Different tether proteins of the same membrane contact site affect the localization and mobility of each other

The different membranes that are found in eukaryotic cells establish proximity regions with each other known as membrane contact sites. These structures are formed by proteins or protein complexes that tether the membranes together, and they are not an intermediate step in the process of membrane fusion (Eisenberg-Bord et al., 2016). Contact sites play important roles in metabolism and organelle dynamics (Prinz et al., 2019). They can function as platforms that enable the direct exchange of molecules. Lipid-transfer proteins can mediate the exchange of membrane lipids at contact sites, by shielding them from the hydrophilic environment in hydrophobic cavities (Wong et al., 2019). Contact sites can also be involved in the exchange of luminal material, mediated by channels. One well-documented example is their role in Ca²⁺ signaling (Burgoyne et al., 2017). Additionally, the function of membrane contact sites can be mediated by the physical attachment of the membranes. For instance, attachment to the cell cortex controls the positioning of the mitochondrial network and its inheritance during asymmetric cell division (Pernice et al., 2018). Contact sites can also generate specialized niches within an organelle. For instance, sites of interaction with the endoplasmic reticulum (ER) determine the place of mitochondrial fission and fusion (Abrisch et al., 2020; Friedman et al., 2011). Systematic screens designed to uncover all contact sites in the cell have shown the existence of proximity regions between every pair of organelles assessed, indicating that these structures are a prominent way of communication between organelles (Kakimoto et al., 2018; Shai et al., 2018; Valm et al., 2017).

At the structural level, proteins or protein complexes that interact with both membranes establish the apposition, generating the contact site. These proteins are called tethers. Any protein that interacts with both membranes will contribute to their tethering and, for many contact sites, many tethering proteins have been described to co-exist. Membrane contact sites are a crowded molecular niche that involves multiple tethers and non-tethering proteins (Jamecna et al., 2019). Additionally, the region of the membrane involved in contact site formation might have differential properties, as when cells were brought to lower temperatures, lipid phase separation was observed to coincide with organelle contact sites (King et al., 2020).

Many contact sites are regulated by environmental conditions, and they change in extension or protein composition in response to specific cues (Bohnert, 2020). For example, the extension of the nuclear–vacuolar junction (NVJ) in Saccharomyces cerevisiae is regulated in response to glucose availability. This is mediated by changes in the protein levels of its tether Nvj1, through control of both its transcription and degradation (Hariri et al., 2017; Tosal-Castano et al., 2021). Tethering at some contact sites is regulated by Ca²⁺ levels. For instance, in the context of store-operated Ca²⁺ entry, a drop in ER luminal Ca²⁺ levels induces a contact site of this organelle with the plasma membrane, which activates plasma membrane Ca²⁺ channels (reviewed in Derler et al., 2016; Soboloff et al., 2012). In turn, tethering of the ER to the plasma membrane by the extended synaptotagmin E-Syt1 is regulated by cytoplasmic Ca²⁺, as E-Syt1 possesses a C2 domain that interacts with phosphatidylinositol 4,5-bisphosphate [PI(4,5)P₂] on the plasma membrane only in the presence of Ca²⁺ (Giordano et al., 2013). Finally, many changes in the contact site size or proteome involve post-translational modifications of the tethering proteins. For example, the Vps39 protein (encoded by VAM6) of the vacuole–mitochondria contact site in yeast is phosphorylated in response to glucose depletion, resulting in the disassembly of the contact site (Hönscher et al., 2014). The lipid-transfer protein CERT transports ceramide from the ER to the Golgi complex for its further processing into complex sphingolipids, such as sphingomyelin. CERT is also able to tether these two organelles and can be regulated by phosphorylation in response to the cellular requirements of sphingomyelin (reviewed in Kumagai and Hanada, 2019).

Membrane contact sites are a unique molecular microenvironment with two closely apposed membranes separated by a thin layer of cytosol, including numerous proteins with and without tethering capacity. This imposes particular spatial restrictions on the proteins, making it likely that different proteins largely influence each other. One possibility is that tethers that generate a ‘seed’ contact site facilitate the interaction of other tethers. Additionally, the establishment of a narrow contact might exclude large proteins and the crowding of the molecular space might hinder accessibility. In this work, we used an in vivo system in S. cerevisiae to uncover how proteins with the capacity to tether the same membranes in the cell influence each other, regarding localization, distribution between different complexes, and dynamics. As tethers can be regulated in response to stimuli, this has important implications to understand if other tethers are cross-regulated. Our results show that tethers can dictate the distribution of other tethering proteins between different complexes or localizations, restrict their localization to specific regions of an organelle, and change their mobility within the contact site region.

We set out to address how different proteins with the ability to tether the same membranes influence each other in vivo. To do this, we decided to build an artificial inducible system, which would allow us to address the effects mediated by the tethering of the membranes without the effect of specific protein–protein interactions. To achieve this, we used the human FK506-binding protein of 12 kDa (FKBP1A, hereafter referred to as FKBP) and the small FKBP–rapamycin-binding domain (FRB) of the human mTor1 protein (encoded by MTOR). These two polypeptides dimerize with high affinity in the presence of the drug rapamycin (Chen et al., 1995). All experiments involving this system were carried out in a strain that lacks the yeast rapamycin-binding protein Fpr1 and contains a mutant version of Tor1 that renders it insensitive to rapamycin (Haruki et al., 2008). This system has been used extensively to rapidly remove proteins from their site of action by attaching them to an aberrant location (Haruki et al., 2008). Here, we used the FKBP and FRB domains fused to the cytosolic side of transmembrane proteins of different organelles. The addition of rapamycin would then produce an artificial contact between the organelles instead of a change in localization (Fig. 1A). The FKBP–FRB system has been used before to favor the dimerization of two parts of split-biotin ligases directed to different organelles, to allow proximity labeling of proteins found in the contact site (Cho et al., 2020; Kwak et al., 2020).

The first artificial contact site that we established was between the vacuole and the mitochondria. To do this, we attached the FKBP domain to the vacuolar Zinc transporter Zrc1 (MacDiarmid et al., 2000) and the FRB domain together with GFP to the pre-protein receptor Tom70 (Brix et al., 2000), a transmembrane protein of the mitochondrial outer membrane. Tom70 normally localizes homogeneously to the mitochondrial outer membrane (Fig. 1B; Fig. S1A; −rapamycin). Upon addition of rapamycin, Tom70 re-localized, forming a concentrated patch in the proximity of the vacuole (Fig. 1B; Fig. S1A; +rapamycin), indicating the formation of an artificial contact site. The re-localization could already be observed within 10 min of rapamycin addition, and after 30 min, all the protein had re-localized to the contact site region (Fig. S1A). For all subsequent experiments, the time point of 30 min post addition of rapamycin was selected. Simultaneous imaging of the mitochondrial outer membrane protein Om45 showed that rapamycin addition caused the whole organelle to localize in the proximity of the vacuole, but only Tom70 was enriched in this region (Fig. 1B). The morphology of the mitochondrial network remained mostly unaffected. These results show that the rapamycin-induced dimerization of FKBP and FRB can be used to generate inducible artificial contact sites.

As we were able to induce artificial tethers between organelles, we assessed their effect on other tethers of the same interface. The protein Vps39 is a tether of the vacuole–mitochondria contact site, known as the vacuole and mitochondria patch (vCLAMP), as well as a subunit of the homotypic vacuole fusion and protein sorting (HOPS) complex, involved in vesicular fusion processes of the vacuole (Bröcker et al., 2012; Elbaz-Alon et al., 2014; Hönscher et al., 2014; Ostrowicz et al., 2010; Shvarev et al., 2022). Both functions of the protein are genetically separable (González Montoro et al., 2018). The interaction partners involved in these two functions compete for the same pool of Vps39 (González Montoro et al., 2021). At endogenous levels, Vps39 localized in puncta along the vacuole membrane, which were not preferentially localized close to the mitochondria (Fig. 1C,D, −rapamycin) and likely represent sites of HOPS function. When the protein is overexpressed, patches of enrichment at the vacuole–mitochondria interface are observed, which represent extended vCLAMPs (González Montoro et al., 2018; Hönscher et al., 2014). We thus tested whether inducing proximities between the vacuole and mitochondria by the artificial tethers affected the localization of Vps39 at endogenous levels. We observed that upon the addition of rapamycin, Vps39 accumulated next to the artificially induced contact sites (Fig. 1C, +rapamycin). Quantification of the localization of Vps39 accumulations relative to mitochondria showed that the generation of the artificial contact site shifted the dot-like accumulations of the protein from being mostly away from mitochondria to being mostly next to them (Fig. 1D). The drug did not affect the localization of the protein in the absence of the artificial contact sites (Fig. S1B,C).

We have recently described that the protein Cvm1 can act as a tether of the vacuole–mitochondria contact site, acting independently of Vps39 (Bisinski et al., 2022). Cvm1 localizes mostly at a different contact site at endogenous levels, but enriches strongly at the vacuole–mitochondria interface upon overexpression (Bisinski et al., 2022). We thus tested whether the enlarged vacuole–mitochondria interfaces caused by overexpression of Cvm1 would also result in a shift in the localization of Vps39. We observed that, similar to the contact sites formed by the artificial inducible system, overexpression of Cvm1 caused a re-localization of Vps39. Under these conditions, Vps39 accumulated directly next to Cvm1-induced contact sites and thus close to the mitochondria (Fig. 1E,F).

To confirm that dots that re-localize in these systems do not represent the recruitment of endosomes to mitochondria, we analyzed the localization of Vps8 puncta, a late endosome marker (Day et al., 2018), relative to mitochondria before and after induction of the artificial contact sites, and with or without overexpression of Cvm1. The localization of Vps8 relative to mitochondria was not affected by these treatments, unlike that of Vps39 (Fig. S2A–D).

To analyze whether this effect depended on the ability of Vps39 to tether the vacuole to the mitochondria, we made use of a mutant version of Vps39, in which its ability to act as a tether is impaired by mutation of the presumed binding interface with the TOM complex as well as regulatory sites (Vps39^12xM; González Montoro et al., 2018). Vps39^12xM was still affected by the overexpression of Cvm1, or the generation of the artificial tether, but to a lesser extent than Vps39 wild type (wt) (Fig. 1G,H; Fig. S2E,F). A small effect is still expected in the absence of specific recruitment, as these treatments will increase the tethering between the organelles, resulting in a larger surface of the vacuole being in contact with the mitochondria. Thus, a vacuolar protein will be more likely to localize close to the mitochondria by chance. Comparison of the results with Vps39wt and Vps39^12xM by a two-way ANOVA showed a significant interaction between the protein used and the rapamycin treatment or the overexpression of Cvm1 (P<0.05 in both cases), indicating that these treatments affect the two proteins in a statistically significantly different way. The increased effect seen for Vps39wt would represent the specific effect of tethering beyond what would be expected by the increased proximity between the organelles.

Taken together, these results indicate that the generation of a region of membrane proximity by a tether can cause the recruitment of other proteins with the capacity to tether the same membranes.

The second system that we used was the generation of an artificial ER–vacuole contact site. Vacuoles form a well-described contact site with the nuclear ER, called the nuclear–vacuolar junction (NVJ). This is one of the most extended contact sites in yeast under normal growth conditions (Collado et al., 2019). The main tether of this contact site is formed by the interaction of the vacuolar protein Vac8 with Nvj1, a transmembrane protein of the ER containing a luminal domain that interacts with the nuclear membrane, restricting the protein to the nuclear ER (Millen et al., 2008; Pan et al., 2000). Several other proteins with tethering capacity reside in this contact site, including the lipid transfer proteins Lam6 (or Ltc1) and Osh1 (or Swh1), which contain no structural features that should restrict them to a particular domain of the ER (Elbaz-Alon et al., 2015; Levine and Munro, 2001; Murley et al., 2015). However, no contact site between the peripheral ER and the vacuole has been described. We hypothesize that the restriction of Nvj1 to the nuclear ER is the cause of the restriction of other ER–vacuole tethering proteins to this region of the organelle.

To test this hypothesis, we sought to create an inducible artificial peripheral ER–vacuole contact site. Such a contact site would be spatially separated from the NVJ, because they involve different regions of the ER. Thus, it would allow us to assess the recruitment of the different tethering proteins to one or the other contact. To do this, we used as an anchor the protein Rtn1, an ER protein that is excluded from the nuclear region, i.e. exclusively localized to the peripheral ER (Voeltz et al., 2006). Dimerization of this protein with a vacuolar transmembrane protein using the FRB–FKBP system should create specific peripheral ER–vacuole contact sites (Fig. 2A). In the condition without rapamycin, Rtn1 was only found in the peripheral ER, whereas it was excluded from the nuclear ER, visible with the general ER marker Sec63 (Fig. 2B). Drug-induced dimerization of Rtn1 with the vacuolar transporter Zrc1 produced contacts between peripheral ER tubules and the vacuole, as Rtn1 was then observed to surround this organelle (Fig. 2B; Fig. S3A; +rapamycin). Sec63 was also observed to then surround the vacuole, indicating that ER tubules come into contact with the vacuole and not just the protein Rtn1 (Fig. 2B, +rapamycin). Thus, this system allows the induction of peripheral ER–vacuole contact sites.

Next, we assessed the effect of induction of this contact site on proteins that normally reside in the NVJ and have a tethering capacity: the lipid transfer proteins Osh1 and Lam6. Under normal growth conditions, Lam6 is mostly localized at ER–mitochondria contact sites (Elbaz-Alon et al., 2015; Murley et al., 2015). This was also observable in our strains, with Lam6 appearing as dot-like structures away from the vacuole (Fig. 2C, −rapamycin). When Lam6 was enriched next to the vacuole, this structure was negative for Rtn1, indicating that it corresponds to the nuclear ER, i.e. the NVJ (Fig. 2C, white arrowheads). Upon addition of rapamycin, Lam6 formed patches along the surface of the vacuole, which also co-localized with Rtn1, indicating that these were peripheral ER tubules (Fig. 2C, yellow arrowheads).

The lipid-transfer protein Osh1 has a dual localization to the NVJ and the ER–Golgi complex contact site (Levine and Munro, 2001). Similar to Lam6, accumulations of Osh1 next to the vacuole only occurred at the nuclear ER, which is negative for Rtn1 (Fig. 2D, −rapamycin, white arrowheads). However, upon induction of the peripheral ER–vacuole contact site by the addition of rapamycin, Osh1 was recruited to these contact sites (Fig. 2D, +rapamycin, yellow arrowheads).

We reasoned that if a confined tether can cause other tethers to coalesce into this confined domain, an artificial tether with access to the whole ER (both peripheral and nuclear) should form contacts preferentially in the nuclear ER, where Nvj1 is located. We generated such an artificial tether by using the membrane-targeting domain of Tcb3 (1–272 amino acids). This domain tagged with GFP–FRB was located homogeneously in the ER, in both domains (Fig. 3A, −rapamycin). Upon induction of its dimerization with vacuolar Zrc1, we observed that these artificial contact sites were formed mostly in the nuclear ER (Fig. 3A, +rapamycin). To quantify this effect, we measured what percentage of the cellular signal of the Tcb3(1–272) construct was found in the nuclear ER. We observed that this fraction increased from 19.6% to 43.9% when the artificial contact site was induced with rapamycin (Fig. 3B), indicating that it depends on the construct acting as a tether of the contact site and not on an intrinsic preference for the nuclear ER region.

To further confirm that it is the restriction of Nvj1 to the nuclear ER that constrains all tethering proteins of the NVJ to localize in the nuclear ER, we used previously described mutants that disrupt the interaction of Nvj1 with the nuclear membrane (Millen et al., 2008). These mutants, Nvj1^L20E,V23E and Nvj1^{I10A, L13A,L15A}, localize to both the peripheral and nuclear ER and result in the formation of contact sites with the vacuole involving both domains of the ER (Millen et al., 2008) (Fig. 3C). As described previously, under control conditions, Lam6 was mostly localized away from the vacuole, likely in the ER–mitochondria contact site (Fig. 3C, white arrowhead) with some accumulations in the NVJ (Fig. 3C, cyan arrowheads). In the presence of the Nvj1 mutants with access to the peripheral ER, we observed that Lam6 was recruited to the peripheral ER–vacuole contact sites (Fig. 3C, yellow arrowheads; Fig. 3D). These results confirm that the restriction of Nvj1 to the nuclear ER causes the participation of Lam6 in vacuole–ER contact sites to be restricted to the nuclear part of the ER.

On the whole, our results show that, in addition to determining the partition between protein complexes or contact sites, tethers can cause the restriction of other tethers to a specific subdomain of an organelle.

Tethering proteins might not only affect the localization, but also the dynamics of other tethers. The mobility of contact site proteins can be important for their function. For example, the human lipid-transfer protein OSBP, which is also a tether of the ER–Golgi contact site, binds to phosphatidylinositol 4-phosphate (PI4P) on Golgi membranes. It also transports this lipid from the Golgi back to the ER, so it likely creates domains where PI4P is depleted, and it diffuses to richer zones to maintain transport. This was elegantly shown by deleting the local source of PI4P, which created long-distance movements of the protein, which followed a wave-like pattern (Jamecna et al., 2019; Mesmin et al., 2017). For the human protein VAPB, involved in ER–mitochondria contact sites, it was recently shown that a disease-associated mutation changes the mobility of the protein at the contact sites (Obara et al., 2022 preprint).

We thus decided to address whether tethering proteins of the same contact site could affect their mobility. We first analyzed the mobility of the tether Vps39 using fluorescence recovery after photobleaching (FRAP). We found that this tether was highly dynamic, recovering up to 0.75±0.18% (indicated as mean±s.d.) of the original fluorescence, with a half-life of 6 s (Fig. 4A,B). The percentage of the fluorescence that was not recovered, however, did not stem from a true immobile fraction, but rather from the amount of fluorescence that was bleached in the cell, as a second bleaching step resulted in recovery to a similar percentage (0.79±0.19%; the difference was not significant; Fig. S4A,B). The high mobility of this tether was in contrast to what has been reported for the Num1 tether of the mitochondria–cell cortex contact site and components of the ERMES complex, which tether the peripheral ER to the mitochondria. Both of these tethers were found to be almost completely static (Kraft and Lackner, 2017; Nguyen et al., 2012). We were also able to reproduce these findings and we detected almost no recovery for these proteins (Fig. 4A,B). In these cases, the immobile fraction did not stem from the bleaching of a high proportion of the cell signal, as we bleached one dot from approximately 7.6±0.4 dots per cell of ERMES (Fig. S4C), which would thus represent approximately 13% of the fluorescence of the cell. For Num1, to estimate the proportion of the cellular signal bleached, we calculated the percentage of fluorescence found in a region of the same dimensions as that used for bleaching, assuming that the bleaching occurred throughout all z-planes of the cell. We found that such a region contained 2.5±0.2% of the total fluorescence of the cell (Fig. S4D). Thus, in both cases, the bleaching of the signal cannot account for the lack of recovery and the signals are truly immobile. Our results show that contact sites can include tethers that are practically immobile to tethers that are almost completely mobile and highly dynamic (Fig. 4A).

Having established Vps39 as a highly mobile tether, we addressed whether the addition of another more static tether at this interface would influence its mobility. We thus measured the mobility of overexpressed Vps39 at the vacuole–mitochondria interface, in a strain that contains the previously described artificial tether for this contact site. As previously reported, overexpressed Vps39 forms patches of accumulation in the vicinity of mitochondria (Fig. 4C, −rapamycin). Upon the addition of rapamycin, Vps39 co-localized with the formed artificial tether (Fig. 4C, +rapamycin). FRAP analysis in this strain showed that the mobility of Vps39 within the contact sites that included the artificial tether was slower than in the absence of the artificial tether (Fig. 4D). Rapamycin did not affect the mobility of the protein in the absence of the artificial tether (Fig. S4E).

We performed a similar analysis in which we induced vacuole–mitochondria contact sites by overexpression of Cvm1. Under conditions in which both Cvm1 and Vps39 are overexpressed, both tethers are co-enriched in approximately 45% of the vacuole–mitochondria contact sites formed by Vps39 (Bisinski et al., 2022). We thus compared the mobility of overexpressed Vps39 within vCLAMPs without and with Cvm1 overexpression, exclusively in the contacts where they are both enriched. Similar to our previous results, we observed slower mobility in the contact sites that contained the additional Cvm1 tether (Fig. 4E,F).

To address whether the change in mobility caused to Vps39 required tethering to the mitochondria, we sought to disrupt the ability of Vps39 to tether the vacuole with the mitochondria. Ideally, we wanted to do this without manipulating the protein itself, as this could influence its mobility by changing the overall structure of the protein. While working with proteins of the translocase of the outer membrane complex, we observed that tagging the pre-protein receptor Tom20 at the C-terminus inhibited the formation of the extended Vps39-mediated vCLAMPs caused by overexpression of the protein (Fig. 5A). Indeed, quantification of this effect showed almost complete loss of the accumulation patches of Vps39 in the proximity of mitochondria in a strain in which Tom20 was tagged with HaloTag (Fig. 5B). This indicates that tagging Tom20 inhibited the formation of Vps39-mediated vCLAMPs. We have already shown that this modification has no impact on growth under respiratory conditions, indicating that mitochondrial function is not strongly affected (Bhagawati et al., 2021). To confirm that tagging Tom20 also inhibits vCLAMP formation at endogenous levels of Vps39, we addressed the co-purification of mitochondria with vacuoles, which reports on the extent of association between the two organelles (Bisinski et al., 2022; González Montoro et al., 2018). Tagging Tom20 diminished the co-purification of mitochondria with vacuoles, confirming that the formation of contact sites between them is reduced (Fig. 5C).

Next, we tested whether the generation of the artificial contact site affected the mobility of Vps39 in a context in which this protein cannot attach to the mitochondria, owing to the presence of a tagged version of Tom20. We observed that Vps39 was not specifically recruited to the artificial contact sites in the presence of tagged Tom20 (Fig. 5D) and the effect of the artificial tether on its mobility was almost completely abolished (Fig. 5E; Fig. S6A). This indicates that the generation of an artificial contact site with the mitochondria does not affect the mobility of Vps39 if it is not a tether of the contact site itself. Thus, this contact site does not impose a mobility restriction for some non-tethering proteins.

In the case of the vCLAMPs induced by overexpression of Cvm1, we were not able to perform the equivalent control, as tagging Tom20 with HaloTag resulted in the exclusion of Vps39 from the Cvm1-mediated vCLAMPs (Fig. S6B). The Vps39^12xM vCLAMP-impaired mutant was also excluded from these regions (Fig. S6C). In this case, the contact site appears to impose an accessibility restriction on these non-tethering proteins. This indicates that the cases in which Vps39 was co-enriched with Cvm1 depended on its ability to act as a tether.

In this study, we used different strategies to artificially modify membrane contact sites to address how different proteins with the ability to tether the same organelles influence each other.

First, we observed that the presence of a tether can promote the recruitment of other tethers of the same interface. In the case of proteins that participate in more than one contact site, such as Lam6 or Osh1, this effect can cause a shift in their distribution between different contact sites. This suggests that other tethers are factors that influence the overall subcellular localization of multi-contact-site-tethering proteins under a specific cellular state. A different case is that of Vps39, which is not present in multiple contact sites but in different protein complexes with different functions. In this case, the induction of a vacuole–mitochondria contact site shifted its subcellular distribution from dots found mostly away from mitochondria to being mostly in their proximity. As the dots found away from mitochondria likely represent sites of HOPS function, this suggests that the generation of a vacuole–mitochondria interface can regulate the partition of Vps39 between different protein complexes with different functions. Many proteins involved in contact sites are dual-function proteins (Harper et al., 2020), and our results indicate that their partition between different functions could be regulated by other tethers. Moreover, the complementation of cells lacking a tethering protein with an artificial tether is a classical approach when studying contact sites. Our results show that this type of experiment should be interpreted with caution as the artificial tether might influence other endogenous proteins with tethering capacity at a contact site.

In addition, we observed that Lam6 and Osh1 can be recruited to peripheral ER–vacuole contact sites by generating such a contact site with an artificial tether. This implies that several proteins with tethering capacity that normally localize to the NVJ can dynamically access the peripheral ER. In other words, the restriction of only one tether, Nvj1, to a specific subcompartment causes the steady-state enrichment of other tethers in this compartment. The generation of another contact in the peripheral ER was sufficient for these proteins to enrich there.

We then addressed the mobility of Vps39 at vCLAMPs and found that it is highly dynamic. This high mobility could explain why these contact sites are small and rare in control cells, and only observed frequently when tethers are overexpressed, as the interfaces are likely transient (Hönscher et al., 2014). On the contrary, ER–plasma membrane contacts, mitochondria–ER contacts and NVJs are readily visible in wild-type cells by electron and fluorescence microscopy (Collado et al., 2019; Hoffmann et al., 2019; Millen et al., 2008). Interestingly, the introduction of an artificial tether was sufficient to bring endogenous Vps39 to the contact. This suggests that in this type of highly dynamic contacts, the overall landscape could be regulated by one tether as other tethering proteins would follow.

Another interesting point is that the phenotypes caused by the deletion of Num1 and subunits of the ERMES complex could be suppressed by the expression of a static artificial tether (Klecker et al., 2013; Kornmann et al., 2009; Lackner et al., 2013). Both these tethers are static (Fig. 4A,B) (Kraft and Lackner, 2017; Nguyen et al., 2012). In contrast, the effects caused by removing the tethering function of Vps39 in vCLAMPs could not be suppressed by an artificial tether (González Montoro et al., 2021), and we found that this is a very mobile tether (Fig. 4A). This suggests that, in the cases in which tethers cannot be replaced by artificial ones, this might reflect either the requirement of a specific protein or of the dynamic nature of the contact site.

Our results also indicate that tethers of the same contact site can influence the mobility of each other. This reinforces what we previously discussed about cross-regulation among tethers of the same contact site: if the mobility of a tether is altered by a post-translational modification, this can affect the mobility of other proteins of the same interface. In our case, we observed that the introduction of an artificial tether or a second endogenous tether made Vps39 less mobile in the contact site. For the lipid-transfer protein and tether OSBP, it was shown that its disordered N-terminus acts as an entropic barrier that limits the density of the protein at the contact site and makes it more mobile (Jamecna et al., 2019). Disordered regions like this one are present in many other contact site components (Jamecna et al., 2019). An interesting possibility is that this type of component at contact sites might serve to maintain the mobility of all proteins within the interface.

An observation that was apparent throughout our different experiments was that some contact sites were inaccessible to some tethering proteins. When Vps39 was expressed at endogenous levels, it was recruited only to the periphery of both Cvm1 and FKBP–FRB-induced vacuole–mitochondria contacts, i.e. it was excluded from the region where the other tethers were enriched. However, under overexpression conditions, Vps39 co-enriches with the FKBP–FRB tether, but only with approximately 25% of Cvm1-mediated contact sites (Bisinski et al., 2022). This suggests that Vps39 and Cvm1 are more likely to segregate from each other than Vps39 and the FRB–FKBP tether. The contacts formed by the FRB–FKBP tether were plastic enough to accommodate Vps39 when it is overexpressed, even though they exclude it under endogenous levels. One characteristic that could influence this behavior is the intermembrane space generated by a tether. In mammalian cells, overexpression of VPS13C and PDZD8 generates extended ER–endolysosomal contact sites, in which these two tethers segregate from each other (Cai et al., 2022). Electron microscopy analysis of these regions showed that the intermembrane distance between membranes generated by PDZD8 were much smaller than for VPS13C (Cai et al., 2022).

Contact site regions of the membrane might also represent a barrier for non-tethering proteins. When Vps39 was not able to be part of the vCLAMP as a tether, its mobility was unaffected by the artificial FRB–FKBP generated contact site. This indicates that this contact site does not pose a mobility restriction to some non-tethering proteins. These results are consistent with the recent observation in mammalian cells that, although the tethering protein VAPB showed reduced mobility at contact sites, a control construct without the ability to tether organelles did not show regions of reduced mobility and is thus likely not affected by the presence of contact sites (Obara et al., 2022 preprint). In contrast, some contact sites do act as accessibility barriers for non-tethering proteins. For example, we observed that Vps39 was excluded from the Cvm1-mediated vCLAMPs when it could not act as a tether. A more physiological example of this is the exclusion of the vacuolar V-ATPase from the NVJ (Dawaliby and Mayer, 2010). These differences in the restriction of access to non-tethering proteins might reflect contact sites with different geometries, including intermembrane distances as described above or different levels of protein crowding, a factor that could be influenced by the presence of proteins with disordered regions, as discussed previously. Finally, when organelle membranes were swollen by incubation in hypotonic medium and cooled down, they displayed phase separation, which coincided with membrane contact sites (King et al., 2020). This suggests that these membrane regions have differential characteristics to the rest of the organelle membrane, a factor that might also influence the partition of both tethering and non-tethering proteins.

Here, we also uncovered the tagging of Tom20 with HaloTag as a way of inhibiting Vps39-mediated vCLAMPs. Although it is still not clear why the presence of this tag inhibits the formation of the contact, a likely explanation is that it generates steric hindrance for the interaction of Vps39 with the TOM complex. This suggests that a population of the TOM complex that contains Tom20 is involved in the formation of the contact site. We believe that this will be an important genetic tool for the dissection of the function of this contact site, which is still unresolved. As all proteins involved in this contact site are dual-function proteins, determining phenotypes of lack of the contact site is particularly tricky. The use of Tom20–HaloTag in combination with our previously described Vps39^12xM vCLAMP-impaired variant will provide a solution to this problem, as phenotypes related to the lack of vCLAMPs should be shared by these two alleles. This should allow distinguishing vCLAMP-related phenotypes from phenotypes stemming from affecting the function of the TOM or HOPS complexes.

In summary, our study shows that contact site tethers at the same interface can strongly influence the localization and dynamics of each other. On the one hand, this level of cross-influence must be taken into account when analyzing experiments. Deletions of tethers could result in decreased recruitment of other tethers or changes in their localization, and thus observed phenotypic consequences could be indirect. Likewise, the incorporation of artificial tethers will alter the localization and dynamics of other tethering proteins of the interface. On the other hand, in a physiological scenario, this means that the overall degree of localization of proteins at contact sites as well as their mobility will be partly determined by other tethers of the interface, without the need for specific molecular interactions. As this recruitment and dynamics are ultimately what determines the functionality of the contact site, these cross-influences will be a key component of the physiology of the structure. Furthermore, these effects could be exploited by cells to regulate tethers indirectly. Regulation of the protein levels of a tether or regulation of its interactions by post-translational modifications might result in a change in the overall proteome and dynamics of a multi-tether contact site. Our data suggest that the regulation and functionality of contact sites are determined by the interplay between direct specific interactions and cross-regulation dictated by the spatial characteristics of these structures.

Saccharomyces cerevisiae strains were based either on BY4741, SEY6210 or AGMY1210, a strain resulting from crossing BY4732 and HHY 110 (Y40342, Euroscarf) (Haruki et al., 2008). Genetic manipulations were carried out by homologous recombination of PCR-amplified cassettes as described in Janke et al. (2004). All yeast strains used in this study are listed in Table S1.

The plasmids pRS416-RTN1pr-RTN1-eGFP-FRB and pRS416-TCB3pr-TCB3(1-272)-eGFP-FRB were generated by PCR amplification of RTN1pr-RTN1 or TCB3pr-TCB3(1-272) from the genomic DNA of a BY4741 strain. The primer pairs used were oAGM148/149 and oAGM150/151, respectively. The PCR products were digested with XbaI and SalI, and ligated into a pRS416-based plasmid containing the eGFP and FRB sequences (a gift from Christian Ungermann, Osnabrück University, Germany), generating the plasmids pAGM033 and pAGM034, respectively. The plasmid pRS416-RTN1pr-RTN1-mKate2-FRB (pAGM042) was generated by inserting the mKate2 sequence between the NheI and SalI sites in the plasmid pAGM033.

The plasmid pAGM083 (pRS403-NVJ1pr-NVJ1-mNeonGreen) was generated by PCR amplification of the NVJ1pr-NVJ1 sequence from the genomic DNA of a BY4741 strain, using the primers oAGM399/443. The PCR product was digested with EcoRV and XhoI and inserted into plasmid pRS403 (Sikorski and Hieter, 1989). The mNeonGreen sequence was amplified by PCR using the primers oAGM403/584, digested with with EcoRI and BamHI and inserted in the same sites. The plasmids pAGM087 [pRS403-NVJ1pr-NVJ1(I10A, L13A, L15A)-mNeonGreen] and pAGM088 [pRS403-NVJ1pr-NVJ1(L20E, V23E)-mNeonGreen] were generated by PCR amplification of plasmid pAGM083 using overlapping primers containing the desired mutations (oAGM461/462 and oAGM463/464, respectively).

For the plasmid pAGM168 (pRS406-VPS39pr-VPS39^12xM-2xmKate2-Vps39ter), the plasmid pRS406-VPS39pr-VPS39^12xM-VPS39ter (González Montoro et al., 2018) was combined with a fragment containing 2×mKate2 sequence by Gibson Assembly using the oligonucleotides oAGM710/713. This plasmid was integrated into the terminator region of VPS39.

pRS403-based plasmids were integrated into the HIS3 locus. All plasmids were confirmed by sequencing. All plasmids and oligonucleotides used in this study are listed in Table S1.

Cells were grown to logarithmic phase in yeast extract peptone medium containing glucose (YPD), or synthetic medium, supplemented with essential amino acids (SDC). The vacuolar membrane was stained by adding 30 μM FM4-64 (Thermo Fisher Scientific) for 20 min, followed by washing and incubation in medium without dye for 30 min. The lumen of the vacuole was stained by adding 20 µM 7-amino-4-chloromethylcoumarin (CMAC) dye (Invitrogen) for 15 min, followed by one washing step in SDC medium. Proteins tagged with the HaloTag were labelled with the Janelia Fluor (JF) 646 ligand (Promega). The yeast cells (0.5 OD Units) were incubated with 2.5 µM of JF646 for 30 min, followed by ten washing steps with SDC medium (Day et al., 2018). Where indicated, cells were incubated for 30 min with 10 μM rapamycin (LC Laboratories, R-5000) prior to imaging.

Cells were imaged live in SDC medium on an Olympus IX-71 inverted microscope equipped with 100× NA 1.49 and 60× NA 1.40 objectives, an sCMOS camera (PCO, Kelheim, Germany), an InsightSSI illumination system, 4′,6-diamidino-2-phenylindole (DAPI), GFP, mCherry and Cy5 filters, and SoftWoRx software (Applied Precision, Issaquah, WA, USA). We used z-stacks with 200, 250 or 350 nm spacing for constrained-iterative deconvolution with the SoftWoRx software. All further image processing and quantification were performed using ImageJ (National Institutes of Health, Bethesda, MD, USA). One plane of the z-stack is shown in each figure.

For the quantification of the recruitment of Vps39, Vps39^12xM or Vps8 to mitochondria (Fig. 1D,F–H; Fig. S2B,D), we used z-stacks taken for the whole volume of the cells, with steps of 0.25 or 0.35 µm. The single dots of signal of the proteins were classified as located next to the mitochondria or not. Around 30 cells were analyzed for each of three independent experiments, and the diamonds in the graphs indicate the number of dots in each category for each cell. Comparisons were done with an unpaired two-tailed t-test when comparing two samples, and one-way ANOVA with a Tukey post hoc test when more than two samples were compared. For comparing the effect of overexpression of Cvm1 or induction of the artificial tether in Vps39wt and Vps39^12xM, a two-way ANOVA was used.

For the analysis of the presence of peripheral ER–vacuole contact sites, the percentage of cells containing patches of Rtn1-mKate2-FRB along the edges of the vacuole (CMAC-positive compartment) were quantified (Fig. 2B). To analyze the recruitment of Lam6 or Osh1 to the peripheral ER–vacuole contact sites (Fig. 2C,D), a cell was considered positive if a patch of Lam6 or Osh1 at the edges of the vacuole was also positive for Rtn1, indicating that it was not the nuclear ER. The cells were analyzed from z-stacks through the whole volume of the cell. Around 30 cells were analyzed for each category (with or without rapamycin) for three independent experiments. Statistical comparison was made between the samples with or without rapamycin, using an unpaired two-tailed Student's t-test.

For the analysis of the recruitment of Lam6 to peripheral ER–vacuole contact sites generated by the Nvj1 mutants, we quantified the percentage of cells that contained patches of Lam6 along the edges of the vacuole that were positive for Nvj1 (wt or mutant), and not part of the nuclear ER. Cells were analyzed through their whole volume, using z-stacks with 0.35 μm steps. The region corresponding to the nuclear ER was assessed morphologically, using the Sec63–HaloTag signal. Around 30 cells were analyzed for each of three independent experiments and condition. Statistical comparison was performed using a one-way ANOVA test and a post hoc Tukey test to compare the means.

For the analysis of the partition of Tcb3(1-272)–GFP to the nuclear and peripheral ER (Fig. 3B), we used z-stacks taken for the whole volume of the cells, with steps of 0.25 μm. The background of the images in the mNeon channel was subtracted using the ‘Subtract background’ function of ImageJ, with standard parameters. We then used the signal of the Elo3–mKate2 channel to generate elliptical regions of interest (ROIs) that encompassed the whole cell or the nuclear ER for each z-plane. The integrated intensity in these ROIs for each z-plane was added and used to calculate the fraction between the intensity of the nuclear ER and that of the whole cell. Five cells were analyzed for each of three independent experiments; each diamond in the graph represents one cell. Statistical comparison was performed between the means of each independent experiment (circles in the graph), using an unpaired two-tailed Student's t-test.

To illustrate the exclusion of Vps39^12xM or Vps39wt in the presence of Tom20–HaloTag from Cvm1-mediated vCLAMPs, enrichment factors were calculated (Fig. S6B,C). To do this, in the z-plane where the Cvm1-containing vacuole–mitochondria contact site was the largest, a line profile along the vacuole membrane in the contact site region was drawn. A second line profile was drawn along the vacuole membrane, outside the contact side region. The enrichment factor represents the ratio between the mean intensity value for Vps39 in the first and second line profiles.

Statistical tests were performed either with OriginPro 9.0 or with the R software package. Statistical comparisons were always performed between experimental means, to address reproducibility.

FRAP was performed with a Zeiss LSM880 with fast AiryScan microscope, equipped with a HXP 120 C fluorescence lamp; GFP, DsRed, CFP and DAPI filters, operated by Zen 2.1 software, and using a C Plan-Apochromat 63×/1.4 Oil objective. Cells were grown as described before, and FRAP assays were carried out at room temperature. Photobleaching was performed with an Argon laser (488 nm) on circular areas of 700 nm diameter within enriched regions positive for mNeonGreen signal, which was attached to the corresponding tether. Images were taken every second; three images were taken before the bleaching and 90 after bleach to analyze the recovery of the fluorescence.

The fluorescence intensity in bleached and non-bleached regions was measured using ImageJ. The different measured values were normalized, setting the intensity before bleaching as 1 and immediately after bleaching as 0. The plotted curves represent the average from three independent experiments, each including at least five cells; error bars represent the standard deviation.

For the two bleaching steps of mNeonGreen–Vps39, bleaching was performed twice in the same region, the second after 60 s of recovery. Each curve was normalized separately to its respective value before and after bleaching, and curves are shown overlaid for comparison. A comparison of the recovery of the first and second bleaching steps was performed by unpaired two-tailed Student's t-test.

Vacuoles were isolated as described in Haas (1995). To prepare whole-cell lysates, 30 OD units of cells in mid-log phase were harvested and resuspended in 300 µl of lysis solution (PBS containing 1% Triton X-100), and 200 μl of acid-washed glass beads (Thermo Fisher Scientific) were added. The samples were lysed by vigorous shaking using a Disruptor Genie (Scientific Industries) two times for 4 min, with 5 min of incubation on ice in between. The lysates were centrifuged for 20 min at 20,000 g and 4°C.

Protein concentration was measured by the Bradford method (Bio-Rad), and the same amount of protein for each sample was loaded for SDS-PAGE. Three independent purifications were performed and all samples were run together in the same gel. Samples were mixed with 4× Laemmli buffer (4% SDS, 0.05% bromophenol blue, 0.0625 M Tris pH 7.4, 400 mM DTT and 10% glycerol) and incubated for 10 min at 80°C. Proteins were separated by SDS-PAGE in 10% Bis-Tris acrylamide/bisacrylamide gels and transferred to nitrocellulose membranes (GE healthcare) using established protocols. After transfer, the membranes were blocked for 30 min in PBS containing 5% milk. Subsequently, the primary antibody [mouse monoclonal anti-Vph1 (10D7A7B2) (Abcam, ab113683, RRID: AB113683, dilution 1:1000) or mouse monoclonal anti-Por1 (Invitrogen, 459500; RRID: AB_2532239, dilution 1:1000)] was added and incubated for 1.5 h at room temperature with gentle shaking. The membranes were washed four times with PBS for 5 min, followed by incubation with a fluorescent dye-coupled secondary antibody (DyLight 800 goat anti-mouse IgG, Invitrogen, SA5-35521, RRID: AB_2556774, dilution 1:20,000). Fluorescence detection of the signal was performed in a LI-COR Odyssey scanner. The signal intensity of the different bands was quantified using ImageJ. The ratio between the Por1 signal and the Vph1 signal was calculated, and all samples were normalized by the average of the ratios obtained for the three samples of the wt strain. The complete scanned western blot membranes for the whole-cell lysates and the purified vacuoles are shown in Fig. S5.

We are thankful to Christian Ungermann, Florian Fröhlich and Lars Langemeyer for feedback on the manuscript, and to Aldana González-Montoro for help with statistical analysis. We thank Christian Ungermann for providing plasmids and strains. We are thankful to all members of the González Montoro lab for helpful discussions. Light microscopy was performed in the iBiOS facility of Osnabrück University.