Comparative sphingolipidomic analysis reveals significant differences between doxorubicin-sensitive and -resistance MCF-7 cells

Drug resistance is responsible for the failure of many available anticancer drugs. Several studies have demonstrated the association between the alteration in sphingolipids (SPLs) and the development of drug resistance. To investigate the association between SPLs metabolism and doxorubicin (dox)-resistance in MCF-7 cells, a comparative sphingolipidomics analysis between dox-sensitive (parental) and -resistant MCF-7 cell lines along with validation by gene expression analysis were conducted. A total of 31 SPLs representing 5 subcategories were identified. The data obtained revealed that SPLs were clustered into two groups differentiating parental from dox-resistant cells. Eight SPLs were significantly altered in response to dox-resistance including SM (d18:1/16), SM (d18:1/24:2), SM (d18:1/24:0), SM (d18:1/20:0), SM (d18:1/23:1), HexCer (d18:1/24:0), SM (d18:1/15:0), DHSM (d18:0/20:0). The current study is the first to conclusively ascertain the potential involvement of dysregulated SPLs in dox-resistance in MCF-7 cells. SPLs metabolism in dox-resistant MCF-7 cells is oriented toward the downregulation of ceramides (Cer) and the concomitant increase in sphingomyelin (SM). Gene expression analysis has revealed that dox-resistant cells tend to escape from the Cer-related apoptosis by the activation of SM-Cer and GluCer-LacCer-ganglioside pathways. The enzymes that were correlated to the alteration in SPLs metabolism of dox-resistant MCF-7 cells and significantly altered in gene expression can represent potential targets that can represent a winning strategy for the future development of promising anticancer drugs.


Introduction
Breast cancer is one of the most common leading causes of death in women across the world [1]. Although it represents a major health problem worldwide, survival rates continue to rise, and women are living longer. To keep this outcome, it is necessary to continue advancing our research study about the disease. The research significantly helps in the prevention, and treatment of breast cancer, and hence can improve the quality of a woman's life. For that purpose, cell lines particularly MCF-7 are usually employed as an in vitro model to accomplish a certain level of experimental evidence. MCF-7 is the best representative for in vitro breast studies mainly because of its mammary epithelium nature that can process estrogen hormone through estrogen receptors, sensitivity to cytokeratin, and ability to form domes and monolayers [2]. It is also the first hormone-responding breast cancer cell line [2]. Although MCF-7 has been used by many research laboratories for more than 45 years, data are still generated to better understand breast cancer development and for the proper development of therapeutic strategy [3]. Thus, it constitutes a ground base for comparative research studies and data analysis not only to study the cancer pathology but also to suggest an appropriate therapy.
Drug resistance, on the other hand, is a major barrier in the efficient treatment of cancer. It is responsible for the failure of many available drugs and hence may lead to their disappearance from the market [4]. Doxorubicin (dox) is currently one of the most effective chemotherapeutic drugs used in breast cancer therapy [5]. However, a recent report has shown that approximately 50% of breast cancer patients have developed dox resistance [4]. Despite all the studies on Dox-resistance mechanisms, it is still a major unresolved problem in cancer therapy.
Recently, the role of cellular lipids in both effective therapy and resistance is drawing scientists attention [4]. Sphingolipids (SPLs) are a class of cellular lipids that play an important role in the structural integrity and fluidity of mammalian cell lipid bilayer [4]. SPLs including ceramide (Cer), sphingomyelin (SM), sphingosine-1-phophate (S1P), hexosylceramide (HexCer), sphingosine (So), and glucosylceramide (GlcCer) act as signaling molecules that contribute in the regulation of several biological processes of a cell [4]. These include cell proliferation, apoptosis, cell differentiation, cell migration, angiogenesis, autophagy and inflammation [6]. Furthermore, SPLs metabolic pathways influence cancer pathogenesis, drug resistance, and chemotherapeutics efficacy [4,7]. The biochemical role of SPLs has been previously studied in cancer progression and development. Additionally, SPLs have been implicated in the mechanism of action of many chemotherapeutic agents [8,9].
The core of SPLs metabolism is Cer, which is formed of a sphingosine base containing 18 carbons (d18), and an amide-linked fatty acyl chain with different number of carbons (C14-C26) [10]. Cer acts as a precursor for the synthesis of complex SPLs, such as SM, and GlcCer, which contain hydrophilic head groups [11]. Several studies have demonstrated a strong connection between the alteration in SPLs metabolism and drug resistance in human cancer cells [12,13]. Enzymes in SPLs metabolic pathway are also involved in the regulation of many cancerous processes. Glucosylceramide synthase (GCS) is proved to be a key player in dox-resistance in various cancer types, importantly by the enzymatic conversion of Cer to GlcCer [14]. The generation of GlcCer acts as the precursor for the synthesis of other glycosphingolipids and gangliosides [15]. Low Cer levels are correlated with a higher degree of malignant progression and severity of prognosis in tumor cells [16]. Many studies proved that drug-resistant cells had 8 to 10-fold higher capacity to convert the precursor [3H]-palmitic acid to Cer and further to GluCer, than non-drug resistant counterparts [17]. This can be explained by the ability of Cer to mediate anti-proliferative pathways or inhibits pro-survival mechanisms [12]. In addition to that, Cer is shown to regulate gene expression, such as upregulates MMP-1 and hTERT [18,19], activates COX-2 promoter [20], inhibits NF-κB activation [21], as well as induces GCS promoter by Sp1 [14]. However, the effect of Cer on multidrug resistance is still not well understood [22]. A strong association between low Cer levels and the elevation in GCS has been reported in dox-resistant cancer cells [23,24].
Uchida et al examined the effect of dox on drug-sensitive HL-60 cells and drug-resistant HL-60/ADR (Adriamycin) cells [25]. Treatment with dox induced apoptosis and Cer production in drug-sensitive HL-60 cells, but not in drug-resistant HL-60/ADR cells. In dox-treated HL-60/ADR cells, the levels of mRNA, and protein of GCS were upregulated [25]. In a more recent study on MCF-7 cells, the effects of different doses of dox were tested [26]. They reported dose-dependent changes in SPLs levels, which include an increased level of Cer, dihydroceramide, S1P, and So, while reduced the levels of HexCer [26]. Moreover, UGCG silencing in Dox-resistant MCF-7 has restored cell sensitivity and increased endogenous ceramide and caspase-3 [27]. In contrast, UGCG-overexpressing MCF-7 cells have increased the cellular proliferation and dox-resistance accompanied by stimulation of Akt and ERK1/2 signaling pathways as well as upregulation of anti-apoptotic genes and multidrug resistance protein 1 (MDR1) [28]. Similarly, it has been shown by a sphingolpidomics analysis on ovarian cancer cells that the levels of cell membrane SPLs have been significantly altered in resistant cells when compared to sensitive cells, both treated with Taxol, although the target is β-tubulin [29]. Collectively, growing evidence suggests that alteration in SPLs metabolism is critical in the dox-resistance mechanism [27, 30, 31]. However, until now, there is no sphingolipidomics analysis study that explore the other potential pathways (other than GCS) in SPLs metabolism in dox-resistant MCF-7 cells. Therefore, a comprehensive sphingolipidomics study may help to elucidate the role of SPLs in dox-resistance in MCF-7 cells.
Based on this, we hypothesize that multiple SPLs metabolic pathways may play a role in dox resistance. Therefore, a comparative sphingolipidomics analysis between dox-resistant and parental P-MCF-7 breast cancer cells was conducted to identify the changes in SPLs metabolism that may be associated with dox-resistance mechanisms. Further, the critical genes and enzymes involved in the alteration of SPLs metabolism were investigated by qRT-PCR. The results of this study can be used for the effective development of cancer therapy.

Methylthiazolyldiphenyl-tetrazolium bromide (MTT) viability assay
MTT assay was performed to verify the dox resistance in dox-resistant MCF-7 cell line. Doxresistant MCF-7 cells were maintained at 37˚C in a humidified incubator inclusive of 5% CO 2 in RPMI media with 10% fetal bovine serum and 1% penicillin/ streptomycin. For the assessment of cell viability, parental and dox-resistant MCF-7 cells were seeded in 96-well plate at 5× 10 3 cells/well and allowed to adhere for 24 h. 50 μL of MTT reagent with 50 μL serum-free media was mixed and then added to each well and incubated at 37˚C for 3 h. After the removal of the solution, 150 μL DMSO was added and shaken for 15 min. The absorbance was then measured at 590 nm using a microplate reader (Varioskan Flash, Thermo Scientific). The sensitivity of parental and dox-resistant MCF-7 cell lines to dox was assayed. The IC 50 for both cell lines were 0.5 μM and 18 μM, respectively. Dox-resistant MCF-7 is 36-fold resistant to dox when compared to parental MCF-7 cells.

Sphingolipids (SPLs) extraction
SPLs extraction was adapted from a previously published methodology [29,33]. First, MCF-7 cells were seeded into T-75 flasks and grown to confluence (3 million cell/mL). Cells were rinsed twice with ice-cold PBS, then scraped into a glass tube. Second, 0.5 mL MeOH, 0.25 mL CHCl 3 and 10 μL internal standards cocktail (2.5 μM) were added consecutively. Afterward, the mixture was sonicated at room temperature for 30s and then incubated at 48˚C for 12 h to extract SPLs. Third, 75 μL of KOH in MeOH (1M) was added and incubated in a shaking water bath for 2 h at 37˚C to cleave any glycerolipids. After cooling and neutralization with 5% acetic acid, the solution was centrifuged, and the supernatant was dried and re-dissolved prior to LC-MS analysis [29, 33].

SPLs data analysis
The data obtained from LC-MS was imported into MetaboScape 4.0 software and library (Bruker Daltonik GmbH, Bremen, Germany) for the identification of SPLs. Then, Microsoft Excel was used to collect and classify the data into tables. In order to investigate the overall differences between the parental and dox-resistant MCF-7, two multivariate analyses were carried out, namely, partial least squares-discriminant analysis (PLS-DA) and hierarchical cluster analysis (HCA). The PLS-DA was used to efficiently differentiate between the parental and dox-resistant MCF-7 cells to identify the significantly different SPLs, while HCA is an unsupervised analysis technique that classified the data into clusters. The most significantly different SPLs between the two cell lines were selected according to Variable Importance in Projection (VIP) value using MetaboAnalyst 4.0 software. VIP values higher than 1.00 were considered significant. Kyoto Encyclopedia of Genes and Genomes (KEGG) LIPIDS PATHWAY database was used to study the SPLs metabolic pathways.

Quantitative real-time polymerase chain reaction (qRT-PCR)
Dox-sensitive (parental) and -resistant MCF-7 cell pellets were placed on ice and treated with 0.6 mL of lysis buffer and vortexed until dispersion. Total RNA was extracted using PureLink RNA Mini Kit (Thermo Fisher Scientific, Waltham, Massachusetts, United States) following the manufacturer protocol. The RNA samples were treated with DNase-I treatment (On-column PureLink1 DNase, ThermoFisher, USA) solution for 15 min at room temperature to remove contaminated DNA. RNA was then eluted by adding 30 μL of RNase-free water at room temperature for 1 min followed by centrifugation at 12,000 RPM. The purity and yield of RNA were assessed by nanodrop2000 spectrophotometer (Thermo-Scinetific, USA). The resulting RNA was stored at -80˚C until used for cDNA synthesis. cDNA was synthesized according to the protocol provided by SensiFAST™ cDNA Synthesis Kit (Bioline Reagents Ltd., London, United Kingdom). RNA (1 μg) was mixed with 4 μL 5x TransAmp Buffer, 1 μL Reverse Transcriptase and DNase/RNase-free water up to a final volume 20 μL. The samples were placed in T100™ Thermal Cycler. Cycling conditions were annealing at 25˚C for15 min, reverse transcription at 42˚C for 30 min, inactivation at 85˚C for 15 min. QRT-PCR was performed to quantify the expression level of 14 genes encoding rate-liming enzymes known to be critical in SPLs metabolic pathways and were identified following our SPLs analysis. GADPH transcript was used as a housekeeping gene. Primers were designed and custom ordered from Microgen Medical Equipment Est (UAE). The sequences of the designed primers are listed in Table 1. Ensemble Genome Browser (https://asia.ensembl.org/index.html) was used for primer design. The quality parameters of the designed pPrimers were checked using Primer-BLAST available at https://www.ncbi. nlm.nih.gov/tools/primer-blast/index.cgi?ORGANISM=9606&INPUT_SEQUENCE=NM_ 001618.3. Oligo software (version 9.1) was used for checking the primer dimer formation. QRT-PCR was carried out on Quant Studio 3 (Thermo Fisher) using SensiFAST™ SYBR Hi-ROX kit. PCR reaction mixture was prepared by mixing 6 μL of SensiFAST TM SYBR Hi-ROX, 0.48 μL forward primer (400 nM), 0.48 μL reverse primers (400 nM), 3.04 μL water, and 2 μL template. Initial steps of qRT-PCR were 2 min at 50˚C for polymerase activation, followed by a 10-min hold at 95˚C. Cycles (n = 40) consisted of 15 secs melt at 95˚C, followed by 1-min annealing at 60˚C and 20 secs extension at 72˚C. The final incubation step was set at 60˚C for 1 min. All samples were amplified in triplicates. Ultra-pure RNA-free water was included in the run as a negative control. The average threshold cycle (Ct) values were obtained from each reaction, and the relative expression was quantified using the 2 (−ΔΔC(T)) method [34].

Statistical analysis
The reports were extracted from MetaboScape 4.0 software (Bruker Daltonik GmbH, Bremen, Germany) to Microsoft Excel, where the compound concentration was calculated. SPLs identification was performed using MetaboScape 4.0 software and library (Bruker Daltonik GmbH, Bremen, Germany). The resulting table, including SPLs names, sample names, and intensity levels, was imported into MetaboAnalyst 4.0 software for multivariate statistical analysis. Partial least-squares discriminant analysis (PLS-DA) and hierarchical cluster analysis (HCA) were used to efficiently differentiate between the parental and dox-resistant MCF-7 cells. Variable Importance in Projection (VIP) value was calculated using MetaboAnalyst 4.0 software. VIP values higher than 1.00 were considered significant. A student's t-test was applied to identify SPLs with statistically significant differences in intensity level between the two cell lines assuming that95% is the confidence level and 5% is false positive (false discovery rate FDR). P-values less than 0.05 were considered significant.

Multivariate analysis indicated a significant alteration in unique SPLs associated with dox-resistance in MCF-7 cells
To investigate the overall differences between the parental and dox-resistant MCF-7 cells, multivariate analyses were carried out including HCA, and PLS-DA. HCA is an unsupervised analysis technique that was used to identify the natural patterns in the samples; thus, avoiding overfitting the sample data, while the PLS-DA was used to identify the key biomarkers that can distinguish between the two cell lines. HCA was carried out to explore the overall differences, similarities, and hidden patterns between the two cell lines. HCA classified the data into two clusters corresponding to parental and resistant cells. SPLs were clustered in the dendrogram according to their intensity levels to a hierarchical relationship that differentiated both cell types (Fig 3). The dendrogram showed variation in SPLs, which is most likely associated with a distinctive pattern related to each cell type and accordingly to a dox-resistance in MCF-7 cells (Fig 2). SM (d18:1/16:1) and SM (d18:1/24:0) showed higher abundance in dox-resistant MCF-7 compared to parental cells, indicating a significant association between alteration in SPLs and resistance mechanism in MCF-7 cells due to dox (Fig 2).
PLS-DA was used to extract the features that can be used to efficiently differentiate between the two cell lines. Pareto scaling and generalized log transformation function (glog) were applied to the data sets. As shown in Fig 3A, the variables were well-separated between the two cell lines, suggesting that this model strongly discriminates the parental from the resistant cells. Furthermore, 15 SPLs showed VIP score values greater than one; most of them were SMs  (Fig 3B). SM d18:0/16:1, and SM d18:1/24:2 were the highest, suggesting that these SPLs could play an important role in the mechanism of dox-resistance in MCF-7 cells.  Table 3) illustrated the agreement between the results obtained from the univariate analysis (student's t-test) and the  multivariate analysis (VIP score) (Fig 3B). Collectively, this significant variation in SPLs between parental and dox-resistant MCF-7 cells suggests that the alteration in SPLs metabolic pathways was most likely involved in MCF-7 resistance to dox.

Proposed model of SPLs dysregulation due to dox-resistance in MCF-7 cells
Kyoto Encyclopedia of Genes and Genomes (KEGG) LIPIDS PATHWAY database in association with our current data were used to propose a model of variation in genes related to SPLs metabolism due to dox-resistance (Fig 5). The synthesis of Cer via de novo pathway was altered via the upregulation of dihydroceramide desaturase 1 generating Cer. The observed downregulation of Cer in dox-resistant MCF-7 cells was oriented toward two different dysregulated pathways. First, the SM-Cer pathway including the upregulation of sphingomyelin synthesis (sphingomyelin synthase 2) and the downregulation of sphingomyelinase 2/3 occurs. Second, the Cer-GluCer-ganglioside including the upregulation of glucosylceramide synthesis (glucosylceramide synthase) and the downregulation of glucosylceramidase 1 occur. This is followed by the production of ganglioside from glucosylceramide. The galactosylceramide pathway showed no significant difference, although galactosylceramidase was decreased.
Besides SM, deregulated Cer levels have shown an association with different aspects of cancer signaling and progression [37,38]. The biological activity of Cer appears to not only rely on the fatty acid chain length, but also on the ratio of different SPLs metabolites [39]. Generally, low Cer levels have been reported as a feature of many drug-resistant cancers [40]. Here, we have identified three different Cer compounds where two of them including Cer d18:1/ 16:0, and Cer d18:1/22:0 were exclusively found in parental MCF-7 cells but with no statistical significance. A similar observation by Mullen et al. has confirmed the accumulation of Cer and DHCers of carbon chain length between C18-22 in MCF-7 cells [41]. Furthermore, Taxol resistant A549T cells showed lower levels of Cer 16:0 when compared to A549 cells. Markedly, upregulation of Cer 16:0 was associated with apoptosis in human colon cancer cells [42].
Other studies also showed that certain Cer compounds including Cer 16:0, accompanied with deregulation in HexCer 24:1, and SM 24:1 levels seem to affect many cellular biological processes in colon cancer, such as a cellular switch from differentiation to apoptosis [43,44].
Dihydroceramide DHCer, on the other hand, is placed in an intermediate step in the de novo Cer synthetic pathway, catalyzed by dihydroceramide desaturase to produce Cer (Fig 5). Although DHCer is found in the tissues in lower concentrations, the added double bond in its structure significantly affects the membrane composition [45], fluidity, and subsequent signaling [37]. Recently, DHCer has been linked to cancer signaling and progression [46]. Many studies have confirmed the significant role of DHCer in autophagy [47]. Treatment of cancer cells with DHCer analogs or dihydrocermaide desaturase inhibitor has led to the accumulation of many endogenous DHCer compounds, mainly DHCer d18:0/16:0 and induced autophagy in cancer cells [47]. We have been able to identify the same compound in both cell lines in equal amounts. Another study has also highlighted the importance of DHCer C16:0 in glioma cells treated with DES1 inhibitor [48]. Thus, leading to a significant increase in DHCer C16:0 that resulted in ER stress and subsequent autophagy [48]. Collectively, the role of DHCer in autophagy is well-studied unlike in cancer resistance where there is sparse evidence.
Cer glycosylation is known as a crucial step in controlling Cer levels and has been highly associated with cancer resistance [49]. Four HexCer; d18:1/16:0, d18:1/18:0, d18:1/24:0, and d20:1/24:1 were identified in our study only in the parental MCF-7 cells. A study performed in human breast cancer patients has agreed with our results by showing an increase in certain compounds of HexCer (C14: 0, C16: 0, C18: 1, C18: 0, C20: 0, C22: 0, C24: 1, and C24: 0) [38]. In general, the balance between Cer and HexCers as well as the rate-limiting enzymes in Cer glycosylation such as UGCG and UGT8, are actively contributing to many aspects of cancer signaling, proliferation, and resistance [49]. Furthermore, we have reported that HexCer d18:1/24:0 was the most abundant in parental MCF-7. This can be related to a study that reported a significant decrease by 76% in HexCer d18:1/24:0 in Dox-treated MCF-7 cells and an overall dose-dependent reduction in HexCers, despite of the length of the N-acyl chain [26]. Accordingly, HexCer d18:1/24:0 seems to be targeted by Dox in MCF-7 and hence, doxresistant cells may develop a depletion mechanism of this metabolite as a unique strategy to overcome the dox effect. Further studies are needed to confirm this hypothesis.
Collectively, dox-resistant MCF-7 cells showed significant variation in SPLs compared to parental cells indicating the importance of SPLs in cancer cell resistance mechanisms. This is considered the first report to indicate specific changes in SPLs of MCF-7 cells due to dox resistance.
To validate our findings and explore the importance of SPLs in the dox resistance mechanism, gene expression of 14 transcripts encoding rate-liming enzymes in SPLs biosynthesis, as indicated in Fig 5 was studied. The expression of sphingomyelin synthases (SMS1, SMS2) was variable, where SMS2 was significantly upregulated in dox-resistant MCF-7 cells, while SMS1 did not show a significant difference (Fig 4A). SMS2 has been reported to stimulate breast cancer cell proliferation by suppressing apoptosis through a Cer-associated pathway [50]. Accordingly, SMS1 and SMS2 may hold a differential activity toward different SMs, and this can explain the significant difference in the expression of both isozymes. On the other hand, a significant downregulation was exhibited with neutral sphingomyelinase (SMPD2, and SMPD3) in dox-resistant MCF-7 cells. A study reported that the upregulation of SMPD2 was differentially induced the levels of very-long-chain (C24:1 and C24:0) Cer, which is correlated with a decrease in C24:0-and C24:1-sphingomyelins in MCF-7 cells [51]. Based on this, we suggest that SMPD 2/3 play an important role in the metabolism of SM (d18:1/24:0, and d18:1/24:2) in dox-resistant MCF-7 cells (Fig 3 and Table 3). Collectively, this may explain the distinctive sphingomyelin synthesis and the consumption of Cer in dox-resistant cells, which is associated with inhibiting the pro-apoptotic effect of Cer.
The expression of three Cer synthase genes (CERS 2, 4, 5) was significantly higher in parental MCF-7 cells (Fig 4B). Previous studies have shown that CERS4 generates C18-C20 Cer, CERS5 and CERS6 generate C14-C16 Cer, and CERS2 selectively generates C22-C24 Cer [52]. Consistently, our results suggest that Cer d18:1/22:0, which were only found in parental cells, are mainly synthesized by the highly expressed CERS2 in parental cells. Consequently, the significant decrease of CERS in dox-resistant cells supports the dominant depletion of CERS. Gene expression of both isoforms of dihydroceramide desaturase (DeS) transcripts was analyzed because of their crucial role in controlling the balance between SPLs and dihydrosphingolipids [53]. Interestingly, our qRT-PCR results demonstrated that DeS1 was remarkably expressed in dox-resistant cells, while DeS2 was completely absent in this cell line. Consistently, resveratrol-induced autophagy in HGC-27 cells was correlated with an increase in the intracellular DHCer levels caused by the inhibition of Des1 activity [54]. Therefore, we can propose that each isoform of DeS enzyme has a distinctive action that requires further investigation to explore their roles in cancer resistance.
Several studies have demonstrated the correlation between multidrug resistance and cer glycosylation [30,55]. UGCG converts ceramide to glucosylceramide (GluCer), which displayed elevated levels in multidrug-resistant cancer cells [56]. Similarly, we can correlate the depletion of HexCer in dox-resistant cells to the increase in UGCG gene expression, which can be explained by the activation of the ganglioside pathway [57]. This in turn enables the cancer cells to convert GluCer to gangliosides to bypass the Cer-induced apoptosis. GBA opposes the action of UGCG (Fig 5). siRNA for GBA1 gene was shown to induce resistance to Taxol in three different cancer cell lines [58]. In contrast, our results did not show any significant difference in the GBA gene expression in both cell lines. GALC converts GalCer to Cer, while until today, there is no report exploring the association between cancer and GALC in tumor cells. We have observed a significant downregulation of GALC expression in dox-resistant cells (Fig 4B), which suggests the potential correlation between GALC activity and dox resistance.
As concluding remarks, the findings from this study have conclusively ascertained the involvement of SPLs in dox resistance in MCF-7 cells. Collectively, SPLs metabolism in doxresistant MCF-7 cells is oriented toward the downregulation of Cer and HexCer with the concomitant increase in SM. We propose that dox-resistant cells tend to escape from the Cerrelated apoptosis by the activation of two different pathways: SM-Cer and GluCer-LacCer-ganglioside. The enzymes that were correlated to SPLs and significantly altered in gene expression may represent potential targets that need further investigation. Further studies on adjusting the SPLs metabolism purposively may represent a winning strategy for the future development of anticancer drugs.
Supporting information S1