Mitochondrial Bioenergetics and Dynamics in Secretion Processes

In ERMCs we can also find SAR1 and ARF. These small GTPases involved in COPII and COPI vesicle assembly are also required to maintain mitochondrial morphology and function (145, 146). SAR1 regulates the size of mitochondrial ER contact sites by affecting the curvature of the membrane (146). While ARF1 and its guanine nucleotide exchange factor (GBF1) interact with mitochondrial protein MIRO, and their depletion affects mitochondrial morphology, autophagy, positioning, and movement within the cell (145, 147). These observations point to an important role of secretory pathway components in dynamics and physiology of mitochondria.

Overall, mitochondria are dynamic organelles whose function, morphology, distribution, and contacts with other organelles (49, 109, 148) undergo changes in response to intracellular energy requirements, as well as other physiological and pathological stimuli (109).

As mentioned before, many of the steps in the conventional secretion pathway demand energy (Table 1 and Figures 1, 2). Chaperones involved in protein translocation to the ER and protein folding consume ATP (5–7), as does protein quality control by the ERAD (8), transport and fusion of vesicles with target membranes (2, 9, 51). Calcium signaling during regulated exocytosis also consumes ATP to maintain the levels of IP3 precursors and to pump calcium from the cytoplasm into the ER or to the extracellular space (13, 14, 149). Overall, an increase in secretion of proteins is bound to challenge the cell's bioenergetics capacity; and inhibition of ATP synthesis, using inhibitors or uncouplers of oxidative phosphorylation, prevents the secretion of many proteins (7, 150–152).

The mechanism for ATP transport into the ER in mammalian cells was elusive for many years, and different possibilities, ranging from the existence of specific transporters to non-specific transport through anion channels or leaky membranes, have been proposed (4). However, recently protein SLC35B1 was identified as an ATP/ADP exchanger present in the ER membrane, and named AXER (153). Silencing the expression of SLC35B1/AXER reduced ATP levels in the ER of different cell lines and affected BiP dependent protein import activity and Ca2+ levels inside the ER (153, 154). A decrease in ATP in the ER was observed upon inhibition of SERCA or exposure to high cytosolic Ca2+ concentrations, suggesting that a Ca2+ gradient across the ER membrane is required for ATP transport into the ER (154). Inhibitors and uncouplers of oxidative phosphorylation lead to a decrease in ATP levels inside the ER as well, suggesting mitochondria is a relevant source of ATP for the ER (154, 155). Interestingly, several components of the mitochondrial ATP synthase complex immunoprecipitate with SLC35B1/AXER; along with proteins involved in ER protein import and folding such as BiP, calnexin, and Sec61 complex (153). Transfer of ATP from mitochondria to the ER probably occurs at ERMCs, where both organelles are in close proximity, yet this is still to be established (4).

As we mentioned above, Ca2+ signals for exocytosis in many secretory cells. The increase in cytosolic Ca2+ concentration triggers vesicle fusion with the plasma membrane leading to the release of the cargo to the extracellular space. Calcium enters the cytosol through channels in the plasma or the ER membranes and several processes can affect the magnitude and duration of the signal for exocytosis, among them is Ca2+ uptake by mitochondria (13). Mitochondria are located in close proximity to Ca2+ channels in the ER and plasma membrane and can take up the ion through the VDAC and MCU channels in the OMM and IMM, respectively (144). This process is known as Ca2+ buffering and is driven by the negative potential across the IMM (144). Positioning of mitochondria near the sites of Ca2+ entry to the cytosol is also a requirement for the efficient uptake of the ion and effective Ca2+ signaling for exocytosis. This depends on proteins involved in mitochondrial dynamics and movement, such as MFN2, myosin, kinesin, TRAK, and MIRO (14, 156–158). Of note, MIRO is both an adaptor protein, that links mitochondria to TRAK and kinesin, and a Ca2+ sensor which can regulate mitochondrial trafficking in response to changes in Ca2+ concentrations (157).

Ca2+ buffering has different effects on exocytosis, depending on the channel and cell type. Voltage-gated Ca2+ channels (Cav) are present in the plasma membrane. These are activated by membrane depolarization and mediate Ca2+ influx in response to action potentials and other depolarizing signals (159). The increase in cytosolic Ca2+ stimulates exocytosis by binding syaptogamins that promote SNARE mediated vesicle fusion with the membrane (65). Mitochondria can be found in the cell cortex regions, near the sites of exocytosis and in this setting Ca2+ sequestering by mitochondria reduces vesicle release (157, 160, 161).

In the plasma membrane we also find store-operated channels, known as Ca2+ release-activated (CRAC) channels. These are composed by Orai pore proteins, present in the plasma membrane, and STIM proteins from the ER and are activated by depletion of calcium stores in the ER (13, 149, 158). When Ca2+ levels fall in the ER STIM proteins migrate to ER-plasma membrane junctions where they activate Orai channels, allowing the entry of Ca2+ into the cytosol. In turn, high cytosolic Ca2+ concentrations result in the closure of the pore and both a rapid and a slow Ca2+-dependent inactivation have been described (158). Mitochondrial translocation to the plasma membrane is essential to sustain Ca2+ influx through CRAC channels (162), since it prevents the slow inactivation of the CRAC channel without affecting the fast Ca2+-dependent inactivation (158). Inhibitors and uncouplers of electron transport and oxidative phosphorylation that dissipate the proton gradient prevent Ca2+ uptake by mitochondria and promote channel inactivation (163). Inhibition of kinesin- dependent transport of mitochondria to the plasma membrane also results in channel inactivation (162).

In the ER membrane the IP3R and ryanodine-sensitive receptors are the channels responsible for Ca2+ release to the cytosol (13, 14). IP3R is activated by IP3, a second messenger formed by hydrolysis of phosphatidylinositol 2,5-biphosphate by phospholipase C. While ryanodine- sensitive receptors in secretory processes are activated by Ca2+ to release Ca2+ stored in the ER, in a process known as Ca2+ induced Ca2+ release that amplifies Ca2+ signaling (149). For both channels the response to cytosolic Ca2+ concentrations is bell shaped, since Ca2+ can act both as a positive or negative regulator of the channel depending on the concentration of the ion (14, 164). Thus, high concentrations of cytosolic Ca2+ result in a decrease in the activity of both I3PR and ryanodine sensitive receptors (14, 164). Uptake of Ca2+ by mitochondria prevents its accumulation precluding channel inhibition by high concentrations of Ca2+ (14).

From what we just described it is clear that mitochondria contribute to the regulation of intracellular Ca2+ concentrations, and therefore impact on the regulated exocytosis of many proteins and other signaling molecules such as neurotransmitters (Figure 1).

The NLRP3 inflammasome is a multiprotein complex that participates in innate immunity. It is activated by multiple signals of infection, cellular damage, or stress, to produce inflammatory cytokines that trigger innate immune responses (71, 165). The sensor molecule NLRP3, the adaptor protein ASC and the effector caspase-1 form the complex, and recently the NIMA-related kinase 7 was identified as a fourth component. Upon stimulation, NLRP3 oligomerizes and recruits ASC that in turn recruits caspase 1. The assembly of the NLRP3 inflammasome leads to the activation of caspase 1 that then cleaves pro- IL-1β and pro-IL-18, producing the mature cytokines IL-1β and IL-18 (71). As mentioned above the assembly of the NLRP3 inflammasome requires energy that is obtained from ATP hydrolysis catalyzed by NLRP3 NACHT domain (71). Mitochondria supplies ATP for this process but contributes also in other ways to the activation of the inflammasome.

NLRP3 inflammasome activation occurs in two steps, priming and NLRP3 activation. Priming involves the recognition of pathogen-associated molecular patterns (PAMPs, such as LPS) by pattern recognition receptors or binding of cytokines to receptors. These events lead to the activation of NF-κB and consequent upregulation of NLRP3, caspase 1 and IL-1β and IL-18 gene expression (71). The second step is the activation of NLRP3 by PAMPs or damage-associated molecular patterns (DAMPs, such as ATP); which involves multiple signaling events that result in the assembly of the complex [for exhaustive reviews see (71, 165)]. Mitochondrial ROS and oxidized mtDNA released into the cytosol can trigger NLRP3 activation (16, 17).

Moreover, mitochondria are required for the assembly of the inflammasome. In absence of stimuli the NLRP3 is found in the cytoplasm associated with the ER, while its activation results in an association with mitochondria and enrichment in ERMCs (17). ASC and caspase 1 also migrate to mitochondria upon activation and are enriched at ERMCs (17, 18). In a recent report Elliot et al. suggest that priming results in NLRP3 and capase 1 linkage to the OMM, by association with the phospholipid cardiolipin, and mtROS are proposed to mediate cardiolipin transfer from the IMM to the OMM. Instead the association of ASC with mitochondria occurs in the activation step and does not require cardiolipin, but is dependent on NLRP3 and regulated by cytosolic Ca2+ (18). Interestingly in absence of VDAC, a relevant OMM channel and tether molecule in mitochondria-ER contact sites, activation is not observed (17).

Other mitochondrial proteins that are required for NLRP3 inflammasome activation are mitochondrial antiviral signaling proteins (MAVS). These proteins form aggregates in the OMM, associate with NLRP3 and promote its oligomerization, and activation of the inflammasome during RNA virus infections (166). The fusion protein MFN2 also participates in inflammasome activation and its silencing reduces IL-1β secretion in response to viral infection (167). Finally, dynein dependent transport of mitochondria along microtubules, toward the ER, is required for the activation of the inflammasome (168). Acetylation and deacetylation of tubulin by acetyltransferase MEC17 and the NAD+ dependent deacetylase sirtuin 2 regulate microtubule transport of mitochondria to the ER and inflammasome activation (168).

In sum, mitochondria provide energy, signaling molecules and a structural scaffold for the assembly and activation of the NLRP3 inflammasome (Figure 1).

Considering the relevant roles of mitochondria in secretion it is not surprising that mitochondrial dysfunction underlies the pathogenesis of certain diseases. Herein we discuss the role of this organelle in some relevant physiological secretion processes, and present evidence linking bioenergetic failure to the development of disease.

Pancreatic β-cells are secretory cells located in the islets of Langerhans in the pancreas. Their main function is to synthesize and secrete insulin, a peptidic hormone responsible for regulating levels of glucose in the blood. Upon the increase in glucose concentrations in plasma, β-cells secrete insulin stored in the secretory granules and increase the synthesis of the hormone (169, 170).

Mitochondria play a key role in glucose-stimulated insulin secretion (GSIS) in pancreatic β-cells. In these cells glucose is metabolized in the glycolytic pathway and tricarboxylic acid cycle and results in ATP synthesis by oxidative phosphorylation. ATP promotes the closure of ATP-sensitive potassium channels and depolarization of the plasma membrane, leading to the opening of voltage-gated Ca2+ channels. Calcium influx then triggers the exocytosis of insulin granules (170). This process is termed KATP- dependent GSIS. A KATP- independent GSIS that depends on mitochondrial GTP and phosphoenolpyruvate metabolism has also been described (171, 172). Therefore, GSIS is extremely sensitive to genetic interventions and pharmacological agents that affect mitochondrial metabolism (22). This is clearly exemplified by the work of Kennedy et al. showing that depletion of mitochondria in the INS-1 insulin-secreting cell line precludes insulin secretion upon exposure to glucose (173). Likewise impairment of autophagy in pancreatic β-cells affects both mitochondrial respiration and GSIS, affecting glucose metabolism in vivo (174). As does inhibition of Ca2+ uptake by mitochondria after silencing of MCU or MICU1 gene expression, that also reduces ATP synthesis and insulin secretion (175).

Pancreatic β-cells present an active mitochondrial network where mitochondria constantly undergo fusion and fission events (176); and functional mitochondrial dynamics are required for GSIS in INS-1 cells and primary β-cells (23). Overexpression of FIS1 in human pancreatic β-cells resulted in fragmentation and clustering of the mitochondrial network in the perinuclear area, a decrease in mitochondrial membrane potential and ATP synthesis, along with impairment of GSIS (177). Similar results where observed upon silencing of FIS1 expression that affected respiration rates and insulin secretion (178), overexpression of a dominant negative form of fission protein DRP1 and DRP1 silencing (178, 179). Alteration in the fusion machinery also affected insulin secretion. Overexpression of MFN1 resulted in hyperfusion and aggregation of mitochondria, reduced ATP levels and GSIS (177); while OPA1 ablation impaired glucose-stimulated oxygen consumption rate, ATP production and insulin secretion (116).

It is clear that mitochondria play relevant roles in pancreatic β-cell secretion and a growing body of evidence links mitochondrial dysfunction to impaired insulin release in diabetes mellitus. Diabetes mellitus are a group of metabolic disorders characterized by hyperglycemia, resulting from defects in insulin secretion, insulin sensitivity, or both (180). In type 1 diabetes the cause is a total deficiency of insulin secretion, while in type 2 diabetes the cause is a combination of resistance to insulin action of the liver, skeletal muscle, and adipose tissue and an inadequate compensatory insulin secretory response (180, 181).

To start with, diabetes mellitus (type 1 and 2) is frequently observed in patients with inherited mitochondrial diseases, which are caused by defects in mtDNA or in nuclear genes encoding mitochondrial proteins that affect mitochondrial ATP synthesis (27). Several point mutations as well as deletions and rearrangements in mtDNA are strongly associated with the onset of diabetes (27, 182). As are mutations in nuclear genes encoding proteins involved in maintenance of the mitochondrial genome (27). The most common defect is the mutation m.3243A>G in MT-TL1, which encodes a mitochondrial tRNA for leucine (183) and can lead to maternally inherited mitochondrial diabetes and deafness (MIDD) (27). Other point mutations in mitochondrial tRNA genes MT-TL1, MT-TK, MT-TS2, MT-TE, and MT-TT as well as in MT-ND6, a core subunit of complex I, have also been associated with diabetes mellitus (27, 184). Patients, carrying the m.3243A>G mutation presented reduced pancreatic insulin secretion (185–187), yet increases in insulin resistance and alterations in glucose metabolism cannot be discarded as causes of the disease (188).

Additionally profound changes in mitochondrial metabolism have been observed in pancreatic islets of type 2 diabetes patients and animal models. These include down regulation of components of the tricarboxylic cycle, ETC, ATP synthase and proteins involved in transport across the OMM and IMM (29, 189); as well as a decrease in glucose oxidation and a loss of glucose-stimulated increases in NADH, ATP, and oxygen consumption (29, 190–192). Other authors have observed an upregulation of uncoupling protein 2 (UCP-2), an IMM protein that dissipates the proton gradient; in pancreatic islets in type 2 diabetes (190, 193, 194). Genetic deletion of UCP-2 in mice enhances islet ATP generation and insulin secretion during glucose stimulation and reduces hyperglycemia in the ob/ob mice (a well-known diabetes type 2 mouse model), providing further support for the role of UCP-2 in the pathogenesis of the disease (194). Regardless of the mechanism, the impairment in mitochondrial function in pancreatic β cells precludes the glucose-induced increase in ATP levels. Lack of ATP will prevent the closure of ATP-sensitive K+ channels (that is required to activate insulin secretion), impair the steps of the secretion pathway that require energy, resulting in lower levels of the circulating hormone and hyperglycemia.

Neurons are excitable cells that communicate with other cells at synapses. At the chemical synapse neurotransmitters are released by pre-synaptic neurons and bind their receptor at the post-synaptic terminal of target cells. Vesicle discharge at the synaptic cleft occurs in response to an influx of Ca2+ ions through voltage-gated Ca2+ channels, triggered by the action potential. Three different pools of synaptic vesicles can be found in neurons readily releasable pool, the recycling pool, and the reserve pool. The vesicles in the readily releasable pool are released under moderate or intense neuronal activity, while vesicles from the reserve pool are recruited only upon intense stimulation. The latter constitute the majority of vesicles in presynaptic terminals (28, 160).

Maintaining the electrochemical gradients required for action potentials, transporting, discharging, recycling, and refilling synaptic vesicles, and regulating Ca2+ concentrations are energy demanding processes, that require functional mitochondria (28, 195). Regarding secretion in particular, loading of vesicles with neurotransmitters requires ATP to fuel the vacuole type H+-ATPase (72). Besides, the mobilization of the reserve pool requires mitochondrial ATP that is used by myosins to move the vesicles along actin filaments (196). Additionally, Ca2+ signaling for vesicle exocytosis requires a constant export of the ion from the cytosol to the extracellular space, the ER and mitochondria. Plasma membrane Ca2+ ATPase (PMCA) and SERCA at the ER actively consume ATP (149). Mitochondria provide ATP for these pumps and directly remove Ca2+ from the cytosol through the MCU, as described previously (28, 196). Therefore, mitochondria are very abundant in neurons, where they form a highly dynamic network and their position and morphology change continually (28). In particular we find a high number of mitochondria at chemical synapses, where they provide ATP and buffer local rises in Ca2+ concentrations (156, 160, 197). Since mitochondrial biogenesis and elimination (by mitophagy) occurs mainly in the neuronal soma, mitochondria travel between the axon terminals and the cell body (156) via the microtubule network. Motor proteins kinesin (anterograde) and dynein (retrograde), adaptor proteins TRAK, MIRO, and syntabulin as well as fission protein DRP1 are involved in mitochondrial trafficking (28, 156, 196). Docking to microtubules, by syntaphilin, or to actin filaments limits mitochondrial mobility, retaining these organelles near to the sites where energy is required (156, 160).

Due to their high-energy demands, neurons are extremely affected in mitochondrial diseases, caused by mutations in mtDNA, and in other pathologies where mitochondrial function is compromised (30, 198). An example of the latter is Alzheimer's disease (AD), a neurodegenerative disease that involves progressive synaptic dysfunction and death of neurons in cortical and subcortical regions. AD is characterized by the accumulation of extracellular plaques of the peptide amyloid-β (Aβ) and the formation of intracellular tangles of the microtubule-associated protein tau (198, 199).

Alterations in energy metabolism can be found in patients and animal models of AD. Reduced glucose metabolism was observed in brain regions affected by AD and the reduction correlated with cognitive decline in AD patients (200, 201). Analysis of brain samples from patients with AD showed abnormalities in mitochondrial cristae of hippocampal neurons (202); and reduced activity of enzymes involved in oxidative metabolism, namely pyruvate dehydrogenase (PDH), α-ketoglutarate dehydrogenase and complex IV (203–205). Studies of brain mitochondrial function in a transgenic AD mouse model also revealed decreases in PDH, complex IV and respiratory control ratios (RCR) and the decay in PDH and neuronal mitochondrial function preceded plaque formation (206). Moreover, in recent reports mitochondrial dysfunction was detected in hippocampal synaptosomes from transgenic AD rat model (207, 208). Decreased complex I activity, oxygen consumption rate, and ATP synthesis, along with an increase in hydrogen peroxide formation were measured, evidencing bioenergetics defects as well as oxidative stress at the presynaptic level (208). Alterations in mitochondrial dynamics and motility have also been observed in transgenic animal models of AD, and result in lower number of mitochondria at the presynapses (28, 209, 210). Regarding the causes behind bioenergetic failure in neurons, the Aβ peptide has been found to accumulate in mitochondria (206, 211), where it directly impairs complex IV activity and ATP synthesis, and increases ROS formation (212). Aβ can also directly affect mitochondrial dynamics, promoting mitochondrial fragmentation (213, 214). Defects in Ca2+ handling by mitochondria have also been observed in AD and Aβ can alter Ca2+ homeostasis at several levels (28, 73).

Hyperphosphorylated tau also affects mitochodrial respiration and ATP synthesis, through the inhibition of complex I, impairs mitochondrial trafficking and induces mitochondrial fission (215). Overall in AD patients impaired mitochondrial energy metabolism, dynamics, Ca2+ homeostasis and oxidant formation are affected in neurons, at presynaptic terminals where neurotransmitter secretion takes place (28, 198).

Synapse loss is an early event in AD animal models and patients, and exhibits a strong correlation with cognitive deficits (73). Presynaptic release of neurotransmissors is affected in AD, and Aβ can affect various steps of the secretion process including SNARE mediated fusion, recovery of vesicle membranes by endocytosis and Ca2+ homeostasis (73). The decrese in mitochondrial ATP synthesis and Ca2+ buffering, observed in AD at the presynaptic level, probably affects neurotransmitter release as well, and should be considered in the design of pharmacological strategies to target the disease (215). Supporting this concept are results obtained in a transgenic AD rat model, where treatment with a pharmacological inducer of mitochondrial biogenesis helped to recover mitochondrial function and protected against cognitive impairment (208).

In this section we present two examples of mitochondrial involvement in secretion processes that underlie pathology. This is not an extensive list, but rather a couple of interesting examples that support the idea that mitochondria might be interesting targets in the design of drugs to modulate secretory processes.

Mast cells are cells of hematopoietic origin that participate in host defense and immunity as well as tissue repair, wound healing, angiogenesis and are also responsible for the development of allergies (216). Mast cells can be activated by foreign pathogens, toxins chemical agents and particles, as well as by mediators of the immune response, such as immunoglobulin E (IgE) and complement components (216, 217). Upon activation mast cells release pre-stored mediators, and/or de novo synthesize and secrete a wide range of biologically active molecules (216). Pre-stored mediators include proteases, histamine, proteoglycans, and cytokines such as tumor necrosis factor (TNF), these are released immediately upon activation and are critical for initiating mast cell-mediated innate immune responses. Differently from pre-stored mediators, other molecules are synthesized de novo (i.e., lipids, cytokines, chemokines, antimicrobial peptides, growth and angiogenic factors) and play multiple roles in the immune response and other physiological and pathological processes (216, 217).

Mast cell degranulation and TNF secretion requires Ca2+ and mitochondrial ATP (31, 218). Interestingly stimuli that induce mast cell degranulation, promote the translocation of mitochondria to sites of exocytosis, close to the cell surface; and pharmacological or genetic inhibition of DRP1 inhibits both TNF secretion and mitochondrial translocation, in a cell line derived from human mast cell leukemia (24). In this report authors also show that calcium is required for DRP1 activation, as well as for mitochondrial translocation, and postulate that calcineurin is involved in DRP1 recruitment to mitochondrial surface (24). The relevance of this observation was further supported by studies in human mast cells from patients with atopic dermatitis where increased degranulation and mitochondrial translocation was also observed, along with an increase in DRP1 and calcineurin expression (24). Mast cell degranulation also results in the secretion of mitochondrial particles outside cells including mtDNA, that in turn promote cytokine release from other mast cells, keratinocytes, and endothelial cells promoting inflammation (219).