Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

1 Introduction

Metabolomics is the study of the totality of small-molecule metabolites in an organism, cell or disease state. Unlike the genome, the metabolome of a cell can change from minute to minute, depending on factors such as its stage in the cell cycle or its environment. Similarly, when a cancer cell responds to an anticancer drug its metabolome is likely to show changes that could be used to decide whether to continue treatment or, in the context of a drug trial, to indicate whether the drug is working and perhaps its mechanism of action. In this review, we will focus on the use of (nuclear) magnetic resonance spectroscopy (NMR/MRS) methods for monitoring the effect of anticancer drugs on the cancer metabolome. Following the standard convention, the nomenclature we will use is as follows. The term “NMR” will be used for the magnetic resonance spectroscopy technique when used ex vivo on tissue extracts, biopsies, etc. The term “MRS” will be used to denote studies by the same technique on living animals or patients.

Conventional high-resolution nuclear magnetic resonance (NMR) can be used ex vivo to analyse cultured cells or biopsies from tumours, either by first extracting the sample into perchloric acid or chloroform/methanol or by high-resolution magic angle spinning (HR-MAS) NMR of solid samples. It is also possible to use magnetic resonance spectroscopy (MRS) noninvasively in vivo to obtain spectra from tumours in living animals or patients (note that when NMR-based methods are used in vivo it is conventional to drop the word “nuclear”). This ability to measure metabolites repeatedly and noninvasively in a living subject is a unique advantage of MRS, and although the number of metabolites detected is small, this method can be exploited in metabolomics.

In fact, all NMR and MRS methods are inherently insensitive and detect only a relatively small fraction of the metabolome. However, they have several compensating advantages in comparison with the more usual mass spectrometry or gas chromatography methods. When conventional high-resolution NMR is used on cultured cells or biopsies, the fact that it is not necessary to derivatize the metabolites within the sample, or to ionize them, removes two major sources of imprecision that impair the quantitative use of mass spectrometric data. Consequently the relative concentrations of the metabolites in a sample can be established by ex vivo NMR with a high degree of precision, making it easy to detect differences between pre- and post-drug spectra. When MRS is used in vivo on tumours in patients or experimental animals, it has an additional advantage, since repeated spectra can be obtained, giving a time-course. Time-course studies on the effect of a drug on its site of action – in this case the cancer – are termed pharmacodynamics.

Most MRS work on pharmacodynamics in vivo has been done using 1H and/or 31P MRS, while ex vivo studies with high-resolution NMR on cell or tissue extracts or by HR-MAS NMR on solid tissue biopsies are usually conducted by 1H NMR, although 31P NMR is sometimes also used. In vivo 31P MRS can be used to obtain markers for tissue bioenergetics (nucleotide triphosphate [NTP], inorganic phosphate [Pi]), intracellular pH (pHi) and membrane turnover (phosphorus-containing components of phospholipid membrane metabolism: phosphomonoester [PME] and phosphodiester [PDE] compounds). The composition of the PME and PDE peaks was established by in vitro 31P NMR of extracts of cancer cells or tumours, since these metabolite peaks are more readily resolved in vitro than in vivo (Evanochko et al. 1984; de Certaines et al. 1993). In tumours, PMEs are made up mainly of phosphocholine (PC) and phosphoethanolamine (PE), which are precursors of the membrane phospholipids phosphatidylcholine (PtdC) and phosphatidylethanolamine (PtdE). The PDEs are comprised of glycerophosphocholine (GPC) and glycerophosphoethanolamine (GPE), breakdown products of PtdC and PtdE, respectively. Thus, although membrane components such as PtdC and PtdE give peaks that are too broad to be detected, one can monitor their precursors and breakdown products. In addition, by 1H MRS in vivo it is possible to detect metabolites such as lactate, lipids and a peak that is usually assigned to total choline-containing compounds (tCho) (Shungu et al. 1992). The choline signals in the tCho peak are mainly from free choline, PC and GPC, but resonances from myo-inositol and taurine (Sitter et al. 2002); and from PE (Govindaraju et al. 2000) and GPE (Nelson et al. 1996) are also present in this region.

2 Pharmacodynamic Markers

Most of the published pharmacodynamic work in cancer has been done on routinely used chemotherapeutic drugs, but MRS can also offer important insights into novel anticancer agents in order to accelerate the drug development process. In addition, some classes of anticancer agents currently under development (e.g. antiangiogenics) are designed to be used in combination with other drugs and may not cause tumour shrinkage when used as single agents in Phase 1 clinical trials. Thus, MRS may have a special role in monitoring the pharmacodynamic actions of such drugs in early-phase clinical trials.

The following section demonstrates the use of metabolomic methods based on in vivo MRS and in vitro 1H and 31P NMR to assess tumour response to selected examples of conventional cytotoxic agents and to some novel drugs with specific molecular targets. MR-based biomarkers such as these could potentially provide surrogate pharmacodynamic markers for use in clinic trials.

Figure 1 shows some different patterns of metabolic change that are found by 31P MRS in tumours following treatment. The examples are drawn from four different classes of novel anticancer drug. Each of these drugs induces a different pattern of change in the 31P MR spectrum (indicated by the asterisks over the peaks), reflecting different alterations in the subset of the metabolome that is detected by this analytical modality. Although single spectra are shown in Fig. 1, the alterations in the peak areas indicated by the asterisks were found to be statistically significant in larger experiments.

Fig. 1
figure 1

In vivo 31P MRS profile of HT29 tumours following treatment with examples of four different classes of novel anticancer drug. The examples are: MN58b—choline kinase inhibitor (Al-Saffar et al. 2006); CYC202—CDK inhibitor (Troy et al. 2002); LAQ824—histone deacetylase inhibitor (Chung et al. 2007); 17-AAG—Hsp90 inhibitor (Chung et al. 2003)

Full size image

3 Conventional Cytotoxic Drugs

3.1 5-Fluorouracil

5-Fluorouracil (5-FU) is an antimetabolite drug that is widely used in medical oncology (Chen and Grem 1992). In vivo and in vitro 31P MRS were used to examine the pharmacodynamic effect of 5-FU on a mouse mammary carcinoma model (Street et al. 1997). Increased NTP to Pi and PCr to Pi ratios were observed in vivo 48 h after 5-FU treatment, implying an improvement in tumour energy metabolism. This increase in high-energy phosphate metabolites relative to Pi (which is also seen in response to a number of other anticancer drugs; see Sects. 3.2 and 3.3) could be due to a decrease in cell number allowing more effective access of oxygen and dissolved nutrients to the remaining cells. The PE to PC ratio was also elevated after 5FU treatment; this was found to be due to an increase in PE (as confirmed by in vitro 31P NMR). In addition, increases in GPC and GPE were also observed (Street et al. 1997).

3.2 Ifosfamide

Tumour growth inhibition was observed in a mouse xenograft model of paediatric embryonal rhabdomyosarcoma (Rd), 7 days after ifosfamide treatment. 31P MRS of Rd tumours in vivo showed significant increases in PME to Pi and β-ATP to Pi ratios after ifosfamide treatment when compared with pretreatment values. The rise in the PME to Pi ratio is due to an increase in PE and unchanged Pi levels (as confirmed by in vitro 31P NMR of tumour extracts) (Vaidya et al. 2003).

3.3 Cyclophosphamide

Increases in the NTP to Pi and PCr to Pi ratios were found in vivo in cyclophosphamide-treated mouse mammary carcinomas. The PE to PC ratio was also elevated after treatment, and this effect was found to be due to a drop in PC (as confirmed by in vitro 31P NMR of tumour extracts). In addition, rises in GPC and GPE levels were also found in extracts of cyclophosphamide-treated tumours when compared with controls (Street et al. 1995).

4 Cyclin-Kinase Inhibitor

CYC202 (R-roscovitine) inhibits the cyclin-dependent kinases 1, 2 and 7, and thus blocks cell cycle progression. Its action in vivo was monitored by 31P MRS in human colon xenografts to gain insights into the biochemical changes associated with cell cycle disruption. Following 4 days of CYC202 treatment, in vivo 31P MRS of HT29 tumours showed a significant decrease in intracellular pH, and the NTP to total phosphorus signal (TotP) ratio, and increases in the Pi to TotP and Pi to NTP ratios. In vitro 1H NMR of CYC202-treated tumour extracts showed falls in GPC, glycine and glutamate when compared with controls. CYC202 treatment caused reduction in tumour proliferation and tissue pH, and impairment in tumour bioenergetics (Chung et al. 2002).

5 HSP90 Inhibitor

The heat-shock protein HSP90 is of interest as an anticancer target (Necker 2002) because it helps maintain the shape of many oncogenic proteins. Inhibiting single oncogenes with “magic bullet” drugs has proved somewhat disappointing as the cancers often become resistant, so a “magic shotgun” that hits many oncogenes simultaneously is an attractive concept. The HSP90 inhibitor 17AAG is currently in clinical trial.

The actions of 17AAG were monitored on several cultured human colon cancer cell lines (HCT116, HT29 and SW620) and on a human colon tumour xenograft model (HT29) (Chung et al. 2003). In the HCT116, HT29 and SW620 cell lines, there were significant increases in PC and GPC. In vivo, after 4 days of 17AAG treatment, the HT29 tumours showed significant growth delay as well as increased ratios of PME to PDE, PME to TotP and PME to β-NTP, and a decrease in the β-NTP to TotP ratio. The rises in the PME ratios were due to increases in PC and PE, as confirmed by in vitro 1H and 31P NMR studies of 17AAG-treated tumour extracts when compared with controls. A significant inverse correlation was found between the percentage change in the PME to PDE ratio and the percentage change in tumour size following 17AAG treatment (Chung et al. 2003)

6 Choline Kinase Inhibitor

Choline kinase (ChoK) is a cytosolic enzyme that catalyses the phosphorylation of choline to form PC, which is involved in cell membrane synthesis. Elevated levels of PC and ChoK found in tumours are associated with cell proliferation and malignant transformation. MN58b is an inhibitor of ChoK.

A significant growth delay was observed in the MN58b-treated HT29 xenografts when compared with controls. In vivo, 31P and 1H MRS of the HT29 xenografts showed a decrease in the ratio of PME to TotP and a decrease in total choline (tCho) concentration after 5 days of MN58b treatment. Extracts of drug-treated HT29 tumours showed significant decreases in PC when compared with controls. No changes in the other phospholipid metabolites (PE, GPC and GPE) were observed. Similar metabolite changes to those in HT29 tumours were also found in MN58b-treated MDA-MB-231 tumours (Al-Saffar et al. 2006). A drop in PC was also found in MN58b-treated HT29 cell extracts when compared with controls (Al-Saffar et al. 2006).

7 HDAC Inhibitors

7.1 LAQ824

LAQ824 is a novel anticancer drug that inhibits histone deacetylase (HDAC), resulting in growth inhibition, cell cycle arrest and apoptosis. Significant tumour growth inhibition was observed in HT29 xenografts following 2 days of LAQ824 treatment when compared with vehicle-treated controls. In vivo, the ratio of PME to TotP was significantly increased in LAQ824-treated HT29 xenografts and this ratio was found to correlate inversely with tumour response. This PME increase is confirmed by the significant rises in PC and PE levels observed in 1H- and 31P-NMR spectra of LAQ824-treated tumour extracts when compared with controls. These increases in PC and PME metabolites could potentially be used as biomarkers of HDAC inhibition (Chung et al. 2007).

Marked decreases in NTP and PCr, and an increase in Pi were also found in vivo by 31P-MRS of LAQ824-treated tumours; in addition, significant decreases in intracellular pH, and in the β-NTP to TotP and β-NTP to Pi ratios, and an increase in Pi to TotP ratio were observed. These observations indicate that the tumour's bioenergetics are severely compromised following treatment, in contrast to the rises in bioenergetic state observed with 5-FU, ifosfamide and cyclophosphamide (see Sects. 3.1, 3.2, and 3.3). Elevated free choline, leucine, iso-leucine and valine levels and reduced GPC, GPE, glutamate, glutamine, glucose PCr and creatine levels were found in LAQ824-treated HT29 tumour extracts when compared with controls. These metabolite changes are also consistent with impaired tumour bioenergetics. A marked reduction of CD31 staining was found in LAQ824-treated tumours, indicating reduced vessel density in the LAQ824-treated group when compared with controls. The metabolite changes found in treated tumour extracts, the vascular changes and the effects that were observed in tumour bioenergetics, are consistent with the known antiangiogenic effect (Qian et al. 2004, and see Sect. 8.1) of LAQ824 on solid tumours (Chung et al. 2007).

Inhibition of HDAC by LAQ824 resulted in altered phospholipid metabolism and compromised tumour bioenergetics. The changes in phospholipid metabolism might function as noninvasive biomarkers of HDAC inhibition per se, whereas the alterations in energy-associated metabolites could be used as biomarkers of the drug's antiangiogenic effects (Chung et al. 2007).

7.2 SAHA

SAHA is an anticancer drug that acts by inhibition of histone deacetylase (HDAC). Sankaranarayanapillai et al. used in vivo 1H MRS and in vitro 31P and 13C NMR to study the effects of SAHA on a prostate cancer line (PC3). Increased PC and total choline levels were found in SAHA-treated cell extracts and these changes were inversely correlated with HDAC activity (Sankaranarayanapillai et al. 2005, 2007a). Increases in tCho to total signal and tCho to lipid ratios were found in PC3 tumours after 2 days of SAHA treatment. However, tCho normalized to the internal water signal remained unchanged, in contrast to the cell extract data (Sankaranarayanapillai et al. 2007b).

7.3 Phenylbutyrate

Phenylbutyrate is also an HDAC inhibitor. Increases in GPC, tCho and NMR-visible lipids were found in DU145 human prostatic carcinoma cells treated with phenylbutyrate. These effects were accompanied by significant increases in cytoplasmic lipid droplets and intracellular lipid volume fraction as observed by morphometric analysis of Oil Red O-stained cells. Phenylbutyrate treatment of cells caused cell cycle arrest in the G1 phase and induction of apoptosis. The simultaneous accumulation of mobile lipid and GPC suggests that phenylbutyrate induces phospholipid catabolism via a phospholipase-mediated pathway (Milkevitch et al. 2005).

8 Vascular Disruption Agents

Vascular disruption agents (VDAs) are novel anticancer drugs that destroy the blood vessels supplying a tumour, eventually causing massive necrosis.

8.1 DMXAA

DMXAA is a VDA. Following treatment with DMXAA, HT29 tumours showed dose-dependent decreases in both β-NTP to Pi and PDE to PME 6 h after treatment, when compared with vehicle-controls (MacPhail et al. 2005). A significant decrease in tCho in vivo was found 24 h after treatment with 21 mg/kg DMXAA; this was associated with a significant reduction in the concentration of the membrane degradation products GPE and GPC measured in tissue extracts. Elevated free choline was found in DMXAA-treated tumour extracts when compared with vehicle controls. These reductions in tumour energetics and membrane turnover are consistent with the vascular-disrupting activity of DMXAA. In vivo 31P MRS revealed tumour response to DMXAA at doses below the maximum tolerated dose for mice (MacPhail et al. 2005), so this method might have use as a surrogate biomarker for this class of agent.

8.2 ZD6126

Radiation-induced fibrosarcoma 1 (RIF-1) tumours treated with ZD6126, another VDA, showed a significant reduction in tCho in vivo, 24 h after treatment, whereas vehicle-treated control tumours showed a significant increase in tCho. Ex vivo HRMAS NMR of tumour tissues and 1H NMR of tumour extracts revealed significant reductions in PC and GPC in ZD6126-treated tumours; this confirmed the in vivo tCho finding. ZD6126-induced reduction of the amount of choline-containing compounds is consistent with a reduction in cell membrane turnover associated with necrosis and cell death following disruption of the tumour vasculature (Madhu et al. 2006).

8.3 Combretastatin A4 Phosphate

Combretastatin A4 phosphate is a VDA. Significant drops in the β-NTP to Pi ratio and intracellular pH were observed in C3H murine mammary tumours 1 h after combretastatin A4 phosphate treatment. An increase in lactate level was also found after treatment, but this effect was not observed consistently. The reduction in tumour energetics and pH was consistent with a reduction in tumour blood flow, but this occurred before any significant incidence of haemorrhagic necrosis was detected (Maxwell et al. 1998).

The acute effects of the antivascular drug combretastatin A4 phosphate were further investigated by Beauregard et al. The tumour bioenergetics of five tumour models – LoVo, RIF-1, SaS, SaF and HT29 – were examined by in vivo 31P MRS following treatment with combretastatin A4 phosphate. A significant increase in the Pi to NTP ratio was observed by in vivo localized 31P MRS in LoVo and RIF-1 tumours 3 h after treatment. SaS, SaF and HT29 tumours did not respond to the same degree. This tumour susceptibility to combretastatin A4 phosphate was found to correlate with vascular permeability (Beauregard et al. 2001).

9 HIF-1α Inhibitor

The hypoxia-inducible transcription factor HIF-1 plays an important role in the development of many tumours. PX-478 is an inhibitor of HIF-1α, one of the two subunits of the HIF-1 protein. In vivo 1H MRS showed a significant decrease of the tCho in HT29 xenografts after 12 and 24 h of PX478 treatment. These changes were due to decreases in PC and GPC, as confirmed by high-resolution 1H and 31P NMR of tumour extracts. Reductions of PE, GPE and myo-inositol were also found in PX-478-treated tumour extracts when compared with controls. Significant reductions in cardiolipin, PtdE and phosphatidylinositol (PtdI) were also observed in lipid extracts of tumours after PX478 treatment when compared with vehicle controls. The significant change in tCho could potentially be used as an in vivo MRS biomarker for drug response following HIF-1α inhibition. The in vitro metabolic profiles of tumour indicated that GPC, PC, myo-inositol, PE, GPE, CL, PtdE and PtdI are potential ex vivo response biomarkers (Jordan et al. 2005).

10 PI3K Inhibitor

10.1 LY294002 and Wortmannin

LY294002 and Wortmannin are inhibitors of the PI3 kinase (PI3K) pathway. In vitro 31P NMR of MDA-MB-231, MCF-7, and Hs578T cell extracts showed significant decreases in PC and increases in GPC levels following LY294002 treatment. A significant drop in the NTP level was also found in Hs578 cells following treatment, but not in MCF-7 or MDA-MB-231 cells. The drop in PC and rise in GPC levels were also observed by 31P NMR of intact MDA-MB-231 cells following exposure to LY294002. A significant decrease in PC was also observed in extracts of MDA-MB-231 cell following treatment with another PI3K inhibitor, wortmannin, and no significant changes in the other metabolite levels were found. This study indicates that PI3K inhibition in human breast cancer cells by LY294002 and wortmannin is associated with a decrease in PC levels (Beloueche-Babari et al. 2006).

10.2 PI103

Treatment of PC3 cells with PI103, a PI3K inhibitor, caused a dose- and time-dependent decrease in PC, PE and NTP levels. These metabolite changes were associated with the drop in AKT phosphorylation and choline kinase activity (Al-Saffar et al. 2007).

11 MAPK Inhibitor

U0126 is a mitogen-activated protein kinase (MAPK) signalling inhibitor. Treatment of MDA-MB-231, MCF-7 and Hs578 cells with U0126 caused inhibition of extracellular signal-regulated kinase (ERK1/2) phosphorylation and a significant drop in PC, as shown by 31P NMR of cell extracts. Similar changes were also observed in colon carcinoma HCT116 cells following exposure to U0126. The reductions in PC level in MDA-MB-231 and HCT116 cells were significantly correlated with the drop in P-ERK1/2 levels. This study showed that MAPK signalling inhibition with U0126 is associated with a time-dependent decrease in cellular PC levels (Beloueche-Babari et al. 2005).

12 Fatty Acid Synthase Inhibitor

Fatty acid synthase (FASE) is a key enzyme that catalyses the terminal steps in the synthesis of saturated fatty acids. FASE expression is low in normal human tissues because most lipids are obtained from the diet. Over-expression of FASE has been found in a wide variety of human cancers and is associated with a poor prognosis. Hence, FASE is an attractive therapeutic target for developing novel anticancer drugs such as orlistat, a fatty acid synthase inhibitor.

Treatment of PC3 cells with orlistat caused a drop in FASE activity and inhibition of cell proliferation (Ross et al. 2007). In vitro 31P and 1H NMR of cell extracts and 13C NMR of extracts of cells treated with labelled choline show reduced levels of fatty acids, PtdC and PC following orlistat treatment. These data indicated the inhibition of de novo synthesis of these metabolites after treatment. Correlations were found between inhibition of FASE and inhibition of de novo synthesis of fatty acids, PtdC and PC (Ross et al. 2007).

13 Antimicrotubule Drug

Docetaxel is an antimicrotubule agent. Significant decreases in tumour PC levels were observed in two breast tumour models, MCF-7 and MDA-MB-231, 2–4 days after docetaxel treatment. A significant decrease in PC was found in vivo after docetaxel treatment, and this observation was confirmed by in vitro NMR of tumour extracts. An increase in GPC was also found in docetaxel-treated tumour extracts. These changes occurred in parallel with tumour growth delay, cell-cycle arrest and cell death (Morse et al. 2007). Since PC is a precursor and GPC is a breakdown product of PtdC in phospholipid membranes, these results would be consistent with a decreased synthesis and increased degradation of cell membranes.

14 Discussion

Table 1 summarizes some of the responses of cancer cells and tumours to the drugs mentioned in this review. The responses highlighted here are those that might be used in vivo, and might therefore be useful for monitoring clinical trials. They comprise cellular bioenergetic parameters that can be monitored by 31P MRS, changes in the PME peak that can also be monitored by 31P MRS, and tCho compound changes, which can be monitored by in vivo 1H MRS. Not all of the responses noted here were actually measured in solid tumours in vivo. Some were measured in extracts of cultured cells and others by ex vivo NMR of tumour biopsies, but in principle it should be possible to monitor such drug actions by noninvasive MRS using currently available 1.5T and 3T instruments, at least if the tumours are in superficial bodily sites such as the lymph nodes, the breast, etc. Thus, this limited subset of metabolomic features (bioenergetic metabolites, phosphomonoesters and choline compounds), which show characteristic changes in response to anticancer drugs, could be monitored in some clinical trials. In the brain, it is possible to monitor several more metabolites by 1H MRS (lactate, N-acetylaspartate, creatine and myo-inositol) so more sophisticated studies might be possible; however, very few anticancer drugs cross the blood–brain barrier (Murphy et al. 2004), so there is at present little need for a way to monitor chemotherapy in brain cancer. Tumours in the pelvis—particularly the prostate (Heerschap et al. 1997; Zakian et al. 2003) and also the cervix (Mahon et al. 2004a, b)—also give quite good 1H MRS spectra.

Certain common features can be seen in the responses tabulated in Table 1. Increased bioenergetic metabolites were observed when tumours in animals were treated with the antimetabolite 5-FU and the alkylating agents ifosfamide and cyclophosphamide, indicating an apparently paradoxical improvement in the tumour's bioenergetic status. This phenomenon, which is not usually observed when tumours in patients (rather than animals) are treated, may perhaps be due to improved blood flow to the remaining tumour tissue as the tumour shrinks rapidly in response to the high drug doses that can be given to animals. Another factor may be the poor blood supply seen in the implanted subcutaneous tumours that are usually studied in animals. In contrast, drugs of several classes have been seen to cause decreases in tumour bioenergetic metabolites: a cyclin kinase inhibitor, an HSP90 inhibitor, a histone deacetylase inhibitor, two VDAs and two PI3 kinase inhibitors. The principle mechanism of action of the VDAs is obvious: they destroy the blood vessels and thus block the access of oxygen and nutrients. LAQ824 was the only one of the three histone deacetylase inhibitors to cause a decrease in bioenergetic metabolites, and this drug is known to have a vascular disruption action as well as inhibition of histone deacetylase (Qian et al. 2004). The mechanism(s) by which the other drugs reduce bioenergetic metabolites is currently unclear.

Table 1 Summary of responses to drugs that can be detected by 31P or 1H MRS in vivo and ex vivo. In some cases, the assays reported were performed ex vivo, but in principle the metabolites in question could be detected noninvasively in vivo. Unless otherwise stated, measurements were performed on tumours in animals
Full size table

The drugs that caused increases or decreases in the PME peak in response to drug treatment are shown in the next column. Increased PME was observed in response to the antimetabolite 5-FU and the alkylating agents ifosfamide and cyclophosphamide, the HSP90 inhibitor 17AAG and the histone deacetylase inhibitors LAQ824 and SAHA. In contrast, decreased PME was observed in response to the choline kinase inhibitor MN58b and the antimicrotubule agent docetaxel.

The next column shows the drugs that caused increases and decreases in the tCho signal following drug therapy. Increased tCho was observed in response to the histone deacetylase inhibitor SAHA. Reduced tCho was found following treatment with the choline kinase inhibitor MN58b; the HIF-1α inhibitor PX478, and two of the vascular disrupting agents DMXAA and ZD6126.

The penultimate column lists the changes in phospholipid metabolites (i.e. PC, PE, GPC and GPE) and free choline following therapy, assessed by ex vivo 1H or 31P NMR. The changes of these metabolites confirmed the in vivo modulation of PME and/or PDE (by in vivo 31P MRS) and tCho (by in vivo 1H MRS). A reduced PC level was found following treatment with many different classes of anticancer drug. These drugs include the alkylating agent cyclophosphamide, the choline kinase inhibitor MN58b, the VDA ZD6126, the HIF-1α inhibitor PX478, the PI3K inhibitors—PI103, LY294002 and wortmannin—the MAPK inhibitor U0126, the fatty acid synthase inhibitor orlistat, and the antimicrotubule agent docetaxel. In cases where both in vivo and ex vivo measurements were carried out, a drop in PC is associated with decreased PME and/or tCho. Reduction in the PC level is generally associated with decreased cell membrane synthesis and proliferation, which is consistent with the expected response following therapy (Ackerstaff et al. 2003; de Certaines et al. 1993).

PME increases due to an elevated level of PE were found in response to the alkylating agent ifosfamide. A rise in the PE/PC ratio (due to a drop in PC) was seen in response to the antimetabolite 5-FU and the alkylating agent cyclophosphamide. However, increases in PME and/or tCho were also observed following treatment with the HSP90 inhibitor 17AAG and two of the histone deacetylase inhibitors, LAQ824 and SAHA. In these cases the PME change was due to increases in PC and PE or PC alone. A rise in PC following a positive response to an anticancer drug treatment is unusual, as elevated PC is normally associated with tumour growth (Ackerstaff et al. 2003; Podo 1999; Aboagye and Bhujwalla 1999). A rise in PC was found in response to the Hsp90 inhibitor and two of the HDAC inhibitors; these two classes of drug both exert inhibitory effects on HSP90 but the mechanism behind this PC increase remained unclear. However, Sreedhar et al. (2003) reported that HSP90 plays an important role in the maintenance of cellular integrity. Hence, one might speculate that the rise in PC may be caused by the release of PC from the cell membrane following the compromised cellular integrity due to inhibition of HSP90.

GPC and GPE are associated with cell membrane breakdown. Changes of GPC and GPE are found to occur following responses to many different classes of anticancer drug but the mechanisms behind these changes remain unclear and require further investigation.

31P MRS studies have been conducted for many years on the effect of classical anticancer drugs on tumours in vivo (reviewed by Negendank 1992; de Certaines et al. 1993). In general, it has been found that the PME signals are the most useful for monitoring drug treatments by 31P MRS, as they fall in response to most forms of therapy. The measurements of PME reported in these two earlier reviews were mainly performed by 31P MRS in vivo. This method has the advantage that it is noninvasive, but in most cases only a single PME peak can be resolved in vivo. In contrast, the studies reviewed in the present work include many measurements performed using 1H MRS in vivo and 1H and/or 31P NMR ex vivo on extracts of cultured cells and tumour biopsies, or by HR-MAS NMR on solid tumour biopsies. In recent years, in vivo 1H MRS to measure tCho for monitoring drug therapy became more widely used and a fall in tCho has been found following a number of types of drug treatment. The ex vivo methods that are also reported in the present review resolve more metabolites because of improved spectral resolution. They are complementary to the in vivo measurements, as they can provide additional information and help in the interpretation of the in vivo data. For instance, ex vivo methods can be used to pinpoint the mechanisms underlying the modulation of the PME or tCho signals.

This review has demonstrated that metabolomics by MRS and NMR methods has many applications for monitoring pharmacodynamics of novel anticancer drugs. Much work remains to be done, however, on the metabolic mechanisms underlying the effects observed.