Detecting treatment response in a model of human breast adenocarcinoma using hyperpolarised [1-13C]pyruvate and [1,4-13C2]fumarate

Background: The recent introduction of a dynamic nuclear polarisation technique has permitted noninvasive imaging of tumour cell metabolism in vivo following intravenous administration of 13C-labelled cell substrates. Methods: Changes in hyperpolarised [1-13C]pyruvate and [1,4-13C2]fumarate metabolism were evaluated in both MDA-MB-231 cells and in implanted MDA-MB-231 tumours following doxorubicin treatment. Results: Treatment of MDA-MB-231 cells resulted in the induction of apoptosis, which was accompanied by a decrease in hyperpolarised 13C label flux between [1-13C]pyruvate and lactate, which was correlated with a decrease in the cellular NAD(H) coenzyme pool. There was also an increase in the rate of fumarate conversion to malate, which accompanied the onset of cellular necrosis. In vivo, the decrease in 13C label exchange between pyruvate and lactate and the increased flux between fumarate and malate, following drug treatment, were shown to occur in the absence of any detectable change in tumour size. Conclusion: We show here that the early responses of a human breast adenocarcinoma tumour model to drug treatment can be followed by administration of both hyperpolarised [1-13C]pyruvate and [1,4-13C2]fumarate. These techniques could be used, therefore, in the clinic to detect the early responses of breast tumours to treatment.

There has been substantial growth and investment in drug discovery and development programmes in oncology in recent years, with the aim of producing therapeutics targeted at the spectrum of genetic abnormalities found in cancer (Hu et al, 2000;Semenza, 2003;Bryant et al, 2005). This rapid expansion has moved the field of clinical oncology closer to the realisation of personalised medicine, in which patients receive drugs, individually or in combination, that are tailored to their specific disease. The delivery of this approach should be facilitated by noninvasive imaging methods that allow an early assessment of treatment response, identifying unsuccessful treatments at an early stage and enabling the selection of more effective treatment (reviewed in Brindle (2008a)).
Currently, tumour treatment responses are assessed from reductions in tumour size (Eisenhauer et al, 2009). However, this approach lacks sensitivity and many weeks may elapse before a change in size is detected (Neves and Brindle, 2006;Weissleder and Pittet, 2008;Brindle, 2008a). In some cases, for example with cytostatic therapies, there may be no change in size despite a positive response to treatment (Brindle, 2008a). Measurements of tumour physiology or biochemistry, however, can give a much earlier indication of treatment response. Positron emission tomography (PET) measurements of the uptake of the glucose analogue [ 18 F] 2-fluoro-2-deoxy-D-glucose (FDG) are being used increasingly in the clinic to stage disease and provide early evidence of treatment response (Eisenhauer et al, 2009). MRS, particularly 1 H MRS, can also be used to detect the metabolic changes that accompany a positive response to treatment (Aboagye and Bhujwalla, 1999;Glunde et al, 2006). The relative lack of sensitivity of MRS, however, limits the spatial and temporal resolution of 1 H spectroscopic imaging experiments and in many cases only single voxel studies are performed . Although 1 H MRS experiments are performed widely in the clinic, this lack of sensitivity has limited their routine use. This situation may change with the recent introduction of a dynamic nuclear polarisation (DNP) technique that can increase sensitivity in the solution state 13 C MRS experiment by 410 000-fold (Ardenkjaer-Larsen et al, 2003). The enormous gain in sensitivity means that, following injection of a hyperpolarised is B30 s in vivo, which means that the material must be injected and imaged within B5 min. Thus, in order to image metabolism using this technique, the hyperpolarised 13 C-labelled substrate must be taken up rapidly by the cell and its subsequent metabolism must be very fast. Even then, it is often only possible to monitor a single enzyme-catalysed reaction (Gallagher et al, 2009a). Nevertheless, the technique has already shown promise for detecting treatment response in tumours (Day et al, 2007;Witney et al, 2009).
The exchange of hyperpolarised 13 C label between [1-13 C]pyruvate and lactate, in the reaction catalysed by lactate dehydrogenase (LDH), was shown to decrease in a drug-treated murine lymphoma in vivo (Day et al, 2007), where decreased label flux was due to a number of factors, including: DNA damagemediated activation of polyADP-ribose polymerase (PARP) and consequent depletion of the NAD(H) coenzyme pool, a loss of LDH activity, and a reduction in tumour cellularity (Day et al, 2007;Witney et al, 2009). A similar study in the same tumour model using [1,4-13 C 2 ]fumarate, showed that the rate of the fumarase catalysed conversion of fumarate to malate was a measure of subsequent drug-induced cellular necrosis (Gallagher et al, 2009b). With a clinical trial using hyperpolarised [1-13 C]pyruvate about to start in prostate cancer (Brindle, 2008b), there is a reasonable expectation that the technique could translate to the clinic, where it offers a new functional imaging approach to detect early tumour responses to treatment.
We show here, in a model of human breast adenocarcinoma, that a combination of hyperpolarised [1-13 C]pyruvate and [1,4-13 C 2 ]fumarate can be used to detect response to doxorubicin treatment before there is any detectable change in tumour size. There was a decrease in labelled lactate production, reflecting a DNA damage response and initiation of the apoptotic program, and an increase in labelled malate production, reflecting the onset of cellular necrosis.

Flow cytometry
Apoptosis and necrosis were measured by flow cytometry. Cells were trypsinised (0.25% trypsin; 1 mM EDTA) and harvested by centrifugation (1300 g, 3 min). Detached cells present in the media before trypsinisation were retained and pooled with the trypsinised cells. Cell pellets (1 Â 10 6 cells) were washed once in ice-cold HEPES-buffered saline (10 mM HEPES, 140 mM NaCl, 2.5 mM CaCl 2 , pH 7.4) and resuspended in 100 ml of the same buffer. Annexin V-Pacific Blue (5 ml per 100 ml of cell suspension) and SYTOX Red (Invitrogen) (5 nM) were added to the cells and incubated for 15 min at 201C. The resulting mixture was washed once in ice-cold HEPES-buffered saline, kept briefly on ice and then analysed in an LSRII cytometer (BD Biosciences, Rockville, MD, USA), with 20 000 cells counted per event. Viable cells were identified as cells that stain with neither annexin V nor SYTOX Red; apoptotic cells as cells that stain with annexin V, but not SYTOX Red; and necrotic cells as cells that stain with both annexin V and SYTOX Red. Cellular NADH content was assessed from measurements of autofluorescence at 455 nm following excitation at 350 nm (Day et al, 2007;Witney et al, 2009). Cytosolic NADH has been shown previously to make a significant contribution to this autofluorescence (Aubin, 1979) and it is the cytosolic, rather than the mitochondrial NAD(H) pool, that is depleted the following activation of PARP (Ying et al, 2005).

C MRS measurements on cells
Cells (2 -6 Â 10 7 ) were examined in a 10-mm NMR tube, using a broadband probe (Varian NMR Instruments, Palo Alto, CA, USA) in a 9.4-T vertical wide-bore magnet (Oxford Instruments, Oxfordshire, UK) interfaced to a Varian INOVA console (Varian NMR Instruments, Paulo Alto, CA, USA). The sample temperature was maintained at 371C. Hyperpolarised [1-13 C]pyruvate (75 mM), hyperpolarised [1,4-13 C 2 ]fumarate (20 mM) and nonhyperpolarised, unlabelled lactate (75 mM) were injected into the cell suspension and single transient 13 C spectra were acquired every second for 240 s, using a 61 flip angle pulse and a spectral width of 32 kHz. The area under each peak was used to calculate concentration of the labelled metabolites. The two inequivalent malate peaks, 13 C-labelled at the C-1 and C-4 position, were combined to produce a single measurement of intensity (Gallagher et al, 2009b). The peak intensities of hyperpolarised metabolites were fitted to two-site exchange models describing the fluxes of hyperpolarised 13 C label between pyruvate and lactate (k P ), and fumarate and malate (k F ; Day et al, 2007).

Tumour implantation
Female severe combined immunodeficiency mice were purchased at 6 -8 weeks of age from Harlan UK Ltd (Bicester, UK). Tumour cells (1 Â 10 7 ) were injected subcutaneously into the shaved flanks of mice. Tumour size was measured using callipers and is reported as the product of the two largest perpendicular diameters (mm 2 ). Mice were imaged 24 h before treatment with an i.v. injection of 10 mg doxorubicin per kg body weight and at 24 and 48 h post-treatment. All procedures were carried out in accordance with the Animals (Scientific Procedures) Act of 1986 (UK) and were designed with reference to the UK Co-ordinating Committee on Cancer Research Guidelines for the Welfare of Animals in Experimental Neoplasia.

C MRS and 1 H MRI measurements on tumours
Mice were anesthetised with an intraperitoneal injection of 10 ml per kg body weight of a 5 : 4 : 31 mixture of Hypnorm (VetaPharma Ltd, Leeds, West Yorkshire, UK), Hypnovel (Roche, Basel, Switzerland) and saline and a catheter inserted into a tail vein. A 25 mm diameter surface coil tuned to 13 C (100 MHz) was positioned over the tumour. The entire assembly was placed in a quadrature 1 H-tuned volume coil (Varian), in a 9.4-T vertical wide-bore magnet. Transverse 1 H images were acquired from the tumour using a spin-echo pulse sequence (repetition time, 1.5 s; echo time, 30 ms; field of view, 32 mm Â 32 mm; data matrix, 256 Â 256; slice thickness, 2 mm; 11 slices). Following injection of hyperpolarised [1-13 C]pyruvate, which was accomplished within 3 s, 128 single transient spectra were collected from a 6 mm tumour slice over a period of 128 s (every sixteenth spectrum was collected from the entire sensitive volume of the surface coil). Spectra were acquired using a slice-selective 600 ms sinc pulse, with a nominal flip angle of 51. C-1 lactate and pyruvate peak intensities were fitted to the modified Bloch equations for two-site exchange to obtain an apparent rate constant for label exchange between pyruvate and lactate (k P ), as described previously (Day et al, 2007). Following [1-13 C]pyruvate injection, [1,4-13 C 2 ]fumarate was polarised and injected into the animal, which was still anesthetised. This was achieved within 1 h of the previous injection. Owing to the low-malate signals observed, dynamic measurements of fumarate conversion to malate could not be performed in vivo. Instead, a single 13 C spectrum, collected from a 6 mm tumour slice, was acquired 20 s after injection of 0.2 ml fumarate using a nominal flip angle of 201; a time at which the maximum malate peak intensity had been observed in a previous dynamic study in a drug-treated murine lymphoma tumour (Gallagher et al, 2009b). Data were expressed as the malate/fumarate ratio, calculated from the intensities of the [1,4-13 C 2 ]fumarate resonance and the combined resonances from [1-13 C]malate and [4-13 C]malate.

Statistics
Results are expressed as the mean±s.d. Significant differences between values were determined using a Mann -Whitney nonparametric test. Statistical analysis was performed using GraphPad Prism version 5.0 for windows (GraphPad Software, San Diego, CA, USA). Differences between treatment groups were considered significant if Po0.05.

Measurement of cell death following drug treatment in vitro
Drug-induced apoptosis and necrosis of human breast cancer cells (MDA-MB-231) were assessed by flow cytometry, following staining of the cells with Annexin V-Pacific Blue and SYTOX Red, respectively ( Figure 1). There were significant increases in apoptosis and necrosis by 48 h after treatment with doxorubicin, which is a commonly used chemotherapeutic drug (Gewirtz, 1999; Figure 1A). Annexin V binding to apoptotic cells correlated with the induction of caspase-3 activity at 48 h, as indicated by PARP cleavage (Kaufmann et al, 1993). There was a small decrease in the level of intact PARP at 48 h, with a 26 and 20% decrease compared with cells at 0 h in two independent experiments, with the level of uncleaved PARP becoming almost undetectable at 72 h ( Figure 1A, inset). The apoptotic fraction peaked at 72 h, before declining at 96 h, whereas the necrotic fraction was increased at 72 h and further increased at 96 h.

DISCUSSION
Imaging currently has an important role in the management of breast cancer patients by enabling assessment of treatment response following primary or neoadjuvant therapy. Therapeutic response is assessed primarily by measurements of tumour shrinkage, using a combination of computed tomography (CT) and/or MRI (reviewed in Ollivier et al (2005)). However, there are problems with these imaging modalities. CT has the advantages of simplicity, speed and availability, however, changes in tumour size are frequently overestimated in diffuse or multinodular tumours (Ollivier et al, 2005). Changes in tumour size can also be overestimated by MRI, particularly with smaller tumours and if there is surrounding tissue oedema (Merchant et al, 1993). Dynamic contrast-enhanced MRI can provide further diagnostic information, with the intensity of contrast enhancement shown to predict the presence of residual tumour after preoperative chemotherapy (Gilles et al, 1994). There is a need, however, to develop imaging methods that can detect treatment response more sensitively and before there is any change in tumour size. FDG-PET is being used increasingly in the clinic as a functional imaging technique for detection, staging and early response monitoring in breast cancer (Eubank and Mankoff, 2005). Although currently most useful as a staging tool, initial studies have shown FDG-PET to be capable of predicting pathological response in patients with locally advanced breast cancer following a single course of neoadjuvant chemotherapy (Biersack et al, 2004;Krak et al, 2004). However, the presence of infected or inflamed tissue, and high uptake by infiltrating immune cells can mask changes in tumour FDG uptake (Strauss, 1996;Zhuang et al, 2005). Patients also receive a high radiation dose, with scans being costly, time-consuming and are only available in a few selected centres. Complementary methods for the detecting treatment response in  breast cancer are therefore still required that provide alternative biological readouts that can be used in conjunction with existing methods in a multimodality approach. Treatment of murine EL-4 lymphoma cells with a topoisomerase II inhibitor and DNA damaging agent, etoposide, was shown previously to inhibit the rate of hyperpolarised 13 C label exchange between [1-13 C]pyruvate and lactate, catalysed by the enzyme LDH. The decreased exchange was explained by induction of a DNA damage response and consequent induction of PARP, for which the coenzyme NAD þ is a substrate (Day et al, 2007;Witney et al, 2009). We have shown here that doxorubicin treatment of the oestrogen receptor negative-human breast cancer cell line, MDA-MB-231, in vitro resulted in a marked decrease in the rate of hyperpolarised 13 C label exchange between [1-13 C]pyruvate and lactate and that this was correlated with a decrease in the concentration of the cellular NAD(H) coenzyme pool. Doxorubicin, like etoposide, is a topoisomerase II inhibitor and a DNA damaging agent that induces PARP activation (Munoz-Gamez et al, 2005). Therefore, this hyperpolarised [1-13 C]pyruvate method for detecting response to genotoxic agents would appear to be generally applicable to other tumour types.
The hyperpolarised [1-13 C]pyruvate experiment shows that the tumour cell has responded to treatment, but not that it has died. Moreover, it provides 'negative contrast', in that there is a decrease in lactate signal intensity in those tumours that have responded to treatment (Day et al, 2007;Witney et al, 2009). Such a decrease in lactate signal could also be due to other factors, such as reduced delivery of pyruvate to the tumour. In contrast, hyperpolarised 13 C-labelled [1,4-13 C 2 ]fumarate provides positive contrast and appears to be a marker of cell death (Gallagher et al, 2009b). Fumarate shows relatively slow plasma membrane transport, and thus, in viable cells there is no detectable conversion of fumarate to malate within the relatively short lifetime of the 13 C polarisation. However, in necrotic cells, where the plasma membrane permeability barrier has been compromised, fumarate is rapidly  converted to malate and the hyperpolarised [1,4-13 C 2 ]malate signal provides a positive marker of tumour cell necrosis. Treatment of MDA-MB-231 cells with doxorubicin resulted in the production of significant levels of labelled malate by 72 h after drug treatment, a time at which there was also a significant decrease in the rate of labelled lactate production. There was no detectable labelled malate produced in untreated cells. The rate of labelled malate production showed a good correlation with the level of tumour cell necrosis, as has been observed previously (Gallagher et al, 2009b). As with the hyperpolarised [1-13 C]pyruvate method for detecting response to genotoxic agents, these experiments have demonstrated that the hyperpolarised [1,4-13 C 2 ]fumarate experiment could be a generic method for detecting tumour cell death posttreatment. Experiments with hyperpolarised [1-13 C]pyruvate and [1,4-13 C 2 ]fumarate were also performed in implanted MDA-MB-231 tumours, where we have shown a response to doxorubicin treatment in the absence of any detectable change in tumour size. Co-hyperpolarised [1-13 C]pyruvate and [1,4-13 C 2 ]fumarate preparations could not be used in vivo as malate production was masked by overlapping signals from lactate and pyruvate hydrate formed from the labelled pyruvate in these lower resolution 13 C spectra. When the pyruvate was injected alone, a significant decrease in pyruvate -lactate exchange was observed at 48 h post-therapy at the maximum tolerated dose of 10 mg kg À1 doxorubicin. This relatively slow response to therapy is consistent with that observed in cultured cells and reflects the relative resistance of MDA-MB-231 tumours to genotoxic agents (Fang et al, 2007). At 24 h, after treatment with doxorubicin and 20 s after the injection of hyperpolarised [1,4-13 C 2 ]fumarate the hyperpolarised 13 C malate/fumarate ratio increased 5.4-fold ( Figure 5B), indicating the onset of tumour cell necrosis. At this time point, a decrease in pyruvate -lactate exchange was also observed, but did not reach significance. Therapy response was observed earlier in tumours than in the cell experiments, with the larger number of tumour cells in vivo giving an increase in sensitivity.
The level of tumour cell death after drug treatment has been shown, in preclinical and clinical studies, to be a good prognostic indicator for treatment outcome, including in breast cancer (Chang et al, 2000). However, early evidence of treatment response does not necessarily indicate a favourable long-term outcome. For example, an early decrease in FDG uptake in oesophageal adenocarcinoma in the clinic following neoaduvant treatment did not correlate with a histopathological response in the longer term (Smithers et al, 2008) and in head and neck cancers treated with radiotherapy uptake of l-[methyl-11 C]methionine was reduced, but there was no correlation with treatment outcome (Nuutinen et al, 1999). However, in the case of the MDA-MB-231 human breast cancer model used in this study, a single treatment with doxorubicin, at a slightly lower dose than used here, has been shown to produce a long-term treatment response, as indicated by a tumour growth delay of B20 days (Mason et al, 2008). Thus, the imaging methods that we have described here can detect early evidence of a treatment response that is correlated with outcome in the longer term.
In summary, we have shown that hyperpolarised 13 C experiments for detecting early evidence of treatment response, which had previously only been demonstrated in an implanted murine lymphoma tumour model that responds very rapidly to treatment with a genotoxic agent (Day et al, 2007;Witney et al, 2009), also work in a human breast tumour xenograft that is much more resistant to treatment with these agents. The hyperpolarised [1-13 C]pyruvate experiment provided early evidence of a DNA damage response, whereas the [1,4-13 C 2 ]fumarate experiment indicated the presence of tumour cell necrosis post-treatment. Both experiments indicated the presence of a treatment response in the absence of any change in tumour size. Translation of this technique to the clinic could offer a new method for monitoring treatment response in breast cancer.