In situ characterizing membrane lipid phenotype of breast cancer cells using mass spectrometry profiling

Lipid composition in cell membrane is closely associated with cell characteristics. Here, matrix-assisted laser desorption/ionization- Fourier transform ion cyclotron resonance mass spectrometry was employed to in situ determine membrane components of human mammary epithelial cells (MCF-10 A) and six different breast cancer cell lines (i.e., BT-20, MCF-7, SK-BR-3, MDA-MB-231, MDA-MB-157, and MDA-MB-361) without any lipid extraction and separation. Partial least-square discriminant analysis indicated that changes in the levels of these membrane lipids were closely correlated with the types of breast cell lines. Elevated levels of polyunsaturated lipids in MCF-10 A cells relative to six breast cancer cells and in BT-20 cells relative to other breast cancer cell lines were detected. The Western blotting assays indicated that the expression of five lipogenesis-related enzymes (i.e., fatty acid synthase 1(FASN1), stearoyl-CoA desaturase 1 (SCD1), stearoyl-CoA desaturase 5 (SCD5), choline kinase α (CKα), and sphingomyelin synthase 1) was associated with the types of the breast cells, and that the SCD1 level in MCF-7 cells was significantly increased relative to other breast cell lines. Our findings suggest that elevated expression levels of FASN1, SCD1, SCD5, and CKα may closely correlated with enhanced levels of saturated and monounsaturated lipids in breast cancer cell lines.


Differences in lipid levels
). It should be noted that 8 simultaneously up-or down-regulated lipids in breast cancer cells relative to control were observed, and they are PC(32:1), SM(34:0), C 22 PLS-DA core plot of 180 common membrane lipids from six different breast cancer cell lines could generate six obvious clusters, with the PRESS of 0.1885 (Fig. 2C). BT-20 (ER-/PR-) cells derived from a primary tumor 20 was located in the lower left hand quadrant, MCF-7 (ER+ /PR+ ) cells from a metastatic tumor with a weakly invasive ability was in the lower right hand quadrant, highly invasive MDA-MB-231 (ER-/PR-) cells 21,22 and MDA-MB-157 (ER-/PR-) cells 23 derived from metastatic pleural effusions of human breast carcinoma were found in the upper left hand quadrant, and MDA-MB-361 (ER+ /PR+ ) cells from a brain metastasis with weakly luminal epithelial-like phenotype 24 were found in the upper right hand quadrant. Interestingly, SK-BR-3 (ER-/PR-) cells with weakly luminal epithelial-like phenotype 20 were located at the right side of horizontal axis. It was found that a total of 49 lipids with the VIP values of > 1.0 have contributed to characterize their differences. The detailed information on their identification is shown in Supplementary Table S3 (36:4), and PC(36:1) in MCF-7 cells relative to other five cancer cells were obviously observed, the levels of PI(34:1) and PC(28:0) in SK-BR-3 relative to other five cancer cells were significantly increased, the levels of PI(36:1), PC(36:1), and PC(36:2) in MDA-MB-231 relative to other five cancer cells were obviously increased, the levels of PI(36:1), PC(28:0), and SM(34:1) in MDA-BR-157 relative to other five cancer cells were significantly decreased, and the levels of PE(O-38:5) and SM(34:1) in MDA-MB-361 relative to other five cancer cells were significantly increased. Their detailed change trends are shown in Supplementary Fig. S2B, and the p values of the above-mentioned 15 lipids between six different breast cancer cell lines are listed in Supplementary Table S4.
Correlation and cluster analysis of characteristic lipids in seven different breast cancer cells. To confirm further the associations between the levels of the membrane lipids with specific fatty acyl chains and saturation with different breast cancer cells, Spearman correlation analysis was carried out to explore their relationships. Fig. 3 show the correlation coefficients (r) between lipid species highlighted in colors for different cell lines, and Spearman correlation analysis indicated that r > 0.35 or < − 0.35 indicate statistically significant difference (p < 0.05) between lipid species. The detailed information is listed in Supplementary Table S5-S11. For MCF-10 A cells as a control, saturated FAs (i.e.,C 16:0 and C 18:0 ) were positively correlated with monounsaturated FAs (i.e.,C 16:1 and C 18:1 ), and both were negatively Hierarchical cluster analysis for the lipids of interest was described that used 'Ward' s linkage' algorithms to arrange lipids according to similarity measure. As shown in and PC(30:0)), detailed cluster analysis has further showed obvious associations of the their expression levels with the types of breast cell lines (Fig. 4B). In addition, a heatmap of 8 FAs has indicated that their levels were also associated with the types of different breast cell lines (Fig. 4C).
Association of expression levels of FASN1, SCD1, SCD5, CKα, and SMS1 with the types of breast cells. To determine the link of the expression levels of lipogenesis-related enzymes with the types of breast cells, the expression of FASN1, SCD1, SCD5, CKα , and SMS1 in seven different breast cell lines were detected using Western blotting. As shown in Fig. 5 and Supplementary Table S12, significantly increased levels of FASN1 in BT-20, MCF-7, SK-BR-3, and MDA-MB-361were detected relative to MCF-10A, MDA-MB-231, and MDA-MB-157. It should be noted that the SCD1 level in MCF-7 was significantly increased relative to other six breast cell lines (p < 0.001), while difference in the SCD1 level in

Discussion
Conventional lipidomic analytical approaches usually involve a complex sample preparation, including lipid extraction, separation by chromatography followed by mass spectrometric detection. In this study, mass spectrometry profiling was performed to in situ detect membrane lipids of seven breast cell lines  without any lipid extraction and separation. It was found that direct biological analysis not only maintained cell integrity but also contributed to obtain rapidly lipid profiles. In addition, the application of FTICR MS could offer ultrahigh accurate mass measurements 25 . Finally, we have observed 180 common membrane lipids in seven different breast cell lines (Supplementary Table S1). In addition, 23  , CKα (E) and SMS1 (F) relative to the level of GAPDH in seven breast cell lines. The data are expressed as mean ± standard deviation (SD) in the triplicate, and statistically significant differences (p values) of these enzyme intensity between these breast cell lines are listed in Supplementary Table S12. PE (36:2), and PC(32:0 ) were again analyzed using electrospray ionization (ESI) technique, and most of them have similar change trends between MALDI and ESI. The detailed information is shown in Supplementary Fig. S2.
PLS-DA score plot has revealed that a combination of 8 membrane lipids (i.e., PC(32:1), SM(34:0), C 22:4 , PC(38:4), PI(38:4), PC(30:0), C 18:2 , and C 16:1 ) could be used to obviously differentiate non-malignant human breast epithelial cells from six different breast cancer cell lines, suggesting that differences in membrane lipid components between non-malignant cells and malignant cells exist, and that significantly increased polyunsaturated lipids (i.e., C 22:4 , PC(38:4), PI(38:4), and C 18:2 ,) in the non-malignant cells relative to the malignant cells revealed the close link of the unsaturated degree of membrane lipids with the types of breast cancer cells. In addition, PLS-DA score plot of 180 lipids from six different breast cancer cell lines has remarkably classified them into six clusters (Fig. 2C), indicating that the membrane lipid components of six different breast cancer cells might also reflect differences in their membrane components. It is worth noting that the highest expression of polyunsaturated lipids (i.e., PE(38:4), PE(P-38:5), PE(36:4), and C 20:4 ) in BT-20 cell membrane from a primary tumor, along with a weakly invasive ability and negative for estrogen and progesterone receptors were observed relative to other five breast cancer cell lines derived from metastatic breast cancer tumors (Fig. 2D) 20 , indicating that the unsaturated degree of membrane lipids are closely associated with the types of breast cancer cells.
Substantial studies have indicated that FASN as a key enzyme in lipogenesis catalyzes de novo synthesis of palmitate (C 16:0 ) from acetyl-CoA, and that the over-expression of FASN can suppress tumor necrosis factor-α production 26 and contribute to poor prognosis in various types of cancer 27,28 . As shown in Fig. 5B, our data show increased level of FASN1 in BT-20, MCF-7, SK-BR-3, and MDA-MB-361 cells relative to MCF-10A, MDA-MB-231, and MDA-MB-157 was observed, but no correlation with its corresponding products, such as C 16:0 and C 18:0 , was observed ( Supplementary Fig. S3). One reason may be that free C 16:0 and C 18:0 in cells are usually located in cytoplasm, and they are difficultly detected by mass spectrometry and another reason may be that they are rapidly consumed to generate membrane lipids (e.g. PC(30:0) and PC(32:1)) or converted to other long-chain FAs by elongases and desaturases 29 , resulting in an increase in the levels of lipids (e.g. PC (34:1), PI(34:1) and PI(36:1)). Thus, the over-expression of FASN1 in cancer cells sustains an increasing demand for saturated and monounsaturated lipids during cell proliferation. It was found that the levels of polyunsaturated species (e.g., PC(38:4) and PI(38:4) ) were down-regulated in most cancer cells. Some polyunsaturated FAs (e.g. C 22:6 and C 20:5 ) have been recognized as potential adjuvants against breast cancer cell proliferation, migration, and invasion 30 .
SCDs mainly catalyze the synthesis of C 16:1 and C 18:1 from C 16:0 and C 18:0 , respectively. Previous studies have indicated that SCD1 (one isoform of SCD) played an important role in cancer progression 31 . In this study, it is found that high expression of SCD1 was detected in MCF-7 relative to BT-20, SK-BR-3, MDA-MB-231, MDA-MB-157, and MDA-MB-361 (Fig. 5C), which is in agreement with the high ratios of C 16:1 /C 16:0 and C 18:1 /C 18:0 (Fig. 6), and that significant increase in the SCD5 level (another isoform of SCD) in SK-BR-3 and MDA-MB-231cells might be another factor to increase the ratios of C 16:1 /C 16:0 and/ or C 18:1 /C 18:0 , but we could not explain what caused the high ratio of C 16:1 /C 16:0 in MDA-MB-361 cells (Fig. 6), which may be associated with other SCD enzymes.
PCs can be synthesized through cytidine diphosphate-choline pathway and choline kinase (CK) catalyzes the first step reaction in the choline pathway for de novo synthesis of PCs 32 . It was found that CKα , one isoform of CKs, was associated with the uncontrolled growth of cancer cells 33 . Our findings indicate that increased CKα level in BT-20, MCF-7, SK-BR-3, and MDA-MB-231 cells could lead to generate higher levels of PC(28:0), PC(30:0) and PC(32:1) relative to MCF-10 A cells ( Supplementary Fig. S2). Specifically for MDA-MB-231 cells, it is found that the highest CKα level (Fig. 5E) was remarkably associated with the highest level of PC(36:1) and PC(36:2) relative to other six different breast cells. In the Golgi apparatus, SMS1 can utilize PC species as its substrate to produce SMs 34 . In this study, it is worth noting that the expression level of SMS1 in BT-20, MCF-7, SK-BR-3, and MDA-MB-361 (Fig. 5F) was positively and closely correlated to change in the SM(34:1) level, specially for MDA-MB-361, but not related to the SM(34:0) level.
Gene microarray analysis is often used to be diagnostic biomarkers and distinguish metastatic ability of breast cancer cells 35 . Our results indicate that direct membrane lipid profiling coupled with the expression detection of lipogenesis-related enzymes may provide another insight to understand differences in molecular mechanisms of different breast cells, and to help distinguish different breast cell lines. Altogether, the correlations of changes in the levels of membrane lipids with the expression levels of five lipogenesis-related enzymes of six different breast cancer cell lines are shown in Fig. 7. It is found that polyunsaturated lipids in cancer cell membrane from a primary tumor, such as BT-20, were significantly increased relative to other five breast cancer cell lines from metastatic breast cancer tumors. It is worth noting that polyunsaturated lipids in breast cell membrane were really synthesized by themselves, not from exogenous fat, and that the unsaturated degree of membrane lipids for highly invasive breast cancer cells from metastatic breast cancer tumors was significantly decreased relative to healthy cells or breast cancer cells from a primary tumor. Most importantly, our results indicate that the membrane lipid phenotype and lipogenesis-related enzymes of breast cell lines might be associated with their malignancy. It should be noted that in this study some questions still remain unanswered, including the relationships between changes in the levels of membrane lipids, the lipogenesis-related protein expression, and cell properties for seven different cell lines. Further studies are need to determine these associations for each individual cell lines, such as by the addition of different exogenous lipids as cell food 36 .

Conclusions
In this study, we have performed an in situ detection of membrane lipid profiles of seven different breast cell lines using an ultrahigh resolution MALDI-FTICR MS, without the requirement of separation and extraction process prior to mass spectrometric analysis. Our data show that a combination of 8 lipids could differentiate non-malignant MCF-10A cells from six different breast cancer cell lines and that a combination of 15 lipids could differentiate six different breast cancer cells. It is worth noting that significantly increased levels of monounsaturated lipids might be associated with the malignant degree of breast cancer cells. In addition, the presence of the high levels of polyunsaturated lipids in MCF-10A and BT-20 cell membranes relative to MCF-7, SK-BR-3, MDA-MB-231, MDA-MB-157, and MDA-MB-361 strongly suggest that polyunsaturated FAs were self-synthesized, not from dietary fat intake. Center, Chinese Academy of Medical Sciences. MCF-10A and BT-20 cells were grown in minimum essential medium, MCF-7 cells was cultured in Dulbecco's modified Eagle's medium, and other four cell lines were grown in Roswell Park Memorial Institute-1640 medium. All media were supplemented with 10% fetal bovine serum (FBS) and 1% antibiotics (100 U/mL penicillin and 100 μ g/mL streptomycin). Two conductive indium tin oxide (ITO)-coated glass slides were placed at the bottom of each of 100-mm culture dishes and cells were seeded in these dishes 37 , followed by the incubation until 80% confluence. All experiments were performed in accordance with relevant guidelines and regulations.
Cell sample preparation. Cell-coated ITO-glass slides were washed with phosphate buffer saline three times and then air-dried. Matrix sublimation deposition was performed as our own previous study 38 . Briefly, the cell-coated ITO-slide was mounted onto the bottom of the condenser using thermally conductive double-sided adhesive tape and the slide was kept at approximately 4 °C using a water cooling system. 10 mg of DHB (Sigma-Aldrich, St. Louis, MO, USA) was sublimed under a fixed vacuum  Supplementary Table S4 and S12.
Scientific RepoRts | 5:11298 | DOi: 10.1038/srep11298 of 5 mTorr at an optimized sublimation temperature of 120 °C for 20 min. Finally, approximately 3.5 mg of DHB was coated at the surface of one cell-coated ITO-slide followed by mass spectrometric analysis in a positive ion mode. 20 mg of DMAN (Sigma-Aldrich, St. Louis, MO, USA) was sublimed under the above-mentioned vacuum condition at 50 °C for 9 min, and approximately 7.2 mg of DMAN was coated at the surface of another cell-coated ITO-slide from the same dish for mass spectrometric analysis in a negative ion mode. Each of the matrix-coated slide was placed at -20 °C for 30 min, and then the recrystallization was performed in a Petri dish (100 mm diameter × 15 mm depth) at a saturated solution atmosphere of methanol/water (1:1, v/v). After isotope deconvolution, monoisotopic peaks and their respective corresponding intensities were obtained using DAssis software (Bruker Daltonics). Peaks were selected with a signal-to-noise ratio of > 3, relative intensity of > 0.1%, and absolute intensity of > 10,000. The resulting peaks between different samples were aligned within a narrow mass tolerance window of ± 0.001 Da as a single lipid (or one variable). The peaks from [M+ H] + , [M+ Na] + , and [M+ K] + ions in the positive ion mode were further combined as one variable. The peaks observed at least in two-thirds of seven cell lines were selected, and the half of the baseline strength in each spectrum was adopted as their intensities of missing lipids. Finally, the resulting datasets were exported to Microsoft Excel. The intensities of lipids (or variables) from each mass spectrum were normalized to a constant number of 1000. Resulting datasets in the positive and negative ion modes were further combined into one dataset before statistical analysis.

Mass spectrometry profiling.
Partial least-square discriminant analysis (PLS-DA) was performed to evaluate differences between control and different breast cancer cell lines. Univariate analysis was performed using non-parametric Wilcoxon-Mann-Whitney test. Correlation and cluster analysis were performed to analyze the correlation between saturated and unsaturated lipids. A p value of < 0.05 was considered as statistically significant. Statistical analyses were performed using SAS software (version 9.2, SAS Institute Inc.) and SPSS software (version 16.0, SPSS Inc., Chicago, IL).

Identification of membrane lipids of interest.
Significantly changed lipids were identified as our own previous studies 38,39 . Briefly, lipids were identified with the aid of the available databases (the Lipid maps (http://www.lipidmaps.org/) and the METLIN database (http://metlin.scripps.edu/)), as well as their observed accurate m/z values relative to theoretical vales of < 2 ppm, isotopic abundance distribution relative to theoretical distribution of < 5%, and tandem mass spectra.
Western blotting. Cells were harvested and suspended in RIPA lysis buffer (Solarbio Science & Technology Co., Beijing, China) which contains an antiprotease cocktail. After centrifugation, the amount of proteins in the supernatant was determined using bicinchoninic acid protein assay. Five aliquots of total proteins (30 mg) from each cell line were isolated by SDS-PAGE followed by transferring onto a polyvinylidene fluoride microporous membrane, respectively. These membranes were blocked with 5% (w/v) non-fat powdered milk in Tris-buffered saline plus 0.1% Tween-20 for 2 h at room temperature, and then incubated overnight at 4 °C with CKα -, SMS1-, SCD1-, SCD5-, and FASN1-specific antibodies (Abcam, San Francisco, CA, USA), respectively. These membranes were assayed against glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as loading control. Secondary goat anti-mouse (Southern Biotechnology Assoc., Asbach, Germany) or goat anti-rabbit (Abcam) antibodies were used at a dilution of 1:2000 with the incubation for 2 h at room temperature. Signals were detected by chemiluminescent HRP substrate (Millipore, Billerica, MA, USA) and images were acquired using ImageQuant TM LAS 4000 mini (GE Healthcare Uppsala, Sweden). The chemiluminescent signals were quantified using Quantity One software (Bio-Rad Laboratories) and were normalized to GAPDH, respectively.
One-way analysis of variance (ANOVA) was performed to determine statistically significant differences between non-malignant control and six different breast cancer cell lines. A p value of < 0.05 was considered as statistically significant.