ELSEVIER
Contents lists available at ScienceDirect
Molecular and Cellular Endocrinology
journal homepage: www.elsevier.com/locate/mce
-
Molecular and Celular Endocrinology
.
The impact of acyl-CoA:cholesterol transferase (ACAT) inhibitors on biophysical membrane properties depends on membrane lipid composition
Check for updates
Huong To ª, Peter Reinholdtb, Mohammad Bashawatª, Meike Luck ª, Line Lauritsen ”, Vibeke Akkerman , Matthias Kroiss ª, Daniel Wüstner, Jacob Kongsted b, Peter Müller a, ** , Holger A. Scheidt e,*
ª Humboldt University Berlin, Department of Biology, Invalidenstr. 42, 10115, Berlin, Germany
b Department of Physics, Chemistry and Pharmacy, University of Southern Denmark, DK-5230, Odense M, Denmark
” Department of Biochemistry and Molecular Biology, University of Southern Denmark, DK-5230, Odense M, Denmark
d LMU University Hospital, LMU Munich, Department of Internal Medicine IV, Ziemssenstr. 5, 80336, München, Germany
e Leipzig University, Institute for Medical Physics and Biophysics, Härtelstr. 16-18, 04107, Leipzig, Germany
ARTICLE INFO
Keywords:
Acyl-coenzyme A:cholesterol O-Acyltransferase Sterol-O-acyl transferases Acyl-coenzyme A Diacylglycerol O-Acyltransferases Nevanimibe Sandoz 58-035 AZD 3988 Lipid membrane Lipid-drug interaction
ABSTRACT
Acyl-coenzyme A: cholesterol acyltransferases are enzymes which are involved in the homeostasis of cholesterol. Impaired enzyme activity is associated with the occurrence of various diseases like Alzheimer’s disease, atherosclerosis, and cancers. At present, mitotane is the only inhibitor of this class of enzymes in clinical use for the treatment of adrenocortical carcinoma but associated with common and severe adverse effects. The thera- peutic effect of mitotane depends on its interaction with cellular membranes. The search for less toxic but equally effective compounds is hampered by an incomplete understanding of these biophysical properties. In the present study, the interaction of the three ACAT inhibitors nevanimibe, Sandoz 58-035, and AZD 3988 with membranes has been investigated using lipid model membranes in conjunction with biophysical experimental (NMR, ESR, fluorescence) and theoretical (MD simulations) approaches. The data show, that the drugs (i) incorporate into lipid membranes, (ii) differently influence the structure of lipid membranes; (iii) affect membrane structure depending on the lipid composition; and (iv) do not cause hemolysis of red blood cells. The results are discussed with regard to the use of the drugs, in particular to better understand their efficacy and possible side effects.
1. Introduction
Acyltransferases play important roles in synthesizing neutral lipids. Acyl-coenzyme A: cholesterol acyltransferases (ACAT, also named sterol-O-acyl transferases, SOAT) are membrane-bound enzymes that mediate the transfer of long-chain fatty acyl chains from coenzyme A (CoA) to cholesterol (Chol) leading to the formation of cholesteryl esters (Chang et al., 2009; Hai and Smith, 2021). In mammals, two isoenzymes, ACAT-1 and ACAT-2, exist which are encoded by two different genes. ACAT activity is tightly regulated to maintain intracellular Chol con- centration at a low level by incorporating excess Chol into cholesteryl esters at the endoplasmic reticulum (ER) membrane thereby enabling the storage of such esters in cytoplasmic lipid droplets (Xu et al., 2019). Owing to this significant role in Chol homeostasis, ACAT are involved in the pathophysiology of several diseases like Alzheimer’s disease,
atherosclerosis and cancer and, therefore, have been intensively inves- tigated as drug target to treat these diseases by specifically inhibiting enzyme activity (Hutter-Paier et al., 2004; Löhr et al., 2022; Yagyu et al., 2000; Yang et al., 2016). Importantly, mitotane, the only approved drug for the treatment of adrenocortical carcinoma, has been shown to act as a SOAT1 inhibitor (Sbiera et al., 2015).
Another family of acyltransferases, i.e. acyl-coenzyme A: diac- ylglycerol O-acyltransferases (DGAT), is responsible for catalyzing the last reaction of the glycerol phosphate pathway during triacylglycerol synthesis (Chen et al., 2022). There mainly exist two isoforms, DGAT-1 and DGAT-2, which are endoplasmic reticulum (ER)-membrane-bound proteins. Whereas DGAT-1 is most highly expressed in the small intes- tine, DGAT-2 is primarily expressed in the liver. Likewise, these enzymes play an important role in triggering various diseases, which is why their inhibition of relevant proteins is also being considered for therapeutic treatments (Naik et al., 2014; Zammit et al., 2008). The selective
* Corresponding author.
** Corresponding author. E-mail addresses: peter.mueller.3@rz.hu-berlin.de (P. Müller), holger.scheidt@medizin.uni-leipzig.de (H.A. Scheidt).
https://doi.org/10.1016/j.mce.2024.112385
| Abbreviations | NBD-PE | 1-palmitoyl-2-(12-[N-(7-nitrobenz-2-oxa-1,3-diazol-4-yl) amino]dodecanoyl]-sn-glycero-3-phosphoethanolamine; | |
|---|---|---|---|
| ACAT | acyl-coenzyme A: cholesterol acyltransferases | NBD-PS | 1-palmitoyl-2-(12-[N-(7-nitrobenz-2-oxa-1,3-diazol-4-yl) |
| DGAT | diacylglycerol O-acyltransferases | amino]dodecanoyl]-sn-glycero-3-phosphoserine; | |
| CF | 6-carboxyfluorescein | POPC | 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine; |
| Chol | cholesterol | POPC-d31 | the sn-1 chain perdeuterated analog of POPC |
| DiIC18 | 1,1'-dioctadecyl-3,3,3',3'-tetramethylindocarbocyanine per-chlorate | POPE POPE-d31 | 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine; the sn-1 chain perdeuterated analog of POPE |
| ESR | electron spin resonance | POPS | 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoserine; |
| GUVs | giant unilamellar vesicles | SL-PC | 1-palmitoyl-2-(4-doxylpentanoyl)-sn-glycero-3- |
| HBS | HEPES buffered saline; | phosphocholine; | |
| HEPES | 2-(4-(2-Hydroxyethyl)-1-piperazinyl)-ethanesulfonic acid | SL-PE | 1-palmitoyl-2-(4-doxylpentanoyl)-sn-glycero-3- |
| Ld | liquid disordered domain | phosphoethanolamine; | |
| LUVs | large unilamellar vesicles | SL-PL | spin-labeled phospholipids |
| MAS | Magic Angle Spinning | SL-PS | 1-palmitoyl-2-(4-doxylpentanoyl)-sn-glycero-3- |
| MD | molecular dynamics | phosphoserine; | |
| NBD-PC | 1-palmitoyl-2-(12-[N-(7-nitrobenz-2-oxa-1,3-diazol-4-yl) amino]dodecanoyl]-sn-glycero-3-phosphocholine; | SSM | N-Stearoyl-D-sphingomyelin; |
inhibition of DGAT-1 is investigated for the treatment of obesity and type II diabetes (Tsuda et al., 2014). The manipulation of DGAT-2 ac- tivity is considered for the treatment of DGAT-2-related liver diseases like hepatic steatosis, hepatic injury, and fibrosis (Naik et al., 2014).
In order to design efficient drugs for any disease, their mechanism(s) of action have to be understood comprehensively. This knowledge covers a multitude of complex physiological, biochemical and pharma- cological processes operating on all levels of the organism, i.e. organs, tissues, and cells which are e.g. the uptake of respective molecules by the organism, their tissue distribution and accumulation, their cellular up- take and intracellular distribution, as well as biochemical conversion and degradation processes. The optimal understanding of all these as- pects is a prerequisite for developing and applying drugs with a high efficacy and low side effects. One important mechanism of drug impact concerns the interaction of the agents with cellular membranes, once the drug has entered cells. However, despite the significance of this process, the knowledge about the impact of many applied drugs on membranes is comparatively insufficient.
With regard to ACAT inhibitors, we have characterized the interac- tion of mitotane (o,p’-dichlorodiphenyldichloroethane, o,p’-DDD, Lysodren®, structure see Fig. 1) with lipid membranes (Scheidt et al., 2016). In that study, we found that mitotane inserts into lipid
membranes and causes a disturbance of bilayer structure and an increased permeability of the membrane for polar molecules. Notably, these effects were lipid specific, in that a membrane perturbation was especially observed in the presence of phosphatidylethanolamine (PE) and Chol.
In the present study, we have focused on the membrane impact of further drugs which inhibit acyl transferase activity and which are candidates for a medical application: nevanimibe (1[[1[4(Dimethyla- mino)phenyl]cyclopentyl]methyl] 3[2,6di(propan2yl)-phenyl]urea hy- drochloride, also named ATR 101 or PD-132301), Sandoz 58-035 (3- [Decyldimethylsilyl]-N-[2-(4-methylphenyl)-1-phenethyl]propanamid) and AZD 3988 (trans-4-[-[[[5-[(3,4-Difluorophenyl)amino]-1,3,4-oxa- diazol-2yl]carbonyl]amino] phenyl]cyclo-hexaneacetic acid). The structures of the molecules are shown in Fig. 1.
Nevanimibe and Sandoz 58-035 are selective inhibitors of ACAT-1 activity (Ross et al., 1984; Saxena et al., 1995). Nevanimibe has been developed initially for the treatment of atherosclerosis and investigated as a therapy of adrenocortical cancer (Dominick et al., 1993a,b; LaPensee et al., 2016; Smith et al., 2020) but further development has been discontinued. AZD 3988 inhibits DGAT-1 (McCoull et al., 2012). Since it has been shown that mice lacking this enzyme are viable and resistant to weight gain when fed a high-fat diet, this molecule was
A
B
CI
CI
M
CH3
CI
Si
CH2(CH2)}CH3
O
CH2
CH3
CI
C
D
O
HŅ
ZI
HO
NH
N
N
F
O
O
H
F
H-CI
proposed for the therapeutic treatment of diabetes and obesity (Birch et al., 2010; McCoull et al., 2012; Zammit et al., 2008).
To the best of our knowledge, there are no studies that have addressed the interaction of these molecules with membranes. There- fore, here the effect of the drugs on membranes at the molecular level was investigated by applying different experimental (NMR, ESR, fluo- rescence) and theoretical (MD simulations) approaches.
2. Materials and methods
2.1. Materials
All labeled and unlabeled lipids, i.e. 1-palmitoyl-2-oleoyl-sn-glycero- 3-phosphocholine (POPC), 1-palmitoyl-2-oleoyl-sn-glycero-3-phos- phoethanolamine (POPE), 1-palmitoyl-2-oleoyl-sn-glycero-3-phospho- serine (POPS), N-Stearoyl-D-sphingomyelin (SSM), Cholesterol (Chol), 1-palmitoyl-2-(12-[N-(7-nitrobenz-2-oxa-1,3-diazol-4-yl)amino]dodec- anoyl]-sn-glycero-3-phosphocholine (NBD-PC), 1-palmitoyl-2-(12-[N- (7-nitrobenz-2-oxa-1,3-diazol-4-yl)amino]dodecanoyl]-sn-glycero-3- phosphoethanolamine (NBD-PE), 1-palmitoyl-2-(12-[N-(7-nitrobenz-2- oxa-1,3-diazol-4-yl)amino]dodecanoyl]-sn-glycero-3-phosphoserine (NBD-PS) and the sn-1 chain perdeuterated analogs of POPC and POPE (POPC-d31, POPE-d31) were purchased from Avanti Polar Lipids, Inc. (Alabaster, AL, USA) and 1,1’-dioctadecyl-3,3,3’,3’-tetramethylindo- carbocyanine per-chlorate (DiIC18) from Thermo Fisher (Darmstadt, Germany). Spin-labeled phospholipids (SL-PL) 1-palmitoyl-2-(4-dox- ylpentanoyl)-sn-glycero-3-phosphocholine (SL-PC), 1-palmitoyl-2-(4- doxylpentanoyl)-sn-glycero-3-phospho-ethanolamine (SL-PE), and 1- palmitoyl-2-(4-doxylpentanoyl)-sn-glycero-3-phosphoserine (SL-PS) were prepared as described previously (Fellmann et al., 1994). Neva- nimibe was purchased from MCE (Monmouth Junction, USA), Sandoz 58-035 from Merck (Darmstadt, Germany), and AZD 3988 from Tocris (Bristol, UK). All other chemicals were purchased from Sigma-Aldrich (Taufkirchen, Germany) and used without further purification.
2.2. Preparation of large unilamellar vesicles (LUVs)
LUVs were prepared by extrusion (Mayer et al., 1985) as described (Scheidt et al., 2016). For that, aliquots of the respective lipids were dissolved in chloroform, dried in a rotating round-bottom flask under vacuum and resuspended in a small volume of ethanol (final ethanol concentration was below 1% (v/v)) and, subsequently, in HBS (HEPES buffered saline, 145 mM NaCl and 10 mM HEPES, pH 7.4). For preparing the vesicles for the leakage experiments (see below), HBS additionally contained 6-carboxyfluorescein (CF). The mixture was vortexed and LUVs were prepared by subjecting the suspension to five freeze-thaw cycles. Subsequently, the lipid suspension was extruded 10 times through 0.1 um polycarbonate filters at 40 ℃ (extruder from Lipex Biomembranes Inc., Vancouver, Canada; filters from Costar, Nucleo- pore, Tübingen, Germany). LUVs for the leakage experiments (see below) were prepared accordingly, but using a mini-extruder from Avanti Polar Lipids with 11 extrusions.
2.3. Preparation of giant unilamellar vesicles (GUVs)
GUVs were prepared using the electro swelling method (Angelova et al., 1992) as described (Haralampiev et al., 2015). Briefly, the lipid mixtures were prepared from stock solutions in EtOH. Subsequently, 100 nmol of the lipids and the liquid disordered domain (Ld) marker DilC18 were spotted onto custom-built titan chambers. These were placed on a heater plate at 50 ℃ to facilitate solvent evaporation, and lipid-coated chambers were assembled using a spacer of Parafilm (Pechiney Plastic Packaging, Chicago, IL, USA) for insulation, subse- quently put under high vacuum for at least 1 h in order to evaporate remaining traces of solvent. The electro swelling chamber was filled with 1 ml sucrose buffer (250 mM sucrose, 15 mM NaN3, osmolarity of
280 mOsm/kg) and sealed with plasticine. An alternating electrical field of 10 Hz rising from 0.02 V to 1.1 V in the first 56 min was applied for 2.5 h at 55 ℃. To detach the vesicles from the slides the electrical field was changed to 4 Hz and 1.3 V and applied for 30 min.
2.4. Sample preparation for NMR experiments
For the NMR measurements, mixtures of the drugs and phospholipids were dissolved in chloroform (for AZD 3988 methanol/chloroform mixture) at the respective molar ratios. After evaporating the solvent, the samples were re-dissolved in cyclohexane and lyophilized overnight at high vacuum to obtain a fluffy powder. After hydration with 50 wt% D20 for 1H MAS NMR or with H2O-buffer (10 mM HEPES, 100 mM NaCl, pH 7.4) for 2H NMR experiments, the samples were equilibrated by freeze-thaw cycles and gentle centrifugation and finally transferred into 4 mm HR MAS rotors with spherical Kel-F inserts.
2.5. 2H NMR measurements
2H NMR measurements were conducted on a Bruker DRX300 NMR spectrometer (Bruker Biospin GmbH, Rheinstetten, Germany) operating at a resonance frequency of 46.1 MHz for 2H using a double channel solids probe equipped with a 5 mm solenoid coil. The 2H NMR spectra were obtained using a phase-cycled quadrupolar echo sequence (Davis et al., 1976) and a relaxation delay of 1 s. The two /2 pulses of~3.2 us were separated by a 50 us delay. All spectra were measured at a tem- perature of 30 ℃. After depeaking the spectra (Sternin et al., 1983) the smoothed order parameter profiles where calculated according to (Lafleur et al., 1989).
2.6. 31P NMR measurements
The static 31P NMR spectra were acquired at a Bruker Avance III 600 MHz spectrometer using a Hahn echo sequence with a relaxation delay of 2.5 s, a 1/2 pulse length of about 10 us, and low power 1H decoupling. All spectra were measured at a temperature of 30 ℃. The line shape of the spectra was numerical simulated using a self-written script in Mathcad (2001) to obtain the chemical shift anisotropy (CSA).
2.7. 1H MAS NMR spectroscopy
1H MAS NMR measurements were also carried out on a Bruker Avance III 600 MHz spectrometer using a 4 mm HR MAS probe at a MAS frequency of 6 kHz. A 2H lock was used for field stability. The 1/2 pulse length was 4 ps. In all 1H NMR spectra the chemical shift of the terminal methyl group of the lipid chains was calibrated at 0.885 ppm, which represents a referencing relative to Tetramethylsilane. All measure- ments were conducted at a temperature of 30 ℃.
Two-dimensional 1H MAS NOESY spectra (Jeener et al., 1979) were acquired at five mixing times (between 0.1 ms and 500 ms). In the in- direct dimension, 256 data points were acquired. The sum of relaxation delay and mixing times was always 3.3 s.
The volume of the respective diagonal and cross peaks was inte- grated using the Bruker Topspin 4.1 software package. NOE build-up curves were fitted according to the spin pair model yielding cross- relaxation rates (Øij) (Scheidt and Huster, 2008).
2.8. Reduction of spin-labeled phospholipids by ascorbate
For measuring the permeability of membranes, the reduction of SL- PL by ascorbate was measured as described in (Greube et al., 2001). LUVs of different lipid (2.5 mM) and SL-PL (50 uM) composition were prepared. Vesicles were mixed with the drugs or the respective solvent of the drugs (i.e. ethanol or DMSO with the same volume used for addition of the drug). Subsequently, sodium ascorbate in HBS was added (final concentration 20 mM) from a 100 mM stock solution adjusted to pH 7.4.
ESR spectra were recorded at room temperature at different times using an EMX spectrometer (Bruker, Karlsruhe, Germany). Measuring pa- rameters were as follows: modulation amplitude 4 G, power 20 mW, scan width 80 G, 1x accumulated. The decrease of ESR signal intensity was estimated by relating the intensities of the low field signal to those in the absence of ascorbate.
2.9. Reduction of fluorescent phospholipids by dithionite
As an alternative assay for determining membrane permeability, the transmembrane diffusion of the anion dithionite was measured (McIntyre and Sleight, 1991; Pomorski et al., 1994). LUVs of different lipid (1 mM) and NBD-labeled lipid (5 uM) composition were prepared. Vesicles were mixed in a cuvette with the drugs or the respective solvent of the drugs (i.e. ethanol or DMSO with the same volume used for addition of the drug). Subsequently, HBS50 (HEPES buffered saline, 145 mM NaCl and 50 mM HEPES, pH 7.4) was added and fluorescence of the NBD group was monitored continuously at 540 nm (Nex = 470 nm, slit width for excitation and emission, each 4 nm) at 37 ℃ using an Aminco Bowman Series 2 spectrofluorometer (Urbana, IL). After 30 s, sodium dithionite was added from a 1 M stock solution (freshly prepared in 100 mM Tris (pH 10)) to give a final concentration of 50 mM. Finally, after 300 s, Triton X-100 was added to a final concentration of 0.5% (w/v) enabling complete reaction of dithionite with the respective NBD-labeled lipid resulting in a loss of fluorescence. The curves were normalized to the fluorescence intensities before addition of dithionite and after addition of Triton X-100.
2.10. Measurement of 6-carboxyfluorescein (CF) leakage
The leakage of CF was measured as described (Fischer et al., 2022; Haralampiev et al., 2016). LUVs prepared in the presence of 70 mM CF were given onto a PD-10 column (GE Healthcare, Freiburg) and eluted with HBS in order to separate CF-filled LUVs from CF solved in the buffer. The vesicles were mixed in a fluorescence cuvette with HBS and the time-dependent fluorescence intensity (FI) (Nex = 490 nm; Nem = 520 nm; slit width for excitation and emission, each 4 nm) was recorded at 37 ℃ using an Aminco Bowman Series 2 spectrofluorometer. After 30 s, the drugs or the respective solvent of the drugs (i.e. ethanol or DMSO with the same volume used for addition of the drug) was added. An increase of FI reflects the release of CF from the vesicles causing a decline of CF self-quenching within the vesicles. The maximal leakage was triggered by adding 0.65% (w/v) Triton X-100 at the end of the measurement. The degree of leakage (A FI) was determined according to:
ΔΕΙ
FI330 - FI0 FImax - Flo (2)
with FI0 being the initial fluorescence intensity of the vesicles before addition of the drug/solvent, FI330 the intensity before addition, and FImax the intensity after addition of Triton X-100.
2.11. Measurement of fluorescence lifetime
The fluorescence lifetime of vesicles labeled with NBD-PL was measured as described (Fischer et al., 2023; Haralampiev et al., 2020). LUVs containing 1 mM lipid and 0.5 mol% NBD-PL were mixed with the drugs or the respective solvent of the drugs (i.e. ethanol or DMSO with the same volume used for addition of the drug), followed by dilution of the mixture in a fluorescence cuvette with HBS. The NBD-fluorescence lifetime was measured at room temperature using a FluoTime200 time-resolved spectrometer (Picoquant, Berlin, Germany). The acquisi- tion of the data was performed by time-correlated single-photon counting (excitation with a 467 nm laser, emission recording at 540 nm). Data were collected up to a level of 20.000 counts as defined by the
maximum amplitude of the fluorescence lifetime decay kinetics. The decay kinetics of lifetimes were fitted using two exponential compo- nents, from which an average fluorescence lifetime (Tav) was calculated based on the intensity weighted ratio of the two time constants that contribute to the biexponential kinetics.
2.12. Measurement of vesicle size
The size of LUVs was measured using dynamic light scattering (DLS) as described (Fischer et al., 2022, 2023). LUVs were mixed with the drugs or the respective solvent of the drugs (i.e. ethanol or DMSO with the same volume used for addition of the drug) and transferred into a cuvette (UVette, Eppendorf, Hamburg, Germany). Light scattering measurements were performed using a DynaPro NanoStar dynamic light scattering instrument (Wyatt Technologies, Dernbach, Germany) at 37 ℃. The average radius size was calculated from the mean of at least eight single experiments.
2.13. Fluorescence microscopy
For microscopic observation of GUVs containing 0.1% of the Ld marker DilC18, a Leica DMIRBE microscope equipped with an Andor IxonEM blue EMCCD camera operated at -75 ℃ and a Lambda SC smart shutter (Sutter Instrument Company, USA) as illumination control was used. The vesicles were mixed with a buffer consisting of 280 mM glucose, 11.6 mM potassium phosphate, pH 7.2, osmolarity 300 mosm on the glass bottom of a culture dish (P35GC-1.5-14-C, MatTek life sciences, USA). The drugs were added at two concentrations, giving a L/ D ratio of 5 and 10. Alternatively, the solvents ethanol or DMSO were added to the vesicles at the same volume as used for addition of the drugs. DilC18 was imaged with a 100 x 1.3 NA oil immersion Fluotar objective (Leica Lasertechnik GmbH) using a standard red excitation filter (535 nm, 50 nm bandpass), a 565 nm dichromatic mirror, and a 610 nm emission filter (75 nm bandpass). The vesicles were imaged directly after the addition of the drugs/solvents and 30 min later. The resulting images were post-processed by deconvolution using the ImageJ plugin DeconvolutionLab. The Richard-Lucy algorithm was applied with 30 iterations, and a theoretical point spread function (PSF) was utilized for deconvolution, generated using the Diffraction PSF 3D plugin in ImageJ (https://imagej.net/plugins/diffraction-psf-3d).
2.14. Measurement of hemolysis
The hemolysis of human red blood cells was measured as described (Fischer et al., 2022, 2023). Citrate stabilized human blood was ob- tained from the German Red Cross (Berlin, Germany). The buffy coat of the blood was removed after centrifugation, and erythrocytes were washed twice in HBS at 4 ℃. One mL of the washed cells was diluted with 8 mL of HBS adjusting a cell hematocrit of about 10%. For every measurement, 200 µL of the cell suspension was mixed with 200 µL of HBS containing the drugs or the respective solvent of the drugs (i.e. ethanol or DMSO with the same volume used for addition of the drug) and incubated at 37 ℃ for 10 min. Subsequently, the cells were centrifuged (1 min at 13,000g) to separate cells from the released he- moglobin. A volume of 150 µL of the supernatant was transferred to a cuvette and mixed with 50 µL of Triton X-100 (2% (w/v)). The ab- sorption of released hemoglobin was determined at a wave length of 540 nm using a spectrophotometer (BioSpectrometer basic, Eppendorf, Hamburg). The ratio of hemolysis was calculated and related to com- plete hemoglobin release, which was obtained by lysis of the cells after addition of Triton X-100.
2.15. MD (molecular dynamics)
All molecular dynamics simulations were carried out with the Gro- macs program (Abraham et al., 2015), version 2023.1. Three types of
simulations were performed. In the first set, we placed ten drug mole- cules randomly in the water phase outside a POPC membrane to study the drug uptake into the membrane. In the second set of simulations, we repeated the same setup but with a single nevanimibe molecule. In a third set of simulations, drug molecules were initially randomly distributed inside the POPC model membrane. We use this third set of simulations to study the effects the drug molecules impart on the global membrane structure.
For the first and second set of simulations, we considered pure POPC membranes with 100 POPC lipids in each leaflet, solvated in 20000 TIP3P water molecules and 44 K+ and Cl- ions (corresponding to a KCI concentration of about 0.15 M), along with ten drug molecules placed randomly in the water phase. We used a similar setup for the third set of simulations but with a smaller water layer of 10000 TIP3P waters, 22 K+, and 22 Cl” ions. The membranes in the second set of simulations contained 80 POPC lipids and 20 drug molecules in each leaflet, corre- sponding to a 20 mol% drug concentration. The initial structures were assembled with the Packmol program (Martínez et al., 2009). The lipids were described by the amber lipid17 force field (Dickson et al., 2014). Parameters for each of the drug molecules were derived using the QForce program (Sami et al., 2021), derived from a r2scan-3c (Grimme et al., 2021) quantum mechanical hessian computed using the Orca program, version 5.0.0 (Neese et al., 2020). Charges were assigned from an electrostatic potential fit (Bayly et al., 1993). Flexible dihedrals were fitted with dihedral scans. Lennard-Jones parameters were assigned from OPLS parameters (Jorgensen et al., 1996).
The membranes were equilibrated in three steps. First, the structures were minimized for 5000 steps to remove steric clashes. Next, a short 200 ps NVT dynamics run was carried out. After this, a 2 ns NPT dy- namics simulation was carried out using a semi-isotropic Berendsen barostat (Berendsen et al., 1984). Finally, a 1000 ns production run was carried out. We used a time step of 2 fs, and bonds involving hydrogens were constrained with the LINCS algorithm (Hess et al., 1997). Long-range electrostatics were treated with the particle mesh Ewald method (Darden et al., 1993) with a 12 Å short-range cutoff. The tem- perature was controlled by a Nose-Hoover thermostat (Nosé, 1984), towards 298.15 K, while the pressure was coupled with a semi-isotropic Parrinello-Rahman barostat (Parrinello and Rahman, 1981).
The resulting trajectories were analyzed with a combination of the CPPTRAJ program (Roe and Cheatham, 2013) and in-house Python scripts relying on the MDAnalysis Python library (Michaud-Agrawal et al., 2011).
2.16. Statistical analysis
Statistical analysis was conducted using IBM SPSS Statistics 24 (SPSS Inc., IBM, NY, USA). In order to determine whether the vesicle type, the drug species have a direct or interactive effect on the ascorbate permeation, a generalized linear mixed model was applied. The distri- bution of ascorbate permeation parameters (Kruskal-Wallis rank sum test) and the medians of CF leakage parameters (non-parametric median test) were compared after treatment with the different drugs. Significant differences were identified for P ≤ 0.05.
3. Results
3.1. Influence of drugs on membrane structure of LUVs membranes
To determine the impact of the drugs on membrane structure, their influence on the permeability of polar molecules across lipid bilayers was determined, which is in general comparatively low for such model membranes. However, in case of a perturbation of bilayer structure, e.g. caused by incorporation of guest molecules, the permeability may in- crease. For determining this parameter, the ascorbate-mediated reduc- tion of the ESR signal intensity of spin-labeled lipids in lipid vesicles was measured (Fischer et al., 2022; Greube et al., 2001; Haralampiev et al.,
2020)). From the kinetics of signal decrease, the rate constants kp were calculated reflecting the permeation of ascorbate across the membrane. If the kp values measured in the presence of drugs are normalized to those without drugs, it is possible to quantify the impact of the mole- cules. Since we found in a previous study that the membrane impact of mitotane depends on the lipid composition (Scheidt et al., 2016), we also studied the influence of the main lipids present in eukaryotic cell membranes, i.e. PC, PE, PS, and Chol.
Fig. 2 shows that the different drugs have distinct impacts on membranes depending on the lipid composition. In order to determine, whether the drugs species and/or the vesicle type have a general in- fluence on the ascorbate permeation, a generalized linear mixed model was applied. The data show, that (i) the drugs, (ii) the lipid composition of the vesicles, and (iii) the combination of both parameters has a sig- nificant impact on the ascorbate permeation (Table SI 1). Specifically, for nevanimibe a significant increase of kp values was found in POPC, POPC/POPE, POPC/Chol and POPC/POPE/Chol containing mem- branes. The extent of this impact was comparatively similar in these lipid mixtures and depended on the drug concentration in the investi- gated L/D range from 5 to 20. Notably, also at the comparatively low L/ D of 20, nevanimibe caused an acceleration of ascorbate reduction. In contrast, nevanimibe had a much lower impact on POPC/POPS mem- branes. AZD 3988 caused a significant membrane perturbation in POPC and POPC/POPE membranes, however to a lesser extent than observed for nevanimibe. In the presence of Chol or POPS, AZD 3988 had no impact on ascorbate permeation. Sandoz 58-035 did not show any effect in the investigated membrane mixtures. We also repeated these mea- surements for mitotane confirming the results of our previous study (Scheidt et al., 2016). Specifically, we found that the impact of this drug depends on the lipid composition of the membrane with the strongest perturbation in POPC/POPE/Chol membranes (Fig. 2). Pairwise com- parisons of the drug effects in each vesicle species were preformed using the Kruskal-Wallis rank sum test. Significant differences for P ≤ 0.05 are shown above each figure.
The influence of the drugs on membrane permeation was charac- terized by an additional fluorescence-based approach. The principle of this assay is similar to the ascorbate test but relies on measuring the fluorescence decrease of NBD-lipids in LUVs membranes upon addition of dithionite (Langner and Hui, 1993). It also allows for detecting pu- tative membrane disturbances (Scheidt et al., 2016; Tannert et al., 2007). Similar to the ascorbate assay, LUVs of similar lipid compositions (except POPC/POPS vesicles) were prepared, and the signal reduction was measured in the absence and in the presence of the drugs. From the kinetics, the kp values representing dithionite permeation in the pres- ence of drugs were determined and normalized to those measured without drugs (see Fig. SI 1). The drugs show a similar pattern of membrane impact depending on lipid species as found with the ascor- bate assay. While Sandoz 58-035 has no influence on kp values in all lipid compositions investigated, nevanimibe causes a similarly strong increase in permeability in all membranes, and AZD 3988 was solely active in POPC LUVs. The experiments were also performed using mitotane finding similar results as previously published in that the drug mainly affects POPC/POPE/Chol membranes (Fig. SI 2) (Scheidt et al., 2016). Comparing the extent of the membrane disturbance measured by the two approaches, i.e. the dithionite and the ascorbate assay, it is obvious, that the latter one reflects drug-mediated disturbances much more sensitively which is in line with former studies (Fischer et al., 2022). These differences could be caused by a different flexibility and/or position of the respective label moieties within the membrane resulting in a different accessibility of the respective quencher to the label group (Chattopadhyay and London, 1987; Huster et al., 2001; Vogel et al., 2003).
Next, we asked whether the drugs enhance the permeation of a liposome content marker, i.e. the escape of the fluorophore CF from the lumen of LUVs. To test that, the aqueous dye CF was incorporated into the vesicles at rather large concentration which causes a self-quenching
A
B
100
100
80
80
rel k
60
rel k
60
40
40
20
20
0
Neva 5:1
Neva 10:1
Neva 20:1
Sandoz 5:1
Mito 5:1
AZD 5:1
0
Neva 5:1
Neva 10:1
Neva 20:1
Sandoz 5:1
Mito 5:1
AZD 5:1
C
D
100
100
80
80
rel k,
60
rel k
60
40
40
20
20
0
Neva 5:1
Neva 10:1
Neva 20:1
Sandoz 5:1
Mito 5:1
AZD 5:1
0
Neva 5:1
Neva 10:1
Neva 20:1
Sandoz 5:1
Mito 5:1
AZD 5:1
E
40
30
rel k,
20
10
0
Neva 5:1
Sandoz 5:1
Mito 5:1
AZD 5:1
of its fluorescence. A (time-dependent) leakage of CF, relieving its self- quenching and resulting in a significant fluorescence increase was solely observed upon addition of nevanimibe in POPC LUVs (P ≤ 0.05, non-parametric median test) whereas the effect of the other drugs was in the range of the respective pure solvent used for drug solubilization (Fig. 3). Typical measuring kinetics of fluorescence increase are shown in Fig. SI 3.
Additionally, the influence of acyltransferase inhibitors on mem- brane structure was characterized by employing fluorescence lifetime measurements. This parameter measured for fluorescent lipids in membranes gives information about the molecular environment around the labeled lipid and its putative changes e.g. upon incorporation of external molecules into the membrane (Bastos et al., 2012). The fluo- rescence lifetime of NBD-PC incorporated into POPC LUVs was quanti- fied in the absence and in the presence of the drugs. From the
bi-exponential kinetics, average lifetimes (Tav) were calculated which are shown in Fig. SI 4. The respective solvents of the drug, i.e. ethanol (for nevanimibe and Sandoz 58-035) and DMSO (for AZD 3988) served as controls. The figure shows that in the presence of the solvents as well as of Sandoz 58-035, the Tav values are quite similar having a value of around 4.95 ns. After addition of AZD 3988, a slight decrease of the life time with 4.87 ns was measured. Solely, nevanimibe caused a decrease of the life time to about 4.44 ns, which indicates changes of molecular membrane structure around the NBD moiety in the presence of this drug.
3.2. Interaction of drugs with membranes measured by 31P NMR spectroscopy
Next, the interaction of the drugs with lipid bilayers was character- ized using various approaches of NMR spectroscopy allowing to
0.4
0.3
L
4
0.2
00 0
0.1
8
8
8
O
0.0
0
e
EtOH
Neva
Sandoz
Mitotane
DMSO
AZD
determine further parameters of lipid structure and mobility. 31P NMR spectra, reflecting the behaviour of lipid headgroups, were recorded for membranes having different lipid compositions without and with the drugs (See Fig. SI 5). The spectra exhibit a typical line shape charac- teristic for a lamellar bilayer phase. Only a very small amount of isotropic contributions (<3%), which also occur for pure lipid systems due to the sample preparations, were observed in some cases. This means, even in the presence of 20 mol% of the different molecules (which was necessary to obtain a reasonable signal to noise ratio in the 1H MAS measurements below), the bilayers were not significantly disturbed in their phase state. Moreover, for Sandoz 58-035 and AZD 3988 no substantial change in the CSA, the chemical shift anisotropy, was measured obtained from numerical simulations of the line shape (error auf measurement ± 1 ppm) (Table SI 2). As reported previously, mitotane had no influence on the CSA (Scheidt et al., 2016). In contrast, nevanimibe caused some increase of the CSA of all investigated phos- pholipids compared to the pure lipid systems indicating a certain decrease in the head group mobility.
3.3. Influence of the drugs on membrane order measured by 2H NMR
To characterize the effect of the different molecules on the hydro- carbon core of the lipid membranes, 2H NMR experiments using per- deuterated phospholipid molecules were conducted. In pure POPC-d31 membranes, all three molecules decreased the 2H NMR chain order parameter especially in the middle and lower chain region (Fig. 4). While the presence of AZD 3988 and Sandoz 58-035 only led to a quite small decrease, nevanimibe caused a substantial disordering of the lipid chains of POPC. This effect is also reflected in the calculated lipid chain extent (Petrache, Dodd and Brown, 2000), which is decreased in the presence of nevanimibe by 0.8 Å (Table SI 3).
The results shown above using the ascorbate and dithionite assay reflect an influence of the lipid composition on the perturbation of bilayer structure for AZD 3988 and nevanimibe. Therefore, for these two drugs, their influence on lipid order was also measured in further lipid mixtures. For AZD 3988 the data reflect smaller changes of the lipid chain order in POPC/Chol membranes compared to its effect in pure POPC membranes (Fig. SI 6).
0.25
0.20
Order parameter S
0.15
0.10
pure POPC-d31
0.05
20 mol% Nevanimibe/POPC-d3
20mol% AZD/POPC-d31
20 mol% Sandoz /POPC-d31
0.00
2
4
6
8
10
12
14
16
Carbon number n
In contrast, for nevanimibe a membrane perturbation was observed in POPC-, but also in POPE- and Chol-containing membranes. Therefore, 2H NMR chain measurements were also performed for those membrane compositions. Using POPC/POPE membranes (molar ratio 2:1), the decrease of the chain order parameters in the presence of nevanimibe (Fig. SI 7) was even more pronounced than in pure POPC membranes (compare to Fig. 4). This effect was observed for both lipids, i.e. POPC and POPE, which were separately chain deuterated in the different samples. A similar impact of nevanimibe on the order parameters was found in POPC/POPE/Chol membranes (molar ratio 1:1:1), again measured with deuterated PC and PE (Fig. SI 8). These effects are also reflected in the calculated lipid chain extents (Table SI 3). In addition, the influence of another ACAT inhibitor, i.e. mitotane, on the membrane order parameters in these membrane systems was measured (see Figs. SI 7 and SI 8) confirming the results of our previous study (Scheidt et al., 2016).
3.4. Interaction of drugs with membranes measured by 1H MAS NOESY NMR spectroscopy
1H MAS NMR measurements allow to characterize the embedding of molecules within the lipid bilayer. The respective 1D 1H MAS spectra (including the signal assignment) of the investigated drugs incorporated into POPC membranes are shown in Fig. SI 9. While the signal arising from aromatic protons are clearly distinguishable from the POPC sig- nals, some substantial signal overlap is observed for the other protons. In addition, the different proton signals of the three substances themselves overlap in some cases which makes a definite signal assignment difficult.
The 2D 1H-1H NOESY MAS spectra provide cross signals between the different protons of the membrane bound molecule and the surrounding lipid molecules. A mixing time depended measurement of these spectra allows a quantitative analysis of the cross relaxation rate between the respective molecular groups (Scheidt and Huster, 2008). Due to the strong distance dependence of cross relaxation rate and the high mobility in lipid membranes, NOESY cross relaxation rates between molecular groups of the drug molecule and POPC can be interpreted as a contact probability between them. Plotting the obtained cross relaxation rates of a proton of the drug molecule to all molecular groups of POPC the long axis of the lipid molecule provides therefore a distribution function of the drug molecule in the membrane. As shown before for several molecules (Galiullina et al., 2019; Scheidt et al., 2004, 2016; Scheidt and Huster, 2008; Weizenmann et al., 2012) this can be used to obtain the membrane location and orientation of small membrane-bound molecules. Further, these results can be cross checked
to the electron densities obtained from MD simulations.
3.4.1. Nevanimibe
The NOESY cross relaxation rates between nevanimibe and POPC are shown in Fig. 5. The signals from the aromatic protons are difficult to assign and to distinguish. Due to signal overlap, both aromatic rings contribute to the quite broad signal between 7.1 and 7.6 ppm (see Fig. SI 9). As a consequence, the obtained cross relaxation rates and therefore the distributions functions along the membrane normal, which exhibit their respective maximum in the interface/head group region of the membrane, have to be attributed to both aromatic rings. Also, cross relaxation between the four methyl groups bound to one of the aromatic rings (at 1.1 ppm) and POPC could be observed. The obtained cross relaxation rates lead to a distribution function, which place these pro- tons deeper into the membrane (upper chain region).
3.4.2. AZD 3988
The NOESY cross relaxation rates between AZD 3988 and POPC are shown in Fig. SI 10. For this drug, only broad signals of the two aromatic rings could be found well-resolved in the 1H MAS spectrum. Using chemical shift predictions (https://neural.dq.fct.unl.pt/spinus/), the signals at 7.3 and 7.5 ppm could be assigned to the respective aromatic ring. The signals of other protons (cyclohexane ring) are hidden under the POPC signals. The obtained distribution functions are very similar and both located to the aromatic ring in the upper chain glycerol region of the POPC membranes.
3.4.3. Sandoz 58-035
Sandoz 58-035 exhibits a number of proton peaks (Fig. SI 9), which are difficult to analyse due to signal overlap with the POPC signals and further on due to possible intramolecular contributions to the 2D NOESY cross peaks. In the end, only one signal of one aromatic ring and the methyl groups bound to the silicon atom could be analyzed. The
aromatic ring proton has the maximum of its distribution function in the upper chain/glycerol region of the membrane (Fig SI 11). The distri- bution function of the methyl groups exhibits its maximum for the acyl chain region - which shows that the acyl chain of Sandoz 58-035 is inserted into the hydrocarbon core of the membrane.
3.5. Influence of drugs on integrity of lipid vesicles
The impact of the drugs on lipid membranes found by the various biophysical assays (see above) may reflect (local) membrane perturba- tions without affecting the structural integrity of the vesicles. Alterna- tively, the data could also reflect a disintegration of the liposomes followed by their conversion into other lipid assemblies, such as mixed micelles. To differentiate between these options, various assays were employed which are sensitive to the integrity of (lipid) vesicles. First, the size of LUVs was measured using dynamic light scattering. POPC LUVS were prepared, and their size was measured in the presence of the drugs (using a rather large concentration, L/D = 2.5) as well as of the respective drug solvents. From the preparation method, the vesicles are expected to have a diameter of 100 nm. Indeed, for the LUVs a diameter of around 100 nm was measured at all conditions demonstrating that the vesicles kept intact (Fig. SI 12). As control, we also analyzed the vesicles after addition of Triton X-100 finding a particle size of about 18 nm. This low diameter reflects the solubilization of LUVs into micellar structures in the presence of the detergent.
Next, GUVs in the presence of the drugs were observed by fluores- cence microscopy, which allows to detect any effect of the molecules on the vesicle structure. Moreover, by preparing vesicles with lipid do- mains, an impact of the drugs on the phase behavior of the membrane can be visualized. For that, GUVs consisting of POPC, SSM, Chol, and the fluorescent DilC18, a marker for liquid disordered (Ld) domains, were prepared. Fig. 6 shows that the GUVs remained intact, both immediately and after 30 min of addition of nevanimibe, Sandoz 58-035, mitotane, or
1.2
1.1
HŅ
IZ
Cross Relaxation Rate /s-1
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
Z
0.0
CH,
(CH2)
CH2-C=
C-3
C-2
G-1
G-3
CL
A
”
H-CI
0.3
0.07
Cross Relaxation Rate /s-1
0.06
Cross Relaxation Rate /s 1
0.2
0.05
0.04
0.03
0.1
0.02
0.01
0.0
CH3
(CH2),
CH2-C=
C-3
C-2
G-1
G-3
CX
₿
0.00
7
CH,
(CH2),
CH2-C=
C-3
C-2
G-1
G-3
Cx
B
7
Nevanimibe
Sandoz 58-035
Mitotane
AZD3988
A
B
C
D
10 um
10 pm
10 μm
10 pm
E
F
G
H
10 μπ
10 pm
10 pm
10 pm
AZD 3988
Nevanimibe
100
POPC Membrane
Distance (Å)
A
E
Replica 1
100
z-position (Å)
50
50
0
-50
0
Mitotane
Replica 2
100
Distance (Å)
100
B
z-position (Å)
50
0
50
-50
0
Replica 3
100
Nevanimibe
50
Distance (Å)
z-position (Å)
100
C
0
-50
50
0
200
400
600
800
1000
Time (ns)
F
723 ns
724 ns
0
Sandoz 58-035
Distance (Å)
100
D
50
725 ns
726 ns
0
0
250
500
750
1000
Time (ns)
AZD 3988 (L/D = 5 in each case) in that (i) no destructed vesicles could be observed and (ii) the domain organization of the membranes remained intact. However, for some GUVs we observed local protrusions or deformations in the presence of Sandoz 58-035 and AZD 3988 (Fig. SI 13). In the presence of the drug’s solvents ethanol and DMSO, no al- terations were detected (Fig. SI 14).
3.6. Influence of drugs on hemolysis of human red blood cells
To determine the impact of the inhibitor molecules on the intactness of a biological membrane, their influence on the hemolysis of human red blood cells was measured. Erythrocytes were incubated with the drugs (L/D = 10) or with the respective drug solvents (same volume as added with the drugs) at 37 ℃. After centrifugation of the cells, the amount of released hemoglobin in the supernatant was measured and related to complete hemolysis (which was induced by addition of Triton X-100) reflecting the extent of cell lysis. Fig. SI 15A shows that after a 10 min- incubation, in the presence of the respective drugs as well as the respective solvents, no increase of hemolysis was detected. Since it can be expected that under physiological conditions the drugs are in contact with the cells for longer times, the degree of hemolysis was also measured after 60 min incubation. Despite a small increase of hemolysis in all samples (including the control), again no increased hemolysis in the presence of the drug molecules was observed (Fig. SI 15B).
3.7. Modeling of drug-membrane interactions using MD simulations
To obtain further insight into the molecular mechanisms underlying the observed drug effects on membranes molecular dynamics (MD) simulations were used. For comparison, such simulations were also carried out for mitotane, thereby complementing our previous experi- mental analysis of membrane activity of this inhibitor (Scheidt et al., 2016).
First, the membrane insertion of all drug molecules from the water phase into a POPC bilayer was followed over the course of MD simula- tions (Fig. 7). Ten drug molecules were initially placed in the water phase at random positions and, subsequently, allowed to move freely in the system. Notably, an initial rapid aggregation process occurs for all drug molecules, leading to small drug aggregates in the water phase, as can be seen from the intermolecular center-of-mass-distances (Fig. 7 colored lines) coalescing to values of molecular contact. For AZD 3988, the aggregate rapidly merges with the membrane, and the drug becomes dispersed within the membrane. For mitotane and Sandoz 58-035, a similar process was observed, but slightly delayed at around 500 and 600 ns, respectively. Notably, absorption into the membrane does not occur for nevanimibe during the timescale of the simulations indicating that the kinetics of membrane incorporation is somewhat slower for this drug and not accessible on the time scale of MD simulations.
The latter finding could potentially be an artifact of the self- aggregated nevanimibe molecules. Therefore, we carried out addi- tional MD simulations which included only a single nevanimibe mole- cule, using three independent replica simulations. As shown in Fig. 7E, we find that insertion into the membrane occurs in two of the three MD runs at 260 ns (replica 1) and 765 ns (replica 2). In the first replica, embedding proceeds with the carbonyl group initially inserting, fol- lowed by the dimethylaniline ring, after which the xylene ring finally enters. In replica 2 (see Fig. 7F), the insertion of nevanimibe proceeds with adsorption to the membrane surface. The molecule remains on the surface for a brief period of about 2 ns, after which complete insertion into the membrane occurs with the xylene ring first. From these single- molecule insertion MD runs, it is also clear that many close contacts between nevanimibe and the membrane surface occur without any insertion, suggesting that the barrier towards membrane insertion is somewhat higher than for the remaining drug molecules. Thus, the absence of absorption into the membrane in the earlier higher- concentration simulations could likely be explained by a slower
membrane incorporation.
Next, the effect of the membrane-incorporated drug molecules on membrane structure was determined by calculating chain order pa- rameters of the palmitic and the oleic acid of POPC within a pure POPC and a POPC/Chol membrane (at different Chol concentrations) (Fig. 8 and Fig. SI 16). The latter were used to act as a point of comparison for estimating the strength of any ordering effect. For that, a set of simu- lations was used in which the drug at a concentration of 20 mol% is initially placed randomly within the bilayer. Fig. 8 shows that the sim- ulations for all investigated drug molecules yield an increase of the order parameters of both POPC fatty acyl chains. The strongest ordering effect was found for Sandoz 58-035, with an extent similar to that of a POPC membrane having 15 mol% Chol (see Fig. SI 16). The other drugs, i.e. AZD 3988, mitotane and nevanimibe, also increase the order parame- ters, but to a slightly lesser extent, especially in the “high carbon segment” region near the membrane center.
We also considered the partial electron density traces along the membrane normal (Fig. SI 17). The traces show that the global density and the membrane thickness are only weakly influenced by the inclusion of drug molecules, with almost no discernible difference between the different POPC membrane components. The drug molecules occupy re- gions near the PC headgroups, and are only weakly exposed to the water solvent. Additionally, parts of the drug molecules extend across the entire membrane. Sandoz 58-035, in particular, has a high density in the membrane center.
4. Discussion
Acyltransferases are enzymes involved in the metabolism of lipids. Malfunctions of their activity are linked to several pathological condi- tion including metabolic syndrome, atherosclerosis, neurodegeneration and cancer. To treat those diseases, one strategy is to apply molecules which inhibit the respective enzyme activity. During the treatment of a disease, the drugs, after administration to the organism, necessarily interact with (lipid) membranes to exert their cellular effect, i.e. with plasma membranes and intracellular membranes, and any impact on these might be linked with the efficacy of the drugs. However, for most of the molecules, their interplay with membranes is not well understood. This is particularly important for acyl transferase inhibitors, as these enzymes are membrane embedded proteins, accessing their substrate from the lipid phase (Long et al., 2020, 2021). In this study, we have focused on the membrane interactions of various acyl transferase in- hibitors, which are potential candidates for medical applications.
The main results of the study are: (i) the investigated drugs AZD 3988, nevanimibe and Sandoz 58-035 incorporate into lipid mem- branes, both as monomer and as aggregate; (ii) the drugs differently influence the structure of lipid membranes; (iii) the impact of the drugs on lipid membrane structure depends on the lipid composition; and (iv) the drugs do not cause hemolysis of red blood cells. The data are compared with those of another ACAT inhibitor, i.e. mitotane, which have been published earlier (Scheidt et al., 2016).
The molecule position within the membrane was determined using NMR measurements and MD simulations. This position is a result of different physical interactions (electrostatics, dipolar interactions, hydrogen bonding) between the molecule and the lipids of the mem- brane in the very complex environment of the membrane-water- interface, in which several physical properties are largely change over a small distance (White et al., 2001). By 1H MAS NOESY NMR experi- ments, the interaction of certain drug segments with the different lipid protons was determined allowing to specify the embedding and orien- tation of the molecules within the membrane. The spectra reveal, that the basic structures of all three drugs insert mainly into the glycerol region, i.e. the upper chain region of POPC membranes. In addition, the acyl chain of Sandoz 58-035 is oriented, as one could expect, into the hydrocarbon core of the membrane. For nevanimibe, part of the mole- cule, i.e. the methyl groups which are linked to one aromatic ring, are
0.225
0.200
0.200
0.175
0.175
0.150
0.150
0.125
POPC
0.125
0.100
POPC/AZD 3988
0.075
POPC/Mitotane
0.100
POPC/Nevanimibe
0.075
0.050
POPC/Sandoz 58-035
0.025
5
10
15
5
10
15
PA carbon segment
OL carbon segment
somewhat deeper embedded into the hydrophobic region of the mem- brane. This deeper insertion of nevanimibe might be the cause for the disturbance of membrane structure observed for this drug (see below). The experimental data can be compared to the electron densities of the respective molecules obtained from MD simulations (see Fig. SI 17) in that the drug molecules are mainly also found at the edge of the bilayer with Sandoz 58-035 and nevanimibe additionally exhibiting some density in the membrane center.
By using several assays, it was investigated which influence the membrane-embedded drugs have on the structure and integrity of lipid membranes. For that, the order parameter of fatty acyl chains was measured (2H NMR) and calculated (MD). The NMR experiments revealed that all three drugs decrease the chain order parameter (only determined for the palmitic acid) especially in the middle and lower chain region with nevanimibe causing the largest disordering. In contrast, the MD calculations describe a different situation. Here, the drugs increase the order parameters (calculated for both fatty acyl chains of POPC) with the strongest ordering effect for Sandoz 58-035. One potential explanation for this discrepancy is that the computa- tional and experimental setups describe different physical scenarios. In the MD simulations, the drugs are initially distributed in a random configuration, and due to the rather short time-scale accessible in the membrane simulations, the drugs will remain relatively close to their initial positions in the membrane. In the experimental setup, the mem- branes undergo extensive equilibration with freeze-thaw cycles, and this can affect equilibrium distribution of the membrane embedded drugs in various ways. For example, it is possible that the drugs are forming oligomeric states inside the bilayer, which are not captured in the sim- ulations. Indeed, our simulations show that small drug aggregates can form already in the water phase, and such aggregates, once inserted into the bilayer, could affect membrane properties very differently from drug monomers. Drug oligomerization inside the bilayer could also take place, but likely on a much longer time scale than accessible by atomistic MD simulations.
As another approach reflecting membrane structure, the influence of the drugs on the permeation of anions was measured. The results of the ascorbate and the dithionite assay reflect the impact of the molecules on membrane integrity and its dependence on the lipid composition in a comparable way. Especially for nevanimibe, a strong increase in anion permeation was found in all lipid mixtures tested, except in PS- containing membranes, which showed only a small permeation in- crease, may due to the negative charge of POPS. AZD 3988 also disturbs membrane integrity to a lesser extent. But its effect seems to depend on the presence of PC and is absent in the presence of Chol and/or PS. The increase of the reduction constants in the presence of AZD 3988 was largest in pure PC membranes, smaller in PC/PE vesicles and completely suppressed in Chol- or PS-containing membranes. In contrast, for Sandoz 58-035 no effect at all was measured in all membranes tested, which is a may a result of an easy insertion of Sandoz 58-035 with its acyl chain as anchor into the hydrocarbon core of the membrane. Repeating those
measurements for mitotane, we obtained similar results as previously published, in that this drug showed the strongest effect in PC/PE/Chol- containing membranes (Scheidt et al., 2016). Of note, the results of the ascorbate and dithionite assay could also be explained by an accelerated lipid flip flop in the presence of drugs, in that an enhanced flop of (labeled) lipids would also cause an accelerated ESR (electron para- magnetic resonance) or fluorescence signal decrease. In that scenario, the drug molecules could accelerate transbilayer migration of NBD-versus spin-labeled lipids to a different degree. With our approach, we cannot differentiate between both explanations. However, in any case, the results reflect a drug-mediated membrane disturbance in that the trans-bilayer permeation of a polar molecule, i.e. dithioni- te/ascorbate or phospholipid head group, is increased.
The profound impact of nevanimibe on membrane integrity may indicate a dissolution of the bilayer structure in that the drug molecules trigger the formation of other supramolecular, non-bilayer assemblies. However, several lines of evidence argue against this assumption. First, the line shapes of the NMR spectra reflect the existence of mainly lipid bilayers in the presence of nevanimibe and also the other drugs. Second, the size of LUVs in the presence of the drugs was not altered which contradicts a dissolution of the vesicles. Third, nevanimibe did not alter the morphology or domain structure of GUVs. Therefore, it is assumed that the observed perturbing influence of nevanimibe on membrane integrity can be explained by (local) disturbances of the bilayer struc- ture, like e.g. (specific) interactions of the drug with lipids. This assumption is supported by fluorescence life time measurements which show that the fluorescent NBD-PC detects in the presence of nevanimibe a different surrounding compared to the control as seen from a decreased life time. For the other drugs, no influence on the fluorescence life time of NBD-PC was observed.
The various experiments reflect an increased membrane perme- ability in the presence of nevanimibe for molecules having a low-weight (dithionite, ascorbate, CF), but not for large molecules, such as proteins (hemoglobin). This observation could play an important role for un- derstanding the mechanism(s) by which this ACAT inhibitor acts in cells; apart from the known direct interaction with the ACAT enzyme (Long et al., 2020, 2021). The drug might increase the permeability of sub- cellular membranes for low-molecular weight chemicals (e.g. ions and metabolites) thereby affecting metabolism and overall cell homeostasis. For example, the permeability of mitochondrial membranes for metab- olites is tightly controlled, and an integrate part of several metabolic pathways, such as fatty acid oxidation, ATP production or the coupling between glycolysis and citric acid cycle. Drug-induced alterations in mitochondrial membrane permeability could change such coupling be- tween transport and metabolism, e.g., by affecting the activity of the malate-aspartate shuttle, thereby altering glycolysis of tumor cells (Wang et al., 2016).
For understanding the interaction of the drugs with lipid bilayers on a molecular level, their different physico-chemical properties of the drugs and their influence to the physical interactions in the complex
membrane environment have to be considered. These properties are described among others by the logP value, which quantitatively reflects the molecular hydrophobicity suggesting that molecules with a larger logP values have a stronger propensity to incorporate into membranes. With regard to the molecules investigated in the present study, their logP values are: AZD 3988 = 4.16; nevanimibe = 5.66; mitotane = 6.00; Sandoz 58-035 = 8,93; (calculated with https://molinspiration.com/). Considering only this parameter, one would expect the strongest inter- action with membranes for Sandoz 58-035 and the lowest for AZD 3988, which is not reflected by our data (note, that a difference in logP value of one reflects a distribution difference of one order of magnitude between n-octanol and water). This emphasizes that in order to understand the complex interaction of drugs with membranes, additional aspects must be taken into account, like steric orientations, dipolar moment, number and spatial arrangement of hydrogen bond donor and acceptor atoms, and topological polar surface area (Lipinski et al., 2001).
Regarding the experimental procedure, the assays in which the molecules were added to existing lipid bilayer membranes are closer to the in vivo (cellular) situation. For the NMR measurements, the drugs were already mixed in organic solvent with the lipids before membrane preparation (necessary due to the low water content in solid-state NMR sample). However, it can be presumed that the addition of drugs to existing membranes should lead to similar membrane effects as the incorporation of the molecules during membrane formation, if the molecules have sufficient time to incorporate into the membrane reaching a thermodynamic equilibrium. Therefore, we have assessed membrane-incorporated drugs as well. Considering the transfer of drugs from buffer into the membrane, the solubility of the molecules in water has to be taken into account. Hydrophobic molecules with large logP values should tend to form supramolecular structures in buffer, like micelles or aggregates.
To check the formation of drug aggregates in buffer, we have determined the light scattering in HBS at increasing drug concentrations (Fig. SI 18) showing that the scattering is increasing for AZD 3988, mitotane, or Sandoz 58-035 measured up to 400-500 uM drug. For nevanimibe, the curve reaches a plateau at about 200 µM. Moreover, using a pyrene-based assay (Kalyanasundaram and Thomas, 1977) we have determined the critical micellar concentration (cmc) of the drugs in HBS (Fig. SI 19) giving values in the range from about 2 µM (Sandoz 58-035) to about 100 µM (AZD 3988). Especially, the low cmc of Sandoz 58-035 is certainly caused by the alkyl chain that this drug has (see Fig. 1).
The presence of the drugs in buffer as micelles and/or aggregates may (i) hamper their transfer into membranes or (ii) cause detergent- like effects that can lead to a disturbance of membrane structure. With regard to the first point, drug molecules may insert into the bilayer either as monomers or the aggregates merge with the membrane. However, the physiological situation is probably different, since (hy- drophobic) drugs in the serum are often localized in lipid vesicles (Jang et al., 2023; Yáñez-Mó et al., 2015) or bound to plasma proteins (Deeks, 2016; Peng et al., 2005) and only a low amount is available as free molecules. It can be surmized that those compartments serve as a reservoir for continuous release of drug molecules for an incorporation into plasma membranes. With regard to the second point, we believe, that detergent-like effects of the drugs are less probable as a cause of the observed membrane impacts, especially observed for nevanimibe, since we did not find any significant influence of the drugs on the size of LUV which was measured at a comparatively large concentration (400 µM).
Indeed, for all investigated drugs the MD simulations initially found the formation of aggregates that - with the exception of nevanimibe - subsequently merge with the membrane. The observation that nevani- mibe was not found in the membrane in the simulations indicates a slower membrane incorporation for this drug. One has to note, that the simulated accessible time scale of 1 us is rather short compared to experimentally relevant conditions. Notably, aggregate formation was most effective for nevanimibe, whereas the rise of the curve was the
smallest for Sandoz 58-035. For the latter, having a long hydrophobic acyl chain, one could expect the formation of micelles in water.
We note, that our study is not aimed to characterize drug-membrane interactions with a therapeutic background, since the investigated molecules have been, in contrast to mitotane, either discontinued or have not been authorized for a human use. In addition, the data only contribute to a limited understanding of the molecular processes of ACAT inhibition (Long et al., 2020, 2021; Websdale et al., 2022). However, studies like this one improve the knowledge about basic mechanisms of interaction of ACAT inhibitors with cells/membranes compared with the medically used mitotane. For example, it is conceivable that inhibition of ACAT, which are membrane-spanning proteins (Chang et al., 2009), leads to a local increase in membrane cholesterol content that profoundly alters local membrane properties. Perturbation(s) of membrane structure/integrity might be linked to side effects of the drugs, like triggering of cell apoptosis or lysis. The lipid composition of the vesicles used here was chosen to mimic the main lipids present in biological membranes. The composition of the latter is, however, much more complex and contains a multitude of lipid and protein species. According to our results, the plasma membrane, at least that of red blood cells, is not damaged by ACAT inhibitors. The composition of intracellular membranes profoundly differs from that of plasma membranes. E.g. Chol accounts for up to 50 mol% of plasma membrane lipids, whereas it contributes to endoplasmic reticulum lipids (which is the organelle harbouring ACAT) only to about 5% (Radhakrishnan et al., 2008). Therefore, the impact of a drug on intra- cellular membranes might differ from that on plasma membranes. Changes of membrane structure and dynamics upon interaction with a drug may modify the activity of membrane-bound/integrated enzymes. Various ACAT inhibitors have been shown to cause toxic effects on he- patocytes and other cells (Dominick et al., 1993a,b; Kellner-Weibel et al., 1998; Yang et al., 2016), which could be caused by increased levels of intracellular free cholesterol due to absence of cholesterol ester formation. For example, nevanimibe is toxic towards adrenal cortex cells (LaPensee et al., 2016) while Sandoz 58-035 is to a much lesser extent although being also a potent ACAT inhibitor (Sbiera et al., 2015). Those processes might be influenced by effects of the drugs on plasma/in- tracellular membranes.
Collectively, our investigation of drug-membrane interactions using lipid model membranes demonstrates that the physicochemical prop- erties of drugs that target lipid metabolism must be considered to disentangle on-target effects caused by ACAT inhibition from physico- chemical effects that may be linked both to efficacy and side effects.
CRediT authorship contribution statement
Huong To: Writing - original draft, Investigation, Data curation. Peter Reinholdt: Writing - original draft, Investigation, Data curation. Mohammad Bashawat: Writing - original draft, Investigation, Data curation. Meike Luck: Writing - original draft, Supervision, Investiga- tion. Line Lauritsen: Writing - original draft, Investigation, Data curation. Vibeke Akkerman: Writing - original draft, Investigation, Formal analysis, Data curation. Matthias Kroiss: Writing - review & editing, Supervision, Conceptualization. Daniel Wüstner: Writing - review & editing, Supervision, Conceptualization. Jacob Kongsted: Writing - review & editing, Supervision, Conceptualization. Peter Müller: Writing - review & editing, Writing - original draft, Project administration, Conceptualization. Holger A. Scheidt: Writing - review & editing, Writing - original draft, Investigation, Data curation.
Funding
This work was supported by the Deutsche Forschungsgemeinschaft to P.M. (MU 1017/14-1) and D.W. (MU 1017/14-1, Mercator Fellow).
Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
This study was supported by a grant of the Deutsche For- schungsgemeinschaft to P.M. (grant MU 1017/14-1) and D.W. (grant MU 1017/14-1, Mercator Fellow). We thank Sabine Schiller (Humboldt University Berlin) for technical assistance.
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi. org/10.1016/j.mce.2024.112385.
Data availability
Force-field parameters for the drug molecules are available at https://zenodo.org/records/13746548. All other data will be made available on request.
References
Abraham, M.J., Murtola, T., Schulz, R., Páll, S., Smith, J.C., Hess, B., Lindahl, E., 2015. GROMACS: high performance molecular simulations through multi-level parallelism from laptops to supercomputers. SoftwareX 1-2, 19-25.
Angelova, M.I., Soleau, S., Meleard, P., Faucon, J.F., Bothorel, P., 1992. Preparation of giant vesicles by external AC electric fields. Kinetics and applications. Progr. Colloid Polym. Sci. 89, 127-131.
Bastos, A.E.P., Scolari, S., Stöckl, M., de Almeida, R.F.M., 2012. Chapter three - applications of fluorescence lifetime spectroscopy and imaging to lipid domains in vivo. In: conn, P.M. (Ed.), Methods in Enzymology. Academic Press, pp. 57-81.
Bayly, C.I., Cieplak, P., Cornell, W., Kollman, P.A., 1993. A well-behaved electrostatic potential based method using charge restraints for deriving atomic charges: the RESP model. The Journal of Physical Chemistry 97, 10269-10280.
Berendsen, H.J.C., Postma, J.P.M., Gunsteren, W.F.v., DiNola, A., Haak, J.R., 1984. Molecular dynamics with coupling to an external bath. J. Chem. Phys. 81, 3684-3690.
Birch, A.M., Buckett, L.K., Turnbull, A.V., 2010. DGAT1 inhibitors as anti-obesity and anti-diabetic agents. Curr Opin Drug Discov Devel 13, 489-496.
Chang, T.Y., Li, B.L., Chang, C.C., Urano, Y., 2009. Acyl-coenzyme A:cholesterol acyltransferases. Am. J. Physiol. Endocrinol. Metab. 297, E1-E9.
Chattopadhyay, A., London, E., 1987. Parallax method for direct measurement of membrane penetration depth utilizing fluorescence quenching by spin-labeled phospholipids. Biochemistry 26, 39-45.
Chen, G., Harwood, J.L., Lemieux, M.J., Stone, S.J., Weselake, R.J., 2022. Acyl-CoA: diacylglycerol acyltransferase: properties, physiological roles, metabolic engineering and intentional control. Prog. Lipid Res. 88, 101181.
Darden, T., York, D., Pedersen, L., 1993. Particle mesh Ewald: an N·log(N) method for Ewald sums in large systems. J. Chem. Phys. 98, 10089-10092.
Davis, J.H., Jefferey, K.R., Bloom, M., Valic, M.I., Higgs, T.P., 1976. Quadrupolar echo deuteron magnetic resonance spectroscopy in ordered hydrocarbon chains. Chem. Phys. Lett. 42, 390.
Deeks, E.D., 2016. Ceritinib: a review in ALK-positive advanced NSCLC. Target Oncol 11, 693-700.
Dickson, C.J., Madej, B.D., Skjevik, A.A., Betz, R.M., Teigen, K., Gould, I.R., Walker, R.C., 2014. Lipid14: the amber lipid force field. J. Chem. Theor. Comput. 10, 865-879.
Dominick, M.A., Bobrowski, W.A., MacDonald, J.R., Gough, A.W., 1993a. Morphogenesis of a zone-specific adrenocortical cytotoxicity in Guinea pigs administered PD 132301-2, an inhibitor of acyl-CoA:cholesterol acyltransferase. Toxicol. Pathol. 21, 54-62.
Dominick, M.A., McGuire, E.J., Reindel, J.F., Bobrowski, W.F., Bocan, T.M., Gough, A. W., 1993b. Subacute toxicity of a novel inhibitor of acyl-CoA: cholesterol acyltransferase in beagle dogs. Fundam. Appl. Toxicol. 20, 217-224.
Fellmann, P., Zachowski, A., Devaux, P.F., 1994. Synthesis and use of spin-labeled lipids for studies of the transmembrane movement of phospholipids. In: Graham, JM, Higgins, J (Eds.), Methods in Molecular Biology. Biomembrane Protocols: II. Architecture and Function. Humana Press Inc, Totowa NJ, pp. 161-175.
Fischer, M., Luck, M., Werle, M., Vogel, A., Bashawat, M., Ludwig, K., Scheidt, H.A., Müller, P., 2023. The small-molecule kinase inhibitor ceritinib, unlike imatinib, causes a significant disturbance of lipid membrane integrity: a combined experimental and MD study. Chem. Phys. Lipids 257, 105351.
Fischer, M., Müller, P., Scheidt, H.A., Luck, M., 2022. Drug-membrane interactions: effects of virus-specific RNA-dependent RNA polymerase inhibitors remdesivir and favipiravir on the structure of lipid bilayers. Biochemistry 61, 1392-1403.
Galiullina, L.F., Scheidt, H.A., Huster, D., Aganov, A., Klochkov, V., 2019. Interaction of statins with phospholipid bilayers studied by solid-state NMR spectroscopy. Biochimica et Biophysica Acta (BBA) - Biomembranes. 1861, 584-593.
Greube, A., Müller, K., Töpfer-Petersen, E., Herrmann, A., Müller, P., 2001. Influence of the bovine seminal plasma protein PDC-109 on the physical state of membranes. Biochemistry 40, 8326-8334.
Grimme, S., Hansen, A., Ehlert, S., Mewes, J.M., 2021. r(2)SCAN-3c: a “Swiss army knife” composite electronic-structure method. J. Chem. Phys. 154, 064103.
Hai, Q., Smith, J.D., 2021. Acyl-coenzyme A: cholesterol acyltransferase (ACAT) in cholesterol metabolism: from its discovery to clinical trials and the genomics era. Metabolites 11.
Haralampiev, I., Alonso de Armiño, D.J., Luck, M., Fischer, M., Abel, T., Huster, D., Di Lella, S., Scheidt, H.A., Müller, P., 2020. Interaction of the small-molecule kinase inhibitors tofacitinib and lapatinib with membranes. Biochimica et Biophysica Acta (BBA) - Biomembranes. 1862, 183414.
Haralampiev, I., Mertens, M., Schwarzer, R., Herrmann, A., Volkmer, R., Wessig, P., Müller, P., 2015. Recruitment of SH-containing peptides to lipid and biological membranes through the use of a palmitic acid functionalized with a maleimide group. Angew. Chem. Int. Ed. 54, 323-326.
Haralampiev, I., Scheidt, H.A., Abel, T., Luckner, M., Herrmann, A., Huster, D., Müller, P., 2016. The interaction of sorafenib and regorafenib with membranes is modulated by their lipid composition. Biochim. Biophys. Acta 1858, 2871-2881.
Hess, B., Bekker, H., Berendsen, H.J.C., Fraaije, J.G.E.M., 1997. LINCS: a linear constraint solver for molecular simulations. J. Comput. Chem. 18, 1463-1472. Huster, D., Muller, P., Arnold, K., Herrmann, A., 2001. Dynamics of membrane penetration of the fluorescent 7-nitrobenz-2-oxa-1,3-diazol-4-yl (NBD) group attached to an acyl chain of phosphatidylcholine. Biophys. J. 80, 822-831.
Hutter-Paier, B., Huttunen, H.J., Puglielli, L., Eckman, C.B., Kim, D.Y., Hofmeister, A., Moir, R.D., Domnitz, S.B., Frosch, M.P., Windisch, M., Kovacs, D.M., 2004. The ACAT inhibitor CP-113,818 markedly reduces amyloid pathology in a mouse model of Alzheimer’s disease. Neuron 44, 227-238.
Jang, Y.J., Cha, B.S., Kim, D., Lee, E.S., Kim, S., Han, J., Shin, J., Kim, S., Park, K.S., 2023. Extracellular vesicles, as drug-delivery vehicles, improve the biological activities of astaxanthin. Antioxidants 12.
Jeener, J., Meier, B.H., Bachmann, P., Ernst, R.R., 1979. Investigation of exchange processes by two-dimensional NMR spectroscopy. J. Chem. Phys. 71, 4546-4553. Jorgensen, W.L., Maxwell, D.S., Tirado-Rives, J., 1996. Development and testing of the OPLS all-atom force field on conformational energetics and properties of organic liquids. J. Am. Chem. Soc. 118, 11225-11236.
Kalyanasundaram, K., Thomas, J.K., 1977. Environmental effects on vibronic band intensities in pyrene monomer fluorescence and their application in studies of micellar systems. J. Am. Chem. Soc. 99, 2039-2044.
Kellner-Weibel, G., Jerome, W.G., Small, D.M., Warner, G.J., Stoltenborg, J.K., Kearney, M.A., Corjay, M.H., Phillips, M.C., Rothblat, G.H., 1998. Effects of intracellular free cholesterol accumulation on macrophage viability: a model for foam cell death. Arterioscler. Thromb. Vasc. Biol. 18, 423-431.
Lafleur, M., Fine, B., Sternin, E., Cullis, P.R., Bloom, M., 1989. Smoothed orientational order profile of lipid bilayers by 2H-nuclear magnetic resonance. Biophys. J. 56, 1037-1041.
Langner, M., Hui, S.W., 1993. Dithionite penetration through phospholipid bilayers as a measure of defects in lipid molecular packing. Chem. Phys. Lipids 65, 23-30.
LaPensee, C.R., Mann, J.E., Rainey, W.E., Crudo, V., Hunt 3rd, S.W., Hammer, G.D., 2016. ATR-101, a selective and potent inhibitor of acyl-CoA acyltransferase 1, induces apoptosis in H295R adrenocortical cells and in the adrenal cortex of dogs. Endocrinology 157, 1775-1788.
Lipinski, C.A., Lombardo, F., Dominy, B.W., Feeney, P.J., 2001. Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings. Adv. Drug Deliv. Rev. 46, 3-26.
Löhr, M., Härtig, W., Schulze, A., Kroiß, M., Sbiera, S., Lapa, C., Mages, B., Strobel, S., Hundt, J.E., Bohnert, S., Kircher, S., Janaki-Raman, S., Monoranu, C.M., 2022. SOAT1: a suitable target for therapy in high-grade astrocytic glioma? Int. J. Mol. Sci. 23.
Long, T., Liu, Y., Li, X., 2021. Molecular structures of human ACAT2 disclose mechanism for selective inhibition. Structure 29, 1410-1418.e4.
Long, T., Sun, Y., Hassan, A., Qi, X., Li, X., 2020. Structure of nevanimibe-bound tetrameric human ACAT1. Nature 581, 339-343.
Martínez, L., Andrade, R., Birgin, E.G., Martínez, J.M., 2009. PACKMOL: a package for building initial configurations for molecular dynamics simulations. J. Comput. Chem. 30, 2157-2164.
Mayer, L.D., Hope, M.J., Cullis, R.P., Janoff, A.S., 1985. Solute distributions and trapping efficiencies observed in freeze-thawed multilamellar vesicles. Biochim. Biophys. Acta 817, 193-196.
McCoull, W., Addie, M.S., Birch, A.M., Birtles, S., Buckett, L.K., Butlin, R.J., Bowker, S.S., Boyd, S., Chapman, S., Davies, R.D., Donald, C.S., Green, C.P., Jenner, C., Kemmitt, P.D., Leach, A.G., Moody, G.C., Gutierrez, P.M., Newcombe, N.J., Nowak, T., Packer, M.J., Plowright, A.T., Revill, J., Schofield, P., Sheldon, C., Stokes, S., Turnbull, A.V., Wang, S.J., Whalley, D.P., Wood, J.M., 2012. Identification, optimisation and in vivo evaluation of oxadiazole DGAT-1 inhibitors for the treatment of obesity and diabetes. Bioorg Med Chem Lett 22, 3873-3878.
McIntyre, J.C., Sleight, R.G., 1991. Fluorescence assay for phospholipid membrane asymmetry. Biochemistry 30, 11819-11827.
Michaud-Agrawal, N., Denning, E.J., Woolf, T.B., Beckstein, O., 2011. MDAnalysis: a toolkit for the analysis of molecular dynamics simulations. J. Comput. Chem. 32, 2319-2327.
Naik, R., Obiang-Obounou, B.W., Kim, M., Choi, Y., Lee, H.S., Lee, K., 2014. Therapeutic strategies for metabolic diseases: small-molecule diacylglycerol acyltransferase (DGAT) inhibitors. ChemMedChem 9, 2410-2424.
Neese, F., Wennmohs, F., Becker, U., Riplinger, C., 2020. The ORCA quantum chemistry program package. J. Chem. Phys. 152, 224108.
Nosé, S., 1984. A molecular dynamics method for simulations in the canonical ensemble. Mol. Phys. 52, 255-268.
Parrinello, M., Rahman, A., 1981. Polymorphic transitions in single crystals: a new molecular dynamics method. J. Appl. Phys. 52, 7182-7190.
Peng, B., Lloyd, P., Schran, H., 2005. Clinical pharmacokinetics of imatinib. Clin. Pharmacokinet. 44, 879-894.
Petrache, H.I., Dodd, S.W., Brown, M.F., 2000. Area per lipid and acyl length distributions in fluid phosphatidylcholines determined by (2)H NMR spectroscopy. Biophys. J. 79, 3172-3192.
Pomorski, T., Herrmann, A., Zachowski, A., Devaux, P.F., Müller, P., 1994. Rapid determination of the transbilayer distribution of NBD-phospholipids in erythrocyte membranes with dithionite. Molec. Membr. Biol. 11, 39-44.
Radhakrishnan, A., Goldstein, J.L., McDonald, J.G., Brown, M.S., 2008. Switch-like control of SREBP-2 transport triggered by small changes in ER cholesterol: a delicate balance. Cell Metabol. 8, 512-521.
Roe, D.R., Cheatham 3rd, T.E., 2013. PTRAJ and CPPTRAJ: software for processing and analysis of molecular dynamics trajectory data. J Chem Theory Comput 9, 3084-3095.
Ross, A.C., Go, K.J., Heider, J.G., Rothblat, G.H., 1984. Selective inhibition of acyl coenzyme A:cholesterol acyltransferase by compound 58-035. J. Biol. Chem. 259, 815-819.
Sami, S., Menger, M., Faraji, S., Broer, R., Havenith, R.W.A., 2021. Q-force: quantum mechanically augmented molecular force fields. J Chem Theory Comput 17, 4946-4960.
Saxena, U., Ferguson, E., Newton, R.S., 1995. Acyl-coenzyme A:cholesterol- acyltransferase (ACAT) inhibitors modulate monocyte adhesion to aortic endothelial cells. Atherosclerosis 112, 7-17.
Sbiera, S., Leich, E., Liebisch, G., Sbiera, I., Schirbel, A., Wiemer, L., Matysik, S., Eckhardt, C., Gardill, F., Gehl, A., Kendl, S., Weigand, I., Bala, M., Ronchi, C.L., Deutschbein, T., Schmitz, G., Rosenwald, A., Allolio, B., Fassnacht, M., Kroiss, M., 2015. Mitotane inhibits sterol-O-acyl transferase 1 triggering lipid-mediated endoplasmic reticulum stress and apoptosis in adrenocortical carcinoma cells. Endocrinology 156, 3895-3908.
Scheidt, H.A., Haralampiev, I., Theisgen, S., Schirbel, A., Sbiera, S., Huster, D., Kroiss, M., Müller, P., 2016. The adrenal specific toxicant mitotane directly interacts with lipid membranes and alters membrane properties depending on lipid composition. Molec. Cell. Endocrin. 428, 68-81.
Scheidt, H.A., Huster, D., 2008. The Interaction of small molecules with phospholipid membranes studied by 1H NOESY NMR under Magic-Angle Spinning. Acta Pharmacol. Sin. 29, 35-49.
Scheidt, H.A., Pampel, A., Nissler, L., Gebhardt, R., Huster, D., 2004. Investigation of the membrane localization and distribution of flavonoids by high-resolution magic angle spinning NMR spectroscopy. Biochimica et Biophysica Acta (BBA) - Biomembranes. 1663, 97-107.
Smith, D.C., Kroiss, M., Kebebew, E., Habra, M.A., Chugh, R., Schneider, B.J., Fassnacht, M., Jafarinasabian, P., Ijzerman, M.M., Lin, V.H., Mohideen, P., Naing, A., 2020. A phase 1 study of nevanimibe HCl, a novel adrenal-specific sterol O-
acyltransferase 1 (SOAT1) inhibitor, in adrenocortical carcinoma. Invest New Drugs 38, 1421-1429.
Sternin, E., Bloom, M., Mackay, A.L., 1983. De-pake-ing of NMR spectra. J. Magn. Reson. 55, 274-282.
Tannert, A., Topfer-Petersen, E., Herrmann, A., Muller, K., Muller, P., 2007. The lipid composition modulates the influence of the bovine seminal plasma protein PDC-109 on membrane stability. Biochemistry 46, 11621-11629.
Tsuda, N., Kumadaki, S., Higashi, C., Ozawa, M., Shinozaki, M., Kato, Y., Hoshida, K., Kikuchi, S., Nakano, Y., Ogawa, Y., Furusako, S., 2014. Intestine-targeted DGAT1 inhibition improves obesity and insulin resistance without skin aberrations in mice. PLoS One 9, e112027.
Vogel, A., Scheidt, H.A., Huster, D., 2003. The distribution of lipid attached spin probes in bilayers: application to membrane protein topology. Biophys. J. 85, 1691-1701.
Wang, C., Chen, H., Zhang, M., Zhang, J., Wei, X., Ying, W., 2016. Malate-aspartate shuttle inhibitor aminooxyacetic acid leads to decreased intracellular ATP levels and altered cell cycle of C6 glioma cells by inhibiting glycolysis. Cancer Lett. 378, 1-7. Websdale, A., Kiew, Y., Chalmers, P., Chen, X., Cioccoloni, G., Hughes, T.A., Luo, X., Mwarzi, R., Poirot, M., Røberg-Larsen, H., Wu, R., Xu, M., Zulyniak, M.A., Thorne, J. L., 2022. Pharmacologic and genetic inhibition of cholesterol esterification enzymes reduces tumour burden: a systematic review and meta-analysis of preclinical models. Biochem. Pharmacol. 196, 114731.
Weizenmann, N., Huster, D., Scheidt, H.A., 2012. Interaction of local anesthetics with lipid bilayers investigated by 1H MAS NMR spectroscopy. Biochim. Biophys. Acta 1818, 3010-3018.
White, S.H., Ladokhin, A.S., Jayasinghe, S., Hristova, K., 2001. How membranes shape protein structure. J. Biol. Chem. 276, 32395-32398.
Xu, Y., Du, X., Turner, N., Brown, A.J., Yang, H., 2019. Enhanced acyl-CoA:cholesterol acyltransferase activity increases cholesterol levels on the lipid droplet surface and impairs adipocyte function. J. Biol. Chem. 294, 19306-19321.
Yagyu, H., Kitamine, T., Osuga, J., Tozawa, R., Chen, Z., Kaji, Y., Oka, T., Perrey, S., Tamura, Y., Ohashi, K., Okazaki, H., Yahagi, N., Shionoiri, F., Iizuka, Y., Harada, K., Shimano, H., Yamashita, H., Gotoda, T., Yamada, N., Ishibashi, S., 2000. Absence of ACAT-1 attenuates atherosclerosis but causes dry eye and cutaneous xanthomatosis in mice with congenital hyperlipidemia. J. Biol. Chem. 275, 21324-21330.
Yáñez-Mó, M., Siljander, P.R., Andreu, Z., Zavec, A.B., Borràs, F.E., Buzas, E.I., Buzas, K., Casal, E., Cappello, F., Carvalho, J., Colás, E., Cordeiro-da Silva, A., Fais, S., Falcon- Perez, J.M., Ghobrial, I.M., Giebel, B., Gimona, M., Graner, M., Gursel, I., Gursel, M., Heegaard, N.H., Hendrix, A., Kierulf, P., Kokubun, K., Kosanovic, M., Kralj-Iglic, V., Krämer-Albers, E.M., Laitinen, S., Lässer, C., Lener, T., Ligeti, E., Line, A., Lipps, G., Llorente, A., Lötvall, J., Manček-Keber, M., Marcilla, A., Mittelbrunn, M., Nazarenko, I., Nolte-‘t Hoen, E.N., Nyman, T.A., O’Driscoll, L., Olivan, M., Oliveira, C., Pállinger, É., Del Portillo, H.A., Reventós, J., Rigau, M., Rohde, E., Sammar, M., Sánchez-Madrid, F., Santarém, N., Schallmoser, K., Ostenfeld, M.S., Stoorvogel, W., Stukelj, R., Van der Grein, S.G., Vasconcelos, M.H., Wauben, M.H., De Wever, O., 2015. Biological properties of extracellular vesicles and their physiological functions. J. Extracell. Vesicles 4, 27066.
Yang, W., Bai, Y., Xiong, Y., Zhang, J., Chen, S., Zheng, X., Meng, X., Li, L., Wang, J., Xu, C., Yan, C., Wang, L., Chang, C.C., Chang, T.Y., Zhang, T., Zhou, P., Song, B.L., Liu, W., Sun, S.C., Liu, X., Li, B.L., Xu, C., 2016. Potentiating the antitumour response of CD8(+) T cells by modulating cholesterol metabolism. Nature 531, 651-655.
Zammit, V.A., Buckett, L.K., Turnbull, A.V., Wure, H., Proven, A., 2008. Diacylglycerol acyltransferases: potential roles as pharmacological targets. Pharmacol. Ther. 118, 295-302.