Using a biologically annotated library to analyze the anticancer mechanism of serine palmitoyl transferase (SPT) inhibitors

Mechanistic understanding is crucial to anticancer drug discovery. Here, we reveal that inhibition of serine palmitoyl transferase (SPT), the rate‐limiting enzyme in sphingolipid synthesis, induced death in a lung cancer cell line via a necrosis‐dependent pathway. To elucidate the mechanism of cell death induced by SPT inhibition, a biologically annotated library of diverse compounds was screened with an SPT inhibitor. This analysis identified suppressors of SPT inhibitor‐mediated cell death. Further analysis using hit compounds from this screening revealed that SPT inhibitors induce COX‐2 expression, leading to necrosis‐dependent cell death. SPT inhibitors might therefore represent novel candidates for cancer therapy via necrosis pathway regulation. Our data illustrate that compound combination screening of biologically annotated libraries could be used for mechanistic elucidation.

Mechanistic understanding is crucial to anticancer drug discovery. Here, we reveal that inhibition of serine palmitoyl transferase (SPT), the rate-limiting enzyme in sphingolipid synthesis, induced death in a lung cancer cell line via a necrosis-dependent pathway. To elucidate the mechanism of cell death induced by SPT inhibition, a biologically annotated library of diverse compounds was screened with an SPT inhibitor. This analysis identified suppressors of SPT inhibitor-mediated cell death. Further analysis using hit compounds from this screening revealed that SPT inhibitors induce COX-2 expression, leading to necrosis-dependent cell death. SPT inhibitors might therefore represent novel candidates for cancer therapy via necrosis pathway regulation. Our data illustrate that compound combination screening of biologically annotated libraries could be used for mechanistic elucidation.
Cancer is a major public health program worldwide, and accordingly, pharmaceutical companies aim to develop novel anticancer-related drugs. Although both genetic and environmental factors are closely related to cancer development and are responsible for some portion of cancer progression, many as yet undescribed factors are also involved in cancer progression. In addition to conventional medical and radiation treatments, molecular targeted therapy has recently become popular in the drug discovery process. Especially, developing anticancer drugs regulating metabolic pathways that are selectively activated in cancer cells represent a new promising approach to cancer therapy.
Cancer metabolism is the focus of current and emerging therapeutic approaches to anticancer drug discovery [1][2][3]. The best known example of a metabolic shift in cancer cells is the Warburg effect [4]. Cancer cells tend to depend on the glycolytic pathway rather than the tricarboxylic acid (TCA) cycle in order to generate energy more efficiently in a hypoxic microenvironment. Recent metabolomics technology research has revealed additional metabolic pathways that are closely related to cancer cell growth. Newly identified cancer metabolism-related targets, such as isocitrate dehydrogenase 1 (IDH1) and HMG-CoA reductase, are now considered promising anticancer drug targets [5,6].
Serine palmitoyl transferase (SPT) mediates the conjugation of serine and palmitoyl CoA to form ceramide and represents a rate-limiting step in sphingolipid synthesis. Ceramide is a well-known cytotoxic lipid that under normal conditions, is readily transferred from the endoplasmic reticulum (ER) to the Golgi by the ceramide transfer protein CERT, where it undergoes further synthesis to glucosylceramide, sphingomyelin, and sphingosine-1-phosphate [7,8]. Abnormal sphingolipid metabolism has been observed in several types of cancer cells. In head and neck cancer, ceramidase overexpression enhances resistance to Fas ligandmediated apoptosis, and in various solid cancers, sphingosine kinase 1 overexpression leads to enhanced proliferation [9][10][11].
Although the metabolomics approach to cancer drug discovery works well and has led to the identification of new anticancer drug targets, the relationship between metabolic alterations and cancer cell growth is not always clear, and it is therefore critical for drug discovery researchers to understand the mechanisms of action (MOA) for such drugs. There are two commonly accepted methods for analyzing the MOA of cancer drugs. One method involves a targetspecific, hypothesis-based approach that combines known information with newly obtained data from transcriptome and metabolome analyses. The other method involves a discovery-based approach involving a functional genomics analysis of whole-genome siRNA or shRNA [12,13]. Functional genomics analyses have provided many novel possibilities with regard to target relationships and thus represent a powerful approach for the identification of relatively novel targets. However, these unbiased siRNA-or shRNA-based approaches share the fundamental challenges of off-target effect [14,15], knockdown efficiency, protein turnover, and compensatory reactions [16]. The establishment of alternative methods would be valuable to our understanding of the MOA of anticancer drugs.
Biologically annotated library screening is currently attracting considerable interest as a straightforward approach to phenotypic drug discovery [17][18][19][20][21][22]. This approach allows us to easily link target molecules with disease phenotypes and to generate hypotheses regarding the underlying biological mechanisms. Unlike siRNA or shRNA, small molecules directly inhibit or activate target protein, independent from expression level and turnover rate of target protein. Moreover, it is noteworthy that tool compounds collected from a biologically annotated library are optimized to enhance not only the potency against target protein but also target selectivity. Therefore, we hypothesized that MOA analysis of anticancer drugs by using biologically annotated library could become complementary methods for functional genomics.
In this study, we demonstrate that the inhibition of SPT, the rate-limiting enzyme in sphingolipid synthesis, inhibits the growth of lung cancer cells. We also describe the MOA of SPT inhibitors through a combination study involving a biologically annotated library and SPT inhibitors.

Compounds
Biologically annotated compounds were collected to create a screening compound library. SPT inhibitors were synthesized at Takeda Pharmaceutical Company, Ltd. (Fujisawa, Japan) [23].
Preparation of human SPT2 enzyme PCR with specific primers was used to generate cDNAencoding human SPT2, and the transcript was subsequently subcloned to generate expression vectors. For preparation of the SPT2 enzyme, FreeStyle293 cells were transfected with human SPT2 expression plasmids and cultured for 3 days. Cells were then homogenized in 50 mM HEPES buffer (pH 7.5) containing 250 mM sucrose, 5 mM EDTA, 5 mM DTT, and Complete, EDTA-free (Roche Applied Science, Penzberg, Upper Bavaria, Germany). Cell homogenates were centrifuged, and supernatants were harvested. Total membrane fractions were isolated by ultracentrifugation. Pellets were resuspended in 50 mM HEPES buffer (pH 7.5) containing 5 mM EDTA, 5 mM DTT, and Complete, EDTA-free. Cell lysates were stored at À80°C. The protein concentration was determined with using the CBB Protein Assay.

Enzyme assay
Enzyme reactions were run in 20 lL volumes with assay buffer comprising 100 mM HEPES (pH 8.0), 2.5 mM EDTA, 5 mM DTT, and 0.01% bovine serum albumin (fatty acid-free), and conducted in a 384-well assay plate. Briefly, 5 lL of a tested compound and 10 lL of 100 lgÁmL À1 SPT2-expressed membrane dissolved in assay buffer were mixed and incubated for 60 min. Subsequently, 5 lL of a substrate solution containing 2 mM -serine and 20 lM palmitoyl-CoA in assay buffer was added to start the enzyme reaction. After a 15-min incubation period at room temperature, the reaction was terminated by adding 20 lL of 2% formic acid. Finally, 40 lL of acetonitrile containing 600 nM C17-sphinganine was added as an internal standard. High-throughput online solid-phase extraction was performed using a RapidFire Ò 300 device (Agilent Technologies, Santa Clara, CA, USA). Mass spectrometric analysis was performed using an API-4000 TM triple quadrupole mass spectrometer (AB SCIEX, Framingham, MA, USA) in positive SRM mode. The SRM transitions for 3ketodihydrosphingosine (reaction product) and C17-sphinganine were set to 300.5/270.3 and 288.4/60.2, respectively. Analytical data were acquired using ANALYST software, version 1.5.0 (AB SCIEX), and 300. 5

Growth inhibition assay
HCC4006 cells were dispensed into a 384-well culture plate at a density of 250 cells per well in 40 lL of culture medium and cultured overnight. Subsequently, the cells were treated with 10 lL of a tested compound and cultured for 5 days. The medium was then removed and replaced with 30 lL of CellTiter Glo Luminescent Cell Viability Assay reagent (Promega, Fitchburg, WI, USA). Luminescence was measured on an EnVision device (PerkinElmer, Waltham, MA, USA). The IC 50 values for test compounds were calculated using GRAPHPAD PRISM 5.0 (GraphPad Software, San Diego, CA, USA).

Lactate dehydrogenase release
HCC4006 cells were seeded in black 384-well plates and treated with compounds for 96 h. From each well, 20 lL of cell culture medium was transferred to a 384-well clearbottomed plate (#3680, Corning Corp.); CytoTox 96 Non-Radioactive Cytotoxicity Assay (Promega Corp.) reagent was added to each well, followed by a 30-min incubation period; cell variability was measured on a Spectramax Paradigm multiplate reader (Molecular Device Corp., Sunnyvale, CA, USA).

Combination screening
HCC4006 cells were seeded in black 384-well plates and pretreated with Biologically annotated compounds for 1 h, after which SPT inhibitors were added for 120 h. CellTiter Glo was added to each well, and cell variability was determined by measuring the firefly luciferase intensity on an EnVision device.

Pathway enrichment analysis
We used the Ingenuity Pathway Analysis (IPA) system for canonical pathway enrichment analysis to perform functional enrichment tests of target candidate genes linked to the hit compounds.

Statistical analysis
Values are presented as means AE SD. Statistical significance among groups was determined using an ANOVA followed by Dunnett's test. A P value <0.05 was considered statistically significant.

SPT inhibitors attenuate lung cancer cell growth
Previous studies suggested that SPT inhibition suppressed the growth of both melanoma and lung cancer cells [24,25]. We found that the lung cancer cell line HCC4006 was sensitive to myriocin, a known SPT inhibitor (Fig. 1A). Therefore, we synthesized 137 pyrazolopyridine derivatives as SPT inhibitors and used these to validate the relationship between in vitro SPT activity inhibition and cancer cell growth. We confirmed that the inhibition of HCC4006 cancer cell growth correlated well (R 2 = 0.87) with the in vitro inhibition of SPT2 enzyme activity, suggesting that SPT inhibition is responsible for HCC4006 cancer cell growth inhibition (Fig. 1B). In this study, we utilized compound 1 as a chemical probe against SPT. inhibited SPT2 with an IC 50 value of 0.8 nM in an in vitro enzyme assay and suppressed HCC4006 cell growth with an IC 50 value of 59 nM (Fig. 1C).

SPT inhibitor induces necrosis-dependent cell death in HCC4006 cells
Although SPT inhibition was shown to induce growth inhibition in HCC4006 cancer cells, the underlying MOA remained unclear. Cell death can be largely classified as follows, according to morphological and biochemical characteristics: apoptosis or programmed cell death; nonapoptotic cell death such as necrosis; and ferroptosis, an increasingly recognized and well-regulated cell death mechanism [26,27]. To understand the MOA, we examined which types of cell death were induced by SPT inhibitors using a well-characterized assay system and specific inhibitors against each pathway. First, effects against the apoptotic pathway were examined, as previous reports suggested that SPT inhibition induced apoptotic signals [25]. However, under our assay conditions, we confirmed that caspase 3/7 cleavage was activated by compound 1 only at concentrations exceeding 3 lM, but was not activated with myriocin treatment, suggesting that the observed caspase 3/7 activity at high concentrations of compound 1 might represent an off-target effect (Fig. S1A, Supporting Information). We also confirmed that treatment with the pan-caspase inhibitor z-vad did not attenuate SPT inhibitor-induced growth inhibition (Fig. S1B). Taken together, these observations suggest that apoptosis is not involved in SPT inhibitor-induced cell growth inhibition. Second, we evaluated whether SPT inhibitor treatment would induce necrosis. Necrosis is an apoptosis-independent cell death mechanism characterized by a disruption of the cell membrane structure and subsequent release of cellular components to the extracellular medium. Treatment with compound 1 and myriocin induced lactate dehydrogenase (LDH) release in a dose-dependent manner with respective EC 50 values of 47 nM and 0.4 nM ( Fig. 2A), indicating good agreement with the IC 50 values for cell growth (59 nM and 4 nM, respectively) and suggesting that SPT inhibition leads to necrosis. We further confirmed that the known necrosis inhibitor IM-54, which was originally identified as a suppressor of hydrogen peroxide-induced necrosis [28], attenuated SPT inhibitor-induced cell death (Fig. 2B). Third, we examined whether SPT inhibitor treatment would induce ferroptosis. Ferroptosis is a newly identified type of cell death involving the irondependent accumulation of reactive lipid species [29]. We confirmed that treatment with SPT inhibitors induced the generation of reactive oxygen species (ROS), a hallmark of ferroptosis, whereas treatment with ferrostatin-1, a well-characterized ferroptosis inhibitor, did not attenuate Compound 1-induced cell growth inhibition (Fig. S1C,D). These data suggest that ROS generation is a secondary effect of SPT inhibitor treatment and that SPT inhibitor-induced cell growth inhibition is independent of ferroptosis. These results collectively indicate that SPT inhibitors suppress cell growth via the necrotic pathway.

Compound combination screening using a biologically annotated library with SPT inhibitors
A recent study illustrated that functional genomics studies involving siRNA or shRNA could be a useful  approach to the elucidation of unknown MOA of targeted compounds [30]. However, suppression of a single gene might be overwhelmed by the compensatory activity of structurally related subtypes [16] and, for siRNA studies in particular, the efficiency of knockdown varied according to the target protein and, in most cases, partial knockdown did not affect the desired phenotype; in addition, off-target effects of siRNA are frequently observed [14,15]. To overcome these obstacles, we performed an unbiased combination study using a biologically annotated library with SPT inhibitors. The concept of a biologically annotated library has been proposed by several pharmaceutical companies [17][18][19][20]. We collected approximately 3000 compounds to form our biologically annotated library. Our criteria for the selection of compounds were in vitro pharmacological activity with IC 50 or EC 50 value of less than or equal to 1 lM on each target protein, which is based on the results of cell-free and cell-based assays with multiple types such as functional and binding assays, as shown in Fig. 3A and Table S1. Consequently, our biologically annotated compound library targets approximately 1500 unique proteins, each of which is often annotated by multiple compounds to avoid the misinterpretation of the results caused by off-target effects of small molecules. In fact, 70% of target protein information is annotated by two or more compounds. The remaining 30% covered by a single compound for each is still included, since functional genomics approaches (e.g., CRISPR, shRNA) would strengthen the hypothesis derived from the small molecule-based approaches. In our combination study, the SPT inhibitor concentration was set to 1 lM, which was expected to exhibit maximal growth inhibitory activity when used with biologically annotated library compounds at a concentration of 3 lM; the latter was expected to fully regulate the target protein activity. We screened biologically annotated library at 3 lM concentration with or without 1 lM SPT inhibitor (Fig. 3B). We identified 33 hit compounds that mitigated SPT inhibitor-induced cell death (Fig. 3C and Table 1).

Upregulation of COX-2 expression triggers necrosis in SPT inhibitor-treated cells
Pathway enrichment analysis, using IPA pathway enrichment software, was performed to reveal essential pathways related to SPT inhibitor-induced cell death, and 18 pathways were nominated as candidate pathways (Fig. 3D). We focused on the prostanoid biogenesis pathway because we noticed that 4 of the 33 hit compounds were related to COX-2, which catalyzes the conversion of arachidonic acid to prostanoid. Two selective COX-2 inhibitors, celecoxib and rutaecapine, are included in this category [31], and were found to dose-dependently attenuate Compound 1-mediated growth inhibition (Fig. 4A). COX-2 is an inducible family protein that is expressed at low levels under basal conditions; expression of this protein can be induced by a particular stimuli, leading to the generation of prostaglandin products [32,33]. We examined whether treatment with SPT inhibitors would induce COX-2 expression, thus validating our combination library-screening findings, and confirmed that treatment with SPT inhibitors induced COX-2 expression after 96 h (Fig. 4B). We also confirmed that beclomethasone and flumethasone suppressed SPT inhibitor-mediated cell death (Fig. 4C). These two compounds were previously reported as suppressors of COX-2 expression [34]. These results strongly suggest that Compound 1-induced cell growth inhibition is mediated by COX-2 function. Next, we reanalyzed the results of our biologically annotated library screening and found that JZL184 suppresses Compound 1-induced cell growth inhibition. JZL184 is an irreversible inhibitor for monoacylglycerol lipase (MAGL), the primary enzyme responsible for degrading the endocannabinoid 2-arachidonoylglycerol (2-AG) to arachidonic acid [31]. We measured the inhibitory activities of two other lipase inhibitors, CHEMBL130098 and CEHMBL1082517 against MAGL because arachidonic acid metabolism is likely to be the key pathway for Compound 1-induced cell growth inhibition. As expected, both compounds inhibited in vitro MAGL enzyme activity (Table S2).
These results indicate that seven compounds out of 33 hit compounds identified via combination library screening were related to arachidonic acid metabolism. These observations strongly support the validity of our biologically annotated library-screening strategy. Finally, we performed a knockdown experiment using COX-2 or MAGL siRNA to exclude the possibility of off-target effect of annotated compounds. Treatment of COX-2 or MAGL siRNA suppressed Compound 1-induced cell growth inhibition (Fig. S2), and these results also support the importance of COX-2-and MAGL-related pathway for SPT inhibitor-mediated cancer cell death. In summary, we conducted a combination screening of compounds using a biologically annotated library to reveal the MOA of SPT inhibition. Accordingly, we found that COX-2 expression was upregulated by SPT inhibition. Although the mechanism by which COX-2 expression is induced remains unclear, COX-2 induction was critical for SPT inhibition-induced cell death (Fig. 4D). Our results present the possibility that the expression level of MAGL or COX2 in lung cancer patients could be one of the candidates of the patient stratification marker. A more detailed analysis will be the subject of further study. Finally, we emphasize that our combination approach involving a biologically annotated library could be widely applicable to the investigation and discovery of the MOA of other types of anticancer drugs. Our compound combination screening using a biologically annotated library for anticancer drug MOA analysis could provide novel findings on target-related pathways and be applicable as a complementary method for functional genomicsbased MOA analysis.

Supporting information
Additional Supporting Information may be found online in the supporting information tab for this article: Fig. S1. (A) HCC4006 cells were treated with various concentrations of Compound 1 or myriocin for 96 h. Caspase 3/7 activity was measured using a Caspase 3/7 Glo assay. (B) HCC4006 cells were treated with various concentrations of Compound 1 or myriocin and 20 µM z-VAD for 120 h. Cellular viability was measured using CellTiter Glo. (C) HCC4006 cells were treated with various concentrations of Compound 1 or myriocin for 96 h. Intracellular reactive oxygen species (ROS) production was measured using a ROS Glo assay. (D) HCC4006 cells were treated with various concentrations of Compound 1 or myriocin with 10 µM Ferrostatin-1. Cellular viability was measured by CellTiter Glo. Fig. S2. (A) HCC4006 cells were cotreated with 6 nM COX-2, MAGL, or control siRNA with 3 µM Compound 1 for 72 h. Cellular viability was measured using CellTiter Glo. (B) HCC4006 cells were treated with 6 nM COX-2, MAGL, or control siRNA for 48 h. Cells lysates were subjected to measure expression level of COX-2 and MAGL by qPCR. Relative knockdown efficiency was calculated by delta-delta CT method. Table S1. Composition of library used for combination screening. Table S2. Summary of inhibitory activity against monoacylglycerol lipase (MAGL). Appendix S1. Materials and methods.