Reversible interactions between smooth domains of the endoplasmic reticulum and mitochondria are regulated by physiological cytosolic Ca²⁺ levels

An association between the ER and mitochondria was first noted soon after electron microscopy enabled the morphological visualisation of the ER (Dempsey, 1953; Porter and Kallman, 1953). More recently, confocal imaging, using recombinant proteins specifically targeted to the ER and mitochondria, has confirmed the association of these two organelles and demonstrated the importance of this interaction in Ca²⁺ homeostasis (Montero et al., 2000; Rizzuto et al., 1993; Rizzuto et al., 1998; Simpson et al., 1997; Szabadkai et al., 2003). Mitochondria are able to sense domains of elevated Ca²⁺ concentrations generated by the IP3R and/or plasma membrane Ca²⁺ channel activation (Rizzuto et al., 1993; Rizzuto et al., 1998). These local elevated concentrations enable rapid mitochondrial accumulation of Ca²⁺ that stimulates mitochondrial metabolism (Montero et al., 2000). Mitochondrial Ca²⁺ uptake is greater when triggered either by quantal incrementation of the Ca²⁺ concentration (Csordas et al., 1999) or by sustained Ca²⁺ release from the ER (Szabadkai et al., 2003). Such increases might arise from the simultaneous activation of several Ca²⁺ channels, such as IP3R and the ryanodine receptor (RyR), clustered at sites of close apposition between the ER and mitochondrial membranes (Csordas et al., 1999). In addition, uptake of Ca²⁺ by mitochondria has been documented to feedback, through a currently undefined mechanism, on the ER Ca²⁺-release process by controlling the frequency or intensity of IP3R or RyR channel activation (Csordas et al., 1999; Jouaville et al., 1995; Straub et al., 2000; Zimmermann, 2000). Ca²⁺ exchange between the two organelles is therefore a finely tuned cellular process that impacts on cellular Ca²⁺ homeostasis as well as Ca²⁺ signaling.

Changes in ER morphology during cellular processes such as fertilisation and oocyte maturation affect its capacity to release free cytosolic Ca²⁺ ([Ca²⁺]_cyt) in response to increases in Ins(1,4,5)P₃ (Shiraishi et al., 1995; Terasaki et al., 1996; Terasaki et al., 2001). Other studies have shown that artificial increases of [Ca²⁺]_cyt dramatically reorganise the ER (Pedrosa Ribeiro et al., 2000; Subramanian and Meyer, 1997). The ER is heterogeneous with respect to its Ca²⁺ storage function owing in part to the presence of microdomains enriched in Ca²⁺-binding proteins (Papp et al., 2003). Mitochondria are also functionally heterogeneous (Collins et al., 2002; Park et al., 2001) and intracellular segregation of mitochondria is crucial for the spatial regulation of Ca²⁺ signaling and the control of secretion in pancreatic acinar cells and chromaffin cells (Montero et al., 2000; Tinel et al., 1999). The heterogeneous nature of the ER and mitochondria together with the importance of the ER-mitochondria association in regulating cellular Ca²⁺ dynamics suggest that specific, local interactions should necessarily respond to subtle changes in intracellular Ca²⁺ levels.

The 3F3A monoclonal antibody (mAb) raised against AMFR, also called gp78 (Nabi and Raz, 1987), stimulates cell motility and metastasis, mimicking ligand activation, and competes with autocrine mobility factor (AMF) for receptor binding (Nabi et al., 1990; Silletti et al., 1991). In cells transfected with FLAG-tagged AMFR, 3F3A labelling of AMFR is increased, but the label only partially colocalises with the exogenously expressed protein, indicating that it is identifying a subpopulation of cellular AMFR (Registre et al., 2004). We previously identified a smooth ER (SER) subdomain that is labelled with the 3F3A mAb against AMFR that selectively associates with mitochondria (Benlimame et al., 1995; Goetz and Nabi, 2006; Wang et al., 1997; Wang et al., 2000). The association between this SER subdomain and mitochondria in digitonin-permeabilised cells is dependent on [Ca²⁺]_cyt (Wang et al., 2000); however, the nature of this interaction in intact cells has yet to be determined. Here, we show that the 3F3A-labelled ER domain is distinct from the tubular ER labelled for reticulon (Voeltz et al., 2006), does not label the nuclear membrane and only partially colocalises with the rough ER marker Sec61α. Importantly, we demonstrate that physiological modulation of [Ca²⁺]_cyt induces the reversible dissociation of this ER domain from mitochondria.

The 3F3A mAb was originally generated against purified gp78/AMFR (Nabi et al., 1990). Judged by immunoelectron microscopy, in addition to plasma membrane labelling, the 3F3A mAb also labels smooth ER tubules that frequently extend from the ribosome-studded rough ER (Benlimame et al., 1998; Benlimame et al., 1995; Wang et al., 1997). In cells rapidly fixed with very cold methanol-acetone, 3F3A immunofluorescent labelling defines an ER domain distinct from the calnexin- and calreticulin-labelled ER and ER-Golgi intermediate compartment (ERGIC) (Benlimame et al., 1998; Benlimame et al., 1995; Wang et al., 1997; Wang et al., 2000). 3F3A labelling is highly stable and persists after 16 hours of cycloheximide treatment (Benlimame et al., 1995). We therefore used a Dicer siRNA approach to knockdown AMFR, resulting in an ∼60-70% reduction in 3F3A labelling judged by western blot. By quantitative immunofluorescence, we observed a similar reduction in 3F3A labelling of the mitochondria-associated ER domain, demonstrating that 3F3A labelling of ER tubules is specific for gp78/AMFR (Fig. 1).

In contrast to the distribution of 3F3A labelling, which is restricted to mitochondria-associated ER tubules, exogenously expressed FLAG- or GFP-tagged AMFR is localised throughout the ER (Fang et al., 2001; Registre et al., 2004). Reticulon4a/NogoA (Rtn4a) defines and maintains tubular ER domains preferentially associated with the peripheral ER, as opposed to the saccular or perinuclear ER that includes the nuclear membrane (Voeltz et al., 2006). In Cos7 cells, the 3F3A-labelled ER does not localise with transfected MYC-tagged Rtn4a or the perinuclear GFP-Sec61β-expressing ER and 3F3A does not label the nuclear membrane (Fig. 2A). 3F3A-labelled tubules do, however, exhibit partial colocalisation with the translocon component Sec61α (Fig. 2B). Upon transfection of FLAG-tagged AMFR, the antibody against FLAG, but not 3F3A, labels the nuclear membrane (Fig. 2C) and colocalises with the ER marker calnexin (data not shown). In the cell periphery, anti-FLAG and 3F3A labelling colocalise extensively and show only minimal association with Sec61α (Fig. 2C). They also show extensive colocalisation with Rtn4a (Fig. 2D). Importantly, both anti-FLAG and 3F3A still label mitochondria-associated ER domains (Fig. 2E).

In digitonin-permeabilised cells, association of the 3F3A-labelled ER domain and mitochondria is favoured at high [Ca²⁺]_cyt and reduced at low [Ca²⁺]_cyt (Wang et al., 2000). To study this interaction in intact cells, we modulated [Ca²⁺]_cyt by using the Ca²⁺ ionophore ionomycin in the presence of varying extracellular Ca²⁺ concentrations ([Ca²⁺]_ex). While we were able to disrupt the morphology of the calnexin-labelled ER in MDCK cells exposed to 10 μM ionomycin in 10 mM extracellular Ca²⁺ ([Ca²⁺]_ex; data not shown), these conditions were found to be highly toxic and resulted in cell rounding and detachment. Reducing the ionomycin concentration led to conditions that selectively affected the 3F3A-labelled SER domain and not the calnexin-labelled ER. As demonstrated in Fig. 3, 3F3A-labelled SER tubules remained associated with mitochondria in MDCK cells incubated with 1 μM ionomycin and 200 μM EGTA (Fig. 3A-C). Increasing [Ca²⁺]_ex to 1 mM disrupted the tubular pattern of the 3F3A-labelled ER, which became punctate, dispersed throughout the cell and exhibited reduced colocalisation with mitochondria (Fig. 3D-F). Increasing [Ca²⁺]_ex to 10 mM resulted in mitochondrial rounding and close association with the 3F3A-labelled SER domain. Relative to total cytoplasmic 3F3A pixel intensity, the intensity of 3F3A-labelled pixels that do not overlap with mitochondria provides a relative measure of dissociation of this SER domain from mitochondria. In parallel, [Ca²⁺]_cyt was measured using the cell-permeable Ca²⁺ probe Fura-2-AM (Fig. 3J). Cells incubated with 1 μM ionomycin in a buffer containing EGTA, essentially R_min, presented no measurable [Ca²⁺]_cyt. The extensive dissociation of the two organelles detected in the presence of 1 μM ionomycin and 1 mM [Ca²⁺]_ex corresponded to [Ca²⁺]_cyt of 115±17 nM. In the presence of 1 μM ionomycin and 10 mM [Ca²⁺]_ex, [Ca²⁺]_cyt increased to 1651±62 nM and resulted in the reassociation of the two organelles (Fig. 3J).

Thapsigargin is a specific and irreversible inhibitor of the sarco/endoplasmic reticulum Ca²⁺-ATPases (SERCA), and its application results in the depletion of intracellular ER Ca²⁺ stores (Thastrup et al., 1990). MDCK cells treated with 10 μM thapsigargin in a buffer containing 200 μM EGTA present a dissociation of the 3F3A-labelled SER domain from mitochondria similar to that of cells treated with 1 μM ionomycin and 1 mM [Ca²⁺]_ex (Fig. 3K). We did not observe alteration in the distribution of the calnexin-labelled ER under these conditions (data not shown), as previously reported (Pedrosa Ribeiro et al., 2000). Incubation of the cells in an EGTA-containing buffer in the presence of both 10 μM thapsigargin and 1 μM ionomycin also resulted in SER tubule-mitochondria dissociation after 20 minutes (Fig. 3K). The [Ca²⁺]_cyt of cells treated with thapsigargin alone for 20 minutes was measured to be 211±42 nM, and the concomitant addition of 1 μM ionomycin and 10 μM thapsigargin caused a slight reduction of [Ca²⁺]_cyt to 148±19 nM owing to the efflux of some Ca²⁺ into the extracellular medium. Incubation for 60 minutes in 1 μM ionomycin plus 10 μM thapsigargin-EGTA resulted in a further decrease of [Ca²⁺]_cyt to 35±7 nM and reassociation of the 3F3A-labelled SER tubules and mitochondria (Fig. 3K). Therefore, depletion of free ER Ca²⁺ with thapsigargin does not modify the response of the 3F3A-labelled SER domain to a reduction in [Ca²⁺]_cyt.

Using electron microscopy, cells incubated with 1 μM ionomycin in EGTA buffer displayed elongated smooth and rough ER tubules, the latter defined by the presence of a linear array of membrane-bound ribosomes, often closely associated with mitochondria (Fig. 4A,E). Incubation of cells with 1 μM ionomycin and 1 mM [Ca²⁺]_ex reduced the association between SER tubules and mitochondria (Fig. 4B,F). Increasing [Ca²⁺]_ex to 10 mM increased the association between mitochondria and ER (Fig. 4C,G). The shortest distance of ER tubules from the nearest mitochondria was measured from cells plated on plastic, fixed, scraped and embedded as a cell pellet (Wang et al., 2000) and from cells grown on filters and fixed and embedded in situ. As demonstrated in Fig. 4H,I, the minimal distance of each ER tubule from mitochondria varies greatly. Nevertheless, a cluster of ER tubules in close proximity (<50 nm) to mitochondria was observed predominantly for SER tubules in cells treated with 1 μM ionomycin and EGTA and disrupted in cells treated with 1 μM ionomycin and 1 μM [Ca²⁺]_ex. Addition of 10 mM [Ca²⁺]_ex enhanced the number of both smooth and rough ER tubules in close proximity to mitochondria. Similar results were obtained for cells fixed as pellets (n=3) and on filters (n=2). We therefore counted the number of ER tubules within 50 nm of individual mitochondria and combined data from all five experiments. The number of SER tubules in proximity to mitochondria was significantly greater (4-5 fold) than RER tubules in the presence of 1 μM ionomycin plus EGTA and selectively reduced upon treatment with 1 μM ionomycin plus 1 mM [Ca²⁺]_ex. Treatment with ionomycin plus 10 mM [Ca²⁺]_ex resulted in increased association of both SER and RER tubules with mitochondria (Fig. 4J). These results confirm the Ca²⁺ sensitivity of the SER tubule-mitochondria association detected by 3F3A labelling of AMFR (Fig. 3) and identify 3F3A labelling as a valid marker for a Ca²⁺-sensitive mitochondria-associated SER domain.

ATP is a physiological agonist that stimulates purinergic receptors in the plasma membrane, thereby inducing production of Ins(1,4,5)P₃ and IP3R activation, resulting in a transient increase in [Ca²⁺]_cyt (Jan et al., 1999). Upon application of 10 μM ATP to MDCK cells, [Ca²⁺]_cyt measured with Fura-2-AM reached 190±13 nM in less than 20 seconds and then rapidly decreased to a plateau in the 60 nM range after 1 minute (Fig. 5J). The transient increase was followed at 2 minutes by dissociation of the 3F3A-labelled SER from mitochondria (Fig. 5A-F′), comparable to what we observed in the presence of ionomycin plus 1 mM [Ca²⁺]_ex or 10 μM thapsigargin (Fig. 3). After 5 minutes, the mitochondrial association of 3F3A-labelled SER tubules was restored (Fig. 5G-I′). Quantification of the extent of 3F3A SER tubule-mitochondria association shows clearly that transient dissociation of the two organelles follows the IP3R-mediated Ca²⁺ transient (Fig. 5J). Pretreatment of MDCK cells with 2 μM xestospongin C, an inhibitor of IP3R (Gafni et al., 1997), for 15 minutes at 37°C before ATP application, limited the ATP-induced [Ca²⁺]_cyt increase to below 65 nM. In the presence of xestopongin C, basal dissociation levels were reduced relative to untreated cells and the ATP-induced dissociation of the two organelles was prevented (Fig. 5J). Dissociation of 3F3A-labelled SER tubules and mitochondria is therefore a physiological response to ATP-stimulated release of Ins(1,4,5)P₃-sensitive Ca²⁺ pools.

In MDCK cells, the ER tubules labelled by 3F3A immunoelectron microscopy include smooth extensions of RER tubules as determined by electron microscopy (Benlimame et al., 1995; Wang et al., 1997). Continuities between the 3F3A- and calnexin-labelled ER are observed under conditions where the 3F3A-labelled SER domain remains associated with mitochondria (Fig. 6A-B). After a two minute ATP treatment and dissociation of the 3F3A-labelled SER domain from mitochondria, the fragmented 3F3A labelling remains associated, but does not overlap, with the calnexin-labelled ER (Fig. 6C-D). SERCA colocalises with 3F3A labelling even following ATP-mediated dissociation from mitochondria (Fig. 6E-H). Upon changes in [Ca²⁺]_cyt, 3F3A-labelled ER elements therefore retain SERCA but do not mix with the calnexin-labelled ER.

Labelling for the IP3R exhibited extensive overlap with mitochondria-associated 3F3A-labelled SER tubules (Fig. 7), consistent with the previously reported distribution of IP3R to the SER and its presence in a mitochondria-associated ER domain (Rizzuto et al., 1993; Ross et al., 1989; Sharp et al., 1992; Takei et al., 1992). Quantification of the extent of overlap between the two ER domains by using the Pearson coefficient confirmed that the 3F3A- and IP3R-labelled ER domains colocalise with mitochondria (Fig. 7C). After 2 minutes ATP treatment, or 1 μM ionomycin and 1 mM [Ca²⁺]_ex (data not shown), the 3F3A- and the IP3R-labelled SER exhibited reduced overlap with mitochondria, which was restored after 5 minutes of treatment (Fig. 7A,C). Dissociation of those two SER domains from mitochondria was also associated with reduced colocalisation between 3F3A and IP3R (Fig. 7B,C).

Expression of both AMF and AMFR is generally upregulated in tumour cells, resulting in autocrine activation of this motility factor receptor (Yanagawa et al., 2004). By western blot, AMFR expression is increased in RAS-transformed NIH-3T3 cells relative to untransformed NIH-3T3 cells (Fig. 8A). Similarly, quantitative immunofluorescence shows a significant increase in 3F3A labelling of AMFR in RAS-transformed NIH-3T3 cells (Fig. 8A). By immunofluorescence, 3F3A labelling in RAS-transformed NIH-3T3 cells no longer presents a tubular mitochondrial-associated morphology and is dispersed throughout the cytoplasm, extending to the pseudopodia of these cells (Fig. 8B). Dissociation of the 3F3A-labelled ER domain from mitochondria corresponds to an increase in [Ca²⁺]_cyt from ∼40 nM to ∼100 nM, similar to the [Ca²⁺]_cyt ranges associated with changes in SER tubule-mitochondria association in MDCK cells (Fig. 8B,C). In the presence of ionomycin and 1 mM [Ca²⁺]_ex, 3F3A-labelled SER tubule-mitochondria dissociation was observed in NIH-3T3 cells in a fashion similar to that of MDCK cells. Depletion of [Ca²⁺]_cyt by incubation of RAS-transformed NIH-3T3 cells with ionomycin and EGTA buffer resulted in increased overlap of the 3F3A-labelled ER domain and mitochondria (Fig. 8D).

As observed for MDCK cells, both 3F3A labelling and IP3R labelling colocalised extensively with mitochondria in NIH-3T3 cells (Fig. 9). In RAS-transformed NIH-3T3 cells, the IP3R-labelled ER remained associated with mitochondria and showed reduced dissociation from mitochondria relative to the 3F3A-labelled ER, and these two ER domains showed limited overlap. This suggests that 3F3A and IP3R labelling might define ER domains that differentially associate with mitochondria.

The ER is a continuum of multiple domains, including the morphologically distinct SER and ribosome-studded RER. Recent work identified a group of integral membrane proteins, the reticulon and DP1 protein families, responsible for the high membrane curvature observed in ER tubules (Voeltz et al., 2006). Reticulon and DP1 do not label the nuclear membrane and segregate from the perinuclear ER, containing translocon components such as Sec61. The 3F3A mAb against AMFR has been characterised as a marker for a mitochondria-associated SER domain (Benlimame et al., 1998; Benlimame et al., 1995; Wang et al., 1997; Wang et al., 2000). Here, we show that 3F3A labelling of SER tubules diminishes upon AMFR siRNA treatment, does not overlap with perinuclear ER markers, such as GFP-Sec61β or calnexin, does not label the nuclear envelope and does not colocalise with reticulon-labelled ER tubules. Interestingly, partial colocalisation of 3F3A with endogenous Sec61α is observed, consistent with immunoelectron microscopy 3F3A labelling of part-rough part-smooth ER tubules and smooth extensions of the rough ER (Benlimame et al., 1998; Benlimame et al., 1995; Wang et al., 1997). Drug treatment and overexpression of select ER resident proteins, such as GFP-Sec61β, can induce the formation of organised SER domains connected to the RER (Snapp et al., 2003; Sprocati et al., 2006; Wang et al., 1997). Interestingly, 3F3A does not overlap with overexpressed GFP-Sec61, suggesting that the 3F3A-labelled SER domain is distinct from the stacked smooth ER structures induced by overexpression of GFP-Sec61 (Snapp et al., 2003). However, upon overexpression of FLAG-AMFR, the 3F3A mAb labels the peripheral, reticulon-expressing tubular ER network in addition to the mitochondria-associated ER domain (Fig. 2), indicative of a relationship between these two ER domains.

The 3F3A antibody against AMFR therefore appears to recognise an AMFR conformation distinct from that of newly synthesised AMFR protein retained in the ER, thereby explaining the distinct ER distribution of 3F3A labelling and transfected AMFR (Fang et al., 2001; Registre et al., 2004). AMFR is an E3 ubiquitin ligase that interacts with components of the ER-associated degradation (ERAD) complex such as the VCP/p97 ATPase and derlin and is implicated in the degradation of various substrates, including CD3-delta, the T-cell receptor, ApoB lipoprotein and HMG CoA reductase (Fang et al., 2001; Liang et al., 2003; Song et al., 2005; Ye et al., 2005; Zhong et al., 2004). The structural and functional specialisation of the AMFR conformation within the mitochondria-associated SER might be related to its function in ERAD. In this study, we define the Ca²⁺-sensitive interaction of the 3F3A-labelled ER domain with mitochondria.

As reported previously, in digitonin-permeabilised cells (Wang et al., 2000), intermediate (100-200 nM) [Ca²⁺]_cyt levels obtained by cell treatment with ionomycin under various [Ca²⁺]_ex conditions were found to disrupt the SER tubule-mitochondria complex, whereas high (>1500 nM) levels tended to favour complex formation in intact cells. By electron microscopy, SER tubules were found preferentially located within 50 nm of mitochondria in single sections (Fig. 4), consistent with the existence of direct linkages between the two organelles (Csordas et al., 2006) and with our previous report of the close association of mitochondria with 3F3A-labelled SER tubules (Wang et al., 2000). Electron microscopy of ionomycin-treated MDCK cells paralleled the results obtained by 3F3A immunofluorescence, showing reduced SER tubule-mitochondria association at intermediate levels of [Ca²⁺]_cyt. We also observed an increase in the number of RER tubules in proximity to mitochondria at high [Ca²⁺]_cyt, consistent with previously reported effects of elevated [Ca²⁺]_cyt on general ER morphology (Pedrosa Ribeiro et al., 2000; Subramanian and Meyer, 1997). It is important to note that SER and RER in cultured cells do not present the dramatic morphological differences observed in tissue. Indeed, we define ER tubules as rough based on the presence of a linear array of more than three membrane-bound ribosomes (Benlimame et al., 1998; Benlimame et al., 1995). Using this approach, our data argue that smooth ER tubules lacking ribosomes, unlike ribosome-studded rough ER tubules, show a preferential Ca²⁺-dependent association with mitochondria. Whether SER and RER in tissues present differential mitochondria association remains to be determined.

Levels of [Ca²⁺]_cyt between 110 and 210 nM were associated consistently with SER tubule-mitochondria dissociation, whereas levels of [Ca²⁺]_cyt below 60 nM were found to correlate with SER tubule-mitochondria association. The two organelles were observed to reassociate when [Ca²⁺]_cyt was decreased in MDCK cells treated with ionomycin and thapsigargin or following the ATP-induced Ca²⁺ transient or in RAS-transformed NIH-3T3 cells treated with ionomycin-EGTA. The SER tubule-mitochondria interaction is therefore a reversible process closely tied to [Ca²⁺]_cyt. Reduced [Ca²⁺]_cyt was not associated with SER tubule-mitochondria association in digitonin-permeabilised cells (Wang et al., 2000), suggesting that the mechanisms underlying SER tubule-mitochondria association under conditions of high or low [Ca²⁺]_cyt are distinct. Those prevailing at low [Ca²⁺]_cyt might not be functional in cytosol-replenished digitonin-permeabilised cells as a result of disruption of the cytoskeletal network and because of the overall cellular architecture (Wang et al., 2000).

The temporal dissociation of SER tubules and mitochondria following ATP-induced release of Ca²⁺ from Ins(1,4,5)P₃-sensitive Ca²⁺ pools clearly demonstrates that dissociation of the two organelles is a physiological response to changes in [Ca²⁺]_cyt. The specific role of Ca²⁺ release from ER stores was shown by the ability of xestospongin C, a specific inhibitor of Ins(1,4,5)P₃ activation (Gafni et al., 1997), to prevent SER tubule-mitochondria dissociation. The SER tubule-mitochondria reassociation observed following extended incubation with thapsigargin and ionomycin (Fig. 3) indicates that depletion of free Ca²⁺ from the ER does not modify the response of 3F3A-labelled SER tubules to changes in [Ca²⁺]_cyt. The interaction of this SER domain with mitochondria can therefore respond to changes in [Ca²⁺]_cyt independently of the replenishment state of ER Ca²⁺ stores.

Both the 3F3A- and IP3R-labelled ER domains dissociated from mitochondria upon stimulation with ATP (Fig. 7). This is consistent with the overall reduction in SER tubule-mitochondria association observed upon elevation of [Ca²⁺]_cyt by electron microscopy (Fig. 4). Compared with the IP3R-labelled ER domain, the 3F3A ER domain shows increased dissociation from mitochondria in RAS-transformed NIH-3T3 cells relative to nontransformed NIH-3T3 cells. IP3R gating occurs at a [Ca²⁺]_cyt of 200-300 nM (Bezprozvanny et al., 1991; Hirata et al., 1984; Iino, 1990). This is slightly higher than reported here for mitochondrial dissociation of the 3F3A SER domain (100-200 nM) but sufficient for increased clustering of IP3R in the ER adjacent to mitochondria (Shuai and Jung, 2003; Wilson et al., 1998). Mitochondrial and ER mobility are controlled by physiological [Ca²⁺]_cyt, with an IC₅₀ of ∼400 nM, that might recruit these organelles to cell regions of elevated [Ca²⁺]_cyt (Brough et al., 2005; Yi et al., 2004). While comparison of absolute [Ca²⁺]_cyt based on Fura-2 measurements must be done with caution, local, physiological changes in [Ca²⁺]_cyt would appear to have variable effects on the integrity and mitochondrial interaction of SER domains. Together with the segregation of the IP3R- and 3F3A-labelled ER domains upon elevation of [Ca²⁺]_cyt, this suggests that multiple mechanisms regulate the association of SER tubules with mitochondria.

PACS-2 is a sorting protein that regulates the amount of MAM-localised lipid-synthesising enzymes, ER homeostasis, Ca²⁺ signalling and the translocation to mitochondria of the pro-apoptotic factor Bid following an apoptotic stimulus (Simmen et al., 2005). The voltage-dependent anion channel (VDAC or porin) has been proposed as the mitochondrial partner for IP3R-mediated Ca²⁺ transport from ER to mitochondria; VDAC expression has been localised to both ER and mitochondrial outer membranes and is enriched in the zone of apposition, where it enhances Ca²⁺ transfer to mitochondria (Hajnoczky et al., 2002; Rapizzi et al., 2002; Shoshan-Barmatz et al., 2004; Szabadkai et al., 2006). AMFR-dependent recruitment of the VCP/p97 ATPase and its p47 cofactor, implicated in ER membrane fusion (Kano et al., 2005; Roy et al., 2000), might impact on the integrity and morphology of mitochondria-associated SER domains. Whether [Ca²⁺]_cyt impacts on specific mechanisms of ER-mitochondria interaction or generally affects the interaction between ER and mitochondria by impacting on ER domain morphology remains to be determined.

The 3F3A rat IgM mAb against gp78/AMFR was as described (Nabi et al., 1990). Antibodies to mitochondrial heat shock protein 70 (mtHSP70; clone JG1) and SERCA-2 were purchased from Affinity Bioreagents, to calnexin from Sigma, to IP3R from Calbiochem, to the mitochondrial ATP-synthase (Complex V, Subunit α) from Molecular Probes and to Sec61α from Upstate. Cross-absorbed secondary antibodies conjugated to FITC (Rat IgM) were purchased from Jackson ImmunoResearch Laboratories and to Alexa (568, 647) from Molecular Probes. Plasmid for expression of FLAG-AMFR was as previously described (Registre et al., 2004) and for expression of Rtn4a-MYC and GFP-Sec61β (Voeltz et al., 2006) were kindly provided by G. Voeltz and T. Rapoport (Harvard University, MA). Xestospongin C was from Calbiochem and ionomycin, thapsigargin, ATP, Fura-2-AM and other chemical reagents from Sigma.

MDCK (Wang et al., 2000), Cos7 (Registre et al., 2004), NIH-3T3 and RAS-transformed NIH-3T3 cells (Le et al., 2002) were grown as previously described. Cos7 cells were transfected using Effectene transfection reagent (Qiagen) and fixed 24 hours after transfection (Registre et al., 2004). Ionomycin and thapsigargin were used at 1 μM and 10 μM, respectively, in a 125 mM NaCl, 20 mM HEPES, 5 mM KCl, 1.5 mM MgCl₂ and 10 mM glucose buffer adjusted to pH 7.4 (normal buffer). Normal buffer was supplemented with 1 or 10 mM Ca²⁺, or 200 μM EGTA as indicated. For ATP treatments, cells were incubated with normal buffer containing 10 μM ATP and, where indicated, pretreated with 2 μM xestospongin C in culture medium for 15 minutes at 37°C.

About 815 bp of C-terminal AMFR was amplified by PCR using the following set of primers, AMFR-C-For (5′ AGA CAC CTC CTG TCC AAC 3′), AMFR-C-Rev (5′ GGA GGT CTG CTG CTT CTG 3′), and TA cloned. Positive clones were screened for the cis and trans orientation of fragments. Using T7 RNA polymerase (Invitrogen) and linearised cis and trans positive clones, complementary mRNAs were synthesised in vitro. These mRNA molecules were annealed, diced and the resulting 19-21-bp d-siAMFR molecules were purified using Block-it RNAi kit (Invitrogen). Transfection of NIH-3T3 with d-siAMFR or control siRNA was performed using Lipofectamine 2000 (Invitrogen).

Two methods were used to process cells for ultrastructural analysis. Cells grown on Petri dishes were washed rapidly twice with 0.1 M sodium cacodylate (pH 7.3) before fixing in the same solution containing 2% glutaraldehyde for 60 minutes at room temperature. The fixed cells were rinsed in 0.1 M sodium cacodylate, scraped from the Petri dish and collected by centrifugation. The cell pellet was post-fixed for 1 hour with 2% osmium tetroxide in sodium cacodylate buffer, dehydrated and embedded in LR-White resin. The sections were stained with 2% uranyl acetate and visualised in a Hitachi H7600 transmission electron microscope. Alternatively, cells were grown on polycarbonate filters, fixed in 0.1 M sodium cacodylate, 1.5% paraformaldehyde and 1.5% glutaraldehyde (pH 7.3), post-fixed for 1 hour on ice in 1% osmium tetroxide in 0.1 M sodium cacodylate (pH 7.3), stained for 1 hour with 0.1% uranyl acetate and then dehydrated. The filters were left in a 1:1 solution of propylene oxide:Polybed overnight and then embedded in 100% Polybed. Sections were viewed and photographed on a Philips 300 electron microscope operating at 60 kV.

MDCK cells plated for 2 days were fixed by the direct addition of pre-cooled (–80°C) methanol-acetone (80-20%, vol-vol) and labelled as described previously (Wang et al., 1997). Endogenous Sec61α labelling was performed following treatment with 0.1 mg/ml RNAse A for 30 minutes at room temperature. No labelling was observed in the absence of RNAse treatment. Confocal images were obtained using 488, 568 and 633 laser line excitation with 63× (NA 1.4) and 100× (NA 1.3) or 60× (NA 1.4) and 100× (NA 1.35) planapochromat objectives of Leica TCS SP1 and Olympus FV1000 confocal microscopes, respectively. Mask overlay quantification of 3F3A labelling with mitochondria was determined from 15 8-bit confocal images of mtHSP70 and 3F3A labelling acquired just below saturation with the 100× objective at zoom 2 and quantified with Northern Eclipse software (Empix Imaging, Mississauga, Ontario, Canada), as described previously (Wang et al., 2000). Complete cells within the field were selected and non-specific nuclear 3F3A labelling (Benlimame et al., 1995) cut from both images. The mtHSP70 image was saturated (0-155 grey levels) to ensure that mitochondria-associated 3F3A labelling was covered by the mitochondrial mask and the 3F3A image thresholded (50-255 grey levels) to eliminate background pixels. The ratio of 3F3A pixel intensity after mask subtraction relative to total cytoplasmic 3F3A pixel intensity provides a relative measure of dissociation of the SER from mitochondria for samples for which cell treatment, labelling and image acquisition were performed in parallel. For IP3R labelling, 12-bit confocal images were acquired with an Olympus FV1000 confocal microscope using the same acquisition parameters, determined using the Hi-Lo function, for all samples and Pearson's coefficients determined using ImagePro image analysis software (Media Cybernetics).

[Ca²⁺]_cyt was estimated at pH 7.4 at room temperature using the Ca²⁺-sensitive fluorophore Fura-2-AM (2 μM potassium salt) and a dual-excitation spectrofluorometer (Spex Industries) with excitation at 350 and 380 nm and emission at 505 nm, and [Ca²⁺]_cyt was calculated as described previously (Grynkiewicz et al., 1985). The maximum fluorescence ratio (R_MAX) was determined using a 10 mM CaCl₂, 10 μM ionomycin (pH 8) solution to saturate Fura-2, and R_MIN with a Ca²⁺-free solution containing 2.5 mM EGTA and 10 μM ionomycin (pH 8).

This study was supported by a grant from the Canadian Institutes of Health Research (MT-15132). I. R. Nabi is an Investigator of the Canadian Institutes of Health Research. J.G.G. holds a doctoral fellowship from the Ministère de la Recherche et des Technologies for his doctoral studies to be submitted jointly to the Université de Montréal and the Université Louis Pasteur de Strasbourg (UMR CNRS 7175).