Innovation in Metabolomics to Improve Personalized Healthcare

Introduction

The Human Genome Project and the subsequent publication of a reference human genome sequence[1] represent milestones in biology research in general and in cancer in particular. In the last decade, metabolomics has contributed greatly to the understanding of cancer biology. In analogy to genomics and proteomics, the term metabolomics can be defined as the study of the complete ensemble of all small molecules (molecular weight (MW) < 1500 Da) formed by numerous biosynthetic and catabolic pathways within a biological system or originating from host-specific microbes and the intake of food nutrients and pharmaceuticals, which are present in a cell, tissue, or biofluids such as urine,[2] blood,[3] or saliva,[4] in the context of a physiological or pathological condition.[5

In 2004, the Human Metabolome Project (HMP), the equivalent of the Human Genome Project for metabolomics, was created as an inventory of 2500 small molecules produced by metabolic reactions in the body's tissues and biofluids.[6] The publication of the third version of the Human Metabolome Database (HMDB)[7] presented a comprehensive, web-accessible metabolomics database that brings together quantitative chemical, physical, clinical, and biological data on about 40,000 experimentally detected and biologically expected human metabolites.[7]

The major challenges of metabolomics stem from its advantages. While genomics and proteomics involve the study of molecules that are chemically similar or at least comparable, metabolomics deals with structurally heterogeneous and physico-chemically diverse small molecules. The range of their concentration spans up to nine orders of magnitude,[8] posing additional technical obstacles in terms of dynamic range and comparability. These small molecules or metabolites include compounds differing in chemical properties and function. These include, but are not limited to, lipids, sugars, ions, metabolic intermediates, and products of biochemical reactions, as well as building blocks for all other biochemical species including proteins, nucleic acids, and cell membranes.

Technology development, as well as new methods of data analysis,[10] has played a key role in driving the field of metabolomics. New methods and instrumentations, as well as incremental improvements in efficiency and sensitivity, have been fundamental for achieving the remarkable throughput and performance of this technology. Different techniques have been available to investigate the metabolome, distinguishing the different metabolites on the basis of their chemical and physical properties.

This review focuses on the rapidly developing analytical technologies, such as nuclear magnetic resonance (NMR), mass spectrometry (MS), Fourier-transform infrared (FT-IR), and Raman spectroscopy, but also discusses the most important steps in the workflow of metabolomic research.

Metabolic profiling in cancer is discussed using prostate cancer (PCa) as a paradigm. Specifically, cross-validation across high-throughput platforms has allowed findings from one type of biological data, expression profiling, to be confirmed with metabolomics, whereby specific genetic alterations are shown to drive unique metabolic programs. Thus, metabolites can be used as discovery tools for the identification of targetable metabolic enzymes or as biomarkers of genetic subtypes of PCa.

Techniques utilized in the assessment of metabolites

Metabolic profile of normal prostate cells and prostate cancer

Otto Warburg reported that, in normoxic conditions, cancer cells convert glucose to lactate at a higher rate than normal cells (the so-called Warburg effect).[17] Although glycolysis is less efficient for energy supply than aerobic respiration, it is 100 times faster and provides intermediate metabolites from amino acid and pentose phosphate production necessary for highly proliferating cancer cells.[18] Although increased aerobic glycolysis is a common feature in several malignances, in PCa it is not often present, except in tumors that are purely PI3K driven.[19] Altered lipid metabolism is instead a common feature of both primary and advanced PCa.[20] Finally, increased lipids, whether synthetized de novo or taken up from the circulation, can be utilized for fatty acid oxidation as an important bioenergetics source to support cell proliferation and growth.[21, 22] Recently, Raman spectroscopy was used to trace the metabolism of PCa in single living cells with high spatial–temporal resolution,[23] confirming the dominant fatty acid uptake over glucose uptake in prostate cells.[24

The terminal steps in the de novo biogenesis of fatty acids are catalyzed by the key lipogenic enzyme fatty acid synthase (FASN), a metabolic oncogene in PCa.[25, 26] Higher concentrations of protein and mRNA were found in PCa[20, 27] and associated with aggressive biological behavior.[27] Importantly, the highest levels of FASN expression are present in androgen-independent bone metastasis.[20] As mentioned above, unlike most tumors, PCa does not show the characteristic glycolytic switch. Thus, primary tumors are not efficiently detectable with analogs of glucose like ¹⁸F-fludeoxyglucose (FDG) in PET. Alternative tracers for PET imaging such as ¹¹C-acetate and ¹¹C-choline have been successfully used to detect increased lipid synthesis in PCa cells.[28

A peculiar feature of normal prostate cells is the accumulation and secretion of citrate due to the inhibition of the mitochondrial enzyme m-aconitase, which catalyzes the first step of citrate oxidation. This enzyme is inhibited by high concentrations of zinc in the prostate (which usually present 10- to 20-fold higher than other organs). The concentration of zinc decreases when prostate cells undergo neoplastic transformation with a resultant activation of m-aconitase and citrate oxidation. Lower concentration of citrate in PCa was reported both in seminal/ prostatic fluid[29, 30] by NMR and in prostatic tissue[31] by MRS spectroscopy. Citrate showed a stronger correlation with PCa and Gleason score when associated with other biomarkers (e.g., choline, creatine, or spermine) by in vivo MRSI[32, 33] and by ex vivo HR-MAS.[33, 34] However, stroma surrounding primary PCa and bone-enveloping metastatic disease contribute to the difficulty in analyzing PCa samples by HR-MAS. Recently, hyperpolarized ¹³C-pyruvate has been used in MRSI to characterize metabolic alteration in PCa patients.[14] The researchers evaluated the distribution of [1-¹³C]pyruvate and its metabolic products lactate, alanine, and bicarbonate in a time range of seconds. The results were extremely promising, showing elevated [1-¹³C]lactate/[1-¹³C]pyruvate in regions of biopsy-proven cancer.

Sarcosine has been found greatly increased by FT-MS during progression from normal tissue to PCa and metastatic disease, suggesting a key role in cancer cell invasion and aggressiveness.[35] Although the initial controversy surrounding the use of sarcosine as a biomarker was due to the difficulty in differentiating it from alanine in GC-MS analyses and the absence of internal analytical validation in the first studies,[36] sarcosine has been demonstrated to have a role in promoting PCa growth and progression, using both in vitro and in vivo models.[37] Quantitative measures of sarcosine, together with alanine, glycine, and glutamate, are used in the new commercially available test Prostarix™ to stratify PCa risk. More recently, sarcosine levels have been associated with MYC-driven tumors, suggesting it may be a biomarker of a molecularly defined category of PCa.[19

Formalin-fixed paraffin-embedded metabolomics

The availability of frozen tissues is often limited. Formalin-fixed paraffin-embedded (FFPE) tissue is routinely used in the diagnostic setting, is clinically annotated in databases with long-term follow-up and outcome, and exhibits long-term stability when stored at room temperature. The ability to use these samples would be of great benefit for PCa studies and could help in the discovery and validation of clinically useful biomarkers. FFPE tissue collections, accompanied by patient information as well as other molecular determinants, indeed represent invaluable resources for translational studies in oncology and other fields.

Yuan et al. recently demonstrated that stored FFPE tissue samples could be used for tumor classification and metabolic pathway analysis.[38] The authors described the use of a mass spectrometer to profile endogenous polar metabolites by methanol extractions from biological samples, without further sample manipulation. The target metabolites covered most metabolic pathways, such as glycolysis, the tricarboxylic acid cycle, the pentose phosphate pathway, and the metabolism of amino acids and nucleotides. Another study examined the technical feasibility and reproducibility of using targeted LC coupled with tandem MS (LC-MS/MS) to profile FFPE samples.[39] In this study, the authors profiled a set of five FFPE soft tissue sarcoma specimens and five paired normal tissue samples by LC-MS/MS. We recently were able to detect almost 400 metabolites in FFPE PCa cells and tissues, including lipids, amino acids, carbohydrates, nucleotides, cofactors, and vitamins (Cacciatore et al., unpublished observation). Interestingly, metabolomics signatures obtained from these samples were able to discriminate tumor from adjacent normal tissues.

However, the metabolic analysis of FFPE is limited by sensitivity of the mass spectrometer, and it does not allow the detection of some classes of metabolites. Moreover, steps in the procedures of FFPE, such as the time used to fix the tissue, can strongly affect the concentration of some metabolites. Unfortunately, the time of fixation is usually not available. The ability to perform such analyses in FFPE tissues paves the way for metabolomics-based biomarker discovery and validation using large retrospective and clinically well-defined FFPE sample collections.

Oncogenes and metabolic profiles

Cancer cells may overcome growth factor dependence by deregulating oncogenic and/or tumor suppressor pathways that affect their metabolism, or by activating metabolic pathways de novo with targeted mutations in critical metabolic enzymes. The oncogene MYCregulates several aspects of cellular biology including anaerobic glycolysis[40] and glutaminolysis.[41, 42] Recently, distinct lipid profiles in high and low MYC expression states were observed in lymphomas,[43] suggesting a relationship between the lipid metabolism and the overexpression of MYC.

In a recent study,[19] the metabolic reprogramming of two different oncogenes (AKT and MYC) was characterized in cells, murine models, and human PCa. MS-based metabolite profiling was performed on immortalized human prostate epithelial cells transformed by AKT1 or MYC, transgenic mice driven by the same oncogenes under the control of a prostate-specific promoter, and human prostate specimens characterized for the expression and activation of these oncoproteins. Integrative analysis of these metabolomic datasets revealed that AKT1 activation was associated with accumulation of aerobic glycolysis metabolites, whereas MYC overexpression was associated with dysregulated lipid metabolism. These data show how prostate tumors undergo a metabolic reprogramming that reflects their molecular phenotypes, with implications for the development of metabolic diagnostics and targeted therapeutics.

De novo fatty acid synthesis at the mitotic exit

Cellular metabolism plays a key role in cancer cell proliferation. The high rate of growth in cancer cells is accomplished by the activation of growth signaling and metabolic pathways allowing greater nutrient uptake and an increase in macromolecular biosynthesis. Transition from G₁ to S is regulated by the complex interplay between G₁ cyclins, cyclin-dependent kinases (CDKs) and their inhibitors, the retinoblastoma proteins (pRBs), and E2F-dependent events, but the effects of the metabolic reprogramming on the cell cycle remain elusive.[44]

Using MS-based metabolomics, a decrease in lysophospholipids was observed during the transition from G₂/M to G₁ phase despite an increase in the de novo synthesis of fatty acids and phosphatidylcholine,[45] suggesting that enhanced membrane production was related to a decrease in its turnover. The arrest of the cell at G₂/M for the inhibition of fatty acid synthesis indicates that the membrane production starts before G₁ and it is necessary to complete cellular division. Importantly, it also highlights evidence of closer cross talk between metabolism and the cell cycle, showing that the inhibition of fatty acid synthesis can block cell growth at G₂/M with high concentration of fatty acids in the media. This study suggested that the cell cycle completion is fundamental for de novo lipogenesis. This “lipogenic checkpoint” should be investigated further for therapeutic application in PCa. This study underlines the importance of metabolomics as a tool to view the biochemical changes during cell proliferation in cancer.

Conclusion

Metabolomics aims to acquire robust and reproducible quantitative information, but there is not a unique platform able to capture the entire metabolome.

MS is the most sensitive technique and can be used for the quantification of a wide spectrum of known molecules. Ultrahigh-accuracy mass spectrometers allow the detection and quantification of thousands of metabolites in a single-run experiment. NMR spectroscopy has a very limited sensitivity but it can extract information from a range of molecules normally difficult to assess by MS, such as lipoprotein particles in blood. In addition, NMR is a nondestructive technique and it can be applied successfully to the measure of metabolites in intact tissue or even in vivo in the whole body. NMR spectroscopy and MS are complementary techniques, and both methods are needed to obtain a comprehensive view of the metabolome.

On the other hand, FT-IR and Raman spectroscopy are rapid, robust, and highly reproducible analytical techniques, which detect only a subset of the metabolome without providing the concentration of single metabolites. Emerging techniques for imaging biological tissues based on Raman spectroscopy (e.g., stimulated Raman scattering microscopy) show considerable potential for the detection of specific classes of components (e.g., lipids) in vivo in single cells with high spatial–temporal resolution.

Most metabolomic research on PCa has focused on relatively small groups of metabolites that have long been known to be relevant to the prostate, such as citrate, choline, and polyamines. Efforts at more global profiling of the prostate metabolome are increasing, and are benefitting from advances in bioinformatics that accompany high-dimensional genomics and proteomics data analysis. Although metabolomic technologies have improved and evidence is accumulating to support their use in clinical decision making, the discipline is still in its infancy. The integration of the metabolomic with other high-throughput approaches will offer incredible opportunities, and large retrospective and clinically well-defined FFPE sample collection will be useful to bring us closer to the goal of personalized medicine.