Current view of cancer metabolism

The seminal work by Otto Warburg defined glycolysis under aerobic conditions as the basis of cancer metabolism (Warburg, 1956a). His primary basis for excess glucose metabolism was postulated to be a result of impaired mitochondria metabolism. While correct in the context of his working models, Warburg dismissed parallel studies from Sidney Weinhouse that accurately described oxidation of carbohydrates and fatty acids within tumors (Weinhouse et al., 1951). Weinhouse had discovered that tumors had active enzymes within the TCA cycle comparable to normal tissue (Wenner et al., 1952). These studies suggested that tumors in fact rely on metabolic pathways intrinsic to mitochondria. The two scientists - Warburg and Weinhouse disagreed on the basis of cancer metabolism and articulated these differences in a historic publication in Science in 1956 (Warburg, 1956b; Weinhouse, 1956). Their disagreement raised an important question - what metabolites do tumors use to sustain rapid anabolic growth? Today our understanding of cancer metabolism is that it is a heterogeneous process classically delineated by aerobic glycolysis but also equally dependent on mitochondrial metabolism for tumor growth (Desjardins et al., 1985; Morais et al., 1994; Cavalli et al., 1997; Weinberg et al., 2010).

Our view of tumor metabolism defined by Otto Warburg (Warburg et al., 1927) and coined by Efraim Racker in 1972 as the “Warburg Effect” (Racker, 1972) has been much refined, deepened and broaden enough that we may accept both Warburg and Weinhouse were correct, given the metabolic heterogeneity of their tumor models. For more reading on the life of Otto Warburg and the impact his research has had on the field of cancer metabolism please refer to an excellent review by Chi Dang and colleagues (Koppenol et al., 2011). The last two decades have marked a wave of studies dissecting the genetic, molecular and environmental basis of altered metabolism in cancer cells. Given that tumors show tremendous mutational heterogeneity, it is not surprising that they would likewise show diverse metabolic heterogeneity. Tumors utilize a host of metabolites that include glucose, amino acids and fatty acids (Vander Heiden and DeBerardinis, 2017). However, the bulk of cancer metabolism studies have been confined to in vitro cell culture systems and require in vivo translation. Given our newly emerging vision of tumor metabolism as a multi-faceted and heterogeneous process how might we begin studying tumor metabolism through a different set of lenses? One answer lies in real time imaging of tumors.

In this review we will describe the use of multi-modality imaging (MMI) techniques that have been combinatorially used to quantitatively measure and assess tumor metabolism in vivo in both animal models of cancer and in cancer patients. We will describe noninvasive imaging modalities that include positron emission tomography (PET), magnetic resonance imaging (MRI) and magnetic resonance spectroscopy (MRS), computed tomography (CT) and optical imaging. We will also describe methods such as liquid chromatography and gas chromatography mass spectrometry (LC-MS and GC-MS) and quantitative immunohistochemistry (qIHC) that can be utilized to complement noninvasive imaging techniques.

Positron emission tomography

Positron Emission Tomography (PET) imaging is a widely used imaging modality employed in both clinical and basic research settings. PET imaging works by detection of gamma rays from positron emitting radionuclides that have been injected into a patient or animal. The most commonly used radionuclide is ¹⁸F but there is a wide range of radionuclides available - the more commonly used isotopes include ¹⁸F, ¹¹C and ¹⁵O. The high consumption of glucose by advanced tumors makes PET imaging with [¹⁸F]fluro-2-deoxyglucose (¹⁸F-FDG) an ideal probe to detect glycolytic tumors. ¹⁸F-FDG works by entering the cell through glucose transporters (GLUTs) (Fig. 1) where it is rapidly phosphorylated by hexokinase into ¹⁸F-FDG-6-phosphate where it can no longer be metabolized. Here the radiotracer remains trapped in the early stages of glycolysis allowing the detection of emitted gamma rays as the probe decays. Importantly, the 3 dimensional detection of gamma emission from a single source enables 3 dimensional reconstruction of the tumor. ¹⁸F-FDG is used to successfully diagnose a broad range of tumors that include cancers of the lung, liver, bone and soft tissue (Minn et al., 1997; Oshida et al., 1998; Shiomi et al., 2001; Shi et al., 2015; Hwang et al., 2016).

¹⁸F-FDG PET is particularly useful for metabolic studies because it measures glucose flux into the tumor and activity in early steps of glycolysis. While it also serves to detect a tumor mass, importantly it provides valuable functional information about the metabolic needs of the tumor. Moreover, when imaging cancer metabolism, it is critical to identify probes and modalities that provide, in addition to anatomical registration, functional information about the metabolic activity within the tumor(s). The noninvasive nature of PET imaging allows for repeat scans to be performed on patients during the course of treatment. The PET Response Criteria in Solid Tumors or PERCIST is a set of criteria that utilizes PET imaging with ¹⁸F-FDG to determine therapeutic response in patients. The efficacy of PERCIST criteria was demonstrated in a clinical trial, which used ¹⁸F-FDG uptake in tumors to evaluate breast tumor response to the PI3K inhibitor Buparlisib (Mayer et al., 2014). Likewise, longitudinal PET imaging of small animals is feasible and provides an accurate means to study tumor metabolism and measure therapeutic response over time. This is represented in Fig. 2A showing the longitudinal ¹⁸F-FDG PET and CT imaging of Kras^G12D driven lung tumors in genetically engineered mouse models (GEMMs).

Tumors do not solely rely on glucose. Therefore radiolabeling of additional metabolites such as acetate, choline, methionine and glutamine with either ¹⁸F or ¹¹C provide opportunities to perform broad profiling of cancer metabolism with PET imaging. ¹¹C-acetate is converted to acetyl-CoA and used in mitochondria in TCA cycle or incorporated into cell membranes (Vavere et al., 2008; Yoshimoto et al., 2001) and along with ¹¹C- and ¹⁸F-choline both are used in management of prostate cancer (Testa et al., 2016; Wibmer et al., 2016). ¹¹C-methionine is used as a marker of amino acid uptake and protein synthesis primarily in glioma where uptake of the radiotracer correlates with tumor grade (Pirotte et al., 2004; Kim et al., 2005; Van Laere et al., 2005). ¹¹C and ¹⁸F-glutamine have been used in both clinical and basic research to evaluate reliance of tumor cells on glutamine uptake and glutaminolysis (Lieberman et al., 2011; Ploessl et al., 2012; Wu et al., 2014; Venneti et al., 2015; Hassanein et al., 2016; Momcilovic et al., 2017; Schulte et al., 2017; Zhou et al., 2017).

Computed tomography

Computed tomography (CT) imaging utilizes multi-positional X-ray imaging to generate a 3 dimensional view of the imaged area. Tomographic reconstruction of the X-ray images provides detailed anatomical information of the imaged patient or animal. CT imaging can be performed with contrast agents that register vasculature within the tumor and perfusion within the tumor(s). CT imaging with iodine based contrast agents such as iohexol, iodixanol and ioversol are advantageous because it renders a quantitative measure of blood vessels supplying the tumors (Kao et al., 2003; Mukundan et al., 2006; Karathanasis et al., 2009; de Vries et al., 2010).

Tumor vasculature directly impacts the nutrients accessible to the tumors and can be used to differentiate well-perfused regions from hypoxic regions. Recent study has evaluated contrast-enhanced CT with HX4-PET, PET probe specific for hypoxia. This study showed that in lung cancer patients both modalities were able to classify tumors as normoxic or hypoxic (Even et al., 2017). These imaging technologies are widely available for basic research and clinical use. However, since CT imaging exposes both patients and animals to ionizing radiation, the frequency of repetitive imaging must be taken into consideration for both clinical evaluation and studies in animal models (Brenner and Hall, 2007). Excessive exposure to radiation can be avoided by opting for low dose CT scans that provide accurate anatomical registration and patient benefit (National Lung Screening Trial Research Team et al., 2011, 2013; Rampinelli et al., 2013; Kubo et al., 2016).

Magnetic resonance imaging and spectroscopy

Magnetic Resonance Imaging (MRI) is based on the physical phenomenon called nuclear magnetic resonance (NMR). NMR is based on quantifying changes in nuclear spin, a property of atomic nuclei, in response to a strong external magnetic field. NMR signal is detected upon relaxation of nuclear spins. For imaging metabolism most useful nuclei are ¹H, ¹³C and ³¹P. Like CT imaging, MR imaging is most frequently used to determine anatomical registration of the tumor. In addition, MRI can be performed with contrast agents to determine perfusion within the tumor (Yankeelov and Gore, 2009).

Most commonly Magnetic Resonance Spectroscopy (MRS) has been used to evaluate endogenous ¹H signals from choline-containing molecules. Signal intensity of the ¹H MRS peak has correlated with proliferation in brain, prostate, breast, colon and cervical cancers (Nelson et al., 2002; Geraghty et al., 2008; Kurhanewicz and Vigneron, 2008; Haddadin et al., 2009; Mountford et al., 2009). One of the limitations of 1H MRS imaging is low sensitivity, with concentration of endogenous metabolites required to be in the high micromolar to millimolar range (Di Gialleonardo et al., 2016). In order to address the relatively low sensitivity of MRS imaging, a novel method for producing hyperpolarized probes has been developed (Ardenkjaer-Larsen et al., 2003; Golman et al., 2006). Hyperpolarization of ¹³C nuclei has allowed for increase in the signal to noise ratio and detection sensitivity with up to 10,000-fold enhancement compared to conventional MRI (Ardenkjaer-Larsen et al., 2003; Keshari and Wilson, 2014).

Hyperpolarized [1-¹³C]-pyruvate has been used to detect levels of lactate and alanine as well as total amount of hyperpolarized ¹³C in preclinical models of prostate cancer (Albers et al., 2008). Furthermore, hyperpolarized ¹³C-pyruvate was successfully applied to 31 patients with prostate cancer with excellent safety profile and spectral quality (Nelson et al., 2013). Data obtained from ¹³C-pyruvate was able to accurately delineate the presence, location and size of cancer relative to surrounding non-cancer tissue. Moreover, MRS imaging is able to detect not only ¹³C-pyruvate itself, but also metabolites downstream of pyruvate such as lactate and alanine. This greatly expands usefulness of the probe in measuring metabolism. In addition to detection and quantification of uptake of the initial probe, quantifying downstream metabolites allows for the estimation of enzyme activity and the estimation of activity of entire pathways in tumors (Gutte et al., 2015).

As with other modalities of non-invasive imaging, monitoring response to therapy is one of the most promising aspects of MRS imaging. In a pre-clinical model of lymphoma, ¹³C-pyruvate was used to show that loss of flux from pyruvate to lactate correlated with response to chemotherapy (Day et al., 2007). Increased lactate signal was observed at sites of cancer recurrence in a MYC-driven murine breast cancer model (Shin et al., 2017). Additionally, ¹³C-labeled acetate was used in pre-clinical models of glioblastomas and in brain tumor patients illustrating that tumors can oxidize acetate in addition to glucose (Mashimo et al., 2014). As with each imaging modality and technology, MRS imaging with hyperpolarized probes will benefit from further development of technology to improve generation and delivery of the probe to patients.

Optical imaging

Optical imaging is a widely used modality in basic research laboratories that measures and quantifies emission of visible light and wavelengths in near infrared spectrum. Optical imaging can be done in both fluorescent and bioluminescent models. Fluorescence imaging of dyes and proteins such as green fluorescent protein (GFP) has a low signal to noise ratio, due to the emission of light within the visible spectrum by cellular proteins and DNA. This has been improved upon with the development of fluorescent probes with emission in the near infrared spectrum (Zhang et al., 2012). The problem of signal to noise is absent in bioluminescent imaging (BLI) of cells and rodent models of disease since mammalian tissue does not spontaneously generate light.

Bioluminescence is a natural phenomenon observed across numerous species from bacteria, to worms to beetles to fireflies in which the luciferase enzyme catalyzes a reaction that releases photons of light. First taking advantage of the luciferase enzyme from fireflies, researchers were able to clone luciferase and introduce it as a reporter in cell lines and transgenic animals (Fan and Wood, 2007; Woolfenden et al., 2009). BLI is highly sensitive and can detect fewer than 10 injected cells in a mouse using enhanced firefly luciferase (Rabinovich et al., 2008; Kim et al., 2010). BLI has most frequently been used as a noninvasive means to detect implanted tumors and to follow metastases in mouse xenografts. In addition, stable luciferase reporters have been introduced into genetically engineered mouse models (GEMMs) in which the luciferase gene is placed under a lox-stop-lox regulated endogenous promoter such as the Rosa26 locus (Safran et al., 2003). The lox-stop-lox regulated luciferase reporters are activated simultaneously with tumor induction and serve as an effective reporter of tumor growth, where the amount of light emission by BLI is proportional to the tumor burden (Fig. 2A).

The sensitivity of BLI enables small tumor lesions to be detected at early stages of tumorigenesis that fall below the level of detection for CT, MRI and PET imaging (Shackelford et al., 2013; Stollfuss et al., 2015). This is highlighted in Fig. 2A comparing ¹⁸F-FDG PET and BLI of lung tumor lesions at 2–10 weeks post tumor induction. A limitation of BLI is the lack of 3 dimensional resolution. Therefore, the BLI can give a general estimate of tumor mass, but it cannot discern between single vs multiple tumor nodules. In contrast PET, CT and MR imaging provide an accurate 3 dimensional spatial mapping of the tumor and surrounding tissue but is limited in its ability to detect early stage small neoplasias. In a study comparing both ¹⁸F-FDG PET and BLI in animal models of metastatic breast cancer the authors concluded that both imaging modalities were complementary (Kang et al., 2015).

Imaging with D-luciferin can detect the presence of tumors, but can BLI be used to study functional metabolism? BLI using luciferase can be used to measure ATP levels, allowing for the estimation of the energetic charge within the cell. This is based on the fact that reaction catalyzed by luciferase requires ATP, thus emitted light can be correlated to the amount of ATP available (Kimmich et al., 1975; Morciano et al., 2017). A limited number of bioluminescent and fluorescent probes have been developed to study metabolic pathways in cancer in more detail. Both glucose and free fatty acid analogs have been developed to study both glucose and fatty acid metabolism shown in Fig. 1 (Zhou et al., 2009; Henkin et al., 2012). Moreover, the use of bioorthogonal chemistry also known as ‘click’ chemistry enables the design of functional luciferin probes that can be activated by cellular metabolites and enzymes. Several recent studies utilized a luciferin probe conjugated to the free fatty acid palmitate to measure in vivo uptake and reduction in brown adipose tissue (Jang et al., 2016; Lynes et al., 2017). Here a cleavable disulfide bond links palmitate to luciferin. In the uncleaved form, palmitate acts as a cage to restrict luciferin from being catalyzed by the luciferase enzyme, which is expressed as a reporter gene. The caged luciferin is rendered inert until entry into the cell. Upon import of the probe into the cell, the disulfide bond is reduced and cleaved thus resulting in uncaging of the luciferin and subsequent catalysis by the luciferase reporter (Henkin et al., 2012). The amount of light detected following activation of the probe is proportional to the uptake of palmitate by the tissue being imaged. The development of caged luciferins that measure import of metabolites and enzyme activity represent a potential emerging technology that may integrated into in vivo studies of cancer metabolism.

Mass spectrometry-based isotopomeric metabolic tracing

Metabolomics refers to the global study of metabolism and can be examined using multiple techniques. While technologies such as gene set enrichment analysis (GSEA) and protein expression can be used to profile the tumor’s global metabolic profile (Nielsen, 2017), this review will focus on direct analysis of metabolites using Stable Isotope Resolved Metabolomics (SIRM). This technique utilizes liquid chromatography mass spectrometry (LC-MS) or gas chromatography mass spectrometry (GC-MS) as a direct means to measure the distribution of labeled metabolites. It takes advantage of low abundance in the cells and tissues of 13C and 15N isotopes compared to high abundance of natural isotopes 12C and 14N. This selectivity is coupled to extremely high sensitivity of detection of stable 13C and 15N isotopes offered by mass spectrometry. Employing SIRM allows tracing of the metabolic fate of the individual atoms from the labeled molecule. This means that as labeled molecule undergoes transformation in a metabolic pathway its contribution to each step of the pathway can be accurately quantified. A number of 13C- and 15N-labeled molecules are commercially available (for example ¹³C-glucose, ¹³C- and ¹⁵N-glutamine, ¹³C-palmitate) making it possible to study the fate of an individual metabolite as it is processed through multiple enzymatic steps in the cell.

The Fan laboratory has published a number of pioneering studies using SIRM technology in lung cancer cell lines (Fan et al., 2004, 2006) and more importantly on human lung cancer isolated from resected tumors (Fan et al., 2009). These studies demonstrated the translatability of the SIRM from basic research models into human patients. Several studies from the Vander Heiden lab have performed in vivo SIRM analysis using mouse models that identified distinct metabolic dependencies in Kras driven tumors (Davidson et al., 2016; Mayers et al., 2016). The DeBerardinis lab extended these findings by performing SIRM analysis on human non-small cell lung cancer. Quantification of glycolysis in tumors identified that metabolic heterogeneity within tumors can lead to different preference for carbon courses (Hensley et al., 2016). These findings bring forward an important question: do areas of the tumor that rely on the different carbon sources have different responses to standard therapy, whether it be chemotherapy, immunotherapy or targeted therapy?

Additional mass spectrometry based approaches have taken advantage of ‘click’ chemistry to design probes such as MitoClick that measure mitochondrial membrane potential (Logan et al., 2016). The mitochondrial membrane potential is maintained by the electron transport chain and serves to drive the generation of ATP by complex V. The MitoClick probe is voltage dependent and will accumulate in the inner mitochondrial membrane when a mitochondrial membrane potential is present. Quantification of MitoClick was performed using LC-MS of cardiac tissue in mice and although this probe cannot be analyzed by noninvasive methods, it does highlight a novel use of click chemistry and LC-MS to study mitochondrial bioenergetics in vivo.

Quantitative immunohistochemistry of signal transduction pathways

The use of bright field microscopy has been a staple of histological analysis of tumor tissue for decades and is widely used by clinical pathology as well as basic research laboratories. Quantitative immunohistochemical (qIHC) staining of key signaling proteins and metabolic enzymes that regulate metabolism can readily be performed on formalin fixed paraffin embedded (FFPE) tumor sections in both human and mouse tumors. Fig. 1 represents signaling kinases within pathways such as the PI3K/AKT/mTOR signaling axis that function as critical regulators of glycolysis (Elstrom et al., 2004; Duvel et al., 2010; Hu et al., 2016). The bulk of these kinases are activated through phosphorylation of serine, threonine and tyrosine residues making them amenable to detection using phospho-specific antibodies. This allows for on-off rates of these enzymes to be measured in primary, FFPE tumor tissue.

To quantify IHC staining, slides are digitally scanned and staining intensity can be quantified using morphometric software. Next, druggable targets can be identified and treated with targeted therapies. Fig. 2B shows IHC staining of the mTORC1 substrate phospho-S6 (P-S6) is elevated in lung tumors isolated from Kras/Lkb1 mutant GEMMs. The signaling pathway and several druggable targets are summarized in Fig. 2C. Our laboratory has performed pre-clinical in vivo studies in GEMMs targeting lung tumors that have hyper-active mTORC1 signaling with the mTOR inhibitor MLN128 (Momcilovic et al., 2015). This study was conducted using a combined analysis of ¹⁸F-FDG PET to measure glucose metabolism and IHC staining of mTOR substrates P-S6 and P-4EBP1 to measure inhibition of mTOR signaling. We detected lower ¹⁸F-FDG signal in tumors that received MLN128 and this correlated with lower P-S6 staining in the same tumors compared to untreated tumors (Fig. 2D). Further validating this approach, ¹⁸F-FDG PET was successfully used to guide a study targeting NF1 mutant tumors. Here the authors mapped both the mTOR and MAPK signaling pathways as critical regulators of tumor metabolism and survival (Malone et al., 2014). NF1 mutant tumors were highly responsive to combined inhibition of both these signaling nodes resulting in reduced tumor growth and glucose uptake in tumors as measured by ¹⁸F-FDG PET.

Coupling PET and CT imaging to mass spectrometry

While noninvasive ¹⁸F-FDG PET imaging provides both spatial and temporal assessment of glycolysis at different stages of tumorigenesis, this modality does not inform us of the fate of glucose beyond early steps of glycolysis. Therefore, by coupling ¹⁸F-FDG PET imaging to SIRM analysis using LC-MS of ¹³C-glucose one can attain a comprehensive picture of glucose uptake and glucose utilization within the cell. Examples of this type of approach are highlighted in a study examining the role of Kras dependent metabolism in pancreatic ductal adenocarcinoma (PDAC) (Ying et al., 2012). Here the authors performed ¹⁸F-FDG PET on pancreatic tumors in a Kras^G12D driven PDAC mouse model followed by LC-MS analysis of ¹³C-glucose on mouse derived PDAC cell lines. Additional analysis using qIHC identified the MAPK signaling pathway as a critical driver of glycolysis in PDAC. Subsequent studies have shown the feasibility of in vivo metabolite analysis by infusing either mice or patients with ¹³C-glucose and isolating tumors for ex vivo LC-MS analysis (Davidson et al., 2016; Hensley et al., 2016).

Imaging mice with two different PET probes on consecutive days was performed using ¹⁸F-FDG and ¹⁸F-Glutamine in order to evaluate glucose and glutamine uptake in tumors. Venetti et al., demonstrated usefulness of ¹⁸F-Glutamine over ¹⁸F-Glucose in imaging gliomas in both mouse models and in human patients (Venneti et al., 2015). ¹⁸F-Glutamine had superior tumor-to-background ratio compared to ¹⁸F-FDG, allowing clear tumor delineation. Recent reports showed that both ¹⁸F-FDG and ¹⁸F-Glutamine or ¹¹C-Glutamine are consumed by EGFR mutant lung adenocarcinomas suggesting metabolic dependency on both glucose and glutamine in this subset of lung cancer (Hassanein et al., 2016; Momcilovic et al., 2017). Moreover, ¹⁸F-Glutamine and ¹¹C-Glutamine uptake correlated with levels of the glutamine transporter ASCT2 (also known as SLC1A5), which was detected by IHC.

Imaging tumor metabolism with hyperpolarized MRS

Hyperpolarized ¹³C-Pyruvate has been used in multiple preclinical studies of lymphoma (Day et al., 2007; Dutta et al., 2013), prostate cancer (Keshari et al., 2013), glioblastoma (I. Park et al., 2011; Chaumeil et al., 2012), pancreatic cancer (Rajeshkumar et al., 2015; Serrao et al., 2016) and breast cancer (Shin et al., 2017). Hyperpolarized ¹³C-Glutamine has also been used in pre-clinical models of prostate cancer (Canape et al., 2015), liver cancer (Cabella et al., 2013; Gallagher et al., 2008), glioma (Qu et al., 2011) and IDH1/2 mutant chondrosarcoma (Salamanca-Cardona et al., 2017). Similarly, use of hyperpolarized ¹³C-Glucose has been reported in mouse lymphoma cell line (Timm et al., 2015), mouse lymphoma and lung tumors (Rodrigues et al., 2014). While still early in development, MRS is a new technique that has great potential in clinical application. This was demonstrated by successful application in patients with prostate cancer in which ¹³C-pyruvate and its metabolic products lactate, alanine and bicarbonate were measured in tumors using MRS imaging (Nelson et al., 2013). Please refer to an excellent review by Salmanca-Cardona and Keshari (2015) for an in depth summary of nuclear polarization-enhanced MR imaging.

Profiling metabolic phenotypes in tumors with PET and optical imaging

Multi-modality imaging in preclinical mouse studies has routinely combined PET/CT with BLI bringing together sensitivity of PET to detect metabolic activity within tumors with sensitivity of BLI for detection of small number of cells. Pioneering studies by the Gambhir lab, demonstrated that combining ¹⁸F-FDG/CT imaging with BLI was an effective and sensitive method to study tumor metabolism (Deroose et al., 2007). Their laboratory has recently extended the application of PET and BLI multi-modality imaging in an orthotopic mouse model of human glioma to investigate a novel radiotracer, ¹¹C-DA-SA-23 that directly measures activity of pyruvate kinase M2 isoform (PKM2) in vivo (Witney et al., 2015). Our lab has successfully utilized ¹⁸F-FDG/CT and BLI imaging to identify early stage tumors in GEMMs and to monitor therapeutic response (Shackelford et al., 2013; Momcilovic et al., 2015). The addition of bioluminescence or fluorescence imaging probes to PET imaging can help delineate changes in tumor metabolism in response to therapy before any changes in tumor volume become apparent (Tseng et al., 2017).

Optical imaging represents a cost effective alternative to more costly approaches such as PET and MRS imaging. Moreover, recent studies have demonstrated that caged luciferin probes can be designed to detect metabolite or enzyme activity (Wehrman et al., 2006; Yao et al., 2007; Dragulescu-Andrasi et al., 2009; Hickson et al., 2010; Van de Bittner et al., 2010; Scabini et al., 2011; Jang et al., 2016; Lynes et al., 2017; Park et al., 2017). Optical imaging with bioluminescence requires the presence of reporters in cell lines and mouse models of cancer and at first glance appears not directly translatable into clinical applications. However, intraoperative fluorescent imaging is used in surgical oncology to aid in visualization of tumors and complete resection of the tumors and may have applications in cancer metabolism in the future (Vahrmeijer et al., 2013; van Oosten et al., 2013; Mondal et al., 2014; Landau, Gould, & Patel, 2016; Dworsky et al., 2017).

Multi-modality imaging of tumor metabolism in cancer patients

Is it possible to translate MMI imaging from pre-clinical studies in mice to cancer patients? Yes. This was recently demonstrated in a seminal study by the DeBerardinis laboratory in which they profiled glucose metabolism in human lung cancer patients using MMI coupled to detailed metabolic and molecular analysis (Hensley et al., 2016). Here the authors coupled ¹⁸F-FDG PET and MRI with LC-MS and IHC to comprehensively profile tumor metabolism in patients undergoing surgical resection of tumors. The work-flow of this MMI analysis is represented in Fig. 3. The authors utilized ¹⁸F-FDG PET to inform them of the tumor’s glucose uptake and MRI with contrast to differentiate well-perfused regions from avascular hypoxic regions. The patients were infused with ¹³C labeled glucose followed by LC-MS metabolomic analysis of resected tumor. In addition, IHC staining and gene expression analysis were performed on the tumor to score parameters such as expression of metabolic enzymes, tumor histology, tumor grade and tumor proliferation index. What the authors discovered was that tumor metabolism is heterogeneous within a single tumor and across individual tumors. They found that well perfused lung tumors utilized glucose and alternative nutrients while tumors with low perfusion relied primarily on glucose (Fig. 3).

Multi-modality imaging to guide therapy

Can profiling tumor metabolism improve cancer treatment? Tumor metabolism is recognized as a hallmark of cancer therefore characterizing the metabolic phenotype of the tumor is an important first step in diagnosis and treatment. Tumor metabolism informs us of the anabolic growth rate of the tumor and its capacity for continued growth. Importantly, metabolic profiling provides functional analysis of the tumor. Furthermore, coordinating PET with LC-MS and qIHC will serve to identify metabolic dependencies and druggable proteins within tumors that can be inhibited with targeted therapies or even combined with immune based therapies. Metabolic profiling performed on human NSCLC revealed tumor metabolism is heterogeneous, suggesting that combination treatments will be required. However, in order for metabolic therapies to successful target a tumor, first the metabolic dependencies must be identified. If a tumor uses a multitude of metabolites as its fuel source then a treatment strategy that inhibits only one metabolite such as glucose is predicted to fail. By profiling tumors noninvasively with MMI and validating with metabolomics, researchers and clinicians will be able to take a snap shot of the tumor’s metabolic dependencies. This will allow for design of precise therapies that inhibit key metabolic nodes. The noninvasive nature of PET and MRS imaging means that repeat imaging will be possible with the goal of this imaging informing oncologists and radiologists of treatment efficacy and course of action.

This was demonstrated recently in Kras^G12D driven GEMMs of lung cancer treated with the glutaminase (GLS) inhibitor CB-839. As a single therapy, CB-839 failed to have any impact on tumor growth because these tumors are not dependent upon glutamine metabolism (Davidson et al., 2016). Kras^G12D driven tumors readily adapt their metabolism to treatment and oncogenic stress as shown in a study examining metabolic response to Kras^G12D ablation in a PDAC mouse model. Here the authors demonstrated that tumors adapt to inhibition of glucose metabolism by upregulation of oxidative mitochondrial metabolism following suppression of oncogenic Kras^G12D (Viale et al., 2014). Conversely, Kras^G12D driven tumors treated with biguanides, which are mitochondria complex I inhibitors that suppress oxidative phosphorylation (OXPHOS), induced up regulation of glycolysis as a compensatory mechanism that is readily detected by ¹⁸F-FDG PET imaging (Dykens et al., 2008; Shackelford et al., 2013). Combined inhibition of both glycolysis and oxidative phosphorylation potently suppressed Kras^G12D driven lung and pancreatic tumors (Momcilovic et al., 2015; Viale et al., 2014).

Distinct populations of tumors such as lung adenocarcinomas driven by EGFR mutations or triple negative breast cancer have shown dependency on glutamine metabolism in vivo and responsiveness to CB-839 treatment (Gross et al., 2014). PET imaging based profiling of EGFR mutant lung tumors with ¹⁸F-FDG and ¹⁸F-Glutamine or ¹¹C-Glutamine radiotracers suggested a dual dependency on both glucose and glutamine in vivo and predicted sensitivity to combined inhibition of glucose and glutamine metabolism (Qu et al., 2012; Hassanein et al., 2016; Momcilovic et al., 2017). Treatment of EGFR mutant lung tumors with the EGFR inhibitor erlotinib in combination with the GLS inhibitor CB-839 induced a potent anti-tumor response and tumor regression in mouse xenografts (Momcilovic et al., 2017). Additionally, a recent study showed that ¹⁸F-glutamine was a predictive therapeutic biomarker of response in mouse models of BRAF mutant colorectal cancer (CRC) to single or combination inhibitors against the BRAF and PI3K signaling pathways (Schulte et al., 2017).

The importance of PET imaging in determining response to therapy was elegantly demonstrated using ¹⁸F-FAC, a deoxycytidine kinase analog that is phosphorylated by deoxycytidine kinase (dCK) and incorporated in DNA synthesis pathway (Radu et al., 2008; Laing et al., 2009). ¹⁸F-FAC had a better selectivity for lymphoid organs compared to ¹⁸F-FDG and allowed for stratification and precise targeting of acute lymphoblastic leukemia (ALL) using targeted therapies against the ataxia telangiectasia and Rad3-related (ATR) protein and the dCK enzyme (Le et al., 2017). Two additional probes can be used to determine activity of enzymes involved in DNA salvage pathway. Tumor uptake of ¹⁸F-FMAC relies on activity of cytidine deaminase (CDA) and uptake of ¹⁸F-FLT is dependent on activity of thymidine kinase (TK) (Grierson and Shields, 2000; Lee et al., 2012). The importance of ¹⁸F-FAC probe was further demonstrated in metabolomics based study in which LC-MS and ¹⁸F-FAC uptake in liposarcomas identified dependency on nucleoside metabolism and successfully predicted sensitivity to gemcitabine treatment (Braas et al., 2012).

Open questions, challenges and future directions

Both PET and MRS imaging are costly ventures and often are not feasible on an academic budget. Can we identify and develop cost effective strategies to expand imaging in cancer metabolism in both pre-clinical and clinical studies? Development of optical imaging probes may provide a cost effective alternative to measure metabolism in tumors. However, development and careful validation of these probes will be required. Since caged luciferin probes rely on reporter-based systems their applications are restricted to pre-clinical studies. Combining multiple imaging platforms such as PET and MRI into single imaging units will likely increase the demand and throughput for pre-clinical imaging. Lastly, the availability of affordable radiotracers and hyperpolarized probes as well as micro-PET and MRS scanners hold promise to expand probe development and usage of MMI in the field of cancer metabolism. Miniaturized radiotracer development platforms such as ELIXYS may provide critical first steps towards wider availability of probes for basic and clinical research (Lazari et al., 2013).

Regardless of the imaging modality used, imaging of bulk tumors reflects a total sum from all cells in the tumor. Quantitative analysis within a heterogeneous tumor microenviroment presents a cadre of challenges. For example, what are the contributions of probe uptake for tumor supportive cells, such as stroma and tumor infiltrating leukocytes (TILs)? The recent success of immune based therapies in cancer have ushered in a wave of studies that have begun exploring the intersection of metabolism between the tumor and its microenvironment. It will be important to consider what proportion of the imaging signal is contributed by the tumor microenvironment contribute when imaging tumors with PET or MRS and validating with SIRM. Comparing the signal from bulk tumor to that of isolated tumor and immune cells may provide answers to these important questions.

Metabolism is a fundamental cornerstone of PET and MRS imaging and represents a powerful tool to study tumor metabolism in basic research and clinical settings. Tumors are heterogeneous by nature and their metabolism reflects this. Since tumors utilize a host of metabolites beyond glucose it is conceivable that the future of PET and MRS imaging in patients will utilize a multitude of radiotracers and ¹³C labeled metabolites to provide a detailed and real time signature of the tumor’s metabolic profile. This signature promises to provide valuable information to guide precise and personalized treatments. The likely future of cancer treatment is to manage the disease chronically with cocktail therapies similar to what has been done for HIV patients. Here, MMI guided metabolic profiling would provide oncologists, radiologists and pathologists with a roadmap to how tumors adapt and how we can stay one step ahead with effective therapies.