Datasets:

Hantao

/

ChemImages

like 1

2985_chem201901961-fig-0005.jpg

Figure 5Open in figure viewerPowerPointSurfactants controlling swarming behavior of P. aeruginosa. a) Biosynthesis of rhamnolipids and swarming pattern of genetic knockout strains of the indicated biosynthesis genes. b) Synthetic surfactants modulating or inhibiting swarming motility of P. aeruginosa.

2985_chem201901961-fig-0006.jpg

Figure 6Open in figure viewerPowerPointFlagellar motor assembly of H+- and Na+-driven flagella and compounds interfering with motor function causing swarming inhibition.

2985_chem201901961-fig-0007.jpg

Figure 7Open in figure viewerPowerPointSwarming and bioavailability of ferric iron. a) Avaroferrin produced by Shewanella algae blocks iron-dependent swarming motility of Vibrio alginolyticus. b) Chelation of ferric iron by the ICDH-coumarin 42 switches off RssAB two-component system signaling and thereby triggers swarming.

2529_fig1.jpg

Figure 1Open in figure viewerPowerPointMechanism of (a) competing hydrolysis and (b) transphosphorylation involving a covalent phospho-enzyme intermediate catalyzed by PhoN-Se.26

2529_fig2.jpg

Figure 2Open in figure viewerPowerPointSubstrates not converted by PhoN-Sf, PhoN-Se, PiACP and phytase [300 mM substrate concentration except for rac-14 (200 mM) and 17 (15 mM)].

2529_fig3.jpg

Figure 3Open in figure viewerPowerPointEffect of P-donor type and pH on the phosphorylation of 2 (product formation over time). Conditions: 500 mM 2, 250 mM PPi or PPPi or 100 g L–1 polyP, 50 μg mL–1 PhoN-Sf, 1 mL volume, 1 % DMSO as internal standard, 30 °C, 600 rpm shaking.

11611_cssc201903139-fig-0001.jpg

Figure 1Open in figure viewerPowerPointa) Schematic diagram of the investigated 18650-type cell, with the positioning of the incoming neutron beam (blue) and the diffracted neutrons (green). The cell was positioned inside a He-filled cryostat to maintain a constant operating temperature. b–d)) Excerpt of the recorded diffractogram measured using the fresh cell in b) a charged state (blue); c) discharged state (back); and d) in the discharged state and without the shrinking tube (purple) and the empty cryostat (green–brown).

However, there is a dearth of knowledge on the charging and aging mechanisms of Si/C composite anodes containing micrometer-sized particles in full cells, even though these are key components that need to be understood to improve low-temperature aging, fast-charging capability, and battery cycle life. In particular, there is a lack of knowledge about Li deposition on Si/C composite anodes. Therefore, in this study we used operando neutron diffraction as a powerful method to obtain deeper insights into the processes that occur in Si/C anodes during the operation of Li-ion cells. In our experiments, incident neutrons were scattered from the electrodes inside a cylindrical Li-ion cell with a Si/C composite anode (Figure #cssc201903139-fig-0001#1 a; particle diameter≈10 μm) during charging and subsequent relaxation. This allowed simultaneous investigation of the LiC12 and LiC6 phase fractions, their changes with time, and electrochemical data. The influence of temperature, charging C-rate, and state-of-health (SOH) on the relaxation processes inside the active material of Si/C composite anodes was investigated in detail to unravel the effects of Si compounds on the charging and aging mechanisms. Post-mortem analysis of the cells was conducted as a complimentary method for analysis of the aging mechanisms. As shown in Figure #cssc201903139-fig-0001#1 a, the incident neutrons were scattered from the Si/C anode and the LixNi0.86Co0.10Al0.04O2 (NCA) cathode of the 18650-type cell. The neutron diffraction data in the discharged and charged state are shown in Figure #cssc201903139-fig-0001#1 b, c. For the discharged state, the angular position of the reflexes were 2θ of NCA (003)=25.923(7)°, 2θ of C (002)=37.025(6)° and for the charged state, 2θ of NCA (003)=26.210(4)°, 2θ of LiC6 (001)=33.553(3)°, 2θ of LiC12 (002)=35.3179°. During electrochemical operation, the diffraction angle of the NCA reflex changed owing to the oxygen repulsion, which is typical for layered oxides. However, in the break after charging, neither its intensity nor its position changed (see Supporting Information, Figures S2 and S3). All other reflexes were attributed to the active materials, casing, and current collectors, except one low intensity peak at 2θ=30.5°. This reflex was not observed in a measurement of the empty cryostat (Figure #cssc201903139-fig-0001#1 d) and did not change with charging. Therefore, it is likely to be caused by an inactive material.

11611_cssc201903139-fig-0002.jpg

Figure 2Open in figure viewerPowerPointEvolution of the integral intensities of the a) LiC12 and b) LiC6 reflections after charging at 0.1 C, 0.5 C, and 0.75 C and c) respective voltage (left axis) and differential voltage curves (right axis) at −21±2 °C. The transitions between the different relaxation phases (part I&II) are marked by the vertical dashed lines. Outliers are displayed in faded colors and were not taken into account for the fitting. The inset in (c) shows a magnification for better visibility of the curve for 0.5 C charging (see also Figure S1).

To obtain insights into the charging mechanism of a fresh cell, operando neutron scattering experiments were performed with a fresh cell at −21±2 °C for different C-rates. The evolution of the integrated intensities for the LiC12 and LiC6 reflexes during relaxation after the charging are shown in Figure #cssc201903139-fig-0002#2 a, b. The voltage relaxation and differential voltage curves that were recorded simultaneously are shown in Figure #cssc201903139-fig-0002#2 c. There were no significant changes in the integral intensities of the LiC6 and LiC12 phase in the neutron diffraction data after 0.1 C charging (Figure #cssc201903139-fig-0002#2 a,b). The lithiation of the anode from the electrode surface to the current collector was expected to be homogeneous for 0.1 C charging because polarization effects are minor (see Figure #cssc201903139-fig-0003#3 a, b for model). The voltage and the differential voltage relaxation curves (Figure #cssc201903139-fig-0002#2 c) did not contain a voltage plateau for 0.1 C charging. Therefore, no Li deposition was expected for this case. The integrated intensities, voltage curves, and differential voltages during the relaxation after charging at a rate of 0.75 C between −10 °C and −23 °C are shown in Figure #cssc201903139-fig-0007#7. The integrated intensities of LiC12 and LiC6 during relaxation for the fresh cell are shown in Figure #cssc201903139-fig-0007#7 a, b. Similar to that in Figure #cssc201903139-fig-0002#2, the relaxation was divided into two parts. Again, the transitions observed in the neutron diffraction data were coincided with the electrochemical data, that is, the transitions occurred at the inflection points of the voltage curves (see vertical dashed lines in Figure #cssc201903139-fig-0007#7 a–c). This was also consistent with the data in Figure #cssc201903139-fig-0006#6 recorded during charging.

11611_cssc201903139-fig-0003.jpg

Figure 3Open in figure viewerPowerPointa) Schematic diagram of the charging, relaxation, and Li redistribution process in Si/C composite anodes at −21±2 °C and different charging C-rates. a) Discharged cell; b) charging at 0.1 C; c–e) 0.5 C; f–h) 0.75 C. i) SEM of the anode surface after continuous 0.75 C cycling at −21 °C. The amounts of Li are intended to show the observed trend rather than quantitative values.

There were no significant changes in the integral intensities of the LiC6 and LiC12 phase in the neutron diffraction data after 0.1 C charging (Figure #cssc201903139-fig-0002#2 a,b). The lithiation of the anode from the electrode surface to the current collector was expected to be homogeneous for 0.1 C charging because polarization effects are minor (see Figure #cssc201903139-fig-0003#3 a, b for model). The voltage and the differential voltage relaxation curves (Figure #cssc201903139-fig-0002#2 c) did not contain a voltage plateau for 0.1 C charging. Therefore, no Li deposition was expected for this case. During charging of the Si/C composite, the graphite particles are most likely preferentially lithiated before the Si compound (see Figure #cssc201903139-fig-0003#3 a, c–e for model). Thermodynamically, Si should be lithiated before graphite. However, charging at a high rate of 0.5 C moves the system away from the equilibrium situation. In contrast, charging at a low rate of 0.1 C is closer to the equilibrium situation and therefore no relaxation effects were observable (see above). A similar time evolution of the LiC6 integrated intensity was observed during the rest period after charging at a rate of 0.75 C compared with that at 0.5 C. In contrast to charging at 0.5 C, we observed a decrease in the LiC12 integrated intensity in part I, which was in accordance with the data reported by Zinth et al. Again, for 0.75 C charging we observed a coincidence of the turning point of the differential voltage curve and a change from part I to part II of the relaxation in the neutron diffraction data. However, part I (re-intercalation of deposited Li) for 0.75 C was longer compared with that of 0.5 C charging; the reason was stronger polarization of the anode at higher C-rates, which was most likely caused by a higher amount of deposited Li (see Figure #cssc201903139-fig-0003#3 a, f–h for model). An SEM image of an anode after disassembly of a cell cycled 100 times with a charging C-rate of 0.75 C at −21 °C is shown in Figure #cssc201903139-fig-0003#3 i. The graphite particles were covered by a Li film, which was in agreement with the neutron diffraction and electrochemical data obtained under similar conditions (−21 °C, 0.75 C charging). The deposited Li was not dendritic, which has previously been observed for another cell type, most likely because of the pressure in the cylindrical cells and reaction with electrolyte. Simultaneous operando neutron diffraction and electrochemical assessment combined with post-mortem analysis were used to gain valuable new insights into the charging and aging mechanisms of Si/C composite anodes (3.5 wt % Si, ≈10 μm particle size of the Si compound) of Li-ion batteries. At −21±2 °C, the charging mechanism of a fresh cell depended on the charging C-rate (see Figure #cssc201903139-fig-0003#3 for model). In particular, slow charging (0.1 C) most likely led to simultaneous lithiation of graphite and the Si compound. No significant redistribution of Li was observed after charging. During charging at a moderate C-rate (0.5 C), small amounts of Li metal were deposited as a parallel reaction to lithiation of graphite and alloying of the Si compound. The relaxation after charging could be divided into two parts. The transition of both parts occurred at the inflection point of the differential voltage curve. In part I of the relaxation, deposited Li re-intercalated into graphite (and likely also into the Si compound). Most interestingly, in part II of the relaxation, Li was redistributed from lithiated graphite to the Si compound. For faster charging (0.75 C), the charging and relaxation mechanisms were similar to charging at 0.5 C; however, the Li deposition effect was stronger, that is, more Li metal was deposited on the anode surface.

11611_cssc201903139-fig-0004.jpg

Figure 4Open in figure viewerPowerPointCharge (black triangles upwards), discharge capacity (red triangles downwards) and coulombic efficiency (blue rhombohedra, right y-axis) vs. cycle number of the aged cell (25 °C, 1 C). 80 % SOH is marked with the grey dashed horizontal line. For better visibility, only every 10th data point is displayed.

The cycling stability of the investigated 18650-type cells during cycling at 25 °C and 1 C is shown in Figure #cssc201903139-fig-0004#4. The strong initial capacity loss was in accordance with the low coulombic efficiency within the first 100 cycles. Therefore, the capacity loss over the first 100 cycles (0.239 Ah) was nearly as high as the cumulative capacity loss during the next 600 cycles (0.242 Ah). The absence of a sudden capacity drop observed for other cell types indicated that the main aging mechanism does not change drastically during cycling. In contrast, the capacity fade decelerated, indicating that the amount of cyclable Li decreased in the aged cell whereas the usable active materials stayed relatively constant. The decelerated capacity fade was consistent with the increase in coulombic efficiency, indicating fewer side reactions with increasing cycle number. The post-mortem analysis results after disassembly of a fresh cell and a cell cycled at 25 °C are shown in Figure #cssc201903139-fig-0005#5. A different Si distribution was detected for the aged cell (Figure #cssc201903139-fig-0005#5 b) compared with the fresh cell (Figure #cssc201903139-fig-0005#5 a). A lower maximum and a broader distribution was observed for the Si distribution of the aged cell, which indicated that the formed film extended more into the depth of the anode compared with the anode from the fresh cell. These results for the cell cycled at 25 °C were consistent with depth profiling results for the same cell type cycled at 0 °C and 45 °C reported previously. The most likely formed Li silicates consume cyclable Li, which was consistent with the decelerated capacity fade curve in Figure #cssc201903139-fig-0004#4. The loss of cyclable Li has also been observed for other cells without Si compound in the anode. Therefore, a part of the capacity loss in the cells was likely to originate from the growth of a solid electrolyte interphase (SEI) layer on the graphite particles in addition to the aging effect of the Si compound; however, a large part of the film growth and the low coulombic efficiency can most likely be attributed to the formation of Li silicates.

11611_cssc201903139-fig-0005.jpg

Figure 5Open in figure viewerPowerPointPost-mortem analysis of the fresh and aged cell with Si/C composite anode cycled 700 times at 25 °C at a rate of 1 C. a, b) GD-OES depth profiles of a) fresh and b) aged Si/C electrode. c, d) SEM images of the anode cross-section of c) fresh and d) aged Si/C electrode.

The post-mortem analysis results after disassembly of a fresh cell and a cell cycled at 25 °C are shown in Figure #cssc201903139-fig-0005#5. A different Si distribution was detected for the aged cell (Figure #cssc201903139-fig-0005#5 b) compared with the fresh cell (Figure #cssc201903139-fig-0005#5 a). A lower maximum and a broader distribution was observed for the Si distribution of the aged cell, which indicated that the formed film extended more into the depth of the anode compared with the anode from the fresh cell. These results for the cell cycled at 25 °C were consistent with depth profiling results for the same cell type cycled at 0 °C and 45 °C reported previously. The most likely formed Li silicates consume cyclable Li, which was consistent with the decelerated capacity fade curve in Figure #cssc201903139-fig-0004#4. The loss of cyclable Li has also been observed for other cells without Si compound in the anode. Therefore, a part of the capacity loss in the cells was likely to originate from the growth of a solid electrolyte interphase (SEI) layer on the graphite particles in addition to the aging effect of the Si compound; however, a large part of the film growth and the low coulombic efficiency can most likely be attributed to the formation of Li silicates. The GD-OES depth profile of the aged cell (Figure #cssc201903139-fig-0005#5 b) shows high Li and O values at the anode surface, representing additional electrolyte consumption and SEI formation; furthermore, the aged cell shows no P maxima at the electrode surface, indicating the consumption of conductive salt and the dissolution of P-containing species form the SEI. Aging of the Si compound was also visible at the border of the particles (compare Figure #cssc201903139-fig-0005#5 c for fresh cell and Figure #cssc201903139-fig-0005#5 d for aged cell). A detailed comparison of the images of the particle cross-sections supported consecutive film formation predominantly on the Si particles. Additionally, the surface film on the particles seemed to be de-contacted, and the particle itself showed no sharply separated but more frayed interface to the formed film.

11611_cssc201903139-fig-0006.jpg

Figure 6Open in figure viewerPowerPointOperando neutron diffraction data recorded during charging. a) Integrated intensities of the LiC12 reflection during charging with 0.75 C at −10 °C (black squares), −18 °C (red circles) and −23 °C (purple triangles) at 100 % SOH (filled data points) and 80 % SOH (empty data points) and c) the LiC6 integrated intensities. b) Excerpt of diffraction data series recorded every 3 min during the charging at −23 °C and 100 % SOH. d) Excerpt of diffraction data series recorded every 3 min during the charging at −23 °C and 80 % SOH.

Insights into the charging mechanism as a function of temperature is a major key for understanding both the main aging mechanism and fast-charging capability of a cell. The neutron diffraction experiments during charging at a fixed rate of 0.75 C in the range of −23 °C to −10 °C is shown in Figure #cssc201903139-fig-0006#6. In these experiments, a fresh (discharge capacity at 25 °C: 2.56 Ah, 100 % SOH) was compared with an aged cell (aging conditions: 700 cycles at 1 C/25 °C, discharge capacity at 25 °C: 2.04 Ah, 79.6 % SOH). For charging of the fresh cell at −10 °C (Figure #cssc201903139-fig-0006#6 a and c, black squares), a mostly homogenous lithiation of the anode was expected as the LiC12 integrated intensity linearly decreased and the LiC6 integrated intensities linearly increased, corresponding to Equation (1 b). At −10 °C, the slope of the LiC6 phase seemed to flatten for the last data point, which weakly indicated the starting point for Li metal deposition during galvanostatic charging. The integrated intensities, voltage curves, and differential voltages during the relaxation after charging at a rate of 0.75 C between −10 °C and −23 °C are shown in Figure #cssc201903139-fig-0007#7. The integrated intensities of LiC12 and LiC6 during relaxation for the fresh cell are shown in Figure #cssc201903139-fig-0007#7 a, b. Similar to that in Figure #cssc201903139-fig-0002#2, the relaxation was divided into two parts. Again, the transitions observed in the neutron diffraction data were coincided with the electrochemical data, that is, the transitions occurred at the inflection points of the voltage curves (see vertical dashed lines in Figure #cssc201903139-fig-0007#7 a–c). This was also consistent with the data in Figure #cssc201903139-fig-0006#6 recorded during charging.

11611_cssc201903139-fig-0007.jpg

Figure 7Open in figure viewerPowerPointa, d) Integrated intensities during the relaxation time of LiC12 after charging with 0.75 C at −10 °C (black squares), −18 °C (red circles), and −23 °C (orange triangles) at a) 100 % SOH and d) at 80 % SOH. b, e) Integrated intensities during relaxation time of LiC6 after charging with 0.75 C at −10 °C (black squares), −18 °C (red circles), and −23 °C (orange triangles) at b) 100 % SOH and e) at 80 % SOH. c, f) Relaxation voltage profiles (left axes, solid lines) and differential voltage (right axes, dotted lines) at −10 °C (black), −18 °C (red), and −23 °C (orange) at c) 100 % SOH and f) 80 % SOH. Outliers are displayed in faded colors.

The integrated intensities, voltage curves, and differential voltages during the relaxation after charging at a rate of 0.75 C between −10 °C and −23 °C are shown in Figure #cssc201903139-fig-0007#7. The integrated intensities of LiC12 and LiC6 during relaxation for the fresh cell are shown in Figure #cssc201903139-fig-0007#7 a, b. Similar to that in Figure #cssc201903139-fig-0002#2, the relaxation was divided into two parts. Again, the transitions observed in the neutron diffraction data were coincided with the electrochemical data, that is, the transitions occurred at the inflection points of the voltage curves (see vertical dashed lines in Figure #cssc201903139-fig-0007#7 a–c). This was also consistent with the data in Figure #cssc201903139-fig-0006#6 recorded during charging. The integrated intensities of LiC12 and LiC6 during relaxation for the aged cell are shown in Figure #cssc201903139-fig-0007#7 d,e. In contrast to the fresh cell, there was no relaxation in two parts for the aged cell. Instead, the LiC6 integrated intensities only decreased during relaxation, indicating only a transfer of Li from lithiated graphite to the Si compound. However, no re-intercalation of Li into graphite was observed for the aged cell. This was in very good agreement with the voltage and differential voltage curves in Figure #cssc201903139-fig-0007#7 f, which do not show plateaus or minima, respectively.

11611_cssc201903139-fig-0008.jpg

Figure 8Open in figure viewerPowerPointOverview of the processes that occur during low-temperature charging of fresh and aged cells with Si/C anodes. Discharged Si/C anodes of a) fresh and h) aged cells. Charging and relaxation of the fresh cell at b–d) −10 °C/0.75 C and e–g) −23 °C/0.75 C. i) Charging and j) relaxation of the aged cell with a silicate film10 in which Li deposition was not observed. The amounts of Li are intended to show the observed trend rather than quantitative values.

Regarding the influence of temperature and charging at 0.75 C, we found the following trends: Li deposition on the anode occurred in the range of −10 °C to −23 °C for the fresh cells (see Figure #cssc201903139-fig-0008#8 a–g for model). Lower temperatures led to longer relaxation times of part I and most likely to higher amounts of deposited Li. In contrast to the fresh cells, cells cycled at 25 °C/1 C cells did not show any evidence of Li deposition under the same conditions (see Figure #cssc201903139-fig-0008#8 h–j for model). The reason for the absence of Li deposition in the aged cell was the lower amount of cyclable Li, which prevented higher lithiation states of the graphite. Fully consistent with this, in the case of charging at −23 °C and 0.75 C, neutron diffraction did not show any LiC6 phase contribution. The loss of cyclable Li as the main aging mechanism at 25 °C in the investigated cell type was consistent with the capacity fade curve and with post-mortem analysis.

11423_cmdc202100569-fig-0001.jpg

Figure 1Open in figure viewerPowerPointWorkflow of the computational screening and experimental testing steps, with the respective compound counts. Three virtual screening concepts (docking, pharmacophore and shape screening) were used in parallel. These were validated retrospectively and applied prospectively to our in-house compound database. Compounds that are predicted as actives by at least two out of three screening concepts were considered as virtual hits and selected for primary in vitro testing, followed by an SAR analysis of the [1,2,4]triazolo[1,5-b]isoquinoline scaffold.

To discover new MELK inhibitors, we have assembled a consensus-based virtual screening workflow, employing three orthogonal computational methods: ligand docking, pharmacophore screening and shape screening. Prior to applying the workflow for prospective purposes, each of the screening methods were thoroughly validated retrospectively, on a dataset of 50 known, potent MELK inhibitors (<100 nM IC50) extracted from the ChEMBL database, and 2544 decoy molecules. For ligand docking, seven PDB structures were selected based on the diversity of their binding site conformations, and the geometric mean of the Glide SP docking scores was used as the data fusion rule for ensemble docking. For pharmacophore screening, a consensus model was developed based on 40 known actives and 10 known inactives. For shape screening, the most active inhibitor with an experimentally determined binding mode was used as the query. The workflow is summarized in Figure #cmdc202100569-fig-0001#1, and retrospective performances of the models are reported in the Experimental section.

11423_cmdc202100569-fig-0002.jpg

Figure 2Open in figure viewerPowerPointStructure-guided selection of further analogues for testing. A) Predicted binding mode of compound 18 (green sticks) overlaid on the binding mode of dorsomorphin (white sticks, PDB: 6GVX31). In addition to the good overlay of the respective cores, the 4-OMe group of 18 mimics the larger, solvent-exposed substituent of dorsomorphin, while the 10-ethyl group extends inside the binding pocket. B) Predicted binding mode of compound 11 (green sticks) overlaid on the binding mode of the MELK inhibitor NVS-MELK8F (white sticks, PDB: 5IHA32). Here, the heterocyclic nitrogen of the pyridyl unit can act as the anchoring group against the backbone NH group of the hinge residue C89, while the advantageous position of the 10-methyl group is supported by the hydrophobic group of NVS-MELK8F in the same position.

To further evaluate the possible role of the ethyl group in position 10 and to explore other substitutions of the 2-phenyl unit, we have carried out a substructure search of our compound database, which revealed 21 additional compounds with the same core scaffold. From these, we have selected eight further compounds for in vitro testing, with a structure-guided approach, based on their predicted binding poses in the active site of MELK, overlaid on 14 PDB structures containing sub-micromolar MELK inhibitors with experimentally determined binding modes (Figure #cmdc202100569-fig-0002#2). Interestingly, many of the co-crystallized MELK inhibitors, as well as our inhibitor series, exhibit binding modes with only one hydrogen bond acceptor acting as an anchoring group toward the hinge region, instead of the 3-point donor-acceptor-donor motif that is typical for other kinases. From the first series, only compound 12 exhibited a significant inhibitory activity, suggesting that a heteroatom can only contribute to the binding affinity from position 2 of the 2-phenyl unit (Figure #cmdc202100569-fig-0002#2B). This contrasts with the increased affinity of methoxy analogues in positions 3 (5) or 4 (4), suggesting that the methoxy group does not act as a hinge-binding moiety (Figure #cmdc202100569-fig-0002#2A). As for the 10-ethyl series, the chloro scan reveals a preference for the 4 position of the 2-phenyl unit as the ideal point of substitution, with compound 17 being the only chloro-substituted analogue with a significant (single-digit micromolar) inhibitory activity. Additionally, the 10-ethyl-2-(4-methoxyphenyl) analogue (18) is highlighted as the lead compound of this series with a sub-micromolar activity, presenting 2.5- and 4.5-fold activity boosts over its 3-methoxyphenyl (19) and 10-methyl (4) counterparts, respectively. These observations, along with the binding mode comparisons in Figure #cmdc202100569-fig-0002#2 provide ample support for positions 10 (main core) and 4 (2-phenyl unit) as the main drivers of the SAR and potential growing vectors of this MELK inhibitor series.

11423_cmdc202100569-fig-0003.jpg

Figure 3Open in figure viewerPowerPointROC curves with AUC values (and 95 % confidence intervals47) of the retrospective docking. Ensemble docking based on the minimum, maximum and geometric mean of the seven docking scores is better than using any single structure. Based on our earlier results, we have selected the geometric mean from the available data fusion rules.26

Ligand docking. Twenty-three X-ray crystal structures of MELK with resolutions below 3 Å were downloaded from the PDB database. Hierarchical clustering of the binding site residues with max. 5 Å distance from the ligands was performed with Scipy (version 1.5.2) based on the Euclidean distance with complete linkage. Seven representative structures (4BKZ, 4D2T, 4D2 W, 4UMP, 5MAH, 5TWZ, 5TX3) were selected for further application in docking as the centroids of each cluster defined at a distance threshold of 0.5 Å (the dendrogram is reported in Figure S2). Schrödinger Glide (with the SP scoring function) was applied for ensemble docking, and geometric mean, minimum and maximum values of the seven docking scores were calculated for the final ranking of the ligands. AUC values of the ROC curves were calculated for the retrospective study. Figure #cmdc202100569-fig-0003#3 shows the result of the retrospective docking in the case of four different protein structures as examples and the three data fusion methods. From these, the geometric mean was chosen based on its AUC value of 0.93 and our earlier results on data fusion rules, Glide docking scores were multiplied by −1 to allow for the calculation of geometric means. Compounds were classified as active above a fused Glide docking score of 6.09 (based on the minimum Euclidean distance from the [0;1] point, which represents ideal classification in the ROC plot with an AUC value of 1).

14151_celc201700219-fig-0001.jpg

Figure 1Open in figure viewerPowerPointSchematic diagram of PL and ECL processes. i) and ii) represent reductive and oxidative charging, respectively, which occur before electron-hole recombination. Solid dot=electron and hollow dot=hole. Modified with permission form Ref [21]. Copyright (2002) AAAS.

ECL of semiconductor nanoparticles (NPs) or quantum dots (QDs) was first reported by Bard and coworkers in 2002. They synthesized Si NPs with core diameters of ∼2 to 4 nm. ECL of Si NPs was generated by ion annihilation and use of co-reactants, oxalate and peroxydisulfate. Interestingly, all ECL spectra of Si NPs showed a maximum ECL intensity at 640 nm, which is considerably red-shifted from the wavelength (420 nm) of their maximum photoluminescence (PL) intensity by 220 nm. They also observed that ECL of Si NPs was not sensitive to NP size and the capping agent used, even though PL of Si NPs was size-dependent. The large energy difference (0.8 eV) between PL and ECL indicates that the emission process for ECL would be different from that for PL. Electrochemical analysis of the Si NPs led the Bard group to surmise that the potential difference between hole and electron injections could provide sufficient energy for optically radiative electron-hole recombination (ECL from core states). However, such ECL was not observed because the carrier energies are still not enough. According to their optical spectroscopic study, Si NPs requires much higher energy than the band gap energy for efficient PL (Figure #celc201700219-fig-0001#1). The suggested ECL process is that charge would be injected to the surface states of a Si NP and then radiative electron-hole recombination for ECL occurs (Figure #celc201700219-fig-0001#1). Since charge injection in a Si nanoparticle is considered to occur through its surface states, ECL is more sensitive to the presence of surface states rather than core size.

14151_celc201700219-fig-0002.jpg

Figure 2Open in figure viewerPowerPointECL spectrum of Ge NPs in dimethylformamide containing 0.1 M tetrabutylammonium perchlorate by stepping electrode potential between +1.5 and −2.5 V at 10 Hz rate for 30 min. Inset: PL spectrum obtained from the Ge NPs dispersed in CHCl3. Excitation wavelength: 380 nm. The large spikes represent the effect of cosmic ray events on the CCD. Reproduced with permission from Ref. [22]. Copyright (2004) American Chemical Society.

The similar ECL-emitting process was observed from Ge NPs (∼4 nm in diameter). The maximum intensity of ECL produced by the annihilation reaction between oppositely charged Ge NPs was observed at about 200 nm longer wavelength than PL (Figure #celc201700219-fig-0002#2). No ECL emission from the core states was produced. When the electrode potential was cycled between +1.5 V and −2.3 V, the higher ECL intensity was observed in the positive potential region than in the negative potential. This implies that the reduced (negatively charged) Ge NPs formed in the negative potential region are stable enough to maintain their charge states until the electrode potential is sufficiently positive to generate the oxidized (positively charged) ones. The early ECL studies of CdS and PbS NPs were not successful, because these NPs are chemically unstable upon electron transfer. For example, Bard and co-workers proposed the electrochemical process of thioglycerol-capped CdS NPs as an (EC)n reaction, in which multiple electrons are transferred by fast coupled chemical reactions due to the decomposition of NP cores. However, Zhu and co-workers could observe ECL from the CdS spherical assemblies dispersed in CH3CN containing 0.1 M tetra-n-butylammonium perchlorate (TBAP). As shown in Figure #celc201700219-fig-0003#3, the spherical assemblies are secondary particles aggregated by small CdS NPs (∼5 nm in diameter). Because their CdS NPs have no capping agents, aggregation could occur as prepared due to the small dimension and high surface energy of NPs. Electrochemistry of these assemblies showed similar processes, e. g., (EC)n reaction, which the Bard group observed. Although some of CdS NPs in the assemblies could be decomposed upon cathodic and anodic electron transfer (C1 and A1 in Figure #celc201700219-fig-0004#4 (a)), overall assemblies could be stable enough to observe ECL. When the electrode potential was cycled between −2.5 V and +2.5 V, four ECL peaks (I–IV) were observed as shown in Figure #celc201700219-fig-0004#4 (a). ECL peaks I and II were resulted from annihilation reaction between oppositely charged CdS NPs, while ECL peaks III and IV were from reactions between charged Cd NPs and the product of decomposed NPs (e. g., Cd+). However, ECL peaks III and IV were not observed in the first cycle of potential scan (Figure #celc201700219-fig-0002#2 (b)), indicating the product of decomposed NPs was involved in the ECL process. All ECL generated at different potentials (peaks I–IV) was emission from CdS NPs, showing a maximum intensity at 700 nm.

14151_celc201700219-fig-0003.jpg

Figure 3Open in figure viewerPowerPointTEM image of the as-prepared CdS spherical assemblies. Reproduced with permission from Ref. [24]. Copyright (2005) Elsevier.

The early ECL studies of CdS and PbS NPs were not successful, because these NPs are chemically unstable upon electron transfer. For example, Bard and co-workers proposed the electrochemical process of thioglycerol-capped CdS NPs as an (EC)n reaction, in which multiple electrons are transferred by fast coupled chemical reactions due to the decomposition of NP cores. However, Zhu and co-workers could observe ECL from the CdS spherical assemblies dispersed in CH3CN containing 0.1 M tetra-n-butylammonium perchlorate (TBAP). As shown in Figure #celc201700219-fig-0003#3, the spherical assemblies are secondary particles aggregated by small CdS NPs (∼5 nm in diameter). Because their CdS NPs have no capping agents, aggregation could occur as prepared due to the small dimension and high surface energy of NPs. Electrochemistry of these assemblies showed similar processes, e. g., (EC)n reaction, which the Bard group observed. Although some of CdS NPs in the assemblies could be decomposed upon cathodic and anodic electron transfer (C1 and A1 in Figure #celc201700219-fig-0004#4 (a)), overall assemblies could be stable enough to observe ECL. When the electrode potential was cycled between −2.5 V and +2.5 V, four ECL peaks (I–IV) were observed as shown in Figure #celc201700219-fig-0004#4 (a). ECL peaks I and II were resulted from annihilation reaction between oppositely charged CdS NPs, while ECL peaks III and IV were from reactions between charged Cd NPs and the product of decomposed NPs (e. g., Cd+). However, ECL peaks III and IV were not observed in the first cycle of potential scan (Figure #celc201700219-fig-0002#2 (b)), indicating the product of decomposed NPs was involved in the ECL process. All ECL generated at different potentials (peaks I–IV) was emission from CdS NPs, showing a maximum intensity at 700 nm.

14151_celc201700219-fig-0004.jpg

Figure 4Open in figure viewerPowerPointa) Cyclic voltammogram and ECL response of CdS spherical assemblies obtained at a scan rate of 0.1 V/s. b) ECL response measured at the first scan of cyclic voltammetry. Reproduced with permission from Ref. [24]. Copyright (2005) Elsevier.

The early ECL studies of CdS and PbS NPs were not successful, because these NPs are chemically unstable upon electron transfer. For example, Bard and co-workers proposed the electrochemical process of thioglycerol-capped CdS NPs as an (EC)n reaction, in which multiple electrons are transferred by fast coupled chemical reactions due to the decomposition of NP cores. However, Zhu and co-workers could observe ECL from the CdS spherical assemblies dispersed in CH3CN containing 0.1 M tetra-n-butylammonium perchlorate (TBAP). As shown in Figure #celc201700219-fig-0003#3, the spherical assemblies are secondary particles aggregated by small CdS NPs (∼5 nm in diameter). Because their CdS NPs have no capping agents, aggregation could occur as prepared due to the small dimension and high surface energy of NPs. Electrochemistry of these assemblies showed similar processes, e. g., (EC)n reaction, which the Bard group observed. Although some of CdS NPs in the assemblies could be decomposed upon cathodic and anodic electron transfer (C1 and A1 in Figure #celc201700219-fig-0004#4 (a)), overall assemblies could be stable enough to observe ECL. When the electrode potential was cycled between −2.5 V and +2.5 V, four ECL peaks (I–IV) were observed as shown in Figure #celc201700219-fig-0004#4 (a). ECL peaks I and II were resulted from annihilation reaction between oppositely charged CdS NPs, while ECL peaks III and IV were from reactions between charged Cd NPs and the product of decomposed NPs (e. g., Cd+). However, ECL peaks III and IV were not observed in the first cycle of potential scan (Figure #celc201700219-fig-0002#2 (b)), indicating the product of decomposed NPs was involved in the ECL process. All ECL generated at different potentials (peaks I–IV) was emission from CdS NPs, showing a maximum intensity at 700 nm.

14151_celc201700219-fig-0005.jpg

Figure 5Open in figure viewerPowerPointPL (black dashed line) and ECL (red solid line) spectra of a) PbS-BDY and b) PbS-OA. The PL spectra were obtained with excitation at 532 nm, whereas the ECL spectra were measured during two potential scans between −0.4 and 1.4 V vs. SCE. The inset shows the zoom-in part of the PL spectrum. Reproduced with permission from Ref. [30]. Copyright (2015) American Chemical Society.

Attachment of ECL-active dye molecules on the surface of NPs could produce dual ECL emissions. Ding et al. prepared a hybrid system (PbS-BDY) consisting of PbS NPs and BODIPY dye (BDY) ligands and observed strong dual ECL emissions at 680 and 984 nm (Figure #celc201700219-fig-0005#5a) resulted from BDY and PbS moieties, respectively, when ECL was produced with tripropylamine (TPrA) as a co-reactant. Since both PbS and BDY moieties are oxidized at ∼0.744 V, the dual ECL emissions are simultaneously produced. Relative to the emission from the BDY moiety, the ECL emission from the PbS moiety is much stronger than its PL (Figure #celc201700219-fig-0005#5a). This implies that the excited state of the PbS moiety is more efficiently produced by electrochemical and subsequent chemical reactions. As shown in 5b, such enhancement was also observed in the PbS-OA/TPrA system, which shows weak ECL by the annihilation route. The observed ECL efficiencies for PbS-OA and the PbS moiety of PbS-DBY were reported ∼96 % relative to that of Ru(bpy)32+/TPrA.

14151_celc201700219-fig-0006.jpg

Figure 6Open in figure viewerPowerPointPlots of ECL vs. potential measured from a) a CdSe-NP film, b) a CdSe/MWCNTs composite, and c) a CdSe/NCNTs composite cased on GCEs. Inset: corresponding cyclic voltammograms measured at 100 mV/s (vs. Ag/AgCl). Reproduced with permission from Ref. [32e]. Copyright (2009) Wiley-VCH.

ECL of CdSe NPs could be enhanced by preparing a composite with secondary materials. Ju et al. prepared CdSe-NP composites with multiwalled carbon nanotubes (MWCNTs) and nitrogen-doped carbon nanotubes (NCNTs). The films of CdSe NPs and these composites (CdSe/MWCNTs and CdSe/NCNTs) were casted on glassy carbon electrodes (GCEs). ECL was measured in pH 9.0 PBS containing 10 mM H2O2 as a co-reactant. As shown in Figure #celc201700219-fig-0006#6, more intense ECL was clearly observed from the composite-modified electrodes; ECL measured from CdSe/NCNTs was 3-fold higher than that from CdSe/MWCNTs and 5-fold higher than that from the CdSe NP film. This enhancement was resulted from the electrocatalytic reduction of CdSe NPs by carbon nanotubes. Obviously, NCNTs displayed better electrocatalytic activity than MWCNTs (inset of Figure #celc201700219-fig-0006#6). Owing to the overpotential for electrocatalytic reduction of CdSe NPs, ECL from CdSe/NCNTs began at more positive potential: 0.22-V more positive from CdSe/MWCNTs and 0.33-V more positive from the CdSe film. Wang et al. also reported similar enhancement from a composite made with CdSe NPs, graphene oxide, and chitosan.

14151_celc201700219-fig-0007.jpg

Figure 7Open in figure viewerPowerPointECL intensity of bare GCE (black line), bulk ZeSe (red line), ZB ZnSe (green line), and WZ ZnSe (blue line) modified electrodes after surface modification in a 0.1 M KOH solution containing 0.1 M K2S2O8. Scan rate: 100 mV/s. Inset: photograph showing the good solubility of WZ ZnSe NPs. Reproduced with permission from Ref. [35]. Copyright (2016) American Chemical Society.

ECL of ZnSe NPs is highly dependent on the crystal structure of NPs. The stable crystal structure of ZnSe is zinc-blende (ZW), where cubic-close-packed lattices grow in the (111) direction, but it is also possible to possess the wurtzite (WZ) structure, where hexagonal-close-packed lattices grow in the (0001) direction. The WZ structure is metastable at room temperature and thus it is difficult to obtain. Recently, Dai et al. developed a facile method to synthesize WZ ZnSe NPs (∼15 nm in diameter) showing good solubility in aqueous solution (inset of Figure #celc201700219-fig-0007#7). ECL of WZ ZnSe produced in a 0.1 M KOH solution containing 0.1 M K2S2O8 displayed λECL at 793 nm significantly red-shifted by 368 nm from its λPL (425 nm), while λECL of ZB ZnSe was red-shifted by 344 nm. This difference came from the degrees of symmetry-breaking in the NP crystal lattices. As shown in Figure #celc201700219-fig-0007#7, it is notable that WZ ZnSe produced most intense ECL with onset voltage at −1.35 V vs. Ag/AgCl.

14151_celc201700219-fig-0008.jpg

Figure 8Open in figure viewerPowerPointTEM and corresponding HRTEM images of ZnO nanostructures with different morphologies: a, b) nanopyramid; c, d) nanoflakes; and e, f) nanocolumn. Reproduced with permission from Ref. [42f]. Copyright (2012) Science China Press and Springer.

One of the interesting ECL properties of ZnO NPs is that they display shape-dependent ECL emissions. Liu et al. synthesized ZnO nanostructures with three different shapes: nanopyramids, nanoflakes, and nanocolumns, of which the structures are all hexagonal-based (Figure #celc201700219-fig-0008#8). The base size of nanopyramids is ∼300 nm–1.5 μm; the size of nanoflakes is ∼200–500 nm in width and ∼ 20–30 nm in thickness; and the size of nanocolumns are ∼2–10 μm in length and ∼40–50 nm in diameter. For ECL measurements, indium tin oxide (ITO) electrodes were modified with these ZnO nanostructures. ECL was produced with a co-reactant, K2S2O8 in 0.1 M KOH solution containing 0.1 M KCl by scanning the electrode potential between 0 and −2.0 V vs. Ag/AgCl. As shown in Figure #celc201700219-fig-0009#9, they obtained the broad visible ECL spectra with λECL at 610, 596, and 584 nm for nanopyramids, nanoflakes, and nanocolumns, respectively. These are all red-shifted from their PL, implying that ECL emissions were from the surface state. Since the size of ZnO nanostructures is sufficiently large enough to exclude the size effect, such difference in ECL is most likely results from the different surface structures of ZnO. The mainly exposed facets of nanopyramids are the O-terminated polar planes for the side surfaces and the O-terminated polar plane for the base; those of nanoflakes are the Zn-terminated polar (0001) plane for the top surface and the O-terminated polar plane for the bottom surface; and those of nanocolumns are the nonpolar planes, composed of the equivalent amount of Zn2+ and O2−, for all sides.

14151_celc201700219-fig-0009.jpg

Figure 9Open in figure viewerPowerPointECL spectra of ZnO nanostructures with various shapes: a) nanopyramids, b) nanoflakes, and c) nanocolumns. Reproduced with permission from Ref. [42f]. Copyright (2012) Science China Press and Springer.[a] Versus Ag/AgCl. [b] Versus SCE, where SCE is a saturated calomel electrode.

One of the interesting ECL properties of ZnO NPs is that they display shape-dependent ECL emissions. Liu et al. synthesized ZnO nanostructures with three different shapes: nanopyramids, nanoflakes, and nanocolumns, of which the structures are all hexagonal-based (Figure #celc201700219-fig-0008#8). The base size of nanopyramids is ∼300 nm–1.5 μm; the size of nanoflakes is ∼200–500 nm in width and ∼ 20–30 nm in thickness; and the size of nanocolumns are ∼2–10 μm in length and ∼40–50 nm in diameter. For ECL measurements, indium tin oxide (ITO) electrodes were modified with these ZnO nanostructures. ECL was produced with a co-reactant, K2S2O8 in 0.1 M KOH solution containing 0.1 M KCl by scanning the electrode potential between 0 and −2.0 V vs. Ag/AgCl. As shown in Figure #celc201700219-fig-0009#9, they obtained the broad visible ECL spectra with λECL at 610, 596, and 584 nm for nanopyramids, nanoflakes, and nanocolumns, respectively. These are all red-shifted from their PL, implying that ECL emissions were from the surface state. Since the size of ZnO nanostructures is sufficiently large enough to exclude the size effect, such difference in ECL is most likely results from the different surface structures of ZnO. The mainly exposed facets of nanopyramids are the O-terminated polar planes for the side surfaces and the O-terminated polar plane for the base; those of nanoflakes are the Zn-terminated polar (0001) plane for the top surface and the O-terminated polar plane for the bottom surface; and those of nanocolumns are the nonpolar planes, composed of the equivalent amount of Zn2+ and O2−, for all sides.

14151_celc201700219-fig-5001.jpg

Scheme 1Open in figure viewerPowerPointSchematic representation of key assay formats for the measurement of target analyte: a) competitive format and b) non-competitive (sandwich) format.

14151_celc201700219-fig-5002.jpg

Scheme 2Open in figure viewerPowerPointSchematic illustration of the procedure for preparation of NP-based ECL immunoassay. AFP: α-fetoprotein; MNPs: Fe3O4 magnetic nanoparticles; QDs: CdTe nanoparticles; and BSA: bovine serum albumin to block nonspecific binding sites of MNPs and QDs. Reproduced with permission from Ref. [50]. Copyright (2017) Elsevier.

14151_celc201700219-fig-5003.jpg

Scheme 3Open in figure viewerPowerPointSchematic illustration of ‘ECL off-on’ biosensing system via target-induced structure switching. Reproduced with permission form Ref. [53]. Copyright (2014) American Chemical Society.

14151_celc201700219-fig-5004.jpg

Scheme 4Open in figure viewerPowerPointA) Schematic illustration of ECL aptasensor for detection of thrombin and B) fabrication procedures of rGO−H-AuNRs-G4H. Reproduced with permission from Ref. [54]. Copyright (2016) American Chemical Society.

14151_celc201700219-fig-5005.jpg

Scheme 5Open in figure viewerPowerPointSchematic illustration of multiplex ECL DNA sensor for determination of target DNAHBV and target DNAHCV based on Au NPs-probe DNA/target DNA/CdTe QDs-capture DNA/GNs/GCE composite film. Reproduced with permission from Ref. [55]. Copyright (2016) Elsevier.

14151_celc201700219-fig-5006.jpg

Scheme 6Open in figure viewerPowerPointa) The conventional ‘double antibody sandwich’ electrochemiluminescence immunoassay; b) the ‘single antibody sandwich’ electrochemiluminescence immunoassay developed with molecularly imprinted technology. Reproduced with permission from Ref. [56]. Copyright (2013) Elsevier.

16606_cmdc202200027-fig-0001.jpg

Figure 1Open in figure viewerPowerPointTKIs with P-gp and BCRP inhibitory activity, gefitinib and erlotinib, and third generation P-gp inhibitors, tariquidar and elacridar.

Recently, quinazoline-4-amine based tyrosine kinase inhibitors (TKIs), such as gefitinib and erlotinib, have been identified as ABC transporters modulators (Figure #cmdc202200027-fig-0001#1), and many studies reported on compounds with the 4-anilino-quinazoline scaffold as potent BCRP inhibitors. In addition, the quinazoline moiety was introduced into compounds that proved to be P-gp inhibitors. In the present study, we designed and synthesized a new series of 2,4-disubstituted quinazoline derivatives with the aim of discovering new P-gp and/or BCRP inhibitors (Figure #cmdc202200027-fig-0002#2). For this purpose, secondary or tertiary protonable amines were inserted in position 4 of the quinazoline scaffold in place of the aniline residues typical of TKIs. The selected amines were 4-(2-(6,7-dimethoxy-3,4-dihydroisoquinolin-2(1H)-yl)ethyl)aniline (I), 2-phenylethan-1-amine (II), morpholine (III), 1-methylpiperazine (IV) and 6,7-dimethoxy-1,2,3,4-tetrahydroisoquinoline (V) (Figure #cmdc202200027-fig-0002#2). The amine I was chosen since it is present in two of the most interesting third generation P-gp inhibitors tariquidar and elacridar that are also able to bind the BCRP transporter (Figure #cmdc202200027-fig-0001#1). The other amines (II–V) were chosen to vary both the steric hindrance and the electronic proprieties. Position 2 of the quinazoline nucleus was substituted with aryl residues able to improve the P-gp interaction. Therefore, aromatic groups, such as anthracene or methoxy-substituted aryl moieties (Figure #cmdc202200027-fig-0002#2), were chosen because of their presence in our previously synthesized compounds, with different scaffolds, that have proved to be potent and efficacious P-gp dependent MDR reversers. In particular, the hydrogen bond acceptor methoxy group is considered important for the MDR-reversing activity and is present in many well-known P-gp modulators.

16606_cmdc202200027-fig-0002.jpg

Figure 2Open in figure viewerPowerPointGeneral structure of derivatives 1–7.

In the present study, we designed and synthesized a new series of 2,4-disubstituted quinazoline derivatives with the aim of discovering new P-gp and/or BCRP inhibitors (Figure #cmdc202200027-fig-0002#2). For this purpose, secondary or tertiary protonable amines were inserted in position 4 of the quinazoline scaffold in place of the aniline residues typical of TKIs. The selected amines were 4-(2-(6,7-dimethoxy-3,4-dihydroisoquinolin-2(1H)-yl)ethyl)aniline (I), 2-phenylethan-1-amine (II), morpholine (III), 1-methylpiperazine (IV) and 6,7-dimethoxy-1,2,3,4-tetrahydroisoquinoline (V) (Figure #cmdc202200027-fig-0002#2). The amine I was chosen since it is present in two of the most interesting third generation P-gp inhibitors tariquidar and elacridar that are also able to bind the BCRP transporter (Figure #cmdc202200027-fig-0001#1). The other amines (II–V) were chosen to vary both the steric hindrance and the electronic proprieties. Position 2 of the quinazoline nucleus was substituted with aryl residues able to improve the P-gp interaction. Therefore, aromatic groups, such as anthracene or methoxy-substituted aryl moieties (Figure #cmdc202200027-fig-0002#2), were chosen because of their presence in our previously synthesized compounds, with different scaffolds, that have proved to be potent and efficacious P-gp dependent MDR reversers. In particular, the hydrogen bond acceptor methoxy group is considered important for the MDR-reversing activity and is present in many well-known P-gp modulators.

16606_cmdc202200027-fig-5001.jpg

Scheme 1Open in figure viewerPowerPointReagents and conditions: I) ArCOCl, dry pyridine, rt, 4 h; II) NH4OH (33.0 %), EtOH, 80 °C, 20 h; III) ArCHO, CuCl2, EtOH, reflux, 16 h, or ArCOOH, HATU, DIPEA, dry CH2Cl2, 50 °C, 16 h, then NaOH (10.0 M), EtOH, rt, 2 h; IV) SOCl2, dry DMF, CHCl3 (free of ethanol), 50 °C, 6 h or POCl3, reflux, 5–12 h; V) amines, CH3SO3H, abs. EtOH, reflux, 4 h (Method A), or amines, K2CO3, dry DMF, 60 °C, 5 h (Method B). For the structure of final compounds 1–7 see Table 1.

As can be seen in Figure #cmdc202200027-fig-0003#3 i, there is not a common recognizable binding pattern for the studied compounds, which span nearly throughout the whole transmembrane region. Nevertheless, some structural features are worth noting for selected compounds. In particular, the most potent compounds, 1 d, 1 e, 2 a, 2 c and 2 e, bearing aryl groups A or B, exhibited some common peculiarities with binding poses in a quite delimited region, thus suggesting a possible key for their potency (Figure #cmdc202200027-fig-0003#3 ii). The interactions with receptor residues are mostly hydrophobic with few or, in some cases, no polar contacts. The transmembrane domains (TM) mainly involved in binding are TM6, TM7 and TM12 (Figures #cmdc202200027-fig-0004#4 i-v).

16606_cmdc202200027-fig-0003.jpg

Figure 3Open in figure viewerPowerPointCollective picture of binding poses of the entire set of studied compounds (i) and of the most active ones (1 d, 1 e, 2 a, 2 c and 2 e) (ii) within the P-gp binding region.

As can be seen in Figure #cmdc202200027-fig-0003#3 i, there is not a common recognizable binding pattern for the studied compounds, which span nearly throughout the whole transmembrane region. Nevertheless, some structural features are worth noting for selected compounds. In particular, the most potent compounds, 1 d, 1 e, 2 a, 2 c and 2 e, bearing aryl groups A or B, exhibited some common peculiarities with binding poses in a quite delimited region, thus suggesting a possible key for their potency (Figure #cmdc202200027-fig-0003#3 ii). The interactions with receptor residues are mostly hydrophobic with few or, in some cases, no polar contacts. The transmembrane domains (TM) mainly involved in binding are TM6, TM7 and TM12 (Figures #cmdc202200027-fig-0004#4 i-v).

16606_cmdc202200027-fig-0004.jpg

Figure 4Open in figure viewerPowerPointBinding pose of the most active compounds 1 d (i), 1 e (ii), 2 a (iii), 2 c (iv) and 2 e (v) within the P-gp binding region.

A further consideration deserves compound 7 c, which is the only ligand of the set showing a behaviour of pure inhibitor, with a BA/AB ratio lower than 2, even if with a low potency. This compound is characterized by a binding pose in a very apical position inside the internal cavity of P-glycoprotein (Figure #cmdc202200027-fig-0005#5). This localization enables the compound to reach four different domains (TM1, TM6 and TM12) with weak hydrophobic interactions. The peculiar pose of the compound and its “cross-linking” ability could be an explanation of the functional profile of this ligand that is able to block the protein, rather than being transported.

16606_cmdc202200027-fig-0005.jpg

Figure 5Open in figure viewerPowerPointBinding pose within the P-gp binding region of compound 7 c, behaving as a pure inhibitor.

Starting from the results reported in Table 1, we selected the highly active and selective P-gp ligand, compound 1 e, to study its ability to restore the cytotoxic activity of the antineoplastic drug doxorubicin in a co-administration assay in the “pure” model of cells overexpressing P-gp (MDCK-MDR1 cells), representative for resistant cancers. Doxorubicin, being a P-gp substrate, is effluxed by the pump out from the cell membranes, with a consequent reduction of its cytotoxic activity. Thus, we tested compound 1 e alone, to evaluate its intrinsic cytotoxicity, and in the presence of the chemotherapeutic drug at 10 μM, to evaluate its ability to increase the doxorubicin cytotoxic activity. As depicted in Figure #cmdc202200027-fig-0006#6, compound 1 e alone shows an intrinsic cytotoxicity of around 20–30 % at each tested dose. In the co-administration assay, the doxorubicin cytotoxicity increased by 50 % already at the dose of 500 nM of compound 1 e, reaching an increase of 80 % with 10 μM of compound 1 e.