Fenretinide binding to the lysosomal protein saposin D alters ceramide solubilization and hydrolysis

Fenretinide is a synthetic retinoid pharmaceutical linked to ceramide build-up in vivo.

Abstract

Fenretinide is a synthetic retinoid pharmaceutical linked to ceramide build-up in vivo. Saposin D is an intralysosomal protein necessary for ceramide binding/degradation. We show, via electronic absorption spectroscopy, fluorescence spectroscopy, and ceramide hydrolysis assays, that fenretinide is bound by saposin D {K_a = (1.45 ± 0.49) × 10⁵ M^–1}, and affects ceramide solubilization/degradation.

Proteolysis of the human lysosomal glycoprotein prosaposin (PSAP) produces a family of four sphingolipid activator proteins termed saposin A, B, C and D (sapA, B, C and D; Fig. 1).1 Saposin proteins are non-enzymatic, heat-stable glycoproteins that traffic and solubilize specific sphingolipids and facilitate hydrolysis in-tandem with specific lysosomal hydrolases.1–4

Fig. 1. Synthetic scheme of sapD + acid ceramidase degradation of ceramide.

More specifically, sapD promotes the intra-lysosomal hydrolysis of the sphingolipid secondary messenger ceramide, by acid ceramidase (AC), via solubilization of the lipid, and presentation of such to the enzyme, as demonstrated by the accumulation of α-hydroxyl fatty acid ceramides in the kidneys, and in the cerebellum of sapD^–/– mice.5 Supra-physiological build-up of ceramide in humans is termed Farber's disease, an extremely rare lysosomal storage disease (LSD) caused by deficiency in AC and/or in total saposin deficiency.6,7 Despite extensive research into PSAP, sapA, B and C, considerably less is known about sapD, the most abundant saposin in human tissues evaluated.1,8

Fenretinide [4-hydroxy(phenyl)retinamide (4-HPR); Fen] is a synthetic retinoid pharmaceutical (Fig. 2), structurally related to vitamin A, and investigated for use in the treatment of cancer,8–10 particularly of the breast,8 as well as cystic fibrosis (CF).11 In human cancer research, and studies in patients with CF, Fen was found to increase ceramide levels, by activation of the de novo synthesis pathway or sphingomyelinase enzyme driven conversion of sphingomyelin to ceramide.9 This up-regulation of ceramide is linked to the drugs method of action, either by facilitating tumour cell apoptosis (in the treatment of cancer), or replenishing depleted ceramide levels in CF,12 aiding in bacterial clearance from CF affected organs. This elevation of ceramide levels associated with Fen administration has been tied to an increase in production of the sphingolipid,9 with no mechanism to date describing a possible route to decreased degradation of such.

Fig. 2. The structure of the retinoid pharmaceutical fenretinide [4-hydroxy(phenyl)retinamide (4-HPR); Fen].

Herein, we demonstrate that sapD, important in facilitating ceramide solubilization and AC dependent hydrolysis, binds Fen. The formation of the [sapD–Fen] complex subsequently prevents ceramide solubilization. We also demonstrate that binding of ceramide to sapD prior to Fen addition, results in the formation of a ternary complex of [sapD–ceramide–Fen], similar to that observed for [sapB–sulfatide–chloroquine] and [sapB–CoQ10–A2E] ternary complexes, previously reported by us. Together, these observations suggest the possibility that multi-substrate binding may occur across all saposin family proteins.13,14 The formation of [sapD–ceramide–Fen] does maintain ceramide solubility, but results in dissipated ceramide hydrolysis (in the presence of AC).

Results and discussion

Initially, we sought to measure binding with isothermal titration calorimetry, however issues with obtaining sufficient concentrations of Fen in a suitable aqueous buffer meant attempts at such produced variable results, which supported binding, but gave differing binding values. We therefore moved to electronic absorption spectroscopy (Fig. 3) and fluorescence quenching spectroscopy (Fig. 4) experiments to investigate the ability of sapD to bind Fen, as previously demonstrated with sapB.13,14 Both spectroscopic techniques produced fully reproducible results that coincided with an average K_a of (1.45 ± 0.49) × 10⁵ M^–1; K_d = 6.9 μM. It should be noted, the discontinuity (i.e. the break in the linearity of the absorbance and fluorescence changes versus the concentration of the drug) represents the binding stoichiometry between sapD and Fen, which for our assays is ∼1 drug molecule per sapD monomer. Noticeably, sapB displayed ∼1 ligand molecule per sapB dimer.13,14

Fig. 3. Differential absorbance changes as a function of Fen concentration (left panel). Double reciprocal plot (i.e. Benesi–Hildebrand plot) of the differential absorbance changes as a function of the reciprocal concentration of Fen (right panel). Here, ΔA is the change in absorbance values between sapD and Fen, and Δε is the change in molar extinction coefficients. Conditions: 3–9 μM sapD in 50 mM sodium phosphate buffer, pH 5.5, 2 μl injections of 250–750 μM Fen in DMSO. Similar results were obtained when Fen is dissolved in methanol and sapD in 50/50 DMF/PBS buffer pH 5.5. Break in linearity describes stoichiometry and is consistent with 1 Fen per sapD monomer.

Fig. 4. Fluorescence quenching of sapD upon Fen binding (left panel), and plot of the fluorescence intensity change as a function of Fen concentration (left panel, inset). Logarithmic plot of the fluorescence quenching of sapD with different concentrations of Fen (right panel). Conditions: 15 μM sapD in 50 mM sodium phosphate buffer, pH 5.5, 2 μl injections of 1.13 mM Fen in DMSO. λ_exc = 280 nm, λ_em = 300–450 nm with excitation and emission slits opening of 5 nm each. Similar results were obtained using 50/50 DMF/PBS buffer, pH 5.5.

Upon confirmation of sapD binding of Fen, we sought to investigate the implications of this interaction on the binding, solubilization and/or AC dependent hydrolysis of ceramide. A fluorescent analog in the form of BODIPY TR ceramide (BDC) was used to allow tracking of fluorescent emission in the 610–670 nm range (Excitation@589 nm) (Fig. 5B and D). The [sapD–Fen] complex was preformed (Fig. 5A) and used for subsequent BDC titrations. A control of BDC into 50 mM sodium phosphate buffer pH 5.5 (PB) (Fig. 5B and F) did not solubilize, as expected, and congruent with un-modified ceramide. Titration of BDC into sapD prepared in PB resulted in lipid binding and solubilization (Fig. 5C and E), again congruent with the ceramide data presented herein. With the controls in place, we then titrated BDC into preformed [sapD–Fen] in PB, which revealed a lack of solubilization/binding of the BDC (Fig. 5D), as evidenced by the presence of insoluble, fluorescent BDC droplets (Fig. 5F). The presence of BDC droplets at lysosomal pH can prevent binding and subsequent solubilization of ceramide moieties, not observed in the absence of Fen (Fig. 5E).

Fig. 5. (A). Fluorescence quenching of sapD + Fen in PB (Excitation@280 nm). (B). Titration of BDC into PB (Excitation@589 nm). (C). Fluorescence quenching of sapD + BDC in PB (Excitation@280 nm). (D). BDC addition to preformed [sapD–Fen] (Excitation@589 nm). (E). 20 μM sapD in PB titrated with 20 μM BDC (five injections of 2 μl each), imaged after 30 minutes showing the solubilization of the lipid. (F). Titration of 20 μM BDC into precomplexed 20 μM [SapD–Fen] in PB, imaged after 30 minutes, showing the loss of lipid solubilization due to pre-complexation of Fen with sapD.

Subsequent to the solubilization experiments, we then asked whether preformed [sapD–Cer] could still bind Fen, an order of addition effect we had observed and reported on before for sapB, where we showed the sap protein could have multiple binding sites.13 We therefore titrated Fen into [sapD–Cer] and observed evidence for Fen binding (Fig. 6A and B). To further investigate this, we assayed whether this new [sapD–Cer–Fen] complex interfered with AC dependent hydrolysis of the ceramide. Equimolar concentrations of either [sapD–Cer] or [sapD–Cer–Fen] complexes were exposed to AC, and hydrolysis monitored using reverse-phase high-performance liquid chromatography (Fig. 6C and D) over 24 h.

Fig. 6. (A) Binding of BDC by [sapD]. (B) Binding of Fen to [sapD–BDC]. Binding monitored via fluorescence quenching of sapD with excitation at 280 nm and emission scan range of 300–400 nm. Fen binding to sapD or its presence hinders AC dependent turn-over of sphingolipid hydrolysis, (C) CerOH (N-palmitoyl-d-erythro-sphingosine) and (D) CerPO₄ (N-palmitoyl-d-erythro-sphingosine). Data in (C and D) were analyzed with repeated measurements one-way ANOVA test. *P < 0.005, **P < 0.001. Conditions: 25 μM sapD in 50 mM sodium acetate buffer at pH 4.0, with equimolar concentrations of Fen and/or ceramide (25 μM), 0.5 μM AC and incubated at 37 °C (AUC = area under the curve; mAU.s = molar absorbance units second, peak area).

Ceramides, CerOH (N-palmitoyl-d-erythro-sphingosine) and CerPO₄ (N-palmitoyl-d-erythro-sphingosine), were both evaluated pre-complexed to sapD with and without Fen and challenged against a preformed sapD–Fen complex (Fig. 6C and D). In both studies when the sapD–Cer complex was challenged against the presence of Fen, each assay resulted in less hydrolysed ceramide over 24 h. Furthermore, pre-complexed [sapD–Fen] resulted in ∼50% less ceramide hydrolysed, in both studies. These results show Fen's ability to compete with ceramide moieties for binding that can directly affect ceramide hydrolysis by AC in a 24 h circadian cycle, which may hint at a possible mechanism for Fen's functionality in cancer, and/or CF therapy.

To determine the amount of ceramide hydrolysed, multiple wavelengths were monitored: 206, 220, 280, and 365 nm. Initially the sapD–ceramide complex was monitored to follow degradation of the complexation, followed by peak production and growth of the hydrolysed product. Multiple wavelengths were monitored as numerous complexations were possible post-hydrolysis with free sapD able to bind the hydrolysis product, Fen and/or AC.

It is evident that Fen interaction with sapD interferes with the binding and subsequent hydrolysis of ceramide moieties, but further evaluation was needed to determine whether Fen could inhibit ceramide hydrolysis via direct AC inhibition as well. Tamoxifen, used to treat breast cancer, has been shown to inhibit AC's ability to hydrolyse ceramides contributing to the therapeutics' pharmacodynamic profile.15–17 We sought to investigate if Fen had similar inhibition properties to tamoxifen by pre-incubating Fen with sapD and AC, independently, for 30 minutes prior to performing the hydrolysis assay and monitored hydrolysis for 72 h (Fig. 7). In both studies, Fen pre-incubation with either sapD or AC nearly completely inhibited ceramide hydrolysis. The absence of Fen resulted in measured hydrolysis over 72 h. Experiments were performed in duplicate and monitored at 206 and 220 nm.

Fig. 7. Inhibition assays to evaluate the pre-binding of Fen to either sapD or AC demonstrated Fen may inhibit ceramide hydrolysis at both the sapD and AC interface during sapD assisted enzymatic hydrolysis of ceramide moieties. Mean and standard error of mean presented, R² values for sapD–CerPO₄ 0.861, sapD–CerPO₄ + AC is 0.871, sapD–CerPO₄ + AC–Fen 0.901, and sapD–Fen + AC–CerPO₄ 0.685. Conditions: 25 μM sapD in 50 mM sodium acetate buffer at pH 4.0, with equimolar concentrations of Fen and/or ceramide (25 μM), 0.5 μM AC and incubated at 37 °C.

There is limited information of the sapD active site since there is no structure of sapD bound with a ligand published. We used SwissDock molecular modeling18 with the known crystal structure of human sapD (PDB:; 2RB3)19 to simulate the docking of three targeted ligands, Fen, tamoxifen, and 18 : 1 ceramide-1-phosphate (Fig. 8). The docking simulations displayed a hydrophobic cleft viable for the binding of hydrophobic/lipophilic molecules. Polar headgroups of residues surrounding the hydrophobic cleft appear to accommodate polar moieties of the hydrophobic/lipophilic substrates like the sapD binds, such as the ceramide core or hydroxyaniline moiety of Fen.

Fig. 8. SwissDock molecular modeling for sapD binding of targeted ligands. Docking images were generated using the SwissDock modeling in Chimera 1.14 (A) crystal structure of human saposin (PDB:2RB3) used as protein input for docking simulations, (B) residues involved in the active site of sapD determined via docking simulations, (C) 18 : 1 ceramide-1-phosphate (from PDB:; 4K8N) docked in sapD, (D) active site for 18 : 1 ceramide-1-phosphate docking, (E) Fen docked in sapD, (F) active site for Fen docking, (G) tamoxifen docked into sapD, (H) active site for tamoxifen docking. SwissDock estimated ΔG (kcal mol^–1) (C) –6.84, (E) –6.40, (G) –7.26. For summary of docking interactions see Table S1.† .

Additional docking simulations were evaluated with sapB (PDB:4V2O) and Fen to investigate the possible interactions Fen may have with the other saposins. Previous investigations into sapB and chloroquine produced a crystal structure detailing a ligand docking site.13 SwissDock molecular modelling software docked Fen into the same binding region for chloroquine (Fig. S4†). The lipophilic tail of Fen docked into a hydrophobic pocket between the monomers of the dimeric sapB structure, with the hydroxyaniline moiety sitting between R38 (monomer A), M61, M65, and L73 (monomer B), including a hydrogen bond between the Fen phenol and R38.

Conclusions

Herein, we demonstrate that sapD can bind the bisretinoid pharmaceutical Fen with an average K_a of ∼1.5 × 10⁵ M^–1, as measured by electronic absorption and fluorescent spectroscopy. The impact of Fen binding to sapD is seen in the loss of sapD dependent ceramide solubilization in aqueous buffer in the presence of the [sapD–Fen] complex. In addition, we demonstrate that a preformed [sapD–ceramide] complex can still subsequently interact with Fen, resulting in a decrease in AC dependent hydrolysis turn-over in 24 h. Our results suggest a possible mechanism of action for Fen, along with direct inhibition of AC, in terms of the empirically observed build-up of ceramide noted in human studies. These data suggest also that new analogues of Fen with greater affinity for sapD (and greater solubility) might produce even greater positive outcomes in terms of treating breast cancer or CF, and we have begun to explore this novel approach.

Experimental

Expression and purification

A pPROEX vector containing human sapD (PSAP residues 405–486, corresponding to sapD residues 1–81) was kindly provided by the Nadar group at McGill University in Montreal.19 SapD was recombinantly expressed in SHuffle® T7 Competent E. coli cells (New England Biolabs), yielding ∼12 mg per L of LB media. Cultured media was incubated and shaken in a C24 Incubator Shaker (New Brunswick Scientific) at 175 rpm and 37 °C until an OD₆₀₀ of 0.8 was achieved. Protein expression was induced with isopropyl β-d-1-thiogalactopyranoside (IPGT) at a final concentration of 400 μM, after which, the temperature was lowered to 20 °C for overnight expression. Post expression, cultures were pooled and centrifuged (Sorvall Legend RT, Thermo) at 4150 rpm for 15 min at 4 °C. Cell pellets were collected in a minimal amount of 50 mM PB at pH 5.5 and placed on ice. Pierce™ Protease Inhibitor Mini Tablets, EDTA-Free (Thermo Scientific) were added to the cellular solution and stirred on ice until pellets dissolved. The cellular solution was sonicated with a Sonic Dismembrator Model 100 (Fisher Scientific). The lysate was pooled and centrifuged at 4150 rpm for 15 min at 4 °C to collect the supernatant containing crude sapD.

Protein purification was performed by fast protein liquid chromatography (FPLC) using a 5 mL HisTrap Excel (GE Healthcare) on an AKTA Prime Plus liquid chromatograph with an automated fraction collector and UV detection at 280 nm (GE Healthcare). The system was equilibrated with 50 mM PB at pH 5.5 to load crude extract, followed by 5× column volume washes with 50 mM PB, 20 mM imidazole at pH 5.5. The sapD product was eluted and collected using 50 mM PB, 200 mM imidazole at pH 5.5, then dialyzed in 3 times against 50 mM PB at pH 5.5. SDS-PAGE electrophoresis was performed to track sapD and verify purity (Fig. S1†). Mass spectrometry and circular dichroism spectroscopy were used to confirm identity and folded state. (Fig. S2 and S3†)

Electronic absorption20 and fluorescence quenching measurements were carried out on a Varian Cary 50 Bio spectrophotometer and a Varian Cary Eclipse fluorimeter from Agilent Technologies, respectively. Data analysis was performed as described by Priyadarsini et al.21 All experiments were conducted at 25 °C using a QNW Quantum Northwest TC125 temperature controller, in 50/50 DMF/PBS buffer, pH 5.5. Final reagents concentrations are indicated in the figure captions. All experiments were repeated three to five times using independent protein preparations to ensure reproducibility. The absorbance and fluorescence traces shown in the figures represent one of multiple individual runs.

Ceramide hydrolysis assay was adapted from the literature.22–24 SapD was precomplexed to either Fen, CerOH, or CerPO₄ at equimolar concentrations in 50 mM sodium acetate buffer at pH 4.0. The counter substrate, either Cer or Fen was added, respectively, at equimolar concentrations, followed by the addition of AC and incubated at 37 °C. Hydrolysis was monitored using an Agilent InfinityLab Poroshell 120 EC-C8 2.7 μm 4.6 × 100 mm analytical LC column on an Agilent 1200 Series High-Performance Liquid Chromatography instrument. The ceramide hydrolysis was monitored at 206 and 220 nm to monitor the depleted ceramide levels over a 24 h period. Fen inhibition of AC was monitored over 72 hours, with complexation of targeted mixtures incubated for 30 minutes at 37 °C, prior to performing the hydrolysis assay.

Conflicts of interest

RPD holds shares in Lysoclear, Inc. (LaFayette, NY, USA), a wholly owned subsidiary of Ichor therapeutics, Inc.