Liposomal FRET Assay Identifies Potent Drug‐Like Inhibitors of the Ceramide Transport Protein (CERT)

Abstract Ceramide transfer protein (CERT) mediates non‐vesicular transfer of ceramide from endoplasmic reticulum to Golgi apparatus and thus catalyzes the rate‐limiting step of sphingomyelin biosynthesis. Usually, CERT ligands are evaluated in tedious binding assays or non‐homogenous transfer assays using radiolabeled ceramides. Herein, a facile and sensitive assay for CERT, based on Förster resonance energy transfer (FRET), is presented. To this end, we mixed donor and acceptor vesicles, each containing a different fluorescent ceramide species. By CERT‐mediated transfer of fluorescent ceramide, a FRET system was established, which allows readout in 96‐well plate format, despite the high hydrophobicity of the components. Screening of a 2 000 compound library resulted in two new potent CERT inhibitors. One is approved for use in humans and one is approved for use in animals. Evaluation of cellular activity by quantitative mass spectrometry and confocal microscopy showed inhibition of ceramide trafficking and sphingomyelin biosynthesis.

Abstract: Ceramide transferp rotein (CERT) mediates nonvesicular transfer of ceramide from endoplasmic reticulum to Golgi apparatus and thus catalyzes the rate-limiting step of sphingomyelin biosynthesis. Usually,C ERT ligands are evaluated in tedious binding assays or non-homogenous transfer assays using radiolabeled ceramides. Herein, af acile and sensitivea ssay for CERT,b ased on Fçrsterr esonancee nergyt ransfer (FRET), is presented. To this end, we mixed donor and acceptorv esicles, each containing a different fluorescent ceramide species. By CERT-mediated transfero ff luorescentc eramide, aF RET system was established, whicha llows readout in 96-well plate format,d espite the high hydrophobicity of the components. Screening of a2 000 compound library resulted in two new potent CERT inhibitors. One is approved for use in humansa nd one is approved for use in animals. Evaluation of cellulara ctivity by quantitative mass spectrometry and confocal microscopy showedi nhibition of ceramide trafficking and sphingomyelin biosynthesis.
Cellularh omeostasis is the result of multiple, regulated metabolic pathways that may be over-or under-represented in pathological contexts. Pharmacological influenceo nm etabolite concentrationsi su sually achieved by inhibition of rate-limiting enzymes of ag iven pathway.H owever,r ate-limiting en-zymesf or most metabolic pathways have been identified in the past from cell lysates or with purified enzymes,a tt he expenseo ft opological information. While topologicalc onsiderationsm ay be simple for pathwaysc onverting soluble substrates, they are crucial and complex for lipid substrates. By definition, lipids are not water-soluble and transitions between different membranes require either complex vesicular transport mechanisms or specific transport proteins. [1] Like many other biosynthetic pathways, the biosynthesis of sphingolipids extends throughout the membranes of different organelles, which makess pecific transport processes essential components of the individual pathway. [2] Ag rowing body of evidence suggestst hat such transport processes are the rate-limiting steps of the whole sphingolipid biosynthetic pathway and sphingolipid transport proteins are now considered as pharmacological target molecules. [3] Sphingolipidsc onstitute an abundant and ubiquitous class of lipids in eukaryotic organisms. Their biosynthesis starts with the condensation of serine and palmitoyl CoA in the cytosolic leafleto ft he endoplasmic reticulum (ER) membranes. [4] In three further steps, ceramide is formed, which serves as the membrane anchor fort he more complex sphingolipids,l ike glycosphingolipids (GSL) or sphingomyelin (SM), am ajor component of eukaryotic plasma membranes.H owever,s ince synthesis of the latter takes place in different sub-compartments of the Golgi apparatus (Golgi),c eramide must be transported from the ER membrane to the membranes of the Golgi appara-tus. Recently,i th as been shown that non-vesicular transport of ceramide from ER to Golgi is mediatedb yt he ceramide transport protein (CERT) and that this transporter represents the rate-limiting step for the biosynthesis of sphingomyelin. [5] In fact, malfunctioning of the CERT protein leads to various pathologies [6] and inhibition of this transporter has shown promising results in experimental therapies for the treatment of cancer, [7] hepatitis Cv irus (HCV) [8] or chlamydiai nfections. [9] Some time ago, the ceramide analogueH PA-12 was developed as ap otent inhibitor of CERT. [10] Co-crystal structures of the START domain of CERT together with ceramide or HPA-12 have been published [11] andn umerouss tudies have been carried out to find an efficient synthesis for the inhibitor [12] or an improvement of its structure. [13] Indeed, improvements of affinity to the START domain have been achieved in cell-frees ystems. In most cases,c omplex assays weree mployed, in which af luorescence-labeled ceramide was used for competitive binding experimentsw ith subsequentp ull-down of the CERT protein.
Binding affinity is defined by the quotiento ft he ceramide fluorescencef rom the supernatant and from the CERT-pulldown fraction. [14] Alternatively,l iposomal transfer of radiolabeled ceramide between donor and acceptorl iposomes is possible. [15] The acceptorl iposomes are pulled down by ab eadbound lectin, since only the acceptor liposomes contain a lectin ligand, glucosylceramide. Subsequently,q uantification of the transfer is calculated by radiometric measurements of the two different fractions. Recently,afluorescence-based assay of ceramide transfer has also been described,i nw hich af luorescently labeled ceramide is transferred to liposomes containing aq uencher lipid. [16] This assay was performed in cuvettes, but miniaturization has not been described yet, probably due to low transfer rates. The fact that ceramide is one of the most hydrophobic mammalian lipids is ap rimary obstacle and the use of detergents in liposome-based applications is contraindicated. Because of these difficulties and the urgentn eed for more drug-like CERTi nhibitors, av irtuals creening has recently been performed, taking advantage of the well-known 3D binding mode of the HPA-12 inside CERT'sS TART domain. Al ead identified in this process showed aK D of 11 mm,which after optimization could be improved to 60 nm.T he cellular activity of the final compound wasc omparable to HPA-12, making it the first non-ceramide analoguei nhibitor of CERT. [17] However, nothingi sk nown aboutt he compounds toxicity or whether it is applicable to experiments in mice or humans. Despite the great success of this approach, we reasoned that the search for compounds that mimic an already known binding mode would revealo nly af raction of the possible inhibitors. We therefore aimed at developing as creening assayt hat should be as realistic as possible in order to be able to detect compounds with completely new binding modes. We reasoned that in anyc ase, at ransfer assay would be more predictive than ab inding assay.B eing inspiredb yo ur recent success in developing assayso fs phingolipid metabolizing enzymes based on Fçrster Resonance Energy Transfer (FRET), [18] we wanted to apply this principle to aC ERT-mediated lipid transfer assay.T he goal was to monitort he activity of CERT in ah omogeneous assay system in multi-well format and making use of the inherentn ormalization of fluorescenceb yd ual wavelength readout. [19] The idea wast hat perhaps ab i-directional transfer of dye-labelled ceramides coulda chieve the necessary transfer rate. For this purpose, we used fluorescent dyes, which we had proven to be effective in our previous FRET probes. Hence, ac oumarin ceramide (MCC-Cer) and an NBD ceramide (NBD-Cer) weres ynthesized, [20] respectively,a nd embedded into separate liposomes (Figure 1).
When coumarin was excited (l Ex = 347 nm), the corresponding fluorescence (l Em = 405 nm) could be observed, whereas no significant fluorescencef or NBD (l Em = 536 nm) could be observed. However,a fter addition of recombinant human CERT,a ni ncrease in NBD fluorescenceo ccurred rapidly. The curve changed to aplateau after some time, indicating atermination of the net transfer.T he final increasei nN BD fluorescence was up to 4fold. Somewhat unexpectedly,t he coumarin fluorescenced id not change significantly,i ndicating that overall only as mall fraction of the coumarin fluorescencee nergy was transferred to NBD. The specificity of the NBD fluorescence change was verifiedb yd ifferent concentrationso fC ERT and the possibility of inhibition by HPA-12 (see Supporting Information for details). Transfer of this assay into 96-well plates showedt he same results without significant loss of lipid transfer activity.T ot est the hypothesis of bidirectional transfer,o ne of the two ceramide derivativesw as successivelyr eplaced by a non-transferable lipid carrying the same dye. Replacement of coumarin ceramide showed the same resultsa sa bove,w hile the replacemento fN BD ceramide by non-transferrable NBDlipid resulted in the absence of any fluorescence change (data not shown). We therefore concluded that NBDc eramide (NBD-Cer) but not coumarin ceramide (MCC-Cer) is transferred by CERT.H avinge stablished af easible miniaturized CERT assay,we aimed to perform al imited screening for possible inhibitors. However,t he absence of transfer the short excitation wavelength of coumarin seemed to us unfavorable, since many aromatic compounds show as ignificanta bsorption at such a short wavelength.T his would interferew ith our screening assay.S ince we showedt he transferability of NBD ceramide, we considered making this lipid the FRET donor and providing it with as uitable acceptor lipid. For this scenario, Nile red (NR) emerged as as uitable partner and therefore, aN R-ceramide was synthesized,a ccordingly. [21] Fortunately,a lso with this ceramide FRET pair an assay in 96 well format could be successfully established. Moreover,i na ne xperiment, similar to the one described above,w eo bserved transfer of NR-ceramide in presence of non-transferrable NBD-lipid, which was inhibited in the presence of the inhibitor HPA-12. Finally,alibraryo f 2000 compounds with proven pharmacological and/or biological activity was tested at ac oncentration of 10 mm.S ignificant inhibition-defined by inhibition like 1 mm HPA-12 or betteryielded at otal of 45 primary hits. Dose-response evaluationf or these hits and sorting out of compoundsw hose structures did not indicate any promising furtherd evelopment,r esulted in a total of 9c ompounds that were considered promising and which were further investigated. The inhibitory effects were found to be highly reproducible, with the exact values showing somevariation ford ifferent batches of the liposome prepa-ration. For each assay,H PA-12 was therefore used as ar eference substance. The assay established by us is in principle an enzyme-catalyzed mixingo fs eparatel ipid pools and is therefore potentially interferedb ys ubstances that stimulate vesicle fusion. Such compoundsw eree xpectedt og ive false negative results (no inhibition), although an umber of compounds with detergent-like properties showed inhibition, contrary to this hypothesis. Nonetheless, we aimed at further investigation, using orthogonal assays devoid of liposomes in order to proof the validity of our approach. The binding assays described so far seemed to us very complex and tedious for dose-response studies. Therefore, we considered using the environmental sensitivity of the existing dye-labeled ceramides for ac ompetitive assay.N ile red is knownt os how strongestf luorescence in a non-polar environment, therefore we suspected that Nile red ceramide bound to ah ydrophobic pocket of CERTw ould be much more fluorescent than in the unbound state. To test this hypothesis, a1mm micellar solution of Nile red ceramide was treated with increasing concentrations of CERT.I nf act, an up to 20-fold increasei nf luorescencew as observed, which could be reverted to approximately the initial value using HPA-12. Notably,t his effect was much less pronounced using the less active R,R diastereomer of HPA-12 (see Figure S8). Hence, the fluorescenceo fN ile red ceramide is therefore as uitable tool to evaluatec ompetitive CERT binders in ah omogenous fashion. After the NR ceramide competition assay,f our compounds were selected for further evaluation in cells. Three compounds, 8E8, 20D5 and 6B11s howed high activity in both, the transfer and the competition assay ( Figure 2B&C). Furthermore, we se-lected 17C9, which showed aw eak dose-response in the transfer assay, but av ery strong competitiont ot he binding of NR ceramide. Finally,w ef luorescently labeled the recombinant START domain of CERT and monitored binding to the candidate compounds, using microscale thermophoresis (MST).F or HPA-12,aK D of % 40 nm wasd etermined, which is in rather good agreement with K D values determinedi narecently developed surface plasmonr esonance assay,i nw hich the START domain was immobilized on solid phase. [17] Our new solution phase MST-based binding assay confirmed binding for all compounds except 17C9, but the affinitiesf or 8E8 (K D % 5 mm), 20D5 (K D % 150 mm)a nd 6B11( K D % 5 mm)w ere significantly lower than for HPA-12 ( Figure S6). This could reflect different binding modes to CERT,w hichm ay include components other than the START domain.
Finally,w et ested the activity of these compounds in cellbased assays (Figure 3). For this purpose, HeLa cells were incubated with the identified compounds or HPA-12 for 48 hours in as erum-reduced medium, and the effecto ni ndividual sphingolipid species was analyzed by quantitative mass spectrometry. [22] Indeed,8 E8 and2 0D5 showed ad ose-dependent reduction of total sphingomyelin, while treatment with 6B11o r 17C9 showedn os ignificant changes in total sphingomyelin. Compared to HPA-12, 20D5 showedas imilart otal sphingomyelinf or all concentrations used. In contrast, 8E8 showed lower effects than HPA-12 at lower concentrations, but ad rastic reduction of sphingomyelin at 5 mm.I na ddition, relationo f total sphingomyelin to ceramide showed much clearer dose dependencies for both 20D5 and 8E8 and an increased poten-   cy for 8E8 over all concentrationsc ompared to HPA-12 and similar or higher potency for 20D5. For the compounds finally investigate, the transfer assayw as more predictive for the reductioni nS Mb iosynthesis, than the NR-ceramide competition assay.S urprisingly,6 B11a nd 17C9 led to ad ramatically in-creasedS M/Cer rate, which was due to ar eduction of almost all ceramide speciesi nvestigated. At present we do not have a conclusive explanation for this behavior,b ut af uture investigation of this phenomenon could be interesting.
Because all substances used were compounds with known biological activity,i tc annotb ee xcluded that the observed changes in the sphingolipidome are due to indirect effects. To test the hypothesis that the identified compounds actually inhibit transport of ceramide in living cells, we treated HeLa cells with fluorescence-labelled ceramide (BODIPY-Cer) in the presence of the four final compoundsa nd examined the effects by confocalm icroscopy.I nf act, 6B11a nd 17C9 showedo nly minor effects and resembled the untreated controls, but compounds 8E8 and 20D5 significantly influencedt he intracellular distribution of BODIPY-Cer (Figure3Ca nd Figures S10&S11). While the phenotype induced by 20D5 resembled that of HPA-12, there was am ajor change in phenotype induced by 8E8. 20D5 (Fluralaner) is used in veterinary medicinea nd targets arthropod parasite chloride channels and has no known target in mammals. [23] In contrast, 8E8 (lomitapide) is targeting the human microsomal triglyceride transfer protein (MTTP). [24] This protein is located in the ER lumen and catalyzes the loading of triglycerides, cholesterol esters and phospholipids into lipoproteins. [25] It has been shown that MTTP deficiency reduces plasma levels of ceramide and to al esser extend of sphingomyelin and that the purified protein is capable of transferring radioactively labeled sphingomyelin and ceramide between liposomes. [26] Based on our results that these molecules also interact with purified human CERT and inhibitC ERT-mediated ceramide transfer,w ee stablished binding models for thesec ompounds by molecular docking and moleculard ynamics (MD) simulations.
MD simulations revealed that the CERT START domain in apo form could close-open at the W1l oop ( Figure 4A). The W1 loop showed high temperature factors in the crystal structure [27] and might act as ag ate for ligand binding. Both, 8E8 (lomitapide) and 20D5( Fluralaner) maintained stable binding in the ceramide-binding pocket, with some fluctuationsn ear the highly mobile W1l oop. Their amide group mimicked that of the ceramide to form ah ydrogen bond with residue Y553 (Figure 4B and 4C). However,t he affinity was mostly attributed to hydrophobic interactions, especially with Y576 and F579 (see Figure S7 for key interactions and binding energies). In addition, 8E8 bound ionically with E446 and had significantly lower bindingf ree energy (À26.3 AE 14.5 kcal mol À1 )t han 20D5 (À8.3 AE 6.8kcal mol À1 ), which maya ccount for the stronger inhibitions by 8E8.
In summary,w eh ave developedanew FRET-based ceramide transfer assay to identify new CERTinhibitors. For two of these compounds, we showede ffective inhibition of CERT-mediated transfer in vitro, replacemento fC ERT-bound Nile red ceramide and interaction with the fluorescently labeled START domain of CERT by MST.F inally,t wo compounds resulted in an increase of cellularc eramide at the expense of sphingomyelin concentrations. These two compounds were significantly more active than HPA-12 at higher concentrations( 5mm). Moreover,c onfocal microscopy of treated cells revealed that the compounds altered ceramide trafficking.N oteworthy,t hese compounds are approved for pharmacological use in humans (Lomitapide)o r animals( Fluralaner). By modelling the identified structures into the START domain of CERT,w eestablished ac onclusive binding model, which may be used for structure-guided design of future CERT inhibitors with increased affinity and selectivity. Our results motivatet os creeningo fl arger libraries and to apply the principles developedh ere to further lipid transferases of biomedical interest.