Mitochondrial calcium uptake

Mitochondrial [Ca²⁺]_i Influx in the Heart.

The [Ca²⁺]_i transient in heart cells produces repetitive Ca²⁺ elevations that envelope the mitochondria with every heartbeat. This Ca²⁺ is then removed from the cytosol via the sarcoplasmic reticulum/endoplasmic reticulum Ca²⁺-ATPase (SERCA) and the sarcolemmal Na⁺-Ca²⁺ exchanger (NCX), resulting in cell-wide reduction of [Ca²⁺]_i within 500 ms (36). During this time, mitochondria have the opportunity to sequester Ca²⁺ from the cytosol through open MCUs. To facilitate the comparison of these Ca²⁺ fluxes, we scaled all data to a liter of cytosol (see Table S1). Conventional wisdom is that 70–80 μmol of Ca²⁺ enters a liter of cytosol during a [Ca²⁺]_i transient in ventricular myocytes and that this Ca²⁺ is subsequently removed during a contraction–relaxation cycle (25, 45).

A fair comparison of mitochondrial Ca²⁺ uptake reported by various sources using diverse experimental techniques requires that the data be converted to the same units (i.e., micromoles of Ca²⁺ per liter of cytosol per second; details are provided in Eqs. S1–S3 and Figs. S1–S3). Although the opinions of different groups are substantially at odds with regard to the role of the ensemble of mitochondria as a dynamic Ca²⁺ buffer in the heart, the measurements of mitochondrial Ca²⁺ fluxes appear to be in good agreement with each other when compared quantitatively (Fig. 2). Importantly, the whole-cell mitochondrial Ca²⁺ uptake flux (whole-cell MCU flux) derived from experiments conducted in cells (Fig. 2, filled circles) does not differ appreciably from uptake measured in suspensions of isolated mitochondria (Fig. 2, open circles) once scaled appropriately. The solid black line in Fig. 2A (“Cardiac MCU”) represents an empirical best-fit line to the experimental results of the form y = mx^p, where m = 0.67 and p = 1.7, and this line is subsequently used for a comparison between the whole-cell MCU flux in other cell types and uptake from other Ca²⁺ transport systems (i.e., SERCA and NCX). For example, whole-cell MCU flux in the liver (Fig. 2B, colored data points) is qualitatively similar to that in the heart (Fig. 2B, black line) and is small over the physiological range of [Ca²⁺]_i. Note that when [Ca²⁺]_i is experimentally elevated beyond the physiological range (44) to levels greater than 200 μM, the measured flux is still consistent with the cardiac best-fit line (see above). Interestingly, the whole-cell MCU flux in skeletal muscle (Fig. 2C, colored data points) (28) can exceed that seen in heart (Fig. 2C, solid line) in some cases but is still relatively small over the physiological range of [Ca²⁺]_i. A detailed discussion regarding mitochondrial Ca²⁺ uptake across mitochondria from different tissues is provided below.

When the whole-cell MCU flux is compared with other major cytosolic Ca²⁺ fluxes (i.e., SERCA and NCX) present in the heart over the physiological range of [Ca²⁺]_i (Fig. 2D), both theoretical (46, 47) (Fig. 2D, solid lines) and experimental (25) (Fig. 2D, filled circles) measurements suggest that MCU flux is small by comparison. This dramatic difference in magnitude (Fig. 2D, compare red and blue lines with the black line) suggests that most of the Ca²⁺ that is added to the cytosol during the contraction or during a Ca²⁺ spark is resequestered into the SR by SERCA or extruded from the cell by NCX. This becomes even more apparent when each flux is scaled by all cytosolic removal fluxes (Fig. 2E). Although these values are for small rodents, the result is similar (Fig. S4) for larger animals (e.g., rabbit) that have less SERCA and more NCX activity. Importantly, Fig. 2 D and E clearly shows that fluxes attributed to SERCA and NCX are significantly larger than those of MCU over the physiological range of [Ca²⁺]_i (i.e., 0.1–20 μM). This comparison strongly suggests that mitochondrial fluxes have a minor effect on [Ca²⁺]_i dynamics. Note that this conclusion is consistent with earlier work by Bers and colleagues (48, 49), who showed that mitochondrial uptake accounted for approximately 1% of cytosolic Ca²⁺ “extrusion.” It is, however, important to note that the transport fluxes (SERCA and NCX) do saturate at high [Ca²⁺]_i levels and that this would theoretically allow the whole-cell MCU flux to exceed even SERCA when [Ca²⁺]_i is “superphysiological.” However, fluxes of this magnitude would depolarize the IMM (18), thereby reducing the driving force for Ca²⁺ entry into the mitochondria. Importantly, under normal conditions, even the microdomain [Ca²⁺]_i that bathes the ends of an IFM with high [Ca²⁺]_i is unlikely to exceed 20 μM (39–41) and the remainder of the IFM experiences much less. We believe this to be true for both heart and skeletal muscle.

Biophysical Properties of MCU Ca²⁺ Uptake.

The cardiac whole-cell MCU flux, measured experimentally over a wide range of [Ca²⁺]_i values, can be visualized (Fig. 3) alongside a theoretical flux (blue line) of the form

where N_mito is the number of mitochondria per cell (N_mito = 10,000) (50), N_mcu is the number of MCUs per mitochondrion (N_mcu = 200) [estimated from a single-channel MCU current (i_mcu) and a whole-mitoplast MCU current (I_mcu) by Fieni et al. (21); details regarding this estimate are provided in SI Material], ∆Ψ_m is the mitochondrial inner membrane potential, is the Nernst reversal potential for the IMM, F is Faraday’s constant, V_myo is the cytosolic volume of the cell (18 pL), and P_O is the open probability (P_O = 0.9). Additionally, the single-channel MCU conductance (g_mcu) is assumed to follow a Michaelis–Menten type relationship, with a K_m of 19 mM and a maximal g_mcu of 8.1 pS (17, 18, 20, 51) (Eqs. S7 and S8 and Fig. S7). This relationship between ion channel conductance and ion concentration is a known biophysical feature of selective ionic channels (more information is provided in ref. 52). Additionally, although electrophysiological measurements of MCU currents have been carried out by different investigators using different experimental approaches, the maximal single-channel g_mcu estimated from these measurements appears to be similar (17, 18, 51). Therefore, we assumed a maximal single-channel g_mcu of 8.1 pS in order to estimate the number of MCU channels per cardiac mitochondrion. We calculate that the N_mcu is approximately 200 (details are provided in Eqs. S5–S10 in SI Material). It is important to note that our whole-cell MCU flux formulation (Eq. 1) is robust. Specifically, a change in measured single-channel g_mcu would be balanced by a corresponding change in the estimated N_mcu (Eq. S9). For example, if a twofold increase in single-channel g_mcu were to be observed (i.e., 8.1 to 16.2 pS), a twofold reduction in the N_mcu would be implied. In this scenario, no other elements of our formulation (i.e., Eq. 1) require any additional adjustment. In Eq. 1, the i_mcu (measured in picoamperes, as given by Ohm’s law and the Nernst reversal potential) is converted to a whole-cell MCU flux (J_mcu, in μM s⁻¹) scaled to the cytosolic volume of a ventricular myocyte. When the IMM is polarized, this flux formulation will yield results similar to the Goldman–Hodgkin–Katz formulation. Interestingly, over the physiological range of [Ca²⁺]_i (0.1–20 μM), the theoretical MCU flux (Fig. 3A, blue line) clearly supersedes the measured experimental fluxes (Fig. 3A, gray circles). Additionally, the experimental fluxes appear to increase nonlinearly (Fig. 2A) with [Ca²⁺]_i, suggesting the possibility of Ca²⁺ regulation of MCU uptake. To investigate this issue, we fit the experimental data with the same flux equation scaled by a P_O function given by

where P_max is the maximal open probability, P_min is the minimum open probability, η is the cooperativity factor, and K_m,act is the half-saturation constant for MCU open probability.

The red line in Fig. 3A shows the resulting best-fit line when P_max = 0.9 (18) and η = 1.45, P_min = 0.03, and K_m,act = 108 μM (Fig. 3B). The nonlinear increase in experimental fluxes (Fig. 2A, Cardiac MCU line) and the remarkable agreement between the uptake measurements (Fig. 2A, gray circles) and the red line in Fig. 3A suggest that [Ca²⁺]_i may be responsible for modulating (i.e., activating) MCU Ca²⁺ uptake. Although the details are not clear, such a mechanism has been proposed before (53, 54). One possible mechanism for this Ca²⁺ regulation is mitochondrial calcium uptake 1 (MICU1), which encodes a mitochondrial EF hand protein (55) and appears to regulate MCU uptake. In fact, a recent study showed that MICU1 altered mitochondrial Ca²⁺ uptake at low Ca²⁺ levels (56). There are, however, other possible causes of nonlinear MCU uptake, including the following:

• i) Changes in the ∆Ψ_m. Although large Ca²⁺ uptake fluxes certainly influence ∆Ψ_m (33, 53), these fluxes should depolarize the IMM, causing the slope of Ca²⁺ uptake to decrease with larger [Ca²⁺]_i, rather than increasing as seen in Fig. 2A.

• ii) Active transport pathways:

  • a) The mitochondrial NCX transporter (NCLX) that extrudes Ca²⁺ from the mitochondrial matrix (57), is a possible cause of apparent nonlinear uptake. However, NCLX efflux should cause the slope of the influx to decline, rather than increase. Note that NCLX-based Ca²⁺ influx is also possible under rare conditions (58, 59) but is very unlikely in the conditions shown here.

  • b) Leucine zipper-EF-hand containing transmembrane protein 1 (Letm1), an electrogenic Ca²⁺/H⁺ exchanger that operates at nanomolar [Ca²⁺]_i, is another possible cause of nonlinear uptake. At elevated matrix Ca²⁺ ([Ca²⁺]_m) levels or low cytosolic pH, Letm1 is a Ca²⁺ extrusion mechanism, whereas at low [Ca²⁺]_m, it provides additional Ca²⁺ influx accompanied by matrix alkalinization (60). As such, Letm1 reverses transport direction very early in the Ca²⁺ transient and might contribute to nonlinearity of Ca²⁺ uptake. Of note, Letm1 is also inhibited by Ruthenium Red (RuR) and is not present in excitable cells.

• iii) Extramitochondrial Ca²⁺ buffers. In the isolated mitochondria experiments, the presence of unaccounted Ca²⁺ buffers (e.g., membrane-bound proteins, remnant Ca²⁺ buffers used during mitochondrial isolation) may account for the nonlinear removal of Ca²⁺ from the extramitochondrial solution, resulting in an overestimation of Ca²⁺ uptake. However, studies usually take special care to avoid such artifacts.

• iv) Multiple g_mcu levels. Recent work (51) has indicated the presence of multiple MCU unitary Ca²⁺ conductance levels. The activation of larger conductance levels could create the nonlinear uptake seen in Fig. 2A.

Although we believe that the nonlinearity of the MCU uptake flux visible in Fig. 2A is likely due to Ca²⁺ regulation of the MCU, elucidating the cause of this nonlinearity and the molecular mechanism of Ca²⁺-regulated MCU uptake represents a compelling focus for future experimental investigations.

When experimental conditions push [Ca²⁺]_i beyond the physiological range, the influx of Ca²⁺ causes a substantial degradation of the voltage gradient across the IMM (∆Ψ_m). This lowers the driving force for Ca²⁺ influx and is thought to occur because ∆Ψ_m is sustained in respiring mitochondria primarily by proton pumping systems and not by fast conductance K⁺ channels (53). Experimental measurements under these conditions require ∆Ψ_m to be sustained pharmacologically with a K⁺ ionophore, such as valinomycin (53), or by voltage-clamping a mitoplast (2- to 5-μm vesicles representing the entire inner membrane of a single mitochondrion) (18, 20, 21, 51). A study carried out with mitoplasts from COS-7 cells showed that MCU was a highly selective ion channel. The I_mcu was observed to have a Michaelis–Menten type saturation with a K_m of 19 mM (18). A later study with human cardiac mitoplasts (51) and murine cardiac mitoplasts (20) showed inward Ca²⁺ currents (termed I_mCa1 and I_mcu, respectively) consistent with the same known features of the MCU as first reported in COS-7 cells (18). More recently, MCU activity was measured in mitochondria from different tissue types, and I_mcu in the heart was shown to be significantly lower than in other tissue types (including skeletal muscle, liver, and kidney cells) (21). For comparison, we have converted the heart mitoplast current values to a whole-cell MCU flux and plotted them in Fig. 3A (filled circles). We have also converted the single MCU currents from a study by De Stefani et al. (17) (Eq. 1 and Fig. 3A, blue circle). Again, note the agreement between the theoretical Ca²⁺-activated whole-cell MCU flux values (Fig. 3A, red line) and Ca²⁺ uptake experiments (Fig. 3A, gray circles) conducted at physiological [Ca²⁺]_i levels, and even mitoplast studies (Fig. 3A, green circles) conducted at high [Ca²⁺]_i levels.

I_mcu Magnitude in Cardiac and Skeletal Muscle.

Variation in tissue-dependent mitochondrial Ca²⁺ influx is reported in a recent paper by Kirichok and colleagues (21). Fieni et al. (21) showed that the I_mcu density (measured as picoamperes per picofarads, pA/pF) in mouse skeletal muscle mitoplasts was approximately 28-fold greater than the I_mcu density in cardiac mitoplasts when measured in 100 μM [Ca²⁺]_i at −160 mV. This finding raises an obvious question: Does the larger skeletal muscle I_mcu density mean that skeletal muscle mitochondria have a larger effect on the skeletal muscle [Ca²⁺]_i transient than cardiac mitochondria have on the cardiac [Ca²⁺]_i transient under physiological conditions? Because the surface area of the IMM is proportional to its capacitance (1 μF/cm²) and the capacitances of skeletal muscle mitoplasts were the same as those of cardiac mitoplasts (0.65 pF and 0.67 pF, respectively), one would predict from the results of the study by Fieni et al. (21) that skeletal muscle mitochondria would take up much more Ca²⁺ than cardiac mitochondria. However, as shown in Fig. 2C, over the physiological range of [Ca²⁺]_i (0.1–20 μM), skeletal muscle and cardiac mitochondria take up similar amounts Ca²⁺ (at the whole-cell level). Although this finding seems to be at odds with the findings of the study by Fieni et al. (21), the small mitochondrial volume fraction in skeletal muscle (and possibly the Ca²⁺ dependence of I_mcu) may provide an explanation. There is significant recent evidence that [Ca²⁺]_i-dependent “activation” of the MCU occurs by multiple Ca²⁺-dependent regulatory mechanisms (the physical location of some of these proteins is uncertain), including MICU1 (55, 56), MICU2 (61), mitochondrial calcium uniporter regulator 1 (MCUR1) (62) and Ca²⁺/calmodulin-dependent protein kinase II (CaMKII) (20). Presumably these (and possibly other MCU regulators) underlie the [Ca²⁺]_i-dependent modulation of MCU P_O shown in Fig. 3B. This regulation is likely to be tissue-specific and time-dependent. The actual Ca²⁺ flux will also depend on the number of functional MCUs per square micron of IMM; this is also likely to be tissue-dependent and a function of developmental stage and other factors, such as dominant-negative isoforms (63).

Mitoplast Properties.

Fieni et al. (21) reported a surface area of ∼0.7 pF or 70 μm² for cardiac mitoplasts prepared using a “French press” methodology. This value is in good accordance with measurements of the IMM surface area by Page (64) and Smith and Page (65) using EM, as well as with estimated measurements from “idealized” cardiac mitochondria that are “brick-shaped,” 500 nm in width and height, and 1.5 μm long, and have maximally packed cristae (Fig. S6). This suggests that the mitoplasts from the study by Fieni et al. (21) are likely the product of a single mitochondrion. In contrast, another study used mitoplasts with capacitances nearly 13-fold larger (9 pF; ref. 20). The membrane origins of such large “mitoplasts” remain to be determined and should be considered when interpreting these MCU currents.

Tissue-Dependent MCU Ca²⁺ Influx Under Physiological Conditions.

There are many factors to consider regarding the influence of mitochondrial Ca²⁺ uptake on cytosolic Ca²⁺ levels under physiological conditions. Although small amounts of Ca²⁺ entering into the mitochondria may limit the capability of whole-cell MCU flux to modulate normal [Ca²⁺]_i transients, these small amounts of Ca²⁺ play a critical role in regulating mitochondrial function. In addition to I_mcu magnitude, one must consider the mitochondrial volume fraction of the cell when considering the likelihood of MCU influencing [Ca²⁺]_i. The relationship between I_mcu density and mitochondrial volume fraction (see Table S2) is shown in Fig. 4. One of the highest mitochondrial volume fractions is in mammalian heart (∼33%), whereas that in skeletal muscle is among the lowest (∼5%) and that in the liver is somewhere in between (∼20%). Tissues that display significantly higher MCU activity levels (i.e., skeletal muscle) also seem to be associated with a lower mitochondrial volume fraction. In fact, there appears to be an inverse relationship between I_mcu magnitude and mitochondrial volume fraction (Fig. 4, red line). Given the small fraction of cellular volume comprising mitochondria in skeletal muscle and the small magnitude of I_mcu in the heart, it seems that the whole-cell MCU flux will be modest in both tissues and unlikely to influence [Ca²⁺]_i dynamics. It is also important to consider morphological differences between tissues. Cardiac mitochondria are unique in their abundance and close proximity to the CRUs (36). Skeletal muscle mitochondria are a close second in terms of proximity but are far less abundant (66). Mitochondria in other tissues (e.g., liver) have more variability but certainly experience less [Ca²⁺]_i compared with heart and skeletal muscle, possibly necessitating a larger I_mcu to generate a similar whole-cell flux. Many studies have used a diverse set of experimental approaches to measure this influence with varying conclusions. Importantly, under otherwise physiological conditions, the acute loss of mitochondrial Ca²⁺ uptake seems to have a small yet measurable effect on [Ca²⁺]_i transients in both heart and skeletal muscle (38, 49, 67–70).

MCU Contribution to Cytosolic Ca²⁺ Signaling.

The leading MCU candidate (encoded by NP_001028431) is expressed (at varying levels) in mitochondria across all tissues (19, 71). The phylogenic distribution of the uniporter’s membrane-spanning pore subunit (MCU) and regulatory partner (MICU1) (55) shows that homologs of both proteins tend to be coexpressed in all major branches of eukaryotic life (71). Evidence so far suggests that although other Ca²⁺ influx pathways into the mitochondrial matrix may exist, MCU is the dominant one. The heart may well be the ideal place to study MCU uptake, given that cardiac mitochondria are abundant and see regular, repetitive large elevations in extramitochondrial Ca²⁺ due to the cardiac Ca²⁺ release cycle. This would seem to create the ideal set of conditions for large whole-cell MCU fluxes. In the heart, however, where fast and abundant nonmitochondrial Ca²⁺ transporters prevail, MCU fluxes have only a modest effect on [Ca²⁺]_i, consistent with recent experimental results (49, 72). However, when the heart rate increases, each mitochondrion is bathed more frequently in elevated Ca²⁺ and mitochondria will progressively accumulate more Ca²⁺ even though Ca²⁺ uptake per beat is modest. In this sense, [Ca²⁺]_m dynamics could be characterized as a low-pass filter version of [Ca²⁺]_i dynamics. Thus, although cardiac myocytes may be excellent for examining mitochondrial Ca²⁺ movement, the Ca²⁺ handling systems in other cells are different and require that quantitative experiments be carried out whenever mitochondrial Ca²⁺ signaling is thought to be critical. Examples of such systems include dorsal root ganglion neurons (4, 6), liver cells (7), motor nerve terminals (8, 10), gonadotropes (9), Sertoli cells (13), olfactory sensory neurons (14), lymphocytes (43), chromaffin (44), and various model cells. Importantly, in cells that lack significant Ca²⁺ extrusion pathways (i.e., SERCA, NCX), mitochondrial Ca²⁺ uptake is more likely to influence cytosolic Ca²⁺ signaling, especially during rapid and large changes in [Ca²⁺]_i.

Biophysical Properties of MCU: g_mcu and P_O.

The ∆Ψ_m dependence and Ca²⁺ dependence of g_mcu has been highlighted as a critical feature (18, 20, 51). Kirichok et al. (18) showed that g_mcu saturated at high [Ca²⁺]_i (millimolar range) and was largely open (P_O = 0.99) when ∆Ψ_m = −200 mV but was generally closed (P_O = 0.11) when ∆Ψ_m = −80 mV. Note that these measurements were obtained at very high [Ca²⁺] levels (around 105 mM). In contrast, over the physiological range of [Ca²⁺]_i (0.1–20 μM), there appears to be a significant Ca²⁺ dependence of P_O (Figs. 2 and 3), which suggests a low P_O (<0.1). When combined, a consistent story emerges from the data related to the Ca²⁺ influx into the mitochondria under physiological conditions: the influx is small. In fact, whole-cell MCU fluxes are largely consistent in the heart, skeletal muscle, and liver cells based on the findings of 17 independent studies. Importantly, the findings discussed here show that whole-cell MCU flux (over the physiological [Ca²⁺]_i range; Fig. 2) is consistent with the data from mitoplast studies (18, 20, 21) and the latest characterizations of the MCU (17, 19, 51) (Fig. 3) when [Ca²⁺]_i- and ∆Ψ_m-dependent changes in P_O (Eq. 2 and Fig. 3B) are taken into account.

Quantitative Aspects of Mitochondrial Ca²⁺ Fluxes.

Computational modeling of Ca²⁺ signaling at all levels can provide valuable insights into the interactions within complex systems. Such work, for example, has sharpened the discussion on Ca²⁺ sparks, arrhythmogenic Ca²⁺ waves, and intra-SR Ca²⁺ movement (41, 73–75). When constrained by experimental findings, such modeling helps formulate new critical experiments and can test signaling relationships that otherwise remain beyond current experimental reach. Mitochondrial Ca²⁺ modeling, however, has been less successful at conforming to experimental observations. Our analysis of many models (76–84) suggests a wide range in mitochondrial Ca²⁺ handling kinetics. Some models display physiological mitochondrial Ca²⁺ fluxes (76, 77, 81), whereas others (78, 82) display MCU fluxes that are orders of magnitude greater than those shown in Fig. 2A and represent a cytosolic Ca²⁺ sequestration mechanism superseding even that of SERCA. The unsubstantiated magnitude of the whole-cell MCU fluxes in these models calls into question computational results that claim mitochondrial Ca²⁺ uptake can significantly influence cytosolic Ca²⁺ signaling (79). It is imperative that future models include modern MCU formulations [i.e., formulate MCU as a highly selective Ca²⁺ channel (82)] and be constrained by the latest experimental observations (17–19, 21, 27, 51).

Future Challenges.

There are three clear challenges for future experiments that are motivated by the findings of this review:

• i) Reexamine Ca²⁺ influx and efflux from mitochondria in intact cells to provide quantitative information under both physiological and pathophysiological conditions.

• ii) Develop mitochondrial computational models of Ca²⁺ movement that take into account cellular Ca²⁺ signaling data, metabolic characteristics of the mitochondria, subcellular anatomy, and the nanoscopic mitochondrial organization, including cristae.

• iii) Use quantitative experimental finding to constrain the computational models and the model results to provoke new experiments.