Purpose:

Preclinical studies show that antiangiogenic therapy exacerbates tumor glycolysis and activates liver kinase B1/AMP kinase (AMPK), a pathway involved in the regulation of tumor metabolism. We investigated whether certain metabolism-related in situ biomarkers could predict benefit to regorafenib in the phase II randomized REGOMA trial.

Patients and Methods:

IHC and digital pathology analysis were used to investigate the expression in glioblastoma (GBM) sections of monocarboxylate transporter 1 and 4 (MCT1 and MCT4), associated with OXPHOS and glycolysis, respectively, phosphorylated AMPK (pAMPK), and phosphorylated acetyl-CoA carboxylase (pACC), a canonical target of AMPK activity. The status of each biomarker was associated with clinical endpoints, including overall survival (OS) and progression-free survival (PFS) in patients with relapsed GBM treated either with regorafenib or lomustine.

Results:

Between November 2015 and February 2017, 119 patients were enrolled (n = 59 regorafenib and n = 60 lomustine) and stratified for surgery at recurrence, and baseline characteristics were balanced. Biomarker analysis was performed in 84 patients (71%), including 42 patients of the regorafenib arm and 42 patients of the lomustine arm. Among all markers analyzed, only pACC showed predictive value in terms of OS. In fact, median OS was 9.3 months [95% confidence interval (CI), 5.6–13.2] for regorafenib and 5.5 months (95% CI, 4.2–6.6) for lomustine for pACC-positive patients, HR, 0.37 (95% CI, 0.20–0.70); log rank P = 0.0013; test for interaction = 0.0453. No statistically significant difference was demonstrated for PFS according to pACC status.

Conclusions:

We found that AMPK pathway activation is associated with clinical benefit from treatment with regorafenib in relapsed GBM.

Translational Relevance

Regorafenib has recently shown promising therapeutic activity in relapsed glioblastoma (GBM). However, predictive biomarkers of response to antiangiogenic drugs are not available. On the basis of preclinical studies showing a role of the AMP kinase (AMPK) pathway in modulating tumor response to anti-VEGF drugs, we report here that tumor expression of phosphorylated acetyl-CoA carboxylase, a marker of AMPK activation, correlates with improved survival in patients with GBM treated with regorafenib. This metabolism-associated biomarker can be useful to identify patients with GBM who will best benefit from regorafenib in future studies.

Regorafenib is a small molecule that targets angiogenic kinases [VEGFRs (VEGFR1–3) and TIE2], stromal kinases [platelet-derived growth factor receptor (PDGFR) and FGFR], and cancer cell–associated kinases (KIT, RET, RAF1, and BRAF; refs. 1–3). Regorafenib is approved for the treatment of several advanced or metastatic tumors (4–6). Preclinical studies indicate therapeutic activity of regorafenib in glioblastoma (GBM) models, associated with decreased tumor vascularization and inhibition of PDGFR pathways (7), as well as direct cytotoxic effects on human GBM cells, which were further increased by its combination with the ERBB1/ERBB2 inhibitor, lapatinib (8). On the basis of these characteristics and preclinical results, we recently concluded a randomized, comparative, multicenter phase II trial (REGOMA) designed to investigate for the first time the role of regorafenib in the treatment of patients with recurrent GBM (9). Results of the REGOMA trial showed encouraging activity of regorafenib in patients with relapsed GBM, indicating that this drug might be a new potential treatment for these patients. A translational research program was associated with the REGOMA trial and included evaluation of several predefined in situ biomarkers in tumor sections. Such biomarkers included some metabolism-associated biomarkers, which were chosen based on preclinical studies by our and other groups and indicated that the glycolytic activity of tumors or the activity of the liver kinase B1 (LKB1)/AMP kinase (AMPK) pathway could modulate response to antiangiogenic drugs (10, 11). Although there are several methods to investigate tumor metabolism, expression of certain transporters associated with pyruvate/lactate transport, such as MCT1 and MCT4, is considered a reliable surrogate of the OXPHOS/glycolytic activity of tumor cells, which can be investigated in formalin-fixed, paraffin-embedded (FFPE) sections (12–14). With regard to LKB1/AMPK, activation of this pathway has been observed in experimental models following treatment of tumors with the antiangiogenic drug, bevacizumab (10, 15). AMPK activation depends on LKB1 and is interrogated by measuring the expression of the phosphorylated Thr172 residue or, alternatively, by measuring downstream phosphorylation of the Ser79 epitope of acetyl-CoA carboxylase (ACC), one of the best known targets of AMPK (16). LKB1/AMPK activation can be increased by antiangiogenic treatment but a certain level of basal activation is common in tumors, due to the low oxygen and ATP levels present in central, poorly vascularized tumor regions. Therefore, interrogating AMPK or ACC phosphorylation in tumors ahead of treatment can indicate whether the pathway is active in a given tumor sample. Moreover, recent studies uncovered an oncogenic role of AMPK in brain tumors. Briefly, oncogenic stress putatively accounts for chronic AMPK activation in GBM, leading to cooption of the CREB1 pathway to coordinate tumor bioenergetics through the transcription factors HIF1α and GABPA (17). Therefore, tumor samples showing phosphorylated AMPK or ACC expression might underscore HIF1α expression and an angiogenic profile. On the basis of these theoretical and experimental considerations, we report here the correlation between expression of these biomarkers and outcome to regorafenib in the REGOMA trial.

Samples and IHC staining

This translational study was approved by the institutional review board and was conducted in accordance with the Declaration of Helsinki and Good Clinical Practice guidelines. All patients had to sign an informed consent form approved by the ethical committee of the enrolling institution according to national regulations. REMARK guidelines were followed. IHC was conducted in 5-μm-thick FFPE tumor sections at first surgery in all cases but four, which were obtained at relapse, dewaxed in a specific solution (Bond Dewax Solution, Leica Microsystems), and rehydrated in progressively less concentrated alcohol solutions. To investigate expression of MCT1, MCT4, phosphorylated ACC (pACC), and phosphorylated AMPK (pAMPK), slides were incubated with rabbit anti-MCT1 polyclonal antibody (1:1,500, ab85021, Abcam), rabbit anti-MCT4 polyclonal antibody (1:300, H-90, sc-50329, Santa Cruz Biotechnology), rabbit anti-pACC polyclonal antibody (1:100, Ser79, Cell Signaling Technology), and rabbit anti-pAMPK monoclonal antibody (1:100, Thr172, Cell Signaling Technology). Microvessels were stained by using the anti-human CD31 mouse monoclonal antibody (1:100, JC/70A, Thermo Fisher Scientific). Citric acid buffer (pH 6.0, 10 mmol/L) was used for antigen retrieval in all cases. IHC was performed using a Leica Autostainer and the Bond Polymer Refine Detection Kit (Leica Microsystems). Immunostaining was visualized following substrate chromogen incubation: 3,3-diaminobenzidine tetrahydrochloride hydrate for 10 minutes, followed by hematoxylin counterstaining (5 minutes). Positive control tissue samples were used as recommended by the manufacturer of the primary antibodies.

Image acquisition and analysis

Tumor representation and quality of staining were initially evaluated by an experienced pathologist (G. Esposito) and a biologist (M. Verza) blinded to the results of the REGOMA trial. Slides were digitally acquired at 10× magnification by the Aperio CS2 (Leica Biosystems) and the evaluation of IHC score was assessed through the Scanscope Image Analysis Software (ImageScope v12.4.0.708). On the basis on their localization, the different markers were analyzed by using the Aperio membrane algorithm v9 (MCT1 and MCT4), the Aperio cytoplasmic algorithm v2 (pACC and pAMPK), the Aperio nuclear algorithm (pACC), and the microvessel analysis v1 (CD31), with slight refinements. Microvessel density (MVD) was calculated for the entire tumor section. Aperio Genie Classifier was trained to recognize tumor tissue, stroma, and background (glass), and then combined with Aperio Membrane v9 and Aperio Cytoplasmic v9. Results provided the percentage of cells with different expression of proteins classified as 3+ (highly positive), 2+ (intermediate positive), 1+ (low positive), and 0 (negative). The sum of percentage of marker-positive cells from these four tiers equals 100%. For each sample, digital quantification performed by the software was confirmed by the pathologist (G. Esposito). For statistical analysis, for each sample, we grouped the percentage of cells with 0/1+ and 2+/3+ values. Samples were scored positive when >5% of tumor cells in the section were positive for the marker analyzed.

Statistical analysis

Details of the REGOMA trial have been described previously (9). Analysis of the expression of metabolic biomarkers as potential prognostic or predictor factors was an exploratory endpoint identified a priori in the REGOMA protocol. Sample size was not calculated as it was a secondary endpoint of the study.

Continuous and ordinal variables were summarized as median and interquartile range (IQR), categorical variables were reported as counts and percentages. Association between variables was assessed with χ2 or Fisher exact test, as appropriate. Pairwise associations between biomarkers were assessed by means of Cramer V ranging from 0 (no association) to 1 (variables completely concordant). Correlation of pACC expression among primary and relapse samples was assessed by means of t test for paired data.

Outcome was analyzed in terms of progression-free survival (PFS) and overall survival (OS). OS was defined as the time from randomization until death from any cause; PFS was defined as the time from randomization to tumor progression according to RANO criteria or death. Patients without progression or death were censored at their final follow-up visit. The survival probability was computed using the Kaplan–Meier method and compared between levels (negative vs positive) of biomarkers using the log-rank test.

To assess the possible predictive role of the different considered biomarkers as modifier of treatment effect to regorafenib, Cox models were estimated including the interaction between treatment and each biomarker as covariate and PFS and OS as dependent variables, respectively.

P values less than 0.05 were considered statistically significant.

Statistical analyses were performed using the SAS Statistical Package (SAS, release 9.4; SAS Institute Inc.).

The REGOMA trial is registered with the EU Clinical Trials Register database, number 2014-003722-41, and with ClinicalTrials.gov, number NCT02926222.

Patient characteristics

Clinical data of the 119 patients enrolled in the REGOMA trial have been published elsewhere (9). Protein biomarker analysis was performed in 84 patients (70.6%), including 42 and 42 patients in the regorafenib and lomustine arms, respectively. Baseline characteristics of patients who had at least one biomarker analyzed are shown in Table 1.

Table 1.

Baseline characteristics at randomization in patients with at least one biomarker analyzed.

RegorafenibLomustine
(n = 42)(n = 42)
Age (years) 52.9 (46.2–60.9) 59.2 (51.8–64.4) 
Sex 
 Men 27 (64.3%) 29 (69.1%) 
 Women 15 (35.7%) 13 (30.9%) 
ECOG performance status 
 0 18 (42.8%) 18 (42.8%) 
 1 24 (57.2%) 24 (57.1%) 
Surgery at the time of recurrence 7 (16.7%) 10 (23.8%) 
Corticosteroid use 23 (54.8%) 28 (66.7%) 
MGMT status, n (% of tested) 
 Methylated 21 (50.0%) 19 (46.3%) 
 Unmethylated 21 (48.8%) 22 (53.7%) 
 Not done/unknown — 
IDH status, n (% of tested) 
 Unmutated 28 (93.3%) 28 (100%) 
 Mutated 2 (6.7%) — 
 Not done/unknown 12 14 
RegorafenibLomustine
(n = 42)(n = 42)
Age (years) 52.9 (46.2–60.9) 59.2 (51.8–64.4) 
Sex 
 Men 27 (64.3%) 29 (69.1%) 
 Women 15 (35.7%) 13 (30.9%) 
ECOG performance status 
 0 18 (42.8%) 18 (42.8%) 
 1 24 (57.2%) 24 (57.1%) 
Surgery at the time of recurrence 7 (16.7%) 10 (23.8%) 
Corticosteroid use 23 (54.8%) 28 (66.7%) 
MGMT status, n (% of tested) 
 Methylated 21 (50.0%) 19 (46.3%) 
 Unmethylated 21 (48.8%) 22 (53.7%) 
 Not done/unknown — 
IDH status, n (% of tested) 
 Unmutated 28 (93.3%) 28 (100%) 
 Mutated 2 (6.7%) — 
 Not done/unknown 12 14 

Note: Data are median (IQR) or n (%), unless otherwise indicated.

Abbreviations: ECOG, Eastern Cooperative Oncology Group; IDH, isocitrate dehydrogenase; MGMT, O6-methylguanine-DNA-methyltransferase.

Biomarkers analysis and association with prognosis

Because of inadequate representation of tumor or loss of tissue section during IHC analysis, the number of samples with evaluable IHC biomarkers was 38–42 in the regorafenib arm and 39–42 in the lomustine arm (Table 2). Digital pathology data for all biomarkers and tumor samples analyzed are shown in Supplementary Table S1. We performed IHC of pAMPK to investigate the LKB1/AMPK pathway activation in tumor sections. Phosphorylation of ACC, a direct downstream target of AMPK, was also analyzed. In general, pAMPK and pACC expression were detected in the cytoplasm of tumor cells, but 17 GBM samples showed nuclear positivity for pACC (Fig. 1). This nuclear pattern of positivity for pACC has been previously reported by another study in gliomas (18), and we scored these samples positive for this biomarker as those with canonical cytoplasmic positivity. Quantification of pACC expression in representative samples by the digital pathology technology exploited in this study is shown in Supplementary Fig. S1.

Table 2.

Evaluation of in situ biomarkers (IHC) in the REGOMA trial.

Treatment group
RegorafenibLomustine
N%N%
MCT1 38 39 
 0 = negative 13 34.2 23.1 
 1 = positive 25 65.8 30 76.9 
MCT4 42 41 
 0 = negative 15 35.7 10 24.4 
 1 = positive 27 64.3 31 75.6 
pAMPK 39 42 
 0 = negative 30 76.9 34 80.9 
 1 = positive 23.1 19.1 
pACC 41 42 
 0 = negative 17 41.5 13 30.9 
 1 = positive 24 58.5 29 69.1 
Treatment group
RegorafenibLomustine
N%N%
MCT1 38 39 
 0 = negative 13 34.2 23.1 
 1 = positive 25 65.8 30 76.9 
MCT4 42 41 
 0 = negative 15 35.7 10 24.4 
 1 = positive 27 64.3 31 75.6 
pAMPK 39 42 
 0 = negative 30 76.9 34 80.9 
 1 = positive 23.1 19.1 
pACC 41 42 
 0 = negative 17 41.5 13 30.9 
 1 = positive 24 58.5 29 69.1 
Figure 1.

Patterns of pACC expression in GBM samples from the REGOMA clinical trial. Representative microphotographs of pACC expression in one negative (133-8), one positive (133-7) sample with cytoplasmic pACC expression, and one sample with predominantly nuclear pACC expression (133-13); magnifications 200×, are shown.

Figure 1.

Patterns of pACC expression in GBM samples from the REGOMA clinical trial. Representative microphotographs of pACC expression in one negative (133-8), one positive (133-7) sample with cytoplasmic pACC expression, and one sample with predominantly nuclear pACC expression (133-13); magnifications 200×, are shown.

Close modal

In addition to pAMPK and pACC, we studied the expression of MCT1 and MCT4, which have a membrane pattern of staining and have been associated with the respiratory and the glycolytic metabolism of tumors, respectively (12–14). Among all the patients analyzed for these biomarkers, we reported a median OS of 6.4 months [95% confidence interval (CI), 5.5–7.9] and a median PFS of 1.9 months (95% CI, 1.8–2.1). Regardless of the treatment received, we found no impact of any of the biomarkers analyzed with either PFS or OS (Table 3), suggesting that these markers do not have prognostic significance in relapsed GBM.

Table 3.

Association of in situ biomarkers (IHC) with clinical outcome in the REGOMA trial.

PFSOS
6-month PFS (95% CI)Log-rank test P12-month OS (95% CI)Log-rank test P
MCT1 0 = negative 9.1 (1.6–25.1) 0.4634 40.9 (20.8–60.0) 0.2045 
 1 = positive 10.9 (4.4–20.7)  25.4 (14.9–37.4)  
MCT4 0 = negative 8.0 (1.4–22.5) 0.4104 36.0 (18.2–54.2) 0.3281 
 1 = positive 12.1 (5.3–21.8)  25.9 (15.4–37.5)  
pAMPK 0 = negative 10.9 (4.8–19.9) 0.7317 29.7 (19.1–41.1) 0.9020 
 1 = positive 16.7 (6.1–31.8)  33.3 (17.5–50.0)  
pACC 0 = negative 16.7 (6.1–31.8) 0.2290 33.3 (17.5–50.0) 0.3032 
 1 = positive 7.5 (2.4–16.6)  26.4 (15.4–38.7)  
PFSOS
6-month PFS (95% CI)Log-rank test P12-month OS (95% CI)Log-rank test P
MCT1 0 = negative 9.1 (1.6–25.1) 0.4634 40.9 (20.8–60.0) 0.2045 
 1 = positive 10.9 (4.4–20.7)  25.4 (14.9–37.4)  
MCT4 0 = negative 8.0 (1.4–22.5) 0.4104 36.0 (18.2–54.2) 0.3281 
 1 = positive 12.1 (5.3–21.8)  25.9 (15.4–37.5)  
pAMPK 0 = negative 10.9 (4.8–19.9) 0.7317 29.7 (19.1–41.1) 0.9020 
 1 = positive 16.7 (6.1–31.8)  33.3 (17.5–50.0)  
pACC 0 = negative 16.7 (6.1–31.8) 0.2290 33.3 (17.5–50.0) 0.3032 
 1 = positive 7.5 (2.4–16.6)  26.4 (15.4–38.7)  

With regard to the distribution of the markers in the tumor sections, we observed that MCT4, pAMPK, and pACC expression were stronger in peri-necrotic areas (Supplementary Fig. S2) possibly due to hypoxia and low ATP levels, which are typically found in these tumor regions (10).

Figure 2.

Kaplan–Meier curves of OS (top) and PFS (bottom) according to pACC status (negative, graphs on the left; positive, graphs on the right). Statistically significant interaction between treatment and pACC tumor status was demonstrated in terms of OS (interaction test P = 0.0453).

Figure 2.

Kaplan–Meier curves of OS (top) and PFS (bottom) according to pACC status (negative, graphs on the left; positive, graphs on the right). Statistically significant interaction between treatment and pACC tumor status was demonstrated in terms of OS (interaction test P = 0.0453).

Close modal

Moreover, we also assessed the possible role of the biomarkers as predictors of regorafenib treatment benefit. Among the biomarkers considered, only pACC status resulted in a significant modifier of the effect of treatment (interaction test P = 0.0453). Indeed, the HR for death of patients with pACC-positive tumors who received regorafenib was 0.37 (95% CI, 0.20–0.70) compared with those who received lomustine (P = 0.0020). Differently, the HR for patients with pACC-negative tumors treated with regorafenib was 1.1 (95% CI, 0.48–2.53) compared with those who received lomustine (P = 0.6927). The different impact of regorafenib treatment on the survival for the status of pACC was also highlighted in the Kaplan–Meier curves (Fig. 2). Patients with pACC-positive tumors reported a median OS of 9.3 months (95% CI, 5.6–13.2) compared with 5.5 months (95% CI, 4.2–6.6) for patients treated with lomustine (log-rank test P = 0.0013). OS at 12 months was 45.8% (95% CI, 25.6–64.0) in the patients with pACC-positive tumors treated with regorafenib with respect to 10.3% (95% CI, 2.6–24.3) for patients treated with lomustine. Vice versa, OS was not statistically different in patients with pACC-negative tumors according to treatment received.

No statistically significant interaction between treatment and pACC tumor status was demonstrated in terms of PFS (interaction test P = 0.2531), however, as showed in Fig. 2, PFS was longer in patients with pACC-positive tumors treated with regorafenib with respect to lomustine (HR, 0.54; 95% CI, 0.31–0.94). Test for interaction was negative for all other markers analyzed (data not shown), although pAMPK-positive patients had significantly improved OS when treated with regorafenib with respect to lomustine (Supplementary Fig. S3).

To investigate the variability between initial and recurrent setting, we analyzed pACC expression in a set of 16 matched diagnosis/relapse GBM samples including two pairs from the REGOMA clinical trial and 14 additional pairs from our previous study (19). Results showed no statistically significant difference in the expression level of pACC at diagnosis and relapse time (t test for paired data P = 0.62).

AMPK activation regulates ACC phosphorylation in GBM

Speculatively, one possible mechanism upstream of pACC expression could be activation of LKB1/AMPK, and activation of this pathway has an established role in reducing metabolic demand and cell proliferation under nutrient-starving conditions (20). To investigate this possibility, we correlated pAMPK with pACC expression in tumor sections from patients with GBM. Moderate pairwise association was found between pAMPK and pACC (Cramer V = 0.38; P = 0.0003). We also investigated the association between pACC and LKB1 expression. Because of the limited amount of tissue available for LKB1 staining, this analysis was limited to 67 samples. Statistical analysis disclosed a significant association between pACC status and LKB1 expression (Cramer V = 0.32, P = 0.0166). Finally, we investigated whether pACC-positive tumors could be more angiogenic and therefore better respond to regorafenib. We, therefore stained tumor sections from n = 52 patients with the endothelial cell marker, CD31, and calculated MVD by digital pathology (Supplementary Table S2). Statistical analysis disclosed poor association between pACC expression and MVD (categorized according to median value) in the samples analyzed (Cramer V = −0.11; P = 0.4296). In conclusion, pACC-positive tumor samples were not more angiogenic compared with pACC-negative samples.

Many studies sought to identify biomarkers correlating with activity of antiangiogenic drugs in patients with cancer (21, 22). Most of the biomarkers included in these studies focused on components of the VEGF/VEGFR or related angiogenic pathways. In contrast, the metabolic biomarkers reported in our study have never been previously analyzed in the context of antiangiogenic treatments of GBM. We found a possible predictive role of pACC expression for regorafenib benefit in patients with recurrent GBM. In our study, we showed that pACC expression correlated with survival, while no association was demonstrated in terms of PFS. Likely, this may be due to many factors including the lower effect of regorafenib on PFS compared with OS (9) and the small number of GBM samples studied in exploratory analysis. Moreover, REGOMA study was structured to have OS as primary endpoint.

Our hypothesis, based on preclinical data and previous retrospective translational studies in other cancer types (10, 23), poses that AMPK activation is increased by antiangiogenic therapy and triggers metabolic rewiring, which restrains cell proliferation. AMPK is indeed known to inhibit biosynthesis kinases, such as mTOR and ACC. Thus, according to this model, AMPK plays a tumor suppressor role in cancer cells. However, several studies indicate that AMPK can behave as oncogene in certain contexts, providing survival advantage critical for tumor growth (24–26). In brain tumors, Chhipa and colleagues recently observed that AMPK promotes GBM bioenergetics and tumor growth (17). In this study, oncogenic stress accounts for chronic AMPK activation in GBM stem cells and tumors, followed by cooption of the CREB1 pathway to coordinate tumor bioenergetics through the transcription factors HIF1α and GABPA. Moreover, independent studies reported that AMPK enhanced glioma cell viability by inducing lipid import in vitro (27), and was required to maintain cancer cell proliferation in astrocytic tumors (28). Consistent with tissue- and species-specific effects of AMPK, AMPKα1 suppressed lymphomagenesis in a Myc mouse model (29), whereas it protected leukemia-initiating cells in myeloid leukemias from metabolic stress in the bone marrow (30). Altogether, these studies underscore a complex and nonunivocal function of AMPK in cancer cell survival and tumor growth.

How do these considerations relate to our findings? First, it is conceivable that pACC positivity reflects AMPK activation in GBM samples, as supported by the moderate pairwise association between pAMPK and pACC expression. Notably, there was improved OS in the pAMPK-positive cohort of patients treated with regorafenib, but this biomarker yielded negative results in the interaction test, putatively due to the poorer quality of pAMPK compared with pACC staining or the limited power of this trial. Certainly, this biomarker deserves further investigation in future investigations. Second, both direct and indirect mechanism could hypothetically explain the better response of pACC-positive GBM to regorafenib. On the basis of previous studies, we speculated that pACC-positive samples might be more angiogenic compared with pACC-negative samples, through a mechanism involving increased HIF1A expression, which on its own sustains VEGF expression and angiogenesis in these tumors (31). Because regorafenib blocks VEGFR activity, this could account for higher antiangiogenic activity of the drug in pACC-positive compared with pACC-negative tumors. However, this hypothesis was not supported by results of MVD quantification, which suggested similar angiogenesis in pACC-positive and pACC-negative tumors. Apart from angiogenesis-related mechanisms, it cannot be excluded that regorafenib, which targets several tyrosine kinase receptors expressed by tumor cells, could have more pronounced direct effects on pACC-positive tumors, as AMPK activation in GBM might reflect oncogenic stress (17).

In addition to pAMPK and pACC, we studied expression of MCT1 and MCT4, which have been associated with the respiratory and the glycolytic metabolism of tumors. In previous studies, hypoxic areas are characterized by HIF1A-driven increased expression of MCT4, increased glycolytic metabolism, and lactate production, whereas well-vascularized tumor areas preferentially express MCT1, which predominantly acts as monocarboxylate importer and supports OXPHOS (reviewed in ref. 32). These biomarkers were considered of interest because a previous publication indicated that the glycolytic phenotype of tumors, indicated by high expression levels of MCT4, associated with improved clinical outcome in patients with colorectal cancer treated with the antiangiogenic tumor kinase inhibitor cediranib in a phase III clinical trial (11). However, in our study MCT1 and MCT4 did not predict efficacy of regorafenib, nor where they prognostic. Taking into consideration the recent evidence that tumor metabolic traits reflect, in part, those of the tissue of origin (33), this result could reflect a minor role of glycolysis in normal astrocytes and GBM cells.

Finally, a weakness of our study could be that, we used primary tumor tissue for biomarker analyzes, whereas the recurrent tumor was treated. However, we performed an exploratory analysis in 16 matched pairs of tumor samples and did not find statistically significant variations in pACC expression between the primary tumor and relapse, suggesting that the status of this molecular marker does not substantially change between primary and recurrent disease.

In conclusion, we demonstrated that pACC expression on tumoral tissue could be a valid predictive biomarker for patients with recurrent GBM treated with regorafenib. However, given the relatively small study population, our findings need to be validated in a larger population, prospectively.

S. Indraccolo is a co-inventor on a provisional patent application on the development of a biomarker predictive of response to regorafenib in glioblastoma patients that is owned by Istituto Oncologico Veneto IOV-IRCCS. G.L. De Salvo reports grants and non-financial support from Bayer SpA during the conduct of the study and is a co-inventor on a provisional patent application on the development of a biomarker predictive of response to regorafenib in glioblastoma patients that is owned by Istituto Oncologico Veneto IOV-IRCCS. T. Ibrahim reports personal fees from Eisai and Sanofì, grants from Novartis (institution), and other from PharmaMar (participation at meeting), Ipsen (participation at meeting), and Novartis (participation at meeting) outside the submitted work. V. Zagonel reports grants from Bayer during the conduct of the study as well as grants and personal fees from BMS, Roche, and Lilly, and personal fees from Astellas, Servier, and Astra Zeneca outside the submitted work. G. Lombardi reports grants and non-financial support from Bayer during the conduct of the study as well as personal fees and non-financial support from Bayer, personal fees from AbbVie and Brainfarm outside the submitted work, and is a co-inventor on a provisional patent application on the development of a biomarker predictive of response to regorafenib in glioblastoma patients that is owned by Istituto Oncologico Veneto IOV-IRCCS. No potential conflicts of interest were disclosed by the other authors.

S. Indraccolo: Conceptualization, formal analysis, supervision, funding acquisition, writing-original draft. G.L. De Salvo: Conceptualization, formal analysis, writing-original draft. M. Verza: Formal analysis, investigation, methodology. M. Caccese: Resources, investigation. G. Esposito: Formal analysis, methodology. I. Piga: Formal analysis, investigation, methodology. P. Del Bianco: Data curation, formal analysis, writing-review and editing. M. Pizzi: Resources, writing-review and editing. M.P. Gardiman: Resources, writing-review and editing. M. Eoli: Resources. R. Rudà: Resources. A.A. Brandes: Resources. T. Ibrahim: Resources. S. Rizzato: Resources. I. Lolli: Resources. V. Zagonel: Resources, writing-review and editing. G. Lombardi: Resources, writing-review and editing.

This work was funded by IOV Intramural Grant 5 × 1000 Genomica dei Tumori (to S. Indraccolo), Associazione Luca Ometto (to S. Indraccolo and V. Zagonel), and AIRC (IG18803 to S. Indraccolo).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Abou-Elkacem
L
,
Arns
S
,
Brix
G
,
Gremse
F
,
Zopf
D
,
Kiessling
F
, et al
Regorafenib inhibits growth, angiogenesis, and metastasis in a highly aggressive, orthotopic colon cancer model
.
Mol Cancer Ther
2013
;
12
:
1322
31
.
2.
Wilhelm
SM
,
Carter
C
,
Tang
L
,
Wilkie
D
,
McNabola
A
,
Rong
H
, et al
BAY 43-9006 exhibits broad spectrum oral antitumor activity and targets the RAF/MEK/ERK pathway and receptor tyrosine kinases involved in tumor progression and angiogenesis
.
Cancer Res
2004
;
64
:
7099
109
.
3.
Wilhelm
SM
,
Dumas
J
,
Adnane
L
,
Lynch
M
,
Carter
CA
,
Schutz
G
, et al
Regorafenib (BAY 73-4506): a new oral multikinase inhibitor of angiogenic, stromal and oncogenic receptor tyrosine kinases with potent preclinical antitumor activity
.
Int J Cancer
2011
;
129
:
245
55
.
4.
Bruix
J
,
Qin
S
,
Merle
P
,
Granito
A
,
Huang
YH
,
Bodoky
G
, et al
Regorafenib for patients with hepatocellular carcinoma who progressed on sorafenib treatment (RESORCE): a randomised, double-blind, placebo-controlled, phase 3 trial
.
Lancet
2017
;
389
:
56
66
.
5.
Demetri
GD
,
Reichardt
P
,
Kang
YK
,
Blay
JY
,
Rutkowski
P
,
Gelderblom
H
, et al
Efficacy and safety of regorafenib for advanced gastrointestinal stromal tumours after failure of imatinib and sunitinib (GRID): an international, multicentre, randomised, placebo-controlled, phase 3 trial
.
Lancet
2013
;
381
:
295
302
.
6.
Grothey
A
,
Van Cutsem
E
,
Sobrero
A
,
Siena
S
,
Falcone
A
,
Ychou
M
, et al
Regorafenib monotherapy for previously treated metastatic colorectal cancer (CORRECT): an international, multicentre, randomised, placebo-controlled, phase 3 trial
.
Lancet
2013
;
381
:
303
12
.
7.
Daudigeos-Dubus
E
,
Le Dret
L
,
Lanvers-Kaminsky
C
,
Bawa
O
,
Opolon
P
,
Vievard
A
, et al
Regorafenib: antitumor activity upon mono and combination therapy in preclinical pediatric malignancy models
.
PLoS One
2015
;
10
:
e0142612
.
8.
Hamed
HA
,
Tavallai
S
,
Grant
S
,
Poklepovic
A
,
Dent
P
. 
Sorafenib/regorafenib and lapatinib interact to kill CNS tumor cells
.
J Cell Physiol
2015
;
230
:
131
9
.
9.
Lombardi
G
,
De Salvo
GL
,
Brandes
AA
,
Eoli
M
,
Ruda
R
,
Faedi
M
, et al
Regorafenib compared with lomustine in patients with relapsed glioblastoma (REGOMA): a multicentre, open-label, randomised, controlled, phase 2 trial
.
Lancet Oncol
2019
;
20
:
110
9
.
10.
Nardo
G
,
Favaro
E
,
Curtarello
M
,
Moserle
L
,
Zulato
E
,
Persano
L
, et al
Glycolytic phenotype and AMP kinase modify the pathologic response of tumor xenografts to VEGF neutralization
.
Cancer Res
2011
;
71
:
4214
25
.
11.
Pommier
AJ
,
Farren
M
,
Patel
B
,
Wappett
M
,
Michopoulos
F
,
Smith
NR
, et al
Leptin, BMI, and a metabolic gene expression signature associated with clinical outcome to VEGF inhibition in colorectal cancer
.
Cell Metab
2016
;
23
:
77
93
.
12.
Halestrap
AP
,
Price
NT
. 
The proton-linked monocarboxylate transporter (MCT) family: structure, function and regulation
.
Biochem J
1999
;
343 Pt
2
:
281
99
.
13.
Mikkilineni
L
,
Whitaker-Menezes
D
,
Domingo-Vidal
M
,
Sprandio
J
,
Avena
P
,
Cotzia
P
, et al
Hodgkin lymphoma: a complex metabolic ecosystem with glycolytic reprogramming of the tumor microenvironment
.
Semin Oncol
2017
;
44
:
218
25
.
14.
Sonveaux
P
,
Copetti
T
,
De Saedeleer
CJ
,
Vegran
F
,
Verrax
J
,
Kennedy
KM
, et al
Targeting the lactate transporter MCT1 in endothelial cells inhibits lactate-induced HIF-1 activation and tumor angiogenesis
.
PLoS One
2012
;
7
:
e33418
.
15.
Bonanno
L
,
De Paoli
A
,
Zulato
E
,
Esposito
G
,
Calabrese
F
,
Favaretto
A
, et al
LKB1 expression correlates with increased survival in patients with advanced non-small cell lung cancer treated with chemotherapy and bevacizumab
.
Clin Cancer Res
2017
;
23
:
3316
24
.
16.
Shackelford
DB
,
Shaw
RJ
. 
The LKB1-AMPK pathway: metabolism and growth control in tumour suppression
.
Nat Rev Cancer
2009
;
9
:
563
75
.
17.
Chhipa
RR
,
Fan
Q
,
Anderson
J
,
Muraleedharan
R
,
Huang
Y
,
Ciraolo
G
, et al
AMP kinase promotes glioblastoma bioenergetics and tumour growth
.
Nat Cell Biol
2018
;
20
:
823
35
.
18.
Wakamiya
T
,
Suzuki
SO
,
Hamasaki
H
,
Honda
H
,
Mizoguchi
M
,
Yoshimoto
K
, et al
Elevated expression of fatty acid synthase and nuclear localization of carnitine palmitoyltransferase 1C are common among human gliomas
.
Neuropathology
2014
;
34
:
465
74
.
19.
Indraccolo
S
,
Lombardi
G
,
Fassan
M
,
Pasqualini
L
,
Giunco
S
,
Marcato
R
, et al
Genetic, epigenetic, and immunologic profiling of MMR-deficient relapsed glioblastoma
.
Clin Cancer Res
2019
;
25
:
1828
37
.
20.
Hardie
DG
,
Ross
FA
,
Hawley
SA
. 
AMPK: a nutrient and energy sensor that maintains energy homeostasis
.
Nat Rev Mol Cell Biol
2012
;
13
:
251
62
.
21.
Jayson
GC
,
Hicklin
DJ
,
Ellis
LM
. 
Antiangiogenic therapy–evolving view based on clinical trial results
.
Nat Rev Clin Oncol
2012
;
9
:
297
303
.
22.
Jubb
AM
,
Harris
AL
. 
Biomarkers to predict the clinical efficacy of bevacizumab in cancer
.
Lancet Oncol
2010
;
11
:
1172
83
.
23.
Zulato
E
,
Bergamo
F
,
De Paoli
A
,
Griguolo
G
,
Esposito
G
,
De Salvo
GL
, et al
Prognostic significance of AMPK activation in advanced stage colorectal cancer treated with chemotherapy plus bevacizumab
.
Br J Cancer
2014
;
111
:
25
32
.
24.
Bungard
D
,
Fuerth
BJ
,
Zeng
PY
,
Faubert
B
,
Maas
NL
,
Viollet
B
, et al
Signaling kinase AMPK activates stress-promoted transcription via histone H2B phosphorylation
.
Science
2010
;
329
:
1201
5
.
25.
D'Amico
D
,
Antonucci
L
,
Di Magno
L
,
Coni
S
,
Sdruscia
G
,
Macone
A
, et al
Non-canonical Hedgehog/AMPK-mediated control of polyamine metabolism supports neuronal and medulloblastoma cell growth
.
Dev Cell
2015
;
35
:
21
35
.
26.
Laderoute
KR
,
Calaoagan
JM
,
Chao
WR
,
Dinh
D
,
Denko
N
,
Duellman
S
, et al
5′-AMP-activated protein kinase (AMPK) supports the growth of aggressive experimental human breast cancer tumors
.
J Biol Chem
2014
;
289
:
22850
64
.
27.
Rios
M
,
Foretz
M
,
Viollet
B
,
Prieto
A
,
Fraga
M
,
Garcia-Caballero
T
, et al
Lipoprotein internalisation induced by oncogenic AMPK activation is essential to maintain glioblastoma cell growth
.
Eur J Cancer
2014
;
50
:
3187
97
.
28.
Rios
M
,
Foretz
M
,
Viollet
B
,
Prieto
A
,
Fraga
M
,
Costoya
JA
, et al
AMPK activation by oncogenesis is required to maintain cancer cell proliferation in astrocytic tumors
.
Cancer Res
2013
;
73
:
2628
38
.
29.
Faubert
B
,
Boily
G
,
Izreig
S
,
Griss
T
,
Samborska
B
,
Dong
Z
, et al
AMPK is a negative regulator of the Warburg effect and suppresses tumor growth in vivo
.
Cell Metab
2013
;
17
:
113
24
.
30.
Saito
Y
,
Chapple
RH
,
Lin
A
,
Kitano
A
,
Nakada
D
. 
AMPK protects leukemia-initiating cells in myeloid leukemias from metabolic stress in the bone marrow
.
Cell Stem Cell
2015
;
17
:
585
96
.
31.
Schito
L
,
Semenza
GL
. 
Hypoxia-inducible factors: master regulators of cancer progression
.
Trends Cancer
2016
;
2
:
758
70
.
32.
Morandi
A
,
Indraccolo
S
. 
Linking metabolic reprogramming to therapy resistance in cancer
.
Biochim Biophys Acta Rev Cancer
2017
;
1868
:
1
6
.
33.
Muir
A
,
Vander Heiden
MG
. 
The nutrient environment affects therapy
.
Science
2018
;
360
:
962
3
.