Strategic Positioning and Biased Activity of the Mitochondrial Calcium Uniporter in Cardiac Muscle*

Control of myocardial energetics by Ca2+ signal propagation to the mitochondrial matrix includes local Ca2+ delivery from sarcoplasmic reticulum (SR) ryanodine receptors (RyR2) to the inner mitochondrial membrane (IMM) Ca2+ uniporter (mtCU). mtCU activity in cardiac mitochondria is relatively low, whereas the IMM surface is large, due to extensive cristae folding. Hence, stochastically distributed mtCU may not suffice to support local Ca2+ transfer. We hypothesized that mtCU concentrated at mitochondria-SR associations would promote the effective Ca2+ transfer. mtCU distribution was determined by tracking MCU and EMRE, the proteins essential for channel formation. Both proteins were enriched in the IMM-outer mitochondrial membrane (OMM) contact point submitochondrial fraction and, as super-resolution microscopy revealed, located more to the mitochondrial periphery (inner boundary membrane) than inside the cristae, indicating high accessibility to cytosol-derived Ca2+ inputs. Furthermore, MCU immunofluorescence distribution was biased toward the mitochondria-SR interface (RyR2), and this bias was promoted by Ca2+ signaling activity in intact cardiomyocytes. The SR fraction of heart homogenate contains mitochondria with extensive SR associations, and these mitochondria are highly enriched in EMRE. Size exclusion chromatography suggested for EMRE- and MCU-containing complexes a wide size range and also revealed MCU-containing complexes devoid of EMRE (thus disabled) in the mitochondrial but not the SR fraction. Functional measurements suggested more effective mtCU-mediated Ca2+ uptake activity by the mitochondria of the SR than of the mitochondrial fraction. Thus, mtCU “hot spots” can be formed at the cardiac muscle mitochondria-SR associations via localization and assembly bias, serving local Ca2+ signaling and the excitation-energetics coupling.


Control of myocardial energetics by Ca 2؉ signal propagation
to the mitochondrial matrix includes local Ca 2؉ delivery from sarcoplasmic reticulum (SR) ryanodine receptors (RyR2) to the inner mitochondrial membrane (IMM) Ca 2؉ uniporter (mtCU). mtCU activity in cardiac mitochondria is relatively low, whereas the IMM surface is large, due to extensive cristae folding. Hence, stochastically distributed mtCU may not suffice to support local Ca 2؉ transfer. We hypothesized that mtCU concentrated at mitochondria-SR associations would promote the effective Ca 2؉ transfer. mtCU distribution was determined by tracking MCU and EMRE, the proteins essential for channel formation. Both proteins were enriched in the IMM-outer mitochondrial membrane (OMM) contact point submitochondrial fraction and, as super-resolution microscopy revealed, located more to the mitochondrial periphery (inner boundary membrane) than inside the cristae, indicating high accessibility to cytosol-derived Ca 2؉ inputs. Furthermore, MCU immunofluorescence distribution was biased toward the mitochondria-SR interface (RyR2), and this bias was promoted by Ca 2؉ signaling activity in intact cardiomyocytes. The SR fraction of heart homogenate contains mitochondria with extensive SR associations, and these mitochondria are highly enriched in EMRE. Size exclusion chromatography suggested for EMRE-and MCU-containing complexes a wide size range and also revealed MCU-containing complexes devoid of EMRE (thus disabled) in the mitochondrial but not the SR fraction. Functional measurements suggested more effective mtCU-mediated Ca 2؉ uptake activity by the mitochondria of the SR than of the mitochondrial fraction. Thus, mtCU "hot spots" can be formed at the cardiac muscle mitochondria-SR associations via localization and assembly bias, serving local Ca 2؉ signaling and the excitation-energetics coupling.
Mitochondrial Ca 2ϩ uptake in response to cytosolic calcium signals has been emerging as a pivotal means for tuning oxidative ATP production to energy demands but also as risk factor for Ca 2ϩ overload, oxidative stress, and ignition of cell death pathways. The inside negative membrane potential across the IMM 3 (⌬⌿ m ) is a tremendous driving force for Ca 2ϩ uptake that is implemented via the tightly regulated Ca 2ϩ -activated Ca 2ϩ channel, mtCU (1,2). Recent research identified mtCU as a hetero-oligomeric complex of subunits, the "uniplex" (3). Of these subunits, the pore-forming MCU and the single-pass transmembrane protein EMRE have proven essential for functional channel formation in mammalian cells (4 -7). MICU1 and -2 are Ca 2ϩ -sensing EF-hand proteins in the intermembrane space that regulate Ca 2ϩ gating of the channel, and MICU1 seems to be required for the association of MICU2 with the mtCU complex (uniplex) (see Ref. 8 for a review). MICU1 has been proposed to require EMRE to associate with the mtCU complex (3). In this regard, EMRE has been recently shown to interact with MCU via its transmembrane domain and with MICU1 via its C-terminal acidic tail in the intermembrane space. This would help to retain MICU1 with the MCU-EMRE complex; otherwise, MICU1 can bind to MCU without EMRE (5). The functionally dominant negative MCU homologue MCUb can directly hetero-oligomerize with MCU (9). Although tissue variations occur, optimal activation of mtCU to decode rapid signaling events usually requires cytosolic [Ca 2ϩ ] ([Ca 2ϩ ] c ) exposure that exceeds the global peak levels (10). Such exposure is established at focal associations with the SR in close proximity of activated Ca 2ϩ release channels (RyR2 in the heart). The myocardial contractile cycle is predominantly fueled by mitochondrial oxidative metabolism and driven by RyR2-derived Ca 2ϩ signals. Propagation of these signals to the mitochondria as a feed-forward metabolic tuning loop (excitation-energetic coupling) has long been established (11)(12)(13)(14). It has been recently suggested that mitochondrial Ca 2ϩ uptake during "baseline" pacing is very small; it does not shape [Ca 2ϩ ] c spikes but becomes significant upon stress (15). Germ line ablation of MCU in mice also suggested dispensability for baseline myocardial activity and even for stress tolerance (16,17). In conditional, short term cardiac MCU KO, on the other hand, impairment of metabolic and contractile adaptation to adrenergic stress revealed the need for mitochondrial Ca 2ϩ uptake in the adaptive response (18,19). The above mentioned effects of short term cardiac MCU ablation were also phenocopied by cardiac muscle-specific transgenic expression of a dominant negative MCU (20).
With a large mitochondrial population perpetually being exposed to rapid [Ca 2ϩ ] c oscillations, precisely controlled mitochondrial delivery of Ca 2ϩ signals is critical in the ventricular cardiomyocytes to support excitation-energetic coupling without Ca 2ϩ overload. In line with a tighter Ca 2ϩ uptake control, cardiac muscle mitoplasts presented the lowest mtCU activity (current density) among a range of different tissues (21). Because adult cardiac mitochondria are relatively large with greatly enhanced IMM surface by abundant cristae invaginations, the relatively scarce mtCU, if evenly distributed, would have little chance to effectively mediate local delivery of Ca 2ϩ from RyR2-derived nanodomains. We have shown earlier that junctional SR (jSR or dyadic SR)-mitochondria associations align with IMM-OMM contact points (hereafter referred to as "contact points") and reasoned that this would shorten the path for Ca 2ϩ from the RyR2 to the IMM (22). This implied that mtCU would be present at the contact points, with which mitochondria-jSR associations are aligned. With the molecular building blocks revealed (23), it is now testable whether location-specific variances in mtCU could contribute to effective Ca 2ϩ signal reception by cardiac mitochondria. Here, via fractionation and immunovisualization approaches as well as Ca 2ϩ uptake assays, we identify biases in the localization and subunit composition of the mtCU that lead to more effective Ca 2ϩ uptake at the area of mitochondria-jSR associations.

Results
Distribution Bias of mtCU toward Contact Points-Mitochondria-jSR associations have been shown to form over contact point-containing mitochondrial surface segments (22). We speculated that SR-derived Ca 2ϩ release would expose mtCU optimally if it was located at contact points and/or the inner boundary membrane as opposed to being deep inside the cristae. To this end, submitochondrial membrane fractions (Fig.  1A) enriched in OMM, contact points, and IMM were probed for the mtCU constituents MCU and EMRE. For reference, the OMM porin voltage-dependent anion channel (VDAC) was most abundant in the OMM-rich fraction (severalfold enrichment against the contact point-rich fraction; Fig. 1, B and C). The IMM protein mitofilin/Mic60, a constituent of cristae junctions (and contact points) showed similar enrichment in the contact point-rich fraction against the IMM-rich fraction but was practically absent from the OMM-rich fraction. The SR-resident calsequestrin (CSQ; marking mitochondria-associated jSR) was most (ϳ2-fold) enriched in the contact pointrich fraction (Fig. 1, B and C). Importantly, MCU and EMRE both showed enrichment similar to that of CSQ and mitofilin/ Mic60 in the contact point-rich fraction when compared with the IMM-rich fraction, consistently with a contact point-directed location bias. Finally, another IMM-resident prohibitin with no reported contact point preference was similarly abundant in contact point-rich and IMM-rich fractions while essentially missing from the OMM-rich fraction. To verify a preferred localization to the inner boundary membrane in situ, MCU immunofluorescence distribution was referenced to the matrix protein mtHsp70 in glass-mounted mitochondria (mitochondrial fraction) by means of 3D super-resolution/single-molecule microscopy (Fig. 2). mtHsp70 immunofluorescence was broadly almost completely overlapping with the matrix probe MtTrRed in isolated rat cardiomyocytes when visualized via confocal microscopy ( Fig. 2A). In the glassmounted mitochondrial fraction, the mtHsp70 immunofluorescent particles, imaged via 3D super-resolution microscopy, essentially "filled" the individual mitochondrial volumes (Fig.  2B). The much scarcer anti-MCU particles were ϳ2.5 times more prevalent at the periphery than inside the volume filled by mtHsp70 labeling (Fig. 2B, supplemental Video S-1), consistent with a distribution bias to the inner boundary membrane. Similar results could be obtained when MCU immunofluorescence was referenced to mtHsp70 in primary cardiomyocytes by means of a different super-resolution approach (see Fig. 4D). Taken together, the above results show that mtCU preferentially locates close to the cytosolic boundary of the mitochondrion (Figs. 1, 2, and 4D) at the contact points ( Fig. 1) and so is highly accessible to Ca 2ϩ derived from SR release when entering through the OMM.
Distribution Bias of MCU toward Mitochondria-jSR Associations in Isolated Cardiac Mitochondria-To correlate MCU distribution to mitochondria-jSR associations, its colocalization with RyR2 was evaluated using IF assays. Consistent with earlier published data (17), MCU protein was absent in the MCU KO cells as confirmed by Western blot (data not shown). The mitochondrial fraction is known to contain mitochondriaassociated jSR fragments that occur in confocal images as discrete RyR2 immunofluorescent particles over the mitochondrial surface and that are capable of RyR2-dependent Ca 2ϩ delivery to the mitochondrial matrix (24). First, we used this preparation for IF experiments to minimize cross-reactions that were apparent in primary cardiomyocytes when using the Sigma HPA016480 MCU antibody. For the MCU KO mitochondrial fraction, the IF labeling by this antibody was Ͻ10% of that in the control (Fig. 3A). Also, specific anti-MCU labeling for the mitochondrial fraction required permeabilization (not shown), confirming that the antibody had to cross organelle membrane(s) (OMM and IMM) to bind its epitope. Anti-MCU and anti-RyR2 IF colocalization was examined in the mitochondrial fraction of rat and mouse heart. The corresponding IF patterns counterlabeled with the mitochondrial matrix probe MtTrRed are shown in Fig. 3, B (rat) and C (mouse). Colocalization analysis of MCU with RyR2 is schematized in Fig. 3D, depicting a mitochondrion associated with a jSR and two MCUs located at or away from the jSR interface (top). The association with jSR is mostly confined to the transversal side (one or both). In the middle, a mitochondrion is represented by a cylinder with equal height and diameter (not elongated). For such a shape one transversal side would represent ϳ17% (onesixth) of the total surface, which is also the portion of MCU that would locate there if evenly distributed. This percentage would be even lower for an elongated cylinder that would correspond to the majority of intermyofibrillar mitochondria. The colored discs at the bottom illustrate the fluorescence signals emitted by the IF fluorophores or MtTrRed with consideration of the ϳ250 -330 nm diffraction limit in resolution. After binarization (thresholding), anti-MCU and anti-RyR2 immunofluorescent spots with overlap (crossing, x) were accounted as colocalized (MCU x and RyR x ; non-colocalized are ⅐ MCU and ⅐ RyR). The percentage overlap between MCU and RyR2 IF spot areas (yellow segment) was ϳ12-13%, consistent with the two mole-cules residing in different compartments. Notably, most RyR2 spots did display overlap with MtTrRed (regardless of colocalization with MCU), suggesting that they resided in mitochondria-associated jSR. MCU and RyR2 IF spot counts per field (of ϳ1,300 m 2 ) were 30.5 Ϯ 2 and 57 Ϯ 3.6, respectively, in the rat, whereas they were 42.5 Ϯ 3.5 and 134 Ϯ 7 in the mouse. Importantly, the percentage of MCU IF spots colocalizing with RyR2 spots was 50 Ϯ 4.5% for the rat and 56 Ϯ 3% for the mouse (Fig. 3E). Thus, over half of the distinct MCU-positive structures had overlap with RyR2 labeling, which is more than what would be expected upon even distribution. Of note, in murine cardiac muscle, not all mitochondria have close contact with dyads, and some of the anti-MCU IF spots that did not colocalize with RyR2 must have derived from those mitochondria.
Distribution Bias of MCU toward Mitochondria-jSR Associations in Primary Adult Cardiomyocytes-To further establish MCU distribution in the cardiac muscle mitochondria in situ in the context of the myocyte architecture, colocalization of MCU FIGURE 1. MCU and EMRE are enriched in the mitochondrial contact points. A, submitochondrial membrane fractions obtained via osmotic and ultrasonic rupture of mitochondria from rat heart homogenates were separated on a 30 -60% quasicontinuous sucrose gradient (layered as shown with fraction numbers aligned on the left). On the right, enzyme profiles (top) of the OMM-resident MAO and the IMM-bound SDH along with the protein profile (bottom) of the submitochondrial fractions are shown. Fractions from the two protein peaks richest in OMM (OMM, first MAO peak before SDH increase), contact points (CP, second MAO peak at high SDH), and IMM (IMM, MAO minimum after its second peak) are labeled accordingly and selected by gray rectangles. B, representative WB of the indicated mitochondrial and SR proteins in the indicated submitochondrial fractions. Some of the blots also contain non-relevant lanes (sample prepared/treated differently) that are labeled with a hyphen. C, relative abundances (normalized to CP) of OMM proteins (the voltage-dependent anion channels VDAC1/2), IMM proteins (mitofilin/Mic60, MCU, EMRE, prohibitin), and SR-resident proteins (calsequestrin, CSQ). Note the enrichment of mitofilin/ Mic60, MCU, EMRE, and CSQ but not prohibitin in the CP (n ϭ 3 rats, for the enzyme assays 3 technical replicates each). Error bars, S.E. *, p Յ 0.05; **, p Յ 0.01. and RyR2 IF in freshly isolated primary adult cardiomyocytes was investigated. To better resolve the constellation of the proteins, two super-resolution approaches were employed. The primary anti-MCU antibody had to be changed from the Sigma HPA016480 to the Cell Signaling Technologies D2Z3B because the former displayed too much nonspecific labeling in the MCU KO myocytes (Fig. 4A, i-iv versus v-x). First, using the same fluorophores as for the isolated mitochondria (Fig. 3), the IF distributions were imaged using the Airyscan detector of the Zeiss LSM880 confocal system. This technology allows us to go ϫϳ1.7 beyond the diffraction limit in an otherwise conventional laser-scanning confocal setup. Fig. 4B shows the overview of MCU and RyR2 IF distribution over a myocyte (iii-v). The enlarged area (vi) reveals numerous MCU particles (red) in close vicinity of RyR2 (green). For reference, MCU IF was also examined in the context of the matrix protein mtHsp70 (Fig. 4, B (vii-x) and D). Due to the higher resolution, colocalization was evaluated based on proximity and not on actual overlapping fluorescence pixels; the binarized RyR2 fluorescence masks were expanded radially by 66 nm (1 pixel), and the MCU masks overlapping with the expanded RyR2 mask were counted (Fig. 4C). As the bar graph in Fig. 4C shows, 45.7 Ϯ 1.3% of the MCU spots were in the close vicinity of RyR2, whereas Ͼ85% colocalized with mtHsp70. MCU IF was also referenced to the outlines of clearly distinguishable intermyofibrillar mitochondria defined by the mtHsp70 fluorescence. Here, over individual mitochondrial (mtHsp70) areas, MCU immunofluorescent particles were grouped by whether or not they were in contact with the boundary line (peripheral) and also by how many peripheral MCUs were located on transversal and longitudinal sides (Fig. 4D). Consistent with a location preference toward the inner boundary membrane/contact points, ϳ85% of the MCU particles were peripheral (Fig. 4E). Importantly, 52 Ϯ 4% of the MCU IF particles were in contact with transversal sides (where dyad association can occur), whereas only 33 Ϯ 4% were in contact with longitudinal sides despite their elongated shape (Fig. 4E, T versus L).
Finally, MCU and RyR2 IF distribution were also evaluated via true nanoscopy using the Vutara350 single-molecule imaging system. Fig. 5A shows example images of the point cloud representation of the secondary antibody-conjugated fluorophores CF TM 568 (RyR2; green) and Alexa Fluor (AF) 647 (MCU; red) without further image filtering (i and ii) and after applying a 5 ϫ 5-pixel Gaussian blur (iii and iv). Colocalization of MCU and RyR2 was evaluated via nearest neighbor distance analysis in the non-filtered point cloud. Ͼ55% of the MCUbound antibodies had an RyR2-bound neighbor in Յ150-nm distance, and ϳ40% had a neighbor in Յ100-nm distance (Fig.  5B, top bar chart), numbers that are quite consistent with the colocalization data obtained by the confocal and Airyscan approaches. For reference, ϳ95% of the MCU-bound antibodies had an mtHSP70-bound neighbor in Յ100-nm distance and 100% in Յ150-nm distance (Fig. 5B, bottom bar chart). Collectively, these colocalization data obtained from primary adult cardiomyocytes further reinforced an MCU distribution bias toward mitochondria-dyad associations, consistent with the idea of mtCU hot spots in these areas.
MCU Localization to the Dyads Is Promoted by Ca 2ϩ Signaling Activity-As for what would drive the above established MCU distribution bias toward the dyad areas, one obvious candidate was the Ca 2ϩ signaling activity attributed to the dyads. To this end, the colocalization of MCU and RyR2 IF was (re)evaluated in freshly isolated adult ventricular myocytes that were superfused either by Ca 2ϩ -containing ([Ca 2ϩ ] e ϳ1 mM) or in quasi-Ca 2ϩ -free ([Ca 2ϩ ] e ϳ2 M, just enough to prevent detachment) incubation buffer under 2-Hz electric field stimu-FIGURE 2. MCU distribution and clusterization in glass-mounted mitochondria resolved by 3D super-resolution microscopy. Anti-MCU IF distribution was correlated with that of the matrix marker mtHsp70 in coverglass-mounted mitochondrial fraction isolated from rat heart. Super-resolution imaging was carried out as described under "Experimental Procedures." A, reference confocal images showing closely overlapping mtHsp70 IF (green) and (matrix-loaded) MtTrRed fluorescence (red) in a primary cardiomyocyte. B, low magnification overview (i) of a group of mitochondria defined by discrete 3D (no rotation, z axis is perpendicular to the image plane) clouds of mtHsp70 immunofluorescent particles (Alexa 647-conjugated antibodies). ii, IF control without primary antibody. The mitochondrion in the gray rectangle (i) is enlarged (iii) and shown in two additional rotation angles (iv and v; see the red-green-blue x-y-z arrows for orientation) to reveal the position of the anti-MCU labeling (green dots representing Cy3b-conjugated secondary antibodies) in the volume filled by the mtHsp70. In the different rotational views, the numbers label MCU-positive "peripheral" particles that in the given plane could be moved away without having to pass/cross mtHsp70 particles (cyan numbers) and "subperipheral or internal" particles that are deeper in the mtHsp70 point cloud (white numbers). Note that in the different rotational views peripheral particles may appear inside the red point cloud when they are positioned in front or behind the cloud (e.g. particles 2 and 4 in rotation panel iii look subperipheral (white), whereas in rotation panel iv, they turn out to be peripheral (cyan)). An anti-MCU particle was classified as "peripheral" if in any rotation plane it was found to be peripheral. The bar chart summarizes these evaluations (n ϭ 1 rat). Error bars, S.E. lation before fixation. Contractile activity was suspended pharmacologically (2,3-butanedione monoxime (BDM)). As Fig. 6A shows, there was no significant difference in the RyR2 IF pattern and metrics (sarcoplasm area occupied was 8.7 Ϯ 0.2 versus 8.6 Ϯ 0.7%) at the different [Ca 2ϩ ] e levels. The size of MCU IF particles did not change either (mean Feret's diameter was 0.33 m for both). However, the extracellular Ca 2ϩ removal that eliminated the pacing-associated [Ca 2ϩ ] c oscillations (Fig. 6B) also significantly decreased the percentage of MCU particles in close vicinity of RyR2 (Fig. 6C). These data thus strongly suggest that Ca 2ϩ signaling activity is one of the driving mechanisms that recruit MCU to the jSR interface areas.

The SR Fraction Contains Mitochondria with Extensive Reticular Associations and with a Distinct Protein Profile with High
Levels of EMRE-The tissue homogenates, from which mitochondrial and SR fractions were isolated, had been prepared without enzymatic digestion to preserve SR-mitochondrial tethers. We postulated that although the SR fraction contained  Fig. 7C) mitochondria. However, a larger portion of the mitochondrial cross-sections displayed associations with membrane vesicles in the SR fraction than in the mitochondrial fraction (Fig. 7E, top). Also, for vesicle-associated mitochondria, the number of vesicles per mitochondrion and even more per the mitochondrial perimeter length was substantially larger in the SR than in the mitochondrial fraction (Fig. 7E).
Glass-attached MtTrRed-loaded mitochondria of the SR fraction had more RyR2 IF-labeled particles than those of the mitochondrial fraction; however, the number of RyR2 particles per area unit of the MtTr labeling was disproportionally larger, consistently with more extensive mitochondrial-jSR associations (Fig. 8A).
If the mtCU complex was concentrated to mitochondria-jSR associations, its constituents would probably be more represented among the mitochondrial proteins of the SR than the iii and iv, an overview of the Sigma antibody distribution over a myocyte and a ϫ8 zoom of the indicated area, respectively. Note the extensive IF labeling throughout the cell. vii and viii, overviews of RyR2 (green; AF 488) and the Cell Signaling MCU antibody (red; AF 647) IF distribution over a myocyte separately and then overlaid in ix and enlarged (ϫ8) from the indicated area in x. Note that the bulk of the sarcoplasmic area is clear of MCU IF (also compare with B, iv-vi) with the exemption of the nuclear areas (nuc) and some aggregates mostly at the cell edges. B, IF co-labeling of MCU (Cell Signaling) with RyR2 (iii-vi) and MCU with mtHsp70 (vii-x) in control cardiomyocytes. i and ii, show MtTrRed distribution and negative IF control, respectively. iii and iv, overviews of RyR2 and MCU IF distribution over a myocyte separately and then overlaid in v, and enlarged (ϫ8) in vi. Following the same order of image arrangement, mtHsp70 and MCU IF distributions in a myocyte are shown in vii-x. C, colocalization analysis of MCU ϫ RyR2 and MCU ϫ mtHsp70 IF. Due to the higher resolution, colocalization was determined by proximity and not actual IF overlap as depicted in the scheme and explained further under "Results." The ϳ75-nm radius for RyR fluorophore (ϳ520-nm emission 3 ϳ260-nm diffraction limit for particle diameter) is calculated by dividing the estimated diffraction-limited value by 1.7. The bar chart shows the percentage of MCU IF particles at Յ66-nm distance from the edge of RyR2 (black bar) and mtHsp70 particles (gray bar). D, MCU distribution analysis over individual mitochondria. MCU IF particle distribution was referenced to masks drawn over individual mitochondria (mtHsp70 IF). Thresholded MCU particles over each mask were classified according to whether or not they contacted the contour (peripheral (peri.) versus internal (int.)) and whether they contacted the transversal (T) or longitudinal (L) contour side. The bar chart in E summarizes the analysis. n ϭ 4 control mice, 2 MCU KO mice; technical replicates, 10 cells each. Error bars, S.E. mitochondrial fraction. To this end, the relative abundance of different mitochondrial proteins was determined in SR and mitochondrial fractions of rat and mouse heart homogenates. As shown in Fig. 8, B and C (rat) and D and E (mouse), SR-resident proteins calsequestrin and calnexin were strongly enriched in the SR fraction, whereas some IMM and matrix proteins (Tim23, NCLX, and mtHsp70) were enriched 2-fold to severalfold in the mitochondrial fraction. Strikingly, the mandatory mtCU component EMRE and the cristae junction constituent mitofilin/Mic60 were strongly (Ն2-fold) enriched in the SR fraction, in line with the idea that the more "SRwrapped" mitochondria in the SR fraction hosted more mtCU. Notably, a substantial pool of mitochondrial proteins displayed no or Ͻ50% (1.5-fold) enrichment in either fractions. MCU was falling in this pool with just slightly (not significantly) higher abundance in the rat heart mitochondrial fraction (Fig. 8, B and C). MICU1 enrichment among the fractions of mouse heart followed that of MCU (Fig. 8, D and E). MICU1 was tested only for the mouse tissue, where the antibody specificity could be referenced by MICU1 KO (25).
Interestingly, the enrichment of EMRE and mitofilin/Mic60 in the cardiac SR fraction relative to the mitochondrial fraction was more conspicuous in mice than it was in the rat (Fig. 8, compare B and C with D and E). On the other hand, when the crude mitochondrial and SR fractions of skeletal muscle (rat hind limb) were compared, both EMRE and MCU were enriched in the mitochondrial fraction to similar extents (Fig. 8,  F and G). The extent of enrichment in the skeletal muscle mitochondrial versus SR fraction for EMRE and MCU was still much smaller than for mtHsp70 (Ͼ40-fold), suggesting that there was contamination of the SR by mitochondrial derivatives but without particular concentration of EMRE or MCU to them (Fig. 8,  F and G).
The notion that MCU abundance was not higher in the SR versus the mitochondrial fraction led us to speculate that a significant portion of the MCU could have still located to SR-associated areas in the (crude) mitochondrial fraction. Indeed, isolation of mitochondria-associated reticular membranes (MAMs) is frequently done by separating the crude mitochondrial fraction on a Percoll gradient into a "light" (MAM-enriched) and "heavy" (purified mitochondrial) band (26). We thus decided to compare the protein abundance profiles of the SR versus crude mitochondrial versus Percoll-purified mitochondrial fractions (Fig. 9). As expected, SR-resident calsequestrin levels were severalfold lower in Percoll-purified versus crude mitochondrial fractions (Fig. 9, B (i) and C). Strikingly, the Percoll purification that diminished the SR-resident protein content also significantly reduced the MCU levels (Fig. 9B, ii), thus turning the direction of enrichment toward the SR fraction (Fig. 9C, note the crude mitochondrial fraction-directed downward deflection for MCU in the top panel versus the SR fraction-directed upward deflection in the bottom panel). This way the ratio of MCU abundance in the SR versus the Percoll-purified mitochondrial fraction became similar to that of VDAC (Fig. 9C). EMRE levels also decreased significantly upon Percoll purification of the mitochondrial fraction (Fig. 10B, iii). On the other hand, IMM-resident prohibitin levels displayed higher values in the Percoll-purified versus crude mitochondrial frac- , and the right panels show a 5 ϫ 5 Gaussian blur filtering (iii and iv). ii and iv, ϫ4 enlargements of the indicated areas from i and ii, respectively. B, colocalization (proximity) analysis between MCU and RyR2 IF (top bar chart) and between MCU and mtHsp70 (bottom bar chart) based on nearest neighbor distance ranges. The scheme illustrates approximate particle sizes (corresponding to fluorophores conjugated to individual antibodies). n ϭ 3 mice; technical replicates, 5 cells each. Error bars, S.E. tion (Fig. 9B, iv). Although the mean density values themselves were not significantly different, the extent by which they surpassed the prohibitin levels in the SR fraction was significantly larger in the Percoll-purified mitochondria (normalized density differences 0.32 Ϯ 0.1 versus 0.11 Ϯ 0.07, means Ϯ S.E., n ϭ 3, p Ͻ 0.05).  Collectively, these data suggest that in the cardiac muscle mitochondria, both functionally mandatory mtCU constituents EMRE and MCU concentrate to the mitochondria-jSR associations but to a different extent, with EMRE displaying stronger preference for the extensively SR-associated mitochondria in the SR fraction. Also, the contact points aligned with the interface with jSR (22) might have a distinct composition (e.g. host more EMRE and mitofilin). This arrangement is not apparent in the skeletal muscle, in which mtCU expression levels are high (21) and the Ca 2ϩ release units (triads) are more distant from the mitochondrial surface (27).  OCTOBER 28, 2016 • VOLUME 291 • NUMBER 44 MCU and EMRE Contribute to Differently Sized Complexes in the SR and Mitochondrial Fractions-The higher enrichment of EMRE than MCU in the SR fraction raised the possibility of differential molecular composition of the mtCU complex inside and outside the area of mitochondria-jSR associations. To this end, protein complexes from non-denaturing detergent-solubilized mitochondrial and SR fractions had been separated via size exclusion (gel filtration) chromatography, and their MCU and EMRE content profiles were established. Of the 22 fractions collected, fraction 9 corresponded to the void volume, where large molecules completely excluded from entering the column beads were eluted. For a crude reference (separation is based on size and shape, not mass), the ϳ2-MDa blue dextran elutes in this fraction, whereas the ϳ450-kDa apoferritin complex elutes at fraction 17 (arrows in Fig. 10B). The MCU content profile of the variously sized complexes was multimodal in both mitochondrial and SR fractions. In the SR fraction, a steep rise to a plateau of maximum abundance (75-100%) at the void volume (fraction 9) was followed by a steep decline to Ͻ20% from fraction 14 to 18 (Fig. 10B, red trace). In the mitochondrial fraction, on the other hand, maximum MCU abundance was reached in a peak at fraction 10 with the neighboring fractions containing ϳ45% abundance levels, and from there it gradually decreased to 23% by fraction 20, after which it dropped to its minimum (Fig. 10B, black trace). The EMRE content profile of the differently sized complexes in the SR fraction closely followed the MCU content profile throughout the fractions (Fig. 10, C and D, red traces). By contrast, the EMRE content profile of the complexes from the mitochondrial fraction was unimodal. The peak was closely overlapping with the MCU peak but steeply diminished down to Ͻ8% abundance levels by fraction 14 (Fig. 10C, black), thus strongly diverging from both the MCU profile of the mitochondrial fraction (Fig.  10D, black traces) and EMRE profile of the SR fraction (Fig.  10C, black versus red). Overall, these observations suggest the following. (i) The principal constituents of the mtCU complex, MCU and EMRE, appear in a wide range of differently sized protein complexes, with a substantial portion being built into very large complexes, probably in or close to the 10 6 Da range. (ii) Importantly, for the SR fraction, the MCU content and In similar comparative assays as in Fig. 8, mitochondrial and (for reference) SR protein levels were compared between crude and Percoll-purified mitochondrial fraction as well as the SR fraction isolated from rat heart. A, representative WB of mitochondrial and SR-resident proteins in the three fractions as labeled (cMt and pMt are crude and Percoll-purified mitochondrial fractions, respectively). B, bar charts showing the the relative abundance of the SR-resident calsequestrin (i) and the mitochondrial proteins MCU (ii), EMRE (iii), and prohibitin (iv). C, -fold enrichment for all proteins analyzed in SR versus crude mitochondrial (top) and SR versus Percoll-purified mitochondrial (bottom) fractions expressed as abundance ratios as in Fig. 8C. n ϭ 3 rats. *, p Յ 0.05; **, p Յ 0.01.

Calcium Uniporter Hot Spots at Mitochondrion-SR Association
EMRE content profiles are similar over the resolved size range of complexes, whereas for the mitochondrial fraction, MCU and EMRE contents diverge toward the smaller sized complexes, some of which may contain MCU but lack EMRE. Hence, a part of the pool of MCU outside the mitochondria-jSR association area may exist in (relatively small) functionally compromised or silent complexes that incorporate suboptimal or no EMRE. mtCU-mediated Ca 2ϩ Uptake Is More Effective into the Mitochondria of the SR Fraction-To test whether the above identified biases in MCU and EMRE distribution impacted mtCU function, mitochondrial 45 Ca 2ϩ isotope uptake assays were carried out in suspensions of mitochondrial and SR fractions isolated from the rat heart. As mentioned before (Fig. 7, A-D), the SR fraction contained fewer and smaller mitochondria per total protein. Cristae details also appeared less sharply defined in the mitochondria of the SR fraction (e.g. compare the enlarged images in Fig. 7A versus those in Fig. 7B). ⌬⌿ m -dependent accumulation of MtTrRed appeared also weaker in the coverglass-attached mitochondria of the SR fraction (Fig. 8A, bar  chart ii). For comparative assessment of mtCU activity, compensation had to be applied for these differences. Activity assays for the matrix enzyme, citrate synthase, have been commonly utilized as a quantitative reference for the content of intact mitochondria in suspensions. Respiratory enzyme assays are frequently normalized to the citrate synthase activity (e.g. see Ref. 28). We thus decided to normalize the measured mtCU-mediated Ca 2ϩ uptake activities to the citrate synthase activities in the SR and mitochondrial fractions.
MtCU-mediated Ca 2ϩ uptake was defined as the Ru360-sensitive Ca 2ϩ accumulation at pharmacologically blocked mito-chondrial Ca 2ϩ efflux pathways and disabled/depleted SR Ca 2ϩ store. The initial uptake was determined after elevating [Ca 2ϩ ] o by a pretitrated CaCl 2 / 45 CaCl 2 bolus for 15 s. The titration was done by cuvette fluorometry using the ratiometric high and low affinity indicators Fura-2 and Fura-loAff. Recent direct Ca 2ϩ measurements in the dyadic cleft of rat cardiomyocytes estimated local systolic [Ca 2ϩ ] c peaks inside the clefts to be Ն1.1-1.5 M (29). This was used as crude estimate for the local [Ca 2ϩ ] peak in the mitochondria-jSR cleft (which is supposed to be smaller than in the dyadic cleft, given the larger distance from RyR2s). Two [Ca 2ϩ ] o levels were tested: ϳ300 nM and ϳ1.5 M. Most of the initial Ca 2ϩ accumulation happened via mtCU with a relatively small Ru360-insensitive component (Fig. 11A). Without correction, initial Ca 2ϩ accumulation at [Ca 2ϩ ] o ϳ1.5 M was significantly greater in the mitochondrial fraction (1.61 Ϯ 0.11 versus 0.76 Ϯ 0.11 nmol/mg protein; Fig. 11B), whereas at [Ca 2ϩ ] o ϳ300 nM, the difference was negligible. Citrate synthase activity was on average ϳ7.5 (7.55 Ϯ 0.6, n ϭ 4)-fold higher in the mitochondrial than in the SR fraction at equal total protein concentrations (Fig. 11C, left). This -fold difference, individually determined for each experiment, was used as the correction factor for the mitochondrial Ca 2ϩ uptake. For reference, increasing the total protein of the SR fraction by 10-fold increased its citrate synthase activity level proportionally, slightly surpassing that in the mitochondrial fraction (Fig. 11C, ii (red bar) versus i (black bar)). Increasing the total protein of the mitochondrial fraction by 10-fold using heat-inactivated (boiled) suspension did not change significantly its citrate synthase activity (Fig. 11C, i versus ii (black  bars)). IMM integrity was also assessed by measuring the portion of citrate synthase activity that did not require membrane permeabilization. This portion ("leaky mitochondria") was ϳ20% in the mitochondrial, whereas it was ϳ55% in the SR fraction (Fig. 11C, iii); this -fold difference was not included in the correction to avoid overshooting. After correcting to the relative citrate synthase activities, the initial Ca 2ϩ uptake into the mitochondria of the SR fraction became significantly larger than that into the mitochondria of the mitochondrial fraction at both tested [ Fig. 11D). FIGURE 11. Greater mitochondrial Ca 2؉ uptake efficacy in the SR fraction than in the mitochondrial fraction of the cardiac but not skeletal muscle. Ca 2ϩ uptake assays based on 45 Ca 2ϩ isotope retention. Ca 2ϩ uptake activation was determined from the initial (15-s) 45 Ca 2ϩ accumulation in suspensions of mitochondrial (Mt; black) and SR (SR; red) fractions of cardiac (A-D) and skeletal (E) muscle. The SR Ca 2ϩ store was depleted, and SR Ca 2ϩ uptake and mitochondrial Ca 2ϩ extrusion pathways (mitochondrial Na ϩ /Ca 2ϩ exchanger and the permeability transition pore) were pharmacologically blocked (by thapsigargin, CGP-37157, and cyclosporine A, respectively). A, initial Ca 2ϩ uptake at [Ca 2ϩ ] o ϳ0.3 and ϳ1.5 M in the absence (left) and presence (right) of Ru360 (10 M) in cardiac mitochondrial and SR fractions. B, mtCU-mediated (Ru360-sensitive) initial Ca 2ϩ uptake (Ca 2ϩ uptake mtCU ) calculated from A. C, citrate synthase activity in the mitochondrial and SR fractions at equal protein amount per assay (left) or, for reference, with 10 times more unaltered protein in the SR than in the mitochondrial fraction (middle). In the latter case, to maintain equal proteins, the unaltered mitochondrial fraction was supplemented with nine portions of heat-denatured/boiled mitochondrial fraction. Right, percentage portion of citrate synthase activity in the two fractions that did not require membrane permeabilization, corresponding to mitochondria with compromised IMM integrity (leaky/broken mitochondria). D, initial Ca 2ϩ uptake mtCU adjusted by the -fold difference in citrate synthase activity per total protein between the mitochondrial and SR fractions (from C, as the arrow shows). E, the same experiments as in A-D were done using skeletal muscle mitochondrial and SR fractions. For all, n ϭ 3 rats; 3 technical replicates for each point. Error bars, S.E.
The same assays were also performed using mitochondrial and SR fractions of the skeletal muscle, in which EMRE and MCU were both strongly enriched in the mitochondrial fraction (Fig. 8, F and G). By strong contrast to the cardiac muscle, the mtCU-mediated initial mitochondrial Ca 2ϩ uptake was severalfold larger at [Ca 2ϩ ] o in the micromolar range even after correcting to the difference in citrate synthase activities (which was smaller than in the heart, 3.4 Ϯ 0.75-fold) (Fig. 11E).
Together, these functional assays are consistent with more extensive jSR-associated mitochondrial regions in the cardiac muscle SR fraction to host mtCU in higher density and/or with more competence than those in the mitochondrial fraction with less associated jSR. Such functional bias is not present in the skeletal muscle, where the mtCU components EMRE and MCU are not concentrated to the SR-associated mitochondrial segments.

Discussion
We tested the hypothesis that in cardiac muscle, mitochondria mtCU-mediated Ca 2ϩ uptake activity is biased toward the mitochondria-jSR associations, the sites of local exposure to RyR2-derived Ca 2ϩ signals. Multiple levels of biases were found when the distribution of the essential channel constituents MCU and EMRE were interrogated. Both proteins displayed preference for the contact points and inner boundary membrane as opposed to being deep in the cristae folding, as resolved via membrane fractionation and super-resolution approaches in both isolated mitochondria and freshly isolated primary cardiomyocytes. mtCU in some non-muscle cells has been proposed to also operate inside the cristae so that opening and tightening cristae junctions would regulate mtCU exposure to Ca 2ϩ signals (30,31). Such topology in adult myocardial mitochondria would require a lot of mtCU to populate all of the numerous cristae for an effective local Ca 2ϩ signal reception. This seems unlikely, because mtCU current densities in cardiac mitoplasts have been measured to be the lowest of a large range of mouse tissues (21). Being placed close to the cyto/sarcoplasmic boundary presumably offers mtCU better exposure to Ca 2ϩ entering the intermembrane space.
The immunovisualization approaches to resolve mtCU distribution were focused around anti-MCU IF. Suitable EMRE antibody for IF and knock-out animals for tissue-specific verification are still awaited. Homogenous mitochondrial membrane distribution of anti-MCU IF has been recently reported in rabbit cardiomyocytes (32). All of our cellular models (rat and mouse ventricular myocytes, MEFs, HeLa) showed clustered anti-MCU labeling over individual mitochondria. Also, the antibody used in the referenced work (Sigma HPA016480) produced strong nonspecific cross-labeling all across the cytoplasm of MCU KO cardiomyocytes so that we could use it only to label isolated mitochondria. Here we validated another MCU antibody (CST D2Z3B) that has been used before to determine MCU expression levels in cardiac muscle but not for IF (18), and it performed well across the bulk of the sarcoplasmic area. Immunofluorescent particle analysis using various confocal and super-resolution/single-molecule imaging of crude mitochondrial fraction or freshly isolated cardiomyocytes found ϳ50 Ϯ 5% of the MCU colocalized with RyR2. We interpret this as a distribution bias, considering that the mitochondrial interface formation with dyad(s) in most of the cases engages less than one-third of the entire mitochondrial surface. For this, the resolution had to be sufficient to distinguish anti-MCU IF spots along the four cardinal directions over a mitochondrion. Even with the confocal microscope, the mean anti-MCU IF particle area was Ͼ10-fold smaller than that of the diffusely distributed MtTrRed or cytochrome c, thus sufficiently small to resolve a polarized distribution.
Originally, MCU was suggested to exist in a ϳ480-kDa complex (33) that still remained close to 400 kDa upon ablation of EMRE (3). Higher molecular mass (ϳ720 kDa) complexes have also been described, and their existence depends on EMRE and MICU1/2 (3,4,34). Consistently, size exclusion chromatography suggested very large complexes incorporating a major portion of the MCU and EMRE pool to be on a par with the blue dextran (2 MDa). Nevertheless, MCU contributed to a wide size range of complexes, some of which were clearly smaller than the originally predicted size of the "uniplex" (ϳ480 kDa, which would probably correspond to fractions 16 and 17 of the 22, based on where the apoferritin complex of similar molecular mass elutes). The precise molecular composition and stoichiometry of this wide size range of MCU-containing complexes remain to be elucidated. Nevertheless, the lack or diminishment of EMRE was characteristic of smaller complexes in the mitochondrial fraction. This probably reflected the proposed role of EMRE to connect MICU1/2 oligomers to the mtCU complex (3) via its C-terminal acidic tail in the intermembrane space (5,7). Without EMRE or MICU1, MCU has been shown to stay in a ϳ300-kDa complex (3) that might correspond to MCU-containing complexes from the mitochondrial fraction eluted in fractions 18 and 19. Considering the recently revealed pentameric structure, the MCU oligomer without any accessory proteins would be ϳ175 (5 ϫ 35) kDa (35). Because size exclusion fractions cannot be directly translated to molecular mass ranges, it would be difficult to tell if the end of the size range (fraction 20) could correspond to this pentamer. Loss of MICU1/2 has been connected to enhancement of basal channel activity because of impaired gatekeeping and decreased [Ca 2ϩ ] activation threshold (34, 36 -38). Nevertheless, in vertebrates, the channel activity itself fundamentally requires EMRE (3-6), without which the presence or absence of other regulatory subunits becomes functionally irrelevant. In any case, with the notion that one of the recently proposed functions of EMRE is to retain MICU1 (5), one would expect MICU1 to follow EMRE in its distribution pattern. When MICU1 relative abundance was determined for the mitochondrial and SR fractions, it was similar to MCU, not EMRE. However, because the overall mitochondrial membrane integrity was more compromised in the SR fraction, as revealed by the citrate synthase assays and electron microscopy, the possibility cannot be excluded that a portion of this peripheral/membrane-associated protein of the intermembrane space had been lost during preparation (to a larger extent from the SR fraction). Regardless, because MICU1 seems to have the ability to bind individually to both MCU and EMRE (5), it is not mandated to EMRE-containing complexes only. Thus, further studies with alternative approaches will be needed to establish the potential location-specific contri- OCTOBER 28, 2016 • VOLUME 291 • NUMBER 44 bution of MICU1 to the mtCU complex in cardiac muscle mitochondria.

Calcium Uniporter Hot Spots at Mitochondrion-SR Association
A large portion of the imaging, biochemical, and functional assays presented here to capture the bias(es) in mtCU distribution relied on comparing mitochondrial and SR or submitochondrial fractions. Fig. 12 shows our mechanistic model of how recruitment of EMRE-complemented competent but not EMRE-deprived incompetent MCU complexes to the mitochondria-jSR association area would explain the experimental observations. As shown in Fig. 12 (top), mitochondria in the cardiomyocyte may exist in various shapes/sizes with or without an interface with a dyad. Mechanical homogenization of the ventricular wall yields various mitochondrial and SR formations (Fig. 12, middle). Some mitochondria may exit the tissue without major structural alteration (Whole). Some, especially larger ones or those interconnected via "nanotunnels" (39) could undergo fragmentation (Fragmented). Others may rupture, yielding various IMM-, OMM-, and IMM-OMM-SR-containing (submitochondrial) vesicles (Ruptured). When this homogenate is subjected to differential centrifugation (bottom), extensive jSR associations would make smaller mitochondria (whole or derived from fragmented larger ones) more buoyant and so sediment in the SR fraction. These mitochondria with a large surface portion interfacing with jSR would harbor high levels of competent MCU per mitochondrial protein. Larger "whole" mitochondria with a smaller segment of their surface associated with jSR and mitochondria without a jSR connection would sediment in the mitochondrial fraction and host competent mtCU complexes at the jSR association area and miscellaneous complexes outside of it. We consider complete (whole or fragmented) mitochondria as the main Ca 2ϩ -handling mitochondrial elements in the suspensions, although some contribution of submitochondrial IMM vesicles from ruptured mitochondria was also likely in both fractions.
Verification of biased mtCU-mediated Ca 2ϩ uptake activity at the submitochondrial level in situ in a permeabilized cardiomyocyte is difficult due to various tightly packed Ca 2ϩ diffusion limiters (e.g. sarcomeres, Ca 2ϩ binding sites at/around the mitochondria-jSR interface) that may interfere with globally applied Ca 2ϩ to uniformly expose the mitochondrial surface. T-tubules also tend to be less sensitive to permeabilization (40); thus, the added Ca 2ϩ may better access the bulk of the sarcoplasm than the mitochondrial interface with dyads. Recently, in paced rabbit cardiomyocytes, [Ca 2ϩ ] m signals were found larger at the Z lines (where dyads locate) than at the M lines (32), indicating effective local mitochondrial Ca 2ϩ uptake.
The molecular mechanism(s) that drives MCU and even more EMRE to the mitochondria-jSR association area remains to be elucidated in its entirety. Importantly, here we identified the Ca 2ϩ signaling activity as a significant factor in recruiting MCU into close proximity with RyR2 in freshly isolated ventricular myocytes. Further studies will be needed to identify the Ca 2ϩ sensor that would mediate the recruitment. Potential candidates could be the acidic tail of EMRE, which has been proposed to be analogous to Ca 2ϩ -sensing motifs ("Ca 2ϩ bowl" and "Ca 2ϩ clasp") of some ion channels (6) and the EF-hand regulatory proteins MICU1/2. One would also expect direct or indirect contribution by mitochondria-jSR tethers that secure the organelles together. Cardiac ablation of mitofusin 2, a multifunctional OMM protein that has been proposed to be (41) or not to be (42, 43) a mitochondria-ER tether in non-muscle cells, causes a decrease in mitochondria-jSR associations and Ca 2ϩ transfer, suggesting a tethering role in cardiomyocytes (44). However, the -fold enrichment of EMRE in the SR versus mitochondrial fractions was not affected significantly by cardiac mitofusin 2 ablation (n ϭ 4 mice), suggesting that either EMRE recruitment to the mitochondria-jSR association did not require mitofusin 2 or alternative mechanism(s) compensated for its loss. Of note, EMRE and mitofilin/Mic60, the two IMM proteins found strongly enriched in the SR fraction in our assays, are both single-pass transmembrane proteins, and as such, they probably share membrane insertion (e.g. stop-transfer) (45) and possibly degradation pathways. Because Tim23 and mtHsp70 proteins that assist IMM insertion (45) were highly enriched in the mitochondrial fraction, a location bias in the IMM insertion of EMRE and mitofilin toward mitochondria-jSR association is not likely. Further studies will be needed to evaluate whether the mitochondria-jSR association area could be protective against EMRE degradation or to elucidate possible other mechanisms underlying the strategic mtCU positioning. Because many mitochondrial membrane proteins are multifunctional (e.g. mitofusin 2 in the OMM, Letm1, MCUR1 in the IMM), a possibility remains that the disproportionally high enrichment of EMRE versus MCU in the SR-associated mitochondrial regions might also serve yet to be identified role(s) for EMRE other than being an mtCU constituent.
Strategic mtCU positioning that we have found here in the cardiac muscle of rat and mouse is not necessarily a ubiquitous phenomenon for the mitochondrial-reticular associations that serve local Ca 2ϩ transfer. An arrangement that makes biological sense at low mtCU abundance, large mitochondrial size, and extensive cristae folding in the cardiac muscle might make no sense in other tissue like skeletal muscle with high channel abundance (21), smaller mitochondria, less extensive cristae, and larger distance (Ͼ100 nm (46)) between RyR and the mitochondrial surface. Indeed, when mitochondrial and SR fractions isolated from skeletal muscle of rat were compared, EMRE showed no enrichment in the SR as it did in the cardiac muscle; instead, it showed enrichment in the mitochondrial fraction similar to that of MCU (ϳ2-fold). Consistently, in the mtCU-mediated mitochondrial 45 Ca 2ϩ uptake assays, the mitochondria of skeletal muscle SR fraction underperformed those of the mitochondrial fraction.
Collectively, here we present evidence for multiple levels of biases in the location and molecular composition of the mtCU in the cardiac muscle. Selective concentration of EMRE-complemented mtCU complexes to the area of jSR association would represent a Ca 2ϩ uptake "hot spot" where the [Ca 2ϩ ] nanodomains even at the 30 -50-nm distance from the Ca 2ϩ release units (RyR2s) (13) could effectively activate Ca 2ϩ uptake and support excitation-energetics coupling. Meanwhile, outside the mitochondria-jSR association areas, the global [Ca 2ϩ ] c oscillations would expose a less effective, more dispersed mtCU population that is partly silent or impaired due to missing or suboptimal EMRE module(s), probably as a protective measure from overload, as has been proposed (21). FIGURE 12. Schematics for the mtCU distribution and activity bias toward mitochondria-jSR associations in the cardiac muscle and for the related differences between mitochondrial and SR fractions. Top (Mitochondria in tissue), TEM images of rat ventricular wall segments illustrate the context of differently sized jSR-associated mitochondria (mt) in the tissue. Dyad (dy)-forming SR and T-tubule lumens are shaded in light red (two in the left and one in the right image). Below, two exemplary mitochondrion-jSR associations (involving either a small or a large mitochondrion) are depicted with the proposed bias in mtCU distribution and a third without dyad association. MCU complexes (black rectangles) complemented with EMRE (green rectangles) are more prevalent at the interface with the jSR, whereas mtCU lacking EMRE stays outside the interface. Hence, the short lasting RyR2-derived high [Ca 2ϩ ] nanodomain (depicted as light red shade) exposes an IMM segment containing MCU at higher density and always complemented by EMRE. By contrast, diluted global [Ca 2ϩ ] o rises (not depicted) would expose IMM surfaces outside the dyad interface hosting MCU at smaller density, some of which may not be complemented by EMRE (and so incapacitated). The middle and bottom panels illustrate the potential fates of the mitochondria upon tissue homogenization and isolation via differential centrifugation of the mitochondrial and SR fractions, respectively. In the bottom panel, the mitochondrial derivatives in the two fractions are shown in an elevated [Ca 2ϩ ] o environment, and Ca 2ϩ influx through the activated MCU is depicted by red arrows to explain the differences observed during the 45 Ca 2ϩ uptake assays (Fig. 11). Although there are more active mtCU per mitochondrion in the SR fraction, those mitochondria are fewer (not depicted) with probably weaker ⌬⌿ m and more compromised membrane integrity (paler gray shade in the matrix). Membrane vesicles derived from mitochondrial rupture are also shown for both fractions but not shown as active in Ca 2ϩ uptake because in many of them the ⌬⌿ m generation is probably compromised. See "Results" for more details. tion, the supernatant from the first 9,500 ϫ g centrifuge step was collected and centrifuged again at 40,000 ϫ g (45 min). This pellet was resuspended in 200 l of Basic Medium.
Protein Analysis and WB-A biuret-based reagent kit from Bio-Rad (DC protein assay kit) was used to measure the protein concentrations for the isolated mitochondria and SR. WB was done using a Bio-Rad setup, including a Trans-Blot Turbo transfer system. Equal amounts of total protein were loaded (unless otherwise noted), separated electrophoretically by SDS-PAGE (12-15% gels), and transferred to a nitrocellulose membrane. After a blocking step in LI-COR Odyssey blocking solution (1 h at room temperature), the membrane was incubated overnight at 4°C with primary antibodies. The LI-COR infrared fluorescent secondary antibodies (IRDye 800CW) were used for visualization by means of a LI-COR Odyssey scanner. Band quantification ("densitometry") was performed using ImageJ software (National Institutes of Health).
Isolation of Submitochondrial Membrane Fractions from Rat Heart-Isolation of submitochondrial membrane fractions was performed as described before (22) with slight modification in the sonication parameters for better IMM yield. Briefly, Percoll-purified "heavy" mitochondria were resuspended and incubated in a hyposmotic phosphate buffer ("swelling" buffer: 10 mM NaH 2 PO 4 /Na 2 HPO 4 , pH 7.4) for 20 min. Then a "shrinking buffer" (60% sucrose in 29% of final volume) was added. In turn, mitochondrial membranes were disrupted by sonication (Branson 102C CE converter, 3 ϫ 30 s of irradiation at 22% amplitude, 1-min interval), centrifuged at 6,500 ϫ g for 10 min. The supernatant was loaded on top of a 30 -60% sucrose gradient (see Fig. 1A) and spun at 200,000 ϫ g for 13 h). Fractions of 1 ml were collected. All procedures were done on ice, and centrifuge steps were at 4°C.
Enzyme Assays-Monoamine oxidase (MAO) and succinate dehydrogenase (SDH) activity assays were performed and analyzed as described previously (22) except that all enzymatic reactions were measured in a hybrid microplate reader (Biotek Synergy 4). To evaluate citrate synthase activity and IMM intactness, mitochondrial or SR fractions were pretreated with 0.1% Triton X-100 and diluted in Tris-HCl buffer (0.1 M, pH 8) supplemented with 0.1 mM 5,5Ј-dithiobis(2-nitrobenzoic acid), 0.3 mM acetyl-CoA, and 0.5 mM oxaloacetic acid. Absorbance at 405 nm was tracked for 5 min (10-s intervals). Enzyme activity was expressed as citrate synthase activity units per mg of mitochondrial protein (28). The mitochondrial inner membrane intactness was calculated as the ratio between the citrate synthase activity of the mitochondria preparation in Tris-HCl buffer (0.1 M, pH 8) versus the activity of the same mitochondria preparation treated with 0.1% Triton X-100 lysis buffer (giving the percentage of "leaky" or broken mitochondria).
IF of Cells and Isolated Mitochondria-Membrane organelles from the mitochondrial or SR fraction were glued with Cell-Tak TM to coverglasses and loaded with MitoTrackerRed CMXRos (MtTrRed; 50 nM) during attachment (20 min, 20 -23°C) and then fixed (40 g/liter paraformaldehyde/PBS, 10 min). Intracellular medium (120 mM KCl, 10 mM NaCl, 1 mM KH 2 PO 4 , 20 mM HEPES/Tris, 2 mM NaATP, 2 mM MgCl 2 ) was used for attaching the mitochondria. Fixed mitochondria and cardiomyocytes were subjected to a standard IF staining protocol. 5% goat serum and 0.2% Triton X-100 in PBS were used for blocking and permeabilization. SlowFade was used for mounting on microscope slides. Secondary antibodies were conjugated with Alexa Fluor488 or 647 (to minimize spectral overlap with MtTrRed that is excited at 568 nm). To assess nonspecific binding of secondary antibodies, negative controls without primary antibody were used. Specificity of the anti-MCU primary antibodies was tested in knock-out samples as described under "Results" (Figs. 3A and 4A). Immunofluorescence was imaged using either a Zeiss LSM780MP or LSM510MP confocal microscopes or the LSM880MP system equipped with the Airyscan super-resolution (ϫ1.7 beyond the diffraction limit) detection system. A ϫ63 Zeiss plan-apochromat oil, 1.4 numerical aperture, differential interference contrast lens was used to obtain all images. Image and colocalization analyses were done using FIJI/ImageJ (National Institutes of Health). After subtracting background using a fluorescence threshold mask, the resulting particles with Ͼ0.05-m 2 area were used. The threshold varied between antibodies, but for each antibody, it was the same. Particle count and areas were determined. Next, overlaps of Ͼ0.015 m 2 between different fluorophores were counted, and their area was measured. Colocalizing particles per all particles and the overlapping area per total area were calculated (for each antibody). For the Airyscan images, the same threshold was applied, but then the RyR2 signal was dilated by 66 nm (1 pixel) from the original labeling. Using this new mask, the same colocalization criterion between MCU and RyR2 was applied. 5-10 microscope fields of 1275 m 2 were analyzed for each condition.
3D Super-resolution Imaging-3D super-resolution imaging was done using a Vutara 350 microscope equipped with a biplane module, a Hamamatsu ORCA Flash version 4.0 V2 sCMOS camera, and an Olympus Super-Apochromat ϫ60 water objective with numerical aperture 1.2. Two-color channels were detected sequentially at 50 frames/s (20-ms acquisition time). Laser powers (647-and 561-nm wavelength) were adjusted to provide sufficient fluorophore switching for localization and to minimize photobleaching. Secondary antibodies were conjugated with either Alexa Fluor647, CF TM 568, or Cy3b fluorophores. The IF labeling was performed as described above for cells except that the cells were not mounted but kept at 4°C in PBS until just before the imaging experiments were done. Imaging was performed in an oxidizing buffer containing 10% (w/v) glucose in 10 mM PBS-Tris, pH 7.5, with 10 mM mercaptoethylamine combined with 50 mM ␤-mercaptoethanol, 2 mM cyclooctatetraene that promoted the reversible photoswitching of the fluorophores conjugated with the secondary antibodies. Data analysis was performed using the Vutara SRX software (version 5.22). Single molecules were identified by their brightness frame by frame after removing the background. Identified particles were then localized in three dimensions by fitting the raw data in a customizable region of interest (typically 16 ϫ 16 pixels) centered on each particle in each plane with a 3D model function that was obtained from recorded bead data sets. Fit results were stored as data lists for further analysis. The image resolution capable of experimentally being achieved is 20 nm laterally (x and y) and

Calcium Uniporter Hot Spots at Mitochondrion-SR Association
50 nm axially (in z). The nearest neighbor particle analysis was done automatically by the Vutura SRX software. Peripheral versus internal localization MCU immunolabeling over the 3D cloud of mtHsP70 (matrix) immunolabeling was established based on xyz rotations of individual mitochondria. The positioning of anti-MCU-positive particles in relation to the surface of the mtHsp70-filled volume was determined for 24 randomly selected mitochondria from two different low magnification fields. The software allowed 3D trimming, so information from neighboring mitochondria did not interfere with the rotation views.
Electrical Field Stimulation on Isolated Adult Cardiomyocytes-Electrical field stimulation was performed as described before (48). Briefly, isolated calcium-tolerant cardiomyocytes from adult mouse hearts were placed in 25-mm diameter coverslips pretreated with laminin and mounted in an electrical stimulation chamber equipped with two Pt electrodes (RC-47FSLP, Warner Instruments). Cells were continuously perfused with field stimulation buffer, containing 150 mM NaCl, 5.4 mM KCl, 10 mM HEPES acid, 1 mM glucose, 2 mM pyruvate, 5 mM creatine, and 5 mM taurin, supplemented with 2 M or 1 mM CaCl 2 and BDM (10 mM) to prevent contractions. Electrical biphasic pulses of 20-mV amplitude, 2-Hz frequency, and 5-ms duration were applied to the cells for 10 min (MyoPacer Stimulator, IonOptix).
To record cytosolic Ca 2ϩ transients during field stimulation, cardiomyocytes were loaded with 5 M fura-2 acetoxymethyl ester dissolved into the plating buffer in the presence of 0.003% Pluronic F127 (Sigma-Aldrich P-2443) for 20 min at room temperature, washed, and briefly postincubated to allow hydrolysis of the acetoxymethyl ester within the cells. Imaging was performed using an Olympus IX81 inverted microscope (Fluo/ 340 ϫ40 oil immersion objective, numerical aperture 1.35) fitted with a Lambda DG-4 ultra-high speed wavelength-switch illuminator (Sutter Instruments) and a 512 ϫ 512 EMCCD camera (ImagEM, Hamamatsu, Japan). 340/30-nm and 380/ 20-nm band pass excitation filters, beam splitter 59022Bs, and dual dichroic 59022M were from Chroma Technologies. Cells were excited alternately with 340-nm and 380-nm wavelengths, and changes in the Ͼ500-nm emission were captured. Recordings from each cell were obtained at a rate of 45 ms/frame, and changes in the average ratio (340 nm/380 nm) for previously defined regions of interest were analyzed throughout the stimulation period (Metamorph version 7.8.6.0, Molecular Devices, LLC). Next, cardiomyocytes were fixed and subjected to a standard IF staining protocol and imaged/analyzed as described above for the Zeiss LSM880 Airyscan system. 45 Ca 2ϩ Retention Assays-Mitochondrial or SR fractions (250 g/ml) were rapidly resuspended in 350 l of intracellular medium (120 mM KCl, 10 mM NaCl, 1 mM KH 2 PO 4 , 20 mM HEPES/Tris, 2 mM NaATP, 2 mM MgCl 2 ) supplemented with protease inhibitors (leupeptin, antipain, and pepstatin; 1 g/ml each), thapsigargin (Tg; 2 M), caffeine (10 mM), EGTA/Tris (pH 7.4) (50 M), and malate and pyruvate (1 mM concentration of each to energize mitochondria). Na ϩ -dependent mitochondrial Ca 2ϩ extrusion and the permeability transition pore were blocked by CGP-37157 (20 M) and cyclosporine A (2 M), respectively. After a 20-s incubation at 37°C, the suspension was transferred to an Eppendorf tube containing a CaCl 2 bolus with 8 Ci of 45 Ca 2ϩ to elevate [Ca 2ϩ ] o to 0.3 or 1.5 M. CaCl 2 boluses were pretitrated in fluorometric assays (see below). After a 15-s, 30-s, and 4-min incubation at 37°C, 100-l aliquots were rapidly diluted in 5 ml of ice-cold washing buffer (140 mM KCl, 10 mM Hepes/Tris, 0.5 mM EGTA/Tris, pH 7.2) and filtered through 0.3-m nitrocellulose filters (Millipore). Filters were rinsed once with washing buffer, and 45 (36). Despite the Ca 2ϩ -free isolation procedure at 0 -4°C, significant Ca 2ϩ residue remained in the SR store of the SR fraction. This was discharged by the Tg and caffeine without increasing the (EGTA-stabilized) baseline (which remained Ͻ40 nM) but decreasing the CaCl 2 bolus that was needed to reach the target [Ca 2ϩ ] o .
Gel Filtration (Size Exclusion) Chromatography-Rat mitochondrial and SR fractions were solubilized in an ice-cold buffer containing 3% CHAPS, 120 mM KCl, 20 mM MOPS, and 0.5 mM EDTA for 5 min. Then the resulting sample was centrifuged at 40,000 rpm (ϳ57,000 ϫ g) at 4°C. The supernatant was loaded on top of a Sephacryl-S400 HR chromatography column (GE Healthcare, 17-0609-05). The column was equilibrated in 120 mM KCl, 20 mM MOPS, 0.5 mM EDTA, and 0.75% CHAPS. 24 0.5-ml fractions were collected, and 25 l of each were analyzed by WB. WB profiles were determined from equal volumes of the fractions because there was no significant difference in their protein content over the relevant range (fractions 9 -22; Fig. 10A) TEM-Rat heart was fixed and processed for TEM as described earlier (22). High pressure freezing and freeze-substitution were carried out in the University of Pennsylvania Electron Microscopy Resource Laboratory. Samples (mitochondrial and SR fraction suspension droplets from one rat heart) were frozen vitreously in an Abra HPM010 machine and dehydrated at Ϫ100°C in 100% acetone, 2% OsO 4 , 0.1% uranyl acetate for 48 h. Samples were warmed to room temperature, washed in acetone, and embedded in EPON resin before polymerization. Material containing sample was sectioned at 60-nm thickness, mounted on TEM grids, and secondarily stained with 2% aqueous uranyl acetate and Reynold's lead citrate before observation at 80 keV on a JEOL JEM1010 or FEI Tecnai 12 transmission election microscope. Images were digitally collected on a Hamamatsu CCD camera with AMT software. Organelle and membrane particle metrics were taken by FIJI/ ImageJ. Mitochondria with no major disruption in the OMM line and recognizable IMM/cristae were considered "well defined."