Calcium Regulates the Association between Mitochondria and a Smooth Subdomain of the Endoplasmic Reticulum

A close association between ER and mitochondria has been described in various cell types. Close contacts between the ER membrane and the mitochondrial outer membrane have been visualized by EM (Franke and Kartenbeck 1971; Morre et al. 1971; Montisano et al. 1982), and recent electron tomography studies have identified 15-nm diameter sites of close contact between the membranes of these two organelles (Perkins et al. 1997). Subcellular fractionation of cells identified an ER membrane fraction that copurifies with mitochondria (Lewis and Tata 1973; Shore and Tata 1977) and ER-derived mitochondria-associated membrane (MAM) fractions have been implicated in phospholipid transfer between the two organelles (Vance 1990; Shiao et al. 1998; Achleitner et al. 1999).

Interaction between ER and mitochondria has also been implicated in the sequestration and regulation of free cytosolic calcium (Ca²⁺; for review see Pozzan et al. 1994). Ca²⁺ release due to the generation of inositol 1,4,5-triphosphate (IP3) in response to receptor activation results in the formation of high local concentrations of Ca²⁺ at sites of close contact between ER and mitochondria that results in increases in mitochondrial Ca²⁺ (Rizzuto et al. 1993, Rizzuto et al. 1998; Simpson et al. 1997). ER-mediated Ca²⁺ uptake by mitochondria may activate mitochondrial metabolic activity by stimulating Ca²⁺-dependent mitochondrial dehydrogenases (Denton and McCormack 1990; Rutter et al. 1996; Jouaville et al. 1999), as well as regulate the subcellular pattern of Ca²⁺ oscillations generated by receptor activation (Rizzuto et al. 1994; Landolfi et al. 1998; Robb-Gaspers et al. 1998; Hajnoczky et al. 1999). Close contacts between ER and mitochondria, particularly under conditions of high local free cytosolic Ca²⁺, are therefore of functional importance.

We have identified a smooth ER subdomain specifically labeled for autocrine motility factor receptor (AMF-R; Benlimame et al. 1995). By EM, smooth AMF-R–labeled tubules exhibit continuity with the ribosome-studded tubules of the rough ER, and after treatment with ilimaquinone, form a highly fenestrated network of smooth tubules morphologically equivalent to the smooth ER of the hepatocyte (Benlimame et al. 1995; Wang et al. 1997). AMF-R tubules can be distinguished from the ER-Golgi intermediate compartment (ERGIC) by fluorescence microscopy and therefore represent a distinct smooth ER subdomain (Wang et al. 1997). We show here that smooth ER AMF-R tubules are intimately associated with mitochondria. Using a semipermeabilized cell assay, we further show that this association can be disrupted by cytosol in the presence of low cytosolic free Ca²⁺ levels. Elevated cytosolic free Ca²⁺ inhibits the ability of cytosol to dissociate the two organelles identifying a mechanism that facilitates the close association between the ER and mitochondria necessary for mitochondrial Ca²⁺ uptake.

MDCK II cells were grown in DME supplemented with 10% FCS, nonessential amino acids, vitamins, glutamine, and a penicillin–streptomycin antibiotic mixture in an air 5% CO₂ incubator at constant humidity.

Monoclonal IgM antibody against AMF-R was used in the form of concentrated hybridoma supernatant (Nabi et al. 1990). Polyclonal antibody to calreticulin was kindly provided by Dr. Luis Rokeach (Department of Biochemistry, Université de Montréal, Quebec, Canada) and to calnexin by Dr. John Bergeron (Department of Cell Biology and Anatomy, McGill University, Montreal, Quebec, Canada). Antibody to mitochondrial heat shock protein 70 (Mt-HSP70; clone JG1) was purchased from Affinity Bioreagents, Inc. Secondary antibodies conjugated to either Texas red or FITC were purchased from Jackson ImmunoResearch Laboratories. DTT was purchased from Sigma-Aldrich, and digitonin was purchased from ICN. The K₂H₂EGTA and K₂CaEGTA solutions (Calibration Buffer Concentrate Kit) used to prepare the CaEGTA buffers. 5,5′-dimethyl BAPTA-AM, BAPTA-AM, 5,5′-difluoro BAPTA-AM, and 5,5′-dibromo BAPTA-AM were purchased from Molecular Probes.

Immunofluorescence labeling of 80/20% (vol/vol) methanol/acetone fixed MDCK cells for AMF-R, calnexin, calreticulin, and Mt-HSP70, and postembedding immunoelectron microscopic labeling for AMF-R were performed as previously described (Benlimame et al. 1995; Wang et al. 1997). Confocal images were obtained with a BioRad MRC-600 confocal microscope and EM images with a Phillips 300 electron microscope. The minimal ER-mitochondria distance was quantified by measuring from 14 randomly taken EM images (21,000×), the distance between the point on each ER tubule, smooth or rough, labeled or not for AMF-R, closest to a mitochondrion within the same cell. Rough ER tubules were identified by the presence of a linear array of membrane-associated ribosomes.

Rat liver cytosol was prepared from cleaned rat liver cut into small pieces by scissors in buffer A (25 mM Tris, 25 mM KCl, 1 mM DTT, pH 7.4) supplemented with 85 mM sucrose and protease inhibitors at 4°C and homogenized at 4°C in 1 vol of the same buffer. The homogenate was centrifuged at 11,060 g for 20 min and the supernatant recentrifuged at 100,000 g for 90 min at 4°C. The supernatant was collected and assayed for protein concentration. Cytosol was heat-inactivated by boiling for 5 min.

MDCK cells plated at 25,000–50,000 cells/35-mm plate on glass cover slips for 2 d were quickly washed twice with cytoskeleton stabilizing buffer (CSB; 130 mM Hepes, 2 mM MgCl₂, 10 mM EGTA, pH 6.9) prewarmed to 20°C and then incubated with 10 μg/ml digitonin in buffer A for 1 min at 20°C. The cells were then washed with CSB and incubated in buffer A, supplemented as indicated in the text, for 1 h at 37°C. The cells were washed twice with CSB containing 0.2% BSA, fixed with methanol/acetone and labeled for AMF-R (Texas red) and Mt-HSP70 (FITC). The gain and black level of the confocal images were adjusted for each slide for both the FITC and rhodamine channels such that the full dynamic range of the instrument was used. For each experiment, confocal images from all slides were obtained in one sitting using equivalent pinhole settings and the extent of overlap of AMF-R tubule labeling with mitochondria was determined using Northern Eclipse software (Empix Imaging). The nonspecific AMF-R labeling of the nuclei was first cut from the image and the total number of cytoplasmic pixels labeled for AMF-R determined. Using a convolution filter, the mitochondrial labeling was enlarged to ensure that shifts between the FITC and Texas red images and cellular variations in the intensity of the mitochondrial image did not influence the measure of the extent of AMF-R tubule overlap with mitochondria. Using the enlarged mitochondrial labeling as a mask, AMF-R labeling that coincided with the enhanced mitochondrial labeling was deleted and quantification of the remaining AMF-R–labeled pixels provided a measure of smooth ER AMF-R tubules that had dissociated from mitochondria. This value was divided by the total AMF-R–labeled cytoplasmic pixels to generate the percent dissociation of AMF-R tubules from mitochondria.

To prepare the CaEGTA buffers, 100 mM stock solutions of K₂H₂EGTA and K₂CaEGTA were diluted in buffer A to 25 mM. The resulting buffers were then mixed to generate 8 different 25-mM CaEGTA buffers containing 0, 5, 10, 15, 20, 22.13, 22.92, and 25 mM total Ca²⁺ (see Table). Predicted free Ca²⁺ concentrations were calculated using WEBMAXCv1.10 at the CPatton MaxChelator website.

The free Ca²⁺ concentration of the Ca²⁺-EGTA buffer solutions ([Ca²⁺]_SOL) was estimated with the Ca²⁺ sensitive fluorophor Fura-2 used as a potassium salt at a concentration of 2.5 μM. Fluorescence measurements were performed using a dual-excitation spectrofluorometer (Spex Fluorolog II; Spex Industries Inc.). The excitation wavelengths were set at 350 and 380 nm and emission was monitored at 505 nm with a standard bandpass filter (Andover Corporation 500FS10).

[Ca²⁺]_SOL was calculated using the following equation (Grynkiewicz et al. 1985): [Ca²⁺]_SOL = K_d × (R − R_MIN)(R_MAX − R) × Sf₂/Sb₂; with K_d equal to 224 nM, R the ratio of the fluorescence measured at 350 and 380 nm, respectively, and Sf₂/Sb₂, the ratio of fluorescence at 380 nm in low and high Ca²⁺, respectively. The maximum fluorescence ratio (R_MAX) was determined using a 10 mM CaCl₂ solution to oversaturate the Fura-2, whereas R_MIN was obtained with a Ca²⁺ free solution containing 10 mM EGTA. Measurements of [Ca²⁺]_SOL were performed at a pH of 7.35 at room temperature (21–23°C).

Specific labeling of smooth ER tubules with the anti–AMF-R antibody has been previously reported in MDCK and other cell types by postembedding immunoelectron microscopy (Benlimame et al. 1995, Benlimame et al. 1998; Wang et al. 1997). Using confocal microscopy, we compared the distribution of smooth ER AMF-R tubules with that of the ER markers calnexin and calreticulin. As seen in Fig. 1, the majority of the AMF-R label can be distinguished from that of the calnexin- and calreticulin-labeled ER although partial overlap of AMF-R tubules with the ER can be observed. AMF-R–labeled tubules are always seen in proximity of calnexin- or calreticulin-labeled elements of the ER and frequently appear as projections of calnexin- or calreticulin-labeled tubules (Fig. 1C and Fig. F, arrowheads), consistent with AMF-R labeling of smooth projections of rough ER tubules observed by EM (Benlimame et al. 1995; Wang et al. 1997).

Double labeling of MDCK cells for AMF-R and Mt-HSP70 revealed a high degree of colocalization between the two signals (Fig. 2, A–C). In contrast, the calnexin- and calreticulin-labeled ER exhibits only partial overlap with the mitochondrial signal (Fig. 2, D–I). Postembedding EM labeling of smooth ER tubules for AMF-R is shown in Fig. 3. Smooth ER AMF-R tubules are observed to exhibit direct interaction with mitochondria and to associate with more than one mitochondrion (Fig. 3B and Fig. C). Quantification of the proximity of ER tubules to mitochondria on 14 EM micrographs of AMF-R–labeled MDCK cells taken in series with Fig. 3 A, revealed that AMF-R–labeled smooth ER tubules, and indeed smooth ER tubules in general, are more frequently observed within 200 nm of mitochondria than are rough ER tubules (Fig. 4). The fact that almost half of the AMF-R–labeled smooth ER tubules are located within 200 nm, the limit of optical resolution, of mitochondria in 80-nm ultrathin sections is consistent with the extensive colocalization of smooth ER AMF-R tubules with mitochondria observed by confocal microscopy. By subcellular fractionation, AMF-R is exclusively found within a high-speed membrane pellet and not within the 12,000 g mitochondrial pellet that contains Mt-HSP70, demonstrating that membranous smooth ER AMF-R tubules can be physically dissociated from mitochondria, similar to a previous study (Benlimame et al. 1995). Smooth ER AMF-R tubules are therefore a subdomain of the ER that is selectively associated with mitochondria.

To better study the relationship between AMF-R tubules and mitochondria, we established a semipermeabilized assay based on digitonin permeabilization of subconfluent cultures of MDCK cells. Under the conditions of digitonin permeabilization used, upwards of 90% of the cells were permeabilized as determined by the ability to label cells with antibodies to cytoplasmic keratin filaments before fixation. Antibodies to calreticulin and to the lumenal domain of LAMP-2 did not label digitonin-permeabilized cells, indicating that intracellular organelles remained intact. Labeling with antibodies to tubulin revealed that the majority of microtubules were no longer intact. The association between smooth ER AMF-R tubules and mitochondria is maintained in semiintact MDCK cells since the two organelles exhibit essentially complete overlap in digitonin-permeabilized cells incubated only with buffer A (Fig. 5A, Fig. B, and Fig. C). However, in cells incubated with buffer A supplemented with 5 mg/ml rat-liver cytosol, the AMF-R label was no longer completely associated with mitochondria and distinct regions of nonoverlap between the two organelles were observed (Fig. 5D, Fig. E, and Fig. F) demonstrating the ability of cytosol to induce the dissociation of smooth ER from mitochondria.

Using confocal images of cells double-labeled with anti–AMF-R (Texas red) and anti–Mt-HSP70 (FITC), a mask of the mitochondria labeling was generated and used to quantify the extent of overlap of AMF-R tubule labeling with mitochondria (see Materials and Methods). Using this protocol we obtained consistent data of the extent of dissociation of smooth ER AMF-R tubules from mitochondria (Fig. 6 A). In the presence of buffer A alone, the majority of AMF-R labeling overlapped with mitochondria and AMF-R labeling not associated with mitochondria was limited, normally from 5–10%. The addition of 5 mg/ml rat liver cytosol resulted in a significant increase in the extent of AMF-R labeling that did not overlap with mitochondria (40–50%). Heat inactivated cytosol was unable to dissociate mitochondria and smooth ER implicating cytosolic proteins in the dissociation activity. The ability of cytosol to dissociate AMF-R tubules and mitochondria is not energy dependent and occurs equally well in the absence or presence of 2 mM ATP, 2 mM GTP, or a creatine kinase based ATP regenerating system (ARS; 2 mM ATP, 37 U/ml creatine kinase, 8 mM creatine phosphate; data not shown).

The addition of millimolar concentrations of Ca²⁺ inhibited the ability of cytosol to dissociate AMF-R tubules from mitochondria (Fig. 6 B). The effect was concentration-dependent and the addition of 0.5 mM Ca²⁺ inhibited cytosol-induced dissociation by ∼90%. In the absence of cytosol, Ca²⁺ alone did not affect the extent of association between smooth ER and mitochondria, demonstrating that Ca²⁺ specifically regulates the cytosolic dissociation activity. To determine the role of free Ca²⁺ in the regulation of the cytosol-dependent dissociation of smooth ER and mitochondria, we used BAPTA Ca²⁺ chelators with varying K_ds for Ca²⁺ (Fig. 6 C). 5-5′-dimethyl BAPTA (K_d = 40 nM) reversed by 70%, and BAPTA (K_d = 160 nM) by ∼30% the ability of 0.5 mM Ca²⁺ to inhibit cytosol-dependent smooth ER-mitochondria dissociation. 5-5′-difluoro BAPTA (K_d = 635 nM) and 5-5′-dibromo BAPTA (K_d = 1.6 μM) reversed only minimally smooth ER–mitochondria dissociation. The ability of BAPTAs of various K_d's to differentially influence smooth ER mitochondria association demonstrates that free Ca²⁺ levels regulate ER–mitochondria association. The fact that these effects are observed in the presence of 100 μM BAPTA indicates that free Ca²⁺ in the presence of cytosol and cells must be below 100 μM, in spite of the fact that 0.5 mM Ca²⁺ was added to the reaction. Indeed, no effect was observed using 50 μM of the various BAPTAs, indicating that protein Ca²⁺ chelators in the cells and in the cytosol reduce free Ca²⁺ to between 50–100 μM.

To further define the free Ca²⁺ concentrations that regulate cytosol-mediated dissociation of smooth ER and mitochondria, we added cytosol in the presence of CaEGTA buffers of defined free Ca²⁺ concentrations. Free [Ca²⁺]_SOL was predicted using the WEBMAXCv1.10 program and measured using Fura-2 (Table). Compared with the high degree of dissociation observed in the presence of free [Ca²⁺]_SOL of less than 50 nM (CaEGTA buffers 2 and 3), free [Ca²⁺]_SOL of ∼100–400 nM (CaEGTA buffers 4 and 5) were associated with partial inhibition and free [Ca²⁺]_SOL concentrations >1 μM (CaEGTA buffers 6–8) associated with a significant inhibition of cytosol-mediated dissociation of smooth ER and mitochondria. Graphic representation of the data presented in Table (CaEGTA buffers 2–7) reveals a logarithmic relationship between free [Ca²⁺]_SOL measured with Fura-2 and cytosol dissociation activity (Fig. 6 D). In particular, free [Ca²⁺]_SOL below 100 nM are associated with extensive dissociation of the two organelles, whereas increasing the free [Ca²⁺]_SOL above 100 nM results in the progressive inhibition of cytosol-mediated dissociation.

The ER constitutes the largest intracellular membrane organelle of the cell, extends throughout the cytoplasm, and is composed of multiple functional domains (Sitia and Meldolesi 1992). Confocal images of MDCK cells double-labeled for AMF-R and calnexin or calreticulin show that while the AMF-R–labeled tubules can be clearly distinguished from the calnexin- and calreticulin-labeled ER, partial overlap is apparent and regions of close interaction between the two are evident. These images are therefore consistent with the identity of the AMF-R tubule as a smooth subdomain of the ER (Benlimame et al. 1995; Wang et al. 1997). The presence of AMF-R, the select sensitivity of smooth ER AMF-R tubules, but not the rough ER to ilimaquinone (Wang et al. 1997), and the fact that AMF-R labeling can be clearly distinguished from that of calnexin and calreticulin identify the smooth ER AMF-R tubule as a molecularly distinct subdomain of the ER. The high degree of association of AMF-R–labeled tubules with mitochondria by confocal and EM further identifies the AMF-R tubule as a mitochondria-associated subdomain of the ER.

The existence of ER subdomains that interact with mitochondria and which function to sequester cytosolic Ca²⁺ (Simpson et al. 1997; Rizzuto et al. 1998), together with the demonstrated role of cytosolic Ca²⁺ in regulating AMF-R tubule–mitochondria association, suggest a role for this smooth ER subdomain in the regulation of cytosolic Ca²⁺ levels. The diversity of Ca²⁺ concentrations and expression of Ca²⁺ binding proteins in different subdomains of the ER is well-established (Takei et al. 1992; Button and Eidsath 1996; Rooney and Meldolesi 1996; Golovina and Blaustein 1997; Montero et al. 1997; Pezzati et al. 1997). Not unlike the smooth ER AMF-R tubule, a smooth membrane-bound organelle of 50–250 nm apposed to the ER and mitochondria, called the calciosome, has been described to be enriched in a nonmuscle calsequestrin-like protein and to represent the major IP3-sensitive Ca²⁺ storage organelle (Volpe et al. 1988). Both AMF-R and the IP3 receptor are associated with cell surface caveolae (Fujimoto et al. 1992; Benlimame et al. 1998), and the caveolar-like clathrin-independent internalization pathway to smooth ER defined by AMF-R (Benlimame et al. 1998; Le et al. 2000) may serve to regulate the delivery of Ca²⁺ regulatory proteins to this mitochondria-associated smooth ER subdomain.

In semipermeabilized MDCK cells in the absence of cytosol or energy and independent of Ca²⁺, smooth ER AMF-R tubules and mitochondria maintain their close association, suggesting that interorganellar adhesion mechanisms are associated with the organelles. The ability of cytosol to dissociate the two organelles demonstrates the existence of a cytosolic mechanism that regulates ER–mitochondria association. The transfer of phosphatidyl serine between mitochondria and MAMs is prevented by proteolysis of mitochondrial surface proteins, but not of MAMs, suggesting that a mitochondrial surface protein is involved in the association between mitochondria and ER (Shiao et al. 1998). Whereas the relationship of smooth ER AMF-R tubules to MAMs remains unclear, cytosol may disrupt smooth ER AMF-R tubule–mitochondria association due to the presence of a Ca²⁺-regulated cytosolic protein that inhibits the adhesive function of such a mitochondrial surface protein.

Our results using BAPTAs of various affinities and CaEGTA buffers of varying free [Ca²⁺]_SOL indicate that free Ca²⁺ regulates the ability of cytosol to stimulate dissociation of AMF-R tubules from mitochondria. While we hesitate to attribute precise free Ca²⁺ levels to our data, both the BAPTA and CaEGTA experiments are consistent with the ability of low free [Ca²⁺]_SOL (<100 nM) to favor dissociation and of high free [Ca²⁺]_SOL (>1 μM) to favor association of smooth ER tubules and mitochondria (Fig. 6; Table). Therefore, these data argue that at physiological cytosolic Ca²⁺ levels, cytosol stimulates smooth ER-mitochondria dissociation. Whereas addition of cytosol increases the amount of smooth ER AMF-R tubules that do not colocalize with mitochondria, the two organelles still exhibit extensive association (Fig. 5 F) such that even in the presence of cytosol the two organelles still interact. Dissociation of AMF-R tubules from mitochondria induced by cytosol therefore reflects a partial disruption of direct contacts between the two organelles. The increased overlap of the two fluorescent signals observed in the presence of increasing Ca²⁺ may be indicative of the increased formation of close contacts between the two organelles.

Ca²⁺ transients of 1.6–1.9 μM have been identified at high density patches of the ER enriched in the IP3 receptor and calreticulin (Simpson et al. 1997), and the mitochondrial surface has been shown to be exposed to 3.4 μM Ca²⁺ after IP3 stimulation (Rizzuto et al. 1998). Our data showing increased ER–mitochondria association at millimolar cytosolic Ca²⁺ indicates that the elevated Ca²⁺ levels present at these ER–mitochondria close contacts can actively regulate ER–mitochondria association. Ca²⁺-dependent smooth ER–mitochondria association could therefore contribute to the establishment of segregated cytoplasmic domains between the two organelles in which the local elevated Ca²⁺ concentrations necessary for mitochondrial uptake can be established. A cellular mechanism that senses elevated cytosolic Ca²⁺ levels and then facilitates the establishment of ER–mitochondria contacts is therefore proposed to be implicated in the efficient sequestration of cytosolic Ca²⁺ to mitochondrial stores.

We thank Kevin Coonan and Nigel Banner of Empix Imaging (Mississuaga, Ontario) for developing the Northern Eclipse overlay program used to quantitate AMF-R tubule–mitochondria overlap. The photographic and digital reproductions were the work of Jean Leveillé and Gaston Lambert.

This study was supported by grants from the Medical Research Council of Canada.