Mitochondrial cAMP exerts positive feedback on mitochondrial Ca²⁺ uptake via the recruitment of Epac1

Aldosterone, which is secreted by adrenal glomerulosa cells, is the principal regulator of salt–water balance. As such, it plays a significant role in the control of blood pressure, and participates in the pathogenesis of cardiovascular, inflammatory and renal diseases (Briet and Schiffrin, 2010; De Mello, 2017; Prabhu and Frangogiannis, 2016; Rossier et al., 2017; Zhang and Lerman, 2017). The most important physiological stimuli of aldosterone secretion are angiotensin II (AngII) and extracellular K⁺, their actions are mediated chiefly by the cytosolic Ca²⁺ signal (Hattangady et al., 2012; Spät and Hunyady, 2004). Cytosolic Ca²⁺, beside inducing and activating StAR, the steroidogenic acute regulatory protein that transports cholesterol to the inner mitochondrial membrane (IMM) (Cherradi et al., 1997), evokes mitochondrial Ca²⁺ signalling (Pitter et al., 2002; Spät and Pitter, 2004), resulting in enhanced reduction of pyridine nucleotides (Pralong et al., 1992; Rohács et al., 1997) and production of ATP (Tarasov et al., 2012). Moreover, the mitochondrial Ca²⁺ signal and NAD(P)H formation within mitochondria are essential for the hypersecretion of aldosterone (Spät et al., 2012; Wiederkehr et al., 2011).

The soluble adenylyl cyclase (sAC, encoded by ADCY10), a cAMP-generating enzyme activated by HCO₃⁻ (Chen et al., 2000; Lefkimmiatis et al., 2013; Steegborn et al., 2005b) and Ca²⁺ (Jaiswal and Conti, 2003), is present in the mitochondrial matrix (Acin-Perez et al., 2009b). Mitochondrial Ca²⁺ signals result in enhanced sAC-mediated formation of cAMP within the mitochondrial matrix of HeLa cells and rat cardiomyocytes (Di Benedetto et al., 2013, 2014), as well as in human adrenocortical H295R cells (Katona et al., 2015). In addition to the increased formation of ATP (Acin-Perez et al., 2009b; Di Benedetto et al., 2013; Wang et al., 2016), mitochondrial cAMP (mt-cAMP) may have cell type-specific roles, as exemplified by the reduced aldosterone production in adrenocortical cells after the knockdown of sAC (Katona et al., 2015). Mitochondrial Ca²⁺ is one of the key regulators of aldosterone synthesis (Wiederkehr et al., 2011), and it also boosts pyridine nucleotide reduction, which may also favour steroid production (Pralong et al., 1992; Spät et al., 2012). Therefore, in the present study, we examined whether the hormone secretion-promoting action of mt-cAMP reflects an effect on mitochondrial Ca²⁺ handling itself. We show that mt-cAMP enhances mitochondrial Ca²⁺ uptake and boosts the subsequent hormonal response in H295R human adrenocortical cells. The effect of mt-cAMP on mitochondrial Ca²⁺ handling is not mediated by changes in the mitochondrial membrane potential (ΔΨ_m) or Ca²⁺ efflux, but requires the cAMP-regulated guanine nucleotide exchange factor Epac1 (also known as RAPGEF3). Importantly, this feedback mechanism was also detected in HeLa cells, showing that it is not confined to steroid-producing cells.

First, we examined the effects of 2-OHE, a membrane-permeable sAC inhibitor (Steegborn et al., 2005a), on Ca²⁺ signalling in AngII-stimulated H295R cells. The drug significantly reduced the mitochondrial Ca²⁺ uptake rate without a similar effect on the cytosolic response (Fig. S1A). Further experiments, however, revealed that 2-OHE reduces the fluorescence of the membrane potential-sensitive dye TMRM both in the mitochondria and the nuclei (Fig. S1B). 2-OHE failed to influence ΔΨ_m in permeabilized cells (Fig. S1C,D), suggesting that the reduction of TMRM fluorescence in intact cells was due to plasma membrane depolarization. Nevertheless, we regarded the results obtained with 2-OHE in AngII-stimulated intact cells as inconclusive.

H295R cells show a voltage-dependent Ca²⁺ influx with an ensuing mitochondrial Ca²⁺ signal when stimulated with K⁺ (Szanda et al., 2006). Knocking down sAC with a verified siRNA sequence (MR2; Di Benedetto et al., 2013; Katona et al., 2015 and Fig. S2D) significantly reduced the net mitochondrial Ca²⁺ uptake rate without any effect on cytosolic Ca²⁺ response in K⁺-stimulated cells (Fig. 1A,B). Importantly, knockdown of sAC slightly hyperpolarized mitochondria both under resting conditions and during Ca²⁺ signalling as compared to what was seen in control cells (Fig. 1C). Since hyperpolarization of the mitochondria would increase, rather than decrease, the mitochondrial Ca²⁺ uptake rate, these findings strongly suggest that attenuating mt-cAMP formation impedes mitochondrial Ca²⁺ uptake in a ΔΨ_m-independent manner.

With the aim of minimizing the effect of extramitochondrial factors, next we analysed the mt-cAMP dependence of mitochondrial Ca²⁺ uptake in permeabilized cells in which perimitochondrial [Ca²⁺] can be adjusted. 2-OHE was used to inhibit sAC, and erythro-9-(2-hydoxy-3-nonyl)adenine (EHNA), a selective inhibitor of PDE2A (Acin-Perez et al., 2011) and, as such, an enhancer of mt-cAMP formation in H295R (Katona et al., 2015) and HeLa cells (Di Benedetto et al., 2013), was applied to increase [mt-cAMP]. Ca²⁺ accumulation was significantly dampened by 2-OHE (Fig. 2A) and EHNA markedly accelerated the uptake (Fig. 2B). These observations were corroborated by the finding that the knockdown of sAC reduced mitochondrial Ca²⁺ uptake into the mitochondria of permeabilized H295R cells (Fig. 2C), and we observed a similar reduction of Ca²⁺ uptake in permeabilized sAC-silenced HeLa cells (Fig. 2D).

The sAC–Ca²⁺ interplay was further investigated by combining treatment with a cAMP analogue and sAC silencing. The membrane-permeable cAMP analogue 8-Br-cAMP strongly intensified mitochondrial Ca²⁺ accumulation (Fig. 3A) without any measurable effect on ΔΨ_m (Fig. 3B), supporting the notion that mt-cAMP enhances mitochondrial Ca²⁺ accumulation independently of ΔΨ_m. Moreover, 8-Br-cAMP increased mitochondrial Ca²⁺ uptake to comparable levels in both the control and sAC-silenced permeabilized H295R cells (Fig. 3C), implying that cAMP is a downstream effector of sAC.

In order to study the effect of increased mt-cAMP formation on Ca²⁺ handling, we generated wild-type (WT) and enzymatically inactive mitochondrially targeted versions of sAC. H295R cells express the truncated (∼48 kDa) form of sAC, which localizes predominantly to the particulate fraction (Katona et al., 2015). Although the full-length sAC mRNA (NM_018417.5; 4832 bp CDS) can be found in these cells (Fig. S2A), for our mitochondria-targeted constructs, we used the sequence of the physiologically occurring and enzymatically enhanced truncated version of sAC (Buck et al., 1999; Steegborn, 2014) (Figs S2B, S3 and S4). Interestingly, two sAC isoforms were consistently amplified from H29R cells; one corresponding to the canonical sequence and the other lacking a short section (Δ215–276; Fig. S5A) from the C1–C2 region, which links the two catalytic domains (Steegborn, 2014).

In order to obtain functionally compromised mutants, two amino acid changes were introduced (Fig. S2B). In the double mutant sAC the D99A mutation aims to eliminate Mg²⁺ binding, and N412A to abolish ribose binding (Steegborn, 2014). As expected, HeLa cells expressing double mutant mitochondrial sAC (mt-sAC) exhibited significantly reduced mt-cAMP production during Ca²⁺ signalling or when stimulated with HCO₃⁻, a direct activator of sAC (Chen et al., 2000; Steegborn, 2014) (Fig. S2C).

If mt-cAMP intensifies mitochondrial Ca²⁺ uptake then one would expect a reduced Ca²⁺ accumulation when an enzymatically inactive sAC is overexpressed in mitochondria. This was indeed the case, as the overexpression of the double mutant mt-sAC markedly reduced the rate of mitochondrial Ca²⁺ accumulation in both permeabilized H295R and HeLa cells as compared to what was seen upon overexpression of the WT cyclase (Fig. 4A,B). [We obtained identical results with the adrenocortical isoforms of sAC (Fig. S5B), strongly suggesting that the linker region is indeed dispensable for the enzymatic activity.] Moreover, the difference in Ca²⁺ uptake rate between WT and mutant sAC was also preserved in the presence of the Na⁺/Ca²⁺ exchanger (NCLX) blocker CGP-37157 (Fig. S6). In order to test the effect of mt-cAMP-dependent enhancement of Ca²⁺ uptake on the biological response of adrenocortical cells, namely steroid synthesis, we measured basal and AngII-stimulated aldosterone production in cells transfected with WT or mutant mt-sAC over 2 h (Fig. 4C). Cells expressing mutant mt-sAC showed less steroidogenesis (as compared to WT mt-sAC-expressing counterparts) highlighting the contribution of mt-cAMP to the hormonal response.

Next we studied the involvement of recognized cAMP effectors, protein kinase A (PKA) and the guanine nucleotide exchange (GEF) factor Epac1, in conveying the enhancing effect of mt-cAMP on Ca²⁺ uptake. ESI-09, a pan-Epac inhibitor (Almahariq et al., 2013) reduced mitochondrial Ca²⁺ uptake rate in permeabilized H295R cells (Fig. 5A). In view of data confirming (Zhu et al., 2015) or challenging the specificity of ESI-09 (Rehmann, 2013), we also examined the effect of the structurally unrelated Epac1 inhibitor CE3F4 (Courilleau et al., 2012). This drug also decelerated mitochondrial Ca²⁺ uptake in permeabilized as well as in AngII-stimulated intact H295R cells without affecting the cytosolic Ca²⁺ signal (Fig. 5B,C). Importantly, the conventional PKA inhibitor H-89 failed to affect the Ca²⁺ uptake rate but did cause a decrease in the steady-state [Ca²⁺]_m in permeabilized cells (Fig. 5B). And finally, the enhancing effect of mt-sAC overexpression on Ca²⁺ uptake was lost if Epac1 was inhibited with CE3F4 (Fig. 5D) strongly suggesting that Epac is downstream of mt-sAC and mt-cAMP. Taken together, these data strongly suggest that Epac has an essential role in the control of initial Ca²⁺ uptake whereas PKA, acting with a longer lag-time, may contribute to maintaining elevated [Ca²⁺]_m.

Ca²⁺ uptake into mitochondria occurs by diffusion through the voltage-dependent anion channels (VDACs) in the outer mitochondrial membrane followed by the transport through the IMM by the mitochondrial Ca²⁺ uniporter (MCU) multiprotein complex, the structure of which has been recently elucidated (MCU, Baughman et al., 2011; De Stefani et al., 2011; MCUb, Raffaello et al., 2013; MICU1, Perocchi et al., 2010; MICU2, Plovanich et al., 2013; EMRE, Sancak et al., 2013; for reviews see, for example, De Stefani et al., 2016; Mammucari et al., 2016). The transport is driven by the 150–180 mV (inside negative) mitochondrial membrane potential (Ψ_m). The ensuing mitochondrial Ca²⁺ signal modulates important mitochondrial processes as pyridine nucleotide reduction (McCormack et al., 1990), ATP synthesis (Jouaville et al., 1999), apoptosis (Hajnóczky et al., 2000), and even cell type-specific functions including insulin and aldosterone production (Wiederkehr et al., 2011). However, the short-term control of Ca²⁺ uptake under physiologically relevant conditions has remained elusive (De Stefani et al., 2016). In the present study, we reveal a hitherto unrecognized mechanism of this control, specifically that mt-cAMP enhances mitochondrial Ca²⁺ uptake.

Ca²⁺ influx into the mitochondrial matrix activates mt-sAC which, in turn, produces mt-cAMP, a process first described in HeLa cells and cultured cardiac myocytes (Di Benedetto et al., 2013). In our previous study on H295R adrenocortical cells (Katona et al., 2015), we observed that the inhibition or the knockdown of sAC as well as the buffering of mitochondrial Ca²⁺ by the targeted expression of a Ca²⁺-binding protein (S100G) significantly reduced AngII-elicited mt-cAMP signalling whereas the inhibition of mitochondrial PDE2A with EHNA (Acin-Perez et al., 2011) enhanced cAMP production within the organelle. The biological significance of mt-cAMP signalling was demonstrated by the reduced aldosterone production after the inhibition or knockdown of sAC. Since [Ca²⁺]_m is one of the key factors regulating aldosterone secretion (Spät et al., 2012), in the present study we examined whether the secretagogue effect of mt-cAMP could be mediated by an action of this messenger on mitochondrial Ca²⁺ uptake itself. Moreover, we extended our study of mitochondrial Ca²⁺ handling to HeLa cells so as to assess whether the mt-cAMP–Ca²⁺ interplay is unique to endocrine cells or is a more general phenomenon.

Applying experimental approaches that manipulate mt-cAMP signalling we found that: (1) knockdown of sAC resulted in decelerated mitochondrial Ca²⁺ uptake in intact H295R cells; (2) increasing the mt-cAMP activity through the inhibition of PDE2A, treatment with the cAMP analogue 8-Br-cAMP or by the overexpression of mt-sAC all accelerated Ca²⁺ uptake in permeabilized H295R and HeLa cells; (3) mt-cAMP-dependent changes in mitochondrial Ca²⁺ uptake were not dependent on changes in ΔΨ_m or Ca²⁺ efflux; (4) the augmenting effect of mt-cAMP on mitochondrial Ca²⁺ accumulation was sensitive to the inhibition of the Rap GEF Epac1; and (5) overexpressing WT sAC within mitochondria intensified aldosterone production as compared to what was seen for an enzymatically compromised sAC mutant.

Knockdown of sAC decelerated mitochondrial Ca²⁺ accumulation during Ca²⁺ signalling in K⁺-stimulated intact H295R cells. Although sAC in H295R cells locates predominantly to the particulate fraction (Katona et al., 2015), by studying mitochondrial Ca²⁺ uptake in permeabilized cells one may minimize the influence of extramitochondrial factors (Bernardi et al., 1999), and cytosolic cAMP (and quite possibly cytosolic sAC) are also lost under such conditions. Importantly, data obtained in permeabilized cells confirmed the findings in intact cells, as both inhibition and knockdown of sAC impeded Ca²⁺ uptake in H295R and HeLa cells. Moreover, overexpression of WT sAC in the mitochondrial matrix accelerated Ca²⁺ accumulation into the organelle in both cell types.

We found that H295R cells, besides expressing the canonical sAC variant, also express an isoform lacking amino acids 215–276 (Δ215–276) within the linker region of the enzyme. This adrenocortical isoform had identical effects on mitochondrial Ca²⁺ uptake as the canonical 48-kDa version, confirming the notion that the linker region is dispensable for enzymatic activity (Steegborn, 2014). (The deletion in this mutant respects exon–intron boundaries and thus the Δ215–276 sAC isoform can probably be regarded as a naturally occurring variant.)

Theoretically, accelerated mitochondrial Ca²⁺ accumulation may reflect increased Ca²⁺ conductance via the MCU, inhibited Ca²⁺ efflux through NCLX, a larger driving force and, naturally, the combination thereof. The maximal rate of mitochondrial Ca²⁺ uptake far exceeds that of Ca²⁺ efflux in H295R cells (A.S. and G.S., unpublished observation) rendering the NCLX an unlikely target of mt-cAMP. Nevertheless, the possibility that mt-cAMP inhibits the NCLX, and thereby accelerates Ca²⁺ accumulation, had to be considered. The enhanced Ca²⁺ uptake in sAC-overexpressing cells was preserved in the presence of CGP-37157, a conventional inhibitor of NCLX. As to the driving force, the membrane-permeable cAMP analogue 8-Br-cAMP increased mitochondrial Ca²⁺ uptake rate but failed to influence ΔΨ_m. Moreover, knockdown of sAC slightly hyperpolarized mitochondria, yet it reduced the rate of Ca²⁺ accumulation. Taken together, changes in ΔΨ_m or Ca²⁺ efflux are not required for the mt-cAMP-dependent modulation of mitochondrial Ca²⁺ handling.

Apart from cyclic nucleotide-activated cation channels in the plasma membrane, two downstream signalling pathways of that are activated by cAMP are generally accepted: activation of PKA and the Epac family of Rap1 exchange factors (Seino and Shibasaki, 2005). The location of PKA within the mitochondrial matrix has not been unambiguously demonstrated. Owing to binding of PKA to several anchoring proteins in the outer mitochondrial membrane and possibly also to the outer surface of the IMM (see Di Benedetto et al., 2013; Lefkimmiatis et al., 2013), data on the exact location of PKA in the ‘inner’ mitochondria (Sardanelli et al., 2006; Schwoch et al., 1990) are not conclusive, and reports on the location of PKA within the matrix are also conflicting (Acin-Perez et al., 2011; Monterisi et al., 2017). There is also no agreement on whether PKA activates the electron transport chain and ATP production (Acin-Perez et al., 2009a; Di Benedetto et al., 2013; Lefkimmiatis et al., 2013). In our present study, the conventional PKA inhibitor H89 failed to exert a significant effect on the initial mitochondrial Ca²⁺ uptake. Nevertheless, with some delay, H89 decreased the steady-state [Ca²⁺]_m. The real biological significance of this latter effect warrants further studies, but the reported off-target effects of H89 should also be kept in mind (Lochner and Moolman, 2006).

Besides PKA, the cAMP-activated GEFs Epac1 and Epac2 (also known as RAPGEF4) are possible mediators of the mt-cAMP effect on mitochondrial Ca²⁺ uptake. The presence of Epac1 in mitochondria (Qiao et al., 2002), in mitoplasts (Wang et al., 2016), in the IMM and also in the matrix (Fazal et al., 2017) has already been documented and an N-terminal mitochondrial target sequence on Epac1 has been identified (Fazal et al., 2017). As to the role of Epac family proteins on mitochondrial Ca²⁺ metabolism, we are aware of two reports on myocardial cells. However, the data are conflicting inasmuch as those of Wang and co-workers (Wang et al., 2016) indicate an inhibitory whereas those of Fazal et al. (Fazal et al., 2017) show a stimulatory effect for the Epac proteins. It should be emphasized that non-physiological conditions were applied in these experiments: mitochondria were isolated from hypertrophic rat hearts (Wang et al., 2016) and mitochondrial Ca²⁺ accumulation in murine cardiac cells was observed in the presence of supraphysiological (50 µM) extramitochondrial Ca²⁺ (without data on the uptake rate) (Fazal et al., 2017). In our experiments a pan-Epac inhibitor and a specific Epac1 inhibitor reduced mitochondrial Ca²⁺ accumulation in intact H295R cells and also in permeabilized H295R and HeLa cells. Moreover, the overexpression of mt-sAC failed to accelerate the mitochondrial Ca²⁺ response in the presence of Epac inhibitors, indicating that Epac1 is the downstream effector of mt-cAMP in this respect. Taken together, whereas the data on cardiac cells (Fazal et al., 2017; Wang et al., 2016) contribute to our knowledge on Epac-dependent cell death, our experiments show that intramitochondrial cAMP and Epac signalling support mitochondrial Ca²⁺ accumulation also at physiological Ca²⁺ concentrations.

Only a small amount of data is available about the role of Epac proteins in the adrenal cortex. A report showing that only ∼60% of the cAMP-mediated effects of adrenocorticotropic hormone (ACTH) on transcription were dependent on PKA in Y1 mouse adrenal cells (Schimmer et al., 2006) gave rise to the idea that the PKA-independent actions were mediated by Epac proteins (Lewis et al., 2016). In murine Y1 cells, the expression of Epac2 could be detected via immunocytochemistry, yet the cell-permeable Epac activator 8-pCPT-2′-O-Me-cAMP failed to simulate the effect of cAMP either on the expression of steroidogenic factors or on steroid secretion over 24 h (Aumo et al., 2010). In bovine glomerulosa cells, both 8-pCPT-2′-O-Me-cAMP and AngII activated the Epac substrate Rap1 but neither of them stimulated aldosterone production over 1 h (Gambaryan et al., 2006). Considering that 8-pCPT-2′-O-Me-cAMP is not resistant to phosphodiesterases (Laxman et al., 2006), its breakdown during cell incubation cannot be ruled out. Owing to well-known species differences in the control of aldosterone secretion (Spät and Hunyady, 2004; Spät et al., 2016), these data should not be extrapolated to human cells. Indeed, our findings showing that knockdown of sAC inhibits (Katona et al., 2015) whereas overexpression of mt-sAC augments aldosterone production, together with the observation that mt-sAC fails to affect mitochondrial Ca²⁺ uptake in the presence of Epac inhibitors, strongly suggest that mitochondrial Epac is involved in the regulation of the steroidogenic response in the human adrenal cortex.

The combination of work described here and in past studies (Katona et al., 2015) reveals a positive-feedback loop controlling mitochondrial Ca²⁺ handling: Ca²⁺ triggers the formation of mt-cAMP which, in turn, recruits Epac and enhances Ca²⁺ uptake (Fig. 6). This positive feedback is a new example for the convergence of Ca²⁺ and cAMP signalling (Spät et al., 2016). Although excessive function of this system may lead to cell death, it may also have great significance in situations of emergency, especially when a rapid cellular response is required. Changes in basal mt-cAMP metabolism modified the mitochondrial Ca²⁺ uptake rate without a detectable delay, suggesting that even basal mt-sAC activity has a role in triggering the biological response. In case of glomerulosa cells, it cab be assumed that during severe salt-water loss (e.g. haemorrhage, diarrhoea and strenuous physical exercise in hot environment) aldosterone secretion is triggered by the simultaneous actions of AngII and corticotrophin (Spät and Hunyady, 2004) and is enhanced by the here-described Ca²⁺‒mt-cAMP‒Ca²⁺ positive-feedback system. In addition, given that mt-cAMP also enhances mitochondrial Ca²⁺ uptake in HeLa cells, the present observations may be of broader cell physiological significance.

OPTI-MEM, Lipofectamine LTX, RNAiMax, Fluo-4 AM, Rhod-2 AM, tetramethyl rhodamine methylester (TMRM) and Mitotracker Deep Red, as well as the BCA assay kit were purchased from Invitrogen (Thermo Fisher Scientific, Waltham, MA). Fura-2 AM was from Tocris (Elliswille, MO) and Ultra-MEM was from Lonza (Basel, Switzerland). UltroSer G was from Bio-Sepra (Cergy-Saint-Christophe, France). siRNA for silencing sAC (MR2, Di Benedetto et al., 2013) and control siRNA (Universal Negative Control, SIC001) were obtained from Sigma-Aldrich (St Louis, MO). 4mt-H30 was constructed by Giulietta Di Benedetto and Tullio Pozzan (Institute of Neuroscience, Italian National Research Council, Padova, Italy). Mitochondria-targeted inverse Pericam (mt-i-Pericam) was a gift from Atsushi Miyawki (RIKEN, Saitama, Japan). 4mt-D2 was prepared as described previously (Fülöp et al., 2011). CE3F4 and 8-Br-cAMP was obtained from Cayman (Ann Arobor, MI); other chemicals were obtained from Sigma-Aldrich.

H295R and HeLa cells were cultured as previously described (Fülöp et al., 2011). Briefly, H295R cells (CRL-2128, ATCC, Manassas, VA) were cultured in Dulbecco's modified Eagle's medium (DMEM)/Ham's F12 (1:1, v/v) supplemented with 1% insulin-transferrin-selenium (ITS⁺), 2% UltroSer G, 100 U ml⁻¹ penicillin and 100 µg ml⁻¹ streptomycin. HeLa cells (CLL-2, ATCC) were grown in DMEM containing 10% heat-inactivated fetal bovine serum (FBS), 100 U ml⁻¹ penicillin and 100 µg ml⁻¹ streptomycin. Passages numbered 3–20 were used. Before experiments, H295R cells were serum-starved overnight (14–16 h) and HeLa cells were starved for 3–5 h. Incubation conditions for aldosterone experiments are described below.

Cells (2.5×10⁴–5×10⁴ H295R or 2×10⁴–5×10⁴ HeLa cells) were plated onto 25-mm diameter circular glass coverslips on day 1 and transfected on day 2 with plasmid DNA (0.5 μg/coverslip except for HeLa cells transfected with 1 μg/coverslip 4-mt-H30 DNA) using Lipofectamine LTX (H295R) or Fugene HD (HeLa) according to the manufacturer's protocol. For silencing sAC, on day 2, cells were transfected with 50 pmol MR2 siRNA or control RNA in Ultra-MEM using Lipofectamine RNAiMax. If H295R cells were co-transfected with plasmid DNA and siRNA, RNAiMax was used as the reagent. Experiments were conducted on day 4 or, in case of mt-cAMP measurements, on day 5.

RNA isolation and cDNA synthesis were performed as described previously (Ella et al., 2016). For the determination of GAPDH expression by quantitative real-time PCR (qPCR) the LightCycler 480 system (Roche) with the FastStart DNA Master SYBR Green (Roche) was used. Primers were: forward: 5′-AAGGTGAAGGTCGGAGTCAACG-3′; reverse: 5′-GACGGTGCCATGGAATTTGC-3′.

A Zeiss LSM710 confocal laser scanning microscope (operated with ZEN 11.0 software) and a 40*/1.3 water immersion objective (Plan-Apochromat, Zeiss) were used. For monitoring cytosolic and mitochondrial Ca²⁺ signals, the cells were preloaded with Fluo-4 and Rhod-2 or transfected with mt-i-Pericam, as specified in the figure legends. Fluorescence was monitored in multitrack mode; excitation and emission detection wavelengths were set as described previously (Fülöp et al., 2011; Katona et al., 2015). The optical slice was 5 µm in the cytosolic and 3 µm in the mitochondrial channels, but the latter was set at 1.5 µm in the colocalization studies (Figs S2 and S3). In kinetic studies, fluorescence intensity was normalized to the average intensity measured for 20–60 s before stimulation (F₀). The linear section of the normalized Ca²⁺ curves was regarded as rate of Ca²⁺ uptake. Uptake rate was expressed as (ΔF/F₀)/Δt for Rhod-2, or (F₀/ΔF)/Δt in case of mt-i-Pericam.

For the measurement of the mitochondrial membrane potential (ΔΨ_m), TMRM was used. TMRM (15–25 nM) was present throughout the entire experiment including drug pre-incubation periods. At the end of each run, ΔΨ_m was dissipated with 10 μM FCCP+8 μg ml⁻¹ oligomycin and TMRM fluorescence was normalized to that measured after FCCP+oligomycin addition (F_FCCP). Excitation and emission were as described previously (Fülöp et al., 2011).

An inverted microscope (Axio Observer D1, Zeiss) equipped with a 40×1.4 Plan-Apochromat oil immersion objective (Zeiss) and a Cascade II camera (Photometrics) was used for epifluorescence measurements. Mitochondrial [Ca²⁺] ([Ca²⁺]_m) and cytosolic [Ca²⁺] ([Ca²⁺]_c) in intact cells were monitored by using 4mt-D₂-cpV (4mt-D₂) and Fura-2, respectively, and [Ca²⁺]_m in permeabilized cells was monitored with mt-i-Pericam. The mitochondrial Ca²⁺ uptake rate was calculated as described previously (Fülöp et al., 2011). The cAMP level in the mitochondrial matrix was monitored by means of fluorescence resonance energy transfer (FRET), using 4mt-H30 as described previously (Katona et al., 2015).

Microscopic experiments were carried out at room temperature. For kinetic studies in intact cells, the coverslips were superfused with a modified Krebs–Ringer solution containing 140 mM Na⁺, 4.5 mM K⁺, 1.2 mM Ca²⁺, 0.5 mM Mg²⁺, 5 mM HEPES and 2 mM HCO₃⁻ (pH 7.4). Cells were permeabilized with 25 μg ml⁻¹ digitonin in Ca²⁺-free cytosol-like medium (see below) for 6 min. For superfusion of permeabilized cells, a pH 7.1 cytosol-like solution was used containing 129 mM K⁺, 10 mM Na⁺, 1.13 mM Mg²⁺, 120 mM HEPES, 2 mM EGTA, 2 mM ADP, 2 mM pyruvate and 2 mM malate (for H295R) or succinate (for HeLa). To adjust the [Ca²⁺] and [Mg²⁺] of the cytosol-like medium, EGTA, HEDTA, Ca²⁺ and Mg²⁺ were added as calculated by the chelator software (Fabiato, 1988). The flow rate of superfusion was ∼1 ml min⁻¹. The solutions were applied with a solenoid valve-equipped gravity-driven superfusion system, terminating at ∼2 mm from the selected cells.

Cells were transfected with the appropriate sAC constructs by means of electroporation (2×10⁶ cells in 100 μl resuspension buffer, 1600 V, 20 ms; Neon Transfection System, Thermo Fisher Scientific) and plated onto 24-well plates (5×10⁵ cells/well) on day 1. Following incubation in the complete medium (see section ‘Cell culture and transfection’), on day 3 the medium was changed to one with 0.1% UltroSer G content for overnight incubation. On day 4, after a 30-min pre-incubation in serum-free medium (supplemented with 0.01% bovine serum albumin), the cells were incubated at 37°C for 2 h in a similar medium, with or without AngII. The aldosterone content of the supernatant and cellular protein amount were determined with a DERCW100 RIA kit (Demeditec, Kiel, Germany) and with a BCA assay (Thermo Fisher Scientific, Waltham, MA), respectively.

To create the 5mt-sAC-mRFP constructs, the truncated version of soluble adenylate cyclase (t-sAC, amino acids 1–469) was amplified from serum-deprived human adrenocortical H295R cell cDNA. The following primers were used: forward, 5′-ATATGTCGACGATGAACACTCCAAAAGAAGAATTCC-3′ and reverse, 5′-ATATACCGGTCCTGCTCCGACTTTCTCAGTACGGCCC-3′. PCR products were digested and cloned into the pmRFP-N1 vector at SalI and AgeI sites. Five consecutive mitochondrial target signals from the human cytochrome c oxidase VIII (5 mt) were subcloned at the N-terminal end of sAC. V5- and His-tagged constructs were generated by amplifying the 5mt-sAC section from the above plasmid and cloning into the pcDNA3.1/V5-His TOPO vector (Thermo Fisher Scientific, Waltham, MA). Amino acid changes (D99A, N412A) were introduced by site-directed mutagenesis using the following primers: sAC_D99A forward, 5′-GCAGGTGCTGCACTGCTAGCC-3′; sAC_D99A, reverse 5′-GCAGTGCAGCACCTGCAAATTTCAGG-3′; sAC_N412A forward, 5′-GTCGCCTTAGCTGCCAGGATGATGATGTACTACC-3′; and sAC_N412A reverse, 5′-GGCAGCTAAGGCGACTTTTTGACCAATGACTG-3′. Every construct was verified by sequencing (Microsynth, Balgach, Switzerland).

Means±s.e.m. are shown unless otherwise specified. In some experiments, in cases of high-impact outliers, minimal and maximal values have been uniformly excluded in all groups (trimmed mean). In some cases, for better visibility, mean±s.e.m. are shown, and in microscopy experiments with 1 Hz≤acquisition frequency, only every second or third data points are marked by actual symbols. For estimating significance of differences, a Student's unpaired t-test, Mann–Whitney and Kolmogorov–Smirnov test, one- and two-way ANOVA and post-hoc tests were used, as appropriate. Data were analysed with Statistica 13, Microsoft Excel 2016 and GraphPad Prism 5 software.

We gratefully acknowledge the valuable discussions with Professors Erzsébet Ligeti, László Tretter and Miklós Geiszt; the excellent technical help of Miss Ágnes Südy and Miss Laura Szalai in the qPCR measurements is highly appreciated.