β2-adrenergic receptor regulates ER-mitochondria contacts

Interactions between the endoplasmic reticulum (ER) and mitochondria (Mito) are crucial for many cellular functions, and their interaction levels change dynamically depending on the cellular environment. Little is known about how the interactions between these organelles are regulated within the cell. Here we screened a compound library to identify chemical modulators for ER-Mito contacts in HEK293T cells. Multiple agonists of G-protein coupled receptors (GPCRs), beta-adrenergic receptors (β-ARs) in particular, scored in this screen. Analyses in multiple orthogonal assays validated that β2-AR activation promotes physical and functional interactions between the two organelles. Furthermore, we have elucidated potential downstream effectors mediating β2-AR-induced ER-Mito contacts. Together our study identifies β2-AR signaling as an important regulatory pathway for ER-Mito coupling and highlights the role of these contacts in responding to physiological demands or stresses.

www.nature.com/scientificreports/ apposition between ER and Mito-generates a fluorescent product (red) that can be detected in situ ( Fig. 2A). Isoproterenol-treated cells showed significant increase in PLA fluorescent signal, demonstrating that this drug indeed facilitates ER-Mito contacts ( Fig. 2A,B). Of note, different treatment time of isoproterenol resulted in different level of responses in PLA signal (Fig. S3C), suggesting a dynamic nature of the drug effect on ER-Mito contacts. When PLA was performed in HeLa cells treated with CCPA (PubChem CID 123807; adenosine A1 receptor agonist) or PGE (PubChem CID 5280723; prostaglandin E2 receptor agonist), both of which were iden-  www.nature.com/scientificreports/ tified as hits (Fig. 1F), similar results were obtained (Fig. S3A, B), even though their target receptor expression levels are low according to the available mRNA-seq data (Table S4). Next, we assessed the contacts at the ultrastructural level by transmission electron microscopy (TEM). The HEK293T cells treated with or without isoproterenol were fixed and analyzed by TEM. The relative number of mitochondria closely juxtaposed (< 30 nm) to ER over the total number of mitochondria in isoproterenol-treated cells was significantly higher (DMSO, 22.5%; isoproterenol, 41.7%; p < 0.0001; Chi-square test) (Fig. 2C,D). Furthermore, the isoproterenol-treated cells showed a significant increase in total contact coverage, calculated as the percent of the total contact length divided by the total Mito perimeter measured among all mitochondria (5.6 ± 2.1%, control; 14.6 ± 2.7%, isoprot; p = 0.0286) (Fig. 2E). Interestingly, when only the mitochondria having a tight coupling with ER were compared, the percent of contact length did not show a statistically significant difference (Fig. 2F). We speculate that isoproterenol might increase the proximity of the ER and Mito (contact width) or the number of contacts, with little or no effect on contact length.
Finally, the ascorbate peroxidase (APEX)-proximity labeling method was employed, which biotinylates proteins within 20 nm proximity of the enzyme. Mito-APEX, targeted to Mito outer membrane, can label and identify ER proteins in close proximity to the Mito 16 . HEK293T cells transfected with mito-APEX were treated with DMSO or isoproterenol, and the biotinylated proteins (by incubating with biotin-phenol and with H 2 O 2 ) were affinity-purified using streptavidin-magnetic beads and subjected to Western blot analysis. In these affinity purified samples (which were within 20 nm proximity to the Mito membrane) prepared from the isoproterenoltreated cells, we found that ER proteins IP3R1 and SEC61b were increased, but not cytosolic protein GAPDH ( Fig. 2G; Fig. S6). Together, these data from three independent assays corroborate our results with the split-Rluc assay, strongly supporting that the β2-AR agonist isoproterenol enhances physical contacts between the ER and Mito.
Isoproterenol facilitates mitochondrial Ca 2+ uptake and bioenergetics. We next investigated if drug-induced physical contacts created functional coupling between the ER and Mito. Given the role of ER-Mito contact sites as hotspots for Ca 2+ influx to mitochondria (Mito) 9 , we explored if agonist treatment increases Mito Ca 2+ influx, using a fluorescent Ca 2+ sensor mito-R-GECO1 18 . HeLa cells were pre-treated with DMSO or isoproterenol, and then the Mito Ca 2+ uptake was measured every 2.5 s in response to histamine-stimulated ER Ca 2+ release (Fig. 3A). Upon ER Ca 2+ release, isoproterenol elevated Mito Ca 2+ uptake (Fig. 3A), resulting in a higher maximum peak (Fig. 3B) and faster uptake rate (Fig. 3C). Also, when ionomycin (Ca 2+ ionophore) was used to trigger cytoplasmic Ca 2+ influx, a similar increase in Mito Ca 2+ uptake was observed (Fig. 3D). Together these data support that the β2-AR agonist isoproterenol enhances, not only physical, but also functional coupling between the ER and Mito. CCPA and PGE1 treatment also showed similar results ( Fig. S4A-C).
ER-Mito coupling contributes to Mito bioenergetics by supplying Ca 2+19-21 , required for the function of proteins in the tricarboxylic acid (TCA) cycle and oxidative phosphorylation [21][22][23] . We thus investigated the effect of the agonist on Mito bioenergetics. The Mito membrane potential, an indicator of mitochondrial metabolism 24 , was measured with the tetramethyl rhodamine methyl ester (TMRM) fluorescence dye (Fig. 3E,F). Isoproterenoladministration significantly increased the intensity of the TMRM signal (Fig. 3E,F). These findings imply that drug-induced ER-Mito contacts enhance Mito metabolism probably by facilitating ER-to-Mito Ca 2+ transfer. Interestingly, isoproterenol led to a reduction in average Mito length in HeLa cells (Fig. 3G,H). Given that ER-Mito contacts are known to mark Mito fission sites 25 , our data raise the interesting possibility that ER-Mito contact enhancement by β2-AR activation may facilitate Mito fission events. The adenosine receptor agonist CCPA also showed similar results ( Fig. S4D-G) despite the low expression level of adenosine receptors (Table S4).

Gαs/cAMP mediates β2-AR signal-induced ER-Mito contacts.
To gain insight into the downstream factors of the β2-AR signaling involved in ER-Mito contact modulation, we evaluated the involvement of the G s pathway because G αs is the primary Gα coupled to β2-AR ( Fig. 4A) 13 . First, we examined the G αs pathway activation using a cAMP response element (CRE) reporter gene assay (a proxy for cAMP measurement). Compared to DMSO, isoproterenol increased CRE reporter activity in HEK293T cells (fold change: 4.79 ± 1.66) (Fig. S5A), confirming G αs pathway activation by isoproterenol. A similar result was observed in HeLa cells with a lower increase in CRE reporter activity (Fig. S5A). When β2-AR itself was overexpressed, it induced much higher CRE reporter activity (fold change: 249.60 ± 19.90) (Fig. S5B) than isoproterenol treatment, arguing that the isoproterenol-mediated β2-AR stimulation in our assay is not as strong as a full activation via receptor overexpression, as expected.
Next, we asked if G αs pathway activation influences the level of ER-Mito contacts. A constitutively active construct for G αs (G αs CA), which activates the signaling cascade even in the absence of ligand 26 , augmented split-Rluc activity (Fig. 4B). Corroborating these results, an adenylyl cyclase (AC) activator forskolin (thus increasing cAMP level) enhanced PLA signal (Fig. 4E). Together these results demonstrate that G αs /AC/cAMP pathway activation leads to an increase in ER-Mito contacts.
Because cAMP can transduce signal through two different downstream effectors, EPAC and PKA, we next examined which of these factors were involved in mediating the signal for contact regulation. Using two different tools, a dominant negative (DN) construct and a pharmacological inhibitor (ESI 09), we determined if EPAC activity is required for G αs /AC/cAMP pathway-induced contacts. The EPAC DN construct significantly reduced G αs CA-induced split-Rluc activity (fold change: 3.49 ± 0.164, G αs CA; 2.28 ± 0.074, G αs CA/EPAC DN; p < 0.0001) without affecting basal level of split-Rluc activity (fold change: 1.00 ± 0.069, pCIG (vector); 0.97 ± 0.035, EPAC DN; p = 0.9131) (Fig. 4B). Similarly, a chemical inhibitor of EPAC, ESI 09, caused a significant reduction in isoproterenol-induced split-Rluc activity (Fig. 4C, left). Because it also reduced the split-Rluc activity in DMSOtreated cells (Fig. 4C, left), we compared the fold changes of the isoproterenol-induced activity (over DMSO)    4D). Consistent with these data, ESI 09 also blocked forskolin-induced PLA signal enhancement (forskolin increases cAMP by stimulating adenylyl cyclase) (Fig. 4E). Of note, we found that RAP1A (EPAC downstream effector) is also involved, as RAP1A DN decreased the isoproterenol response (Fig. 4F), albeit its effect is not www.nature.com/scientificreports/ as robust as EPAC inhibition (Fig. 4D); this implicates that EPAC-mediated effect is only partially transmitted through RAP1A. In contrast to EPAC, the inhibition of PKA activity by Rp-cAMPS and H-89 (Fig. 4G,H), or of MEK activity by PD 032509 (Fig. S5C) did not show any blocking effect on isoproterenol response. Together these results support that EPAC, but not PKA, is a downstream factor mediating β2-AR signaling for ER-Mito contacts. In accordance with our data, a recent study has demonstrated that EPAC is necessary for increasing the interaction between VDAC1, IP3R1, and GRP75, a well-established complex at ER-Mito contact sites 27 . Elevated cytosolic Ca 2+ participates in β2-AR signal-induced ER-Mito contacts. In addition to cAMP, cytosolic Ca 2+ level is also known to increase upon β2-AR activation by isoproterenol (Fig. 5A) 28 . We thus set out to evaluate the potential role of cytosolic Ca 2+ in isoproterenol-induced ER-Mito interactions. First, we ensured an increase in cytoplasmic Ca 2+ level upon isoproterenol application to the cells by using fluorescent Ca 2+ sensor GCaMP6s (Fig. 5B). Next, we investigated if intracellular Ca 2+ increase itself could enhance ER-Mito interactions. For this, we assessed the effects of the three different drugs, histamine, ionomycin and cyclopiazonic acid (CPA), all of which are known to elevate cytosolic Ca 2+ albeit by distinct mechanisms [28][29][30] . Histamine stimulates cytosolic Ca 2+ release from the ER through IP3 receptor activation 28 ; ionomycin, a membrane permeable Ca 2+ ionophore, allows both Ca 2+ release from intracellular stores and extracellular Ca 2+ influx to the cytoplasm 28 ; and CPA, a sarco-endoplasmic reticulum Ca 2+ -ATPase (SERCA) inhibitor, blocks the uptake of Ca 2+ ions back to the sarcoplasmic reticulum (SR)/ER lumen after their release into the cytosol, thus accumulating Ca 2+ in the cytoplasm 29 . All three drugs elevated Split-Rluc activity (Fig. 5C), suggesting that increasing cytoplasmic Ca 2+ concentration is sufficient to enhance ER-Mito contacts. We corroborated these findings with the PLA results; histamine or ionomycin treatment increased PLA signal (Fig. 5D).
Finally, we examined if cytoplasmic Ca 2+ increase is required for isoproterenol-induced ER-Mito coupling. A recently identified calcium mobilization inhibitor, ebselen 31 , eliminated the positive effect of isoproterenol ( Fig. 5E), proving the requirement of cytoplasmic Ca 2+ in isoproterenol-induced contacts. However, considering its multiple inhibitory functions, the possibility of ebselen inhibiting other factors or pathways could not be excluded. Unlike ebselen, the PLC inhibitors edelfosine and U73122 exhibited a rather mild inhibition on isoproterenol-induced ER-Mito contacts (Fig. 5F,G). When the fold changes of the isoproterenol-induced luciferase activity (over DMSO) were compared between the vehicle-vs. each drug-treated cells, edelfosine showed a trend toward blocking isoproterenol response (p = 0.0674) although the effects of both drugs did not reach statistical significance (Fig. 5F,G; right). These data argue for a significant involvement of Ca 2+ but a negligible role of PLC in β2-AR signaling-induced ER-Mito coupling.

Actin polymerization is required for β2-AR signal-mediated ER-Mito contact regulation. Finally,
we explored the potential mechanism by which cytosolic Ca 2+ and cAMP could modulate ER-Mito interaction. It has been shown that cytosolic Ca 2+ spike by multiple stimuli such as ionomycin and histamine, triggers a transient surge in cytosolic actin polymerization 32,33 , which is required for ER-Mito contacts 28 . Thus, we postulated that an actin polymerization (regulated by cytosolic Ca 2+ rise and cAMP upon β-AR activation) would be required for isoproterenol-induced ER-Mito interaction. To test this hypothesis, the cells were treated with latrunculin A, an actin polymerization inhibitor, and its effect on isoproterenol activity was assessed. Supporting our hypothesis, latrunculin abolished the isoproterenol-induced split-Rluc activity (Fig. 6B). This finding supports a model that actin filament assembly triggered by β2-AR activation may facilitate a close contact between the ER and Mito.
In the cell, actin reorganization is mainly regulated by RHO GTPases; CDC42, RAC, and RHO are the most prominent members among them (Fig. 6A). It has been well established that the activation of CDC42, RAC, and RHO promotes actin polymerization (Fig. 6A) 34 . These RHO GTPases are regulated by GEFs (guanine nucleotide exchange factors), GAPs (GTPase activating proteins) and GDIs (guanine nucleotide dissociation inhibitors); the GPCR signaling is linked to the activation of these regulators 35,36 . In addition to G α12 -mediated pathway which is well documented for regulating RHO GTPases through RHO GEFs, increased intracellular Ca 2+ (via activating calmodulin) as well as cAMP (via activating EPAC) play roles in activating RHO GTPases [36][37][38] . This prompted us to test if CDC42, RAC, and RHO participate in β2-AR-induced ER-Mito contact formation.
To address this, CDC42, RAC, or RHO activity was pharmacologically inhibited in the presence of isoproterenol and then ER-Mito contacts were assayed. The CDC42 inhibitor ZCL 278 completely eradicated the effect of isoproterenol (Fig. 6C), supporting a role for CDC42 in β2-AR-induced ER-Mito contacts, likely by triggering actin polymerization. This result was corroborated by PLA results where ZCL 278 counteracted the positive effect of forskolin on ER-Mito contacts (Fig. 4E). The RAC inhibitor EHop 016 also reduced split-Rluc activity when treated with or without isoproterenol (Fig. 6D, left). However, its inhibitory effect appeared to be independent of isoproterenol response, considering that the fold increase by isoproterenol over DMSO was in fact higher in EHop 016-treated cells when compared to vehicle-treated ones (Fig. 6D, right). Finally, when RHO activity was inhibited by rhosin, no changes in isoproterenol response was detected (Fig. 6E). These results strongly support that actin polymerization triggered by CDC42, but not by RAC or RHO, is a downstream event facilitating ER-Mito coupling, in response to β2-AR signal.
Of note, another cytoskeletal drug nocodazole, which interferes microtubule polymerization and blocks mitochondrial motility 39 , caused a dose-dependent increase in split-Rluc activity; this suggests that microtubule polymerization normally hinders ER-Mito contacts probably due to excess mitochondrial dynamics along the microtubule ( Fig. S5D; see "Discussion"). Together these results implicate an important role of cytoskeletal dynamics in ER-Mito contact modulation.

Discussion
Communication between the ER and Mito via membrane contact sites is crucial for many cellular functions. Although recent studies have focused on identifying the molecular components of the contacts, the signaling pathway(s) involved in contact regulation remains unknown. Herein we have identified β2-AR signaling as a significant pathway modulating ER-Mito interactions. Upon ligand binding, β2-AR transduces the signal through G αs /cAMP and Ca 2+ to induce physical contacts between the ER and Mito likely by stimulating a rapid, transient actin filament assembly through CDC42 activation. As a result, mitochondrial Ca 2+ uptake is enhanced, thus activating mitochondrial matrix proteins to generate more ATPs (Fig. 7) 20,23 . Together our data provide significant insights into the roles of ER-Mito contacts in maintaining cellular homeostasis and in responding to physiological demands or stresses, under the influence of the GPCR signaling.
Our data support a model where EPAC activation via G αs /AC/cAMP, and cytosolic Ca 2+ increase lead to CDC42 activation [36][37][38] and subsequently to actin filament assembly 34 , a driving force for ER-Mito interaction (Fig. 7). In addition to CDC42, ER-bound inverted formin2 (INF2), could be a good candidate protein modulating actin filament assembly, given that it mediates actin polymerization in response to cytosolic Ca 2+ burst 28 upon GPCR activation 40 . Furthermore, INF2-mediated actin polymerization at the ER increases ER-Mito contacts,   28 , all of which were also increased with isoproterenol treatment (Figs. 2, 3). In addition to stimulating actin dynamics, cAMP and cytosolic Ca 2+ surge may also exert their effects through other mechanisms. For instance, they might facilitate the formation of ER-Mito coupling complexes. In fact, a recent study demonstrated that under stress condition, EPAC (activated by cAMP) enhances ER-Mito coupling by promoting the interaction between VDAC1/IP3R1/GRP75, an ER-Mito tethering complex 27 . In a similar way, Ca 2+ might affect the tethering by controlling either the tethering proteins directly or modulators of the complex. For example, the presence of Ca 2+ -responsive tethering proteins such as E-Syt1 (extended synaptotagmin 1), at the ER-plasma contact sites 41 , raises the possibility of the existence of similar proteins at the ER-Mito contact sites. In this regard, it will be interesting to investigate whether IP3Rs (IP3R1, 2, and 3) represent such tethering proteins responding to cytoplasmic Ca 2+ surge, given their structural roles independent of their channel functions 42 and the presence of Ca 2+ binding sites on them 43 . Another possible mechanism by which cytosolic Ca 2+ regulates ER-Mito coupling, is through modulating mitochondrial dynamics. It has been documented that IP3R-mediated cytosolic Ca 2+ rise arrests mitochondrial movement along the microtubule, increasing the (2) inducing tethering complex formation (e.g., VDAC1/IP3R1/GRP75). Increased cytosolic Ca 2+ can also increase ER-Mito interactions by influencing actin polymerization through CDC 42 or potentially through INF2 (inverted formin2) (ref. 29 ). It is also plausible that cAMP and Ca 2+ might employ other mechanisms (e.g., Mito movement arrest on microtubule by Ca 2+ ) and that another downstream pathway of β2-AR might also participate in ER-Mito coupling. β2-ARmediated ER-Mito coupling increases mitochondrial Ca 2+ uptake, which enhances mitochondrial bioenergetics by activating proteins involved in TCA cycle and oxidative phosphorylation (OXPHOS).  39 . Consistent with this, nocodazole (an inhibitor of microtubule polymerization shown to block mitochondrial motility) increased ER-Mito contacts in a dose dependent manner (Fig. S5D). Accumulating evidence indicates that ER-Mito dysfunction contributes to a variety of neurodegenerative diseases 4,44 . In Parkinson's disease models, β2-AR agonists have been documented to be beneficial 45 . Salbutamol and metaproterenol (β2-AR agonists), both of which were also identified in our screen, reduce α-syn transcription levels and appears to reduce risk of developing PD 14,15 . Although their efficacy was linked to α-syn transcription regulation 14 and anti-inflammation 46 , our data indicate that these drugs might also exert therapeutic effects by increasing ER-Mito contacts since some PD-associated mutation insults disrupt these contacts 47 . This opens up the possibilities for therapy of other conditions associated with ER-Mito contact dysregulation. When considering ER-Mito contact enhancement drugs for therapeutic options, it is important to recognize that both increased and decreased contacts can be detrimental. Therefore, careful titration of these drugs will be crucial.
Analysis of our screen results suggest that the reason why β2-AR agonists, among other GPCR targeting drugs, were identified as the major modulators for the ER-Mito contacts, is probably because β2-AR is one of the most prominently expressed GPCRs in H293T cells used in our drug screen (see Table S4). Given that G αs CA expression itself, as well as ionomycin or CPA treatment, elevated ER-Mito contacts (Figs. 4B, 5C), it is likely that any GPCRs that can trigger cAMP-or cytosolic Ca 2+ -increase may be able to modulate ER-Mito coupling. The implication that a broad range of GPCRs may be able to regulate ER-Mito contact is that the contact dynamics would be cell-type specific, because different cells will respond differently to the same stress/stimuli depending on their specific GPCR availability. Furthermore, considering the role of GPCRs in sensing various extracellular signals and subsequently eliciting appropriate cellular responses, our data provide strong evidence for the involvement of ER-Mito contacts in the adaptation of the cells to acute physiological conditions (e.g., starvation, high stress, etc.).
Our results also demonstrate that the split-Rluc assay can serve as a valid assay to screen for modulators of ER-Mito contacts, yet we recognize its limitation. Given that it takes approximately 1.5 h for EnduRen substrate to generate stable luminescence, which then will last for 24 h, the split-Rluc assay is not suitable to detect realtime dynamics of ER-Mito contacts. This explains the discrepancy between the PLA and split-Rluc assay when comparing the signal intensities in time course experiments for isoproterenol treatment; PLA signal shows increase (up to 3 h) then decrease (at 6 h) (Fig. S3C) probably due to negative feedback in GPCR signaling, while the split-Rluc activity shows continuous increase up to 16 h (maximum time measured) (data not shown). The split-Rluc assay appears to detect accumulated changes (irreversible), despite the relatively short half-life of each split fragment (Fig. S2D, E).
While our current screen focused on an FDA-approved drug library, further screens with other compound libraries will likely provide additional biologic insights into this important inter-organelle communication (e.g., identifying small molecules that could directly target ER-Mito contact sites). Furthermore, ORF or siRNA library screens would be warranted to identify additional molecular machinery functioning for the contact modulation. Finally, while our current screening system is most suitable for enhancer identification due to low basal level of split-Rluc activity, modification to our system could permit the screening for inhibitors in future studies. Furthermore, the detailed mechanisms of the GPCR pathway-regulated ER-Mito contacts and any distinctions, among different classes of GPCRs, in downstream pathways leading to ER-Mito contact modulation will need to be addressed.
Split-Renilla luciferase (split-Rluc) reconstitution assay. HEK293T cells were used for all split-Rluc assays in this study unless mentioned otherwise, via a transient transfection of the split-Rluc reporter plasmids (Mito-Rluc N and Rluc C -ER) in 12 well plates. Four to six hours after transfection, cells were dissociated and re-plated into a 96-well plate coated with poly-D-lysine (50 μg/mL). Twenty-four hours post-transfection, live cell substrate EnduRen (30 μM, Promega) was added to the culture media and incubated for 2-5 h. Luciferase activity (luminescence) was measured by a POLARstar Omega microplate reader (BMG LABTECH, Ortenberg, Germany) and normalized to the mean of the control.
Library screen. HEK293T cells were transfected with split-Rluc constructs in a 100 mm dish and used for split-Rluc assay. 1 μM of each compound (final concentration) was added to each well together with Enduren and incubated for 5 h followed by luminescence measurement. The initial screen was carried out once. The split-Rluc activity of each drug was normalized by quantile normalization using R software (version 4.0). Hits were picked using the cut-off value of 3SD higher than the average of the control (DMSO), and further tested for their activity with one more round of split-Rluc assay. For drug enrichment analysis, we employed the over representation analysis 51 . False discovery rate (FDR) was calculated by R software. All dendrograms were created using R with packages 'ggraph' , 'igraph, 'tidyverse' and 'RcolorBrewer' . Transmission electron microscopy (TEM). HEK293T cells cultured in 100 mm dish were treated with DMSO or isoproterenol for 30 min. These monolayers of cells were fixed with glutaraldehyde/paraformaldehyde mixture and post-fixed in 2% aqueous OsO 4 in 0.2 M S-collidine buffer, pH 7.4. One-micron thick sections were cut on an ultra-microtome and stained on the glass slide, and thin sections were mounted on 200 mesh copper grids and stained as previously described 52,53 . The stained grids were examined on a Jeol JEM-100CX electron microscope equipped with a digital camera. With randomly selected EM images (160 images for control and 158 images for isoproterenol), the distance between the outer mitochondrial membrane (OMM) and ER membrane was measured. If the distance (measured in Image J) was within 30 nm, it was counted as 'contact' , whereas if greater than 30 nm, 'no contact' . The number of mitochondria with ER contact was counted and its percentage over the total number of mitochondria was calculated. For the total contact coverage, the partial mitochondrial perimeter showing a close contact with ER was measured and added all together. This sum was divided by the sum of the mitochondrial perimeter measured among all mitochondria regardless of their contact status. For the contact length per mitochondrion, the partial perimeter of each mitochondrion showing contact with ER membrane (less than 30 nm distance) and the entire perimeter were measured for each mitochondrion in ImageJ and the ratio was plotted.

Proximity labeling assay (PLA
Mito-APEX proximity labeling. HEK293T cells were transfected with pcDNA3-mito-APEX in a 6-well plate as described in supplementary Methods. For proximity labeling of proteins, previously reported method 16 was used with a slight modification. Briefly, cells were treated with DMSO or isoproterenol (1 μM) for 1 h, before incubated with biotin-phenol (500 μM) and treated with H 2 O 2 for 3 min 16 . After the biotinylation reaction was quenched 16 , cells were lysed with RIPA buffer and centrifuged at 13,000 rpm for 10 min at 4 °C. To get rid of the free biotin-phenol, the sample was applied to Zeba Spin Desalting Column (Thermo Fisher Scientific). The recovered lysates were incubated with streptavidin magnetic beads 16 , and the beads were washed five times with RIPA buffer. Biotinylated proteins of mito-APEX pull downs were eluted by boiling with 2X protein loading buffer (Bio-Rad) containing 2 mM biotin and 20 mM DTT at 95 °C for 10 min and subjected to 4-15% SDS-PAGE followed by Western blot analysis described below.
Calcium level measurement. To monitor the mitochondrial and cytoplasmic calcium dynamics, mito-R-GECO1 and GCaMP6s constructs respectively were transfected into HeLa cells in an 8-well chamber slide. Sixteen hours after transfection, the cells were treated with DMSO or indicated drugs and imaged on Zeiss Observer Z1 inverted microscope using Zeiss Zen Pro software, every 2.5 or 5 s. Approximately 7.5 s (histamine) or 15 s (ionomycin) after the start of the experiment, histamine (100 μM) or ionomycin (3 μM) was superfused over transfected cells. The boundary of each cell was set as ROIs (regions of interest) and randomly selected based on the presence of R-GECO1 or GCaMP6s fluorescent signal. ∆F/F 0 and (∆F/F 0 )/ ∆t (R-GECO1) (from 7.5 to 12.5 s) were calculated as described previously 54 .
Mitochondrial transmembrane potential and mitochondrial length. Mitochondrial membrane potential (Ψ mit ) was measured by tetramethyl rhodamine methyl ester (TMRM) fluorescent dye (I34361, Thermo Fisher Scientific). HeLa cells were loaded with 10 nM TMRM for 30 min at RT and then treated with DMSO or isoproterenol (1 μM) for 1 h. TMRM fluorescence was imaged using Zeiss Observer Z1 inverted microscope, and the signal intensity was analyzed in Image J. The lengths of mitochondria were analyzed in Image Analyst MKII software (version 4.0) with default parameter [largest mitochondria size (width in pixel) = 5, Sensitivity (top range scaling, percentile) = 99.5 and minimum size (area, pixels) = 5].
GPCR expression level analysis using mRNA-seq data. The raw sequence files were downloaded from NIH GEO data sets with each GEO set access number (GSE99249 for HEK293T, GSE122986;SRR8247560 for HeLa, GSE148617;SRR10125054 for C2C12, GSE84481;SRR3927297 for Neuro2a ). The downloaded fastq files were aligned and sorted with GRCh38 or GRCm38 genome index by HISAT2 (version 2.1.0) and samtools (version 1.9), respectively. We calculated TPM with TPMCalculator 55 .