Silencing of the glycerophosphocholine phosphodiesterase GDPD5 alters the phospholipid metabolite profile in a breast cancer model in vivo as monitored by ³¹P Magnetic Resonance Spectroscopy

Abstract

Abnormal choline phospholipid metabolism is an emerging hallmark of cancer, which is implicated in carcinogenesis and tumor progression. The malignant metabolic phenotype is characterized by high levels of phosphocholine (PC) and relatively low levels of glycerophosphocholine (GPC) in aggressive breast cancer cells. Phosphorus Magnetic Resonance Spectroscopy (³¹P MRS) is able to noninvasively detect these water-soluble metabolites of choline as well as ethanolamine phospholipid metabolism. Here we have investigated the effects of stably silencing glycerophosphoester diesterase domain containing 5 (GDPD5), which is an enzyme with glycerophosphocholine phosphodiesterase activity, in MDA-MB-231 breast cancer cells and orthotopic tumor xenografts. Tumors in which GDPD5 was stably silenced with GDPD5-specific shRNA contained increased levels of GPC and phosphoethanolamine (PE) compared to control tumors.

Introduction

Altered choline phospholipid metabolism is an emerging hallmark of cancer, which is associated with oncogenesis and tumor progression (1). Oncogenic transformation of cells changes the expression and activity of enzymes that control anabolic and catabolic pathways in membrane phospholipid metabolism, thereby altering the levels of choline- and ethanolamine-containing precursors and breakdown products (1,2) as shown in Figure 1. These choline- and ethanolamine-containing metabolites can be monitored in vivo by ³¹P magnetic resonance spectroscopy (MRS), which detects phosphomonoesters (PME) predominantly in the anabolic pathway and phosphodiesters (PDE) in the catabolic pathway (Fig.1) (3-5). In breast and ovarian cancer cells, a switch from high glycerophosphocholine (GPC) and low phosphocholine (PC) in normal cells to high PC and low GPC in malignant cells has been observed, and the PC/GPC ratio increases with cancer aggressiveness (6,7).

Figure 1.

Network of ethanolamine and choline phospholipid metabolism. MRS-detectable metabolites are shown in red (ethanolamine) and blue (choline) boxes. Anabolic enzyme reactions are depicted by solid arrows and catabolic enzyme reactions by dotted arrows. Enzymes are shown in green boxes. GDPD5, which is examined in this study, is a GPC-PDE. Silencing of GPDP5 is hypothesized to increase glycerophosphocholine levels and decrease free choline levels.

Abbreviations of metabolites: PtdSer, phosphatidylserine; PtdEth, phosphatidylethanolamine; PtdCho, phosphatidylcholine; 1-acyl-GPE, 1-acyl-glycerophosphoethanolamine; 1-acyl-GPC, 1-acyl-glycerophosphocholine. Abbreviations of enzymes: ek, ethanolamine kinase; ck, choline kinase; CTP, phosphocholine cytidylyltransferase; DG-CTP, diacylglycerol choline phosphotransferase; DG-ETP, diacylglycerol ethanolamine phosphotransferase; PLC, phospholipase C; PLC, phospholipase D; PLA2, phospholipase A2; lyso-PL, lysophospholipase; GPC-GPE, glycerophosphocholine phosphodiesterase; GPE-GPE, glycerophosphoethanolamine phosphodiesterase; GDPD5, glycerophosphodiester phosphodiesterase domain containing 5.

Increased PC levels are associated with increased proliferation, and complex reciprocal interactions exist between oncogenic signaling and choline phospholipid metabolism (1). Since MRS noninvasively detects choline-containing compounds, an increase in these compounds can be used as a noninvasive biomarker of transformation and staging (6,8). A decrease in these compounds can indicate response to therapy, including novel therapies that target oncogenic signaling pathways (9,10).

Enhanced levels of PE and GPE have also been detected in various tumor tissues (4,8). However, the role of ethanolamine phospholipid metabolism in cancer cells has not yet been studied in great molecular detail, although phosphatidylethanolamine (PtdEth) constitutes 20%–40% of all phospholipids in mammalian cell membranes (11). Apart from its structural roles in cellular membranes, PtdEth also acts as the donor of the ethanolamine moiety that covalently links glycosylphosphatidylinositol membrane anchors to terminal carboxyl groups of proteins attached to the surface of cells (12). Choline kinase alpha and beta have the ability to also use ethanolamine as a substrate to produce PE (13), which could be a reason for the high PE levels in tumors.

The metabolites in the choline and ethanolamine pathways (red and blue shaded boxes in Fig.1) can be studied with ¹H and ³¹P MRS. When using ¹H MRS, these signals overlap even when studied ex vivo with ¹H MRS at high field strengths (14). When using ³¹P MRS at high field strengths, the larger spectral separation of the ³¹P-containing choline and ethanolamine metabolites makes it possible to study them individually in vivo, however, at the cost of lower sensitivity (15,16).

Several enzymes in phospholipid metabolism (green arrows in Fig.1) could serve as potential targets for image-guided cancer therapy. The goal of phospholipid-metabolism-targeted therapy of breast cancer would be to revert the malignant metabolic phenotype of high PC and low GPC back to the normal metabolic phenotype of low PC and high GPC. Silencing and inhibition of choline kinase alpha has been shown to reduce PC levels in aggressive breast cancer cells and tumor xenografts and thereby reduce proliferation and tumor growth (10,17,18). A drug that inhibits choline kinase activity is currently being tested as cancer therapy in clinical trials (19).

Another possibility to lower the PC/GPC ratio of aggressive breast cancer cells would be to target enzymes with glycerophosphocholine phosphodiesterase (GPC-PDE) activity (20). Enzymes with GPC-PDE activity break down GPC into choline and glycerol-3-phosphate, thereby supplying the cancer cell with free choline, which is then quickly metabolized to PC due to the high activity of choline kinase alpha (20,21). We have recently shown that glycerophosphodiester phosphodiesterase domain containing 5 (GDPD5) is significantly overexpressed in highly malignant estrogen receptor negative breast cancer cells and breast tumors from patients (21). GDPD5 positively correlated with the total choline-containing metabolite level and PC/GPC ratio in human breast tumors, and showed a trend towards a negative correlation with the GPC level (21). We therefore hypothesize that silencing GDPD5 could lead to higher GPC levels and simultaneously to a reduction of free choline in the cell and therefore a reduced availability of substrate for choline kinase alpha to produce PC. This could in turn lead to a lower PC/GPC ratio, which is characteristic of the non-malignant metabolic phenotype of breast epithelial cells. To this end, we have studied the effects of stably silencing GDPD5 on the phospholipid metabolite levels in MDA-MB-231 breast cancer cells and MDA-MB-231 breast tumor xenografts in mice by using ³¹P MRS.

Materials and methods

Cells and cell culture conditions

Triple negative MDA-MB-231 breast cancer cells, which were derived from a metastatic lesion of a breast adenocarcinoma in the mammary gland, were purchased from the American Type Culture Collection (ATCC, Manassas, VA) and used within 6 months of obtaining from ATCC. This cell line was tested and authenticated by ATCC using two independent methods: the ATCC cytochrome C oxidase I PCR assay, and short tandem-repeat profiling using multiplex PCR. Cells were cultured in RPMI-1640 supplemented with 10% fetal bovine serum (FBS; Sigma), 100 units/mL penicillin, and 100 μg/mL streptomycin (Invitrogen, Calabasas, CA). RPMI-1640 contains 21.43 μmol/L of choline and no ethanolamine. Cells were maintained in a humidified atmosphere of 5% CO₂ in air at 37C°.

Cloning and lentivirus preparation

A pLKO.1 vector containing lentiviral shRNA against GDPD5 with the hairpin sequence 5’CCGGGCTCTCCGTATGTTCAGACAACTCGAGTTGTCTGAACATACGGAGAGCTTTTT-3’ was digested with SacII and Nde1. The isolated and purified shRNA insert against GDPD5 was cloned between Nde1 and EcoRV into a human U6 promoter-driven pRRL vector containing enhanced green fluorescent protein (EGFP) as a reporter gene driven by a phosphoglycerate kinase (pGK) promoter as previously described (10). An empty vector control lacking any shRNA, but expressing pGK-driven EGFP, was used as control (vector control) (22). Infectious viral supernatants (DMEM (Mediatech, Manassas, VA) with 1% FBS) were derived by transient cotransfection of 293T cells (6 × 10⁶ in 100-mm cell culture dishes) using Lipofectamine 2000 (Invitrogen, Calabasas, CA). A total of 19.5 μg of plasmid in the proportion of 12 μg of lentiviral vector carrying shRNA, 6 μg of packaging plasmid pCMVDR8.2 DVPR (VPR deleted; (23)), and 1.5 μg of pCMV-VSVG were used, and viral supernatant was collected at 48, 72, and 96 h after transfection. Pooled supernatants were concentrated using an Amicon Ultra-15 100K cutoff filter device (Millipore, Billerica, MA). The viral titer of the concentrated supernatant was determined by performing a p24 ELISA kit (Cell Biolabs, Inc. San Diego, CA) to detect the HIV-p24 core protein of the vector.

Generation of stably GDPD5-silenced breast cancer cell

For the transduction of MDA-MB-231 cells, 1×10⁶ cells were plated in 100-mm cell culture dishes. Viral supernatants were centrifuged at 3,000 × g for 20 min at 4°C. Five mL of viral supernatant with 1 mg/mL polybrene (Sigma, St. Louis, MO) was added to the cells for 4 h. This procedure was repeated for three consecutive days. The transduction efficiency was assessed by EGFP expression. Photomicrographs were taken with a Nikon TS100 inverted microscope. GDPD5 mRNA expression in transduced MDA-MB-231-GDPD5-shRNA cells relative to vector control cells was assessed by qRT-PCR as described below. The expression of target RNA was calculated relative to the housekeeping gene hypoxanthine phosphoribosyltransferase 1 (HPRT1).

Generation of stably GDPD5-silenced breast cancer tumor xenografts

Approximately 2×10⁶ MDA-MB-231-GDPD5-shRNA cells or empty vector control MDA-MB-231 (MDA-MB-231-vector control) cells in 50 μL HBSS (Mediatech) were inoculated in the upper right thoracic mammary fat pad of female athymic nude mice. Tumor volumes were measured weekly by caliper measurements. All surgical procedures and animal handling was performed in accordance with protocols approved by the Johns Hopkins University Institutional Animal Care and Use Committee, and conformed to the Guide for the Care and Use of Laboratory Animals published by the NIH.

RNA isolation, cDNA synthesis, and quantitative RT-PCR

MDA-MB-231-GDPD5-shRNA cells and empty vector control (MDA-MB-231-vector control) cells were cultured in 75-mm cell culture dishes in triplicate. When the cells were 70% confluent, they were washed quickly with DEPC-treated water twice, followed by RNA isolation using QIAshredder and RNeasy mini kit (Qiagen, Valencia, CA). cDNA was synthesized using qScript cDNA supermix (Quanta Biosciences, Gaithersburg, MD). 2μl of 1:10 diluted cDNA was used for real-time PCR performed in an iCycler IQTM real-time PCR machine (Bio-Rad). The GDPD5 specific primers used for PCR were as follows: sense strand 5’-CTACAACCCTGAGCAGAT-3’; anti-sense strand 5’-AACATACGGAGAGCACAT-3’. Normalization was performed with respect to the housekeeping genes hypoxanthine phosphoribosyltransferase 1 (HPRT1). The Primers for HPRT1 were as follows: sense strand 5’-CCTGGCGTCGTGATTAGTGATG-3’, antisense strand 5’-CAGAGGGCTACAATGTGATGGC-3’. The acquired data were analyzed in MS-Excel using the ΔΔct method. The relative fold change in gene expression of MDA-MB-231-GDPD5-shRNA cells was calculated based on the threshold cycle (ct) as R=2^-Δ(Δct), where Δct=ct( GDPD5 - hPRT) and ΔΔct= Δct(GDPD5_{MDA-MB-231-GDPD5-shRNA})- Δct(GDPD5_{MDA-MB-231-vector control}).

Dual-phase extraction of cells and tumors

For cell extracts, cells were grown to 70% confluence in standard cell culture medium, which contained 21.43 μmol/L of choline and no ethanolamine. Approximately 10⁸ cells per extract were harvested by trypsinization with 0.25 % Trypsin-EDTA Solution (Sigma-Aldrich). Cells were counted in a dilution of trypan blue as a vital stain for quantification and subjected to dual-phase extraction as described below. For tumor extracts obtained directly after the in vivo measurements, mice were sacrificed and tumors were removed immediately. The entire tumor was quickly freeze-clamped, pulverized by grinding over liquid N₂, and homogenized with a tissue homogenizer in 4 mL of ice-cold methanol. Both lipid and water-soluble cell and tumor extract fractions were obtained using a dual-phase extraction method based on methanol/chloroform/water (1:1:1; v/v/v) as previously described (18,24).

High-resolution ³¹P MRS (³¹P HR MRS) studies

The water-soluble and lipid fractions were dissolved in deuterated solvents containing 0.24×10^-6 mol 3-(trimethylsilyl)propionic-2,2,3,3-d4 acid (TSP; Sigma-Aldrich) and 6.0×10^-6 mol phenylphosphonic acid (PPA; Sigma-Aldrich) as internal concentration and chemical shift standards. ³¹P HR-MRS was performed on a Bruker Avance 500 (11.7 T) spectrometer (Bruker BioSpin Corp.) using a 10-mm broadband probe tuned to the ³¹P frequency. The MR spectra were acquired using the following acquisition parameters: 90° flip angle for tumor extracts, 60° flip angle for cell extracts, 10,162 Hz sweep width, 0.8 s time domain, 8K data points, repetition time of 3 seconds for cell extracts and 15 seconds for tumor extracts, 2000 averages for cell extracts, and 300 averages for tumor extracts. All ³¹P HR MR spectra were processed using the MestReC 4.9.9.6 software (MestReLab Research). Lorentzian lines were fitted to the signals of PPA, PE, PC, GPE, and GPC. The areas under the curve were corrected for differences in T1 relaxation time and possible saturation effects owing to the relatively large flip angle. Metabolite levels were normalized to cell number or tumor weight. T1 relaxation times of the metabolites were measured with a progressive saturation series in phantom solutions of PC, PE, GPC in D₂O.

Non-invasive in vivo ³¹P MRS studies

In vivo ³¹P MRS was performed on a 9.4T Bruker Biospec spectrometer. A double tuned (¹H and ³¹P frequency) solenoid coil with an inner diameter of 12 mm was used (MRcoils BV, Drunen, The Netherlands). Mice were anesthetized by breathing a mixture of air and isoflurane (2%) through a nose cone. The tumor was hanging into the coil while the animal lay on a cradle with an opening for the coil. Body temperature was maintained during the experiment by using a blanket with circulating warm water. Breathing rate was monitored throughout all MR measurements with a movement sensor, which was attached at the mouse’s abdomen. A 3D RARE image was acquired with the following parameters: echo time (TE) of 7.2ms, repetition time (TR) of 500ms, RARE factor of 4, flip angle of 90⁰, field of view (FOV) of 1cm×1cm×1cm, 64 phase encode steps (64×64×64), and number of averages (NA) of 4. Total acquisition time was 13 minutes. Shimming of 1^st and 2^nd order B₀ gradient fields was performed manually by iteratively assessing the water line width obtained from the entire tumor. Non-localized ³¹P MR spectra were acquired with an adiabatic excitation (BIR4 45°, 200μs, 120ppm band width), repetition time of 1 second, and 1800 averages. A saturation slab (adiabatic full passage pulse driven at half the amplitude to achieve excitation with fully dispersed phase (25)) covered the mouse body to eliminate signals from muscles in the body. The combination of the drop off of ³¹P radiofrequency (RF) field strength perpendicular to the solenoid coil in which the tumor was placed, and the saturation slab positioned on the mouse body ensured that we acquired signal from tumor tissue only.

Analysis of in vivo ³¹P MRS data

Lorentzian lines were fitted to the ³¹P MRS data using JMRUI 4.0 software (26) and the AMARES algorithm (27). The resonance of phosphocreatine (PCr) was set to 0 ppm. In the fitting, the line widths of phosphomonoesters and phosphodiesters were constrained to the line width of PCr, and the frequency difference between PC and phosphoethanolamine (PE) and between GPC and glycerophosphoethanolamine (GPE) was fixed to 100 Hz. Metabolite levels were quantified as ratios with respect to β-nucleotide triphosphate (NTP). Metabolite levels were corrected for differences in T1 relaxation. Metabolite T1 values were measured in vivo by progressive saturation series in normal MDA-MB-231 tumors (n=4).

Statistical analysis of all data

A two-sided t-test assuming unequal variances between the MDA-MB-231-GDPD5-shRNA group (N=5) and the vector control group (N=5) was performed to test for statistically significant differences.

Results

The knockdown of GDPD5 was verified by qRT-PCR analysis of the mRNA expression of GDPD5. MDA-MB-231-GDPD5-shRNA cells had significantly decreased gene expression levels of GDPD5 mRNA compared to the MDA-MB-231-vector control cells (Fig. 2a). In ³¹P HR MR spectra of cell extracts (n=3), no signals for GPE and GPC could be detected in MDA-MB-231-vector control cells, whereas MDA-MB-231-GDPD5-shRNA cells clearly displayed GPC and GPE signals at a concentration of 0.057 fmol per cell and 0.048 fmol per cell respectively, resulting in an increase of both upon GDPD5-silencing (p=0.18 for GPC, p=0.24 for GPE). We also observed a borderline significant decrease in PC levels (p=0.06) in MDA-MB-231-GDPD5-shRNA cells compared to vector controls.

Figure 2.

(a) qRT-PCR demonstrating significant GDPD5 knockdown in MDA-MB-231-GDPD5-shRNA cells compared to vector controls. MDA-MB-231 cells containing empty vector, which lacked any shRNA expression served as vector controls. (b) High resolution ³¹P MRS of MDA-MB-231-GDPD5-shRNA and MDA-MB-231 vector control cells showed a trend towards decreased PC and PE levels in MDA-MB-231-GDPD5-shRNA cells compared to vector controls. GPE was not detectable in MDA-MB-231 vector control cells. Average and standard error of 3 cell extracts are shown for each group.

In vivo ³¹P MR measurements detected PME and PDE signals, as well as inorganic phosphate (Pi), phosphocreatine (PCr), and α-, β-, and γ-NTP in both the MDA-MB-231-GDPD5-shRNA and MDA-MB-231-vector control tumors (Fig. 3a). Since the magnetic susceptibilities of water and fat are different from each other, the frequently found water-lipid transitions in heterogeneous tumor tissue lead to broad line widths in in vivo MR spectra. The average and standard deviation of the line width of PCr was 94 ± 8 Hz and 98 ± 16 Hz respectively in MDA-MB-231-GDPD5-shRNA and MDA-MB-231-vector control tumors. However, it was still possible to distinguish the individual signals of PE, PC, GPE, and GPC in the in vivo ³¹P MR spectra. We detected a trend towards higher ratios of GPC/β-NTP and PE/β-NTP in the MDA-MB-231-GDPD5-shRNA tumors, which did not reach statistical significance (Fig. 3b).

Figure 3.

In vivo data of MDA-MB-231-GDPD5-shRNA and vector control tumor xenografts. (a) In vivo ³¹P MR spectra at 9.4 Tesla. (b) Quantified metabolite levels of PE, PC, GPE and GPC. Average and standard error of 6 tumors per group is shown. Metabolite levels were normalized to β-NTP.

The PME and PDE signals were clearly separated in the ³¹P HR MR spectra of tumor extracts (Fig. 4a). When normalized to tumor weight and the internal reference PPA, significantly increased levels of PE were detected in MDA-MB-231-GDPD5-shRNA tumors as compared to vector controls (Fig. 4b). This confirmed the in vivo findings of elevated PE and GPC levels in MDA-MB-231-GDPD5-shRNA tumors compared to vector control tumors.

Figure 4.

(a) Ex vivo ³¹P HR MRS of water-soluble extract fractions obtained from MDA-MB-231-GDPD5-shRNA and vector control tumor xenografts. (b) Corresponding metabolite quantification from MDA-MB-231-GDPD5-shRNA and vector control tumor (average and standard error, n=6 each).

Discussion

Stable silencing of GDPD5 in MDA-MB-231 breast cancer cells resulted in higher levels of GPC and PE in extracts of MDA-MB-231-GDPD5-shRNA tumors compared to MDA-MB-231-vector control tumors. The trends towards increased PE and GPC levels in MDA-MB-231-GDPD5-shRNA tumors xenografts relative to vector controls measured in vivo were consistent with the significant increases observed in tumor extracts using ex vivo ³¹P HR MRS. Ex vivo ³¹P HR MRS provided a good method to validate in vivo ³¹P MRS results. Such changes in phospholipid metabolite levels observed by in vivo ³¹P MRS might be large enough to be clinically relevant.

Silencing of GDPD5, which is an enzyme with GPC-PDE activity (28,29), was expected to increase GPC levels. This was indeed the case as confirmed by a higher GPC/β-NTP ratio in MDA-MB-231-GDPD5-shRNA tumors and a higher GPC level in MDA-MB-231-GDPD5-shRNA tumor extracts (see Fig. 3b and 4b). We also observed an increased GPC level in the MDA-MB-231-GDPD5-shRNA cells, which did not reach statistical significance (Fig. 2). The increase in GPC was not as drastic as expected when considering the good efficiency of the GDPD5 knockdown, which resulted in about 68% reduction in GDPD5 mRNA levels in stably transduced MDA-MB-231-GDPD5-shRNA cells. However, there is a possibility that 32% of GDPD5 is still sufficient to hydrolyze the majority of the produced GPC. It is likely that GDPD5 is not the only GPC-PDE enzyme responsible for the breakdown of GPC. For example, it was recently shown that GDPD6, i.e. EDI3, is also responsible for breaking down GPC in breast cancer cells (30). This was demonstrated by transient siRNA silencing of GDPD6 in MDA-MB-231 breast cancer cells, which significantly increased the GPC/PC ratio by a factor of 3.3 (30). Our GDPD5-knockdown data are from stably silenced cells, in which shRNA against GDPD5 is continuously produced in cells that are able to survive. This approach results in adaptations in stably silenced cells that cannot be observed during transient siRNA knockdown, which is why compensatory mechanisms in phospholipid metabolism might play an important role in our stably GDPD5-silenced cells, which nevertheless displayed an increase of GPC levels from undetectable to 0.057 fmol/cell. GDPD isozymes such as GDPD1, 2, 3, 4, or 6 may partially compensate for the reduction in GDPD5 and, to some extent, counteract the GPC increase caused by GDPD5 silencing. It is also possible that breast cancer cells have other compensatory mechanisms that are counteracting the knockdown of GDPD5. The significant increase in the level of PE/β-NTP in the MDA-MB-231-GDPD5-shRNA tumors compared to vector control tumors could be due to ethanolamine phosphorylation by choline and/or ethanolamine kinases as compensatory mechanism for the reduced PC levels in these cells. However, the absolute increase in GPC by 0.057 fmol/cell following stable GPDP5 silencing is relatively small compared to the total cellular PC content of 2.75 fmol/cell. Unaltered PC levels in GDPD5-silenced versus vector control tumors in vivo, which in turn resulted in unaltered PC/GPC ratios in vivo, do not necessarily indicate that GDPD5-silencing does not influence in vivo tumor aggressiveness. An increased PC/GPC ratio was shown to be associated with elevated ovarian and breast cancer cell aggressiveness (6,7). However, more recent studies have shown that patient-derived animal models of basal-like breast cancer and tumor tissue from patients with triple-negative breast cancer contained high GPC levels and, as a consequence, relatively low PC/GPC levels (31,32). This emphasizes the fact that GPC levels in breast cancers are not yet well understood and require further investigation. In addition, the unaltered PC/GPC ratios in our GDPD5-silenced tumor xenografts may also be due to compensatory mechanisms in these stably silenced cell lines that where selected for survival.

In cultured cells, PC decreased as a consequence of GDPD5 silencing, which was not the case when growing the same cells as orthotopic tumor xenografts. This finding indicates that the tumor microenvironment, e.g. stromal cells (33), tumor pH (34), or hypoxia (35), may have increased tumor PC and PE levels by modulating enzymes in choline phospholipid metabolism such as choline kinase alpha (35). Choline kinase alpha and beta both have the ability to also use ethanolamine as a substrate to produce PE (13). It is well known that phospholipid metabolite levels in cells are also controlled by culture conditions (36-39). Our cell culture experiments were performed with MDA-MB-231 cells growing in logarithmic growth phase with 21.43 μM choline and no ethanolamine. However, since mammalian cells are unable to synthesize ethanolamine de novo, ethanolamine must be provided from the diet or from degradation of PtdEth made by the phosphatidylserine decarboxylase pathway (40), which could explain the elevated PE levels in tumor xenografts compared to the same cells in cell culture.

In vivo ³¹P MRS at 9.4T enabled the assessment of the individual phosphorylated metabolites in choline and ethanolamine phospholipid metabolism. Shimming of the tumor tissue was challenging because frequent water-lipid transitions in the tumor caused changes in magnetic susceptibility on a microscopic scale. These susceptibility changes on a microscopic scale led to an inhomogeneous B₀ field that could not be corrected by shimming, where smooth first, second, and third order magnetic field gradients were applied to compensate for the variations in B₀. Therefore, the line width obtained in in vivo MRS of the tumors in our study likely reflected the heterogeneity of the tumor tissue. The line width was broader in tumors with more necrotic tissue.

Tumor heterogeneity is commonly found in different types of tumors and could result from genetic as well as microenvironmental differences within the tumor tissue (41). Therefore, a next step would be to use ³¹P chemical shift imaging (CSI) to investigate the spatial distribution of phospholipid metabolites inside the tumor tissue. So far, mostly ¹H CSI has been used in tumor models to study total choline levels (tCho = PC+GPC+Choline) (42). In general, higher tCho levels have been associated with increased breast cancer aggressiveness (6,43). Treatment response to chemotherapy (44) or targeted anticancer therapies that disrupt oncogenic signaling pathways (45-49) can be detected by means of a decrease in tCho level within a short time period after treatment, e.g. after 24 hours of treatment in some cases (50). However, some novel molecular anticancer treatments that inhibit Hsp90 (51) or histone deacetylase (HDAC) (52) were shown to increase tCho, emphasizing the fact that the molecular pathways that lead to a detectable change in tCho need to be investigated for every drug for which tCho will be monitored as a biomarker of treatment response. With the use of ³¹P MRS, a change in tCho can be narrowed down to changes in PC and/or GPC, which is molecularly specific. In regard to the question if GDPD5 could be a potential anticancer target in breast cancer cells, it would be premature to draw any conclusions based on the presented data on stable GDPD5 silencing. Further molecular studies are ongoing in our laboratory to answer this question.

In conclusion, silencing of GDPD5 increased the PE and GPC levels in MDA-MB-231-GDPD5-shRNA tumors, indicating that GDPD5 is an enzyme with GPC-PDE activity whose expression is important in choline and ethanolamine phospholipid metabolism of breast cancer, which can be studied by ³¹P MRS in vivo. Phosphorus HR MRS of tumor extracts can be used to validate in vivo ³¹P MRS measurements and may provide prior knowledge for in vivo data analysis. Phosphorus MRS is a powerful tool for studying cancer metabolism and can potentially be used for monitoring novel therapies that target enzymes in choline/ethanolamine phospholipid metabolism

Alphabetical list of nonstandard abbreviations

B₁

applied magnetic field

BIR

B1 independent rotation

cDNA

complementary deoxyribonucleic acid

ck

choline kinas CTP: phosphocholine cytidylyltransferase

DG-CTP

diacylglycerol choline phosphotransferase

DG-ETP

diacylglycerol ethanolamine phosphotransferase

EDTA

ededic acid (ethyl-enediaminetetraacetic acid)

EGFP

enhanced green fluorescent protein

ek

ethanolamine kinase

FBS

fetal bovine serum

FOV

field of view

GPC

glycerophosphocholine

GPC-PDE

glycerophocholine phosphodiesterase

GDPD

glycerophosphodiester phosphodiesterase domain containing

GPE

glycerophosphoethanolamine

HPRT 1

hypoxanthine phosphoribosyltransferase

HR-MRS

high resolution magnetic resonance spectroscopy

HSP

heat shock protein

lyso-PL

lysophospholipase

MR

Magnetic Resonance

MRI

Magnetic Resonance Imaging

MRS

Magnetic Resonance Spectroscopy

NMR

Nuclear Magnetic Resonance

NA

number of averages

NTP

nucleotide triphosphate

PC

phosphocholine

PCr

phosphocreatine

PCR

polymerase chain reaction

PDE

phosphodiesters

PE

phosphoethanolamine

pGK

phosphoglycerate kinase

Pi

inorganic phosphate

PME

phosphomonoesters

PLC

phospholipase C

PLC

phospholipase D

PLA2

phospholipase A2

ppm

parts per million

PtdEth

phosphatidyl ethanolamine

qRT-PCR

quantitative Reverse Transcription PCR

RARE

rapid acquisition with relaxation enhancement

RF

radiofrequency

RNA

Ribonucleic acid

ROI

region of interest

RPMI

Roswell Park Memorial Institute

shRNA

stably silenced Ribonucleic acid

SNR

signal to noise ratio

T

tesla

tCho

total choline

TE

echo time

TR

repetition time