Lytic reactions of drugs with lipid membranes

The involvement of drugs in direct chemical reactions with lipids may be linked to toxic effects in liver cell lines.


Introduction
Lipids are the key component of many materials, including liposomes used in technology applications and the membranes of cells and organelles in biological systems. In this article we examine whether organic molecules that partition into membranes, such as propranolol (1), can undergo direct chemical reactions with lipids (Scheme 1). Demonstration of such reactivity between membrane lipids and membraneassociated drugs in vivo would constitute a signicant addition to our understanding of membrane chemistry. Reactivity of this nature is likely to lead to biological effects that may account for the adverse activities of some drugs and the unusual pharmacokinetic proles of others. By contrast, exploiting this reactivity could offer opportunities for the purposeful design of new membrane-active compounds. The direct transfer of acyl groups from membrane lipids to suitably disposed molecules embedded within the membrane has a precedent, having been demonstrated for peptides in vitro. [1][2][3] In these cases, peptide lipidation results from the high affinity of the peptide for the membrane, combined with the appropriate positioning of reactive groups in the membrane interface. These aminolysis reactions involve nucleophilic attack on a lipid ester group by a suitably disposed nucleophile on the peptide, typically a lysine 3-amino group or the N-terminal amino group, and lead to the formation of a lysolipid alongside the lipidated peptide. 1 When the nucleophile is a serine hydroxyl group, the process is formally a transesterication. More recently, evidence has emerged to suggest that some membrane proteins may also be lipidated by direct transfer from the membrane. For example, the lipidation prole of the lens protein aquaporin-0 has been found to be highly complex. 4 Aquaporin-0 has two lipidation sites that do not correspond to known consensus sequences for lipidation enzymes and are partially lipidated with numerous different acyl groups, with typically up to eight identiable. The relative abundance of the acyl modications at each site correlates with the acyl ester content of the lipids in the plasma Scheme 1 Lipidation reactions involving propranolol. Key: HG, headgroup.
membrane leaet proximal to the site. Whilst these observations are not direct proof that this protein is lipidated by transfer from the membrane, they do give a strong indication that acyl transfer from the membrane is likely to be signicant in vivo. Fundamentally, aquaporin-0 is a moderately high molecular weight protein that is permanently membraneembedded in vivo, whilst peptides such as melittin that are lipidated in vitro are of medium molecular weight with modest membrane affinity. It is therefore of key interest to establish whether low molecular weight organic molecules, with concomitantly lower membrane affinity, are able to undergo the same process. We selected the b-blocker propranolol to probe this reactivity as it has both alcoholic and amine functionalities alongside well-characterised membrane binding characteristics. [5][6][7] Furthermore, synthetic versions of O-and N-acyl propranolol derivatives have been described as prodrugs, [8][9][10][11] providing precedents for understanding their properties in vitro and synthesising reference compounds for this work. In addition, propranolol is known to induce phospholipidosis, 12,13 an adverse activity associated with disorders in lysosomal phospholipid storage, characterised by the formation of lamellar bodies visible by microscopy of susceptible cells. [14][15][16][17][18][19][20] In principle therefore, propranolol may undergo both aminolysis and transesterication reactions, both of which lead to the formation of lysolipids, as shown in Scheme 1 for reaction with the acyl group at the sn-1 position of the lipid (although reaction can occur at either the sn-1 or sn-2 position). The lysolipid product would be formed as an equilibrium mixture of the 1and 2-acyl species. 1 All of the products could be reasonably expected to display signicant biological activity as a consequence of the presence of a fatty acyl chain. Lysolipids in particular, are capable of inducing signicant deleterious biological effects at levels as low as 1 mol%. 21 Lysolipids can also be formed by lipid hydrolysis, with concomitant formation of a fatty acid. Membrane-associated drugs have the potential to inuence the rate of this hydrolysis by either changing the bulk properties of the membrane, mediated by secondary effects on interfacial water activity, or by direct involvement in acid or base catalysis. 22,23 This article describes the propensity of a number of membrane-active drugs to undergo direct lipidation reactions with membrane lipids, or promote other lytic reactions of lipids.

Materials
Phospholipids and lysolipids, including E. coli Extract Polar (catalogue number 100600P) and Liver Polar Lipid Extract (Bovine, catalogue number 181108P) were purchased as powders from Avanti Polar Lipids (via Instruchemie B.V., The Netherlands). Propranolol was used as a racemic mixture.

Liposome preparation and sample set up
Liposomes were prepared by extrusion of lipid dispersions 10Â through polycarbonate lters (Whatman) with 100 nm tracketched pores, using a LIPEX thermobarrel extruder (Northern Lipids Inc., BC, Canada). All experiments were conducted at concentrations of 1.27 mM for lipids and (when present) 0.127 mM for membrane binding compounds. Using published partitioning data 5 these concentrations produce a bound propranolol to lipid ratio of about 1 : 30. Samples at pH 7.4 were buffered using bicarbonate at a NaCl concentration of 90 mM. Calibration curves for analyte concentration were generated by tting a logistic model to data obtained using authentic standards of lysolipids, propranolol and acyl propranolol derivatives at known concentrations.

Hep G2 culture and extraction
Hep G2 cells from ATCC were grown to conuence at 37 C, 5% CO 2 , and 95% humidity in Dulbecco's Modied Eagle Medium (DMEM) with 10% foetal bovine serum (Gibco/ThermoFisher). Cells (10 6 ) were incubated at 37 C, 5% CO 2 , and 95% humidity overnight to adhere. The medium was removed and replaced with either fresh medium (5 mL) for two controls, or medium containing 30 mM propranolol (5 mL). Following incubation for 72 h, the medium was removed from asks and replaced with phosphate buffered saline (5 mL). Cells were collected by centrifugation for 10 min at 1000 Â g, and decanted into a glass tube before extraction with CHCl 3 : MeOH (2 : 1; 3 mL). The CHCl 3 : MeOH solution was washed with H 2 O (0.6 mL), isolated, and the solvent removed in vacuo. Samples were resuspended in IPA : MeCN : H 2 O for liquid chromatographymass spectrometry (LC-MS) analysis.

Propranolol lipidation in liposomes composed of single lipids and binary mixtures
Our initial objective was to establish that incubation of propranolol (1) with liposomes could lead to the generation of lipidated propranolol products. We were able to obtain unambiguous evidence for propranolol lipidation through the use of liposomes with well-dened lipid compositions and comparison of the lipidated products with authentic standards prepared chemically. At 37 C and pH 7.4, using liposomes composed of either 1-palmitoyl-2-oleoyl-sn-glycero-3phosphocholine (POPC, Fig. 1a) or 1,2-dioleoyl-sn-glycero-3phosphocholine (DOPC, Fig S1b †), lipidated propranolol (2 and 3, Scheme 1) was detectable in both cases, alongside increased levels of lysolipid. Lysolipids were found in both samples as a mixture of 1-and 2-oleoyl-sn-glycero-3phosphocholine (OPC), in addition to 1-and 2-palmitoyl-snglycero-3-phosphocholine (PPC) in the case of POPC. The predominant products in both cases arose from a trans-esterication reaction involving the alcoholic group of propranolol to form the O-acyl product (2). O-and N-acyl modications to propranolol were distinguishable by the shorter retention times for O-acyl propranolol derivatives in comparison to their N-acyl counterparts ( Fig. 1a and S1 †) and by their fragmentation patterns in tandem mass spectra produced by collision-induced decay (CID). The CID MSMS spectra of Oacyl propranolol (2) yielded a fragment of m/z 260, with species of m/z 155 and m/z 183 having the highest relative abundance ( Fig. 1b and S1-S3, Table S1 †). In contrast, N-acyl propranolol homologues (3) were particularly notable for the high relative abundance of fragments with an intact acyl chain, at m/z 338 and m/z 380 for N-oleoyl propranolol, alongside a high relative abundance for the fragment at m/z 260 corresponding to the loss of the acyl chain (Fig. 1c, S4, Tables S2 and S3 †).
Aer instrument calibration, using data from authentic standards to convert raw ion intensities to concentrations (Fig. S5, Table S4 †), the concentrations of both total lipidated propranolol and lysolipid are seen to increase steadily over the rst 48 h following mixing (Fig. 2). At these early time points, the predominant lipidated product is the O-acyl ester (Fig. S6 †). The concentration of the N-acyl amide exhibits a small latency period before concentrations begin to increase, which indicates that the N-acyl product may be formed by O to N migration of the acyl group rather than direct aminolysis of the lipid. Such a migration has been described in the literature for propranolol ester derivatives in solution. 8,10 The data are reproducible at time points up to 48 h, but show greater sample-to-sample variability over longer periods, particularly for the concentration of lysolipid. Aer 72 h the total content of lipidated propranolol decreases, the reasons for which are not currently understood, although hydrolysis or insolubility of one or both of the acyl derivatives are likely causes. Consequently, only data obtained within the rst 48 h have been used for comparative analyses. As indicated in Fig. 1a, acyl transfer to propranolol from the lipid is expected to lead to the formation of a stoichiometric equivalent of lysolipid, which should lead to identical concentration changes for acyl propranolol and lysolipid, assuming that there are no competing processes. For DOPC, a linear t to the data over the rst 48 h gave a rate of acyl propranolol formation of 8.7 (AE2.8) Â 10 À5 mM h À1 , which was 0.6 (AE0.3) times the rate of lysolipid formation in the same period. This indicates that either propranolol binding changes bilayer properties to facilitate hydrolysis, or that propranolol is able to catalyse lipid hydrolysis in addition to undergoing transesterication reactions. The lipidated propranolol products themselves may also promote lipid hydrolysis, for example by inducing changes to membrane properties that modify interfacial water activity, or by direct involvement in acid-base catalysis.

Propranolol lipidation in lipid extracts
Having demonstrated lipidation of propranolol in vitro, it was of fundamental interest to probe whether this reaction also occurred in cellulo. In order to address this question propranolol was rst incubated with liposomes composed of commercial lipid mixtures extracted from E. coli or liver cells (Fig. S7 †) so that the likelihood of lipidation occurring in complex membranes could be assessed and the scale of the analytical challenge understood, given the likely complexity of the products.
In both of these experiments, the ability to identify lipidation products was conrmed, with each extract yielding a series of acyl-modied derivatives (Fig. 3a, S8, Tables 1 and S5 †) with   (Tables S6 and S7 †). Some products anticipated on the basis of the relative fatty acid abundances were not observed, including modication with C22:5. A number of reasons could account for the failure to observe these products, including their presence at concentrations below the detection limit, differences in the fatty acid prole of the commercial mixture from published data, peak overlap leading to ion suppression, and their presence in a form that is not reactive.

Propranolol lipidation in cellulo
With knowledge that acylated propranolol derivatives could be identied in complex mixtures of natural lipids, propranolol lipidation in cellulo was probed. The Hep G2 cell line was selected for this work in order to build on the data described above for liver cell extracts and because propranolol is known to induce phospholipidosis (PLD) in Hep G2 cells with an EC 50 between 12.6 and 16 mM. 13 In our hands, incubation of Hep G2  a Proportion (mol%) of each fatty acyl chain found in total bovine liver phospholipids 26 or in lipids isolated from Hep G2 cells cultured in the absence of propranolol in a medium containing 10% foetal bovine serum. 27   cells with 30 mM propranolol induced PLD as expected (Fig. S9 †). By performing tests with POPC liposomes beforehand (Fig. S10 †), the use of a chloroform/methanol mixture for the chemical extraction of Hep G2 cells was veried as the best approach for recovering both O-and N-acyl propranolol derivatives. When compared with the starting liposome mixture, a partial reduction in the relative intensity of O-oleoyl propranolol was found in the extracts, indicating a reduced extraction efficiency for propranolol modied with unsaturated fatty acyl groups. However, in the worst-case scenario all species could at least be partially recovered. Extraction of Hep G2 cells by this approach, followed by LC-MS analysis, identied a range of lipids (Fig. S11, Table S8 †) alongside 14 different lipidated propranolol species (Fig. 3b, Table 1). The predominant fatty acyl modications to propranolol observed in the chromatograms, palmitoyl and oleoyl (Table 1), were those anticipated on the basis of published fatty acid proles for Hep G2 cells cultured in the same medium, considered alongside the differences in extraction efficiency discussed above. The identities of some of the lipidated propranolol species were further conrmed by LC-MSMS analysis of the Hep G2 extract, with the palmitoyl-and oleoyl-modied propranolol parent species targeted for fragmentation (Fig. 4). As evidenced by the relatively early retention times and a high relative abundance of product ions with m/z 155 and m/z 183, the predominant products were O-acyl propranolol derivatives. Only trace quantities of the N-acyl species could be detected (Fig. 4, peaks iii and vii). The CID target mass window was sufficiently wide to conrm the presence of other related species in the mixture, including O-palmitoleoyl (Fig. 4, peak i), O-linoleoyl (peak iv) and O-stearoyl propranolol (peak vi).
In contrast to the experiments with extracted lipid mixtures, the reactivity of propranolol in Hep G2 membranes might be expected to produce a response that is inuenced by the presence of lipid homeostasis in the cell. Indeed signicant differences were found in the intensities of some lipids when comparing cells cultured with or without propranolol (Table  S9 †). These included decreases in the levels of some triglycerides and phosphatidylserine, and striking increases in the levels of ether-linked lipids, which would have the effect of reducing the ester content of the membrane that is transferrable in intrinsic lipidation reactions.

The biological effects of drug lipidation
As we had evidence that propranolol reacts with membrane lipids both in vitro and in vivo, leading to the generation of lysolipids and a lipidated drug, it was of fundamental interest to establish whether this reactivity had the potential to induce detectable physiological effects. Previous work has established that some classes of drug linked with phospholipidosis are able to promote lipid hydrolysis in model systems in vitro, potentially by acting as phase transfer catalysts. 22,23 We therefore explored whether propranolol and other compounds known to induce phospholipidosis (Fig. 5) were subject to lipidation in a model  DOPC system or otherwise inuence the levels of lysolipid, and examine the correlation between these activities and phospholipidosis activity. Interestingly, only propranolol underwent any lipidation reactions in DOPC membranes. However, all of the compounds with appreciable phospholipidosis activity yielded detectable increases in the initial rate of lipid hydrolysis over 24 h (Fig. 6a) compared to a control in the same conditions without compound. It may be the case that the amino group of propranolol is positioned in such a manner that it can promote the transesterication reactions of propranolol, but in the absence of proximal acyl group acceptor, the amino groups of the other drugs favour reactions involving interfacial water. Although there is a correlation between phospholipidosis activity and hydrolysis rates, with for example uoxetine producing the fastest rate of lipid hydrolysis and having the lowest EC 50 (Fig. 6b), the relationship is not simple, as indicated by the non-linearity in Fig. 6b.
It might be expected that both the lipidated drug and the lysolipid have the potential to disrupt lipid membranes as a consequence of their shape, which favours detergent-like activity. Both N-oleoyl and N-palmitoyl propranolol were found to have measurable micelle-forming behavior, with CMCs of 9 mM and 10 mM respectively (Fig. S12 †). Addition of both of these N-acyl compounds at a concentration of 1 mol% to POPC liposomes loaded with the markers 8-aminonaphthalene-1,3,6trisulfonic acid (ANTS, 12.5 mM) and p-xylene-bis-pyridinium bromide (DPX, 45 mM) resulted in the loss of membrane integrity with concomitant increase in uorescence (Fig. S13 †). Similar results were obtained for monooleoyl and monopalmitoyl PC. Parallel analyses using O-acyl species proved to be problematic due to the rapid hydrolysis of these species in the absence of membranes.

Conclusions
This work has demonstrated the fundamental point that low molecular weight organic molecules are capable of generating lytic chemical processes in lipid membranes, including direct intermolecular reactions such as transesterication, or the promotion of other reactions such as hydrolysis. However, as evidenced by the reactivity of propranolol in comparison to drugs such as uoxetine and sertraline, the selectivity for lipidation vs. hydrolysis is complex and most likely linked to a number of factors, including the molecular disposition in the membrane interface, the distribution prole (log D) and the pK a values of any ionisable groups. Furthermore, it is clear that the activities observed in vitro are also likely to occur in vivo. This has been demonstrated directly by the observation of the products of lipidation reactions of propranolol in Hep G2 cells and indirectly by correlations between the hydrolysis promoting activity of drugs in vitro and their EC 50 for phospholipidosis. In the case of propranolol it is also striking that the O-acyl esters, which normally rapidly rearrange to the N-acyl amide counterpart in vitro, exhibit increased stability in the membrane environment. This observation highlights the fact that there is much concerning the chemistry that occurs in the membrane interface that remains to be understood.

Conflicts of interest
There are no conicts to declare.