The Unfolded Protein Response and Membrane Contact Sites: Tethering as a Matter of Life and Death?

Abstract

The endoplasmic reticulum (ER) is the most extensive organelle of the eukaryotic cell and constitutes the major site of protein and lipid synthesis and regulation of intracellular Ca²⁺ levels. To exert these functions properly, the ER network is shaped in structurally and functionally distinct domains that dynamically remodel in response to intrinsic and extrinsic cues. Moreover, the ER establishes a tight communication with virtually all organelles of the cell through specific subdomains called membrane contact sites. These contact sites allow preferential, nonvesicular channeling of key biological mediators including lipids and Ca²⁺ between organelles and are harnessed by the ER to interface with and coregulate a variety of organellar functions that are vital to maintain homeostasis. When ER homeostasis is lost, a condition that triggers the activation of an evolutionarily conserved pathway called the unfolded protein response (UPR), the ER undergoes rapid remodeling. These dynamic changes in ER morphology are functionally coupled to the modulation or formation of contact sites with key organelles, such as mitochondria and the plasma membrane, which critically regulate cell fate decisions of the ER-stressed cells. Certain components of the UPR have been shown to facilitate the formation of contact sites through various mechanisms including remodeling of the actin cytoskeleton. In this review, we discuss old and emerging evidence linking the UPR machinery to contact site formation in mammalian cells and discuss their important role in cellular homeostasis.

Key Functions of the Endoplasmic Reticulum and the Unfolded Protein Response

Maintaining homeostasis is an important aspect of cellular physiology. This also applies to the different cellular organelles, all of which require optimal conditions in order to function and are under strict control. The endoplasmic reticulum (ER) is one of the largest cellular organelles and besides being a major hub of organelle biogenesis, membrane trafficking, lipid synthesis, and Ca²⁺ storage, it represents the major site of protein folding activity in the cell. Approximately one third of the cell’s proteome is synthesized in the ER, which constitutes the first compartment of the secretory pathway (English, Zurek, & Voeltz, 2009; Gorlach, Klappa, & Kietzmann, 2006; Levine & Rabouille, 2005). ER folding activity is mainly regulated by specific ER chaperones that require the highly specialized environment of the ER lumen (Gorlach et al., 2006; Stevens & Argon, 1999).

In resting conditions, the steady-state ER Ca²⁺ concentration fluctuates between 0.1 and 1 mM (Verkhratsky & Toescu, 2003). This ER luminal homeostasis is maintained by specialized enzymes like the sarco/ER Ca²⁺ ATPases (SERCA), which help maintain high Ca²⁺ level in the ER lumen by shuttling cytosolic Ca²⁺ into the ER at the expense of ATP (Gorlach et al., 2006).

Many intrinsic and extrinsic cues can cause disturbances in the folding machinery of the ER and trigger a condition known as ER stress, which is hallmarked by the incapacity of the ER to cope with the increased folding burden (Ron & Walter, 2007). The ER has evolved an intricate system that can sense changes in ER homeostasis and take steps to resolve them. Collectively known as the unfolded protein response (UPR), the three main ER stress sensor proteins are the pancreatic ER kinase-like ER kinase (PERK), a major eukaryotic initiation factor 2 α (eIF2α) kinase, the inositol-requiring enzyme 1 (IRE1), the most conserved UPR sensor with bifunctional endoribonuclease and kinase activity, and the activating transcription factor 6 (ATF6; Ron & Walter, 2007). These three ER transmembrane proteins each harbor an ER luminal domain and a cytosolic effector domain. The luminal domain acts as the canonical sensor of ER stress through the inhibitory-binding activity of the ER chaperone glucose-regulated protein 78 (Grp78). Upon the accumulation of unfolded proteins in the ER lumen, Grp78 is titrated away from its binding to the ER stress sensors, allowing their dimerization and consequent activation. When activated, each arm of the UPR will initiate a signaling cascade that initially aims to relieve folding stress on the ER by causing a translational stop while simultaneously trying to expand the ER folding capacity by the transcriptional upregulation of key ER chaperones and lipid synthesis. When this response proves inadequate, the UPR signaling switches from a prosurvival to a prodeath role (Hetz, 2012; Jager, Bertrand, Gorman, Vandenabeele, & Samali, 2012). The main outlets for pro-death signaling are PERK-mediated upregulation of the activating transcription factor 4 (ATF4), the transcription factor CCAAT/enhancer binding protein homologous protein (CHOP), and finally, the scaffolding and ribonuclease (RNase) role of IRE1. CHOP is an important transcription factor in UPR-mediated ER stress signaling and is a main effector of mitochondrial cell death through the downregulation of the antiapoptotic B-cell lymphoma 2 (Bcl2) and upregulation of the Bcl‐2 homologous 3‐only (BH3-only) proapoptotic Bcl-2-like protein 11 (also known as BIM) (Kim, Xu, & Reed, 2008). In a second mechanism of cell death induction, increased levels of ATF4 and CHOP induce protein translation through the dephosphorylation of P-eIF2α, mediated by the downstream target growth arrest and DNA damage-inducible protein 34, leading to ATP depletion and an increase in reactive oxygen species levels that kill the cell (Han et al., 2013). CHOP also works in tandem with p38 kinase to induce cell death (Maytin, Ubeda, Lin, & Habener, 2001). IRE1 is mainly a prosurvival UPR mediator through X-box-binding protein 1 (XBP1) cleavage. However, the RNase activity of IRE1 has been recognized to be able to cleave other messenger RNA (mRNA) sequences through a mechanism named regulated IRE1-dependent decay of mRNA (RIDD). While cleavage of XBP1 is generally cytoprotective, IRE1-associated RIDD activity mostly enacts apoptosis by degrading mRNAs encoding growth promoting factors or by the cleavage of microRNA’s that repress the translation of proapoptotic proteins (Maurel, Chevet, Tavernier, & Gerlo, 2014).

The canonical UPR is traditionally described as being mainly a transcriptional program, as both the prosurvival and the prodeath signaling pathways are regulated through XBP1, ATF6, and ATF4/CHOP transcription factors, downstream of the main ER stress sensors (for comprehensive reviews, see Hetz, 2012; Ron & Walter, 2007; Schroder & Kaufman, 2005). Although these are the main outlets of UPR signaling, they may not encompass the full breadth of the mechanisms restoring ER homeostasis or inducing cell death. An intriguing and largely unexplored mechanism regulating ER homeostasis may involve the dynamic rearrangement of ER membrane contact sites with other organelles, most notably the plasma membrane (PM) and mitochondria. In the next sections, we discuss some relevant studies supporting this exciting possibility.

Dynamic Reshaping of the ER: From Homeostasis to ER Stress

The ER is the largest intracellular organelle made up of a single continuous membrane (Westrate, Lee, Prinz, & Voeltz, 2015). Due to its dynamic and plastic nature, the ER is ideally equipped to act as a signaling hub at the center of a tightly regulated network. This network encompasses other subcellular compartments and organelles in the cell, and the ER is able to rearrange its shape according to the cell’s needs (Voeltz, Rolls, & Rapoport, 2002).

The ER can be broadly divided into the nuclear ER, or rough ER (RER), and the smooth ER (SER), defined by the presence or absence of ribosomes, respectively (Park & Blackstone, 2010; Voeltz et al., 2002). The RER has a flat shape and large lumen to accommodate the binding of ribosomes and the translation of large amounts of cellular proteins. Conversely, the SER has a rounded, tubular shape, and is largely devoid of ribosomes. The SER is the primary site of lipid biosynthesis and, in specialized cells (i.e., muscle cells), a primary location of Ca²⁺ signaling. Interestingly, ER architecture seems to be determined and molded by a variety of integral membrane proteins and interactions with other organelles and the cytoskeleton (Phillips & Voeltz, 2016). In particular, the characteristic curvature of ER tubules is maintained by wedge-shaped proteins specifically localized to the ER tubules and to the edges of the ER sheets (Voeltz, Prinz, Shibata, Rist, & Rapoport, 2006). These proteins, such as the reticulons (Di Sano, Bernardoni, & Piacentini, 2012) and DP1/REEP5/Yop1 (Hu et al., 2008), are “morphogenic” proteins which assess and stabilize the highly curved tubules. The complex ER network, extending through a typical reticular appearance throughout the entire cytoplasm, is due to the formation of three-way junctions where by sliding along the microtubules, adjacent ER regions contact and fuse through mechanisms that are regulated by the Atlastin family of Guanosine-5'-triphosphatases (GTPases) (Friedman, Webster, Mastronarde, Verhey, & Voeltz, 2010; S. Wang, Tukachinsky, Romano, & Rapoport, 2016). To guarantee homeostatic regulation, the ER needs to maintain close communication with other key organelles. To this end, the ER forms several nonhomotypic sites of close apposition with almost all membrane-bound organelles including the PM, mitochondria, endosomes, lysosomes, lipid droplets, Golgi, and peroxisomes (Phillips & Voeltz, 2016). These contact sites constitute dynamic subdomains that permit efficient nonvesicular exchange of small metabolites, such as lipids and Ca²⁺, thus allowing the ER to rapidly sense and coordinate cellular responses to extrinsic as well as intrinsic cues (Stefan, Manford, & Emr, 2013; van Vliet, Garg, & Agostinis, 2016). In line with this, in response to an increased protein folding demand, the ER undergoes rapid and extensive remodeling hallmarked by an expansion of its lumen and an increase in dynamic tubules (Schuck, Prinz, Thorn, Voss, & Walter, 2009; Sriburi et al., 2007).

The exact role of the UPR in this ER reshaping, however, is still unclear. It has been shown that different arms of the UPR are involved in regulating ER morphology upon ER stress. The signaling cascade initiated by the IRE1-XBP1 pathway is able to induce ER expansion by increasing lipid synthesis and thus the supply of membrane lipids. However, such a response can also be modulated by the ATF6 pathway in an XBP1-independent way (Bommiasamy et al., 2009). Interestingly, membrane expansion itself was found to alleviate ER stress independent of the total amount of chaperones (Schuck et al., 2009). ER size is thus important in regulating ER homeostasis, allowing it to tolerate more misfolded proteins and to promote more folding. Changing the ratio between ER sheets and tubules had no effect on the cell’s ability to overcome ER stress (Schuck et al., 2009).

Further studies revealed that upon ER stress, the ER does not only undergo changes in its size but is also able to rearrange its contacts with other organelles in order to potentiate the communication between subcellular compartments so as to ameliorate ER stress (Bravo et al., 2011; Feske et al., 2006; Liou et al., 2005; Prakriya et al., 2006; Roos et al., 2005; van Vliet et al., 2017; Vig et al., 2006).

In the next sections, we will illustrate known and emerging roles of ER-PM and ER–mitochondria contact sites. We will then discuss how key molecular elements, such as tethering proteins and signaling mediators occurring at contact sites, interface with and shape the output of the UPR machinery during ER stress.

ER–Plasma Membrane Contact Sites

It is becoming increasingly evident that ER-PM juxtapositions have broad physiological roles and are able to regulate PM lipid-mediated signaling, migration, and exocytosis, among other processes (Daniele & Schiaffino, 2014). Recent research has uncovered a multitude of specialized proteins mediating the different roles of ER-PM contacts. In yeast, three families of ER-anchored proteins assist in the formation of ER-PM contacts by interacting with the PM through their cytoplasmic lipid-binding and protein-binding domains; the vesicle-associated membrane proteins Scs2/22 (vesicle-associated membrane protein-associated protein (VAP) in mammalian cells), Ist2 (related to the mammalian transmembrane protein 16 (TMEM16) ion channels), and the tricalbin proteins Tcb1/2/3 (orthologs of the extended synaptotagmin-like proteins 1/2/3 (E-Syt1/2/3) in mammals) (Stefan et al., 2013). In mammalian cells, the nature and function of ER-PM contacts remained elusive until research in recent years led to the discovery of key proteins and pathways regulating the dynamic interface between the ER and the PM.

ER-PM Contacts and Ca²⁺ Homeostasis

The best characterized function of ER-PM junctions is the regulation of Ca²⁺ homeostasis and store-operated Ca²⁺ entry (SOCE). As mentioned earlier, the ER requires high levels of luminal Ca²⁺ in order to properly fold proteins. This elevated ER-Ca²⁺ level is maintained by the active import of cytosolic Ca²⁺ by the SERCA ATPases. Intraluminal Ca²⁺ transport is constantly offset by the leakage of Ca²⁺ from the ER. In mammalian cells, the primary causes of leakage are the inositol 1,4,5-trisphosphate (IP₃) receptor (IP₃R) and the translocon (Flourakis et al., 2006; Van Coppenolle et al., 2004). A well-established pathway for this is through IP₃-triggered ER Ca²⁺ release, which is triggered by agonist-mediated activation of phospholipase C (Berridge, 2016). To sense alterations in Ca²⁺ levels, the ER harbors a subset of specialized proteins. The ER transmembrane protein stromal interaction molecule 1 (STIM1) is one such protein and is able to sense luminal Ca²⁺ levels through its EF hand domain and will form high-order oligomers when intraluminal Ca²⁺ levels drop (S. L. Zhang et al., 2005). This event causes STIM1 to translocate to the PM where it interacts with Ca²⁺ release-activated channel proteins, of which the Ca²⁺ release-activated calcium modulator 1 (ORAI1) is the most prominent. Opening of the ORAI1 channel causes extracellular Ca²⁺ influx through SOCE (Varnai, Hunyady, & Balla, 2009; S. L. Zhang et al., 2005). The exact mechanism of STIM1 redistribution is still unclear but has been shown to go hand in hand with large-scale reorganization of the ER (Orci et al., 2009). This process is associated with morphological changes in the ER and the expansion of cortical ER (cER), which is composed of specialized thin ER structures that are devoid of chaperones but are highly enriched in STIM1 and make close contacts with the PM (8–11 nm), thus coupling formation of cER to store refilling (Orci et al., 2009; Shen, Frieden, & Demaurex, 2011; S. L. Zhang et al., 2005). However, the exact nature of these ER-PM microdomains is not fully resolved. An important observation was that SERCA has been found to be enriched at sites where STIM1 forms puncta and makes contact with the PM, which would enable Ca²⁺ entering the cell to be taken up by the ER almost simultaneously. This mechanism further highlights the relevance of ER to PM communication through contact sites in order to maintain physiological ER luminal Ca²⁺ levels (Manjarres, Rodriguez-Garcia, Alonso, & Garcia-Sancho, 2010).

ER-PM Contacts in Lipid Homeostasis

Over the past years, a lot of work has gone into exploring the role of ER-PM contacts in lipid trafficking and signaling. The ER has long been identified as a major site of phospholipid synthesis (Stone & Vance, 2000). Phosphatidylserine synthases 1 (PSS1) and 2 (PSS2) are localized to the ER, as is the phosphatidylethanolamine N-methyltransferase, which converts phosphatidylethanolamine (PE) to phosphatidylcholine (PC). Establishing contact sites with other organelles, such as the PM, is thus a prime way for the ER to exchange lipids required for membrane homeostasis (Stefan et al., 2013). Highlighting the link between ER contact sites and cellular homeostasis, studies mostly performed in yeast have shown that proper organelle contact and lipid transport are required to prevent ER stress. Deletion of all six ER-PM tethering proteins (ist2, scs2/22, and tcb1/2/3) in yeast (a strain termed Δtether) causes a substantial reduction in cER in contact with the PM. This observation was associated with collapsed ER structures in the cytoplasm (Manford, Stefan, Yuan, Macgurn, & Emr, 2012). Interestingly, these Δtether cells had increased ER stress and a constitutively activated UPR, indicating that ER-PM contact sites were required for normal ER homeostasis. Another study in yeast has uncovered that upon ER stress, the ER protein Nvj2p (lipid-binding ER protein, enriched at nucleus-vacuolar junctions) promotes ER-Golgi contact sites and may transport ceramide from the ER to the Golgi, to be processed there into inositolphosphorylceramide. Nvj2p is able to form a contact between the ER and medial-Golgi directly and contains a synaptotagmin-like, mitochondrial-lipid-binding protein (SMP) domain through which it may directly bind ceramide. This pathway prevents the toxic buildup of ceramides at the ER (L. K. Liu, Choudhary, Toulmay, & Prinz, 2017). Recently, a specialized set of proteins was identified that countertransport lipids between the ER and the PM (Chung et al., 2015; Moser von Filseck et al., 2015; Stefan et al., 2011). Termed oxysterol-binding protein-related protein (ORP, Osh proteins in yeast), these ORPs are able to shuttle phosphatidylserine (PS), synthesized at the ER, to the PM, while at the same time, transporting phosphatidylinositol 4-phosphate from the PM back to the ER. In a similar vein, a new class of lipid transporting proteins is now being characterized, the E-Syts (Perez-Lara & Jahn, 2015). E-Syts (1 and 2) are ER localized proteins tail-anchored through their N-terminus. They consist of an SMP domain and various Ca²⁺ sensitive C2 domains (five for E-Syt1 and three for E-Syt2). The SMP domain of E-Syt1, which belongs to the tubular-lipid-binding superfamily (Kopec, Alva, & Lupas, 2010, 2011; Wong & Levine, 2017), may transport diacylglycerol (DAG) from the PM to the ER. E-Syt1 was also shown to be able to transfer glycerolipids between bilayers in vitro (Saheki et al., 2016; Yu et al., 2016). E-Syts are able to form ER-PM contact sites through their C2 domains through an interaction with phosphatidylinositol 4,5-bisphosphate (Giordano et al., 2013). In the case of E-Syt1, the interaction with phosphatidylinositol 4,5-bisphosphate was dependent on the elevation of cytosolic Ca²⁺, which relieves the autoinhibition on its lipid-binding C2E domain (Bian, Saheki, & De Camilli, 2018). A recent finding also implicated E-Syt1-driven ER-PM contact sites in the formation of the autophagosome, the double membrane vesicles that deliver intracellular components to lysosomes for degradation, which is a hallmark of macroautophagy (for an extensive review on autophagy, see Lamb, Yoshimori, & Tooze, 2013). E-Syt1 contributes to the regulation and formation of phosphatidyl-inositol-3-phosphate, an essential lipid for autophagosome formation (Nascimbeni et al., 2018). E-Syt2 and E-Syt3 were also found to interact with the vacuole membrane protein 1 and beclin-1, two important proteins linking the phosphatidylinositol-3 kinase (PI3K) Complex III to the ER at ER-PM contact sites (Nascimbeni et al., 2018). This also makes sense of the finding that nascent autophagosomes receive crucial molecular precursors from both compartments, the ER and the PM (Ravikumar, Moreau, Jahreiss, Puri, & Rubinsztein, 2010).

Although all the signaling elements of the process of lipid transport at the ER-PM contacts and the resulting implications are still being uncovered, these observations expose the potential broad scope of ER-PM cross talk in interorganellar lipid transfer.

ER-PM Contacts in ER Stress and UPR

Only a few studies have hinted at a direct link between SOCE and UPR in terms of ER stress relief. Of the three UPR signaling proteins, only PERK has been linked to SOCE in previous research. One of the earliest studies on the relation between PERK and Ca²⁺ signaling found that genetic ablation of PERK in secretory and muscle cells caused a reduced efflux of Ca²⁺ through the IP₃R due to a disturbance in the Ca²⁺ signaling complex formed by PM Ca²⁺ ATPase-IP₃R-SERCA (Huang et al., 2006). A later study found that ablation of PERK caused a defect in SOCE after ER Ca²⁺ depletion (Verfaillie et al., 2012). This observation was accompanied by a reduction in phospholipase C-mediated IP₃R Ca²⁺ release in PERK^−/− cells; a condition not observed when cells were first permeabilized before adding exogenous IP₃ ligand directly (Verfaillie et al., 2012). This led to the hypothesis that there could be a disturbance in the communication between the ER and PM mediated by PERK (Verfaillie et al., 2012). Subsequent studies further showed that in rodent pancreatic β-cells PERK regulates cellular Ca²⁺ dynamics, including SOCE, through the downstream action of calcineurin (CN), a Ca²⁺ sensitive phosphatase shown previously to interact with PERK (Bollo et al., 2010; R. Wang et al., 2013). Beyond a role in SOCE, this study also uncovered a potential link between PERK and SERCA, indicating that PERK may play a role in ER refilling after store depletion. Treating cells with a pharmacological PERK inhibitor (PkI; Axten et al., 2012) caused decreased ER Ca²⁺ refilling upon store depletion and an increased inhibitory interaction between SERCA and calnexin (CNX). A potential link between PERK and SERCA had already been reported, as it was shown that CN, besides interacting and activating PERK, dephosphorylates CNX, leading to SERCA activation and increased ER Ca²⁺ refilling (Bollo et al., 2010). Altogether these studies uncovered a complex loop between SERCA activity and PERK-regulated SOCE, mediated by CN (Bollo et al., 2010). The precise molecular mechanism through which PERK mediates SOCE, and if this was caused by the previously observed phenotype of a potential disturbance in ER-PM distance (Verfaillie et al., 2012), remained though largely unknown.

Research performed in our lab has linked PERK mechanistically with ER-PM regulation through the F-Actin cytoskeleton (van Vliet et al., 2017). PERK is able to interact with filamin-A (FLNA) to regulate the F-actin network directly beneath the PM to efficiently coordinate the relocalization of STIM1 and E-Syt1. Interestingly, this newly found role of PERK was independent of its luminal domain sensing the accumulation of unfolded proteins during the UPR. Instead, elevation of cytosolic Ca²⁺ after ER luminal Ca²⁺ depletion was found to mediate PERK dimerization (van Vliet et al., 2017). This mechanism could neatly fit together with previous studies and help tie together SOCE and UPR signaling. Following Ca²⁺-induced dimerization, PERK interaction with FLNA and remodeling of the actin cytoskeleton were found to drive efficient ER-PM contact site formation and SOCE (van Vliet et al., 2017). The extracellular Ca²⁺ influx would then be taken up through SERCA, potentially present at the site of STIM1-ORAI1 contacts. Importantly, SERCA activity could be increased through the action of PERK and CN, described in previous studies, a hypothesis that needs urgent validation (Bollo et al., 2010; R. Wang et al., 2013). This mechanism could replenish ER stores in order to ensure sufficient Ca²⁺ levels for ER chaperones to function properly and alleviate the unfolded protein burden of the ER. Sensing rapid changes in cytosolic Ca²⁺ levels—well before the activation of the UPR—and ensuring proper ER Ca²⁺ levels through the modulation of ER-PM contact sites could thus be an additional role of PERK in a broader response to ER stress.

This Ca²⁺ modulated functional link between PERK, known to have a tethering role at the mitochondria-associated membranes (MAMs) (Verfaillie et al., 2012), and FLNA, a major actin-binding protein, is rather intriguing given the crucial role of the actin cytoskeleton in regulating ER morphology and mitochondria trafficking and constriction/fission mechanics. Recent studies show that the tubular ER is associated with mitochondrial fission and that formation of constriction sites occurs at the MAMs (Friedman et al., 2011). The ER-localized protein inverted formin 2 enables the polymerization of actin filaments at fission foci, thereby supplying the force required for mitochondrial constriction, and allowing the recruitment of the mitochondria fission machinery at the constriction sites (Korobova, Ramabhadran, & Higgs, 2013). Thus, upon ER-Ca²⁺ depletion and subsequent elevation in cytosolic Ca²⁺, the strengthening of the interaction of PERK with FLNA (van Vliet et al., 2017) may not only propagate ER-PM contact formation but also generate the actin cytoskeleton remodeling required to form mechanical constriction sites. This could potentially regulate the mitochondrial fusion/fission balance, an interesting possibility, with yet to be defined functional implications in ER stress (more information on ER stress and ER–mitochondria contact sites in the “ER–Mitochondria Contact Sites” and “ER–Mitochondria Contact Sites: A Signaling Platform Harnessed by the UPR” sections).

Notably, IRE1 and ATF6 have so far not been linked to SOCE or Ca²⁺ homeostasis in general. In the case of IRE1 especially, there is some scope to hypothesize a potential function in regulating SOCE, since IRE1 has been linked with both the microtubular as with the F-actin cytoskeleton (Ishiwata-Kimata, Yamamoto, Takizawa, Kohno, & Kimata, 2013) However, whether IRE1 may elicit ER-PM appositions through STIM1-mediated tethering and a mechanism involving remodeling of the actin cytoskeleton (Grigoriev et al., 2008) has not been explored yet.

Beyond STIM1, PERK-FLNA-mediated F-actin remodeling also favors E-Syt1 translocation to the PM, something that was largely abolished in cells lacking PERK and rescued by reexpressing either full length PERK or its kinase-dead variant in PERK^−/− cells (van Vliet et al., 2017). While these observations highlight a general scaffolding role of the PERK-FLNA axis in ER-PM contacts formation, this event could also link lipid homeostasis at the ER-PM contact sites with the UPR and could implicate this pathway in ER stress. Lipid homeostasis as such has strong links with ER stress, since UPR signaling will aim to increase lipid synthesis with the main purpose of favoring ER expansion, prevalently through the IRE1 arm (Sriburi et al., 2007). Thus, a pathway at the ER-PM contact sites shuttling lipids may contribute to a positive outcome after ER stress. A previous study has shown that PERK possesses an intrinsic lipid kinase function (Bobrovnikova-Marjon et al., 2012). Using DAG as a substrate, PERK was found to be able to generate an important signaling lipid, phosphatidic acid (PA), which correlates with the phosphorylation of Akt on Ser473. Furthermore, the phosphatidylinositol-3 kinase class IA p85α subunit interacts with PERK and enhances its lipid kinase function (Bobrovnikova-Marjon et al., 2012). These results are exciting when placed next to the notion that E-Syt1 was shown to shuttle DAG from the PM to the ER in a pathway that may constitute a loop, activated after ER Ca²⁺ depletion or early ER stress. PERK activation would then promote the relocalization of E-Syt1 to the PM, where E-Syt1 shuttles lipids (including DAG) to the ER to be used as substrates or signaling molecules.

This is an intriguing hypothesis which, if proven true, could connect a key sensor of the UPR machinery to the spatiotemporal regulation of lipid and Ca²⁺ signaling through the rapid formation of ER-PM contact sites.

ER–Mitochondria Contact Sites

Beside ER-PM juxtapositions, the ER is able to make contact with the mitochondrial network, which is regulated in a spatial and temporal manner in response to various inputs. Contacts between the ER and mitochondria are home to a host of important cellular processes, including, but not limited to, mitochondrial fission, bioenergetics, cell death, and inflammation, and have therefore been a topic of intense study in recent times (van Vliet, Verfaillie, & Agostinis, 2014). ER–mitochondria contact sites, also referred to as MAMs, were the first discovered interorganelle contact site (Vance, 1990; van Vliet et al., 2014). Further studies via electron tomography established that the precise distance of interaction between ER and the mitochondria is around 10 nm across the SER and 25 nm across the RER (Csordás et al., 2006), and the percentage of the mitochondrial surface physically in contact with the ER is estimated to be 5% to 20% of the total (Rizzuto et al., 1998).

Early work unveiled the crucial role of MAMs in lipid synthesis and transfer (Vance, 1990, 1991, 2014) and highlighted a novel, nonvesicular mechanism for the exchange of lipids at the ER–mitochondria interface (Vance, 2014). Later studies on the functional role of the MAMs uncovered that this dynamic platform constitutes specific Ca²⁺-enriched microdomains (Rizzuto et al., 1998).

ER–Mitochondria Contact Sites in Ca²⁺ Transfer

Despite possessing a Ca²⁺ uptake system in the form of the mitochondrial calcium uniporter, mitochondrial calcium uniporter has a very low affinity for Ca²⁺ (Kd ∼ 15–20 M). However, this seeming paradox is explained by IP₃-dependent stimuli, which cause mitochondria to be exposed to a much higher Ca²⁺ concentration than the bulk cytosol, allowing Ca²⁺ import. This finding suggested that mitochondrial Ca²⁺ uptake highly relies on close ER–mitochondria proximity, highlighting the importance of the ER architecture and the mitochondrial distribution in the regulation of Ca²⁺ signaling (Marchi et al., 2018; Rizzuto et al., 1998).

Several Ca²⁺ handling proteins are enriched at MAMs. Chief among them are the IP₃Rs, with the IP₃R3 isoform particularly targeted to MAMs (Mendes et al., 2005). Interestingly, IP₃Rs allow Ca²⁺ transfer from the ER to the mitochondria through voltage-dependent anion channel 1 (VDAC1), and IP₃R-VDAC1 interaction has been shown to require the MAM-associated Grp75 (Szabadkai et al., 2006). This Ca²⁺ transferring axis at MAMs is important, and many studies have shown that a host of other proteins enriched at MAMs could regulate Ca²⁺ transfer within this platform either directly, by modulating IP₃R/VDAC1 stability or interaction, or indirectly, by modulating ER–mitochondria tethering itself. An important direct regulator of IP₃R Ca²⁺ transfer is the sigma 1 receptor (Sigma1R; Hayashi & Su, 2007). Indirect regulators of ER–mitochondrial Ca²⁺ transfer include the ER transmembrane protein VAP and the mitochondrial localized protein tyrosine phosphatase-interacting protein 51 (PTPIP51). Another indirect regulator is mitofusin 2 (MFN2), a protein found both on the mitochondria and on the ER. Both these interactions are able to modulate ER–mitochondria proximity and thus promote IP₃R3-VDAC1-mediated Ca²⁺ cross talk (Gomez-Suaga et al., 2017). The exact role of MFN2 in ER–mitochondria tethering, however, still needs to be fully understood (de Brito & Scorrano, 2008; Filadi et al., 2015; Schrepfer & Scorrano, 2016).

The ER Ca²⁺ pump SERCA is regulated by several MAM-enriched proteins, indicating that MAMs also constitute a site of regulation of Ca²⁺ transport into the ER (Vandecaetsbeek, Vangheluwe, Raeymaekers, Wuytack, & Vanoevelen, 2011). Specifically, the SERCA2b isoform is particularly enriched at the ER–mitochondria contacts and is a target of CNX and thioredoxin-related transmembrane protein 1, among others. The latter two proteins function as a positive and negative regulator of SERCA2b activity, respectively, further evidencing MAMs as a hub of fine-tuned regulatory mechanisms of Ca²⁺ fluxes between ER and mitochondria (Lynes et al., 2012).

ER–Mitochondria Contact Sites in Phospholipid Trafficking

As mentioned earlier, ER–mitochondria interactions were first described as a location of lipid trafficking and synthesis (Vance, 1990). The ER being the most important site of lipid production in the cell, while mitochondria, although able to synthesize some of their required lipids, mostly rely on phospholipids synthesized in the ER (Futerman, 2006; Van Meer, Voelker, & Feigenson, 2008). Since their initial characterization, subsequent studies have indicated that the MAMs are sites of not only nonvesicular lipid trafficking but also phospholipid synthesis themselves (Holthuis & Levine, 2005). For instance, PS is synthesized at the MAMs from phosphatidylcholine or PE in a reaction catalyzed by PSS1 and PSS2 (Kuchler, Daum, & Paltauf, 1986; Vance, 1990). Once PS is transferred to mitochondria, it is decarboxylated by the PS decarboxylase in order to form PE. Interestingly, it was shown that the inhibition of PS decarboxylase causes a strong buildup of PS at MAMs, thereby supporting the active role of this platform in mediating the transfer of phospholipids such as PS (Ardail, Lerme, & Louisot, 1991).

The lipid composition of the mitochondria is fundamental for mitochondrial homeostasis and could affect mitochondrial dynamics and morphology (Mårtensson, Doan, & Becker, 2017). More specifically, two phospholipids have been highly studied: cardiolipin (CL) and its precursor PA. Curiously, while mitochondria are able to produce PA, the bulk of PA used for the synthesis of CL is produced in the ER before being transferred via the MAMs (Osman, Voelker, & Langer, 2011; Potting et al., 2013). Functionally, CL could facilitate the binding of the dynamin-related protein 1 (Drp1) to the mitochondria. Altering CL levels may thus be a mechanism to regulate or fine-tune mitochondrial fission (Bustillo-Zabalbeitia et al., 2014; Macdonald et al., 2014; Montessuit et al., 2010). In addition, increased levels of PA can support MFN1/2-mediated mitochondrial fusion, thus creating a potential regulatory mechanism for fusion and fission (Choi et al., 2006).

The exact role of phospholipids in mitochondrial metabolism is not well understood, but certain studies have indicated that CL and PE are master regulators of mitochondrial respiratory capability, as suggested by the evidence that both the absence of CL and PE impairs the activity of the respiratory chain (Böttinger et al., 2012; Pfeiffer et al., 2003; Wenz et al., 2009). In particular, it has been shown that upon increased respiratory demand, the mitochondrial membrane becomes enriched with CL, which has been shown to be fundamental for the stability of the mitochondrial Complex III and Complex IV as well as for the oligomerization state of the ADP/ATP carrier (Claypool, Oktay, Boontheung, Loo, & Koehler, 2008; M. Zhang, Mileykovskaya, & Dowhan, 2002, 2005).

In addition, the lack of CL affects the biogenesis of Fe-S, thereby impairing not only the stability but also the activity of the mitochondria supercomplexes, in particular Complexes II and III (Patil, Fox, Gohil, Winge, & Greenberg, 2013). Furthermore, the absence of PE was shown to bind and affect the activity of the mitochondrial complexes such as Complex IV (Otsuru et al., 2013; Tasseva et al., 2013). Combined, this evidence unravels the crucial role of CL and PE in mitochondrial respiratory metabolism.

Notably, the ER-PM localized lipid transfer proteins ORP5/8 have been found to also localize to ER–mitochondria contact sites, where they interact with the mitochondrial outer membrane PTPIP51 and regulate mitochondrial respiration (Galmes et al., 2016), thus uncovering a potential novel mechanism regulating lipid transport between the ER and mitochondria. This also reinforces the idea that PS transfer at the MAMs is important to maintain mitochondrial metabolism.

Taken together, these findings shed light on the importance of ER–mitochondria contacts as proteinaceous channels responsible for lipid transfer between these two organelles. However, more research is required to fully appreciate the impact of MAM-regulated phospholipid trafficking in mitochondrial bioenergetics.

ER–Mitochondria Contact Sites: A Signaling Platform Harnessed by the UPR

As stated above, upon ER stress, one of the first modifications the ER undergoes is a change in morphology. To restore ER proteostasis, the adaptive phase of the UPR will initiate the expansion of the ER lumen as well as the plasticity and dynamics of ER tubules, which are ideally suited to establish contacts with various organelles (Sriburi et al., 2007; van Vliet & Agostinis, 2017).

Given that mitochondria are the primary source of energy production in the cell, these organelles will be tightly involved in ER homeostasis, since protein folding is one of the most energetically demanding cellular processes and ER homeostasis relies on a substantial supply of ATP (Gorlach et al., 2006; Hoseki, Ushioda, & Nagata, 2009). Contact between ER and mitochondria has been shown to increase the level of mitochondrial Ca²⁺ uptake (Rizzuto et al., 1998), which in turn stimulates the consumption of mitochondrial oxygen and the production of ATP (Jouaville, Pinton, Bastianutto, Rutter, & Rizzuto, 1999). In line with this, after ER stress, mitochondria relocate to the perinuclear region to aid the formation of ER–mitochondria contacts or tighten existing ones (Bravo et al., 2011; Csordás et al., 2006).

Other studies support the strong link between mitochondrial bioenergetics, MAMs and the adaptive response to ER stress (Carreras-Sureda, Pihan, & Hetz, 2017; Sassano, van Vliet, & Agostinis, 2017). Interestingly, MAMs are enriched for several ER chaperones and ER stress sensors, thus suggesting that ER–mitochondria contacts contribute to shape the cellular responses evoked by ER stress.

ER stress caused by glucose deprivation stimulates an increase in ATP levels (Y. Liu, Liu, Song, & Zuo, 2005). However, when nutritional stress becomes too severe cells will undergo apoptosis. Interestingly, this cell death could be prevented by the overexpression of the MAM-enriched chaperone Grp75 (Yang et al., 2008), which regulates ER–mitochondria tethering and Ca²⁺ transfer by acting as a bridge between IP₃Rs and VDAC1 (Szabadkai et al., 2006).

An increasing number of ER chaperones, including CNX, Grp78 (Bip), Calreticulin and Sigma1R, have been shown to be present at MAMs, suggesting their possible roles in regulating Ca²⁺ signaling between the ER and mitochondria (Hayashi & Su, 2007). In line with this, CNX, which is recruited to the MAMs upon palmitoylation (Lynes et al., 2012), is a positive regulator of SERCA2b (Roderick, Lechleiter, & Camacho, 2000), a SERCA isoform strongly enriched at the MAM, and is thus able to modulate ER Ca²⁺ levels. In addition, Sigma1R, whose expression is upregulated upon ER stress, serves as a Ca²⁺ sensor and after depletion of ER, Ca²⁺ dissociates from Grp78 to interact with IP₃R3 at the MAMs. This interaction promotes the stability of IP₃R3 and its proper functioning during ATP-mediated repetitive physiological stimuli, ensuring a prosurvival mitochondrial Ca²⁺ flux (Hayashi & Su, 2007).

Taken together, ER–mitochondria contact sites form strategic chaperone-enriched platforms. At these ER subdomains, chaperones control ER proteostasis upon store depletion and indirectly fine-tune Ca²⁺ signaling from the ER to mitochondria, thus regulating either prosurvival or proapoptotic signals.

ER–Mitochondria Contacts and ER Stress-Mediated Cell Death

Chronic, unresolved ER stress promotes UPR-mediated proapoptotic signals leading to cell death (Deniaud et al., 2008; Ferreiro et al., 2008). During ER stress, the PERK-eIF2α-ATF4 branch of the UPR can induce the expression of a truncated version of SERCA1 (S1T; Chami et al., 2008). This S1T mutant is a splice variant characterized by exon 11 splicing, which has only one of the seven Ca²⁺-binding residues and is, therefore, unable to pump Ca²⁺ into the ER (Chami et al., 2001). Interestingly, Chami et al. (2018) uncovered that S1T is highly enriched at MAMs and increases ER–mitochondria contact sites by inhibiting mitochondrial movement, which leads to mitochondrial Ca²⁺ overload, triggering apoptosis. In addition, another indirect mechanism through which the PERK-ATF4 pathway is able to modulate the ER–mitochondria interface is through the upregulation of the E3 ubiquitin ligase Parkin, which prevents ER stress-induced mitochondrial fragmentation and cell death (Bouman et al., 2011). Furthermore, by increasing ER–mitochondria contact sites Parkin can also promote limited Ca²⁺ transfer from the ER to mitochondria, thus enhancing prosurvival IP₃R stimulation and ATP production (Calì, Ottolini, Negro, & Brini, 2013). Other studies have defined a role of PERK in protecting cells from ER stress-induced cell death through its interaction with MFN2 at the MAMs (Muñoz et al., 2013). MFN2^−/− cells display a sustained activation of the PERK pathway, which causes defective mitochondrial morphology and a reduced sensitivity to ER stress-mediated apoptosis, suggesting that MFN2 is able to regulate PERK activation through protein interaction.

Taken together, these studies disclose that the transcriptional program activated by the PERK arm of the UPR regulates crucial cellular responses (cell survival vs. cell death) by modulating the molecular nature and plasticity of ER–mitochondria contact sites. However, as mentioned previously, PERK has also been shown to localize at the ER–mitochondria contact sites and has a tethering function there (Verfaillie et al., 2012). Under conditions of loss of SERCA function and consequent ER-Ca²⁺ depletion caused by oxidative damage, the cytosolic domain of PERK promotes mitochondrial apoptosis by facilitating the rapid transfer of reactive oxygen species (possibly also in the form of lipid peroxides) by preserving the close apposition between the ER and mitochondria (Verfaillie et al., 2012).

Considering that PERK also modulates ER-PM contact sites, as discussed earlier (van Vliet et al., 2017), this UPR mediator seems to serve as an apical sensor of perturbations of ER-Ca²⁺ homeostasis. Depending on the intensity and type of ER stressor, PERK might help to recover ER function or precipitate cell death through its dual function as coordinator of interorganellar contact sites and master regulator of the UPR. How the recruitment of PERK at the MAMs, or other subdomains of the ER membrane, is dynamically modulated and what are the interacting partners or molecular modulators favoring this partitioning are still open questions that need to be urgently solved.

Very little is known about the role of the other kinase of the UPR, that is, IRE1, at the MAMs. IRE1 has been found to have a cytoprotective role at the ER–mitochondria contacts through UPR-related mechanisms. In line with this, the major downstream mediator of the IRE1 arm, XBP1, was found to reduce the sensitivity toward apoptosis due to its interaction with Sigma1R, which promotes survival in response to ER stress (Mori, Hayashi, Hayashi, & Su, 2013). In line with this, after ER-Ca²⁺ store depletion induced by the SERCA pump inhibitor thapsigargin, Sigma1R deficient cells display a reduction in XBP1 splicing which results in enhanced apoptotic cell death (Mori et al., 2013).

However, whether IRE1 could play a role at the ER–mitochondria platform independently of the classical UPR pathway still remains to be investigated.

Finally, the link between lipid transfer at the ER–mitochondria interface and ER-stress-mediated cell death is still largely obscure. An interesting study showed that the buildup of Monosialotetrahexosylganglioside (GM1-ganglioside) at the ER–mitochondria contact sites and, more specifically, the glycosphingolipid-enriched microdomain fraction of ER–mitochondria contact sites led to hyperstimulation of IP₃R-mediated ER Ca²⁺ depletion. This led to mitochondrial Ca²⁺ overload and ER stress-induced mitochondrial apoptosis due to mitochondrial membrane permeabilization (Sano et al., 2009). This suggests that altering the transfer of MAMs-associated lipids to the mitochondria may sensitize them to undergo ER stress-mediated cell death, a conjecture that needs further investigation.

Concluding Remarks

The study of ER contact sites has resulted in many new discoveries in recent years. From Ca²⁺ signaling to lipid transfer, new and important cellular functions regulated by contact sites are progressively arising. An important aspect of this is the identification of the molecular nature of the tethering proteins and the mechanisms allowing their dynamic compartmentalization at the contact sites in mammalian cells. This is providing us with new insights into a possible role in terms of ER homeostasis or signaling. The UPR is often seen as being primarily a transcriptional response to ER stress, with important roles for the RNase activity of IRE1 and the kinase activity of PERK. Other results have pointed at a broader role of UPR signaling in ER stress. Studies performed in the last few years are indicating that the rearrangement in ER membrane contact sites as a dynamic response to ER stress could also constitute an ER stress response, in many cases, directly mediated by the UPR proteins themselves.

As discussed here, the concept that certain ER-associated tethers are not restricted to one contact site but are able to connect and establish contact sites between the ER and other organelles is emerging (Figure 1). This is particularly relevant for Ca²⁺ and UPR signaling, since PERK has a role in both ER–mitochondria and ER-PM contact sites. This does raise several questions about the mechanisms of partitioning allowing these same subsets of proteins to connect the ER to multiple organelles, for example, PM and mitochondria. Beyond this, more studies should explore the cross talk between ER stress and interorganellar contacts in cell nonautonomous processes that are increasingly known to be regulated by the UPR and membrane contact sites including, but not limited to, inflammation and immunity (Bassoy et al., 2017; Rufo, Garg, & Agostinis, 2017). Future research will help answer these outstanding questions and allow us to propose and explore new paradigms to decipher how interorganellar coordination shapes homeostasis and ER stress.

Figure 1.

Schematic representation of the intersection between ER–plasma membrane (PM) and ER–mitochondria contacts with responses elicited by ER stress and the unfolded response (UPR). Upon accumulation of unfolded proteins and consequent ER stress, the unfolded protein response (PERK, IRE1, and ATF6) will activate their major transcriptional program in the form of ATF4, spliced XBP1 (sXBP1), and cleaved ATF6 (cATF6). ER stress will also lead to a rearrangement of the ER structure and the formation of contact sites with mitochondria (mitochondria-associated membrane, MAM) and plasma membrane (ER-PM). Especially, the UPR kinase PERK plays a role in both these events. At the MAM, by aiding tethering and through an interaction with MFN2, and at the ER-PM contact site, by interacting with filamin A (FLNA) and inducing the formation of ER-PM contact sites through F-actin remodeling. Some of the major roles of each contact site are depicted here: At the ER-PM contact site, lipids are transported through E-Syt1 and ORP5/8, and extracellular Ca²⁺ is transported through store-operated Ca²⁺ entry (SOCE). At the MAM, lipids are synthesized through the action of phosphatidylserine synthase (PSS1/2) and potentially transported through ORP5/8. The interaction between IP₃R and VDAC1, mediated by Grp75 (not shown here), constitutes the major Ca²⁺ signaling pathway from ER to mitochondria, while SERCA represents the main pathway for Ca²⁺ uptake into the ER. One of the major tethering complexes at the ER–mitochondria contact site, vesicle-associated membrane protein-associated protein–protein tyrosine phosphatase-interacting protein 51 (VAP-PTPIP51) is also depicted here. ER = endoplasmic reticulum; PM = plasma membrane; ORP = oxysterol-binding protein-related protein; SERCA = sarco/endoplasmic reticulum Ca²⁺ ATPase; STIM1 = stromal interaction molecule 1; ORAI1 = the Ca²⁺ release-activated calcium modulator 1; SOCE = store-operated Ca²⁺ entry; PTPIP51 = protein tyrosine phosphatase-interacting protein 51; VDAC1 = voltage-dependent anion channel 1; IP₃R = inositol 1, 4, 5-trisphosphate receptor; VAP = vesicle-associated membrane protein-associated protein; MFN = mitofusin; PERK, pancreatic ER kinase-like ER kinase; PS = phosphatidylserine; PSS1 = phosphatidylserine synthase 1; MAM = mitochondria-associated membrane; FLNA = filamin-A; CNX = calnexin; Grp78 = glucose-regulated protein 78; IRE1 = inositol-requiring enzyme 1; ATF6 = activating transcription factor 6; XBP1 = X-box-binding protein 1; eIF2α = eukaryotic initiation factor 2 α.