ORP3 phosphorylation regulates phosphatidylinositol 4-phosphate and Ca²⁺ dynamics at plasma membrane–ER contact sites

Membrane contact sites (MCSs) between various organelles are emerging as key structural elements where important communication between organelles takes place (Gatta and Levine, 2017; Wu et al., 2018). They have been primarily featured in non-vesicular lipid transfer and Ca²⁺ signal propagation, but their importance is likely to reach beyond these two processes. An MCS is defined as the membrane apposition between membranes of two organelles with a distance no greater than 30 nm. While MCSs can be dynamic, they are stabilized by tethering proteins that also have functional roles. Several proteins have been identified that function in contact sites, including the extended synaptotagmins (ESyts) (Giordano et al., 2013; Manford et al., 2012) and many lipid transfer proteins (Wong et al., 2018). The ER Ca²⁺ sensor, STIM1 and the plasma membrane (PM) Orai1 channels also cluster in contact sites between the ER and PM to promote Ca²⁺ entry (Deng et al., 2009). Tethering proteins between the ER and mitochondria have also been shown to be important for Ca²⁺ transfer between these two organelles (Csordás et al., 2006; Szabadkai et al., 2006).

An important class of molecules that function at MCSs are the ORPs (oxysterol-binding protein-related proteins) (Kentala et al., 2016; Pietrangelo and Ridgway, 2018), which are mammalian orthologues of the yeast Osh proteins (Beh et al., 2012). The Osh/ORP proteins mediate the transport of specific lipids between cellular membranes, with their lipid cargo preference defined by their lipid transfer ORD (OSBP-related domain) domain (Pietrangelo and Ridgway, 2018). The best characterized members of this family are OSBP and its relatives, ORP5 (also known as OSBPL5), ORP8 (OSBPL8) and ORP2 (OSBPL2). Osh4 and OSBP were the first lipid transfer proteins in which the paradigm of phosphatidylinositol 4-phosphate (PI4P) gradient-driven cholesterol–PI4P counter-transport between the endoplasmic reticulum (ER) and the Golgi complex has been described (de Saint-Jean et al., 2011; Mesmin et al., 2013). ORP2 has been shown to mediate cholesterol transport between the ER and the PM controlled by PI(4,5)P₂ (Wang et al., 2018), whereas ORP5 and ORP8 proteins carry out the bidirectional transport of phosphatidylserine (PS) and PI4P between the ER and the PM (Chung et al., 2015). Less is known about the functions of other ORPs, particularly at ER–PM contact sites. Three additional ORPs, ORP3, ORP4 and ORP6 (OSBPL3, OSBPL4 and OSBPL6, respectively) have been reported to transfer lipids in membrane contact hotspots between the ER and PM (Mochizuki et al., 2018).

Generally, ORPs are anchored to the ER either with a C-terminal membrane spanning domain (in the case of ORP5 and ORP8) or via interaction of their FFAT (two phenylalanines in an acidic stretch) motifs with the ER-resident vesicle-associated membrane protein (VAMP) associated proteins (VAPs) (Loewen et al., 2003). The PM association of some of the ORPs is mediated by the interaction of their N-terminal PH domains with adjacent polybasic sequences with PM phosphoinositides, primarily with PI(4,5)P₂ and PI4P (Chung et al., 2015; Ghai et al., 2017; Sohn et al., 2018). Notably, however, not all yeast Osh proteins contain PH domains (Stefan et al., 2011).

While ORP3 has been shown to be associated with ER–PM contact sites (Mochizuki et al., 2018), recent studies also showed its presence in Rab7-positive endosomes, promoting their fusion with the nuclear envelope membrane (Santos et al., 2018). Binding of ORP3 to the PM requires protein kinase C activation, whereas its ER association is VAP protein dependent (Weber-Boyvat et al., 2015). Previous data showed that ORP3 has an important role in regulating cell adhesion due to its interaction with the small G-protein R-Ras (Weber-Boyvat et al., 2015). ORP3 overexpression also rescued the cellular phenotype associated with the P56S VAPB mutation linked to familial amyotrophic lateral sclerosis (Darbyson and Ngsee, 2016). In addition, ORP3 protein can form homo- or heterodimers with ORP6 while increasing PI4P turnover at the PM (Weber-Boyvat et al., 2015). Although these few selected cellular processes have been linked to ORP3, many details about its functions remain elusive.

In the present study, we investigated the role of inositol lipids in the control of ORP3 function and conversely, the possible role of ORP3 in the organization of ER–PM contacts with respect to PI4P status and store-operated Ca²⁺ entry (SOCE). Our studies showed that association of ORP3 with the PM depends on PI4P and PI(4,5)P₂ levels at the PM, and on the ability of the protein to reduce PM PI4P levels and inhibit Ca²⁺ entry via SOCE. Our results also revealed an important role of the C-terminal portion of ORP3 in its ability to transport lipids and inhibit Ca²⁺ influx.

The lipid transport processes mediated by OSBP and several ORP proteins use the chemical gradient of PI4P for the counter-transport of selected lipid cargoes at MCSs (Antonny et al., 2018). This PI4P gradient partly relies on the ER-resident Sac1 phosphatase, which keeps PI4P levels low in the ER (Del Bel and Brill, 2018; Zewe et al., 2018) (Fig. 1A). To investigate the effect of ORP3 activation on PM phosphoinositide levels, we used HEK293 cells overexpressing ORP3. Since ORP3 requires ER attachment via the VAP proteins, we created an expression vector that contained two coding sequences expressed as two separate polypeptides at fixed stoichiometry: the ORP3 protein fused to the N-terminus of a fluorescent protein and the ER-resident protein VAPB separated by the viral T2A sequence (Fig. 1B). This approach allows simultaneous expression of two proteins with fixed stoichiometry (Tóth et al., 2012; Várnai et al., 2017). Transfection of HEK293-AT1 cells (a HEK293 cell line expressing the AT1 angiotensin II receptors; Hunyady et al., 2002) with this construct yielded the expected products, ORP3-Venus and VAP-B separately (Fig. 1C). To alter the lipid binding affinity of the ORD, we introduced three separate mutations (H631/632A, K829A and K837A) that were designed based on homology with the yeast OSH3, OSH4 and OSH6 proteins for which crystal structures were available (Tong et al., 2016). The H631/632A (HH/AA) mutation was shown to eliminate PI4P binding of the yeast OSH4 (de Saint-Jean et al., 2011; Maeda et al., 2013) and the mammalian ORP5 and ORP8 proteins (Chung et al., 2015; Ghai et al., 2017). We also generated a truncated form of ORP3 in which the C-terminus distal to the ORD domain was deleted (ORP3-ΔC-terminal-mCherry) or key hydrophobic residues within this domain were mutated (ORP3^FF/IL-mCherry and ORP3^WY/LL-mCherry) (Fig. 1B).

Confocal analysis of transiently transfected HEK293-AT1 cells was used to assess the ER association of the ORP3-Venus protein using mCherry-Sec61B as an ER marker (Fig. 1D) (for the single Venus channel of the same images, see Fig. S1). In agreement with published data (Weber-Boyvat et al., 2015), ORP3 localization to the ER was more pronounced in the presence of VAPB but no PM attachment was observed in unstimulated cells (Fig. S1A). Only after protein kinase C (PKC) activation by PMA (100 nM) treatment was the PM association visible (Fig. 1D and Fig. S1A). Notably, the ER–PM patches after PMA treatment were more pronounced with the ORP3^HH/AA mutant (Fig. 1D and Fig. S1A). The PM association of these constructs was next analyzed by total internal reflection fluorescence (TIRF) microscopy. As shown in representative TIRF images in Fig. 2A, the number of PM patches developed after PMA treatment was again higher with ORP3^HH/AA protein compared with the wild type. Although it appeared that puncta formation was more pronounced in several cells when VAPB was also expressed (Fig. 2A), this difference was not substantiated by analysis of a larger number of cells (Fig. 2B). Membrane association was quantified by measuring the average pixel intensity change following PMA treatment after background subtraction.

Phosphorylation of both endogenous ORP3 (Fig. 1C, lanes 1 and 2) and the Venus-tagged protein (Fig. 1C, lanes 4 and 5) was indicated by the slower mobility of the phosphorylated protein in SDS gels, although this difference was less marked with the larger, Venus-fused protein. This result suggested that PKC activation was necessary for PM association and that lipid cargo binding may exert an inhibitory effect on the PM binding process, as observed with other lipid transfer proteins (Kim et al., 2015).

Given the ability of OSBP and some ORPs to transfer PtdIns4P, we wanted to know whether activation of ORP3 by PKC has any impact on PM phosphoinositide levels. For this, we monitored PM PI4P, PI(4,5)P₂ and PI(3,4,5)P₃ levels by bioluminescence resonance energy transfer (BRET) analysis in HEK293T cells expressing ORP3. Note that for most subsequent studies, we used HEK293-AT1 cells because of their utility to use receptor-mediated activation, but this early experiment was carried out in HEK293T cells, which expressed a higher level of the BRET constructs. The PMA-induced PI4P increase is more pronounced in these cells than in the HEK293-AT1 cells used in most of the current studies. The reason for this difference is not clear at present and was not pursued in this study, but it does not affect the conclusions regarding the effects of ORP3.

In control, mock-transfected cells, PMA treatment caused a slight elevation of PM PI4P and PI(3,4,5)P₃ levels, as observed earlier (Tóth et al., 2016). These changes were attributed to the activation of PI4K and Ras, respectively (Tóth et al., 2016). However, in cells expressing ORP3 (together with VAPB), PI4P showed a substantial decrease without any change in the other two phosphoinositides (Fig. 1E). The effects of ORP3 were similar but smaller in magnitude without VAPB co-expression, whereas VAPB alone had no effect (Fig. S1B). The PMA-induced PI4P decrease was greatly reduced in cells expressing ORP3^HH/AA, but it was not completely eliminated (Fig. 1E). Other mutants targeting PI4P binding (K829A or R837A) also reduced the effect of PMA, but neither has completely prevented the PI4P decrease (Fig. S1B,C). These results showed that PKC activation is required for triggering ORP3 function and that activated ORP3 can extract PI4P from the PM (Darbyson and Ngsee, 2016; Weber-Boyvat et al., 2015).

Next, we investigated the mechanism by which the ORP3 protein binds to the PM. To determine if phosphoinositides are required for PM binding of ORP3, we acutely altered phosphoinositide levels in the PM of HEK293-AT1 cells using the recruitable phosphoinositide-modifying enzymes, pseudojanin (PJ)-Sac1 and PJ 5-phosphatase (PMA did not increase PI4P levels in these cells as opposed to HEK293T cells). These tools can be used to selectively reduce PM PI4P or PI(4,5)P₂ content or both (Gulyás et al., 2017; Hammond et al., 2012). To probe the potential role of the phosphoinositide 3-kinase (PI3K) products such as PI(3,4)P₂ and PI(3,4,5)P₃ in ORP3 PM targeting, a low concentration of wortmannin (Wm, 100 nM) was used to inhibit class I PI3Ks. These manipulations were combined with the expression of ORP3-Venus and its PM association was monitored by TIRF microscopy. We used the lipid-transport-deficient ORP3^HH/AA mutant to mitigate its own effects on PM PI4P levels. In simultaneous experiments, the changes in PM phosphoinositides were followed with previously described BRET sensors to verify the effects of the recruited lipid-modifying enzymes (Fig. S2).

Reducing PI4P levels in the PM with the recruitable PJ-Sac1 enzyme significantly decreased both the rate and the extent of ORP3^HH/AA PM attachment compared with cells where the inactive PJ enzyme was expressed (Fig. 2C). Parallel BRET measurement showed that PI4P was selectively reduced in the PM after the recruitment of PJ-Sac1 (Fig. S2A). Even more profound inhibition of ORP3^HH/AA PM binding was observed when the PJ 5-phosphatase or the fully active PJ (both 5- and 4-phosphatase active) was recruited to the PM prior to PMA stimulation (Fig. 2C). Again, the BRET results confirmed the expected phosphoinositide changes after PM recruitment of the respective enzymes (Fig. S2A). In contrast, Wm pre-treatment had no effect on PMA-induced ORP3^HH/AA PM attachment (compared with the PJ-dead control), suggesting that PI(3,4)P₂ and PI(3,4,5)P₃ do not play an important role in ORP3 PM targeting (Fig. 2C). Confocal experiments showed that both of these phosphoinositide lipids were completely eliminated from the PM by Wm treatment (not shown). In the BRET experiment, the complete elimination of the lipids translates to a 60–80% decrease in BRET ratios (Fig. S2B) since the luciferase construct still had a small remaining effect on the membrane-bound Venus protein even when it is completely released from the membrane (also see Gulyás et al., 2017; Tóth et al., 2019, 2016). Overall, these data showed that PI(4,5)P₂ has the most important role in the PM binding of ORP3, while PI4P has a smaller, but still significant contribution. Our results did not show a significant role of the PI3K lipid products in ORP3 PM attachment.

So far, we have used PMA to stimulate PKC. Next, we wanted to examine whether activation of cells via Ca²⁺-mobilizing receptors was able to affect ORP3 localization or function. This was of interest because receptor-mediated PLC activation stimulates PKC by a combination of cytoplasmic Ca²⁺ increase and diacylglycerol (DAG) production, whereas PMA only mimics DAG. Moreover, at full activation, PLC also decreases the levels of PM PI4P and PI(4,5)P₂ (Bojjireddy et al., 2014), both of which we just showed to be important for the PM binding of ORP3. For this reason, we used HEK293-AT1 cells and a range of concentrations of angiotensin II (AngII). Lower agonist concentrations elicit a Ca²⁺ signal and activate PKC without the massive transient depletion of PI4P and PI(4,5)P₂ that is observed when using high concentrations of the agonist.

These studies showed that low AngII concentrations that did not appreciably decrease PI4P or PI(4,5)P₂ levels in mock-transfected cells, caused a decrease of PI4P in cells expressing ORP3 and VAPB (Fig. 3A, red curves). Higher agonist concentrations caused a transient decrease followed by a slow recovery in both PI4P and PI(4,5)P₂ levels in control cells (Fig. 3A, blue curves). In cells expressing ORP3 (with VAPB), the recovery of PI4P [but not of PI(4,5)P₂] was strongly inhibited (Fig. 3A, red versus blue curves). These data indicated that even at the slightly reduced steady-state PI(4,5)P₂ levels reached at later time points (>7 min) of 100 nM AngII stimulation, activated ORP3 can extract PI4P from the PM. The fact that we were unable to detect any changes in PI(4,5)P₂ levels due to ORP3 expression, suggested that ORP3 is unable to extract PI(4,5)P₂ from the PM. These data confirmed earlier observations that PI(4,5)P₂ levels can be maintained even at reduced steady state PI4P levels (Hammond et al., 2012; Bojjireddy et al., 2014; Nakatsu et al., 2012).

Since PKC activation by receptor stimulation results from the combined effects of DAG production and cytoplasmic Ca²⁺ elevation, we wanted to compare the effect of PMA treatment with that of the combination of PMA and thapsigargin (Tg) on PI4P changes in cells expressing ORP3 with VAPB. Tg activates the SOCE pathway by decreasing the ER luminal Ca²⁺ concentration via inhibition of the SERCA Ca²⁺ pump of the ER. As shown in Fig. 3B, Tg treatment, which had no effect of its own (red trace), clearly enhanced the effect of PMA in the activation of ORP3-mediated PI4P transport (green versus blue traces).

After testing the effects of receptor stimulation on the ability of ORP3 to influence phosphoinositide levels, we wanted to examine the PM association of ORP3 when stimulating with PMA alone or via AngII receptors. We, again, used TIRF microscopy and ORP3^HH/AA, which has a largely reduced effect on PM PI4P levels (Fig. 3B). Also, we used low AngII concentrations (5 nM) that do not reduce PM PI4P or PI(4,5)P₂ levels (Fig. 3A). Stimulation with 5 nM AngII caused a rapid association of ORP3 with the PM, a response that was faster and larger than that observed with PMA (Fig. 4A) (compare Fig. 2B and Fig. 4A, blue traces). Again, the ORP3^HH/AA protein showed significantly stronger and faster PM binding than the wild-type variant (Fig. 4A, red trace). The larger PM-translocation response of ORP3^HH/AA to low level AngII stimulation compared with the response to PMA was consistent with a stronger PKC activation in case of receptor stimulation. Since PMA activation has no effect on cytoplasmic Ca²⁺, whereas AngII stimulation evokes a Ca²⁺ increase, the role of Ca²⁺ was studied next.

To buffer the cytoplasmic [Ca²⁺] elevation during AngII stimulation we pre-treated cells with BAPTA/AM (10 μM, 10 min) before stimulation. BAPTA loading caused a significantly slower and reduced PM-translocation response of ORP3^HH/AA (Fig. 4A, green trace). In a reverse experiment, we examined whether increasing Ca²⁺ concentration by Tg on its own or combined with PMA would affect the PM-translocation response of ORP3^HH/AA. Treating cells with Tg alone resulted in a small and transient PM binding, which was essentially eliminated in cells pre-treated with BAPTA/AM (Fig. 4B, tan and light green traces). However, the PM attachment became more pronounced after stimulation with PMA and Tg together, with the response comparable in size to that observed with AngII treatment (Fig. 4B, red trace). Pre-treatment of the cells with BAPTA effectively inhibited the PM translocation of ORP3^HH/AA upon combined stimulation with PMA and Tg (Fig. 4B, green trace). Taken together, these data suggest an important role of the cytoplasmic [Ca²⁺] increase in the activation process of ORP3, presumably because of the faster and more complete PKC activation compared with PMA treatment alone.

In addition to releasing Ca²⁺ from the ER, both agonist stimulation and Tg treatment activates store-operated Ca²⁺ entry (SOCE) (Liou et al., 2005; Zhang et al., 2005). Therefore, we investigated the role of SOCE and STIM1 in ORP3 PM binding, and contact site formation. First, we used Ca²⁺-free medium (nominally Ca²⁺-free medium, containing 0.1 mM EGTA) to eliminate Ca²⁺ influx during stimulation with Tg and PMA and studied the translocation response of the ORP3^HH/AA mutant using TIRF analysis. As shown in Fig. 4B (orange trace), ORP3^HH/AA translocation to the PM was reduced in Ca²⁺-free medium, especially in the earlier phase. However, this difference was not statistically significant when calculated for the entire time period as the translocation response slowly approached the same level as seen in the presence of Ca²⁺. Parallel cytoplasmic Ca²⁺ measurements using a BRET-based Ca²⁺ sensor (Gulyás et al., 2015) showed that the cytoplasmic Ca²⁺ increase was only transient in response to Tg and PMA in Ca²⁺-free medium, reflecting Ca²⁺ release but no Ca²⁺ influx (Fig. 4C). In contrast, BAPTA loading buffered the early Ca²⁺ rise and slowed down the Ca²⁺ increase, which eventually reached similar plateau levels to those of Tg- and PMA-treated cells in normal medium (Fig. 4C, green trace). The strong effect of BAPTA on the association of ORP3^HH/AA with the PM suggested that both the early rate of Ca²⁺ rise during combined Tg and PMA stimulation and increased Ca²⁺ influx had an important role in ORP3 activation.

Next, we examined whether the contact sites marked by STIM1 activation are the same or different from those occupied by activated ORP3. For this, we turned to COS-7 cells, which have a larger flat footprint and an elaborate network of tubular ER. Cells were transfected with ORP3-mCherry (together with VAPB) and a thymidine kinase-driven YFP-STIM1 construct and examined with confocal microscopy. Stimulation of cells either via the Gq-coupled endogenous P_2Y receptor by ATP, or Tg and PMA (Fig. 5A and B, respectively) caused clustering of both STIM1 and ORP3 and they showed good colocalization, suggesting that they were enriched in the same contact sites. The effect of STIM1 overexpression was also examined on ORP3^HH/AA translocation to the PM in HEK293-AT1 cells (Fig. 5C). STIM1 overexpression had no effect when ORP3 translocation was examined after PMA stimulation (Fig. 5D, blue). Although STIM1 expression slightly enhanced the small response to Tg, this effect did not reach statistical significance (Fig. 5D, pink). Only when the cells were stimulated with Tg and PMA did STIM1 increase the translocation response of ORP3^HH/AA (Fig. 5D, red). Again, BAPTA treatment largely reduced the ORP3^HH/AA translocation response regardless of the presence of STIM1 (Fig. 5D, green).

Lastly, we wanted to determine the contribution of endogenous STIM1 to ORP3 PM attachment. For this, we knocked down STIM1 by siRNA and determined ORP3^HH/AA PM translocation by TIRF microscopy. We also followed cytoplasmic Ca²⁺ changes using the BRET-based Ca²⁺ reporter during Tg and PMA stimulation. These experiments showed that STIM1 knockdown (Fig. 5F) significantly reduced the PM translocation of ORP3^HH/AA following TG and PMA treatment (Fig. 5E) and also reduced the sustained phase of Ca²⁺ increase, consistent with its well-documented role in SOCE (Fig. 5G).

Next, we explored whether lipids other than PI4P can be transported by the ORP3 ORD domain. For this, we used an ORP3 construct in which the PH domain was replaced with the FKBP module (Fig. 1B). This approach allows acute activation of ORP3 protein by acutely establishing its PM association using the PM-targeted FRB fragment of mTOR, as previously described for the ORP5 and ORP8 proteins (Sohn et al., 2018). To monitor changes in specific lipids in the PM after ORP3 activation, we used our previously described BRET-based PM lipid sensors for PI4P, PI(4,5)P₂, PI(3,4,5)P₃, PS and cholesterol (Sohn et al., 2018; Tóth et al., 2016).

As shown in Fig. 6A,B, rapamycin treatment caused a rapid decrease in PM PI4P, confirming the ability of the protein to extract PI4P from the PM. However, no decrease was observed when using the D4H-based cholesterol probe; in fact, a slight increase was observed in PM cholesterol levels after rapamycin treatment (Fig. 6A,B, red). In contrast, a slight decrease was observed with the PA-specific BRET probe (Fig. 5A,B, green). No changes were detected in PS, PI(4,5)P₂ or DAG levels using this approach (Fig. 6AB).

Given the presence of ORP3 and STIM1 in the same ER–PM contact sites, we investigated whether this colocalization has a functional consequence on Ca²⁺ signaling. For cytoplasmic [Ca²⁺] measurements, we used the BRET-based Ca²⁺-sensor in HEK293-AT1 cells. Cells were stimulated with increasing concentrations of AngII, which evoked bigger and bigger plateau Ca²⁺ elevations in control cells (Fig. 7A, blue curves). ORP3 expression together with VAPB reduced the Ca²⁺ plateau at higher AngII concentrations, and notably, the HH/AA mutant version was as effective as the wild-type protein (Fig. 7A, red and green curves). Other ORP3 mutants (K829A, R837A) that also impaired PI4P transport by these proteins (see Fig. S1B and C) had similar inhibitory effect (Fig. 7B). Importantly, no inhibition was detected when the VAPB protein was not expressed with the ORP3 proteins (Fig. S3B). We repeated these experiments using AngII together with Tg to prevent the refilling of the ER Ca²⁺ stores during AngII receptor desensitization. This manipulation allows better evaluation of any potential difference between the mutant and wild-type ORP3 proteins on Ca²⁺ signals. However, no difference was observed between the inhibitory effects of ORP3 or its mutants on these larger Ca²⁺ signals (Fig. 7C). Importantly, overexpression of the various ORP3 constructs did not prevent association of STIM1 with the PM (not shown).

Lastly, to determine if the inhibitory effect of ORP3 required its activation by PKC, we performed similar Ca²⁺ measurements using only Tg or Tg together with PMA in cells overexpressing ORP3 or ORP3^HH/AA together with VAPB. These experiments showed that only when PMA was added together with Tg were the ORP3 proteins able to inhibit the Tg-induced Ca²⁺ elevations, and, again, the ORP3^HH/AA protein was as effective as the wild type (Fig. 7D).

The ORD domains of the Osh and ORP proteins show a large degree of conservation (Tong et al., 2016). The X-ray structures of yeast OSH3 and OSH4/Kes1 proteins show a long C-terminal segment that follows the conserved helix, the latter forming one side at the top of the lipid binding pocket (Fig. 8A). Interestingly, this C-terminal segment (labeled pink in Fig. 8A) is highly conserved among the various ORP3 orthologues (Fig. 8B). We investigated whether this region of the molecule has any importance for the function of the ORP3 protein.

First, we deleted the last 35 amino acid residues of ORP3 (ORP3ΔC) (Fig. 8B). This truncated ORP3 protein was also combined with the mutations that affected the lipid binding of the ORD domain. None of these truncated proteins was able to extract PtdIns4P from the PM following PMA and Tg treatment, even in the presence of VAPB (Fig. 8C). Moreover, none of these truncated ORP3 proteins were able to modulate Ca²⁺ signaling in Tg- and AngII-stimulated cells (Fig. 8D). Importantly, the C-terminally deleted ORP3 was still able to associate with the ER and increase its ER association after activation by Tg and PMA treatment (Fig. 8E). Yet, it showed no punctate structures after activation, suggesting that it no longer interacted with the PM (Fig. 8E). Although these features argued against a major folding problem, we generated single-residue mutations targeting conserved hydrophobic residues that were predicted to provide interaction with the core of the ORD domain (labeled cyan in Fig. 8A). These mutants [F854I, F855L (FF/IL) and W863L, Y869L (WY/LL)] completely reproduced the behavior of the C-terminally deleted form.

Taken together, these data suggest that this C-terminal region of ORP3 (and perhaps other ORPs) plays an important role most likely by stabilizing the position of the long, conserved helix and thereby allowing retention of the lipid cargo within the lipid binding pocket.

The present study was designed to investigate the properties and possible regulatory role of the ORP3 protein in PM–ER contact sites. The unique feature of ORP3 compared with OSBP and some other ORPs is that it requires PKC activation to induce a molecular rearrangement in order to display its membrane interaction and lipid transport ability (Weber-Boyvat et al., 2015). Little is known about the functions and regulation of ORP3, although the protein has been linked to focal adhesions, by regulating β1-integrin and R-Ras activity (Lehto et al., 2008; Weber-Boyvat et al., 2015), and its depletion causes aneuploidy and possibly contributes to cancer progression (Njeru et al., 2019).

In this paper, we found that the primary determinants of ORP3 PM binding are PI4P and PI(4,5)P₂ in the PM, but PM binding only occurs after PKC activation. A previous study using the isolated PH domain of ORP3 found its binding preference to PI(3,4,5)P₃ and PI(3,4)P₂, but our experiments using intact cells could not confirm the role of these lipids in bulk PM binding of ORP3. We cannot rule out the possibility that the 3-phosphorylated lipids play a role in localized enrichment of the protein in specific PM compartments, such as in focal adhesions, or in other locations within the cell. For example, recent studies have shown an important role of ORP3 in entry of late endosomes to the nucleoplasmic reticulum (Santos et al., 2018), a defect that may be related to the aneuploidy observed in ORP3-depleted cancer cells (Njeru et al., 2019). Although PI(4,5)P₂ has been shown to be present in late endosomes (Baba et al., 2019), it is possible that other phosphoinositides, such as PI(3,4)P₂ control ORP3 function in those compartments.

It has also been shown earlier that PKC-mediated phosphorylation regulates the attachment of ORP3 to VAPA (Weber-Boyvat et al., 2015). While we did not focus on this aspect of ORP3 regulation, we confirmed that PMA treatment caused a mobility shift in the migration of ORP3 in SDS gel electrophoresis consistent with its phosphorylation, and our studies also showed that binding of ORP3 to the PM phosphoinositide was also regulated by PKC activation. Based on these data, it is most likely that phosphorylation induces a major conformational change in ORP3 and opens the protein from a closed conformation that masks interaction of both the N-terminal PH domain and the FFAT motifs with their respective binding partners.

Recent studies have shown that phosphoinositides, particularly PI(4,5)P₂ and PI(3,4)P₂, control the lipid cargo loading of the ORD domains of ORP1 and ORP2 (Dong et al., 2019; Wang et al., 2018). In fact, the presence of a PH domain is not necessary for the functions of some of the yeast Osh proteins, such as Osh4/Kes1 (Maeda et al., 2013). The role of the PH domain (and adjacent sequences) in metazoan and some yeast OSBP homologues is to facilitate membrane interaction and localization of these proteins to membrane contact sites. Consistent with these ideas, replacing the PH domains with FKBP preserves the cargo transfer ability of the ORP3 protein and makes it inducible after establishing PM binding with rapamycin-mediated heterodimerization. Using this approach, we showed that the main phospholipid cargo of ORP3 protein is PI4P. Of the other lipids examined, we found a slight decrease in PM PA after ORP3 activation and a slight elevation in PM cholesterol. This latter finding might indicate a PI4P-coupled counter-transport of cholesterol to the PM by ORP3, but most likely, the main lipid cargo of this protein still awaits identification. The observed PA changes could reflect a direct ORP3-mediated PA transport or could be, again, an indirect consequence of PI4P transfer.

The most important and robust lipid change observed after ORP3 activation was the substantial depletion of PI4P in the PM. This suggested that ORP3 belongs to the increasing number of ORPs that use PI4P as a lipid cargo. Interestingly, mutations (HH/AA, K829A and R837A) within the ORD domain that have been shown to grossly decrease, if not eliminate, PI4P binding (Chung et al., 2015; de Saint-Jean et al., 2011; Maeda et al., 2013), have greatly reduced but did not fully eliminate the PI4P-depleting effect of ORP3. This may indicate a residual PI4P binding of the ORD domains, but also raised the possibility that ORP3 protein could stabilize ER–PM contacts and facilitate a possible action of Sac1 on PM PI4P in trans as speculated for Osh3 (Stefan et al., 2011). While the trans activity of Sac1 is not likely to be a major factor at the PM under physiological conditions (Zewe et al., 2018), it may become appreciable when promoting ER–PM contacts with expressed proteins. Curiously, a modification that completely eliminated all the activities of ORP3 was a truncation at its C-terminus or mutations that are expected to weaken its interaction with the ORD proper. This region is highly conserved among ORP3 orthologues, including Osh3 (Fig. 7B). Based on the structure of Osh3, this part of the molecule is not directly involved in cargo binding and contains solvent exposed positively charged and hydrophobic amino acids that may confer a binding surface for either an intramolecular interaction or interaction with another protein partner. The intramolecular interaction that is mediated by the hydrophobic residues mutated in this study could determine and stabilize the position of the long helix that borders the cargo binding hydrophobic pocket (Fig. 7A). More studies will be needed to characterize the role of this segment of the molecule in its activation and lipid transfer activities.

An important finding of our study was the synergistic effect of Ca²⁺ and PMA in evoking a maximal activation of ORP3. This explains the more robust activation of ORP3 during stimulation of receptors that activate PLC. Even at concentrations of agonists that reduce steady-state PI(4,5)P₂ and PI4P levels, there was enough phosphoinositide in the PM to support ORP3 activation. Our studies comparing the effects of BAPTA loading, removal of external Ca²⁺ or knockdown of STIM1 on ORP3 membrane association suggested that both the rate of Ca²⁺ rise and the sustained Ca²⁺ influx through the SOC channels were important for the full activation and PM association of the protein. The fact that STIM1 knockdown had a larger effect on ORP3 PM association than removal of external Ca²⁺, raised the possibility that STIM1-mediated PM–ER contacts, which occur after ER store depletion even without external Ca²⁺, could help ORP3 accumulation in this compartment. Conversely, overexpressed STIM1 increased ORP3 PM association simply by promoting and enlarging those ER–PM contact areas.

Lastly, our data revealed a potent inhibitory effect of ORP3 on Ca²⁺ signaling via SOCE. A previous study using targeted aquorin as a Ca²⁺ sensor showed that ORP5 and ORP8 overexpression increased Ca²⁺ concentration in caveolar microdomains and in mitochondria but was without effect on cytoplasmic Ca²⁺ changes evoked by histamine stimulation (Pulli et al., 2018). Since ORP3 activation leads to PM PI4P depletion, an obvious explanation for this inhibitory effect was a change in PM PI4P. Previous data suggested that SOCE is sensitive to PM phosphoinositide levels, as the C-terminal polybasic domain of STIM1 binds to acidic phospholipids such as PI(4,5)P₂ (Liou et al., 2007). Notably, the role of the different phosphoinositide levels has been a matter of debate, because PI4P, rather than PI(4,5)P₂, was suggested by three studies as the more important lipid in this context (Broad et al., 2001; Korzeniowski et al., 2009; Nakatsu et al., 2012). However, the inhibitory effect of ORP3 was also observed with mutants that were unable to achieve the same PI4P depletion as the wild type, and only the C-terminally-deleted or mutated ORP3 mutants were without effect. This raised the possibility that the ORP3 effect on SOCE may not be solely mediated by changes in phosphoinositides. The close physical proximity of the ORP3 proteins to the components of the STIM1–Orai1 machinery raise the question of whether there is a direct interaction between ORP3 and one or more of these proteins. Alternatively, a secondary change in the lipid composition of the PM caused by the increased presence of ORP3 could be responsible for this effect. While the exact mechanism of how this inhibition takes place is a matter of further investigation in our laboratory, it is intriguing that the sensitivity of ORP3 to SOCE-mediated Ca²⁺ increases is associated with an inhibitory effect of ORP3 on the very same process.

In summary, our studies have revealed several features of ORP3 activation; its binding to PM PI(4,5)P₂ and PI4P, the synergistic role of combined Ca²⁺ and DAG signals in ORP3 activation and the intimate reciprocal relationship between ORP3 and SOCE. The importance of the C-terminal domain of ORP3 activation and lipid transport suggests an evolutionarily conserved feature that will require further studies to be fully understood.

All molecular biology reagents were from New England Biolabs unless specified otherwise. For the polymerase chain reaction, the Pfu Turbo Hotstart polymerase was used (Agilent Technologies). Lipofectamine 2000 was purchased from Thermo Fisher Scientific. White flat bottom 96-well plates obtained from Corning were used in the BRET experiments. Eight-well µ-Slides for confocal experiments were ordered from IBIDI (Cat. No. 80826), while glass bottom single dishes for TIRF experiments were from CellVis (Mountain View, CA, USA). Coelenterazine h was purchased from Regis Technologies (Morton Grove, IL, USA). Fura2-AM was from Thermo Fisher Scientific. Rapamycin and human AngII were obtained from EMD Millipore (Burlington, MA). Unless otherwise stated, all other chemicals and reagents were purchased from Sigma (St Louis, MO, USA).

Polyclonal rabbit antibodies against: VAP-33 (Cat. No. A304-366A-T); VAPB (Cat. No. A302-894A-T); ORP3 (Cat. No. A304-557A-1), were all obtained from Bethyl Laboratories (Montgomery, TX, USA). The polyclonal rabbit antibody against STIM1 (Cat. No. 4916S) and the monoclonal mouse antibody against α-tubulin (Cat. No. 2144S), were from Cell Signaling Technology (Danvers, MA). The secondary fluorescent antibodies (IRDye^® 680LT goat anti-rabbit and IRDye^® 800CW goat anti-mouse) were both from LI-COR. All primary antibodies were used at 1:1000 and the secondary antibodies at 1:5000 dilutions.

The ORP3-mCherry construct was created by cloning the amplified ORP3 cDNA sequence (NM_015550.3, purchased from Dharmacon, clone ID: 3909164) into GFP-N1 plasmid (for primers, see Table S1) using HindIII and KpnI restriction sites. The GFP sequence was then replaced with mCherry using AgeI and BsrGI sites. The resulting construct has a linker sequence AVPRARDPPVAT between ORP3 and GFP. The mCherry-VAPB construct was a gift from Christopher Stefan (UCL, London, UK) and the mCherry-Sec61B-C-18 was Addgene plasmid #55129 (deposited by Michael Davidson). To create the T2A-based vector, which contained the ORP3 and VAPB sequences in the same plasmid, the NES-Sluc-GRP1PH sequence was replaced with the PCR-amplified sequence of VAPB in our recently introduced L10-mVenus-T2A-NES-Sluc-GRP1PH construct (Tóth et al., 2019) using EcoRI and BamHI enzyme sites. In the second step, the T2A-VAPB piece was PCR-amplified, and then inserted after the mCherry sequence in the ORP3-mCherry plasmid using Bsp1407I and NotI enzymes. The FKBP-ORP3-mCherry construct was created by amplifying the FKBP sequence with PCR and inserting it into the ORP3-mCherry construct by replacing the PH domain of the protein (30–154 amino acids), using the SalI and EcoRI sites. The SalI site is present in the ORP3 sequence, whereas the EcoRI site was generated by site directed mutagenesis. Mutations (H631/632A, K829A, K837A, FF854/855IL and WY863/869LL) were introduced in the ORP3-mCherry, as well as in the ORP3-mCherry-T2A-VAPB constructs using the primers (see Table S1) using the Quik-change mutagenesis kit (Promega, Madison, WI, USA). To generate the C-terminal truncated mutants, the ORP3 sequence lacking last 35 amino acids was amplified by PCR and ligated back to the original template constructs using the NheI and AgeI restriction enzymes. In the case of the VAPB containing variants, the original ORP3 sequence was replaced with the truncated version. Venus tagged ORP3 proteins were made by replacing the mCherry sequence with the monomeric Venus (N1-mVenus, containing the A206K mutation) using AgeI and Bsp1407I enzymes. The mRFP-STIM1 construct (Várnai et al., 2007), as well as the BRET-based lipid-sensors for PIPs (Tóth et al., 2016), cholesterol and PtdSer (Sohn et al., 2018), or PA and DAG (https://www.biorxiv.org/content/10.1101/677229v1) were described earlier. The BRET-based cytoplasmic-Ca²⁺-sensor was described in Gulyás et al. (2015) and the single-plasmid versions of the recruitable lipid-phosphatase system in Tóth et al. (2012).

HEK293T, COS-7 (both from ATCC) and the HEK293-AT1 cells (a line that stably expresses the type 1A rat angiotensin II receptor (rAT1_AR) (Hunyady et al., 2002) were maintained in Dulbecco's modified Eagle's medium (DMEM, Gibco 11995-065) supplemented with heat inactivated 10% fetal bovine serum, 50 U/ml penicillin and 50 µg/ml streptomycin (Gibco, 15140-122) together with 5 μg/ml Plasmocin (InvivoGen) as a mycoplasma prophylactic (culture media) in a 5% humidified CO₂ incubator at 37°C in tissue culture flasks (Thermo Fisher Scientific, 178905). Cell lines were regularly tested for mycoplasma contamination.

HEK293-AT1 cells were plated at a density of 3×10⁵ cells/well in 6-well plates in culture medium at 37°C. One day later, the cells were transfected with 200 µl transfection mixture containing the indicated DNA constructs and 2 µl/dish Lipofectamine 2000. After 24 h, the plate was transferred on ice, and the cells were scraped into 300 μl RIPA buffer. After centrifuging the samples (15,000 g, 10 min at 4°C), SDS sample buffer was added to the supernatant and samples boiled at 95°C for 5 min. Samples were loaded onto 8–16% precast SDS gels (Invitrogen, Carlsbad CA) and run for 90 min at 130 V at room temperature. The proteins were transferred to nitrocellulose membranes and incubated with the respective primary and secondary antibodies. The blots were then scanned with an Odyssey fluorescence imager (LI-COR, NE).

For BRET measurements, HEK293 cells were seeded at a density of 7×10⁵ cells/ml in culture medium. The transfection mixture contained 0.5 μl/well Lipofectamine 2000 and 0.2 µg/well total DNA from the indicated constructs in 50 μl/well Opti-MEM (Gibco, 31985-070). After incubation, the mixture was gently mixed with 100 μl/well cell suspension and seeded onto poly-L-lysine-treated white 96-well, flat bottom microplates (7×10⁴ cells/well). Poly-L-lysine (Sigma, P8920, diluted to 0.001%) treatment was for 1 h and then removed to dryness before cell plating. BRET measurements were performed 25–26 h after transfection. Cells were washed once with a modified Krebs–Ringer buffer (50 µl) (120 mM NaCl, 4.7 mM KCl, 1.2 mM CaCl₂, 0.7 mM MgSO₄, 10 mM glucose and 10 mM Na-HEPES, pH 7.4) then incubated in the same medium for 30 min at 37°C. Coelenterazine h (the cell-permeable luciferase substrate) was dissolved in the modified Krebs–Ringer buffer (5 µM final concentration) and added in 40 µl/well volume. Measurements were then made at 37°C using a Tristar2 LB 942 Multimode Microplate Reader (Berthold Technologies). The emitted intensities were recorded using 475/20 and 540/40 nm emission filters. Detection time was 500 ms for each wavelength. The indicated reagents were added manually in 10 µl after dissolving in modified Krebs–Ringer buffer. For the stimulations, the plates were unloaded, which resulted in interruption of the recordings. To calculate changes in the PM lipid levels, BRET ratios were calculated by dividing the intensities recorded in the 540/40 nm filter by those recorded at the 475/20 wavelengths. These ratios were then normalized to the baseline and the percent changes were calculated (Tóth et al., 2019). In case of Ca²⁺ measurements, the intensities were normalized to the average values of the last three data points recorded before stimulation.

COS-7 and HEK293-AT1 cells were seeded at a density of 3×10⁴ cells/well on IBIDI 8-well µSlides and cultured at 37°C for 1 day. For transfection, the culture medium was changed to 200 µl transfection mixture containing the DNA constructs (0.2 µg total DNA/dish) and 0.33 µl/dish Lipofectamine 2000. 6 h later, the medium was changed to 300 µl supplemented DMEM medium. Cells were inspected 24–26 h after transfection using a Zeiss LSM 710 laser confocal microscope with a 63×/1.4 oil-immersion objective. Cells were kept in modified Krebs–Ringer buffer (described above) at room temperature during the experiments. Post-acquisition picture analysis was performed using Fiji and Photoshop (Adobe) software. Only linear changes were made during picture analysis and processing.

HEK293-AT1 cells were plated on 35 mm dishes at a density of 3×10⁵ cells/dish. One day later, the cells were transfected with 200 µl transfection mixture containing the indicated DNA constructs (0.6 µg each) and 2 µl/dish Lipofectamine 2000. Cells were analyzed 20–24 h after transfection using a Nikon Eclipse TiE inverted microscope equipped with a TIRF module. Excitation was provided by a Nikon TIRF illuminator with a fiber-coupled LUN-V four-line laser launch (Nikon). Emitted intensities were acquired by a Zyla 5.5 EMCCD camera (Andor) through a Lambda 10-2 emission wheel (Sutter Instruments). mVenus and mRFP fluorescence were imaged with 488 nm and 561 nm excitation respectively, with a dual-emission filter collecting 505–550 nm and 570–620 nm (Chroma). Cells were kept in the modified Krebs–Ringer buffer at 37°C. For this, the stage and the 1.45 NA 100× plan-apochromatic oil-immersion objective (Nikon) were heated. For calculating the normalized PM intensities, images were background subtracted, then the average intensities were estimated on the entire image excluding the pixel values with 0. For normalization, each average intensity value was divided with the average value of the last three points acquired before the stimulation. Images were acquired every 10 s. The addition of stimuli was carried out manually after the fifth frame. Post-acquisition picture analysis was performed using Fiji and Photoshop (Adobe) software.

HEK293-AT1 cells were plated on IBIDI 8-well µSlides or on 96-well microplates at a density of 10⁴ cells/well. One day later, cells were transfected with 20 µl (for TIRF) or 50 µl (for BRET) transfection mixture containing the indicated siRNA (20 pmol/well) [r(GGGAAGACCUCAAUUACCA)dTdT sense sequence for Stim1; AllStars negative control siRNA, Cat. No. 1027280, both from Qiagen] and 0.75 µl/dish Lipofectamine RNAiMAX (Thermo Fisher Scientific, Cat. No. 13778150). On the next day, the medium was changed to fresh culture medium. On the fourth day, cells were transfected with the indicated DNA constructs with a mixture that contained 0.15 μg or 0.05 μg DNA and 0.75 μl or 0.5 μl Lipofectamine 2000 for TIRF and BRET experiments, respectively. For STIM1 gene silencing, a duplex customized siRNA was used (Qiagen). Cells were analyzed 24–26 h after transfection, as detailed above for the TIRF and BRET measurements.

For data plotting and statistical analysis, the GraphPad Prism (San Diego, CA) software was used. Specific statistical analyses are given in the figure legends for each experiment.

Confocal imaging was performed in the Microscopy Core of NICHD with the kind assistance of Drs Vincent Schram and Lynne Holtzclaw. The technical assistance of Kata Szabolcsi in the early phase of this work is highly appreciated.

The peer review history is available online at https://jcs.biologists.org/lookup/doi/10.1242/jcs.237388.reviewer-comments.pdf