Acid ceramidase as a therapeutic target in metastatic prostate cancer.

Acid ceramidase (AC) catalyzes the hydrolysis of ceramide into sphingosine, in turn a substrate of sphingosine kinases that catalyze its conversion into the mitogenic sphingosine-1-phosphate. AC is expressed at high levels in several tumor types and has been proposed as a cancer therapeutic target. Using a model derived from PC-3 prostate cancer cells, the highly tumorigenic, metastatic, and chemoresistant clone PC-3/Mc expressed higher levels of the AC ASAH1 than the nonmetastatic clone PC-3/S. Stable knockdown of ASAH1 in PC-3/Mc cells caused an accumulation of ceramides, inhibition of clonogenic potential, increased requirement for growth factors, and inhibition of tumorigenesis and lung metastases. We developed de novo ASAH1 inhibitors, which also caused a dose-dependent accumulation of ceramides in PC-3/Mc cells and inhibited their growth and clonogenicity. Finally, immunohistochemical analysis of primary prostate cancer samples showed that higher levels of ASAH1 were associated with more advanced stages of this neoplasia. These observations confirm ASAH1 as a therapeutic target in advanced and chemoresistant forms of prostate cancer and suggest that our new potent and specific AC inhibitors could act by counteracting critical growth properties of these highly aggressive tumor cells.

chemotherapeutic regimes ( 19,20 ). Two major challenges in PC are to fi nd predictive markers that identify those tumors most likely to follow a hormone-independent, aggressive clinical course as aids to decide early intervention and to identify molecular targets for improved therapies of castration-resistant cases that respond poorly to conventional chemotherapeutic regimes. Here, we provide new evidence to reinforce the notion that the acid ceramidase ASAH1 is a valid therapeutic target in advanced prostate cancer, and we characterize new potent and specifi c inhibitors of AC.

Cells and reagents
PC-3/Mc and PC-3/S cells ( 21 ) were grown in RPMI1640 medium supplemented with 10% fetal bovine serum, nonessential amino acids, 2 mM glutamine, 1 mM sodium pyruvate, 100 U/ml penicillin, and 100 µg/ml streptomycin (all from PAA, Ontario, Canada). Fibroblasts from a Farber patient (FD, wild-type) and FD transformed to stably overexpress AC (FD10X) were grown in a humidifi ed 5% CO 2 atmosphere at 37°C in DMEM medium supplemented as above.

Transient transfection of fi broblasts
Twenty-four hours before transfection, cells were plated at a density of 2.5 × 10 5 cells in 35 mm diameter plates. Cells were then transfected with the specifi c constructs or empty vectors using lipofectamine 2000 (Invitrogen, Carlsbad, CA). Twenty-four hours after transfection, cells were either processed immediately or collected by trypsinization, washed twice with PBS, and centrifuged. Pellets were stored at Ϫ 20°C until use. In transfections with neutral ceramidase (NC), the success of transfection was confi rmed by activity assays with CerC12NBD in intact cells.

Fluorogenic ceramidase assay
Cells were collected by trypsinization, washed with PBS, resuspended in 0.25 M sucrose and lysed by ultrasonication in an ultrasonic bath. For the assay, 75 µl of reaction buffer (100 mM sodium acetate buffer, pH 4.5, for acid ceramidase activity), containing 40 µM RBM14C12 fl uorogenic substrate ( 22 ) (with or without test compounds), was mixed with 25 µl of the cell lysates (20 µg of total protein content) and incubated at 37°C for 3 h. For time dependence of inhibition, incubations were carried out for 0.5, 1, 2, and 3 h with different amounts of protein. The reaction was stopped by addition of methanol followed by NaIO 4 (2.5 mg/ml in 200 mM glycine/NaOH buffer, pH 10.6). After 1 h at 37°C, 100 l of 200 mM glycine/NaOH buffer were added and fl uorescence detected at 355/460 nm excitation/emission wavelengths on a SpectraMax Microplate Reader (Molecular Devices, Sunnyvale, CA). To determine ceramidase activity in intact cells, 2 × 10

Papain activity
Papain activity was determined in 96-well plates by a modifi cation of the reported procedure ( 23 ). The reaction mixture contained 250 µl of 0.1 M phosphate buffer (pH 6.5) with 0.3 M KCl, 0.1 mM EDTA, and 3 mM DTT; 30 µl of substrate solution (L-pyroglutamyl-L-phenylalanyl-L-leucine-p -nitroanilide; 2.2 mM many neoplastic cells activate to offset the toxic accumulation of palmitate is its peroxisome proliferator-activated receptor (PPAR) ␥ -dependent funneling to eventually form triglycerides, which can be further used as energy stores ( 4 ). A second pathway followed by palmitate is the condensation of palmitoyl CoA with L-serine, leading to the synthesis of ceramides ( 5 ). The accumulation of ceramides also poses a problem for cell survival because of its proapoptotic consequences (6)(7)(8). This can be counteracted by activities that convert ceramides to a variety of metabolites, including sphingosines through deacylation catalyzed by ceramidases. Sphingosines can be converted into the growth-and survival-promoting sphingosine-1-phosphate (S1P) through the action of sphingosine kinases ( 8 ). S1P can be either irreversibly degraded by S1P lyase or reutilized by sequential dephosphorylation and acylation for ceramide synthesis. The importance of ceramidases in the context of cancer is supported by the observation that their inhibition by drugs or RNAi severely compromises the growth and survival under stress of tumor cells ( 9 ).
De novo ceramide biosynthesis requires the coordinate action of serine palmitoyl transferase and ceramide synthase to generate ceramide. This process begins with the condensation of serine and palmitoyl-CoA to form 3-ketosphinganine ( 5 ), which is reduced to the sphingoid base sphinganine and acylated by ceramide synthase to generate dihydroceramide. This compound is oxidized to ceramide by introduction of a trans -4,5 double bond. This pathway can be stimulated by drugs and ionizing radiation and usually results in a prolonged ceramide accumulation ( 10 ).
Once generated, ceramide may amass or be converted into a variety of metabolites. Phosphorylation by ceramide kinase ( 11 ) generates ceramide 1-phosphate, while deacylation by alkali ne, neutral or acid ceramidases (the products of different genes) ( 12 ) yields sphingosine, which may be phosphorylated by sphingosine kinase to S1P. Two distinct sphingosine kinases have been cloned. These two isoforms differ in temporal patterns of expression during development, are expressed in different tissues, and possess distinct kinetic properties ( 13 ), implying that they perform different cellular functions. Ceramide may also be converted back to SM by transfer of phosphorylcholine from phosphatidylcholine via SM synthases ( 14 ). Alternatively, it can be glycosylated by glucosylceramide synthase to form glucosylceramide, which may be further modifi ed by various enzymes in the Golgi apparatus to form complex glycosphingolipids ( 15 ).
Many tumor types express high levels of acid ceramidase (AC). Specifi cally, the expression levels of AC in prostate cancer have been reported to be elevated relative to normal prostate tissue ( 16,17 ). Prostate cancer (PC) is the most prevalent neoplasia in men in industrialized nations ( 18 ). Although PC is frequently initially sensitive to hormonal deprivation therapies and follows indolent clinical courses, a signifi cant proportion of cases eventually become resistant to such therapeutic approaches, accompanied with aggressive growth, establishment of metastasis, and tumors that are highly resistant to conventional 480 Software release 1.5.0. The amplifi cation levels of RN18S1 and HMBS were used as internal references to estimate the relative levels of specifi c transcripts, and relative quantifi cation was determined by the ⌬ ⌬ Cp method. All determinations were done in triplicate.

Cell cycle analysis
Cells were seeded in 6-well Corning plates (Corning, NY), detached with Trypsin/EDTA/1% BSA, washed twice, resuspended in PBS, and fi xed at 4°C for at least 2 h by dropwise addition of 70% ethanol. Subsequently, cells were washed with PBS/50 mM EDTA/1% BSA and incubated with 1 mg/ml RNase A (Sigma) at 37°C for 1 h and 0.1 mg/ml propidium iodide (Sigma, Alcobendas, Madrid, Spain). DNA content was determined in a Cytomics FC500 instrument (Coulter, Hialeah, FL), and cell cycle distribution analyzed with Multicycle. All determinations were done in triplicate.

Anchorage-independent growth
For soft-agar colony formation assays, 0.5% agar in complete culture medium was placed at the bottom of 12-well plates, allowed to solidify, and overlayed with a suspension of 3 × 10 3 cells in 0.3% agar in complete medium. After solidifi cation, wells were fed with complete medium twice a week. After three weeks, they were fi xed with 0.5% glutaraldehyde, stained with 0.025% crystal violet, and visualized under a Leica magnifying glass (Wetzler, Germany) coupled to an Olympus digital camera (Olympus, Hamburg, Germany). Colonies у 0.2 mm diameter were scored with the ImageJ software (National Institutes of Health, MD). Each experimental condition was performed in triplicate.

In vivo tumor formation
For localized growth, 1 × 10 3 to 1 × 10 5 cells with stably integrated fi refl y luciferase were injected in a volume of 50 µl of RPMI 1640 (without FBS) intramuscularly in each hind limb of anesthetized six-week-old male NOD-SCID mice. Tumor growth was monitored by luminometry on an ORCA-2BT instrument (Hamamatsu Photonics, Hamamatsu, Japan), 5 min after intraperitoneal injection of luciferine (100 mg/kg in 150 µl of PBS). For lung colony formation, 5 × 10 5 cells in 150 µl RPMI1640 were injected through the dorsal caudal vein. Mice were imaged immediately after injection, and thereafter, tumor development was monitored by weekly imaging. For bioluminescence plots, photon fl ux was calculated relative to background values from luciferin-injected mice with no tumor cells and normalized to the value obtained immediately after xenografting. In lung colonization free survival analysis, lesions that had an increased photon fl ux value above day 0 were counted as events.

Production and transduction of lentiviral particles
Constructs based on the pLK0puro vector and bearing ASAH1targeting shRNAs or control sequences were purchased from Sigma-Aldrich. The lentivirus packaging cell line HEK293T was cotransfected with these DNAs, together with pCMVdeltaR8.91 and pVSV-G (Clontech, Mountain View, CA) for 12 h using Fugene HD (Roche). Supernatants were collected for the following 48 h and fi ltered through 0.45 µm methylcellulose fi lters (Millipore, Billerica, MA). Lentiviral particles were concentrated by ultracentrifugation at 27,000 rpm for 90 min on 20% sucrose density gradients. Viral particles were resupended with medium and added to the cells together with 8 µg/ml polybrene (Sigma). Cells were infected for 24 h and allowed to recover in fresh medium for 24-48 h. Selection for cells with integrated sequences was carried out for three days in medium supplemented with 5 µg/ml puromycin (Biomol, Exeter, UK). in DMSO, 0.22 mM fi nal concentration); 20 µl of enzyme solution (30 µg/ml in reaction buffer); and 3 µl of inhibitor solution or vehicle. Chymostatin at 1 µM and 10 µM was used as positive control of inhibition of papain activity. The reaction was stopped by the addition of 20 µl of 1 N HCl, and the OD was measured at 410 nm.

Sphingolipid analysis by UPLC/MS
The liquid chromatography-mass spectrometry equipment consisted of a Waters Aquity UPLC system connected to a Waters LCT Premier orthogonal accelerated time of fl ight mass spectrometer (Waters, Millford, MA), operated in positive electrospray ionization mode. Full scan spectra from 50 to 1,500 Da were acquired, and individual spectra were summed to produce data points every 0.2 s. Mass accuracy and reproducibility were maintained by using an independent reference spray by the LockSpray interference. The analytical column was a 100 mm × 2.1 mm id, 1.7 µm C8 Acquity UPLC BEH (Waters). The two mobile phases were phase A: MeOH/H 2 O/HCOOH (74:25:1 v/v/v); phase B: MeOH/HCOOH (99/1 v/v), both also containing 5 mM ammonium formate. A linear gradient was programmed as follows: 0.0 min: 80% B; 3 min: 90% B; 6 min: 90% B; 15 min: 99% B; 18 min: 99% B; 20 min: 80% B, at 0.3 ml/min fl ow rate. The column was held at 30°C. Quantifi cation was carried out using the extracted ion chromatogram of each compound, using 50 mDa windows. Linear dynamic range was determined by injecting standard mixtures, and positive identification of compounds was based on accurate mass measurement (< 5 ppm error) and LC retention time compared with that of a standard (± 2%).

HPLC/fl uorescence detection
These analyses were carried out in an Alliance Waters 2695 HPLC system coupled to a Waters 2475 Multi fl uorescence detector (Waters, Milford USA) equipped with an Atlantis T3 C18 (50 mm × 4.6 mm) column (Waters). The mobile phase was composed of a mixture of acetonitrile/H 2 O (80: 20), and the fl ow rate was set at 1 ml/min. All solvents contained 0.1% trifl uoroacetic acid. Fluorescent compounds were monitored at 420/483 nm excitation/emission wavelengths. Peak quantifi cation was carried out using the Empower Pro 2.0 software (Waters).

Real-time RT-PCR
Total RNA was extracted with the RNeasy Kit (Qiagen, Venlo, Netherlands). Complementary DNAs were synthesized with the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA). Real-time quantitative PCR assays were performed on a LightCycler 480 instrument (Roche Diagnostics, Mannheim, Germany) and analyzed with the LightCycler

Immunohistochemistry
The procurement of human tissues complied with Spanish legislation regarding informed consent, privacy, and all legal requirements after approval by the Hospital Clínic Institutional Ethics Committee. Sections (2 µm thick) were obtained for immunohistochemistry either from formalin-fi xed and paraffi n-embedded tissue blocks or from tissue microarrays (TMA) built with a Manual Tissue Arrayer 1 (Beecher Instruments, Sun Prairie, WI). A total of 33 samples, containing normal, prostate intraepithelial neoplasia, and carcinomatous glands were analyzed. Tissue sections were mounted on xylaned glass slides (Thermo Scientifi c, Braunschweig, Germany) and used for immunohistochemical staining using the Bond Polymer Refi ne Detection System (Leica Microsystems, Wetzlar, Germany). Samples were deparaffi nized, antigen retrieval was performed at pH 6 for 20 min in citrate buffer, and primary antibody was incubated for 1 h at room temperature. Rabbit anti-ASAH1 (BD Transduction Laboratories, Franklin Lakes, NJ) was used at a dilution of 1/100. Staining was scored as the percentage of cells with clear positivity and the predominant staining intensity. Images were captured with an Olympus BX-51 microscope equipped with an Olympus DP70 camera.

Statistical analysis
Signifi cance was determined by the two-tailed unpaired t -test using the Graph Pad Prism 4.0 software.

Increased expression of acid ceramidase in highly metastatic clones derived from PC-3 prostate cancer cells
The PC-3 prostate cancer cell line was used to generate two distinct clonal populations. PC-3/S cells were isolated in vitro by single-cell cloning from luciferaseexpressing PC-3 cells. A second single-cell progeny, hereafter designated PC-3/Mc, was isolated from luciferaseexpressing PC-3/M cells, a PC-3 subline that had been selected in vivo for its high metastatic potential ( 21 ). Intramuscular grafting in NOD-SCID mice of 2.5 × 10 5 PC-3/Mc cells quickly produced large tumors ( Fig. 1A ) with the appearance of abdominal lymph node metastases by 19 days in 50% of mice ( 21 ). In vitro, PC-3/Mc cells grew much faster than PC-3/S cells ( Fig. 1B ). Moreover, PC-3/Mc cells were highly clonogenic, whereas PC-3/S cells showed limited anchorage-independent growth ( Fig. 1C ).
To investigate the sphingolipid profi les of both cell lines, cells were seeded (0.25 × 10 6 cells/ml) and grown under standard conditions, and their sphingolipid composition determined after 48 h of culture. LC/MS analysis showed that total ceramide abundance in PC-3/S cells was 1.3-fold that of PC-3/Mc cells ( Fig. 1D ). Interestingly, this difference increased to 2.2 for the C14 and C16 N -acyl species (PC-3/S versus PC-3/Mc ratio: 2.2). Likewise, the cell content of SM and ceramide monohexosides (CMH, including both glucosylceramides and galactosylceramides) was 1.3-1.5 times higher in PC-3/S cells than in PC-3/Mc cells ( Fig. 1E, F ), and this difference was similar for all the differently N -acylated species. No signifi cant differences between free bases and Cell viability assay MTT (3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide, Sigma) was added to cultured cells at a fi nal concentration of 0.5 mg/ml, incubated at 37°C for 3 h, and the resulting precipitates were solubilized with dimethyl sulfoxide. Absorbance was measured at 570 nm on a SpectraMax Microplate Reader (Molecular Devices, Sunnyvale, CA).
Spectroscopic data for the synthesized compounds: SABRAC :         protein and activity levels were not paralleled by signifi cant differences at the mRNA level (not shown), suggesting a posttranscriptional regulation of ASAH1 expression.

Requirement of ASAH1 for optimal growth and metastatic potential of PC-3/Mc cells
Given the above differential ceramidase activity and expression of ASAH1 protein between metastatic and nonmetastatic prostate cancer cell clonal populations and to determine its importance for the growth and metastatic properties of PC-3/Mc cells, we proceeded to stably knock down its transcript. Cells were transduced with fi ve different lentiviral constructs expressing shRNAs targeting fi ve distinct sequences on the ASAH1 mRNA. Three of these shRNAs were effective at specifi cally decreasing ASAH1 mRNA levels with variable effi cacies, ranging from 90 to 60% ( Fig. 2A ). This gene knockdown was paralleled by a corresponding inhibition of the AC activity in these cells ( Fig. 2A ). The silencing effi cacies long-chain phosphates were detected between the two cell lines (not shown).
Enzyme activity determination showed that the highly aggressive and metastatic PC-3/Mc cells displayed levels of AC activity that were 2.5 to 4 times higher than those of PC-3/S cells, determined either in cell lysates at acidic pH or in intact cells ( Fig. 1G ) using a fl uorogenic assay ( 22 ). Several fi ndings support that the bulk of the ceramidase activity detected in these cells is due to AC. First, the fl uorogenic substrates were accepted by neutral ceramidase but not by alkaline ceramidases (not shown). Second, RBM14-C8, which is a substrate of NC but not of AC, was not hydrolyzed by PC-3/Mc cells. Third, the hydrolytic activity detected in these cells was lost after treatment with SABRAC, which inhibits AC but not NC (see below). Furthermore, PC-3/Mc cells expressed 2-to 4-fold higher levels of the AC ASAH1 than PC-3/S cells, as determined by Western blotting ( Fig. 1H ). A comparative transcriptomic survey showed that the observed differences in ASAH1 that PC-3/Mc cells are resistant to the apoptotic and growth-inhibitory effects of ceramide accumulation under standard growth conditions. Consistently, the growth rate of ASAH1-knockdown PC-3/Mc cells in standard growth medium (10% FBS) did not differ signifi cantly from that of control cells when seeded at initial densities of 1,000 cells/cm 2 ( Fig. 3A ). However, when seeded at a density of 500 cells/cm 2 , ASAH1-knockdown cells showed a significantly reduced growth rate compared with control cells ( Fig. 3B ). This suggests that ASAH1 might be critically required to sense factors dependent on cell density, including paracrine factors or cell-cell interactions. In order to know whether ASAH1 knockdown sensitized PC-3/Mc cells to limiting concentrations of exogenous growth factors, their rate of proliferation was determined in medium containing 0.5% FBS. Under these conditions, ASAH1knockdown PC-3/Mc cells grew at signifi cantly slower rates of shRNAs 399 and 402 were confi rmed by Western blotting ( Fig. 2B ).
The sphingolipid content of ASAH1-knockdown PC-3/ Mc cells was analyzed by UPLC-TOF. Both ASAH1-specifi c shRNAs caused the accumulation of ceramides, SM, and CMH compared with cells transduced with a control lentiviral vector ( Fig. 2C-E ), indicating an impairment of ceramide catabolism, which confi rms the functionality of the knockdowns. Unexpectedly, sphingosine was increased in both knockdown clones ( Fig. 2F ), which suggests that other ceramidases are upregulated upon chronic knockdown of ASAH1 ( 24 ).
We next tested if this accumulation of ceramides provoked an impairment of the growth or viability of ASAH1knockdown PC-3/Mc cells. Neither of the two knockdowns had any effect on the PC-3/Mc population, nor did they cause the accumulation of sub-G1 populations, suggesting PC-3/Mc cells to form tumors in NOD-SCID mice, with independence of the initial inoculum size ( Fig. 4A ), and it signifi cantly delayed their ability to colonize lungs upon intravenous injection ( Fig. 4B ).
Therefore, although PC-3/Mc cells are resistant to the effects on survival or the cell cycle caused by ceramide accumulation following knockdown of ASAH1, they display a strong dependence on ASAH1 for the maintenance of key properties associated with tumor-initiating cells, namely, anchorage-independent growth in vitro and tumorigenesis and lung colony formation in immunodefi cient mice.

Development of novel, highly specifi c acid ceramidase inhibitors
The above results support the importance of ASAH1 for the self-renewal and metastatic phenotypes of PC-3 prostate cancer cells, and validate it as a potential therapeutic than control cells at all initial seeding densities, with more pronounced effects at the lowest initial seeding density studied ( Fig. 3C, D ).
In agreement with the weak effect of ASAH1 knockdown on the growth of PC-3/Mc cells in standard culture conditions, only one of the two ASAH1-specifi c shRNAs caused a modest decline of the G1 population with an increased G2-M population ( Fig. 3E ). The ability of cells to form colonies in anchorage-independent conditions is a measure of their self-renewal potential and is closely related to their tumorigenic and metastatic capacities in vivo ( 25 ). As described above, the metastatic PC3/Mc cells were strongly clonogenic, as opposed to the nonmetastatic PC-3/S cells. Knockdown of ASAH1 completely abolished the clonogenic capacity of PC-3/Mc cells ( Fig. 3F ), indicating that ASAH1 is required for their self-renewal. Consistently, knockdown of ASAH1 strongly inhibited the ability of Neither SABRAC nor RBM1-12 inhibited the activity of the cysteine protease papain, whereas chymostatin completely blocked papain activity at the standard concentrations (supplementary Fig. IIA). On the other hand, at conditions under which SABRAC inhibited AC, AC activity was unaffected by the cysteine protease inhibitor E64d (supplementary Fig. IIB).

Effects of AC inhibitors on the growth of PC-3/Mc cells
We next tested the effects of RBM1-12, RBM1-13, and SABRAC on the growth properties of PC-3/Mc cells. First, the inhibitory potency of these compounds on AC was determined by incubating the cells with increasing doses for 48 h, and then determining AC activity in cell lysates using a fl uorogenic assay. Under these conditions, the inhibition of AC by RBM1-12 and SABRAC was dose-dependent (RBM1-12: 1 µM < IC 50 < 5 µM; SABRAC: IC 50 < 1 µM) ( Fig. 6A ). Surprisingly, RBM1-13, which had shown to be a good AC inhibitor in FD10X intact cells and cell lysates ( Fig. 5B ), did not display any AC inhibitory effect in PC-3/ Mc cells. In fact, RBM1-13 tended to enhance the activity of AC in these cells ( Fig. 6A ). We do not know the reason for this striking difference in the effects of RBM1-13 between these two cell lines, although differential metabolization or uptake of the compounds in the prostate cancer cell lines tested may contribute to the discrepancy. The sphingolipid profi les of PC-3/Mc cells after treatment with these compounds refl ected their AC inhibitory activities. Thus, both RBM1-12 and SABRAC induced an accumulation target in hormone-insensitive, metastatic prostate cancer. Several AC inhibitors have been synthesized and tested with potencies in the low micromolar range. Considering that AC is a cysteine hydrolase, a small family of ceramide analogs modifi ed at the amide linkage with thiol reactive functions was generated and tested. These compounds were inspired in reported cysteine protease inhibitors ( 26 ) and include two ␣ -haloamides and several ␣ , ␤ -unsaturated amides as Michael acceptors. Their structures are shown in Fig. 5A .
The RBM1 series of compounds was synthesized by Nacylation of dihydrosphingosine following standard procedures. All compounds were tested in intact FD10X cells as well as in cell lysates at pH 4.5 following the reported fl uorogenic assay ( 22,27 ). The best inhibitors in intact cells were compounds RBM1-12, RBM1-13, RBM1-18, and SABRAC, with percentages of inhibition ranging from 50 to 70% ( Fig. 5B ). Whereas RBM1-12, RBM1-13, and SABRAC maintained their inhibitory activities in the in vitro assay, RBM1-18 had no inhibitory activity in cell lysates ( Fig. 5B ). Therefore, compounds RBM1-12, RBM1-13, and SABRAC were selected for further studies. To assess their specifi city, the three compounds were tested for their effects on NC using the standard substrate ( 28 ) and on FD cells transiently transfected with the ASAH2 gene. None of the compounds inhibited NC ( Fig. 5C ), attesting to their specifi city as inhibitors of acid ceramidase. In vitro dose-response determinations showed that SABRAC was the best inhibitor, with an IC 50 value of 52 nM, followed by RBM1-12 (IC 50 = 0.53 µM) and RBM1-13, which exhibited the lowest potency (IC 50 = 11.2 µM) ( Fig. 5D ). Furthermore, both in the presence of SABRAC or RBM1-12, the enzyme activity facilitate the removal of the ceramides that accumulate as a consequence of ASAH1 silencing. The short-term inhibition of AC by chemical inhibitors, such as in the above experiments, would not allow suffi cient time to trigger such hypothetical long-term adaptive responses.
These compounds were assessed for their effects on the growth of PC-3/Mc cells. At a concentration of 5 µM and in medium containing 10% FBS, the growth of PC-3/Mc cells was significantly slowed by all the compounds, although RBM1-13 showed the least growth inhibitory activity at 1 µM ( Fig. 6C ). At 5 µM, the two AC inhibitors active on PC-3/Mc cells, RBM1-12 and SABRAC, completely abolished the ability of PC-3/Mc cells to form colonies in anchorage-independent conditions, while RBM1-13 did not show signifi cant effects in these assays ( Fig. 6D ). None of the compounds had signifi cant effects at a concentration of 1 µM, and similar to knockdown of ASAH1, none of ceramides ( Fig. 6B ), whereas, as expected for its lack of AC inhibitory activity in PC-3/Mc cells, RBM1-13 did not signifi cantly alter the abundance of ceramides.
The abundance of complex sphingolipids such as SM or CMH was not signifi cantly affected by any of the compounds, except for the levels of CMH in cells treated with RBM1-12 at 5 µM, which increased 50% over controls ( Fig. 6B ). This is in contrast with the effect of stably knocking down ASAH1 in PC-3/Mc cells, which caused the accumulation of these complex sphingolipids ( Fig. 2D, E ). In a second discrepancy between transcript knockdown and chemical inhibition of AC, levels of sphingosine did not change by treatment of PC-3/Mc cells with either RBM1-12 or SABRAC (data not shown), whereas they increased upon ASAH1 knockdown. We speculate that prolonged knockdown of ASAH1 may permit an adaptive response of cells by upregulating other ceramidases to In summary, similar to ASAH1 knockdown, two of the newly developed AC inhibitors, RBM1-12 and SABRAC, showed strong inhibitory activities on AC activity and on the growth and clonogenic capacities of the highly metastatic PC-3/Mc cells. Interestingly, SABRAC displayed strong growth inhibitory effects on PC-3/Mc cells, while exhibiting very limited cytotoxicity. A third compound, RBM1-13, which had AC inhibitory activity in other cells, did not inhibit AC in PC-3/Mc cells, and although it exerted cytotoxic activity on these cells, it had less potent inhibitory activity on the growth of these cells on plastic and showed no effect on their capacity to grow in anchorage-independent conditions.

Expression of ASAH1 in nonmetastatic and metastatic prostate cancer
It has previously been reported that ASAH1 is expressed in a number of tumor types, including PC, at levels higher than in normal tissues ( 17 ). However, thus far its expression in PC has not been reported by immunohistochemistry, which permits to correlate staining intensities (as surrogates of expression levels) with morphological parameters. We used an ASAH1-specifi c antibody ( Figs. 1H and 2B ) to analyze by immunohistochemistry a total of 33 samples from prostate cancer patients that contained tumoral glands, glands with normal morphologies, and of the inhibitors caused signifi cant changes in the cell cycle profi le of PC-3/Mc cells (data not shown).
Next, the dose-dependent cytotoxic activity of these compounds on PC-3/Mc cells was evaluated after 72 h incubations. The most toxic compound was RBM1-13 (CC 50 = 10.0 µM), followed by RBM1-12 (CC 50 = 28.2 µM) and SABRAC (CC 50 > 300 µM) (supplementary Fig. III). Because RBM1-13 did not show a detectable AC inhibitory activity in PC-3/Mc cells, its cytotoxic activity on these cells must be caused by off-target effects. The nonmetastatic PC-3/S clone was more sensitive to the cytotoxic effects of SABRAC than were the isogenic, metastatic PC-3/Mc cells, (CC 50 = 25.3 µM). In contrast, both clones exhibited similar sensitivities to RBM1-12 and RBM1-13 (CC 50 = 37.4 µM and 14.6 µM, respectively) (supplementary Fig. III). The cytotoxic activity of these compounds was tested on additional pairs of cell lines with different tumorigenic or metastatic potentials, including the transformed prostate epithelial cells RWPE-2 (tumorigenic) versus RWPE-1 (nontumorigenic), the lung cancer cell lines HAL8 (metastatic) versus HAL24 (nonmetastatic), and the breast cancer cell lines MDA-MB-231 (metastatic) versus MCF-7 (nonmetastatic). The less aggressive cell lines were generally more sensitive to the cytotoxic activity of these acid ceramidase inhibitors than were their more aggressive partners (supplementary Fig. III).  (3 × 10 3 ) were cultured in soft agar in the presence of 5 µM of the test compounds or vehicle alone, and then stained with crystal violet after three weeks. Images were captured and processed with ImageJ, and colonies у 0.2 mm diameter were scored. Data are represented as the mean of triplicates ± SD. with histological grade (Gleason score), stage, or the presence of lymph node metastasis. However, the observation that strong expression of ASAH1 was associated with the preneoplastic PIN lesions more than with normal glands and in turn with prostate adenocarcinoma more than with PIN lesions suggests that ASAH1 expression levels tend to increase during PC progression.

DISCUSSION
Advanced, hormone-independent, castration-resistant prostate cancer represents a devastating form of the disease that frequently develops from initially less aggressive tumors and shows resistance to conventional chemotherapeutic agents (18)(19)(20). In this evolutive process, the progressive dominance in the tumor of cancer stem cells, endowed with high survival and low drug sensitivity ( 25,29 ), is emerging as crucial.
Here, we studied the clonal population PC-3/Mc, derived from prostate cancer cells and highly enriched in tumor-initiating cells ( 21 ), and found that its growth and clonogenic potential are extremely sensitive to the knockdown of ASAH1 or chemical inhibition of AC activity. In addition, knockdown of ASAH1 in these cells strongly inhibited their capacity to grow tumors upon local implantation or to colonize lungs after intravenous injection. Although the growth of PC-3/Mc cells under adherent conditions was also affected by ASAH1 knockdown, it was made most evident only upon deprivation of growth factors, suggesting that ASAH1 is required for optimal glands with characteristics of the preneoplastic lesion prostate intraepithelial neoplasia (PIN). All epithelial structures stained for ASAH1, albeit at varying intensities ( Fig. 7A -C ). In a range of staining intensities from 1 to 3, most normal or neoplastic structures were given an intensity score of 2 (72.7% of normal glands, 75% of PIN, and 63.6% of tumoral glands; Fig. 7D and supplementary Table I). However, only 1 out of 22 evaluable cases with normal glands was scored as intensity 3 (4.5%), while 4 of 24 (16.7%) of PIN glands and 10 of 33 (30.3%) of tumoral glands were scored as intensity 3 ( Fig. 7D and supplementary Table II). In those cases in which normal and tumoral glands could be evaluated simultaneously in the same sample (22 cases), the ASAH1 staining intensities of normal versus tumoral glands were scored as equal in 10 cases (45.5%), the staining intensity was scored as stronger in tumoral glands versus normal glands in 8 cases (36.4%), and in 4 cases (18.2%), the maximum intensity in tumoral glands was scored as lower than the maximum intensity in normal glands ( Fig. 7E and supplementary Table II). In several cases, the staining intensity for ASAH1 in tumoral areas was clearly stronger than the staining in adjacent normal glands ( Fig. 7C, E ).
Thus, strong ASAH1 immunostaining in human prostate tissues tended to be associated with prostate adenocarcinoma ( Fig. 7D ), an observation that complements previous reports in which immunohistochemical analysis was not performed, and thus the specifi c cell type, either epithelial or stromal, expressing ASAH1 was not determined ( 17 ). On the other hand, with the cases studied here, stronger ASAH1 staining did not show correlations Because of this different enzyme compartmentalization, the increase in SM and CMH is unlikely to arise from augmented synthesis from the lysosomal ceramide. A plausible explanation involves downregulation of lysosomal acid sphingomyelinase and glucocerebrosidase as a result of intralysosomal ceramide buildup. Interestingly, the sphingolipid profi le of the PC-3/Mc cells knocked down for AC is similar to that seen in PC-3/S cells. This fi nding supports that the accumulation of sphingolipids inside the lysosome as a result of different AC activities in both clones is related to the different aggressiveness of the two phenotypes.
The growing recognition of AC as a potential therapeutic target in cancer has encouraged the development of AC inhibitors ( 12 ). AC belongs to the N -terminal nucleophile (Ntn) hydrolase family ( 48 ). This family of enzymes shares the common feature of having an N -terminal nucleophile, which is generated by autoproteolytic processing ( 49 ). The Cys143 nucleophilic thiol in AC is exposed at the N -terminus of the ␤ -subunit after cleavage of the precursor protein. Therefore, AC belongs to the same subcategory of Ntn hydrolases as the cysteine proteases. Based on this relationship, AC has been recently found to be inhibited by the cysteine protease inhibitors, cystatins ( 50 ). During the course of this study, we identifi ed novel potent and specifi c inhibitors of AC within a series of small molecules inspired in reported irreversible cyteine protease inhibitors (51)(52)(53). These compounds feature either an ␣ -halocarbonyl unit or an ␣ , ␤ -double-bond Michael acceptor moiety. The fi rst screening in cells overexpressing AC showed that, among the compounds prepared, the ␣ -bromoamides RBM1-12 and SABRAC were the most potent inhibitors. Within the ␣ , ␤ -unsaturated amides, only the methacrylamide (RBM1-13) elicited AC inhibitory activities both in intact cells and cell lysates. None of the compounds was active on NC, consistent with their conception as thiol-targeting molecules. The low potencies of the other ␣ , ␤ -unsaturated amides relative to RBM1-13 suggest that substitution at the ␤ -position hinders the attack of the enzyme nucleophilic thiolate. Surprisingly, acrylamide RBM1-19 had no activity as AC inhibitor. Although the reasons for this fi nding have not been investigated, it is possible that the high reactivity of unsubstituted acrylamides results in alternative reactions of RBM1-19 before reaching the AC target. The activity of the bromoamides is especially relevant. SABRAC and RBM1-12 are among the most potent AC inhibitors so far reported ( 54 ), with IC 50 values of 52 nM and 530 nM, respectively, as assayed with lysates of cells overexpressing AC. Time dependence of inhibition supported that both SABRAC and RBM1-12 are irreversible inhibitors, which agrees with the expected mechanism considering their structure and the involvement of a nucleophilic cysteine residue in the catalytic site. Furthermore, they are inactive over both NC and papain, a cysteine protease, thus supporting their selectivity for AC.
Both SABRAC and RBM1-12 were also potent inhibitors of AC in intact PC-3/Mc cells. Surprisingly, RBM1-13, which inhibited AC in FD10X cells, failed to inhibit AC in growth of PC-3/Mc cells under limited growth factor supply but is dispensable for adherent growth when growth factors are present at high concentrations. These results, together with the strong phenotype that we observed in anchorage-independent growth assays, which correlates with selfrenewal, tumorigenic, and metastatic potentials ( 21,25 ) for both ASAH1 knockdown and AC inhibition, suggest that AC is required for self-renewal and growth factor signaling in drug-resistant and metastatic PC-3/Mc prostate cancer cells.
That ASAH1 knockdown and inhibition of AC produces a specifi c effect on a particular growth property of PC-3/ Mc cells, namely, anchorage-independent spheroid formation, is further supported by the absence of signifi cant cell death or the accumulation of sub-G1 cell populations, in spite of the expected accumulation of ceramides. Additionally, neither ASAH1 knockdown nor AC inhibition caused a signifi cant sensitization of PC-3/Mc cells to drugs used in advanced prostate cancer therapy, including docetaxel, doxorubicin, and etoposide (supplementary Fig. IV), which confi rms the chemoresistance of this highly aggressive subpopulation of prostate cancer cells enriched in tumor-initiating cells unaffected by the accumulation of ceramides. The failure of AC inhibition or ASAH1 knockdown to chemosensitize PC-3/Mc cells differs from the sensitization found in various other tumor cells (30)(31)(32)(33)(34)(35)(36)(37)(38)(39). In PC-3/Mc cells, AC inhibition or ASAH1 knockdown may upregulate specifi c pathways to metabolize the resulting ceramide excess. Such pathways include phosphorylation, glycosylation, and conversion into SM. The possibility of phosphorylation as a tumor death escape route ( 40 ) does not seem plausible, as no ceramide-1phosphate was ever detected in PC-3/Mc cells, regardless of any genetic manipulation or chemical treatment. In contrast, most SM and CMH species are signifi cantly higher in ASAH1-knockdown cells than in mock cells. Although SM synthases have been seldom addressed as targets to overcome resistance ( 41,42 ), the usefulness of inhibition of glucosyltransferases as a means to exploit ceramide as an antitumor agent has been extensively documented ( 43,44 ). Furthermore, chemoresistance is not only the result of ceramide clearance by glycosylation but also the increased glycosylation products themselves have been reported to upregulate the expression of multidrug resistant protein 1 (MDR1) through c-Src kinase and ␤ -catenin signaling ( 45 ). Whether UDP-glucose ceramide glucosyltransferase (UGCG) or MDR1 is upregulated upon ASAH1 silencing in PC-3/Mc cells has not been investigated. However, the UGCG transcript levels are higher in PC-3/S than in PC-3/Mc cells (unpublished observations), as opposed to their AC activity.
Nevertheless, the signifi cant increase in SM and CMH in PC-3/Mc cells knocked down for AC deserves comment. Because AC is a lysosomal enzyme and ceramides cannot exit from the lysosome ( 46 ), a lack of AC activity leads to an intralysosomal accumulation of ceramide. However, sphingomyelin synthase 2 is localized in the plasma membrane, whereas sphingomyelin synthase 1 and glucosylceramide synthase reside in the Golgi apparatus ( 14,47 ). the PC-3/Mc cell line. This difference is likely due to the different incubation times (FD10X, 4 h; PC3Mc, 48 h) pointing to the metabolization of RBM1-13 after long-term incubations. In contrast, the bromoamides maintained their inhibitory activity at long incubation times. In agreement with their AC inhibitory activity, both SABRAC and RBM1-12 induced a buildup of ceramides, which was signifi cant compared with controls for all the different N -acyl species. However, levels of sphingosine remained unaffected. The latter results, although unexpected for AC inhibitors, are not unprecedented. Bielawska et al. ( 55 ) reported that the sphingolipid profi les of MCF-7 cells treated with different ceramidase inhibitors derived from B13 and D-e-MAPP followed different patterns depending on the chemical function substituted for the original amide. In the case of an N -alkyl analog of D-e-MAPP, namely, compound LCL284, ceramides accumulated, but long chain bases did not change signifi cantly. A plausible explanation is the triggering of compensatory mechanisms to keep bioactive sphingolipids at nonlethal levels. One such mechanism could involve an increased activity of other ceramidases. A similar scenario might lead to the slight increase of sphingosine found in PC-3/Mc cells knocked down for AC. In support of this hypothesis, Hu et al. ( 24 ) demonstrated that knockdown of the alkaline ceramidase 3 upregulated the expression of the alkaline ceramidase 2 with increases of both sphingosine and its phosphate.
Our fi ndings support the notion that ceramidase metabolism ( 56 ) and, more specifi cally, acid ceramidase activity, are critical regulators of the self-renewal, tumorigenic, and metastatic potentials of cancer stem cells, represented by the PC-3/Mc population ( 21 ), beyond their known regulation of signals that tilt the balance between cell survival and death. Furthermore, we argue that our cellular model, specifi cally selected for subpopulations with strong selfrenewal and aggressive phenotypes, allows to better address therapeutic strategies against advanced cancer ( 45 ). The precise mechanism linking the elevated AC activity and the increased aggressiveness of PC-3/Mc cells compared with the PC-3/S clone is under investigation.