Watching amorphous solid dispersions phase separate in water with a fluorescent polarity probe

Background and motivation

Many marketed drugs and many more development compounds are poorly soluble (Biopharmaceutics Classification System classes II and IV). Making the drug amorphous raises its apparent solubility, but the amorphous form has a higher free energy than the crystal and tends to crystallize, which cancels the benefit. The usual remedy is to disperse the drug molecularly in an amorphous polymer, giving an amorphous solid dispersion (ASD). The polymer slows crystallization by raising the glass transition temperature (Tg), coupling the motion of drug and polymer, lowering the chemical potential of the drug and forming specific interactions such as hydrogen bonds; a hydrophilic polymer also improves wetting and dissolution (Introduction, refs 1–15). C22

The weak point of this picture is water. ASDs absorb water during storage and, unavoidably, during dissolution. Absorbed water plasticizes the matrix, raises molecular mobility and lowers viscosity, and it can drive amorphous–amorphous phase separation (AAPS): the one-phase dispersion splits into drug-rich and polymer-rich domains while both stay amorphous, at least at first. The drug-rich domains carry little polymer and so crystallize faster; in an earlier felodipine–PVP study, crystals grew from a phase-separated dispersion as fast as from the pure amorphous drug and more slowly from a miscible one (Rumondor 2009, ref 49). C1·C23·C32·R25 Earlier Taylor-group work had shown that PVP dispersions of several drugs, probucol among them, separate under moisture unless the drug hydrogen bonds very strongly to PVP (Rumondor 2009, ref 18; Rumondor 2011, ref 17), while drug–HPMCAS dispersions resisted at high relative humidity (Rumondor 2010, ref 41; ref 49). R24·R26·R27

Detecting AAPS is difficult. Infrared and solid-state NMR spectroscopy, X-ray scattering, DSC and atomic force microscopy have all been used, and each has limits: low sensitivity, long analysis times or demanding sample preparation (refs 19–26, 43, 44). C2 Samples that contain water are the hardest case, and so, the authors point out, nobody had been able to check whether AAPS happens while an ASD hydrates during dissolution, the moment that matters most for performance (Discussion, p.17). C2

The paper's idea comes from polymer science and from the group's own work on supersaturated solutions. Environment-sensitive fluorescent probes change their emission with the polarity of their surroundings and have been used to follow phase separation in polymer blends, water uptake in polymer films, micelle polarity and critical micelle concentration, and, recently, the amorphous solubility of drugs (Ilevbare 2013, ref 37; Jackson 2014, ref 38). C4 The authors' hypothesis ("We hypothesize") is that a trace of such a lipophilic probe, dissolved in an ASD film, will partition into the drug-rich phase when AAPS happens, sense a more hydrophobic environment and change its spectrum, so that the separation can be followed in situ as the film hydrates. C3 They tested this first on probucol–PVP, a system known to separate under moisture, then applied it to ritonavir–PVP, whose behaviour in water had not been reported, with ritonavir–HPMCAS as a control expected not to separate, and checked every step with IR, DSC, X-ray diffraction and microscopy. C11

Background concepts

Amorphous–amorphous phase separation. In a miscible ASD drug and polymer are mixed at the molecular level. When water enters, the authors describe what happens as "a classic case of solvent-induced immiscibility": the polymer (PVP here) prefers water, the hydrophobic drug interacts unfavourably with it, so polymer–water contacts replace drug–polymer contacts and, "given sufficient mobility", a drug-rich phase forms (Discussion, p.17). C24 For ritonavir, whose NH groups hydrogen bond to the PVP carbonyl, sorbed water is said to break exactly these drug–polymer hydrogen bonds (p.18). C27 The resulting domains are still amorphous, but the drug-rich ones have lost most of the polymer that inhibited their crystallization. C1·C32

Environment-sensitive probes and blue and red shifts. The emission of a polarity-sensitive dye moves to shorter wavelength (a blue shift) when its surroundings become less polar and to longer wavelength (a red shift) when they become more polar (p.4). C4 Two probes carry the paper. Prodan has a single broad emission band, so it is read by the wavelength of its maximum, ${\lambda }_{\mathrm{e}\mathrm{m}}$ (excited at 360 nm). Pyrene has five vibronic emission peaks (excited at 332 nm; Figure 2b), and the relative heights of the first and third depend on polarity. P5·R2 The paper defines the read-out in words (p.10):

Here $I_I$ and $I_{\mathrm{I}\mathrm{I}\mathrm{I}}$ are the emission intensities of pyrene peaks I and III. In plain terms: the ratio of the first to the third pyrene peak falls when the probe sits in a less polar, more hydrophobic environment and rises in a more polar one. E1 For a two-component film the useful frame is the pair of pure-component values. Pyrene reads about 1.30 in amorphous ritonavir and about 1.81 in PVP (read from Figure 5, approximate); a ritonavir–PVP film that starts in between and moves to about 1.30 must hold its probe in ritonavir-like surroundings. R6

Solubility parameter. To rank the polarity of polymers the authors use the solubility parameter $\delta$, a measure of cohesive energy: the higher $\delta$, the more polar the material. Where no experimental value existed they calculated it with the Fedors group-contribution method (ref 40); the paper names the method but prints no equation, and the unit of $\delta$ is not printed in Table 2. E2·C6

Reading phase separation in DSC and IR. A molecularly mixed binary amorphous solid shows one glass transition between those of its components; two separate glass transitions mean two amorphous phases, and one close to the Tg of the pure drug points to a drug-rich phase. The authors used modulated DSC and read the Tg steps in the reversing heat flow. P12·C19·C20 In IR spectroscopy, a hydrogen bond between a drug donor (NH or OH) and the polymer carbonyl shifts the carbonyl band; a new band in the dispersion that disappears after water exposure, while the bands of the pure components return, is read as loss of drug–polymer contacts. P10·C17 Crystals, finally, show up as birefringence under crossed polarizers, as sharp peaks in X-ray diffraction and as a melting endotherm in DSC. P8·P11·P12

Methods in detail

The study has three parts: calibrating the probes in single-component films, following ASD films over time in humid air and in buffer, and checking the humidity samples with independent techniques. The flow below summarizes them.

Materials and sample preparation

The model drugs were probucol (PB), a very hydrophobic compound (log P 10), and ritonavir (RTV); procaine (log P 1.3) served as a hydrophilic reference host for the probes. The ASD polymers were PVP K-90 and HPMCAS (MF grade); eight more polymers (PAA, HPMC 606, PVPVA 64, PVAc, polyvinyl-2-pyridine, polyvinyl-4-pyridine and the two cellulose esters CAP-482 and CAB-ADP) were used only to calibrate the probe (Materials, p.5). M1·M2·M3·M4·M5·M6·M7·M8·M9·M10·M11·M12·M13 Films were spin-coated: drug and polymer at a set weight ratio were dissolved in methanol at 50 mg/mL total, with 0.1% w/w probe (1 µg probe per mg dispersion) for the fluorescence films; 100 µL was spun at 3000 rpm for 30 s onto quartz coverslips (fluorescence) or KRS-5 discs (IR) and dried under vacuum for at least 24 h. Pure polymer films were spun from 20 mg/mL polymer in methanol:dichloromethane 1:1, pure amorphous drug films from methanol (p.6). P1·P2 Powders for DSC and XRD were made by rotary evaporation of methanolic solutions at 45 °C, vacuum dried for 24 h and stored over phosphorus pentoxide; their drug:polymer ratio is not printed. P3·X8

Probe calibration

Four probes (Table 1 lists pyrene, 4-Di-2-ASP, prodan and dansyl glycine) were dissolved in amorphous probucol and procaine films, and prodan in films of ten polymers, and their emission was recorded on a Shimadzu RF-5301PC spectrofluorophotometer (pp.6–7). P4·R1·R3 The methods section names pyrenecarboxyaldehyde instead of 4-Di-2-ASP among the screened probes; Table 1 reports 4-Di-2-ASP and gives no pyrenecarboxyaldehyde data. X1

Humidity and immersion experiments

For the humidity series, PB–PVP films at 50:50 and 20:80 (with prodan) and RTV–PVP films at 20:80, 50:50 and 80:20 (with pyrene) were stored at 97% RH in desiccators over saturated K₂SO₄ and read in a solid-sample holder at times from 1 h to 5 days; 50:50 RTV–HPMCAS films with pyrene and films of the pure components were treated the same way (p.7). P5 The storage temperature and the number of replicates behind the error bars are not stated. X8 For the immersion series, 50:50 films of PB–PVP, RTV–PVP and RTV–HPMCAS were immersed in 100 mM pH 6.8 phosphate buffer for times from 10 s to 30 min (RTV–HPMCAS to 15 min), taken out and measured on the undissolved film material; pellets of the two PVP dispersions were also dissolved and measured, but these data are not shown (pp.7, 13–14). P6·P7·R12

Independent techniques

Films were imaged by fluorescence microscopy (Olympus BX51; excitation 330–380 nm, emission from 420 nm, so pyrene can be imaged; 40X objective in Figure 6) and checked for crystals under crossed polarizers (Nikon Eclipse E600 Pol, 10X). P9·P8 IR spectra (400–4000 cm⁻¹, 64 scans, 4 cm⁻¹) were recorded in transmission; films that had been at 97% RH were dried under vacuum and flushed with dry air until the OH and NH regions stopped changing, so the spectra describe the dried films, not the wet ones (p.8). P10 DSC used about 5 mg of powder in pinholed pans, a 10 h nitrogen purge, then heating at 5 °C/min to 200 °C with ±2 °C modulation every 60 s; amorphous PB and RTV were made in the pan by melting and quench cooling (pp.8–9). P12 XRD covered 5–35° 2θ on the 250–425 µm sieve fraction of the powders (p.8). P11 The exposure time at 97% RH is given for the PB–PVP DSC (6 h and 30 h) and XRD (1, 16 and 30 h) samples but not for the IR spectra or the RTV–PVP DSC and XRD samples. X8

Experimental analysis

The probes read polarity in amorphous solids

All four probes registered amorphous probucol as less polar than amorphous procaine (Table 1). Prodan showed the largest difference, a blue shift of 30 nm from procaine to probucol, and the pyrene ratio fell from 1.4 to 1.2. R1·R2·C5 In polymer films, prodan's emission maximum spanned 33 nm, from 427.0 nm in PVAc to 460.0 nm in PAA, and the ten polymers fall in exactly the same order as their solubility parameters (Table 2; Table A). The authors take this as support ("providing support for the supposition") that prodan reports the polarity of its local polymer environment. R3·C6 The two ASD polymers sit at different points of this scale: PVP K-90 at 450.7 nm with $\delta$ 25.60, HPMCAS-MF at 441.5 nm with $\delta$ 22.44. R3

Table A. Prodan emission in single-component amorphous films, from Tables 1 and 2 (unit of δ not printed). R1·R3·E2

Probucol–PVP at 97% RH: separation, then crystallization

The dry 50:50 PB–PVP film read 437 nm, between pure PVP (451 nm) and amorphous probucol (418 nm), which the authors read as a miscible film with intermediate properties (p.11). R4·C7 At 97% RH the emission fell to about 426 nm at the first reading (1 h) and stayed at 425–427 nm up to 28 h; the minimum in Table 3 is 425 nm (Figure 3, read approximately). R4·R15 The text calls this a "gradual" blue shift, but nearly all of the 12 nm fall is already there at 1 h, so the onset of separation is faster than the sampling can resolve. C8·X6 The authors' explanation is that water uptake separates the film into drug-rich and polymer-rich domains and the inherently hydrophobic probe partitions into the drug-rich phase ("This can be explained by the absorption of water…"). C8

After 21 h the trend reversed. The emission rose to about 430.5 nm at 40 h, about 433 nm at 64 h and about 434.5 nm at 5 days (Table 3 end point 435 nm; the TOC graphic adds a 7-day point near 435 nm). R4·R15·R23 Brightfield and cross-polarized images of the film after 21 h show branched, birefringent crystals (Figure 4). R5 The authors explain the red shift by probucol crystallization, which expels the probe into the hydrophilic PVP-rich environment; the end value stays below pure PVP because some probucol "might be" left amorphous, which they connect with a DSC Tg matching amorphous probucol after 30 h. C9·R19 The 20:80 film behaved similarly from a higher start (443 nm, more polymer): its minimum was 437 nm and its end point 445 nm, with about 447 nm plotted at 5 days. R4·R15·C10·X2

Note by the AI writer (not discussed in the paper): water alone also moves prodan to longer wavelengths. In a pure PVP film the emission rose from about 450.5 to about 455.3 nm over 5 days at 97% RH (Figure 7). Part of the late rise in the PB–PVP films may therefore come from continued hydration of the PVP-rich phase rather than from crystallization alone, and the 20:80 film, for which no crystallization evidence is shown, ends above its starting value. R9·X9

Ritonavir–PVP at 97% RH: down to the ritonavir value

All three RTV–PVP films started between pure PVP (about 1.81) and pure ritonavir (about 1.30) and dropped at the first reading (Figure 5; Table B). The 50:50 film kept falling and plateaued at 1.30, "very close" to amorphous ritonavir; the 80:20 film reached 1.29 (about 1.275 at 5 days in Figure 5); the 20:80 film stayed near 1.50–1.55, with a lowest value of 1.48 in Table 3 (p.12). R6·R15·C12·X2 Ritonavir is a very slow crystallizer and did not crystallize during the experiment, so no red shift followed. C12 Pyrene was chosen for this system because it responded more than the other dyes tried (data not shown). R6

Table B. Probe values of ASD films at 97% RH, from Table 3; reference values of the pure components read from the time-0 points of Figures 3, 5 and 9 (approximate). The time of the "end point" is not stated, and some Table 3 values differ slightly from the plotted points. R15·R4·R6·R13·X2

To test the assumption that the probe goes where the drug goes, the 80:20 RTV–PVP film was imaged before and after humidity (Figure 6). Before, the fluorescence was low and uniform; after, the probe sat in many bright, roughly round domains a few micrometres across (about 5 µm against the 100 µm scale bar) with dark gaps between them, which the authors identify as "presumably" the drug-rich amorphous domains. R7·C13 Films of pure PVP and pure ritonavir stayed uniform under the same treatment (images not shown). R8

Controls: water alone, and the HPMCAS dispersion

The key control is that water alone moves the probes the other way. Over 5 days at 97% RH, prodan in pure PVP rose from about 450.5 to 455.3 nm and pyrene in amorphous ritonavir from about 1.305 to 1.365 (Figure 7), and pyrene in HPMCAS from about 1.515 to 1.565 (Figure 9). The authors take this as support ("supports the supposition") that the blue shift of the PVP dispersions comes from moisture-induced AAPS and not from the absorption of water per se. R9·R13·C14

The 50:50 RTV–HPMCAS film never moved toward ritonavir: its pyrene ratio went from 1.42 to 1.47 (Table 3; about 1.49 at days 4 and 5 in Figure 9), which the authors attribute to the film becoming more hydrophilic as it absorbs water. R13·R15·C16 Its IR spectrum, with a carbonyl band at 1733 cm⁻¹ that the authors assign to RTV–HPMCAS interactions, was unchanged after humidity and drying (Figure 12). R18·C18 Note by the AI writer (not discussed in the paper): the 1733 cm⁻¹ band lies between the HPMCAS carbonyl maximum (about 1743 cm⁻¹) and the ritonavir band at 1717 cm⁻¹, so an overlap of the two component bands could also produce it; the unchanged spectrum after humidity does not depend on that assignment. X10

Immersion in buffer: fast for probucol, slower for ritonavir

The central question of the paper is whether AAPS also happens when an ASD meets bulk water. Table C collects the immersion data. For 50:50 PB–PVP, prodan fell from 437 nm to about 428 nm after 10 s and to about 421.5 nm at 7–10 min, then rose slightly to about 424.5 nm at 30 min (Figure 8a). The text gives the final position as "around 425 nm", the same as after 97% RH, and Table 4 lists 425 nm as the lowest value, although the plotted minimum is lower. R10·R16·C15·X3 For 50:50 RTV–PVP, the pyrene ratio fell from 1.57 to about 1.49 at 10 s, stayed between about 1.45 and 1.51 for 15 min, reached about 1.405 at 20 min and 1.29 at 30 min, close to amorphous ritonavir (Figure 8b); the authors read the final 1.3 as "suggesting" an RTV-rich environment. R11·R16·C15 For both films the signal disappears once most of the film has dissolved, so every reading is taken on the material that remains. R10 The 50:50 RTV–HPMCAS film again moved away from ritonavir, from 1.42 to about 1.50–1.53 within 2–15 min, close to pure HPMCAS (Figure 10). R14·R16

Table C. Probe values of films immersed in 100 mM pH 6.8 phosphate buffer, from Table 4 and read from Figures 8 and 10 (approximate). R10·R11·R14·R16·X3

From these data the authors conclude that "AAPS is very fast following immersion, with the largest shifts in the emission peak occurring within a few seconds" (p.13). C15 That description fits PB–PVP, where more than half of the fall is complete at 10 s. For RTV–PVP the largest single step comes between 20 and 30 min, just as the film disappears. X4 Dissolving pellets of the same two PVP dispersions gave the same trends, but those data are not shown. R12

IR, DSC and XRD of the humidity samples

IR. Dry amorphous ritonavir has carbonyl bands at 1630 and 1717 cm⁻¹ and dry PVP at 1681 cm⁻¹. The dry 80:20 RTV–PVP film shows a new band at 1654 cm⁻¹, assigned to hydrogen bonds between the PVP carbonyl and ritonavir's donors. After exposure to 97% RH and drying, this band was gone and the bands at 1630 and 1717 cm⁻¹ (an RTV-rich phase) and at 1681 cm⁻¹ (a PVP-rich phase) were present (Figure 11), which the authors call "indicative of AAPS" and in good agreement with the fluorescence data. R17·C17 The PB–PVP system was not studied by IR here because its moisture behaviour had been characterized that way before (ref 17). R24

DSC. Both dry dispersions showed a single Tg, which the authors read as molecular mixing (Figures 13, 14). Read from the figures, the dry PB–PVP powder had one step at about 130–150 °C and the dry RTV–PVP powder one at about 79–104 °C (midpoint about 94 °C); the drug:polymer ratio of these powders is not printed. R19·R20·C19·C20·X8 After 6 h at 97% RH the PB–PVP powder showed two Tg steps (near 40 and 100 °C), "suggesting phase separation", before probucol starts to crystallize at 21 h. After 30 h it showed a Tg at 25 °C, which corresponds to amorphous probucol, and a large endotherm close to the melting point of crystalline probucol, which the authors attribute to crystallization "either due to continued moisture exposure or during the DSC heating cycle". R19·C19 The exposed RTV–PVP powder showed two Tgs with midpoints of 50 °C, an RTV-rich phase (amorphous ritonavir read about 48 °C), and 140 °C, a PVP-rich phase. R20·C20

XRD. Both powders gave halos before exposure. PB–PVP still gave halos after 1 h and 16 h at 97% RH and sharp crystalline peaks at 30 h; the text puts their emergence at exposure "longer than 30h", while the 30 h pattern in Figure 15 already shows them. RTV–PVP stayed amorphous (Figure 16). R21·R22·C21·X11 Taken together, probucol crystals were seen at 21 h in films and at 30 h in powders, in step with the red shift of the probe. R5·R19·R21

Discussion

The authors connect the findings in a short causal chain (Discussion, pp.16–20). The flow below sets it out, with the polymer deciding which branch a dispersion takes.

On this basis the authors argue that, because both PVP systems separate in humid air by fluorescence, IR and DSC, "it is reasonable to expect" the same in bulk water, and that the immersion results "strongly suggest" that both dispersions undergo AAPS as the film hydrates during dissolution, "albeit with faster kinetics" (pp.17–18). C25 They see drug content as a factor: in PB–PVP the larger fall at 50:50 (12 nm) than at 20:80 (6 nm) is read as "suggesting" more extensive AAPS, as found before for this system, and in RTV–PVP the 50:50 and 80:20 films "appear to undergo the largest extent" of AAPS because their ratios reach the ritonavir value. C26·C28·R15 Polymer type matters more: PVP dispersions "seem to be particularly susceptible" to moisture-induced AAPS, with only drugs that hydrogen bond very strongly to PVP resistant, while HPMCAS dispersions showed no water-induced immiscibility here and in earlier work (p.19). C29·C30·R26·R27

The consequences for performance are argued, not measured. AAPS "is likely to be important" because it "would be expected to" shorten shelf life, by making crystallization more likely, and to impair dissolution, both lowering bioavailability. Packaging can keep water away during storage but not during dissolution, so understanding AAPS on hydration "may thus help to explain" why ASD dissolution depends on polymer type and amount, and "it might be anticipated" that an ASD that separates while dissolving in the gastrointestinal tract crystallizes more within the matrix and reaches lower supersaturation (pp.19–20). C31·C33 The authors "believe" that the fast, in situ fluorescence method will help researchers evaluate drug–polymer combinations during formulation development. C34·C36

Note by the AI writer (not discussed in the paper): for formulation the practical message is about timing. In these films, a hydrophobic drug in PVP moved the probe most of the way toward the pure-drug value within seconds to minutes of wetting, before the film had dissolved, while HPMCAS kept ritonavir away from it in both vapour and buffer. Whether that matters for release depends on what the drug-rich domains do next, which this paper does not measure.

Limitations

Limits the authors state. The authors give few explicit limitations. They note that the end value of the crystallized PB–PVP film does not reach pure PVP because some probucol "might be" left amorphous, that the DSC melting endotherm may come from crystallization "either due to continued moisture exposure or during the DSC heating cycle", and that the signal of immersed films disappears once most of the film has dissolved. C9·C19·R10 The links from separation to shelf life, dissolution, supersaturation and bioavailability are put as expectations ("would be expected", "might be anticipated"). C31·C33

The AI writer's own judgement (1): the dissolution evidence. The claim that AAPS happens during dissolution rests on one technique applied to what is left of the film after immersion. If PVP leaves the film faster than the drug, the residue becomes drug-rich and the probe would report a drug-like environment even without phase separation inside a hydrated matrix; the final RTV–PVP value equals pure ritonavir just when the film has mostly dissolved. No second technique was used on immersed films, the composition of the residual film was not measured, and the pellet data are not shown. "Within a few seconds" also holds for PB–PVP only. X5·X4·R12 The humidity evidence is much stronger: four independent read-outs agree there. R4·R6·R7·R17·R19·R20

The AI writer's own judgement (2): numbers and conditions. The summary values in Tables 3 and 4 do not always match the plotted points: the PB–PVP 20:80 end point is 445 nm in Table 3 and about 447 nm at 5 days in Figure 3, the RTV–PVP 20:80 lowest value is 1.48 in the table and about 1.50 in the figure, and the PB–PVP lowest value in buffer is 425 nm in Table 4 and about 421.5 nm in Figure 8a; the time of the "end point" is never stated. X2·X3 The drug:polymer ratio of the DSC and XRD powders, several exposure times, the temperatures of storage and immersion, and replicate numbers are not given. X8 Different probes were used for the two drugs and probe partitioning was not calibrated, so the "extent" of separation can be compared only within one drug–probe pair. X12 Only one HPMCAS dispersion (ritonavir, 50:50) was tested, although the discussion speaks of "the corresponding HPMCAS dispersions" in the plural. X7 There are also small textual slips: the methods name a probe (pyrenecarboxyaldehyde) that Table 1 does not report, and the text cites Figure 1 for the pyrene spectrum, which is in Figure 2. X1·X13

Conclusion

Environment-sensitive probes at 0.1% w/w turn phase separation of an ASD into a spectral shift that can be followed over time: prodan and pyrene report the polarity of amorphous solids, move toward the pure-drug value when PVP dispersions of probucol and ritonavir separate at 97% RH, and move the other way with water alone or in a ritonavir–HPMCAS dispersion that stays mixed. R1·R3·R4·R6·R9·R13·C35 The humidity results are confirmed by imaging, IR and DSC, and the later red shift of PB–PVP coincides with probucol crystals. R7·R17·R19·R20·R5·R21 In buffer the same shift appears within seconds for PB–PVP and within 30 min for RTV–PVP, which the authors read as AAPS during dissolution ("strongly suggest"); this rests on the probe signal of the remaining film alone. R10·R11·C25·X5 The consequences for crystallization, supersaturation and bioavailability are proposed, not measured. C31·C33