots) P8·M1.

Charge and size. Drug at 45, 70 or 120 μg/mL was added to buffer with 0–100 μg/mL polymer, stirred at 300 rpm and 37 °C for 5 min, and measured by DLS and zeta potential; the viscosity of each polymer solution was measured and used in the size calculation, and every sample had a polydispersity index below 0.5 P5.

Fluorescent label. Dansyl-labelled HPMCAS at 50 and 100 μg/mL was mixed with 20–120 μg/mL enzalutamide by brief vortexing, and emission spectra were recorded with excitation at 380 nm P9. Spectra in buffer and in dichloromethane first checked that the attached label still reports polarity R16.

Contribution 4: one chain on a drug cluster

How. A 12-unit Eudragit E PO chain (four repeats of its three monomer units) and a 10-monomer HPMCAS-MF chain with the substituent pattern of Table 1, ionized at its three succinate groups with three Na⁺ counterions, were built, minimized and equilibrated in SPC/E water at 310 K and 1 bar with GROMACS and the CHARMM force field P10. Ten enzalutamide molecules, taken from the crystal structure, were first aggregated in the gas phase; the cluster was then placed 1–2 nm from each equilibrated chain, solvated, equilibrated (5 ns NVT, 1 ns NPT) and run for 15 ns P11. From the trajectories the authors extracted snapshots, polymer–drug RDFs, an estimated free energy of interaction from the Lennard-Jones and Coulomb terms, and the $R_g$, SASA and interfacial-area measures defined above P11·E4.

Note by the AI writer (not discussed in the paper). The paper does not state a protonation state or counterions for the Eudragit E PO chain, although the discussion describes its amino groups as ionized at pH 6.5 and the measured zeta potentials are positive; the commercial monomer ratio is not given either, and each polymer is represented by one short chain and a 10-molecule cluster over 15 ns X14.

Experimental analysis

Main result 1: without polymer, droplets make no difference. Enzalutamide alone crystallized within about 5–15 min at every concentration from 30 to 120 μg/mL (Figure 2) R1. The Kruskal–Wallis test gave chi-square 0.4386 and p = 0.5078 for droplet-free against droplet-containing solutions; the main text reads this as no significant difference, which the p value supports, while the caption of Table S3-1 says the null hypothesis was rejected R5·C7·X2. Without polymer, then, crystallization is already fast enough that the droplets cannot shorten it measurably.

Table A. Induction times in minutes, mean of n = 3, read from Figures 2 and 3 (approximate). Columns 30–40 μg/mL contain no droplets; 45, 70 and 120 μg/mL contain about 3, 28 and 78 μg/mL of drug in droplets. NC = no crystallization by the 12 h end (720 min); — = not tested R1·R3.

Main result 2: droplets make inhibition much harder. With polymer present, crossing 42 μg/mL shortened induction times sharply in several series: HPMCAS at 2.5 μg/mL from about 129 to 43 min, P177 at 2.5 μg/mL from about 453 to 73 min, P177 at 5 μg/mL from no crystallization to about 589 min, and Eudragit E PO at 50 μg/mL from no crystallization to about 91 min, between 40 and 45 μg/mL drug R3.c·R3.d·R3.af·R3.ag·R3.ai·R3.aj·R3.y·R3.z. More droplets needed more polymer: 5 μg/mL HPMCAS kept 45 and 70 μg/mL drug free of crystals for 12 h but held 120 μg/mL for only about 100 min, and 50–100 μg/mL HPMCAS were needed to reach about 431–530 min R3.h·R3.i·R3.j·R3.k·R3.l. In the regression the "below GLPS" term had an estimate of +5.61 (p = 0.0017) R6.h. The authors write that their results "demonstrate" that inhibition is much harder with droplets, that this "supports the conjecture" that droplets favour crystallization, and that the droplets "tend to enhance" the crystallization tendency C46·C13.

Note by the AI writer (not discussed in the paper): the effect is a step at 42 μg/mL rather than a steady trend with droplet amount. Without polymer the times rise slightly from 45 to 120 μg/mL; with 2.5 μg/mL P177 they are longer at 70 than at 45 μg/mL; and with 2.5 or 5 μg/mL Eudragit E PO they barely change between 40 and 45 μg/mL, because they are short on both sides X20.

Main result 3: the polymers rank HPMCAS, then P177, then Eudragit E PO. Table A shows the ranking with droplets. At 120 μg/mL drug only HPMCAS at 50 and 100 μg/mL passed the 180 min benchmark (about 431 and 530 min), while P177 reached about 71–125 min R3.k·R3.l·R3.al·R3.ao·R3.ap. Eudragit E PO was not tested at 120 μg/mL because the drug crystallized rapidly; at 45 and 70 μg/mL it stayed below about 91 min at every concentration except 100 μg/mL with 45 μg/mL drug (about 417 min) R3.m–ac·C30. Without droplets the picture is different: Eudragit E PO at 50 μg/mL prevented crystallization at 30–40 μg/mL for 12 h, and at 5 μg/mL it held 30 μg/mL for about 267 min R3.w·R3.x·R3.y·R3.r. In the regression, against Eudragit E PO as base case, HPMCAS (+7.63, p = 5 × 10⁻⁷) and P177 (+6.78, p = 4 × 10⁻⁶) raised the probability of inhibition, while "no polymer" did not differ from Eudragit E PO (−0.71, p = 0.70) R6.j·R6.l·R6.k. The authors call Eudragit E PO "the least-effective polymer", P177 "particularly effective" without droplets, and HPMCAS more effective than the other polymers C10·C11·C12.

Two statements in the discussion go beyond the plotted data. "Above the GLPS concentration, only HPMCAS is effective" holds at 120 μg/mL drug, but at 45 and 70 μg/mL P177 at 50 μg/mL also prevented crystallization for 12 h, and Eudragit E PO at 100 μg/mL reached about 417 min at 45 μg/mL C48·X8. And HPMCAS at 50 μg/mL is said to be effective without droplets, "as is" Eudragit E PO, although HPMCAS at 50 and 100 μg/mL was tested only with 120 μg/mL drug; its effectiveness without droplets rests on the 2.5 and 5 μg/mL data C39·C12·X9. The regression itself has collinear terms: above/below GLPS is fixed by drug concentration, and two rows of Table S3-2 are identical; its coefficients are best read qualitatively X3.

Main result 4: Eudragit E PO makes up a larger share of the droplets, by NMR. Table B gives the distribution data. By NMR the droplets contained 40.8 and 41.0 w/w% Eudragit E PO at 50 and 100 μg/mL polymer, and 12.7–24.7 w/w% HPMCAS at 25–100 μg/mL, the HPMCAS share rising with concentration R7. The HPMCAS distribution was the same after 15 and 75 min of stirring, which the authors read as equilibrium within 15 min ("suggesting") R7.a·R7.b·C18. A paired t-test found no difference between 50 and 100 μg/mL polymer (p = 0.76) and a difference between Eudragit E PO and HPMCAS (p = 3.4 × 10⁻⁵) R10. The authors find it "apparent" that Eudragit E PO distributed more into the droplets than HPMCAS, noting that "there were less drug-rich nanodroplets present in the Eudragit E PO system" C19·C20.

Table B. Polymer distribution between solution and droplets, from Table 2 (1H NMR) and Table S4-1 (colorimetry); standard deviations in parentheses, n = 3 R7·R9.

Two features of Table B weaken the comparison. First, the Eudragit E PO systems contain 70 μg/mL drug and the HPMCAS systems 120 μg/mL. In μg/mL the polymer leaving solution was similar for the two polymers: subtracting the NMR solution values from the added amounts gives about 27.1 and 32.1 μg/mL for Eudragit E PO and 21.6 and 32.6 μg/mL for HPMCAS at 50 and 100 μg/mL (calculation by the AI writer, using the paper's own depletion method); the higher Eudragit E PO percentage mainly reflects the smaller amount of drug in its droplets X4. Second, the colorimetric assay, which the paper presents as "independent confirmation" that polymer is in the droplets, gives 22.5 and 18.9 w/w% for Eudragit E PO and 19.7 and 20.4 w/w% for HPMCAS at 50 and 100 μg/mL, so it confirms that polymer is present but not that Eudragit E PO is enriched; its supernatant values also differ from the NMR values, and the caption of Table S4-1 describes only the 120 μg/mL HPMCAS case C16·X5.

Main result 5: both polymers lower the drug in solution equally. In the same NMR samples, enzalutamide in solution fell from 42 μg/mL to about 36.5, 29.5 and 20.5 μg/mL with 25, 50 and 100 μg/mL HPMCAS, and to about 30.5 and 23.5 μg/mL with 50 and 100 μg/mL Eudragit E PO (Figure 5) R8. This is the expected effect of a second component mixing into the drug-rich phase, which lowers the drug's activity there and with it the amorphous solubility C21·C23. The authors had considered that a polymer mixing more strongly into the droplets might lower the driving force for crystallization more ("One possible explanation"); because the reduction is about the same for both polymers, they conclude that the thermodynamic driving force is similar ("this suggests") and that other factors separate the polymers C41·C42. The 42 μg/mL bar in Figure 5 is the amorphous solubility, not an NMR value, and its source is not given R8·X11.

Table C. Drug in solution read from Figure 5 (approximate); polymer leaving solution calculated by the AI writer as polymer added minus polymer in solution from Table 2 (1.25 h for HPMCAS 25, 50 and 100; 15 min for Eudragit E PO) R8·R7·X4.

Main result 6: the polymers sit at the droplet surface. In TEM of 120 μg/mL drug with 100 μg/mL HPMCAS (Figure 6a), approximately spherical droplets, given in the text as about 100–200 nm, sat within HPMCAS-rich regions; EDX found 24 ± 6% fluorine in the spherical regions and 4 ± 2% in the surroundings, and the FFT of the droplets showed no order, so they were amorphous (Figure S5-1) R11·C24. After crystallization the drug showed lattice fringes and FFT spots in several orientations, and the authors note that no polymer was visible around the crystallized regions (Figures 6b, 6c) R11. At 500 μg/mL HPMCAS, droplets of 100–150 nm were surrounded by smaller polymer aggregates of about 20–50 nm (Figure 6d) R11. With Eudragit E PO (70 μg/mL drug, 100 μg/mL polymer), the droplets were larger and agglomerated, with polymer at their edges (fluorine 31.1 ± 14% against 6.3 ± 5%), and crystalline agglomerates formed (Figure 7) R12·C25. The authors read the images as polymer associated with the droplets at the drug–water interface ("suggest") C26. The basis of the fluorine percentages is not stated, and both droplet values exceed the fluorine mass fraction of pure enzalutamide, so they are best read as a contrast between regions X12.

The zeta potential (Figure 8) supports the same picture. Without polymer the droplets were negative (about −16, −18 and −9 mV at 45, 70 and 120 μg/mL drug); with HPMCAS they became more negative (about −15.5 to −30 mV); with Eudragit E PO they became positive (about +9.5 to +15 mV) R13. The authors attribute the change to the polymer charge ("could be attributed") C26. Because nearly all values lie within ±30 mV, they conclude that the polymers "most likely" stabilize the droplets against ripening sterically rather than electrostatically; ripening itself was not followed over time C28.

Table D. Zeta potential (mV) and initial hydrodynamic diameter (nm) at 70 μg/mL enzalutamide, read from Figures 8 and 9 (approximate) R13·R14.

Main result 7: HPMCAS keeps the droplets small. Without polymer the droplets grew from about 480 nm at 45 μg/mL to about 720 nm at 70 μg/mL and 1665 nm at 120 μg/mL, which the authors attribute "most likely" to agglomeration of the glassy droplets R14·C29. With HPMCAS the droplets were about 115–350 nm, smallest at 50–100 μg/mL; with Eudragit E PO two populations appeared at most conditions, the larger one reaching about 1500 nm (Figure 9, Table D) R14. The authors conclude that HPMCAS stabilized the droplets better than Eudragit E PO against both crystallization and size enlargement C56. Their text gives different ranges from the plotted bars (HPMCAS 200–400 nm with a smallest size of 80 nm, Eudragit E PO 200–1100 nm, size roughly inverse to Eudragit E PO concentration), and the larger Eudragit E PO sizes do not appear as peaks in the size distributions of Figures S6-3 and S6-4, which peak at about 170–270 nm with a minor population at several micrometres; the figure values are used here X6·X7·R15.

Main result 8: the label sees a less polar environment once droplets form. Dansyl-HPMCAS emitted weakly near 535 nm in buffer and strongly near 500 nm in dichloromethane (Figure 10), so the attached label still reports polarity R16·C32. With enzalutamide (Figure 11), every spectrum had a peak near 437 nm; below 42 μg/mL a broad band near 528–535 nm remained, and above it new peaks appeared at 469 and 487 nm with higher overall intensity, while the 535 nm band became a shoulder R17. The authors read this as two environments for the label, a less polar one at the droplets and a polar aqueous one ("This suggests"), consistent with HPMCAS distributing between both phases, and later say the spectra "confirm" close interaction of HPMCAS with the droplets C33·C51. Note by the AI writer: the new peaks are distinct only from 70 μg/mL drug, not at 45 μg/mL, and the intensity at 400 nm also rises steeply with drug concentration; no spectra of drug dispersions without the label are shown, so scattering or drug emission are not excluded X19.

Main result 9: in simulation HPMCAS lies extended over the drug cluster. Both chains stayed in contact with the 10-molecule cluster during the 15 ns runs, and the polymer–drug RDFs peaked at about the same distance (about 1.0–1.1 nm), which the authors read as a comparable polymer–drug distance R20·C34. The snapshots (Figure 12) show Eudragit E PO as a compact coil beside the cluster and HPMCAS stretched across it R19. Table E collects the measures.

Table E. Simulation results read from Figures 13, 14 and S7-1 (approximate); one chain per polymer, 10 enzalutamide molecules, 15 ns R20·R21·R22.

The authors attribute the more negative interaction free energy of HPMCAS to the larger surface its extended conformation offers ("explaining"), and in the SI they "propose" that the difference in interfacial area comes from the polymer conformation, since the drug-cluster SASA and the RDFs are similar C35.

Discussion

The central contrast. For HPMCAS and Eudragit E PO, the authors "do not observe a link" between the amount of polymer associated with the droplets and inhibition: Eudragit E PO associated more but was "far less effective" with droplets, which differs from the correlations reported by Ueda et al. and Wang et al. C37. In the abstract this becomes a general statement: in supersaturated solutions containing amorphous nanodroplets, "the extent of association of a polymer with the drug nanodroplet does not solely predict crystallization inhibition" C38. The word "solely" matters: the authors keep distribution between the phases as a necessary first condition, in line with Raina et al., and argue that it is not sufficient C36·C58.

The authors' chain of reasoning. Crystallization can start in either phase or at their interface, so an effective inhibitor must "associate with the drug present in each phase" and then inhibit crystallization in each C3·C36. Without droplets, Eudragit E PO at 50 μg/mL works, so its weakness with droplets "appears to be related" to inhibition in the droplet phase or at the droplet/bulk interface C39·C40. A difference in driving force is ruled out by the equal drop in solution concentration C42. The droplet/solvent interface is "likely" a highly favourable site for heterogeneous nucleation, and all three surface probes, TEM, zeta potential and MD, put the polymers at that interface C45·C43. What remains is how the polymer sits there:

The HPMCAS side of the chain draws on earlier work: HPMCAS owes its effectiveness to hydrophobic regions that interact with drugs plus carboxylic acid groups that ionize above pH 5–6 and keep it solvated, and ionized HPMCAS adsorbs at crystal surfaces in a more extended conformation (Schram 2015) C50. These features "should" also draw HPMCAS to the droplets, and the labelled-polymer spectra are taken to confirm that they do C51. For Eudragit E PO, an amino methacrylate copolymer with hydrophobic groups and an ionized amino group, the authors write that HPMCAS-like behaviour "might be anticipated" C52. The simulation offers "insight into possible reasons" for the difference: the globular Eudragit E PO chain limits its contact with the drug cluster, while the extended HPMCAS chain covers more of it, which agrees with the group's earlier finding that polymers with strong intramolecular interactions in water "tend to be" poorer inhibitors (Mosquera-Giraldo 2016, 2018) C53·C54. Finally, citing the idea that adsorbed additives must desorb before solute molecules can join an embryo, the authors argue that this desolvation barrier "is expected to be larger" for HPMCAS, given its more negative interaction free energy C55. The paper cites this idea as "Liu et al." with reference 71, which is a paper by Chen, Liu et al. on drug–polymer–water interactions; reference 60 (Liu, interfacial effect of molecules on nucleation kinetics) appears to be the intended source X16.

The proposal. In the conclusions the authors write that stabilization "may be mediated" by drug–polymer interactions at the droplet/water interface, that the simulations show that conformation there "may be critical", and "Thus, we propose" that an inhibitor for droplet-containing solutions must first distribute between the aqueous and drug-rich phases and second adopt a certain conformation at the interface to prevent the formation or growth of embryonic crystals; they add that "additional mechanisms may also be important" C57·C58·C59.

What it suggests for formulation. The authors' practical message is that inhibitors should be tested at drug and polymer concentrations like those produced by a dissolving ASD, because droplets change both the polymer amount needed and the ranking C14·C13. Against their 180 min benchmark, droplet-free systems "tend to" stay supersaturated with enough polymer, while with droplets only HPMCAS held 120 μg/mL enzalutamide long enough C48·R3.k·R3.l. Note by the AI writer (not discussed in the paper): a screen run only below the amorphous solubility would have rated Eudragit E PO at 50 μg/mL as good as any polymer here, yet it failed within about 91 min once a few droplets were present R3.y·R3.z. For an ASD that releases drug beyond the amorphous solubility, the polymer that dissolves with the drug has to protect the droplet interface, and the amount of polymer found on the droplets is not by itself a sufficient selection criterion; the paper offers no measurable descriptor for the interfacial conformation it proposes.

Limitations

Limits stated by the authors. The authors note that "there were less drug-rich nanodroplets present in the Eudragit E PO system" when comparing association C20, that Eudragit E PO could not be tested above 70 μg/mL drug because of rapid crystallization C30, and that "additional mechanisms may also be important, and further studies are clearly warranted" C59. Their mechanistic statements are hedged: steric stabilization is "most likely", the interface is "likely" a favourable nucleation site, the desolvation barrier "is expected" to be larger for HPMCAS, and the role of conformation "may be critical" C28·C45·C55·C57.

The AI writer's own judgement (1): the association comparison. The headline contrast rests on Table 2, where Eudragit E PO is measured with 70 μg/mL drug after 15 min and HPMCAS with 120 μg/mL after 1.25 h. In μg/mL the two polymers leave solution to a similar extent; the larger Eudragit E PO percentage follows from the smaller droplet mass it is divided by X4. The colorimetric assay offered as confirmation gives about 19–23 w/w% for both polymers at 50 and 100 μg/mL, so it does not support the ranking X5. The robust part of the conclusion is therefore weaker but still useful: Eudragit E PO, which clearly reaches the droplets, inhibits poorly with droplets present, so reaching the droplets is not enough. Whether it reaches them more than HPMCAS at equal conditions was not measured.

The AI writer's own judgement (2): the mechanism. The step from "both polymers are at the interface" to "conformation decides" rests on a simulation of one short chain per polymer with a 10-molecule cluster for 15 ns, without a stated charge state for Eudragit E PO, and on an interaction energy used as a proxy for a desolvation barrier that was not measured X14·C55. Where nucleation actually starts, in the droplets, at the interface or in solution, was not measured either C45. The induction-time data themselves are solid: they are in triplicate, cover many conditions, and show the droplet effect for all three polymers R3.

The AI writer's own judgement (3): reporting. Several printed statements do not match the plotted data or each other: the polymer levels in the methods (Figure 3 uses 2.5–100 μg/mL) X1; the Kruskal–Wallis caption X2; the droplet sizes in the text against Figure 9 X6·X7; "only HPMCAS" above the amorphous solubility X8; the NMR peak position of enzalutamide X10; SI cross-references X17; and a bar in Figure 3 without an error bar that did crystallize (Eudragit E PO 5 μg/mL with 35 μg/mL drug, about 52 min) X21. The 42 μg/mL amorphous solubility, which defines every droplet amount, is given without method or reference, and induction-time temperature and stirring are not stated X11·X13. None of these changes the main finding that droplets make inhibition harder and that HPMCAS performed best, but they matter for anyone reusing the numbers.

Conclusion

For enzalutamide in pH 6.5 buffer, crossing the amorphous solubility of 42 μg/mL changed the job of a crystallization inhibitor: once drug-rich nanodroplets formed, induction times dropped sharply, and only HPMCAS at 50–100 μg/mL kept 120 μg/mL drug supersaturated beyond three hours R3·C46. Both HPMCAS and Eudragit E PO went into the droplets and lowered the drug in solution to the same extent, yet the cationic Eudragit E PO, with a larger share of the droplets by NMR, failed within about an hour and a half at most droplet-containing conditions R7·R8·C37. The authors conclude that association with the droplets "does not solely predict" inhibition and "propose" that the polymer must also adopt a suitable conformation at the droplet/water interface, extended in the case of HPMCAS, a link drawn from simulation rather than measured at the droplet surface C38·C58·C53.