Identification of dysregulation of sphingolipids in retinoblastoma using liquid chromatography-mass spectrometry

Abstract

Retinoblastoma (RB) is a rare ocular cancer seen in children that counts for approximately 3% of all childhood cancers. It is found that mutation in RB1, a tumour Suppressor Gene on chromosome 13 as the cause of malignancy. Retinoblastoma protein is the target for ceramide to cause apoptosis. We studied lipidomics of two RB cell lines, one aggressive cell line (NCC–RbC-51) derived from a metastatic site and one non aggressive cell line (WERI-Rb1) in comparison with a control cell line (MIO-M1). Lipid profiles of all the cell lines were studied using high resolution mass spectrometer coupled to high performance liquid chromatography. Data acquired from all the three cell lines in positive mode were analyzed to identify differentially expressed metabolites. Several phospholipids and lysophospholipids were found to be dysregulated. We observed upregulation of hexosyl ceramides, and down regulation of dihydroceramides and higher order sphingoglycolipids hinting at a hindered sphingolipid biosynthesis. The results obtained from liquid chromatography-mass spectrometry are validated by using qPCR and it was observed that genes involved in ceramide biosynthesis pathway are getting down regulated.

1. Introduction

Retinoblastoma (RB) is an ocular cancer mostly prevalent in children below 5 years of age. It is estimated that, 8000 new cases of RB are seen globally. Asia and Africa reported 40–70% mortality of children with retinoblastoma (Jain et al., 2019). Middle-income countries share largest burden (69%) of RB compared to high-income (11%) and low-income countries (20%) (Dimaras et al., 2012). The RB tumour progression is associated with inactivation or complete loss of house-keeping gene RB1. The mutation could be germ line or somatic (Hong FD et al., 1989; Knudson, 1971). Recently, role of MYC N leading to disease progression is elucidated in a subset of RB tumours (Lee WH et al., 1984). Several studies were done to understand retinoblastoma using Genomics (Kooi et al., 2016; Sradhanjali S et al., 2021), Transcriptomics (Alvarez-Suarez DE et al., 2020; Winter et al., 2020; Elchuri et al., 2018) and single cell sequencing (Collin J et al., 2021). Studies on comparative proteomics revealed mitochondrial dysfunction and dysregulated lipid metabolism in RB (Naru J et al., 2017; Galardi A et al., 2020). The lipid binding proteins like APOA1, CRABP2, CRABP1 and FABP5 were up regulated in RB tumours (Mallikarjuna et al., 2010). Phosphorylation/Dephosphorylation of RB protein regulates cell growth (Chao R et al., 1992) RB protein was identified as a target for ceramide to cause cell death (Dbaibo GS et al., 1995). In retinoblastoma, as the RB gene is mutated it would be worth studying ceramide pathway. Sphingolipid pathway has been a topic of research globally. Association of C₁₆ and C_24:1 ceramide with metastasis in pancreatic cancer was reported (Jiang Y et al., 2013). Reduction in the levels of C₁₈ ceramide in Head and Neck squamous cell carcinoma was correlated to lymphovascular invasion and nodal metastasis (Karahatay S et al., 2007). mRNA expression of ceramide synthase family in breast cancer was reported to play an important role in diagnosis (Erez-Roman R et al., 2010). Low ceramide levels were found and were attributed to malignant progression in high grade glial tumours (Riboni L et al., 2002). We chose two cell lines to study lipidomic profiles of RB. NCC-Rbc-51 is an established cell line from a metastatic site. This has been derived from cervical lymph nodes of a bilateral RB patient (Inomata et al., 1994). WERI-Rb1 is established from an enucleated RB tumour (McFall et al., 1977). WERI-Rb1 was characterized and extensively studied (Busch et al., 2015; McFall et al., 1978; Schwermer et al., 2019). Intra ocular injection of WERI-RB1 cells to immune suppressed rabbit resulted in tumor growth (Kim et al., 2017). NCC-Rbc-51 is relatively new and less explored (Ravishankar H et al., 2020). In this study we compared lipidomic profiles from these two cell lines with a control retinal cell line, MIO-M1.

2. Materials and methods

2.1. Cell culture

NCC-RbC-51 was isolated from a metastatic site of bilateral RB patient (Ravishankar H et al., 2020). These cells (obtained from RIKEN BioResource Centre, Ibaraki, Japan) were maintained in RPMI 1640 (Roswell Park Memorial Institute 1640), with 10% Fetal bovine serum (FBS) and supplemented with antibiotic-antimycotic solution containing penicillin 100 μg/mL, streptomycin 100 μg/mL, and amphotericin B 250 ng/mL. WERI-Rb1cells were isolated from the primary site of a tumor. These cells were cultured in RPMI. Müller Glial MIO-M1cell line was a kind gift from Professor Astrid Limb, London, UK, and were cultured in DMEM (Dulbecco’s Modified Eagle’s Medium) containing 10% FBS. MIO-M1 cell line is derived from neural retina and is used in various studies before as control cell line (Jayashree et al., 2016; Maradani et al., 2022). Once the cells are harvested and centrifuged, the supernatant media is discarded completely. Further, they are washed with PBS three times to make sure all the media is removed. Before the lipid extraction, all the samples appeared clear and visually no sample showed any traces of media.

2.2. Lipid extraction from cell lines

We have established the required cell count for the optimum extraction of lipids in our laboratory. Cell pellet from approximately 5 × 10⁶ cells were subjected to repeated freeze thaw cycles for lysis followed by lipid extraction by Folch method (Folch et al., 1957). Lipids were extracted from biological triplicates of NCC-Rbc-51, WERI-Rb1 and MIO-M1. Briefly, 300 μL of distilled water was added to the cells and vortexed for 60 s. Further, 400 μL of methanol and 800 μL of chloroform were added and vortexed for 60 s. Lipids were extracted by shaking the samples for 30 min on an overhead rotor. Samples were centrifuged at 10000 RPM at 4 °C for 10 min and the supernatant aqueous layer was carefully drawn out. 400 μL of methanol was added to the remaining solution and vortexed for 60 s. The organic solvent was carefully withdrawn after the protein part settled down by centrifugation. The extract was evaporated to dryness by speed vac and was resuspended in 100 μL Methanol: Chloroform in 9:1 ratio.

2.3. Liquid chromatography-mass spectrometry (LC-MS/MS) analysis

Lipids were separated on a C₁₈ reversed phase HPLC column (100 mm × 3 mm, 3 μm, Machery Nagel, Germany). Samples were loaded in 95% mobile phase A (10 mM ammonium formate with 0.1% formic acid in 60:40 Acetonitrile/water) and 5% Mobile phase B (10 mM ammonium formate with 0.1% formic acid in 90% Isopropyl alcohol and 10% Acetonitrile) with 250 μL/min flow rate. Both the mobile phases were adjusted to pH 4. The column temperature was maintained at 30 °C. The sample compartment was maintained at 5 °C. Pump program maintained at 5% B from 0 to 1 min. B was increased to 20% in 3 min, 40% in 6 min, 70% in 12 min and 100% in 24 min and maintained the flow with mobile phase B alone up to 35 min and next 5 min the column was kept for equilibration for the next run. MS and MS/MS data was acquired on Q Exactive plus Mass Spectrometer (Thermo Scientific, Bremen, Germany) in positive mode. MS1 data was acquired at 70000 resolution and MS2 data at 35000 resolutions. MS2 data was collected at CE:10,30, and 40. Isolation window to select MS/MS precursor was set as 1.0 m/z. Top 5 ions from MS1 were selected for MS/MS fragmentation. Mass spectrometry conditions were Spray voltage:4350 V, Capillary temperature; 330 V, Sheath gas; 35, Auxillary gas; 12, Sweep gas; 3, Capillary Temperature: 350 °C Probe heater temperature; 325 °C, S-Lens: 55. Each sample was analyzed in quadruplicates with 4 μL of injection volume. Each biological replicate is analyzed in technical triplicates. Thus, data is obtained from 9 LC-MS analyses for each cell type. The data was acquired in full scan mode to identify the differentially regulated lipids. Further, each biological replicate is analyzed in data dependant MS/MS mode to characterize the lipids.

2.4. Data analysis post data acquisition

Initial processing of positive mode raw data files was done by Compound Discoverer™ (version 3.2.0.421) software that included alignment, detection, annotation and statistical analysis as described in the workflow (Fig. 1). When samples are run in a long sequence in HPLC, peaks from some samples may show some shift in their retention time. This is corrected by the alignment algorithm in the software. Raw data files were aligned with adaptive curve setting with 5 ppm mass tolerance and 0.2 min retention time shift. Software detected protonated ions [M+H] ⁺, sodiated ions [M+Na] ⁺and ammoniated ions[M + NH₄]⁺ of the molecules and grouped all the ions related to each molecular species. Parameters for grouping the features were being mass tolerance: 5 ppm, S/N ratio:3, Relative intensity tolerance for isotope search: 30%, Minimum peak intensity: 1 × 10⁶. Background subtraction and noise removal were done using a procedural blank sample during the pre-processing step. Peaks with less than a 5-fold increase compared to blank samples, peaks detected in less than 50% of QCs and Peaks with relative standard deviation (%RSD) of the QCs greater than 20% were discarded. Peak areas, across all samples, were subsequently median normalized. Annotation of the compounds was done by searching against ChemSpider™ chemical structure database, Endogenous metabolites mass list and mzCloud spectral library with a mass tolerance of 5 ppm. Following data sources were selected via the ChemSpider database: BioCyc, Human Metabolome Database (HMDB), Kyoto Encyclopaedia of Genes and Genomes (KEGG), LipidMAPS, LGC Standards, and WikiPathways. Statistical analysis was performed using t-test, ANOVA, Tuckey’s post hoc test and p-value adjustment using Benzamini-Hochberg procedure. Further, annotation of the metabolites was also manually verified based on both accurate mass and diagnostic fragment ions from MS/MS spectra wherever applicable.

Fig. 1. Workflow to analyze the LC-MS data in compound discoverer software.

2.5. RNA extraction and qPCR

WERI-Rb 1, NCC-RbC-51 and MIO-M1cells were collected and used for RNA extraction by TRIzol method (sigma). The RNA concentration was quantified using nano drop spectrometer. Further cDNA synthesis was performed using an iscript cDNA conversion kit (Thermo scientific). Quantitative real time PCR was carried out using the relative concentration of mRNA expression, normalized with the house keeping gene (GAPDH). The primers used for the mRNA study are given in Table 1. The Ceramide pathway gene expression in RB cells was normalized to the gene expression in the control cells and fold change in gene expression with respect to control was calculated. Statistical analysis was done using Graphpad prism. Sidak’s multiple comparisons test was performed in the software and the p adjusted values were calculated for Ceramide pathway gene expression in 1) control versus WERI-Rb1, 2) Control versus NCC-Rbc-51 and 3) WERI Rb1 versus NCC-Rbc-51. However, NEK expression was calculated in NCC-Rbc-51 using WER1-Rb1 as control. Paired t-test was used for determining the statistical significance of fold change in gene expression. The data is an average of three independent experiments in replicates.

Table 1. Details of primers used to check gene expression.

Gene	Primer Sequence
SPTLC1	FP RP	CTCCTCCCAGAGGAAGAACTGG TCTTGAGTCCTCTCTGCGTG
KDSR	FP RP	TACCCACCAGACACAGACAC CCGAGAGCATGTACCCATCT
CerS2	FP RP	CCAGGTAGAGCGTTGGTT CCAGGGTTTATCCACAATGAC
CerS6	FP RP	AAGCAACTGGACTGGGATGTT AATCTGACTCCGTAGGTAAATACA
UGCG	FP RP	TGCTCAGTACATTGCCGAAGA TGGACATTGCAAACCTCCAA
ST3GAL5	FP RP	GCCGAGCAATGCCAAGTGAG CAGCGCCATTGATGTCTTGG
NEK2	FP RP	GATTGGAGCAGAAAGAACAG CTGAGGATGGAAGATTAAGAAG
GAPDH	FP RP	GAAGGTCGGAGTCAACGGATT CGCTCCTGGAAGATGGTGA

2.6. Drug treatment

The WERI-Rb 1and NCC-RbC-51 cells were grown for 24 h and then treated with CDC25 cell cycle inhibitor NSC663284 at IC₅₀ value of 100 nM for 4 h and then Immunofluorescence experiments and gene expression by qPCR were performed. The WNT signalling inhibitor LiCl was treated at 40 mM for 48 h according to the published protocol by Silva et al. (Silva AK et al., 2010). The NEK2 gene expression was performed according to the protocol given above. The gene expression was first normalized to GAPDH in WERI-Rb 1and NCC-RbC-51 cells. Fold change in the expression level after the drug treatment was calculated with respect to 0-h expression in both cell lines. WERI-Rb1was treated as control for these experiments. The data is an average of three independent experiments done in replicates.

2.7. Immunofluorescence

The localization of NEK2 protein was studied using immunofluorescence technique by Fluorescence microscopy. The experiment was performed according to the protocol of Chao et al., 1992; Amit C et al., 2020). The NEK antibody used was (Catalog # PA5-31259, from Thermofisher).

3. Results

3.1. Metabolite analysis of retinoblastoma cell lines

Lipidomic data from MIO-M1(normal retinal cell line), WERI-Rb1 and NCC-RbC-51 cells was acquired in positive ion mode using LC-MS/MS approach. We identified 4898 metabolite features in positive ion mode. Out of that, 786 features could be annotated using publicly available databases (Supplementary Table 1). Post statistical analysis of the data, 88 metabolites were found to be up regulated (log ₂fold change ≥2 and p value ≤ 0.05) and 147 metabolites found to be down regulated (log₂fold change ≥2 and p value ≤ 0.05) in NCC-RbC-51 Vs MIO-M1cells (Fig. 2) and 72 metabolites were up regulated and 151 metabolites were down regulated in WERI-Rb1Vs MIO-M1cells (Fig. 3). Tentative structures are proposed to 45 dysregulated molecules after manually verifying the MS and MS/MS (Table 2). Pearson coefficient of QC samples was found to be 0.99 (Table 3). Tentative structures are identified by accurate mass information and further confirmed by MS/MS data.

Fig. 2. Volcano plot of metabolites from NCC-RbC-51 Vs MIO-M1.

Green coloured circles represent down regulated entities and red coloured circles represent upregulated entities. In both the directions entities below the vertical scale 2 (light green coloured and thick red coloured circles) are not considered. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 3. Volcano plot of metabolites from WERI-Rb 1Vs MIO-M1.

Green coloured circles represent down regulated entities and red coloured circles represent upregulated entities. In both the directions entities below the vertical scale 2 (light green coloured and thick red coloured circles) are not considered. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Table 2. List of dysregulated lipids in retinoblastoma.

Name of the compound	Observed Mass (m/z)	Theoretical mass (m/z)	Error (ppm)	RT (min)	Ratio: (NCC–RbC- 51)/(MIO-M1)	Ratio: (WERI- Rb1)/(MIO-M1)	Adjusted		Adjusted
							P-Value (NCC–RbC- 51)/MIO-M1		P-Value (WERI- Rb1)/(MIO-M1)
PC(P-17:0/0:0) *	494.3603	494.3605	−0.36	15.76	4.359	3.092	2.03E-13		2.27E-13
PC (22:0/0:0) #	580.4336	580.4337	−0.10	16.69	0.037	0.022	2.03E-13		2.27E-13
PC (24:1/0:0) #	606.4492	606.4493	−0.13	16.71	0.034	0.006	2.03E-13		2.27E-13
Cer(d18:0/14:0) *	512.5035	512.5037	−0.39	16.95	0.007	0.007	2.03E-13		2.27E-13
Cer(d18:0/16:1) #	538.5191	538.5194	−0.56	17.02	0.004	0.005	2.03E-13		2.27E-13
Hexcer(d18:1/16:0) *	700.5723	700.5722	0.14	17.47	0.049	0.230	2.03E-13		2.27E-13
Ganglioside GM3 (d18:1/16:0) #	1152.7207	1152.713	6.68	17.47	0.082	0.277	2.60E-13		2.03E-13
Cer(d18:1/18:1) #	564.5351	564.535	0.18	17.49	0.005	0.004	2.03E-13		2.67E-12
HexCer(d18:1/16:0) *	700.5720	700.5722	−0.29	17.82	0.080	15.487	2.03E-13		2.27E-13
Ganglioside GA1 (d18:1/16:0) *	1227.7575	1227.7572	0.20	17.85	0.046	0.644	2.03E-13		4.66E-08
Cer(d18:0/16:0) *	540.5347	540.535	−0.56	18.03	0.001	0.002	2.03E-13		2.27E-13
Cer(d18:0/18:1) #	566.5501	566.5507	−1.06	18.05	0.001	0.001	2.03E-13		2.27E-13
PC (24:0/0:0) #	608.4646	608.465	−0.66	18.19	0.006	0.006	2.03E-13		2.27E-13
SM(d18:0/14:0) *	677.5588	677.5592	−0.53	18.44	0.100	0.126	2.03E-13		2.27E-13
LacCer(d18:1/16:0) #	862.6241	862.625	−1.04	18.61	0.146	1.582	2.03E-13		3.39E-13
Cer(d18:1/16:0) #	538.5187	538.5194	−1.30	18.62	0.094	1.480	2.03E-13		2.80E-8
Cer(d18:0/18:0) *	568.5651	568.5651	0.00	19.10	0.115	0.097	2.03E-13		2.27E-13
DG(P-14:0/18:1) *	551.5031	551.5034	−0.63	19.11	6.033	3.742	2.03E-13		2.27E-13
HexCer(d18:1/16:0) *	700.5718	700.5722	−0.57	19.16	15.576	25.495	2.03E-13		2.27E-13
Cer(d18:1/14:0) #	510.4874	510.4881	−1.37	19.35	9.794	20.313	2.03E-13		2.27E-13
SM(d18:0/16:0) #	705.5900	705.5905	−0.75	19.56	0.022	0.037	2.03E-13		2.27E-13
PE(P-18:0/22:6) *	776.5571	776.5589	−2.34	19.85	0.171	0.042	4.46E-11		2.27E-13
PC(P-36:4)/PC (O- 36:5) *	766.5733	766.5745	−1.57	19.95	4.068	0.980	2.03E-13		2.27E-13
Ganglioside GM3 (d18:1/24:1) *	1263.8296	1263.83	−0.32	20.01	0.136	0.155	2.03E-13		2.27E-13
PC (42:6) #	856.5871	856.5851	2.37	20.03	12.034	1.290	1.19E-09		0.75
PC(P-30:0)/PC (O-	690.5426	690.5432	−0.87	20.04	4.944	2.674	2.03E-13		2.27E-13
30:1) * PC (40:6) #	834.5997	834.6007	−1.16	20.06	5.598	2.629	7.23E-07		0.001
Cer(d18:1/18:0) #	566.5502	566.5507	−0.88	21.37	1.363	13.237	0.002		2.27E-13
PE (P-18:0/20:3) #	754.5738	754.5745	−0.94	21.43	5.700	7.345	1.12E-11		1.28E-09
Cer(d18:0/24:0) *	652.6586	652.6602	−2.45	21.56	0.201	0.183	2.03E-13		2.27E-13
PE(P-18:0/22:4) #	780.5898	780.5902	−0.54	21.72	6.294	6.888	1.38E-08		4.24E-07
HexCer(d18:1/24:1) #	810.6806	810.6817	−1.36	21.93	13.577	15.058	2.03E-13		2.27E-13
HexCer (d18:1/22:0) #	784.6651	784.6661	−1.27	22.00	5.569	7.317	2.03E-13		2.27E-13
SM(d18:0/20:0) #	761.6527	761.6531	−0.51	22.01	0.073	0.073	2.03E-13		2.27E-13
DG (18:3/20:1/0:0) #	645.5454	645.5452	0.33	22.03	0.186	0.366	2.03E-13		3.52E-13
DG (20:3)/20:4/0:0) *	667.5272	667.5296	−3.57	22.03	0.182	0.465	2.03E-13		2.18E-09
DG (20:2/20:3/0:0) *	671.5604	671.5609	−0.71	22.04	0.122	0.090	2.03E-13		2.27E-13
HexCer(d18:1/24:1) #	810.6817	810.6817	0.00	22.57	604.579	0.860	2.03E-13		2.27E-13
DG (20:3/20:3/0:0) *	669.5424	669.5452	−4.15	22.70	0.226	1.250	2.03E-13		0.131
DG (18:3/20:0/0:0) *	647.5604	647.5609	−0.83	22.70	0.154	1.152	2.03E-13		0.87
Cer(d18:1/22:0) #	622.6129	622.6133	−0.64	23.20	5.797	2.893	2.03E-13		2.27E-13
Cer(d18:0/24:1) #	650.6445	650.6446	−0.15	23.48	0.002	0.117	2.25E-13		0.0002
Cer(d18:0/22:0) *	624.6290	624.6289	0.16	23.54	0.070	0.202	2.03E-13		2.27E-13
Cer(d18:1/23:0) *	636.6287	636.6289	−0.31	23.62	5.199	0.743	2.03E-13		2.27E-05
Cer(d18:1/26:0) #	678.6756	678.6759	−0.44	24.77	1.862	0.073	6.32E-12		2.27E-13

PC: Phosphatidyl choline, HexCer: Hexosyl ceramide, Cer: Ceramide, PS: Phosphatidyl serine, SM: Sphingomyelin, LacCer: Lactosyl ceramide, DG: Diglyceride, PE: Phosphatidyl ethanolamine,

^*Compounds identified by accurate mass information alone,

^#Compounds identified by accurate mass information and confirmed by diagnostic ions from MS/MS spectra.

Table 3. Pearson Coefficient between the QC samples.

	RB-QC-1	RB-QC-2	RB-QC-3
RB-QC-1	1	0.995	0.9992
RB-QC-2	0.995	1	0.9963
RB-QC-3	0.9992	0.9963	1

3.2. Dysregulation of phospholipid metabolites

Multiple classes of glycerophospholipids were identified including Phosphatidylcholines (PC), Phosphatidylethanolamine (PE) and Phosphatidylserine (PS) from the lipidomic data. We observed significant changes in the abundance of lysophospholipids, plasmalogen PEs and plasmalogen PCs. Lyso lipids, PC (22:0/0:0), PC (24:0/0:0) and PC (24:1/0:0) were down regulated in NCC-RbC-51 compared to MIO-M1cells with a log₂ fold change value of 4.75, 7.49 and 4.88 respectively. These three lipids were also downregulated in WERI-Rb 1 cells with a log2 fold change value of 5.52, 7.31 and 7.45 respectively. Identification of Lyso PCs was done by accurate mass and confirmed by the presence of characteristic ion at m/z 104 in the MS/MS spectra (Xu F et al., 2009). Similarly, we identified ether lipids. A search in the lipid Maps database against the mass, 494.3603 resulted in only one lipid, PC (P-17:0/0:0) with in 10 ppm error. A search for the mass 766.5733 in Lipid Maps database resulted in 4 ether lipids with same mass. Two are alkyl lipids and two are plasmalogen lipids. They are PC(O-16:0/20:5), PC(O-16:1/20:4), PC(P-16:0/20:4) and PC(P-18:0/18:4). Hence, tentatively mass 766. 5733 can be correlated to the structure PC(P-36:4)/PC (O-36:5). A search for the mass, 690.5426 resulted in 2 alkyl and 2-plasmalogen lipids with same mass. They are PC(O-14:0/16:1), PC (O-16:0/14:1), PC(P-16:0/14:0) and PC(P-18:0/12:0). Hence, tentative structure PC(O-30:1)/PC(P-30:0) can be correlated to the mass 690.5426. These ether lipids are 4-fold high abundant in NCC-RbC-51 cells compared to MIO-M1cells (Fig. 4). WERI-Rb 1 cells showed similar trend but levels of PC(P-36:4)/PC(O-36:5) are comparable with control. Plasmalogen phosphoethanolamines were also found to be dysregulated. PE(P-18:0/22:4) and PE (P-18:0/20:3) were high abundant whereas PE(P-18:0/22:6) was low abundant in NCC-RbC-51 cells compared to MIO-M1cells (Fig. 4). A similar trend was observed in WERI-Rb 1 cells also. The head group in PE(P-18:0/20:3) was identified by a neutral loss of 141 Da from the protonated ion, 754.5738 giving rise to a fragment ion at m/z 613.5 in the MS/MS spectra. A characteristic fragment ion at m/z 392 indicated the presence of 18:0 alkenyl chain at the sn-1 position (Berry and Murphy, 2004). Fragment ion at m/z 363 confirmed the presence of acyl chain 20:3 at sn-2 position. Head group in PE(P-18:0/22:4) was identified by a neutral loss of 141 Da from the protonated ion, 780.5898 giving rise to a fragment at m/z 639.5 in the MS/MS spectra. Characteristic fragment ion at m/z 392 indicated the presence of 18:0 alkenyl chain and ion at m/z 389 indicated acyl chain 22:4 at sn-2 position. PE(P-18:0/22:4) is isobaric with PE(P-20:0/20:4) and so coeluted. Though the MS/MS spectra is a mixture of ions from both the parent ions, as the acyl chains are different, they gave rise to different product ions. PE(P-18:0/22:6) was identified by accurate mass and MS/MS spectra of the corresponding sodium adduct of the compound showed a neutral loss of 141 Da from the parent [M+Na]⁺ ion confirming the phosphoethanolamine head group.

Fig. 4. Heatmap of phospholipids observed in MIO-M1, WERI-Rb 1and NCC-RbC-51 cells.

“Row min” from the heat map scale corresponds to the lowest value for that compound in all the samples and coded dark blue. “Row max” corresponds to the highest value for that compound in all the samples and coded dark red. Each biological replicate mentioned is the average of three technical replicates. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

3.3. Dysregulation of sphingolipids

Another major class of metabolites dihydroceramides, ceramides and glycated ceramides from ceramide pathway were also observed to be dysregulated in NCC-RbC-51 and WERI-Rb 1 cells (Fig. 5A, Fig. 6). Glucosyl ceramides and galactosyl ceramides are referred to as monohexosyl ceramides, as they cannot be separated and co-eluted in the current experimental conditions. Dihydroceramides and ceramides are identified by the diagnostic ions at m/z 266 and 264 respectively in their MS/MS Spectra (Hsu, 2018). In total, 8 dihydroceramides with d18:0 backbone were down regulated in cancer samples (Fig. 5A). Subsequent downstream metabolite, ceramide with d18:1 backbone was also found containing a variety of acyl chains C14:0, C16:0, C18:0, C18:1, C22:0, C23:0, and C26:0. Ceramides with acyl chains C14:0, C18:0 and C22:0 were up regulated both in NCC-RbC-51 and WERI-Rb 1 cells. Ceramides with acyl chains C23:0 and C26:0 were high abundant in NCC-RbC-51 cells whereas WERI-Rb 1 cells showed an opposite trend for these molecules (Fig. 7). Ceramides with common acyl chain C18:1 was down regulated in both NCC-RbC-51 and WERI-Rb 1 cells. Surprisingly, ceramide with acyl chain C16:0 is down regulated in NCC-RbC-51 but up regulated in WERI-Rb 1. Hexosyl ceramides were identified by a characteristic loss of 180 Da from the protonated parent ion (Fig. 5B). Identical mass corresponding to hexosyl ceramide (d18:1/16:0) was observed in 3 peaks eluted at 17.47, 17.81 and 19.16 min. This could be due to the presence of isomeric species. Similarly, another hexosylceramide (d18:1/24:1) at m/z 810.68 was eluted at 21.93 and 22.57. Hexosyl ceramide at 22.57 min was highly dysregulated with a fold change of 604 in NCC-RbC-51 cells compared to MIO-M1cells. Surprisingly, this metabolite has comparable levels in WERI-Rb 1and MIO-M1cells. Box and whisker plot of this molecule clearly showed the higher abundance of hexosyl ceramide in NCC Rbc-51 compared to the other tw o samples (Fig. 9). Complex glycosphingolipids like lactosylceramide and gangliosides were also found in these three cell lines. Lactosyl ceramide was identified by a loss of 324 Da (sequential loss of two hexose moieties of 162 Da each) from the protonated parent ion (Fig. 5C). Lactosyl ceramide was upregulated in WERI-Rb 1and down-regulated in NCC-RbC-51 cells. GM3 Ganglioside was identified by the initial loss of 291 Da corresponding to Sialic acid from protonated parent ion followed by sequential loss of 162 Da and 162 Da (Fig. 5D). All higher order glycosphingolipids (3 gangliosides) were downregulated in NCC-RbC-51 cells and WERI-Rb 1 compared to MIO-M1cells. Extract ions chromatograms of three gangliosides showed higher abundance in control MIO-M1 cells compared to cancer cells (Fig. 8). We observed three ions at m/z 677. 5844, 705.5899 and 761.6527 in all the cell lines. Lipid maps identified these masses as sphingomyelins. Accurate mass data showed less than 1 ppm error with theorical mass (Table 1). All the three ions are up regulated in WERI-RB1 compared to control and NCC-Rbc-51.

Fig. 5. Heatmap and MS/MS spectra of sphingolipids from MIO-M1, Weri-Rb1 and NCC-Rbc-51 cells.

A. Heatmap of sphingolipids identified across MIO-M1, Weri-Rb1 and NCC-Rbc-51 cells. B. MS/MS spectra of Hexosyl Ceramide (d18:1/22:0) C. MS/MS spectra of Lactosyl Ceramide (d18:1/16:0) and D. MS/MS spectra of Ganglioside (d18:1/22:0). “Row min” from the heat map scale corresponds to the lowest value for that compound in all the samples and coded dark blue. “Row max” corresponds to the highest value for that compound in all the samples and coded dark red. Each biological replicate mentioned is the average of three technical replicates. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 6. Box-Whisker plots of dysregulated lipid species (***P < 0.0001) Cer: Ceramide, SM: sphingomyelin, PC: Phosphatidyl choline.

Fig. 7.

Representative extract ion chromatogram of ion at m/z (A) 494.3 corresponding to PC(P-17:0/0:0) and (B) 622.6 and 636.6 corresponding to Ceramide (d18:1/22:0) and ceramide (d18:1/23:0) from all the three different cellular samples each indicated by a different colour. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 9. Box-whiskers plot of Hexosyl ceramide observed in MIO-M1, WERI-Rb 1 and NCC-Rbc-51 cells.

Fig. 8.

Representative extract ion chromatogram of ion at m/z 1157.7, 1227.7 and 1263.8 corresponding to GM3 (d18:1/16:0) and GA1 (d18:1/16:0) and GM3 (d18:1/24:1) from all the three different cellular samples each indicated by a different colour. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

3.4. Comparison between NCC-RbC-51 and WERI-Rb 1

Current study identified differences between aggressive (NCC–RbC-51) and nonaggressive (WERI-Rb 1) cells. Glucosyl ceramide (d18:1/24:1) was highly dysregulated in NCC-RbC-51 cells with a fold change of 604 compared to WERI-Rb1with a fold change of 0.86 (Fig. 9). Sphingolipids with long and uncommon acyl chains are high abundant in NCC-RbC-51 than in WERI-Rb 1. Fold change for Ceramide(d18:1/22:0) in NCC-RbC 51 is 5.79 and it is 2.89 in WERI-Rb 1. Similarly, fold change of Ceramide(d18:1/23:0) in NCC-RbC-51 is 5.19 and 0.743 in WERI-Rb1. Contrary to this, sphingolipids with more frequent acyl chains were low abundant in NCC-RbC-51 cells compared to WERI-Rb 1 cells. Fold change of Ceramide (d18:1/18:0) is 13 in WERI-Rb 1and in 1.3 in NCC-RbC-51. Fold change of Ceramide(d18:1/16:0) is 1.48 in WERI-Rb 1and 0.09 in NCC-RbC-51.

3.5. Down regulation of ceramide pathway gene expression

We investigated if the observed down regulation of ceramides is due to the transcription control of the ceramide pathway genes. Expression of the 5 genes studied was low in NCC-Rbc-51 (p < 0.001) and WERI-Rb1 (p < 0.004) cancer cells compared to MIO-M1cells. (Fig. 10). However, the fold change comparison in gene expressions between two types of cancer cells was not significant (p > 0.05).

Fig. 10. Differential gene expression of sphingolipid pathway genes in WERI-Rb-1 and NCC-RbC- 51.

The ceramide mediated apoptosis regulation is identified recently as the cause for drug resistance phenomenon in neuroblastoma cells (Çoku et al., 2022). We envisaged a similar differential drug resistance/-susceptibility between the RB cell lines due to differential modulation in Ceramide pathway. Therefore, we studied drug resistance phenomenon in the RB cell lines. NIMA related kinase (NEK) was used as oncogenic and drug resistant marker (Marina and Saavedra, 2014). NEK2 was upregulated in RB tumours compared to the retina (Fig. 12). The up regulated NEK2 expression led us to study if there was differential response to drugs in RB cell lines. A well-known cell cycle inhibitor NSC663284 was used to test the drug resistance in NCC-RbC-51 cells compared to WERI-Rb1cells. The cell cycle inhibitor was chosen as this process is dysregulated in RB tumours (Rajasekaran S et al., 2019). The inhibitor dysregulated CDC25 proteins, which effects the cell cycle progression, and was proposed as drug molecule to regulate tumorigenesis (Pu L et al., 2002; Guo et al., 2007). The two cell lines were treated with NSC663284 and NEK2 gene expression was used as drug resistance/susceptibility marker. Interestingly, significant lowered expression of NEK2 gene, which is known to be associated with drug sensitivity, was observed in WERI-Rb1cells after the drug treatment when the gene expression was normalized to 0 h gene expression (P < 0.05). However, the gene expression remained unaltered in NCC-RbC-51 cells after the drug treatment (p > 0.05). (Fig. 13). We then tested if the NEK2 protein localization was altered in response to the cell cycle inhibitor as this could indicate drug resistance phenomenon. NEK2 protein is predominantly expressed in nucleus in NCC-RbC-51, whereas the expression is cytosolic in WERI-Rb1(Fig. 13B). The cytosolic expression of the protein is known to be associated with less aggressive phenotype, whereas the nuclear localization is known to be observed in more cancer cells with aggressive phenotype. Therefore, we checked if the localization will be altered after the drug treatment. The drug treatment resulted in unaltered protein localization in NCC-RbC- 51 cells. Further, WNT signalling could play an active signalling role for the observed drug resistance. Therefore, we studied if WNT signalling is involved in the drug sensitive/resistance phenotype. The LiCl treatment, a WNT signalling inhibitor was tested on both control (Weri-Rb-1) and NCC-RbC-51 cells (Supplementary Fig. 1). The NEK2 gene expression was significantly decreased in Weri-Rb1 cells (p < 0.05) after normalized to 0hr value (P < 0.05) whereas no significant decrease in the NEK2 gene expression was observed in NCC-RbC-51 cells (P > 0.05).

Fig. 12.

A. Expression levels of NEK2 in RB tumours and retina. The gene expression data is down loaded from GSE125903 from GEO website. The NEK2 gene expression is plotted from 3 retina samples and 7 RB tumour samples. B. Gene Expression of NEK2 in WERI-Rb1(control) and NCC-RbC-51.

Fig. 13.

(A). Gene Expression of NEK2 in WERI-Rb1(control) and NCC-RbC-51 on treatment with CDC inhibitor, NSC663284. The expression levels in both cell lines were normalized to the expression to GAPDH housekeeping gene expression. The effect of drug treatment on NEK2 gene expression is studied after normalizing to 0 h drug treatment. **P < 0.05 (B). Differential NEK2 protein localization using Immunofluorescence in WERI-Rb1and NCC-RbC-51 cells. Bottom panel- The nuclear localization of NEK2 protein expression is observed after CDC inhibitor drug treatment in NCC-RbC-51 cells.

4. Discussion

Phospholipid dysregulation in cancer biology is extensively studied and several potential lipid biomarkers are reported (Perrotti F et al., 2016). Monoacyl glycerophospholipids are also called as lyso phospholipids. Lyso phospholipids are vital cellular components and often mediate signalling. Lyso PCs are the substrates for lysophosphatidyl acyl choline transferase (LPCAT1) which transfers an acyl chain to make it a diacyl PC. Higher expression of LPCAT1 was correlated to the progression of breast cancer and prostate cancer (Zhou X et al., 2012; Abdelzaher and Mostafa, 2015). Additionally, low abundant lysoPCs are reported in clear cell renal cell carcinoma (Du Y et al., 2017). Down-regulation of lysoPCs can be attributed to its faster conversion into diacyl PC in the RB cells. Another possibility is the faster exchange of pre-existing PC/PE with PS (Kim et al., 2014). However, the mechanism needs further investigation in RB cells.

Sphingolipids play a vital role in the functioning of different cell types present in the retina and thus a dysregulation in sphingolipid metabolism can contribute to multiple retinal pathologies (Mondal and Mandal, 2019). Sphingolipid biosynthesis is initiated by the reaction between serine and palmitoyl CoA and catalyzed by Serine-palmitoyl transferase (SPTLC) Palmitic acid contains fatty acyl chain with 16 carbons. Reaction of Palmitoyl CoA with serine in presence of SPTLC gives rise to ketodihydosphinganine with a hydrophobic chain containing 18 carbons, one keto group and one amino group. This entity is further reduced by ketodihydro reductase (KDSR) to generate dihydrosphingosine. This 18:0 carbon skeleton is (0 denotes number of double bonds) often termed as sphingolipid backbone. Ceramide synthase (CerS) transfers a fatty acyl chain to the amino group of dihydrosphingosine. Condensation reaction between the amine and fatty acid yields to the generation of dihydroceramide with newly formed amide bond. Finally, ceramide will be formed with the insertion of a double bond in the sphingolipid backbone by dihydroceramide desaturase (DEGS1) (Fig. 11). Any dysregulation in the upstream enzymes SPTLC, KDSR, CERS and DEGS1 can lead to hampered sphingolipid pathway. In the current study, low expression of KDSR and SPTLC could be the reason behind low amounts of 8 dihydroceramides. Low expression of CERS2 and CERS6 also might have played a vital role in the low abundance of some of these dihydroceramides. Ceramide synthase family till date known to have 6 members (CERS1-6). Fatty acyl chains with C24:0 and C24:1 are preferred targets for CERS2 and short chain fatty acyl chains like C16:0 are preferred targets for CERS6 (Mizutani et al., 2005). Gene expression of CERS2 and CERS6 is low in both NCC-RbC-51 and MIO-M1 cells. Low abundance of Cer(d18:0/24:0), Cer (d18:0/24:1), Cer(d18:1/16:0) and Cer(d18:1/16:1) in both WERI-Rb-1 and NCC- RbC-51 compared to MIO-M1is complementing the gene expression results.

Fig. 11. Pathway map of sphingolipid biosynthesis.

CerS2 is also known as longevity assurance gene, LASS2. Expression levels of CerS2 was significantly lower in bladder cancer samples. CerS2 mRNA levels correlated well with clinical stage and invasion. CerS2 negative carcinomas were found to have less survival in bladder cancer (CerS2-BC) (Aldoghachi AF et al., 2019). The longer survival of some ovarian cancer patient’s post-progression was attributed to high levels of CERS2 (Zhang et al., 2021). Further, a xenograft model showed metastasis when CERS2 is knocked down by siRNA in parental SKVO3 cells. Moreover, importance of ceramide synthases is highly pronounced in retinal function. Electroretinograms (ERG) and retinal morphology were studied in mice deficient in CERS1, CERS2 and CERS4. Though all the mice showed reduced amplitudes in ERG recordings, CERS2 deficient mice showed the most severe phenotype (Brüggen et al., 2016). These evidences highlight the pivotal role played by CERS2 in the healthy status of retinal cells. In the wake of down regulation of upstream enzymes, sphingomyelinase pathway helps the cell to regenerate ceramides mediating growth, differentiation and apoptosis (Vento R et al., 1998). Sphingomyelins are hydrolysed by sphingomyelinase releasing phosphocholine and ceramide. This could be the reason for the high abundance of few ceramides and low abundance of sphingomyelins. Down regulation of sphingomyelins in RB cells agrees with lowered levels of the lipid compounds observed in prostate cancer cells (Goto et al., 2015).

Ceramide-glucosylceramide rheostat in sphingolipid metabolism plays a crucial role in tumor progression, drug resistance, and chemo-therapeutic response. Shift towards glucosylceramide synthesis serves as a metabolic sink that can suppress the anti-proliferative actions of ceramides. Multi drug resistant cell lines as well as patients who have failed chemotherapy found to have increased levels of Glucosyl ceramide (Lavie Y et al., 1996; Lucci et al., 1998). Higher levels of Hexosyl ceramide (d18:1/24:1) in aggressive cell line, NCC-RbC-51 are in coherence with literature. But the expression of gene, UGCG which participates in the biosynthesis of glucosyl ceramide is low in the current study. Hence, biosynthesis of glucosyl ceramide may not be the reason behind higher levels of hexosyl ceramide. This observation hints at a hampered catabolic pathway for the accumulation of sphingoglycolipids in RB. Glucosylceramidase hydrolyzes glucosylceramide to yield ceramide (Futerman and Platt, 2017). Saposin C activates glucosylceramidase facilitating access to the glycosphingolipid substrate (Sun Y et al., 2003). Inactive or defunct glucosyl ceramidase (GBA) or its binding partner Saposin C could be reasons behind the accumulation of hexosyl ceramide in the present study. Elevated amounts of hexosyl ceramides in a prosaposin deficient fetal kidney are reported (Bradovfi et al., 1993). Decreased expression of glucosyl ceramide synthase UGCG will affect the biosynthesis of all complex sphingoglycolipids as attachment of glucose moiety to the sphingosine backbone is the first step in the biosynthesis. Down regulation of lactosyl ceramide and higher order sphingoglycolipids can be correlated to this. Additionally, Low expression of gene ST3GAL5 which synthesizes GM3 lipids confirms an impaired biosynthesis of higher order sphingoglycolipids.

Gangliosides are bioactive signalling molecules and are essential in the maturation, differentiation and activation of T cells. Ganglioside repertoire helps in the maturation of different subsets of T cells playing a crucial role in human immune system. Gangliosides and ST3GAL5 are known to inhibit Epithelial-Mesenchymal transition, a process by which cells gain migratory and invasive properties (Nagafuku M et al., 2012). Both ST3GAL5 and gangliosides are down regulated in lung adenocarcinoma and bladder cancer (Zhang J et al., 2023; Ouyang S et al., 2020). Tumour cells escape from T cell surveillance. Down regulation of gangliosides may be playing a role in such an escape. A similar scenario of defective ST3GAL5 and dysregulated lipids were observed in neurological disorders (Gordon-Lipkin E et al., 2018; Boccuto L et al., 2014). The accumulation of hexosyl ceramides is a common feature in multidrug resistant cancer cells from different tissues of origin (Lavie Y et al., 1996; Veldman RJ et al., 2002). Oxaliplatin resistant colorectal cancer cells exhibited upregulated glucosyl ceramide and corresponding enzyme (Madigan JP et al., 2020). Glucosyl ceramide accumulation is observed in drug resistant ovarian cancer cells (Lavie Yet al., 1996). Hence, up regulation of glucosyl ceramide seen in NCC-RbC-51 could indicate drug resistant phenotype in RB. From our earlier studies, we noticed that NCC-Rbc-51 has higher proliferation rate and migration rate compared to WERI-RB1. The differential response of these cells to stiffness and cancer drugs reveals that NCC-RBc-51 is more aggressive cell line with drug resistance phenomenon (Ravishankar H et al., 2020; Divya et al., 2022; Maradani et al., 2022)). Whereas WERI-Rb1 cells are reported to be drug sensitive(Oronowicz et al., 2021; Reinhard et al., 2020; Sen et al., 2022; Subramanian et al., 2012). Further, NEK2, a prominent oncogene, marker for drug resistance is overexpressed in RB tumours (Fig. 12). Unaltered NEK2 in NCC-RbC-51 compared to WERI-Rb1 upon the treatment of cell cycle inhibitor indicated drug resistance phenotype. The nuclear localization of NEK2 in NCC-Rbc-51 compared to predominant cytosolic localization of this protein in WERI-Rb1 indicated overactive NEK2 mediated cell cycle progression. (Fig. 13). The nuclear localization of NEK2 was observed in several tumours (Barbagallo F et al., 2009). The drug mediated response could be through WNT signalling in the NCC-RbC-51 cells as these cells did not respond to the LiCl treatment. The NEK2 mediated drug resistance in breast cancer cells was through stabilizing WNT signalling mediated through β catenin (Shen H et al., 2019). A similar drug resistance mechanism is possible in NCC-RbC-51. However, the precise mechanism of drug resistance with altered metabolism in NCC-RbC-51 needs further investigation.

5. Conclusion

We conclude that there is a metabolic shift in RB cancer cells. Sphingolipid synthesis is significantly hampered in RB.

Supplementary Material

Supplementary data to this article can be found online at https://doi.org/10.1016/j.exer.2024.109798.

Data availability

All the data is given in the manuscript. Additionally supporting information in the tabular form is also provided