Microenvironmental control of glucose metabolism in tumors by regulation of pyruvate dehydrogenase

Abstract

During malignant progression cancer cells undergo a series of changes, which promote their survival, invasiveness and metastatic process. One of them is a change in glucose metabolism. Unlike normal cells, which mostly rely on the tricarboxylic acid cycle (TCA), many cancer types rely on glycolysis. Pyruvate dehydrogenase complex (PDC) is the gatekeeper enzyme between these two pathways and is responsible for converting pyruvate to acetyl-CoA, which can then be processed further in the TCA cycle. Its activity is regulated by PDP (pyruvate dehydrogenase phosphatases) and PDHK (pyruvate dehydrogenase kinases). Pyruvate dehydrogenase kinase exists in 4 tissue specific isoforms (PDHK1–4), the activities of which are regulated by different factors, including hormones, hypoxia, and nutrients. PDHK1 and PDHK3 are active in the hypoxic tumor microenvironment and inhibit PDC, resulting in a decrease of mitochondrial function and activation of the glycolytic pathway. High PDHK1/3 expression is associated with worse prognosis in patients, which makes them a promising target for cancer therapy. However, a better understanding of PDC’s enzymatic regulation in vivo and of the mechanisms of PDHK-mediated malignant progression is necessary for the design of better PDHK inhibitors and the selection of patients most likely to benefit from such inhibitors.

Tumor hypoxia and glucose metabolism

Hypoxia, defined as deficiency in the amount of oxygen reaching a tissue, is a well-known phenomenon found in solid tumors. As the tumor grows, the oxygen that reaches it is limited due to the poor vascularization. In order to overcome the limitations brought on by lack of oxygen, hypoxia inducible factors (HIFs) initiate the primary response of cells to the hypoxic environment^{1, 2}. This response involves gene expression changes facilitating adaptation to hypoxia. Hypoxic metabolic reprogramming is under the control of HIF-1, which is a heterodimer consisting of subunits α and β. HIF-1α’s stability is regulated by oxygen levels, while the expression of the β subunit is constitutive.

In the presence of oxygen, HIF-1α is modified by hydroxylation on prolines 564 and 402. A family of prolyl hydroxylases (PHD; prolyl-4-hydroxylase domain enzymes), which use oxygen as substrate, catalyzes these reactions³. Hydroxylated HIF-1α is recognized by the tumor suppressor pVHL which acts as an E3 ubiquitin ligase and targets HIF-1α for degradation by the proteasome⁴. Hydroxylated HIF is very unstable in vitro, with a half-life of less than 5 minutes⁵. Under hypoxia, HIF-1α hydroxylation decreases, it escapes degradation and the protein accumulates.

In general, HIF-target genes mediate adaptation to moderate levels of hypoxia by increasing oxygen delivery (inducing angiogenesis and erythropoiesis), by decreasing oxygen consumption (decreasing mitochondrial function, increasing glucose uptake and glycolysis), and by escaping environments with limited oxygen availability (increasing cell motility and invasion). Under conditions of very severe oxygen deprivation (anoxia) these adaptive responses are unable to sustain survival and growth, and HIF-independent cell death ensues⁶.

Altered glucose metabolism is a common cancer characteristic. Glucose serves as a main energy source, and under physiologic conditions it is oxidized in mitochondria to generate ATP by oxidative phosphorylation (OXPHOS). Under hypoxic conditions, OXPHOS decreases, and the glycolytic pathway is activated (Figure 1). However, in cancer cells, reliance on glycolysis can occur even in normoxia, a phenomenon known as aerobic glycolysis or the Warburg effect⁷. Due to the inefficiency of glycolysis in ATP production, there is a higher demand for glucose uptake in cancer cells⁸, and glucose transporters GLUT1 and GLUT3 are also some of the genes induced by HIF^{9, 10}. Pyruvate is the end-product of cytoplasmic glycolysis and crosses the outer mitochondrial membrane through VDAC¹¹. Two obligatory subunits of the elusive inner mitochondrial membrane pyruvate carrier (MPC1,2) have been recently discovered^{12, 13}, and their expression is downregulated in many cancers¹⁴. Moreover, hypoxic tumor cells actively divert pyruvate from mitochondria by inducing pyruvate dehydrogenase kinase (PDHK) 1 and 3^15–17. The abbreviation PDK(s) is frequently used to describe Pyruvate Dehydrogenase Kinases, however, to avoid confusion with the 3-phosphoinositide-dependent protein kinase-1 (PDK-1/PDPK1), the newer PDHK nomenclature will be used throughout here. These HIF-responsive kinases phosphorylate pyruvate dehydrogenase (PDH) and inactivate the whole pyruvate dehydrogenase complex (PDC). When unphosphorylated, PDC converts pyruvate to acetyl-CoA in the mitochondrial matrix for the TCA cycle. The phosphorylation of serines 293, 300, or 232 (sites 1, 2, 3, respectively) on PDH by PDHKs inhibits this irreversible decarboxylation¹⁸ and makes pyruvate available for lactate dehydrogenase (LDH, another HIF-target¹⁹) to convert to lactate. This reaction regenerates NAD⁺ for subsequent glycolysis rounds. Why cancer cells modify their metabolism this way is believed to conserve oxygen²⁰, support lipid and nucleotide synthesis and also redox balance necessary for sustained proliferation²¹, while maintaining sufficient, albeit lower, energy production. Thus energy-inefficient pathways are selected for in order to sustain the provision of metabolic intermediates and redox equivalents for mass biosynthesis.

Figure 1: Glycolytic HIF-1 target genes.

(1) GLUT1/3, glucose transporter 1 and 3^{9, 10, 119}, (2) HK1/2, hexokinase 1 and 2^{120, 121}, (3) PGI, phosphoglucose isomerase^{121, 122}, (4) PFK1/2, phosphofructokinase 1 and 2^{121, 123}, (5) ALDO A/C, aldolase A and C^{119, 121, 124}, (6) TPI, triosephosphate isomerase^{121, 125}, (7) GAPDH, glyceraldehyde-3-phosphate dehydrogenase^{121, 126}, (8) PGK, phosphoglycerate kinase^{119, 121, 124}, (9) PGM, phosphoglycerate mutase^{121, 127}, (10) ENO A, enolase A^{121, 128}, (11) PKM, pyruvate kinase M^{121, 124}, (12) PDHK1/3, pyruvate dehydrogenase kinase 1 and 3^15–17, (13) LDHA, lactate dehydrogenase A^{19, 121, 128}, (14) MCT4, monocarboxylate transporter 4¹²⁹. Pyruvate enters mitochondria through VDAC (voltage-dependent anion channel) in the outer and MPC 1/2 (mitochondrial pyruvate carrier 1/2) in the inner mitochondrial membrane, followed by conversion to Acetyl-CoA by the pyruvate dehydrogenase (PDH) complex in the mitochondrial matrix.

Mechanistic regulation of PDC

PDC is a 9.5 MDa multi-unit complex. The central core of PDC, forming a pentagonal dodecahedron consists of 48 copies of the E2 subunit and 12 copies of an E2 homolog, the E3 binding protein (E3BP) (Figure 2). The E1 and E3 components of PDC bind non-covalently to the core, as well as the 4 regulatory PDH kinases and 2 phosphatases, creating a flexible shell on top of the complex core. Mobile E2 and E3BP lipoyl domains shuffle substrates and products between active sites of the catalytic E1, E2, and E3 subunits²². Three serine residues of the E1α subunit (293, 300, and 232) are the targets of PDHKs (serine 232 can be phosphorylated only by PDHK1²³), and phosphorylation of any one of them inactivates the whole PDC complex in vitro, while their dephosphorylation by PDPs restores PDC activity²⁴.

Figure 2: Pyruvate dehydrogenese complex structure (9.5 MDa).

(A) PDC core consisting of 48 E2 and 12 E3BP components is assembled into a pentagonal dodecahedron (3.5 MDa). (B) 20–30 E1 (α₂β₂ tetramer) components attached to the subunit binding domains of E2 and 6–12 E3 (α₂ dimer) units on E3BP make up a plastic shell on the PDC core. Regulatory PDHKs (α₂ dimers) and PDPs (αβ dimers) are also part of this shell, binding to lipoyl domains (mostly inner lipoyl L2 domains) on the flexible arm regions of E2 and E3BP. Such formation facilitates easy movement of individual enzymes and reaction intermediates over the complex.

E1 (or PDH) catalyzes the rate-limiting reaction of oxidative decarboxylation of pyruvate and reductive acetylation of the E2 lipoyl moieties, using thiamine pyrophosphate (TPP) as co-enzyme (Figure 3). E2/dihydrolipoyl acetyltransferase (DLT) and E3/dihydrolipoyl dehydrogenase (DLDH) are the two other catalytic subunits of PDC. E3BP is a structural E3 binding subunit, and PDHKs and PDPs are the regulatory components of PDC. PDHKs are tightly associated with mostly the inner E2 lipoyl domain²⁵. To complete PDC reactions, E2 transfers the lipoate-activated acyl residue to CoA, producing acetyl-CoA and reduced dihydrolipoamide. Finally, E3 regenerates oxidized lipoate, while reducing its prosthetic FAD group to FADH₂, which is then back-oxidized in a reaction reducing NAD⁺ to NADH+H⁺. The structural organization of the complex offers an efficient way of moving reaction intermediates between subunits, executed by “swinging arms” of E2 and E3BP, and a “hand-over-hand” interchange of PDHKs on the surface of the PDC core allowing rapid phosphorylation of E1²⁶. A recent simulation of PDC, where layers of the complex were added successively, visualizes the complexity of this efficient system²⁷.

Figure 3: PDC reaction sequence.

E1 (PDH, pyruvate dehydrogenase) first catalyzes thiamine pyrophosphate (TPP)-dependent irreversible decarboxylation of pyruvate (1) and reductive acetylation of lipoate on the lipoyl domains of E2 (and E3BP), the rate-limiting step (2). E2 (DLT, dihydrolipoyl transacetylase) then catalyzes the transfer of the acetyl group from dihydrolipoate to CoA producing Acetyl-CoA (3). E3 (DLDH, dihydrolipoyl dehydrogenase) oxidizes dihydrolipoate back to lipoate by reducing its FAD cofactor to FADH₂ (4), which is oxidized back to FAD by NAD⁺ (5). Movement of the lipoyl domains between active sites of the individual enzymes is executed by the mobile arms of E2 and E3BP.

Since E1α phosphorylation-dephosphorylation controls PDC activity²⁸, PDHKs and PDPs play crucial roles in the regulation of carbohydrate flux through the complex. The product-to-substrate ratios of the main PDC reaction greatly affect the activity of PDHKs (Figure 4A). They are activated by high NADH+H⁺/NAD⁺ and acetyl-CoA/CoA ratios and inhibited by pyruvate. Pyruvate is an allosteric PDHK inhibitor and binds to PDHKs directly²⁹. NADH and acetyl-CoA on the other hand stimulate PDHKs by reducing and acetylating the E2 lipoyl moieties in reverse E2 and E3 reactions³⁰. Acetyl-lipoamide then induces direct allosteric conformational changes in the lipoyl domain-bound PDHK structure that generate a more active PDHK state, most likely accelerating ADP dissociation. In contrast, full oxidation of the lipoyl groups significantly reduces PDHK activity³¹.

Figure 4: Regulation of PDC activity.

(A) PDC activity is regulated by reversible phosphorylation and dephosphorylation of its E1α subunit executed by PDHKs and PDPs. PDHKs perform inhibitory phosphorylation of three E1α serine residues (Ser 293 – site 1, Ser 300 – site 2, Ser 232 – site 3) and PDPs activate PDC by dephosphorylating these three sites. High PDC reaction endproduct/substrate ratios activate PDHKs by feedback inhibiton, while pyruvate or its synthetic analogue dichloroacetate inhibit PDHKs allostericaly (details in text). PDPs are activated by Ca²⁺ and Mg²⁺ ions, and insulin. (B) Post-translational regulation of PDC activity (details in text). (1)⁹⁷, (2)¹⁰⁰, (3)¹⁰¹, (4)¹⁰², (5)¹⁰³, (6)¹⁰⁴, (7)¹⁰⁷, (8)⁹⁸, (9)⁹⁹, (10)¹⁰⁵, (11)¹⁰⁶, (12)¹³⁰, (13)⁶⁰.

An important feature of PDHKs is that they are active as dimers. The C-terminal tail of each monomer interacts with the lipoyl-binding pocket in the N-terminal domain of the other monomer (summarized in³²). The binding to the E2 lipoyl domain not only co-localizes PDHK to E1 but also opens its active site due to a crossover configuration of its C-terminal tail. PDHK dimers interact with two lipoyl domains simultaneously, producing very high binding affinity to PDC and supporting the hand-over-hand migration over the complex. The PDHK active site is at the C-terminus and is folded similarly to the catalytic domains of the ATPase/kinase superfamily (such as histidine protein kinases), unlike other Ser/Thr-specific protein kinases³³. PDHKs are thought to form homodimers but a recombinant PDHK1-PDHK2 catalytically active heterodimer has also been reported³⁴.

The three inhibitory serine phosphorylation sites on the E1α subunit were identified 40 years ago³⁵. Since E1 is an α₂β₂ tetramer with two active sites positioned at the interface of E1α and E1β, there are six serines in total whose phosphorylation could inhibit PDC. Interestingly, phosphorylation of any one site inactivates E1 in vitro²⁴. The reason for the existence of three sites per subunit is probably flexibility in the modulation of PDC inhibition under different conditions. Site 1 (Ser293) is preferentially phosphorylated first, and sites 2 (Ser300) and 3 (Ser232) are sequentially phosphorylated, leading to slower rates of the complex³². However, the dynamics of these post-translational modifications in vivo are not known yet.

Under hypoxia all three regulatory E1α serine residues become phosphorylated, and mitochondrial function decreases. HIF1-dependent transcription of PDHK1^{15, 16} and PDHK3¹⁷ regulates PDC activity and mitochondrial oxygen consumption, however, the contribution of post-transcriptional and post-translational mechanisms to enzymatic control was not known³⁶. With the development of phospho-serine-specific E1α antibodies³⁷, it became possible to monitor the phosphorylation status of each of the three regulatory serines. We were able to confirm in vivo that phosphorylation of site 3 required PDHK1, as PDHK1 knockout cells completely lost the ability to phosphorylate Ser232³⁸. This incapacity caused a significant decrease in pancreatic model tumor growth³⁸. In vitro studies on purified proteins had shown that phosphorylation of any of the three regulatory serines inactivates PDH²⁴. It is still unknown however, why hypoxia specifically regulates certain PDHK isoforms, and why phosphorylation of sites 1 and 2 is not sufficient for maximum tumor growth, at least in the models tested to date. Additionally, we showed that hypoxic PDH phosphorylation is not entirely dependent on HIF; HIF-1α knockout mouse embryonic fibroblasts were unable to upregulate PDHK1 protein, however, they did maintain some ability to phosphorylate Ser232 in hypoxia, albeit not to the extent of their HIF-expressing counterparts³⁸. Hence hypoxia exerts a dual control over PDHK1, by increasing both the protein levels and its enzymatic activity. Another intriguing, but as of yet untested, possibility is that hypoxia regulates PDHK1 heterodimer formation³⁴ as an additional mechanism of activity control.

The earliest recognition of PDH site 3 oxygen sensitivity came from the parasitic nematode Ascaris suum³⁹. This species undergoes aerobic/anaerobic transition during development and has two PDH isoforms, specifically an anaerobic adult PDH-I isoform with two phosphorylation sites (serine 232 is changed to alanine) and an aerobic larval PDH-II with all three conserved phosphorylation sites, analogous to mammalian PDH. Anaerobic adult PDH-I lacking site 3 is a better substrate for PDHK but less sensitive to inactivation. This suggested that the expression of the less PDHK-sensitive PDH-I prevents its complete inactivation during transition to anaerobiosis. Similarly, the presence of Ser232 in mammalian PDH may make PDC regulation more flexible during fluctuations in oxygen and fine-tune pyruvate oxidation rates.

Most recently, chronic hypoxia was shown to reduce expression of PDC’s E1β subunit possibly through induction of the mitochondrial matrix LONP protease⁴⁰. This mechanism maintains glycolytic metabolism even in reoxygenated areas of tumors, further supporting the Warburg effect. Additionally, prolyl hydroxylase PHD3 stabilizes PDC by physically interacting with the E1β subunit, while depleted PHD3 in hypoxia leads to destabilization and reduced PDC activity⁴¹.

PDC and clinical implications

The most studied isoform in cancer is PDHK1, due to its involvement in oncogenic transformation^{15, 16, 36}. In addition to PDHK1/3 regulation by tumor hypoxia, other PDHK isoforms are reported to participate in cancer progression^42–45. Metabolic control by PDHK extends beyond this hypoxic activation. Oncogene-induced senescence of primary melanocytes by oncogenic BRAF^V600E expression requires increased PDC activity, which is mediated by PDHK1 downregulation and PDP2 upregulation⁴⁶. Forced inhibition of mitochondrial pyruvate metabolism bypassed senescence and permitted tumorigenesis⁴⁶. Conversely, in established melanomas with acquired resistance to the BRAF^V600E inhibitor vemurafenib, combination treatment with PDHK inhibitor dichloroacetate (DCA) and vemurafenib produced more than additive cell killing in vitro⁴⁷. Additionally, DCA sensitized glycolytic melanoma cell lines to the pro-oxidant elesclomol⁴⁸.

A growing number of works points to PDH activity downregulation by PDHKs as important for tumor growth (Table 1). Conversely, it has been reported that, for certain cancer models, PDH activity per se is not essential for survival and growth. For example, E1α ablation in cell lines caused significant changes to central carbon metabolism as expected, however, in vitro growth was supported by reliance on extracellular lipids and glutamine-dependent fatty acid synthesis⁴⁹. In a mouse model of hepatocellular damage and carcinogenesis, liver-specific E1α ablation did not impact tumor growth, despite the decrease in acetyl-CoA levels that inhibition of pyruvate decarboxylation caused⁵⁰. In contrast to autochthonous liver cancer, liver metastases of breast cancer contain elevated phospho-E1α, and PDHK1 downregulation impaired the ability of mouse breast tumors to specifically colonize the liver but did not impact colonization of the lungs⁵¹. Thus, it is becoming evident that alternative pathways exist to coordinate the supply of electron donors to the respiratory chain with TCA metabolite production for biosynthesis.

Table 1:

Summary of published clinical studies evaluating impairments in the pyruvate dehydrogenase complex components and their impact on patient prognosis.

PDC Impairment	Tumor Type	Patient Prognosis	Predicted PDC Activity
PDHK1 upregulation	Nasopharyngeal carcinoma131–133	Poor prognosis
	Oropharyngeal carcinoma38	Poor prognosis
	Neuroblastoma134	Poor prognosis
	Retinoblastoma135
	Head&Neck squamous cell carcinoma136	Poor prognosis
	Non-small cell lung carcinoma54, 55	Poor prognosis
	Breast cancer137	Poor prognosis
	Breast cancer liver metastases51
	Gastric carcinoma138	Poor prognosis
	Renal cell carcinoma139	Lower tumor stage
PDHK1 activation	Glioblastoma98, 99	Poor prognosis
PDHK1 downregulation	Renal cell carcinoma139	Poor prognosis
PDHK2 upregulation	Gastric carcinoma140	Poor prognosis
PDHK3 upregulation	Colon carcinoma43	Poor prognosis
PDHK4 upregulation	Bladder carcinoma141	High grade invasive
Increased phospho-E1α	Oropharyngeal carcinoma38	Poor prognosis
	Non-small cell lung carcinoma54	Poor prognosis
	Ovarian carcinoma142	Poor prognosis
Decreased total E1α	Esophageal squamous cell carcinoma143	Poor prognosis
	Gastric carcinoma144	Poor prognosis
	Ovarian carcinoma145	Poor prognosis
	Hepatocellular carcinoma146	Poor prognosis
Increased total E1α	Prostate carcinoma57, 147	Poor prognosis
Downregulated E1β	Gastric cardia carcinoma148	Poor prognosis
Downregulated E2/E3BP	Non-small cell lung carcinoma56	Good postoperative outcome
PDP1 upregulation	Prostate carcinoma57	Poor prognosis
Poor prognosis	Epidermal carcinoma149	Poor prognosis

Conversely, lung and prostate cancers are characterized by high PDC activity. Metabolic profiling of mouse K-ras driven lung tumors and human NSCLC patients showed significant pyruvate flux through PDC in vivo^{52, 53}, and genetic loss of PDHA1 (E1α gene) in Kras^G12D-derived cell lines inhibited subsequent tumor growth⁵². On the other hand, immunohistochemical detection of PDC components on patient biopsies associated their predicted low PDC activity with either poor^{54, 55} or good⁵⁶ outcome, for reasons that are not clear yet. In prostate, E1α inactivation inhibited xenograft tumor growth by affecting lipid biosynthesis⁵⁷. Intriguingly, both mitochondrial and nuclear PDC contributed to lipogenesis by supplying citrate for fatty acid synthesis and acetyl-CoA for epigenetic activation of lipid-biosynthesis genes, respectively⁵⁷. The mechanism for the nuclear translocation of PDC is not entirely clear but it can be mediated by heat-shock protein 70⁵⁸.

PDHKs are implicated in other medical conditions in addition to cancer. Therefore, deeper understanding of PDC regulation could find clinical applications in the treatment of cancer and other diseases associated with deregulated pyruvate metabolism. For instance, PDHK2 and PDHK4 are upregulated in diabetes and under fasting conditions, and insulin is a negative regulator of PDHK4 expression, resulting in impairment in glucose oxidation⁵⁹. Hence, this isoform could be a promising target for therapy of type 2 diabetes. Additionally, insulin stimulates PDC activity through PDP phosphorylation by mitochondrially translocated PKCδ⁶⁰ (Figure 4B). The highly ubiquitous relevance of PDC function in the organism translates into many medical conditions beyond the scope of this review. Further examples include but are not limited to: acute liver failure⁶¹, non-alcoholic steatohepatitis⁶², obesity⁶³, cardiovascular diseases^64–67, viral and bacterial infection^{68, 69}, and even autism⁷⁰.

PDC defects are the causal factors in pyruvate dehydrogenase complex deficiency, a disease classified as an inborn error of metabolism. The symptoms can range in clinical severity and include congenital lactic acidosis, poor coordination, intellectual disabilities, recurrent ataxia, and delay in development⁷¹. PDC inability to convert pyruvate into acetyl-CoA impairs mitochondrial ATP generation by OXPHOS, forcing the cell to rely on glycolysis for energy production. ATP shortage can cause severe damage to the nervous system and other tissues with high-energy demand. The failure to convert pyruvate to acetyl-CoA can result in lactate accumulation, causing lactic acidemia^{72, 73}. In most cases, PDC deficiencies are caused by genetic defects in E1α and impact enzymatic activity and/or E1α protein production. The PDHA1 gene is located on the X chromosome, which accounts for the higher incidence of PDC deficiencies in men⁷⁴. Mosaicism in the PDHA1 gene due to lyonization of the X chromosome can impeach the diagnosis of PDC deficiency in females⁷⁵, and also in males^{76, 77}. Defects in other PDC subunits are infrequent^78–81. E3 deficient patients suffer additional complications because this PDC subunit is also involved in 2-oxoglutarate dehydrogenase (OGDH) and branched-chain α-ketoacid dehydrogenase (BCKDH) complexes⁸². Treatment for patients with PDC deficiency is mostly symptomatic. DCA, the common inhibitor of pyruvate dehydrogenase kinases has been used to restore PDC activity in these patients. However, DCA is not a potent and specific PDHK inhibitor⁸³ and can have serious adverse effects⁸⁴. Therefore, patients with PDC deficiency would benefit from new and more specific approaches to inhibiting PDHKs.

Metabolic regulation of immune function by PDHKs

Another microenvironmental component with important implications in cancer growth is the metabolic reprogramming in innate and adaptive immunity. Cells of the immune system encounter oxygen- and nutrient-deprived conditions in inflamed tissues and must therefore modify their metabolism to cope with the environmental changes. At the same time, metabolic reprogramming can be triggered by pathogens or other activating stimuli, supporting immune cell activation and function^{85, 86}.

It is generally accepted that activated pro-inflammatory M1 macrophages and inflammatory T helper Th17 cells undergo a switch to glycolytic metabolism similar to cancer cells, and OXPHOS is a hallmark of anti-inflammatory M2 macrophages and anti-inflammatory regulatory T cells (T_regs)⁸⁷. PDHK1 has been shown to be required for CD4⁺ Th17 cell function⁸⁸. Silencing of PDHK1 or DCA treatment inhibited IL-17 production, and through decreased Th17 cell expansion and function, suppressed autoimmunity in vivo⁸⁸.

The possible role of PDHKs in macrophage metabolic reprogramming and function under different polarization states is an active area of investigation. PDHK1 has been shown to be required for the induction of M1-specific markers in murine bone marrow derived macrophages⁸⁹. Also, inhibition of PDHK function by DCA was shown to suppress hypoxia-induced migration οf murine peritoneal macrophages in vitro and in vivo⁹⁰. In contrast to the above studies, prolyl hydroxylase PHD2 knockout macrophages, which cannot hydroxylate HIF-1α and target it for degradation, and thus express HIF-dependent genes like PDHK1 even under normoxia, showed reduced migratory capacity that could be reversed by DCA treatment⁹¹. Additionally, stable isotope metabolic flux analyses found glucose flux through PDH to be maintained in activated macrophages, providing carbons for the biosynthesis of the antimicrobial metabolite itaconate and lipids⁹². The reasons for these discrepancies are not clear yet. The source of cells (bone marrow derived vs elicited peritoneal vs RAW 264.7) and/or differences between genetic and pharmacologic inhibition of PDHKs may be partly responsible. Genetically engineered mouse models of PDHK ablation should help determine their roles in antimicrobial and antitumoral defenses in clinically relevant settings.

PDC regulation by the extracellular matrix

Loss of communication between cells and their extracellular microenvironment has significant ramifications for cell function and physiology, including metabolism. Extracellular matrix (ECM) is the non-cellular scaffolding within tissues and can vary in composition at different sites. During cancer progression, tumor cells must undergo ECM detachment to migrate and then attach again at distant sites. Interestingly, anchorage-independent growth can cause changes in PDC flux^{93, 94}. During ECM-attached growth, ERK activity suppresses the expression of PDHK4⁹³. Loss of ECM attachment reduces ERK activity and results in increased PDHK4 levels and reduced PDC activity in normal mammary epithelial MCF-10A cells⁹³. Mechanistically, ERK signaling was shown to control ring-finger protein 126 (RNF126) expression, which acted as an ubiquitin ligase for PDHK1,3,4 (at conserved lysine 258), resulting in their proteasomal degradation and increased PDC activity in MDA-MB-231 breast metastatic cancer cells⁹⁴. RNF126 depletion profoundly affected cell growth in suspension but not attached growth, indicating its role in anoikis-resistance⁹⁴. Anoikis is a form of apoptotic cell death caused by loss of ECM-attachment, typical for normal cells, while malignant cancer cells are usually resistant to this stress. ERK phosphorylation and RNF126 are lost under detached conditions in normal cells but are maintained in most cancer cells⁹⁴, which suggests that oncogenic signaling may determine the balance between aerobic glycolysis and mitochondrial pyruvate oxidation during anchorage-independent growth.

Emerging mechanisms of post-translational control of PDC activity

In recent years, emphasis is being given to post-translational mechanisms of regulation of metabolic enzymes^{95, 96}. PDHK1 is subject to post-translational phosphorylation on several different residues by different kinases, which can alter PDHK1 activity and consequently impact tumor growth (Figure 4B). Among them, the first report showed that PDHK1 could be phosphorylated on at least three different tyrosine residues (136, 243, and 244) by oncogenic tyrosine kinases, such as FGFR1, BCR-ABL, JAK2, and FLT3-ITD⁹⁷. The authors suggested these Tyr modifications increased ATP binding to the active site of PDHK1, as well as its association with PDC. Nonetheless, Tyr243 phosphorylation of PDHK1 wasn’t altered under hypoxic conditions, so this post-translational modification most likely does not play a role in the induction of PDHK1 activity by hypoxia.

Two reports from 2016 investigated hypoxic PDHK1 activity, although unfortunately none of them measured phosphorylation of the PDHK1-specific target Ser232 of E1α. Li and colleagues proposed an elaborate mechanism, where hypoxia or oncogenic mutations activate ERK, which then phosphorylates the cytosolic glycolytic enzyme PGK1 (phosphoglycerate kinase 1, the first ATP-generating enzyme in the glycolytic pathway)⁹⁸. Phosphorylated PGK1 subsequently translocates into mitochondria, where it assumes a role of protein kinase and phosphorylates PDHK1 on threonine 338, thereby enhancing its activity⁹⁸. In addition, they showed that Thr338 phosphorylation promotes tumorigenesis and that high phospho-Thr338 correlates with poor prognosis in glioblastoma patients. On the other hand, Chae and colleagues identified a pool of activated phosphorylated Akt that translocates to the mitochondria during hypoxia, where it phosphorylates PDHK1 on threonine 346 in the ATP lid⁹⁹. Similar to phospho-Thr338, Thr346 phosphorylation activates PDHK1 and is a negative prognostic factor in glioblastoma patients.

PDHK1 is not the only component modified post-translationally to deactivate the complex (Figure 4B). The same oncogenic tyrosine kinases that phosphorylate PDHK1 on Tyr243 have been shown to phosphorylate E1α at tyrosine 301¹⁰⁰. This modification blocks pyruvate binding and hampers tumor growth¹⁰⁰. E1α Tyr301 phosphorylation is not subject to hypoxic regulation, so its importance in hypoxic PDH suppression is likely minor. PDP1 is also phosphorylated by FGFR1 on two different residues. Tyrosine 94 phosphorylation inhibits PDP1 by blocking lipoic acid binding on E2¹⁰¹. The second FGFR1-phosphorylated residue on PDP1 is tyrosine 381¹⁰², which acts as a primer for a series of post-translational lysine acetylations ultimately leading to downregulated PDC activity. When oncogenic tyrosine kinases phosphorylate PDP1 Tyr381, Sirt3 deacetylase is dissociated and ACAT1 acetyltransferase recruited to PDC. This leads to acetylation of lysine 321 on E1α, which recruits PDHK1 and lysine 202 on PDP1 dissociating it from PDH¹⁰². Sirt3 knockout mice have decreased PDC activity¹⁰³, which is associated with hyper-acetylated and hyper-phosphorylated E1α, and increased fatty acid and aminoacid metabolism in skeletal muscle¹⁰⁴.

Src is yet another kinase post-translationally modifying the E1α subunit of PDC on tyrosine 289¹⁰⁵. Suppression of Src decreases E1α Tyr289 phosphorylation and increases PDC activity independently of Ser-E1α status. Any potential role of Src-induced phospho-Tyr289 under hypoxia is not known.

A different type of post-translational modification affecting protein function is lysine-succinylation. Multiple sites on PDC have been identified as hypersuccinylated in Sirt5 knockout cells¹⁰⁶. Mitochondrial sirtuin Sirt5 removes malonyl and succinyl moieties from target lysines. By desuccinylating PDC, Sirt5 suppresses its activity, however, the physiological significance of PDC (de)succinylation in tumor metabolism is not known.

Furthermore, hypoxia reduces E2 lipoylation and PDC activity via Poldip2 (polymerase-δ interacting protein 2) downregulation¹⁰⁷. Poldip2 indirectly protects the lipoic-acid activating enzyme ACSM1 from degradation resulting in lipoylated E2 and active PDC¹⁰⁷. Additionally, Poldip2 loss led to decreases in α-ketoglutarate levels and subsequent normoxic stabilization of HIF-1α, PDHK induction and PDC E1α phosphorylation, further contributing to low mitochondrial function.¹⁰⁷.

Conclusions

The PDHKs’ involvement in metabolic disorders and cancer warrants the development of specific inhibitors with favorable pharmacokinetic/pharmacodynamic profiles. So far, little success has been made (a new generation of small molecule PDHK inhibitors is far from proving their clinical value¹⁰⁸) and only limited early phase clinical data on the chemotherapeutic efficacy of DCA have been published. Most, but not all, trials reported acceptable toxicities but not very compelling improvements in response endpoints^109–112. The highly complex regulation of PDC by the microenvironment requires its detailed characterization for the successful design of specific targeting strategies. Tumor hypoxia as an example, regulates PDC by multiple and selective mechanisms, but it is not clear what the adaptive advantage of each additional level of regulation is (Figure 5). It has been known for a long time that HIFs control the expression of metabolic enzymes. Technological advances, such as proteomics and metabolomics, have made it evident that hypoxia can control not only the amount of proteins encoded by HIF-inducible genes but also their activity. This may offer flexibility in responding to acute changes in oxygenation by rapidly modulating the activity of enzymes already present. The array of proteins whose activity is controlled by oxygen concentrations is still emerging, and no single mechanism is apparent from the reports. In addition to PDHK1, it includes lactate dehydrogenase A¹¹³, carbonic anhydrase IX¹¹⁴, sphingosine kinase 2¹¹⁵, or ornithine decarboxylase¹¹⁶. It is very likely that hypoxia exerts its acute effects on enzymatic activity not only through post-translational modifications but also through fluctuating metabolite levels. An important consideration for the clinical success of PDHK inhibitors will be the selection of tumors that rely on PDHK activity for growth. In preclinical settings, it should be possible to establish murine models of mutant PDHK and constitutively active or partially regulatable PDH forms to assess tumor physiology, hypoxic metabolism and response to PDHK inhibitors. In clinical trials, it may be necessary to monitor the phosphorylation state of all three E1α serines targeted by PDHKs in parallel with emerging technologies such as in vivo stable isotope metabolic flux analyses^{53, 117}. Although such integrative approaches are not standard clinical practice at present, they are offering great insight into the fate of metabolic substrates, including pyruvate, in human tumors.

Figure 5: Mechanisms involved in regulation of PDC activity by tumor hypoxia.

Hypoxia decreases PDC activity (more details in text) by: HIF-1-mediated transcription of PDHK1 and PDHK3, which inhibit PDC by phosphorylation of E1α (1)^15–17; HIF-1-induced transcription of LONP protease that degrades E1β (2)⁴⁰; downregulation of Poldip2, resulting in increased degradation of the lipoic-acid activating enzyme ACSM1, which leads to decreased E2 lipoylation and PDC activity, accompanied by decreased α-ketoglutarate (αKG) levels that additionally inhibit prolyl-hydroxylases (PHDs) and stabilize HIF-1α (3)¹⁰⁷; activating PDHK1 by post-translational phosphorylation of its threonine residues either by ERK-phosphorylated and mitochondrially-translocated phosphoglycerate kinase 1 (PGK1) (4)⁹⁸ or by mitochondrially-translocated phosphorylated Akt (5)⁹⁹; prolyl-hydroxylase PHD3 depletion that leads to PDC destabilization due to decreased PHD3-E1β interaction (6)⁴¹.

Abbreviations:

Akt

protein kinase B

ACSM1

lipoate-activating enzyme acetyl-CoA synthetase medium-chain family member 1

ACAT1

acetyl-CoA acetyltransferase 1

ADP

adenosine diphosphate

ATP

adenosine triphosphate

BCKDH

branched-chain α-ketoacid dehydrognease

CoA

Coenzyme A

DCA

dichloroacetate

DLDH

dihydrolipoyl dehydrogenase, E3 subunit of PDC

DLT

dihydrolipoyl acetyltransferase, E2 subunit of PDC

E3BP

E3 binding protein

ECM

extracellular matrix

ERK

extracellular signal-regulated kinase

FAD

flavin adenine dinucleotide

GLUT

glucose transporter

HIF

hypoxia-inducible factor

LDH

lactate dehydrogenase

MPC

mitochondrial pyruvate carrier

NAD

nicotinamide adenine dinucleotide

OGDH

2-oxoglutarate dehydrogenase

OXPHOS

oxidative phosphorylation

PGK1

phosphoglycerate kinase 1

PDC

pyruvate dehydrogenase complex

PDH

pyruvate dehydrogenase

PDHA1

pyruvate dehydrogenase E1α subunit gene

PDHK, PDK

pyruvate dehydrogenase kinase

PDP

pyruvate dehydrogenase phosphatase

PHD

prolyl-4-hydroxylase domain enzymes

Poldip2

polymerase-δ interacting protein 2

pVHL

Von Hippel Lindau tumor suppressor protein

RNF126

ring-finger protein 126

Sirt

sirtuin, NAD⁺-dependent deacetylase

TCA

tricarboxylic acid cycle

TPP

thiamine pyrophosphate

VDAC

voltage-dependent anion channel