Bile salts as nucleation inhibitors: how sodium taurocholate and SDS change the lifetime of supersaturated drug solutions

Background and motivation

Many drug candidates dissolve too poorly in water to be absorbed well. The paper opens with the estimate that up to 80% of candidates in current discovery and development pipelines are poorly water-soluble, which limits both their bioavailability and their further development (Babu 2011, ref 1). R12·C1 Formulation can rescue some of them: by solubilizing the drug in surfactant micelles and lipids, with cosolvents, as amorphous solid dispersions, or as nanocrystals (refs 2–9). C1

The amorphous route works through supersaturation. The theoretical solubility of an amorphous drug is usually several times that of its crystal (refs 10–13, including the group's own work on amorphous solubility, Almeida e Sousa 2014, and on telaprevir, Mosquera-Giraldo 2014). Under nonsink conditions an amorphous drug can therefore dissolve to a concentration far above the crystalline solubility. That state is thermodynamically unstable: the solution tends to nucleate the stable crystal, and once nucleation starts the concentration falls back toward the crystalline solubility. The practical goal is to slow nucleation enough that the higher concentration lasts long enough for absorption. C2 Supersaturation can also arise in the body from the drug's own properties, its ionization behaviour; the abstract counts supersaturated solutions, however generated, among the most promising ways to improve solubility-limited absorption. C3

The standard tool for holding supersaturation is a polymer. PVP-based polymers and cellulose derivatives are the primary additives used both to stabilize solid dispersions and to inhibit nucleation from solution, a field in which the Taylor group had published induction-time studies (Ilevbare 2012, refs 14 and 15). C3

Bile salts entered the picture through two earlier observations that pointed in opposite directions. In the group's own work on celecoxib amorphous solid dispersions, STC inhibited nucleation of celecoxib from supersaturated solutions (Chen 2015, ref 16); in that system STC was also known to be below its critical micelle concentration in the presence of the drug. R9·R11 In another study STC promoted the crystallization of carbamazepine dihydrate during dissolution of anhydrous carbamazepine (Murphy 2002, ref 17). R10 Whether STC is a general crystallization inhibitor, or whether the celecoxib result was a special case, was open.

The question matters beyond formulation additives because bile salts are always present in the intestine. They are cholesterol derivatives made in the liver; micelle-forming bile salts raise the solubility of poorly soluble compounds, and together with phospholipids and their hydrolysis products they form mixed micelles that solubilize drugs well (Fuchs 2014, ref 18). They lower interfacial tension and so help emulsify and absorb fats. C4 STC is present in human bile, although it is not the most prevalent bile salt (Wuyts 2013, ref 24), and it is widely used in media that simulate intestinal fluid. R13·C21

The paper therefore sets out to measure, for a structurally diverse group of poorly soluble compounds, whether STC delays nucleation from supersaturated solutions, by comparing each drug's nucleation induction time in a medium with and without the bile salt. As a comparison it repeats the experiment with SDS, a common anionic surfactant, and in a smaller set of runs it tests two further bile salts. The authors describe the results as "exciting preliminary observations". A1·A2·A3·C26

Background concepts

Supersaturation and the ratio C/Ceq. A solution is supersaturated when it holds more dissolved drug than the crystalline solubility allows. The paper expresses the degree of supersaturation as a ratio, defined in words on p.2595:

In plain terms: $C$ is the drug concentration generated at the start of an experiment, and $C_{\mathrm{e}\mathrm{q}}$ is the equilibrium solubility of the crystalline drug in the same medium. A ratio of 10 means ten times more drug in solution than the crystal can hold at equilibrium. The important detail is "in the same medium": $C_{\mathrm{e}\mathrm{q}}$ was measured separately in buffer, in buffer with 0.1% STC and in buffer with 0.1% SDS, and the ratio was held fixed across the three. E2·P2·R1

Nucleation induction time. Nucleation, the birth of the first stable crystal nuclei, cannot be observed directly, because a nucleus can only be detected once it has grown to a measurable size. Following Söhnel and Mullin (ref 21), the paper therefore defines an experimental induction time (Eq. 1):

In plain terms: the measured induction time $t_{\mathrm{i}\mathrm{n}\mathrm{d}}$ is the true nucleation time $t_n$ (the time to form a critical nucleus) plus the growth time $t_g$ (the time for the nucleus to grow to a detectable size). The paper reports $t_{\mathrm{i}\mathrm{n}\mathrm{d}}$ and calls it the nucleation induction time; it does not separate the two parts. E1·C5 A longer induction time means that the supersaturated solution stays clear for longer.

Detecting crystals by extinction. The paper follows each solution with a fiber-optic UV/vis dip probe. At a wavelength where the drug does not absorb (350–450 nm), any loss of transmitted light comes from particles scattering it; this "extinction" rises when crystals appear. At the same time the absorbance at an absorption maximum of the drug falls as dissolved drug is lost to the crystals. The induction time is taken from the rise in extinction. Figure S1 shows the two signals for telaprevir in 0.1% SDS (Section "Experimental analysis" below). P3·R8

Critical micelle concentration (CMC). Surfactants form micelles above a threshold concentration, the CMC. The paper cites CMC values in water of 0.24% for SDS (ref 22) and 0.97% for STC (ref 23); the 0.1% used here is below both. M1·M4 A drug can lower a surfactant's CMC, which the authors did not measure. C12

What a bile salt looks like. Figure 1 draws the three bile salts used. Each has a rigid four-ring steroid skeleton carrying hydroxyl groups on one face and a side chain ending in an anionic group. In STC the side chain is a taurine conjugate ending in a sulfonate; in sodium glycocholate (SGC) and sodium glycodeoxycholate (SGDC) it is a glycine conjugate ending in a carboxylate. STC and SGC carry three hydroxyl groups; SGDC carries two. M1·M2·M3 The authors stress the facial polarity of this structure: one edge of the steroid is hydrophilic and the other hydrophobic, which sets bile salts apart from chain surfactants such as SDS. C4

Working with un-ionized drugs. Ionization changes both solubility and crystallization. To keep the comparison clean, the medium for each drug was chosen to keep it un-ionized: HCl at pH 3 for celecoxib, phosphate buffers at pH 2 or 6.5, citric buffers at pH 3.5 or 4, and a carbonate–bicarbonate buffer at pH 9.6 for the base bifonazole (Table 1). C6·P2 Note by the AI writer (not discussed in the paper): with the pKa values printed in Table 1, naproxen (pKa 4.15) at pH 3.5 is about 18% ionized, indomethacin (pKa 4.5) at pH 3.5 about 9%, and compound 1 (basic pKa 3.4) at pH 4 about 20% protonated, so for these three drugs "un-ionized" is approximate. X3

Methods in detail

The study has a simple, deliberately broad design: eleven drugs, three media per drug (buffer, buffer with 0.1% STC, buffer with 0.1% SDS), and two read-outs (crystalline solubility and induction time), plus a short series with two further bile salts. The flow below shows the order of the work. A1·A2·A3

The eleven drugs and four additives

The drugs were chosen to differ in structure and physicochemical properties (Figure 2, Table 1): five neutral compounds (celecoxib, griseofulvin, felodipine, tadalafil, telaprevir), three acids (warfarin, pKa 5.0; naproxen, pKa 4.15; indomethacin, pKa 4.5) and three bases (bifonazole, pKa 6.22; nevirapine, pKa 2.8; and compound 1, a Merck development compound with pKa values of 1.9 and 3.4). M5–M15 Compound 1 appears in the SI tables as "MK2022"; the SI never names compound 1, and MK2022 is the only drug there that is not one of the other ten, so the two names refer to the same compound. X2 The additives were STC, SGC and SGDC (each at least 97% pure, Sigma) and SDS (at least 99.0%, Sigma), used as received. M1–M4

Table A. Conditions of the induction-time experiments: ionization and medium from Table 1, C/C_eq from Table 2, stock solvent from SI Table S3. All at 25 °C except telaprevir (37 °C). M5–M15·P2

Contribution 1: crystalline solubility in every medium

Problem. SDS and STC raise drug solubility by very different amounts. If the same total amount of drug were added to buffer and to a surfactant solution, the surfactant solution would simply be less supersaturated, and a longer induction time could not be credited to the additive. The solubility in every medium is therefore needed first. C7

What they did. An excess of crystalline drug was added to 15 mL of medium, with or without 0.1% surfactant, stirred and equilibrated for 48 h in a water bath at 25 °C (telaprevir at 37 °C). The excess solid was removed by ultracentrifugation at 35 000 rpm for 20 min at the same temperature (Beckman Optima L-100 XP, swinging-bucket rotor SW 41 Ti). The supernatant was assayed by HPLC for felodipine, warfarin, bifonazole, naproxen and compound 1 (C18 columns, detection at 210–261 nm; SI Table S1), or directly with the UV dip probe at 251–300 nm for celecoxib, griseofulvin, indomethacin, nevirapine, tadalafil and telaprevir (SI Table S2). The number of replicates and the meaning of the ± values in Table 2 are not stated. P1·R1

Contribution 2: induction times at matched supersaturation

What they did. Supersaturation was created by solvent shift: a concentrated stock of the drug in an organic solvent (methanol, DMF, DMSO or ethanol, SI Table S3) was titrated into 50 mL of the test solution, stirred at 300 rpm with a cross-shaped magnetic stirrer. P2 The amount added set the starting concentration $C$ at a fixed multiple of the crystalline solubility in that same medium, between 1.5 and 15 depending on the drug (Table A). The ratios were chosen so that the induction time in plain buffer was "reasonably short", under 100 min. P2·E2 The extinction was then followed with the dip probe at a nonabsorbing wavelength between 350 and 450 nm, with a path length of typically 10 mm (shorter for more concentrated solutions). Each experiment was repeated three times and the average induction time reported. P3

Why the matching matters. The authors state the reason directly: a fixed C/C_eq was used "to compensate for the potential influence of the differences in solubility enhancement on drug nucleation rate". C7 Under this design a solution in SDS contains more drug in total than one in buffer, but both are equally far above the crystalline solubility in their own medium. Note by the AI writer (not discussed in the paper): the difference in total drug can be large. Computed from Table 2, felodipine at C/C_eq 8 corresponds to 4.8 μg/mL in buffer, 8.8 μg/mL in 0.1% STC and 173.6 μg/mL in 0.1% SDS. With STC the total drug concentration was only 1.1 to 3.5 times that in buffer; with SDS it was up to about 36 times. X4

Gaps in the matrix. Telaprevir was run at 37 °C because it crystallized extremely slowly from buffer at 25 °C; all other drugs were run at 25 °C. R6·P3 SDS runs were not done for compound 1, because the target concentration in SDS could not be reached without adding an excess of stock solution and therefore of organic solvent, nor for bifonazole, because of experimental issues at pH 9.6. P3 The duration of each run is not stated; runs in which no crystals appeared are reported as "greater than" the experimental duration. X7

Contribution 3: two more bile salts

To see whether the effect belongs to bile salts as a family, "limited studies" added 0.1% SGC to celecoxib and nevirapine, and 0.1% SGDC to nevirapine, at 25 °C and otherwise the same conditions. SGDC could not be tested with celecoxib because it is insoluble at pH 3, the pH of the celecoxib medium. P4·R5·C26

Experimental analysis

STC barely changes solubility; SDS often changes it a lot

All eleven compounds are poorly soluble in their blank media, from 0.2 μg/mL (bifonazole and compound 1) to 81.3 μg/mL (nevirapine) (Table 2). Both surfactants raised solubility above that in buffer, but by very different amounts. R1·C8

Table B. Equilibrium solubility of the crystalline drugs (from Table 2; 25 °C unless noted; n and the type of ± not stated). R1

The authors summarize that SDS "In general" gave greater solubility enhancement than STC; that the maximum enhancement with STC was "approximately a factor of 2"; that SDS raised the solubility of compound 1 by a factor of 250 and that of the others by factors between about 1.5 and 37; and that the very insoluble compounds (below 1 μg/mL in buffer) "appeared" to be solubilized most by SDS but not by STC. R2·C9·C10 Note by the AI writer (not discussed in the paper): the stated factors do not all follow from Table 2. The STC factor is 3.5 for bifonazole (0.2 to 0.7 μg/mL) and 2.5 for telaprevir; the SDS factor is 1.2 for nevirapine, below the stated lower bound of about 1.5, and about 266 for compound 1. Because the blank solubilities of bifonazole and compound 1 are 0.2 ± 0.1 μg/mL, their factors are poorly defined. Celecoxib is the one drug for which STC (1.8 μg/mL) solubilized slightly more than SDS (1.6 μg/mL). The table values should be used rather than the rounded factors. X1

Both surfactants were used below their CMC in water. From the low solubilization by STC the authors conclude that "it is likely" that STC is also below its CMC in the presence of the drug, as is known for celecoxib, while noting that the effect of the drug on the CMC was not determined. C11·C12·R11 This matters for the next result: in the induction-time runs, an STC solution contained almost the same amount of drug as a buffer solution at the same C/C_eq, so any longer induction time with STC cannot be explained by a lower supersaturation. X4

STC lengthens the induction time for every drug

Figure 3 is the central result. In buffer, the chosen supersaturations gave induction times of about 7 min (bifonazole) to about 87 min (celecoxib). With 0.1% STC, every one of the eleven drugs took longer to crystallize (Table C). For celecoxib, griseofulvin, warfarin and naproxen no crystallization event was detected during the measurement; Figure 3 labels them as more than 1000, 960, 1000 and 250 min. R3·C13

Table C. Nucleation induction times in minutes, read from Figure 3 (approximate; mean of three runs). ">" values are the labels printed on the bars for runs without a detected crystallization event; "<1" is the printed label for naproxen in SDS. R3

The authors call the increase "quite dramatic" for most compounds, and in the Conclusions write that STC "was demonstrated to prolong" the induction times of all eleven drugs. They also stress that "there is still considerable variation" in how far STC extends the induction time from compound to compound, and that "more detailed investigations are necessary to gain mechanistic insight". C13·C14·C15 The effect was seen at 25 °C for ten drugs and at 37 °C for telaprevir. C16·R3

Note by the AI writer (not discussed in the paper): the variation is wide. Computed from the bar readings, the STC induction time is about 2 times the buffer value for bifonazole (about 7 to 15 min), about 5 times for telaprevir and about 9 times for tadalafil, and more than 10 times for the other eight drugs, reaching more than 100 times for felodipine (about 8 to 930 min). No statistical test is reported, the type of error bar is not stated, and the ">" values are lower bounds set by the length of each run, so ratios built on them are minimum values. The 37 °C statement rests on telaprevir alone. X6·X7·X10

SDS shortens the induction time for most drugs

SDS behaved very differently (Table C, Figure 3). Of the nine drugs tested in 0.1% SDS, only two crystallized later than in buffer: griseofulvin, with "significant inhibition" (more than 500 min against about 78 min), and nevirapine, with "a much lower extent of inhibition" (about 38 against about 20 min). For the other seven, celecoxib, felodipine, tadalafil, telaprevir, warfarin, naproxen and indomethacin, the induction time was shorter in SDS than in buffer; for celecoxib it fell from about 87 to about 5 min and for naproxen to under 1 min. The authors read this as "suggesting that the nucleation of these compounds is enhanced in the presence of this surfactant". R3·C17·C18

Note by the AI writer (not discussed in the paper): because supersaturation was matched as C/C_eq, every SDS run held more drug in total than the corresponding buffer run (felodipine about 36 times more). The paper does not separate an effect of SDS on nucleation from an effect of the higher total concentration, or of drug held by surfactant aggregates. The statement that SDS "enhances" nucleation is therefore defined relative to the crystalline solubility in the SDS medium. X4·X5

Two further bile salts

Figure 4 compares STC with SGC and SGDC on two drugs at 25 °C. For celecoxib, crystals appeared after about 86 min in buffer, and not within the run with STC (more than 1000 min) or with SGC (more than 990 min); SGDC was not tested. For nevirapine, crystals appeared after about 21 min in buffer and about 502 min with STC (error bar about 400 to 610 min), but not within the run with SGC or SGDC (both more than 990 min). R4·R5 The authors describe this as "a similar or better nucleation inhibition effect relative to that observed for STC", and conclude that SGC and SGDC "also showed promise as nucleation inhibitors". C23 Note by the AI writer: for celecoxib both STC and SGC are lower bounds at the run length, so only nevirapine shows a measurable advantage over STC. X12

What a single experiment looks like

SI Figure S1 shows one run in full: telaprevir in 0.1% SDS at C/C_eq 13. The apparent concentration, from the absorbance at an absorption maximum, stays at about 255 μg/mL for about the first 22 min, then falls steeply to about 45 μg/mL by 50 min and about 43 μg/mL at 60 min. The extinction at 370 nm stays near zero until about 20 min and then rises to about 0.28, mirroring the loss of dissolved drug. R8 The starting value agrees with the design: 13 times the crystalline solubility of telaprevir in 0.1% SDS (19.0 μg/mL) is 247 μg/mL. E2·R1 The induction time for this system in Figure 3, about 22 min, falls where the extinction begins to rise. R3

Discussion

How the authors connect the results

The paper proposes no mechanism. Its argument runs from the two observations to their consequences for formulation and testing, as sketched below. C13·C18·C15

The authors draw three consequences. First, "it is clearly important to evaluate the impact of the surfactant employed in a given formulation on the tendency of the system to undergo crystallization", particularly in formulations meant to generate supersaturation; on these data "SDS is expected to be problematic in this context for many compounds, and STC may provide a viable alternative". C19·C20 Second, STC is often present in media that simulate intestinal fluids and are used to test how long supersaturation lasts, so "the choice of test medium thus may influence the crystallization outcome", leading to over- or underprediction of crystallization in vivo. C21 Third, because STC is present in human bile, "its properties as a crystallization inhibitor may be important in influencing the amount of drug absorbed from the GI tract" for systems that supersaturate in vivo. C22·R13 In the Conclusions they generalize, with a hedge, that bile salts "potentially represent a family of additives that can delay the nucleation of a variety of poorly water-soluble drugs" and deserve further study for supersaturating formulations. C25

Where the result applies

The evidence covers eleven structurally diverse drugs in simple aqueous buffers or HCl, each kept largely un-ionized, at one surfactant level (0.1%, below the CMC of both surfactants in water), at supersaturations of 1.5 to 15 times the crystalline solubility, and mostly at 25 °C. P2·P3·M1·M4 Within these conditions the direction of the STC effect is uniform across neutral, acidic and basic drugs, and its size varies from small (bifonazole) to more than 100-fold (felodipine). R3·C14·X6

Notes by the AI writer (not discussed in the paper)

• The contrast between the two surfactants separates two properties that are often treated as one. SDS is the stronger solubilizer (up to about 36-fold among the drugs it was tested on) yet shortened the supersaturated period for most drugs; STC barely solubilizes (1.1- to 3.5-fold) yet lengthened it for all. Solubilization data alone would rank the two surfactants the wrong way round for holding supersaturation. R1·R3·X1
• The matched-C/C_eq design answers the question "at the same distance from equilibrium, which medium nucleates faster?". A formulation works at a fixed dose instead; there, SDS would lower the supersaturation by raising solubility, and the net effect of SDS is not predicted by these data. X4·X5
• Because STC alone already delayed nucleation by hours for most drugs, induction-time rankings of polymeric inhibitors measured in plain buffer may differ from those in bile-salt-containing media. The paper does not test bile salts combined with polymers.

Limitations

Limits the authors state

• The work is presented as "exciting preliminary observations", and the tests with SGC and SGDC as "Limited studies". C26
• "Clearly more detailed investigations are necessary to gain mechanistic insight into the observed effects." C15
• "The impact of the drug on the CMC was not determined", so the micellar state of STC with each drug is inferred ("it is likely"), not measured. C11·C12

The AI writer's own judgement

The central observation is strong in direction: all eleven drugs crystallized later with STC, eight of them by more than 10-fold, and the STC runs held nearly the same amount of drug as the buffer runs, so the effect is not a solubilization artefact. R3·R1·X4·X6 The evidence is weaker in its details and in its reach:

• Precision. All induction times are bar-chart readings; individual runs, the error-bar type and run durations are not given, and the ">" values are lower bounds. X7
• Small effects. No statistical test is reported; the bifonazole effect (about 7 to 15 min) is small. X6
• The SDS comparison. At matched C/C_eq the SDS runs held far more drug, so "SDS enhances nucleation" is not separated from a concentration effect. X4·X5
• Ionization. Naproxen, indomethacin and compound 1 are partly ionized at the chosen pH. X3
• Temperature. The claim that STC works at both 25 and 37 °C rests on one drug at 37 °C. X10
• Stated numbers. The solubility enhancement factors in the text do not all match Table 2. X1
• Cosolvent. The amount of organic solvent added with the stock is not reported, and the stock solvent differs between drugs. X9
• Surfactant basis. Whether "0.1%" is w/v or w/w is not stated. X8
• Reach. The relevance to simulated intestinal media and to absorption is argued, not tested: STC was used alone at 0.1% in simple buffers, without phospholipids or other bile salts. X13

Two small source details are worth knowing when reading the paper: the abstract graphic shows induction times of 505 and 23 min for an unnamed drug, which match nevirapine with STC and in buffer; and the SI tables call compound 1 "MK2022". X11·X2

Conclusion

This short communication makes one clear empirical point and several hedged proposals. The empirical point: at 0.1% and matched C/C_eq, STC lengthened the nucleation induction time of all eleven drugs tested, often from under 100 min to many hours, and SDS shortened it for seven of nine; SGC and SGDC did at least as well as STC on the two drugs tested. R3·R4·C13·C17·C23 The proposals, that bile salts "potentially represent a family of additives" for supersaturating formulations, that STC in test media "may" change the measured crystallization outcome, and that STC in bile "may be important" for absorption, remain to be tested in biorelevant media, at other bile salt levels and in vivo, and the mechanism of the inhibition is open. C25·C21·C22·C15·X13 For a formulation scientist the practical message, which the authors state themselves, is that a surfactant in a supersaturating formulation should be checked for its effect on the tendency to crystallize. C19