Inhibition of lysosomal phospholipase A2 predicts drug-induced phospholipidosis

Phospholipidosis, the excessive accumulation of phospholipids within lysosomes, is a pathological response observed following exposure to many drugs across multiple therapeutic groups. A clear mechanistic understanding of the causes and implications of this form of drug toxicity has remained elusive. We previously reported the discovery and characterization of a lysosome-specific phospholipase A2 (PLA2G15) and later reported that amiodarone, a known cause of drug-induced phospholipidosis, inhibits this enzyme. Here, we assayed a library of 163 drugs for inhibition of PLA2G15 to determine whether this phospholipase was the cellular target for therapeutics other than amiodarone that cause phospholipidosis. We observed that 144 compounds inhibited PLA2G15 activity. Thirty-six compounds not previously reported to cause phospholipidosis inhibited PLA2G15 with IC50 values less than 1 mM and were confirmed to cause phospholipidosis in an in vitro assay. Within this group, fosinopril was the most potent inhibitor (IC50 0.18 μM). Additional characterization of the inhibition of PLA2G15 by fosinopril was consistent with interference of PLA2G15 binding to liposomes. PLA2G15 inhibition was more accurate in predicting phospholipidosis compared with in silico models based on pKa and ClogP, measures of protonation, and transport-independent distribution in the lysosome, respectively. In summary, PLA2G15 is a primary target for cationic amphiphilic drugs that cause phospholipidosis, and PLA2G15 inhibition by cationic amphiphilic compounds provides a potentially robust screening platform for potential toxicity during drug development.

Abstract Phospholipidosis, the excessive accumulation of phospholipids within lysosomes, is a pathological response observed following exposure to many drugs across multiple therapeutic groups. A clear mechanistic understanding of the causes and implications of this form of drug toxicity has remained elusive. We previously reported the discovery and characterization of a lysosome-specific phospholipase A2 (PLA2G15) and later reported that amiodarone, a known cause of drug-induced phospholipidosis, inhibits this enzyme. Here, we assayed a library of 163 drugs for inhibition of PLA2G15 to determine whether this phospholipase was the cellular target for therapeutics other than amiodarone that cause phospholipidosis. We observed that 144 compounds inhibited PLA2G15 activity. Thirtysix compounds not previously reported to cause phospholipidosis inhibited PLA2G15 with IC 50 values less than 1 mM and were confirmed to cause phospholipidosis in an in vitro assay. Within this group, fosinopril was the most potent inhibitor (IC 50 0.18 μM). Additional characterization of the inhibition of PLA2G15 by fosinopril was consistent with interference of PLA2G15 binding to liposomes. PLA2G15 inhibition was more accurate in predicting phospholipidosis compared with in silico models based on pKa and ClogP, measures of protonation, and transport-independent distribution in the lysosome, respectively.
In summary, PLA2G15 is a primary target for cationic amphiphilic drugs that cause phospholipidosis, and PLA2G15 inhibition by cationic amphiphilic compounds provides a potentially robust screening platform for potential toxicity during drug development.
Supplementary key words Acyltransferase • 1-O-acylceramide • lysosome • phospholipase A2 group XV • drug-induced phospholipidosis • drug toxicity • cationic amphiphilic drugs • drug development • high-throughput screening • amiodarone Phospholipidosis is the excess storage of phospholipids within lysosomes. Drug-induced phospholipidosis (DIP), in distinction to inherited forms of lysosomal phospholipid accumulation such as those associated with disorders such as Niemann-Pick C disease, represents an acquired lysosomal disorder (1,2). DIP most often involves the lung, liver, or kidney where it is associated with pulmonary fibrosis, hepatic steatosis or steatohepatitis, and acute or chronic kidney injury, respectively. Phospholipidosis often, but not always, results from exposure to basic cationic amphiphilic drugs (CADs). DIP is measured experimentally by use of in vitro or in vivo assays and is often observed in clinical settings. It is among the most common forms of drug toxicity as it is associated with exposure to more than 50 FDA-approved agents. When DIP is detected in preclinical screening studies, an otherwise promising compound may be abandoned. If DIP is found in patients under treatment with a specific drug, then the therapeutic is often discontinued. Research on DIP has been dominated by three overarching questions. First, what are the mechanisms responsible for DIP? Second, what chemical properties of a candidate compound can be used to predict phospholipidosis and used as a guide for further development? Third, what significant shortand long-term toxicities are the specific consequences of DIP?
With regard to the first question, several mechanisms have been proposed as the basis DIP. These include the stimulation of phospholipid synthesis (3), the direct binding of CADs to lysosomal phospholipases with inhibition of these enzymes by competitive or allosteric mechanisms (4), the inhibition of lysosomal trafficking to lysosomes (5), the displacement of phospholipases from the lysosomal membrane with secondary degradation by lysosomal proteases (6), and the binding of CADs to phospholipids with prevention of their degradation (7). Lysosomal phospholipase A1, A2, and C activities have been previously associated with DIP. However, to date only three phospholipases are known to be lysosome-based. They include acid sphingomyelinase (8), phospholipase D3 (9), and lysosomal phospholipase A2 (10).
With regard to the second question, efforts to predict DIP have generally followed two strategies. The first approach has employed analyses in which the physical properties of drugs are correlated with empirically observed phospholipidosis (11)(12)(13)(14). The second strategy has used the development of novel in vitro assays that can be applied to the screening of individual drug candidates or chemical libraries to predict phospholipidosis potential. These assays include those that detect lipid accumulation in cell lines or that specifically measure lysosome associated lipid biomarkers such as bis(monoacylglycerol)phosphate (15) or gene expression profiling (16). The assessment of these various in silico and in vitro strategies is limited by the absence of proof of a mechanism responsible for DIP.
With regard to the third question, a determination of the pathological significance of phospholipidosis has been limited by the lack of identification and characterization of a specific target or targets of compounds that cause DIP. Identifying the cellular target or targets responsible for DIP as distinguished from toxicities resulting from separate off-target effects would represent a significant step in understanding and managing this form of drug toxicity.
Our group identified an enzyme with 1-O-acyl-ceramide synthase activity and subsequently characterized a purified enzyme as lysosomal phospholipase A2 (LPLA 2 ), now designated PLA2G15 (17)(18)(19). LPLA 2 has an acidic pH optimum and colocalizes with lysosomes and late endosomes. Loss of function of LPLA 2 in mice results in alveolar macrophage foam cell formation and surfactant accumulation, a phenotype similar to that observed with amiodarone-associated phospholipidosis (20). In subsequent work we reported that amiodarone is a potent inhibitor of LPLA 2 , but does so by inhibition of electrostatic charge interactions between the hydrolase and anionic phospholipids (21). This mechanism of action was further substantiated by our determination of the crystal structure of LPLA 2 and the identification of critical residues in the lipid membrane-binding domain (22).
Based on these studies we considered whether the inhibition of LPLA 2 by cationic amphiphilic compounds is a more general mechanism for DIP. We assayed two libraries of small molecules for their ability to inhibit LPLA 2 and correlate this inhibition with physical properties of these compounds used by others as the basis for predictive models of phospholipidosis. The first library consisted of drugs known to cause phospholipidosis in either in vitro or in vivo studies based on published reports. The second library consisted of compounds for which DIP has not been previously reported. We observed that inhibition of LPLA 2 strongly correlates with drugs reported to cause phospholipidosis and have identified drugs that have not previously known to cause phospholipidosis, not all of which are cationic amphiphiles.

Transacylase activity of LPLA 2
The LPLA 2 activity assay is based on the following principles (23). LPLA 2 is uniquely characterized as having an acidic pH optimum and as a transacylase recognizing short-chain lipophilic alcohols as acceptors. Based on these properties, short-chain 1-O-acyl-ceramides are unique products of this reaction. Because LPLA 2 binds preferentially to negatively charged liposomes, sulfatide was included in the liposomes but is not itself a substrate and does not function as a cofactor for lysosomal hydrolases. The transacylase reaction is based on the unique property of LPLA 2 to transfer an acyl group from the sn-2 or sn-1 position of a glycerophospholipid to Nacetyl-sphingosine (NAS) forming 1-O-acyl-N-acetylsphingoine (1-O-acyl-NAS) (18,22,24). 1-O-acyl-NAS is not known to be a product of any other enzyme. The reaction mixture included 50 mM sodium citrate buffer (pH 4.5), 10 μg/ml bovine serum albumin, and liposomes consisting of 38 μM N-acetyl-sphingosine, 127 μM DOPC, 12.7 μM sulfatide, and test compound in a total volume of 0.5 ml. The test compounds were dissolved in DMSO. The final DMSO concentration in the reaction mixture was 0.125%. The reaction was initiated by the addition of recombinant LPLA 2 protein (30 ng) and carried out at 37˚C for 10 min. The reaction was terminated by the addition of 3 ml chloroform/methanol (2/ 1, v/v), followed by 0.3 ml of 9% (w/v) NaCl. After centrifugation for 7 min at 1800 × g, the resulting lower layer was transferred to new tube and dried under stream of nitrogen gas. The dried lipid was dissolved in 40 μl of chloroform/ methanol (2/1, v/v) and applied to HPTLC plates. HPTLC plates were run in chloroform/acetic acid (9/1, v/v). The plates were dried and soaked in 8% (w/v) CuSO 4 .5H 2 O, 6.8% (v/v) H 3 PO 4 , and 32% (v/v) methanol and then charred for 15 min in an oven at 150˚C. Scanned plates were analyzed by NIH ImageJ 1.651j8 (National Institutes of Health). LPLA 2 esterase assay pNPB was used to directly measure the activity of LPLA 2 . pNPB is a water-soluble substrate that can directly access the catalytic site in the absence of liposomes (25). A reaction mixture of pNPB (0.2 mM) and cationic amphiphilic compounds at varying concentrations in sodium citrate buffer (pH 4.5) was prepared and prewarmed to 37 • C for 5 min in a total volume of 500 μl. The reaction was initiated by the addition of recombinant LPLA 2 (5 μg). At predetermined times, 120 μl of the reaction mixture was transferred to a tube containing 120 μl of 0.2 M NaHCO 3 and kept on ice. The cold reaction product was subsequently warmed to 37 • C, and the absorbance of the reaction product, p-nitrophenoxide, was measured at 400 nm with a Beckman Du-640 spectrophotometer.

Liposome LPLA 2 cosedimentation assay
Liposomes consisting of DOPC and sulfatide (10:1 M ratio, 127 μM total lipid) were incubated with 5 μg of LPLA 2 in 500 μl 50 mM sodium citrate, at pH 4.5 for 30 min on ice. The reaction mixture was then centrifuged for 1 h at 150,000 g at 4 • C. The resulting precipitate was rinsed with cold 50 mM sodium citrate pH 4.5 and dissolved with 40 μl of SDS-PAGE sample buffer. The sample was separated by using 10% SDS-PAGE. After electrophoresis, LPLA 2 was detected with Coomassie brilliant blue. Band quantification was performed with ImageJ software I1.651j8 (25).

LPLA 2 thermal stability measurement
A thermal stability assay was employed to determine the melting point (T m ) of LPLA 2 (26). An incubation mixture consisting of 2.5 μl of 8x SYPRO Orange, 1 μg of LPLA 2 in 50 mM Na citrate at pH 4.5, and ddH 2 O in a final volume of 20 μl was added to wells of a 48-well thin-wall PCR plate. The plates were sealed with Optical-Quality Sealing Tape (Bio-Rad) and heated in a Real-Time PCR Detection System Life Technology (Thermo Fisher, Ann Arbor, MI) from 20 to 90 • C in steps of 0.2 • C. T m values were calculated as the inflection point of the melting curve using the instrument software.

Screening phospholipidosis assay
The assay was modified from one reported previously (27). MDCK cells were seeded in 100 μl culture medium at cell density 3,000 cells per well in 96-well black-walled clear bottom Greiner micro plates (Sigma-Aldrich) and were allowed to adhere overnight. Cell culture medium was replaced with phospholipidosis staining solution (1:1,000 dilution) of Lip-idTOX Red Phospholipidosis detection reagent (Invitrogen), and simultaneously with different concentrations of fosinopril or amiodarone in total volume of 100 μl. Compounds were prepared as stock solutions at 200-fold higher concentration than the desired top concentration (solvent concentration maintained at 0.5%). Compound treatment was performed for 24 h with 5% CO 2 at 37 • C. Then the culture medium was removed and cells were fixed with 100 μl, fixation solution consisting of 4% formaldehyde in phosphate buffered saline (PBS). After washing, cells were incubated with 1 drop of NucBlue Live (NBL) for 20 min. Cells were washed three times with PBS, and fluorescence image acquisition was performed using the Molecular Devices (San Jose, CA) spectrophotometer. Cell nuclei fluorescence was detected using a 410-480 nm emission filter, red phospholipidosis detection was performed using 549-615 nm emission filter.

Image acquisition and processing
Ninety six-well plates were visualized under a Leica DM IRB microscope and images acquired with an Olympus DP70 camera via Olympus DP Manager software. All images were identically adjusted in GNU Image Manipulation Program to improve background and overall image clarity postacquisition.

LipidTOX red particle quantification
Images were quantified utilizing ImageJ as follows. Images were initially processed with the Subtract Background feature with a rolling ball radius of 50 pixels. Following conversion to 8 bit, images were subjected to Auto Local Threshold processing using the Bernsen algorithm with a radius of 15. Particles were subsequently quantified and analyzed utilizing the Analyze Particle feature. A total of six 10x fields (2 per triplicate) were quantified with an average of over 4,000 cells per field.

Statistical analysis
Data from at least three independent experiments were analyzed with a paired t test in GraphPad Prism 7 and expressed as mean ± SD. The differences between control and treated samples were considered statistically significant at P < 0.05.

RESULTS
A library of 163 compounds was assembled and assayed for inhibition of LPLA 2 . One hundred and nine compounds were identified via literature review as causing phospholipidosis based on either in vitro or in vivo assays (Table 1). In the latter case, the animal species employed is indicated. These compounds were chosen represent a wide spectrum of therapeutic indications, having a range of pKa and ClogP that fell within and outside of values commonly associated with DIP and in which the lysosomal pathology is observed across a range of organs. Most, but not all, of the compounds are cationic amphiphiles, and several are central nervous system penetrant. A second set of 54 compounds was assayed representing drugs for which no reports of phospholipidosis were found but which were representative of a similar spectrum of chemical properties ( Table 2). Included in this set were metabolites chosen as negative controls (glucose, leucine, and uridine). The primary clinical indications listed in these tables are consistent with a wide range of cellular targets for these compounds.
Two primary physical properties of a drug have been used to predict whether a compound may be lysosomotropic. These are the ClogP and pKa (basic). ClogP, a measure of partitioning between octanol and water, predictive of transport independent distribution across cell membranes. pKa (basic) is a determinant of the protonation of an amine at lysosomal pH. ClogP and pKa (basic) were employed by Ploemen and colleagues to generate an in silico model that is predictive of phospholipidosis (Table 3) (115). In a subsequent paper, a modification was proposed to improve the positive and negative predictive value of the model (11). In contrast, the assay used for the measurement of LPLA 2 activity is cell-free and thus not dependent on the ability of a particular compound to enter a target cell and distribute into late endosomes or lysosomes. A comparison between the physical properties and    Lysosomal phospholipase A2 and phospholipidosis   Lysosomal phospholipase A2 and phospholipidosis     The generic name, International Union of Pure and Applied Chemistry (IUPAC) designation, chemical abstracts registry (CAS) number, and clinical indication are provided. Chemical properties, including ClogP and pKa (basic) were obtained from the Pubchem and chEMBL databases of bioactive molecules and used to calculate the Ploemen value for each compound. By convention, negative pKa values were assigned a value of 0 for calculation of the Ploemen number and would be predicted negative based on pKa values of less than 8 or 6 for the Ploemen and modified Ploemen models respectively ( Table 3). The species for in vivo data are designated as human (H), dog (D), rat (R), mouse (M), and hamster (Ha). LPLA 2 IC 50 denotes the concentration at which 50% of the LPLA 2 dependent 1-O-acyl N-acetylsphingosine synthase activity is observed.  inhibitory effects on LPLA 2 of the compounds tested was therefore assessed as independent variables. Under these assay conditions, 112 compounds from the entire library of known phospholipidotic and control compounds inhibited LPLA 2 acyl transferase activity at IC 50 values less than 250 μM. Twenty-eight compounds inhibited with IC 50 s greater than 250 μM, and 19 compounds showed no inhibition (Fig. 1). Surprisingly, 35 compounds not previously reported to cause phospholipidosis inhibited LPLA 2 with 22 compounds inhibiting at IC 50 s of 250 μM or less (black circles). The measured IC 50 s were continuous between 3.8 μM and 2 mM. The compounds assayed did not segregate based on therapeutic use or whether phospholipidosis had been reported as a result of in vitro (red circles), in vivo (blue circles) or both in vitro and in vivo studies (yellow circles). The most potent inhibitor among this group was fosinopril, an angiotensinconverting enzyme inhibitor not previously reported to cause phospholipidosis.
The library of compounds previously reported to cause phospholipidosis and assayed for LPLA 2 inhibition were plotted based on their calculated ClogP and pKa basic values (Fig. 2). Following the convention employed in the Ploemen study, negative pKa basic values were assigned a value of 0 for purposes of calculating a Ploemen value. Eight compounds inhibited LPLA 2 at concentrations greater than 500 μM. Three compounds (proparacaine, vinblastine, and clenbuterol) inhibited LPLA 2 at millimolar concentrations, and two compounds, spiperone and chloroquine, inhibited LPLA 2 with IC 50 s slightly greater than 500 μM. Four compounds (betaxolol, methapyrilene, ropinirole, and sulpiride) had no inhibitory activity against LPLA 2 and could thus be considered to be true false-negatives for predicting phospholipidosis.
Importantly, 23 compounds in the library reported to cause phospholipidosis did not meet either the Ploemen or modified Ploemen criteria. Of these 23 compounds, 17 inhibited LPLA 2 , 9 of these compounds having IC 50 values less than 50 μM. Thus in this limited library of compounds previously reported to cause phospholipidosis, LPLA 2 inhibition was observed for almost threequarters of the drugs that would have been considered false-negatives by the Ploemen or modified Ploemen criteria.
The second library consisting of compounds not reported to cause phospholipidosis was similarly graphed (Fig. 3). Thirty of 55 compounds inhibited LPLA 2 at IC 50 values less than 250 μM, and 15 of these compounds inhibited LPLA 2 at less than 50 μM. Only half or 15 of the 30 inhibitors would have been identified by the modified Ploemen criteria. An in vitro assay using the LipidTOX red detection reagent was used to determine whether exposure of cells to these compounds was associated with lysosomal phospholipid accumulation (Table 4). This assay was validated using fosinopril, the  most potent inhibitor of LPLA 2 activity (supplemental Figs. S1 and S2). In addition, none of the 163 compounds shifted the melting temperature of LPLA 2 more than 2 • C when assayed for thermal stability, consistent with a lack of direct binding of any compound to the phospholipase (supplemental Table 1). In contrast the fluorophosphonate inhibitors, isopropyl dodec-11-enyl fluorophosphonate and methyl arachidonyl fluorophosphonate, covalently bind to the catalytic serine of LPLA 2 and increase the melting temperature by 10 and 12 • C, respectively (22). Fosinopril is an angiotensin-converting enzyme inhibitor not previously known to cause phospholipidosis. Fosinopril inhibited LPLA 2 activity at an IC 50 of 180 nM, considerably lower than that observed for any other compound ( Fig. 4A and B). The basis for LPLA 2 inhibition by this compound was therefore studied in greater detail. We had previously shown that the inhibition of electrostatic binding of liposomes to LPLA 2 could be measured by loss of cosedimentation. Fosinopril partially inhibited the cosedimentation of liposomes and recombinant LPLA 2 when centrifugation was performed at 150,000 g (Fig. 4C). As was previously reported with amiodarone (25), no inhibition of the soluble esterase activity of LPLA 2 was observed in the presence of fosinopril. The transacylase activity of LPLA 2 toward p-NPB as substrate was first confirmed by the formation of 1-O-butanoyl-N-acetylsphnigosine when present in the fully constituted LPLA 2 assay (Fig. 4D). The formation of 1-O-butanoyl-N-acetylsphingosine was also observed when LPLA 2 activity was assayed only in the presence of LPLA 2 , N-acetylsphingosine, and p-NPB as a monodispersion (Fig. 4E) consistent with accessibility of the substrate and acceptor within the catalytic domain of LPLA 2 . However, in the presence of 250 nM fosinopril, no inhibition of 1-O-butanoyl-N-acetylsphingosine formation was observed. This is consistent with the absence of direct inhibition of LPLA 2 by this drug.
Compared with untreated controls, 325 nM fosinopril significantly increased the number of LipidTOX Red particles as assessed by fluorescence microscopy (supplemental Fig. S1). Quantification of particles revealed an approximate 20-fold increase in particle number in fosinopril-treated cells compared with controls (supplemental Fig. S2), which was similar to that observed with amiodarone at the same concentration. Concomitant increases in percent area and mean fluorescence intensity (MFI) were also documented.
Fosinopril is a prodrug for the active metabolite fosinoprilat. The measured IC 50 value for fosinoprilat (5.8 μM) was more than 50 times greater than that observed for fosinopril (0.18 μM). However, fosinoprilat also generated a positive signal in the LipidTOX assay (Table 4). Thus metabolism of fosinopril to its active metabolite cannot explain the previously reported absence of phospholipidosis in this case.

DISCUSSION
There are three important findings in this study. First, inhibition of LPLA 2 , as measured by a decrease in 1-O-acylceramide formation in a cell-free assay, is observed in the presence of most of 110 drugs studied previously reported to cause phospholipidosis. None of the drugs assayed shifted the melting temperature of LPLA 2 consistent with the absence of the direct binding of any compound to LPLA 2 . Thus the inhibition of LPLA 2 activity by these compounds occurs within a concentration range and likely by a mechanism similar to amiodarone, namely interference with the electrostatic charge interaction between cationic residues in the lipid-binding domain of LPLA 2 and anionic phospholipid head groups. This mechanism, however, would not explain the inhibition of LPLA 2 by compounds that cause phospholipidosis but lack a functional group that would be protonated at lysosomal pH including fosinopril, mitotane, and mannitol.
Second, the LPLA 2 inhibition assay identified several CADs known to cause phospholipidosis but that are not predicted to do so by use of in silico models based on the pKa and ClogP of CADs. The measurement of LPLA 2 inhibition as a stand-alone assay for prediction of DIP is associated with a greater sensitivity and accuracy than models based on ClogP and pKa alone but slightly less than when these models are combined with an in vitro assay (28). Specifically, LPLA 2 inhibition with an observed IC 50 < 500 μM is 86% accurate in predicting phospholipidosis compared with the 58 and 79% accuracies of the Ploemen and modified Ploemen models, respectively. The accuracy in predicting phospholipidosis is greater than 90% for any observed inhibition of LPLA 2 (Table 5). Individual drugs that cause phospholipidosis may do so synergistically (116), and such drugs may achieve concentrations within the lysosome that are up to 50,000-fold greater than that measured extracellularly (117). It is therefore possible that compounds that inhibit LPLA 2 may do so at concentrations significantly greater than those associated with their therapeutic activity.
Third, LPLA 2 inhibition may identify chemical entities currently approved by regulatory agencies that cause phospholipidosis but not previously identified as such. This is exemplified in the current study by the potent inhibition of LPLA 2 by fosinopril and its active metabolite fosinoprilat and by validation of their phospholipidotic potential in the LipidTOX Red phospholipidosis assay. The IC 50 value for fosinopril is significantly lower than that of the other drugs assayed in this study. Fosinopril is unique among the larger class of ACE inhibitors in that it contains a phosphinic-acidcontaining ester that serves as the binding group as opposed to the more common carboxyl or sulfhydryl  Table 1 are graphed in relation to pKa (basic) and ClogP. The exclusion limits of the Ploemen and modified Ploemen models are delineated by the red lines. The IC 50 s for LPLA 2 -dependent 1-O-acyl N-acetylsphingosine synthase activity are indicated as follows: greater than 1 mM (red circles), less than 100 μM (green circles), greater than 100 μM and less than 1 μM (blue circles).  Table 2 are graphed in relation to pKa (basic) and ClogP. The exclusion limits of the Ploemen and modified Ploemen models are delineated by the red lines. LPLA 2 IC 50 s for LPLA 2 -dependent 1-O-acyl N-acetylsphingosine synthase activity are indicated as follows: greater than 1 mM (red circles), less than 100 μM (green circles), greater than 100 μM and less than 1 μM blue circles.
functions that characterize other ACE inhibitors (118). It is also among the most lipophilic of this class of drugs. Like amiodarone, the mechanism of inhibition by fosinopril appears to occur by interference of binding between LPLA 2 and liposomes as supported by observed inhibition of cosedimentation of LPLA 2 and liposomes in the presence of fosinopril. However, fosinopril is an amide and as a weak base not protonated at lysosomal pH unlike cationic amphiphilic drugs. Thus a different mechanism of inhibition that is distinct from amiodarone is likely in this case.
While there is general agreement that phospholipidosis results from the lysosomal accumulation of CADs, there is less agreement regarding the cause. LPLA 2 is a good candidate for a cellular target by drugs that cause phospholipidosis. Although LPLA 2 was first characterized as a phospholipase with transacylase activity toward short-chain ceramide acceptors (17), it was later recognized to be a phospholipase A2 with an acidic pH optimum (18). The further characterization of the enzymatic activity revealed broad substrate specificity to several glycerophospholipids including phosphatidylcholine, phosphatidylethanolamine, phosphatidylserine, and phosphatidylglycerol. Subsequent work characterized LPLA 2 as having both PLA 1 and PLA 2 activity (119).
The earliest observed phenotype of a transgenic mouse knocked out for LPLA 2 was the presence of alveolar macrophages with a foam cell appearance. Lipid analyses of both the macrophages and bronchoalveolar lavage fluid demonstrated increased levels of glycerophospholipids that were substrates for LPLA 2 (20). The pulmonary toxicity associated with amiodarone is consistent with several of these functions. In its classic form, amiodarone toxicity is manifest as the accumulation of lipid-laden alveolar macrophages. The ultrastructure of these foam cells is characterized by the presence of lamellar bodies within lysosomes. Because the knockout mouse phenotype bore a strong resemblance to that seen with pulmonary amiodarone toxicity, the possible inhibition of LPLA 2 by amiodarone was further studied (21). We observed at that time that amiodarone was not a direct inhibitor of LPLA 2 , but appeared to block the electrostatic interaction between liposomes and enzyme. This mechanism was further supported by the loss of activity in the presence of buffers of higher ionic content.
More recently, we determined a structure of LPLA 2 by X-ray diffraction (22). The presence of the catalytic triad and the disulfide bond previously characterized was confirmed (120). Two tracks could accommodate the phospholipid head groups of a broad range of substrates. In the current model, sn-1 and sn-2 fatty acyl groups of these phospholipids can be oriented in track A within the catalytic domain and be recognized as the scissile fatty acyl group. This model is supported by the formation of 1-O-acyl-ceramides that are products of either sn-1 or sn-2 acyl groups on phospholipid substrates (121). The structural studies identified a distinct lipid binding domain and four cationic residues within the domain that are required for liposome binding. The observation that each of these residues was necessary for LPLA 2 activity lent further support for the  The reaction mixture was then centrifuged for 1 h at 150,000 g at 4 • C. The resulting precipitate was rinsed with cold 50 mM sodium citrate pH 4.5 and dissolved with 40 μl of SDS-PAGE sample buffer. The sample was separated by using 10% SDS-PAGE. Following electrophoresis, LPLA 2 was detected with Simply Blue. Band quantification was performed with the Image J software I1.651j8. D: LPLA 2 transacylase activity against comparing DOPC to p-NPB as substrates. Liposomes containing DOPC-sulfatide (10:1 M ratio) were incubated with recombinant LPLA 2 (30 ng/ml) with or without p-NPB (200 μM) in the presence or absence of 10 μM NAS at 37 degrees C in 500 μl Na-citrate buffer (50 mM, pH 4.5). E: LPLA 2 transacylation activity toward p-NPB comparing liposomes to a monodispersed substrate. Fosinopril (250 nM) was present in lanes 5 and 6. The reactions as detailed in panels E and F were terminated by the addition of 3 ml chloroform/methanol (2/1, v/v), followed by 0.3 ml of 9% (w/v) NaCl. After centrifugation for 7 min at 1,800 g, the resulting lower layer was transferred to new tube and dried under stream of nitrogen gas. The dried lipid was dissolved in 40 μl of chloroform/ methanol (2/1, v/v) and applied to HPTLC plates. HPTLC plates were run in chloroform/acetic acid (9/1, v/v). The plates were dried and soaked in 8% (w/v) CuSO 4 .5H 2 O, 6.8% (v/v) H 3 PO 4 , and 32% (v/v) methanol and then charred for 15 min in an oven at 150˚C. Scanned plates were analyzed by NIH ImageJ 1.651j8 (National Institutes of Health). proposed mechanism of inhibition by amiodarone. Acid sphingomyelinase has also been identified as another target for DIP. A role for acid sphingomyelinase is also supported by the possibility that the substrate recognition of this phospholipase C may extend to phospholipids and beyond sphingomyelin. However, while the phospholipase C activity of acid sphingomyelinase may extend to phosphatidylcholine as well as sphingomyelin, the comparative activity is an order of magnitude greater for sphingomyelin (122,123).
Although the inhibition of both LPLA 2 and lysosomal acid sphingomyelinase by drugs that cause phospholipidosis has been proposed to occur by inhibition of electrostatic interactions between the respective phospholipases and anionic lipids, this model would not explain the phospholipidosis observed by compounds that are not basic drugs. This is exemplified in the present study by mitotane and mannitol, which lack amines and thus have no assignable pKa.
DIP has been an active focus of regulatory agencies including the FDA for more than 20 years. In 2004, the FDA announced that it formed an initiative named the Phospholipidosis Working Group under the auspices of the Center for Drug Evaluation and Research (124). The overarching goal of this group was to establish regulatory guidance for drugs that were observed to cause phospholipidosis. Significant research was fostered by this initiative leading to new in silico and in vitro tests, studies on the relation The table compares the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy of the Ploemen, modified Ploemen, and LPLA2 assays. Compounds associated with false-positive and false-negative results are listed. The numbers in parentheses denote the μM IC50 for LPLA2 inhibition. between CAD and DIP, efforts to understand potential connections between phospholipidosis and other toxicities such as QT prolongation and protein trafficking defects, and biomarker development including bis(monoacylglycerol) phosphate. However, these efforts did not provide a consensus as to a common mechanism or cellular target for CADs that cause phospholipidosis.
The identification here that LPLA 2 inhibition is a primary basis for DIP provides further insight into the toxicological significance of DIP. While over 50 inherited monogenic lysosomal disorders have been identified, no clinical phenotype has yet to be described for an inherited loss of LPLA 2 activity. However, a variety of potentially important biological roles for LPLA 2 have been reported suggesting that long-term LPLA 2 inhibition may be a pathological significance. These include a role for LPLA 2 in surfactant degradation (20,125), catabolism of oxidized phospholipids (126), ocular inflammation (127), host response to tuberculosis (128), and lipid antigen presentation through CD1d (129). While LPLA 2 is expressed ubiquitously, the high activity of the phospholipase A2 in macrophages and other antigen presenting cells is consistent with an important role in host defense and antigen processing. Whether or not prolonged exposure to CADs that inhibit LPLA 2 confers increased risk to loss of these functions will require further evaluation. Importantly, the recognition that LPLA 2 is a primary target for DIP should aid in discerning drug-specific toxicities that are independent of LPLA 2 inhibition and the result of other off-target effects.
Finally, the recognition that LPLA 2 is the primary target for DIP raises the possibility that variants in the LPLA 2 gene may account for differences in susceptibility to drugs that cause phospholipidosis within the population. Numerous sequence and splice variants have been identified for LPLA 2 , several of which are in the open reading frame of the LPLA 2 gene and would predictably change the activity of the lipase either by resulting in the loss of catalytic activity or by conformational changes affecting the lipid-binding domain. Amiodarone, a highly effective antiarrhythmic, would be an obvious agent to study as DIP often limits its use. Future questions of interest might focus on establishing whether intrinsic differences in LPLA 2 activity due to these variants account for susceptibility to amiodarone toxicity and whether structure activity studies of amiodarone might identify analogues that eliminate LPLA 2 inhibition while maintaining antiarrhythmic activity.

Data availability
All data are contained within the article and supplemental data.