Why HPMCAS slows felodipine crystal growth less in acid: the shape of the polymer on the crystal surface

Background and motivation

Many new drug candidates dissolve poorly in water, and the paper cites an estimate that up to 80% of investigational drugs have suboptimal aqueous solubility (Babu 2011). One answer is an amorphous drug–polymer blend. The amorphous form has higher free energy than the crystal and no long-range order, so on dissolution it gives a supersaturated solution, one that holds more drug than the crystal's equilibrium solubility (Pouton 2006; Hancock 2000). Supersaturation raises the flux of drug across the gut membrane and therefore passive absorption (Takano 2010; Davis 1991). The authors name telaprevir (Kwong 2011) and vemurafenib (Bollag 2010) as approved drugs developed as amorphous formulations because the crystalline form could not give adequate clinical efficacy.

The difficulty is that a supersaturated solution carries a strong thermodynamic driving force for crystallization, and the solutions generated from amorphous solids "will typically crystallize very rapidly" (Hancock 2000). Crystallization returns the concentration toward the crystal solubility and the absorption advantage is lost, so additives that slow it are critical. They either disrupt nucleation, the birth of new crystals, or inhibit growth of existing crystals by adsorbing to growth sites and acting as a mechanical barrier (Kubota 2001; Ziller 1988; Black 1986). This paper is about the second route, growth inhibition by an adsorbed polymer.

What makes a polymer a good growth inhibitor is not settled. The introduction notes that polymer effects on inorganic crystals such as calcium carbonate (Kulak 2007; Colfen 1998) and the effects of small-molecule additives on organic crystals (Lahav 1987; Weissbuch 1991) have been widely studied, while polymer effects on small organic molecules crystallizing from water are "not widely understood". Two factors had been proposed: a match in hydrophobicity between polymer and drug (Zimmermann 2009; Ilevbare 2012, ref 20) and the ability of the polymer to form specific hydrogen bonds with the drug (Addadi 1982; Raghavan 2001). One observation fitted neither. In studies from the same group on ritonavir (Ilevbare 2012, ref 23; Ilevbare 2013, ref 41), several ionizable, carboxylated cellulose polymers were consistently more effective growth inhibitors at higher pH, where they were highly ionized, although they adsorbed to the ritonavir crystal to a similar extent at both pH values. Those authors speculated that a change in polymer conformation might be the reason.

There was good reason to suspect conformation. When a polymer is ionized, its charged groups repel each other and the chain extends; in the un-ionized state the chain coils, held together by intramolecular hydrogen bonds (Etika 2010; refs 24–26). Roiter and Minko (2005) had imaged such pH-driven transitions of poly(2-vinylpyridine) chains in water by atomic force microscopy (AFM). What had not been done was to image an inhibitor polymer adsorbed on a drug crystal in water, at pH values where its inhibition differs, and ask whether its shape on the surface accounts for the difference.

The stated objective is to investigate the conformation of polymers on the surface of a crystalline drug as a function of pH. The hypothesis is introduced with "It is hypothesized": pH changes the conformation of the adsorbed polymer at the solid–liquid interface, and this change affects its ability to inhibit crystal growth. The test system is felodipine, a poorly water-soluble drug, with hydroxypropyl methylcellulose acetate succinate (HPMCAS) grade LF. The paper gives the pKa of this grade as 5.5, so the authors chose pH 3 and pH 6.8, on either side of it; the two values also stand for the acidic stomach and the near-neutral small intestine. At both pH values they measured felodipine growth with and without HPMCAS and imaged the adsorbed polymer by AFM in liquid; AFM coupled with infrared spectroscopy (AFM-IR) confirmed its identity.

Background concepts

Supersaturation and the growth law. The driving force for crystallization from solution is the difference in the solute's chemical potential between the supersaturated and the saturated solution (Mullin and Sohnel 1977). The paper expresses it as the supersaturation ratio (eq 1):

In plain terms: $S$ says how many times more drug is dissolved than the crystal can hold at equilibrium. $C$ is the actual, supersaturated concentration and $C_{\mathrm{e}\mathrm{q}}$ the equilibrium solubility of the crystal. For felodipine, $C_{\mathrm{e}\mathrm{q}}$ is about 0.5 µg/mL at 25 °C, a value taken from the group's earlier felodipine study (Alonzo 2012), so the starting concentration of 4 µg/mL used in the growth experiments gives $S$ = 8. The overall growth rate is then written as a power law (eq 2):

In plain terms: growth speeds up as supersaturation rises, and the growth order sets how steeply. $R_G$ is the overall growth rate, $k_G$ the growth rate coefficient and $g$ the overall growth order, an empirically fitted number typically between 1 and 2 depending on whether transport of growth units to the surface (diffusion) or their integration into the lattice is the slower step (Judge 1995). The paper reports neither $k_G$ nor $g$; eq 2 serves to argue that a fixed rotation speed keeps $k_G$ constant between experiments, so that differences between runs come from the polymer.

Steps, growth units and growth sites. Growth requires a stable nucleus of critical size (Mullin 2001). Many growth mechanisms build on Kossel's step model (Kossel 1934), in which a crystal grows one molecular layer at a time: dissolved molecules, the growth units, attach at steps and kinks, the growth sites, and the steps sweep across the face. A polymer can slow this by competing for those sites. The paper takes one point from Mullin's textbook that matters for everything that follows: a polymer need not adsorb to every growth site; it only has to disturb the flow of growth layers.

Step pinning. Cabrera and Vermilyea (1958) proposed the pinning mechanism. Impurities, here adsorbed polymer, sit on the growth layer at an average distance $l$ apart and impede the advance of steps, which have to squeeze between them. How strongly growth is held back depends on the radius of the critical nucleus, ${\rho }_c$, relative to $l$. Following the Kubota–Mullin model (Kubota and Mullin 1995; Kubota 2001), the paper writes the growth rate with polymer as (eq 3):

In plain terms: the share of the original growth rate that survives falls as the adsorbed polymer units come closer together. $R_P$ and $R_0$ are the overall crystal growth rates with and without polymer. When $l$ is much larger than ${\rho }_c$, the ratio is close to 1 and the polymer does almost nothing; as $l$ shrinks toward ${\rho }_c$, the ratio falls toward zero. The paper draws the conclusion it needs from this: a smaller $l$ means a more effective polymer.

The effectiveness ratio. Inhibition is reported as (eq 4):

In plain terms: $E_g$ says how many times slower the drug grows when the polymer is present. A value of 1 means no effect, and the polymer is considered effective when $E_g$ is above 1. Because $E_g$ is the inverse of the ratio in eq 3, a smaller spacing $l$ between adsorbed polymer units gives a larger $E_g$. This is the bridge the authors later use between the images and the growth data.

Ionization and conformation of HPMCAS. HPMCAS is a cellulose backbone carrying methyl, hydroxypropyl, acetyl and succinoyl substituents (Figure 1b). The succinoyl groups end in a carboxylic acid, which makes the polymer ionizable. The pKa is the pH at which about half of these groups are ionized; for grade LF the paper gives 5.5, and from it the authors take HPMCAS to be "nearly completely ionized" at pH 6.8 and un-ionized at pH 3. The degree of ionization in the buffers was not measured. The paper also notes that HPMCAS is much less soluble at pH 3 than at pH 6.8, so concentrated pH 3 solutions are cloudy. The picture borrowed from polyelectrolyte studies is simple: ionized chains stretch out to keep their charges apart, un-ionized chains coil. The molecular weight of the grade used is 18 000 g/mol.

AFM imaging in liquid. In tapping-mode AFM a sharp tip oscillates over the surface and touches it briefly in each cycle. The height image records topography; the phase image records the lag of the tip's oscillation, which responds to material properties such as viscoelasticity (Schmitz 1997), so soft adsorbed polymer appears darker than hard crystal. The authors call phase imaging "ideal" for detecting adsorbed polymer. Root-mean-square (RMS) roughness, the root-mean-square height variation over an area, puts a number on how bumpy the layer is.

AFM-IR. AFM-IR couples an AFM with infrared spectroscopy. A pulsed, tunable infrared laser (an optical parametric oscillator, OPO) heats the spots that absorb; they expand briefly, and the cantilever resting on them oscillates in proportion (Harrison 2013). Scanning the wavelength gives a spectrum from a sub-micrometre domain; scanning the surface at one wavelength gives a chemical image. Here 1700 cm⁻¹ marks felodipine and 1720 cm⁻¹ the carbonyl band the authors assign to HPMCAS.

Methods in detail

The study splits into four contributions. First, the authors measured how strongly HPMCAS slows felodipine crystal growth at pH 3 and pH 6.8, using seed crystals on a rotating disc. Second, they imaged HPMCAS adsorbed on felodipine crystals in liquid at the same two pH values and turned the images into numbers: layer height, globule size and spacing, and molecules per globule. Third, they raised the pH over a sample adsorbed at pH 3 and imaged the same area again. Fourth, they confirmed by AFM-IR that the adsorbed material is HPMCAS. The growth data and the images are then joined by a qualitative reading of the pinning relation, eq 3.

Contribution 1: Measuring growth inhibition at pH 3 and pH 6.8

① Problem. The effect to be explained has to be measured on the same drug and polymer that are imaged; the ritonavir results came from a different drug. The read-out also has to be growth on an existing crystal surface, because the pinning argument concerns steps on a crystal. The authors therefore used seeded growth; whether some nucleation also took place during the runs is not addressed.

② What they did. Seed crystals were made by melting felodipine as supplied and recrystallizing the melt in an atmosphere of 75% relative humidity (RH) (Section 3.2).Card check: the paper describes the crystals three ways: melt recrystallized at 75% RH (Section 3.2), "recrystallized amorphous films" (Section 3.3), and recrystallization of amorphous material on exposure to buffer (Figure S1 caption); the route of the imaged crystals is unresolved (X2). The seeds were mounted on a rotating disc apparatus (RDA) at a constant 1000 rpm, which the authors say keeps $k_G$ of eq 2 constant between experiments (Bourne 1976; refs 36–38). A stock of 10 mg/mL felodipine in methanol was added to 50 mM phosphate buffer at pH 3 or pH 6.8 to give 4 µg/mL felodipine, an initial $S$ of 8. HPMCAS solutions at 5 µg/mL were prepared by dissolving the powder in the buffer and mixing for 24 h. Felodipine concentration was followed by its ultraviolet–visible (UV–vis) absorbance at 360 nm, every 10 s for 1 h, starting immediately after the supersaturated solution was made, at 25 °C; calibration solutions were made in methanol. Each condition was run in triplicate, and Figure 2 shows means with error bars of one standard deviation.

The desupersaturation profile, the fall of dissolved drug with time, "is considered to be" directly proportional to the overall growth rate of felodipine. Its slope was recorded as $R_0$ without polymer and $R_P$ with HPMCAS, and the two were combined by eq 4 into $E_g$. Because HPMCAS dissolves poorly at pH 3, the authors checked the growth solution by eye: concentrated pH 3 solutions were cloudy, while the 5 µg/mL solution was clear, "indicating" that the polymer was fully dissolved and not agglomerated at the test concentration (Section 3.2).

③ Why it works. Fixing the disc speed fixes the flow over the seed, so a change in slope can be put down to the polymer rather than to mixing. The ratio of two slopes measured the same way removes factors common to both runs, such as seed area, provided they really are the same. At 5 µg/mL, HPMCAS stays dissolved even at pH 3, so the comparison is between dissolved polymer in two ionization states.

④ How it differs from earlier work. The ritonavir studies (Ilevbare 2012; Ilevbare 2013) compared several carboxylated cellulose polymers and measured both growth and adsorption. This paper does not measure adsorption; it pairs the growth measurement with images of the adsorbed layer at the same two pH values. Felodipine growth had been studied before by the group with HPMC (Alonzo 2012), which supplies the solubility used here.

Contribution 2: Imaging and measuring the adsorbed polymer in liquid

① Problem. If the amount adsorbed does not explain the pH effect, the arrangement of the polymer on the crystal might. It has to be seen in water, because a dried layer can rearrange (the AFM-IR experiment later shows that drying made HPMCAS agglomerate), and it has to be described by numbers such as height and spacing to connect with eq 3.

② What they did. HPMCAS was dissolved at 0.2 mg/mL in phosphate buffer (pH 3 or 6.8) by stirring for up to 24 h. Melt-grown seed crystals were exposed to this solution on the RDA at 200 rpm for 2 h, removed without being allowed to dry, and imaged at once in the same liquid. Imaging used a MultiMode 8 AFM in tapping mode in fluid, with NPG-10 silicon nitride triangular probes (spring constant 0.24 N/m, radius of curvature 30 nm), a scan rate of 0.4 Hz and 512 × 512 pixels. Height and phase images were recorded together at room temperature, and incubation times, a term the paper does not define further, ranged from 3 to 5 h. A felodipine crystal without polymer served as the control (Figure 3c). AFM showed the same felodipine crystal morphology at both pH values with and without polymer, and no change in the drug surface over the time of the experiments.

The images were then measured. On a 1.5 µm × 1.5 µm area at pH 3, ImageJ gave the average radius $r$ of all globules and the average distance between them (Figure 5); cross-sectional analysis gave their average height $h$. Each globule was modelled as a spherical cap, the dome left when a sphere is cut by a plane (eq 5, after Walker and Beebe 2002):

In plain terms: a low dome with base radius $r$ and height $h$ has volume $V$; for a flat dome, with $h$ much smaller than $r$, the $3r^2$ term dominates and the volume grows mainly with the base area. From $V$, the polymer density given by the manufacturer and the molecular weight of 18 000 g/mol, the authors computed the number of HPMCAS molecules per globule; this step is described in words only, and the density value is not printed. RMS roughness was computed on 1.5 µm × 1.5 µm areas. Finally, angles between crystal faces were measured on AFM height cross-sections and compared with predicted angles between crystallographic planes, to identify the faces on which the polymer sat (Figure S1, Table T1).

③ Why it works. Phase contrast separates soft polymer from hard crystal, and the polymer-free control shows the bare crystal under the same conditions, so dark features can be assigned to adsorbed material. Height cross-sections resolve sub-nanometre differences, enough to compare a layer's thickness with one sugar ring. Imaging in the adsorption liquid keeps the polymer in the state it adopted at that pH.

④ How it differs from earlier work. Roiter and Minko imaged single synthetic polyelectrolyte chains. Here the substrate is a drug crystal and the polymer a pharmaceutical growth inhibitor, imaged in the buffers of the growth test. The ritonavir work could only speculate about conformation; this paper looks at the adsorbed polymer and attaches dimensions to it.

Contribution 3: Raising the pH under the microscope

① Problem. Two samples prepared separately can differ for reasons other than pH. The authors also wanted to know whether the adsorbed conformation is reversible: once HPMCAS has adsorbed as globules at pH 3, does raising the pH change its shape, or is the shape fixed at adsorption?

② What they did. HPMCAS was adsorbed at pH 3, and the surface was imaged right away in the pH 3 liquid. Sodium hydroxide (NaOH) was then added to raise the medium to pH 6.8, and the same 2 µm × 2 µm location was imaged again, in phase and in three-dimensional height (Figure 6; incubation 5 h). RMS roughness of a 1.5 µm × 1.5 µm area was compared at the two pH values.

③ Why it works. Imaging the same area removes sample-to-sample variation, so a change can be attributed to the pH step, and roughness turns it into a number. Because nothing was washed off between the images, the experiment also bears on whether chains move away from where they first adsorbed.

④ How it differs from earlier work. Polymer adsorption is usually regarded as irreversible, and the earlier work on ionizable polymers compared samples prepared at each pH. This experiment asks a different question: whether an adsorbed polymer can still change its shape in place when the medium around it changes.

Contribution 4: Identifying the adsorbed material by AFM-IR

① Problem. Dark spots in a phase image show that something soft lies on the crystal, not what it is. Chemical evidence that the adsorbed material is HPMCAS was needed.

② What they did. AFM-IR could not be run in liquid, so a separate sample was made directly on a zinc selenide (ZnSe) prism. A small drop of 5 mg/mL felodipine in methanol was left to evaporate on the prism, giving a smooth, thin crystalline film. The prism was then suspended in pH 6.8 buffer containing 0.2 mg/mL dissolved HPMCAS, with the film in contact with the solution, and the solution was stirred for 1 h; this was designed to mimic the adsorption step of the fluid-cell AFM samples. After removal, the prism was dried with nitrogen to remove the remaining liquid and any HPMCAS that had not adsorbed. A nanoIR AFM with C-450 silicon cantilever probes in contact mode recorded spectra from 1620 to 1800 cm⁻¹ in 4 cm⁻¹ steps with 128 laser pulses per wavelength, and chemical images at 1700 and 1720 cm⁻¹ at a scan rate of 0.1 Hz (Section 3.4).

③ Why it works. In this window felodipine shows bands between 1650 and 1700 cm⁻¹, while HPMCAS has a carbonyl band at 1720 cm⁻¹. A spot that absorbs at 1720 cm⁻¹ therefore contains polymer. Imaging the same area at both wavenumbers should give inverse patterns if the surface features are polymer lying on a drug layer, and a before-and-after image at 1700 cm⁻¹ shows what the film looked like without polymer.

④ How it differs from earlier work. The method follows the group's earlier AFM-IR study of griseofulvin nanocrystals (Harrison 2013). The authors are explicit about its limit in this paper: nitrogen drying made the HPMCAS agglomerate, so the AFM-IR result is used for chemical identification only and carries no information on conformation (Section 4.3).

Experimental analysis

Experimental setup. All work used one drug, felodipine, and one polymer, HPMCAS-LF, in 50 mM phosphate buffer at pH 3 or pH 6.8, under the conditions given above. The paper reports no adsorbed-amount measurement, no pH between 3 and 6.8, and no second polymer or drug.

Table A. Felodipine desupersaturation on rotating seed crystals (initial S of 8, 25 °C, 1000 rpm, 50 mM phosphate buffer). Concentrations read from Figure 2 (approximate); E_g from Table 1.

Main result 1: HPMCAS inhibits growth at both pH values, but much less at pH 3. Both $E_g$ values in Table A are above 1, "indicating" that HPMCAS reduces felodipine growth at both pH values (Section 4.1). The effectiveness at pH 3 is lower than at pH 6.8 by a factor of about 1.8, which the conclusions call "considerably diminished". The trend matches the earlier ritonavir results with several carboxylated cellulose polymers, where the lower effectiveness at low pH was not due to lower adsorption. A rough check from the end points in Figure 2, made by the card's extractor, gives about 1.3 and 2.3, close to the Table 1 values, so the ratios are consistent with the plotted curves. The absolute growth rates $R_0$ and $R_P$ are not printed.Card check: Figure 2 shows one no-polymer curve without a pH, so the pH comparison assumes that R₀ is the same at pH 3 and pH 6.8 (X3). Table 1 gives E_g without SD or n although runs were in triplicate, the time window used for the slopes is not stated, and Figure 2 ends at 50 min although data were recorded for 1 h (X11).

Note by the AI writer (not discussed in the paper): the pH 6.8 curve stays near its starting value for about 14 min before it begins to fall, while the other two curves fall from the start. A single slope over the run therefore mixes a period with almost no growth and a period of slower growth, and the size of $E_g$ at pH 6.8 depends on which window is used. The paper does not comment on this delay, and because the slope window is not stated, the reader cannot tell how the delay entered the value of 2.29.

Main result 2: the adsorbed polymer looks different at the two pH values. In the 2 µm × 2 µm phase images taken in liquid after about 3 h of incubation (Figure 3, phase scale 0–10°), HPMCAS adsorbed at pH 3 appears as distinct dark spots spread over the crystal surface, while HPMCAS adsorbed at pH 6.8 gives dark shading over the entire surface with no discrete spots. The polymer-free crystal shows no dark shading. The authors say the images "qualitatively confirm" a pH-dependent conformation at the solid–liquid interface: the pH 3 spots are distinct adsorbed globules, and the pH 6.8 shading is "indicative of" extended chains dispersed across the drug surface. Their explanation is charge repulsion in ionized HPMCAS and its absence in un-ionized HPMCAS. Table B puts numbers on the difference.

Table B. HPMCAS adsorbed on felodipine from 0.2 mg/mL and imaged in liquid at room temperature (Section 4.2, Figures 3–5). Values with "about" are from the text or read from the figures; the felodipine control without polymer shows no dark shading (Figure 3c).

What Table B means. At pH 6.8, the features of up to about 0.8 nm match the molecular diameter of glucose, about 0.75 nm, taken as a stand-in for an HPMCAS monomer; from this "it can be deduced" that the chains lie parallel to the surface and "may" each cover several growth sites. At pH 3, the polymer forms low domes holding about 46 molecules each, about 44 nm apart on average, roughly 50 times the felodipine molecular diameter of about 0.9 nm. The paper's central picture follows: at pH 3 most of the crystal surface between the globules is bare.Card check: the printed globule volume, 1.05 × 10⁻³ nm³, has a wrong exponent. Eq 5 with r = 15.4 nm and h = 2.98 nm gives about 1.12 × 10³ nm³, and only a volume near 10³ nm³ can hold about 46 molecules of 18 000 g/mol (at a density near 1.3 g/cm³, an extractor back-calculation); the remaining 1.05 versus 1.12 difference is unexplained (X1).Card check: the spacing is not defined as centre-to-centre or edge-to-edge. The Figure 4c profiles show peak-to-peak spacings of about 55 and 88 nm, outside the stated 25–50 nm range if read centre-to-centre, so 44 nm is best used as an order of magnitude (X9).

One instrument parameter limits how far the lateral numbers can be trusted. The probe tip has a radius of curvature of 30 nm, larger than the average globule radius of 15.4 nm. In AFM a tip wider than a feature broadens that feature's apparent width, while heights are not broadened in the same way. The paper does not mention a correction, so the card's extractor treats $r$, $V$ and the count of about 46 molecules per globule as probable overestimates (X6). The heights, 2.98 nm for the globules and up to about 0.8 nm for the pH 6.8 layer, are the more robust part of the comparison.

Main result 3: raising the pH flattens the globules in place. On the sample adsorbed and imaged at pH 3, the phase image shows distinct dark spots and the three-dimensional height image shows discrete raised globules (Figure 6a, c). After NaOH raised the medium to pH 6.8, the height image of the same area shows no compact globular features, while the phase image still shows dark spots, though less obvious ones (Figure 6b, d). RMS roughness of a 1.5 µm × 1.5 µm area fell by 29.3%; the absolute roughness values are not printed. The authors say the change is "especially clear" in the height images and that the roughness drop "quantitatively confirms" that the chains are no longer coiled at pH 6.8.

The remaining dark spots lead to a more careful argument. Adsorption is usually considered irreversible, but "it is possible" that chains coiled within a globule without being bound to the crystal diffuse away after the pH rise and adsorb elsewhere. Against this, Figure 6b still shows dark spots where the globules were, unlike a sample adsorbed directly at pH 6.8 (Figure 3b). They read these spots as dense, extended chains at the former globule sites ("It can be deduced"), and conclude that on ionization "it is not likely" that the chains migrate away within the time of the experiment; rather, they extend to cover growth sites near where they first adsorbed. Note by the AI writer (not discussed in the paper): the contrast between Figure 6b and Figure 3b suggests that the pH history of a crystal surface may matter, since a layer adsorbed in acid and then brought to pH 6.8 did not end up looking like a layer adsorbed at pH 6.8 from the start. Its effect on growth was not measured.

Table C. AFM-IR on a methanol-cast felodipine film on ZnSe exposed to 0.2 mg/mL HPMCAS at pH 6.8 and dried with nitrogen (Section 4.3; spectra from Figure 7, peak positions read from Figure 7, approximate; 5 µm × 5 µm chemical images from Figure 8).

Main result 4: the adsorbed material is HPMCAS. Table C shows that only the agglomerates carry the 1720 cm⁻¹ HPMCAS carbonyl band (Figure 7). Their weaker felodipine bands "can be attributed" to the laser passing through the drug layer under the polymer, which damps the drug signal reaching the tip. In the chemical images (Figure 8), the 1720 cm⁻¹ image is the inverse of the 1700 cm⁻¹ image after exposure, and the smaller features without 1720 cm⁻¹ signal are attributed to their being thinner than the drug layer. The authors call the result "clear evidence" that HPMCAS adsorbs to felodipine under their conditions. The test, however, was done only at pH 6.8 and on a dried film, so the identity of the pH 3 globules rests on phase contrast (X12).

Table D. Angles between faces of the felodipine crystallite in Figure S1, measured by cross-sectional analysis of the AFM height image and compared with predicted angles (from SI Table T1).

Further observation: adsorption on several crystal faces. The face angles in Table D match the predicted angles within 2.5°. With the faces indexed, HPMCAS was seen at pH 3 on the (1 1 −1), (1 −2 −2) and (1 −1 1) faces (Figure S1; a (0 1 1) face is also labelled). These faces expose different chemistry in the crystal-structure views of Figures S2–S4: methyl groups and oxygen atoms on (1 1 −1); chlorine atoms and methyl groups on (1 −1 1); methyl groups, amine groups and oxygen atoms on (1 −2 −2). No difference in adsorption density could be discerned between the three faces, "suggesting" that adsorption is nonspecific. This is a visual judgement without per-face densities.Card check: in Figure S1 the extractor sees discrete dark spots on the (1 −2 −2) and (1 1 −1) regions but a streaked texture without comparable spots on the (1 −1 1) region, so nonspecific adsorption is a qualitative observation (X8).

How the authors link the findings. The paper joins the growth data and the images through a chain of steps. The ionization states are taken as given, the conformations are imaged, and the link from conformation to growth rate is argued through eq 3.

The hedges along this chain differ. That coils combine into globules is stated as "It is now known"; that a larger $l$ in eq 3 lowers $E_g$ is "clear"; that extended chains block more than one site is something they "should" do, by adsorbing at several sites or by forming a barrier at neighbouring sites. The authors add that several molecules per globule "explains" why adsorption did not fall at low pH in the ritonavir study: the polymer is still on the surface, but bunched. The abstract states the conclusion with "it appeared".

Controls and what they rule out. Each control closes off one alternative. The no-polymer growth curve sets the reference rate $R_0$. The polymer-free AFM image shows that the dark shading at pH 6.8 is not a property of the bare crystal. The unchanged crystal morphology makes a change of crystal habit an unlikely cause of the growth-rate difference (it "is likely not" the cause). The stable drug surface in buffer argues against etching artifacts, and the clear 5 µg/mL pH 3 solution against precipitated polymer in the growth test. AFM-IR shows that the agglomerates at pH 6.8 are HPMCAS. The controls do not rule out a difference in adsorbed amount between the two pH values, or a solution origin of the pH 3 globules.

Comparison with the ritonavir results. The felodipine data reproduce the direction of the earlier ritonavir finding, where adsorption was similar at both pH values. The felodipine study offers a reason for that combination, that the same mass of polymer covers the surface differently, but borrows the adsorption result rather than measuring it for felodipine (X7). The pH effect is thus seen for two drugs, while the conformational explanation has been examined only for felodipine with HPMCAS-LF.

Discussion

Where the findings apply. The results come from one drug, one HPMCAS grade (LF) and two pH values in 50 mM phosphate buffer, with growth at 25 °C, one supersaturation (initial $S$ of 8) and one polymer concentration (5 µg/mL). The pinning argument concerns step growth on a crystal surface and says nothing about nucleation, which was not measured. The conformations were imaged after adsorption from 0.2 mg/mL at 200 rpm without supersaturated felodipine, while growth was measured at 5 µg/mL and 1000 rpm (X4), so the images need not show the exact layer present during the growth test.

What the paper shows and what it infers. Three findings rest on direct measurement: HPMCAS inhibits felodipine growth less at pH 3; the adsorbed polymer forms globules at pH 3 and a flat layer at pH 6.8; and the globules flatten when the pH is raised on the same sample. The step that joins them, that the globular shape causes the weaker inhibition, is argued through eq 3. Since ${\rho }_c$ was not estimated and no spacing was measured at pH 6.8, the equation gives only the direction of the effect: any pair of ${\rho }_c$ and $l$ with the same ratio gives the same $E_g$ (X7).

What is unresolved. First, the adsorbed amount of HPMCAS on felodipine was measured at neither pH, so a lower adsorbed amount at pH 3 cannot be excluded. Second, whether the pH 3 globules form on the surface or are aggregates already present in the 0.2 mg/mL solution was not tested; clarity was checked only at 5 µg/mL (X5). Third, the authors raise the question of what happens after a pH rise: do adsorbed chains extend over nearby sites, as they favour, or do chains that were not themselves bound desorb and readsorb elsewhere ("it is possible")? Fourth, the paper attributes chain extension to repulsion among a chain's own charged groups in Section 4.2, but to charge repulsion "between molecules" in the conclusions, so intra- versus intermolecular repulsion is unspecified (X10). Fifth, no pH near the pKa was tested, so it is unknown whether the change in conformation and in $E_g$ is gradual or abrupt across the ionization range.

Note by the AI writer (not discussed in the paper): the paper explains the coiling at pH 3 by the absence of charge repulsion and by intramolecular bonding, but it also reports that HPMCAS is poorly soluble at pH 3. Poor solvent conditions by themselves favour compact coils and aggregation of chains. The globules may therefore reflect the polymer's limited solubility as much as the loss of charge; the present data cannot separate the two.

What it suggests for further work. Three measurements would strengthen the argument most: the adsorbed amount per unit area at each pH, to show whether the same mass of polymer is simply arranged differently; a coverage or spacing measure at pH 6.8, so that eq 3 can be applied at both pH values; and imaging at the growth-test conditions. A series of pH values across the pKa, other HPMCAS grades and other carboxylated polymers would then test whether conformation tracks inhibition beyond two points, and a size measurement of the 0.2 mg/mL pH 3 solution would settle whether the globules come from solution.

What it means for formulation. The authors' own recommendation is that the insights "can be used" to identify polymers that both adsorb to the crystal surface and give a high degree of surface coverage, which "will ultimately" improve the delivery of poorly soluble drugs that rely on supersaturation; they also say the insights "can be applied" to other drug–polymer systems. The following points are the AI writer's reading and are not stated in the paper. For HPMCAS and other carboxylated polymers, growth inhibition may be weaker in acidic media, where the polymer is un-ionized, than in intestinal media, so a supersaturated solution that meets crystals in the stomach is less protected. Adsorption alone is a poor screen for a growth inhibitor: the ritonavir work found similar adsorption at both pH values, and the felodipine images suggest the difference lies in how the polymer is spread, so coverage per mass of polymer is the quantity to look for. Finally, the absolute effect here is modest ($E_g$ of 1.28 versus 2.29 at 5 µg/mL polymer), and whether the pH effect matters at the polymer levels reached when an amorphous dispersion dissolves was not tested.

Limitations

Limitations stated by the authors. The paper has no separate limitations section, but the authors state several caveats in the text. AFM-IR samples "could not be characterized in liquid", and drying with nitrogen agglomerated the HPMCAS, so the AFM-IR experiments serve chemical identification only, not conformation (Section 4.3). Polymer adsorption is "typically considered to be irreversible", yet after the pH rise "it is possible" that chains that were coiled in a globule without being adsorbed diffuse away and adsorb elsewhere (Section 4.2). The small differences between measured and predicted face angles are put down to instrument error and the limitations of top-down cross-sectional analysis (Table T1). HPMCAS is poorly soluble at pH 3; the authors address this for the growth test only, by noting that the 5 µg/mL solution was clear (Section 3.2). Their main claims are also hedged: the images "qualitatively confirm" the conformational difference, and the link to inhibition is introduced with "it appeared".

The AI writer's own judgement (1): the causal link is a two-point correlation. The card rates the paper as moderate overall. The growth data and the in-liquid images are direct, point the same way and have a polymer-free control, and the pH switch and AFM-IR add support. The central claim, that adsorbed conformation explains the pH effect, is weak to moderate: a correlation across two pH values, without adsorbed amount, pH 6.8 spacing or ${\rho }_c$ (X7). The images were made at 40 times the polymer concentration of the growth tests, at a different rotation speed and without supersaturated drug (X4). The pH 3 globules may be solution aggregates that adsorbed (X5), so the phrase "It is now known" overstates what the images show about how the globules formed.

The AI writer's own judgement (2): the numbers carry more uncertainty than the text suggests. The growth comparison rests on one no-polymer curve shown without a pH, $E_g$ values without reported uncertainty, and an unstated slope window, even though the runs were in triplicate (X3, X11). The 30 nm tip radius exceeds the 15.4 nm globule radius, so the lateral globule size and the count of about 46 molecules per globule are probably overestimates (X6). The printed globule volume has a wrong exponent, and even after correction a 7% difference from eq 5 remains unexplained (X1). The inter-globule spacing of about 44 nm is not defined, and the single profiles in Figure 4c do not fit the stated 25–50 nm range if read centre-to-centre (X9). The seed-crystal route is described three ways (X2), nonspecific adsorption and the solution state at pH 3 rest on visual checks (X8, X13), and AFM-IR was done only at pH 6.8 on a dried film (X12). None of this reverses the qualitative picture, but the numbers are best used as orders of magnitude and the causal link as a well-motivated hypothesis.

Conclusion

The paper shows, for felodipine and HPMCAS-LF in phosphate buffer, that the polymer slows crystal growth much less at pH 3 than at pH 6.8 ($E_g$ 1.28 versus 2.29), and that the adsorbed polymer looks different at the two pH values when imaged in liquid. At pH 3, below the pKa, HPMCAS sits on the crystal as separate globules about 15 nm in radius and about 44 nm apart; at pH 6.8 it spreads as a layer no thicker than about one sugar ring. Raising the pH over an adsorbed pH 3 sample flattens the globules in place and lowers the surface roughness by 29.3%, and AFM-IR confirms that the adsorbed material is HPMCAS. The authors connect these observations through step pinning: wide gaps between globules leave growth sites open, while flat, extended chains each block several sites.

The images and the growth data are direct; the causal link between them is an argument based on two pH values. The paper is most useful as a change of question: from whether a polymer adsorbs to how it covers the crystal, with that coverage depending on pH through the ionization of the polymer. For formulators using ionizable polymers, the practical message is to expect weaker growth inhibition where the polymer is un-ionized, and to look for polymers that cover the drug surface evenly at the pH where protection is needed.