The E-Syts represent tethers that intrinsically span the gap between membranes, potentially forming an intermembrane lipophilic channel (Schauder et al., 2014). The mammalian E-Syts, plant SYT proteins, and the homologous yeast Tricalbins generally contain three regions: (i) an N-terminal hairpin (or, for Arabidopsis SYTs, a type I transmembrane domain (Yamazaki et al., 2010), which inserts into the cytoplasmic leaflet of the ER; (ii) single or multiple SMP (synaptotagmin-like mitochondrial lipid-binding protein) domains that share the physical attributes of TULIP domains containing deep hydrophobic channels capable of lipid binding; and (iii) a variable number of C-terminal C2 domains, which directly bind the opposing PM (Craxton, 2004; Groer et al., 2008; Kopec et al., 2010; Toulmay and Prinz, 2012; Giordano et al., 2013; Pérez-Sancho et al., 2016; Reinisch and De Camilli, 2016). C2 domain interactions with membranes are often potentiated by cytosolic Ca²⁺ and/or facilitated by the presence of phospholipids, such as PI(4,5)P₂ (Rizo and Südhof, 1998; Chang et al., 2013; Giordano et al., 2013; Lee et al., 2019).

E-Syts are implicated in the possible Ca²⁺-dependent non-vesicular transfer of neutral glycophospholipids, such as DAG, between membranes and two different structural models have been proposed to explain this process (Saheki et al., 2016; Yu et al., 2016). The “tunnel model” proposes that E-Syts facilitate direct lipid transfer between membranes by forming a hydrophobic tunnel between membranes (Schauder et al., 2014). In this model, the E-Syt SMP domains dimerize to form a long singular channel that binds lipids. Although this model is supported by the observed perpendicular orientation of E-Syts at contact sites between the ER and PM, the length of the SMP dimer tunnel (∼9 nm) is less than the observed distances between the ER and PM (∼19–24 nm; Fernández-Busnadiego et al., 2015; Collado et al., 2019); seemingly too short to span between the ER and PM. However, a presumed obligatory intermembrane distance is a somewhat contrived concept, and only relevant if the lipid tunnel model does not faithfully represent the in vivo conformation of E-Syts (Saheki and De Camilli, 2017a; West et al., 2011; Schauder et al., 2014; Quon and Beh, 2015; Reinisch and De Camilli, 2016; Saheki and De Camilli, 2017a; Saheki and De Camilli, 2017b; Quon and Beh, 2015; Schauder et al., 2014; Reinisch and De Camilli, 2016). Alternatively, in the “shuttle model,” E-Syts act as membrane-anchored lipid transfer proteins that bind and move lipids between the two closely aligned ER and PM bilayers (Schauder et al., 2014; Bian et al., 2019). However, these two structural options might confer different functions and are not necessarily mutually exclusive (Collado et al., 2019).

In mammalian cells, E-Syt2 and E-Syt3 contain an SMP domain followed by three C2 domains whereas E-Syt1 contains five C2 domains (Giordano et al., 2013; Reinisch and De Camilli, 2016; Saheki and De Camilli, 2017b). Previous work has shown that E-Syt2 and E-Syt3 form ER-PM contacts independent of Ca²⁺ while E-Syt1 is recruited to the contact sites when Ca²⁺ levels rise (Giordano et al., 2013; Fernández-Busnadiego et al., 2015; Idevall-Hagren et al., 2015; Yu et al., 2016). Hence, ER-PM junctions are dynamically regulated by Ca²⁺ signaling. In fact, increased Ca²⁺ levels not only promote E-Syt1-mediated ER-PM association in mammalian cells, but also shortens the distance between the ER and the PM and activates SMP-dependent lipid transfer (Fernández-Busnadiego et al., 2015; Saheki et al., 2016; Bian et al., 2018). At the “resting state,” a small pool of E-Syt1 resides at ER-PM contact sites in the absence of Ca²⁺ and the distance between the ER and the PM at these sites is larger than E-Syt2 or E-Syt3 associated ER-PM contact sites (Fernández-Busnadiego et al., 2015; Pérez-Lara and Jahn, 2015). This difference in distance is likely due to different chain length of cytosolic segments of E-Syt1 and E-Syt2/3, given that they contain different numbers of C2 domain units. E-Syt1 C2E domain is highly similar to C2C E-Syt2/3 domain, which enables constitutive interaction to PM in the normal conditions (Fernández-Busnadiego et al., 2015). As Ca²⁺ levels increase, the E-Syt1 C2C domains interact with the PM to lessen the ER-PM gap (Fernández-Busnadiego et al., 2015).

The Ca²⁺ dependency of the yeast Tricalbin E-Syt homologs and plant SYTs is less clear. Recruitment of the Tricalbins and SYTs is Ca²⁺ independent, but Ca²⁺ does play a role both in increasing SYT phospholipid binding affinity (Schapire et al., 2008), and shortening Tcb3p-mediated ER-PM MCS distances (Hoffmann et al., 2019). At sites of Tcb3p-mediated cER-PM contact, high cytosolic Ca²⁺ shortened the distance between ER and PM by ∼2 nm (Hoffmann et al., 2019). This observation might indicate that the shortening of the SMP-C2 bridge at these sites resulted in more C2 proteins coating the PM than in wild-type cells. A denser protein coat across the PM was observed under these conditions. An intermediate density is also associated with E-Syt ER-PM contact sites (Fernández-Busnadiego et al., 2015).

The VAPs represent a different mechanism in membrane tethering. Given that they do not contain any domain capable of making contact in trans with an opposing membrane, VAPs necessarily form multi-subunit bridges. The domain architecture of eukaryotic VAPs, including yeast Scs2p/22p, consists of three major regions: (i) a MSP domain, which includes the binding site for proteins containing the FFAT (“two phenylalanines in an acidic tract”) motif (Loewen et al., 2003); (ii) a single C-Terminal ER transmembrane domain; (iii) an intermediate variable linker region, which is ∼27 nm in VAP-A and Scs2p (Loewen et al., 2003, 2007; Loewen and Levine, 2005; Murphy and Levine, 2016). Scs2p and Scs22p have both been shown to bind phosphoinositides, such as PI4P, through their N-terminal MSP domain which extends out toward to PM (Kagiwada and Hashimoto, 2007). Considering that the linker region length between the transmembrane domain and the MSP domain in Scs2p could conceivably span the two membranes, it is tempting to postulate that Scs2p directly associates the two membranes. However, Scs2p or Scs22p have not been demonstrated to directly interact with the PM. Furthermore, mutations in the MSP domain that disrupt FFAT motif binding inhibit Scs2p-dependent ER-PM membrane association, suggesting that FFAT protein-MSP domain interactions are required for contact (Manford et al., 2012).

In terms of promoting ER-PM MCSs, tether proteins are not functionally equivalent. In yeast, the elimination of Scs2p (the VAP-B homolog) and its paralog Scs22p (most homologous to VAP-A) results in a ∼50% reduction in ER-PM association within yeast cells, which is unmatched in its severity when compared to the removal of any other tether protein (Loewen et al., 2003, 2007; Manford et al., 2012). The many functional and physical interactions involving VAPs and Scs2p (and likely Scs22p also) suggest they act as scaffolding proteins to assemble additional subunits at MCSs to physically connect membranes (Murphy and Levine, 2016). Unlike Scs2p, however, the function of Scs22p is more ambiguous. Within Scs2p, a linker region of ∼20 nm separates the MSP and transmembrane domains, which contributes to the overall distance Scs2p can span between the ER and PM (Murphy and Levine, 2016). Within Scs22p, this linker region is all but absent (Loewen and Levine, 2005), though it is unclear if this segment is even relevant to a multi-subunit intermembrane bridge. It should be noted that Scs22p and Scs2p are not equivalent as tethers, though there appears to be a minor functional overlap (Craven and Petes, 2001; Loewen and Levine, 2005). Scs2p and Scs22p might share specific interactions with some lipid transfer proteins (Riekhof et al., 2014; Weber-Boyvat et al., 2015). However, if Scs2p and Scs22p have shared roles in membrane tethering, it is not manifested in any observed differences in cER-PM association when comparing scs2Δ and scs2Δ scs22Δ cells (Loewen et al., 2007; Manford et al., 2012). Although SCS22 is routinely deleted along with SCS2 to remove all yeast VAP activity, the specific in vivo functions of Scs22p deserve additional attention.

Arabidopsis VAP27 proteins represent plant homologs of both mammalian VAPs and yeast Scs2p (Sutter et al., 2006; Saravanan et al., 2009). The 10 identified VAP27 proteins in A. thaliana (VAP27-1 to VAP27-10) can be grouped into 3 clades (Wang et al., 2016). Members of two of the clades have a single TM domain that determines their localization to ER, whereas members of the third lack the TM domain, and are localized to the PM (Wang et al., 2016). The best studied of these plant proteins is VAP27-1, which consists of a C-terminal ER-spanning transmembrane domain, a coiled-coil domain, and the MSP domain (Wang et al., 2014, 2016). Although VAP27-1 can bind phosphoinositides in vitro, VAP27-1 likely forms an ER-PM bridge in a protein complex with NET3C, which belongs to the plant-specific NET proteins that link actin filaments to different endomembranes within the cell (Deeks et al., 2012; Wang et al., 2014, 2016; Stefano et al., 2018). As a multi-subunit tether complex, VAP27-1 and NET3C form ER-PM MCSs, in which NET3C also binds actin filaments and VAP27-1 might bind microtubules (Wang et al., 2014; Siao et al., 2016; Stefano et al., 2018). However, VAP27-1/NET3C MCSs are not affected by cytoskeletal disruption, even though vap27-1 vap27-3 double mutant plants exhibit a disrupted ER and actin morphology, which is linked to growth aberrant establishment of cell polarization (Wang et al., 2016; Stefano et al., 2018).

The structural architecture of Anoctamin/TMEM16/Ist2p homologs is unique among membrane proteins (Brunner et al., 2016). They have multiple transmembrane helices forming dimers embedded within membranes. Between subunits, a cavity forms at dimeric interfaces and it is potentially filled with lipids. In the PM or ER, this dimer cavity provides a possible passage route for lipids from one leaflet of the bilayer to the other. This lipid scramblase activity is one major function ascribed to Anoctamin/TMEM16/Ist2p homologs, but some act as Cl^– ion transporters, while others have one or none of the activities (Brunner et al., 2016; Whitlock and Hartzell, 2017). A select number of these proteins are found at ER-PM MCSs.

The yeast TMEM16 homolog, Ist2p is a primary tether that has extended sequences long enough to bridge the gap between the ER and PM (Maass et al., 2009; Kralt et al., 2015). Ist2p consists of an ER transmembrane region that shares homology with the Anoctamin (ANO/TMEM16) protein family, a C-terminal cortical sorting sequence, and a cytosolic inter-membrane linker. The cortical sorting sequence is a lysine-rich domain of low complexity that mediates direct PM binding by interacting with PI(4,5)P₂ (Ercan et al., 2009; Fischer et al., 2009; Maass et al., 2009; Wolf et al., 2012). Ist2p plays a major role at PM-ER MCSs because the deletion of IST2 alone causes a significant reduction in membrane association (Manford et al., 2012; Wolf et al., 2012). Indeed, in scs2Δ scs22Δ cells, the additional deletion of IST2 has a greater impact on cER-PM association than deleting all three Tricalbins (Manford et al., 2012). Unlike the Tricalbins and Scs2/22p, the domain of Ist2p that extends from the ER to the PM has little defined structure, but it serves the simple requirement of being long enough to bridge the inter-membrane distance, but it may have another specific function.

The bulk of Ist2p homology to ANO/TMEM16 proteins is located within its multi-pass transmembrane domain region (Ercan et al., 2009; Hartzell et al., 2009). Of the ten human ANO proteins (ANO1-ANO10), Ist2p shares the greatest similarity to ANO10 (TMEM16K) (Kunzelmann et al., 2016). Most of the human ANO proteins, such as ANO1/2 (TMEM16a/b), localize to the PM and act as Ca²⁺-activated Cl^– ion channels. Other ANO homologs, such as human ANO6 (TMEM16f) or A. fumigatus and N. haematococca TMEM16, appear to have phospholipid scramblase activities (Caputo et al., 2008; Schroeder et al., 2008; Yang et al., 2008; Hartzell et al., 2009; Malvezzi et al., 2013; Brunner et al., 2014; Yu et al., 2015). However, when directly assayed Ist2p does not seem to have lipid scramblase activity (Malvezzi et al., 2013). Because Ist2p shares a similar Ca²⁺-dependent sorting pathway to the cER with the mammalian ER Ca²⁺ sensor stromal interaction molecule 1 (STIM1), it has been proposed that Ist2p might be a functional equivalent of STIM1 for yeast regulation of cytoplasmic Ca²⁺ (Kunzelmann et al., 2016; Wanitchakool et al., 2017). If Ist2p or other ER-PM tethers play a role in yeast Ca²⁺ regulation, one might predict that eliminating these proteins would affect Ca²⁺ homeostasis. However, Saccharomyces cerevisiae lacks SERCA (sarco/endoplasmic reticulum Ca²⁺-ATPase)-family Ca²⁺ transporters and both the yeast Golgi and vacuole are the major Ca²⁺ storage organelles, not the ER as in metazoans (Strayle et al., 1999; Cunningham, 2011). Thus, the budding yeast ER and ER-PM MCSs seemingly play a lesser role in affecting intracellular Ca²⁺.

In terms of basic protein structure, Ist2p and all yeast tethers are ER integral membrane proteins (Figure 1A). However, only Ist2p and the Tricalbins are predicted to directly span between cER and PM to make contact. In contrast, Scs2p/22p lack obvious PM interaction domains, and it is unclear if the yeast-specific tether Ice2p (see below) has cytoplasmic domains long enough to link the ER to the PM (Figure 1A). To provide membrane attachment, these proteins would seem to require additional binding subunits. Bridging between membranes through an extended protein complex is an established mode of membrane tethering. In the multi-subunit ER-mitochondria encounter structure (ERMES), ER-mitochondria contact requires several subunits (Lang et al., 2015). In a different context, the multi-subunit exocyst complex also tethers vesicle membranes to the PM during the last steps of exocytosis (Lepore et al., 2018). Whether a single intrinsic tether, or a multi-subunit bridge complex, the varied mechanisms of tethering defies simplistic definitions as involving only a single protein that spans between closely apposed membranes.

In yeast, the distance between cER and the PM is variable (∼16 to 59 nm; West et al., 2011), therefore no absolute domain length is required for a tether to reach between membranes. In fact, the transverse distance between the cER and PM appears to be highly dynamic and morphologically adaptable (Collado et al., 2019). In yeast, an artificially constructed ER-PM staple can functionally complement growth defects in cells lacking ER-PM tethers, despite being predicted to extend between membranes by only ∼10 nm (Quon et al., 2018). Therefore, estimated trans-membrane linker length does not necessarily reflect tethering function of which some can be conferred by non-specific ER-PM physical contact.

Analysis of ER-PM MCSs by cryo-electron tomography and correlative light and electron microscopy has shown that the association of different tether protein families with the PM involves distinct ER shapes (Fernández-Busnadiego et al., 2015; Saheki et al., 2016; Collado et al., 2019; Hoffmann et al., 2019). In wild-type cells, cER has regions that lay flat across the PM with other regions where the ER is tubular (West et al., 2011; Nixon-Abell et al., 2016). In cells lacking six primary tethers, expression of the yeast anoctamin/TMEM16 homolog Ist2p generates cER membrane that is tubular and sheet-like, similar to wild-type cER in width and distance from the PM (Collado et al., 2019). This result suggests that Ist2p plays an important role in shaping cER morphology (Collado et al., 2019). When Scs2p/Scs22p are expressed in cells lacking many ER-PM tethers, cER morphology is significantly altered and is mostly extended ER sheets that are narrow in width, though the distance of cER to the PM is unaffected (Collado et al., 2019). Scs2p/Scs22p appears important for maintaining ER width and does not determine cER-PM distance. The expression of only the yeast E-Syt homologs, the Tricalbins Tcb1p/Tcb2p/Tcb3p, in cells lacking other tethers resulted in cER with mainly tubular structure, and the ER membrane curves toward the PM (Collado et al., 2019; Hoffmann et al., 2019). This curvature brings ER in very close proximity to the PM, where the cER forms peaks of radius ∼10 nm at the base and height ∼7 nm.

Because ER tubulation is facilitated by reticulons, it is surprising that Tricalbin-generated ER curvature is reticulon-independent (Hoffmann et al., 2019). In the absence of reticulon-like proteins Yop1p and Rtn1p the ER loses its tubular structure and expands, but Tcb3p still forms highly curved tubules at ER-PM MCSs (Voeltz et al., 2006; West et al., 2011; Hoffmann et al., 2019). An in situ structural analysis of Tricalbins showed that rod-shaped structures, comprising SMP and C2 domains, connect the ER to the PM where Tcb3p seems to occupy a dense coat along the PM (Hoffmann et al., 2019). These results indicate that tethers not only link the ER and PM together, but in doing so they alter local membrane structure around contact sites.

Ice2p is an ER Type III transmembrane protein, which also contributes to ER-PM tethering (Quon et al., 2018). Ice2p facilitates the movement and inheritance of cER along the PM from yeast mother cells into daughter cells (Estrada de Martin et al., 2005; Loewen et al., 2007). Yeast cER inheritance involves myosin-directed movement of cER tubules along the PM into emerging daughter buds via actin cables positioned parallel to the mother-daughter axis (Du et al., 2001).

Apart from its role in cER inheritance, Ice2p is recognized for multiple events in inter-organelle contact (Estrada de Martin et al., 2005; Loewen et al., 2007; Toulmay and Prinz, 2011; Tavassoli et al., 2013; Markgraf et al., 2014; Rogers et al., 2014; Quon et al., 2018). Ice2p plays distinct roles in organelle associations depending on cellular growth phase. When cells exit from stationary phase, Ice2p facilitates a direct physical interaction between the ER and lipid storage droplets, where it regulates the flow of neutral lipids from lipid droplets into the ER (Markgraf et al., 2014). As cells re-enter exponential growth, Ice2p rapidly re-localizes to the cER at punctate sites along the PM (Estrada de Martin et al., 2005; Markgraf et al., 2014; Quon et al., 2018). The deletion of ICE2 in scs2Δ cells also has considerable impact on mitotic cell growth and PC synthesis (Loewen et al., 2007; Tavassoli et al., 2013). These results are consistent with Ice2p being physically located at ER-PM MCSs and ER-lipid droplet contacts, depending on growth state.

Based on possible secondary structure predictions, Ice2p likely contains no large cytoplasmic domain or defined lipid binding domains capable of bridging across membrane gaps (Markgraf et al., 2014). Nonetheless, the third cytoplasmic loop of Ice2p is predicted to have four amphipathic helices with unstructured linker regions that might in fact extend between membranes at particularly close distances. This third Ice2p loop appears to be particularly important for its linking to LDs, and when expressed alone the Ice2p cytoplasmic loop does bind LDs and to a lesser degree the ER (Markgraf et al., 2014). Whether this loop is also pertinent for ER-PM association remains to be seen, though its estimated length is just under the smallest observed distance (16 nm) between cER and the PM (West et al., 2011).

By deleting ICE2 in the strain lacking the six other tethers, Δ-s-tether cells were generated in which ER-PM association is further reduced to calculated levels of random/stochastic contact (Quon et al., 2018; Figures 1B−E). Although in Δ-s-tether cells ER-PM membrane contact is all but eliminated, these cells are viable and grow slowly, at least under standard growth conditions. The deletion of ICE2 not only reduces cER-PM association beyond that conferred by eliminating the other six tethers, but it also significantly exacerbates phospholipid regulatory defects (Quon et al., 2018). In Δ-s-tether cells, phospholipid levels (PE, PC, and PA) are substantially reduced as are sphingolipid levels, whereas DAG levels are elevated. The addition of the PC precursor choline rescues Δ-s-tether growth defects, suggesting that a major function of yeast ER-PM MCSs involves the regulation of phospholipid synthesis. Curiously, sterol levels are unaffected in Δ-s-tether cells, even though sterol depletion causes cER expansion to nearly encompass the entire PM (Quon et al., 2018). Perhaps the assembly of ER-PM MCSs is regulated by membrane structure changes or stresses, as triggered by sterol depletion, which also affects phospholipid composition.

At pre-established MCSs, secondary tethers and ancillary regulators might induce contact in response to specific stimuli or might fortify and expand existing membrane association. As opposed to primary tethers required for ER-PM contact under standard conditions, secondary tethers would be sufficient (but not necessary) to promote membrane association when overexpressed or induced (Quon and Beh, 2015). Ancillary regulators include those that affect ER structure and ER shape, which influences cER association with the PM and other membrane organelles. Examples of conserved and cell type-specific modulators of ER morphology include: the dynamin-like GTPase atlastin-1 (Sey1p in yeast), reticulons, the ER/Ca²⁺/synaptic vesicle regulator secernin-1, and “receptor expression-enhancing protein” (REEP; homologs of yeast reticulon-interacting protein Yop1) (Park et al., 2010; Beetz et al., 2013). Several of these proteins sculpt ER shape through interactions with cytoskeletal elements such as microtubules (Park et al., 2010; Lindhout et al., 2019).

In addition to cytoskeletal interactions, many ER-PM MCS effectors make direct contact with tether proteins. Around half of VAPs/Scs2p/Scs22p interactions involve binding to “FFAT-motif proteins,” including a subset of the family of ORPs homologs (Murphy and Levine, 2016). Despite their homology to the canonical mammalian “OSBP,” not all ORPs actually bind oxysterols or sterols (Pietrangelo and Ridgway, 2018). However, ORPs act as lipid transfer proteins in vitro and are implicated in both non-vesicular and vesicular transport pathways in vivo (Alfaro et al., 2011; de Saint-Jean et al., 2011; Beh et al., 2012; Mesmin et al., 2013; Chung et al., 2015; Moser von Filseck et al., 2015a, b; Kentala et al., 2016; Tong et al., 2016). Most ORPs in mammals are associated with membrane contact sites between organelles, which appears true for some if not all yeast ORPs homologs as well (Kentala et al., 2016; Pietrangelo and Ridgway, 2018). In yeast, Scs2p binds three ORP homologs (Osh1p-3p) via their FFAT motif, which is found within an N-terminal region aside a PI4P-binding PH domain (Roy and Levine, 2004; Stefan et al., 2011; Tong et al., 2013; Weber-Boyvat et al., 2015). All yeast Osh proteins are soluble but in addition to the cytoplasm, Osh2p and Osh3p can localize to the cell cortex and Osh1p is targeted to sites at the NVJ (Levine and Munro, 2001; Kvam and Goldfarb, 2004). Osh2p and Osh3p localization with cER at ER-PM MCSs is dependent on PI4P and is significantly reduced by repression of Stt4p, the PM PI4P kinase (Stefan et al., 2011). Scs2p/22p binding to FFAT motifs also contribute to Osh2p/3p cortical localization, as well as to Osh1p NVJ localization (Loewen et al., 2003). However, in scs2Δ scs22Δ mutant cells, Osh3p colocalization with cER and PM-ER MCSs is disrupted, though it remains associated with the PM (Stefan et al., 2011). These results suggested that Osh3p acts as a PM-binding adaptor subunit for a Scs2p tethering complex that bridges between the ER and PM (Stefan et al., 2011). However, OSH3 deletion has not been reported to disrupt ER-PM association and Scs2p binding is not required, though localization of Osh3 to the cortex is needed for its functionality (Stefan et al., 2011).

A recent paper reported transient interactions between Osh2p and nascent endocytic sites on the PM near associated “cER rims” (Encinar del Dedo et al., 2017). Through interactions with the type I myosin Myo5p and Scs2p, it is proposed that Osh2p regulates cER-endocytic site associations to facilitate actin patch assembly for endocytic internalization (Encinar del Dedo et al., 2017). These studies raise the possibility that Osh2p and Scs2p couple Myo5p-dependent actin polymerization with the clearance of cER from endocytic sites undergoing membrane invagination. Even in this case, however, there is no evidence that Osh2p provides a stable link for ER-PM contact. Given the lack of functional data, the role of Osh proteins and other FFAT motif proteins in promoting general MCS assembly is still an important but open question.

In yeast cells, ER-PM MCSs are also major determinants of phospholipid regulation. Akin to ORP5/8 in mammalian cells, yeast Osh6p/7p is recruited to ER-PM MCSs by the tether Ist2p, where Osh6p transfers PS in the ER to the PM for the reciprocal exchange of PM PI4P back to the ER (Maeda et al., 2013; Moser von Filseck et al., 2015a; D’Ambrosio et al., 2020). PE transported to the PM is converted to PC by the ER-localized Opi3p (Kodaki and Yamashita, 1987; McGraw and Henry, 1989). Another Osh protein, Osh3p, facilitates Opi3p recruitment to ER-PM MCSs where Opi3p acts in trans by reaching from the ER to generate PC in the PM (Figure 2A; Tavassoli et al., 2013; Pawlik et al., 2020). In Δ-s-tether cells lacking ER-PM MCSs, Opi3p cannot synthesize PC in trans and PC synthesis is disrupted (Quon et al., 2018). Supplying Δ-s-tether cells with supplemented choline improves cell growth likely by producing PC via the alternate Kennedy pathway (Quon et al., 2018).

In the PM, the PLD homolog Spo14p converts PC to PA but unlike mammalian cells, yeast do not have an established Nir2 homolog that transports PA back to the ER (Sreenivas et al., 1998; Xie et al., 1998). Instead, in the PM the lipid phosphatase App1p converts excess PA to DAG. In the ER, the PA phosphatase Pah1p uses PA as a substrate to produce DAG, which can be utilized for PE and PC synthesis by the Kennedy pathway (Figure 2A; Ganesan et al., 2015). At the same time, the lipid kinase Stt4p converts PI in the PM to PI4P, which is thereafter changed to PI(4,5)P₂ by the PI5 kinases, Mss4p (Carman and Han, 2011).

For its role in PI4P regulation, it was suggested that the ER-localized Sac1p phosphatase acts in trans to eliminate PI4P in the PM (Stefan et al., 2011). The deletion of SAC1 and the elimination of ER-PM tethers both aberrantly accumulate PI4P in the PM (Stefan et al., 2011; Quon et al., 2018). One possibility is that Osh3p and Scs2p recruit and activate Sac1p at MCSs in order to reach out to the PM from the ER (Stefan et al., 2011). In conjunction with ORP5-mediated transfer PI4P from the PM, an alternative model proposes that Sac1 in mammalian cells turns over PM PI4P delivered to the ER in cis (Zewe et al., 2018). In the same study, Sac1 was able to act in trans when fused with an additional 6−7.5 nm linker peptide between the catalytic domain and transmembrane domain of Sac1p, which the enabled Sac1 to bridge the distance between the ER to the PM to turn over PI4P (Zewe et al., 2018). Consistent with these findings, the deletion of SAC1 in Δ-s-tether yeast cells results in synthetic lethality, which suggests that SAC1 and ER-PM tethers act in parallel but functionally independent pathways (Quon et al., 2018). Moreover, when ER-PM contact is reintroduced in Δ-s-tether cells using an artificial ER-PM staple, the PM is still enriched in PI4P and does not re-establish the normal cellular distribution of PI4P, which is absent in the PM of mother cells (Quon et al., 2018). This would argue that forcing membrane contact between the ER and PM is not sufficient for Sac1p to act in trans, at least not effectively.

As with mammalian ORPs, the seven yeast ORPs Osh1p-Osh7p together also affect PI4P levels (Stefan et al., 2011). The Osh family of proteins share at least one overlapping essential function and eliminating all Osh proteins increases cellular PI4P by ∼20-fold over wild-type cells (Beh and Rine, 2004; Stefan et al., 2011). This accumulation is even greater than in sac1Δ cells where PI4P levels only increase ∼6 fold (Rivas et al., 1999). Give that more PI4P accumulates in Osh-depleted cells, the Osh protein family affects PI4P regulation beyond just Sac1p-dependent PI4P dephosphorylation. Osh proteins might also regulate other PI4P phosphatases or facilitate PI4P consumption and conversion into greater phosphorylated phosphoinositide species.

In Δ-s-tether cells, the lack of ER-PM MCSs leads to increased DAG levels, decreased levels of PA and PA-derived phospholipids, and decreased amounts of sphingolipids (Quon et al., 2018). This result suggests that DAG generation of PA is inhibited, and some aspect of PA biosynthesis is dependent on ER-PM MCSs. Because DAG is readily consumed in the synthesis of phospholipids and TAG, DAG generally represents a minor fraction of total membrane lipids (Ejsing et al., 2019). In growing yeast cells, however, DAG is enriched in the vacuolar membrane and a minor DAG pool resides in the ER membrane (Ganesan et al., 2019). The pool size and trafficking of DAG is also affected by other lipids including PS and also sphingolipid synthesis, which generates DAG (Figure 2A; Ganesan et al., 2019). In yeast strains defective in sphingolipid synthesis, DAG is primarily localized to the PM rather than the vacuolar membrane (Ganesan et al., 2019). As such, ER-PM MCSs might affect DAG pools by modulating PS trafficking or sphingolipid synthesis, and/or by repressing the normal regulation of DAG-PA interconversion. In general, this regulatory control over phospholipid synthesis and distribution impacts membrane structure and functional activities.

The original finding that PAM fractions are enriched in ER sterol biosynthetic enzymes, also suggests an intimate relationship between ER sterol synthesis and those areas of the ER associated with the PM (Pichler et al., 2001). Moreover, the inactivation of sterol synthesis in yeast induces ER-PM MCSs, resulting in the nearly complete coverage of the PM with cER (Quon et al., 2018). In addition to potential sterol exchange at ER-PM MCSs, additional mechanisms clearly contribute to sterol transport.

Through exocytosis, some sterol is sorted and transferred to the PM by vesicular transport (Surma et al., 2011). In the absence of vesicular transport, however, sterols are still delivered to the PM from their site of production in the ER (Kaplan and Simoni, 1985). A conserved aspect of this “non-vesicular” transfer mechanism is that it involves equilibrative exchange between the ER and PM (Kaplan and Simoni, 1985; Menon, 2018). Even though sterol transport is non-directional, sterols ultimately concentrate within the PM because its lipid composition is preferable for sterol inclusion (Baumann et al., 2005; Menon, 2018). As presumptive sites for ER-PM sterol exchange, ER-PM MCSs might reduce the intermembrane gap necessary for membrane-bound lipid transfer proteins to transport sterols. In yeast, the necessity of ER-PM MCSs for sterol exchange has been directly tested and, despite a defect in sterol internalization, mutant cells lacking the seven primary tethers still retain normal equilibrative sterol exchange between the two membranes (Quon et al., 2018).

Plasma membrane fractions from either wild-type or Δ-s-tether cells indicate that the subcellular distribution or ergosterol (the yeast equivalent of cholesterol) is unaffected by the absence of ER-PM contact sites (Quon et al., 2018). However, retrograde transport of sterols from the PM back to the ER is moderately defective in Δ-s-tether cells (Quon et al., 2018). Using transport-coupled esterification of exogenous sterols as an assay, retrograde transport is slowed ∼4-fold as the imported sterol moves from the PM to the ER and then lipid droplets, following esterification. Measuring sterol transport in the opposite direction, a pulse-chase analysis of de novo synthesized [³H]ergosterol indicated that the half time of sterol transfer from the ER to the PM is ∼10 min in both wild-type and Δ-s-tether yeast (Quon et al., 2018). These results suggest a limited role of ER-PM MCSs in equilibrative sterol exchange.

ER-PM MCSs and membrane-bound sterol transfer proteins might facilitate sterol exchange if they act in concert with other independent transport mechanisms. For instance, an overlapping mechanism for sterol delivery to the PM is vesicular transport. However, blocking exocytosis in △-s-tether cells has no significant effect on the transfer of sterols from the ER to PM (Quon et al., 2018). As another redundant exchange pathway, soluble sterol transfer proteins might act as a compensatory mechanism in the absence of sterol exchange at ER-PM MCSs.

In yeast cells, Osh4p is the most abundant yeast ORP that binds both sterols and PI4P (Im et al., 2005; de Saint-Jean et al., 2011). In liposome transfer assays, Osh4p was shown in vitro to exchange sterols and PI4P (de Saint-Jean et al., 2011; Moser von Filseck et al., 2015b). However, in vivo sterol transfer assays have not confirmed this activity in living cells, at least between the ER and PM (Georgiev et al., 2011; Quon et al., 2018). Osh4p associates with the Golgi and post-Golgi exocytic vesicles and its transfer activities are likely more pertinent in that cellular context (Fairn et al., 2007; Alfaro et al., 2011). In fact, the entire family of Osh proteins is dispensable for ER to PM sterol transfer, though specific Osh proteins affect phosphoinositide metabolism at ER-PM MCSs (Georgiev et al., 2011; Stefan et al., 2011). The role of Osh4p in phosphoinositide regulation, as opposed to sterol transfer, is the probable basis for the lethality of deleting OSH4 in Δ-s-tether cells (Quon et al., 2018). In sterol transfer assays using a conditional osh4 Δ-s-tether mutant, no compounding defect in ER-PM sterol exchange was detected (Quon et al., 2018). The fact that the absence of ER-PM contact sites has no impact on ER to PM sterol transfer is not due to compensatory transport by vesicular transport or Osh4p, which is the predominant sterol-binding protein in yeast. Although the equilibrative mechanism of sterol exchange within yeast is unclear, the import of exogeneous sterols is affected by ER-PM MCSs and associated sterol transfer proteins.

The StARkin family possess StART domains, or StART-like domains, which mediate lipid and sterol transfer (Gatta et al., 2018; Horenkamp et al., 2018; Jentsch et al., 2018; Tong et al., 2018). In budding yeast, there are six StART domain homologs that include Lam1p-4p, which are ER membrane proteins that can be found at various organelle contact sites (Gatta et al., 2015). In Lam proteins, within the StART-like domain, a hydrophobic cavity can accommodate a sterol ligand (Tong et al., 2018). Yeast cells lacking Lam2p (also known as Ysp2p) exhibit reduced esterification rates of imported sterol, indicating a role in exogenous sterol uptake (Roelants et al., 2018; Tong et al., 2018). Lam2p localizes to spots both in cytoplasmic ER and where cER and PM are associated, though distinct from where other functional tethers are observed (Gatta et al., 2015; Quon et al., 2018). In fission yeast, the Lam homolog Ltc1p is also localized to sites on perinuclear ER and the cell cortex, where it is involved in retrograde sterol transfer (Marek et al., 2020). At the cell cortex, Ltc1p facilitates an actin-dependent sterol transfer either directly from the PM to endosomes, or indirectly through the ER as an intermediate (Marek et al., 2020).

The mammalian homologs of the Lam/Ltc protein family are GRAMD1a, b, and c, GRAMD2a and b (also known as GRAM3). GRAMD1a, b, and c contain both StART-like sterol transfer domains and PH-like GRAM (Glucosyltransferases, Rab-like GTPase activators and Myotubularins) sterol-sensing domains, while GRAMD2a and b contain only GRAM domains (Horenkamp et al., 2018; Sandhu et al., 2018; Naito et al., 2019). At ER-PM contacts, GRAMD2a co-localizes with the E-Syt2 and E-Syt3 tethers (Besprozvannaya et al., 2018). In response to PM cholesterol accumulation, the N-terminal GRAM domain of GRAMD1b (also known as Aster-B protein) binds PS and cholesterol, which recruits GRAMD1b to ER-PM MCSs where it mediates retrograde cholesterol transport (Sandhu et al., 2018; Naito et al., 2019). PM cholesterol accumulation cannot be cleared when GRAMD1 proteins are eliminated, indicating that they facilitate cholesterol transport out of the PM and into the cell (Naito et al., 2019). As sites of lipid transfer or regulation, ER-PM MCSs affect the distribution and metabolism of both sterols and phospholipids, which directly impacts PM and ER function and morphology.

In Arabidopsis, knock-out mutations in SYT1 render cells hypersensitive to environmental conditions that trigger membrane stress such as mechanical, cold, and ionic damage (Schapire et al., 2008, 2009; Yamazaki et al., 2008; Pérez-Sancho et al., 2015, 2016). These phenotypes suggest that ER-PM MCSs function is required for structural support further buttressed by the ER and the cytoskeleton. This hypothesis is further supported by dynamic phosphoinositide-mediated changes in ER-PM MCSs observed in Arabidopsis in response to ionic stress and rare earth elements (Lee et al., 2019, 2020; Figure 3).

At the molecular level, the ionic stress response induces PI(4,5)P₂ accumulation, which promotes the electrostatic binding of the C-terminal C2 domains of ER-anchored SYT1 to the PM (Pérez-Sancho et al., 2015). This slow, but highly dynamic process effectively increases ER-PM association, which promotes a protective response that enables the PM to withstand ionic stress. Changes in phosphoinositide composition and ER-PM MCS assembly are specifically induced by stresses that inhibit PM Ca²⁺ channels. Perturbations of Ca²⁺ channel activity by rare earth elements trigger the PM accumulation of PI4P and results in a three- to four-fold increase in SYT1/SYT5 ER-PM MCSs (Lee et al., 2020). This mechanism demonstrates that the coordination of Ca²⁺ and phosphoinositide signals, together with ER-PM MCSs, sustains plant growth under suboptimal stress conditions (Bayer et al., 2017).

The stress-induced changes in phosphoinositides also highlight a role for ER-PM MCS as molecular platforms for other phosphoinositide-dependent processes. As first shown in mammalian cells, E-Syts ER-PM at MCSs become sites for PI3P synthesis where they promote autophagosome biogenesis (Nascimbeni et al., 2017). In plants, the presence of an autophagy-dependent pathway starting at ER-PM MCSs involves the interaction among the Arabidopsis AtEH/Pan1 (homologs of the yeast ARP2/3 complex activator, Pan1p), the actin cytoskeleton, and the ER-PM MCS resident proteins VAP27-1 and VAP2-3 (Stefano et al., 2018; Wang and Hussey, 2019; Wang et al., 2019). These studies feature an important role for plant and animal ER-PM MCS in the targeted removal and recycling of endocytic machinery as a cellular adaptive mechanism to nutrient deprivation.