How the HPMCAS grade and the intestinal pH set the balance between release and crystallization of posaconazole

Background and motivation

Amorphous solid dispersions (ASDs) hold a poorly soluble drug dispersed at the molecular level in a polymer. When they dissolve under non-sink conditions, as in the gut, they produce supersaturated solutions: the dissolved drug has a higher chemical potential than in a solution saturated with crystals, which raises the flux across the intestinal membrane and so absorption (Section "Introduction", refs 10–17). Two things can happen to such a solution. If the concentration exceeds a limit, the solution undergoes liquid–liquid phase separation (LLPS) and a second, drug-rich liquid phase forms, usually at first as droplets a few hundred nanometres across. And at any time the drug can crystallize, which removes the supersaturation. In a solution that has phase-separated, crystals could start in the bulk aqueous phase, inside the drug-rich droplets, at the droplet surface, or in several places at once (refs 23–26) C2·C3.

Posaconazole, an antifungal weak base, is a fast crystallizer: it crystallizes faster than indomethacin or ritonavir (refs 27, 28) C4. It is marketed as an ASD with hydroxypropyl methylcellulose acetate succinate (HPMCAS) in Noxafil, and HPMCAS is known to delay posaconazole crystallization (refs 27, 43–45) C5. HPMCAS comes in three grades, L, M and H, that differ in how many of the cellulose hydroxyls carry acetyl and succinoyl groups. The acetyl group is hydrophobic; the succinoyl group carries a carboxylic acid that ionizes as the pH rises, which is what makes HPMCAS dissolve at intestinal but not at gastric pH. The ratio of acetyl to succinoyl (A/S ratio) rises from L to M to H (Figure 1), so H is the most hydrophobic grade and the one that needs the highest pH to dissolve M2–M4.

Earlier work had compared HPMCAS grades for ASD dissolution and recrystallization (Ueda 2014, ref 33; refs 39–42). The authors identify two open points. First, how the polymer acts on crystallization once the solution has phase-separated, so that drug sits both in solution and in droplets, was poorly understood. Second, inhibition and release had not been examined together while both the grade and the degree of ionization were varied. Their working hypothesis was that hydrophobic interactions between HPMCAS and posaconazole, in the bulk solution and in the drug-rich phase, are what delay crystallization C6. To test this they measured, at pH 5.5, 6.5 and 7.5 with each grade, the amorphous solubility, the nucleation induction time, the fluorescence of posaconazole, the size and charge of the drug-rich droplets and how much polymer goes into them, and then the release of drug and polymer from 10 and 25 wt% drug-loaded ASDs.

Background concepts

Amorphous solubility and LLPS. For a poorly soluble drug, two solubilities matter. The crystalline solubility is the concentration in equilibrium with crystals; for posaconazole in the buffer used here it is about 0.3 µg/mL R1. The amorphous solubility is the highest concentration of dissolved drug before a drug-rich liquid phase separates; drug added above it does not raise the dissolved concentration but goes into droplets. The authors measure it as the "LLPS concentration": they make a solution far above the limit, spin the droplets out with an ultracentrifuge and measure what remains dissolved P3. In this paper the amorphous solubility was about 30 to 40 times the crystalline solubility C10. The ratio between the dissolved concentration and the crystalline solubility is the supersaturation, and it is this excess that drives both faster absorption and crystallization.

Nucleation induction time. Crystallization from a supersaturated solution starts with nucleation, the formation of tiny ordered clusters that can grow. The induction time is the time from creating the supersaturated solution until crystals are first detected. Here it was read from the onset of a rapid rise in light extinction at 400 nm, a wavelength posaconazole does not absorb, so the rise comes from scattering by new particles P4. A longer induction time means better inhibition; a polymer that doubles or triples it buys time for absorption, which the authors take as 2–4 h in the small intestine C31.

HPMCAS ionization. HPMCAS behaves as a weak acid with a pKa of about 4.9 (ref 55). Below this pH the succinoyl acids are mostly unionized and the polymer is insoluble; above it they ionize, the polymer becomes charged and dissolves. Because L carries the most succinoyl groups (about 14–18% against about 4–8% for H, read from Figure 1B), it ionizes most at a given pH and dissolves at the lowest pH; H dissolves only at higher pH M2–M4. Raising the pH from 5.5 to 7.5 therefore makes each grade more charged and more hydrophilic. Posaconazole, with pKa values of 3.6 and 4.6, is almost entirely unionized over this range, so pH changes the polymer, not the drug M1·C9.

Reading the measurements. Posaconazole is fluorescent, and its emission maximum moves to shorter wavelength (a blue shift) when its surroundings become less polar (ref 43) C7. Dynamic light scattering (DLS) gives a z-average size, the intensity-weighted mean hydrodynamic diameter computed from the droplets' diffusion through the Stokes–Einstein equation, and a count rate, the amount of scattered light; the paper names the Stokes–Einstein equation but does not print it E1. The count rate rises with both the number and the size of droplets, so at similar size a higher count rate means more droplets. The ζ-potential is the electric potential near a particle surface; posaconazole droplets on their own carry almost no charge, so a negative ζ-potential in the presence of HPMCAS points to ionized polymer at the droplet surface C8.

Methods in detail

All experiments used the buffer of fasted state simulated intestinal fluid V2 (maleic acid, sodium hydroxide and sodium chloride, made to the Biorelevant recipe), adjusted to pH 5.5, 6.5 or 7.5, at 37 °C. No bile salt or lecithin is mentioned, so the medium is a plain buffer. Dissolving HPMCAS lowered the pH, and for the solution samples it was brought back with 1 M sodium hydroxide M5·P1. The work has two arms run side by side: solution tests that separate the effects of grade and pH on solubility, nucleation and the drug-rich phase, and release tests on real ASDs.

Solution tests: solubility, amorphous solubility and nucleation

Problem. To see how each grade and each pH act on posaconazole, the drug has to be put into solution at a known supersaturation without the complication of dissolving an ASD. What they did. The crystalline solubility was measured by stirring excess crystals (Form I) in buffer for 48 h, ultracentrifuging at 35,000 rpm for 30 min at 37 °C and assaying the supernatant by HPLC with fluorescence detection P2. For the amorphous solubility, a concentrated methanolic stock of posaconazole was added to buffer containing 900 µg/mL HPMCAS to give 100 µg/mL, about ten times the expected limit (a "solvent shift"); the droplets were spun out under the same ultracentrifuge conditions and the dissolved drug was measured P3. Without polymer this was not possible, because the drug crystallized too fast, and H could not be tested at pH 5.5 because it does not dissolve there C56. Nucleation was followed at 10 µg/mL posaconazole with 90 µg/mL HPMCAS or with no polymer: extinction at 400 nm was recorded for 4 h and the onset of its rapid rise taken as the induction time, three samples per condition P4. Why it works. Solvent shift creates the supersaturation in seconds, so the clock for nucleation starts at a defined moment, and the polymer is already dissolved, so the test separates the polymer's action in solution from its dissolution.

Note by the AI writer (not discussed in the paper): the 1:9 drug-to-polymer ratio of these tests is that of a 10 wt% DL ASD, and the 100 µg/mL drug with 900 µg/mL polymer of the phase-separated samples is the composition reached when a 10 wt% DL ASD dissolves completely in the release test P3·P4·P6. The induction time was measured at 10 µg/mL with 90 µg/mL polymer, while the LLPS concentrations of about 9.9–12.5 µg/mL were measured with 900 µg/mL polymer R2·R3. The nucleation solutions therefore sat close to, and for some conditions possibly above, the amorphous solubility, and whether they contained droplets was not checked; the discussion describes them as "close to the amorphous solubility" X12.

Spectroscopy and the drug-rich droplets

What they did. Fluorescence spectra (excitation 310 nm, emission 250–500 nm) were recorded below the LLPS concentration at pH 6.5, from 0 to 16 µg/mL posaconazole in buffer alone or with 900 µg/mL of each grade (slits 15 and 5 nm). Above the LLPS concentration, 100 µg/mL samples at all three pH values were measured (slits 5 and 3 nm), then ultracentrifuged and the supernatant, which holds the drug at exactly the amorphous solubility, measured again (slits 15 and 5 nm) P7·P8. The same 100 µg/mL samples went into DLS (37 °C, 173° backscatter) for size and count rate, with extra runs at 50 and 150 µg/mL at pH 6.5, and into a folded capillary cell for the ζ-potential P9·P10. Because two slit settings were used, absolute intensities can be compared only within one set, not between the 100 µg/mL samples and their supernatants X3.

How much polymer goes into the droplets

Problem. If the polymer acts inside or on the droplets, one needs to know how much of it is there. What they did. After ultracentrifugation of the 100 µg/mL samples, the HPMCAS left in the supernatant was measured by HPLC with an evaporative light scattering detector; the polymer missing from the starting 900 µg/mL was taken to have gone into the sedimented drug-rich phase P11·E2. The paper states the estimate of droplet composition in words: from this depletion and the known amorphous solubility, the relative amounts of drug and polymer in the droplets follow, with water ignored because it was not determined E3.

Note by the AI writer (not discussed in the paper): one way to reproduce the printed fractions is to take the drug in the droplets as the 100 µg/mL total minus the LLPS concentration. For H at pH 6.5 this gives 900 − 861 = 39 µg/mL polymer and 100 − 12.5 = 87.5 µg/mL drug, so a polymer fraction of 39/(39 + 87.5) = 0.31, as printed; the same arithmetic gives 0.05 for L and 0.14 for M E3·R16·R17. The "about 1.7 mol%" quoted for H needs the molar mass of HPMCAS, which the paper does not print.

Release from ASDs

What they did. ASDs at 10 and 25 wt% DL were made with each grade by dissolving drug and polymer in dichloromethane–methanol (1:1, 10 wt% solids), evaporating at 45 °C under vacuum, drying overnight, cryo-milling and sieving to 106–250 µm P5. Release was tested in 100 mL of buffer at 37 °C and 50 rpm in a USP II apparatus for 3 h, at a stated target of 100 µg/mL posaconazole on complete dissolution. Samples of 0.5 mL were taken and replaced with fresh buffer, centrifuged at 3000 rpm for 5 min to remove undissolved particles, and both drug (HPLC) and polymer (HPLC with ELSD) were measured, so the drug curve can be compared directly with the polymer curve P6. No solid-state characterization of the ASDs or of any precipitate is reported X10.

Note by the AI writer (not discussed in the paper): the methods give 100 mg of the 10 wt% DL ASD and 25 mg of the 25 wt% DL ASD per 100 mL. The first gives 100 µg/mL drug and 900 µg/mL polymer as stated; the second would give only 62.5 µg/mL drug. The 25 wt% DL curves in Figure 4 reach about 97–98 µg/mL and the polymer axis ends at 300 µg/mL, which fits about 40 mg of ASD, so the printed 25 mg is probably a misprint X1.

Experimental analysis

Amorphous solubility and nucleation in solution

The crystalline solubility hardly changed with pH: 0.32 ± 0.02, 0.34 ± 0.03 and 0.31 ± 0.02 µg/mL at pH 5.5, 6.5 and 7.5, as expected for a drug that is unionized above pH 5 R1·C9. The amorphous solubility, measured in the presence of each grade, was 30 to 40 times higher (Table A, Figure 2) R2·C10. It rose slightly from L to M to H and fell slightly as the pH rose; at pH 6.5 the values were 10.1, 10.6 and 12.5 µg/mL. The authors read both trends as an effect of ionizable succinoyl: less of it (H) raises the amorphous solubility, more ionization (higher pH) lowers it ("indicating") C11·C12. Later they take the slight rise as consistent with drug–polymer interactions in solution C39.

Table A. Amorphous solubility (with 900 µg/mL HPMCAS) and induction time (10 µg/mL posaconazole with 90 µg/mL HPMCAS), 37 °C. Values with ± are from the text; "about" values are read from Figures 2 and 3 (approximate). Replicate numbers are stated only for the induction time (three samples) R2·R3.

The induction times show the paper's central effect much more strongly than the solubilities. Without polymer posaconazole nucleated within about half an hour at every pH. L added little, especially at pH 7.5 (about 29 against about 24 min). M delayed nucleation beyond the 4 h window at pH 5.5, to 208 min at pH 6.5 and to 136 min at pH 7.5. H prevented nucleation over 4 h at pH 5.5 and 6.5, and at pH 7.5 one of three samples nucleated at 219 min R3. The authors rank the grades H > M > L and read the shorter times at higher pH as weaker inhibition by the more ionized polymer ("indicating") C13·C14. The two variables act in the same direction: both a higher A/S ratio and a lower pH mean fewer charged succinoyl groups on the polymer.

Release from the ASDs: three grades, three outcomes

Figure 4 shows drug and polymer in solution over 3 h for each grade at 10 and 25 wt% DL and three pH values. The patterns are summarized in Table B R4·R5.

Table B. Posaconazole in solution (µg/mL; target 100) during release from the ASDs, read from Figure 4 (approximate) R4.

L. At every pH the L polymer dissolved within 10–15 min, and the drug concentration peaked at the first or second sample and then fell steadily to 3–11 µg/mL by 3 h R4·R5. The authors describe this as almost immediate recrystallization after complete release, "as inferred from the drop in solution drug concentration", and call the supersaturation from L ASDs "extremely short-lived" C18·C34. The text says that at pH 7.5 L and M released the drug "completely" within 15 min. That holds for the polymer; the drug from L never exceeded about 88.5 µg/mL (10 wt% DL) or about 80.5 µg/mL (25 wt% DL), because it was already being lost while the ASD dissolved C15·X14.

H. At pH 7.5 the H polymer and the drug dissolved over about 30 min and the drug then stayed at about 92–98 µg/mL for the rest of the test, with no sign of crystallization R4.c·R4.f·C20. At pH 6.5 the polymer reached only about 365 of 900 µg/mL (10 wt% DL) and about 200 of 300 µg/mL (25 wt% DL) by 3 h, and the drug reached about 31 and about 22.5 µg/mL. At pH 5.5 no polymer was detected in solution and the drug stayed at about 5 µg/mL R4·R5. The authors attribute the incomplete release to the low solubility of the H (and M) polymer at the lower pH values ("could be attributed") and note that the patterns are "consistent with polymer-controlled dissolution" C16·C17.

M. M is the intermediate case. At 10 wt% DL it kept the drug at about 92–99 µg/mL for the full 3 h at pH 6.5 and for 60 min at pH 7.5 before the concentration fell. At 25 wt% DL the drug fell almost at once at pH 7.5 and after 10–15 min at pH 6.5. At pH 5.5 M released only about 25 (10 wt% DL) and 14.5 µg/mL (25 wt% DL), with no fall R4·C19. The authors conclude that in this study M "offers the optimum balance of properties in terms of crystallization inhibition versus release", but "still performs relatively poorly at the lowest pH studied" C36.

Note by the AI writer (not discussed in the paper): in the incomplete releases the drug lagged the polymer. At pH 6.5 and 25 wt% DL, H had released about 67% of its polymer but only about 22.5% of the drug; at pH 5.5, M had released about 53% (25 wt% DL) and about 36% (10 wt% DL) of its polymer against about 14.5% and 25% of the drug; and at pH 5.5 H gave about 5 µg/mL drug with no measurable polymer R4·R5·X11. Release was therefore not strictly congruent, even though the order of release followed polymer solubility.

What fluorescence says about the drug's surroundings

In buffer at pH 6.5, molecularly dissolved posaconazole emitted with a maximum at 380 nm. Its intensity grew gradually with concentration up to 10 µg/mL (from about 22 to about 87.5 a.u.) and jumped to about 138 at 11 µg/mL, where the emission maximum also began to move to shorter wavelength (Figure 5A, B, D) R6·R7·R9. The authors assign the jump and the shift to drug in the drug-rich phase, a more lipophilic environment than water C21. In the presence of each grade the same pattern appeared, but the emission maximum below the LLPS concentration already depended on the grade: 380 nm in buffer, 379 nm with L, 376 nm with M and 372 nm with H, constant from 2 to 10 µg/mL. Above the LLPS concentration it settled at 372, 371, 370 and 368 nm R9. The authors write that this blue shift, in the order no polymer, L, M, H, is "suggesting" (results) and "signifies" (discussion) that drug–polymer interactions form in bulk solution and give the drug a less polar environment; they note that the 1 nm difference between buffer and L is within method variability C22. At equal concentration the polymer also raised the intensity, for example at 8 µg/mL about 59 a.u. in buffer against about 67.5, 75.5 and 88.5 a.u. with L, M and H (Figure 5B, C) R7·R8.

At 100 µg/mL, where most of the drug is in droplets, emission peaked near 370 nm. Its intensity decreased from L to M to H at each pH (at pH 6.5 about 313, 296 and 268 a.u.) and increased with pH (with L about 240, 313 and 335 a.u. at pH 5.5, 6.5 and 7.5). In the supernatants the order reversed: intensity rose from L to H (pH 6.5: about 77, 96 and 160 a.u.), the peak moved from about 380 to about 372 nm, and intensity fell with pH (Figure 6) R10. The authors read lower intensity at 100 µg/mL as less drug in the drug-rich phase ("likely indicated") and higher supernatant intensity as more dissolved drug, in line with the higher LLPS concentration ("suggested") C24·C25·C26.

Note by the AI writer (not discussed in the paper): these intensity readings are weaker than the wavelength shifts. The polymer itself raises intensity at equal drug concentration, so the supernatant differences mix drug amount and polymer environment; at pH 7.5 the supernatant intensity with H is about twice that with L, while the LLPS concentrations differ by about 20% X4. At 100 µg/mL the drug in the droplets, by mass balance, changes by only about 2–3% between grades or pH values, while the intensity changes by 15–40% X5. In addition, the text calls the 100 µg/mL spectra "solid lines" and the supernatants "dotted lines", whereas the legend of Figure 6 draws the 100 µg/mL samples as markers and the supernatants as lines X2.

Size, charge and composition of the drug-rich droplets

Table C. Drug-rich droplets at 100 µg/mL posaconazole and pH 6.5, 37 °C (Figures 7–9). Values with ± are from the text; "about" values are read from the figures (approximate). The weight fraction ignores water R11·R12·R15·R16·R17.

Without polymer the droplets were about a micrometre across and nearly uncharged. Each grade made them smaller, in the order L > M ≈ H, and raised the count rate in the order L < M < H at pH 6.5, which the authors read as fewer, larger droplets with L and more, smaller droplets with H R11·R12·C28. The L droplets also grew with pH (291 ± 10, 462 ± 40 and 512 ± 65 nm at pH 5.5, 6.5 and 7.5) and with drug concentration (about 415, 462 and 520 nm at 50, 100 and 150 µg/mL), whereas M and H droplets stayed at about 265–305 nm under all conditions R11·R14·C29. The ζ-potential was negative with all grades and least negative with H, matching the number of acid groups; the authors conclude that every grade sits on the droplet surface and report no notable change with pH R15·C27·C43. Polymer depletion from solution was largest with H: 861 ± 7 µg/mL remained against 895 ± 6 (L) and 886 ± 8 (M), which gives droplet compositions of about 31, 5 and 14 wt% polymer, water ignored. Partitioning fell slightly with rising pH R16·R17·C30. In the authors' words, some polymer "appears to mix" with the drug-rich phase, in particular H C45.

Note by the AI writer (not discussed in the paper): two details limit what Table C can carry. The L and M depletions (5 and 14 µg/mL) are within about one and two standard deviations of the measurement, so only H clearly takes up polymer X7. And M and H droplets are the same size even though H takes up about three times more polymer, so the abstract's statement that more partitioning results in smaller droplets is a co-occurrence across grades rather than a tested link X6. Some summary statements also do not fit every point: at pH 5.5 L had a higher count rate (about 222,000 cps) than M (about 215,000), and M's ζ-potential was about −8.5 mV at pH 5.5 against −10.8 mV at pH 6.5 X9.

Discussion

The authors' explanation

The authors build their explanation from the balance of three pairwise affinities: drug–water, polymer–water and drug–polymer. Posaconazole is lipophilic (log P 4.6), so it has little affinity for water and phase-separates at about 10 µg/mL. HPMCAS is amphiphilic and can interact with both drug and water. Its affinity for posaconazole rises with the A/S ratio and with falling pH, and its affinity for water does the opposite; H therefore interacts more with posaconazole in solution and partitions more into the drug-rich phase than L C48. From there the argument runs through the places where nucleation can start:

Two parts of this chain carry the conclusions. In the bulk, the authors hold that the blue shift "signifies" drug–polymer interaction and that these interactions "are expected to" slow nucleation C22·C40. In the drug-rich phase, they write that the high polymer content, especially with H, "serves to decrease molecular mobility, as well as decreasing the drug chemical potential via dilution", and that the extent of partitioning follows the same order as the supersaturation time in release C46·C47. In the conclusions this is put cautiously: increased polymer partitioning "is believed to be one factor important for reducing the drug crystallization tendency" C51. The authors place their results in line with earlier reports that polymer amphiphilicity matters for crystallization inhibition, as does polymer solubility (refs 33, 37, 58, 59) C37.

What it means for formulation

The practical message of the paper is a trade-off. Small-intestinal pH varies with location, subject and dosing, and the authors take pH 5.5–7.5 as the physiological extremes C32. Across this range the effect of pH on release and crystallization "is clearly highly dependent on the polymer grade" and "highlights trade-offs that may need to be made during formulation development" C33. L gives full release but almost no lasting supersaturation; H gives long supersaturation but full release only at pH 7.5; M gave "the optimum balance" in this study but still released poorly at pH 5.5 C34·C35·C36. For posaconazole, which nucleates within about 30 min near its amorphous solubility without an inhibitor, the authors see an "essential role" for a polymer with good inhibitory properties to keep supersaturation over the 2–4 h absorption window C31.

Note by the AI writer (not discussed in the paper): two further points follow from the data. First, the solution induction time ranked the grades in the same order as the supersaturation time in release, so it is a quick screen for inhibition; but it says nothing about release extent, which here was set by whether the polymer dissolves at the test pH. A grade screen therefore needs both an inhibition test and a release test at the lowest pH expected. Second, the inhibition ranking depended on drug loading: M held supersaturation for 3 h at pH 6.5 at 10 wt% DL but for only 10–15 min at 25 wt% DL R4. Ranking at one loading need not transfer to another.

Limitations

Limits stated by the authors. The authors write that further studies are needed on the specific nature of the interactions between HPMCAS and drugs in solution, and on the role of drug-rich droplet size and composition in the performance of ASDs in vivo C55. They also report that H could not be tested in solution at pH 5.5 and that the amorphous solubility without polymer could not be measured because posaconazole crystallized too fast C56. The droplet composition ignores water, which was not determined C45.

The AI writer's own judgement. The measured effects are solid and consistent: at pH 6.5 five independent solution measures (induction time, emission maximum, droplet size, ζ-potential and polymer depletion) rank the grades in the same order, and the release trade-off is visible directly in drug and polymer curves at three pH values and two loadings R3·R9·R11·R15·R16·R4·R5. The mechanism that links them is weaker. Hydrophobic interaction, nucleation at the droplet surface, and lower mobility and chemical potential inside the droplets are all inferred, not measured; the abstract states as a result that ionization "decreased the hydrophobicity of the polymer", which no experiment measured, and the droplet ζ-potential with L did not change with pH C40·C42·C46·C53·X8. Fluorescence intensities do not measure drug amounts here, because the polymer raises intensity and two slit settings were used X3·X4·X5. Only H clearly takes up polymer, and the link between partitioning and slower crystallization is a correlation over three grades that also differ in bulk interaction and solubility X6·X7. Crystallization during release was inferred from the fall in concentration only; no solid-state analysis of the precipitates or of the ASDs is reported X10. Replicate numbers and the type of error bar are not stated except for the induction times, the 25 wt% DL dose is misprinted, and the 10 µg/mL induction-time solutions may have been at or above the amorphous solubility X13·X1·X12. Finally, one drug and one bile-salt-free buffer were studied.

Conclusion

The paper shows, for posaconazole at 37 °C in fasted-state intestinal buffer, that HPMCAS chemistry acts in two opposite directions. A higher acetyl-to-succinoyl ratio and lower pH, both meaning fewer charged succinoyl groups, made HPMCAS a much stronger inhibitor: the induction time at pH 6.5 went from 50 min with L to 208 min with M and beyond 240 min with H, against 29 min without polymer R3. The same chemistry made the polymer less soluble, so H released fully only at pH 7.5 and M released poorly at pH 5.5, while L released at every pH but lost the supersaturation within the hour R4·R5. In the phase-separated solutions, higher-A/S grades gave smaller, less charged droplets and more polymer inside them, and the authors believe that polymer partitioning into the drug-rich phase is one factor in slower crystallization R11·R15·R17·C51. The mechanism behind these trends was inferred rather than measured; the trade-off itself was measured directly, and in the authors' words it "highlights trade-offs that may need to be made during formulation development" C33.