Mitochondrial Ca²⁺ transport in the endothelium: regulation by ions, redox signalling and mechanical forces

Abstract

Calcium (Ca²⁺) transport by mitochondria is an important component of the cell Ca²⁺ homeostasis machinery in metazoans. Ca²⁺ uptake by mitochondria is a major determinant of bioenergetics and cell fate. Mitochondrial Ca²⁺ uptake occurs via the mitochondrial Ca²⁺ uniporter (MCU) complex, an inner mitochondrial membrane protein assembly consisting of the MCU Ca²⁺ channel, as its core component, and the MCU complex regulatory/auxiliary proteins. In this review, we summarize the current knowledge on the molecular nature of the MCU complex and its regulation by intra- and extramitochondrial levels of divalent ions and reactive oxygen species (ROS). Intracellular Ca²⁺ concentration ([Ca²⁺]_i), mitochondrial Ca²⁺ concentration ([Ca²⁺]_m) and mitochondrial ROS (mROS) are intricately coupled in regulating MCU activity. Here, we highlight the contribution of MCU activity to vascular endothelial cell (EC) function. Besides the ionic and oxidant regulation, ECs are continuously exposed to haemodynamic forces (either pulsatile or oscillatory fluid mechanical shear stresses, depending on the precise EC location within the arteries). Thus, we also propose an EC mechanotransduction-mediated regulation of MCU activity in the context of vascular physiology and atherosclerotic vascular disease.

1. Introduction

The endothelium is a cellophane-like membrane lining the circulatory system and its major functions are regulation of vascular tone and vessel wall permeability. Endothelial cell (EC) dysfunction has been implicated in the early stages of many cardiovascular diseases, which renders the modulation of EC functions a key therapeutic target [1–3]. Under normal conditions, EC functions depend on changes in the intracellular calcium concentration ([Ca²⁺]_i). The Ca²⁺ concentration in extracellular biological fluids is in the range of 1.6–2 mM, whereas, within cells, Ca²⁺ is bound to phospholipids, proteins and nucleic acids or is sequestered in organelles, and thus only 0.1% of the total Ca²⁺ content is free in the cytosol. Consequently, [Ca²⁺]_i is kept at approximately 100 nM to regulate cellular processes that function over a wide dynamic range. In most cells, increases in [Ca²⁺]_i are transient and oscillatory [4] and the frequency of oscillation (temporal [Ca²⁺]_i signature) is differentially decoded as cellular signal [5]. At any time, the spatio-temporal complexity in the regulation of [Ca²⁺]_i is determined by a balance between the interplay of multiple counteracting processes divided into ‘on’ and ‘off’ mechanisms. The ‘on’ reaction results in generation of the [Ca²⁺]_i rise due to Ca²⁺ influx from the extracellular medium and Ca²⁺ release from intracellular stores, such as the endoplasmic reticulum (ER) in non-excitable cells. The ‘off’ reaction removes intracellular Ca²⁺ by the combined action of pumps, buffers and exchangers. One important intracellular organelle associated with Ca²⁺ handling is the mitochondrion [6,7]. Although other compartments such as the ER, lysosomes, Golgi apparatus and endosomes are recognized as Ca²⁺ stores [7–9], the mitochondria remain highly specialized in this context because of their sponge-like retentive capacity with regard to Ca²⁺.

In ECs, the mitochondria account for 25% of the cell's Ca²⁺ storage. It has long been ascertained that mitochondrial content in ECs is modest (endothelial mitochondria occupy 2–6% of the cytoplasmic volume) compared with cardiac myocytes (32% of the cytoplasmic volume) and other cell types, as the energy requirements in ECs are relatively low and glycolysis is the major source of ATP production. In vitro and in vivo studies have shown that, apart from the well-known role of mitochondria in bioenergetics, they also play a prominent role in signalling cellular responses to environmental cues [10–12]. Two such important modes of signalling are their regulated production of mitochondrial reactive oxygen species (mROS) and their service as Ca²⁺ sinks. Mitochondrial Ca²⁺ uptake and mROS generation are interdependent phenomena and contribute to cell function in a ‘mutual crosstalk’, with mitochondrial Ca²⁺ concentration ([Ca²⁺]_m) representing the key to deciphering mROS signals [13].

Mitochondria have a large negative membrane potential of 180 mV that facilitates flooding of mitochondria by Ca²⁺ from the cytosolic milieu. During evolution, cells acquired sophisticated mitochondrial Ca²⁺ transport machineries that control Ca²⁺ entry to the mitochondrial matrix and Ca²⁺ exit/redistribution for proper cell function. The mitochondrial Ca²⁺ uptake system is a heterogeneous protein complex with the molecular identity mixed and matched with the mitochondrial Ca²⁺ uniporter (MCU) as the core component [14,15]. The MCU is regulated by an array of proteins including mitochondrial Ca²⁺ uptake 1 (MICU1), 2 (MICU2), 3 (MICU3), essential MCU regulator (EMRE), MCU regulating protein 1 (MCUR1) and MCU dominant negative beta subunit (MCUb) (see §2 for a complete description) [16–24]. Besides protein regulation (positive by MCUR1 and EMRE [19,23] and negative by MICU1 [22]), recent studies from our group and others demonstrated a regulation of mitochondrial Ca²⁺ uptake by ions and redox signalling. In this review, we summarize these regulatory mechanisms of MCU activity and suggest a potential therapeutic utility in controlling endothelial MCU activity during vascular (patho)physiology. We also introduce a novel hypothesis that spatial regulation of MCU activity by local haemodynamic forces (via their effect on ion and redox signalling) in human arteries may, at least in part, explain the focal nature of atherosclerotic vascular disease. Since atherosclerotic vascular disease continues to be the major cause of death in developed nations [25] and EC dysfunction plays a causative role in disease initiation [26,27], this review aims to position the MCU as a potential critical target for intervention against atherosclerosis.

2. Ionic regulation of mitochondrial Ca²⁺ uptake

The MCU contains two transmembrane (TM) domains with coiled-coil regions positioned before and after these domains. Topologically, both the N- and C-terminal regions reside in the matrix constituting a majority of the protein [28]. The regulation of the MCU pore is complex as evidenced by the numerous protein regulators which have been identified. MCUb is a homologue of MCU sharing 50% sequence similarity, but no ability to constitute a Ca²⁺-permeable channel despite the presence of similar TM domains to those found in the MCU. These MCUb TM domains contain two noteworthy amino acid substitutions (R251Q and D256V) compared with the MCU, which may be related to the dominant negative inhibition of MCU activity after interactions with MCUb [20]. MICU1/2/3 proteins play gating roles promoting a closed MCU at low [Ca²⁺]_i [19,21,23,24,29–33]. EMRE is a 10 kDa, single TM protein that stabilizes and/or bridges MCU and MICU1 interactions [19]. Solute carrier family 25 member 23 (SLC25A23) is found on the inner mitochondrial membrane (IMM) and functions as an ATP and phosphate transporter protein, interacting with both MCU and MICU1 and enhancing mitochondrial Ca²⁺ uptake [34]. MCUR1 is a 40 kDa protein containing two putative TM domains; moreover, it has recently been determined that MCUR1 is a vital scaffold factor required for assembly of the heteromeric MCU complex through interactions with both EMRE and the MCU [16,23].

The regulation of the MCU via its numerous protein binding partners has been delineated recently; however, the ability of [Ca²⁺]_i to regulate MCU channel activity was reported approximately 25 years prior to the identification of any of the genes/proteins involved in this process [35–38]. Furthermore, it is now clear that some of the protein regulators exert their modulatory function via interactions with Ca²⁺ ions. MICU1 is a 54 kDa protein that has two Ca²⁺ binding EF-hand domains. Vertebrates express three homologues, but only the functions of MICU1 and MICU2 appear non-redundant with respect to MCU regulation, while MICU3 is predominantly localized to the central nervous system [39]. Studies are in agreement about the gatekeeping function of MICUs at low [Ca²⁺]_i, but localization within the mitochondria, sites of interaction with the MCU and function at higher [Ca²⁺]_i have been debated [19,21,23,29–32]. Knockdown of MICU1 causes augmented [Ca²⁺]_m at basal [Ca²⁺]_i [23,30] due to destabilization of the MICU1 homologue, MICU2, which inhibits MCU channel activity [18]; moreover, MICU1 overexpression in MICU2-depleted cells also increases [Ca²⁺]_m, suggesting that MICU1 is a stimulator of MCU uptake [18]. MICU1 has been reported to be mitochondrial matrix localized, directly interacting with the MCU [21], or intermembrane space localized, indirectly interacting with the MCU via an accessory protein [30,40]. Knockout of MICU1 results in unregulated MCU activity, while knockout of MICU2 only lowers the Ca²⁺ level threshold at which MICU1 gates the MCU; moreover, the MICU1 or MICU2 Ca²⁺ binding-incompetent mutants constitutively inactivate the MCU irrespective of Ca²⁺ levels [31]. Recently, it has been demonstrated that MICU1–MICU2 heterodimer formation establishes a steeply cooperative sub-micromolar Ca²⁺ binding affinity; moreover, the sharp Ca²⁺-binding cooperativity conveys a binary-like ON/OFF MCU activation phenotype over relatively small changes in [Ca²⁺]_i [36].

A crystal structure of MICU1 revealed that two EF-hand pairs (i.e. EF1 and EF2) are each preceded by intervening domains (i.e. ID1 and ID2) (figure 1a). Structural homology analysis using DALI [41,42] shows that no other currently known structure arranges EF-hand domains with similar ID1/2 structures. While EF1/2 show significant homology to calmodulin (CaM)-like proteins, ID1/2 appear relatively unique. Each EF-hand pair coordinates a single Ca²⁺ ion resulting in increased interhelical angles [43], yet the clefts do not show any evidence for intramolecular or homotypic intermolecular protein interactions in the available structures, contrary to other EF-hand proteins [44–48]. Furthermore, structural alignment of ID1/2 domains when EF1/2 are in the Ca²⁺-loaded and -depleted states show only minor structural differences (figure 1b,c). Thus, the structural mechanisms underlying MICU1 modulation of MCU activity and how Ca²⁺ may affect this regulation remain unclear.

Figure 1.

Structural mechanisms of MCU regulation. (a) Crystal structure of human MICU1 residues 108–442 in the Ca²⁺-bound state. The conformation and orientation of the two EF-hand domains (yellow ribbon diagrams; EF1 (residues 219–315); EF2 (residues 374–442)) are shown relative to the two preceding intervening domains (ID1, green ribbon diagram (residues 108–218); ID2, blue ribbon diagram (residues 316–373)). Each EF-hand domain coordinates a single Ca²⁺ atom (orange spheres). (b) All-atom structural alignment of the MICU1 ID1 domain in the presence (green ribbon) and absence (grey ribbon) of Ca²⁺. (c) All-atom structural alignment of the MICU1 ID2 domain in the presence (blue ribbon) and absence (grey ribbon) of Ca²⁺. ID1 and ID2 adopt similar structures with the EF-hand domains in either an open or a closed conformation. (d) NMR- and transmission electron microscopy-driven model of the Caenorhabditis elegans C-terminal domain residues 166–318. The inner pore is made up of TM2 (magenta ribbons; residues 244–260) and is surrounded by TM1 (light blue ribbons; residues 215–234). The DXXE motif (dashed black circle) links TM1 and TM2. The Asp240 (red space fill) and Glu243 (orange space fill) residues which bind Mn²⁺ cooperatively are indicated within the DXXE motif. The outer and inner coiled-coil bundles (CC1, brown ribbons (residues 180–193); CC2, yellow ribbons (293–316)) are labelled. The location of the matrix, IMM and IMS are shown relative to the TM domains. (e) Crystal structure of the human MCU N-terminal domain residues 72–189. The two β-sheets (light blue ribbons; residues 76–80, 83–88, 97–100 for β-sheet 1 and 125–128, 149–153, 156–160 for β-sheet 2) grasping the two centrally located helices (pink ribbons; residues 108–118 and 141–146 for α1 and α2, respectively) are labelled. The dashed black box highlights the location of the loop 2. (f) Zoomed view of the Cys97 and Ser92 residues within loop 2 which are susceptible to post-translational modifications. The close spatial proximity of the Arg93 and Glu95 side chains which form a salt bridge in some of the crystallized forms and stabilize the loop is shown. (g) First shell ligand coordination of the Mg²⁺ atom (yellow sphere) by the Asp147 side chain and five H₂O molecules (blue spheres) in an octahedral geometry. The close spatial proximity of the other Asp residues making MRAP are also indicated (red sticks). (h) Electrostatic surface potential of the human MCU N-terminal domain residues 72–189. The large MRAP region is indicated by a dashed black circle, and the individual acidic residues contributing to the region are labelled relative to the Mg²⁺ atom (yellow sphere). The −2 (red) to +2 (blue) kT/e gradient is shown below the protein. In (a–h) amino and carboxy termini are labelled N and C, respectively. The structures were rendered in PyMOL using the 4NSD (Ca²⁺-loaded) and 4NSC (Ca²⁺-depleted) coordinates for MICU1, 5ID3 for the C. elegans MCU C-terminal domain and 5KUJ for the human MCU N-terminal domain.

Many Ca²⁺ channel proteins exhibit Ca²⁺-dependent feedback mechanisms including the inositol 1,4,5-trisphosphate receptor (IP₃R) [49,50], ryanodine receptor [51] and Ca²⁺ release-activated Ca²⁺ channel [52,53]. In the seminal work by Kirichok et al. [54], which first characterized the electrophysiological properties of the MCU channel using mitoplast patch clamp recordings, no Ca²⁺-dependent inactivation was evident between 0.01 and 10 µM intramitoplast Ca²⁺ concentration at constant energy potentials of −160 mV. However, the same study demonstrated greater than 50% reduction in the Ca²⁺ conductance of the MCU in a 0.2 mM Ca²⁺/0.5 mM Mg²⁺ extramitochondrial divalent cation mixture [54]. Additionally, numerous studies have described different aspects of Ca²⁺-dependent MCU inactivation over the past decades, even prior to the molecular identification of the MCU complex components (e.g. [38,55–58]). More recently, however, a report has suggested that the C-terminal domain of EMRE is responsible for a matrix Ca²⁺-dependent inactivation phenomenon at low [Ca²⁺]_i. Using mitoplast patch clamp experiments, raising matrix Ca²⁺ concentrations to between approximately 0.03 and 0.4 µM markedly inhibited inward rectifying uniporter currents [59]. Consistent with this notion, mutations or truncations within the EMRE C-terminal domain abolished this Ca²⁺-dependent inactivation phenomenon [59]. However, given that the topology of EMRE remains controversial with several papers suggesting that the C-terminal domain lies in the inner membrane space (not the matrix) [16,60,61], the precise molecular mechanisms behind Ca²⁺-dependent MCU inactivation mediated by low matrix Ca²⁺ concentration effects on EMRE remain to be clarified.

Although a high-resolution structure of the pore-forming and coiled-coil regions of human MCU is currently unavailable, a nuclear magnetic resonance (NMR) spectroscopy- and transmission electron microscopy-driven pentameric model of Caenorhabditis elegans MCU with a deleted N-terminal domain has been constructed [62]. This structure shows that the second TM domain is positioned to line the pore of the channel with a short DXXE motif contributed by each subunit lining the entrance to the pore (figure 1d). The DXXE motif links the first and second TM domains. While the Glu is positioned deeper in the pore, the Asp is more solvent exposed, and mutating either of these acidic residues abrogates MCU channel activity [14]. NMR titrations showed that this DXXE motif can bind Mn²⁺, a surrogate for Ca²⁺-binding in NMR titrations, at both the Asp and Glu residues in a cooperative manner [63]. This strongly positive binding cooperativity for Ca²⁺ (i.e. defined by surrogate Mn²⁺ binding experiments) may underlie the ion selectivity mechanism of the MCU. This same site is involved in binding ruthenium red MCU channel inhibitors at the more exposed Asp residues [63], thus blocking the pore entrance.

While the C-terminal region alone can assemble and form an active channel [62], the significance of the N-terminal domain for the regulation of the channel is evident from work showing that self-association of the N-terminal half of the MCU promotes channel assembly and activation [64], oxidative modification of C97 within this N-terminal domain alters the channel architecture and function (see §3 for a complete description) [65], and the S92A mutation in this N-terminal domain dominant negatively disrupts mitochondrial Ca²⁺ uptake in cells [66]. Thus, the MCU N-terminal domain appears to serve as an important sensory hub which regulates MCU channel activity. Several high-resolution crystal structures of the human MCU N-terminal domain have been elucidated in the presence of various ions and concentrations (i.e. see PDB codes 5KUG, 5KUE, 5KUI, 5KUJ, 4XTB, 5BZ6, 4XSJ) [64,66]. The human MCU N-terminal domain forms a β-grasp-like fold made up of two three-stranded β-sheets holding two centrally located α-helices (figure 1e). The shorter α2-helix sits as a cap on the overall globular fold and is arranged approximately perpendicular to the central α1-helix. Interestingly, the extended loop 2, which connects the β1- and β2-strands and contains the S92 and C97 residues previously found to regulate channel activity by phosphorylation [66,67] and S-glutathionylation (SSG) [65], respectively, also contains a salt bridge which stabilizes this functionally important loop (figure 1f). The salt bridge is not present in all crystal forms, suggesting that local environmental conditions, including ionic strength, can modulate the structure and stability of the loop and, thus, the post-translational modifications which regulate channel activity.

The surface electrostatic potential of the MCU N-terminal domain forms two negatively charged acidic and two positively charged basic patches. The largest acidic patch is near the C-terminal region of the domain and is primarily constituted by D131, D142, D147, D148 and D166 residues (figure 1g). These acidic residues are conserved among vertebrate MCU homologues, suggesting an important role affiliated with the surface exposure. Indeed, Mg²⁺ ion has been found bound to this acidic patch in one crystal form [64]. The first shell Mg²⁺ coordination to the protein occurs through a D147 side chain oxygen atom, although the D131 and D148 side chains are also located within approximately 6.5 Å of the ion. The octahedral coordination geometry occurs via ligating of five additional water molecules, two of which are bridged by D123 side chain oxygens in second shell coordination (figure 1h). The single-ligand first shell and double-ligand second shell coordination interactions with the protein are consistent with the weak Mg²⁺ binding (i.e. approx. millimolar affinity) estimated using in vitro experiments with the isolated N-terminal domain in solution [64]. Ca²⁺ was also found to interact weakly with the N-terminal domain in solution using a similar assessment.

While the MCU N-terminal domain crystallized as a monomer, solution investigations by analytical ultracentrifugation and size-exclusion chromatography with inline multi-angle light scattering showed that the domain self-associates. Remarkably, inclusion of Ca²⁺ or Mg²⁺ divalent cations in the experimental buffers shifted the self-association equilibrium towards the monomer and, consistently, destabilized the fold. Furthermore, D131R or D147R charge swapping mutations caused a similar shift in the self-association equilibrium towards the monomer as the divalent cations [64]. Thus, incorporation of a basic residue in the acidic patch by mutation mimics the effects of basic cation coordination in this region of the MCU N-terminal domain. To assess the functional significance of this acidic patch, these mutations were also incorporated into constructs expressing full-length human MCU. In HeLa cells, expression of full-length human MCU harbouring D131R or D147R mutations caused a profound dominant negative inhibition of mitochondrial Ca²⁺ uptake without altering the IMM potential [64]. Furthermore, as observed in the isolated MCU N-terminal domain context, the mutations perturbed the full-length assembly of the protein complex. Consistent with the mutational phenotypes, blockade of Ca²⁺ extrusion from the matrix using a Na⁺/Ca²⁺ exchange inhibitor significantly decreased the Ca²⁺ uptake rates after a series of [Ca²⁺]_i pulses. With each successive [Ca²⁺]_i pulse, the matrix underwent loading concomitant with decreased mitochondrial Ca²⁺ uptake rates. Importantly, the membrane potential remained intact during this pulse train. Congruently, incubation with extramitochondrial Mg²⁺, which is taken up into the matrix via the Mrs2p channel, also significantly decreased mitochondrial Ca²⁺ uptake rates [64]. Collectively, these data identified an MCU regulating acidic patch (MRAP) on the conserved MCU N-terminal domain which binds Mg²⁺ and Ca²⁺, resulting in a destabilization of the self-association equilibrium and perturbation of the full-length MCU complex assembly. Ultimately, this divalent cation-induced perturbation in functional MCU architecture represents an autoregulation mechanism for physiologically vital cellular processes. The weak Mg²⁺ binding affinity of the MCU N-terminal domain is in the range of Mg²⁺ concentrations reported in the matrix and, thus, ideally suited as a sensor of Mg²⁺ levels which would be affected by Mrs2p activity as well as the metabolic state of the cell (i.e. ADP, ATP and PO₄³⁻ levels) [68,69]. On the other hand, the Ca²⁺ binding sensitivity of the MCU N-terminal would depend on a close proximity of the MCU N-terminal domain to the channel pore, where approximately millimolar levels may be reached in nanodomains [70–72]. Future high-resolution structural information of the intact human MCU protein is required to precisely capture the atomic mechanism for transduction from the disruption of the N-terminal domain self-association equilibrium to inhibition of MCU channel activity.

3. Regulation of mitochondrial Ca²⁺ transport by redox signalling

In agonist-stimulated non-excitable cells, it is generally accepted that [Ca²⁺]_i oscillations are triggered by Ca²⁺ release from the ER IP₃R and are fine-tuned by repetitive Ca²⁺ exchange between ER and the mitochondria, but they also depend on store-operated Ca²⁺ entry for long-term sustainability [7,73,74]. Specifically, following the binding of a cytokine to its G protein-coupled receptor (GPCR) on the plasma membrane, GPCR activates phospholipase C (PLC), which hydrolyses phosphatidylinositol 4,5-bisphosphate (PIP₂) to IP₃ and diacylglycerol (DAG) (figure 2a) [77]. IP₃ activates its receptor on the ER and promotes Ca²⁺ release [77] (figure 2a). In addition to regulation by IP₃, IP₃R is regulated by Ca²⁺ in a biphasic manner; low [Ca²⁺]_i stimulates IP₃R, whereas high [Ca²⁺]_i inhibits ER Ca²⁺ release [78,79]. Store depletion activates store-operated Ca²⁺ channels on the plasma membrane allowing Ca²⁺ influx from the extracellular space [77]. The mitochondria release some of their Ca²⁺ via the mitochondrial Na⁺/Ca²⁺ exchanger (mNCX), and part of that Ca²⁺ is taken up by the ER via the sarco-endoplasmic reticulum Ca²⁺-ATPase (SERCA) (figure 2a) [77,80,81]. Analysis of Ca²⁺ dynamics within intracellular stores, ER and mitochondria, simultaneously with [Ca²⁺]_i in stimulated HeLa cells, revealed that the first [Ca²⁺]_i oscillation is due to ER Ca²⁺ release with a fraction of that Ca²⁺ taken up by mitochondria, whereas subsequent [Ca²⁺]_i oscillations are initiated by mitochondrial Ca²⁺ release, which triggers regenerative Ca²⁺ release from the ER [82]. Samanta et al. [83] demonstrated that knocking down the MCU significantly inhibits the generation of [Ca²⁺]_i oscillations in stimulated mast cells, verifying the important role of the MCU in regenerative ER Ca²⁺ release.

Figure 2.

Ca²⁺ and ROS crosstalk between ER and mitochondria, with emphasis on redox control of the MCU. (a) In a non-excitable cell exposed to a chemical agonist, binding of the agonist to a cell surface receptor activates the GP/PLC/IP₃ pathway. IP₃ activates the IP₃R and causes Ca²⁺ release from the ER. Both mitochondrial Ca²⁺ uptake via the MCU and Ca²⁺ release via the mNCX differentially regulate the activation of IP₃R in the MAM region (because IP₃R is also regulated by Ca²⁺) and, thus, play a role in shaping the [Ca²⁺]_i transients and oscillations in response to the agonist. Mitochondrial Ca²⁺ release is thought to also help refill the ER via the SERCA. High [Ca²⁺]_m stimulates respiratory chain activity, leading to higher amounts of mROS. MROS can target the MCU and, following diffusion in the MAM region, the IP₃R and SERCA, further amplifying the disruption of Ca²⁺ and ROS homeostasis. (Adapted from [13,75].) (b) The MCU senses mROS via oxidation of the matrix-facing Cys-97. S-glutathionylation of Cys-97 induces a conformational change (and causes channel clustering; not shown) leading to increased MCU activity/mitochondrial Ca²⁺ uptake. (Adapted from [65,76].)

During an inflammatory/activated cell state, increases in [Ca²⁺]_i, due to Ca²⁺ release from the ER and/or Ca²⁺ influx from the extracellular space, result in increased [Ca²⁺]_m which, by stimulating enzymes of the Krebs cycle and oxidative phosphorylation and promoting ATP synthesis, enhances the mitochondrial electron transport chain (ETC) activity/mitochondrial O₂^·− production (provided there are no shortages in the supply of O₂ and electron donors [84]) and increases the mROS levels (figure 2a) [13,85]. Mitochondrial O₂^·− production takes place at ETC complexes I, II and III (mainly I and III; figure 2a) [86,87]. Complexes I and II release O₂^·− into the matrix, whereas complex III releases O₂^·− into both the matrix and the intermembrane space [88]. O₂^·− is converted to hydrogen peroxide (H₂O₂) by manganese superoxide dismutase (MnSOD) in the mitochondrial matrix and by copper zinc SOD (CuZnSOD) in the intermembrane space (cytosolic side) [89]. Any H₂O₂ that escapes the mitochondrial antioxidant mechanisms diffuses out and increases the cytosolic ROS levels (especially in the region between ER/mitochondria called mitochondria-associated membranes, MAMs [90]), and can cause oxidative modifications to the IP₃R, SERCA and even the stromal interaction molecule 1 (STIM1)-Orai1 on the plasma membrane, generally resulting in further increase in [Ca²⁺]_i (due to activation of IP₃R and/or inactivation of SERCA) (figure 2a) [13,91–93]. However, the functional consequences of ion channel regulation by the redox state are divergent. For example, a reactive Cys in STIM1 undergoes SSG that promotes Orai1/2 channel opening for Ca²⁺ entry without affecting intracellular store depletion [92], whereas reactive Cys oxidation in Orai1/2 channels inhibits Ca²⁺ entry [94,95]. mROS are involved in a plethora of signalling pathways, such as nuclear factor-κB (NF-κB)- and tumour necrosis factor-α (TNF-α)-induced signalling, leading to expression of cell adhesion molecules and cell dysfunction, but also activation of survival mechanisms [96–102]. ROS were found to exert their effect via SSG of regulatory Cys within NF-κB (inhibited by SSG) and the TNF receptor Fas (activated by SSG) [103–105]. However, the role of SSG of proteins within the mitochondrial compartment in the pathogenesis of cardiovascular disease has not been defined.

Although [Ca²⁺]_m elevation typically stimulates bioenergetics, under pathological stimuli (endotoxin insult), excessive and persistent [Ca²⁺]_m elevation causes [Ca²⁺]_m overload, leading to aberrant electron leak that promotes mROS production and cellular oxidative damage [92,106–108]. Since the mitochondria are one of the main endotoxin-sensitive ROS sources in ECs, it is to be expected that the MCU-mediated mitochondrial Ca²⁺ uptake will be the primary signal for mROS production. [Ca²⁺]_m and mitochondrial redox balance are generally accepted as both critical and interdependent determinants for EC survival and barrier function [2]. Although [Ca²⁺]_m and mROS go hand in hand, until recently, it was unknown whether the mROS can oxidatively modify any components of the MCU complex. Our group provided the first evidence that MCU activity is positively modulated via MCU protein oxidation by the luminal mROS [65]. Priming of ECs with endotoxin (oxidative stress) results in thrombin-induced [Ca²⁺]_m elevation that leads to a further increase in mROS, reduction in ATP production and mitochondrial respiration, upregulation of pro-inflammatory genes, and an increase in apoptosis [65]. mROS appear to only oxidize the core component of the uniporter complex, the MCU itself; other MCU complex components are not susceptible to oxidation under similar conditions [65]. The conserved Cys at position 97 of the human MCU, located in the N-terminal domain of the MCU facing the matrix, is the only reactive thiol within the MCU that undergoes redox modification (SSG) under conditions of high mROS, leading to enhanced MCU activity (figure 2b) [65,76]. Although oxidation of the MCU does not alter the interactions of the MCU with its regulators, oxidation induces a conformational change within the N-terminal β-grasp fold, redistributing the MCU complex to form higher order oligomers and increased channel activity. One of the mitochondrial intermembrane space (IMS) proteins, Mia40, both undergoes oxidative folding and mediates protein import to the IMS [109]; however, it is unknown whether it senses mROS. In summary, the MCU has a moonlighting function as a mitochondrial luminal redox sensor that enables mitochondria to decode mROS pathophysiological signals to metabolic or cell death responses. Overall, a better understanding of the Ca²⁺ and ROS crosstalk at the (sub)cellular level is critical, because disruption of mitochondrial Ca²⁺ and mROS (and, as a consequence, of cytosolic Ca²⁺ and ROS) homeostasis has been implicated as the underlying cause in almost every cardiovascular and neurodegenerative disease, as well as in different stages of cancer development [110–113]. It is noteworthy that the interplay between Ca²⁺ and ROS is not limited to the mitochondria. In most pathological conditions, decreases in ER Ca²⁺ levels lead to ER stress/activation of the unfolded protein response (UPR), which, via ER ROS, contributes to the redox regulation of local Ca²⁺ transport [7,13,114].

4. Regulation of mitochondrial Ca²⁺ transport by mechanical forces

Mechanical forces are known factors in regulating every organ system in the human body, but are crucial in the development and physiology of the circulatory system [115,116]. Fluid mechanical (haemodynamic) forces exerted on the vessel wall are particularly important in vessel structure/remodelling and in vascular pathology, such as atherosclerosis, aneurysms and arteriovenous malformations [117–122]. Atherosclerotic vascular disease continues to be the major cause of death in developed nations [25]. Though influenced by systemic risk factors (hyperlipidaemia, hypertension, diabetes, smoking), atherosclerotic plaques develop preferentially at the outer walls of bifurcations, the inner walls of curvatures and downstream of a stenosis in the carotid, coronary, aortic, renal, iliac, femoral and other arteries [123,124]. It is well accepted that modulation of the vascular EC phenotype by the local haemodynamic forces in those regions contributes to the geometrically focal nature of the disease [26,125–127]. ECs in arteries and arterioles are continuously exposed to three types of haemodynamic stress (force/surface area): normal stress (due to hydrostatic pressure), tensile stress (due to transmural pressure difference) and shear stress (the frictional force exerted by blood flowing over the EC luminal surface) [128,129]. As EC dysfunction is a key and early factor in atherogenesis, and ECs, due to their unique location, experience complex blood flow/shear stress profiles, a lot of research has focused on the effects of shear stress on EC physiology and pathology.

Wall shear stress is the product of blood viscosity and the velocity gradient (slope of the velocity vector in a direction normal to the wall) on the EC luminal surface. For most of the arterial circulation, local wall shear stress can be considered proportional to the volumetric flow rate and inversely proportional to the cube power of the internal radius at each axial position along the blood vessel [130–132]. Using different imaging modalities to measure flow rates and internal diameters, the wall shear stress in straight portions of the arterial vascular network was estimated to vary between 1 and 7 Pa (10–70 dynes cm⁻²). Those regions, which include locations preceding and following arterial bifurcations as well as the inner walls of flow dividers, are spared from atherosclerotic disease. Flow in those regions is undisturbed (characterized by parallel streamlines), laminar (Reynolds number less than 2100), unsteady (due to the cardiac cycle) and unidirectional; the wall shear stress is called pulsatile (PS) and has a relatively high mean value (figure 3a) [119,125,133]. In regions susceptible to disease, such as the outer walls of bifurcations, the local flow is disturbed and reverses during the systole phase of the cardiac cycle (within a flow recirculation zone with high temporal and spatial shear stress gradients), but it is still laminar; the wall shear stress is called oscillatory (OS), has a mean close to zero and varies between ±0.6 Pa (6 dynes cm⁻²) (figure 3a) [119,125,133]. Histology and digital image analysis of microparticle flow in post-mortem arteries, non-invasive (magnetic resonance phase velocity mapping) and invasive imaging modalities (coronary angiography, intravascular ultrasound), together with computational fluid dynamics, have verified the spatial correlation between OS and atherosclerosis initiation and progression in animals and humans [120,124,134,135].

Figure 3.

Fluid shear stress profiles and their effects on Ca²⁺ and ROS signalling in the vascular endothelium. (a) Schematic of axial velocity profiles at different locations in a vascular bifurcation, such as the carotid, coronary, renal and iliac artery flow dividers. (Adapted from [119].) Regions on the vessel wall exposed to PS (blue circles) are spared from atherosclerotic disease, whereas those in the flow recirculation zone (yellow circles) are exposed to OS and are prone to disease. (b) Idealized shear stress profiles versus time are shown for PS, SS, OS and static controls (used in in vitro studies). (c) PS onset raises cytosolic ROS levels; both cytosolic ROS and mROS levels decline/return to baseline at longer times. ROS levels decline due to decreased NOX expression, but any remaining O₂^·−/ROS scavenge NO. NO levels are high due to increased eNOS activation and, at longer times, expression. NO promotes generation of mitochondrial O₂^·−/mROS. MROS regulate the MCU activity (shear-induced changes in expression of the MCU and its regulators are currently unknown) and, via [Ca²⁺]_m, are expected to affect the [Ca²⁺]_i response (and vice versa). [Ca²⁺]_i drives MCU expression. [Ca²⁺]_m increases mitochondrial O₂^·−/mROS. [Ca²⁺]_i and mROS/cytosolic ROS are known to regulate autophagy, which is reported to be fully functional under PS, and, thus, helps maintain mROS levels under control. The combination of high bioavailable NO, low ROS and functional autophagy results in the atheroresistant endothelial phenotype. (d) Under OS, ER stress/UPR contributes to increased mROS (and vice versa) and cytosolic ROS levels. Upregulated NOXs also lead to increased cytosolic ROS levels. eNOS uncoupling further contributes ROS. Increased cytosolic ROS and downregulated eNOS expression result in decreased NO levels. Elevated mROS levels enhance the MCU activity (any changes in mRNA/protein levels are unknown), which leads to increased [Ca²⁺]_m with unknown, but anticipated, effects on [Ca²⁺]_i. Autophagy, under regulation by [Ca²⁺]_i and mROS/cytosolic ROS, is impaired (specifically, the flux is impaired) under OS. The combination of diminished NO, high ROS, and dysfunctional autophagy results in the atherosusceptible endothelial phenotype.

In order to understand the EC phenotypes due to exposure to different shear stress profiles, many groups have subjected cultured ECs from different species and vascular beds (or ex vivo arteries) to controlled flow conditions using perfusion chambers/systems or cone-and-plate viscometers followed by biochemical analysis at both the gene and protein levels [126,136–140]. With few exceptions, most of those studies approximated PS and OS profiles as sinusoidal waves of fixed amplitude and period (figure 3b). For corresponding controls, cultured ECs were either exposed to undisturbed laminar steady shear stress (SS) at a value equal to the PS time-average or were left static (figure 3b). For verification of findings in vivo, the endothelium was isolated from regions in animal arteries known to be exposed to either PS or OS, and analysed for differential gene expression (mRNA, miRNA), and these databases drove further in vitro experimentation on upstream intracellular signalling pathways and downstream cell behaviour [127,141,142]. Changing arterial flow in animals by creating a stenosis through partial ligation (OS develops downstream) established the causal relationship between OS and atherogenesis [143]. In vitro studies showed the existence of metrics of PS and OS, besides the main ones (mean value, amplitude and period), such as shear stress angular direction, harmonic frequencies and oscillatory shear index, that are also sensed by ECs [144–146]; in this review, we will refer to PS and OS based solely on their main metrics. Overall, both in vivo and in vitro studies verified that OS induces EC inflammation and dysfunction, which, in the presence of additional systemic risk factors, primes the OS-exposed arterial regions for disease initiation. It is noteworthy that the detrimental effect of OS is not purely mechanical; prolonged residence time of cytokines and other chemicals released by ECs and trapped blood cells in the flow recirculation zone contributes to the EC dysfunction.

PS (or SS) is known to decrease the EC proliferation rate and extent of apoptosis (the latter, following EC exposure to a stimulus that causes oxidative stress), rendering the endothelium ‘atheroresistant’, whereas OS does the opposite (‘atherosusceptible’ endothelium) [147–150]. The atheroprotective effect of PS (or SS) is, to a large extent, due to shear-induced expression and activation of endothelial nitric oxide synthase (eNOS) that produces the vasodilator NO (figure 3c) [151–154]. By contrast, eNOS expression was decreased under OS in vivo, compared with PS, and in vitro, compared with static controls (figure 3d) [153,155,156]. In agreement with that, isolated arterioles from animals/humans showed NO-dependent dilation under PS or SS; OS did not induce dilation [157,158]. PS (or SS) was found to trigger EC secretion of ATP causing autocrine/paracrine activation of surface purinergic receptors (G_q/G₁₁ protein-coupled P2Y₂Rs), activation of PLC, generation of IP₃, release of Ca²⁺ from the ER IP₃R, and [Ca²⁺]_i transients, which led to sequential phosphorylation of platelet/endothelial cell adhesion molecule-1 (PECAM-1), vascular endothelial growth factor receptor 2 (VEGFR2), Akt and eNOS [159–161]. [Ca²⁺]_i signalling is necessary for eNOS activation, because P2Y₂R deficiency blocked both the shear-induced [Ca²⁺]_i transients and PECAM-1/VEGFR2/Akt/eNOS activation [161]. [Ca²⁺]_i signalling is also important for PS (or SS)-induced eNOS expression, which occurs due to Ca²⁺/CaM-mediated phosphorylation/nuclear export of histone deacetylase 5 (HDAC5) and resultant expression of Krüppel-like factors 2 and 4 (KLF2/4), both of which are transcription factors for eNOS [162]. eNOS is transcriptionally repressed under OS due to miR-92a targeting of KLF2/4 [156,163] and probably also due to increased expression/activation of the tumour suppressor p53 [164,165].

Several groups measured SS-induced EC [Ca²⁺]_i changes, but reported conflicting data regarding the type of the response (single transient versus oscillations) and the relative contribution of Ca²⁺ release from the ER versus Ca²⁺ entry from the extracellular space [166–172]. Most of them spatially averaged fluorescence signals from Ca²⁺-sensitive probes over ECs in a microscope field of view and reported an initial peak followed by a slow decline to baseline in response to SS. By monitoring [Ca²⁺]_i changes versus time in individual cells, our group concluded that a significant percentage of ECs respond to SS with [Ca²⁺]_i oscillations (repetitive transients), and [Ca²⁺]_i peak amplitudes and oscillation frequencies depend on the SS level [75]. Importantly, the SS-induced [Ca²⁺]_i response was due to release of Ca²⁺ from the ER and required [Ca²⁺]_m uptake via the MCU and release via the mNCX [75]. As Ca²⁺ is known to regulate the IP₃R in a biphasic manner (§3), these data suggested that mitochondrial Ca²⁺ uptake via the MCU within the MAMs may suppress the Ca²⁺-inhibitory effect on IP₃R (following prior ER Ca²⁺ release), causing continuation of ER Ca²⁺ release and recurrence of [Ca²⁺]_i oscillations [75]. Not only is [Ca²⁺]_i regulated by mitochondrial Ca²⁺ transport, but also [Ca²⁺]_m is controlled by [Ca²⁺]_i via Ca²⁺-mediated activation of CREB, the transcription factor that induces MCU expression [173]. A group that measured [Ca²⁺]_i under PS versus OS reported [Ca²⁺]_i oscillations in ECs exposed to either flow, an initial peak in spatially averaged [Ca²⁺]_i in PS-exposed ECs, and maintenance of spatially averaged [Ca²⁺]_i at baseline in OS-exposed ECs [174,175]. Prendergast et al. [176] showed enhanced [Ca²⁺]_i signals in ECs from thoracic aorta (PS-exposed) compared with ECs from aortic arch (OS-exposed) in isolated tissues from either wild-type (WT) or ApoE^−/− mice (prior to lesion development) following exposure to carbachol, suggesting that [Ca²⁺]_i responses in OS-conditioned ECs may be blunted compared with those in PS-conditioned ECs. This agrees with recent findings by Taylor et al. [177], who showed that acetylcholine (ACh)-induced EC [Ca²⁺]_i dynamics were diminished in carotid arterial regions proximal to a partial ligation (OS-exposed), compared with PS-exposed control regions, in WT mice. To the best of our knowledge, the role of mitochondrial Ca²⁺ transport in [Ca²⁺]_i has not been examined in ECs under the physiological flow profiles PS and OS. Concurrent recordings of spatio-temporal profiles of [Ca²⁺]_i and [Ca²⁺]_m have shed light on the communication between the two Ca²⁺ pools in different cell types under chemical stimulation [5,82,83,178,179], but these experiments have not been conducted in ECs exposed to shear stress. Interestingly, work by the Pozzan group showed that MCU overexpression, via enhanced Ca²⁺ buffering by mitochondria, reduces the [Ca²⁺]_i peaks both during spontaneous oscillations and following caffeine treatment in cardiomyocytes [180]. A higher EC MCU activity (and/or expression) in OS- versus PS-exposed arterial regions would lead to increased mitochondrial Ca²⁺ uptake and would explain the reported diminished [Ca²⁺]_i responses under OS.

The pro-inflammatory signalling pathways underlying atherogenesis, including c-Jun N-terminal kinase (JNK)/the family of mitogen-activated protein kinases (MAPKs) and the transcription factors NF-κB and activator protein-1 (AP-1), are redox sensitive [181]. OS was found to cause a sustained increase in O₂^·− production/ROS levels, whereas PS (or SS) induced an increase upon the onset of flow, but, after the first couple of hours, the increase in ROS was compensated by upregulation of antioxidant defences, such as haem oxygenase-1 (HO-1) and NADPH quinone oxidoreductase 1 (NQO1), under the control of nuclear factor erythroid 2-related factor 2 (Nrf2) [153,182–186]. NADPH oxidases (NOXs), xanthine oxidase, mitochondria and uncoupled eNOS are the main sources of cytosolic ROS in ECs [187]. PS (or SS) downregulated NOX4 expression; OS upregulated NOX1 and NOX2 expression in parallel with xanthine oxidase levels, with mixed results regarding the OS effects on NOX4 (figure 3c,d) [153,188]. NOX1-derived ROS, in particular, were responsible for activation of NF-κB and upregulation of intercellular adhesion molecule-1 (ICAM-1) in OS-exposed ECs in vitro and in vivo [188]. Shear-induced ROS are expected to play a key role in Ca²⁺ homeostasis, because the second messengers are known to interact in a bidirectional manner (in general, increased [Ca²⁺]_i causes an increase in [Ca²⁺]_m and mROS, whereas increased mROS result in a further increase in [Ca²⁺]_m) (figure 3c,d). In addition, ECs in atheroprone areas in vivo were found to undergo ER stress, and UPR activation is associated with increased ER ROS [114,127,189]. Increased mROS are thought to both activate ER stress/UPR and be the result of ER stress/UPR (the latter due to increased Ca²⁺ release from the ER resulting in elevated [Ca²⁺]_i and [Ca²⁺]_m) (figure 3c,d) [13,190]. Any O₂^·− in the cytosol/organelles that escaped antioxidant defences is consumed by a diffusion-limited reaction with NO to form the reactive nitrogen species (RNS) peroxynitrite (ONOO⁻), thus lowering the bioavailable NO levels (ONOO⁻ can also uncouple eNOS) and promoting atherosclerosis [191,192]. Either OS-exposed ECs in vitro or atheroprone regions in vivo had increased levels of nitrotyrosine, the footprint of ONOO⁻ formation, compared with PS-exposed ECs and atheroresistant regions, respectively (due to increased O₂^·− production and despite decreased NO generation under OS) [193]. Even following short exposure of cultured ECs to SS, low levels of ONOO⁻ formed in mitochondria resulted in partial inhibition of mitochondrial respiratory complexes and increased mitochondrial O₂^·−/ROS levels; the effect was attenuated when SS was performed at lower than atmospheric (closer to arterial) O₂ levels [194,195]. Interestingly, NOX-derived O₂^·−/ROS were found to either activate or inactivate eNOS: under SS (and one might assume the same for PS), activated NOX2/p47phox and resultant ROS were responsible for eNOS activation, whereas, under OS, ROS from activated NOX1/NOXO1 caused eNOS uncoupling leading to eNOS-derived O₂^·−/ROS and a concomitant decrease in NO production/bioavailable levels (figure 3c,d) [196]. Based on the higher levels of mROS/cytosolic ROS in OS-exposed ECs compared with PS (or SS)-exposed ECs and the recently discovered mROS sensing by the MCU [65], one can speculate that there may be increased MCU activity/mitochondrial Ca²⁺ uptake/[Ca²⁺]_m levels under OS versus PS (figure 3c,d), which may be the reason behind the OS-induced EC inflammation and decreased cell resistance to apoptosis.

5. Mitochondrial Ca²⁺ uniporter: a potential therapeutic target in vascular diseases

The speculation regarding altered MCU activity in the endothelium of atheroprone regions, compared with activation in atheroprotective ones, is in agreement with the oxidative stress and inflammation encountered in those locations. Specifically, under OS, increased [Ca²⁺]_m/mROS/cytosolic ROS would dampen NO bioavailability and render the endothelium dysfunctional, which could result in vascular risk factors (figure 3d). It is safe to assume that, even if the global EC [Ca²⁺]_i does not change significantly in OS-exposed regions, [Ca²⁺]_i in specific locations of the cytosol, such as in proximity to the mitochondria (e.g. the MAMs), could influence [Ca²⁺]_m overload. Based on the speculation of regional dependency of MCU activity, graded inhibition of MCU activity (and/or expression), down to levels encountered in atheroprotective regions, may be a great strategy to prevent EC dysfunction and atherosclerotic disease initiation/progression. As MCU activity is elevated following its oxidative modification by mROS [65], administration of mitochondria-targeted antioxidants would have a beneficial effect. In agreement with this notion, Li et al. [197] showed that human aortic ECs (HAECs) treated with the pro-inflammatory lipid lysophosphatidylcholine increased their mROS due to enhanced MCU activation; both were attenuated in the presence of the mROS scavenger mitoTEMPO. MitoTEMPO also suppressed EC activation and monocyte recruitment in the aorta of ApoE^−/− mice fed with a high-fat diet [197]. Although this study did not examine specific aortic locations, the beneficial effect of the drug would be apparent in locations that experience OS. Decreasing mitochondrial Ca²⁺ entry in vitro can also be done by knocking down proteins involved in ER–mitochondria contact sites and, thus, increasing the MAM width. Although this method cannot be translated to clinical settings, nevertheless it was shown that it protects cultured cells from oxidative stress-induced apoptosis [198]. One of the reasons why graded inhibition of MCU activity may be able to maintain EC health in OS-exposed regions is its potential impact on autophagy. Autophagy is a critical homeostatic mechanism in cardiovascular health [199–201]. Functional autophagy was shown to be a prerequisite for the SS-induced increase in NO production [202,203]. Dysfunctional autophagy occurs when autophagy induction (autophagosome formation) becomes highly activated and, as a result, the autophagic flux (rate of autophagosome–lysosome fusion and degradation of autophagosomal cargo) slows down (‘impaired’ flux) leading to inadequate removal of dysfunctional mitochondria and a tonic increase in oxidative stress. Li et al. [204] showed that OS exposure increases the induction of autophagy, but decreases the autophagic flux in cultured HAECs (whereas PS maintains autophagy at levels comparable to static controls); both the increased induction and decreased flux under OS were mediated by JNK activation and elevated mROS. The impaired EC autophagic flux was verified in OS-exposed regions of rabbit aortas [204]. Graded inhibition of MCU activation in OS-exposed regions will reduce mROS, which, by lowering autophagy induction, should allow for, at least, partial normalization of the impaired flux, in agreement with a recent review [205]. In support of the importance of a functional autophagic flux in EC health, it was recently shown that SS upregulates the transcription factor EB (TFEB), master regulator of autophagic flux and lysosomal biogenesis, and that transgenic mice with EC-specific TFEB expression exhibit reduced leucocyte recruitment to ECs and decreased development of atherosclerosis [206]. Since mitochondrial ion homeostasis plays a major role in endothelial function, we put forward a hypothesis that perturbation of mitochondrial Ca²⁺ homeostasis could lead to several vascular pathological conditions.

Data accessibility

This article has no additional data.

Authors' contributions

B.R.A. and M.M. conceived the themes of this manuscript. B.R.A., S.S., A.P., P.B.S. and M.M. drafted the manuscript. All authors gave final approval for publication.

Competing interests

We declare we have no competing interests.

Funding

B.R.A. is supported by an American Heart Association Grant-in-Aid. P.B.S. is supported by the Natural Sciences and Engineering Research Council of Canada, Prostate Cancer Canada Network (Telus Ride for Dad) and the Canadian Foundation for Innovation. M.M. is supported by NIH R01GM109882, R01HL086699, R01HL119306 and 1S10RR027327 grants.