Do intestinal bile salts stop drug crystals from growing? Seeded growth of telaprevir with six bile salts, as monomers and as micelles

Background and motivation

Many new drug candidates dissolve too poorly in water to be absorbed well, and a common formulation answer is to give the drug in a higher-energy form. In an amorphous solid dispersion (ASD), the amorphous solid and the supersaturated solution it produces raise the concentration gradient that drives the drug across the intestinal membrane (refs 1–3). That high energy is also the weakness of these products. The drug tends to crystallize, which threatens both the physical stability and the performance of the formulation (Taylor and Zhang 2016, ref 4). An established response is to add components that interrupt crystallization and so keep the solubility advantage (Guzman 2007, ref 5).

Crystallization from solution consists of two kinetic events. In nucleation the system overcomes the free-energy barrier to form a critical nucleus; in crystal growth, dissolved drug deposits on crystals that already exist (Mullin 2001, ref 6). The authors point out that most research on additives had addressed nucleation. Polymers are well-known nucleation inhibitors (refs 7–10), whereas surfactants were found to promote nucleation (Chen 2015, ref 11). Growth matters for amorphous products in particular, because seed crystals can form during formulation, manufacturing, storage or after the dose is taken. When seeds are present, secondary nucleation and growth face lower energy barriers than the first nucleation step, so supersaturation is consumed faster. The authors argue that slowing growth "may" slow this consumption and lengthen the time available for absorption (Introduction).

For polymers, growth inhibition is reasonably well described. Additives adsorb at growth sites and act as a mechanical barrier (Mullin 2001, ref 6; Kubota 2001, ref 14). The degree of inhibition has been linked to hydrophobic drug–polymer interactions (refs 15–18, including Ilevbare 2012, ref 15), to polymer conformation on the surface (Schram 2015, refs 19 and 20), to hydrogen bonding (refs 21 and 22) and to seed preparation (Schram 2016, ref 23). Surfactants, by contrast, can promote growth and cancel polymer inhibition (refs 23 and 24), so a surfactant cannot be assumed to help once crystals exist.

Bile salts are the surfactants that the body itself supplies to the intestine, and every supersaturating dosage form meets them in biorelevant dissolution testing and in vivo. They had been shown to inhibit nucleation of many structurally diverse drugs (Chen 2015, ref 12; Li 2016, ref 13). In the comparison by Li 2016, monomeric sodium taurochenodeoxycholate (STCDC) had the strongest induction-time inhibition for celecoxib, nevirapine and flibanserin among 13 bile salts. The authors' group had two companion studies on telaprevir. The first measured how bile salts change its solution thermodynamics and supplies the activity calibration used here. The second showed that all biorelevant bile salts slowed crystal formation from supersaturated solutions; monomers were effective, micelles opposed the monomer effect for some bile salts, and preliminary results hinted at growth inhibition. Both appear in the accepted manuscript only as placeholders, not in the reference list. The effect of bile salts on crystallization kinetics, and on growth in particular, had in the authors' words "not been explored in detail".

The paper therefore measures crystal growth of telaprevir in the presence of six bile salts that are common in human intestinal fluid (Riethorst 2015, ref 37; Hofmann and Small 1967, ref 38). These are the trihydroxy salts sodium taurocholate (STC) and sodium glycocholate (SGC), and the dihydroxy salts sodium taurodeoxycholate (STDC), sodium glycodeoxycholate (SGDC), STCDC and sodium glycochenodeoxycholate (SGCDC). Each is tested at 1.86 mM, where most of the bile salt is taken to be monomeric, and at 12 mM, where micelles are taken to dominate. The main tool is a seeded desupersaturation experiment at the same activity-based supersaturation in every medium, with seeds grown in the vessel by a proposed "in-situ common history seeding" method. The authors add the supersaturation left after 15 h, the zeta potential of the seeds, and a calculation that separates growth inhibition from nucleation inhibition.

Background concepts

Supersaturation measured as activity. In plain buffer, supersaturation is the dissolved concentration divided by the crystalline solubility. With micelles present this ratio misleads, because drug inside micelles does not add to the driving force for crystallization; that force is set by the thermodynamic activity, which follows the free drug. The authors therefore compare all media at the same activity-based supersaturation δ, defined relative to a solution in equilibrium with crystalline telaprevir. They measure it in a side-by-side diffusion cell with a regenerated cellulose membrane (6–8 kDa cut-off), across which the drug mass flow rate is proportional to activity (Equation 1):

Here $F$ is the drug mass flow rate across the membrane from the solution of interest, $F^{\circ }$ the mass flow rate for the standard state (a solution in equilibrium with crystalline telaprevir), $a$ and $a^{\circ }$ the corresponding drug activities, and δ the supersaturation (dimensionless). In plain terms: if drug crosses the membrane five times faster from the test solution than from a crystal-saturated solution, its activity is five times higher and δ is 5, whatever extra drug the micelles hold. The flux values are not printed; δ comes from a calibration curve in the first companion manuscript. Table 2 shows the practical effect. To reach δ 2.4, drug was added to 12.4 µg/mL in buffer, with trihydroxy bile salts and with 1.86 mM dihydroxy bile salts, but to 20.2 µg/mL with 12 mM dihydroxy bile salts; for δ 12.5 the figures are 65.2 and 145.2 µg/mL. The ratios 12.4/2.4 and 65.2/12.5 both imply a crystalline solubility in buffer of about 5.2 µg/mL, which the paper does not print. No concentration is given for δ 7.9.

Nucleation, growth and secondary nucleation. Primary nucleation creates the first crystals in a clear solution; growth adds molecules to existing crystals; secondary nucleation creates new crystals because crystals are already present. In a seeded experiment the fall in dissolved concentration (desupersaturation) counts all mass leaving solution, so it equals growth only if no new crystals form. To tell the cases apart, the authors follow the extinction at 370 nm, where telaprevir does not absorb, so a rise means more scattering particles (Ilevbare 2013, ref 17). In an unseeded solution, the induction time is the time until crystals are first detected. It reflects both the formation of nuclei and their growth to a detectable size, so a long induction time can mean few nuclei or slow growth.

Common-history seeds. Crystals prepared in different environments grow differently (Schram 2016, ref 23). White et al. (1998, ref 25) introduced the common-history (CH) seed: a batch of crystals that nucleated at the same time and then shared the same growth history, meaning the same temperature, supersaturation and duration. For such seeds the shape of the crystal size distribution is reported to be kept as the crystals grow (refs 26 and 27). The authors' version nucleates the seeds in the same vessel and buffer that are then used for the growth measurement.

Bile-salt structure and aggregation. All six bile salts share a steroid ring system with a side chain conjugated to taurine or glycine (Table 1). They differ in which hydroxyl groups the steroid carries and in the head group: taurine conjugates end in a sulfonate, glycine conjugates in a carboxylate. Table A reproduces the substitution pattern as printed.

Table A. Substituents of the six bile salts on the steroid scaffold (from Table 1; class from the text).

Bile salts are present as single molecules (monomers) at low concentration and also form micelles above a critical concentration. Citing the companion studies, the authors take 1.86 mM as a concentration at which most bile salt is monomeric, and 12 mM as one at which micelles are "most likely" the dominant species. No critical micelle concentration is given, and the aggregation state was not measured in this paper.

Zeta potential. The zeta potential is the electrical potential of a particle at the plane where it shears against the liquid, obtained from how fast it moves in an electric field. Because telaprevir is neutral and bile-salt monomers are anions in the pH 6.5 buffer, adsorbed bile salt should make the zeta potential of the crystals more negative; the authors use this as a qualitative indication of adsorption (refs 32 and 33), not as an adsorbed amount.

Methods in detail

The work splits into three contributions. The first is an assay: seeds grown in place from a supersaturated solution, regeneration of supersaturation to a fixed activity, and measurement of the initial desupersaturation rate, with turbidity used to decide whether that rate reports growth alone. The second is the bile-salt screen run with this assay. Five bile salts were measured by rate and three by the supersaturation left after 15 h, each at a monomeric and a micellar concentration, and the zeta potential of the seeds was measured with all six. The third is a calculation that combines the seeded rates with unseeded induction times from the companion study, to estimate how strongly each bile salt inhibits nucleation as distinct from growth. The flow below shows how the parts connect.

Contribution 1: In-situ common-history seeds and a seeded desupersaturation assay at equal activity

① Problem. A fair measurement of growth inhibition needs three conditions. The seeds should resemble crystals that form in a supersaturated solution, since seed preparation changes how crystals grow (ref 23). Media with and without micelles must be compared at the same driving force, which a fixed total concentration does not ensure. And the rate must be shown to reflect growth, not growth plus new crystals.

② What they did. Methanolic telaprevir stock (12 mg/mL) was titrated into 50 mL of 50 mM sodium phosphate buffer, pH 6.5, to 60 µg/mL, and the solution was stirred at 300 rpm and 37 °C for a fixed 6 h. Crystals formed and the concentration fell back to around the crystalline solubility; in the example trace of Figure 2 it falls from about 63 to about 8–10 µg/mL. The seeds stayed in the vessel. Bile salt, where used, was then added at 1.86 or 12 mM and left to equilibrate (time not printed), and supersaturation was regenerated with more stock to δ 2.4, 7.9 or 12.5 (Table 2). A fiber-optic UV dip probe (SI Photonics) followed the concentration at 270 nm, scanning every minute for 16 h. Second-derivative spectra (SIMCA 13.0.3) reduced the effect of light scattered by seeds and micelles (standard curve R² above 0.99), and the extinction at 370 nm was recorded alongside as the turbidity signal.

The desupersaturation rate $R$ is the initial slope of the concentration–time curve, taken over the first 30 min at δ 2.4 and over the first 10 min at the higher levels.Card check: the Methods print the 30 min window as applying to "δ=2.7"; no δ 2.7 run exists and every table and figure uses δ 2.4, so 2.7 is read as a misprint (X2). Rates are given in µg/min without a volume basis. The Figure 3 slope at δ 12.5 (about 1.8 µg/mL per min) times the 50 mL volume gives about 90 µg/min, matching the printed 89.2 µg/min, so the rates are per vessel (X12). From the rates the authors define two effectiveness ratios. The first is the growth-inhibition effectiveness (Equation 2):

Here $R_0$ and $R$ are the initial desupersaturation rates without and with bile salt, measured at δ 2.4. In plain terms: $E_G$ says how many times slower the seeds grow when the bile salt is present; a value of 1 means no effect, and a bile salt with $E_G$ above 1 is counted as an effective growth inhibitor. The second is the effectiveness against the combined kinetics (Equation 3):

The expression is the same, but the rates are taken at δ 7.9 or 12.5, where secondary nucleation also occurs. In plain terms: $E_C$ says how much the bile salt slows the total loss of drug by growth and by new crystals together; again, a value above 1 marks an effective inhibitor.

For scanning electron microscopy (SEM), about 10 µL of seeded solution taken after 6 h was dried on glass, platinum-coated and imaged at 5 kV; supplier crystals were imaged as received, and seeding in water served as a control against buffer salts.

③ Why it works. Every run returns to about the same concentration after 6 h, so the authors take the seed mass to be the same in every run. The buffer-only rates were reproducible, which they read as a sign that the method controls seed properties ("imply"). Regenerating to a fixed δ, not a fixed concentration, means that a bile salt which only solubilizes drug in micelles cannot appear to inhibit growth by lowering the free drug. The turbidity check sorts the δ levels: at δ 2.4 no change was seen and desupersaturation is treated as growth ("can be assumed"), while at δ 7.9 and 12.5 turbidity rose, taken as secondary nucleation. Note by the AI writer (not discussed in the paper): seed mass is inferred, not weighed, and seed number, size and surface area were not measured, so rates cannot be expressed per unit crystal area.

④ How it differs from earlier work. Earlier bile-salt studies (refs 12 and 13 and the companion study) reported on nucleation or crystal formation in unseeded solutions, where nucleation and growth are mixed. Here growth is isolated on existing seeds, which are produced in situ in the same buffer rather than taken from a bottle (Figure 4), and media are compared on an activity basis, which matters because micelles change the link between concentration and driving force.

Contribution 2: Screening six bile salts as monomers and micelles by rate, residual supersaturation and surface charge

① Problem. Knowing that bile salts delay crystallization of unseeded solutions does not tell whether they slow the growth of crystals that are already present. Nor does it tell whether all bile salts behave alike, or whether micelles act as the monomers do. A 10 or 30 min initial rate also does not show whether supersaturation survives for the hours that matter for absorption.

② What they did. Five bile salts (STC, SGC, STDC, SGDC and STCDC) were run by the seeded assay at 1.86 and 12 mM, giving $E_G$ at δ 2.4 and $E_C$ at δ 12.5 (Figure 5). The 1.86 mM salts were also run at δ 7.9, and the rate ratio $R_0/R$ was plotted against δ (Figure 7). SGCDC was left out of the rate measurement because it may form a complex with telaprevir (first companion manuscript) and was assessed only by residual supersaturation. For that, solutions regenerated for δ 12.5 crystallized for 15 h and were then placed in a side-by-side diffusion cell (PermeGear). The flux ratio against a crystal-saturated standard gave δ (Equation 1), reported as a "residual effective free drug concentration" in µg/mL for STC, STCDC and SGCDC at both concentrations (Table 3). Cell volumes, sampling times and replicate numbers are not printed.

For the zeta potential (Malvern Nano ZS), seeds grown overnight in 10 mL of buffer at 37 °C received 1.86 mM bile salt 30 min before measurement; seeds without bile salt were the reference, and all six bile salts were measured. The printed method also describes the seed buffer as already containing 1.86 mM bile salt, so the final concentration (1.86 or 3.72 mM) is unclear (X15).

③ Why it works. Running the same bile salt at 1.86 and 12 mM at the same δ separates the aggregation state from the driving force. Any difference then reflects what micelles do to growth, not how much drug they take up. The 15 h residual integrates growth and secondary nucleation over hours rather than minutes. Because it is measured by membrane flux, it counts free drug only, so drug held in micelles does not appear as dissolved. The zeta-potential test rests on charge alone: telaprevir is neutral and bile-salt monomers are anionic in the buffer, so a negative shift when bile salt is added points to bile salt at the crystal surface.

④ How it differs from earlier work. Bile-salt research had focused on nucleation, and the growth inhibitors studied in detail had been polymers (refs 15–20 and 23). This paper extends growth-inhibition measurements to endogenous surfactants and adds the aggregation state as a variable. It also sets the bile-salt values beside published polymer data: the authors call $E_G$ of monomeric STC, SGC and SGDC comparable to reported values for commercial polymers with poorly soluble drugs (refs 17 and 28).

Contribution 3: Separating growth inhibition from nucleation inhibition

① Problem. In the companion study the unseeded induction time of telaprevir was about 30 min without bile salt and more than 16 h with monomeric bile salts. Such a delay can come from fewer nuclei or from nuclei that cannot grow, and the authors write that "it is possible" the long induction times were partly due to strong growth inhibition.

② What they did. Following refs 28 and 36, they estimated the ratio of nucleation rates with and without bile salt from the seeded rate ratio and the unseeded induction times (Equation 4):

Here $J$ and $J_0$ are the nucleation rates with and without bile salt, $R_0/R$ is the seeded desupersaturation rate ratio (taken from Figure 5(b)), and $t_{u,0}$ and $t_u$ are the unseeded induction times without and with bile salt, in minutes. In plain terms: part of the longer induction time is explained by slower growth, which the first factor accounts for; whatever delay remains is attributed to slower nucleation. They then define the nucleation-inhibition effectiveness (Equation 5):

In plain terms: $E_J$ says how many times more slowly nuclei form with the bile salt, and a value above 1 marks an effective nucleation inhibitor. For the monomeric bile salts the authors used $t_{u,0}$ of about 30 min and $t_u$ of 960 min; the measured induction time exceeded 16 h, so 16 h was entered. With these inputs the induction-time ratio is 32, and the two equations reduce to $E_J={32}^4/(R_0/R{)}^3$, with 32⁴ about 1.05 × 10⁶. For STCDC and SGCDC, which showed no desupersaturation, the authors set the rate ratio to 100. Then $E_J$ is about 1.05, which the paper calls "minimal" and which Figure 8(c) plots at about 1. The induction times used for the 12 mM points are not printed.

③ Why it works. Note by the AI writer (not discussed in the paper): the paper does not derive Equation 4 and cites refs 28 and 36 for it. The reasoning in that literature is that crystals become detectable when their total volume reaches a small fixed fraction. That volume rises with the nucleation rate, with the cube of the growth rate and with the fourth power of time. A bile salt that slows growth therefore lengthens the induction time even if nucleation is unchanged, and the formula removes that part; the seeded rate ratio stands in for the ratio of growth rates if seed loading is equal. As a check, $E_C$ of about 14.4 for STC gives about 350, and about 1.5 for SGC gives about 3.1 × 10⁵, close to the printed "around the range of 400-340,000" within plot-reading precision.

④ How it differs from earlier work. Ref 28 (Alonzo 2012) used this decomposition for HPMC and felodipine, and ref 36 (Verdoes, Kashchiev and Van Rosmalen 1992) for calcium carbonate. Here the seeded rates and the induction times come from two studies run "under similar experimental conditions", not from one experiment. Two details matter. First, the rate ratios entered are the δ 12.5 values of Figure 5(b), that is $E_C$, as the plotted $E_J$ values confirm, yet the text speaks of "assuming E_G to be 100" and Figure 8 plots $E_J$ against $E_G$ from δ 2.4 (X13). Second, a note by the AI writer: $t_u$ was capped at 960 min although the true value was longer, and $E_J$ rises with its fourth power, so the 1.86 mM $E_J$ values are lower bounds. For STCDC, the value of about 1 follows from two assumed numbers, the cap and the rate ratio of 100.

Experimental analysis

Setup in brief. One drug (telaprevir, neutral in the buffer), one medium (50 mM sodium phosphate, pH 6.5, 37 °C, 300 rpm) and one kind of seed; varied were the bile salt (none or one of six), its concentration (1.86 or 12 mM) and the regenerated δ (2.4, 7.9 or 12.5). Replicate numbers and error-bar definitions are not printed, and apart from the buffer-only rates and Table 3 all values are read from plots, so they carry "about". HPMCAS-MF is listed under Materials, but no experiment with it is reported (X16).

The reference: seeds without bile salt. Table B collects the buffer-only data that every ratio is divided by.

Table B. Seeded desupersaturation of telaprevir without bile salt. Rates from the text (n not printed); concentrations read from Figure 3 and extinction read from Figure S1, approximate; the per-mL rate is the AI writer's division by 50 mL.

At δ 2.4 the concentration fell slowly, from about 11.3 to 8 µg/mL in 30 min, and the turbidity stayed flat: the drug left solution by growing on the seeds. At δ 7.9 and 12.5 the initial rates were about 11 and 17 times higher, and the extinction at 370 nm rose within about an hour of regeneration. At δ 12.5 more than half the drug left solution in 30 min. So δ 2.4 reports growth, and the higher levels growth plus secondary nucleation. The plotted starting concentrations (about 68.5 and 11.3 µg/mL) differ slightly from the Table 2 targets of 65.2 and 12.4 µg/mL; the plots show measured starts (X6).

What the seeds look like. In Figure 4, the supplier crystals (imaged at 2 400× and 1 200×) are irregular faceted particles and fragments up to about 30–40 µm. The in-situ seeds (5 000× in water, 8 000× in buffer) are bouquet-like aggregates about 15–25 µm across of rod- or blade-like crystals a few µm long.Card check: the Methods state 5000–30000×; the Figure 4 data bars show the values given here (X5). The authors say the difference "can potentially affect crystal growth behavior", but growth on supplier crystals was not measured. Because the seeds aggregate, growth was taken from the fall in solution concentration, not from imaging single crystals (Mullin 2001, ref 6).

Main result 1: the bile salts differ widely, and STCDC stops growth. Table C gathers the effectiveness ratios and the zeta potentials.

Table C. Effectiveness of the bile salts against seeded desupersaturation, and seed zeta potential (read from Figures 5(a), 5(b), 7 and 6, approximate; ± is half the drawn error bar; the −7.7 mV reference is from the text).

Four points follow from Table C. First, at 1.86 mM four of the five bile salts run by rate slowed growth, and STDC showed what the authors call "minimal effectiveness". The $E_G$ of monomeric STC, SGC and SGDC lies around 1.4–3.2, which the authors call comparable to commercial polymers with poorly soluble drugs (refs 17 and 28). Second, with monomeric STCDC no desupersaturation occurred at any δ, so its bars run off the axes and carry no number. The authors first call it "the extreme case", then state that it "blocked" growth, a complete suppression they note is common with impurities or additives (refs 29–31). Third, at 12 mM the four modest inhibitors stayed in a similar range, which "suggested" to the authors that micelles have minimal impact on their growth inhibition, while STCDC lost almost all of its effect ($E_G$ about 1.4, $E_C$ about 2.1); its micelles "appeared to negate" the monomer inhibition. By the paper's own criterion, 12 mM SGDC ($E_G$ about 1.0) is not a growth inhibitor, and 12 mM STC carries a wide error bar, so "comparable" holds only within large error bars (X9).

Fourth, STC behaves differently from the others as δ rises. Its rate ratio climbs from about 3.2 at δ 2.4 to about 9.8 at δ 7.9 and about 14.4 at δ 12.5, while SGC, STDC and SGDC stay at about 1.2–3.8 (Figure 7). The authors read this as "suggesting impact on secondary nucleation". They add that with 1.86 mM STC or STCDC, turbidity at δ 7.9 and 12.5 gave no evidence of secondary nucleation, so the two "appear to inhibit telaprevir secondary nucleation"; the Conclusions upgrade this to "were found to". The turbidity traces with bile salt are not shown (X10).

The zeta potential moved from −7.7 mV for seeds alone to about −38 to −50 mV with each bile salt, roughly a 5–6-fold rise in magnitude, "suggesting that bile salts interact with telaprevir crystal surfaces". The authors found no direct correlation with inhibition, and the table shows why: weakly inhibiting SGDC and growth-blocking STCDC both reach about −50 mV. The mechanism, they write, "is not well understood". They cite two dipyridamole studies: adsorption of taurocholate on active growth sites was speculated (Mithani, ref 34), and 2.2 mM STC shifted dipyridamole growth from mass-transfer control to a hybrid of mass-transfer and surface-integration control (Abbou Oucherif, ref 35).

Main result 2: the time course over four hours. Figure S2 is the only place where whole concentration–time curves with bile salts are shown. Table D reads them out.

Table D. Seeded desupersaturation of telaprevir over 240 min with 1.86 mM bile salt (read from Figure S2, approximate).

The STCDC traces stay flat for four hours, while buffer at δ 12.5 loses most of its drug within the first hour. STC holds the concentration longer but still falls to about 27 µg/mL. At δ 2.4 the STDC trace lies at or below the buffer trace for about the first 120 min, although Figure 5(a) gives it an $E_G$ of about 1.2, so STDC is at most a marginal growth inhibitor (X7). The ranking matches the initial rates, and the STCDC block is not a short delay.

Main result 3: what remains after 15 hours. Table E reproduces Table 3 of the paper.Card check: the printed caption of Table 3 repeats the Table 1 title ("Chemical structure of bile salts"); the table is identified by its content and by the text (X1).

Table E. Residual effective free drug concentration after 15 h of crystallization, measured by diffusion cell (from Table 3; runs regenerated for δ 12.5; n not printed). No buffer-only, SGC, STDC or SGDC residual is reported.

With 1.86 mM STCDC or SGCDC almost all the drug was still free after 15 h; with 12 mM of either, and with STC at both levels, most supersaturation was consumed. The authors conclude ("it is evident") that monomeric STCDC and SGCDC are highly effective inhibitors of telaprevir growth and secondary nucleation, and that their micelles "appear to negate" the monomer effect.

Three points limit how far Table E can be pushed. The authors call residual supersaturation "directly related" to $E_C$, but for STC the higher-$E_C$ condition (1.86 mM, about 14.4) leaves less free drug than the lower one (12 mM, about 5.6); the relation holds for the STCDC and SGCDC contrasts only (X8). The SGCDC runs start from 95.2 µg/mL, which matches no δ level, so SGCDC can be compared with STCDC only qualitatively.Card check: Table 2 gives 65.2 µg/mL (1.86 mM) and 145.2 µg/mL (12 mM dihydroxy) for δ 12.5; 95.2 µg/mL appears nowhere else (X4). And the residuals are free drug concentrations, not δ; the solubility needed for conversion is not printed (X17). Note by the AI writer: the last column divides a free concentration by a total one, so for 12 mM STCDC, whose 145.2 µg/mL includes drug held in micelles, the low percentage partly reflects the basis. The course between about 4 and 15 h is not shown (X18).

Main result 4: growth inhibition against nucleation inhibition. Table F lists the values plotted in Figure 8.

Table F. Nucleation-inhibition effectiveness against growth-inhibition effectiveness (read from Figure 8, log axes, approximate). The last column inverts Equations 4 and 5 with t_u,0 = 30 min; it is the AI writer's arithmetic, since the paper does not print the 12 mM induction times.

For the four monomeric bile salts other than STCDC, $E_J$ ranges from about 360 to 3 × 10⁵, while $E_G$ stays at about 1.2–3.2. The authors' reading, introduced with "That is to say", is that their ability to hold supersaturation by growth inhibition is "much weaker" than their ability to inhibit nucleation. Monomeric STCDC sits at the opposite corner: with its rate ratio set to 100, $E_J$ is about 1, and the long induction times "can be due" to growth inhibition, since "it is possible that nucleation did take place" but the nuclei could not grow to a detectable size. The authors extend this, as "a possible contributing factor", to STCDC giving the longest induction times for three other drugs in Li 2016.

At the micellar level the authors group the bile salts by structure, a grouping they introduce with "In general" and "some structural dependence". Trihydroxy STC and SGC remain strong nucleation inhibitors at 12 mM ($E_J$ about 6 × 10³ and 6 × 10⁵), and their micelles do not appear to change the inhibition significantly (Figure 8(a)). The deoxycholates STDC and SGDC, which lack the hydroxyl at R₃, appear to be strong nucleation inhibitors as monomers, but at 12 mM $E_J$ falls to about 70–85. Their micelles appear to negate the monomer inhibition, as in the companion study (Figure 8(b)). STCDC (and SGCDC), lacking the hydroxyl at R₄, is an extreme growth inhibitor as a monomer, and its micelles likewise negate the effect (Figure 8(c)).Card check: the Figure 8 caption calls panel (c) "trihydroxy bile salt STCDC"; Table 1 (R₄ = H) and the text class STCDC as dihydroxy (X3). Inverting the equations (last column of Table F) gives induction times of roughly 2–3 h for the dihydroxy micelles, against the 960 min cap for STC and SGC; these times are not printed, so the 12 mM $E_J$ values cannot be reproduced from this paper (X14).

Comparisons across bile salts. Note by the AI writer (not discussed in the paper): there is no consistent difference between taurine and glycine conjugates (STC above SGC, STDC close to SGDC, STCDC and SGCDC both holding supersaturation). The hydroxyl pattern separates the bile salts better, and within the dihydroxy group the position of the missing hydroxyl decides: the R₃-lacking deoxycholates barely slow growth, the R₄-lacking chenodeoxycholates block it.

Controls and what they rule out. The buffer-only rates are the denominator of every ratio. Flat turbidity at δ 2.4 rules out significant secondary nucleation under the growth condition; seeding in water rules out buffer salts in SEM; regeneration to equal δ rules out micellar solubilization as the reason for a slower rate. Missing are a buffer-only 15 h residual and residuals for SGC, STDC and SGDC, so Table E has no uninhibited baseline; the turbidity traces with bile salts; any particle count; and growth on supplier crystals, so the case for common-history seeds rests on SEM images alone.

Discussion

Where the approach applies. The assay (seeds grown in place, supersaturation regenerated to a fixed activity, initial rates read where turbidity stays flat) can serve any drug and additive where the drug is UV-detectable and crystallizes readily in the test medium. The equal-activity design matters most when the additive forms micelles. The findings are narrower: one neutral drug, single bile salts, pH 6.5 phosphate buffer at 37 °C, one seed preparation, and two concentrations assigned to monomer and micelle states from the companion studies. Note by the AI writer: the ranking may differ for another drug, because growth blocking by an adsorbed additive depends on the surfaces of the particular crystal.

What is unresolved. The mechanism is open: zeta potential shows that all six bile salts reach the crystal surface but does not separate a salt that blocks growth from one that barely slows it. Why micelles of the dihydroxy salts cancel the monomer effect was not examined, and no critical micelle concentration or adsorbed amount is given. Whether nuclei form and stall with monomeric STCDC was not tested. SGCDC's growth inhibition is inferred from the residual alone, which its possible complexation could also affect.

What the authors draw for formulation. Residual seed crystals are often unavoidable in amorphous formulations, so the ability to slow growth significantly affects the supersaturation available for absorption. For telaprevir, supersaturation "is possible to be maintained" in vivo even with residual crystallinity in the ASD, owing to bile-salt growth inhibition. If surface poisoning, reported for polymers (ref 23), also applies, bile salts used as excipients should have a larger effect ("is expected"). Their concluding statement is that "bile salts are not interchangeable from a crystallization inhibition standpoint", and they propose studying bile-salt mixtures for better biopredictive dissolution media.

What the AI writer reads from it for formulation (not stated in the paper). Taurocholate, the bile salt of most simple biorelevant media, was only a moderate growth inhibitor here, so a single-bile-salt medium may misjudge how fast seeds grow in intestinal fluid, which also contains chenodeoxycholate conjugates. More bile salt is not automatically protective: at micellar levels the strongest inhibitors lost their effect. And because nucleation and growth inhibition differed by orders of magnitude for the same bile salt, an unseeded induction-time screen does not tell how well a medium or excipient will hold supersaturation once crystals are present; a seeded test does.

Further work the data suggest. Note by the AI writer: adsorption isotherms with and without micelles would test whether micelles reduce surface coverage; particle counting in the 1.86 mM STCDC runs would test whether nuclei form and stall; and dissolution of a telaprevir ASD with a known crystal fraction, in single and mixed bile-salt media, would test the formulation claim directly. Other drugs would show whether the chenodeoxycholate effect is specific to telaprevir.

Limitations

Limitations stated by the authors. The paper has no limitations section, but it states caveats along the way. The growth-inhibition mechanism of bile salts "is not well understood" and more detailed studies are "required". At 12 mM, micelles are "most likely" the dominant species, an assignment taken from the companion studies. The differently shaped supplier crystals "can potentially affect crystal growth behavior", their reason for in-situ seeds. SGCDC could not be run by desupersaturation because of possible complexation. The nucleation estimate uses induction times from a separate study under "similar" conditions, 960 min for times beyond 16 h, and an assumed rate ratio of 100 where no desupersaturation was seen. The zeta potential is only a "qualitative indication" of adsorption. Key statements carry hedges: "appear to" for secondary-nucleation inhibition, "appear to negate" for micelles, and "it is possible" for growth inhibition behind long induction times.

The AI writer's own judgement (1): what holds and what is inferred. The extractor rates the evidence as moderate overall. The headline is solid: monomeric STCDC stopped desupersaturation at three δ levels, its 240-min traces are flat, 60.5 of 65.2 µg/mL was still free after 15 h, all against a buffer control, and the loss of the effect at 12 mM shows in $E_G$, $E_C$ and the residual. Most other claims rest on weaker ground. The ratios and zeta potentials exist only as plotted bars without replicate numbers or error-bar definitions. Secondary-nucleation inhibition rests on turbidity traces that are not shown, yet the Conclusions state it as found (X10). SGCDC is called an "extremely effective" growth inhibitor although its growth rate was never measured; the claim rests on a residual that cannot separate growth, nucleation or complexation, from a starting concentration that matches no δ level (X11, X4).

The AI writer's own judgement (2): the derived quantities. The $E_J$ values pair δ 12.5 rate ratios with induction times from another study capped at 960 min, and are plotted against $E_G$ from δ 2.4, so the two axes of Figure 8 come from different supersaturation levels (X13). The 12 mM values cannot be reproduced here (X14), and the $E_J$ of about 1 for STCDC follows from the assumed rate ratio, so it cannot by itself show that STCDC leaves nucleation untouched. The stated relation between residual and $E_C$ fails for STC (X8); 12 mM SGDC is not a growth inhibitor by the paper's own criterion, and STDC at δ 2.4 is at most marginal (X9, X7). Rates are per 50 mL vessel, not per volume or crystal area (X12).

The AI writer's own judgement (3): design and reporting. Equal seed loading is assumed from the return to crystalline solubility, and seed surface area was never measured. The monomer and micelle assignment is cited, not checked in this medium. Adsorption is inferred from zeta potential at a bile-salt concentration that the method leaves unclear (X15). Residuals are free drug concentrations, not δ, and the solubility needed for conversion is not printed (X17). Several misprints are resolved in favour of the data: δ 2.7 for δ 2.4, the Table 3 caption, STCDC called trihydroxy in Figure 8, and the SEM magnifications (X2, X1, X3, X5). HPMCAS-MF appears under Materials but is never used (X16). None of these misprints changes the conclusions. The formulation relevance is argued, not tested: there is no ASD, no dissolution test with residual crystals, no mixed bile-salt medium and no in vivo data.

Conclusion

The paper introduces a seeded desupersaturation assay with seeds nucleated in the test vessel and supersaturation regenerated to a fixed activity. With it, the authors show that bile salts are far from equivalent as inhibitors of telaprevir crystal growth. Monomeric STCDC stopped growth at every supersaturation tested and kept 60.5 of 65.2 µg/mL free for 15 h. Monomeric SGCDC kept 91.7 of 95.2 µg/mL, although its growth rate was not measured. STC, SGC, STDC and SGDC slowed growth only a few-fold ($E_G$ about 1.0–4.6 across both concentrations), and STC alone became much more effective at high supersaturation. At 12 mM, the micellar level, the chenodeoxycholate effect disappeared. Combining the seeded rates with unseeded induction times, the authors infer that the other four monomers act mainly on nucleation, while STCDC acts on growth. These two inferences carry the hedges and assumptions described above.

For formulation, the work turns one question into two. Whether a medium or excipient keeps a clear supersaturated solution clear is a nucleation question. Whether it keeps supersaturation once crystals are present is a growth question, and the answers can differ by orders of magnitude for the same bile salt. Because the result depends on which bile salt is present and at what level, a single-bile-salt medium may misjudge how fast seeds grow in intestinal fluid. The authors' next step, bile-salt mixtures, is the natural test, together with dissolution of ASDs that contain residual crystals.