*Review* **The Multifaceted Pyruvate Metabolism: Role of the Mitochondrial Pyruvate Carrier**

#### **Joséphine Zangari, Francesco Petrelli, Benoît Maillot and Jean-Claude Martinou \***

Department of Cell Biology, Faculty of Sciences, University of Geneva, 30 Quai Ernest Ansermet, 1211 Geneva 4, Switzerland; Josephine.zangari@unige.ch (J.Z.); francesco.petrelli@unige.ch (F.P.); benoit.maillot@unige.ch (B.M.)

**\*** Correspondence: jean-claude.martiou@unige.ch; Tel.: +41-22-3796443

Academic Editor: Ferdinando Palmieri Received: 18 June 2020; Accepted: 14 July 2020; Published: 17 July 2020

**Abstract:** Pyruvate, the end product of glycolysis, plays a major role in cell metabolism. Produced in the cytosol, it is oxidized in the mitochondria where it fuels the citric acid cycle and boosts oxidative phosphorylation. Its sole entry point into mitochondria is through the recently identified mitochondrial pyruvate carrier (MPC). In this review, we report the latest findings on the physiology of the MPC and we discuss how a dysfunctional MPC can lead to diverse pathologies, including neurodegenerative diseases, metabolic disorders, and cancer.

**Keywords:** mitochondria; mitochondrial pyruvate carrier; metabolism; neurodegeneration; metabolic disorders; cancer

#### **1. Introduction**

Mitochondria are essential organelles of endosymbiotic origin, which participate in a multitude of cellular functions in eukaryotic cells, including energy metabolism, biosynthetic reactions, signaling, and the execution of programmed cell death. They host the respiratory chain complexes and ATP synthase, all of which participate in the formation of ATP (adenosine triphosphate) through the process known as oxidative phosphorylation (OXPHOS). Electrons released through oxidation of carbohydrates, amino acids, or lipids in the mitochondrial tricarboxylic acid (TCA) cycle are stored in ATP in the form of two energy-rich phosphoanhydride bonds. Of the molecules that can provide electrons to the respiratory chain, pyruvate, the end-product of glycolysis, is the most critical in many cell types including neurons. In mitochondria, oxidation of pyruvate by pyruvate dehydrogenase (PDH) generates acetyl coenzyme A (acetyl-CoA), which can then combine with oxaloacetate (OAA) to form citrate, the first substrate of the TCA cycle (Figure 1).

Furthermore, pyruvate can be carboxylated by the pyruvate carboxylase (PC) to OAA (Figure 1), which represents a major anaplerotic pathway to replenish TCA cycle intermediates, not only for gluconeogenesis but also for other pathways including the urea cycle and lipid synthesis [1].

In addition to glycolysis, there are other, mostly minor sources of pyruvate such as oxidation of lactate by lactate dehydrogenase (LDH), conversion from alanine by alanine transaminase (ALT), or conversion from malate by cytosolic or mitochondrial malic enzyme (ME) (Figure 1). Once formed, pyruvate can be either reduced to lactate by LDH in the cytosol, regenerating NAD<sup>+</sup> to fuel glycolysis, or fully oxidized within mitochondria, through the TCA cycle. The choice between these two pathways has important consequences for the cell, since glycolysis yields two molecules of ATP/molecule of glucose, whereas oxidative phosphorylation yields >30 molecules of ATP/molecule of glucose [2,3].

**Figure 1.** Metabolic pathways involving mitochondria. In the cytosol, pyruvate is produced through glycolysis, which generates two adenosine triphosphate (ATP) and two reduced nicotinamide adenine dinucleotide (NADH) molecules per molecule of glucose. Pyruvate can also be produced from oxidation of lactate by lactate dehydrogenase (LDH), conversion from alanine by alanine transaminase (ALT), or from malate by cytosolic or mitochondrial malic enzyme (ME). Pyruvate can be imported into mitochondria to be oxidized into acetyl coenzyme A (acetyl-CoA) by the pyruvate dehydrogenase (PDH), which then fuels the tricarboxylic acid (TCA) cycle. Import of pyruvate requires the voltage-dependent anion channel (VDAC) to cross the outer mitochondrial membrane (OMM) and the mitochondrial pyruvate carrier (MPC) to cross the inner mitochondrial membrane (IMM). The TCA cycle can also be fueled by glutamine (Gln) through glutaminolysis or by fatty acids (FA) released from lipid droplets (LD) where they are stored in the form of triglycerides (TG). FAs provide acetyl-CoA through FA β-oxidation. The TCA cycle and β-oxidation both generate the reducing equivalents NADH and flavin adenine dinucleotide (FADH<sup>2</sup> ), which transfer electrons to the electron respiratory chain (ETC), generating more than 30 ATP molecules per molecule of glucose. The last reaction is catalyzed by ATP synthase or Complex V (CV). This process requires the presence of oxygen and is known as oxidative phosphorylation (OXPHOS). In the liver and kidney, pyruvate can be converted into oxaloacetate (OAA) by the pyruvate carboxylase (PC), which is then reduced into malate by the malate dehydrogenase (mMDH). Malate is then exported into the cytosol, converted into OAA by malate dehydrogenase (cMDH) and into phosphoenolpyruvate (PEP) by PEP carboxykinase (PEPCK) and from there into glucose through several steps, including the reversible steps of glycolysis. IMS: intermembrane space; CPT1: carnitine palmitoyltransferase 1; CPT2: carnitine palmitoyltransferase 2; GDH: glutamate dehydrogenase; GLS: glutamase; α-KG: α-ketoglutarate.

To enter mitochondria, pyruvate crosses the outer mitochondrial membrane (OMM) to reach the intermembrane space (IMS), probably through the large, relatively non-specific, voltage-dependent anion channel (VDAC), and it is then transported together with a proton across the inner mitochondrial membrane (IMM) by the mitochondrial pyruvate carrier (MPC) [4] (Figure 1). The existence of MPC was proposed on theoretical grounds several decades ago [4], although the molecular identification of the MPC complex was only achieved in 2012 [5,6]. As the sole point of entry for pyruvate into the mitochondrial matrix, the MPC plays a crucial role in coordinating glycolytic and mitochondrial activities, and it provides a key decision point for modulating cellular energy production and metabolism.

In this review, we report the most recent findings on the physiology of the MPC and its participation in various pathologies, including neurodegenerative diseases, metabolic disorders, and cancer.

#### **2. Structure of the MPC**

The MPC is encoded by three homologous genes *MPC1*, *MPC2*, and *MPC3* in *Saccharomyces cerevisiae*, by two genes *MPC1* and *MPC2* in flies, and by three genes, *MPC1*, *MPC1-like*, and *MPC2* in mammals [5,6]. In yeast, the active MPC complexes are the MPC1-MPC3 heterodimers, which promote pyruvate transport during respiratory growth, and the MPC1-MPC2 heterodimers, which function during fermentable growth [7,8]. In most mammalian cells, the active carrier is composed of an MPC1 and MPC2 heterodimer, with the exception of spermatocytes, which display MPC1-like and MPC2 heterodimers [9]. Loss of one subunit leads to degradation of the other subunit and disruption of the MPC complex. Functional tests with yeast MPC1 and MPC3 following reconstitution of the carrier in liposomes showed that only heterodimers were able to transport pyruvate [10]. In another report, MPC2 homodimers were also reported to be functional [11], although this was not supported by the results of Tavoulari et al. [10] or by other data reporting that mitochondria from ∆MPC1 mutants were unable to import pyruvate [5,6,12]. The reason for this discrepancy remains unclear.

MPC1 and MPC2 are small integral membrane proteins of, respectively, 12 kDa and 14 kDa. Structure predictions using different algorithms suggest that MPCs belong to the semi-SWEET (Sugar Will Eventually be Exported Transporter) domain family (SLC50 family) [13] or to the SWEET family [14], also known as the PQ-loop family of sugar transporters. The semi-SWEET domain is composed of a simple 1-3-2 triple transmembrane helix bundle (THB), whereas the SWEET domain consists of two semi-SWEET domains linked by a transmembrane helix [15] (Figure 2A). Recently, Medrano-Soto et al. proposed that the MPC belongs to the transporter-opsin-G protein-coupled receptor (TOG) superfamily with seven putative TMSs arranged in a 3+1+3 topology [14]. According to these authors, the MPC1 and MPC2 subunits might have originated from duplication of an MPC precursor, composed of four transmembrane segments, which would have lost its N-terminal transmembrane segment. Although the mechanisms via which the MPC imports pyruvate remain unknown, several hypotheses were postulated using in silico docking analyses, based on structural models of the MPC [16–18].

The heterodimeric composition and homology to the SWEET or semi-SWEET sugar transporters, sets the MPC apart from other families of mitochondrial carriers (named MCF or SLC25). The membrane topology of MPC1 and MPC2 is still not fully resolved and Figure 2 proposes a model based on the semi-SWEET motif. Our earlier biochemical approaches based on the accessibility to proteases or to thiol labeling suggest that MPC1 displays at least two transmembrane segments with the N- and C-termini projecting into the mitochondrial matrix, whereas MPC3 in yeast and MPC2 in mammals probably consist of three transmembrane helices, with the N-term in the matrix and the C-term in the intermembrane space [7]. It was recently reported that, despite this topology, yeast MPC2 and MPC3, both of which display an odd number of transmembrane segments with the N-term in the matrix, are nevertheless imported via the carrier import pathway which includes the receptor Tom70, TIM (Translocase of the Inner Membrane) chaperones, and the TIM22 complex [19,20], and not via the flexible presequence pathway as was previously predicted.

‐ ‐ ‐ ‐ ‐ ‐ **Figure 2.** (**A**) Semi-SWEET and SWEET motifs. Semi-SWEET is composed of a triple helix bundle in this specific order 1-3-2. SWEET contains two semi-SWEET motifs linked together by a helix. All the helices are crossing the membrane. (**B**) Model for the topology of MPC1 and MPC2 in the inner mitochondrial membrane (IMM). Here, a semi-SWEET structure is proposed for both MPC1 and MPC2. MA: matrix; IMS: intermembrane space.

‐ Several *MPC1* mutations resulting in disruption of the MPC complex or loss of transporter function were reported [21,22]. All of these mutations are accompanied by severe clinical symptoms and premature death.

‐ ‐ ‐ Solving the three-dimensional structure of the MPC will be key to resolving the remaining uncertainties concerning the structure of the carrier, its membrane topology, and how it transports pyruvate across the IMM.

‐

‐

#### ‐ **3. Regulation of MPC Expression**

#### ‐ *3.1. Transcriptional Regulation of MPC Expression in Yeast and in Mammalian Cells*

‐ As mentioned above, in *Saccharomyces cerevisiae*, MPC1 and MPC2 are expressed under fermentative conditions and form the MPCFERM complex, while MPC1 and MPC3 are expressed under respiratory conditions and form the MPCOX complex [7,23]. This switch is orchestrated at the level of transcription by the activity of the high osmolarity glycerol (HOG) mitogen-activated protein (MAP) kinase pathway. Accordingly, the *MPC3* gene promoter is bound directly by the Sko1 transcriptional repressor/activator, which is one of the core transcription factors mediating the transcriptional osmostress response downstream of the HOG1 MAP kinase [8].

In mammalian cells, although MPC1 and MPC2 subunits are ubiquitously expressed, their expression is particularly abundant in the heart, kidney, liver, brown adipose tissue, muscles, and brain [24–26]. A number of transcriptional regulatory mechanisms of the MPC genes were reported. Wang and colleagues showed that MPC1 is transcriptionally repressed in human prostate cancer cells by the chicken ovalbumin upstream promoter-transcription factor II (COUP-TFII), a member of the steroid receptor superfamily. MPC1 repression is part of the metabolic switch toward increased glycolysis, which promotes prostate cancer cell growth and invasion [27].

In contrast to COUP-TFII, peroxisome proliferator-activated receptor-gamma co-activator (PGC)-1 alpha (PGC-1α) was found to increase expression of MPC1 in human renal cell carcinoma [28] and cholangiocarcinoma [29]. Overexpression of PGC-1α strongly stimulated *MPC1* transcription through binding to the *MPC1* promoter, while depletion of PGC-1α by small interfering RNA (siRNA) suppressed MPC1 expression. The latter study showed that PGC-1α reversed the Warburg effect by upregulating expression of both MPC1 and pyruvate dehydrogenase E1 alpha 1 subunit (PDHA1), thus leading to enhanced mitochondrial metabolism. Transcriptional activation of MPC1 by PGC-1α was shown to be mediated by recruitment of the estrogen-related receptor alpha (ERRα), which bound the ERRα response element located in the proximal *MPC1* promoter region [28,30]. Inhibiting the activity of ERRα decreased expression of MPC1, interfered with pyruvate entry into mitochondria, and increased cellular reliance on glutamine oxidation and the pentose phosphate pathway (PPP) to maintain reduced NAD phosphate (NADPH) homeostasis [30].

More recently, it was shown that colon cancer cells can reprogram cell metabolism to coordinate proper cellular response to interferon-γ (IFNγ), a cytokine that plays a pivotal role in host antitumor immunity. Downregulation of MPC subunit expression via the signal transducer and activator of transcription 3 (STAT3) pathway attenuated IFNγ-mediated apoptosis of the colon cancer cells by preventing production of reactive oxygen species (ROS) [31]. Moreover, inhibition of STAT3-mediated transcription using the inhibitor Stattic partially reversed the inhibition of MPC1 and MPC2 expression and increased the antitumor efficacy of IFNγ.

In prostate adenocarcinoma, the androgen receptor (AR) is a hormone-responsive nuclear receptor transcription factor that coordinates anabolic processes to enable tumor proliferation through transcriptional regulation of metabolic pathways [32]. Massie and colleagues showed that the AR regulated MPC activity via direct transcriptional control of *MPC2*. In AR-driven prostate adenocarcinoma, MPC inhibition led to reduced OXPHOS, activation of the eukaryotic initiation factor 2 α (eIF2α)/activating transcription factor 4 (ATF4) integrated stress response, and increased glutaminolysis [33]. Importantly, in this experimental model, MPC inhibition by the small-molecule inhibitor MSDC-0160 suppressed tumor growth in vivo, suggesting that the MPC could be a potential therapeutic target for this type of cancer.

Finally, it was found that, in pancreatic cancer, MPC1 is transcriptionally suppressed by the histone lysine demethylase 5A (KDM5A) [34]. Elevated expression of KDM5A and downregulation of MPC1 correlated directly with pancreatic ductal adenocarcinoma progression.

Taken together, these data indicate that a number of different transcription factors can regulate the promoter activity of MPC1 and/or MPC2. How these factors regulate expression of the MPC genes, and how this regulation is integrated into the overall activity of cellular signaling pathways are interesting questions that remain to be further investigated.

#### *3.2. Post-Translational Regulation of MPC Expression*

Several studies described the post-translational regulation of MPC activity. Liang and colleagues found that MPC1 is acetylated on lysine residues K<sup>45</sup> and K46, and that deacetylation of MPC1 by Sirt3 resulted in increased carrier activity [35].

MPC2 was also reported to be acetylated in a mouse type 1 diabetic heart model. However, in this case, acetylation appeared to stimulate activity of the transporter [36].

An indirect model of post-translational regulation was reported by the group of Kim et al. [37], who showed that, in hepatocarcinoma cells, the BH3 (Bcl-2 homology region 3)-only protein PUMA (p53 upregulated modulator of apoptosis), whose transcription depends on p53, can associate with MPC to disrupt dimer formation and MPC function. High expression levels of PUMA were correlated with decreased mitochondrial pyruvate uptake and increased glycolysis.

#### **4. The MPC and Cell Metabolism**

Cell metabolism can be defined as the ensemble of chemical reactions occurring in the cell, including anabolic reactions that convert nutrients into molecular building blocks, and catabolic reactions that convert nutrients into energy. By allowing the import of pyruvate into mitochondria, the MPC participates in both anabolic (synthesis of intermediary metabolites by the TCA cycle) and catabolic (OXPHOS) events.

One of the characteristics of cell metabolism is its plasticity, which results from the interconnectivity between different metabolic pathways. Plasticity ensures that, when one pathway is transiently or permanently interrupted due to lack of an essential metabolite or a key enzyme or transporter, the cell is able to switch to an alternative pathway to compensate for the defect. However, bottlenecks exist, involving certain metabolite carriers, which allows metabolites to cross cellular membranes. Pyruvate metabolism is a good example. As mentioned in the introduction, there are several ways to synthesize this metabolite in the cytosol; however, the MPC is the only route via which pyruvate can cross the IMM. Such molecules provide key regulatory steps for the cell, as well as offer attractive targets for therapeutic intervention.

Nevertheless, because of the metabolic plasticity, cells often find ways to compensate for defective carriers or metabolites. For example, when the MPC is deficient, increased glutaminolysis or oxidation of fatty acids (beta-oxidation) can compensate for the deficit in pyruvate-derived carbon to fuel the TCA cycle [25,38,39].

Different cell types adapt their metabolism differently to changes in their cellular environment and physiology. In cells with high energy demands such as muscle cells and neurons, OXPHOS is the preferred source of ATP. For these cells, either glucose/pyruvate or fatty acids (FAs) provide the main substrates, although neurons preferentially oxidize pyruvate because they do not express many of the enzymes required for beta-oxidation. In contrast, cardiomyocytes in the adult seem to favor FA oxidation over pyruvate [40,41]. Thus, in these cell types, metabolic plasticity is more limited, which may explain the high incidence of degenerative pathologies that affect brain and muscle.

In other cell types, ATP is generated mainly through glycolysis rather than OXPHOS, even when oxygen is available. This is the case for many highly proliferating cells, including antigen-activated lymphocytes [42] and most cancer cells in vitro. This type of metabolism is referred to as aerobic glycolysis or the Warburg effect after Otto Warburg who first described this process in cancer cells [43,44].

In highly proliferating cells, an important advantage of glucose fermentation compared to OXPHOS is that carbohydrates are only partially degraded, thus providing intermediary metabolites, particularly nucleotides generated via the pentose pathway [43], as building blocks to maintain rapid growth.

Thus, it is clear from the above that the importance of the MPC in cell metabolism is highly dependent on the cell type and cell context. Below, we review the role of the MPC in different cell types and discuss how a dysfunctional MPC can lead to diverse pathologies.

#### *4.1. The Role of MPC in Neurons and in Neurogenerative Diseases (NDs)*

Neurons rely mainly on glucose as an energy source and to a lesser extent on amino-acid oxidation; however, they lack most enzymes involved in FA oxidation [45]. As a result, the role of the MPC, and of pyruvate metabolism in general, is particularly important in neurons. Pyruvate is generated principally through glycolysis in these cells, but also by conversion from lactate by LDH. Astrocytes provide the major source of lactate which is taken up by neurons by the so-called astrocyte-neuron shuttle [46].

Although most neurons use oxidative phosphorylation to generate the high levels of energy required for neural transmission, some parts of the brain, such as the medial and lateral parietal and prefrontal cortices, were found to rely mainly on aerobic glycolysis [47]. Energy metabolism in the retina is also predominantly through aerobic glycolysis, and only a small fraction of the pyruvate produced by glycolysis is oxidized in mitochondria. Nevertheless, pyruvate oxidation in mitochondria

appears to be essential for retinal function since mice lacking MPC1 in the retina were found to display degeneration of both rod and cone photoreceptors and decline in visual function [48]. MPC-deficient retinas displayed lower ATP and NADH levels although increased glutaminolysis and ketone body oxidation limited degeneration of the photoreceptors.

Neurodegenerative diseases (NDs) are often considered to be metabolic disorders characterized by a decline in the ability to import or metabolize energy sources, resulting either in or from mitochondrial dysfunction [49]. However, even though bioenergetic defects were observed in diverse pathological conditions, both in mice and in patients [50–52], in most cases, it remains unclear whether this is the cause or the consequence of the pathology. Interestingly, it was shown recently that modulation of energy metabolism through MPC inhibition offers a potential pharmacological approach to treatment for NDs, in particular for Parkinson's and Alzheimer's diseases.

Parkinson's disease (PD) is a neurodegenerative disorder resulting in the death of dopaminergic neurons in the substantia nigra pars compacta of the brain. In recent years, epidemiological evidence showed similarities in metabolic dysfunction between type 2 diabetes and PD. Indeed, several clinical trials in PD patients are in progress using anti-diabetic drugs [53–59]. In 2016, Ghosh and colleagues (2016) investigated the activity of the MPC inhibitor MSDC-0160, a derivative of thiazolidinedione, in diverse models of PD [60]. The authors showed that MSDC-0160 protected tyrosine hydroxylase (TH)-positive dopaminergic neurons against the neurotoxicity of MPP<sup>+</sup> or MPTP in vitro and in vivo, as well as showed beneficial effects in the Engrailed heterozygous mutant mice, which undergo loss of dopaminergic neurons at six weeks of age [60,61]. The mechanisms via which MPC inhibition results in neuronal protection are not well understood. Ghosh and colleagues proposed that inhibition of the MPC may protect neurons through modulation of the mammalian target of rapamycin (mTOR) pathway and autophagy, and indeed MSDS-0160 was shown to reduce neuroinflammation by modulating the NF-κB (nuclear factor kappa-light-chain-enhancer of activated B cells)/mTOR pathway [60]. This could be especially relevant for PD since neuroinflammation is considered to play an important role in the pathophysiology of the disease [62,63].

MPC inhibition could also be beneficial in the treatment of Alzheimer's disease (AD). A phase IIa clinical study in non-diabetic subjects with mild to moderate AD demonstrated that a three-month treatment with the MPC inhibitor MSDC-0160 resulted in a significant increase of glucose uptake in the regions of the brain normally affected in this pathology [64]. This again argues in favor of a possible neuroprotective effect of this compound. However, these results should be considered in the light of recent findings in vitro in which the expression of MPC2 was shown to be decreased in AD-related models [65]. This led to decreased calcium and pyruvate uptake into mitochondria and decreased OXPHOS. The reason for the decrease in calcium import remains unclear. However, another study in hepatocytes and embryonic fibroblasts showed that calcium import into mitochondria through the mitochondrial calcium uniporter (MCU) was decreased following inhibition of MPC activity. This effect was mediated by increased expression of the MCU gatekeeper mitochondrial calcium uptake 1 (MICU1) [66]. It would be interesting to test the expression levels of MICU1 in MPC-deficient neurons.

Another study describing the effects of pharmacological inhibition of the MPC reported that two specific inhibitors, MSDS-0160 and UK5099, protected neurons against glutamate-induced excitotoxicity in vitro [67]. This suggests that inhibition of the MPC could be useful in acute pathologies of the brain such as stroke or brain trauma where neurons are frequently exposed to toxic levels of glutamate.

Despite these promising results in PD, AD, and acute neuronal death, the molecular mechanisms underlying these effects remain to be elucidated. Furthermore, most studies showing a beneficial effect of MPC inhibition in these pathologies were based on small-molecule derivatives of thiazolidinediones. Although these compounds were shown to act as MPC inhibitors, off-target effects cannot be excluded. Therefore, it will be important to confirm these results using other chemical classes of MPC inhibitors or using genetic approaches in mouse models in which MPC1 or MPC2 are deleted specifically in neurons and/or glial cells.

#### *4.2. The Role of MPC in Metabolic Disorders*

Many studies reported that disruption of the MPC affects gluconeogenesis, a process known to play a role in the pathogenesis of type 2 diabetes (T2D) [68]. T2D can result from the dysfunction of several organs, including pancreas, liver, muscle, and kidney, all of which express the MPC. It will, therefore, be important to analyze the consequences of MPC downregulation in each of these organs.

#### 4.2.1. MPC in Pancreas

Glucose is an important physiological stimulus for insulin secretion by pancreatic β-cells. Elevated blood glucose triggers increased glucose uptake into these cells, synthesis of ATP by OXPHOS, and closure of the plasma membrane ATP-sensitive potassium channels (KATP channel), which then leads to membrane depolarization, entry of calcium, and insulin secretion. This phenomenon, termed glucose-stimulated insulin secretion (GSIS), requires both pyruvate oxidation and carboxylation in pancreatic β-cells [69–71]. Inhibition of the MPC, either pharmacologically using UK5099 or genetically using siRNAs directed against MPC1 or MPC2, reduced GSIS in the 832/13 cell line derived from INS-1 rat insulinoma cells, as well as in rat and human islets [72]. Oxygen consumption, the ATP/ADP ratio, and the NADPH/NADP<sup>+</sup> ratio were all reduced upon inhibition of the MPC. Similar results were obtained in mice displaying inactive, truncated MPC2 [26], in mice carrying a targeted deletion of MPC2 in pancreatic cells, as well as in *Drosophila* [73]. All these experiments show that the MPC plays an important and evolutionarily conserved role in insulin-secreting cells through mediating glucose sensing, regulation of insulin secretion, and control of systemic glycemia.

#### 4.2.2. MPC in Liver

#### Gluconeogenesis

One of the mechanisms via which the liver participates in T2D is through gluconeogenesis [68,74], a major regulatory process in which non-carbohydrate substrates are converted into either free glucose or glycogen (Figure 1). Gluconeogenesis can also take place in the kidney, albeit to a lesser extent.

Hepatic glucose production is a critical physiological process that is required for maintaining normoglycemia during periods of nutrient deprivation. The major non-carbohydrate precursors are lactate, amino acids, and glycerol. Lactate is of particular importance. It is converted into pyruvate by the action of lactate dehydrogenase LDHB, and pyruvate is then imported through the MPC into mitochondria where it is carboxylated into OAA and reduced into malate. Malate is then exported into the cytosol where it can be used to generate glucose through several steps, including the reversible steps of glycolysis (Figure 1).

Lactate for gluconeogenesis is derived mainly from muscle cells undergoing anaerobic glycolysis. Under these conditions, muscle cells release lactate, which is then taken up by the liver to provide a substrate for gluconeogenesis. The glucose produced by the liver can in turn provide an energy source for muscle cells, thereby completing the cycle. This pathway, known as the Cori cycle or the glucose-lactate cycle, can account for up to 40% of plasma glucose turnover.

As expected, liver-specific deletion of MPC1 or MPC2 in mice impairs lactate/pyruvate-triggered hepatic gluconeogenesis [75,76]. However, gluconeogenesis from alanine was increased in MPC-deficient mice, and McCommis et al. [76] suggested that intramitochondrial transamination of alanine to pyruvate may contribute to gluconeogenesis when mitochondrial pyruvate import is inhibited. Thus, pyruvate-alanine cycling may constitute an alternative pathway for gluconeogenesis, which circumvents the MPC. This interesting hypothesis implies the existence of a mitochondrial transporter for alanine, which remains to be identified.

By decreasing gluconeogenesis, MPC inhibition was found to attenuate the development of hyperglycemia induced by a high-fat diet (HFD) leading to improved glucose tolerance [75,77].

#### MPC in Nonalcoholic Steatohepatitis (NASH)

The rise in the level of obesity in the population dramatically increased the incidence of a variety of related metabolic diseases, including nonalcoholic fatty liver disease (NAFLD). The spectrum of NAFLD ranges from simple hepatic fat accumulation to a more severe disease termed nonalcoholic steatohepatitis (NASH), involving inflammation, hepatocyte death, and fibrosis.

As mentioned above, thiazolidinediones (TZDs) appear to be potent MPC inhibitors [78]. One TZD derivative, MSDC-0602, prevents and reverses stellate cell activation and fibrosis in a mouse model of NASH. Importantly, the effects of this small-molecule inhibitor were duplicated by genetic deletion of MPC2 in hepatocytes and furthermore, the effects of MSDC-0602 were lost when MPC2 was deleted [79]. Thus, the MPC appears to be an attractive target in NASH [80]. Indeed, results from a phase IIb clinical trial on patients with liver biopsy-confirmed NASH [81] showed that MSDC-0602K significantly decreased liver steatosis, although it failed to prevent liver fibrosis.

#### 4.2.3. MPC in Kidney

Diabetic kidney disease (DKD), which is characterized by albuminuria and renal hypertrophy, is the leading cause of kidney failure. It was shown recently that treatment with artemether, a methyl ether derivative of artemisinin used in the treatment of malaria and identified as a possible candidate for treating T2D, prevented kidney hypertrophy and ameliorated the lesions that lead to renal enlargement in T2D db/db mice. Interestingly, the mechanisms underlying this beneficial effect may in part be associated with the ability of artemether to increase MPC1 and MPC2 levels in db/db mice [82]. In particular, podocytes, which play an important role in the development of DKD, undergo increased apoptosis when the MPC is inhibited using UK5099 or RNA interference. These results suggest that enhancing MPC function may reduce injury in high-glucose-treated podocytes and may possibly attenuate DKD.

#### 4.2.4. MPC in Muscle

During T2D, decreased glucose uptake by skeletal muscle significantly drives chronic hyperglycemia [83]. Skeletal muscle-specific MPC knockout in mice (MPC SkmKO) leads to increased glucose uptake in muscle and increased lactate release [84], thus increasing the Cori cycle as explained above. Furthermore, FA oxidation appears to be increased in MPC-deficient cells. Because hepatic gluconeogenesis is energetically supported by FA oxidation and muscle MPC disruption increases muscle FA oxidation, futile Cori cycling is energetically supported by FA oxidation. In conclusion, these findings raise the possibility that selectively decreasing skeletal muscle pyruvate uptake in obese and T2D patients may promote fat loss and restoration of whole-body insulin sensitivity.

In conclusion of this part on metabolism and pathologies, it emerges that, while inhibition of the MPC in complex metabolic diseases, such as T2D, is potentially interesting, it is difficult to predict the outcome in patients for two main reasons: (i) because these pathologies are the result of multiorgan dysfunction, and (ii) because antagonistic effects of MPC inhibition occur in different organs. For example, while MPC inhibition increases glucose uptake in the muscles and decreases gluconeogenesis in the liver, two beneficial effects for T2D, it also decreases insulin secretion, which a priori could be problematic for diabetic patients. Thus, it remains unclear whether the combination of these positive and negative effects will result in an overall benefit for the patient.

Due to embryonic lethality in mice at around E12, it is not possible to generate a constitutive deletion of either MPC1 or MPC2 in all tissues [24–26] and, to date, experiments addressing the role of MPC in the mouse were mainly performed using organ-specific deletion of the carrier. Furthermore, we do not yet know whether a global, conditional knockout of MPC1 or MPC2 in adult mice would be lethal. An alternative approach, therefore, is to address this question pharmacologically by administering small-molecule inhibitors of the MPC, with the caveat that off-target effects cannot be

excluded. Such experiments were performed recently and, as described above, promising results were obtained in mouse models of NASH and T2D. ‐ ‐

‐

#### *4.3. The Role of MPC in Cancer*

As described above, cancer cells in vitro rely mainly on aerobic glycolysis for rapid cell growth (Figure 3), and it is not surprising, therefore, that loss of MPC expression, which favors increased glycolysis, was found to be associated with poor cancer prognosis [85–90].

**Figure 3.** Tumor cell metabolism. In the primary tumor site, cancer cells mainly rely on aerobic glycolysis (Warburg effect) and this metabolism favors cell proliferation and tumor growth. Some cancer cells escape from the primary tumor site through the bloodstream, generating metastasis in other organs. It is thought that, to do this, some cancer cell types switch to a more oxidative metabolism.

One of the main questions in the field is whether loss of functional MPC can initiate tumorigenesis and, conversely, whether increasing MPC activity would result in reduced aerobic glycolysis. Furthermore, if aerobic glycolysis is generally associated with tumor growth, there are a number of reports showing that OXPHOS and ROS production are required for tumor metastasis. In this case, what is the role of the MPC? Does expression of the MPC contribute to poor prognosis, and would inhibition of the MPC be able to reduce cell invasion and metastasis? These are the questions we will address in the next section.

#### 4.3.1. The Role of MPC in Stemness

Re-expression of wild-type (WT) MPC1 and MPC2 in colon cancer cells, which carried mutations or deletions in the MPC1 gene, impaired colony formation in soft agar and spheroid formation in vitro and reduced tumor growth in vivo [89]. In addition, these antitumoral effects were accompanied by a decrease in stem-cell markers, such as aldehyde dehydrogenase (ALDH) A (ALDHA), lin-28 homolog A (LIN28A), leucine rich repeat containing G protein-coupled receptor 5 (LGR5), and homeobox transcription factor Nanog (NANOG), suggesting that a decrease in MPC expression promotes the

Warburg effect and the maintenance of stemness in colon cancer cells. Consistent with these findings, Zhenhe Suo and colleagues showed that pharmacological or genetic inhibition of the MPC stimulated aerobic glycolysis in prostate, esophageal squamous, and ovarian cancer cells in vitro [87,91–93]. This was associated with higher levels of stem-cell markers including organic cation transporter (OCT) OCT3/4, NANOG, hypoxia inducible factor 1 alpha (HIF1α), notch receptor 1 (NOTCH1), CD44 antigen, and ALDH [91,93]. Moreover, MPC inhibition conferred on these cells the ability to migrate and an increased resistance to both chemotherapy and radiotherapy [87,91,93]. Similar observations were reported for lung adenocarcinomas [90]. Taken together, these results highlight a role of MPC in determining the stemness status of cancer cells in vitro.

More recently, MPC was shown to be involved in the initiation of intestinal tumor formation in mice and *Drosophila* [94]. Sporadic colon tumors are believed to follow a typical progression pathway in which the initial tumorigenic event triggers intestinal stem-cell hyperplasia, leading to the formation of a benign adenoma. This initial event was associated with hyperactivation of the Wnt/β-catenin pathway and loss of function of the *APC* tumor suppressor. In this study, expression of both MPC1 and MPC2 was found to be decreased in adenomas in two different mouse models of colon cancer, the azoxymethane and dextran sodium sulfate (AOM-DSS) models, as well as a genetic model of heterozygous loss of *Apc* in intestinal stem cells (*ApcLrig1KO*/+*Mpc1Lrig1KO* mice) [94]. Targeted deletion of *MPC1* in adult LRIG1<sup>+</sup> intestinal stem cells (*Mpc1Lrig1KO*) led to an increased tumor burden and a substantial increase in the incidence of macroscopic tumors in the AOM-DSS model. Furthermore, in the heterozygous *Apc* model, the *ApcLrig1KO*/+*Mpc1Lrig1KO* mice exhibited a higher tumor burden compared to *ApcLrig1KO*/+ mice, and the macroscopic tumor burden was also much higher. These results demonstrate that loss of MPC function is sufficient to promote intestinal tumor initiation in a chemically induced tumor model and in different genetic tumor models. Similarly, in *Drosophila*, loss of the MPC or *Apc* led to hyperproliferation of intestinal stem cells and, importantly, the hyperproliferation following deletion of *Apc* was completely suppressed by ectopic expression of the MPC. Interestingly, the metabolic consequences of MPC loss resulted in all *Mpc1Lrig1KO* adenomas attaining a stem-like phenotype.

In conclusion, decreased mitochondrial pyruvate metabolism through elimination of MPC activity is sufficient to increase the oncogenic susceptibility of both the fly and the mouse intestinal tracts. Thus, constitutive enforcement of the metabolic program found in hyperproliferative colonic lesions predisposes intestinal stem cells to adenoma formation.

#### 4.3.2. The Role of MPC in Epithelial-Mesenchymal Transition (EMT)

During tumorigenesis, cancer cells acquire migratory and invasive properties through induction of the epithelial-mesenchymal transition (EMT). It was proposed that production of ROS can trigger metastasis [95], and some authors proposed that aerobic glycolysis would promote primary tumor formation, while a shift to a more oxidative metabolism would be required for metastasis (Figure 3). In intrahepatic cholangiocarcinoma, known to have a high malignant potential, low *MPC1* expression is correlated with poor prognosis and a significant increase in the percentage of distant metastasis [88]. The most well-known phenomenon associated with metastasis of cancer cells is EMT, which is strongly linked to the function of MPC1. Indeed, MPC1 expression was downregulated in human biliary tract cancer cells undergoing TGF-β (transforming growth factor beta)-induced EMT, and the knockdown of MPC1 expression led to induction of EMT in these cancer cells. These findings support the conclusion that MPC1 functions as a modulator of EMT induction and contributes to the malignant potential of intrahepatic cholangiocarcinoma cells.

Consistent with this conclusion, the study of Takaoka and collaborators found that, in pancreatic and colon cancer cells, EMT was induced following suppression of MPC1 expression [96]. These authors showed that MPC1 and MPC2 knockdown upregulated the glutaminase GLS, inducing EMT. This effect was suppressed in glutamine depletion conditions, revealing glutamine metabolism as an important mechanism inducing EMT. Finally, *MPC1* was found to be downregulated in renal cell carcinoma tissue when compared with adjacent non-cancerous tissue, and lower MPC1 expression correlated with unfavorable prognosis for renal cell carcinoma patients [97]. Functionally, MPC1 suppressed the invasion of renal carcinoma cells in vitro and reduced their growth in vivo by decreasing the expression of the matrix metalloproteases 7 and 9 (MMP7 and MMP9).

#### 4.3.3. MPC and Lactate in Tumor Growth

The consequence for cancer cells of relying on aerobic glycolysis is, firstly, the need to import high levels of glucose to compensate for the loss of ATP production by OXPHOS and, secondly, high amounts of lactate produced in the cytosol by LDH-mediated reduction of pyruvate are released from the cell. The latter leads to acidification of the extracellular microenvironment, which favors metastasis, angiogenesis, and immunosuppression [98]. Thus, lactate can be seen as an oncometabolite in the metabolic reprogramming of cancer cells. However, the role of lactate in cancer is complex. Recently, two groups reported high levels of lactate in the blood of patients with lung cancers [99,100], as well as in a mouse lung cancer model [100]. Interestingly lactate was found to be imported into tumor cells to at least the same extent as glucose, although the consequences of high levels of lactate in cancer cells remains unclear. In particular, it is unclear whether lactate, after conversion into pyruvate, could fuel the TCA cycle, participate in OXPHOS, and/or lead to the synthesis of intermediary metabolites, including acetyl-CoA, which could influence chromatin remodeling and gene expression.

#### 4.3.4. Inhibition of MPC Activity Delays Tumor Growth

Reduced levels and activity of the MPC are associated with the majority of cancer types. Therefore, in most cases, a therapeutic strategy based on MPC would need to promote MPC expression and/or its activity. However, there are two reports to date of tumors in which inhibition of the MPC was shown to delay tumor growth, as well as a third report showing a promising adjuvant effect of MPC inhibition when coupled with radiotherapy.

In one report (see Section 3.1) the androgen-sensitive prostate tumor was shown to require MPC for growth and it should, therefore, be sensitive to MPC inhibition [33]. In a second report, it was shown that liver-specific disruption of MPC in mice decreased development of a hepatocellular carcinoma induced by a low-dose exposure to *N*-nitrosodiethylamine (DEN) plus carbon tetrachloride (CCl4) [101]. In the latter case, MPC-disrupted hepatocytes showed increased glutaminolysis to maintain the TCA cycle, and re-synthesis of glutathione was found to occur at a lower rate because less glutamine was available for glutathione synthesis. These findings raise the possibility of a model in which inducing metabolic competition for glutamine by MPC disruption would impair hepatocellular tumorigenesis by limiting glutathione synthesis. In the third report, MPC inhibition led to decreased oxygen consumption by tumor cells, thereby sparing oxygen locally and reducing hypoxia in the vicinity of the tumor. Importantly, this higher oxygen concentration around cells exacerbated the toxic effects of radiotherapy [102].

All together these results suggest that MPC inhibitors could be useful therapeutically to treat some selected cancer types.

#### **5. Conclusions**

In conclusion, mitochondrial pyruvate transport is essential for normal embryonic development and plays a key role in the function of many organs in the adult. Being at the heart of cell metabolism, MPC activity is solicited in several processes that require the presence of pyruvate inside mitochondria, to drive either cataplerotic reactions, such as gluconeogenesis or lipid synthesis, or anaplerotic reactions, to drive TCA cycle activity and consequently OXPHOS. Although cells can adapt their metabolism to circumvent impaired metabolic changes, this does not always prevent perturbation of cell homeostasis and pathology. We can see that, if preserving or restoring MPC activity is the objective in several pathologies, including certain cancers, we can also show that inhibition of MPC activity could be

beneficial in some pathologies. These pathologies, which include T2D, NASH, and PD, as well as probably others yet to be discovered, may provide diverse therapeutic applications for MPC inhibitors.

**Funding:** This research was funded by the Swiss National Science Foundation (31003A\_179421) and by Oncosuisse (KFS-4434-02-2018).

**Acknowledgments:** We would like to thank all current and recently departed members of the JCM lab for their comments. Special acknowledgment goes to Kinsey Maundrell for helping to revise the manuscript.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Review* **Metabolic Roles of Plant Mitochondrial Carriers**

#### **Alisdair R. Fernie 1, \*, João Henrique F. Cavalcanti <sup>2</sup> and Adriano Nunes-Nesi 3, \***


Received: 24 May 2020; Accepted: 29 June 2020; Published: 8 July 2020

**Abstract:** Mitochondrial carriers (MC) are a large family (MCF) of inner membrane transporters displaying diverse, yet often redundant, substrate specificities, as well as differing spatio-temporal patterns of expression; there are even increasing examples of non-mitochondrial subcellular localization. The number of these six trans-membrane domain proteins in sequenced plant genomes ranges from 39 to 141, rendering the size of plant families larger than that found in *Saccharomyces cerevisiae* and comparable with *Homo sapiens*. Indeed, comparison of plant MCs with those from these better characterized species has been highly informative. Here, we review the most recent comprehensive studies of plant MCFs, incorporating the torrent of genomic data emanating from next-generation sequencing techniques. As such we present a more current prediction of the substrate specificities of these carriers as well as review the continuing quest to biochemically characterize this feature of the carriers. Taken together, these data provide an important resource to guide direct genetic studies aimed at addressing the relevance of these vital carrier proteins.

**Keywords:** amino acid; biological function; ion; inner mitochondrial membrane; mitochondrial carrier family; organic acid; substrate specificity; transport mechanism; vitamin

#### **1. Introduction**

The acquisition of the mitochondrial endosymbiont brought a wide range of novel metabolic capabilities to the ancestral eukaryotic lineage [1]. Alongside efficient synthesis of ATP via the process of oxidative phosphorylation, the mitochondria are also the site of numerous other anabolic and catabolic pathways. The host cell has exploited this and depends on the mitochondria as a source of carbon skeletons for several further metabolic pathways including nitrogen assimilation, photorespiration, C<sup>1</sup> metabolism, photosynthesis in C<sup>4</sup> and crassulacean acid metabolism as well as the utilization of storage pools of carbon and nitrogen during seed germination [2,3]. Mitochondria additionally play roles in the biosynthesis of amino acids, tetrapyrroles, fatty acids and vitamin co-factors [4,5]. In order to achieve this, the mitochondrial matrix needs to be supplied by a wide range of solute transporters. Intriguingly, in a model of enzymes allocated to specific cellular compartments of Arabidopsis, Mintz-Oron et al. [6] revealed that approximately half of the reactions could be assigned to specific subcellular compartments based on experimental evidence. For the remainder, they predicted the most likely subcellular location based on a parsimony principle of minimizing the number of intracellular transporters required to activate the reactions with a known localization in the corresponding compartments [6]. This method predicted that a metabolic network of some 1200 reactions (compartmented among the cytosol, plastid, mitochondrion, endoplasmic reticulum, peroxisome, vacuole and Golgi apparatus) required a phenomenal 772 intracellular transporters. Similarly, at least 228 metabolites and 89 transport processes are required in the minimal human mitochondrial metabolic network [7], suggesting that the

total number of solute transporters currently catalogued to reside in the plant mitochondria may be insufficient to account for all transport steps required [8]. It is, however, important to note that this list contains not only MCF (mitochondrial carrier family) members but also members of other families [8]. Although the outer mitochondrial membrane is permeable to small solutes (with a molecular mass of less than 4–5 Da) [9–11], the inner membrane is impermeable with only very small uncharged molecules such as O<sup>2</sup> and CO<sup>2</sup> able to readily pass through this membrane. The passage of hydrophilic compounds across the inner mitochondrial membrane is mainly catalyzed by the nuclear encoded mitochondrial carrier family (MCF) [12–15]. MCs (mitochondrial carriers) are small proteins ranging in size from 30–34 kDa and possess common defining structural features. Their primary structure is characterized by three tandemly repeated, approximately 100 amino acid long, homologous domains with each repeat containing two hydrophobic segments, which span the membrane, and a characteristic amino acid sequence motif PX[D/E]XX[K/R]/RX[K/R] (20–30 residues) [D/E]GXXXX[W/Y/F][K/R]G (PROSITE PS50920, PFAM PF00153 and IPR00193). Two sub-families, the aspartate/glutamate and ATPMg-Pi carriers, have additional N-terminal regulatory domains of approximately 150 amino acids that usually contain Ca2+-binding motifs [12,16]. The molecules transported by the MCF are highly variable in size and structure, ranging from H<sup>+</sup> and NAD<sup>+</sup> and coenzyme A. They also display a range of ionic charges being either positive, negative or zwitterionic at physiological pH. They often act as antiporters, although uniport transport and H+-compensated anion symport is also mediated by some MCs. Furthermore, MCs can be subdivided on the basis of their electrical nature with for example the ADP/ATP and aspartate/glutamate transporters drive electrogenic reactions (which result in net charge transfer) whereas the carrier subfamilies for phosphate (Pi), glutamate, and GTP/GDP as well as for 2-oxoglutarate and ornithine are electroneutral.

Considerable research has been conducted on characterizing members of the MCF in both yeast and animals (see [13,14,17–27] for reviews). Regarding other eukaryotes, the MCF members of the early-branching kinetoplastid parasite *Trypanosoma brucei* have been studied by sequence and phylogenetic analyses [27]. This study gave new insights into the evolution and conservation of the 24 identified MCF homologues identified in that organism [27]. In recent years the advent and exploitation of systems biology approaches have provided considerable insight into the putative in vivo function of plant MCFs, whilst the adoption of recombinant enzyme approaches have allowed the biochemical characterization of their functions. In this article, we summarize the structure and transport mechanisms of members of the MCF family, discuss its expansion in plants and finally summarize the biochemical characterization of the transport properties of MCF members that have been reconstituted in liposomes. The reader is referred to our other article in this issue [28] for information concerning the, sometimes unusual, subcellular localization of these proteins and the characterization of transgenic loss-of-function lines.

#### **2. Structure and Transport Mechanisms**

Due to its high abundance, the ADP/ATP carrier (AAC) is the member of the family that has been studied the most. It is the first mitochondrial carrier for which a high-resolution X-ray structure was provided [29]. The bovine carrier was crystallized in the presence of a strong inhibitor, the carboxyatractyloside (CATR). The structure gives an insight not only into the overall fold of mitochondrial carriers in general but also into atomic details of the AAC in a conformation that is open toward the intermembrane space (IMS). The three dimensional structure of the ADP/ATP carrier is critical to our understanding in several other ways. First, it exhibits a three-fold pseudo-symmetry in lines with the three-fold sequence repeats mentioned above [30], similar to that observed by electron microscopy [31]. Secondly, it approximately corresponds to the *c* (cytosolic)-state of the ADP/ATP carrier since CATR blocks the carrier in this state [19]. Thirdly, this structure has become a much used template for the building of homology models of various carriers, greatly improving our understanding of MC structure/function relationships [32–38]. The structural fold observed in the bovine transporter was subsequently confirmed by the structures of two yeast isoforms of the

ADP/ATP carrier. Intriguingly, the odd-numbered transmembrane alpha-helices have pronounced kinks at the Pro residue of the highly conserved signature motif Px[DE]xx[KR], with the P kink giving them a pronounced L shape which helps to block access of the central cavity from the mitochondrial matrix in the *c*-state (Figure 1). The charged residues of the signature motifs from inter-domain salt bridges [29,39] are now known as the matrix salt bridge network. Furthermore, residues of the salt bridge interact with a proximal glutamine residue which hydrogen bonds to both salt bridge residues forming a glutamine brace (Q brace [39]). These residues are highly conserved with one to three Q braces typically found in the SLC25 subfamily of the MCF. CATR inhibits the ADP/ATP carrier by binding tightly in the central cavity thus prevent translocation across the membrane. Just last year, the first structure of a mitochondrial carrier in the *m*-state has been solved—the ADP/ATP carrier from the thermotolerant fungus *Thermothelomyces themophilia* inhibited by bongkrekic acid [40]. The *m*-state structure displays the same three-domain architecture, but with the domains rotated compared with the *c*-state, opening the central cavity to the mitochondrial matrix and closing it to the intermembrane space and thereby disrupting the matrix network and Q brace [14]. On the intermembrane side the transmembrane helices are positioned closely together and form the interdomain cytoplasmic salt bridge network (Figure 1). This network is stabilized by hydrogen bonds with the hydroxyl bonds with the hydroxyl groups of the tyrosines of the motif forming a tyrosine brace (Y brace; [39]). Most SLC members have one to three Y braces. Having structural information for both *c*- and *m*-states has significantly advanced our understanding of how these proteins operate at the molecular level. These advances have been excellently covered elsewhere [25], so we will not detail them here except to say that the structural features are likely to be conserved, with few exceptions, throughout the MCF.

**Figure 1.** Mechanism of substrate translocation catalyzed by mitochondrial carriers. Simplified scheme depicting the transition of mitochondrial carriers from the *c*-state to the *m*-state and vice versa as previously proposed [14]. The trapezoid shape on the left is used to illustrate the *c*-state after the release of the substrate towards the cytosol and immediately after the entry of the substrate from the cytosolic side; the trapezoid shape on the right illustrates the *m*-state after the release of the substrate into the matrix and immediately after the entry of the substrate from the matrix side; and the two central hexagonal shape solids depict the transition states (*t*-state) of the carrier with the bound substrate entered from the cytosol and from the matrix. The yellow disk and green rectangle shapes represent the substrates entering from the cytosol and from the matrix, respectively; orange triangles represent closed gates, and dotted orange triangles indicate open or partially closed gates. All transport steps are fully reversible. The positions of the salt bridge networks (cytosolic and matrix gates), P-G level 1, substrate binding site and P-G level 2 are indicated on the right.

#### **3. Extension of the MCF**

Although only six MC proteins were sequenced following their purification from mitochondria or by DNA sequencing (see [41] and references therein) the genomic era has massively expanded our inventories of the MCFs of various species with *S. cerevisiae* encoding 35 [42], the human genome 50 [18] and *Arabidopsis thaliana* 58 [5]. The first step in identifying MC function is to search for the substrates transported by a specific carrier. In order to do so, the primary tools in our arsenal are phylogenetic clustering, genetic information, knowledge of cellular metabolism and complementation of phenotypes. However, such methods remain inconclusive and overly speculative. To date, the most effective strategy has been heterologous expression in *Escherichia coli* (see for example [43]) or *S. cerevisiae* (see for example [44–46]) and reconstitution of the subsequently purified recombinant carriers into liposomes in which direct transport assays are performed. To date, such gene-function studies have been carried out on 32, 40 and 26 of the MCFs of *S. cerevisiae*, human and *A. thaliana*, respectively. Focusing on the green lineage alone, MCs are highly abundant in the genomes of several species of dicots, monocots and algae with a 2020 update (Table 1 and Table S1) suggesting they range in number from 39 to 141, surpassing the 37 to 125 range when we last reviewed this family in 2011 [12]. In turn, a reasonable understanding of the apparently increasing number of predicted MCF genes in green line in comparison with the report of 2011 [12] is the development of powerful tools concerning both next generation sequencing approaches and/or bioinformatics algorithms for annotation and assembly of new, more accurate *de novo* reference genomes. In this review we demonstrate how the function of *A. thaliana* proteins could be reasonably, yet not completely, accurately predicted via examination of their symmetry-related triplets and subsequent comparison to those of MC subfamilies for which substrate specificities were determined in human, yeast or Arabidopsis itself. Those instances in which poor accordance was found between the prediction and experimental results can be split in two: those displaying novel substrate specificity and those residing at different subcellular localization [28]. Table 2 lists the main subfamilies that MCs can be partitioned into on the basis of their substrate specificities. It is, however, important to note that the caveats which we previously mentioned [12] remain valid. In brief: (i) some substrates are transported by more than one subfamily; (ii) the best transported substrate in reconstituted liposomes may not reflect the most important substrate under physiological conditions; (iii) some subfamilies may additionally transport as yet untested substrates (see for example [47]); and (iv) most of the subfamilies presented in this table are present in all eukaryotes. It has been suggested that key amino acids residues important for the transport mechanism are likely symmetrical, and those involved in substrate binding are likely asymmetrical (indicating the asymmetry of the substrates) [32–38]. Hence, scoring the symmetry of residues in the sequence repeats, it is possible to associate the substrate-binding sites and salt bridge networks that are important for the transport mechanism in family members [32–38]. Thus, the substrate specificity defined carrier subfamilies are also characterized by specific amino acid triplets with the number of characterizing triplets ranging from two to eight. Moreover, related subfamilies sharing some triplets, for example the NAD+, PyC and FAD families, share triplet 19 as well as some transport substrates [48,49] whereas the OGC and DTC subfamilies also share two triplets (KLK and GTY) as well as some transported substrates [43,50] substrate specificity. Finally, as in our previous study, the uncoupling protein (UCP) and unnamed transporters have been added despite the fact that the substrates are unknown for the latter. However, in contrast to our previous study [12], as detailed below, the substrate specificities of UCP have recently been characterized [51].


**Table 1.**Mitochondrial carriers (MCs) present in each chromosome of plant genomes recently sequenced.


**Table 1.** *Cont*.

The data were retrieved from comparative genome platform Plaza (https://bioinformatics.psb.ugent.be/plaza/), Ensembl-plants (http://plants.ensembl.org/index.html) and Phytozome (http://www.phytozome.net/). The InterPRO domain used for mitochondrial carrier was 'IPR023395'. All the sequences were validated by protein blast analysis on the non-redundant database (http://blast.ncbi.nlm.nih.gov/Blast.cgi). The number of MCs refers to sequences which are longer than 265 amino acids and non-redundant. chr, chromosome; mbp, mega base pairs.

An additional method for analyzing function has been the deduction from phylogenetic trees—an approach that has also been used to address the evolution of the MCFs [12,52]. Intriguingly, either if all MCF members of a single representative of a kingdom [12] or multiple representatives of each kingdom but only a subset of the MCFs are used [52] similar broad conclusions can be made. These are namely that the MCFs are highly divergent, yet that the fact that the vast majority are conserved across plants, animals and yeast lineages suggests that many MC functions existed before the speciation events that have produced the three kingdoms [12]. Interestingly, however, the comparison of intraspecific paralogs suggests that these originated by gene duplication events that occurred independently in those three lineages [12]. At a finer level the comparison of tricarboxylic acid (TCA) cycle relevant MCFs alone revealed that mitochondrial organic acid transporters formed two distinct clades. In the first clade, dicarboxylate carriers (DICs) and dicarboxylate-tricarboxylate carrier (DTC) grouped with 2-OG carriers (OGCs). The succinate-fumarate carrier (SFC) formed the second organic acid clade with other non-plant organic acid transporters including oxodicarboxylate carriers (ODCs), citrate carriers (CiCs), and yeast suppressor of HM (histone-like proteins in yeast mitochondria) mutant 2 (YHM2). Biochemical data would indicate that DTC and CiC must be closely related as they both transport citrate; phylogenetic analysis revealed that SFC, and not DTC, is more similar to CiC [52]. Based on available biochemical data, it thus appears that the transport functions of CiC and DTC have evolved independently but perhaps convergently. As would be expected, the possibility to use phylogeny to detect orthologs between plants is much greater than across kingdoms [12]. The use of tools such as Orthofinder, which provides phylogenetic inference of orthologs [53], and PlaNet and FamNet, which include co-expression data to refine such searches [54,55], render such searches easier and will likely prove highly informative in improving our understanding of plant MCFs beyond Arabidopsis.

Further evolutionary insight into plant MCF members was attained by studying the location of introns in MC genes and examining the synteny between MCF members in the dicot *A. thaliana*, the monocot *Brachipodium distachyon* and the algae *Osterococcus luminarius* [12]. The first of these strategies took its cue from the observation that introns tend to interrupt the coding sequence of the human citrate, carnitine and dicarboxylate carrier genes at positions corresponding to or in the close vicinity of the hydrophilic loops in the MC amino acid sequences [56–58]. Comparison of all 58 members of the *A. thaliana* MCF revealed that hydrophobic loops host a notable excess in intron density compared with transmembrane helices, suggesting that introns are unrepresented in transmembrane helices due to negative selection [12]. In the second analysis, the co-linearity of regions of the *A. thaliana*, *B. distachyon* and *O. luminarius* genomes were exploited. In *O. luminarius* six MCF genes were found in co-linear regions of chromosomes 13 and 21, consistent with the known origin of chromosome 21 in this species [59]. In spite of several whole genome duplication events (see [60,61] for reviews) the same number of gene pairs were found in Arabidopsis, a fact best explained by high rate of gene loss and gene rearrangement in this species [62,63]. By contrast, the genome of *B. distachyon* contains six pairs of MC paralogs in co-linear segments, likely reflecting a lower rate of gene loss and gene rearrangement in this species. Fascinatingly, however, even though approximately 500 million years separate monocots and dicots from the common ancestor of the angiosperm, 15 MCs in Arabidopsis and 13 in Brachipodium are present in conserved synteny blocks with an over-representation for nucleotide carriers being apparent which has been suggested to reflect either that their preferential expansion is tolerated in angiosperms or, more likely, that they functionally contributed to angiosperm evolution [12].

#### **4. Biochemical Characterization of Plant MCF Members**

Out of 58 MCF members found in Arabidopsis genome, 17 genes have not been fully characterized and therefore the biochemical role of these proteins remains unknown. In the last 10 years, the biochemical functions of 21 MCs from Arabidopsis have been investigated (Table 2) and studies on the characterization of the physiological importance of these carriers in plants have been reported [28].

#### *4.1. Coenzyme A Carriers*

From the subfamily of nucleotides and dinucleotides carriers, two genes encoding for MCF proteins, At1g14560 and At4g26180, based on the presence of sequence motifs (symmetry-related amino acid triplets [12]) were described as potential coenzyme A (CoA) carriers [12]. Comparative genomic analysis allowed the identification of two homologs of these proteins in maize (*Zea mays*; GRMZM2G161299 and GRMZM2G420119) [64]. It was verified that all these proteins from maize and Arabidopsis are targeted to mitochondria and are also able to complement the growth wild type phenotype in the yeast leu5D mutant [65] defective for the mitochondrial CoA carrier [64]. These proteins also restored the mitochondrial CoA level in the same yeast mutant. These results clearly demonstrated that these proteins catalyze the transport of CoA through the mitochondrial membrane. It is noteworthy that, to our knowledge, the substrate specificity of this transporter has not yet been fully investigated. This is particularly important, because in addition to CoA, these transporters might also have capacity to transport other substrate or substrates. In this regard, it was reported that the Arabidopsis peroxisomal NAD carrier PXN, in addition to NAD+, NADH, AMP, ADP and adenosine 3', 5'-phosphate (PAP), is also able to catalyze CoA transport [66]. PXN is encoded in Arabidopsis by the gene At2g39970 and was investigated regarding its substrate specificity and the transport properties by using a wide range of potential substrates [66]. Detailed biochemical analyses demonstrated that PXN catalyzes fast counter-exchange of substrates and much slower uniport [66]. In the same study, it was shown that the transport catalyzed by PXN is saturable with a submillimolar affinity for NAD+, CoA and other substrates. More recently, the physiological function of PXN in plants was further investigated [67]. Interestingly, by using *S. cerevisiae*, uptake analyses indicated that PXN has a low affinity for CoA, which suggests that the PXN function of the CoA transporter might not be possible under physiological conditions. Complementing diverse mutant yeast strains with PXN and investigating the suppression of the mutant phenotypes, the authors provided evidence that PXN is not able to function as a CoA transporter or a redox shuttle by mediating a NAD+/NADH exchange, but instead catalyzes the import of NAD into peroxisomes against AMP in intact yeast cells [67]. This work demonstrated that Arabidopsis PXN supplies the peroxisomes with NAD by importing this coenzyme from the cytosol in exchange with AMP.

#### *4.2. Nicotinamide Adenine Dinucleotide (NAD) Carriers*

Regarding NAD transport in mitochondria and plastids, in addition to PXN, it has been demonstrated that two MCF members in Arabidopsis, named *At*NDT1 and *At*NDT2, are able to catalyze the import of NAD in these organelles [68]. Both carriers are able to complement the phenotype of a yeast mutant lacking NAD<sup>+</sup> transport [68]. Surprisingly, both *At*NDT1 and *At*NDT2 exhibit similar substrate specificity, being able to import NAD<sup>+</sup> against ADP or AMP, and not accepting NADH, NADP+, NADPH, nicotinamide or nicotinic acid as transport substrates [68]. Intriguingly, despite the similarities in terms of biochemical properties, initial localization analysis indicated that *At*NDT1 was located in the plastid membrane while *At*NDT2 was in the mitochondrial membrane [68]. Surprisingly, *At*NDT1 was found in mitochondrial membranes in proteome studies [69] and previously a GFP-tagging and immunolocalization study was not able to find *At*NDT1 targeted to chloroplast membranes [70]. Very recently, both *At*NDT1- and *At*NDT2-GFP fusion proteins were found exclusively located in the mitochondria, clearly indicating their mitochondrial localization [71].

#### *4.3. Adenylate Carriers*

The transport catalyzed by the ADP/ATP carrier plays an important role in sustaining the cellular ATP homeostasis by facilitating the counter exchange of mitochondrial ATP for cytosolic ADP [72]. ADP/ATP carrier proteins have been identified and characterized in different species including organisms of medical and veterinary importance, such as *T. brucei* [27,73,74]. The importance of the efficient adenylate transport systems for intracellular energy partitioning between the cell organelles has

been widely demonstrated in plants (for review see [72]). Adenylate carriers found in different organelles have been previously identified and biochemically characterized in plants. There are three subgroups of MCF responsible for adenylate transport in plants: (1) the well-characterized ADP/ATP carriers, named AAC carriers (*At*AAC1, At3g08580; *At*AAC2, At5g13490; and *At*AAC3, At4g28390), which are required for mitochondrial energy passage (for review see [72]) and represent the most abundant proteins in the inner mitochondrial membrane (*At*AAC1–3; 53,065 protein copies/mitochondria [75]); (2) the mitochondrial ATP-Mg/phosphate carriers, named as APC carriers (*At*APC1, At5g61810; *At*APC2, At5g51050; and *At*APC3, At5g07320); and (3) the adenine nucleotide transporter ADNT1 (At4g01100), which transport AMP instead of ADP as counter exchange substrate of ATP [76].

In Arabidopsis there are three genes encoding putative APC proteins (*At*APC1–3). These proteins belong to the MCF and exhibit high amino acid sequence similarities to their human and yeast counterparts [12,19]. It was demonstrated that all APC proteins from Arabidopsis localize to mitochondria and restore the growth phenotype of APC yeast loss-of-function mutants [77]. Interestingly, these carriers interact with calcium (Ca2+) via their N-terminal EF-hand motifs *in vitro*, suggesting that APC1–3 isoforms represent Ca2+-regulated ATP-Mg/phosphate transporters. Insights into the biochemical characteristics of these APCs were reported based on reconstitution of heterologously expressed proteins into liposomes [16,78]. The obtained results demonstrated that Arabidopsis APCs mediate antiport of ATP, ADP and phosphate and the transport characteristics indicated that the plant APCs preferentially import the Ca2+- and not the Mg2+-complexed form of ATP, at least in an *in vitro* system [78]. It is important to note that recent evidence indicates that not only Mg2<sup>+</sup> and Ca2+, but also other divalent cations and specifically Mn2+, Fe2+, Zn2<sup>+</sup> and Cu2+, are transported together with ATP by human and Arabidopsis APCs [79].




#### **Table 2.** *Cont*.

**Abbreviations:** AAC, ADP/ATP carrier; AGC, aspartate/glutamate carrier; ANT, peroxisomal adenine nucleotide translocator; APC, ATP-Mg/Pi carrier; CAC, carnitine carrier; CoA/PAP, coenzyme A/adenosine 3',5'-diphosphate carrier; BOU, A bout de soufflé (glutamate transporter); CTP, citrate carrier; DIC, dicarboxylate carrier; DTC, di-/tri-carboxylate carrier; FAD, FAD carrier; GC, glutamate carrier, GGC, GTP/GDP carrier; NDT, NAD<sup>+</sup> carrier; OAC, oxaloacetate/sulfate carrier; ODC, oxodicarboxylate carrier; OGC, oxoglutarate carrier; ORC, ornithine carrier; PiC, phosphate carrier; mPT, mitochondrial phosphate carrier; PNC, pyrimidine nucleotide carrier; SAMC, S-adenosylmethionine carrier; SFC, succinate/fumarate carrier; TPC, thiamine pyrophosphate carrier; UCP, uncoupling protein. AXP, adenine nucleotides; dNDP, deoxynucleoside diphosphates; dNTP, deoxynucleoside triphosphates; PEP, phosphoenolpyruvate; Pi, phosphate; ThMP, thiamine monophosphate; ThPP, thiamine pyrophosphate. \* Symmetry-related amino acid triplets are the triplet sets present in the functionally identified mitochondrial carriers of each family.

#### *4.4. Amino Acid Carriers*

Another enigmatic transporter named A BOUT DE SOUFFLE (BOU) was identified in Arabidopsis At5g46800 a long time ago [117]. Previous studies extensively characterized the physiological function of the BOU transporter in plants and revealed that this protein plays important roles related to fatty acid β-oxidation [117], photorespiration and growth of meristem cells [118]. However, the specific substrate for the BOU transporter protein was unknown until recently [104]. Detailed biochemical characterization of Arabidopsis BOU and Ymc2p, the BOU homolog from *S. cerevisiae*, revealed the transport properties and kinetic parameters of these proteins. Both Ymc2p and BOU proteins are able to transport glutamate, and to a lesser extent L-homocysteinesulfinate, but no other amino acids nor many other tested metabolites [104]. This study also revealed that both proteins Ymc2p and BOU catalyze unidirectional transport of glutamate and, as reported for other known MCs, a faster counter exchange mode of transport, and catalyze a transmembrane glutamate<sup>−</sup> + H<sup>+</sup> symport. These results led to the conclusion that for both Ymc2p and BOU, the physiological function of these proteins is to catalyze the import uptake of glutamate into the mitochondria.

In Arabidopsis, two MCs (*At*BAC1, At2g33820; and *At*BAC2, At1g79900) are able to transport basic amino acids [106,119,120]. *At*BAC1 shares a 36% identity with BOU, whereas *At*BAC2 is 40% similar to SLC25A29, although it is also related to BOU (36% identity) and aspartate/glutamate carriers (AGCs, 30–33% identity) [22]. Recombinant proteins from *At*BAC1 and *At*BAC2 were purified and reconstituted in liposomes [106,120]. The results indicated that both proteins transport lysine, arginine, ornithine and histidine [106,120]. Interestingly, it was verified that only *At*BAC2 transports the neutral amino acid citrulline [106,120]. In addition, these studies indicated that *At*BAC1 and *At*BAC2 exhibit differences in terms of substrate specificity, with *At*BAC2 being less specific for L-amino acids. Despite the similar biochemical properties, the physiological roles of *At*BAC1 and *At*BAC2 seem to be different. While *At*BAC1 is likely involved in remobilization of storage compounds after seed germination in

Arabidopsis and rice [106,119,121], *At*BAC2 is more related with stress responses being expressed especially in responses to hyperosmotic stress and also during senescence [106,119,122,123].

#### *4.5. Uncoupling Proteins*

Uncoupling proteins (UCPs) have been described as being involved in dissipation of proton gradients across the inner mitochondrial membrane that is normally used for ATP synthesis [92–124]. Homology analysis with UCP from humans revealed that six genes in the Arabidopsis genome (*At*UCP1–6) encode putative UCPs [124–126]. It was previously demonstrated that the isoform *At*UCP1 (At3g54110) is localized to mitochondria and exhibits the activity of an uncoupling protein similar to the human UCP1 [124–126]. The function of the isoform *At*UCP2 (At5g58970) was less understood until recently because it was detected in the Golgi apparatus [127] and also in the plasma membrane [128]. Recently, it was shown that *At*UCP2 isoform is also a mitochondrial localized protein [51]. Intriguingly, both isoforms *At*UCP1 and *At*UCP2 were shown to transport amino acids (glutamate, aspartate, cysteine sulfinate, and cysteate), dicarboxylates (malate, oxaloacetate, and 2-oxoglutarate), phosphate, sulfate, and thiosulfate [51]. Further biochemical analyses revealed that both isoforms catalyze an electroneutral aspartate/glutamate heteroexchange activity, in contrast to that mediated by the mammalian mitochondrial aspartate glutamate carrier. Three other former members of the *At*UCP subfamily of Arabidopsis MCF (*At*UCP4–6) were renamed as dicarboxylate carriers (DIC) (*At*DIC1, At2g22500, *At*DIC2, At4g24570; and *At*DIC3, At5g09470) since these proteins are able to transport oxaloacetate, malate, succinate, phosphate, sulfate, thiosulfate and sulfite [93].

#### *4.6. Dicarboxylate Carriers*

As mentioned above, in the Arabidopsis genome three potential homologues of yeast and mammalian mitochondrial DICs were found and designated as *At*DIC1–3 (AT5G09470) [93]. *At*DIC3 shares only 55–60% identical amino acids with *At*DIC1 and *At*DIC2, whereas *At*DIC1 and *At*DIC2 share 70% identical amino acids, suggesting that *At*DIC1 and *At*DIC2 are more closely related [93]. Interestingly, a recent Arabidopsis mitochondrial proteomic study verified that *At*DIC3 is not highly expressed in comparison with *At*DIC1–2, as *At*DIC1 is more abundant than *At*DIC2 (59 and 21 protein copies per mitochondria respectively) [75]. Transport experiments with recombinant and reconstituted *At*DIC proteins demonstrated that the substrate specificity of these proteins is unique to plants, indicating the combined characteristics of the DIC and oxaloacetate carrier in yeast [93]. Indeed, the Arabidopsis DICs transport a wide range of dicarboxylates including malate, oxaloacetate and succinate as well as phosphate, sulfate and thiosulfate at high rates, whereas 2-oxoglutarate was revealed to be a very poor substrate. In the same study, the kinetic properties of recombinant *At*DIC1–3 proteins were determined [93]. It was shown that for all *At*DIC proteins, Vmax is not significantly different for the three substrates tested (malate, sulfate and phosphate). Nevertheless, the Vmax for *At*DIC3 was higher than the values observed for *At*DIC1 and *At*DIC2. Regarding the transport affinity (Km) of *At*DIC1–3 proteins, for sulfate it was lower than the Km values for phosphate and malate. For *At*DIC3, it was verified that the Km for sulfate was one order of magnitude lower than the Km values for malate and phosphate; furthermore, the Km of *At*DIC3 for sulfate was 3–4-fold lower than the Km values of *At*DIC1 and *At*DIC2 using the same substrate. The identification and characterization of the biochemical properties of DIC proteins in Arabidopsis led to different questions about the physiological roles of these carriers in plants under distinct physiological conditions. Surprisingly, according to our current knowledge, the isolation and characterization of mutant plants for each *At*DIC isoform still need to be performed.

#### *4.7. Dicarboxylate*/*Tricarboxylate Carrier*

Dicarboxylate/Tricarboxylates carriers (DTCs) are mitochondrial transporters that are able to transport both dicarboxylic acids (such as malate, maleate, oxaloacetate and 2-oxoglutarate) and tricarboxylic acids (such as citrate, isocitrate, *cis*-aconitate and *trans*-aconitate) [50]. In the human parasite *Trypanosoma brucei*, it was demonstrated that a plant-like mitochondrial carrier family protein, named *Tb*MCP12, is able to transport both dicarboxylates and tricarboxylates across the inner mitochondrial membrane (IMM) [129]. Silencing this carrier in *T. brucei* was not lethal, while its overexpression was deleterious. These results indicated that the intracellular abundance of *Tb*MCP12 is involved in the regulation of NADPH balance and mitochondrial ATP-production. In plants, it was recently demonstrated that DTCs are the most abundant mitochondrial carrier proteins in the IMM of Arabidopsis, comprising 0.8% of the total IMM area (6836 protein copies per mitochondria) [75]. Interestingly, unlike the other three more abundant carrier proteins in the IMM, i.e., ADP/ATP carriers (*At*AAC1–3; 53,065 protein copies/mitochondria), mitochondrial phosphate carriers (*At*MPT2–3; 21,325 protein copies/mitochondria) and uncoupling proteins (*At*UCP1–3; 8595 protein copies/mitochondria), only one DTC homolog is found in Arabidopsis (At5g19760). In addition to Arabidopsis, DTCs have been described in several plant species including tobacco (*Nicotiana tabacum*) [50], grapes (*Vitis vinifera*) [130] citrus (*Citrus junos*) [131], Jerusalem artichoke (*Helianthus tuberosus*) [132] (and maize (*Zea mays*)) [133]. Surprisingly, the numbers of DTC homologs found in different plant species vary without a clear pattern, for example, in the Brassica genus, the number of DTC homologs varies from one in *A. thaliana* and *Arabidopsis lyrata*, two in *Brassica oleracea*, and three in *Brassica rapa* [52]. In tobacco, four homologs (*Nt*DTC1–4) were identified [50].

For *At*DTC and *Nt*DTCs, the transport activity involves an obligatory electroneutral exchange of dicarboxylates such as malate and 2-oxoglutarate and tricarboxylates such as citrate [50]. In addition to catalyzing the dicarboxylate/tricarboxylate transport activity, it has been demonstrated that DTCs are able to catalyze homoexchange transport activities, such as dicarboxylate/dicarboxylate and tricarboxylate/tricarboxylate [50]. It is unclear so far which of these modalities are relevant in in vivo plant systems. From *in vitro* transport assays it is possible to conclude that DTCs are promiscuous in terms of transported substrates [50]. In the same study, it was observed that the highest DTC activities are in the presence of internal 2-oxoglutarate, malate, maleate, oxaloacetate, succinate or malonate. Intriguingly, it was observed also that citrate, isocitrate, *cis*-aconitate, *trans*-aconitate, and sulfate were exchanged for external 2-oxoglutarate, although to a slightly lower extent than the dicarboxylates [50]. Any significant exchange was observed using internal fumarate, phosphoenolpyruvate, phosphate, pyruvate, glutamate, aspartate, glutamine, carnitine, ornithine, or ADP [50]. Together, these results demonstrated that DTCs are able to transport several intermediates of the TCA cycle, with the exception of succinyl-CoA and fumarate for which there is no available information. Another interesting characteristic of DTCs is the pH dependence. It was demonstrated that DTC-mediated oxoglutarate and citrate homoexchanges were dependent on pH, as the oxoglutarate/oxoglutarate and citrate/citrate exchanges increased on decreasing the pH from 8.0 to 5.5 for both *Nt*DTC1 and *At*DTC. For *At*DTC, the homoexchange kinetic constants measured for different substrates in two different pH values indicated that regardless of the substrate, the Km and Vmax varies as a function of pH value. Interestingly, the Km values were increased at pH 7, suggesting that the substrate affinities were reduced; Vmax values were also decreased at pH 7. Of note, the modulation of transport kinetics by pH is highly important for plant metabolism because it has been demonstrated for Arabidopsis that in the mitochondrial matrix the pH is around 8.1 and that in the cytosol the pH is close to 7.3 [134].

#### *4.8. Succinate*/*Fumarate Carriers*

In Arabidopsis, one of the MCF members (At5g01340), named as SFC1 carrier, exhibits 35% similarity with the ACR1 transporter from yeast [96]. The yeast SFC1 is able to transport fumarate, succinate, methylfumarate, 2-OG and OAA against [14C]oxoglutarate [96]. The SFC1 transporter was further shown to prefer succinate and fumarate as substrates since the presence of either substrate almost completely inhibits fumarate/[ <sup>14</sup>C]oxoglutarate exchange [96]. The Arabidopsis SFC1 homolog complemented the *arc1* yeast mutant re-establishing the yeast growth in minimal media with ethanol as the sole carbon source [95]. Despite the predictions and preliminary biochemical information in plants, the biochemical evidence in favor of succinate/fumarate transport is still lacking. Moreover, recently the

SFC1 sequence from Arabidopsis was expressed in *E. coli* and protein was purified and reconstituted in liposomes [94]. Surprisingly, the results of transport properties and kinetic parameters revealed that *At*SFC1 transports mainly citrate, isocitrate and aconitate and, to a lesser extent, succinate and fumarate. Furthermore, it was demonstrated that the *At*SFC1 carrier catalyzes a fast counter-exchange transport and low uniport of substrates, as well as exhibiting a higher transport affinity for tricarboxylates than dicarboxylates [94]. Intriguingly, there have been both reports and model predictions in Arabidopsis showing net influx of succinate to the mitochondria, which would have been expected as succinate is the preferred substrate of non-plant SFCs. Thus, it is likely that another unidentified transporter is using succinate as a counter-substrate to facilitate fumarate transport.

#### *4.9. Phosphate Carriers*

Apart from ADP, the transport of phosphate (Pi) through the IMM is essential for the oxidative phosphorylation of ADP to ATP. In Arabidopsis, three genes encodes mitochondrial Pi carriers, namely *At*MPT1 (or PiC3; At2g17270), *At*MPT3 (or PiC1; At5g14040) and *At*MPT2 (or PiC2; At3g48850), and all of them are related to mitochondrial Pi carrier (PiC) from human and yeast [12,105]. Biochemical studies demonstrate that Arabidopsis PiC1 and PiC2 complement yeast mutants deficient in mitochondrial Pi import [106,107], thus confirming that these proteins act as PiCs. Surprisingly, the role of the Arabidopsis PiC3, which is more distantly related to the other PiC1 and 2 plant isoforms [12,135] remains to be elucidated. Interestingly, ADP/ATP carriers (*At*AAC1–3; 53,065 protein copies/mitochondria) and PiC1–2 (or *At*MPT2–3; 21,325 protein copies/mitochondria) are the most abundant proteins in the IMM [75]. In agreement, it has been proposed that PiCs in the inner mitochondrial membrane are able to physiologically interact with AAC transporters, catalyzing a Pi/H<sup>+</sup> symport (or Pi/OH<sup>−</sup> antiport) and thus supplying phosphate required for the ATP synthesis [136,137]. Recently, it was shown in Arabidopsis that a putative Pi transporter interacts with TCA cycle enzymes [138,139]. Notwithstanding, the significance of these protein–protein interactions at physiological levels remains to be elucidated.

#### *4.10. Pyruvate Carriers*

Pyruvate, the final product of glycolysis in the cytosol, must be transported into mitochondria to supply the carbon skeletons for oxidative metabolism through the TCA cycle reactions. The transport of pyruvate through the IMM must be performed by specific carriers. While candidates for mitochondrial pyruvate carriers (MPCs) have not been identified in the classic MCF yet, the identity and functionality of a series of MPCs, non-MCF members, have been reported in yeast, *T. brucei*, drosophila, mouse and humans [140–142]. The biochemical properties of MPCs have been extensively studied and expertly reviewed [143–146] mainly due to the research efforts to understand the importance of MPCs in metabolism-related human diseases. Furthermore, in *S. cerevisiae* it was demonstrated that MPC is a hetero-dimer in its functional state providing the basis for the structure elucidation of the functional complex [147]. In plants, the biological functions and molecular mechanisms involving MPCs are not well understood. Bioinformatics analysis suggests that a protein named NRGA1, a negative regulator of guard cell abscisic acid (ABA) signaling (At4G05590), shares homology with the MPC2 proteins from yeast, drosophila, human and mouse [148]. Besides NRGA1 protein, four other MPC candidates are encoded by the Arabidopsis genome [149]. This family of MPCs from Arabidopsis are phylogenetically classified into three categories: MPC1 (At5G20090), MPC2-like proteins (At4G14695, At4G22310 and At4G05590) and At4G26780 [149]; that said, little is known regarding the functions of these proteins. So far it is known that NRGA1 is located in the mitochondria and its sequence exhibits transmembrane domains [148]. Furthermore, in Arabidopsis this putative MCP2-like protein seems to be involved in stomata ABA signaling [148]. Interestingly, a recent study demonstrated that *At*MPC1 interacts with NRGA1 and plays a role in the regulation of stomatal movement and pyruvate cellular content [150]. In addition, it was demonstrated with yeast MPCs that by mimicking the physiological pH gradient between the mitochondria and the cytosol, a quantifiable pyruvate transport

was observed, whilst in the absence of the pH gradient no transport of pyruvate was observed [147]. Recently, it was demonstrated that the formation of *At*MPC protein complexes is required for cadmium (Cd) tolerance and also prevention of Cd accumulation in Arabidopsis [151]. In the same study, it was demonstrated that *At*MPC complexes are composed of two elements, the *At*MPC1 and *At*MPC2 (*At*NRGA1 or *At*MPC3). Interrupting the formation of *At*MPCs by silencing *At*MPC1 element, the synthesis of acetyl-coenzyme A was supplemented by glutamate and thus sustaining the activity of TCA cycle reactions and glutathione synthesis following exposure to Cd stress [151]. Clearly, more molecular, biochemical and physiological research efforts are still needed to understand the transport mechanism, substrate specificities and physiological roles of mitochondrial pyruvate transporters in plants.

#### *4.11. Iron Transporters (Mitoferrins)*

Initially, mitochondrial iron (Fe) transporters, namely Mitoferrins (mIT), were identified and characterized in drosophila, zebrafish and humans [152–154]. Plants homologs of mIT were first identified in rice [155] and, recently, two genes encoding for mIT were found in Arabidopsis, and named as *At*mIT1 (At2g30160) and *At*mIT2 (At1g07030) [116]. These proteins have an identity of 81% with each other at the amino acid level and share 38% sequence identity with yeast and 32% identity with zebrafish mIT [116]. In addition, both *At*mIT1 and *At*mIT2 proteins exhibit the classical MCF characteristic feature and were predicted to localize to the mitochondria by proteomic study [156] which has been confirmed by subcellular localization experiments with green fluorescent protein (GFP) fusions and Western blot analyses [116]. The rice mIT protein complemented the growth of yeast mutant which was defective in mitochondrial Fe transport [155]. Similarly, the expression of *At*mIT1 or *At*mIT2 can rescue the phenotype of the yeast mutant defective in mitochondrial Fe transport (mrs3mrs4 mutant; [157]). In mammalian and yeast cells, the redundancy in the roles of mITs has been investigated in terms of biochemical properties and kinetic profiles for Fe2<sup>+</sup> uptake [154,158]. Moreover, a recent study demonstrated that a purified recombinant mitoferrin−<sup>1</sup> (TMfrn1), from *Oreochromis niloticus*, catalyzes the transport free Fe and not a chelated Fe complex. In addition, it was shown that it is selective for alkali divalent ions [159]. In the same study, the results indicated that mITs are high-affinity or high-throughput Fe transporters [159]. Of note, mitochondria are known as organelles where there is utilization of other transition metals than Fe, such as manganese, copper, and zinc; however, despite the importance, the mechanisms by which these metal ions are transported through the IMM are not well understood. In addition, it should be mentioned that the possible substrates used by mITs in exchange for the imported Fe are still unknown. In plant systems the biochemical properties of mITs are much less studied than other organisms. Nevertheless, it has been demonstrated that both *At*mIT1 and *At*mIT2 transporters seem to be important for mitochondrial Fe uptake and also for the correct mitochondrial function, and consequently, they are necessary for the proper growth and development of the plant [116,155].

#### **5. Conclusions**

Research into the metabolic roles of plant MCFs has made impressive advances since the last comprehensive reviews were published some eight to nine years ago [12,135]. This was in part due to be expected, given the massive increase in the number of plant species sequenced in the interim as well as the mechanistic insights into MCF function that were facilitated by recent developments in structural biology. Although, as yet, such experiments have not been carried out for plant proteins, their very high homology to their mammalian counterparts renders the findings based on the human ATP/ADP carrier to likely be highly similar to its plant counterpart and indeed to many other plant MCFs. The genome sequencing has additionally expanded the repertoire of MCFs found in any single species thereby reflecting the challenge that remains in their characterization. That said, as we detail above, via use of heterologous expression, the biochemical characterization of a large number of MCF members has been carried out thereby providing the putative metabolic functions of a substantial

number of the family. It is important to state that, as we discuss in the accompanying article [28], experimental proof that these studies do indeed reflect the in vivo role of the proteins remains lacking in some instances. Moreover, a considerable number of MCF proteins remain to be characterized at the biochemical level and such experiments should be a priority for future research. Only once the biochemical potential of each member of the MCF, as well as information concerning their subcellular locations, is acquired alongside that of non-canonical mitochondrial transporters will we be able to accurately model plant mitochondrial function and, for that matter, truly appreciate the importance of this fascinating organelle.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2218-273X/10/7/1013/s1, Table S1: Ortologous genes of plant mitochondrial carrier family retrieved from Plaza server.

**Author Contributions:** J.H.F.C. performed the genomic analysis. A.N.-N. and A.R.F. wrote the manuscript. All authors agreed to the published version of the manuscript.

**Funding:** This research was funded by Collaborative Research Centers, SFB (Sonderforschungsbereich, Grant TRR 175/1) to A.R.F. and Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) (Grant 306818/2016-7) to A.N.-N.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Review* **Biogenesis of Mitochondrial Metabolite Carriers**

#### **Patrick Horten 1,2 , Lilia Colina-Tenorio 1,3 and Heike Rampelt 1,3, \***


#### Academic Editor: Ferdinando Palmieri Received: 10 June 2020; Accepted: 3 July 2020; Published: 7 July 2020

**Abstract:** Metabolite carriers of the mitochondrial inner membrane are crucial for cellular physiology since mitochondria contribute essential metabolic reactions and synthesize the majority of the cellular ATP. Like almost all mitochondrial proteins, carriers have to be imported into mitochondria from the cytosol. Carrier precursors utilize a specialized translocation pathway dedicated to the biogenesis of carriers and related proteins, the carrier translocase of the inner membrane (TIM22) pathway. After recognition and import through the mitochondrial outer membrane via the translocase of the outer membrane (TOM) complex, carrier precursors are ushered through the intermembrane space by hexameric TIM chaperones and ultimately integrated into the inner membrane by the TIM22 carrier translocase. Recent advances have shed light on the mechanisms of TOM translocase and TIM chaperone function, uncovered an unexpected versatility of the machineries, and revealed novel components and functional crosstalk of the human TIM22 translocase.

**Keywords:** mitochondrial carrier; metabolite transport; mitochondrial pyruvate carrier; sideroflexin; TOM; TIM chaperones; TIM22; protein translocation; mitochondrial biogenesis

#### **1. Introduction**

The mitochondrial inner membrane separates two aqueous compartments, the matrix and the intermembrane space, that differ in their protein and metabolite composition and host distinct metabolic pathways. The inner membrane is also the site of oxidative phosphorylation, and its integrity is crucial to maintain the electrochemical membrane potential that fuels ATP synthesis as well as mitochondrial biogenesis and function. Therefore, metabolite transport into or out of the matrix relies on carrier proteins that facilitate diffusion of specific substrates across the membrane or use the membrane potential to transport metabolites.

Most mitochondrial metabolite carriers belong to the mitochondrial carrier family (MCF, in humans SLC25 for solute carrier family 25). It comprises more than 50 members in humans and over 30 in yeast, and includes the most abundant inner membrane proteins [1–4]. MCF substrates range from nucleotides and amino acids to cofactors, intermediates of oxidative metabolism, and inorganic ions. Thus, they perform crucial functions in mitochondrial metabolism, and mutations in carrier genes are associated with a variety of human pathologies [5]. The mitochondrial pyruvate carrier (MPC) belongs to an unrelated protein family and functions as a hetero-dimer that requires both subunits for carrier activity [6,7]. The sideroflexin family constitutes a third metabolite carrier family, members of which have recently been discovered to function as serine transporters in one-carbon metabolism [8–10].

Like the vast majority of mitochondrial proteins, carriers are encoded in the nuclear genome and synthesized by cytosolic ribosomes. Therefore, they have to be specifically recognized and imported into the correct mitochondrial compartment. A multitude of protein translocases cooperates in the biogenesis of proteins destined for the different mitochondrial compartments [11–14]. Most precursors of mitochondrial inner membrane proteins are imported by the presequence translocase of the inner membrane (TIM23). However, carrier precursors generally lack a presequence, although a few contain an N-terminal extension that can improve solubility and translocation across the outer membrane [15–17]. Instead, carriers are targeted to mitochondria by internal signals. The internal targeting signals are not well defined, and carriers apparently contain several such motifs with different properties. For their biogenesis, metabolite carriers utilize a specialized import pathway involving the carrier translocase of the inner membrane (TIM22) [12,13,17–20] (Figure 1). This pathway can be divided into consecutive, biochemically defined stages: Stages I and II take place in the cytosol and on the mitochondrial surface, leading to import of the carrier precursor by the translocase of the outer membrane (TOM). In the intermembrane space (IMS), the precursor is bound by small TIM chaperones (stage III) and handed over to the TIM22 translocase (stage IV) which integrates the carrier into the inner membrane (stage V). The membrane potential across the inner membrane provides the driving force for membrane integration: Positively charged carrier sequences are subject to an electrophoretic force that pulls them into the matrix. Like the vast majority of mitochondrial proteins, carriers are encoded in the nuclear genome and synthesized by cytosolic ribosomes. Therefore, they have to be specifically recognized and imported into the correct mitochondrial compartment. A multitude of protein translocases cooperates in the biogenesis of proteins destined for the different mitochondrial compartments [11–14]. Most precursors of mitochondrial inner membrane proteins are imported by the presequence translocase of the inner membrane (TIM23). However, carrier precursors generally lack a presequence, although a few contain an N-terminal extension that can improve solubility and translocation across the outer membrane [15–17]. Instead, carriers are targeted to mitochondria by internal signals. The internal targeting signals are not well defined, and carriers apparently contain several such motifs with different properties. For their biogenesis, metabolite carriers utilize a specialized import pathway involving the carrier translocase of the inner membrane (TIM22) [12,13,17–20] (Figure 1). This pathway can be divided into consecutive, biochemically defined stages: Stages I and II take place in the cytosol and on the mitochondrial surface, leading to import of the carrier precursor by the translocase of the outer membrane (TOM). In the intermembrane space (IMS), the precursor is bound by small TIM chaperones (stage III) and handed over to the TIM22 translocase (stage IV) which integrates the carrier into the inner membrane (stage V). The membrane potential across the inner membrane provides the driving force for membrane integration: Positively charged carrier sequences are subject to an electrophoretic force that pulls them into the matrix.

Biomolecules 2020, 10, x 2 of 13

Figure 1. The carrier pathway in the yeast S. cerevisiae (A) and in humans (B) handles the recognition, translocation and membrane integration of mitochondrial metabolite carriers into the inner membrane. Carrier precursors are bound by chaperones in the cytosol and recognized at the translocase of the outer membrane (TOM) by the Tom70 receptor. After their transfer through the outer membrane, they are bound in the intermembrane space by the hexameric TIM chaperones, Tim9-Tim10 in yeast (A) or Tim9-Tim10a in humans (B). The TIM chaperones guide the precursor through the aqueous compartment to the membrane-bound TIM chaperone complex consisting of Tim9-Tim10-Tim12 in yeast (A) or Tim9-Tim10a-Tim10b in humans (B). Substrate transfer to the carrier translocase of the inner membrane (TIM22) is aided by interactions with outer membrane proteins (dashed arrows) involving the metabolite channel porin/VDAC in yeast (A), or the TOM complex in humans (B). In humans, VDAC was found in association with TIM22 components (B) and, thus, might participate in carrier biogenesis similarly to porin. The TIM22 carrier translocase integrates the precursors into the inner membrane in a membrane potential-dependent manner. OM, outer membrane; IMS, intermembrane space; IM, inner membrane; ∆ψ, membrane potential; Hsp70, Hsp90, cytosolic ATP-dependent chaperones; Tom40, pore-forming component of the TOM complex; Tom20, Tom22, Tom70, receptors of the TOM complex; porin/VDAC, voltage-dependent anion channel; Tim22, core component of the TIM22 translocase; Tim18, Sdh3 (succinate dehydrogenase 3), Tim54, auxiliary subunits of the yeast TIM22 translocase; Tim29, AGK (acylglycerol kinase), auxiliary subunits of the human TIM22 translocase. **Figure 1.** The carrier pathway in the yeast *S. cerevisiae* (**A**) and in humans (**B**) handles the recognition, translocation and membrane integration of mitochondrial metabolite carriers into the inner membrane. Carrier precursors are bound by chaperones in the cytosol and recognized at the translocase of the outer membrane (TOM) by the Tom70 receptor. After their transfer through the outer membrane, they are bound in the intermembrane space by the hexameric TIM chaperones, Tim9-Tim10 in yeast (**A**) or Tim9-Tim10a in humans (**B**). The TIM chaperones guide the precursor through the aqueous compartment to the membrane-bound TIM chaperone complex consisting of Tim9-Tim10-Tim12 in yeast (**A**) or Tim9-Tim10a-Tim10b in humans (**B**). Substrate transfer to the carrier translocase of the inner membrane (TIM22) is aided by interactions with outer membrane proteins (dashed arrows) involving the metabolite channel porin/VDAC in yeast (**A**), or the TOM complex in humans (**B**). In humans, VDAC was found in association with TIM22 components (**B**) and, thus, might participate in carrier biogenesis similarly to porin. The TIM22 carrier translocase integrates the precursors into the inner membrane in a membrane potential-dependent manner. OM, outer membrane; IMS, intermembrane space; IM, inner membrane; ∆ψ, membrane potential; Hsp70, Hsp90, cytosolic ATP-dependent chaperones; Tom40, pore-forming component of the TOM complex; Tom20, Tom22, Tom70, receptors of the TOM complex; porin/VDAC, voltage-dependent anion channel; Tim22, core component of the TIM22 translocase; Tim18, Sdh3 (succinate dehydrogenase 3), Tim54, auxiliary subunits of the yeast TIM22 translocase; Tim29, AGK (acylglycerol kinase), auxiliary subunits of the human TIM22 translocase.

Mitochondrial carriers of the MCF/SLC25 family are the eponymous substrates of the TIM22 carrier translocase pathway [12,13,17,19] (Figure 1). The best studied carrier is the ADP/ATP carrier AAC (yeast)/ANT (human; adenine nucleotide translocator) that has been employed both as a model substrate to study carrier biogenesis and for the analysis of MCF structure and transport mechanism [4,17,18,21–23]. Mitochondrial carriers of the MCF/SLC25 family have a tripartite organization, with three homologous repeats consisting of two transmembrane segments each. They uniformly possess six transmembrane segments and expose both their N- and C-termini to the intermembrane space (Figure 2) [4,12,17,19,24]. Mitochondrial carriers transport substrates by enabling alternate access of the substrate(s) to the matrix and the intermembrane space while maintaining membrane impermeability for non-substrates [4,25–28]. Divergent members of the MCF with a differing number of TM segments are localized in the outer membrane and have acquired functions distinct from metabolite transport [4,29,30]. substrate to study carrier biogenesis and for the analysis of MCF structure and transport mechanism [4,17,18,21–23]. Mitochondrial carriers of the MCF/SLC25 family have a tripartite organization, with three homologous repeats consisting of two transmembrane segments each. They uniformly possess six transmembrane segments and expose both their N- and C-termini to the intermembrane space (Figure 2) [4,12,17,19,24]. Mitochondrial carriers transport substrates by enabling alternate access of the substrate(s) to the matrix and the intermembrane space while maintaining membrane impermeability for non-substrates [4,25–28]. Divergent members of the MCF with a differing number of TM segments are localized in the outer membrane and have acquired functions distinct from metabolite transport [4,29,30]. Interestingly, the components of the heterodimeric mitochondrial pyruvate carrier (MPC) have recently been discovered as further substrates of the TIM22 pathway (Figure 1) [31,32]. In contrast to the classical mitochondrial carriers, they are related to sugar transporters of the eukaryotic sugars will eventually be exported transporter (SWEET) and prokaryotic semiSWEET families [33,34].

Biomolecules 2020, 10, x 3 of 13

AAC (yeast)/ANT (human; adenine nucleotide translocator) that has been employed both as a model

Interestingly, the components of the heterodimeric mitochondrial pyruvate carrier (MPC) have recently been discovered as further substrates of the TIM22 pathway (Figure 1) [31,32]. In contrast to the classical mitochondrial carriers, they are related to sugar transporters of the eukaryotic sugars will eventually be exported transporter (SWEET) and prokaryotic semiSWEET families [33,34]. SWEET transporters possess seven TM segments that are arranged into two triple-helix bundles connected by another α-helix. SemiSWEETs instead consist of one triple-helix bundle and assemble to dimers, forming a six TM functional unit like the SWEETs. MPC subunits MPC2 (mammals) as well as Mpc2 and Mpc3 (yeast) have three TM segments (Figure 2). For MPC1/Mpc1, the topology is not entirely clear: It has been suggested that they have only two TM segments with both termini in the matrix, or alternatively that they share the same topology as Mpc2/Mpc3 [6,7,35–37] (Figure 2). SWEET transporters possess seven TM segments that are arranged into two triple-helix bundles connected by another α-helix. SemiSWEETs instead consist of one triple-helix bundle and assemble to dimers, forming a six TM functional unit like the SWEETs. MPC subunits MPC2 (mammals) as well as Mpc2 and Mpc3 (yeast) have three TM segments (Figure 2). For MPC1/Mpc1, the topology is not entirely clear: It has been suggested that they have only two TM segments with both termini in the matrix, or alternatively that they share the same topology as Mpc2/Mpc3 [6,7,35–37] (Figure 2). Additionally, recent studies indicate that the sideroflexins, with five transmembrane segments and the N-terminus in the intermembrane space (IMS), also depend on the TIM22 carrier pathway for their biogenesis [38,39] (Figure 2). Thus, the mitochondrial pyruvate carrier components and the sideroflexins with their unique topologies have challenged long-held views of the structural requirements for TIM22 substrates.

Figure 2. Substrates of the TIM22 carrier import pathway. Mitochondrial carriers of the mitochondrial carrier family (MCF)/SLC25 family (red), the components of the mitochondrial pyruvate carrier (brown, orange), as well as sideroflexins (blue) are imported into mitochondria via the TIM22 carrier pathway. MCF/SLC25 proteins have a uniform topology with 6 transmembrane segments [4,12,17,19,24]. In contrast, Mpc2/Mpc3 has only 3 TM segments, and Mpc1 has 2 or 3 TM segments [6,7,35–37]. The third unique family of metabolite carriers, the sideroflexins, has 5 TM domains with the N-terminus in the intermembrane space (IMS) [8,38,39]. Aside from metabolite carriers, the TIM22 pathway also imports the members of the Tim17 protein family including the translocase components Tim17, Tim22 and Tim23 (green). IMS, intermembrane space; IM, inner membrane. **Figure 2.** Substrates of the TIM22 carrier import pathway. Mitochondrial carriers of the mitochondrial carrier family (MCF)/SLC25 family (**red**), the components of the mitochondrial pyruvate carrier (**brown**, **orange**), as well as sideroflexins (**blue**) are imported into mitochondria via the TIM22 carrier pathway. MCF/SLC25 proteins have a uniform topology with 6 transmembrane segments [4,12,17,19,24]. In contrast, Mpc2/Mpc3 has only 3 TM segments, and Mpc1 has 2 or 3 TM segments [6,7,35–37]. The third unique family of metabolite carriers, the sideroflexins, has 5 TM domains with the N-terminus in the intermembrane space (IMS) [8,38,39]. Aside from metabolite carriers, the TIM22 pathway also imports the members of the Tim17 protein family including the translocase components Tim17, Tim22 and Tim23 (**green**). IMS, intermembrane space; IM, inner membrane.

Additionally, recent studies indicate that the sideroflexins, with five transmembrane segments and the N-terminus in the intermembrane space (IMS), also depend on the TIM22 carrier pathway for their biogenesis [38,39] (Figure 2). Thus, the mitochondrial pyruvate carrier components

and the sideroflexins with their unique topologies have challenged long-held views of the structural requirements for TIM22 substrates.

#### **2. Carrier Recognition at the TOM Complex**

Due to their hydrophobic nature, mitochondrial carrier precursors in the cytosol are bound by chaperones to prevent their aggregation (stage I). Since carrier import takes place post-translationally, the soluble stage can be distinguished from stage II that consists in precursor targeting to the translocase of the outer membrane (Figure 1, Table 1) [13,14,17,19]. The TOM complex is the main entry gate by which almost all precursor proteins destined to the different mitochondrial compartments gain access to the organelle [12,14,40,41]. It forms dimers *in vivo* and consists of the Tom40 β-barrel pore and six α-helical membrane proteins: Tom5, Tom6, Tom7 as well as the receptors Tom22, Tom20 and Tom70 [42–44]. Aggregation of the highly hydrophobic carrier precursors in the cytosol is prevented by molecular chaperones. In yeast, carriers are chaperoned mainly by Hsp70, whereas in mammalian cells both Hsp70 and Hsp90 participate in carrier biogenesis, along with several co-chaperones [14,45–49]. Recognition of precursors at the TOM complex is mediated by the receptors Tom20 and Tom70. They can functionally substitute for each other sufficiently well for single deletions to be viable, however Tom20 preferentially recognizes precursors that contain a presequence, while Tom70 preferentially binds precursors with internal targeting sequences including carrier proteins such as the ADP/ATP carrier or the phosphate carrier [12,13,49–60]. Tom70 not only interacts with the precursor, but also with the associated chaperone(s) via tetratricopeptide repeats (TPR) that bind the C-termini of Hsp70 or Hsp90 chaperones [14,46,49,61]. Moreover, one tripartite carrier precursor of the MCF/SLC25 family can recruit three Tom70 dimers, with each of the repeats participating in the interaction [21]. Thus, the interactions of Tom70 with the precursor-chaperone complex likely contribute to prevention of its aggregation. Import of carriers can be stalled at stage II by depletion of ATP (Table 1). ATP binding to Hsp70 triggers substrate release from Hsp70, the carrier precursor is handed over to the central receptor Tom22, and individual helix-loop-helix modules are threaded into the Tom40 pore in a hairpin-like conformation [21,62]. It is currently unclear how mitochondrial pyruvate carrier precursors with their distinct topology are handled by the TOM complex, although it is tempting to speculate that at least the two C-terminal TM segments of Mpc2/Mpc3 may be recognized in a fashion similar to classical carriers. Interestingly, even during translocation through the TOM complex, carriers follow a different route than presequence precursors, involving the distal regions of the Tom40 dimer and the N-terminal extension of Tom40 [42,44]. It was proposed that translocation through TOM is aided by interaction of positively charged regions in the precursors with the negatively charged inner surface of the Tom40 β-barrel [63], which is consistent with the previously reported head-first insertion of helix-loop-helix modules by their positively charged loops [21].

The efficiency of carrier recognition at the TOM complex is subject to metabolic regulation. Tom70 is phosphorylated by protein kinase A specifically during non-respiratory growth of yeast on glucose, resulting in an impaired interaction with Hsp70 and concomitantly reduced carrier import [64]. Thus, the efficiency of carrier biogenesis can be adjusted to the metabolic requirements of respiratory versus non-respiratory growth.


**Table 1.** Stages of carrier biogenesis via the TIM22 carrier import pathway.

#### **3. En Route through the Intermembrane Space**

Unlike the TIM23 translocase that imports presequence proteins into the matrix or the inner membrane, the translocation of carrier precursors through the TOM complex is apparently not tightly coupled to integration into the inner membrane by the TIM22 carrier translocase [12,17,65,66]. Instead, once a carrier precursor has traversed the Tom40 pore and reached the IMS, it is bound by small TIM chaperones to prevent aggregation during its transit through the aqueous IMS environment to the inner membrane (stage III) [67–73] (Figure 1, Table 1). Translocation through the TOM complex is coupled to TIM chaperone binding [71]. The small TIM chaperones form ring-like hetero-hexameric complexes that bind carriers in an extended conformation [73–75]. The predominant TIM chaperone complex consists of alternating Tim9 and Tim10 subunits (Tim9 and Tim10a in humans) and is required for carrier import; the alternative Tim8-Tim13 complex (Tim8a or Tim8b and Tim13 in humans) has partially redundant substrate specificity [73,75–79]. Mutations in Tim8a cause the deafness–dystonia syndrome called Mohr–Tranebjærg syndrome [80], however, novel evidence suggests that the underlying molecular mechanism reflects a new function of Tim8a in cytochrome c oxidase maturation rather than a defective TIM22 carrier pathway [79]. The small TIM chaperones interact with the N-terminal extension of Tom40 that participates in carrier translocation, so they are ideally positioned to receive their cargo from the TOM complex [21,42,44,71]. Stage III of carrier import can be further subdivided, where stage IIIa denotes carriers bound to soluble TIM chaperones that may be associated with TOM or soluble in the IMS (Table 1) [17,19]. A recent comprehensive study demonstrated that an MCF carrier with six TM segments is bound by two TIM chaperone hexamers, and the precursor is chaperoned by interacting with a conserved hydrophobic cleft between the two tentacle-like α-helices of the small TIM proteins [73]. This conserved substrate binding region is also required for chaperoning of the structurally unrelated mitochondrial pyruvate carrier [31]. The soluble TIM complexes transfer carrier precursors to a separate TIM chaperone complex that is associated with the TIM22 carrier translocase of the inner membrane. This membrane-bound TIM hexamer consists of Tim9/Tim10/Tim12 (yeast) or Tim9/Tim10a/Tim10b (human) [67,81–86] (Figure 1). Carrier precursors bound to the membrane-associated TIM chaperones represent stage IIIb of the carrier import pathway where the precursor is tethered to the inner membrane, but not yet inserted (Table 1) [17,19]. Since this is

the last step that is independent of the membrane potential ∆ψ, precursors accumulate in stage III upon dissipation of the membrane potential. The dependence of carrier import on TIM chaperones also allows to experimentally distinguish this import pathway into the inner membrane from the presequence pathway where import into a protease-protected environment depends on ∆ψ due to the coupling of TOM and TIM23 [65,66].

Unexpectedly, transport of carriers to the TIM22 carrier translocase is also aided by porin/voltage-dependent anion channel (VDAC), the major metabolite channel of the outer membrane (Figure 1) [87–90]. In addition to interacting with TIM chaperones as well as carrier precursors, porin recruits the TIM22 translocase and thereby brings the outer and inner membrane in close proximity, which may enhance carrier biogenesis [88,89]. Since porin, unlike the protein translocases, is significantly upregulated upon respiratory growth [91], this novel physiological role may support the higher levels of protein import required for metabolic remodeling of the mitochondria. While a contribution of VDACs to carrier biogenesis in human mitochondria has not been studied directly, they were found to interact with TIM22 [92,93].

#### **4. Membrane Integration by the TIM22 Carrier Translocase**

The TIM22 translocase consists of the Tim22 protein, several auxiliary subunits that have roles in assembly and stabilization of the complex, and TIM chaperones. In yeast, the additional subunits comprise Tim54, Sdh3, which also interacts with Sdh4 as part of the succinate dehydrogenase complex, and Tim18, a homolog of Sdh4 [12,17,19,20,94–100] (Figure 1A). Until recently, the only known membrane integral TIM22 component in humans was Tim22 itself, however, several studies have discovered two new subunits (Figure 1B). The metazoan-specific subunit Tim29, which like Tim54 is exposed to the IMS and can be crosslinked to TIM chaperones, is required for TIM22 assembly and efficient import of some substrates [92,101]. Moreover, Tim29 interacts with Tom40, indicating that there may be coupling between TOM and TIM22 in human cells [92], in contrast to yeast. The most recently identified subunit of human TIM22 is acylglycerol kinase (AGK) that phosphorylates glycerides to generate lysophosphatidic acid or phosphatidic acid [102–104]. AGK has a dual role in human mitochondria: It is required for TIM22 stability and carrier import independently of its lipid kinase activity, and loss of AGK results in TCA (tricarboxylic acid) cycle defects; however, kinase deficiency causes aberrant mitochondrial ultrastructure and concomitantly reduced respiration [103,104]. Mutations in AGK cause the mitochondrial disease Sengers syndrome [105,106], and its novel role as part of the TIM22 translocase appears to account for the disease phenotype [103]. Aside from AGK, mutations in Tim22 that impair carrier import were also recently reported to result in human pathology with neuromuscular defects [32,107]. Interestingly, human TIM22 interacts with the mitochondrial contact site and cristae organizing system (MICOS) [93], an inner membrane protein complex that is crucial for native cristae architecture and that forms contact sites between the two mitochondrial membranes [108–110]. Upon MICOS disruption, carrier import is specifically impaired, indicating that MICOS-mediated membrane contact sites might support efficient carrier biogenesis in human cells [93].

The TIM22 complex is a voltage-gated preprotein translocase that is thought to insert one helix—matrix loop—helix repeat at a time in a hairpin conformation into the inner membrane [19,20,86,97]. At least a low membrane potential is required for transfer of a precursor from the TIM chaperones to TIM22. Precursors can be trapped experimentally at this stage IV by reducing the membrane potential with ionophores (Table 1) [20]. In the presence of a higher membrane potential as well as of an internal targeting sequence, TIM22 integrates the protein into the inner membrane by an unknown mechanism involving lateral release, and the carrier reaches the mature stage V [20].

The modular topology of mitochondrial carriers of the MCF/SLC25 family—with repeats of helix-loop-helix domains, the termini facing the IMS and positively charged loops in the matrix—was long assumed to be a requirement for substrates of the TIM22 carrier pathway. This pathway also imports members of the Tim17 protein family that includes Tim22 itself as well as the TIM23

translocase components Tim23 and Tim17, all of which have four TM segments but otherwise share the topology of classical carriers [73,75–77,111–113] (Figure 2). As mentioned, both TOM and TIM22 translocases act on paired helices during import of classical carriers, including the dicarboxylate carrier, the ADP/ATP carrier and the phosphate carrier, as well as during import of Tim22-related proteins [21,70,77,114–116], in contrast to the linear import of other mitochondrial proteins. Until recently, multi-spanning inner membrane proteins apart from the MCF and Tim17 families were thought to be imported into mitochondria by the other translocases of the inner membrane: By the TIM23 translocase [11–13,65,66,117], by the oxidase assembly (OXA) translocase that is responsible for the membrane integration of mitochondrially encoded proteins [118,119] or by a combination of both machineries [120–122]. Moreover, truncated variants of carrier proteins are no longer recognized as substrates of the TIM22 pathway and instead are imported by TIM23 or remain in the intermembrane space [115,116]. However, recent work has identified the mitochondrial pyruvate carrier proteins with their divergent topology as substrates of the TIM22 pathway [31,32]. What is more, the unrelated sideroflexins also rely on the TIM22 carrier translocase for their biogenesis [38,39]. Of the MPC components, at least Mpc2 and Mpc3 have unpaired TM helices and none of the MPC proteins possess more than three TM segments [6,7,35–37], while the sideroflexins also have an uneven number of TM segments and yet another topology [8,39] (Figure 2). Thus, multiple recent studies have revealed a surprising versatility of the TIM22 pathway. The mechanistic differences in the handling of substrates with 4 or 6 TM segments (classical TIM22 substrates) versus 2/3 (MPC subunits) or 5 TM segments (sideroflexins) are still unclear. The C-terminal helix-loop-helix domain of Mpc2 and Mpc3 might conceivably be treated similarly as a typical carrier repeat, since they share the topology and positively charged matrix loop. In addition, MPC subunits were reported to have an N-terminal α-helix whose function is unknown [37]. For the human TIM22 translocase a differential requirement of the auxiliary subunits for different substrate classes has been reported: AGK is required for efficient carrier import and dispensable for the import of Tim22 and related proteins [103,104], whereas the opposite is the case for Tim29 [92]. Interestingly, sideroflexins, like classical carriers, rely on AGK, while Tim29 is dispensable [38]. It will be very interesting to learn how TIM22 adapts its function to import this range of structurally distinct substrates.

#### **5. Perspectives**

The biogenesis of mitochondrial metabolite carriers is still not fully understood despite the fact that the TIM22 import pathway has been under scientific investigation for decades. The recent discovery of two human TIM22 components and the very limited insight into the mechanism of membrane integration by the carrier translocase exemplify how much fundamental information is still lacking. Novel findings indicate that carrier import may benefit from contact sites between the outer and inner membranes after all. Moreover, the TIM22 pathway has turned out to be unexpectedly versatile regarding its substrate requirements. Finally, it seems likely that the biogenesis of proteins with such a central role in mitochondrial physiology is regulated at different steps. While there is precedent for this notion from yeast, the human TIM22 pathway still awaits characterization of regulatory factors. Thus, the biogenesis of mitochondrial metabolite carriers remains an exciting field of study that is expected to generate important insights into mitochondrial physiology.

**Funding:** This work was supported by the Excellence Initiative/Strategy of the German Federal and State Governments (EXC 2189 CIBSS Project ID 390939984).

**Acknowledgments:** We thank Nils Wiedemann and Nikolaus Pfanner for discussion and helpful suggestions. Work included in this study has been performed in partial fulfillment of the requirements for the doctoral thesis of P.H.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Review*
