Introduction
Alterations of metabolic pathways are a cancer hallmark, resulting in dependencies on specific nutrients for cell proliferation and tumor growth [
1,
2]. In addition to glucose, mammalian cells use glutamine to feed the tricarboxylic acid (TCA) cycle as an alternative source of carbon, and a precursor for proteins, lipids, and nucleic acids. Glutamine is also a key precursor in the synthesis of the antioxidant glutathione, which is important in maintaining the redox balance in cells and tissues [
3]. Furthermore, glutamine can be converted via glutamate to proline, which is found to play an important regulatory role in cancer [
4,
5]. Recent research has identified proline as an important metabolite during adaption to hypoxia [
6].
Although glutamine is a nonessential amino acid that can be synthesized from glutamate, many cancer cells depend on exogenous glutamine supply for proliferation and tumor growth [
7,
8]. Upregulated glutaminolysis is observed in many aggressive forms of human cancer, including colorectal cancer [
9], gliomas [
10], pancreatic cancer [
11], melanoma [
12], and breast cancer [
13,
14], which highlights the importance of this amino acid in tumor metabolism.
The rate-limiting step in glutamine metabolism is the conversion of glutamine to glutamate, which is catalyzed by the enzyme glutaminase [
15]. Glutaminase exists in several tissue-specific variants, encoded by two genes in mammals, kidney-type glutaminase (
GLS1), and liver-type glutaminase (
GLS2). GLS1 plays a central role in tumorigenesis, whereas the role of GLS2 in cancer remains unclear. GLS1 has been found to be higher expressed in TNBC compared to other subgroups of breast cancer [
14] and is essential for the survival of TNBC cells with a deregulated glutaminolysis pathway [
16].
Inhibition of glutaminase has been stated as an attractive therapeutic approach in various cancers [
8,
13,
14]. The glutaminase inhibitor CB-839 (Calithera) is currently being tested in clinical trials for several malignancies including breast cancer (clinicaltrials.gov, ID: NCT02071862). CB-839 is found to be specific to GLS1 and not to GLS2 [
14]. Gross and colleagues found that TNBC cell lines displayed a higher sensitivity to CB-839 compared to ER+ cell lines [
14]. In addition, CB-839 caused a significant antitumor activity in two selected xenograft models representing TNBC and basal-like/HER2+ breast cancer. In a clinical phase I trial, two of nine patients with TNBC treated with CB-839 experienced stabilized disease [
17]. While these studies have shown promising results in the TNBC subgroup, there is a need to identify better predictive response biomarkers for optimal utilization and selection of patients that more likely will respond to CB-839.
The overall aim of this study was to identify metabolic characteristics associated with response to glutaminase inhibitors in breast cancer. First, we showed that CB-839 treatment has a differential effect on the growth of two patient-derived xenograft (PDX) models of breast cancer representing luminal-like, ER+ (MAS98.06) and basal-like, triple-negative (MAS98.12) breast cancer. Then, we investigated the potential causes of this differential response by (i) assessing the expression of selected genes and proteins directly involved in glutamine metabolism on tumor tissue from untreated mice and (ii) measuring downstream glutamine metabolites in tumor samples from CB-839-treated and untreated mice after administration of [5-13C] glutamine using 13C high-resolution magic angle spinning MR spectroscopy (HR MAS MRS).
Discussion
In this study, we have evaluated the treatment response to the glutaminase inhibitor CB-839 in two breast cancer PDX models. Previous studies suggest that basal-like and triple-negative breast cancers are more sensitive to glutaminase inhibitors than luminal-like and ER+ subgroups [
14,
16]. Our findings complement this paradigm; The MAS98.06 xenografts, previously found to belong to the luminal B subtype, were sensitive to glutaminase inhibition whereas the basal-like MAS98.12 xenografts did not respond to treatment.
Previous reports have suggested that TNBC and basal-like breast cancer are particularly dependent on glutamine for proliferation and tumor growth [
14,
16,
44]. Gross and colleagues showed that the TNBC subtype displayed the greatest sensitivity to CB-839 treatment in vitro and that the sensitivity was positively correlated to dependence on extracellular glutamine for growth, high baseline ratio of intracellular glutamate to glutamine (Glu:Gln), and expression of glutaminase (GLS1) [
14]. A high Glu:Gln ratio has been proposed as an independent diagnostic entity. Although other studies indicate significant differences in Glu:Gln between breast cancer subgroups, a high Glu:Gln phenotype is observed across all the molecular subtypes [
45]. In our experiments, we observed that the triple-negative/basal-like MAS98.12 model had significantly higher gene expression of
GLS1 and a trend of higher GLS1 protein expression, and a significantly higher Glu:Gln ratio compared to the luminal B/ER+ MAS98.06 models. Despite the accordance with previously proposed predictive biomarkers, the MAS98.12 model did not respond to CB-839 treatment, suggesting that response and resistance depend on other mechanisms regulating glutamine metabolism. Metabolic redundancy or plasticity can rescue cancer cells from glutaminase inhibition, and our understanding of metabolic characteristics within the molecular breast cancer subtypes remains limited. Several studies have demonstrated metabolic variability within subtypes and a lack of correlation between metabolic and transcriptomic traits [
46,
47]. Although the triple-negative phenotype has been suggested to be particularly dependent on glutaminolysis, it has been shown that as many as 25% of these do not express GLS1 as determined by immunohistochemistry [
48]. In a panel of 26 PDX models, we found more than ten-fold difference in the expression of
GLS1 within the basal-like subtype (Additional file
2). Similar variability was observed for
SLC1A5,
GS, and
GLUD1. Both MAS98.06 and MAS98.12 displayed
GLS1 expression higher than the average for the cohort. No clear distinctions between basal-like and luminal B xenografts were observed for any of the genes. This indicates that although the two subtypes display metabolic differences on the population level, it must be expected that individuals within each subtype display atypical metabolic characteristics.
In order to better understand mechanisms responsible for response and resistance to glutaminase inhibitors, we studied glutamine utilization in the two xenografts by infusion of 13C-enriched glutamine and ex vivo 13C NMR tumor analysis, combined with gene- and protein expression data. From the 13C HR MAS MRS analysis on untreated models, we saw that the two xenograft models utilize glutamine differently. Although the total glutamine consumption was similar in the two models, the main metabolic fate of glutamine in the non-responding MAS98.12 was conversion into glutamate, lactate, and alanine. In contrast, the responding MAS98.06 tumors store significant amounts of glutamine in the tumors, but also use glutamine for synthesis of proline, alanine, and lactate. Both models use a similar fraction of the glutamine to feed the TCA cycle, as seen by the presence of [1-13C] glutamate (MAS98.06) and [1-13C] lactate (MAS98.12).
A possible scenario that may explain why some tumors are insensitive to CB-839 is that they use glycolysis instead of glutaminolysis for anaplerotic feeding of the TCA cycle. Pyruvate can enter the TCA cycle through dehydrogenation of pyruvate into acetyl-CoA (which is catalyzed by pyruvate dehydrogenase (PDH)) or through carboxylation of pyruvate into oxaloacetate (catalyzed by pyruvate carboxylase (PC)). However, we have previously found that following [1-
13C] Glc injection, MAS98.12 tumors feed less glucose into the TCA cycle than MAS98.06 tumors, both via PDH and PC [
49]. Differences in anaplerotic fueling of the TCA by glucose can therefore not explain the differences in response to CB-839 observed in this study.
It is well established that the signaling landscape plays critical roles in regulating proliferation and tumor growth [
50]. The transcription factor Myc affects glutamine metabolism by enhancing glutamine uptake, glutaminase activity, and upregulation of proline metabolism [
4,
26,
34‐
36]. Myc may trigger addiction to glutamine, which is observed both in vitro and in vivo [
8,
50]. Several transporters are capable of transporting glutamine across the plasma membrane [
26]. Among these, SLC1A5 has received increased attention because its expression is upregulated in many cancer types, including triple-negative breast cancer [
27]. We observed that MAS98.12 xenografts display a borderline significant higher expression of MYC than MAS98.06 xenografts. This may explain the higher expression of the glutamine transporter SLC1A5 protein and
GLS1 mRNA in MAS98.12. However, expression of MYC was not associated with increased glutamine consumption or proline synthesis in our models.
In tumor tissue from the responding MAS98.06 model, we saw a depletion of proline and alanine after treatment. Alanine is produced from pyruvate, which can be synthesized using several routes, including glycolysis. Proline, on the other hand, is a conditionally essential amino acid [
51]. In untreated MAS98.06 xenografts, a significant fraction of the injected [5-
13C] glutamine was converted to [5-
13C] proline. This is consistent with gene and protein expression data, showing that ALDH18A1 and PYCR1 were more highly expressed in MAS98.06 than MAS98.12 both on gene and protein expression level. These results indicate that MAS98.06 tumors have a higher flux from glutamate to proline compared to MAS98.12 tumors. Recent research by Tang and colleagues has identified proline as an important metabolite in cancer, linked to adaption to tumor hypoxia [
6]. They showed that hypoxic microenvironments activated proline metabolism via upregulation of
ALDH18A1 in tumor samples from patients with hepatocellular carcinoma. We have previously shown that the luminal-like MAS98.06 tumors are more hypoxic than basal-like MAS98.12 tumors [
20]. It could therefore be speculated that MAS98.06 tumors adapt to hypoxic microenvironments through activation of proline-mediated mechanisms. Depletion of proline following CB-839 treatment to the MAS98.06 models could leave the tumors unable to handle hypoxic stress, consequently inhibiting the growth of these tumors.
Overall, our results indicate that current biomarkers suggested for predicting response to glutaminase inhibition do not fully capture the complexity of glutamine metabolism in cancer and that the response to glutaminase inhibitors depends on the individual tumor’s ability to compensate for reduced glutamate availability. One possible explanation for the glutamine dependence in MAS98.06 tumors is that they use glutamine in proline biosynthesis, for adaption to hypoxic microenvironments.