Reading a drug's own glow: fluorescence lifetime as a detector of drug-rich aggregates and crystals in dilute supersaturated solutions

Background and motivation

Many drug candidates dissolve poorly in water, and formulations such as amorphous solid dispersions, co-crystals and salts are used to raise the concentration in solution above the crystalline solubility (refs 1–3). Such supersaturated solutions are metastable: given time, the drug crystallizes and the concentration falls back to the crystal solubility, at a rate that depends on the drug, the degree of supersaturation and the medium. A second kind of phase change can come first. When the concentration exceeds the solubility of the amorphous drug, the solution undergoes liquid–liquid phase separation (LLPS): a non-crystalline, drug-rich colloidal phase forms in a continuous water-rich phase. These colloids, here called amorphous aggregates, start at 50 to 1000 nm and coalesce with time, and because the drug-rich phase is itself metastable relative to the crystal, it may later crystallize (Raina 2014, ref 4; Indulkar 2015, ref 8; Ilevbare and Taylor 2013, ref 9; Almeida e Sousa 2014, ref 10).

Aggregates matter well beyond formulation. In high-throughput enzyme screens, colloidal ("promiscuous") aggregation is a known source of false positives, through nonspecific binding of the aggregates to proteins (refs 7, 11–13). Trasi and Taylor (ref 14) showed that a second drug present as an aggregated phase lowers the free concentration of another drug. The authors point out that this matters for combination products such as the ritonavir–lopinavir dispersion and for in vitro metabolism assays: if a co-incubated drug forms aggregates and the substrate partitions into them, the lower free substrate concentration could be mistaken for metabolism inhibition. Such non-specific effects, they note, "have not been investigated to date". On the delivery side, aggregates form during dissolution of amorphous dispersions (Alonzo 2011, ref 15), and the drug-rich phase "has been proposed" to act as a reservoir that keeps supersaturation constant (refs 4, 16).

Several tools already detect LLPS: dynamic light scattering and nanoparticle tracking give aggregate size; UV extinction at a wavelength the drug does not absorb gives the onset concentration; UV absorbance follows crystallization; polarized light microscopy confirms that the aggregates are not birefringent; flux measurements across a membrane, combined with ultracentrifugation, report the thermodynamic activity of the drug; 1H NMR has the highest chemical specificity; and the steady-state fluorescence of an environment-sensitive probe changes when the probe partitions into the drug-rich phase (Raina 2015, ref 18). None of these had used fluorescence lifetime. The authors argue that lifetime is attractive because it is an intrinsic property of a fluorophore, largely independent of how it is measured and of concentration, little affected by photobleaching, but very sensitive to the fluorophore's surroundings (ref 19). They expect this to help in two situations where other methods struggle: telling non-crystalline from crystalline particles in turbid samples, and drugs whose LLPS occurs at very low concentration, below about 10 µM, where UV extinction "may lack sufficient sensitivity".

The paper therefore sets out to use fluorescence lifetime, together with steady-state emission spectra, on three model drugs with known LLPS onsets: felodipine (19 µM), clotrimazole (22 µM) and efavirenz (58 µM) (refs 4, 8, 9) M1·M2·M3. Felodipine is strongly autofluorescent, so it can report on its own state, both as amorphous aggregates and as crystals. Clotrimazole and efavirenz do not fluoresce and are read through the probe PRODAN. Finally, felodipine is followed in turbid biological media: rat liver microsomes, the standard system for in vitro metabolism, and phospholipid vesicles (liposomes) as a lipid-only model of the microsomal membranes.

Background concepts

LLPS and the amorphous solubility. A supersaturated solution can hold drug up to the solubility of the amorphous form. Above that concentration the extra drug separates as a second liquid (or glassy) phase rich in drug. The free drug concentration in the water-rich phase then stays at the amorphous solubility, and adding more drug only adds more drug-rich phase. This is the reason the authors give for why, above the onset, more and more probe moves into the drug-rich phase while the free drug concentration does not change. Crystals are different: they grow at the expense of the dissolved drug and pull the concentration down toward the crystalline solubility (2.9 µM for felodipine, ref 5) M1·C5.

Fluorescence lifetime. After a molecule absorbs a photon it stays in the excited state for a short time, typically nanoseconds, before it emits. The average of that time is the fluorescence lifetime. In this paper it is measured by exciting the sample with short pulses from a 370 nm LED and recording how the emitted light decays after each pulse. The decay is then fitted. Because the samples were expected to contain more than one population of molecules (for example free drug and drug in aggregates), every decay was fitted with two exponential terms, giving a short lifetime ${\tau }_1$ and a long lifetime ${\tau }_2$ with amplitude shares $a_1$ and $a_2$ (in percent), and an "amplitude average" lifetime that weights each lifetime by its share (ref 33). A reduced chi-square of 1 or below was taken as a good fit. The paper prints no equations; the fitting model, the average and the goodness-of-fit measure are named in the methods only.

Why lifetime responds to surroundings. The paper cites lifetime as very sensitive to temperature, viscosity, polarity and molecular interactions (refs 19, 22–30). General background added by the AI writer: an excited molecule can also lose its energy without emitting light, and the speed of these losses depends on its environment; in a viscous or rigid environment they slow down and the lifetime gets longer, and many polarity-sensitive dyes emit longer and at shorter wavelength (a blue shift) in non-polar surroundings. The discussion of the paper cites examples: the dye cypate lives about 0.2 ns in water and about 1 ns in chloroform, BADAN about 0.2 ns in water and about 2.4 ns in dimethylformamide, and the pyrene excimer, a rigid structure, 90 ns (refs 22, 25, 38). Binding to proteins or lipid membranes lengthens lifetimes in the same way: nicotinamide adenine dinucleotide phosphate goes from 0.43 ns free to 2.8 ns bound to proteins, and retinol palmitate reaches 2.17 ns in egg phosphatidylcholine liposomes (refs 29, 30) R16.

Steady-state spectra versus lifetime. A steady-state emission spectrum records the intensity of light emitted at each wavelength under continuous excitation. Its intensity depends on concentration, dilution and photobleaching, while the lifetime, within limits, does not. The two read-outs are complementary: the spectrum shows where the emission lies (here a blue shift signals a less polar environment), the lifetime shows how long the molecules stay excited.

PRODAN. 6-Propionyl-2-dimethylaminonaphthalene is a naphthalene-derived dye whose emission is very sensitive to the polarity of its surroundings (ref 41). In polar, hydrogen-bonding solvents such as water and methanol its decay has two components, the shorter one attributed to hydrogen-bonded dye–solvent complexes; in non-polar or aprotic solvents it decays with a single lifetime (ref 42). Its aqueous solubility is about 3.5 µM, and supersaturated PRODAN self-associates, which shows up as an extra shoulder on the blue edge of its spectrum (ref 43). It was used here at 0.9 µM M4·R16.m·R16.n.

UV extinction. At a wavelength where the drug does not absorb (450 nm here), the apparent absorbance of a solution rises only when particles scatter light. Following this "extinction" while drug is added continuously gives the concentration at which a particle phase appears. In turbid suspensions the measurement still works if the suspension without drug is used as the blank.

Liver microsomes and liposomes. Liver microsomes are vesicles formed from the endoplasmic reticulum of liver cells. They carry the enzymes of oxidative metabolism (cytochrome P450 isoforms, esterases) and of conjugation (glucuronosyltransferases), and are the standard in vitro system for metabolism and drug–drug interaction studies (ref 39). Liposomes are vesicles made of phospholipid bilayers; here they serve as a model that takes up a lipophilic drug into its bilayer, mimicking the lipid part of the microsomes without the proteins.

Methods in detail

The study has four parts, all run in 100 mM sodium phosphate buffer at pH 7.4 and 25 °C. Drugs were added from concentrated methanol stocks to 10 mL of buffer stirred at 300 rpm, and, except in the crystallization run, 100 µg/mL of the polymer the paper calls "Hydroxypropyl methylcellulose (HPMC) AS-MF grade" was dissolved beforehand to stop the drug from crystallizing during the test P1·M5. The method flow below shows how the parts fit together.

Instruments, common to all parts. Lifetimes were recorded on an EasyLife V system (Optical Building Blocks) with a nanosecond pulsed 370 nm LED. An emission filter with a cut-off of about 400 nm passed the fluorescence (about 400–650 nm) and blocked scattered excitation light, but the authors note that light scattered by the aggregates and the Raman scattering of water "may contribute somewhat" to the signal. Decays were collected in 500 channels (start and end delays 60 and 90 ns); the instrument response was measured with a Ludox colloidal silica scatterer, and the decays were fitted by reconvolution in DecayFit 1.4 by minimising chi-square. The number of repeat decays per sample is not printed. Steady-state spectra were taken on a Shimadzu RF-5301PC (excitation 370 nm, emission 370–600 nm; slits 1.5/3 nm for felodipine and 1.5/5 nm for PRODAN) P4·P5·C14·C15·C16.

Contribution 1: felodipine as its own indicator of aggregation

The problem. Probe-based fluorescence needs an added dye whose behavior must itself be understood. Felodipine, a 1,4-dihydropyridine calcium channel blocker (Figure 1a), is strongly autofluorescent, so its own emission can report whether it is dissolved or in a drug-rich phase. What they did. Felodipine was prepared at 12, 16, 19, 24, 28 and 38 µM, straddling its reported LLPS onset of 19 µM, and the decay and the emission spectrum of each solution were recorded P1·R1·R3. Why it works, in the authors' explanation. Free felodipine sits in water; in the aggregates it is surrounded by other felodipine molecules, an environment that is less polar and is expected to be more viscous, both of which lengthen lifetimes and shift emission to the blue. The authors call felodipine a "self-indicator" C39·C40. Difference from earlier work. Earlier fluorescence work on LLPS used the steady-state spectra of added probes (ref 18); here the drug's own lifetime is the read-out.

Contribution 2: a lifetime signature for crystals

The problem. At low overall concentration, aggregates and crystals can both be present, and they have opposite consequences: aggregates co-exist with a supersaturated solution, crystals deplete it. What they did. A felodipine solution was made at 108 µM without polymer, far above the crystalline solubility of 2.9 µM, so that it would crystallize, and its decay was recorded after 15 h P2·M1. The authors' reasoning. In a crystal the molecules are held in a rigid lattice; the authors expect, and cite examples of, rigid or rigidified dyes with long lifetimes (pyrene excimer 90 ns; dyes in gelatin or at −185 °C) R16.e·R16.f·R16.l. No microscopy or diffraction of the sample is reported.

Contribution 3: felodipine in liver microsomes and liposomes

The problem. In a metabolism assay the drug sits in a turbid suspension of microsomes that contain hydrophobic proteins and lipid membranes. Light scattering by the suspension hampers optical methods, and the drug may partition into the vesicles. What they did. Three experiments were run with rat liver microsomes (pooled, male Sprague Dawley) or egg-phosphatidylcholine liposomes, each at 0.25 mg/mL M7·M8:

• UV extinction: a methanolic felodipine stock was pumped at 25 µL/min into the suspension (with 100 µg/mL HPMC), the extinction at 450 nm was read every 40 s with a 1 cm dip probe, and the drug-free suspension served as the blank; mean of two runs P6.
• Lifetimes and spectra at 12, 16 and 19 µM, below the felodipine onset in buffer. The authors restricted themselves to this range because there only two felodipine populations (free and vesicle-bound) are expected; above it several populations would be "extremely difficult to model" P3·C17.
• Ultracentrifugation: felodipine (12, 15 and 18 µM added) alone or with vesicles was spun at 112,303 × g for 30 min (alone, microsomes) or 60 min (liposomes), and the felodipine left in the supernatant was measured by HPLC (C18 column, acetonitrile–water 70/30, 210 nm). Multilamellar liposomes were used here because the about 100 nm unilamellar ones pellet poorly; the authors "expect" this not to affect their conclusions, since they use liposomes only to mimic uptake into the microsomes, not to measure encapsulation. Liposome values were also "corrected" for phospholipid left in the supernatant, by an assay the paper does not describe P7·R9·C30.

The liposomes were made by thin-film hydration: 77 mg egg phosphatidylcholine in 5 mL chloroform, evaporated at 40 °C, dried overnight, hydrated with 5 mL water (multilamellar vesicles), incubated at 40 °C for 30 min, and probe-sonicated twice for 3 min at 0 °C to give unilamellar vesicles; sizes were checked by DLS (Figure S2, not supplied) P8·P9. The authors' reasoning. If felodipine enters the hydrophobic parts of the vesicles, its free concentration falls, more total drug is needed to reach the amorphous solubility, and the bound drug should show a longer lifetime and a blue-shifted band; ultracentrifugation tests the uptake directly by removing the vesicles C25·C23.

Contribution 4: a probe for drugs that do not fluoresce

The problem. Most drugs, including clotrimazole (an imidazole antimycotic) and efavirenz (a non-nucleoside reverse transcriptase inhibitor), do not fluoresce. What they did. PRODAN was dissolved at 0.9 µM, below its aqueous solubility of about 3.5 µM, and clotrimazole (0, 5, 11, 16, 27, 33, 38 µM) or efavirenz (0, 20, 40, 49, 69, 89, 109, 128, 148 µM) was added; lifetimes and spectra of the probe were recorded P1·M4. The authors' reasoning. A probe of moderate lipophilicity stays in water while the drug is dissolved but partitions into the drug-rich phase once it forms; there its surroundings become non-polar, so its lifetime lengthens and its emission moves to the blue (ref 18) C46·C37. Difference from earlier work. The same idea had been used with steady-state spectra (ref 18); here the probe's lifetime is added.

Experimental analysis

Felodipine in buffer: the lifetime rises at the LLPS onset. Table A reproduces Table 1 of the paper. From 12 to 19 µM the average lifetime stays at 0.42–0.46 ns and the decay is dominated by the short component (0.34–0.40 ns, 94–96%). From 24 µM on, the share of the long component grows from 14 to 27 to 39%, its lifetime staying at 1.87–2.04 ns, and the average climbs to 0.53, 0.71 and 0.96 ns at 24, 28 and 38 µM (Figure 2a) R1. The authors place the transition from a single-phase solution to one containing amorphous aggregates at 19 µM, "in excellent agreement" with the earlier value of 19 µM, and read the growing long component as "indicating" aggregates co-existing with free drug C18·C19.

Table A. Two-exponential fits of felodipine decays (from Table 1 of the paper; pH 7.4 buffer, 25 °C, 100 µg/mL HPMC except the last row) R1.

The steady-state spectra tell the same story (Figure 2b). Up to 19 µM the band stays near 445 nm; above it the band moves to about 435 nm and grows steeply. The intensity at 435 nm is about 78, 100 and 146 a.u. at 12, 16 and 19 µM and about 254, 367 and 633 a.u. at 24, 28 and 38 µM (read from Figure 2a, approximate) R2·R3. The authors attribute the change to felodipine "being in a less polar environment as a result of the self-association", and see the intensity rise as "in good agreement" with the lifetimes C20·C21.

Note by the AI writer (not discussed in the paper): the text says that the maximum emission wavelength and intensity "remain constant up to a concentration of 19 µM", but in Figure 2 the intensity nearly doubles between 12 and 19 µM, roughly in proportion to concentration, before the steeper rise; the onset rests on the change of slope, not on a flat baseline X1. Also, a long component of 1.58–1.90 ns is already fitted below the onset, at 4–6%; what marks the aggregates is the growth of its share rather than the appearance of a new lifetime, so the "~2 ns" that the abstract assigns to aggregates is not unique to them X3. The tabulated averages equal the share-weighted mean of the two lifetimes within rounding, which the AI writer checked for all 29 table rows E2.

Felodipine crystals: a 9 ns component. In the 108 µM solution without polymer, after 15 h, the decay held a short component of 0.14 ns (83%) and a long one of 9.04 ns (17%), with an average of 1.62 ns (Table A, last row) R1.g. The long component is four to five times the 1.87–2.04 ns seen with aggregates. The authors write that it "can be attributed to the evolving crystalline phase", and that the long crystal lifetime is "most likely due to being present in a rigid framework in the crystal lattice". They value the distinction because aggregates co-exist with a supersaturated solution while growing crystals deplete it C22·C41·C42. The crystallinity of this one sample was not checked by microscopy, diffraction or a concentration drop X6.

Microsomes and liposomes: a higher onset. With 0.25 mg/mL microsomes the extinction at 450 nm stayed near baseline (about 0.01–0.02) up to about 100 µM felodipine and then climbed to about 0.39 at about 280 µM; with liposomes it stayed low to about 150 µM, rose slowly (about 0.05 at 200 µM) and reached about 0.38 at about 313 µM, with wide ranges between the two runs above about 220 µM (Figure 3a, approximate). The text gives the onsets as about 100 µM (microsomes) and about 160 µM (liposomes), against 19 µM in plain buffer R5·M1. The authors call the increase drastic and explain it as "likely due to the drug being absorbed by the hydrophobic portions, reducing its free concentration in solution and requiring higher total concentrations to achieve the amorphous solubility", adding that surfactant micelles have a similar effect (ref 5) C25·C26.

Below the onset, felodipine goes into the vesicles. Three read-outs at 12–19 µM support the uptake (Table B). The average lifetime roughly doubles in microsomes (0.87–0.95 ns) and rises to 0.71–0.78 ns in liposomes, against 0.42–0.46 ns in buffer (Table 2). The emission band is about 2.4–3.2 times brighter and shifted from about 445 nm to about 430–435 nm, while microsomes or liposomes alone emit nothing (Figures 3c, d). And after ultracentrifugation only about 0.6–1.2 µM (microsomes) or about 0.95–1.3 µM (liposomes, corrected) remains in the supernatant, against about 8.6–12.4 µM for felodipine alone (Figure 3b) R6·R7·R8. The authors say the lifetimes "suggest" a more non-polar environment, that the spectra "clearly" indicate it, and the supernatant data as showing removal of the drug from solution; in microsomes they attribute the long lifetime to "interactions with hydrophobic proteins and/or lipid bilayer components", in liposomes only the lipids are available C28·C29·C31·C45.

Table B. Felodipine below its onset in 0.25 mg/mL microsomes and liposomes: lifetimes from Tables 1 and 2 of the paper; band maxima read from Figures 3c, d (approximate) R1·R7·R8.

Table C. Felodipine (µM) left in the supernatant after ultracentrifugation, read from Figure 3b (approximate; mean of two runs). Liposome values corrected for phospholipid remaining in the supernatant R6.

Note by the AI writer (not discussed in the paper): two details qualify this picture. First, in liposomes the long component is not longer than felodipine's own long component in buffer (1.66–1.70 against 1.58–1.90 ns); what changes is its share (26–27% against 4–6%). The statement that the increase is "due to a longer τ₂ component" fits the microsomes (3.14–3.22 ns) but not the liposomes X4. Second, felodipine alone is recovered at only about 70% of the amount added (about 8.6 of 12 µM), although these solutions are below the LLPS onset; the loss is not discussed, so the vesicle results are best compared with the drug-alone bars rather than with the amounts added X12. The paper also does not state which liposome type, unilamellar or multilamellar, was used for the extinction and fluorescence runs X11.

Clotrimazole and efavirenz: the probe reports LLPS. PRODAN alone in buffer decays with 0.50 ns (66%) and 1.77 ns (34%), an average of 0.93 ns, which the authors find "in a good agreement" with the 0.70 ns (60%) and 2.0 ns (40%) reported in water (ref 37) R10.a·R15·C32. With clotrimazole the average stays at 0.89–0.94 ns up to 16 µM and rises to 1.19, 1.28 and 1.32 ns at 27, 33 and 38 µM; with efavirenz it stays at 0.90–0.98 ns up to 49 µM and rises to 1.16, 1.53, 1.77, 1.80 and 1.97 ns from 69 to 148 µM (Tables 3 and 4, Figures 4a, b) R10·R11. The long component lengthens with the drug-rich phase: from 1.89 to 3.40 ns with clotrimazole and from 2.05 to 3.89 ns with efavirenz. In the spectra (Figures 4c, d) PRODAN alone emits at 520 nm; above the onset a second band appears at 435 nm with clotrimazole and at 450 nm with efavirenz and grows with concentration, while 38 µM clotrimazole or 148 µM efavirenz without probe gives no emission R13. The authors attribute the change to "a second phase, i.e., the dye is now also present in the newly-formed drug aggregates", conclude from the unchanged spectra below the onset that PRODAN does not interact with the free drug, and find the onsets "in good agreement" with the earlier UV-extinction values of 22 and 58 µM C33·C36·C35. They also rule out self-aggregation of the probe at 0.9 µM: the extra blue-edge shoulder of supersaturated PRODAN is absent C47.

Table D. LLPS onsets found by fluorescence against the values reported earlier (refs 4, 8, 9). The paper states the onset as the last unchanged concentration; the change is first seen at the next tested one R4·R14·M1·M2·M3.

Note by the AI writer (not discussed in the paper): the agreement in Table D holds within the spacing of the tested concentrations (brackets of 5, 11 and 20 µM), and each series is a single run as far as the paper states X8·X7. The component shares behave less simply than the authors' picture of more probe moving into a growing drug-rich phase: with clotrimazole the share of the long component falls from 30% to 24–26% above the onset while its lifetime lengthens, and only with efavirenz above 109 µM does the share rise (35–42%) X5. The lifetimes plotted in Figure 4 sit about 0.01–0.02 ns above the values in Tables 3 and 4 (for example 11 µM clotrimazole: 0.89 ns in the table, about 0.91 ns in the figure), and the PRODAN-alone fit in Table 3 has a reduced chi-square of 1.26, above the "≤ 1" criterion of the methods; its repeat in Table 4 gives 0.90 ns with a chi-square of 0.85 X9·X10.

Discussion

How the authors tie the results together. In their account the fluorophore reports on its immediate surroundings. Dissolved felodipine, or PRODAN next to dissolved drug, sits in water and has a short lifetime. Once the free concentration reaches the amorphous solubility a drug-rich phase forms; felodipine self-associates in it, and PRODAN partitions into it. The path then splits.

The first branch rests on the authors' statement that "a combination of factors is likely to contribute" to the longer lifetime in aggregates: a microviscosity that "is expected to be higher" than in solution, and the lower polarity seen as the blue shift in Figure 2b. The second rests on the rigidity of the crystal lattice, called "most likely". Neither microviscosity, polarity nor rigidity was measured; the support comes from the cited behavior of other dyes C40·C41·R16.

A third case covers the biological media. In microsomes and liposomes, felodipine partitions into hydrophobic proteins and lipid membranes even below its onset; this lengthens its lifetime and blue-shifts its band, lowers its free concentration, and so raises the total concentration needed for the free concentration to reach the amorphous solubility, which the authors give as about 100 µM with microsomes and about 160 µM with liposomes. They describe the mechanism as "likely" and point to surfactant micelles as an earlier example of the same effect C25·C26·C45.

Where the approach applies. The felodipine experiments show that an autofluorescent drug can report its own aggregation in buffer, with the lifetime and the steady-state intensity changing between the same two concentrations. For non-fluorescent drugs the probe approach worked for two drugs with onsets of about 20 and 60 µM. The authors see value for in vitro metabolism and drug–drug interaction assays, where "it is important to determine at what concentration aggregates form in the presence of microsomes", and they conclude that fluorescence techniques "represent a promising approach to identify phase transformations, even in complex media" C44·C51. In the microsome and liposome media, however, the fluorescence measurements stop below the onset; the onset itself comes from UV extinction C17·X16.

Note by the AI writer (not discussed in the paper), on what this suggests for formulation. The vesicle experiment is a clear demonstration that any medium that takes up drug shifts the LLPS onset to a higher total concentration: here about 5- to 8-fold for felodipine at 0.25 mg/mL of vesicles. An amorphous solubility measured in plain buffer is therefore a limit on the free drug concentration, not on the total concentration, in lipid- or protein-rich media, and dissolution tests in such media should be read with that in mind. A lifetime component far longer than that of aggregates could flag crystallization in turbid samples where UV methods cannot tell particle types apart, but this paper supports that with a single sample, so an orthogonal check of crystals would be needed before relying on it.

Limitations

Limits stated by the authors. Light scattered by the aggregates and Raman scattering of water "may contribute somewhat" to the measured signal C14. In microsomes and liposomes, lifetimes were measured only at 12–19 µM, because above the onset several populations would be "extremely difficult to model" C17. For ultracentrifugation multilamellar vesicles replaced the unilamellar ones; the authors "expect" this not to affect their conclusions, since quantitative encapsulation "is not our primary interest" C30.

The AI writer's own judgement. Overall the evidence is of moderate strength. What holds well: for every drug, lifetime and emission change between the same two tested concentrations, and these brackets contain the onsets reported earlier by UV extinction; felodipine uptake by the vesicles is seen by three independent read-outs. What weakens it:

• Each onset is a bracket set by the tested concentrations (19–24, 16–27 and 49–69 µM), and the number of repeat decays and spectra is not stated; only Figures 3a and 3b give two runs X8·X7.
• The crystal lifetime comes from one sample whose crystallinity was not confirmed by another method X6.
• The two exponential components are not clean populations: a ~2 ns component is fitted below the felodipine onset, the liposome long component is not longer than in buffer, and the PRODAN shares do not follow the amount of drug-rich phase X3·X4·X5.
• Text and figures disagree in small ways: felodipine intensity called constant up to 19 µM though it nearly doubles, PRODAN lifetimes plotted about 0.01–0.02 ns above the tables, a fit chi-square of 1.26 against the stated criterion of 1, and a fraction "higher than 94%" where the table lists 94% X1·X9·X10·X2.
• The claimed advantage for drugs whose LLPS lies below 10 µM was not tested: all onsets found here are 16 µM or higher, and lifetime and UV extinction were not compared on the same solutions X15.
• No lifetime blank of the vesicle suspensions is reported, the vesicle type used for extinction and fluorescence is not stated, and felodipine alone is recovered at only about 70% after ultracentrifugation X14·X11·X12.
• The polymer is printed as "Hydroxypropyl methylcellulose (HPMC) AS-MF grade"; AS-MF is a grade name of hydroxypropyl methylcellulose acetate succinate, so which polymer was used is unclear X13.
• All work was done at 25 °C in one buffer with one polymer level, and the biological media were tested with felodipine only.

Conclusion

The paper shows, for three model drugs in pH 7.4 buffer, that fluorescence lifetime together with steady-state spectra picks up the formation of drug-rich aggregates at concentrations that bracket the LLPS onsets measured earlier by UV extinction: felodipine through its own autofluorescence (average lifetime from about 0.45 to 0.96 ns), clotrimazole and efavirenz through the probe PRODAN. A crystallizing felodipine solution adds a 9.04 ns component, which the authors attribute to crystals. In rat liver microsomes and liposomes felodipine is taken up by the vesicles, which shows as longer lifetimes, a brighter blue-shifted band and much less drug in the supernatant, and the aggregation onset moves from 19 µM to about 100 and 160 µM. The explanations for the longer lifetimes (microviscosity, polarity, lattice rigidity, binding) are drawn from other dyes and were not measured, the crystal signature rests on one sample, and inside the vesicle media the onset rests on UV extinction. The authors "believe" fluorescence techniques are "a promising approach" for phase transformations in complex media; they present the aggregation onset in microsomes as the starting point for studying how aggregates affect drug–drug interaction assays R1·R5·R6·R7·R8·R10·R11·C22·C25·C51.