The Phosphoinositide 3-Kinase Inhibitor PI-103 Downregulates Choline Kinase α Leading to Phosphocholine and Total Choline Decrease Detected by Magnetic Resonance Spectroscopy

Abstract

The phosphoinositide 3-kinase (PI3K) pathway is a major target for cancer drug development. PI-103 is an isoform-selective class I PI3K and mammalian target of rapamycin inhibitor. The aims of this work were as follows: first, to use magnetic resonance spectroscopy (MRS) to identify and develop a robust pharmacodynamic (PD) biomarker for target inhibition and potentially tumor response following PI3K inhibition; second, to evaluate mechanisms underlying the MRS-detected changes. Treatment of human PTEN null PC3 prostate and PIK3CA mutant HCT116 colon carcinoma cells with PI-103 resulted in a concentration- and time-dependent decrease in phosphocholine (PC) and total choline (tCho) levels (P < 0.05) detected by phosphorus (³¹P)- and proton (¹H)-MRS. In contrast, the cytotoxic microtubule inhibitor docetaxel increased glycerophosphocholine and tCho levels in PC3 cells. PI-103–induced MRS changes were associated with alterations in the protein expression levels of regulatory enzymes involved in lipid metabolism, including choline kinase α (ChoK_α), fatty acid synthase (FAS), and phosphorylated ATP-citrate lyase (pACL). However, a strong correlation (r² = 0.9, P = 0.009) was found only between PC concentrations and ChoK_α expression but not with FAS or pACL. This study identified inhibition of ChoK_α as a major cause of the observed change in PC levels following PI-103 treatment. We also showed the capacity of ¹H-MRS, a clinically well-established technique with higher sensitivity and wider applicability compared with ³¹P-MRS, to assess response to PI-103. Our results show that monitoring the effects of PI3K inhibitors by MRS may provide a noninvasive PD biomarker for PI3K inhibition and potentially of tumor response during early-stage clinical trials with PI3K inhibitors.

Introduction

Magnetic resonance spectroscopy (MRS) is increasingly used for the noninvasive assessment of cellular metabolism, both in vitro (tissue cultures, body fluids, tissue extracts, and isolated tissues) and in vivo (small animals and humans; ref. 1). Phosphorous (³¹P)-MRS measures signals from endogenous metabolites such as ATP, indicative of cellular energy status, as well as phosphocholine (PC), phosphoethanolamine (PE), glycerophosphocholine (GPC), and glycerophosphoethanolamine (GPE) arising from the synthetic and degradative pathways of major phospholipids (1). Proton (¹H)-MRS can measure the total choline (tCho) peak consisting of PC, GPC, free choline, and corresponding ethanolamine metabolites (1). The levels of lactate, creatine, glutamine/glutamate, and other amino acids can also be obtained from the ¹H-MR spectra (1). In the field of cancer, MRS has emerged as a promising tool for characterizing disease and assessing response to therapy (2,3). Higher levels of phosphomonoesters comprising PC and PE were observed in tumors compared with the corresponding normal tissues (4,5). Other studies showed that progression from the normal to the malignant phenotype is associated with an increase in PC and tCho (6). Recently, an increase in the MRS-detected PC/GPC ratio was reported in a mouse model of early gastrointestinal tumorigenesis, highlighting its use as a potential biomarker for monitoring disease progression (7). Similar findings have also been reported in human colorectal cancer (8).

With the emergence of promising new molecularly targeted agents for the treatment of cancer (9), new technologies for the screening and early detection of response to therapy are required. The development of noninvasive endpoints such as MRS is desirable because it may avoid the need for tumor biopsy (10,11).

One oncogenic pathway that is highly and frequently activated in a wide range of tumor types is the production of the phospholipid, phosphatidylinositol-3,4,5-triphosphate by phosphoinositide 3-kinases (PI3K), triggering cell growth, proliferation, survival, motility, invasion, and angiogenesis (12). The use of the PI3K inhibitors wortmannin and LY294002, despite their lack of specificity, has provided initial proof of concept for the anticancer activity of PI3K inhibitors; more recently, compounds that inhibit class I PI3K with increased specificity have been discovered (13-17), including the pyridofuropyrimidine lead compound PI-103 that we identified (14). Detailed characterization was carried out on this compound (17-20). PI-103 showed antiproliferative activity against a range of human cancer cell lines in vitro as well as significant antitumor activity in human tumor xenografts in athymic mice (19,20). PI-103 is now used widely as a chemical tool compound (17) and has been recommended (21) as a replacement for the less potent and selective LY294002 when used alongside wortmannin to show the involvement of PI3K in cellular processes.

Using MRS, we have previously reported that blockade of PI3K with LY294002 and wortmannin in human breast cancer cells was associated with a decrease in PC levels as well as (LY294002 only) an elevation in GPC content (22). With the emergence of more selective and drug-like PI3K inhibitors, we have used MRS to test whether the novel class I PI3K and mammalian target of rapamycin (mTOR) inhibitor PI-103 would result in similar metabolic alterations in human carcinoma cell lines with activating genetic abnormalities in the PI3K pathway. MRS-detected changes were compared with the effects of the cytotoxic microtubule inhibitor docetaxel that is widely used in prostate cancer, with the objective of ruling out nonspecific antiproliferative effects associated with cytotoxicity. We have also investigated potential mechanisms underlying the observed metabolic changes. We report a concentration- and time-dependent decrease in PC and tCho levels following PI3K pathway inhibition with PI-103. Furthermore, we discovered that PC concentrations correlated with choline kinase α (ChoK_α) protein expression levels, indicating that ChoK_α deregulation following PI3K inhibition is one of the main mechanisms underlying the reduction in PC detected by MRS.

Materials and Methods

Cell culture and treatment

The human PTEN null PC3 prostate adenocarcinoma and PIK3CA mutant HCT116 colorectal carcinoma cell lines (American Type Culture Collection) were cultured in DMEM (Life Technologies) supplemented with 10% FCS (PAA Labs Ltd), 100 U/mL penicillin, and 100 μg/mL streptomycin (Life Technologies) at 37°C in 5% CO₂. Cells were cultured according to the supplier's instructions and used at passages 6 to 12. Cell viability was routinely >90%, as judged by trypan blue exclusion. Both cell lines routinely tested negative for Mycoplasma by PCR. PTEN status was verified by protein expression using Western blotting in house, and PIK3CA status was determined or verified experimentally by sequencing. In the case of the HCT116 cell line, single nucleotide polymorphism profiling was also used to confirm that the genotypes matched those provided in the Cancer Genome Project Cosmic database.

Both cell lines were treated with PI-103. PC3 cells were also treated with docetaxel (Sigma). GI₅₀ values (concentrations causing 50% inhibition of proliferation of tumor cells) were determined using the sulforhodamine B assay following 96 hours continuous exposure to compounds (20). Cell culture conditions were selected to ensure similar availability of choline per cell in control and treated cells over the time course of treatment. At the required time points, cells underwent trypsinization and trypan blue exclusion assay (23). The effect of treatment on cell number was monitored by counting the number of viable attached cells in a treated flask and comparing that number with the number of attached cells in a control flask.

Flow cytometry

Cell cycle analysis was performed as previously described (24). Cell volume was estimated using the Time of Flight parameter on the Elite ESP Beckman Coulter (High Wycombe) cell sorter. Pulse width is measured as the particle (cell) moves through the laser beam and the width of the laser beam is subtracted, through machine hardware, from the measurement. The calibration of the machine for diameter measurements was performed using beads of known diameters: 5, 10, and 15 μm. Assuming a spherical particle, a quantity proportional to volume was calculated by measuring cell diameter from a pulse width signal, using the formula (4/3 × π × r³).

Immunoblotting

Western blotting was performed as previously described (22). Western blots were probed for pAKT (Ser473), total AKT, pS6RP (Ser240/244), total S6RP (Cell Signaling), and glyceraldehyde-3-phosphate dehydrogenase (Chemicon). Protein expression levels of ChoK_α were detected using the human ChoK_α monoclonal antibody (25). The blots were also probed for fatty acid synthase (FAS; Cell Signaling) as well as phosphorylated (Ser454) and total ATP-citrate lyase (ACL, Cell Signaling). Western blot bands were quantified by densitometry using Image Quant version 5 (GE Healthcare).

In vitro ¹H- and ³¹P-MRS of cell extracts

To obtain an MR spectrum, an average of 4 × 10⁷ cells in logarithmic phase were extracted from cell culture using the dual phase extraction method, as previously described (24,26). Before acquisition of the MRS spectra, the water-soluble metabolites were resuspended in deuterium oxide for ¹H-MRS or deuterium oxide with 10 mmol/L EDTA (pH 8.2) for ³¹P-MRS. ¹H- and ¹H-decoupled ³¹P-MR spectra were acquired and metabolites were quantified as previously reported (24).

Statistical analysis

Data are presented as the mean ± SD (n ≥ 3). Statistical significance of differences was determined by unpaired two-tailed Student's standard t tests with a P value of ≤0.05 considered to be statistically significant. Pearson correlation coefficients were calculated using SPSS version 17.

Results

Concentration-response as well as time-response experiments were performed using the PTEN null PC3 human prostate cell line. MR-detected changes were compared with effects on cell number as well as downstream markers of PI3K pathway inhibition. We have examined consistency using the PIK3CA mutant HCT116 human colorectal carcinoma cell line. Results in PC3 cells were also compared with the action of the cytotoxic microtubule inhibitor docetaxel, widely used for the treatment of prostate cancer, to rule out nonspecific effects associated with cytotoxicity.

¹H- and ³¹P-MRS of cell extracts

MRS of aqueous extracts from cells treated in vitro with the class I PI3K and mTOR inhibitor PI-103 was used to identify potential biomarkers of PI3K pathway inhibition.

The PTEN null human prostate cell line PC3 was treated with PI-103 for 24 hours at pharmacologically active concentrations corresponding to 1×, 2.5×, 5×, and 10× GI₅₀ (GI₅₀ = 100 nmol/L). Examples of the ³¹P-MR spectra of control and PI-103–treated PC3 cells are illustrated in Fig. 1A. As summarized in Table 1, ³¹P-MRS showed that PI-103 treatment for 24 hours led to a significant concentration-dependent reduction in PC, PE, and nucleotide triphosphate (NTP) levels. Levels of PC were reduced from 15.6 ± 1.4 fmol/cell in DMSO-treated control cells to 12.0 ± 0.7 fmol/cell, P < 0.003 using 1× GI₅₀, and to 9.0 ± 0.2 fmol/cell, P < 0.0001 using PI-103 at 10× GI₅₀ (Supplementary Table S1). PI-103 at 5× GI₅₀ caused a time-dependent decrease in PC starting from 6 hours following treatment. For example, at 8 hours, PI-103–treated cells contained 13.2 ± 2.0 fmol/ cell compared with 19.3 ± 3.5 fmol/cell in DMSO-treated control cells (P < 0.02). As shown in Table 2A, the decrease in PC levels in PI-103–treated cells was maintained for up to 24 hours. PI-103 at 5× GI₅₀ also caused a significant time-dependent decrease in PE and NTP levels, but GPC or GPE were not affected (Table 2A).

Figure 1.

Metabolic changes caused by PI-103 in human PTEN null PC3 prostate and PIK3CA mutant HCT116 colon carcinoma cell lines. A, representative in vitro ³¹P-MRS and expansion of the choline-containing metabolites region of ¹H-MR spectra of PC3 aqueous cell extracts following treatment with PI-103 (5× GI₅₀, 500 nmol/L) compared with vehicle-treated (DMSO) control; Gluc, glucose. B, representative in vitro ³¹P-MRS and expansion of the choline-containing metabolites region of ¹H-MR spectra of HCT116 aqueous cell extracts following treatment with PI-103 (5× GI₅₀, 5 μmol/L) compared with vehicle-treated (DMSO) control. C, comparison of cellular effects and ³¹P-MRS–detected metabolic changes caused by PI-103 (5× GI₅₀) in PC3 prostate cancer and HCT116 colon cancer cells at 24 h. Results are expressed as % treated versus control (% T/C); columns, mean of at least three separate experiments; bars, SD. Statistically significantly different from the control; *, P ≤ 0.05; **, P < 0.01; †, P ≤ 0.005; and ‡, P < 0.0005; two-tailed unpaired t test was used for all comparisons.

Table 1.

Concentration-response analysis of inhibition with PI-103 (24 h): effects on cell number, cell volume and MRS-detected metabolite levels in PTEN null human prostate cancer cells PC3.

PI- 103/nM	Cell no.	P	Cell vol.	P	PE (31P)	P	PC (31P)	P	GPC (31P)	P	NTP (31P)	P	PC (1H)	P	GPC (1H)	P	tCho (1H)	P
100	96(±8)	ns	91(±3)	ns	51(±14)	0.04	80(±1)	0.001	135(±29)	ns	76(±4)	0.01	78(±5)	0.02	94(±34)	ns	78(±6)	0.03
250	91(±6)	ns	81(±14)	ns	36(±6)	0.0003	65(±7)	0.003	121(±39)	ns	77(±8)	0.02	62(±6)	0.002	180(±55)	ns	76(±8)	0.01
500	79(±8)	0.05	70(±9)	0.04	38(±8)	0.01	58(±13)	0.04	124(±35)	ns	65(±8)	0.03	65(±11)	0.04	121(±9)	ns	77(±3)	0.006
1000	79(±5)	0.03	72(±12)	ns	43(±5)	0.003	56(±5)	0.007	147(±63)	ns	75(±7)	0.03	59(±4)	0.004	122(±33)	ns	74(±4)	0.01

NOTE:

-Data are expressed as %T/C and presented as the mean ± SD, n ≥ 3.

-Two-tailed unpaired t test was used to compare results in treated cells to controls.

-GPE in PC3 cells was not consistently detectable due to the low levels of this metabolite and was not affected by treatment with PI-103.

-Experiments were not designed to measure levels of other metabolites present in ¹H-MR spectra such as lactate, alanine, glutamine/glutamate or acetate.

Table 2.

Time-response analysis of inhibition with PI-103 (5×GI₅₀): effects on cell number, cell volume and metabolite levels in PTEN null human prostate cancer cells PC3:

A. 31P-MRS
Time/h	Cell no.	P	Cell Vol.	P	PE	P	PC	P	GPC	P	NTP	P
2	104(±4)	ns	98(±3)	ns	77(±16)	ns	93 (±6)	ns	111(±21)	ns	91(±5)	ns
4	103(±10)	ns	97(±14)	ns	78(±20)	ns	87(±9)	ns	70(±16)	ns	84(±6)	0.02
6	96(±10)	ns	94(±9)	ns	87(±20)	ns	79(±10)	0.006	97(±17)	ns	81(±13)	0.02
8	108(±9)	ns	91(±12)	0.006	64(±19)	0.02	69(±8)	0.002	102(±41)	ns	78(±6)	0.002
12	101(±10)	ns	84(±14)	0.003	61(±19)	0.01	57(±10)	0.001	85(±13)	ns	73(±14)	0.02
16	88(±4)	0.01	81(±9)	0.005	62(±17)	0.03	56(±8)	0.002	120(±66)	ns	81(±10)	0.05
24	83(±6)	0.001	82(±12)	0.002	63(±27)	0.03	65(±8)	0.0002	88(±30)	ns	73(±7)	0.0003

NOTE:

-Data are expressed as %T/C and presented as the mean ± SD, n ≥ 3.

-Two-tailed unpaired t test was used to compare results in treated cells to controls within the same time-point.

-GPE in PC3 cells was not consistently detectable due to the low levels of this metabolite and was not affected by treatment with PI-103.

¹H-MR spectra of extracts of control and PI-103–treated PC3 cells were also investigated (Fig. 1A). Concentration-response analysis showed a statistically significant concentration-dependent decrease in PC levels in PI-103–treated PC3 cells compared with controls. As shown in Table 1, changes in PC observed by ¹H-MRS were similar to those detected by ³¹P-MRS. Furthermore, a concentration-dependent decrease in the tCho (PC+ GPC+ choline) levels was also detected (Table 1). Table 2B shows the time-dependent changes in PC and tCho levels following treatment with PI-103 detected by ¹H-MRS. In PC3 cells, tCho levels are dominated by the PC signal; hence, changes in tCho reflected those seen in PC levels over the time course of treatment.

Table 2.

B. 1H-MRS
Time/h	PC	P	GPC	P	tCho	P
4	99(±8)	ns	107(±23)	ns	106(±8)	ns
6	80(±9)	0.01	102(±17)	ns	86(±9)	0.03
8	71(±8)	0.002	98(±17)	ns	78(±10)	0.01
12	62(±7)	0.002	96(±18)	ns	67(±5)	0.002
16	58(±3)	0.0001	134(±50)	ns	61(±4)	0.0006
24	61(±5)	0.00002	106(±21)	ns	69(±8)	0.0003

NOTE: See (A)

The PIK3CA mutant human colorectal cell line HCT116 was treated with PI-103 for 24 hours at a pharmacologically active concentration corresponding to 5× GI₅₀ (GI₅₀ = 1 μmol/L). Analysis of ³¹P-MR spectra of control and PI-103–treated HCT116 cells (Fig. 1B) showed that treatment for 24 hours led to a 54 ± 8% (P = 0.01) decrease in PC levels relative to control, changing from 14.7 ± 1.2 fmol/cell in DMSO-treated control cells to 6.7 ± 0.7 fmol/cell (P = 0.004) following treatment. Levels of NTP were not altered by PI-103, and GPC, GPE, and PE signals were small and not accurately measurable. ¹H-MRS (Fig. 1B) showed a significant decrease in the levels of PC (down to 52 ± 1%, P = 0.0005) as well as tCho (down to 62 ± 4%, P = 0.006). As shown in Fig. 1C, PI-103 caused similar effects on PC levels in HCT116 to those seen in PC3 cells.

Relationship of MRS changes with markers of PI3K pathway inhibition

Similar to our previous results (20), 24-hour treatment with PI-103 (5× GI₅₀) resulted in a decrease in the number of treated cells per flask compared with controls both in PC3 cells (down to 83 ± 6%, P = 0.001) and HCT116 cells (down to 69 ± 11%, P = 0.05), consistent with decreased proliferation. In PC3 cells, PI-103 caused both concentration and time-dependent changes in cell number (Tables 1 and 2A). The decrease in PC levels was detectable using PI-103 for 24 hours at 1× GI₅₀ (Table 1). Treatment at 5× GI₅₀ did not show reduced cell number for up to 12 hours, whereas a significant decrease in PC levels was detected as early as 6 hours after PI-103 treatment (Table 2A and B).

Western blotting analysis of downstream components of the PI3K pathway was performed. Using PC3 cells, PI-103 blocked PI3K signaling as indicated by decreases in pAKT (Ser473) and pS6RP (Ser240/244) in treated compared with control cells. This effect was observed at a PI-103 concentration of 1× GI₅₀ (Supplementary Table S2A; Fig. 2A). At 5× GI₅₀, PI-103 decreased pAKT (Ser473) and pS6RP (Ser240/244), which was detectable at 2 hours following treatment and was sustained over the time course of treatment (Supplementary Table S2B; Fig. 2B). Overall, no effect on total AKT or S6RP was detected, indicating that the effects seen were on the phosphorylation of these proteins. Treatment of HCT116 cells with PI-103 for 24 hours at 5× GI₅₀ reduced levels of pAKT (Ser473) and pS6RP (Ser240/244; Fig. 2C), confirming PI3K pathway inhibition in these cells. These results showed that inhibition of the PI3K pathway was occurring at PI-103 concentrations and time points that resulted in the MRS changes reported above.

Figure 2.

Molecular effects of inhibition with PI-103 on the PI3K signaling pathway in PTEN null PC3 prostate and PIK3CA mutant HCT116 colon carcinoma cell lines. A, representative Western blots showing decreases in molecular biomarkers in the PI3K signaling pathway in PC3 prostate cancer cells following treatment with PI-103 at 1×, 2.5×, and 5× GI₅₀, 24 h posttreatment. B, representative Western blots showing decreases in molecular biomarkers in the PI3K signaling pathway in PC3 prostate cancer cells at selected time points posttreatment with PI-103 (5× GI₅₀). C, representative Western blots showing inhibition of molecular biomarkers in the PI3K signaling pathway and changes in protein expression levels of enzymes involved in lipid metabolism in HCT116 colon cancer cells following treatment with PI-103 (5× GI₅₀).

Cell cycle analysis by flow cytometry showed that 24-hour treatment with PI-103 at 5× GI₅₀ caused an increase in the G₁ cell population and a decrease in S phase fraction in both cell lines (Fig. 3A–C). In PC3 cells, cell cycle effects of PI-103 were both concentration- and time-dependent (Fig. 3A and B). These data show that our MRS-detected PC changes occurred at low concentrations and preceded PI-103–mediated cell cycle effects.

Figure 3.

Effects of inhibition with PI-103 on cell cycle distribution in PTEN null PC3 prostate and PIK3CA mutant HCT116 colon carcinoma cell lines. A, representative flow cytometry analysis histograms showing the cell cycle distribution of PC3 cells with vehicle treatment (DMSO, control, solid) or following treatment with PI-103 at 1×, 2.5×, and 5× GI₅₀, 24 h posttreatment (empty). B, representative flow cytometry analysis histograms showing the cell cycle distribution of PC3 cells with vehicle treatment (DMSO, control, solid) or following treatment with PI-103 (5× GI₅₀, empty) at selected time points posttreatment. C, representative flow cytometry analysis histograms showing the cell cycle distribution of HCT116 cells with vehicle treatment (DMSO, control, black) or following treatment with PI-103 (grey).

Mechanism for PC depletion

PI3K signaling is well known to regulate cell size (27). This will have an effect on cellular metabolism and hence necessitates normalizing metabolite concentrations to cell volume. We carried out cell volume measurements using flow cytometry on cell samples used for MRS analysis, to determine whether cell volume changes caused by PI-103 contribute to our MRS-detected changes. In PC3 cells, PI-103 at 5× GI₅₀ caused a decrease in cell volume compared with controls starting from 8 hours following treatment (Table 2A). After normalization for cell volume, there is no significant effect of PI-103 on PE or NTP (P > 0.1), but a significant average decrease of 30% (P < 0.05) in PC levels is still detected relative to controls starting from 8 hours following treatment (Supplementary Table S3). Treatment of HCT116 cells with PI-103 reduced cell volume to 73 ± 4% (P = 0.01), relative to controls. Normalizing to cell volume, PC levels in treated cells were still decreased compared with control treatment (down to 63 ± 9%, P = 0.03).

To rule out nonspecific antiproliferative effects associated with cytotoxicity, PC3 cells were also treated with the antimitotic agent docetaxel. Docetaxel was selected as this is used clinically in prostate cancer patients (28). Treatment with docetaxel at 5× GI₅₀ (GI₅₀ = 4.25 nmol/L) for 24 hours decreased cell number to 51 ± 6% (P = 0.001). In direct contrast to the decrease observed with PI-103, ³¹P-MRS showed that docetaxel increased concentrations of PC (up to 132 ± 12%, P = 0.02), PE (up to 173 ± 23%, P = 0.01), GPC (up to 667 ± 98%, P = 0.002) and NTP (up to 175 ± 19%, P = 0.006). Docetaxel acts through the inhibition of microtubular depolymerization leading to the inhibition of mitosis and induction of apoptosis. In line with this, docetaxel-treated cells were arrested in G₂-M phase (from 19 ± 2% to 67 ± 4%, P = 0.0001), resulting in an increase in average cell volume to 138 ± 11% (P = 0.01). Normalizing to cell volume, no change in PC concentrations was detected compared with control, but increases in PE (up to 126 ± 12%, P = 0.04), GPC (up to 487 ± 76%, P = 0.003), and NTP (up to 127 ± 10%, P = 0.02) were still evident. An increase in choline-containing metabolites was also detected by ¹H-MRS following treatment with docetaxel, contrasting with the decrease caused by PI-103 (Fig. 4A).

Figure 4.

Investigation of mechanisms underlying MRS-detected changes in PC levels following treatment of PTEN null human prostate cancer cells PC3 with PI-103. A, comparison of cellular effects and ¹H-MRS–detected metabolic changes caused by PI-103 and docetaxel (DTX) in PC3 cells at 5× GI₅₀, 24 h. B, representative Western blots showing changes in protein expression levels of FAS, pACL (Ser454), and ChoK_α at selected time points posttreatment with PI-103 (5× GI₅₀). C, changes in PC (³¹P-MRS) normalized to cell volume and ChoK_α protein expression levels (densitometry) in PC3 cells over the time course of treatment with PI-103 (5× GI₅₀). Results are expressed as % T/C; points, mean of at least three separate experiments; bars, SD. Statistically significantly different from the control; *, P ≤ 0.05; **, P < 0.01; †, P ≤ 0.005; ‡, P < 0.0005; two-tailed unpaired t test was used for all comparisons.

Several reports have identified a link between PI3K signaling pathway and enzymes involved in choline and lipid metabolism including ChoK (29), FAS (30) and ACL (31). Hence, the effect of PI-103 on the expression of these enzymes was determined to investigate potential mechanisms underlying MRS-detected PC depletion. Western blotting of cell extracts obtained from cells used for MRS analysis showed that PI-103 caused a small decrease in the ACL phosphorylation (Ser454) in PC3 cells (Fig. 4B). The decrease in pACL (Ser454) was maintained up to 24 hours posttreatment. Quantification using densitometry showed a significant decrease in pACL (Ser454) levels reaching a minimum of 67 ± 14% (P = 0.03) in 6 hour–treated cells compared with control cells (Supplementary Table S4). However, no correlation (r² = 0.3, P = 0.2) was found between changes in the levels of PC (measured by MRS) and pACL (Ser454; quantified by densitometry). A small statistically significant decrease (down to 84 ± 4%, P = 0.04) in FAS levels was seen only at 4 hours after PI-103 treatment (Supplementary Table S4; Fig. 4B). Treatment of HCT116 cells with PI-103 at 5× GI₅₀ for 24 hours also resulted in a decrease in pACL (Ser454; Fig. 2C) and FAS expression compared with controls. However, no correlation was found between PC and pACL (r² = 0.8, P = 0.1) or FAS (r² = 0.2, P = 0.5).

Levels of ChoK_α were also analyzed by Western blotting, using an antibody against human ChoK_α (25). A decrease in ChoK_α expression was detectable from 4 hours following treatment of PC3 cells with PI-103 at 5× GI₅₀ (Fig. 4B). ChoK_α levels, quantified by densitometry, decreased over the time course of treatment reaching a minimum of 59 ± 13% (P = 0.01) in 16 hour–treated cells relative to control (Supplementary Table S4; Fig. 4C). From 6 hours onwards, a strong correlation was found between PC levels (measured by MRS) and ChoK_α levels (r² = 0.9, P = 0.009; Fig. 4C). ChoK_α levels also decreased in HCT116 cells (Fig. 2C) and correlated with the decrease in PC concentrations (r² = 1, P = 0.02) following PI-103 treatment.

Discussion

PI-103 is a potent inhibitor of class I PI3K and mTOR that was synthesized following a high-throughput screen to identify inhibitors of the enzymatic activity of recombinant PI3K p110_α (13-17). Detailed pharmacologic characterization has been carried out on PI-103 (17-20). PI-103 is widely used as a chemical probe for the PI3K/mTOR pathway (17,21) and has inhibitory properties similar to several clinical candidates (32). Identification of noninvasive biomarkers of target inhibition and potentially of tumor response to this novel treatment would be of value in the clinical development of PI3K inhibitors. In this study, we have used MRS to search for potential noninvasive biomarkers of the effects of PI-103 in two human cancer cell lines, PTEN null PC3 prostate and PIK3CA mutant HCT116 colon, in vitro. Using ³¹P-MRS, a significant decrease in PC concentrations was detected in both tumor lines following PI-103 treatment. A significant reduction in PE and NTP levels was also detected in the human prostate cancer cell line PC3, but no effect on GPC or GPE was observed. In the HCT116 colon cancer cells, NTP levels were not affected by PI-103 treatment while other phosphometabolites including PE, GPE and GPC were below the detection limit of our MR spectrometer.

We have previously reported a similar decrease in PC concentrations following inhibition of PI3K signaling with LY294002 and wortmannin (22). This indicates that the decrease in PC levels is likely to be related to the action of these inhibitors and PI-103 on their common target PI3K, while the increase in GPC observed following treatment with LY294002 (22) may be related to other off-target effects of this weakly potent and nonselective compound. It is important to note, however, that a decrease in PC concentrations was also observed following inhibition of other signaling molecules, for example: mitogen-activated protein kinase (33), FAS (34), HIF-1_α (35) and phospholipase C_γ1 (36). Although this might suggest that these changes may not be specific to the PI3K pathway, we have shown that they are different from those seen with for example, inhibitors of HSP90 (23) and HDAC (37,38). Furthermore, assessing the effect of the antimitotic agent docetaxel, we detected a decrease in cell number but an increase in PE and GPC levels. This further supports the view that our MRS-detected changes are a consequence of the inhibition of cell signaling and not simply a nonspecific antiproliferative effects associated with cytotoxicity.

Although changes in PC and/or tCho may not be specific to drugs acting on a particular molecular target or pathway, they can nevertheless provide valuable biomarkers of response (11). Many biomarkers are not specific to a given pathway, but nevertheless have considerable clinical value (e.g., [¹⁸F]-2-fluoro-deoxy-d-glucose). Significant PC changes have been observed at 1x GI₅₀ concentrations of PI-103, which are achieved in animal studies (20). Furthermore, preliminary evidence suggests that PI3K inhibitors with improved pharmaceutical properties compared with PI-103 also show similar effects (data not shown). We have demonstrated that we can measure choline-containing metabolites in spectra from tumors from patients in the clinic (2,39). Based on our observation that changes in these metabolites parallel target inhibition in cells, it seems reasonable to hypothesize that our findings reported herein may have the potential to translate to animal and human studies.

Using ¹H-MRS, we were able to detect a similar decrease in PC levels to that measured by ³¹P-MRS in both PC3 and HCT116 cells following treatment with PI-103. More importantly, the composite tCho peak decreased significantly, indicating that the effect of PI-103 can be followed in vivo despite the overlap of the choline-containing metabolite peaks in ¹H-MR spectra. This increases the potential for the translation of our work into the clinic as ¹H-MRS can be performed on many current clinical MRS systems, in contrast to ³¹P-MRS, which requires specialized hardware. ¹H-MRS also has the advantage of higher sensitivity compared to ³¹P-MRS, thus reducing measurement times.

The antiproliferative effects of PI-103 were previously investigated using a panel of human cancer cell lines with an activated PI3K pathway caused by different genetic abnormalities (20). Using the PTEN null PC3 cells, PI-103 was found to be ~50-fold more potent than the commonly used PI3K inhibitor LY294002 (20). Similarly, our results show that a low concentration of PI-103 equal to 1x GI₅₀ was sufficient to significantly decrease PC levels, whereas LY294002 required a ~50-fold higher concentration to cause a similar decrease in PC levels in the human breast cancer cell line MDA-MB-231 (22). PI-103 showed a concentration-dependent effect on cell proliferation and G₁ cell cycle arrest as well as the decrease in PC levels. PI-103 treatment for 24 h at 1x GI₅₀ resulted in a significant reduction in PC detected by MRS, which was associated with pAKT depletion. Higher concentrations of PI-103 were required to affect cell number and cell cycle. Furthermore, a time-dependent decrease in PC levels starting from 6 hours following treatment with PI-103 (5x GI₅₀) was observed. This was associated with a reduction in pAKT but preceded G₁ cell cycle arrest and reduction of cell number. These results further confirm that the decrease in PC levels is related to the inhibition of PI3K signaling, rather than being a consequence of the effect of PI-103 on cell growth. Thus, MRS-detected PC may represent a sensitive and early pharmacodynamic biomarker for response to PI3K inhibition.

PC plays important roles in lipid biosynthesis and cell signaling (40). The concentration of PC in cells is, therefore, likely to result from an interplay between both metabolic and signaling networks. Aberrant choline metabolism has been demonstrated in cancer cells in culture in which elevated choline levels may be indicative of membrane turnover (40), increased malignant potential, or upregulation of signaling processes due to oncogenic activation events (41,42).

In the present study, we have investigated possible mechanisms underlying MR-detectable metabolic changes with the focus on the decrease in PC levels. The aim was to establish the link between PI3K signaling and phospholipid metabolism and, in turn, to help optimize the use of MRS-detectable changes as potential biomarkers for inhibitors of this pathway.

First, we have investigated the decrease in cell size following PI3K/mTOR inhibition as a possible cause for the reduction in MRS-detectable lipid biosynthesis. An enhanced lipid synthesis following AKT activation was previously reported, attributed to the acceleration of the de novo membrane synthesis required to allow the increased cell volume (31,43). In line with this report, inhibition of the PI3K pathway using PI-103 resulted in a significant decrease in treated PC3 and HCT116 cells volume compared to controls. However, this reduction in cell volume was not enough to explain the decrease in PC levels detected by MRS in our system.

Second, we ruled out the possibility that the decrease in PC levels was due to nonspecific effects associated with cytotoxicity by comparing MRS-detected changes in PC3 cells with the effects of the clinically relevant microtubule inhibitor docetaxel. Treatment with this antimitotic agent caused a decrease in cell number but had no significant effect on PC levels. We have previously reported similar results using doxorubicin (23). Further work could be performed with other cytotoxics in the future.

Finally, the effects of PI-103 on enzymes involved in lipid metabolism were explored. As discussed in our previous report (22) and references therein, oncogene- and/or mitogen-signal transduction effectors are known to modulate many phosphatidylcholine-specific enzymes. More specifically, the PI3K pathway has been shown to regulate ChoK, the enzyme responsible for phosphorylation of choline into PC (29). Furthermore, activation of AKT has been shown to induce expression of FAS and ACL, enzymes catalyzing the de novo synthesis of long-chain fatty acids, thus increasing membrane lipid biosynthesis (31,43,44). Hence, we hypothesized that modulation of PI3K pathway using PI-103 would affect these enzymes and thus modulate choline metabolism leading to the changes we detected by MRS. Our results showed a consistent decrease in ChoK_α protein expression following treatment with PI-103 in both cell lines. Furthermore, a strong correlation between the decrease in PC levels detected by MRS and ChoK_α expression was observed in the human prostate carcinoma cell line PC3 following treatment with PI-103. In contrast, our results showed no obvious relationship between the reduction in PC levels and FAS or pACL protein expression. This indicates that the decrease in ChoK_α protein expression induced by PI-103 plays a major role in the MRS-detected PC depletion that we detected in our study.

It is important to note that the involvement of other enzymes related to choline metabolism, such as phospholipases D, C and A₂ (45,46) as well as choline transporters (47) cannot be excluded. Further work is required to identify whether these enzymes are also involved in MRS changes detected following PI3K inhibition. We have previously published a genome-wide cDNA microarray profiling study of mRNAs that show altered expression in response to PI-103 in the PTEN null human glioblastoma cells (Supplementary data published therein; ref. 19). Inspection of the data shows changes in the expression of genes involved in glucose and cholesterol biosynthesis. Thus further studies of pathways involved in tumor metabolism should be carried out in the future.

Taken together, our results strongly support a link between ChoK_α and its product PC and the PI3K signaling pathway. Because ChoK_α protein levels are decreased, one possibility is that ChoK_α could be regulated at downstream of mTOR (48). However, GSK-3 could also play a role in lipid metabolism control through ACL (49). Inhibitors that act at the different key components in the PI3K-AKT-mTOR network (32) could be used to further clarify the association between MR-detectable metabolic changes and the PI3K signaling pathway.

In conclusion, our study has shown that MRS can be used to detect a decrease in PC and tCho levels that is associated with the inhibition of PI3K by the class I PI3K and mTOR inhibitor PI-103 in human prostate and colon carcinoma cell lines. We have also identified inhibition of ChoK_α as a factor involved in the observed decrease in PC levels following treatment with PI-103. Monitoring the changes in PC and tCho levels by MRS (using either ¹H or ³¹P) may provide noninvasive pharmacodynamic biomarkers of PI3K pathway inhibition and potentially of tumor response in solid tumors during clinical trials with novel PI3K inhibitors.