High Ca2+ content in the Golgi apparatus (Go) is essential for protein processing and sorting. In addition, the Go can shape the cytosolic Ca2+ signals by releasing or sequestering Ca2+. We generated two new aequorin-based Ca2+ probes to specifically measure Ca2+ in the cis/cis-to-medial-Go (cGo) or the trans-Go (tGo). Ca2+ homoeostasis in these compartments and in the endoplasmic reticulum (ER) has been studied and compared. Moreover, the relative size of each subcompartment was estimated from aequorin consumption. We found that the cGo accumulates Ca2+ to high concentrations (150–300 μM) through the sarco plasmic/endoplasmic reticulum Ca2+-ATPase (SERCA). The tGo, in turn, is divided into two subcompartments: tGo1 and tGo2. The subcompartment tGo1 contains 20% of the aequorin and has a high internal [Ca2+]; Ca2+ is accumulated in this subcompartment via the secretory pathway Ca2+-ATPase 1 (SPCA-1) at a very high affinity (K50=30 nM). The subcompartment tGo2 contains 80% of aequorin, has a lower [Ca2+] and no SPCA-1 activity; Ca2+ uptake happens through SERCA and is slower than in tGo1. The two tGo subcompartments, tGo1 and tGo2, are diffusionally isolated. Inositol trisphosphate mobilizes Ca2+ from the cGo and tGo2, but not from tGo1, whereas caffeine releases Ca2+ from all the Golgi regions, and nicotinic acid dinucleotide phosphate and cADP ribose from none.
The Golgi apparatus (Go) plays a prominent role in the processing and sorting of lipids and proteins and allocates their final destination. Proteins are synthesized and packaged inside transport vesicles in the endoplasmic reticulum (ER), which then reach the cis-Go (cGo) and progress distally through the medial-Go to the trans-Go (tGo). In the tGo they are sorted and packaged into cargo vesicles either for constitutive or regulated secretion, or for being moved to lysosomes and other intracellular destinations. During their passage through the Go, proteins are modified according to their fate and function [1,2].
Studies using the Ca2+ probe aequorin targeted to the Go have suggested that the internal Ca2+ concentration in the Go ([Ca2+]Go) is high, around 300 μM , and that the Go may contribute to Ca2+ signalling by its release via an inositol 1,4,5-trisphosphate (InsP3)-mediated mechanism . The Go is a heterogeneous compartment, not only morphologically but also functionally, with respect to Ca2+ handling . Two different Ca2+-pumping ATPases contribute to the accumulation of Ca2+ within the Go: the sarco plasmic/endoplasmic reticulum Ca2+-ATPase (SERCA) and the secretory pathway Ca2+-ATPase (SPCA) . SPCA expression is restricted to the tGo , where it may be the only Ca2+-ATPase present . InsP3 cannot release Ca2+ from the tGo [8,9]. On the other hand, it has recently been proposed that Ca2+ is accumulated through both SERCA and SPCA in the medial-Go, and that InsP3 results in calcium release .
The contribution of the Go to the generation of cytosolic Ca2+ signals seems to be quantitatively less important than that of the ER [4,5,11]. The Go has been reported to contain only 2.5–5% of the total cellular Ca2+ content , and only a part of it can be mobilized by InsP3-producing agonists [3,4]. It has been speculated that, because of its physical proximity to the nuclear envelope, the Go may participate in nuclear Ca2+ signalling . In addition, maintenance of a high [Ca2+] in the tGo seems necessary for protein processing, and SPCA-1 knockdown caused missorting of secretory cargo, which was rescued by over-expression of tGo-localized SPCA-1 [12,13].
Previous studies on Ca2+ handling by the Go have reported conflicting results [3,8,10]. The discrepancies may be due to: (i) differences in Ca2+-measuring techniques, using either aequorins or fluorescent sensors; (ii) mistargeting of the probes due to over-expression; or (iii) the different cell models used. Therefore, a comparative study of Ca2+ handling in the Go subcompartments, using the same methodology and cell type, and in response to identical stimuli, is needed. Furthermore, quantitative data on the relative size of the different compartments are required to decide whether Ca2+ transport by the Go is relevant only for Go homoeostasis or whether it may also contribute significantly to cytosolic Ca2+ signalling.
In the present study, we directly compared Ca2+ handling by the cGo and the tGo through the use of low-affinity aequorins specifically targeted to each compartment. As our probes were also fluorescent, it was possible to assess the correct subcellular distribution before the functional experiments. In addition, we explored differential Ca2+ handling in each subcompartment by either inhibiting the different Go Ca2+ pumps and/or silencing SPCA-1 expression with siRNA. Furthermore, we measured the relative size of the different subcompartments from aequorin consumption, and found that the cGo has a high Ca2+ content, primarily due to accumulation through SERCA, and that the tGo is divided into two subcompartments: tGo2 with lower [Ca2+] and no SPCA activity, and tGo1 with higher [Ca2+] and high SPCA-1 expression. Finally, the actions of various Ca2+ mobilizers on the different calcium pools have been assessed.
For further details on experimental procedures see the supplementary material.
Gene construction, stable cell lines, and transfection with DNA or siRNA vectors
The enhanced green fluorescent protein (EGFP)–aequorin (GA) plasmid  was targeted to either the cis (and cis-to-medial) or the trans side of the Golgi apparatus, using targeting sequences from β1,6-N-acetylglucosaminyltransferase I [15,16] or β1,4-N-galactosyltransferase [17,18] for cGo and tGo respectively. The sequence of the siRNA against human (h)SPCA-1 was as previously reported . Transfection was performed with Lipofectamine 2000 (Life Technologies). HeLa cell lines expressing ER–GA, tGo–GA or cGo–GA were established by antibiotic selection of clonal stably transfected cells.
Immunolocalization, Western blot and aequorin measurements
For immunolocalization cells were fixed with methanol, incubated overnight at 4°C with the primary antibody, revealed with adequate secondary antibodies, counterstained with 4′,6-diamidino-2-phenylindole (DAPI) and observed by confocal microscopy. For Western blotting of SPCA-1, extracts of cells transfected with the different siRNAs were used and revealed an anti-SPCA-1 antibody (kindly donated by Dr Frank Wuytack). Aequorin measurements were performed in either intact or digitonin-permeabilized cells superfused with different solutions at 22°C, as described previously [14,20]. Results are expressed as the ratio of emitted counts per second to total counts per second at every time value, and calibrated in different Ca2+ concentrations .
The results are expressed as means±S.E.M. The statistical significance was evaluated using Student's t-test or one-way ANOVA with GraphPad InStat3 software.
Targeting of GFP–aequorin to cis- and trans-Golgi regions of the Golgi apparatus
We used the low Ca2+ affinity mutant aequorin (D119A) [22,23], fused to GFP and termed ‘GAD119A’ , to measure [Ca2+] inside the Go ([Ca2+]Go). This indicator was first devised for ER measurements and was used in the present study because of the expected high [Ca2+]Go levels, in the micromolar range [3,5]. The specific localization to the Go was achieved by fusing the GA to the targeting sequences of two resident enzymes that selectively localize to the different regions of Go. β1,6-N-acetylglucosaminyltransferase has been found to locate to cis and cis-to-medial-Go (cGo) by immunolocalization, by confocal microscopy and electron microscopy [15,16,24]. β1,4-N-galactosyltransferase locates to tGo, as shown by both confocal fluorescence microscopy and immunogold electron microscopy [17,18]. Putative mistargeting of the Ca2+ probes due to over-expression was avoided by expressing the Go–GA genes under the herpes simplex virus (HSV) IE4/5 promoter, which achieved low expression levels. To further assure reproducibility of the results, we generated clonal cell lines that stably expressed either cGo–GA or tGo–GA.
Distribution of the GFP fluorescence encoded by the targeted probes indicated correct localization to the Go. Fluorescence accumulated close to the nuclear envelope, with a laterally asymmetrical pattern, whereas the nucleus showed no labelling (Figure 1A) (see Supplementary Video S1). The cGo and tGo labelling was interleaved. Figures 1B and 1C show that GA targeted to cGo or tGo co-localized with the corresponding markers, the 130-kDa Golgi matrix protein (GM130, which locates into cGo and one or two cisternae on the cis side of the Golgi stack)  (Figure 1B) and the 46-kDa tGo network protein (TGN46–a specific marker of tGo)  (Figure 1C), respectively. Figure 1D summarizes the results of several similar experiments. Even though they were interleaved in the same areas, GM130 and TGN46 had non-coincident cytoplasmic locations. In fact, Pearson's r coefficient had a negative value (see Supplementary Table S1), indicating mutual exclusion from the same physical sites. In addition, the specific localization of the targeted GA proteins was assessed by comparing their expression pattern with selective markers of other organelles, such as ER matrix (SERCA-2), or lysosomes [lysosome-associated membrane protein 2 (LAMP2)] (Figure 1A; see Supplementary Video 2). Co-localization was quantified on 2D or 3D confocal images . As shown in Figure 1D, co-localization of cGo and tGo probes with the corresponding selective markers was highly significant (P<0.0001), whereas co-localization with lysosomal markers showed Pearson's r values significantly smaller than 0, indicating exclusion. Supplementary Table S1 shows the results of extensive co-localization tests that confirm the correct subcellular distribution of both Go–GA probes.
Localization of the new Ca2+ probes targeted to the cis- and trans-Golgi apparatus
Go–GAs are functional Ca2+ probes
To test the functionality of Go–GA probes, HeLa cell lines expressing GA targeted to either cGo or tGo were Ca2+ depleted and reconstituted with the aequorin cofactor coelenterazine n. Incubation in medium containing 1 mM Ca2+ allowed refilling of the cellular calcium stores. Figure 2A compares Ca2+ uptake in cGo, tGo and ER, each one monitored by the corresponding targeted GA. The ER refilled up to levels of 686±15 μM (mean±S.E.M., n=3), a higher level than that observed in either the cGo (219±9 μM) or the tGo (289±18 μM). Likewise, the initial rate of Ca2+ uptake was higher in the ER (553±107 μM/min) than in the cGo or the tGo (220±15 and 194±43 μM/min, respectively).
Calcium that accumulated inside the ER, cGo or tGo can be released both by stimulation with InsP3-producing agonists (ATP, carbachol or histamine) and by activation of ryanodine receptors (RyRs) with caffeine. Representative traces are illustrated in Supplementary Figures S1A and S1B, and a summary of results is compiled in Supplementary Figure S1C. In all three organelles, stimulation with ATP released about 80% of the calcium content, whereas caffeine produced an almost 50% release.
Ca2+ uptake by organelles can be studied in a simpler and more controlled cell model by permeabilizing the plasma membrane and clamping cytosolic Ca2+ concentration ([Ca2+]C) at any desired value using calcium buffers. In this model, intracellular messengers or modulators can be directly tested by addition to the incubation medium, which is now directly connected with the cytosol [14,28]. Figure 2B shows that, in digitonin-permeabilized HeLa cells, exposure to 100 nM Ca2+, a concentration similar to that found in the cytosol at rest, produced Ca2+ accumulation in all three compartments. The [Ca2+] reached in the ER at the steady state was 454±6 μM (mean±S.E.M., n=3), whereas the accumulation by the Go was significantly smaller, reaching steady-state levels of 156±15 μM for the cGo, and 135±6 μM for the tGo. These results parallelled the ones observed in intact cells (Figure 2A), suggesting that the pump:leak ratio is higher in the ER than in the Go. The initial rate of uptake was higher in the ER (129±6 μM/min; n=3) compared with the Go (64±5 μM/min in tGo and 14±8 μM/min in cGo).
We also tested the effects of three putative Ca2+-releasing agents, nicotinic acid–adenine dinucleotide phosphate (NAADP), cADP-ribose (cADPR) and InsP3. Whereas InsP3 produced virtually complete release of Ca2+ in all three compartments (87±6% for the ER, 74±7% for the tGo and 92±1% for the cGo), neither cADPR nor NAADP induced measurable Ca2+-releasing effects (Figure 2B).
ER, tGo and cGo Ca2+ uptake and release in HeLa cells
A distinct thapsigargin-resistant compartment, amounting to 20% of the total, is contained in the tGo
The identity of the Ca2+ uptake mechanism for the Go was first investigated by addition of the specific SERCA inhibitor thapsigargin . The effects of 20 nM thapsigargin were very different among the ER, cGo and tGo (Figure 2A). Ca2+ uptake by the ER or the cGo was virtually abolished, with steady-state [Ca2+] levels as low as 4.0±0.3 μM for the ER and 15±1 μM for the cGo (mean±S.E.M., n=3). In contrast, Ca2+ uptake by the tGo was much faster than in the cGo or ER and [Ca2+]Go reached higher levels at the steady state (119±25 μM, n=3–Figure 2A). Similar results were found using 2,5-di(t-butyl)-1,4-benzohydroquinone (TBH), a reversible inhibitor of SERCA  (results not shown).
Kinetics of the thapsigargin inhibition of Ca2+ uptake by the ER and tGo are compared in Figure 2 (C, a and b). At 20 nM, thapsigargin inhibited ER uptake by >99%, whereas as much as 20–30% of the Ca2+ uptake by the tGo was unaffected by this thapsigargin concentration. Increasing the inhibitor concentration up to 200 nM produced a further inhibition, which was close to its maximum at 1 μM (Figure 2C, c; note the logarithmic scale). These results indicate that the tGo Ca2+ uptake is composed of two different mechanisms: one thapsigargin-sensitive and another thapsigargin-resistant. In contrast, only the thapsigargin-sensitive component is present in the cGo (Figure 2A).
Holo-aequorin is consumed during Ca2+-induced bioluminescence emission. The consumption rate depends on the [Ca2+] and follows an exponential decay which can be monitored along each experiment. This property can be exploited to detect the existence of different subcellular compartments and estimate their relative size. If Ca2+ accumulation in one of the compartments is prevented, the consumption of the aequorin contained within the other compartment follows a single exponential. In permeabilized cells, the intracellular Ca2+ stores of which had been refilled by incubation with 100 nM Ca2+, the rate of aequorin consumption inside the ER fitted to a single exponential (a straight line in the logarithmic representation of Figure 3A). The half-decay time was 1.8 min and 80–90% of aequorin was consumed through the experiment. By contrast, consumption by either cGo or tGo was slower (Figure 3A), consistent with the smaller [Ca2+] reached inside the Go (see Figure 2A); aequorin consumption in these compartments was <40% during the whole observation period.
Comparison of aequorin consumption in the ER, tGo and cGo
When SERCA was blocked by treatment with 20 nM thapsigargin, Ca2+ uptake by the ER was completely blocked (see Figure 2A). As a consequence, the aequorin localized inside the ER was not consumed in thapsigargin-treated cells (Figure 3B). In contrast, aequorin consumption in the tGo of thapsigargin-treated cells was fast immediately after addition of Ca2+, but slowed down progressively over the following minutes, and finally reached a plateau with very slow consumption by the end of the experiment (Figure 3B). This result indicates that thapsigargin-resistant Ca2+ uptake is present almost exclusively in a fraction of the tGo space which amounts to about 20% of the total aequorin-containing space (see extrapolation with dotted lines in the Figure). The remaining 80% of aequorin must be inside another pool, the refilling of which was completely blocked by 20 nM thapsigargin, and that is not connected with the first pool. The same result was obtained using the reversible inhibitor TBH rather than thapsigargin to block SERCA (Figure 3C); aequorin consumption was completely inhibited in the ER and almost entirely in the cGo, whereas almost 20% of the total aequorin was quickly consumed in the TBH-treated tGo.
We repeated the experiments using native coelenterazine (rather than coelenterazine n) for reconstituting the apo-aequorin in order to increase the sensitivity to Ca2+. The results obtained were in full agreement with those obtained with coelenterazine n (see Supplementary Figure S2A). Again, aequorin consumption in the tGo was initially fast and stopped at about 20%. In the cGo the consumption of aequorin was much slower. Increasing the [Ca2+]C from 100 nM to 500 nM considerably accelerated the consumption for the cGo, and had a negligible effect on the tGo. In an additional experiment, we incubated thapsigargin-pretreated cGo-expressing cells with 500 nM [Ca2+]C for a long incubation period (30 min) to increase the extent of aequorin consumption (see Supplementary Figure S2B). Under this condition >60% of the aequorin was consumed by the end of the incubation period, and the rate of consumption was essentially constant over the whole period. This result indicates that Ca2+ enters the entire cGo space via the thapsigargin-sensitive mechanism. In summary, thapsigargin (20 nM) completely blocks Ca2+ uptake into the ER and most of it into the cGo. In contrast, about 20% of the tGo compartment is able to refill with Ca2+ in the presence of 20 nM thapsigargin, whereas the Ca2+ uptake in the remaining 80% of the tGo is completely blocked at this concentration of thapsigargin.
Properties of the thapsigargin-resistant and thapsigargin-sensitive Ca2+ pools of the tGo
Ca2+ transport into the thapsigargin-resistant component of the tGo can be readily studied in great detail in cells pre-treated with this irreversible inhibitor. The steady-state Ca2+ concentrations reached by cells treated with 20 nM thapsigargin could be estimated from aequorin calibration and are shown in Figure 4 (note logarithmic scale for [Ca2+]). The steady-state [Ca2+] levels were (mean±S.E.M., n=4): 2.0±0.2 μM in the ER, 9±2 μM in the cGo and 237±16 μM in the tGo (Figure 4A). The initial rates of Ca2+ uptake were also very different among the three organelles studied (mean±S.E.M., n=4): 0.3±0.1 μM/min for the ER, 3.3±0.2 μM/min for the cGo and 19±2 μM/min for the tGo (Figure 4B).
Ca2+ uptake by the ER, tGo and cGo in permeabilized HeLa cells treated with thapsigargin or TBH
Inhibition of SERCA with 10 μM TBH produced essentially the same results as 20 nM thapsigargin (Figure 4C): the [Ca2+]Go attained inside the tGo at the steady state (187±7 μM, mean±S.E.M., n=3) was much larger than in the cGo (2.9±0.1) or the ER (1.9±0.1). Consistent with the steady-state values, the initial rate of Ca2+ uptake in the tGo (103±1 μM/min) was much higher than that obtained in the cGo (0.6±0.1) or the ER (0.14±0.08) (Figure 4D).
Characterization of the thapsigargin-sensitive pools in cGo and tGo
The results of the experiments described above focus on the thapsigargin-resistant component of the tGo Ca2+ pool. Unfortunately, we were unable to find a selective inhibitor of this component in order to study the contribution of the SERCA-mediated component of Ca2+ uptake directly and exclusively. However, taking advantage of aequorin consumption during Ca2+-induced bioluminescence emission and of the reversibility of SERCA inhibition by TBH, we designed a protocol to measure Ca2+ transport selectively in the SERCA-dependent tGo space. We incubated the digitonin-permeabilized cells with 100 nM Ca2+ in the presence of TBH for 10 min. During this time, the TBH-insensitive tGo pool refills with Ca2+ and the aequorin contained in this compartment (tGo1) will burn out. If TBH is then washed out, we can directly follow the uptake into the TBH-sensitive component (tGo2), the aequorin of which was preserved intact, without interference from the TBH-resistant aequorin pool. This approach is illustrated in Figure 5. Incubation of tGo with 100 nM Ca2+ and 10 μM TBH for 10 min resulted in consumption of about 20% of the total aequorin during the first few minutes (Figure 5A–see dotted line). This consumption resulted from quite a fast uptake (initial rate, 190 μM/min) which reached a steady-state level of 270 μM within 2–3 min (Figure 5B). At this time point, aequorin consumption became very slow (11–13 min; Figure 5A). Subsequent washout of TBH at 13 min triggered, after a brief gap, acceleration of the consumption rate (Figure 5A), which corresponded to slow Ca2+ uptake (4–8 μM/min; Figure 5B) into the TBH-sensitive aequorin pool (about 80% of the total aequorin). After 5–8 min of incubation a steady-state level of about 40 μM was reached (Figure 5B). Further increase of [Ca2+]C up to 1000 μM increased the uptake of Ca2+ into the tGo.
Comparison of the effect of TBH on the uptake of Ca2+ (B, D and E) and aequorin consumption (A and C) by the tGo and cGo
The cGo (Figures 5C and 5D) behaved very differently. Ca2+ uptake in the presence of TBH was negligible (Figure 5D) and slowly increased to reach a rate of 11 μM/min on removal of the SERCA inhibitor, without reaching a steady-state level during the experiment. Addition of 1 mM Ca2+ at 22 min increased the uptake (Figure 5D) and the rate of aequorin consumption (Figure 5C). By the end of the experiment, almost 80% of the aequorin had been consumed with no signs of major compartmentalization (note the quite large slope in Figure 5C, at 27–30 min).
Another experimental protocol designed to study Ca2+ handling by the tGo is shown in Supplementary Figure S3. We compared the Ca2+ uptake triggered by the addition of 100 nM [Ca2+]C, either in the presence of TBH (TBH-resistant component–tGo1) or 15 min after removal of TBH (TBH-sensitive component–tGo2). The initial rates of uptake were 98 and 11 μM/min, and the steady-state levels reached were 186 and 40 μM, respectively; these values are consistent with the results shown in Figures 5A–5D.
We next asked whether Ca2+ could be released from these two tGo subcompartments by InsP3 or calcium-induced calcium release mechanisms. This was tested by applying InsP3 or caffeine to digitonin-permeabilized HeLa cells expressing tGo–GA (Figure 5E). First, the digitonin-permeabilized cells were incubated in intracellular-like solution containing 100 nM Ca2+ and 10 μM TBH. Ca2+ was quickly taken up by the tGo reaching a steady-state level of 154±7 μM (mean±S.E.M., n=3). Addition of InsP3 produced no calcium release from this TBH-resistant tGo subcompartment (tGo1), whereas caffeine released 66±1% of the total Ca2+ content (Figure 5E). Incubation with 100 nM Ca2+ was maintained for a period of 15 min to ensure full consumption of the aequorin contained in this TBH-resistant compartment, which amounted to about 20% of the total tGo (results not shown). Then, TBH was washed out and the incubation was continued in 100 nM Ca2+ to allow refilling of the TBH-sensitive compartment (tGo2). Refilling proceeded slowly to reach a steady-state level of 83±7 μM, a concentration well below the one reached in tGo1. The initial rate of uptake in tGo2 was also lower, 14±2 μM/s compared with 76±6 μM/s in tGo1, although it affected a larger fraction of aequorin. By the end of the experiment aequorin consumption had reached 70–80%, with no change in the rate. This suggests that tGo2 was a homogeneous compartment containing all of the remaining aequorin (about 80%). Stimulation with InsP3 produced a large Ca2+ release (79±5%) from this compartment (Figure 5E). In agreement with these results, experiments performed in intact cells confirmed that InsP3-producing stimuli, such as ATP, were unable to cause Ca2+ release from tGo1, whereas caffeine was a potent releaser (first trace in Supplementary Figure S4A). In contrast, ATP released most of the calcium stored by tGo2, once it had been activated by washing out TBH during the second half of the experiment (second trace in Supplementary Figure S4A). The tGo1 pool amounted to about 20% of the total tGo aequorin content, as revealed by aequorin consumption measurements (see Supplementary Figure S4B).
The thapsigargin-resistant Ca2+ uptake component in the tGo shows a high affinity for Ca2+
We repeatedly observed that the rate of Ca2+ uptake by the tGo1 does not change when the [Ca2+]C is increased from 100 nM to 500 nM, in contrast to the uptake by the cGo, which is noticeably increased (see Supplementary Figure S2A). These results suggested that tGo Ca2+ transport might already be close to saturation at [Ca2+]C of 100 nM. On the other hand, it has been reported that SPCA displays a higher affinity for Ca2+ than SERCA [31–33] with a dissociation constant, Kd, of 10–25 nM. We compared the uptake by cGo and tGo in thapsigargin-treated cells at 20 and 100 nM [Ca2+]C (Figure 6A). The rate of uptake by cGo was very low at both concentrations and the [Ca2+]Go reached at the steady state was very low (8±3 μM at 100 nM [Ca2+]C; mean±S.E.M., n=3). This agrees with the results already shown in Figure 2, which illustrate the high sensitivity of Ca2+ uptake to thapsigargin in this compartment. By contrast, in the tGo [Ca2+]Go reached 306±37 μM (mean±S.E.M., n=3) at 100 nM [Ca2+]C and 131±7 μM at 20 nM. It is interesting to note that 20 nM is clearly below the normal resting [Ca2+]C  and that SERCA activity should be close to 0 at this cytosolic [Ca2+] . We next studied the kinetics of Ca2+ uptake in response to [Ca2+]. Figure 6B compares the kinetics of Ca2+ activation of thapsigargin-sensitive Ca2+ uptake by the ER with thapsigargin-resistant uptake by the tGo. The Ca2+ affinity of the tGo system was higher, with a K50 of about 30 nM, compared with 170 nM for the ER. Note that, at 50 nM [Ca2+], Ca2+ uptake was practically negligible in the ER, whereas it was more than 50% in the tGo (Figure 6B).
Ca2+-dependence of thapsigargin-resistant Ca2+ uptake by cGo and tGo
The thapsigargin-resistant Ca2+ uptake into the tGo is mediated by the secretory pathway Ca2+-ATPase
To investigate the mechanisms involved in the accumulation of Ca2+ in the Go, we explored the participation of SPCA  in Ca2+ transport by the cGo and tGo compartments through the silencing of SPCA-1 expression with an siRNA. Figure 7A shows that human SPCA-1 expression was inhibited by 49% using 30 nM siRNA and by 72% at 100 nM siRNA. Using the latter siRNA concentration, the uptake of Ca2+ in the thapsigargin-resistant component of tGo was practically abolished (Figure 7B), [Ca2+]Go steady-state level was decreased from 122±21 to 2±1 μM, and the initial rate of Ca2+ uptake decreased from 52±7 μM/s to 1±1 μM/s (mean±S.E.M., n=3). Figure 7C shows the averaged values from three experiments. In the cGo thapsigargin-resistant Ca2+ accumulation was very weak (steady-state level 4±2 μM) and further decreased by siRNA treatment to 2±1 μM (see averaged values in Figure 7C). The thapsigargin-resistant accumulation of Ca2+ into the ER was almost non-measurable with or without siRNA treatment (Figure 7C; note logarithmic scale).
SPCA-1 knockdown inhibits thapsigargin-resistant Ca2+ uptake by the tGo
We have developed two new bioluminescent Ca2+ sensors for measuring specifically within either cis (and cis-to-medial) or trans compartments of Go. This GA sensor is also fluorescent, which facilitates assessment of its correct subcellular distribution. The probe was targeted to the Go by fusion with two very well-characterized sequences, a domain of β1,6-N-acetylglucosaminyltransferase I for cGo (cis and cis-to-medial) [15,16,24] or β1,4-N-galactosyltransferase for tGo targeting [17,18].
Previous knowledge about Ca2+ homoeostasis in the Go derives from studies using two different probes, aequorin [3,7,9,19,37] and the fluorescent chameleon D1cpv [8,10], with targeting by fusion to the transmembrane domain of sialyltransferase I (tGo) or a 32-residue domain of β1,6-N-acetylglucosaminyltransferase I (cGo). Important discrepancies were found between data obtained with Go–aequorin and Go–D1cpv, even though the targeting sequence was identical and the experiments were performed by the same group. These discrepancies were attributed to mistargeting of Go–aequorin protein, which was distributed throughout all Go compartments and even in the ER of some cells, probably because of overexpression . This poses serious doubts about the validity of the data obtained earlier with Go–aequorin. On the other hand, N-acetylglucosaminyltransferase–D1cpv expression seemed to be restricted to cGo , but quantitative analysis of data obtained with D1cpv probes may be limited by its relatively small fluorescence changes, e.g. maximal agonist stimulation produced a decrease in the fluorescent ratio of only 15% . In contrast, aequorins display a very large dynamic range and a steep Ca2+-dependence, which allows coverage of a wide range of [Ca2+] (10−8 to 10−3 M) . The GA probes are also fluorescent, and this allows assessment of the probe's distribution in living cells .
We used a relatively weak promoter, HSV-IE4/5, to prevent the mislocalization by excessive expression, reported previously . In addition, we generated clonal cell lines expressing the different Ca2+ probes to increase reproducibility of the results. The GA Ca2+ probes targeted to cGo or tGo showed the correct localization (see Figure 1 and Supplementary Table S1). Finally, we devised a simplified model with digitonin-permeabilized cells that allows continuous control of [Ca2+]C and access to non-permeable intracellular messengers. The combination of these approaches and tools decreased variability and allowed more precision in the quantitative measurements.
Aequorins have a unique combination of properties that make them particularly suitable for detecting and studying non-homogeneous Ca2+ distribution within and among organelles . They are consumed during bioluminescence emission on binding Ca2+. This apparent disadvantage can be used to trace the history of compartments in which [Ca2+] has been increased and to quantify their size. In addition, aequorin luminescence emission is usually less sensitive to acidic pH than the fluorescence sensors , and this is an advantage when measurements are being made inside mildly acidic compartments such as the Go [17,40].
Our results reveal both similarities and differences among the three Ca2+-storing endomembrane systems: ER, cGo and tGo. All three compartments accumulate Ca2+ in their matrix at high concentrations, >10−4 M. As a consequence, these compartments may generate cytosolic Ca2+ signals by sudden release of their Ca2+ content, as is well known for the ER . We find higher [Ca2+] in the ER (400–700 μM) than in both Go compartments (≤300 μM). These results agree with previous reports for the tGo [3,8,41]. In the present study, the tGo contains two pools, tGo1 and tGo2, with different [Ca2+]Go levels (see below). In contrast, the maximal [Ca2+] in the cGo and tGo1 is similar, 150–300 μM (Figure 8). Both the cGo and tGo2, but not tGo1, can release Ca2+ rapidly and extensively in response to InsP3. Caffeine mobilizes Ca2+ from all the Go subcompartments, whereas NAADP or cADPR is unable to release Ca2+ (see Figure 2 and Supplementary Figure S1). The relative Ca2+ content of Go is small, 2.5–5% of the total calcium . Therefore, the ability to generate cytosolic Ca2+ peaks must be much smaller for Go than for the ER. Ca2+ homoeostasis in the Go may be more relevant to warrant a high [Ca2+] in the Go matrix, which seems to play an important role in correct protein sorting and cargo in this organelle [12,13,42].
A functional model for Ca2+ homoeostasis in the Golgi apparatus
The size of three distinct functional calcium pools–cGo, tGo1 and tGo2 (Figure 8)–can be quantified by analysis of the bi-exponential aequorin consumption, e.g. the thapsigargin-resistant component of tGo (tGo1) amounted to about 20% of the total tGo aequorin, and tGo2 contained the remaining 80% (Figure 8). These two compartments were diffusionally isolated from each other, with no flow of aequorin or Ca2+ between them. For these reasons, we envision these compartments as physically isolated, perhaps constituted by the tGo stack cisternae (tGo2) and the vesicles (tGo1), respectively.
[Ca2+] in the cGo and tGo1 is high (150–300 μM), whereas [Ca2+]Go inside tGo2 is smaller, about 50–100 μM. The mechanisms for Ca2+ uptake also differ. In the cGo it is mediated by SERCA and blocked by low concentrations of thapsigargin (20 nM) or TBH (10 μM), and the contribution of SPCA-1 cGo is negligible (see Figures 2 and 6). Ca2+ pumping to the matrix in tGo2 takes place through SERCA, but activity is comparatively small. At 100 nM, Ca2+ uptake into tGo2 is 5–10 times slower than into tGo1 (see Figure 5). The uptake into tGo1 is fast and resistant to thapsigargin (20 nM) and TBH (10 μM). It is mediated by SPCA-1 because it is blocked with SPCA-1–siRNA (see Figure 7). The presence of a thapsigargin-resistant component in tGo described here is consistent with previous reports [9,43], although the sensitivity of SPCA to thapsigargin reported in the present study (inhibition constant Ki=250 nM) is somewhat higher than that previously reported for SPCA on the basis of measurements of ATPase activity [7,31]. Although ER uptake is inhibited >99.9% by 20 nM thapsigargin, much higher concentrations (up to 1–2 μM) are often used in routine experiments. Note that these extremely high concentrations could also partly inhibit Ca2+ uptake by tGo1 (see Figure 2C).
The affinity of the uptake mechanism for Ca2+ is much higher for tGo1 than for the ER or cGo (see Figure 6), and this is consistent with the higher Ca2+ affinity reported for SPCA activity compared with SERCA [7,44]. This extra-high affinity would secure adequate Ca2+ filling of the tGo, even at low [Ca2+]C levels, thus preventing alterations of secretory cargo. The distribution of the InsP3 receptor is also asymmetrical, absent from tGo1 but present in tGo2 and cGo. In this way, emptying of the stores by InsP3 during cytosolic Ca2+ signalling would not interfere with the high [Ca2+]Go required for tGo1 function. RyR are found in all the subcompartments, whereas NAADP and cADPR receptors were absent from them.
130-kDa Golgi matrix protein
herpes simplex virus
lysosome-associated membrane protein 2
nicotinic acid-adenine dinucleotide phosphate
sarco plasmic/endoplasmic reticulum Ca2+-ATPase
secretory pathway Ca2+-ATPase
46-kDa trans-Golgi network protein
Francisco Aulestia performed and analysed most of the experiments. María Teresa Alonso thought up the study and designed the molecular biology approach. Javier García-Sancho thought up the study, designed and analysed the aequorin consumption experiments, and wrote the manuscript. The three authors discussed the results and corrected the manuscript.
We thank Dr Frank Wuytack, Leuven, Belgium, for the anti-SPCA-1 antibody, Dr Juan Llopis, University of Castile–La Mancha, Albacete, Spain, for the YFP–Go plasmid and Dr Assou El-Battari, Institut national de la santé et de la recherche médicale (INSERM) U-559, Marseille, France, for the C2 (1–32)–EGFP plasmid. We also thank Alvaro Martín for assistance with cell sorting, Cristina Sánchez for help with the confocal microscope, and Miriam García Cubillas and Jesus Fernández for excellent technical support.
This work was supported by a grant from the Spanish Ministerio de Economía y Competitividad (MEC) [grant numbers BFU2010–17379] and the Instituto de Salud Carlos III [grant numbers RD06/0010/0000 and RD12/0019/0036]. F.J.A. was supported by an FPI fellowship from MEC.