Hyperpolarized 13C-Pyruvate to Assess Response to Anti-PD1 Immune Checkpoint Inhibition in YUMMER 1.7 Melanoma Xenografts

There is currently no consensus to determine which advanced melanoma patients will benefit from immunotherapy, highlighting the critical need to identify early-response biomarkers to immune checkpoint inhibitors. The aim of this work was to evaluate in vivo metabolic spectroscopy using hyperpolarized (HP) 13C-pyruvate and 13C-glucose to assess early response to anti-PD1 therapy in the YUMMER1.7 syngeneic melanoma model. The xenografts showed a significant tumor growth delay when treated with two cycles of an anti-PD1 antibody compared to an isotype control antibody. 13C-MRS was performed in vivo after the injection of hyperpolarized 13C-pyruvate, at baseline and after one cycle of immunotherapy, to evaluate early dynamic changes in 13C-pyruvate–13C-lactate exchange. Furthermore, ex vivo 13C-MRS metabolic tracing experiments were performed after U-13C-glucose injection following one cycle of immunotherapy. A significant decrease in the ratio of HP 13C-lactate to 13C-pyruvate was observed in vivo in comparison with the isotype control group, while there was a lack of change in the levels of 13C lactate and 13C alanine issued from 13C glucose infusion, following ex vivo assessment on resected tumors. Thus, these results suggest that hyperpolarized 13C-pyruvate could be used to assess early response to immune checkpoint inhibitors in melanoma patients.


Introduction
Melanoma is a skin cancer caused by the uncontrolled growth of abnormal melanocytes. It is known for its poor survival rates in its advanced form and increased incidence rates over the last decades [1]. Indeed, early stages of melanoma are curable after surgery; meanwhile, in metastatic diseases, less than 10% of patients will recover [2]. The introduction of targeted therapies in 2012 led to unprecedented progress in melanoma care; however, the benefits are often transient due to the development of resistance [3]. Interestingly, immunotherapies based on immune checkpoint inhibitors do offer a long-term effect with durable clinical benefits [4,5], yet a significant proportion of patients display intrinsic resistance to this therapy. A response rate of 59% was observed following treatment with combined immune checkpoint inhibitors [6]. Immune checkpoints are receptors expressed on the membrane of activated T cells, and control the activation or inhibition of the immune responses [7,8]. Immune checkpoint blockades use antibodies to inhibit the interaction between receptors expressed at the surface of activated T cells and on cancer cells to unleash immune responses [7][8][9]. T cells play a crucial role in anti-tumor immune defense; they become activated after recognizing tumor antigens and consequently migrate to eliminate cancer cells. This activation is affected by the tumor microenvironment (TME), i.e., the interaction of the stimulatory and inhibitory ligand receptors between T cells, dendritic cells, and macrophages, with the tumor cells acting as immunosuppressors [10,11].
Immunotherapy using immune checkpoint inhibitors (ICIs) constituted a major breakthrough in the treatment of cancer over the past several years [12]. Ipilimumab was the first FDA-approved antibody in melanoma that inhibits signaling via cytotoxic T-lymphocyteassociated proteins (CTLA-4) on T cells. Three years later, an anti-programmed death-1 (anti-PD1) monoclonal antibody, pembrolizumab, was also approved for the treatment of patients with metastatic melanoma [13]. In the same year, an anti-programmed cell death protein 1 (anti-PD1), nivolumab, gained approval [14]. These immunotherapies showed clinical responses that last for several years. However, primary resistance mechanisms are still present [15]. Despite the fact that immunotherapies have displayed long-term effects with durable benefits on a subset of patients, there are no tools which can identify those patients that will benefit from immunotherapies [4,5]. The identification of robust biomarkers remains a challenge in the assessment of drug resistance. Several biomarkers have been studied, including PD-L1 expression, mismatch repair deficiency, and tumor-infiltrating lymphocytes (TILs); however, unfortunately, these have failed to show predictive value for treatment responses in the clinical setting [16].
Recently, a metabolic competition between tumor cells and immune cells has been reported. In 2015, different works reported that the aerobic glycolysis (the Warburg effect) promotes the depletion of extracellular glucose, thereby restricting glucose to T cells and rendering T cells dysfunctional [17,18]. Moreover, it has been shown that the PD1 blockade leads to decreased glycolysis in tumors [17]. Besides metabolic competition, tumor cells create an environment characterized by hypoxia, low pH, and the presence of lactic acid, which all play an immune inhibitory role [19,20]. In this context, it is relevant to assess treatment responses to cancer immunotherapy by evaluating potential shifts in the tumor metabolism.
Among several non-invasive molecular imaging techniques currently used in the preclinical and clinical settings, 18 F-fluoro-deoxy-glucose-positron emission tomography (FDG-PET) was suggested to be useful as a marker of melanoma response [21]. Early FDG-PET scans at day 15 in patients treated with targeted therapy (dabrafenib) on phase I trials demonstrated early, complete, or partial metabolic responses, potentially with prognostic significance [22,23]. In the scope of immunotherapy, few studies assessed 18 F-FDG PET/CT as an early predictor of responses to PD-1 or CTLA-4 blockades in patients with non-smallcell lung carcinoma or melanoma [24,25]. In particular, 18 F-FDG PET/CT performed after two cycles of ipilimumab was shown to be highly predictive of the final treatment outcome in patients with progressive and stable metabolic disease in metastatic melanoma [26]. However, the distinction of patients between pseudoprogression and progressive disease has been shown in several consecutive studies to be limited using 18 F-FDG PET/CT [27,28]. More recently, one clinical study highlighted that early scans performed from three to four weeks after the start of ICI treatment could be predictive of the response in patients with advanced melanoma when functional and anatomic imaging parameters were combined, further stressing the need to conclusively develop larger cohorts [24].
There is, therefore, currently a lack of robust noninvasive imaging markers to evaluate early responses to immunotherapy in the clinical setting. The purpose of this study was to investigate the ability of non-invasive metabolic imaging to evaluate responses to immune checkpoint blockades with anti-PD1 using 13 C-MRS (magnetic resonance spectroscopy) metabolic imaging. 3 of 15 13 C-metabolic imaging is a cutting-edge technology that is now considered to assess sensitivity to therapy in the preclinical setting as well as in clinical studies, with a higher specificity than 18 F-FDG, by monitoring the fate of glucose beyond glucose uptake via tumor cells [29,30]. The monitoring of 13 C-enriched metabolic substrates can be monitored using tools such as 13 C magnetic resonance spectroscopy (MRS) or imaging (MRI). MRS can investigate static metabolic processes in vivo after injecting a 13 C-enriched substrate (i.e., 13 C-glucose or 13 C-glutamine) [31]. Moreover, dynamic metabolic processes can be assessed by temporarily boosting the 13 C NMR signal of some key metabolic substrates (such as 13 C-pyruvate) using hyperpolarization, thereby allowing for the dynamic measurement of real-time metabolic conversions in vivo [32]. Notably, dynamic nuclear polarization (DNP) can increase 13 C-MRS sensitivity 10,000-fold, providing an opportunity to detect real-time metabolic fluxes in vivo, such as 13 C-pyruvate-13 C-lactate exchange, in preclinical studies as well as in patients [33][34][35][36].
In this study, we characterized the 13 C-metabolic profile in response to the PD1 blockade in YUMMER1.7 syngeneic melanoma xenografts characterized for BrafV600E/wt Pten−/−Cdkn2−/−somatic mutations [37], allowing for the use of immunocompetent mice to fully integrate all aspects of the tumor microenvironment. We aimed to evaluate the relevance of hyperpolarized (HP) 13 C-pyruvate and 13 C-MRS fluxomic after [U-13 C]glucose injection as potential response markers to immunotherapy in the preclinical setting.

Anti-PD1 Monoclonal Antibody Delays Tumor Growth in Syngeneic YUMMER1.7 Melanoma Xenografts
An anti-PD1 monoclonal antibody was used to evaluate the therapeutic effects of the immune checkpoint blockade in vivo. C57BL/6 mice were subcutaneously inoculated with YUMMER1.7 (1.5 × 10 6 cells) into the right leg and treated with isotype or anti-PD1 antibodies via intraperitoneal injection 3 times per week for 2 weeks (2 cycles of immunotherapy) ( Figure 1A).
Anti-PD1 significantly delayed (p = 0.04) the progression of YUMMER1.7 melanoma xenografts compared to the isotype group, with a significant difference between groups upon day 9 after treatment initiation, as shown in Figure 1B,C. The growth delay was potentially induced by a combined effect of anti-PD1 treatment on tumor cell death and tumor cell proliferation, as shown by the trends observed with IHC of KI-67 and caspase-3, respectively, assessed after two cycles of immunotherapy ( Figure 1D,E). At the same timepoint, the infiltration of T cells and macrophages showed an increasing trend in the anti-PD1 group at the edge of the tumors ( Figure 1F). Of note, there was a lack of change in T cell infiltration and in the number of macrophages in the central regions of the tumors.

Hyperpolarized 13 C-Pyruvate Detects Real-Time Metabolic Changes Induced by Anti-PD1 In Vivo
In order to determine metabolic differences between control and treated mice, we performed intravenous injections of HP [1-13 C] pyruvate to mice bearing YUMMER1.7 xenografts to monitor, in vivo, the real-time exchange between [1-13 C] pyruvate and [1-13 C] lactate using in vivo 13 C-MRS (Bruker Biospec, 11.7T). Figure 2B shows representative evolutions of 13 C pyruvate and 13 C lactate peaks over time, from which the area under the curve and ratios were calculated. Anti-PD1-treated tumors showed a significant decrease in the 13 C lactate over the total 13 C signal (LA/ 13 C) ratio at day 5 compared to day 0, whereas the isotype control tumors did not show any significant change over time ( Figure 2C,D). The anti-PD1-treated tumors also displayed a significant increase in the 13 C pyruvate over total 13 C signals (PA/ 13 C) ( Figure 2E,F). Consequently, a significant decrease was observed in the evolution of the ratio of hyperpolarized (HP) 13 C lactate to 13 C pyruvate (LA/PA) after treatment with anti-PD1 ( Figure 2G). These data point out that HP 13 C-pyruvate administered after one cycle of anti-PD1 treatment was able to detect metabolic changes before any significant change in tumor size could be observed.

Hyperpolarized 13 C-Pyruvate Detects Real-Time Metabolic Changes Induced by Anti-PD1 In Vivo
In order to determine metabolic differences between control and treated mice, we performed intravenous injections of HP [1-13 C] pyruvate to mice bearing YUMMER1.7 xenografts to monitor, in vivo, the real-time exchange between [1-13 C] pyruvate and [1- 2C,D). The anti-PD1-treated tumors also displayed a significant increase in the 13 C pyruvate over total 13 C signals (PA/ 13 C) ( Figure 2E,F). Consequently, a significant decrease was observed in the evolution of the ratio of hyperpolarized (HP) 13 C lactate to 13 C pyruvate (LA/PA) after treatment with anti-PD1 ( Figure 2G). These data point out that HP 13 C-pyruvate administered after one cycle of anti-PD1 treatment was able to detect metabolic changes before any significant change in tumor size could be observed.  In vivo monitoring of hyperpolarized 13 C pyruvate-lactate exchange in response to the PD1 blockade. Representative spectra of the 13 C signal time course obtained from a mouse at baseline (A). Representative evolution of the area under the curve of 13 C pyruvate and 13 C lactate peaks over time with isotype and anti-PD1 treatment (B). Individual changes in calculated 13 C-lactate/total 13 C-signals in treated PD1 and isotypes, respectively (C,D). Individual changes in calculated 13 Cpyruvate/total 13 C signals in treated PD1 and isotypes, respectively (E,F). Evolution of exchange between HP pyruvate and lactate (measured as the ratio of the AUC of [1-13 C] lactate/the AUC of [ 1-13 C] pyruvate) in melanoma xenografts after 1 cycle of immunotherapy (G). t-test n = 4/group. *: p ≤ 0.05; *** p ≤ 0.001. The dots correspond to the data points.

Ex Vivo Metabolic Profiling Using 13 C-Glucose Tracing Is Not Modified in Response to Anti-PD1 Treatment
To further characterize the glucose metabolism in the YUMMER1.7 model in response to anti-PD1 and isotype antibodies, we conducted fluxomic experiments after injecting uniformly-labeled U-13 C glucose at day 6 on the same cohort. The downstream metabolites were detected after tumor resection (75 min after U-13 C glucose infusion) and metabolite extraction. Typical spectra, acquired using high-resolution 13 C-NMR (Bruker Ascend, 600 MHz), are shown in Figure 3A

Expression of Metabolic Transporters and LDH-A in Response to Anti-PD1
Since we observed a lower HP 13 C lactate to 13 C pyruvate ratio in response to anti-PD1 treatment, and since this ratio has been shown to be dependent on the LDH-A activity, as well as on the activity of the monocarboxylate transporters MCT-1 and MCT-4, and the glucose transporter GLUT-1, we assessed their expression via IHC after one cycle of anti-PD1 treatment to match the timing of the metabolic imaging experiments. A lack of change in GLUT-1 staining was observed in response to anti-PD1 treatment ( Figure 4A), yet GLUT-4 was not assessed in this study. Further studies would be needed to better characterize the effect on glucose uptake. Similarly, the monocarboxylate transporters MCT-1 and MCT4 were not modified in response to anti-PD1 ( Figure 4B,C). However, we We did not observe any significant change in the steady-state levels of 13 C-MRS ex vivo detectable metabolites in YUMMER1.7 xenografts, i.e., 13 C-lactate, 13 C-alanine, and 13 C-glutamate, between the isotype and anti-PD1-treated tumors ( Figure 3C).

Expression of Metabolic Transporters and LDH-A in Response to Anti-PD1
Since we observed a lower HP 13 C lactate to 13 C pyruvate ratio in response to anti-PD1 treatment, and since this ratio has been shown to be dependent on the LDH-A activity, as well as on the activity of the monocarboxylate transporters MCT-1 and MCT-4, and the glucose transporter GLUT-1, we assessed their expression via IHC after one cycle of anti-PD1 treatment to match the timing of the metabolic imaging experiments. A lack of change in GLUT-1 staining was observed in response to anti-PD1 treatment ( Figure 4A), yet GLUT-4 was not assessed in this study. Further studies would be needed to better characterize the effect on glucose uptake. Similarly, the monocarboxylate transporters MCT-1 and MCT4 were not modified in response to anti-PD1 ( Figure 4B,C). However, we observed a decreasing trend for LDH-A staining (p = 0.06, n = 4), showing a potential role in the ratio of decreased lactate to pyruvate ( Figure 4D). Further, the flux through a pathway is not dictated by expression levels of the enzymes but rather by their activity, emphasizing the need for in vivo real-time dynamic flux measurements. In addition, the lack of change in these proteins at early time points excludes them as early markers of response to anti-PD1 therapy.

Discussion
Biomarkers are needed to identify patients that resist early treatment in the course of their therapy, allowing them to be moved to potential alternative therapies and/or to avoid any unnecessary side effects. Therefore, identifying patients who will respond to immunotherapy constitutes an important unmet clinical need. In particular, there is currently no consensus to determine which advanced melanoma patient will benefit from

Discussion
Biomarkers are needed to identify patients that resist early treatment in the course of their therapy, allowing them to be moved to potential alternative therapies and/or to avoid any unnecessary side effects. Therefore, identifying patients who will respond to immunotherapy constitutes an important unmet clinical need. In particular, there is currently no consensus to determine which advanced melanoma patient will benefit from immune checkpoint blockers such as PD-L1 expression, mismatch repair deficiency, and tumor infiltrating lymphocytes (TILs), which all failed to be predictors in the clinical setting. Therefore, our aim was to evaluate the relevance of hyperpolarized (HP) 13 Cpyruvate and 13 C-MRS fluxomic after U-13 C-glucose injection as early-response markers to anti-PD1 therapy in syngeneic melanoma xenografts. The YUMMER1.7 melanoma cell line was previously shown to be sensitive to the immune checkpoint blockade, and is, therefore, relevant to the evaluation of the immune checkpoint blockade (ICB) and anti-tumor responses [38]. In the syngeneic YUMMER1.7 model, we confirmed that the PD1 blockade, at a dose of 10 mg/kg, significantly delayed tumor growth in comparison with the isotype control group on day 9 following treatment initiation, potentially induced by the combined effect of an increasing trend in cell death, as assessed using the apoptotic marker caspase-3, and a trend to a decrease in tumor cell proliferation, as shown by the proliferation marker Ki-67, although still with a lack of statistical significance for the sample sizes under study (n = 5-7/group). Nevertheless, our data are in accordance with recent studies showing similar effects in YUMMER1.7 melanomas [39]. In our study, the immunofluorescence analysis of immune cell populations assessed after two cycles of immunotherapy showed an increasing trend in the T cell population at the edge of the tumor and in the macrophage populations in the same area. However, the small sample size in our metabolic imaging study, along with a large variability in the distribution of IHC data, did not allow for significance to be reached. Of note, the immune cell populations were not affected yet after one cycle of immunotherapy. Nevertheless, this observation correlates with a previous study where it has been shown that YUMMER1.7 tumors are infiltrated with CD4 and CD8 T cells [37], and that the infiltration of T cells and macrophages is significant on day nine after implantation in regressing tumors in response to immunotherapy [38].
Although 18 F-FDG PET is a standard clinical metabolic imaging approach for the diagnosis of numerous cancer types, it shows some limitations in the scope of therapy monitoring. In particular, 18 F-FDG PET has not yet been validated thus far as a robust response marker of the response of advanced melanoma patients to immune checkpoint blockers. This might be due to the fact that 18 FDG PET only provides information about the glucose uptake and cannot assess downstream metabolites that are important to understanding cancer metabolism, contrarily to 13 C-glucose tracing and hyperpolarized 13 C pyruvate MRS methods. In this context, DNP was recently translated into the clinical setting to assess tumor metabolism in human tumors, with a first study of metabolic imaging using HP 13 C pyruvate reported in prostate cancer patients in 2013 [39], followed by numerous (ongoing) clinical trials. For instance, Brindle et al. showed that the early response to PI3K inhibition could be detected using metabolic imaging with hyperpolarized 13 C pyruvate, and the assessment of treatment response in breast cancer patients could help to identify patients that benefit from PI3Ka inhibition and design drug combinations to counteract resistance [34]. In addition, it has been shown that hyperpolarized 13 C pyruvate allows for response assessment in breast cancer patients after seven days of chemotherapy, demonstrating prognostic potential [36]. A recent work also illustrated that HP 13 C MRI could differentiate tumor aggressiveness in renal cell carcinoma [35]. Of note, HP 13C-pyruvate measurements are technically challenging in the clinical setting since the hyperpolarization decreases quite fast and MRS acquisitions have to be performed within few minutes [33].
Immunotherapy has shown a major breakthrough in the treatment of melanoma. Immune checkpoint inhibitors stimulate the adaptive immune system in order to produce anti-tumor responses [40]. Glucose metabolism regulates T-cell activation and functions. For instance, the Warburg phenotype or aerobic glycolysis used to describe the metabolic shift in cancer cells is a key process to sustain the activation and differentiation of T cells. In addition, lactate is described to have an immunosuppressive impact in the TME [41,42]. Therefore, hyperpolarized 13 C-pyruvate, a probe used to assess the glycolytic shift in tumors by probing the ratio of 13 C-lactate to 13 C-pyruvate in vivo, is a relevant biomarker candidate to indicate early response to immunotherapy. This ratio can be influenced by the direct conversion of pyruvate to lactate by the lactate dehydrogenase (LDH-A), which is often upregulated in tumors; however, it is also regulated via the expression and activity of monocarboxylate transporters MCT-1 and MCT-4, depending on the tumor model under study [43,44]. In this study, we assessed HP 13 C-lactate to 13 C-pyruvate exchange in YUMMER 1.7 melanoma xenografts. Interestingly, a significant decrease in the ratio of HP 13 C-lactate to 13 C-pyruvate was observed in vivo after one cycle of immunotherapy, which was not observed in the isotype control group. A decrease in the lactate-to-pyruvate ratio is likely due to a decrease in the LDH-A expression in the YUMMER 1.7 melanoma tumors in response to anti-PD1 treatment, although the activity of the enzyme has not directly been studied here. MCT-1, which has been described to rate-limit the conversion of HP 13 C-pyruvate to 13 C-lactate in some models [45], was not modified in response to anti-PD1 in the YUMMER1.7 melanoma xenografts. Importantly, the change in the lactateto-pyruvate ratio preceded any significant change in tumor volume in the anti-PD1 and isotype groups that were significantly different after two cycles of immune therapy. These data illustrate the potential of non-invasive real-time monitoring of the metabolic flux to assess responses before any classic (RECIST) clinical criteria are affected. Further studies using models which are resistant to immunotherapy are required to validate the marker in responsive and non-responsive tumors.
In addition, the observed changes in the metabolic flux reflect a potential crucial role of the tumor microenvironment and the metabolic interactions between tumor cells and immune cells in the sensitivity of tumors to immune checkpoint inhibitors. For instance, it has been shown that tumors can diminish the functions of tumor-infiltrating lymphocytes and that this metabolic competition can lead to T cell hypo-responsiveness. The PD1 expression on T cells inhibits glycolysis and upregulates fatty acid oxidation, leading to impaired energy generation and decreased effector functions [17,46].
Interestingly, the lack of change in metabolite concentrations assessed ex vivo by 13 C-MRS fluxomic tracing experiments after U-13 C-glucose injection illustrates that steady-state metabolite concentrations do not necessarily reflect the activity of a metabolic pathway, contrarily to real-time metabolic measurements. Moreover, differences in the pool of metabolites resulting from a dynamic change in pyruvate lactate exchange can be too subtle to result in a significant change in the steady-state levels of metabolites. Finally, for technical purposes, the timing was slightly different between HP 13 C-pyruvate and 13 Cglucose fluxomic experiments (day six and day seven); still, both data were acquired after one cycle of immunotherapy. Overall, our data illustrate that the dynamic monitoring of metabolic fluxes using HP 13 C-pyruvate can present a significant added value to fluxomic experiments [45,47].
Taken together, these results suggest that metabolic imaging techniques could predict responses to immune checkpoint blockades in the clinical setting. In particular, HP 13 Cpyruvate could be assessed as an early-response maker in melanoma patients treated with immunotherapy.

Mice and Tumors
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/2018/MD/021). Six-week-old C57BL-6 male mice were supplied by Janvier Labs and housed in specific pathogen-free (SPF) environment under standard conditions of temperatures around 20-24 • C and humidity between 45 and 65%. All investigators performing in vivo studies successfully completed FELASA C training.

In Vivo Checkpoint Blockade Treatment
For the in vivo treatment model, 1.5 × 10 6 YUMMER1.7 cells in 50 µL of PBS were subcutaneously inoculated in the right hind paw of 6-week-old C57BL-6 mice. Cells were harvested by trypsinization and resuspended in PBS (pH 7.4) before animals were injected.
During inoculation, mice were kept under inhalational anesthesia with isoflurane 2% in 2 L/min airflow.

Tissue Fixation, Freezing and IHC Stainings
After sacrifice, the individual tumor from each study animal was collected and divided in half. One half of the tumor was immediately snap-frozen in liquid nitrogen and stored at −80 • C until protein extraction. The remaining portion was immediately fixed in 4% paraformaldehyde for 24 h at room temperature. Samples were subsequently transferred into an automated tissue processor and embedded in paraffin. Following deparaffinization, the inactivation of endogenous peroxidases, antigen retrieval in citrate or Tris-EDTA buffer, and non-specific binding blocking, 5 µm sections were incubated overnight at 4 • C with the primary antibodies for MCT1 (Proteintech, Rosemont, IL, USA, #20139-1-AP, 1:1000 dilution) and MCT4 (Sigma, Saint Louis, MO, USA, #HPA021451, 1:500 dilution), LDHA (cell signaling, Danvers, MA, USA, #3558, 1/100 dilution), Ki-67 (cell signaling, Danvers, MA, USA, #12202, 1/300 dilution), and caspase-3 (cell signaling, Danvers, MA, USA, #9661, 1/300 dilution). Consequently, sections were incubated at room temperature for 30 min with the Envision anti-rabbit secondary antibody (Dako Agilent, Santa Clara, CA, #K4003) and stained with diaminobenzidine for 5 min (Dako Agilent, Santa Clara, CA, #K3468). Stained slides were then digitalized using a Pannoramic SCANII slide scanner (3DHistech) at X20 magnification and analyzed using Visiopharm software. The quantification algorithm was run in the viable part of the tissue samples to detect the stained area and the analyzed tumor area. A % of the stained area was calculated as the ratio between the stained area and the analyzed tumor area multiplied by 100.

Computer-Assisted Quantitative Evaluation of Immunostaining in Whole Tissue Sections
Whole multiplex-stained paraffin sections were quantified using the image analysis tool Author (version 2022.01.4) (Visiopharm, Hørsholm, Denmark). Tissue sections were first automatically delineated at a low digital magnification (×5) using a thresholding classification method. Blur areas, damaged tissue, and artefacts were manually excluded. To evaluate the stained area, pixels stained for each marker were detected at high resolution (×20) using a thresholding classification method. Results were expressed as a percentage of the stained area within the analyzed periphery region.

Hyperpolarized 13 C-MRS
Imaging experiments were performed at day 0 (pre-treatment) and day 5. For hyperpolarization experiments, 40 µL of [1-13 C] pyruvic acid (Sigma-Aldrich, Saint Louis, MO, USA) solution, containing 15 mmol/L of trityl radical OXO63 and 2 mmol/L of gadolinium, were hyperpolarized in an Oxford Dynamic Nuclear Polarizer (HyperSense, Oxford, UK) for approximately 45 min at 1.4 K and 3.35 T. The polarized solution was rapidly dissolved in 3 mL of heated buffer containing 100 mg/L of EDTA, 40 mmol/L of HEPES, 30 mmol/L of NaCl, 80 mmol/L of NaOH, and 30 mmol/L of non-HP unlabeled lactate. This solution was quickly injected using a catheter into the tail vein of the mice in the MRI scanner (11.7-Tesla, Bruker, Biospec, NEST Platform, UCLouvain). Mice were scanned using a double-tuned 1 H-13 C-surface coil (RAPID Biomedical, Rimpar, Germany), as previously described [49], which was designed for the spectroscopy of subcutaneous tumors [49,50]. 13 C spectra acquisition and HP [1-13 C] pyruvate infusion processes were started simultaneously. During MR experiments, animals were kept under inhalational anesthesia with isoflurane (2.5% during anesthesia induction, 1-2% during maintenance) in a 2 L/min airflow. The temperature was continuously monitored and kept at 37 ± 1 • C via a warmed water blanket. Anatomic T 2 -weighted images were used to assess the sequence of tumor volume to turbo RARE. After the administration of ±0.2 mL of hyperpolarized pyruvate, 13 C spectra were acquired using a single-pulse sequence every 3 s for 210 s (70 repetitions), a flip angle of 10 • , and an acquisition bandwidth of 50 kHz (10,000 points).
The 13 C exchange between HP [1-13 C] pyruvate and [1-13 C] lactate was measured as the ratio between the corresponding areas under the curve (AUCs) via a homebuilt Matlab routine (The MathWorks Inc., Portola Valley, CA, USA).

U-13 Cglucose Administration
Mice were fasted 6 h before the experiment. Blood glucose levels were measured 15 min before, just prior to intraperitoneal U-13 C-glucose injection (2 mg/g of mouse) (time point 0), and then every 15 min until the sacrifice at the 75th min, whereby the tumor was sampled, according to the protocol published by Yuan and colleagues [51]. Glycemia was measured with a glucose meter (Accu Check, Roche, Switzerland) on blood samples collected from the tail vein. 13 C-glucose was injected intraperitoneally at day 6. Tumors were resected 75 min after U-13 C-glucose administration.

Metabolite Extraction and 13 C-MRS
Tumor biopsies were grinded using a pestle at cryogenic temperature. Following grinding, the fragmented pieces were transferred into a tube and stored at −80 • for further analysis ( 13 C-MRS experiments).
Polar metabolites were extracted from snap-frozen tumor tissue, as previously described [51,52]. Around 250 mg of tissue was homogenized in ice-cold methanol (4 µL/mg) using a mechanical lyser (TissueLyser II, Qiagen, Hilden, Germany). The homogenate was then centrifuged at 14,000 g at 4 • C for 20 min. The supernatant was collected and centrifuged again with the same parameters, and any debris was eliminated. After the second centrifugation, the extract for each sample was collected in glass NMR tubes. The solvent was completely removed using a vacuum concentrator. The sample was reconstituted in a 600 µL sodium phosphate buffer with 10% deuterium oxide containing 0.75 wt% 3-(trimethylsilyl) propionic-2,2,3,3-d4 acid (TSP) (Sigma-Aldrich, Saint Louis, MO, USA). Then, 13 C NMR spectra were acquired on a 600 MHz NMR (Bruker, Biospec, Karlsruhe, Germany), equipped with a broadband cryoprobe, as described previously [53]. The acquisition time was 0.8 s with 2048 repetitions, and an interpulse delay of 10 s (1D sequence with inverse-gated decoupling using a 30 • flip angle). Spectrum analysis, assignment, and quantification were performed with MestReNova software (version 14.2.0-26256) (Santiago de Compostela, Spain). Metabolites were quantified by peak integration relative to internal standards and were corrected for tumor mass per sample.

Statistical Analysis
Unpaired t-test analysis was performed via GraphPad Prism 9.1.2 (software), and p < 0.05 was considered significant. The results are represented as means ± SEM.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.