Acid ceramidase promotes drug resistance in acute myeloid leukemia through NF-κB-dependent P-glycoprotein upregulation

Acute myeloid leukemia (AML) is the most common acute leukemia in adults. More than half of older AML patients fail to respond to cytotoxic chemotherapy, and most responders relapse with drug-resistant disease. Failure to achieve complete remission can be partly attributed to the drug resistance advantage of AML blasts that frequently express P-glycoprotein (P-gp), an ATP-binding cassette transporter. Our previous work showed that elevated acid ceramidase (AC) levels in AML contribute to blast survival. Here, we investigated P-gp expression levels in AML relative to AC. Using parental HL-60 cells and drug-resistant derivatives as our model, we found that P-gp expression and efflux activity were highly upregulated in resistant derivatives. AC overexpression in HL-60 conferred resistance to the AML chemotherapeutic drugs, cytarabine, mitoxantrone, and daunorubicin, and was linked to P-gp upregulation. Furthermore, targeting AC through pharmacologic or genetic approaches decreased P-gp levels and increased sensitivity to chemotherapeutic drugs. Mechanistically, AC overexpression increased NF-κB activation whereas NF-kB inhibitors reduced P-gp levels, indicating that the NF-kappaB pathway contributes to AC-mediated modulation of P-gp expression. Hence, our data support an important role for AC in drug resistance as well as survival and suggest that sphingolipid targeting approaches may also impact drug resistance in AML.

the gene encoding P-gp is ABCB1. P-gp mediates efflux of a diverse range of drugs due to the conformational flexibility of its structure (9). P-gp expression strongly correlates with a drug-resistant phenotype in several different types of cancer (3,10). P-gp was reported to be abundantly expressed in AML blasts from patients with poor prognoses (5,11). Furthermore, high P-gp expression correlated with poor response to chemotherapy (12)(13)(14). While several clinical trials for direct P-gp inhibitors were conducted, the outcomes were not promising due to toxicity of the inhibitors or inadequate dosing to effectively combat drug resistance (3,15). Hence, there is an unmet need to identify novel strategies to combat drug resistance, and sphingolipidtargeting agents that impair AML blast survival as well as drug resistance may represent an alternative and efficacious approach to address this need.
Sphingolipids are a diverse group of lipids that are important for many cellular processes (16). Known functions of sphingolipids include key roles in the structural integrity of cells, signaling pathways, cell cycle progression, migration, and survival (17)(18)(19). Ceramide, a pro-death sphingolipid, is metabolized by ceramidases to sphingosine and free fatty acid. Sphingosine is then phosphorylated by sphingosine kinases to form sphingosine 1-phosphate (S1P), a bioactive molecule involved in survival mechanisms (17). We and others have shown that the dysregulation of the sphingolipid pathway, including the elevated expression of acid ceramidase (AC) and sphingosine kinase 1 in AML patient samples, promotes AML blast survival (20)(21)(22)(23). Targeting AC has been a subject of interest in several different types of cancer, including AML (24)(25)(26)(27)(28)(29). Recent studies have also shown that the sphingolipid enzyme, glucosylceramide synthase (GCS), may upregulate P-gp transcription and that targeting GCS sensitizes cells to chemotherapeutic drugs (30,31). Hence, sphingolipid enzymes may influence drug resistance in addition to cell survival.
The present study explored the hypothesis that elevated AC induces P-gp expression to increase drug resistance in AML. We demonstrated that elevated AC is associated with increased P-gp expression and confers drug resistance to AML chemotherapeutic agents. Genetic or pharmacologic manipulation of AC reduced P-gp levels and sensitized cells to chemotherapeutic drugs. We also showed that AC overexpression increased NF-B activation while exogenous S1P increased MDR1 transcript levels. Furthermore, sphingosine kinase or NF-B inhibitors suppressed P-gp levels, which suggests the involvement of this pathway in AC-mediated P-gp expression. Taken together, these studies show a novel role of AC in regulating drug resistance in AML.

Flow cytometry
To determine P-gp expression, 500,000 cells were washed twice with PBS then incubated with MDR1 antibody (UIC2; Sigma-Aldrich) for 30 min, followed by FITC-conjugated anti-mouse antibody (Ancell Corporation) for 30 min. FITC-tagged normal mouse IgG was used as control. Cells were washed and resuspended with PBS, followed by analysis via flow cytometry. For every sample analyzed, 10,000 events were collected. The FITCpositive population, which is indicative of MDR1-expressing cells, was calculated and reported as a percentage. P-gp efflux assays utilized the MDR1 efflux kit (EMD Millipore, Billerica, MA) according to the manufacturer's protocol. Briefly, cells were incubated with P-gp substrate, 3,3-diethyloxacarbocyanine iodide [DIOC2(3)], for 15 min at 4°C, then transferred to 37°C for 45 min to allow efflux to occur. Cells were washed, resuspended in cold efflux buffer, plated in 96-well plates, and fluorescence intensity was measured (BioTek Cytation 3 plate reader; excitation: 485 nm, emission: 530 nm). A portion of cells kept at 4°C was used as a negative control. Percentage DIOC2(3) efflux was calculated by subtracting the fluorescence intensities of the 37°C sample from the 4°C sample and dividing the difference by the fluorescence intensity of the 4°C sample.

Overexpression and knockdown of AC via lentiviral transduction
The AC-expressing plasmid, pLOC-ASAH1, was purchased from Open Biosystems (Thermo Scientific), and pLKO.1-ASAH1 containing AC shRNA and pLKO.1-GFP control were purchased from Mission (Sigma-Aldrich). Plasmids were transfected into HEK293 T/17 cells with lentiviral packaging plasmids (Invitrogen). Viral supernatant was collected after 48 and 72 h and filtered using a 0.45 m syringe filter. Cells were transduced with viral supernatant plus 6 g/ml polybrene every 12 h for 3 days. Transduced cells with pLOC-ASAH1 were selected with 6 g/ml blasticidin S for 12 days. Lysates were harvested 72 h posttransduction for AC knockdown and 72 h after stable selection for AC overexpression.

In vitro viability assay
All viability assays were measured using CellTiter 96 Aqueous One Solution assay kit (Promega) according to the manufacturer's protocol. The absorbance of the formazan product at 490 nM was determined using a BioTek Cytation 3 plate reader. For cell lines cotreated with LCL204, zosuquidar, cytarabine, daunorubicin, and/ or mitoxantrone, cells were plated at 25,000 cells/well and treated for 24 h. Absorbance from treated wells was compared with absorbance from vehicle controls to calculate percentage viable cells.

Western blot analysis
Cells were cultured at one million cells per milliliter for the drug incubation experiments. Following drug treatment, cells were lysed using RIPA buffer containing phosphatase inhibitor cocktail and protease inhibitor P8340 according to the manufacturer's protocol (Sigma), resolved on 4-12% SDS-PAGE gels (Bolt gels; Invitrogen), and transferred onto PVDF membranes (Bio-Rad). Primary antibodies used in this study were as follows: AC (BD Biosciences), MDR1/ABCB1 (#13342; Cell Signaling), NF-B p65 (#4764; Cell Signaling), and -actin (#3700; Cell Signaling). For secondary antibodies, HRP-conjugated goat anti-mouse or goat anti-rabbit IgG (Cell Signaling) antibodies were used and Clarity™ Western ECL blotting substrate (Bio-Rad) was applied to blots according to the manufacturer's protocol. Image processing and analysis in Java (Image J) software or Image Lab (Bio-Rad) were used to measure band density. Band density was normalized to -actin, and the first lane in each blot was utilized as the reference that was set to 1.

Lipid extraction and analysis by ESI-MS/MS
Lipids were extracted from cell pellets (equivalent to 600 g to 1 mg of protein depending on the experiment) using an azeotropic mix of isopropanol:water:ethyl acetate (3:1:6; v:v:v). Sphingolipids were analyzed by LC/ESI-MS/MS based on the method previously described (22,36). All data reported are based on monoisotopic mass and are represented as picomoles per milligram of protein.

Real-time quantitative RT-PCR
For S1P-treated cells, cells were serum-starved and plated in RPMI media at one million cells per milliliter. S1P (100 nM) or methanol vehicle was added to the cells and incubated at 37°C for 24 h.

NF-B activation assay
Cells were plated at one million cells per milliliter and serum starved overnight. Then, nuclear extract was isolated using the NE-PER nuclear and cytoplasmic extraction reagents kit (Pierce; Thermo Scientific). Nuclear extracts were analyzed for NF-B activation using NF-B p50/p65 EZ-TFA transcription factor assay kit (Millipore) following the manufacturer's protocol.

Statistical analyses and experimental design
Statistical comparisons between two treatment groups utilized Student's t-test. Mann-Whitney U-test was used for the P-gp expression screening analysis. For in vitro assays and Western blot data using cell lines, results were from three independent experiments with three to four replicates, unless otherwise noted in the figure legends. Each graph represents an average of three to four replicates from three independent experiments and error bars represent SEM, unless otherwise noted in the figure legends. Due to the limitation of primary samples, data from primary patient and normal donor samples were from one independent experiment.

P-gp is highly expressed in AML patient samples and drugresistant cell lines
To verify the levels of P-gp in AML, we screened 27 human AML patient samples, six normal PBMC samples, and three cord blood samples for P-gp expression using flow cytometry detection of MDR1 (UIC2) cell-surface epitope (Fig. 1A). Our data showed that P-gp expression did not differ significantly between normal PBMC and cord blood samples. We found that P-gp expression was significantly higher in AML patient samples compared with the nine normal PBMC and cord blood control samples (P < 0.05, Mann-Whitney test). Collectively, the data validate earlier findings that AML patient samples exhibit elevated P-gp expression (5,14). For further mechanistic studies of drug resistance in AML, we used two HL-60-derived cell lines with acquired resistance to vincristine (HL-60/VCR) and Bcl-2 family inhibitor, ABT-737 (HL-60/ABTR) (32,33). HL-60/ VCR cells are known to exhibit increased MDR1 transcription relative to HL-60 (37). Compared with parental HL-60, both of the cell lines exhibited a drug-resistance phenotype with elevated P-gp expression (Fig. 1B) and efflux activity as indicated by a P-gp-specific substrate (Fig. 1C). The drugresistant cell lines also exhibited increased levels of AC (Fig.  1B), which we previously demonstrated is elevated in AML patient samples and contributes to AML pathogenesis (22). Therefore, these cell lines were utilized as models to investigate the role of AC in AML drug resistance.

AC inhibitor, LCL204, reversed drug resistance and induced AC and P-gp loss in HL-60/VCR
One of the main issues plaguing AML patients is the lack of response to standard of care chemotherapeutics, such as cytarabine (ara-C), daunorubicin, and mitoxantrone (38,39). To determine whether AC contributes to drug resistance in AML, we treated HL-60/VCR cells for 24 h with each chemotherapeutic agent without or with the AC inhibitor, LCL204 (7.5 M), which is known to both inhibit AC and promote its degradation via lysosomal proteases (22,35). AC inhibition sensitized HL-60/VCR cells to all three chemotherapeutic agents and significantly reversed drug resistance for cytarabine (P < 0.0005, Fig. 2A), daunorubicin (P < 0.005, Fig. 2B), and mitoxantrone (P < 0.0005, Fig. 2C). Consistent with loss of AC enzymatic activity, 24 h treatment with this dose of LCL204 increased ceramide (2.2-fold) and decreased sphingosine and S1P (2.7-fold and 1.8-fold, respectively) in HL-60/VCR cells (Fig. 2D). Furthermore, HL-60/VCR cells treated with increasing doses of LCL204 exhibited lower levels of both AC and P-gp (Fig. 2E). Taken together, these data suggest a functional relationship between AC and P-gp whereby loss of AC activity is associated with reduced P-gp levels.

AC knockdown decreased P-gp protein expression
Because AC inhibitor treatment decreased P-gp levels in HL-60/VCR cells, we next utilized a genetic approach to knock down AC to determine whether the loss of P-gp expression was related to decreased AC activity or a function of LCL204 acting directly on P-gp. Four different drugresistant cell lines were utilized (HL-60/VCR, HL-60/ABTR, KG1/ABTR, and KG1a/ABTR) that all exhibit robust P-gp expression. The cells were transduced with lentiviral particles containing AC shRNA every 12 h for 3 days. Lysates were harvested 72 h posttransduction and probed for AC and P-gp protein expression levels. AC knockdown corresponded with decreasing P-gp expression levels in all four cell lines (Fig. 3A). The AC and P-gp protein bands were quantified and normalized to -actin to show the decrease in expression levels compared with GFP control vector (Fig.  3B). AC knockdown ranged from 71% to 92% and corresponded to a 45-87% decrease in P-gp expression. Hence, loss of AC protein leads to decreased P-gp expression.
To confirm that the protein expression data translated to functional drug resistance, HL-60 and HL-60/AC cells were treated with daunorubicin (25 M) or mitoxantrone (5 M) for 24 h. HL-60/AC cells ( Fig. 4E; left three bars) were significantly more resistant to both chemotherapeutic agents (P < 0.0005 for daunorubicin and P < 0.005 for mitoxantrone, Student's t-test) compared with parental HL-60 ( Fig. 4F; left three bars). Cotreatment that added the AC inhibitor, Fig. 1. P-gp is highly expressed in AML patient samples and drug-resistant cell lines. A: P-gp cell membrane protein expression of 27 human AML samples, normal PBMC donors (n = 6), and cord blood samples (n = 3) was determined via flow cytometry using P-gp externalepitope antibody, UIC2. Expression bars represent percentage of gated cells that were positive for UIC2. *P < 0.05 compared with normal PBMC and cord blood (Mann-Whitney test). B: AC and P-gp are highly expressed in drug-resistant HL-60/VCR and HL-60/ABTR cells compared with parental HL-60 cells. Western blot showing AC and P-gp/MDR1 protein levels. C: P-gp efflux activity was assayed in HL-60, HL-60/VCR and HL-60/ABTR cells using DIOC2(3) efflux assay. **P < 0.005 compared with HL-60 (Student's t-test). LCL204 (7.5 M), to the chemotherapeutic drugs reversed the observed daunorubicin (P < 0.005, Student's t-test) and mitoxantrone resistance in HL-60/AC cells, but did not affect the response in drug-sensitive HL-60 cells (Fig. 4E, F; middle three bars). Finally, we treated HL-60/AC and HL-60 cells with zosuquidar (1 M) and chemotherapeutic agents to determine whether chemoresistance in HL-60/AC cells requires P-gp activity. P-gp inhibition significantly sensitized  cells to both chemotherapeutic agents in HL-60/AC cells (P < 0.0005, Student's t-test) but not in drug-sensitive HL-60 cells (Fig. 4E, F; right three bars). These studies demonstrate that the drug resistance phenotype of HL-60/AC cells is dependent on P-gp. Thus, AC overexpression in HL-60 cells increased resistance to chemotherapeutic agents by increasing P-gp expression and AC inhibition reversed drug resistance by reducing P-gp expression.

AC overexpression increased NF-B activation, and SPHK and NF-B inhibition reduced P-gp levels
We previously showed that AC overexpression increased S1P levels, and S1P has been shown to activate the NF-B pathway (22,41,42), which is known to induce P-gp/ MDR1 transcription (43). Hence, we investigated whether the NF-B pathway represents a mechanistic link between AC and P-gp expression. Nuclear lysates were prepared from HL-60, HL-60/AC, and HL-60/VCR cells, which showed that HL-60/AC (P < 0.05, Student's t-test) and HL-60/VCR (P < 0.005, Student's t-test) cells had significantly higher nuclear levels of NF-B subunit p65 compared with HL-60 (Fig. 5A). Both HL-60/AC and HL-60/VCR cells also showed higher MDR1 transcript levels compared with parental HL-60 cells (Fig. 5B). HL-60/AC cells are known to exhibit 3.5-fold higher levels of S1P relative to parental HL-60 cells (22). Treatment of parental HL-60 cells with exogenous S1P also increased P-gp (MDR1) mRNA levels (Fig. 5C). To confirm that AC modulates P-gp expression through S1P, HL-60/AC and HL-60/VCR cells were treated with the sphingosine kinase inhibitors, SKI-II (Fig.  5D, E) and SKI-178 (Fig. 5F, G). Both inhibitors decreased NF-B p65 and P-gp levels dose-dependently in HL-60/AC and HL-60/VCR cells. We next investigated whether NF-B promotes P-gp expression in drug-resistant AML cell lines by treating HL-60/AC and HL-60/VCR cells with NF-B inhibitors (Bay-11-7082 and JSH-23). Treatment with both inhibitors decreased the levels of NF-B p65 and P-gp dose-dependently in HL-60/AC (Fig. 5H, I) and HL-60/ VCR cells (Fig. 5J, K). Together with published literature, these data indicate that AC overexpression and the corresponding increase in S1P leads to increased NF-B pathway activation and transcriptional activity that stimulates P-gp expression in drug-resistant AML cells (22,(41)(42)(43).

DISCUSSION
Our data suggest that targeting AC leads to P-gp downregulation and reverses drug resistance. This is a novel finding that potentially enhances the functional significance of AC to include not only AML blast survival but also drug resistance (22). AML patients are frequently unresponsive to induction chemotherapy and may exhibit high levels of P-gp, which we confirmed in our patient cohort (5,12,44,45). This results in poor survival rates and adds to the complexity of treating AML patients, especially those with relapsed disease. Novel treatments including hypomethylating agents and targeted therapies (e.g., venetoclax, midostaurin, and ivosidenib) are currently gaining momentum as AML therapeutics, but these compounds are frequently given in combination with chemotherapy and the issue of MDR still remains a substantial barrier (46)(47)(48)(49). We previously demonstrated that AC is highly expressed in AML patient samples; therefore, targeting AC may represent a novel approach to reverse drug resistance in AML. More studies and larger patient cohorts are needed to more accurately define the AC/P-gp correlation and the role of AC in the many AML subtypes that differ in terms of molecular and cytogenetic alterations (50)(51)(52).
We showed that AC overexpression confers drug resistance to AML cells and that AC inhibition sensitizes cells to chemotherapeutic agents. Furthermore, AC inhibition and knockdown both decreased P-gp expression. Therefore, targeting AC may represent a substantially improved method to decrease AML drug resistance as an alternative to pharmacological agents that only inhibit P-gp. These findings are very relevant to the ongoing drug resistance dilemma, as many P-gp inhibitors have not shown significant benefit in clinical trials of AML patients (3,15,53). AC inhibition canonically increases pro-apoptotic ceramide and depletes pro-survival S1P. Therefore, combination of AC inhibitors with standard induction chemotherapy may create a multipronged approach to eliminate more AML blasts, increase the odds of achieving complete remission, and provide a new and more effective approach for AML chemotherapeutics.
Mechanistically, our data indicate that the NF-B transcription factor contributes to AC-mediated regulation of P-gp expression. Previously, we showed that AC overexpression increases S1P by 3.5-fold (22). As shown by Alvarez et al. (41), S1P induces NF-B by binding to TRAF2 and activating the IB kinase complex, which releases the NF-B dimer from IB thereby freeing it to enter the nucleus and activate transcriptional targets. The P-gp/MDR1/ABCB1 Fig. 5. AC overexpression increased NF-B activation, and SPHK and NF-B inhibition reduced P-gp levels. A: Nuclear NF-B p65 levels in serum-starved HL-60 cells were compared with serum-starved HL-60/AC and HL-60/VCR cells. Calculated percent nuclear NF-B p65 is the percent relative luminescence ± SEM. B: MDR1 expression of HL-60, HL-60/AC, and HL-60/VCR was analyzed by qRT-PCR and normalized to GAPDH. C: Serum-starved HL-60 cells were treated with either methanol vehicle or 100 nM S1P for 24 h, and MDR1 expression was analyzed by qRT-PCR. HL-60/AC and HL-60/VCR cells were treated with sphingosine kinase inhibitors SKI-II (24 h) (D, E) and SKI-178 (48 h) (F, G). SPHK inhibition decreased NF-B p65 and P-gp expression. Protein levels of p65 and P-gp were reduced in HL-60/AC and HL-60/ VCR with NF-B inhibitors Bay-11-7082 (16 h) (H, I) and JSH-23 (24 h) (J, K). Quantification above the blots represents intensities normalized to -actin. *P < 0.05, **P < 0.005, ***P < 0.0005 compared with parental HL-60 or methanol control (Student's t-test).
promoter region contains a consensus binding sequence for NF-B (43,54). Others have also reported that AC deletion downregulated TRAF2 (55). Here, we showed that exogenous S1P increases P-gp/MDR1/ABCB1 transcript levels while sphingosine kinase inhibitors reduce P-gp expression. We also demonstrated that NF-B is upstream of P-gp in AML cell lines by showing that AC expression increased NF-B activation and that NF-B inhibitors reduced P-gp expression.
We previously showed that AC inhibition increased levels of pro-apoptotic ceramide (22). P-gp functions as a flippase for glucosylceramide and ceramide, including short-chain exogenous ceramide species that are cytotoxic toward cancer cells (7,(56)(57)(58)(59). Additional study is needed to fully characterize the substrate specificity, universality, and physiological relevance of P-gp flippase activity for endogenous ceramide species (60,61), Nonetheless, AC inhibition may increase ceramide levels through two complimentary mechanisms: blocking the conversion of endogenous ceramide to sphingosine and preventing P-gp-mediated ceramide and glucosylceramide transport into the Golgi for further glycosylation and detoxification. The coupling of these two pro-ceramide mechanisms with simultaneous S1P depletion makes AC inhibition a particularly appealing therapeutic strategy. Thus far, the contributions of additional sphingolipid metabolizing enzymes, such as ceramide synthases or S1P lyase, to AML pathogenesis or the observed effects of AC inhibitors remain undefined. Further studies will test whether treatment with AC inhibitors in conjunction with exogenous short-chain ceramides may further enhance endogenous ceramide accumulation to synergistically reduce AML viability while simultaneously increasing chemotherapy sensitivity.

SUMMARY
Collectively, our data further highlight AC as an exciting and novel target to reduce cell survival and combat drug resistance in AML. Aberrant AC expression increased P-gp and conferred drug resistance in our model. Furthermore, combinatorial approaches to inhibit AC along with standard AML chemotherapeutic agents reduced survival in AML cell lines. Hence, further development of AC inhibitors for clinical use will enhance potential synergistic treatments to improve clinical outcomes in AML.