Solution methods

How it was shown

Van Eerdenbrugh, Alonzo and Taylor asked whether undissolved drug particles disturb in-situ UV concentration readings only by scattering light, as the dissolution literature assumed, or also by absorbing it vaneerdenbrugh2011b C2, C4. A crystalline felodipine nanosuspension was made by wet media milling: 0.5 g felodipine and 0.5 g PVP in 5 mL water per vial, eight vials, 5 g of 0.5 mm zirconia beads, stirring at 100 rpm until dynamic light scattering showed a mean size near 300 nm; the batches were pooled, left overnight so that large particles and agglomerates settled, and the supernatant was kept vaneerdenbrugh2011b P1. Four coarser suspensions came from jet milling (sieved at 53 µm) and from hand grinding and sieving (three fractions) vaneerdenbrugh2011b P2, P3. All suspensions were diluted in aqueous PVP to 1 mg/mL PVP and 10, 25, 40 or 50 µg/mL total felodipine at 25 °C vaneerdenbrugh2011b P4, A1, A2. A six-channel Ocean Optics system with 1 cm transflectance dip probes scanned 200–700 nm; spectra were smoothed (Savitzky–Golay, which changed neither peaks nor intensities), and the apparent dissolved concentration came from second derivatives over 270–440 nm calibrated with 0.5–20 µg/mL methanolic standards vaneerdenbrugh2011b P9, R11. A correct reading would stay at the solubility whatever the solids content, because total drug is 12 to 62 times above 0.81 µg/mL (synthesis: 10/0.81 = 12.3; 50/0.81 = 61.7) vaneerdenbrugh2011b M1, A2. Three explanations other than particle absorption were checked: higher solubility of small crystals (cited theory and an earlier study put the gain at 100–115% for crystals of this size), tested by ¹H NMR relative solubility in ethanol/water 25/75 for 24 h at 25 °C in triplicate; an amorphous fraction, tested by DSC, XRPD, Raman and mid-IR; and solubilisation by PVP vaneerdenbrugh2011b R16, P8, P5, R9. Particle size came from DLS (nanosuspension) and laser diffraction (all five) vaneerdenbrugh2011b P6, P7.

What the data show

The five suspensions covered three orders of magnitude: the nanosuspension had a Z-average of 300 ± 3 nm (PDI 0.141 ± 0.006) and d(v,0.5) 0.38 µm with a tail to d(v,0.9) 1.90 µm; the others had d(v,0.5) of 10.19 (jetmilled), 38.39, 105.03 and 380.21 µm vaneerdenbrugh2011b R1, R2. Nanosuspension spectra rose at every wavelength with total felodipine: at 276 nm from about 0.24 to 1.28 and at the band maximum near 375 nm from about 0.20 to 1.04, on top of a Tyndall-shaped scattering baseline, against a band of only about 0.015 for 0.81 µg/mL in methanol vaneerdenbrugh2011b R3. The jetmilled suspension peaked at about 0.073 (band maximum at 50 µg/mL); its baseline was almost flat and rose from about 0.011 to 0.046 at 425 nm between 10 and 50 µg/mL vaneerdenbrugh2011b R4. After second-derivative correction the nanosuspension read about 3.4, 5.75, 7.85 and 8.95 µg/mL, and the regression slope was 0.14 (R² = 0.995), that is 1.4 µg/mL of apparent increase per 10 µg/mL of particles vaneerdenbrugh2011b R6.a, R7. A straight line through the four points has an intercept of about 2.1 µg/mL (the extractor's fit), so the reading at 10 µg/mL is about 4.2 times the solubility (synthesis: 3.4/0.81), more than the slope alone accounts for; the slope should not be used alone to correct readings vaneerdenbrugh2011b X9. The coarse suspensions stayed nearly flat with a slight upward trend: jetmilled about 1.0 to 1.6, fraction 1 (38.39 µm) about 0.94 to 1.25 (no marker at 50 µg/mL), fraction 2 about 1.07 to 1.9 at 40 µg/mL (1.75 at 50) and fraction 3 about 1.0 to 2.1 µg/mL vaneerdenbrugh2011b R2, R6.b–e, X11. The highest value is about 2.6 times the solubility (synthesis: 2.1/0.81); the authors put it "most likely" down to the low absolute absorbance of a saturated solution, which turns small absolute errors into large relative ones, and did not pursue it vaneerdenbrugh2011b C22. The three alternative explanations did not hold: the nanosuspension solubility was 89 ± 10% of that of the coarse powder, no amorphous felodipine was detected in any powder or in the freeze-dried nanosuspension, and PVP raised solubility only above 5 mg/mL, five times the 1 mg/mL used vaneerdenbrugh2011b R8, R10, R9. The authors write that the NMR result "independently confirms" that solubility is not the cause and that the excess "strongly suggests" absorption by submicron particles vaneerdenbrugh2011b C15, C24.

Figure

Orange diamonds are the nanosuspension, rising to about 9 µg/mL at 50 µg/mL total; blue, red, green and purple are the jetmilled and sieved suspensions, all near 1–2 µg/mL. A correct reading would be a flat line at the solubility of about 0.81 µg/mL vaneerdenbrugh2011b R6, R7, M1.

Why

The authors' reasoning, step by step (their interpretation): light removed from the beam by a particle is absorbed or scattered, and Mie theory gives both from size, wavelength and refractive indices; for small particles absorption dominates vaneerdenbrugh2011b C3, C26. A second derivative removes constant and smoothly sloping baselines, which is how Tyndall scattering is usually corrected, but it cannot remove a particle absorption band that lies inside the drug's own band vaneerdenbrugh2011b C2, C23, C25. So the probe reports an apparent concentration that rises with the amount of solid vaneerdenbrugh2011b C25. They word it "might absorb" in the abstract, "strongly suggests" in the results and "demonstrate" in the conclusions vaneerdenbrugh2011b C24. What was measured in support: no solubility gain by NMR, no amorphous form, and a Mie calculation with the drug's own absorptivity that reproduces the size of the signal for 300 nm particles (see F2) vaneerdenbrugh2011b R8, R10, R13. What was not measured: the dissolved concentration in the aqueous nanosuspension itself (the NMR test used ethanol/water) and a particle-free aqueous reference spectrum, so absorption by the particles was inferred by exclusion plus modelling, not isolated vaneerdenbrugh2011b X4. A practical sign is the band position: the nanosuspension band peaked near 375 nm, against about 355–365 nm for the methanolic standard and the jetmilled suspensions, and the nanoparticle spectrum shows a red shift of the 361 nm peak vaneerdenbrugh2011b R3, R4; alonzo2012 R20. In the 90:10 HPMC–felodipine dispersion the shift appeared within 5 min of the start of dissolution and was absent for the 50:50 dispersion, pure amorphous felodipine and the dissolved controls alonzo2011 R20. Alonzo and co-workers explain their UV overreading for the 90:10 dispersions by submicron particles of 200–400 nm that absorb like the solution and are not removed by second derivatives alonzo2011 C27. Jackson and co-workers (2015) and Almeida e Sousa and co-workers (2016) say the same of drug-rich nanospecies above the amorphous solubility; Tres and co-workers (2018) add that apparent concentrations are then "not quantitative" jackson2015 C12; almeidaesousa2016 C20; tres2018 C37.

Conditions and limits

The controlled experiment covers one drug (felodipine), crystalline milled or ground particles, 1 mg/mL PVP, water at 25 °C and a 1 cm probe vaneerdenbrugh2011b A1, P9. No suspension between 0.38 and 10 µm was measured, no precipitated or amorphous particle was measured, and the expectation that lower solubility, more solids and smaller particles raise the error was tested only for solids and size vaneerdenbrugh2011b X5, X7, X6, C44. The derivative feature used for quantification is not stated and calibration was in methanol, whose band lies at shorter wavelength than the aqueous nanosuspension band vaneerdenbrugh2011b X8. The later dispersion studies did not test particle absorption directly; they compared UV with flux, NMR or ultracentrifugation (table above).

Counterexamples and conflicts

None found in the library for the existence of the overreading. Open: the coarse suspensions also read above the solubility, up to about 2.6 times (synthesis: 2.1/0.81), which the authors ascribe to noise on a weak band vaneerdenbrugh2011b C22, and the text value of 0.034 a.u. for a saturated solution disagrees with the plotted peak of about 0.016 vaneerdenbrugh2011b R5, X2.

Transfer check

Before trusting an in-situ UV reading of another drug: (1) compare it with a separation-based value (ultracentrifugation followed by HPLC, or flux) on the same suspension; (2) test whether the reading keeps rising in proportion to added solid above the known solubility, which is the signature of this artifact; (3) compare the band position with the solution standard, since a red shift points to particles; (4) measure the particle size (DLS or NTA): the evidence is for particles below about 1 µm, and a flat baseline correction may suffice above about 2 µm, a statement based on calculation only vaneerdenbrugh2011b C33, C36, C38, X5; (5) a drug with lower solubility, higher solids or smaller particles is outside the tested range and the error should be expected to be larger, not smaller vaneerdenbrugh2011b C44.

History

2011: Van Eerdenbrugh and co-workers first show the artifact for a felodipine nanosuspension and explain it by particle absorption vaneerdenbrugh2011b C23, C24. Alonzo and co-workers (2011) use it to read the discrepancy between UV, diffusion and NMR for 90:10 dispersions alonzo2011 C27. Alonzo and co-workers (2012) adopt the second-derivative method and check the 361 nm peak for the red shift alonzo2012 P3, R20. Jackson and co-workers (2015) measure the size of the overreading for a danazol dispersion against flux (16.2 against 8.2 ± 1.2 µg/mL) jackson2015 R2.c, R9.i. Almeida e Sousa and co-workers (2016) and Tres and co-workers (2018) carry the caveat to a weak base and to ritonavir and lopinavir almeidaesousa2016 C20; tres2018 C37.

How it was shown

To obtain the imaginary refractive index, four amorphous felodipine films of 0.52–1.58 µm average thickness (spin-coated from dichloromethane, thickness from weight gain) were read at 361 nm, three readings each; amorphous films were chosen because, unlike crystals, they scatter negligibly vaneerdenbrugh2011b P10, R12. The absorption coefficient follows from k and wavelength,

\alpha = \frac{4\pi k}{\lambda}

vaneerdenbrugh2011b E1, and film attenuation gives α from \alpha h = \ln(I_0/I) vaneerdenbrugh2011b E2. MiePlot v4.2.03 (the BHMIE code) then computed extinction cross-sections for homogeneous spheres of 30, 60, 100, 200, 300, 400, 500, 600, 800, 1000, 1200, 1500, 2000, 3000 and 4000 nm, with a real refractive index of 1.57 for felodipine in water of 1.33, with the film k ("absorbing") or with k = 0 ("nonabsorbing") vaneerdenbrugh2011b P11, A5, M1. The extinction cross-section is the sum of absorption and scattering parts, and without multiple scattering a beam is attenuated exponentially:

C_{ext} = C_{abs} + C_{sca} \qquad \alpha_{ext} = N\,(C_{abs} + C_{sca}) \qquad \frac{I}{I_0} = \exp(-\alpha_{ext}\,h)

vaneerdenbrugh2011b E4, E5, E8. The volume-normalised cross-section was converted to the absorbance a probe would read by

-\log\frac{I}{I_0} = K\,C_{ext,V} \qquad K = \frac{\varphi_{part}\,h}{\ln 10}

vaneerdenbrugh2011b E10, E11; for 50 µg/mL, h = 1 cm and a crystal density of 1.451 read as g/cm³ this gives a particle volume fraction of 3.45 × 10⁻⁵ and K = 1.50 × 10⁻⁵ cm (the extractor's calculation; synthesis: 50 × 10⁻⁶/1.451 = 3.45 × 10⁻⁵ and 3.45 × 10⁻⁵/2.303 = 1.50 × 10⁻⁵) vaneerdenbrugh2011b E11, X10.

What the data show

The films gave α = 4.0 ± 0.4 µm⁻¹ and k = 0.116 ± 0.013 at 361 nm vaneerdenbrugh2011b R12; the extractor's check of Eq. 1 gives 0.115 vaneerdenbrugh2011b E1. For nonabsorbing spheres the extinction at 276 nm rose steeply with size, about 0.01, 0.07, 0.22, 0.61 and 0.93 for 30, 60, 100, 200 and 300 nm, and with shorter wavelength (Tyndall effect) vaneerdenbrugh2011b R13, C28. Absorbing spheres added a felodipine band whose maximum was about 0.57, 0.62, 0.71, 0.88 and 0.94 for the same sizes; at 360 nm the nonabsorbing curve was only about 0.004 at 30 nm and 0.54 at 300 nm vaneerdenbrugh2011b R13. Read at 360 nm, absorption therefore accounts for nearly all the extinction at 30 nm and about 0.40 of 0.94, roughly 43%, at 300 nm (synthesis: 0.94 − 0.54 = 0.40; 0.40/0.94 = 0.43; plot readings), which the authors call "certainly still significant" vaneerdenbrugh2011b R13, C29. The 300 nm curve matched the measured 50 µg/mL nanosuspension in magnitude ("good agreement"), but its band peaked near 350–355 nm against about 375 nm measured; the authors think this "most likely" reflects k taken from amorphous films rather than crystals, since the calculated band shape resembles the film spectrum vaneerdenbrugh2011b R13, R3, C30, C31. From 400 to 1000 nm the absorption contribution shrinks and the pattern becomes complex: for 1000 nm spheres the absorbing curve dips to about 0.56 at 360 nm where the nonabsorbing one reaches about 0.82, and nonabsorbing spheres of 800–1000 nm lose the Tyndall shape vaneerdenbrugh2011b R14, C33. The authors warn that this "may not be easily identified" and that nonabsorbing excipient particles of 600–1000 nm may upset scattering corrections vaneerdenbrugh2011b C34, C35. From 1200 to 4000 nm the extinction at 500 nm falls from about 0.65 to 0.13, and absorbing and nonabsorbing curves converge; above 2000 nm the curve "appears" to be the horizontal baseline assumed in the literature, so offsets or second derivatives "might be appropriate" vaneerdenbrugh2011b R15, C36, C37, C38.

Figure

Solid curves are absorbing spheres, dashed curves nonabsorbing spheres; the gap between the solid and the dashed curve of one colour is the particle absorption. Panel A adds the measured nanosuspension (black) and the amorphous film (dotted); all panels share one y-scale vaneerdenbrugh2011b R13, R14, R15.

Why

Light removed by a sphere is the sum of absorption and scattering; absorption per unit volume is set by k, and scattering grows with size, so absorption dominates for the smallest spheres and scattering takes over as the sphere grows vaneerdenbrugh2011b C3, C26, C29. For spheres of 400–1000 nm the two interfere, so the felodipine band can fall below the scattering baseline vaneerdenbrugh2011b C33. Beyond about 2 µm extinction falls with size and loses its wavelength dependence vaneerdenbrugh2011b C36, C37. The measured coarse suspensions (d(v,0.1) of 3.12 µm for the jetmilled one) read close to the solubility, which is consistent with the calculated flat regime, but they cannot confirm it because no suspension between 0.38 and 10.19 µm exists (synthesis: R2 and R6 against the calculated regimes) vaneerdenbrugh2011b R2, R6.b, X5.

Conditions and limits

The authors list their own simplifications: homogeneous monodisperse spheres, k from amorphous films, a real refractive index estimated and wavelength-independent, and no Kramers–Kronig relation between real and imaginary index; multiple scattering is neglected vaneerdenbrugh2011b C39, C8. The comparison uses a monodisperse 300 nm sphere for a nanosuspension with d(v,0.9) 1.90 µm and span 4.61 (synthesis: (1.90 − 0.15)/0.38 = 4.61, as printed) vaneerdenbrugh2011b R2.a, E12, X3. Calculations were run at one concentration only vaneerdenbrugh2011b X5.

Counterexamples and conflicts

None found in the library; no other paper in the library repeats a Mie calculation (synthesis: the concept evidence table for Mie theory lists one paper).

Transfer check

For another drug: (1) measure k on a scatter-free amorphous film, as done here, and use a real refractive index appropriate to the crystal; (2) run Mie for the measured size distribution, not only the median; (3) compare the predicted extinction at the drug's band with the signal expected at its solubility: if the particle part is a sizeable fraction of that signal, the probe is unreliable vaneerdenbrugh2011b C45; (4) check the prediction against one measured spectrum of the actual suspension, because band position and shape differed by 20–25 nm here vaneerdenbrugh2011b X3.

History

2011: Van Eerdenbrugh and co-workers introduce the calculation and call it a "first evaluation" for a given system vaneerdenbrugh2011b C45. In the papers read for this chapter no later paper repeats it; later authors cite the experimental artifact (see F1).

How it was shown

Raina and co-workers dissolved each dihydropyridine in methanol (5, 10, 20 or 25 mg/mL) and infused it with a syringe pump into 50 mM pH 6.8 phosphate buffer while a 1 cm dip probe read extinction at 450 nm; the concentration of the sharp rise was taken as the onset of a new phase, at 25 and 37 °C, in buffer alone or with HPMC raina2015a P6. A solution with no visible particles, a hazy bluish tinge like the Tyndall effect of colloids and no birefringence under PLM was classed as LLPS; visible particles meant crystallization raina2015a P7, C9. Tres and co-workers pumped a methanolic felodipine stock at 25 µL/min into 10 mL of pH 7.4 buffer with 100 µg/mL HPMC at 25 °C and 300 rpm, read extinction at 450 nm every 40 s, and used the microsome or liposome suspension as the blank tres2017 P6, C24. In the dissolution study, in-situ dip probes (1 cm for ritonavir, 2 cm for lopinavir) recorded 200–450 nm every 60 s, the drug signal at 240 or 260 nm was corrected by the 450 nm signal, and nanoparticle tracking after 1 µm filtration showed the particles tres2018 P8, E4, P10, R19.

What the data show

For nisoldipine at 25 °C extinction was about 0.00 from about 0.3 to about 6 µg/mL, then about 0.05 at about 9.3, 0.10 at about 11.7 and 0.25 at about 16.4 µg/mL; the authors place the onset at "approximately 7" µg/mL, where the solution turned hazy raina2015a R3, C10. In buffer at 25 °C the six onsets were about 1.4 (cilnidipine), 6.0 (felodipine), 38 (nifedipine), 17.8 (nimodipine), 6.0 (nisoldipine) and 15.9 µg/mL (nitrendipine); at 37 °C about 1.9, 8.6, 58, 21.5, 8.0 and 22 µg/mL, and the nifedipine values at both temperatures and the nimodipine value at 37 °C went with visible crystals raina2015a R5, R4. With HPMC the same drugs gave liquid phases at about 49 and 67–73 µg/mL (nifedipine, 25 and 37 °C) and about 18 and 26 µg/mL (nimodipine); the nifedipine value at 37 °C is about 67 in one figure and 73 in the text raina2015a R5, R4, X7. In microsomes extinction stayed at about 0.01–0.02 up to about 100–110 µM and in liposomes up to about 150 µM, then reached about 0.39 at about 280 µM and about 0.38 at about 313 µM tres2017 R5. During dissolution of the 10% ritonavir dispersion extinction was about 0.16 at 1 min, 0.51 at 3 min and a plateau of about 0.55–0.56 from 6 to about 26 min; the 50% dispersion stayed at about 0.01–0.02. The 10% lopinavir dispersion rose steadily to about 0.70 at 54 min, the 50% one stayed at about 0.01 tres2018 R7, R15. The authors read the 10% curves as "indicative of" a second scattering phase and the NTA frames as clear light scattering by the drug-rich phases tres2018 C26, C35, C40. Outside the chapter the same read-out found the onset of other drugs at other wavelengths (table), always at a wavelength where the drug does not absorb: 288 nm is the danazol maximum and it does not absorb at 350 nm jackson2015 M1.

Why

The authors' reasoning (their assumption): a new particulate phase scatters light at a wavelength where the drug does not absorb, so the extinction there is scattering alone and subtracting it also corrects the UV drug signal raina2015a C9; tres2018 C14, E4; jackson2014 C13. Scattering gives the size range of the scattering centres but little about their structure, which is why a second criterion is needed raina2015a C6, C40. (synthesis: because scattering depends on particle size and number, as the Mie results of F2 show, extinction is not a measure of the amount or composition of the new phase; the ritonavir curve plateaued within about 6 min while the lopinavir curve kept rising for 54 min, with NTA sizes of 696 and 294 nm tres2018 R7, R15, R18.) For danazol the early fall of the UV signal at 20 µg/mL without polymer is attributed to rapid coalescence of the droplets, which changes their scattering and absorption, not to crystallization jackson2014 C29.

Conditions and limits

The method is kinetic: it depends on mixing, the addition rate and the detection limit, and gives only the onset ueda2021 C3. In the dihydropyridine study the infusion time scale, the final methanol fraction and the HPMC concentration are not stated, and whether nifedipine and nimodipine crystallize before phase separating depends on that time raina2015a X12, X16. Dynamic addition can overshoot the amorphous solubility before LLPS is detected, and separating the phases by ultracentrifugation can give a value closer to the coexistence concentration; for ritonavir the three methods agreed at about 27–28 µg/mL purohit2017b C19, R1. In the dissolution study the ultracentrifuged amorphous solubilities were 9–10 µM above the cited extinction onsets, an offset the authors do not discuss tres2018 X1. Extinction in turbid media needs a blank of the medium itself tres2017 C24.

Counterexamples and conflicts

None found that extinction misses a phase separation; the converse is the limit above, since the same rise marked crystals for fast crystallizers raina2015a R4, C11.

Transfer check

(1) Choose a wavelength at which the drug has no absorbance (450 nm served for six dihydropyridines, felodipine, ritonavir, lopinavir and CRH1 raina2015a P6; tres2017 P6; tres2018 P8; almeidaesousa2016 R3; 350–400 nm for danazol, ritonavir, telaprevir and posaconazole jackson2014 R4; purohit2017b P1; mosqueragiraldo2015 R4; elkhabaz2019a R2) and confirm it on the drug's spectrum; (2) repeat at two addition rates, because a kinetic onset that moves with rate is not an equilibrium value ueda2021 C3; (3) classify the new phase by clarity and polarized light microscopy, or by one of the probe or particle-counting read-outs of F8 and F9; (4) for a fast crystallizer, repeat with a nucleation inhibitor; (5) if the number is used for design, compare it with an ultracentrifuged supernatant.

History

The extinction method for the onset of a disordered phase and its link to the amorphous solubility came from earlier library work (builds_on: alonzo2010, ilevbare2013a) and is used by raina2015a for six drugs raina2015a P6. Jackson and co-workers (2014) used it with pyrene for danazol jackson2014 R5; Mosquera-Giraldo (2015), Purohit (2017), Almeida e Sousa (2016) and Elkhabaz (2019) used it for telaprevir, ritonavir, CRH1 and posaconazole mosqueragiraldo2015 R4; purohit2017b R1; almeidaesousa2016 R3; elkhabaz2019a R2. Tres and co-workers (2017, 2018) carried it into vesicle suspensions and into ASD dissolution, where it served as the first of several read-outs tres2017 R5; tres2018 R7, R15.

How it was shown

Each probe (stock 1 mg/mL in DMSO) was diluted to 1 µg/mL in water, pH 6.8 buffer, methanol, ethanol, isopropyl alcohol, benzene, toluene and hexane; excitation scans on a SPEX Fluoromax-3 gave the excitation and emission maxima, and emission spectra were recorded at the excitation maximum raina2015a P8. The six probes were pyrene (read as the ratio I1/I3 of its peaks at 373 and 383 nm), PRODAN, dansyl glycine, dapoxyl sulfonate and the two styryl dyes DASPEI and 4-Di-2-ASP raina2015a M8–M13, R1. In drug solutions each probe at 1 µg/mL was exposed to stepwise additions of drug from methanol, and the spectrum was read within 1 min with 3 nm slits and 100 ms integration raina2015a P9. Tres and co-workers used PRODAN at 0.9 µM, excited at 370 nm tres2017 M4, P5.

What the data show

Emission maxima in buffer, methanol and benzene were: PRODAN 520, 498, 463 nm; dansyl glycine 554, 518, 502; dapoxyl 575, 512, 472; DASPEI 580, 575, 564; 4-Di-2-ASP 610, 605, 585; pyrene I1/I3 1.6, 1.2, 1.07 raina2015a R1. The shifts quoted in the short part are the differences between the buffer and benzene values (synthesis: 575 − 472 = 103; 520 − 463 = 57; 554 − 502 = 52; 610 − 585 = 25; 580 − 564 = 16). The spectra add ethanol and toluene: the maximum falls in the order water ≈ buffer > methanol > ethanol > benzene > toluene, with the largest blue shift in toluene (PRODAN about 455 nm against about 521 nm in buffer) raina2015a R2. Pyrene ratios read from the spectra (about 1.47 in buffer, about 0.97 in benzene) are lower than those of Table I raina2015a X6. In a felodipine solution at 37 °C the DASPEI maximum stayed at about 577–579 nm from 1 to 5 µg/mL and fell to about 558 nm at 25 µg/mL while the intensity rose from about 14,000 to about 97,000 cps, about 6.9-fold (synthesis: 97,000/14,000) raina2015a R6. With nifedipine and HPMC at 37 °C the 4-Di-2-ASP maximum fell from about 610.5 nm at 10 µg/mL to about 579 nm at 200 µg/mL and the intensity rose from about 7,100 to about 30,500 cps, about 4.3-fold (synthesis: 30,500/7,100) raina2015a R9. In the same system the pyrene ratio fell from about 1.8 to about 0.76 raina2015a R16. PRODAN, whose band sits at 520 nm in water, showed a second band near 435 nm with clotrimazole and near 450 nm with efavirenz, and about 440 nm with ritonavir and lopinavir tres2017 R13; tres2018 R4, R5, below its benzene value of 463 nm in the dihydropyridine paper (synthesis; the two groups used different excitation wavelengths and instruments) raina2015a R1. Outside the chapter, pyrene gave 1.97 in water, 1.55 in methanol, 1.22 in isopropyl alcohol and 0.66 in cyclohexane, and a danazol drug-rich phase read about 1.5 at about 24 µg/mL against about 1.95 below the onset jackson2014 R6, R8.

Figure

Each normalised spectrum moves to shorter wavelength from water and buffer towards toluene; panel b shows the vibronic peaks of pyrene raina2015a R2.

Why

The authors' explanation (their interpretation): on excitation the probe's dipole moment grows, and before emitting it loses energy by vibrational relaxation and by solvent molecules reorienting around it; the excited state lasts nanoseconds and solvent relaxation picoseconds, so emission comes from the relaxed state; in water much energy is lost and emission is red and weak, in a non-polar medium less is lost and emission is bluer and stronger raina2015a C25. With the photon energy E = h\nu = hc/\lambda, more energy lost before emission means a longer emission wavelength raina2015a E3. Viscosity, relaxation rate, conformational change, charge transfer, proton transfer and probe–probe interaction "may augment" the effect raina2015a C28. The authors expect charge-transfer dyes to be the most polarity-sensitive and so name DASPEI and 4-Di-2-ASP, but Table I shows these two with the smallest solvent shifts, so the ranking should not be read as a ranking of the size of the shift raina2015a C29, X4. For the drug-rich phase they propose that the probe moves from water into the drug-rich liquid, where it is "mainly surrounded" by drug molecules, and that the aminophenyl tail of the styryl dyes, known to embed in hydrophobic membrane regions, "likely" enters it; partitioning was not measured raina2015a C26, C30, X17. For the felodipine shift they add higher local viscosity and possibly stronger hydrogen bonding raina2015a C35. Emission wavelength is preferred to intensity because it does not depend on excitation intensity or probe concentration raina2015a C27. For PRODAN the lifetime study adds that it decays with two components (0.50 and 1.77 ns alone) and that higher microviscosity and lower polarity are "likely to contribute" to the longer lifetime in aggregates; neither was measured tres2017 R10, C40.

Conditions and limits

The solvent series was measured on one instrument at 1 µg/mL probe; isopropyl alcohol and hexane spectra are listed but not shown, and the pyrene value in Fig. 3b differs from Table I raina2015a P8, X6. Because the maxima in the drug-rich phase lie beyond the benzene values and the authors name viscosity and other factors, a probe maximum in the drug-rich phase should not be read as an equivalent solvent polarity raina2015a X5, C28. Pyrene baselines differ between papers and experiments (about 1.8 here, 1.6 in Table I, about 1.95 for danazol), so ratios are compared only within one experiment raina2015a X6; jackson2014 R7.

Counterexamples and conflicts

The authors' ranking of the styryl dyes as the most polarity-sensitive contradicts the smallest shifts in their own Table I raina2015a C29, R1, X4. (synthesis: sensitivity in the sense of the fractional change of intensity at LLPS is a different quantity from the solvent shift and was not tabulated.)

Transfer check

For a new probe or instrument: (1) record the probe in a solvent series of falling polarity with the instrument and excitation to be used; (2) take the aqueous value from a probe-only solution measured the same day; (3) use the emission wavelength, or a ratio, rather than raw intensity raina2015a C27; (4) read the drug-rich-phase result as "non-aqueous environment", not as a polarity.

History

Jackson and co-workers (2014) had used pyrene I1/I3 for danazol solutions jackson2014 R6, R8. Raina and co-workers (2015) tested six probes systematically, borrowing the idea from fluorophores that find micellar critical concentrations raina2015a C7, C24. Tres and co-workers (2017) extended the idea from spectra to lifetime with PRODAN (builds_on) tres2017 C33, C37; the 2018 study kept PRODAN tres2018 M5.

How it was shown

Raina and co-workers compared three onsets for each dihydropyridine in 50 mM pH 6.8 phosphate buffer at 25 and 37 °C: the extinction onset at 450 nm from a methanolic infusion (F3); the fluorescence onset, from stepwise additions of drug stock to buffer holding 1 µg/mL DASPEI (4-Di-2-ASP for nifedipine at 37 °C), each solution read within 1 min; and a thermodynamic estimate of the amorphous solubility raina2015a P6, P9, R11. The estimate multiplies the shake-flask crystalline solubility (48 h, 25 and 37 °C, HPLC) by the free-energy advantage of the amorphous form and by a water-activity term from moisture sorption raina2015a P1–P5:

\sigma_a = \sigma_c \, e^{-I(\alpha_a)} \, e^{\Delta G / RT} \qquad \Delta G = \frac{\Delta H_f \, \Delta T}{T_m}\cdot\frac{T}{T_m}

raina2015a E1, E2, with DSC melting data at 10 °C/min; the inputs are not printed raina2015a P3, P4. Nifedipine and nimodipine were run with HPMC because they crystallize first in buffer (F8). Tres and co-workers (2017) added fluorescence lifetime: decays excited at 370 nm and fitted with two exponentials, reported as the amplitude-weighted average lifetime, and emission spectra, at felodipine 12, 16, 19, 24, 28 and 38 µM, clotrimazole 0–38 µM and efavirenz 0–148 µM in pH 7.4 buffer with 100 µg/mL HPMC at 25 °C; the two dark drugs were read through 0.9 µM PRODAN tres2017 P1, P4, P5, E1, E2. Tres and co-workers (2018) used the PRODAN lifetime and the ratio I1/I2 of its 445 or 450 nm band to its 520 nm band for ritonavir (0, 5, 10, 20, 40, 59, 69 µM) and lopinavir (0, 5, 10, 15, 20, 41, 46, 51, 56 µM) with 100 µg/mL HPMC-AS at 24 °C tres2018 P3, E5.

What the data show

Raina and co-workers' Fig. 8 gives, as estimated amorphous solubility / extinction / fluorescence onset in µg/mL, the values below; the last column is the highest over the lowest of the three (synthesis: arithmetic on the plot readings) raina2015a R11.

The authors call this "a close correlation" for each compound at each temperature; for felodipine at 37 °C they give about 8 (fluorescence), about 9 (extinction) and about 8 µg/mL (estimate), and for nifedipine with HPMC 70, 73 and 73 µg/mL raina2015a C16, R7, R10. Every onset and every estimate was higher at 37 than at 25 °C raina2015a R5, R11. Felodipine at 25 °C gave 6.00 ± 0.57, 5.00 ± 1.00, 5.00 ± 1.32 and 6.00 ± 2.09 µg/mL with pyrene, PRODAN, DASPEI and 4-Di-2-ASP (n = 3) raina2015a R12. The onsets lie far above the crystalline solubility, for example about 6.0 against about 0.7 µg/mL for felodipine at 25 °C, a ratio of about 8.6 (synthesis: 6.0/0.7) raina2015a R5. In the lifetime study the felodipine average lifetime was 0.42, 0.46 and 0.45 ns at 12, 16 and 19 µM and 0.53, 0.71 and 0.96 ns at 24, 28 and 38 µM; the intensity at 435 nm went from about 146 to about 633 a.u. between 19 and 38 µM and the band maximum from 445 to 435 nm tres2017 R1, R2, R3. The PRODAN lifetime stayed at 0.93, 0.93, 0.89 and 0.94 ns up to 16 µM clotrimazole and rose to 1.19, 1.28 and 1.32 ns at 27, 33 and 38 µM; with efavirenz it was 0.90–0.98 ns up to 49 µM and 1.16 to 1.97 ns from 69 to 148 µM; new emission bands appeared at about 435 and 450 nm tres2017 R10, R11, R13. The authors state onsets as the last unchanged concentration (19, 16 and 49 µM) with the next tested points at 24, 27 and 69 µM, so the earlier values of 19, 22 and 58 µM fall inside the brackets and the agreement holds within the concentration spacing tres2017 R14, X8. For ritonavir the PRODAN lifetime was about 1.00, 0.98, 0.99 and 1.03 ns up to 20 µM, then about 1.13, 1.29 and 1.34 ns at 40, 59 and 69 µM, with I1/I2 first rising between 40 and 59 µM; for lopinavir about 0.97–1.03 ns up to 20 µM and about 1.14, 1.19, 1.22 and 1.22 ns at 41, 46, 51 and 56 µM, with no concentration tested between 20 and 41 µM tres2018 R1, R2, X2. Tres and co-workers (2018) also measured the amorphous solubility by ultracentrifugation: 48 µM (ritonavir) and 38 µM (lopinavir) with 800 µg/mL Kollidon VA64, and about 47 and 37.5 µM with HPMC-AS, which is 9–10 µM above the cited extinction onsets tres2018 R9, R17, X1.

Figure

The dark bars are the estimated amorphous solubility, the light bars the extinction onset and the mid-grey bars the fluorescence onset; for each drug the three bars should be of similar height raina2015a R11.

Why

The authors' explanation (their interpretation): above the amorphous solubility the solution splits into water and a drug-rich supercooled liquid, the lipophilic probe moves into it, and its emission shifts to shorter wavelength and brightens, so the start of the shift is the start of LLPS raina2015a C26, C36. For PRODAN they add that further drug adds drug-rich phase without raising the free concentration, so more probe moves in and the lifetime keeps rising tres2017 C34. Probe partitioning was inferred from the spectra and lifetimes and not measured raina2015a X17; tres2017 C33; tres2018 C17. (synthesis: fluorescence onsets lie below the extinction onsets for four of the six drugs at 25 °C in the dihydropyridine study and for both media of the weak base CRH1, which fits the reading that a small amount of drug-rich phase can shift a probe before it scatters enough to be seen; this was not tested, and for nimodipine and nitrendipine the order is reversed raina2015a R11; almeidaesousa2016 R3, R4.)

Conditions and limits

Almost all values are readings of bars or points; n is stated only for the felodipine probe table, and the error bars of the figures are undefined raina2015a X12. Which probe gave the nifedipine and nimodipine bars is not confirmed raina2015a X11. The lifetime onsets are brackets set by the tested concentrations, with no replicate decays and, for the 2018 study, no component tables tres2017 X7, X8; tres2018 X8. The agreement of the lifetime with extinction was not tested on the same dilute solutions, so the claimed advantage of lifetime below 10 µM is untested tres2017 C12, X15. The estimate itself is uncertain: inputs are not printed in the dihydropyridine paper, and in the companion synchrotron paper the printed felodipine value of 6.68 µg/mL does not reproduce (5.93 µg/mL recomputed, a 13% difference (synthesis: (6.68 − 5.93)/5.93)) raina2015a E1; raina2015b E2. The buffers (50 mM pH 6.8 and 100 mM pH 7.4), temperatures (25, 37 and 24 °C) and polymers differ between the papers, so onsets are not comparable across papers except as same-system checks.

Counterexamples and conflicts

Cilnidipine's three values differ by a factor of about 3, and for the dissolution study the two reference values used for "the amorphous solubility" (cited extinction onsets and ultracentrifuged values) differ by 9–10 µM without comment raina2015a X9; tres2018 X1. For the weak base CRH1 the pyrene ratio crossed its threshold during dissolution of the salts at an apparent concentration of about 9 µg/mL, below the solvent-shift onsets of 12.49 and 13.73 µg/mL and below the lowest estimate of 11.2 µg/mL, an offset the authors do not discuss almeidaesousa2016 X3. In danazol the estimate (16 µg/mL) was above the measured 13 µg/mL, and with 10 µg/mL polymers the measured onsets fell to 11, 9 and 8 µg/mL while the estimates rose to 25, 43 and 22 µg/mL jackson2014 R5, R8.

Transfer check

For another drug: (1) measure the extinction onset first (F3); (2) pick the reporter by checking the drug's own emission at the probe's excitation (F6); (3) step the concentration through the expected onset with a probe-only and a drug-only control, and report the onset as a bracket; (4) if the number sets a design limit, confirm it by ultracentrifugation and HPLC of the supernatant; (5) for a fast crystallizer add a nucleation inhibitor and read each point at once (F8); (6) the thermodynamic estimate needs the crystalline solubility, DSC melting data and a water-sorption isotherm of the amorphous form, and should be treated as an estimate, not a measurement (synthesis: the three values of one drug differed by up to 1.65-fold for the five drugs in the table above, and the companion paper's recomputation differed by 13%).

History

2014: Jackson and co-workers compare pyrene fluorescence, UV extinction and an estimate for danazol and find 13 µg/mL by both read-outs against 16 µg/mL estimated jackson2014 R5. 2015: Raina and co-workers extend the comparison to six dihydropyridines and two temperatures and add a drug that is its own indicator raina2015a R11, R15. 2015–2019: Indulkar, Almeida e Sousa, Purohit and Elkhabaz apply probes, pyrene or intrinsic fluorescence to nicardipine, CRH1, ritonavir and posaconazole and find the same kind of agreement indulkar2015 R12; almeidaesousa2016 R3, R4; purohit2017b R1; elkhabaz2019a R3, R4. 2017: Tres and co-workers show that lifetime brackets the same onsets for a fluorescent and two dark drugs tres2017 R14. 2018: Tres and co-workers bring the lifetime to ritonavir and lopinavir and add ultracentrifuged values that are 9–10 µM higher than the cited onsets tres2018 R1, R2, X1.

How it was shown

Raina and co-workers ran felodipine at 25 °C with all six probes (1 µg/mL each, n = 3) and overlaid the drug's emission on the probe bands; they recorded felodipine alone in water, buffer, methanol, ethanol, benzene and toluene and at 0.25–100 µg/mL in buffer at 25 and 37 °C, reading emission maximum and peak intensity, with the excitation wavelength not printed raina2015a P9, P10, R12, R13, R14, R15, X12. Tres and co-workers made the drug-alone controls explicit: 38 µM clotrimazole and 148 µM efavirenz without PRODAN, and PRODAN alone at 0.9 µM, below its aqueous solubility of about 3.5 µM, to look for a blue-edge shoulder of self-aggregation tres2017 R13, M4, C47.

What the data show

With dansyl glycine the felodipine band at about 440 nm grew from about 33,000 cps at 0.5 µg/mL to about 60,000 at 1 and about 117,000 cps at 2 µg/mL while the probe band near 555 nm stayed at about 20,000–25,000 cps; with DASPEI the felodipine band near 443 nm rose from about 10,000 cps at 0.5 µg/mL to about 170,000 cps at 12.5 µg/mL while the DASPEI band near 560–580 nm rose from about 15,000 to about 40,000 cps raina2015a R13.c, R13.d. Pyrene's 373 nm peak fell from about 5.1 × 10⁶ cps at 0.1–0.2 µg/mL to about 1.0–1.7 × 10⁶ at 15–20 µg/mL while the felodipine band grew to about 3.7–4.2 × 10⁶ cps at about 430 nm, so the I1/I3 ratio was unusable and the onset (6.00 ± 0.57 µg/mL) came from the fall of the peak intensity; the authors attribute the fall to quenching by felodipine and say it "confirms" close contact after LLPS, without lifetime data raina2015a R13.a, R12.a, C18, C19. Felodipine's emission band sits at about 438–445 nm in water and buffer, about 435 nm in methanol and ethanol and about 425–435 nm in benzene and toluene raina2015a R14. In buffer at 25 °C its maximum was about 443 nm at 0.25–1 µg/mL, about 440 nm at 5–7.5, about 435 nm at 25 and about 433 nm at 100 µg/mL; at 37 °C it was about 442–444 nm up to 7.5, about 439.5 at 10 and about 432.5 at 100 µg/mL; the peak intensity rose from about 0.01 × 10⁶ cps at 0.25 µg/mL to about 4.9 × 10⁶ (25 °C) and 3.9 × 10⁶ (37 °C) at 100 µg/mL raina2015a R15. The intensity per µg/mL rises later than the wavelength shifts (about 0.024–0.029 × 10⁶ cps per µg/mL between 3 and 12.5 µg/mL at 25 °C, about 0.05 × 10⁶ from 25 µg/mL), so the emission maximum, not the intensity, is the marker of the onset raina2015a X14. In the lifetime study, drug alone at the top concentration gave about 0 a.u. for both dark drugs, and PRODAN alone showed no extra shoulder, so the new bands at 435 and 450 nm belong to the probe; the authors also infer that PRODAN does not interact with the free drug tres2017 R13, C36. Outside the chapter, posaconazole's own fluorescence behaves like felodipine's: in pH 6.5 buffer the maximum moved from 385 nm to shorter wavelength above 8 µg/mL with the intensity rising from about 70 (2 µg/mL) to about 1015 a.u. (15 µg/mL), and in FaSSIF the 368 nm intensity stayed at about 380–395 a.u. up to 19 µg/mL and reached about 900 a.u. at 24 µg/mL elkhabaz2019a R4, R3. For the napadisylate counterion (NAP) of a CRH1 salt the autofluorescence onset (about 9.8 µg/mL) lay below the pyrene and UV onsets in the same medium (12.49 and 13.73) almeidaesousa2016 X13.

Figure

The felodipine band near 430–445 nm grows over the bands of pyrene, PRODAN, dansyl glycine and DASPEI; where it exceeds the probe band the probe cannot be used raina2015a R13.

Why

The authors explain felodipine's fluorescence by its ortho-chloro substitution, which hinders rotation of the phenyl ring and gives a long-lived, more stable excited conformer (bi-exponential decay, high quantum yield, from the literature), plus hydrogen bonding of the ring N–H to solvent raina2015a C32, R18. They explain the absence of fluorescence in nifedipine and the other nitro-substituted dihydropyridines by the nitro group, "possibly" through vibrational coupling or alternatively charge transfer to the nitrobenzene; the structures drawn put the nitro group in position 2 only for nifedipine and nisoldipine raina2015a C33, X2. The felodipine shift on LLPS is attributed to a less polar environment after self-association tres2017 C20 and, in the discussion, to higher local viscosity and possibly stronger hydrogen bonding raina2015a C35. Energy transfer between felodipine and the probes "possibly" matters and was not investigated raina2015a C34.

Conditions and limits

The self-indicator result for felodipine was obtained at 25 and 37 °C in one buffer, without a polymer raina2015a P10. The probe comparison on a fluorescent drug was run only for felodipine; the six-drug comparison used DASPEI raina2015a X10. The PRODAN route needs a probe concentration below its solubility, a probe-alone control and a drug-alone control; the PRODAN-alone baselines differ between series (0.90–0.93 ns, χ² 1.26 in one) tres2017 R10, R11, X10. The statement that the felodipine intensity is constant up to 19 µM is contradicted by the figure, which shows it rising from about 78 to about 146 a.u. tres2017 X1.

Counterexamples and conflicts

None found in the library for the masking of probe emission by the drug's own fluorescence. For the counterion fluorescence the onset depends on the read-out: 9.8 µg/mL by NAP emission against 12.49 and 13.73 µg/mL by pyrene and UV almeidaesousa2016 X13.

Transfer check

For a new drug: (1) record the drug alone at the probe's excitation wavelength; if it emits in the probe's window, change the probe or use the drug as its own reporter; (2) for a fluorescent drug, follow the emission maximum, not the intensity raina2015a X14; (3) for a dark drug, keep the probe below its solubility (0.9 µM PRODAN against about 3.5 µM) and run probe-alone and drug-alone controls; (4) state the onset as a bracket.

History

2015: Raina and co-workers show that felodipine blocks two probes and serves as its own indicator raina2015a R12, R15, C21. 2017: Tres and co-workers repeat the self-indicator idea with lifetime and extend the probe route to two dark drugs with PRODAN tres2017 R1, R10, R11. 2018: the PRODAN route is used for ritonavir and lopinavir tres2018 R1, R2. 2019: Elkhabaz and co-workers read posaconazole's own fluorescence in FaSSIF elkhabaz2019a R3.

How it was shown

In the dihydropyridine paper the extinction onset of each drug was measured by methanolic infusion at 25 and 37 °C in buffer alone and with HPMC (Pharmacoat 606; the concentration is not printed for the extinction series, 100 µg/mL for the DASPEI runs) raina2015a P6, M7, A3. The felodipine DASPEI runs were done at 37 °C in buffer and in 100 µg/mL HPMC raina2015a R6. Tres and co-workers (2018) made the amorphous-solubility references by adding a methanolic stock above the expected onset to buffer holding either 100 µg/mL HPMC-AS or 800 µg/mL Kollidon VA64, ultracentrifuging (209,875 × g, 20 min, 25 °C) and assaying the supernatant by HPLC tres2018 P9, A4. In the ritonavir paper (outside) the extinction, pyrene and ultracentrifuged values were measured in buffer alone and with 512 µg/mL of three polymers purohit2017b P1–P3, R1. The ibuprofen papers measured by ¹H NMR the highest free drug concentration in the aqueous phase once the drug-rich phase had formed, in buffer and with dissolved polymer ueda2020b R1; ueda2021 R5.

What the data show

Extinction onsets in buffer and with HPMC (µg/mL) were, at 25 °C: cilnidipine 1.4 and 1.3, felodipine 6.0 and 7.3, nimodipine 17.8 and 18.0, nisoldipine 6.0 and 6.0, nitrendipine 15.9 and 17.2; at 37 °C: cilnidipine 1.9 and 1.8, felodipine 8.6 and 9.3, nisoldipine 8.0 and 8.0, nitrendipine 22 and 23.5; the largest difference is 1.5 µg/mL (synthesis: 23.5 − 22; nitrendipine, 37 °C) raina2015a R5. For nifedipine and for nimodipine at 37 °C the buffer value marks crystals and the HPMC value a liquid phase, so the two are not comparable (F8) raina2015a R4. The authors state that HPMC changes neither the LLPS concentration nor the probe's emission wavelength; at 1–2 µg/mL felodipine the intensity was about 24,000–26,000 cps with HPMC against about 14,000–19,000 cps without, with overlapping error bars raina2015a C12, X13. With Kollidon VA64 or HPMC-AS the ritonavir values were 48 and about 47 µM and the lopinavir values 38 and about 37.5 µM tres2018 R9, R17. In ritonavir with 512 µg/mL polymer the amorphous solubility was about 27 µg/mL with none of the three polymers changing it ("It is evident") purohit2017b R1, C17. Against this, danazol onsets were 13 µg/mL alone (by fluorescence and by UV), 11 with PVP, 9 and 8 with HPMC and 8 and 8 with HPMC-AS, while the calculated amorphous solubilities were 16, 25, 43 and 22 µg/mL; the authors call the PVP effect minor and the cellulosic effect larger, and note that this differs from ritonavir jackson2014 R5, C23. A later danazol paper cites the earlier work as showing no appreciable effect of polymers up to 1 mg/mL on the crystalline solubility or the LLPS concentration, which is at odds with the lowering seen at 10 µg/mL in the earlier paper (synthesis: the card does not say which data the later statement rests on) jackson2015 R11, C10; jackson2014 R8. For ibuprofen in pH 2.0 buffer at 37 °C, the crystalline solubility was unchanged by 300 µg/mL HPMC or PVP-VA (30.0 and 30.1 against 29.8 µg/mL) but the amorphous solubility fell from 177.1 to 96.9 and 90.7 µg/mL, while the cationic EUD-E raised both (42.1 and 256.3 µg/mL) ueda2020b P1, R1. At 1000 µg/mL polymer the dissolved ibuprofen fell from 154 µg/mL to about 131 (PVP K12), 121 (PVP K25), 118 (PVP K90), 89 (PVP-VA), 106 (HPMC 603) and 100 µg/mL (HPMC 615) ueda2021 R5.

Figure

The emission maximum (black squares, right axis) starts to fall at about the same felodipine concentration in buffer (panel a) and with HPMC (panel b), and the intensity (red, left axis) rises from there raina2015a R6, R7.

Why

For ibuprofen the authors explain the lower value by polymer that has distributed into the drug-rich phase: it lowers the chemical potential of the drug there and so the concentration in equilibrium with it ("This occurs because") ueda2021 C11. For danazol they propose that the cellulosic polymers mix into the drug-rich phase, the supporting NMR being preliminary, not shown and absent for HPMC-AS jackson2014 C48, R26, X15. In the dihydropyridine, ritonavir and lopinavir papers whether the polymer enters the drug-rich phase was not measured (synthesis: this is the missing variable; the cards report no polymer content of those phases) raina2015a C12; tres2018 R9.

Conditions and limits

The null results are for 100–800 µg/mL HPMC, HPMC-AS or Kollidon VA64 (and 512 µg/mL for ritonavir outside) and for onsets read as extinction or probe onsets or as ultracentrifuged supernatants raina2015a R5; tres2018 R9; purohit2017b R1. The tres2018 comparison is between two polymers, not between polymer and none, because a polymer was needed to prevent crystallization tres2018 P9. Polymer concentration alone does not separate the cases: danazol was affected at 10 µg/mL polymer, ritonavir was not at 512 µg/mL (synthesis) jackson2014 R5, R8; purohit2017b R1, C17. The quantities also differ: a solvent-shift onset is a kinetic detection of the new phase, whereas the ibuprofen values are the free concentration at equilibrium with an existing drug-rich phase ueda2021 C3, R5.

Counterexamples and conflicts

Danazol and ibuprofen (above), and the unreconciled statements in the two danazol papers jackson2014 R5; jackson2015 R11.

Transfer check

Before measuring an onset in the polymer-containing medium used to prevent crystallization: (1) measure the onset (or the ultracentrifuged supernatant) at the polymer concentration to be used and, where crystallization allows, without polymer; (2) if the two differ, report the polymer-containing value and not a "buffer" value; (3) test whether polymer sits in the drug-rich phase, for example by ¹H NMR as in the ibuprofen work ueda2021 R4, R5; (4) note that the thermodynamic estimate does not contain the polymer and cannot predict the shift jackson2014 R5.

History

2014: Jackson and co-workers find that polymers lower the danazol onset and note that ritonavir behaved differently jackson2014 C23, R28. 2015: Raina and co-workers find no effect of HPMC for the dihydropyridines raina2015a C12. 2017: Purohit and Taylor find none for ritonavir with three polymers purohit2017b C17. 2018: Tres and co-workers use two polymers for the ritonavir and lopinavir references and find equal values tres2018 R9, R17. 2020–2021: Ueda and co-workers show a large polymer effect for ibuprofen and trace it to polymer in the drug-rich phase ueda2020b R1; ueda2021 C11.

How it was shown

Extinction onsets and the visual and PLM classification were obtained as in F3, at 25 and 37 °C, in buffer and with HPMC raina2015a P6, P7. For the fluorescence read-outs each concentration was a fresh solution prepared from a methanolic stock and measured within 1 min at 1 µg/mL probe, before crystallization could change the spectrum raina2015a P9. Nifedipine at 37 °C was read with 4-Di-2-ASP in 100 µg/mL HPMC, and nifedipine at 25 °C with pyrene in HPMC of unstated concentration raina2015a R9, R16, M7. The ranking of the dihydropyridines by crystallization tendency was taken from earlier work: nifedipine is a very fast crystallizer and nimodipine is faster than the other four but slower than nifedipine raina2015a R17.

What the data show

In buffer the extinction onsets of nifedipine were about 38 µg/mL (25 °C) and 58 µg/mL (37 °C), and that of nimodipine about 21.5 µg/mL at 37 °C, each with visually observed crystals "just below the LLPS onset concentration"; nimodipine at 25 °C (about 17.8 µg/mL) and the other four drugs gave liquid phases raina2015a R5, R4, C11. With HPMC the nifedipine onsets were about 49 and 67 µg/mL (73 in the text for 37 °C) and the nimodipine onsets about 18 and 26 µg/mL raina2015a R5, X7. Without polymer the probes showed "no distinct change" in wavelength or intensity for nifedipine (data not shown) raina2015a R8. With 100 µg/mL HPMC the 4-Di-2-ASP maximum stayed at about 608.5–611 nm from 10 to 70 µg/mL nifedipine, dropped to about 603 nm at 75 µg/mL and reached about 579 nm at 200 µg/mL, with intensity at about 7,100–8,100 cps up to 70 µg/mL and about 30,500 cps at 200 µg/mL; onsets by fluorescence, extinction and estimate were 70, 73 and 73 µg/mL raina2015a R9, R10. Nimodipine crystallizes at 37 °C but not at 25 °C without polymer raina2015a C11. Outside the chapter: felodipine, a slow crystallizer, can exceed its amorphous solubility and phase separate, whereas nifedipine, a faster one, crystallizes just below the coexistence curve unless HPMC, a nucleation inhibitor, is present raina2014a C31, R20. In a study of crystallization from supersaturated solution of the same drugs, the solution-state ranking from slow to fast was cilnidipine < felodipine = nisoldipine < nitrendipine = nimodipine < nifedipine, and only nimodipine and nifedipine recrystallized from the melt on heating above Tg raina2015b C13, C28. A group of drugs classed as slow crystallizers all phase separated before crystallizing ilevbare2013a C27. Ritonavir, with a low crystallization tendency, kept its drug-rich phase for the full 213 min in the later study, whereas lopinavir did not (F9) tres2018 C29, R20, R21. An outside study of weak bases takes the probe response as proof that the second phase is non-crystalline, since dyes cannot partition into a crystalline phase and behave there as in water indulkar2015 C28, C29.

Figure

Black bars are crystalline solubility, light grey bars the onset in buffer and dark grey bars the onset with HPMC; an asterisk marks visually observed crystallization (nifedipine at both temperatures, nimodipine at 37 °C) raina2015a R4, R5.

Why

The authors' explanation (their interpretation): nifedipine is a very fast crystallizer and nimodipine the next, so without polymer they nucleate before reaching the LLPS concentration, no drug-rich liquid forms, the probe stays in water and nothing changes ("As a result, crystallization precedes LLPS") raina2015a C37, C14. HPMC "is known to" impede nucleation of dihydropyridines, which lets both drugs reach LLPS; nimodipine's crystallization at 37 °C only "can be attributed" to faster nucleation at the higher temperature raina2015a C38, C39. The absence of a probe response in crystallizing systems is, in their words, "important verification" that the method detects LLPS specifically and unlike light scattering, which sees any new phase raina2015a C40. Not measured: nucleation rates with and without HPMC raina2015a C38.

Conditions and limits

Six related dihydropyridines in one buffer with one polymer; the HPMC concentration of the extinction series is not printed; the infusion time scale is not stated and the assignment of crystallization before LLPS depends on it raina2015a X12, X16. Faster nucleation at 37 °C appears in the nimodipine result only raina2015a C11, C39. The negative result for nifedipine is described in the text and not shown raina2015a R8.

Counterexamples and conflicts

The solid-state class from a melt does not always predict solution behaviour: papaverine, warfarin and glyburide, classed as slow crystallizers, precipitated as crystals on rapid supersaturation hsieh2014 C30; the crystallization chapter reviews this (F18). Within the dihydropyridine series, however, the ranking in solution and the two drugs that recrystallize from the melt coincide raina2015b C28.

Transfer check

For another drug: (1) place it on a crystallization-tendency scale first (DSC class or growth rate) and expect a fast crystallizer to crystallize before it phase separates; (2) measure the extinction onset in buffer and with a nucleation inhibitor and look for particles or birefringence at the onset; (3) if the onset moves up with the inhibitor, the buffer value marks crystals; (4) keep each measurement short, because crystallization before LLPS depends on time raina2015a X16; (5) report the polymer used with the result (F7).

History

2014: Raina and co-workers show that HPMC lets nifedipine reach LLPS and that felodipine does so unaided raina2014a C31. 2015: Raina and co-workers use the contrast as a negative control for the probes and extend it to nimodipine and to temperature raina2015a C11, C40; the same group ranks the six drugs in solution raina2015b C13. 2018: Tres and co-workers use the lifetime to follow a slow (ritonavir) and a faster (lopinavir) crystallizer after LLPS tres2018 R20, R21.

How it was shown

Raina and co-workers followed pyrene I1/I3 while nifedipine with pre-dissolved HPMC (concentration not printed) was raised to 400 µg/mL at 25 °C, past the LLPS onset and into the range where particles appeared; particles were noted by eye raina2015a P11, R16. Tres and co-workers made 108 µM felodipine in pH 7.4 buffer without HPMC, a driving force of 108 µM against a crystalline solubility of 2.9 µM, and recorded the decay after 15 h; the solutions with aggregates held 100 µg/mL HPMC tres2017 P2, M1, P1. In the third paper supersaturated solutions of ritonavir and lopinavir "above the amorphous solubility" (concentration not printed) were made without HPMC-AS and the PRODAN lifetime followed for about 213 min at 24 °C, with I1/I2 recorded for ritonavir only because the lopinavir spectra were too noisy tres2018 P5, R20, R22, X4.

What the data show

In the nifedipine series I1/I3 was about 1.79–1.84 from about 5 to 60 µg/mL, about 1.72 at 75, 1.13 at 120, 0.84 at 150 and 0.76 at 200 µg/mL; it then rose to about 0.87 (225), 0.94 (300), 1.23 (350) and 1.60 µg/mL (400) raina2015a R16. The authors place LLPS at around 40 µg/mL and crystallization at around 200 µg/mL, where particles appeared, but the plotted ratio first falls between about 60 and 75 µg/mL raina2015a R16, C22, X8. The felodipine sample gave τ1 0.14 ns (83%) and τ2 9.04 ns (17%), an average of 1.62 ns, against τ2 1.87–2.04 ns at 24–38 µM with HPMC tres2017 R1.g, R1.d–f. In the lopinavir solution the lifetime was about 1.24 ns at 22 min, 1.28 at 61 and 1.31 at 94 min, then about 1.26, 1.19 and 1.13 ns at 105, 125 and 161 min and about 1.05–1.06 ns at 193–213 min; in the ritonavir solution it rose from about 1.00 to about 1.28 ns within 3 min and stayed at about 1.31–1.37 ns, while I1/I2 fell from about 1.57–1.64 (3–114 min) to about 1.37 (198–213 min), which the authors call "approximately constant" tres2018 R21, R20, C44, X3. They conclude that the lopinavir drug-rich phase was depleted by crystallization within about 200 min and that of ritonavir was preserved tres2018 C46. Outside the chapter, danazol gave the same signature with pyrene: without polymer the I1/I3 ratio after 15 min equalled the water value at all concentrations, particles were seen after a few minutes jackson2014 R11, R14, C25, and the induction times read from UV absorbance, UV extinction and fluorescence agreed closely: 6 ± 1, 7 ± 1 and 10 ± 2 min at 15 µg/mL danazol, and 88 ± 18, 107 ± 15 and 102 ± 17 min at 20 µg/mL with PVP jackson2014 R17, C35. In danazol dispersions the ratio rose back as the UV concentration fell, for example from about 1.56 at 60 min to about 1.78 at 420 min for the 90:10 PVP dispersion, which the authors read as loss of the drug-rich phase to crystals jackson2015 R2.a, C14.

Figure

The ratio is about 1.8 at low concentration, falls to about 0.76 at 200 µg/mL and then rises towards about 1.6 as particles appear; the arrows mark the authors' LLPS and crystallization onsets, and the first arrow points near 80 µg/mL against the 40 µg/mL stated in the text raina2015a R16, X8.

Panel a: the ritonavir lifetime stays high to 212 min. Panel b: the lopinavir lifetime peaks at 94 min and falls to near the probe-alone value tres2018 R20, R21.

Why

The authors' explanation (their interpretation): as crystals grow the drug-rich phase is depleted, the probe is expelled into the more polar water and its signal returns to the aqueous value, so pyrene I1/I3 rises and the PRODAN lifetime falls raina2015a C23, C42; tres2018 C43, C45. For felodipine, which is its own reporter, the opposite direction is explained by the molecule being located in a rigid crystal lattice, "most likely" tres2017 C22, C41. The lifetime of crystals is therefore long when the drug emits and short when an external probe is expelled (synthesis: the signal reports the environment of the reporter). Not measured: the location of the probe, emission lifetimes of the expelled probe, and the identity of the solid raina2015a X17; tres2018 C45; tres2017 C22. For the ritonavir dispersion, which contains polymer, the authors call recrystallization "unlikely" given the polymer and the drug's low crystallization tendency, and the supernatant of the 10% dispersion matched the amorphous solubility tres2018 C29, C30.

Conditions and limits

The crystallization tests of the third paper had no polymer and an unstated supersaturation, so the 94 min and 213 min figures cannot be compared with other induction times tres2018 X4. The late decline of the ritonavir I1/I2 ratio is unexplained tres2018 X3. The announced induction times by fluorescence were not reported in the dihydropyridine paper raina2015a X15. Steady-state spectra were too noisy to use whenever crystalline or undissolved particles were present tres2018 R14, R22, C34. The 108 µM felodipine sample was measured once, with no microscopy, diffraction or concentration check tres2017 X6.

Counterexamples and conflicts

None found for the direction of the change. The nifedipine onset in Fig. 12 (60–75 µg/mL) differs from the stated 40 µg/mL raina2015a X8.

Transfer check

To use a probe to follow crystallization of a drug-rich phase: (1) run a polymer-free sample at a stated supersaturation and confirm the solid by PLM or XRPD, or by a fall of the supernatant below the amorphous solubility; (2) record the probe-alone and probe-plus-crystals values for the instrument, so that "returned to aqueous" is defined; (3) read the lifetime or the ratio, not the intensity; (4) if the drug is the reporter, expect the lifetime to rise in crystals, not fall.

History

2014: Jackson and co-workers couple the fall of the UV signal with the rise of I1/I3 to define danazol's induction time jackson2014 C30, C35. 2015: Raina and co-workers show the pyrene return for nifedipine raina2015a R16. 2017: Tres and co-workers show a 9 ns crystal component for felodipine tres2017 R1.g. 2018: Tres and co-workers use the PRODAN lifetime to separate a persistent drug-rich phase (ritonavir) from a crystallizing one (lopinavir) tres2018 R20, R21.

How it was shown

Ritonavir or lopinavir and Kollidon VA64 (drug:polymer 10:90 and 50:50 w/w) were dissolved in ethanol:dichloromethane 30:70, rotary-evaporated at 40 °C, dried overnight under vacuum, ground and sieved; the products were amorphous by XRPD (the SI with the patterns was not supplied) tres2018 P1, P2, R10. Enough powder (mass not printed) was added to 50 mL of 100 mM pH 7.4 phosphate buffer at 24 ± 1 °C with 0.9 µM PRODAN and no HPMC-AS for the theoretical maximum drug concentration to exceed the LLPS onset, and the lifetime, the ratio I1/I2, the in-situ UV signal and the extinction at 450 nm were followed for about 60 min tres2018 P4, P6, P7, P8. At the end the medium was ultracentrifuged (209,875 × g, 20 min, 25 °C) and the supernatant assayed by HPLC in duplicate; the amorphous solubility was measured by antisolvent addition with 800 µg/mL Kollidon VA64 or 100 µg/mL HPMC-AS tres2018 P9. The 10% media were filtered (1 µm) and sized by NTA tres2018 P10.

What the data show

Ritonavir, 10% loading: lifetime about 0.95 ns at 0, 1.12 at 2 and 1.54 at 15 min, then about 1.52–1.56 ns to 60 min; I1/I2 about 1.39–1.47 from 5 min; extinction about 0.16 at 1 min, 0.51 at 3 min and a plateau of about 0.55–0.56 from 6 to about 26 min; the UV apparent concentration peaked at 96 µM at 3 min and fell to 44 µM at 59 min; supernatant 49 µM against 48 µM (Kollidon VA64) and about 47 µM (HPMC-AS) tres2018 R6, R7, R8, R9. At 50% loading the lifetime stayed at about 1.01–1.06 ns, I1/I2 at about 0.00–0.02, the extinction at about 0.01–0.02, the UV concentration rose slowly to 21 µM and the supernatant held 19 µM tres2018 R6, R7, R8, R9. Lopinavir, 10% loading: lifetime about 0.95 ns at 0, 1.10 at 2 and 1.23 at 13 min and about 1.19–1.23 to 58 min; I1/I2 from about 0.58 at 5 min to about 1.22 at 58 min; the extinction rose steadily to about 0.70 at 54 min; supernatant 36 µM against 38 and about 37.5 µM tres2018 R11, R15, R17. At 50% loading the lifetime also rose, from about 0.99 to about 1.23 ns by 23 min, but several undissolved particles were seen, the spectra were too noisy for I1/I2, the extinction stayed at about 0.01, the UV concentration reached about 15 µM and the supernatant held 18 µM; after 1 µm filtration the lifetime was 1.00 ns against 0.99 ns for probe alone (F11) tres2018 R11, R12, R13, R14, R15, R16, R17. The authors conclude that only the 10% dispersions dissolved above the amorphous solubility and formed a disordered drug-rich phase ("These data indicate", "revealed") and that the extent of supersaturation "was found to be strongly dependent" on drug loading tres2018 C22, C31, C32, C39, C48. Outside: for ritonavir spin-coated films, fresh 10% and 20% PVP films dissolved rapidly with turbidity (LLPS) and 30–50% PVP films did not, and all HPMCAS films reached LLPS, the 50:50 HPMCAS film giving a free ritonavir concentration of 27 µg/mL and a pyrene I/III of 1.61 purohit2017b C30, C32, C33, C34, C37, R14.u, R10.a, R11.c, C38. For posaconazole in FaSSIF the 10% HPMC-AS dispersion stayed near the amorphous solubility, the 25% one for about 90 min, and the 50% one never reached it and fell to about 5 µg/mL elkhabaz2019a C15. For danazol, the 70:30 films and the 50:50 PVP film exceeded the amorphous solubility and the pyrene response showed LLPS; the 50:50 HPMC film dissolved slowly without a drop in apparent concentration but with a pyrene response above 8 µg/mL, and the 50:50 HPMCAS film did not reach the amorphous solubility jackson2015 R4.

Figure

Panel a: lifetime and I1/I2 rise only for the 10% dispersion; panel b: extinction; panel c: apparent concentration above the dashed onset only for 10%; panel d: supernatant against the amorphous solubility tres2018 R6, R7, R8, R9.

Why

The authors' explanation (their interpretation): at 10% loading dissolution follows the highly soluble hydrophilic polymer ("clearly polymer-controlled"), which carries the drug into solution beyond the amorphous solubility, and a disordered drug-rich phase forms; at 50% loading dissolution follows the poorly soluble hydrophobic drug ("pointing to" drug control) and the concentration stays below the amorphous solubility tres2018 C23, C24. The same low-against-high loading contrast was seen earlier for felodipine and bicalutamide dispersions with Kollidon VA64 at 5% and 50% tres2018 C25, R25. Release rates of drug and polymer were not measured tres2018 X10. The lifetime rise at 50% lopinavir loading is attributed to the dye interacting with undissolved particles, not to a drug-rich phase tres2018 C33. (synthesis: the polymer dependence outside the chapter shows that drug loading alone does not determine whether LLPS occurs; with a polymer that is released quickly together with the drug, such as HPMCAS for the ritonavir films, the drug concentration can exceed the amorphous solubility even at 50% purohit2017b C37, C38.)

Conditions and limits

Two loadings, one polymer, powders from solvent evaporation, 24 °C, pH 7.4 buffer, no bile salts tres2018 P1, P4, X10. The loading at which dissolution stops reaching LLPS lies between 10% and 50% and was not located tres2018 X10. The reference values for the amorphous solubility differ (cited extinction onsets of 39 and 28 µM against ultracentrifuged 48 and 38 µM) tres2018 X1. Several quantities carry no replicate or error bar tres2018 X8.

Counterexamples and conflicts

HPMCAS at 50% loading still reached LLPS for ritonavir films, and PVP at 50:50 reached it for danazol purohit2017b C37; jackson2015 R4.

Transfer check

For another dispersion: (1) test at least three loadings, because the transition lies between the two tested here; (2) record lifetime or extinction together with an ultracentrifuged supernatant and a filtration control; (3) compare the supernatant with the amorphous solubility measured in the same medium; (4) repeat with the polymer of interest, since the loading effect is polymer-dependent; (5) measure drug and polymer release if the mechanism matters.

History

2015: Jackson and co-workers show that polymer type and drug loading decide LLPS during danazol dispersion dissolution (builds_on) jackson2015 R4. 2017: Purohit and Taylor find the same for ritonavir films with three polymers purohit2017b C33, C34, C37. 2018: Tres and co-workers add the PRODAN lifetime and the ultracentrifuged supernatant for ritonavir and lopinavir tres2018 R6, R9, R11, R17. 2019: Elkhabaz and co-workers see the loading effect for posaconazole in FaSSIF elkhabaz2019a C15.

How it was shown

The dissolution protocol is that of F10. In situ UV used dip probes of 1 cm (ritonavir) and 2 cm (lopinavir), the drug signal at 240 or 260 nm corrected by the 450 nm signal (calibration in acetonitrile:water 50:50), duplicate runs with the range as error bar tres2018 P8, E4. The filtration control passed the end-point 50% lopinavir medium through a 1 µm glass-fibre filter and re-measured the lifetime tres2018 P11. NTA used a NanoSight LM10 on 10% loading media after 1 µm filtration, 30 s per sample tres2018 P10. Outside the chapter the same question was asked with diffusion-cell flux (regenerated cellulose membrane, free drug read from a flux standard curve), ¹H NMR and filtration jackson2015 P4; alonzo2011 R14, R15; trasi2016 R8.

What the data show

The ritonavir UV curve: about 9 µM at 0, 65 at 1 min, 88 at 2, 96 at 3, 87 at 5, 72 at 10, 59 at 19, 52 at 30, 48 at 38 and 44 µM at 59 min, against a supernatant of 49 µM tres2018 R8, R9. The lopinavir curve: about 33 µM at 2 min, 140 at 10, 279 at 20, 368 at 30, 421 at 40 and 466 µM at 54 min with the extinction still rising, against a supernatant of 36 µM tres2018 R16, R15, R17. The authors say UV concentrations are "not quantitative" with a drug-rich phase present, because the nanospecies absorb and scatter, and if the 800 µg/mL polymer of the ritonavir case applied to lopinavir the total drug would be about 89 µg/mL, about 141 µM at a molar mass of 628.8 g/mol (the extractor's calculation; the molar mass is not printed), so the UV reading exceeds all the drug added tres2018 C37, X6. In the 50% lopinavir run several undissolved particles were visible, the lifetime rose from about 0.99 to about 1.23 ns by 23 min and the spectra were too noisy for I1/I2; the filtrate gave 1.00 ns against 0.99 ns for probe alone tres2018 R12, R11, R14, R13. NTA after 1 µm filtration gave a mean size of 696.4 ± 213.6 nm for the 10% ritonavir medium and 294.3 ± 67.7 nm for the 10% lopinavir medium, with discrete bright scattering spots in the frames tres2018 R18, R19, C40, C41. Outside the chapter: in 90:10 HPMC–felodipine dissolution the UV peaked at 25–30 µg/mL against about 9 µg/mL for 50:50, whereas diffusion slopes (10.5 and 9.5) and NMR areas (1370 and 1014) were nearly equal alonzo2011 R1, R14, R15; the 90:10 HPMCAS danazol film read about 16.2 µg/mL by UV and 8.2 ± 1.2 µg/mL free drug by flux, and the free drug of the 90:10 PVP and HPMC films was 7.3 ± 1.0 and 8.7 ± 1.2 µg/mL against an amorphous solubility of about 8 µg/mL jackson2015 R2.c, R9.i, R9.c, R9.f, M1; a tacrolimus solution above the amorphous solubility passed a 0.7 µm glass filter and was retained by a 0.22 µm one, with NTA sizes of 220 and 210 nm trasi2016 R8, R11, R12. Other drug-rich phases measured: ritonavir at 30 µg/mL about 360 nm by DLS and a ritonavir HPMCAS dispersion about 500 nm by NTA after 1 h trasi2015a R15; trasi2015b R7.

Figure

The lifetime rises in both runs (at 50% from undissolved particles); extinction and apparent concentration rise only for 10%; the apparent concentration in panel c reaches about 466 µM and is not a concentration profile; the supernatants are 36 against 18 µM tres2018 R11, R15, R16, R17.

Why

The authors' explanation (their interpretation): the fall of the ritonavir UV signal after the peak is "likely" an artefact of aggregates coalescing and growing, which changes their scattering and absorption, and recrystallization is "unlikely" given the polymer and the drug's low crystallization tendency tres2018 C28, C29. For danazol the early UV fall without polymer is likewise attributed to rapid coalescence of the droplets jackson2014 C29. The UV signal includes the absorption of the nanospecies, as in F1 tres2018 C37. The lifetime rise at 50% lopinavir loading is the dye interacting with undissolved ASD particles tres2018 C33; unlike a UV artefact, this one is removed by filtration. Filtration itself is not a safe separation: the drug-rich phase may pass the filter, and filtration may induce crystallization jackson2015 C12. Samples that are only micro-centrifuged can hold non-pelleted nanodroplets of about 220–340 nm, which makes HPLC values higher than the free concentration elkhabaz2019a X3.

Conditions and limits

The filtration control was done once at the end, with a 1 µm filter, and the probe with undissolved particles alone was not measured tres2018 K4. NTA sizes carry a spread of unstated type and n, and the particles are presumed amorphous because the supernatants match the amorphous solubility, not because a solid-state test was made tres2018 X8, X5. Sizes differ between drugs, media and techniques (about 160–700 nm in the rows above), so no size is transferable. The ultracentrifuge speed and time matter: 209,875 × g for 20 min here, and a micro-centrifuge in the posaconazole study tres2018 P9; elkhabaz2019a X3.

Counterexamples and conflicts

None found against the direction of the UV error. For the filter, the tacrolimus data say the 0.7 µm filtrate still holds the drug-rich phase, so a filtration control at 1 µm, as used for the lifetime, would not remove a phase of about 220 nm trasi2016 R8, R11; tres2018 P11 (synthesis: the 50% lopinavir control worked because the interfering particles were undissolved ASD particles, not droplets).

Transfer check

For a dissolving ASD: (1) never read the UV concentration alone once the extinction at a non-absorbing wavelength rises; (2) take a separation-based value (ultracentrifugation with stated speed and time, then HPLC) or a flux-based free concentration and compare it with the amorphous solubility measured in the same medium; (3) for any probe read-out, filter at the end through a pore size larger than the droplets but smaller than the undissolved particles and re-measure, and add a sample of probe with the undissolved solid alone; (4) count and size the particles by NTA or DLS; (5) treat the drug-rich phase as present only if at least two independent read-outs agree.

History

2011: Alonzo and co-workers use the UV artefact to explain a mismatch with flux and NMR alonzo2011 C27. 2014–2015: Jackson and co-workers attribute early UV falls to coalescence and measure the flux-derived free drug jackson2014 C29; jackson2015 R9. 2015–2016: Trasi and co-workers size drug-rich phases and study filtration trasi2015a R15; trasi2016 R8. 2018: Tres and co-workers show both directions of the error in one study and add a particle control for the lifetime tres2018 R8, R16, R13.

How it was shown

Rat liver microsomes (pooled Sprague Dawley, male donor) or egg phosphatidylcholine liposomes at 0.25 mg/mL were suspended in 100 mM pH 7.4 phosphate buffer with 100 µg/mL polymer printed as "HPMC AS-MF grade" at 25 °C tres2017 M7, M8, M5, P1, P3, X13. A methanolic felodipine stock was pumped in at 25 µL/min into 10 mL at 300 rpm while a 1 cm dip probe read the extinction at 450 nm every 40 s, with the vesicle suspension as the blank; values are the mean of two runs with the range as error bar tres2017 P6. Liposomes were made by thin-film hydration (77 mg phosphatidylcholine in 5 mL chloroform, rotary evaporation at 40 °C, hydration with 5 mL water) and, for unilamellar vesicles of about 100 nm, probe sonication tres2017 P8, R9. The buffer reference is the earlier extinction onset of 19 µM tres2017 M1.

What the data show

With microsomes the blank-corrected extinction stayed at about 0.01–0.02 up to about 100–110 µM and reached about 0.20 at about 200 µM and about 0.39 at about 280 µM; with liposomes it stayed at about 0.02 up to about 150 µM, was about 0.05 at about 200 µM and reached about 0.38 at about 313 µM, with ranges widening to about ±0.1 above about 220 µM; the text puts the onsets at about 100 and approximately 160 µM tres2017 R5. The authors call the increase "drastic" and explain it as "likely due to" absorption of felodipine by the hydrophobic parts of the vesicles, which lowers its free concentration so that a higher total concentration is needed to reach the amorphous solubility tres2017 C25. Similar increases are known for surfactant micelles tres2017 C26. In the earlier work, felodipine at 37 °C gave extinction onsets of 8.7 µg/mL in phosphate buffer, 177 µg/mL with 0.04% vitamin E TPGS and 215 µg/mL in FaSSIF, about 20- and 25-fold higher (synthesis: 177/8.7 = 20.3; 215/8.7 = 24.7), in good agreement with the plateau onsets of the flux raina2015c R11, C42. Posaconazole phase separated at 8 µg/mL in pH 6.5 buffer (fluorescence) and at about 20 µg/mL in FaSSIF with 20 µg/mL HPMC-AS (extinction at 350 nm and intrinsic fluorescence), a factor of 2.5 (synthesis: 20/8) elkhabaz2019a R4, R2, R3.

Figure

Panel a: extinction rising only above about 100 µM (microsomes) and about 160 µM (liposomes); panel b: felodipine left in the supernatant after ultracentrifugation; panels c and d: the brighter, blue-shifted spectra with vesicles (see F13) tres2017 R5, R6, R8.

Why

The authors' explanation (their interpretation): LLPS begins when the free drug concentration reaches the amorphous solubility, and anything that takes up drug raises the total concentration at which this happens; here the vesicles take up felodipine (shown below the onset in F13) tres2017 C5, C25. The micelle analogue is the same argument: micelles raise the achievable concentration but not the free drug that drives flux raina2015c C30. What was not measured: the free felodipine concentration at the onset in the vesicle media tres2017 C25. (synthesis: since the amorphous solubility is a free-drug limit, an onset found in plain buffer understates the total concentration at which a lipid-rich medium phase separates.)

Conditions and limits

Felodipine only, 25 °C, one buffer, one polymer level tres2017 P1, P3. No lifetimes or spectra above 19 µM were collected in these media (several populations would be too hard to model), so the aggregation onset there rests on extinction alone and the paper should not be cited for fluorescence-detected onsets in vesicle media tres2017 C17, X16. The vesicle type used for extinction and fluorescence is not stated, no lifetime blank of the vesicle suspensions is reported, and the polymer is printed as HPMC AS-MF, ambiguous between HPMC and HPMCAS tres2017 X11, X14, X13.

Counterexamples and conflicts

None found in the library.

Transfer check

(1) For an in vitro metabolism or drug–drug interaction assay with microsomes, measure the aggregation onset of the co-incubated drug in the assay medium, as the authors advise tres2017 C44; (2) for biorelevant media measure the onset in the medium itself; (3) do not carry a buffer onset over to a lipid-containing medium; (4) check whether the polymer used to stop crystallization in the test is itself taken up.

History

2015: Raina and co-workers show that micelles and FaSSIF raise the felodipine onset about 20- to 25-fold (synthesis: 177/8.7 and 215/8.7) and link it to free drug raina2015c R11, C30. 2017: Tres and co-workers show the same for microsomes and liposomes tres2017 R5. 2019: Elkhabaz and co-workers find a 2.5-fold rise (synthesis: 20/8) for posaconazole in FaSSIF elkhabaz2019a R2, R4.

How it was shown

Felodipine at 12, 16 and 19 µM was added to buffer with 100 µg/mL polymer and 0.25 mg/mL microsomes or liposomes; the lifetime (370 nm LED, two-exponential fit) and the emission spectra (excitation 370 nm) were measured only below the onset, where two populations are expected tres2017 P1, P3, P4, P5, C17. For ultracentrifugation 12, 15 and 18 µM were used, with multilamellar vesicles because the unilamellar ones of about 100 nm pelleted poorly; the run was 35,000 rpm (112,303 × g) for 30 min (microsomes, drug alone) or 60 min (vesicles), and the supernatant was assayed by HPLC (mean of two, range); the liposome values were also corrected for phospholipid remaining in the supernatant by an assay that is not described tres2017 P7, R9, X11, X12.

What the data show

In microsomes the average lifetime was 0.87, 0.91 and 0.95 ns at 12, 16 and 19 µM, from τ1 0.52–0.59 ns (86–87%) and τ2 3.14–3.22 ns (13–14%); in liposomes 0.71, 0.73 and 0.78 ns, from τ1 0.37–0.45 ns and τ2 1.66–1.70 ns at 26–27%; in buffer it was 0.42, 0.46 and 0.45 ns tres2017 R7, R1.a–c. The band maximum was about 253, 311 and 358 a.u. in microsomes and about 256, 314 and 361 a.u. in liposomes against about 80, 107 and 151 a.u. for felodipine alone at 12, 16 and 19 µM (ratios 3.2 to 2.4; synthesis: 253/80 = 3.2; 358/151 = 2.4), with the maximum at about 430–435 nm against about 445 nm; microsomes and liposomes alone gave about 0 a.u. tres2017 R8. In the supernatant felodipine was about 0.6, 1.0 and 1.2 µM with microsomes and about 0.95, 1.2 and 1.3 µM with liposomes (corrected; about 3.5, 4.4 and 5.0 µM uncorrected) for 12, 15 and 18 µM added, against about 8.6, 10.7 and 12.4 µM for felodipine alone tres2017 R6. As fractions of the amount added: about 5–7% with microsomes, 7–8% with liposomes and 69–72% without vesicles (synthesis: 0.6/12 = 5%; 1.2/18 = 7%; 0.95/12 = 8%; 1.3/18 = 7%; 8.6/12 = 72%; 12.4/18 = 69%) tres2017 R6, X12. The authors call the lifetimes "significantly increased" (no statistical test), the spectra a clear indication of a more non-polar environment, and the supernatant "drastically reduced" tres2017 C27, C29, C31, X7.

Why

The authors' explanation (their interpretation): the longer lifetimes show felodipine in a more non-polar environment, from interaction with hydrophobic regions of enzymes and lipid components in microsomes and with the hydrocarbon bilayer in liposomes; the brighter blue-shifted band confirms it and the supernatant shows removal of drug by the vesicles tres2017 C28, C29, C31. Longer lifetimes after binding or absorption are known for other dyes and molecules (NADPH 0.43 ns free against 2.8 ns protein-bound, retinol palmitate 2.17 ns in egg phosphatidylcholine liposomes), which the authors use by analogy tres2017 C45, R16.h, R16.i. Only a lipid interaction is possible in liposomes, whereas proteins may also bind in microsomes; which component binds was not tested tres2017 C45. The liposome result differs in kind: its τ2 is not longer than the 1.58–1.90 ns felodipine shows in buffer, and the higher average comes from a larger amplitude (26–27% against 4–6%) tres2017 X4, R1.a–c.

Conditions and limits

Felodipine only, 25 °C, one vesicle concentration (0.25 mg/mL) tres2017 P1, P3. No lifetime blank of the vesicle suspensions is reported, and scattering by particles and Raman scattering of water "may contribute somewhat" to the signal, a contribution not quantified for the turbid vesicle media tres2017 C14, X14. Replicates of decays and spectra are not stated tres2017 X7. The ultracentrifuge samples used 12, 15 and 18 µM and multilamellar vesicles, the lifetime samples 12, 16 and 19 µM, and felodipine alone was recovered at only about 69–72% for a reason that is not discussed tres2017 X11, X12.

Counterexamples and conflicts

None found in the library.

Transfer check

(1) Measure total and free drug both (ultracentrifugation or flux) in any medium containing vesicles, membranes or micelles; (2) include a vesicle-only lifetime blank; (3) separate lipid from protein binding with lipid-only vesicles and protein-containing preparations; (4) compare lifetimes by amplitude as well as by component value; (5) a drug that is bound to this extent has a much lower free concentration at a given total, which matters for the amorphous solubility (F12).

History

2017: Tres and co-workers first report felodipine partitioning into microsomes and liposomes with three read-outs tres2017 R6, R7, R8. (synthesis: the lifetime approach was then used in ASD dissolution, F10 and F11, and no other paper of this chapter uses vesicles.)