The marketed difluprednate (DFPN) drug product is an oil-in-water (o/w) nanoemulsion (NE) formulation with trace amounts of smaller microemulsion (ME) globules (Fig. 1) [4, 10]. The NE and ME globules are under collision-induced exchange at a millisecond timescale, leading to peak line-broadening in 19F NMR spectra of DFPN [10]. Upon dilution, the exchange slows due to globule density reduction, resulting in phase resolved 19F peaks. Therefore, the F10 formulation was prepared from directly diluting commercial drug product Durezol® ten times (Table 1 and S1). The F10 sample had narrow 19F NMR linewidths for obtaining more accurate drug distribution results. It also permits the creation of PS-80 spiked formulation using F10 as template.

The chemical structure and microstructure of difluprednate (DFPN) nanoemulsion (NE) formulations. The chemical structure of DFPN, castor oil and polysorbate-80 (PS-80) are shown in (A). The NMR observable nuclei of F28 and F29 in DFPN, methyl and methylene protons in castor oil and PS-80, and polyoxyethylene (POE) protons in PS-80 are annotated. The NE globules become smaller upon addition of PS-80 micelle (B) or transition to much smaller microemulsion (ME) after mechanical perturbation (C)

The formulation F10 contained 0.05 mg/mL DFPN, 4 mg/mL PS-80 and 5 mg/mL castor oil, equivalent to an oil/surfactant (o/s) ratio of 1.25 (Table 1) [4, 17]. The 1D 1H NMR spectrum showed strong signals at 3.69 ppm, 3.55–3.65ppm and 0.8–1.4 ppm, corresponding to methylene protons of poly(oxyethylene) (POE) group in PS-80, glycerol, and the overlapped methyl and methylene protons of PS-80 and castor oil, respectively (Figs. 1A and 2A). The resonance of the DFPN fluorine atom at position 28 (F28) was sensitive to nuclei shielding perturbations from the surrounding solvent molecules, making F28 ideal for the phase distribution study. In the 19F NMR spectra, two resolved peaks were observed at -165.5 ppm and − 165.9 ppm, corresponding to the F28 atoms of DFPN distributed in the surfactant phase (s) and oil phase (o), respectively, which were in slow-exchange and agreed well with the findings of a previous report [10] (Figs. 1A and 2B). The peak fitting results revealed the drug phase distribution (DPD) to be 37+/-2% in surfactant phase (s) and 63+/-2% in oil phase (o) (Fig. 2C).

Overlay of 1D 1H NMR spectra (A) and 19F NMR spectra (B) of NE formulation samples F10 (black), F10-PS80 (red) and F10-shake (blue). The peak fitting was conducted on the F28 peak in the 19F NMR spectra of sample F10 (C), sample F10-PS80 (D) and sample F10-shake (E) with green lines showing fit peaks. Phases of surfactant layer (s), oil core (o) and microemulsion (ME) are labeled in 19F NMR spectra (B, C)

The DOSY results based on the methyl peaks of PS-80 and castor oil showed the hydrodynamic diameter of emulsion oil globule size (OGS) to be 81+/-2 nm (Fig. 3A; Table 1). Orthogonal size measurement on 10-fold diluted F10 sample using DLS showed OGS to be 137.3 +/- 0.7 nm with a polydispersity (Pd) of 10 +/- 2% (Table 2). The apparent disagreement between DOSY-NMR and DLS could be due to the sensitivity of DLS towards larger globules and DOSY towards smaller globules, but the OGS values of 81 and 137 nm are both within number based size distribution results of intact drug products, 106 +/- 50 nm [10].

The 2D 1H DOSY NMR spectra of nanoemulsion (NE) formulation samples F10 (A), F10-PS80 (B) and F10-shake (C)

To prepare the formulation F10-PS80, F10 was spiked with additional 5 mg/mL of PS-80 (Table 1). The 1D 1H NMR spectra showed increased peak intensities at 3.7, 1.3 and 0.9 ppm for F10-PS80 than the F10 sample (Fig. 2A), owning to the PS-80 resonance of its polyethylene (POE) group, methylene groups and methyl group, respectively [18]. The 19F NMR analysis demonstrated a significant difference in drug phase distribution (DPD) between the F10-PS80 and F10. In the presence of increased concentration of PS-80 in the solution, a noticeable increase in the intensity of the peak at -165.5 ppm (DFPN in surfactant phase, peak s) was observed, concomitant with a reduction in the peak at -165.9 ppm (DFPN in oil phase, peak o). Specifically, the surfactant phase peak area increased from 37+/-2% in F10 to 59+/-2% in F10-PS80 when extra PS-80 of 5 mg/mL was added (Fig. 2D; Table 1). The trend of DPD increase in surfactant phase continued when extra PS-80 of 10 mg/mL was spiked into F10 formulation, evidenced in the higher peak intensity at peak (s) of -165.5 ppm (green spectrum in Figure S2). Interestingly, despite a significant amount of surfactant PS-80 added to the sample F10-PS80, and with the PS-80 concentration (9 mg/mL) well above the critical micellar concentration (CMC), no observable difluprednate in micelle particles was detected. This was evidenced by the absence of the expected sharp doublet peaks at -165.5 ppm, which was characteristic of difluprednate distributed in pure micelle [10].

DOSY NMR analysis indicated a reduction in OGS to 52+/-2 nm in sample F10-PS80 as compared with the OGS of 81+/-2 nm in sample F10, when the o/s ratio changed from 1.25 to 0.56 in the emulsions (Figs. 1B and 3A and B; Table 1). This observation was consistent with a previous report suggesting that a lower o/s ratio leads to decrease in emulsion droplet size due to the larger surface area of smaller NE globules [4, 22]. The NE microstructure change upon the addition of pure PS-80 is illustrated in Fig. 1B. Orthogonal size measurement on 10-fold diluted F10-PS80 formulation sample using DLS showed OGS to be 135.4 +/- 0.7 nm with a Pd of 11 +/- 3% (Table 2). The DLS measured size difference between F10 (137.3 nm) and F10-PS80 (135.4 nm) was not obvious, owing to the sensitivity of DLS towards larger globules that could remain in F10-PS80.

To further investigate the impact of manufacturing processing on the microstructure of NE formulation, we subjected sample F10 to vigorous shaking for 72 h, resulting in sample F10-shake. Notably, the F10-shake sample exhibited increased visual transparency starting from day 2, aligning with the historical observation and naming of microemulsions (ME) [23]. The 1D 1H spectrum of F10-shake was visually no different from the spectrum of F10 (Fig. 2A), suggesting the 1D 1H NMR signal was only sensitive to short-range chemical structure but not to microstructure properties. In the 19F spectra, a distinctive peak at -165.6 ppm (peak ME in Fig. 2B) with abundance of 74 +/-2%, specific to difluprednate in ME, was observed for sample F10-shake but not in sample F10 or F10-PS80 [10].

The DOSY NMR confirmed a reduction in OGS to 33+/-1 nm in the F10-shake sample, corresponding to a ME globule size [23] (Figs. 1C and 3C; Table 1). Orthogonal size measurement on 10-fold diluted F10-shake sample using DLS showed OGS to be 57.2 +/- 0.2 nm with a Pd of 15 +/- 2% (Table 2). The DLS measured size reduction from F10 (137.3 nm) to F10-shake (57.2 nm) was consistent in trend with DOSY results in OGS reduction from F10 (81 nm) to F10-shake (33 nm), albeit DLS results weigh towards larger globules that could remain in both samples. The increase in Pd from 10% in F10 to 15% in F10-shake suggested more heterogeneity in globule size distribution upon mechanic perturbation, expected.

These observations unambiguously indicate the conversion of NE globules in sample F10 into the thermodynamically stable and smaller ME globules in sample F10-shake. As a result of this conversion, the distribution of difluprednate in the oil phase of NE globules decreased from 63+/-2% to 26+/-2% in sample F10-shake, as shown in Fig. 2E. Notably, the difluprednate within the major ME globule population and in the surfactant phase of the minor NE globule population were in fast exchange and could not be distinguished using the employed methods.