From Glucose to Lactate and Transiting Intermediates Through Mitochondria, Bypassing Pyruvate Kinase: Considerations for Cells Exhibiting Dimeric PKM2 or Otherwise Inhibited Kinase Activity

A metabolic hallmark of many cancers is the increase in glucose consumption coupled to excessive lactate production. Mindful that L-lactate originates only from pyruvate, the question arises as to how can this be sustained in those tissues where pyruvate kinase activity is reduced due to dimerization of PKM2 isoform or inhibited by oxidative/nitrosative stress, posttranslational modifications or mutations, all widely reported findings in the very same cells. Hereby 17 pathways connecting glucose to lactate bypassing pyruvate kinase are reviewed, some of which transit through the mitochondrial matrix. An additional 69 converging pathways leading to pyruvate and lactate, but not commencing from glucose, are also examined. The minor production of pyruvate and lactate by glutaminolysis is scrutinized separately. The present review aims to highlight the ways through which L-lactate can still be produced from pyruvate using carbon atoms originating from glucose or other substrates in cells with kinetically impaired pyruvate kinase and underscore the importance of mitochondria in cancer metabolism irrespective of oxidative phosphorylation.


GLUCOSE AND LACTATE IN CANCER: BACKGROUND
It is a well-known fact that most cancers exhibit increased rates in glucose consumption (Bose and Le, 2018). This is clinically exploited by following radionuclide-labeled glucose analogs for the purpose of tumor imaging in living human beings (Feng et al., 2019). The very same cancers are also known to be major lactate producers, which is important for their survival (de la Cruz-Lopez et al., 2019). The combination of an increased consumption of glucose with an increase in lactate output led to the assumption that cancers exhibit an increase in glycolysis; although this is true, serving the purpose of generating glycolytic metabolites which are diverted toward biosynthetic processes (DeBerardinis et al., 2008) and NADPH by the pentose phosphate pathway (Icard and Lincet, 2012), most tumors express a dimeric form of the M2 isoform of pyruvate kinase which has been reported to be much less active than that found in healthy cells; furthermore, numerous posttranslational modifications and mutations have been reported for this gene product, leading to a much reduced activity but still fueling cancer aggression (see section "Pyruvate Kinase"). Even more so, tumor cells with undetectable levels of pyruvate kinase still producing lactate can be found in vivo (Israelsen et al., 2013). On one hand, the decrease in pyruvate kinase activity is important for maintaining a metabolite "traffic jam, " forcing upstream metabolites toward biosynthetic pathways; on the other hand, it points to a metabolic conundrum because L-lactate may only originate from pyruvate, a metabolite arising from phosphoenolpyruvate (PEP) through pyruvate kinase in glycolysis (see Figure 1). The purpose of this review is to highlight the pathways that can lead to pyruvate and lactate-even commencing from glucose-bypassing pyruvate kinase. This is important because (i) carbon-labeled atoms in glucose may appear in lactate without net ATP production from glycolysis and (ii) hints on the possibility that other pathways leading to pyruvate/lactate could be crucial for cancer cell survival that are perhaps amenable to pharmacological and/or genetic manipulation. The list of pathways appearing below has been assembled by mining the following databases: Kyoto Encyclopedia of Genes and Genomes 1 (Kanehisa and Goto, 2000), BRaunschweig ENzyme Database 2 (Jeske et al., 2019), Metabolic Atlas 3 (Robinson et al., 2020), Biochemical, Genetic, and Genomic knowledge base 4 (King et al., 2016), MetaNetX 5 (Moretti et al., 2016), Human Metabolome Database 6 (Wishart et al., 2018), and Virtual Metabolic Human 7 (Noronha et al., 2019).

PYRUVATE KINASE
Pyruvate kinase generates ATP at the "substrate level" in the absence of oxygen by catalyzing the dephosphorylation of PEP to pyruvate (see Figure 1). There are four isoforms denoted as L, R, M1, and M2. For details regarding kinetic properties, tissue distribution, and regulation, the reader is referred to the review by Israelsen and Vander Heiden (2015). In the present review, the PKM2 isoform will be specifically examined; for a more thorough evaluation, the reader is referred to Li et al. (2014Li et al. ( , 2018, Wong et al. (2015); Yang and Lu (2015), Dayton et al. (2016b), Hsu and Hung (2018), and Alquraishi et al. (2019). The non-enzymatic functions of PKM2 are examined elsewhere (Hoshino et al., 2007;Stetak et al., 2007;Luo et al., 2011;Yang et al., 2012;Yang and Lu, 2013).
Basically, PKM2 exhibits lower enzymatic activity compared to that by PKM1 (Yamada and Noguchi, 1999) and is allosterically regulated by fructose-1,6-bisphosphate (FBP); it exists either as a dimer with low affinity for PEP or as an FBP-bound tetramer with high affinity for PEP (Mazurek et al., 2005;Zhang et al., 2019). Although PKM2 has been branded as "the predominant isoform in cancer cells" (Altenberg and Greulich, 2004;Mazurek et al., 2005), further scrutiny in 25 human malignant cancers, six benign oncocytomas, tissue-matched controls, and several cell lines showed that "PKM2 dominance was not a result of a change in isoform expression, since PKM2 was also the predominant PKM isoform in matched control tissues." Therefore, a switch from PKM1 to PKM2 isoform expression during malignant transformation may not be taking place, as previously postulated (Christofk et al., 2008). Mindful of the controversy surrounding the proposed functions of PKM2 Harris and Fenton, 2019), the group of Vander Heiden characterized the effects of cancer−associated PKM2 mutations on enzyme kinetics and allosteric regulation and reported that a decrease in PKM2 activity supports the rapid proliferation of cells (Liu V. M. et al., 2020). This is in line with earlier reports showing that a decrease in PKM2 activity due to posttranslational modifications (Lv et al., 2011) or inhibition by oxidative stress (Anastasiou et al., 2011) promotes tumor growth (Prakasam et al., 2018). Alternatively, exposure to small molecule PKM2 activators or expression of the constitutively active PKM1 thwarts cancer cell proliferation (Anastasiou et al., 2012). Finally, it has been also shown that PKM2 is not even required for the growth of many cancers (Cortes-Cros et al., 2013;Israelsen et al., 2013;Wang et al., 2014;Lunt et al., 2015;Dayton et al., 2016aDayton et al., , 2018Lau et al., 2017;Tech et al., 2017;Hillis et al., 2018). In aggregate, the consensus seems to be that the lower the pyruvate kinase activity, the greater the stimulation of tumor growth. As discussed in the section below entitled "Evidence Showing That Pyruvate Kinase Inhibition Does Not Lead to a Proportional Decrease in Pyruvate/Lactate Formation, " even those cells exhibiting low-or even undetectable-pyruvate kinase activity still produce lactate, which begs the question: where does this lactate come from?

EVIDENCE SHOWING THAT PYRUVATE KINASE INHIBITION DOES NOT LEAD TO A PROPORTIONAL DECREASE IN PYRUVATE/LACTATE FORMATION
In Cortes-Cros et al. (2013), it was shown that knockdown of both PKM1 and PKM2 (PKM2 knockdown was on the order of > 95%) leading to an approximately fivefold decrease in overall pyruvate kinase activity yielded only a ∼50% decrease in the appearance of 13 C originating from glucose to lactate.
In Chaneton et al. (2012), silencing of both PKM1 and PKM2 to an extent greater than 90% led to only a ∼30% decrease in pyruvate and lactate production, while PEP concentration increased by 100%. FIGURE 1 | Biochemical pathways connecting glucose or other metabolites to pyruvate and L-or D-lactate. The box in magenta represents a mitochondrion. Glycolysis is highlighted in green. Metabolites found both inside and outside the mitochondria that are not connected with an arrow are highlighted in matching striped colors (to avoid arrow clutter). For abbreviations, see Table 1.
In Vander Heiden et al. (2010), it was shown that cancer cell lysates expressing no pyruvate kinase activity produced 50% of pyruvate from PEP compared with the total cell lysates. Although in this work it was postulated that phosphate from PEP is transferred to the catalytic histidine on human PGAM1, this claim was subsequently rejected by the same authors, attributing their earlier findings to contaminating ATP−dependent protein kinases .
In all of the abovementioned studies, it was assumed that, in view of severely diminished pyruvate kinase activity, pyruvate and lactate production is attributed to carbon sources other than glucose. Indeed Yu et al. (2019), determined that, in pancreatic ductal adenocarcinoma cells with PKM1 and PKM2 knockdown, cysteine catabolism generated ∼20% of intracellular pyruvate. The purpose of the present review is to not only outline these pathways but also show additional ways for obtaining 13 C labeling in pyruvate or lactate originating from glucose; furthermore, since some of these pathways involve intermediates that transit through the matrix, the role of the mitochondria is emphasized, which is unrelated to the concept of oxidative phosphorylation.

PATHWAYS LEADING TO PYRUVATE COMMENCING FROM GLUCOSE: INTERMEDIATES NOT TRANSITING THROUGH THE MITOCHONDRIA
The pathways shown in this section refer to Figure 2 (lavender arrows). Multiple arrows imply multiple biochemical steps.
while in rodent hepatomas it was found to be decreased (Weber and Cantero, 1955).
(3) Glc →→→ PEP → pyruvate: the terminal reaction is catalyzed by tartrate-resistant acid phosphatases (TRAP), the molecular identity of which remained unknown well after their biochemical characterization (Helwig et al., 1978;Chen and Chen, 1988;Hayman et al., 1989); they are most likely substantiated by a metalloprotein enzyme with the ability to catalyze the hydrolysis of orthophosphate monoesters under acidic conditions (Bull et al., 2002). The expression of this enzyme (TRAP) is a marker of bone disease in cancer patients (Nguyen et al., 1991;Koizumi and Ogata, 2002;Mose et al., 2003;Terpos et al., 2003;Chao et al., 2005).

PATHWAYS LEADING TO PYRUVATE COMMENCING FROM GLUCOSE: INTERMEDIATES TRANSITING THROUGH THE MITOCHONDRIA
These pathways depend on one or more of three critical parameters: (1) glyoxylate entry into the mitochondria, (2) reversibility of the matrix phosphoenolpyruvate carboxykinase (PCK2), and (3) reversibility of the mitochondrial pyruvate carrier (MPC). Regarding glyoxylate, I was unable to find information on its transport across the inner mitochondrial membrane; however, it is known that it can be processed by the matrix-localized AGXT2 (Kakimoto et al., 1969). PCK2 expression and activity level are critical for many cancer types: in tumor-initiating enriched prostate cancer cell clones, PCK2 was overexpressed, and this correlated with more aggressive tumors and lower survival rates (Zhao et al., 2017); in lung cancer cell lines and in non-small cell lung cancer samples, PCK2 expression and activity were enhanced under low-glucose conditions (Leithner et al., 2015); finally, it was reported that PCK2 is required for glucose-independent cancer cell proliferation and tumor growth in vivo (Vincent et al., 2015). Regarding PCK2 reversibility, the enzyme has been shown to operate in the reaction toward OAA synthesis in mitochondria from rabbit liver (Carlsen et al., 1988), pigeon and rat liver (Wiese et al., 1996), guinea pig liver (Garber and Ballard, 1970;Garber and Salganicoff, 1973), rabbit enterocytes (Wuensch and Ray, 1997), chicken liver (Hebda and Nowak, 1982;Makinen and Nowak, 1983;Wilson et al., 1983;Erecinska and Wilson, 1984), and bullfrog liver (Goto et al., 1980). However, in Vincent et al. (2015), it was shown that a fraction of pyruvate originated from glutamine from PEP through PCK2. With respect to the reversibility of the MPC, this is a working hypothesis because there are no data showing pyruvate release from normally polarized mitochondria. Nevertheless, this is not a far-fetched hypothesis: succinate and other metabolites are effluxed from the mitochondria for non-metabolic roles against a hyperpolarized membrane potential (Mills et al., 2016), demonstrating that this is possible under appropriate conditions. It may be also relevant that pyruvate catabolism through the pyruvate dehydrogenase complex is associated with suppression of tumor growth in vitro and in vivo (Michelakis et al., 2008); relevant to this, genes coding for both the pyruvate dehydrogenase complex and pyruvate carboxylase in certain cancers are usually downregulated (Yuen et al., 2016); furthermore, pyruvate is found in blood plasma, urine, and cerebrospinal fluid, and its presence there is not associated with damage of plasma membranes. Of course, this does not mean that extracellular pyruvate originated from the mitochondria, but it indicates that it can cross the plasma membrane through monocarboxylate transporters, some of which are distributed both in plasma and in the inner mitochondrial membrane (Hussien and Brooks, 2011); indeed monocarboxylate transporter 1, which is one of the four known pyruvate transport mechanisms, was recently shown to export pyruvate from the cell (Hong et al., 2016); however, mitochondrial pyruvate export remains hypothetical especially in view of the fact that its exit is influenced by the membrane potential and →pH. It was also recently reported that loss of an MPC isoform prior to a tumorigenic stimulus doubled the frequency of adenoma formation and produced higher-grade tumors, and this was associated with a glycolytic metabolic phenotype and increased expression of stem cell markers (Bensard et al., 2020). Mindful of the above, these pathways are as shown in Figure 3 (yellow arrows).
(11) Glc →→→ PEP which enters the mitochondria; PEP transport across the inner membrane of mammalian mitochondria has been demonstrated to occur by the tricarboxylate carrier by Robinson (1971) and the group of Soling et al. (1971) and Kleineke et al. (1973) and to a lesser extent by the adenine nucleotide carrier, shown by the Shug and Shrago (1973); Sul et al. (1976) and in Drahota et al. (1983) and reviewed in Passarella et al. (2003). The possibility of a PEP/pyruvate transporter has also been put forward (Satrustegui et al., 2007). More recently, PEP cycling via mitochondrial PEPCK evoking PEP transport across the inner mitochondrial membrane has also been demonstrated by the group of Kibbey (Stark et al., 2009); PEP → OAA by PCK2; OAA → pyruvate by reverse operation of PC. However, this is expected to be a very minor path. Pyruvate may exit the mitochondria through the MPC.
(14) Glc →→→ PEP; PEP enters the mitochondria through the means outlined in pathway 11; PEP → OAA by PCK2; OAA → Mal by MDH2; Mal exits the mitochondria; Mal → pyruvate by ME1 (Zelewski and Swierczynski, 1991;Loeber et al., 1994). ME1 knockdown inhibits the growth of colon cancer cells (Murai et al., 2017), and its overexpression is associated with larger breast tumor size, higher incidence of lymph node metastasis, and higher incidence of lymph-vascular invasion (Liu C. et al., 2020). In the same line, ME1 is associated with tumor budding-a phenomenon representing epithelial to mesenchymal transition-in oral squamous cell carcinomas (Nakashima et al., 2020).
(34) L-cysteine is isomerized to D-cysteine by cysteine racemase (2-amino-3-mercaptopropionic acid racemase) (Soda and Osumi, 1969); D-Cys is converted to 3-mercaptopyruvate by D-amino acid oxidase and, in turn, to pyruvate and H 2 S by 3mercaptopyruvate sulfurtransferase (3MST) (Shibuya et al., 2013) or thiosulfate sulfurtransferase (TST) (Pallini et al., 1991). The possibility of conversion of D-Cys to pyruvate by D-cysteine desulfhydrase (Nagasawa et al., 1985) in mammalian cells is yet to be reported. 3-Mercaptopyruvate can also react with hydrogen cyanide, forming pyruvate and thiocyanate in a reaction catalyzed by 3MST or TST; obviously, this is only a very minor route of pyruvate production due to cyanide toxicity (Bhandari et al., 2014; for further considerations related to cancer, see pathway no. 31).
(35) Ser → dehydroalanine (2-aminoacrylate) by serine dehydratase (SDS), serine dehydratase-like protein (SDSL), or serine racemase (SRR): Dehydroalanine can further hydrolyze to NH 3 and pyruvate through SDS, SDSL, or SRR (Kashii et al., 2005); sometimes this reaction is referred to as hydrolysis by "2-aminoacrylate aminohydrolase." Dehydroalanine can also spontaneously hydrolyze to NH 3 and pyruvate through the intermediate 2-iminopropanoate; the latter later part of this spontaneous hydrolysis can be accelerated by 2-iminopropanoate deaminase (Lambrecht et al., 2012). Dehydroalanine can also FIGURE 4 | Pathways leading to pyruvate but not commencing from glucose, highlighted in green: intermediates not transiting through the mitochondria. For abbreviations, see Table 1. be derived from 2 3,5-diiodo-L-tyrosine or 3,5-diiodo-L-tyrosine by thyroid peroxidase in the process of forming thyroxine and triiodothyronine, respectively (Gavaret et al., 1980). The crucial importance of serine metabolism for the growth and survival of proliferating cells is extensively reviewed in Yang and Vousden (2016) and Newman and Maddocks (2017).

PATHWAYS LEADING TO PYRUVATE BUT NOT COMMENCING FROM GLUCOSE: INTERMEDIATES TRANSITING THROUGH THE MITOCHONDRIA
These pathways are shown in Figure 5 (blue arrows).

INCOMPLETELY CHARACTERIZED REACTIONS FORMING PYRUVATE
In the literature, some reactions have been described to produce pyruvate but are incompletely characterized. These are collectively listed below: (55) O-carbamoyl-L-serine + H 2 O → pyruvate + 2 NH 3 , catalyzed by carbamoyl-serine ammonia lyase (Copper and Meister, 1973). O-Carbamoyl-L-serine is a weak inhibitor of a phosphate-dependent glutaminase (Shapiro et al., 1979); mindful of the crucial importance of glutamine catabolism through glutaminases in many cancer types, this route of pyruvate provision is probably minor.
(57) cystathionine + H 2 O → L-homocysteine + pyruvate + NH 3 or cysteine + H 2 O → sulfide + NH 3 + pyruvate or cystine → thiocysteine + pyruvate + NH 3 , all catalyzed by cystathionine gamma-lyase (Stipanuk et al., 2006;Chiku et al., 2009). Cystathionine gamma-lyase was reported to be upregulated in bone−metastatic PC3 cells, and its knockdown suppressed tumor growth and metastasis . In the same line, this enzyme was shown to be upregulated and played a crucial role in the proliferation and migration of breast cancer cells (You et al., 2017).
(61) L-Alanine → pyruvate + NH 3 catalyzed by glutamate dehydrogenase; this reaction exhibits a weak activity (Silverstein, 1974). The role of glutamate dehydrogenase in cancer cells has been extensively reviewed in Moreno-Sanchez et al. (2020).
(69) Salsolinol can be converted to salsolinol-1-carboxylate by salsolinol synthetase which can then be catabolized to dopamine and pyruvate (by an unknown enzyme); salsolinol is an endogenous catechol isoquinoline detected in humans derived from dopamine metabolism (Sandler et al., 1973;Collins et al., 1979). Salsolinol has been implicated in the initiation and promotion of alcohol-related breast carcinogenesis (Murata et al., 2016).

PATHWAYS LEADING TO L-LACTATE AND D-LACTATE INCLUDING THOSE NOT GOING THROUGH LACTATE DEHYDROGENASE
These pathways are shown in Figure 6 (brown arrows).
Lactate-unlike pyruvate-exhibits chirality; thus, it exists in L-or D-configuration. In humans, a putative D-lactate dehydrogenase is known to exist (Flick and Konieczny, 2002;Ewaschuk et al., 2005;Chen et al., 2015). In metabolomics experiments, it is uncommon to distinguish between L-and D-lactate even although it is possible by using special columns. In this section, D-and L-lactateforming pathways are outlined, including those not going through LDH: (70) D-lactate formation by methylglyoxal and intestinal flora (Chen et al., 2015) (for considerations related to cancer, see pathway no. 2).
(71) Pyruvate + QH 2 → D-lactate + Q, catalyzed by D2HGDH in the mitochondrial matrix (Cammack, 1969(Cammack, , 1970. Mutations in D2HGDH have been reported to be involved in multiple types of cancers but render the enzyme hypoactive or inert (Ye et al., 2018); thus, it is unlikely for this route to be important regarding pyruvate production.

PATHWAYS LEADING TO PYRUVATE COMMENCING FROM GLUTAMINE (GLUTAMINOLYSIS)
It is a well-known fact that most cancer cells grow much better when feeding media contain glutamine; this spurred from the pioneering studies of Eagle et al. (1956), showing the dependence of cancer cells growing in monolayer cultures on glutamine. The many critical roles of glutamine in tumor metabolism is reviewed in Altman et al. (2016). From the energetic point of view it were Reitzer et al. (1979) who first showed that glutamine, not sugars, is the main energy source in cultured HeLa cells and that carbon atoms from glutamine incorporate into lactate, but not more than 13%. Zielke et al. (1980), likewise reported that human diploid fibroblasts metabolize up to 13% of media glutamine to lactate. In the same line of thought, Scott et al. (2011), showed that, in human melanoma cell lines, glutamine did not significantly label lactate, in agreement with the data of Ta and Seyfried (2015) reporting that, in a murine glioblastoma cell line, minimal amounts of lactate derived from glutamine were detected. Le et al. (2012), as well as Son et al. (2013) likewise showed that 13 C-labeled atoms in glutamine appear in lactate also to a minimal extent. However, in a study published by DeBerardinis et al. (2007), ∼60% of the glutamine metabolized by SF188 cells was claimed to be converted to lactate, although they seemed to combine this percentage with that of alanine production. The pathway of converting glutamine to pyruvate (and lactate), referred to by McKeehan (1982) as "glutaminolysis, " has been considered a hallmark of tumor metabolism; however, this is a misconception: in normal tissues, FIGURE 7 | Pathways leading to pyruvate commencing from glutamine (glutaminolysis), highlighted in red. For abbreviations, see Table 1. ∼18% of glutamine carbons appear in lactate (Windmueller and Spaeth, 1974), as opposed to ∼10-13% (or less) in tumor cells (see the references above). Thus, if anything, cancer cells exhibit a decrease in glutamine-to-lactate conversion exactly as anticipated, mindful that glutamine provides both energy and building blocks for several biosynthetic processes of cancer. Although glutaminolysis was originally attributed to the pathway Gln → Glu → aKg → succinyl-CoA → succinate → fumarate → malate (exiting the mitochondria) → pyruvate (through malic enzyme), several other routes may also contribute (outlined below; see Figure 7).

ENERGETICS OF GLYCOLYSIS WITH KINETICALLY INACTIVE PK
Glycolysis yields a net of two ATP molecules per glucose molecule; however, in view of an inactive PK while pyruvate is made through PK-bypass pathways, net ATP production from glycolysis is expected to be zero. Although the importance of high-energy phosphate generation has been downplayed in cancer tissues (Vander Heiden et al., 2009), it cannot be ignored that-according to the BRENDA database-among the 336 enzymatic reactions requiring ATP in a cell (without even considering quantitatively important, non-enzymatic mechanisms such as Na + /K + ATPase), 125 of them occur in the cytosol. Clearly, while it is imperative to prevent phosphofructokinase and hexokinase from ATP-dependent feedback inhibition and allow a high flux of glycolysis for the sake of generating intermediates shuttled toward other pathways, ATP is still needed for many other reactions. Crunching the numbers regarding cytosolic energetics is a daunting task, but what is definite is that a cell with nearly zero ATP production from glycolysis may not harbor ATP-consuming mitochondria, for whatever reason (hypoxia, mtDNA mutations, etc.). This can be solved by maintaining the adenine nucleotide translocase in "forward" mode, i.e., providing ATP to the cytosol which is made by SUCL supported by glutaminolysis (Chinopoulos et al., 2010). Production of pyruvate and, therefore lactate is still maintained by the PK-bypassing pathways so as to thwart a reductive stress as pyruvate-to-lactate by LDH maintains a low NADH/NAD + ratio. Finally, it is important to emphasize that this lack of ATP generation by glycolysis due to PK inhibition does not only occur in neoplastic tissues, but it seems to be a more general pathophysiological mechanism also present in tissue ischemia: it was recently reported that during acute kidney injury, PK was inhibited by oxidative/nitrosative stress for the purpose of diverting glycolytic intermediates toward the pentose phosphate pathway which, in turn, yielded reducing equivalents and mounted a better response during the reperfusion phase where ROS are formed, thus increasing the chances for organ survival (Zhou et al., 2019).

CONCLUSION
The above considerations aim to (i) highlight that L-lactate can still be produced from pyruvate using carbon atoms originating from glucose or other substrates in cells with kinetically impaired pyruvate kinase and (ii) show that the mitochondria may contribute to cancer metabolism irrespective of oxidative phosphorylation by providing means of contributing to pyruvate production. Having said that, it is important to emphasize that none of the aforementioned reactions take into account the potential regulatory effects of metabolites on other reactions such as those occurring on PK by amino acids (Chaneton et al., 2012;Yuan et al., 2018). In addition, each enzyme probably exhibits different kinetic and thermodynamic constraints which control the overall flux, which also means that many of these pathways may not operate simultaneously. Such exponentially increasing complexity of a system precludes the possibility of predictions and modeling, though I would be happy to be proven wrong.

AUTHOR CONTRIBUTIONS
CC wrote and edited the manuscript.

ACKNOWLEDGMENTS
I am indebted to Dr. Dora Ravasz, Dr. David Bui, Jeonghyoun Lee, and Seung Won Jeong for assistance with database mining and to Prof. Thomas N. Seyfried for helpful discussions.