Analyte solutions were prepared at 1 mmol/L in Millipore water or cosolvents as indicated. The MDAP∙CB8 reporter pair solution contained 5 μmol/L of MDAP and 6 μmol/L of CB8 that were dissolved in Millipore water. Two lipid solutions were used, including the commercial Avanti PAMPA lipid blend I and the self-prepared 2% DOPC in dodecane. To proceed with the assay, 5 μL of the lipid solution was pipetted to the microporous filter and allowed to impregnate evenly; 300 μL of the analyte solution was transferred to the lower donor well, and 100 μL of the reporter pair solution was transferred to the upper acceptor well. The acceptor plate was combined with the donor plate after the solutions had been loaded to define the start point of the assay. Fluorescence readings were taken on a Jasco FP-8500 fluorimeter equipped with a microplate reader with an excitation wavelength at 418 nm and monitored at an emission wavelength of 449 nm.

Analyte screening and solvent effect studies were carried out in 96-well microplates; 150 μL of the reporter pair solution was pipetted into each well, followed by the addition of 15 μL of the analyte solution or solvent. For analytes that were insufficiently water-soluble but that dissolved with organic cosolvents, the final organic content was 2% after mixing with the reporter pair solution. Fluorescence was measured with the same parameters as for PAMPA. Intensity changes upon addition of the analyte solutions were recorded, and solvent-induced changes (background) were subtracted.

A recent report on the permeation of analytes across liposomal membranes tested MDAP∙CB8 and other fluorescent complexes as reporter pairs, providing us valuable information on the potential analytes that could be first implemented in the RT-PAMPA format (Biedermann et al., 2020). To establish RT-PAMPA, we started with a small-scale analyte screening procedure, in order to eliminate false-negative results caused by the lack of response to the artificial receptor. These experiments could be performed in homogeneous solution, because only the fluorescence response of the receptor toward the analyte is being determined. A set of 29 analytes prepared in aqueous solution or with organic cosolvents has been examined for their response toward MDAP∙CB8. The fluorescence intensity changes upon addition of the analytes are visualized in Figure 1. Most analytes caused a significant decrease of the fluorescence intensity, and changes >10% were considered to be sufficiently strong to be readily followed by RT-PAMPA, where a lower sensitivity was expected because of the temporal requirement of permeation. Indeed, the concentrations of the analytes in the fluorescence response experiments were conservatively set as 0.1 mmol/L, below the test concentrations used in the subsequent RT-PAMPA (1 mmol/L), such that already the permeation of a fraction of analyte from the donor to the acceptor well should ensure a fluorescence response in the upper.

The compounds that caused the largest fluorescence changes have several common structural features. MDAP possesses a positively charged conjugated organic moiety, which is predestined to get engaged in electrostatic interactions with electron-rich structures (e.g., indoles and coumarins). On the other hand, because of the size selectivity of the CB8 cavity with MDAP inside, large molecules and those with large branched residues showed no or weak associative binding (e.g., quinine, furosemide, primaquine, zolmitriptan, 1-adamantylamine, and 1-adamantanol). In general, an electron-rich, conjugated linear structure is preferred by the selected reporter pair. Naturally, because the sensing in RT-PAMPA is based on a molecular recognition process with an associated selectivity, the range of accessible analytes for a particular FAR is limited. Conversely, the technique is not limited to a single receptor. For example, berberine in combination with cucurbit[7]uril and lucigenin in a pair with p-sulfonato-calix[4]arene are alternative reporter pairs, which respond to different classes of analytes, e.g., adamantyl and polycyclic aliphatic derivatives for the former (Megyesi et al., 2008; Moghaddam et al., 2011) and compounds with tethered trimethylammonium groups for the latter (Arena et al., 2004; Guo et al., 2008, 2011). Through a combination of different receptors, the potential of RT-PAMPA can be expanded, although we concentrate herein on the demonstration for a single one, MDAP∙CB8.

Buffers and organic cosolvents are widely used in traditional PAMPA to increase drug solubility in water. Pharmaceutical libraries routinely involve compound storage in dimethyl sulfoxide (DMSO), and compatibility with this cosolvent always needs to be ensured. Accordingly, we examined the influence of organic cosolvents, including water mixtures with DMSO, methanol, and ethanol. These control experiments were important, because the molecular-recognition process with synthetic macrocyclic cavities, such as that in CB8 or the MDAP∙CB8 complex, is frequently driven by hydrophobic interactions, which are being modulated in different solvents.

The final concentration of the organic proportion ranged from 1 to 10% upon direct addition into the reporter pair solutions. As expected from a leveling of the hydrophobic driving force for receptor–analyte (host–guest) complexation in organic solvents, the fluorescence intensity decreased in all cases, and the effect became significant as the organic proportion increased (Figure 2), with the largest drop (by 29%) being observed for 10% DMSO content. However, the response of the FAR was quite stable at low organic solvent proportion (1%), with the fluorescence intensity being reduced by only 5, 3, and 6% with methanol, ethanol, and DMSO, respectively. Although the results suggested that high proportions of organic cosolvent needed to be avoided, because they would reduce the sensitivity of the assay, the RT-PAMPA application should be sufficiently robust to allow common low DMSO proportions, and most importantly, the effect of organic cosolvents on the fluorescence response of the receptor was found to be predicable and systematic.

In PAMPA, a lipid barrier needs to be crossed when an analyte moves from the donor plate into the acceptor plate. The initial PAMPA used egg lecithin as the lipid bilayer membrane (Kansy et al., 1998; Faller, 2008); other options such as hexadecane (Wohnsland and Faller, 2001) and lipid mixtures (Sugano et al., 2001) have been developed subsequently. Herein, we selected a commercial Avanti lipid solution with good fluidity in all PAMPA experiments. As an alternative common lipid, 2% DOPC in dodecane solution has been used for comparison. Before proceeding to the RT-PAMPA experiments, we monitored the stability of the FAR in the acceptor well when in contact with the donor well in the absence of added analyte. Over a period of 7-h incubation time, a slight decrease of the fluorescence was monitored, which amounted to ca. 5 ± 2% for the 2% DOPC lipid and to 12 ± 3% for the Avanti premixed lipid (see Supplementary Material). Presumably, this decrease reflects some dissolution of the receptor, or a component of it, in the lipid layer. Interestingly, this drop was independent, within error, of the presence of organic cosolvent (up to 20%) in the donor layer, which demonstrates that the diffusion of the polar organic cosolvents through the lipid does not contribute to the fluorescence changes in this time scale.

In the actual RT-PAMPA experiments with MDAP∙CB8 and the Avanti lipid (Figure 3A), the fluorescence decrease recorded from the different analytes needs to be compared with the fluorescence response in the absence of analyte, where also a slight decrease is observed (see previous section). Accordingly, significant fluorescence responses, indicative of analyte permeation through the lipid layer, were observed for umbelliferone, lansoprazole, l-tryptophanamide, coumarin, propanil, thiabendazole, and indole. l-Phenylalanine, serotonin, doxepin, and zolmitriptan showed no significant fluorescence response in comparison to the absence of analyte. Because these analytes show a strong fluorescence response in homogeneous solution (Figure 1), they do not or too slowly permeate through the lipid layer in RT-PAMPA.

For compounds that show insignificant fluorescence response/binding toward the MDAP∙CB8 reporter pair, such as 1-adamantylamine or 1-adamantanol (Figure 1), other reporter pairs can be in principle employed. To demonstrate, we conducted exploratory experiments (see Supplementary Material) with the berberine∙cucurbit[7]uril reporter pair, which showed a strong fluorescence response, indicative of effective permeation, toward 1-adamantylamine and 1-adamantanol. The variability of the reporter pairs presents an additional asset of RT-PAMPA, which at the same time limits its general applicability for screening larger libraries with a single reporter pair. Among the tested reporter pairs, MDAP∙CB8 was, however, found to surpass the other candidates in terms of accessible analyte scope.

The RT-PAMPA was carried out with two kinds of lipid membranes, the commercial Avanti PAMPA blend I and the laboratory prepared 2% DOPC dodecane solution (see Supplementary Material for all results with 2% DOPC). A better resolution between the different analytes was observed with the commercial lipid layer, possibly because it is more permeable by dissolving 1% ethanol and stearic acid in dodecane besides just DOPC. The comparison also afforded differences in permeation profiles of some analytes. For example, umbelliferone, which showed an appreciable permeability through the Avanti lipid (Figure 3A), did not permeate to a significant extent through the 2% DOPC lipid under RT-PAMPA conditions. As a particular advantage of RT-PAMPA, fast and slow permeation characteristics can be immediately differentiated from the permeation curves. Fast permeation is reflected by a sharp drop of the fluorescence during the initial several minutes, e.g., for coumarin, propanil, thiabendazole, and indole, revealing that a significant amount of the analyte has rapidly crossed the membrane model. In contrast, slowly permeating compounds such as umbelliferone, lansoprazole, and l-tryptophanamide showed a fluorescence response on the time scale of several hours in RT-PAMPA. As expected, the kinetics of the permeation profiles (slow, moderate, and fast) nicely correlated with the known permeabilities of the analytes. For instance, the permeability coefficients (logP_e) of indole and coumarin are reported as −4.50 and −4.55, respectively, which are considered fast (Fujikawa et al., 2007). In contrast, doxepin, whose permeability coefficient is < −6 in acidic solutions (which matches our experimental condition for using doxepin hydrochloride in the donor well), is categorized as slow permeation (Oja and Maran, 2015). Meanwhile, the performance of l-tryptophanamide is moderate to slow with a permeability coefficient of −5.56 (Fujikawa et al., 2007). It should also be noted that simple (single-parameter) empirical correlations between the permeability coefficients and other physicochemical parameters of the analytes (such as their molecular size or their lipophilicity as assessed by octanol–water partition coefficients) do not apply, which emphasizes the need to experimentally determine the permeabilities of analytes with dedicated assays, such as RT-PAMPA (Fujikawa et al., 2007; Biedermann et al., 2020).

RT-PAMPA is in principle also suitable to extract absolute permeation coefficients and rates, although these are strongly dependent on the actual experimental conditions (see above) and can only serve as a relative comparison and/or rough method-to-method comparison. The quantitative analysis requires that a plateau value for the final fluorescence intensity has been reached, which is indicative of analyte diffusion having reached equilibrium and/or complete consumption of the reporter pair. To differentiate between the two options, experiments at different analyte concentrations can be performed. For indole, see Figure 3B; these experiments (from 0.01 to 1 mmol/L) revealed variations of the plateau with varying indole concentration, which signifies that the diffusion equilibrium is actually being attained and that the artificial receptor is not being consumed; in the latter case, the same plateau would be reached with different rates. Under those conditions of equilibration and with the knowledge of the binding affinity between the analyte and the reporter pair, the absolute permeability coefficients of the analyte can be extracted (see Supplementary Material). The effective permeability coefficient of indole (log P_e) was calculated as −4.57 (as an average value), which compares favorably with that reported by others (log P_e = −4.50, with 10% egg lecithin solution as the lipid barrier) (Fujikawa et al., 2007). A closer inspection of the permeability coefficients calculated for different time intervals or for different indole concentrations (see Supplementary Material) shows small systematic trends (smaller values at higher analyte concentrations), which may be related to the almost complete consumption of the receptor at higher analyte concentrations and an associated larger error. Important to note, the presence of organic cosolvents in the donor well had a minor influence on the absolute permeation kinetics. To provide a specific example, we prepared 0.1 mmol/L thiabendazole samples in both 2 and 20% ethanol, and the resulting permeation curves overlapped within the time of largest changes (initial hour, see Supplementary Material).