Metabolic imaging using hyperpolarized 13C‐pyruvate to assess sensitivity to the B‐Raf inhibitor vemurafenib in melanoma cells and xenografts

Abstract Nearly all melanoma patients with a BRAF‐activating mutation will develop resistance after an initial clinical benefit from BRAF inhibition (BRAFi). The aim of this work is to evaluate whether metabolic imaging using hyperpolarized (HP) 13C pyruvate can serve as a metabolic marker of early response to BRAFi in melanoma, by exploiting the metabolic effects of BRAFi. Mice bearing human melanoma xenografts were treated with the BRAFi vemurafenib or vehicle. In vivo HP 13C magnetic resonance spectroscopy was performed at baseline and 24 hours after treatment to evaluate changes in pyruvate‐to‐lactate conversion. Oxygen partial pressure was measured via electron paramagnetic resonance oximetry. Ex vivo qRT‐PCR, immunohistochemistry and WB analysis were performed on tumour samples collected at the same time‐points selected for in vivo experiments. Similar approaches were applied to evaluate the effect of BRAFi on sensitive and resistant melanoma cells in vitro, excluding the role of tumour microenvironment. BRAF inhibition induced a significant increase in the HP pyruvate‐to‐lactate conversion in vivo, followed by a reduction of hypoxia. Conversely, the conversion was inhibited in vitro, which was consistent with BRAFi‐mediated impairment of glycolysis. The paradoxical increase of pyruvate‐to‐lactate conversion in vivo suggests that such conversion is highly influenced by the tumour microenvironment.


| 1935
ACCIARDO et Al. mutation: targeted therapies, consisting of BRAFi and MEKi, and immunotherapies. Immunotherapies have displayed long-term effect on a subset of patients but, to date, there are no tools to identify those patients that will benefit from immunotherapies. On the other hand, targeted therapies have immediate effect in term of tumour shrinkage, but the benefit is short-term as resistance occurs after few months in almost every case. As a further matter, some patients do not respond at all to BRAF inhibition because of intrinsic resistance to the treatment. The addition of a MEK inhibitor to the regimen was initially identified as a promising approach to overcome BRAFi-resistance, and hence the combination of BRAFi/MEKi has been approved. [4][5][6] However, in numerous cases the BRAFi/MEKi combination just delays the emergence of resistance as observed when we compare the median progression-free survival of patients receiving the monotherapy (7.3 months) vs the combination (14.9 months). 3 More recently, re-challenging non-responding tumour after a drug holiday period has been suggested as an alternative therapeutic strategy in BRAF-mutated melanomas 8,9 and it has been shown to be beneficial in a subset of patients. [10][11][12][13][14][15] However, safe longitudinal biomarkers to identify the patients who may benefit from the intermittent treatment, or to establish the optimal drug holiday duration, are missing.
Several studies have recently helped deciphering the interplay between oncogenic MAPK signalling, melanoma metabolism and BRAFi-resistance. Apart from a small subset of melanomas, BRAF mutations are associated with increased glycolysis and attenuated oxidative phosphorylation, and such balance is reversed upon treatment with BRAFi. [16][17][18][19][20] Resistance to BRAFi seems to be accompanied by precise metabolic changes as well: higher reliance on oxidative phosphorylation, [21][22][23][24] glutamine dependency 21,25,26 and up-regulated serine metabolism. 26,27 Besides suggesting novel targetable metabolic vulnerabilities, the fast-increasing knowledge in tumour metabolism will hopefully help identifying candidate biomarkers of clinical utility. To date, the lack of validated markers exploiting the aforementioned metabolic alterations to discriminate responding and non-responding melanomas in patients still hinders the long-term effectiveness of current treatment options and represents a barrier to the advancement of personalized medicine.
Metabolic markers such as 13 C magnetic resonance spectroscopy of hyperpolarized substrate may bridge this gap, as they allow to assess crucial metabolic fluxes whose alteration is indicative of treatment response or tumour progression. Notably, hyperpolarization of 13 C-enriched metabolites increases 13 C magnetic resonance spectroscopy (MR) sensitivity by a factor of 10 000, allowing in vivo real-time assessment of metabolic fluxes. In particular, [1-13 C] pyruvate is reduced to [1-13 C] lactate via the enzyme lactate dehydrogenase (LDH). This process results in an altered chemical shift that HP MRI can image at uniquely high-temporal resolution. Hyperpolarized [1-13 C] pyruvate has been safely administered in patients, and its conversion into [1-13 C] lactate was higher in prostate tumours compared with healthy tissue, which was in agreement with previous preclinical studies. 28 The use of hyperpolarized [1-13 C] pyruvate for metabolic imaging of prostate, breast, brain and cervical cancer and other diseases is currently being evaluated in several clinical trials. 29 In addition to pyruvate, several other 13 C enriched substrates have been hyperpolarized, with their conversion being successfully observed in vivo. 29 The aim of this study is to evaluate whether magnetic resonance spectroscopy of hyperpolarized [1-13 C]

| Animal studies
Experiments involving animals were undertaken in accordance with the Belgian law concerning the protection and welfare of the animals and were approved by the UCLouvain ethical committee (agreement reference: UCL/2014/MD/026). All investigators performing in vivo studies successfully completed FELASA C training.
During inoculation, mice were kept under inhalational anaesthesia with 2.5% isoflurane in 2 L/min airflow. Mice were treated with daily intraperitoneal injection of vemurafenib (50 mg/kg, Active Biochem) or vehicle (DMSO, Sigma-Aldrich). Following HP experiments, melanoma xenografts were collected and processed for further analysis (immunohistochemistry, WB, qRT-PCR).

| In vitro EPR spectroscopy
The effect of BRAFi on the oxygen consumption rate was evaluated on A375 and A375R cells via an X-band EPR spectrometer (400 L/h) at 37°C throughout the spectra acquisition. EPR spectra were acquired every 60 seconds, and pO 2 values were obtained by measuring the peak-to-peak EPR signal linewidths, which was then converted into pO 2 by means of a calibration curve. Oxygen consumption rate (OCR) was then calculated as the slope of pO 2 over time curve.

| In vivo EPR spectroscopy
For in vivo EPR experiments, when the shortest tumour diameter reached 5 mm, 50 µL of charcoal suspension (100 mg/mL in 0.9% NaCl containing 3% Arabic gum) was injected intratumorally.
The day after charcoal injection, mice were randomized into two groups (BRAFi-treated or control) and longitudinal EPR measurements were started. Spectra were acquired on a 1.15-GHz EPR spectrometer (ClinEPR). Typical acquisition parameters were as follows: modulation of amplitude 0.4 G, modulation of frequency 21 kHz. During EPR experiments, animals were kept under inhalational anaesthesia with isoflurane (2.5% during anaesthesia induction and 1.2% during maintenance) in 2 L/min airflow. Acquisition was started 5 minutes after setting isoflurane to 1.2%. Tumour pO 2 values were obtained by measuring the peak-to-peak EPR signal linewidth, which was then converted into pO 2 by means of a calibration curve.

| qRT-PCR
Total RNA was extracted from cells and tumour tissue powder   overnight, followed by incubation with anti-rabbit or anti-mouse secondary antibodies (Jackson IR) in tTBS-BSA 1% at room temperature for 1 hour. Detection was performed using the SuperSignal™ West Pico Plus kit (Thermo Scientific) and an ImageQuant LAS 500 camera (GE Healthcare). Quantification was performed on ImageJ by measuring the integral of the optical density profile of the band of the expected molecular weight. No background correction was performed.

| Immunohistochemistry
Melanoma xenografts were fixed in 4% formaldehyde and embedded in paraffin. After rehydration, 5 µm sections were submitted to antigen retrieval using citrate buffer. Sections were then incubated in BSA 5% in TBS/Triton 0,05% to block non-specific binding, then overnight at 4°C with primary antibodies for CD31 (Cell Signaling Technology). Envision anti-rabbit secondary polymer antibody was used (Dako). Stained slides were then digitalized using a SCN400 slide scanner (Leica Biosystems) at 20× magnification and analysed using TissueIA software (Leica Biosystems). The quantification algorithm was run in the non-necrotic part of the tissues.   Figure 1A). 13 C signal arising from alanine was only observed in four mice (33%). Therefore, for the ex vivo analysis, we have compared the two groups at 24 hours after treatment with BRAFi or vehicle.

| Hyperpolarized 13 C MRS detects metabolic changes induced by BRAFi in vivo
In treated xenografts, the glucose transporter GLUT1 was significantly lower both at the mRNA (P = .0010) and protein (P = .0440) level, when compared to control xenografts (Figure 2A-C). Treated xenografts also showed lower mRNA levels of HK2, PDK1 (P = .0037 and P = .0046, respectively) and a significant reduction in c-MYC protein levels (P = .0018) (Figure 2A-C).

| BRAFi impairs glycolysis and oxygen consumption in BRAFi-sensitive, but not in BRAFiresistant, melanoma cells
BRAF inhibition resulted in a decrease of 13 C label exchange between HP pyruvate and lactate in A375 in vitro, but not in A375R cells (P = .0075 and P = .8898, respectively) ( Figure 3A,B). 13 C signal arising from alanine was observed in all but one sample. The label distribution between alanine and lactate was also significantly modified by BRAFi in A375, but not in A375R cells (P = .0333 and P = .8397) ( Figure S1A).
Next, we measured the intracellular and extracellular water-soluble metabolites in melanoma cells incubated with [U-13 C] glucose.
We observed a reduction in extracellular lactate in both sensitive and resistant melanoma cells treated with BRAFi, when compared to their untreated counterpart (P = .0002 and P < .0001, respectively).
Extracellular glucose was significantly higher in BRAFi-treated A375 cells (P < .0001), thus suggesting that treatment affected glucose uptake, as later confirmed via qRT-PCR and Western blot analysis of GLUT1 ( Figure 3C).
In regard to the intracellular metabolites, lactate pool was decreased both in A375 and A375R cells upon inhibition of BRAF (P < .0001 for both A375 and A375R compared with their own controls). Resistant A375R cells; however, had a significantly higher amount of intracellular lactate at baseline (P < .0001) ( Figure 3C).
The glycolytic efficiency, calculated as the ratio of extracellular lactate over consumed glucose, decreased following BRAFi in sensitive A375 cells (P = .0062), but not in A375R cells (P = .4765) ( Figure 3D). In sensitive cells, the intracellular alanine/lactate ratio was significantly modified after treatment as well (P = .0146 for A375 and P = .4765 for A375R) ( Figure S1B).
The decrease in glycolysis was accompanied by an increase in the oxygen consumption rate in sensitive cells treated with the BRAFi, as demonstrated by in vitro EPR spectroscopy ( Figure 3E).

| BRAFi affects the transcription of key glycolytic enzymes and transporters in sensitive melanoma cells
Our qRT-PCR data showed a significant decrease in the mRNA levels of HK2, c-MYC and MCT1 in BRAFi-treated A375 cells, compared with untreated controls (P = .0046, P = .0279 and P = .0076, respectively), but neither of them was affected by the treatment in resistant A375R cells. ALT2 and MPC1 mRNA levels were not modified by treatment neither in A375 nor in A375R cells. However, ALT2 levels were higher at baseline in resistant cells (P = .0490) ( Figure 4A).
F I G U R E 1 HP pyruvate is an early marker of response to BRAF inhibition. A, Representative spectra of the 13

| BRAFi reduces hypoxia in vivo, as assessed by electron paramagnetic resonance (EPR) oximetry
Our in vitro results confirmed that BRAFi reduced the glycolytic activity of sensitive melanoma cells, thus resulting in a decrease 13 C label exchange between pyruvate and lactate. Therefore, we ought F I G U R E 3 BRAFi impairs glycolysis and stimulates oxygen consumption in sensitive melanoma cells. A, 13 C label exchange between HP pyruvate and lactate in live melanoma cells pre-treated with the BRAFi vemurafenib (2 µmol/L, 24 h) or DMSO (two-way ANOVA, Sidak multiple comparisons test, **P < .01, ns: non-significant) (n = 6). B, Representative time course of HP 13 C signal obtained from live A375 (25 × 10 6 ) cells. C, Steadystate concentration of extracellular (left) and intracellular (right) water-soluble metabolites (two-way ANOVA, Sidak multiple comparisons test, **P < .01, ****P < .0001, ns, non-significant) (n = 6). D, For each sample, the ratio between the steady-state level of extracellular lactate and consumed glucose was calculated to assess glycolytic efficiency (two-way ANOVA, Sidak multiple comparisons test, **P < .01, ns: non-significant) (n = 6). E, Oxygen consumption rate measured in live cells (two-way ANOVA, Sidak multiple comparisons test, **P < .01, ns, non-significant) (n = 6). Glc, glucose; Lac, lactate; pyr, pyruvate F I G U R E 4 Molecular markers of response to BRAFi in vitro. A, qRT-PCR analysis of glycolysis-related genes in A375 (black) and A375R (grey) cells following incubation with BRAFi or DMSO (two-way ANOVA, Sidak multiple comparisons test, *P < .05, **P < .01, ***P < .001, ****P < .0001, ns, non-significant) (n = 6). B, Principal component analysis of qRT-PCR data obtained on A375 cells (control: empty circles, BRAFi: black-filled circles) and A375 R cells (control: empty squares, BRAFi: grey-filled squares). C, Western blot analysis of whole cell lysates (right, one representative blot out of three is represented) to understand whether the increased label exchange observed in vivo was due to treatment-induced hypoxia.
EPR oximetry indicated that melanoma xenografts were highly hypoxic (pO 2 < 1.5 mm Hg). BRAFi led to an increase in the oxygen partial pressure (pO 2 ) of treated tumours, as measured by the peakto-peak signal linewidth of the oxygen sensor ( Figure 5A). Such effect occurred after 3 days of treatment, and it was still present after 5 days of treatment (P = .0002 at day 3 and at day 5) ( Figure 5B).
Immunohistochemical analysis of tumour xenografts collected at baseline and after 1 or 5 days of treatment did not show any significant variation in the angiogenesis marker CD31 (Figure 5D,E). These results suggest that the increased pO 2 observed via in vivo EPR experiments may be due to a reduced oxygen consumption in vivo, contrarily to what was observed in vitro. A reduction in the cell density caused by BRAFi could have caused a decrease in oxygen demand and hence, an increase in tumour pO 2 . Another possible explanation is that BRAFi has improved oxygenation without affecting blood vessel density.

| D ISCUSS I ON
Inhibition of BRAF led to a significant increase in the HP pyruvateto-lactate label exchange as soon as 24 hours after treatment administration, before any significant tumour shrinkage.
This effect was paradoxical, as BRAFi is known to inhibit glycolysis in BRAF-mutated melanomas. 17 Within the scope, immunocompetent mice would allow to obtain a more thorough picture of the cross-talk between cancer cells and immune cells. This is all the more important in immunogenic tumours such as melanoma.
Our study points out some aspects that may be taken into account in future research.
The opposite effect of BRAFi in melanoma cells versus xenografts herein observed corroborates recent literature highlighting the importance of animal studies, with the twofold objective of better understanding the links between oncogenic signals and metabolism, while taking into account tumour heterogeneity, and fostering the clinical translation of newly developed hyperpolarized probes. 29,41 Finally, discrepancies between the tumour metabolism of HP probes and steady-state levels of the same metabolites are possible. In fact, contrarily to what happens in physiological conditions, in hyperpolarization experiments the cell metabolism is challenged via the injection of a supra-physiological dose of the hyperpolarized probe. 46 The conversion of a hyperpolarization probe and the measurements of steady-state metabolite levels therefore provide different and complementary information. In our in vitro study, the two approaches provided strikingly similar results, thus suggesting that in this particular model the label exchange between hyperpolarized pyruvate and lactate faithfully described the physiological pyruvate-to-lactate conversion.

CO N FLI C T O F I NTE R E S T
There is no conflict of interest to declare.

AUTH O R S CO NTR I B UTI O N
SA performed the experiments, analysed the data and drafted the manuscript; LM and EL performed experiments; CS and NJ assisted with MRS experiments; FG assisted with qRT-PCR and WB experiments; CB, JFB, and BG made critical revision to manuscript; BFJ supervised the study, contributed to study design and critical revision.
All the authors approved the manuscript for submission.