Advances and Perspectives in Prostate Cancer Biomarker Discovery in the Last 5 Years through Tissue and Urine Metabolomics

Prostate cancer (PCa) is the second most diagnosed cancer in men worldwide. For its screening, serum prostate specific antigen (PSA) test has been largely performed over the past decade, despite its lack of accuracy and inability to distinguish indolent from aggressive disease. Metabolomics has been widely applied in cancer biomarker discovery due to the well-known metabolic reprogramming characteristic of cancer cells. Most of the metabolomic studies have reported alterations in urine of PCa patients due its noninvasive collection, but the analysis of prostate tissue metabolome is an ideal approach to disclose specific modifications in PCa development. This review aims to summarize and discuss the most recent findings from tissue and urine metabolomic studies applied to PCa biomarker discovery. Eighteen metabolites were found consistently altered in PCa tissue among different studies, including alanine, arginine, uracil, glutamate, fumarate, and citrate. Urine metabolomic studies also showed consistency in the dysregulation of 15 metabolites and, interestingly, alterations in the levels of valine, taurine, leucine and citrate were found in common between urine and tissue studies. These findings unveil that the impact of PCa development in human metabolome may offer a promising strategy to find novel biomarkers for PCa diagnosis.


Introduction
Cancer diseases are one of the most important health problems worldwide, prostate cancer (PCa) being one of the most prevalent. Indeed, PCa is globally the second most frequently diagnosed male malignancy and the fifth leading cause of cancer, with more than 1,000,000 new cases and more than 350,000 deaths, each year [1]. PCa is a heterogeneous disease [2] with a broad spectrum of aggressiveness, going from indolent PCa, which is a non-life-threatening cancer, to metastatic PCa with a 5-year survival of 28% [3].
Currently, PCa screening is based in serum prostate specific antigen (PSA) test and digital rectal examination (DRE) [4], whereas prostate biopsy (PB) is mandatory for a final diagnosis [5]. High levels of PSA (>4 ng/mL) are considered a sign of PCa [4]. However, this biomarker shows important limitations [6], due to its reduced accuracy (accu) (62-75%) [7], sensitivity (sens) (20.5%), specificity (spec) (ranging from 51% to 91%) [4,8], and area under the curve (AUC) (varying from 0.53 to 0.83) [7]. These low performance values can be due to interference from other diseases, like benign prostate hyperplasia (BPH) or prostatitis, that may also lead to an increase in serum PSA levels [2,6].  Nicholson et al. (1999) defined metabonomics as "the quantitative measurement of the dynamic multiparametric metabolic response of living systems to pathophysiological stimuli or genetic modification" [19]. Nowadays, the term metabonomics is often used interchangeably with the terms metabolic phenotyping, metabolic profiling, or simply metabolomics in the context of the comprehensive analysis of all metabolites of a biological sample representative of an organism or cell. Metabolomics is the last "omic" platform in the "omics" cascade (genomics-transcriptomics-proteomics-metabolomics), and it focuses on the study of small molecules (<1500 Da) [20] in several complex matrices like serum, saliva, exhaled air, urine, tissue, among others [21]. When compared with other omics, metabolomics shows important advantages: (i) the dynamic feature of metabolome, once it modifies rapidly in response to changes in cell status, allowing a continuous evaluation of the cell state [22]; (ii) minor changes in gene expression or protein synthesis are translated into major alterations in metabolite levels [23]; (iii) the response of metabolome to pathophysiological alterations is much more sensitive than gene or protein response [24]; (iv) the alterations in metabolome are closely related with the observed phenotype; (v) the levels of several metabolites can simultaneously be measured, allowing to establish a pattern of alterations associated with an specific pathophysiological state [22]; (vi) allows to define patterns of disease progression [25].

Metabolomic Approaches to Biomarker Discovery
Human metabolome comprises metabolites of low molecular weight from very different chemical families, such as amino acids, lipids, nucleotides, carbohydrates, organic acids, among others. They are present in a wide range of concentrations and have distinct physicochemical characteristics [26]. When a metabolomics study is designed, the selection of the analytical technique is a critical step, once this choice will restrict the metabolites detected and consequently the obtained results [27]. This selection needs to take into consideration the characteristics of the analytical technique like sensitivity, resolution, limits of detection of the instrumental technique [27], but also the characteristics of the samples and of the metabolites of interest, e.g., metabolite physicochemical properties and abundance [27,28].
Currently, the majority of the metabolic studies are performed using mass spectrometry (MS), frequently coupled with a separation technique like gas or liquid chromatography (GC-MS or LC-MS), and nuclear magnetic resonance spectroscopy (NMR) [26,29]. MS and NMR show several differences, including in the detected range of concentrations, namely, MS allows the detection of metabolites in concentrations ranging from picomolar (pM) to millimolar (mM) [30] and NMR from micromolar (µM) to millimolar (mM) [31]. Table 1 summarizes the advantages and limitations of the three analytical techniques (GC-MS, LC-MS and NMR), for metabolomic studies. As depicted in Table 1, none of these methods are able to cover the entire metabolome. For example, GC-MS is only suitable for the analysis of thermally stable compounds, such as volatile organic compounds (VOCs) [26]. In turn, LC-MS is used for profiling of compounds with medium and low polarities (reversedphase LC) and polar compounds (hydrophilic-interaction LC), but the datasets generated are complex, spectrometer dependent, and require additional MS/MS experiments, as well as spiking with authentic standards, in order to perform metabolite annotation and identification [26,32,33]. NMR shows a lower sensitivity, which compromise the detection of low abundance metabolites. Importantly, due to NMR nondestructive nature, the samples can be recovered after analysis and used in complementary studies (e.g., MS analysis) to obtain a more comprehensive characterization of the metabolome [34]. Indeed, the combination of more than one analytical platform is desirable to allow a more comprehensive analysis of a sample metabolome [26,29]. Metabolomic studies can follow two distinct approaches, namely the untargeted or the targeted approach. In the first, the goal is to cover the maximum of the metabolome detecting as many metabolites as possible in a matrix, and is frequently denominated as hypothesis generation [23]. In the second, a single metabolite or a group of metabolites (e.g., metabolites from a specific metabolic pathway) are previously selected and all the study is designed to detect and quantify these metabolites. This approach can be applied to validate the results obtained through an untargeted approach and is called hypothesis-driven [23,25].
Regarding PCa metabolomic studies, two main goals are recognized: (i) the discovery of biomarkers with high sensitivity and specificity for PCa timely detection and (ii) to understand the metabolic basis of PCa pathogenesis identifying altered metabolic pathways in consequence of PCa development and progression [25]. Nevertheless, the potential application of metabolomic studies is not limited to these two main goals, once metabolomic studies can also be applied to study the effectiveness of treatments, as well as the mechanism of action of therapeutic drugs and the mechanism of drug resistance or contribute to achieve the goal of personalized medicine [38].
Over the years, several independent subareas emerged from metabolomics, like volatilomics, lipidomics, among others. Volatilomics is based on the analysis of VOCs, like aldehydes, ketones, alcohols, hydrocarbons, or aromatic compounds [39], that are produced by human body and released into breath, blood, sweat, urine, feces, or saliva [39,40]. All VOCs share some physicochemical characteristics, such as low molecular weight and low boiling point and/or elevate vapor pressure in normal conditions [41]. The interest to investigate VOCs as potential cancer biomarkers gained strength after the observation that dogs were able to "smell" urine or skin samples of cancer patients with high sensitivity and specificity, indicating that the composition of VOCs is different in cancer individuals [42][43][44]. VOCs are end products of human biological activity and their composition in biological samples can reflect pathological processes [40], alterations in normal biochemical pathways and/or a response to a damage or disease. Indeed, cancer development and progression can lead to the production of new VOCs and/or to change their concentration [41], making them suitable candidates to cancer biomarkers [39]. One of the greatest advantages of VOCs as biomarkers is the possibility to easily, inexpensively and quickly detect them in clinical point of care through the most recent technological developments in biological sensors (e.g., electronic noses (e-nose)) [39].
Lipidomics is the subarea of metabolomics focused on the qualitative and quantitative profile of the lipid species in biological samples [45]. The knowledge of lipid metabolism is crucial to understand cancer development and progression for several reasons: (i) de novo synthesis provide phospholipids for cancer cell proliferation, (ii) fatty acid β-oxidation is important in energetics and redox homeostasis, (iii) lipids play an important role in signaling pathways [46] and, finally, (iv) lipids are extremely dynamic and can reflect physiological, pathological, and environmental alterations [47]. For these reasons, the interest to study the lipid profile of cancer cells has increased in the last years. It is estimated that mammalian cells comprise around 10,000 individual lipid species [48]. These lipids can be classified into different classes: (i) fatty acids, (ii) glycerophospholipids (GPLs), (iii) glycerolipids (e.g., triglycerides (TG)), (iv) saccharolipids, (v) sphingolipids (SL), and (vi) sterols. Each class of lipids show different biological functions. For instance, TG are important for energy storage, while sterols are key elements in cellular membrane and have also hormonal functions [49]. GPLs and SL are important components of cellular membranes and lysophospholipids (LPLs) (a subclass of GPLs) are important molecules for cellular signaling. These three classes (GPLs, SL and LPLs) are the most frequently studied in cancer lipidomic studies. GPLs can still be divided into phosphatidylcholine, phosphatidylethanolamine (major components of human cellular membranes), phosphatidic acid, phosphatidylglycerol, phosphatidylinositol and phosphatidylserine, considering the molecular structure of these molecules. SLs can also be divided into several subclasses like ceramides, sphingomyelins, among others [45]. This summary reflects the importance and the complexity of the lipidome and justify that lipidomics comprises an independent subarea of metabolomics. Furthermore, several studies revealed that cancer cells show alterations in lipidome fingerprint demonstrating the potential of lipids as biomarkers and/or therapeutic targets [46,49].

The Metabolic Phenotype of Prostate Cancer
It is well established that cancer cells suffer profound metabolic alterations that are indispensable for cancer development and progression [50]. One of the most well described metabolic alterations of cancer cells is the Warburg effect, which is characterized by a change in the preferential pathway to produce energy. Indeed, cancer cells preferentially produce ATP via aerobic glycolysis, even in the presence of oxygen, while normal cells produce ATP through oxidative phosphorylation [50,51]. This shift leads to an increase in glucose uptake and in lactate secretion [50,52]. The increase in lactate levels seems to play an important role in cancer development and progression [50]. Lactate can be utilized as fuel for oxidative metabolism, metabolized into alanine and glutamine and can also intervene in cancer cell mobility, immune escape and angiogenesis [50].
To comprehend how the Warburg effect impacts PCa cell metabolism, it is important to revisit the peculiar metabolic phenotype of normal prostate cells. Contrarily to other human cells, prostate cells favor citrate accumulation instead of citrate oxidation for energy production through tricarboxylic (TCA) cycle, also known as Krebs cycle or citric acid cycle [53]. Prostate cells have an increase in the zinc transporter ZIP1 and, consequently, zinc accumulates in prostate tissue [52]. The high levels of zinc are responsible for the inhibition of m-aconitase, which is the enzyme responsible for citrate oxidation in TCA cycle [53]. However, one of the first metabolic alterations associated with PCa development is the loss of cell ability to accumulate zinc and subsequent reduction of citrate levels in PCa cells [53]. Indeed, there is an increment of citrate oxidation in TCA cycle to produce energy in PCa cells [52,53]. For this reason, the Warburg effect and consequent increase in aerobic glycolysis is described mainly in advanced stages of PCa, where the increase in glycolytic pathway is associated with metastases formation and thereafter to a poor prognosis [52,54]. Furthermore, citrate can also be used in PCa cells to produce acetyl-coenzyme A (acetyl-CoA) (important for fatty acids and cholesterol synthesis) and oxaloacetate (amino acid precursor) [55] (Figure 1). Beyond citrate accumulation, normal prostate cells can also accumulate polyamines, such as spermine and spermidine once they are important components of prostatic secretions [56]. Polyamine levels also decrease, similarly to citrate, during cancer development and progression ( Figure 1). Indeed, this reduction in polyamine levels may promote PCa cell survival by preventing apoptosis [55,56].
or choline derivative metabolites, have a very important role in cancer cells proliferation [53,54]. Furthermore, lipids are also essential as energy resource, for energy storage and for intracellular signaling [54]. Therefore, increase in de novo fatty acids synthesis is an initial event in PCa development, which is stimulated by androgen signaling [52], as well as the increase in fatty acids oxidation to produce energy [52,54]. The importance of lipogenesis in PCa is patent in the increase of the expression of lipogenic and lipid-modifying enzymes, occurring in PCa [52,54] and by the accumulation of triglycerides, cholesterol esters and phospholipids (phosphatidylcholine), mainly in aggressive PCa [53]. Furthermore, metastatic PCa cells also show an upregulation of acetyl-CoA synthetase 2, allowing PCa cells to produce acetyl-CoA (essential for fatty acids synthesis) from acetate, while normal cells produce acetyl-CoA essentially from glucose and glutamine [54] (Figure 1). Moreover, PCa cells show the ability to take up exogenous lipids and to synthetize and mobilize lipids storage in other cells, like adipocytes [54].
From this brief explanation, it is reasonable to infer that the study of the metabolic signature of cancer cells has an enormous potential in the discovery of new biomarkers, as well as to elucidate cancer pathophysiological mechanisms, which can be used to define new therapeutic strategies.  Pentose phosphate pathway (PPP) is also altered in PCa cells, once the levels of glucose-6-phosphate dehydrogenase (a key enzyme in PPP) are increased through androgen receptor (AR) signaling [54], which is essential for PCa progression. AR signaling also promotes glycolysis and anabolism [55]. As previously referred, one of the most frequently used treatments for aggressive PCa is androgen deprivation therapy, which is associated to the development of castration-resistant state and consequently alterations in the lipid profile, and to a worse prognosis [55,57]. Furthermore, PCa cells show the ability to synthesize sterols, highlighting the importance of androgen signaling in PCa [57] (Figure 1).
Alterations in different amino acids, such as glutamine, have been associated with PCa and other cancers [50,52]. Glutamine is one of the most abundant amino acids in human plasma and has important roles in human metabolism [54], as it can be converted in glutamate, and subsequently be transformed in α-ketoglutarate, an intermediate in TCA cycle [50,52]. This amino acid can also be used by cancer cells for acetyl-CoA produc-tion [54], for fatty acid synthesis [52] and as a nitrogen and carbon donor for nucleotide, lipids and protein synthesis [50,54]. The glutamate resulting from glutamine is an essential substrate for glutathione synthesis, and therefore important for the protection of the cells against oxidative damage [50] (Figure 1). Arginine is an important amino acid involved in PCa metabolism. Arginine is converted by PCa cells in glutamine and/or proline [52]. The increase in proline levels is needed to the maintenance of the levels of pyridine nucleotides [54]. Arginine has also an important role in nitric oxide (NO) production [52] ( Figure 1). Sreekumar et al. (2009), reported higher levels of sarcosine in urine of PCa patients, which was a milestone in PCa metabolomics [58], but its importance as potential PCa biomarker was refuted in the following years [59][60][61]. Sarcosine is synthesized from other amino acids, glycine, and vice versa. This reaction can be linked to methionine cycle, and the produced methionine can be up-taken to folate cycle. The combination of these two cycles is referred frequently as one-carbon metabolism. One-carbon metabolism fuels building blocks for purines and thymidylates synthesis, which are essential for DNA synthesis and repair [62,63]. Methionine cycle also plays a role in polyamines and glutathione synthesis [54,63] (Figure 1).
Another prominent characteristic of a cancer cell is its ability to proliferate constantly. Lipids are major components of cellular membranes, so alterations in lipids, and in choline or choline derivative metabolites, have a very important role in cancer cells proliferation [53,54]. Furthermore, lipids are also essential as energy resource, for energy storage and for intracellular signaling [54]. Therefore, increase in de novo fatty acids synthesis is an initial event in PCa development, which is stimulated by androgen signaling [52], as well as the increase in fatty acids oxidation to produce energy [52,54]. The importance of lipogenesis in PCa is patent in the increase of the expression of lipogenic and lipid-modifying enzymes, occurring in PCa [52,54] and by the accumulation of triglycerides, cholesterol esters and phospholipids (phosphatidylcholine), mainly in aggressive PCa [53]. Furthermore, metastatic PCa cells also show an upregulation of acetyl-CoA synthetase 2, allowing PCa cells to produce acetyl-CoA (essential for fatty acids synthesis) from acetate, while normal cells produce acetyl-CoA essentially from glucose and glutamine [54] (Figure 1). Moreover, PCa cells show the ability to take up exogenous lipids and to synthetize and mobilize lipids storage in other cells, like adipocytes [54].
From this brief explanation, it is reasonable to infer that the study of the metabolic signature of cancer cells has an enormous potential in the discovery of new biomarkers, as well as to elucidate cancer pathophysiological mechanisms, which can be used to define new therapeutic strategies.

Tissue Metabolomic Studies
The collection of tissue samples is very invasive, hampering their use for PCa screening. However, the study of the tissue metabolome has important advantages, once this is the ideal matrix to establish which metabolic alterations are specific to PCa development and progression. Furthermore, tissue studies have been performed using matched tumoral and nontumoral samples from the same individual, thus minimizing the contribution of confounding factors (e.g., age, comorbidities, lifestyle).
Thirteen metabolomic studies performed in PCa tissue samples were published in the last 5 years, including two lipidomic studies. Table 2 summarizes the study design and main outcomes obtained in those studies. Overall, a total of 98 different metabolites were associated with PCa, indicating that PCa is related with dysregulations in 32 different metabolic pathways (Table 2). Interestingly, 18 metabolites were found to be common among the included studies ( Figure 2). It is important to note that these studies were performed under different analytical conditions, with different sample selection criteria and using different statistical approaches, foreseeing difficulties to compare results across studies. The fact that these metabolites were found common in the various studies, highlights their importance in PCa metabolism and their potential as specific PCa biomarkers.              The most frequent alteration reported among the studies conducted in the last 5 years is the significant decrease in citrate levels in PCa tissue [68,[74][75][76] (Table 2 and Figure 2). This result is not unexpected as the loss of capability to accumulate citrate is one of the first metabolic alterations observed in prostate cells during malignant transformation. This loss of capability to accumulate citrate translates in a profound alteration in energetic metabolism of PCa cells, once PCa cells start to use citrate in TCA cycle more efficiently than normal prostate cells [51,52]. Furthermore, Braadland et al. (2017) compared PCa tissue from men that suffered PCa recurrence after prostatectomy with tissue from men that, until the date of the study, did not show signals of recurrence, unveiling that lower levels of citrate in PCa tissue were associated with shorter time of recurrence [70]. Additionally, lower levels of citrate were also associated with more aggressive PCa [77]. Beyond citrate, two other metabolites involved in energetic metabolism, namely alanine [69,71] and lactate [69,72] showed significant alteration in PCa tissue.
The increased levels of other key metabolites of TCA cycle have also been frequently cited in the reviewed studies, namely succinate [71,76], malate [71,76] and fumarate [71,74,76] (Table 2, Figure 2). The increase in malate and fumarate levels was also correlated with Gleason score [71,78] and tumor stage [71]. Notably, both succinate and fumarate were previously considered oncometabolites [49], once their accumulation leads to cancer progression [79]. The increased levels of succinate and fumarate have been associated in other cancer types (e.g., paraganglioma, pheochromocytoma or kidney cancers [78]) with mutation in the enzymes succinate dehydrogenase (SDH) and fumarate hydratase (FH), respectively [49,80]. However, these results were not observed in PCa studies performed by Shao   The most frequent alteration reported among the studies conducted in the last 5 years is the significant decrease in citrate levels in PCa tissue [68,[74][75][76] (Table 2 and Figure 2). This result is not unexpected as the loss of capability to accumulate citrate is one of the first metabolic alterations observed in prostate cells during malignant transformation. This loss of capability to accumulate citrate translates in a profound alteration in energetic metabolism of PCa cells, once PCa cells start to use citrate in TCA cycle more efficiently than normal prostate cells [51,52]. Furthermore, Braadland et al. (2017) compared PCa tissue from men that suffered PCa recurrence after prostatectomy with tissue from men that, until the date of the study, did not show signals of recurrence, unveiling that lower levels of citrate in PCa tissue were associated with shorter time of recurrence [70]. Additionally, lower levels of citrate were also associated with more aggressive PCa [77]. Beyond citrate, two other metabolites involved in energetic metabolism, namely alanine [69,71] and lactate [69,72] showed significant alteration in PCa tissue.
The increased levels of other key metabolites of TCA cycle have also been frequently cited in the reviewed studies, namely succinate [71,76], malate [71,76] and fumarate [71,74,76] (Table 2, Figure 2). The increase in malate and fumarate levels was also correlated with Gleason score [71,78] and tumor stage [71]. Notably, both succinate and fumarate were previously considered oncometabolites [49], once their accumulation leads to cancer progression [79]. The increased levels of succinate and fumarate have been asso-ciated in other cancer types (e.g., paraganglioma, pheochromocytoma or kidney cancers [78]) with mutation in the enzymes succinate dehydrogenase (SDH) and fumarate hydratase (FH), respectively [49,80]. However, these results were not observed in PCa studies performed by Shao et al. (2018), which suggested the involvement of other mechanisms that could also be related with the increase of the levels of these metabolites in PCa [71]. Once fumarate is also linked with urea cycle [74,76], this metabolic pathway could be res-ponsible for keeping the high levels of fumarate in PCa tissue [74,79]. As previously referred, the accumulation of fumarate leads to cancer progression, this could involve the activation of hypoxia-inducible factor 1-subunit alfa (HIF1α) and NFκB pathways [74]. HIF1α plays an important oncogenic role in PCa once this pathway is responsible for many essential mechanisms to guarantee PCa cell survival, like antiapoptosis, angioge-nesis and increased glycolytic metabolism. Furthermore, HIF1α protects PCa cells against oxidative stress and against the cytotoxicity caused by androgen deprivation therapy, chemotherapy, or radiation [81]. Similarly, NFκB pathways support PCa cell survival, proliferation, and invasion, playing an important role in the development of resistance to castration therapy [82].
The increase in uracil levels is another alteration consistently reported in PCa tissue (Table 2, Figure 2), suggesting that PCa cells have alterations in pyrimidine metabolism [65,71,76]. Pyrimidine metabolism is a complex biochemical pathway that comprises diffe-rent reactions, namely de novo nucleotide synthesis, nucleoside salvage, and pyrimidines degradation [83]. Pyrimidines, like uracil [84], are essential in cells metabolism once they are constituents of nucleotides, nucleic acids, vitamins, proteins, and folates [85]. Furthermore, they are key intermediates in RNA and DNA synthesis, protein and lipids glycosylation, synthesis of phospholipid precursors [84,85], and in reactions of glucuronidation [84]. Cancer cells are dependent on de novo nucleotide synthesis for cell proliferation and consequently for cancer development and progression [83,84]. Importantly, the inhibition of this metabolic pathway is a strategy adopted in the treatment of several cancers (e.g., colorectal cancer and pancreatic cancer) [84,86,87].
One of the main functions of normal prostate is to synthesize polyamines like spermine but apparently this function is impaired with PCa development and progression [70] leading to a decrease in the levels of spermine [68,76]. Indeed, the reduction of spermine levels was proposed as a biomarker able to predict BCR [70]. Interestingly, levels of spermidine, a spermine precursor, were also reported as significantly altered in PCa tissue samples; however, the obtained results were contradictory. Huan et al. (2016) found a significant increase in the levels of spermidine [65], whereas Dudka et al. (2020) showed a significant decrease in the levels of this metabolite in PCa tissue samples [76].
As previously referred, lipid metabolism can be an important source of PCa biomarkers, emphasizing the relevance of lipidomic studies. The major reported lipidic alterations occurring in PCa cells involved phospholipids from cellular membrane [64,66,67,69,[71][72][73]76] which was expected taking into consideration that cancer cells show a high proliferative phenotype. Notably, the significant decrease in the levels of LPC (16:0) was able to predict BCR [64]. This observation is supported by a transcriptomic study, that evaluates the expression of the enzyme LPC transferase 1 (LPCAT1). The increase in the expression of this enzyme was able to discriminate PCa from benign tissue, as well as to differentiate PCa with different GS and to predict BCR and/or metastasis development [88]. Furthermore, phosphocholine also revealed to be able to discriminate PCa tissue with different GS [76].
Finally, alterations in amino acid metabolism have also been widely reported in PCa, mainly in the increase levels of glutamate [72,75,76], tyrosine [74,75], arginine [74,76], and proline [71,74]. Importantly, the significant alteration in the levels of the first three amino acids was associated with GS and consequently PCa aggressiveness [76], making these metabolites potential diagnosis and prognosis biomarkers.

Urine Metabolomic Studies
Urine is the ideal matrix to be used in a screening test, due to its noninvasive nature, along with ease of collection and handling, high volume which allow repeated analysis, and lower complexity when compared with other biofluids (e.g., serum or plasma) [20,37,89]. Furthermore, urinary metabolites are concentrated by the kidneys, which are anatomically close to the prostate [89,90]. However, urine composition can vary due to several external factors, like diet, smoking habits, genetic factors, microbiota, diurnal cycles diabetes, and other diseases which can affect urine metabolome [91].
From 2015 to 2020, 13 studies performed in PCa urine samples were published, including four volatilomics studies. Table 3 summarizes these studies, highlighting study design, altered metabolites, metabolic pathways, as well as candidate biomarkers, whenever available. Overall, 179 different metabolites were associated with PCa, indicating that PCa is correlated with dysregulations in 48 different metabolic pathways. In this section, the metabolites that hold greatest potential as PCa biomarkers will be highlighted, considering different selection criteria: (i) consistency among different urinary studies, (ii) AUC greater than PSA and (iii) translatability between tissue and urine studies.
Despite the great differences (e.g., different analytical platform, samples preparation or different inclusion/exclusion criteria) among the study designs, 15 metabolites have been consistently reported with the same variation among different urinary studies, as represented in Figure 3. Importantly, four metabolites out of the 15 have also been reported with the same alteration in PCa tissue, namely decreased levels of citrate [68,[74][75][76]92,100], increased levels of leucine [74,97,102], increased levels of valine [74,97,102], and increased levels of taurine [96,100,104], suggesting that these alterations may be specific of PCa tumors and suggesting their translatability between tissue and urine samples (Figure 3). Despite the great differences (e.g., different analytical platform, samples preparation or different inclusion/exclusion criteria) among the study designs, 15 metabolites have been consistently reported with the same variation among different urinary studies, as represented in Figure 3. Importantly, four metabolites out of the 15 have also been reported with the same alteration in PCa tissue, namely decreased levels of citrate [68,[74][75][76]92,100], increased levels of leucine [74,97,102], increased levels of valine [74,97,102], and increased levels of taurine [96,100,104], suggesting that these alterations may be specific of PCa tumors and suggesting their translatability between tissue and urine samples (Figure 3). In addition, 12 metabolites stood out once they unveiled similar or even better performance than PSA (AUC ranging from 0.53 to 0.83) for PCa detection [7], namely γamino-n-butyric acid (AUC: 0.  [102]. From these 12 metabolites, the alterations observed in the levels of three metabolites were previously reported in PCa tissues, namely spermine [68,76], ethanolamine [105] and glucose [76]. Importantly, the decrease in glucose levels was also correlated with GS [76]. The significant alterations observed in the other nine metabolites, to the best of our knowledge, were not previously reported in PCa tissue. It is important to highlight that even if it is not possible to prove translatability of a metabolite from tissue to urine, this does not invalidate its potential as PCa biomarker once its alteration can for example be driven from a systemic response to PCa development and progression.
Overall, findings from the reviewed studies showed that PCa development and progression is mainly associated with alterations in amino acid metabolism, energy metabolism, especially in TCA cycle, and membrane metabolism (Tables 2 and 3).

Current Challenges and Future Perspectives
There are no doubts that the scientific community has made enormous efforts to define the impact of PCa in human metabolome with dozens of studies focused on this topic, not only using tissue and urine matrices but also other biological samples as serum, plasma, seminal fluid, prostatic fluid, and even cell lines [18]. However, some biological and technical challenges should be addressed before we can translate all the potentialities of metabolomics into clinical practice.
The traditional paradigm is to find a single biomarker for PCa screening. However, during the last years, the idea of using a panel of biomarkers instead of a single biomarker has gained strength, especially in volatilomic studies. The use of a biomarker panel has important advantages, once a multi-biomarker panel may be able to capture more deeply the various metabolic dysregulations occurring during cancer development and progression than a single biomarker [106]. Hence, a multi-biomarker panel allows the definition of a more robust signature of PCa providing a better evaluation of cancer progression. Furthermore, the use of a biomarker panel avoids that an arbitrary change in a single metabolite leads to a false result [22].
As referred in the previous section, the comparison of the findings from different studies is compromised by the lack of standardized procedures in metabolomic studies, especially in study design which consequently increases interlaboratory variability [107]. Many efforts have been made to accomplish the goal of standardized procedures in metabolomics studies in the last years [107,108]. There is still a long way to go until the desired standardization, but the first steps have already been taken [108].
Other crucial technical challenge is metabolite identification. This is particularly true in volatilomics studies. Additionally, the interpretation of the urinary volatilome signature of PCa is particularly challenging once there is no clear understanding of the biological origin of VOCs [39]. In addition, the volatilome of PCa tissue has not been explored so far, to our knowledge, hindering the elucidation of a potential translatability of VOCs from tissues to urine samples. In future, this issue can be addressed through volatilomics studies of PCa tissue and fluxomic studies. Fluxomic studies allow the understanding of the metabolic origin of endogenous VOCs by labeling and tracking metabolic precursors (e.g., glucose), throughout the metabolic pathways [109].
Perhaps the greatest limitation to biomarker discovery relies on the fact that several metabolomic studies are essentially descriptive and skip the validation step. Indeed, the vast majority of the papers just list which metabolites are statistically different between the groups in study, not proposing candidate biomarkers and/or clearly state the performance of the proposed metabolites/biomarkers (e.g., AUC, sensitivity, specificity and/or accuracy), thus impairing the discussion of which would be the most promising biomar-kers for PCa. Hence, future studies should be less descriptive and more assertive, propo-sing and evaluating potential biomarkers. Furthermore, it is also important to include external sets for model/results validation to improve the robustness of candidate biomarkers and to include unambiguous biochemical and biological interpretation of PCa metabolic dysregulations. Remarkably, some of the studies included in this revision are already fo-llowing this direction. In addition, it is well known that especially urinary metabolic profile can be affected by several factors (e.g., diet, lifestyle, microbiota, race, among others) [107], so it is also crucial to perform studies in large and more heterogeneous populations (e.g., American, African, Caucasian), to ensure that the proposed biomarkers can be applied among different countries and different lifestyles.
To conclude, metabolomics is a powerful tool to uncover the metabolic signature of PCa development and progression. The results obtained so far in tissue and urine metabolomic studies unveiled potential to define new screening/diagnosis biomarkers.