Matrix polyamines regulate bidirectional calcium flux through MCU

SUMMARY

Polyamines, well-known regulators of the mitochondrial calcium (Ca²⁺) uniporter channel, show unexpected effects when binding the channel from within the matrix. Using cryo-EM, molecular dynamics simulations, and mutagenesis experiments, we determine that polyamines achieve such regulation by binding within the pore to a ring of negative residues forming a matrix gate, inhibiting Ca²⁺ conduction. In whole-mitoplast electrophysiology assays, matrix polyamines cause a gradual increase in Ca²⁺ currents during prolonged conduction, due to relief of this inhibition. Notably, this electrostatic binding increases threefold as the inner membrane depolarizes, preventing Ca²⁺ efflux. Additionally, we also identify that phospholipids form part of the Ca²⁺ conduction pathway through MCU. Because we find significant variability in matrix polyamine content across mouse organs, this unexpected mechanism for sculpting the mitochondrial Ca²⁺ waveform suggests a tissue-specific regulation of metabolism.

Graphical Abstract

eTOC BLURB

MCU controls Ca²⁺ entry into mitochondria, regulating metabolism. Wang, Balderas, Ramos et al. show that polyamines in the mitochondrial matrix bind within the MCU pore. This inhibition is relieved during Ca²⁺ influx and strengthened upon membrane depolarization. Organ-specific differences in matrix polyamine content suggest tissue-specific control of mitochondrial Ca²⁺ signals.

INTRODUCTION

The polyamine metabolites spermine, spermidine, and putrescine, present in all eukaryotic cells, are critical for processes such as cell proliferation, macromolecule synthesis, and cell death (Fig. 1A). Their small size and positive charge also allow them to regulate a range of ion channels, best studied in the establishment of inward rectification of potassium channels¹. Polyamines also contribute to mitochondrial Ca²⁺ homeostasis, which modulates respiration, fuel substrate usage, and cell survival across a range of tissues². However, polyamine regulation of mitochondrial Ca²⁺ transport is not well understood³.

Figure 1. External polyamines partially inhibit J_MiCa.

A. Metazoan polyamines. B. Whole-mitoplast electrophysiology of external spermine. Top left, voltage ramp. Bottom, J_MiCa exemplars in the absence (black) and presence (red) of external spermine. Arrows point to the current at −160 mV, for summary in (C). Inset, cartoon, whole-mitoplast recording conditions. K Gluc DVF, potassium gluconate divalent-free solution. C. Dose-response (n=5, 1 μM; 4, 10 μM; 4, 100 μM; 3, 300 μM). D, F. Exemplars, spermidine (D), putrescine (F). E,G. Summary, 34±5% inhibition, spermidine (n = 3); 11±5% inhibition, putrescine (n = 5). H. Polyamine MCU inhibition mechanism. SF, selectivity filter. IMS, intermembrane space. IMM, inner mitochondrial membrane. Dotted line in exemplars is 0 pA/pF level. Current density measured at −160 mV. Bars, mean ± SEM.

The main pathway for Ca²⁺ uptake is via the mitochondrial Ca²⁺ uniporter, a highly-selective Ca²⁺ channel^4,5. Polyamines have long been known to alter mitochondrial Ca²⁺ uptake, though there has been ongoing controversy as to the nature of such regulation³. In the initial report, spermine was found to be a competitive inhibitor of Ca²⁺ transport⁶. However, in subsequent studies, spermine was found to lower the threshold at which Ca²⁺ uptake would occur^7,8. Spermidine and putrescine, which are smaller and less cationic, had weaker effects. Thus, studies over several decades suggested that spermine potentiated uniporter-dependent Ca²⁺ uptake when the level of both of these were low (<50 μM spermine, <5 μM Ca²⁺), whereas at higher concentrations, competitive inhibition between these two occurred⁹. Finally, spermine was also found to inhibit the efflux of mitochondrial Ca²⁺, though this effect was least understood⁸. Because of its ability to lower the threshold for mitochondrial Ca²⁺ uptake, spermine is frequently used as a uniporter activator^7,10–12.

In a recent study, the binding site for spermine potentiation was localized to the gate-keeping accessory subunit MICU1¹³. Under low Ca²⁺ conditions, MICU1 inhibits Ca²⁺ uptake by interactions with the pore-forming MCU subunit, and Ca²⁺ binding to MICU1 rearranges these interactions, leading to a high-conductance, potentiated state^14–16. Spermine binding appears to displace MICU1 even at low Ca²⁺, allowing Ca²⁺ uptake at these subthreshold levels¹³. Higher Ca²⁺ concentrations themselves displace MICU1, so spermine does not potentiate uptake at these levels. Though the spermine binding site could not be clearly defined, this mechanism explains the potentiation of uptake.

These elegant experiments have revealed how polyamines regulate the uniporter from the intermembrane space. However, polyamines are known to be present within the mitochondrial matrix as well, though the concentrations and transporters of these metabolites have not been well-established for mitochondria^17–22. Without clear ways of manipulating matrix polyamine concentrations or transport, prior studies of polyamine regulation were limited in establishing mechanisms at the external face of the uniporter channel.

To examine direct effects of polyamines on MCU, the pore-forming uniporter subunit, we turned to mitochondrial electrophysiology. In this system, a glass pipette is used to rupture the inner membrane of a mitoplast, allowing electrochemical matrix access^23,24. Because mitoplast volumes are small (~0.5 fL) compared to the volume of pipette buffer (several μLs), matrix contents are quickly replaced with pipette buffers. This allows control of the solutions within the matrix and in the bath, and direct measurement of uniporter currents across the entire inner membrane. With this method, we identify a previously undescribed form of polyamine-mediated uniporter regulation. A combination of electrophysiology, cryo-EM, and molecular dynamic simulations establish that these metabolites inhibit Ca²⁺ transport by binding near two rings of glutamates, one set comprising the selectivity filter near the intermembrane space, and a second forming a matrix-facing gate within the uniporter pore. As Ca²⁺ starts permeating the channel, a slow unblocking of the polyamines occurs, leading to increased fluxes in a process we term disinhibition. In addition, because the interaction between polyamines and the uniporter is electrostatic, binding is voltage-dependent, with affinities that increase as the membrane depolarizes. Thus, polyamines can plug the mitochondrial Ca²⁺ uniporter during periods of depolarization, preventing mitochondrial Ca²⁺ efflux that may lead to cytoplasmic Ca²⁺ overload.

RESULTS

External polyamines partially inhibit mitochondrial Ca²⁺ transport through MCU.

To avoid confusion, we refer to the entire multi-subunit channel as the uniporter, and to individual subunits by their respective abbreviations (e.g. MCU). To determine how polyamines affect the uniporter independently of their effect on MICU1, we performed whole-mitoplast electrophysiology, measuring mitochondrial Ca²⁺ currents (J_MiCa) after adding polyamines into the bath (external) solution. In our assays, we use millimolar concentrations of Ca²⁺ to maximize ionic currents and obviate effects from MICUs, which dislodge from the pore at these high Ca²⁺ concentrations¹⁴. For these experiments, we used mouse liver, which has robust J_MiCa. For spermine, we found a dose-dependent inhibition, with total J_MiCa reduced by ~60%, (IC₅₀ of ~120 μM) (Fig. 1B–C, S1A). The residual current was due to MCU, since it was sensitive to ruthenium red (Fig. S1B). In contrast, spermidine was a weaker inhibitor, with a more modest ~40% J_MiCa reduction at 100 μM (Fig. 1D–E). Putrescine was the weakest, with only ~20% inhibition observed at 100 μM (Fig. 1F–G).

This pattern of inhibition closely matches the distribution of positive charge on the polyamine (Fig. 1A), which led us to surmise that blockade may be occurring by occluding access to the highly-negatively charged selectivity filter (SF) encoded by the Asp-Ile-Met-Glu (DIME) motif at the outer face of the uniporter pore (Fig. 1H). To test this directly, we sought to perform mutagenesis. However, mutants of the charge-carrying aspartate and glutamate residues are non-conductive. Nevertheless, previous reports have shown that mutagenesis to the serine residue directly outside the DIME motif (S259A) is permissive for Ca²⁺ conduction, but appears to alter access to the selectivity pore for larger molecules, such as MICU1 and ruthenium inhibitors^23,25,26.

We turned to HEK-293T cells in which endogenous MCU had been deleted, allowing heterologous expression of mutant channels to test structure-function relationships (Fig. S2)²⁷. We rescued uniporter activity by generating stable lines expressing either wild-type human Flag-tagged MCU (MCU-Flag) or a S259A MCU-Flag mutant (Fig. 2). Inhibition of J_MiCa by 100 μM spermine was evident in MCU-Flag cells (~40% block, Fig. 2A–B). In contrast, spermine inhibition of J_MiCa was much less effective in the S259A mutant (~10% block, Fig. 2C–D), suggesting that cytoplasmic polyamines inhibit Ca²⁺ conduction by limiting conduction through the channel pore.

Figure 2. The MCU S259A mutant lacks external spermine inhibition.

A, C. Exemplar J_MiCa MCU-Flag (A) or MCU-S259A-Flag (C). B, D. Summaries, 41±8% inhibition, MCU-Flag (n = 5); 19±15% inhibition, MCU-S259A-Flag (n = 6). Dotted line in exemplar is 0 pA/pF level. Current density measured at −160 mV. Bars, mean ± SEM.

Matrix polyamine concentrations are variable across tissues.

Beyond external regulation by polyamines, mitoplast electrophysiology allows the unique opportunity to examine how these molecules regulate the uniporter from the matrix side. Typical imaging assays have limited ability to analyze matrix effects since the concentrations and transport of polyamines into the matrix are unknown and cannot be easily controlled. Before conducting these experiments, we defined the concentration of polyamines in mitochondria from a range of mouse tissues. We made highly-purified mitochondrial fractions by differential centrifugation and isolation on Percoll gradients. These mitochondria were then subject to assessment of total polyamine species (free and protein bound) via HPLC, quantified against an internal standard (Table 1, Fig S3). We found the distribution of mitochondrial polyamines varied substantially across organs, but appeared to reflect whether the tissue was terminally differentiated. Putrescine levels were detectable only in the heart and spleen, where it was the most abundant polyamine. Levels of spermidine and spermine ranged from undetectable or very low in heart to near millimolar concentrations in the liver and kidney. In skeletal muscle, spermidine appeared more abundant than spermine, while in the liver and kidney, spermine was predominant.

Table 1. Total mitochondrial polyamine content.

Organs from 4 mice processed via HPLC. Averages ± SEM are from 2–4 animals per organ that produced detectable results. Polyamine concentration calculated by normalizing to mitochondrial protein concentration⁴⁶. N.D. Not detectable.

Organ (N)	Putrescine (μM)	Spermidine (μM)	Spermine (μM)
Liver (4)	N.D.	552.2 ± 144.3	746.1 ± 184.8
Spleen (2)	376.4 ± 35.9	201.3 ± 5.4	21.3 ± 6.9
Skeletal Muscle (2)	N.D.	18.7 ± 8.9	10.7 ± 5
Heart (2)	185.1 ± 3.2	N.D.	5.9 ± 1.4
Kidney (3)	N.D.	318.6 ± 41.3	946.8 ± 177.2

Matrix polyamine disinhibition increases J_MiCa during prolonged conduction

To directly assess how matrix polyamines affect J_MiCa, we varied polyamine concentration in our pipette internal solution. Once the whole-mitoplast configuration is established, it is not possible to change the pipette solution, so in these experiments we compared recordings obtained with or without polyamines in the pipette. In mouse liver mitoplasts, J_MiCa showed an unexpected behavior when polyamines were present internally. In the absence of polyamines, J_MiCa shows modest fast and slow inactivation, with a minor decline in current magnitude during a second-long pulse (Fig. 3A)²⁴. In the presence of spermine, however, we observed that after the initial fast inactivation, the J_MiCa amplitude grew during the trace (Fig. 3A–C). The effect was dependent on spermine dose, becoming evident between 10 and 100 μM, with a K_0.5 of ~70 μM.

Figure 3. Matrix polyamines enhance J_MiCa during prolonged pulses.

A. Whole-mitoplast electrophysiology of matrix spermine. Top left, voltage steps. Bottom, J_MiCa exemplars. Arrow, J_MiCa enhancement over time at fixed voltage. Inset, recording conditions. K DVF + Pam, potassium gluconate divalent-free solution with the indicated polyamine. B, C. Summaries using the −150 mV pulse. Current density, 150 ms into the pulse (B); enhancement rate, slope between 150–300 ms (C) (n = 10, Veh; 3, 1 μM; 4, 10 μM; 9, 100 μM; 2, 300 μM). D-G. Exemplars and summaries for spermidine (D, F) and putrescine (E, G). The controls (Veh) are the same for C, F, and G. (Spermidine: n = 2, 100 μM; 4, 300 μM; Putrescine: 4, 300 μM; 5, 1 mM). H. J_MiCa enhancement mechanisms. Dotted line in exemplars is 0 pA/pF level. Bars, mean ± SEM.

The increase of J_MiCa could be reproduced with spermidine and putrescine as well, though with lower efficacy (Fig. 3D–G). The spermidine effect was most obvious >100 μM, whereas with putrescine we were only able to see J_MiCa enhancement above 300 μM. As with the inhibition occurring from the intermembrane space, the degree of J_MiCa increase correlated with the number of positive charges on the polyamine. Thus, more potent effects seen with increased charge may reflect the ease of polyamine binding to the uniporter.

Broadly speaking, there are two possible ways to increase J_MiCa, enhanced opening or disinhibition (Fig. 3H). For enhanced opening, polyamine binding alters uniporter conformation such that its opening is progressively enhanced during the course of a prolonged pulse, leading to increased J_MiCa. For disinhibition, bound polyamines inhibit conduction, and their release from the channel during the voltage step leads to disengagement of this inhibition and increase in J_MiCa. Since we are unable to add or remove polyamines to the pipette solution while recording J_MiCa, the next simplest way to distinguish between these two broad classes is to compare J_MiCa magnitude in the presence or absence of polyamines. If polyamines are enhancing uniporter conduction, the initial current should be similar to the control condition, with subsequent increases in J_MiCa reflecting greater opening. In contrast, if polyamines are inhibiting the channel, the initial J_MiCa magnitude should be lower than control, with increasing currents during the step reflecting release of inhibition. Examining this parameter (Fig. 3B), we found there was a clear decrease in initial J_MiCa magnitude when spermine was present in the pipette (−58±7 pA/pF for control versus −30±2 pA/pF with 100 μM spermine), supporting a disinhibition mechanism.

In this scenario, at our resting holding potential of 0 mV, spermine accumulates within the channel. When the voltage is changed to −150 mV, the negative potential promotes polyamine unbinding and a slow increase in J_MiCa size. Thus, the rate of current increase is a surrogate for the k_off of polyamine dissociation (Fig. 4A, S4A).

Figure 4. Matrix spermine binds rapidly and releases slowly from MCU.

A. Top left, long-step voltage protocol. Bottom, J_MiCa exemplars. Dotted lines, parameters used for binding constants. The red dashed line for spermine is the extrapolated single-exponential fit, that plateaus at the estimated J_max. Inset, spermine binding estimates at 0 and −150 mV. B. Top left, paired-pulse voltage protocol; reference (P1) and test (P2) pulses separated by increasing interpulse intervals (Δ). Above the protocol, parameters for increasing Δ between successive pulse pairs; blue, first pair; orange, 12^th. Bottom, corresponding exemplars. Black arrow, J_MiCa level for reference (first, J₁) pulse in each pair. Blue arrow, J_MiCa level for the test (second, J₂) pulse in the first pulse pair. Orange arrow, J₂ from the 12^th pulse pair. C. Summary of J₂/J₁ versus Δ (n = 6 vehicle, 5 spermine). D. Spermine binding time constant (τ_ON) from exponential fits to the (average spermine J₂/J₁ ratio – average vehicle J₂/J₁ ratio) trace from (C). E. Exemplar extramitochondrial Ca²⁺ imaging traces. Permeabilized MICU1 KO cells loaded with Ca²⁺ and treated with FCCP, antimycin A (AA) and Ru360. Ca²⁺ efflux: increase in fluorescence. F, G. Summary, Ca²⁺ efflux rates (normalized to no spermine) after FCCP + AA (B) and after Ru360 (C) (n = 7, No spermine; 7, 1 mM; 6, 5 mM; 6, 7.5 mM). ns, not significant. One-way ANOVA with post hoc Bonferroni corrected t-tests. Dotted line in electrophysiology traces is 0 pA/pF level. Bars, mean ± SEM.

Next, we examined how quickly polyamines bind back to the channel after disengagement, to obtain an estimate of the binding k_on (Fig. 4B–D). To gauge this parameter, we performed a paired-pulse protocol. Unfortunately, repeated second-long pulses at −150 mV destabilizes the membrane and increases leak, so we limited our efforts to 100 ms on the initial pulse. At the end of this pulse, there will have been partial unbinding of polyamines from the uniporter. Returning the holding potential to 0 mV, polyamines are then able to re-engage the channel. Delivering a second pulse a defined interval after the initial step revealed how inhibited J_MiCa had become in the intervening period. By graphing this fractional blockade of J_MiCa relative to the interpulse interval allows calculation of a rate-constant for polyamine binding to the uniporter. When applied to liver mitoplasts, we found that this recovery of binding occurred much faster than dissociation, with a time constant of ~50 ms for 100 μM spermine (Fig. 4D).

Analyzing J_MiCa during the long pulse to −150 mV allows us to estimate uniporter binding affinities for spermine both at 0 and −150 mV (Table S1). For these estimates, we assume that uniporter density within liver mitoplasts is approximately constant across different experiments, so that differences in J_MiCa in the presence or absence of spermine reflect the relative fraction blocked. We also make simplifying assumptions of a single binding site, one-to-one binding, and lack of any Ca²⁺ conduction through blocked channels. Because the unbinding during the long pulse occurs slowly, we estimate that the current carried at the very beginning of the current step (immediately after fast inactivation, J_spermine(0), Fig. 4A) reflects the fraction of channels lacking spermine at 0 mV. Comparing this initial J_MiCa in experiments with or without spermine (J_spermine(0) or J_Avg,Veh, respectively) allows us to estimate the binding affinity at 0 mV (K_D,OmV ~ 20 μM). Similarly, by mathematically extrapolating the increase in current to its maximal value (J_max, Fig. 4A), we could obtain an estimate for the steady-state binding affinity at −150 mV (K_D,−150mV ~70 μM, in close agreement with the estimate in Fig. 3C).

Taken together, our electrophysiological analysis suggests that matrix spermine blocks the uniporter pore with an affinity that is voltage-dependent, increasing more than three-fold as the membrane depolarizes. Moreover, while binding of spermine to the channel is rapid, disinhibition caused by polyamine unblocking is a slower process.

Spermine prevents Ca²⁺ efflux through the uniporter during mitochondrial depolarization

The data suggested that spermine rapidly blocks the uniporter when the membrane depolarizes, preventing ionic efflux through the uniporter. We identified such blockade by examining outward currents from liver mitoplasts in our voltage-ramp protocol (Fig. S4B–C). When positive voltage was applied, outward current density was lower with 100 μM spermine present in the pipette solution. In addition, ongoing blockade was visible as decreasing current amplitude during the 150 ms plateau at +80 mV. To test this further in intact mitochondria, we measured mitochondrial Ca²⁺ efflux using an imaging assay. This test was performed in MICU1 knockout cells to avoid confounding by spermine binding to MICU1¹³. Unfortunately, because we have no way of directly controlling matrix spermine concentrations, we were limited to incubating cells with high concentrations of extramitochondrial spermine (1–7.5 mM), assuming this would lead to import into the matrix (Fig. S4D). In this assay, low concentrations of digitonin permeabilize the plasma membrane but leave the mitochondrial inner membrane intact. Permeabilized cells were loaded with a 25 μM bolus of Ca²⁺ for 5 minutes in the presence of CGP-37157, a blocker of Na⁺-Ca²⁺ exchange, cyclosporine A, a blocker of the mitochondrial permeability transition, and then incubated with spermine or vehicle for up to 30 minutes. After this period, cells were spun down and resuspended in spermine- and Ca²⁺- free buffer containing an extramitochondrial Ca²⁺ reporter, Oregon Green BAPTA 6F. Baseline fluorescence was recorded, and then mitochondria were depolarized by addition of FCCP and antimycin A, which led to Ca²⁺ efflux through the uniporter, confirmed by blockade with Ru360 (Fig. 4E–G, S4E–F). Notably, the efflux rate was inhibited by increasing concentrations of spermine, suggesting that this polyamine prevents mitochondrial Ca²⁺ release during depolarization events.

Cryo-EM reconstruction of the MCU–EMRE complex in the presence of spermine

To study the structural basis of spermine binding to MCU, we solved a cryo-EM structure of the transmembrane uniporter subunits (MCU and EMRE) in the presence of 300 μM spermine to 2.98 Angstrom (Å) resolution. This structure possesses the highest local resolution of the core transmembrane portions of the uniporter available thus far (Fig. 5A–B, S5–8, Table 2). We chose not to include MICU1 and MICU2 because our focus was on examining polyamine binding to the channel itself, and feared that spermine-mediated disruption of MICU-MCU interactions would impair the ability to obtain resolvable particles. In the presence of spermine, most of the MCU-EMRE complexes were captured in tetrameric form with flexible N-terminal domains, similar to the scheme of the closed uniporter (Fig. 5B, S5B)^28,29.

Figure 5. Spermine binding site below MCU selectivity filter.

A and B. Cryo-EM reconstruction of MCU-EMRE complex with spermine shows 2.4–2.6 Å local resolution in the channel transmembrane region. Color codes, CryoSPARC sharpened map. C. Uncharacterized density in MCU pore. D. IMS view of the density. E. Side view of the density and surrounding MCU residues. IMS, inter-membrane space.

Table 2.

Cryo-EM data processing statistics.

Human MCU-EMRE complex (EMD-77084) (PDB: 13IO)
Data Collection and processing
Microscope	Titan Krios/ Falcon 4i camera
Magnification (kX)	165
Voltage (kV)	300
Electron exposure (e-/Å2)	45
Defocus range (μm)	−0.5 to −2.5
Raw pixel size (Å)	0.7296
Symmetry imposed	C1
Initial particle images (no.)	3,822,371
Final particle images (no.)	84,413
Map resolution (Å)	2.98
FSC threshold	0.143
Map resolution range (Å)	11–2.7
Refinement
Initial model used (PDB code)	ab initio
Model resolution (Å)	3.0
FSC threshold	0.5
Map sharpening B factor (Å2)	−80
Model composition
Non-hydrogen atoms	7,749
Protein residues	884
Ligands	11
Overall model B factors (Å2)
Protein (min - max)	53.41 (13.15 – 110.45)
Ligand (min - max)	59.38 (22.23 – 112.56)
Root mean square deviations
Bond lengths (Å)	0.006
Bond angles (°)	0.581
Validation
MolProbity score	1.70
Clashscore	6.34
Poor rotamers (%)	2.93
Ramachandran plot
Favored (%)	98.50
Allowed (%)	1.27
Outliers (%)	0.23

In the presence of spermine, we identified a non-protein density below the selectivity filter (Fig. 5C). The thread-like density is clearly visible and has similar local resolution to the surrounding amino acid side chains. Based on its shape and chemical environment, we hypothesized that this density represents the added spermine. Thus, we propose that spermine could interact with MCU’s selectivity filter E264 and surrounding residues, such as W260, Y268 and T267 (Fig 5D–E). Due to the flexibility of the spermine molecule and the size of the uniporter pore, we could model more than one configuration of spermine-MCU interactions.

Docking studies and molecular dynamic simulations identify spermine binding sites within uniporter pore

Since spermine and other polyamines are positively charged and show voltage-dependent MCU binding affinity, they may interact with negatively charged pore-lining residues. To test this hypothesis, we performed charge analysis along the MCU pore. Two regions along the pore display significant negative charge—one near the selectivity filter (E264 residues) and another near the E288/E293 residues, which are part of the matrix gate, a section of the uniporter where the pore narrows, requiring interactions with the EMRE accessory subunit to remain conductive (Fig. 6A)^30–32. Despite considerable effort, we were unable to identify non-protein densities near the E288/E293 residues in our cryo-EM structure. Therefore, we turned to molecular docking and molecular dynamics simulations to further investigate the spermine binding to MCU.

Figure 6. Molecular modeling of spermine interactions with MCU.

A. Electrostatic potential along channel pore identifies two potential polyamine binding sites: below selectivity filter (SF) and at the matrix gate (MG). IMS, inter-membrane space. B-E. MD simulations. B. 3D channel structure, showing two representative spermine poses (sphere representation) corresponding to the highest density peak. Sticks, residues within 4 Å of spermine. C. Density profile of spermine at 0 mV and −290 mV plotted along the channel axis (z). D. Interactions of the spermine center of mass with glutamate side chains during simulation time. E. Percentage of contacts between spermine and residues within 5 Å, averaged over all runs. Only contacts greater than 10% of simulation time are shown. IMS, inter-membrane space. SF, selectivity filter. MG, matrix gate. Bars, mean ± SEM.

In concordance with charge analysis, docking predictions identified two potential binding sites for spermine. The first site is located at the selectivity filter, where it interacts with the E264 side chains contributed by all four subunits and W260 from one subunit (Fig. S9). The binding site at the selectivity filter may correspond to density observed in the cryo-EM map of MCU, except the orientation of spermine is more vertical in most docking poses. The second site is positioned near the matrix gate. Spermine binding at this site involves hydrogen bond interactions with E288, E293, and Q287. To address whether restricting the docking search space biased site identification, we performed additional docking with expanded sampling of the matrix-exposed surface of the MCU–EMRE complex. This was achieved by dividing the system into three overlapping regions (selectivity filter, matrix gate, and matrix surface) and performing exhaustive sampling for each region. RMSD-based clustering of the resulting poses revealed a limited number of recurrent binding modes (Fig. S10, Table S2), corresponding to three interaction regions: the selectivity filter (E264), the matrix gate (E288/E293), and a matrix-facing surface site. Importantly, no alternative dominant binding site comparable to the pore-associated E288/E293 site was identified, supporting the robustness of the identified interaction site.

To evaluate the plausibility of the two binding sites, molecular dynamics (MD) simulations were performed using our cryo-EM MCU structure (Table 3). Simulations were conducted at both 0 mV membrane potential and −290 mV. A Ca²⁺ concentration of 75 mM was included by applying a recently-described model³³. Fig. 6B shows the binding modes of spermine, corresponding to the highest time-averaged spatial occupancy density (Fig. 6C) observed in MD simulations. As with docking studies, MD simulations identify both a binding site at the selectivity filter, with hydrogen bond interactions to all four E264 side chains, and a matrix gate binding site, with hydrogen bonds to E288 and E293. The density distribution shows a clear difference between 0 mV and −290 mV runs: in the first, two binding sites similar to the docking poses are seen, whereas in the latter, a high-density peak appears only at the matrix gate site. The matrix gate binding site was consistently favored over the selectivity filter site. In all simulation runs, spermine migrated from the selectivity filter to the matrix gate site within 100 ns (Fig. 6D). In one extended simulation, spermine exited the pore after ~850 ns, consistent with a dynamic interaction rather than a permanently bound state. Importantly, the kinetics of this movement varied between the two conditions. At 0 mV, the movement along the channel axis (z-axis) was slower, whereas at −290 mV, spermine almost instantly relocated to the E288/E293 site, where it remained bound. Fig. 6E compares the spermine interaction frequencies along the channel axis over simulation time. Four simulations included cardiolipin molecules bound at the cryo-EM observed site. Cardiolipin remained stably bound during simulations, interacting with several residues within the MCU transmembrane domains (<5 Å cutoff), but had no substantial effect on spermine behavior within the pore (Fig. S11). Importantly, though spermine remained bound to the E288/E293 matrix gate, its high flexibility, combined with the availability of 8 charged glutamate residues from 4 MCU subunits, meant that spermine formed many different binding poses at this site across simulations, potentially explaining why no consistent density was observed here in the cryo-EM maps.

Table 3.

Simulation overview.

Run	Time (ns)	Conditions	Notes
1	100	0 mV
2	100	0 mV
3	100	−290 mV
4	1000	−290 mV	SPM left channel after ~ 850 ns
5	100	0 mV, 4 cardiolipins
6	100	−290 mV, 4 cardiolipins
7	100	−290 mV, 4 cardiolipins
8	100	−290 mV, 4 cardiolipins
9	500	−290 mV, 4 cardiolipins
Total sampling time: 2.2 μs

The fenestration between TM1 and TM2 of MCU accommodates functional lipids and other non-protein densities

Beyond our investigation of spermine, the much higher local resolution of the transmembrane domains helped highlight non-protein densities (Fig. S8D). Cardiolipin is predominantly found in the mitochondrial inner membrane. With its four tails and inverse-cone geometry, it is preferentially located on the matrix-facing leaflet and contributes to the membrane curvature that MCU homocomplex requires. A cardiolipin binding site on MCU has been reported previously³⁴. In this work, we confirm that MCU R297 forms stable interaction with a lipid phosphate head group (Fig. S12). If we model this density as cardiolipin, the 2^nd phosphate density could also be identified but the density for the glycerol linker between two phosphates is weaker. Cardiolipin interaction to MCU is stabilized by the electrostatic interaction between the two phosphate groups and 3 positive side chains from MCU (R297, K305) and EMRE (K62).

The fenestration between TM1 and TM2 of MCU is also stabilized by a phospholipid that we tentatively modelled as phosphatidylserine (PS) based on the shape of the lipid head group and local chemical environment (Fig. S13). The PS head group directly participates in the formation of MCU ion translation pathway and thus can be considered pore lining (Fig. S13A–C). Thus, the human MCU-EMRE complex has a conductive channel formed by both proteins and lipids. Importantly, the carboxylate from PS inserted between adjacent transmembrane domains of a single MCU subunit and could interact with the matrix gate residue E288 of a neighboring subunit through a cation bridge. Thus, the PS lipid may also contribute to the MCU pore via formation of the matrix gate (E288).

Removing negative charges at the matrix gate within the MCU pore abolishes disinhibition

Both docking and MD simulation analyses suggested that the primary spermine binding site was to the MCU matrix gate. To confirm this effect, we turned again to studies performed using heterologous expression in MCU KO HEK-293T cells (Fig. S2). We created stable lines expressing either wild-type MCU or a matrix-gate double-glutamate mutant (E288A, E293A, MCU EEAA). The E288A/E293A mutant not only neutralized the local negative charge but also altered the lower opening of the pore to be more hydrophobic, which might have implications for polyamine and lipid binding (Fig. 6A, S13D).

Examining Ca²⁺ currents with 100 μM spermine in the pipette, we saw that wild-type MCU had clear J_MiCa enhancement, similar to that observed in liver mitoplasts (Fig. S14). During the long-step protocol, spermine unbinding in this heterologous system occurred with a time constant that remained slow (Fig. 7A). Moreover, spermine binding during the paired-pulse protocol occurred rapidly (τ ~ 70 ms, Fig. 7B–D). The similarity of these findings to those from mouse liver confirmed that disinhibition was a general effect conserved across mice and human MCU isoforms, and was not specific to liver mitoplasts.

Figure 7. Mutation of the matrix gate prevents MCU-spermine binding.

A-D. MCU-Flag. E-H. MCU-EEAA-Flag. A, E. Top left, long-step voltage protocol. Bottom, J_MiCa exemplars, spermine disinhibition for MCU-Flag (A) but not MCU-EEAA (E). Inset, recording conditions. B, F. Top left, paired-pulse voltage protocol. Bottom, J_MiCa exemplars, MCU-Flag (B) and MCU-EEAA (F). C, G. J₂/J₁ ratio (n = 5 vehicle, 6 spermine, MCU-Flag, MCU-EEAA). D, H. Spermine binding time constant calculated from (C) and (G). I. Hypothetical mechanism for matrix polyamine regulation of the uniporter. Dotted line in exemplars is the 0 pA/pF level. Bars, mean ± SEM.

Turning to the mutant construct, we confirmed that these uniporters could still conduct Ca²⁺, in agreement with prior reports³² (Fig. S14B). However, MCU EEAA mutant channels were no longer substantially blocked by inclusion of spermine in the pipette (Fig. S14C–D). Such an effect was not due to the reduced basal MCU EEAA Ca²⁺ current densities, because even wild-type MCU mitoplasts with small currents still showed disinhibition (Fig. S14C–D). Intriguingly, we noted that the MCU EEAA mitoplasts showed a transient increase in current density occurring even without spermine incubation (Fig. S14B). We speculate that this is due to matrix-gate rearrangements that occur as Ca²⁺ permeates through the pore.

To examine the degree to which matrix spermine regulation was lost in the MCU EEAA mutant, we performed our long-pulse and paired-pulse protocols. We failed to find any spermine disinhibition during long pulses (Fig. 7E). In addition, we did not observe any re-establishment of blockade during the paired-pulse protocol, confirming the absence of substantial spermine effects on the mutant (Fig. 7F–H). Taken together, our results reveal that matrix polyamines can dynamically regulate uniporter function by interacting with charged residues in the matrix gate of the channel, producing both a slow increase in influx rates as well as rapidly preventing Ca²⁺ efflux during depolarizations.

DISCUSSION

In this study, we show that polyamines within the mitochondrial matrix can bind within the uniporter channel pore. During Ca²⁺ influx, displacement from the pore increases Ca²⁺ influx over time, which we term disinhibition (Fig. 7I). Such behavior mimics unblocking of polyamines in inward rectifying potassium channels and is also reminiscent of increases in currents through voltage-gated Ca²⁺ channels, though occurring by different mechanisms^35–37. During mitochondrial depolarization, polyamines block the pore, preventing Ca²⁺ efflux. Using structural and MD analyses, we identify two potential spermine binding sites within the pore. Binding to the first site just below the selectivity filter appears transient. Binding to a ring of glutamates that form the lower matrix-facing gate of the channel is more stable. These findings reveal unexpected forms of polyamine regulation on Ca²⁺ uptake, and establish a mechanism by which these ubiquitous metabolites can directly modulate uniporter function from the matrix side.

We also confirmed that polyamines inhibit Ca²⁺ influx from the intermembrane space⁶. The observed pattern of inhibition—strongest with spermine, followed by spermidine and putrescine—directly correlates with the number of positive charges on each polyamine. This charge-dependent effect suggests an electrostatic blockade of the highly negative selectivity filter, supported by the loss of spermine-mediated inhibition in the S259A MCU mutant. The use of the voltage-clamp method in the future may also reveal more subtle time- or voltage-dependent regulation by cytoplasmic polyamines.

The analysis across mouse tissues revealed that total matrix polyamine concentrations are highly variable but can be substantial, with near-millimolar levels of spermine and spermidine in metabolically-active organs like the liver and kidney. The heart, in contrast, had substantial putrescine present, but not spermine or spermidine, a finding of uncertain significance. A limitation is that we are unable to distinguish free from bound polyamines with HPLC, which would require targetable polyamine sensors not currently available. Nevertheless, the heterogenous distribution likely reflects differences in metabolic demand, proliferative capacity, and Ca²⁺ regulation across tissues^38–40. Notably, matrix polyamine regulation may partly explain variability in mitochondrial Ca²⁺ uptake from different tissues noted in earlier studies.

Matrix polyamine regulation of the uniporter likely occurs during normal physiology. The binding constants (20–70 μM) are well within the range of total spermine concentrations measured across tissues (6–1000 μM). Most cellular spermine (90%) is non-covalently bound, primarily to nucleic acids but also to ion channels⁴¹. Thus, even in tissues containing high total spermine, such as the liver or kidney, the putative free spermine pool (70–100 μM) is close to the MCU dissociation constant, suggesting that it can modulate channel function dynamically. In addition, blockade of the pore was stronger at depolarized (0 mV) rather than hyperpolarized (−150 mV) potentials, suggesting that it depends on electrostatic interactions affected by membrane voltage and may be displaced by Ca²⁺ influx. This leads to two consequences for mitochondrial Ca²⁺ transport.

First, during uniporter activation by cytoplasmic Ca²⁺, some fraction of channels may be occupied by spermine. As Ca²⁺ influx drives the polyamines out of the pore, relieving the block, there will be a progressive increase in this flux. The slow dissociation kinetics of spermine suggests that such regulation is a low-pass filter, producing a change in influx rates that would be difficult to quantify in conventional imaging assays. Nevertheless, a current difficulty in investigating such disinhibition further is that there is no way to directly control the matrix concentration of polyamines, as the mitochondrial polyamine transporter has not yet been identified. Thus, though disinhibition may be expected to occur during normal physiology, a limitation of our study is the inability to determine the full extent of this mechanism.

A second form of regulation occurs when mitochondria depolarize, evident as an increase in spermine blockade. Spermine affinity for the channel increases three-fold at 0 mV relative to −150 mV, with a fast time-constant (~50 ms), suggesting that this mechanism prevents Ca²⁺ efflux from the mitochondria. Rapid changes in mitochondrial membrane potential alter cellular homeostasis via processes such as ATP synthesis and reactive oxygen species production⁴². This is most vividly seen in mitoflashes, stochastic bursts of depolarization, which allow mitochondria to rapidly communicate metabolic shifts and orchestrate developmental transitions^43–45. This has important physiological implications, as it means that even transient depolarizations could be sufficient to trap polyamines within the pore, effectively inhibiting Ca²⁺ efflux. Such an effect could be replicated in imaging assays, where spermine incubation led to reduced efflux rates. However, because we cannot infer the matrix concentration of spermine based on its extra-mitochondrial levels, the experiment does not allow a comparison of affinities relative to the electrophysiology experiments.

Most importantly, the cryo-EM structures and MD simulations further specify polyamine interaction within the MCU pore. Spermine interacting with the selectivity filter rapidly relocates to the matrix gate site under physiological membrane potentials, where it remains more stably bound. Interaction at these sites provides a structural basis for the matrix disinhibition we observed experimentally, and was further confirmed by heterologous expression of channels mutated at the matrix gate, where spermine regulation was abolished. Moreover, this mechanism aligns with polyamine pore blockade in other ion channels, such as inwardly rectifying potassium channels. Our electrophysiological, structural, and computational analyses provide a direct explanation for a previously undescribed mode of uniporter regulation.

Taken together, these findings highlight polyamines as versatile modulators of mitochondrial Ca²⁺ uptake, due to their effects on both sides of the inner mitochondrial membrane, tissue-specific abundance, and sensitivity to membrane potential. While our study identifies matrix polyamine-dependent uniporter modulation, it is important to acknowledge that this is one of multiple regulatory mechanisms. The reordering of MICU1 by external polyamines provides a separate means of control. Future studies should aim to investigate how these two modes cooperate to fine-tune mitochondrial Ca²⁺ signaling. In broader terms, our results define a paradigm of small-molecule uniporter modulation from within the matrix. Given the critical role of mitochondrial Ca²⁺ in energy metabolism and cell signaling, such regulation could have important implications in health and disease.

RESOURCE AVAILABILITY

Lead contact

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Dipayan Chaudhuri (dipayan.chaudhuri@hsc.utah.edu).

Materials availability

Plasmids generated in this study have been deposited to Addgene (deposit 87609). All unique/stable reagents generated in this study are available from the lead contact with a completed materials transfer agreement.

Data and code availability:

• The atomic coordinates for the MCU-EMRE complex have been deposited in the protein databank (PDB ID: 13IO) and the map in the EMDB (EMDB ID: EMD-77084) and they are publicly available as of the date of publication.

• Original western blot images are displayed in Fig. S2. Electrophysiology data reported in this paper will be shared by the lead contact upon request.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

STAR★METHODS

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Animal handling

All animal procedures have been reviewed and approved by the Institutional Animal Care and Use Committee at the University of Utah. Adult mice (>10 weeks old) of C57BL/6NJ strain and both sexes were used. Animals were housed under standard conditions and allowed free access to food and water. Mice were euthanized via CO₂ inhalation.

METHOD DETAILS

Plasmid constructs and cell lines

MCU-Flag and MCU-S259A-Flag constructs were a kind gift of Vamsi Mootha^23,25. The MCU-EEAA-Flag construct was synthesized by Twist Biosciences and cloned into a lentiviral vector. The full length human MCU with a cleavable C-terminal GFP-TwinStrep affinity fusion, together with the full length human EMRE with a C-terminal TwinStrep tag, were cloned into a bidirectional PiggyBac transposon vector to co-express these two proteins. All cloning was performed using the NEB HiFi DNA Assembly kit.

Stable HEK293T/17 SF cell lines (ATCC ACS-4500) expressing GFP- and TwinStrep-tagged human MCU and EMRE from the bidirectional plasmid were generated according to a published protocol⁶⁰. MCU and MICU1 KO HEK-293T cell lines were a gift of Yasemin Sancak and Vamsi Mootha. Stable lines expressing MCU constructs were made as described previously⁶¹.

Mitochondria and mitoplast purification

Organs from euthanized mice were removed and washed in ice-cold phosphate buffered saline (PBS) and then placed in ice-cold initial medium containing (mM): 225 mannitol, 70 sucrose, 5 HEPES, 1 EGTA, and 0.1% (w/v) bovine serum albumin (BSA), pH 7.2. For HEK293T cells, media was removed and the cells were washed with phosphate-buffered saline before being scraped into to the same initial medium described above. Tissue or cells were homogenized using a Potter-Elvehjem tissue grinder attached to an overhead stirrer (IKA, Wilmington, NC) for 10–15 strokes at 180 rpm. The homogenate was centrifuged at 700g for 7 minutes and the pellet was discarded. The supernatant was then centrifuged at 8500g for 10 minutes to obtain a crude mitochondrial fraction.

For experiments quantifying polyamines from mouse tissues, highly-purified mitochondrial fractions were obtained as described previously⁴³. The crude mouse organ mitochondrial fraction obtained above was then layered on top of Percoll medium, consisting of (mM): 225 mannitol, 25 HEPES, 1 EGTA, 30% Percoll (pH 7.4). The crude mitochondria and Percoll medium were spun at 95,000g for 30 minutes at 4°C. After the spin, a dense band containing pure mitochondria was located at the bottom of the tube. This band was extracted and spun at 6,300g for 10 minutes at 4°C. The pellet containing the purified mitochondrial fraction was then collected.

For electrophysiology experiments, mitoplasts were generated as described previously⁶². The crude mitochondrial fraction obtained above was incubated for 10 minutes in a hypertonic solution (mM): 140 sucrose, 440 D-mannitol, 5 HEPES, 1 EGTA, pH 7.2. Outer membranes were disrupted by passage through a French Press homogenizer (Thermo Fisher). The eluate was pelleted at 10,000g and resuspended until used for experiments in a hypertonic KCl solution (mM): 750 KCl, 100 HEPES, 1 EGTA, pH 7.2.

Mitochondrial polyamine quantification

Purified mitochondrial fractions were collected as described above. For culture cells, two million HEK293T cells per replicate were permeabilized using a buffer containing (in mM): 125, KCL; 20, HEPES; 10, K₂HPO₄, 1, MgCl₂; 0.5, EGTA; and 62.5 μg/mL digitonin, pH 7.2. Permeabilized cells were either left untreated or incubated with 5 mM spermine for up to 30 mins. Cells were then spun down, washed extensively with spermine-free buffer, and pelleted for HPLC analysis. Purified mitochondrial fractions or permeabilized cells were analyzed for determination of mitochondrial polyamine concentrations using high-performance liquid chromatography (HPLC) based on the gold standard method of Kabra et al., with modifications to improve efficiency enabled by technological advances^63,64. Samples were run in duplicate. In this highly sensitive method, polyamines undergo precolumn dansylation followed by purification using Waters Sep-Pak C18 cartridges, which were activated by washing with 2 column volumes each of 100% and 35% HPLC-grade acetonitrile, and after drawing each sample through, washed with 2 column volumes of 35% acetonitrile. The dansylated polyamines were eluted with 1 mL 100% acetonitrile and detected via HPLC coupled to a fluorescence detector over a range of four sensitivities. Solvents for gradient II, used for the separation of native and acetylated polyamines in tissues, were HPLC-grade acetonitrile and HPLC-grade water. 1,7-diaminoheptane (DAH) was included in all samples and standards as the internal standard. Equilibration time was 6 min and total time was 24 min. The flow rate was 1 mL/min. The column effluent was monitored at an excitation wavelength of 345 nm and an emission wavelength of 505 nm. Final polyamine concentrations were reported relative to total protein in the lysate as nmol polyamine/mg protein. For mouse tissue mitochondrial concentrations, these were converted into molar concentrations assuming 2.5 μL/mg mitochondrial protein⁴⁶.

Western blotting and co-immunoprecipitation

Total protein was extracted from cells with ice-cold radioimmunoprecipitation assay (RIPA) lysis buffer supplemented with protease and phosphatase inhibitors (Thermo) as described previously^62,65. Lysates were centrifuged at 4 °C for 10 minutes at 16000g. BCA assay was used for the measurement of protein concentration. Then, 20 μg of protein was run on 4–12% Bolt SDS-PAGE gels (Thermo) and transferred onto PVDF membrane for blocking and immunostaining at 1:1000 dilution of the corresponding primary antibody.

Electrophysiology

Whole-mitoplast electrophysiology was performed using established protocols. Borosilicate glass pipettes with a resistance of 15–20 MΩ were used. Whole-mitoplast currents were acquired at 5 kHz and filtered at 1 kHz using an Axopatch 200B amplifier (Molecular Devices, San Jose, CA). Mitoplasts had a capacitance of 0.70±0.04 pF (n=117 trials). Pipette solution was composed of (mM): Na-Gluconate 150, HEPES 10, EDTA 1, EGTA 1, pH 7.2, brought up to 320–340 mOsm with D-Mannitol. Bath solution for mitochondrial calcium uniporter (mM): Na-Gluconate 150, HEPES 10, 5 CaCl₂, pH 7.2. Ruthenium red (1 μM) was added to block uniporter-specific currents in particular cases as described in the manuscript. For display purposes, capacitance transients caused by solution exchange in the bath have been removed and the Simplify filter from Adobe Illustrator has been used to reduce the number of points, without altering the shape of the traces. Analysis was performed using pClamp v10 (Molecular Devices). A 150-ms plateau at the beginning and end of the voltage ramp protocols is used to measure the current density at −160 mV and +80 mV reliably. Current density measurements taken at −160 mV were averaged between 80–120 ms in our ramp protocol, and +80 mV measurements were averaged between 1100–1140 ms. Analysis of outwards currents was with Na⁺, not Ca²⁺, as the internal charge carrier, since high levels of internal Ca²⁺ may activate the permeability transition pore.

Binding constant calculations

We can obtain a rough estimate of spermine binding affinity, by assuming the binding reaction is first order (a single molecule of spermine binds the channel, lumping binding at sites along the pore into a single estimate):

$$MC{U}_{\mathit{\text{free}}}+S\underset{k_{\mathit{\text{off}}}}{\stackrel{k_{\mathit{\text{on}}}}{\leftrightarrow }}MC{U}_{\mathit{\text{bound}}}$$

Using this equation, the dissociation constant at equilibrium is:

$$K_D=\frac{k_{\mathit{\text{off}}}}{k_{\mathit{\text{on}}}}=\frac{MC{U}_{\mathit{\text{free}}}\times S}{MC{U}_{\mathit{\text{bound}}}}=S\times \frac{MC{U}_{\mathit{\text{free}}}}{MC{U}_{\mathit{\text{bound}}}}$$

We assume S = 100 μM since the MCU-bound spermine (~10–1000) is a trivial number compared to free spermine in 1–5 μL of pipette solution (~10¹⁴ molecules). Since spermine is a charged molecule, its interaction with MCU will vary depending on the transmembrane voltage gradient (ΔΨ). By examining our electrophysiology data, we can measure the K_D in two relevant states. For these calculations, we make the following assumptions: (1) spermine-bound MCU is non-conductive, (2) during each step pulse to −150 mV, the instantaneous current through the channel is proportional to the number of open channels, since the measured time constants for spermine binding (τ_ON, τ_OFF) are slow compared current flux through the channel, and channels have stable conductance with very high unitary open probability at these external Ca²⁺ concentrations^15,24, (3) the average number of channel complexes per mitochondria is roughly stable for each tissue/cell line, allowing us to compare peak currents between untreated and spermine-treated mitoplasts, (4) we ignore fast and slow inactivation kinetics of vehicle-treated MCU and assume the current is constant over time. Thus,

$$\begin{gathered} MC{U}_{\mathit{\text{free}}}\left(t\right)\propto J_{\mathit{\text{MiCa,Spermine}}}\left(t\right) \\ MC{U}_{\mathit{\text{total}}}\propto J_{\mathit{\text{Average MiCa}},\mathit{\text{Vehicle}}} \end{gathered}$$

where J is the current density (pA/pF).

First, we can estimate the steady-state binding constant when the membrane is fully depolarized, since between test pulses we clamp the transmembrane voltage at 0 mV for several seconds and assume spermine binding to MCU reaches a steady state. In this case, we assume that the initial instantaneous current at V=−150 mV (t = 0, at the beginning of the step pulse to −150 mV) after fast inactivation arises only from the steady-state level of open (unbound) channels at V=0 mV, because there hasn’t been enough time at −150 mV to substantially release spermine from bound channels. In addition, as stated above, we assume that the average current after fast inactivation in vehicle-treated mitoplasts reflects the total pool of MCU.

Therefore,

$$K_{D,0mV}=\frac{MC{U}_{\mathit{\text{free}},0mV}\times S}{MC{U}_{\mathit{\text{bound}},0mV}}\approx S\times \frac{J_{\mathit{\text{MiCa,Spermine}}}\left(0\right)}{J_{\mathit{\text{Average MiCa}},\mathit{\text{Vehicle}}}-J_{\mathit{\text{MiCa,Spermine}}}\left(0\right)}$$

Second, we can estimate the spermine binding constant at steady-state for a membrane held at −150 mV. Here, we fit the slow increase in J_MiCa during long step pulses with a single exponential

$$J_{fit}(t)=J_{max}+J_{\mathit{\text{offset}}}\times e^{\left(-t/{\tau }_{off}\right)}$$

where, again, t = 0 is the beginning of the step pulse to −150 mV. At steady state (t = ∞), the maximal number of unbound channels is then proportional to J_max, obtained from the fit. Therefore,

$$K_{D,-150mV}=\frac{MC{U}_{\mathit{\text{free}},-150mV}\times S}{MC{U}_{\mathit{\text{bound}},-150mV}}\approx S\times \frac{J_{max}}{J_{\mathit{\text{Average MiCa}},\mathit{\text{Vehicle}}}-J_{max}}$$

Ca²⁺ imaging

Ca²⁺ imaging was performed as described previously⁶⁶. MICU1 knockout HEK293-T cells were permeabilized using a buffer containing (in mM): 125, KCL; 20, HEPES; 10, K₂HPO₄, 1, MgCl₂; 0.5, EGTA; and 62.5 μg/mL digitonin, pH 7.2. One million cells in 100 μL buffer were used per assay. Fluorescence imaging was performed on a Cytation5 plate reader (Agilent Technologies) at room temperature. For imaging, cells were suspended in a solution similar to the one above, except containing 10 μM EGTA, and supplemented with 5 mM succinate and 1 μM rotenone. Approximately 30 minutes prior to imaging, cells were incubated with 25 μM CaCl₂, 5 μM CGP-37157 (NCLX inhibitor), 5 μM cyclosporin A (PTP inhibitor) for at least 5 mins, after which we added vehicle (buffer alone) or spermine at room temperature. Approximately five minutes prior to imaging, cells were quickly spun down and resuspended in spermine- and Ca²⁺-free imaging buffer containing 5 μM CGP-37157, 5 μM cyclosporin A, and 1 μM Oregon Green BAPTA 6F hexapotassium salt. Fluorescence was first recorded to establish a baseline, followed by serial injections of 10 μM carbonyl cyanide-p-trifluoromethoxyphenylhydrazone (FCCP), 5 μM antimycin A, and 10 μM Ru360. Mixing was by hand for preincubation steps and by the plate reader after injections. For efflux rates, the slope of fluorescence change during the linear portion of the curve between approximately 1.5 and 3.5 minutes was calculated. For the Ru360 rate, the slope was calculated between approximately 6.5 and 7.5 minutes. Individual efflux and Ru360 rates from each trace were normalized by dividing by the average initial efflux rate from the vehicle-treated condition. For experiments where Ru360 was preincubated, this was added approximately 5 minutes before imaging, and maintained in buffer throughout the remainder of the experiment. In these experiments, the slope was measured in the first minute after FCCP and antimycin A addition, normalized by dividing by the average initial efflux rate from the vehicle-treated condition in the absence of Ru360.

Expression and purification of human MCU-EMRE complex

For protein expression, MCU-EMRE stable cells were grown in suspension in freestyle 293 expression medium (Invitrogen, Carlsbad, CA) at 37 °C in an orbital shaker to a cell density of ~1.5 × 10⁶/mL. Doxycycline was added to the culture to a final concentration of 1 μg/mL to induce protein expression and 5 mM sodium butyrate was employed to enhance protein expression. Cells were harvested 48 h post induction for subsequent affinity purification with Strep-Tactin resin (IBA, Strep-tactin@XT 4flows).

All protein purification steps were carried out at 4 °C unless stated otherwise. Membrane proteins were extracted for 2 h at 4 °C in a buffer composed of 20 mM HEPES (pH7.4), 150 mM NaCl, 2 mM lauryl maltose neopentyl glycol (LMNG-3), 0.4 mM cholesteryl hemisuccinate tris salt (CHS), 5 μg/ml leupeptin, 1.4 μg/ml pepstatin A, and 2 μg/ml aprotinin. The supernatant was collected after centrifugation at 100,000 × g for 30 mins and then incubated with Strep-Tactin resin for 2 h. Contaminant proteins were removed by washing with 15 column volumes of buffer composed of 20 mM HEPES (pH 7.4), 150 mM NaCl, and 0.1 mM LMNG-3, 0.5 mM CaCl₂, 0.5 mM DTT ph 7.4. MCU-EMRE complex was eluted from the resin with a buffer composed of 20 mM HEPES (pH 7.4), 150 mM NaCl, 500 mM biotin, and 0.1 mM LMNG-3, 0.5 mM CaCl₂, 0.5 mM DTT, pH 7.4, and further purified on a Superose 6 column using a buffer composed of 20 mM HEPES (pH 7.4), 150 mM NaCl, and 0.1 mM LMNG-3, 0.5 mM CaCl₂, 0.5 mM DTT, pH 7.4; peak fractions corresponding to the MCU-EMRE complex were pooled and incubated with 300 μM spermine for Cryo-EM analyzes.

Cryo-EM data acquisition, data processing and model building

In all experiments, 5 μl of MCU-EMRE sample at ~3–5 mg/ml, supplemented with 300 μM spermine, was applied to glow-discharged Au 1.2/1.3 holey, 300-mesh gold grids. Grids were plunge frozen in liquid ethane using a Vitrobot Mark IV (FEI) set to 4 °C, 90% relative humidity, 20 s wait time, −1 mm offset, and 10 s blotting time. Data were collected on a Krios (FEI) operating at 300 kV equipped with the K3 Summit direct electron detector at the University of Utah, PNCC, and NCCAT. Movies were recorded using SerialEM⁶⁷, with a defocus range between −1.0 and −3.5 μm. Specifically, movies were recorded with a physical pixel size of 0.7296 Å. The data were collected at a dose rate of 1 e⁻/Å2/frame with a total exposure of 45 frames, giving a total dose of 45 e⁻/Å2.

Movie frames were aligned, dose weighted, and then summed into a single micrograph using MotionCor2⁶⁸. CTF parameters for micrographs were determined using CTFFIND4⁵². For ab initial model generation in CryoSPARC, micrographs with CTF better than 3.5 Å were selected. Initial particle sets were blob picked and 2D classified. Best 2D classes were selected for 3 class ab-initial reconstruction with 0 class similarity in C1 symmetry. One best class showing transmembrane and N-terminal domains were used for 3D alignment. Micrographs with CTF better than 4.5 Å were selected for high resolution structural determination. Approximately 2000 particles were manually boxed out from 50 micrographs in RELION 5 to train a neuronal network model, which was then used to extract particles from all micrographs using TOPAZ^69,70. A total of 3,822,371 particles were extracted from 9373 micrographs in RELION and then imported into cryoSPARC for several rounds of 2D classification⁴⁸. ‘Junk’ particles that were sorted into incoherent or poorly resolved classes were rejected from downstream analyses. The remaining 414,714 particles from well-resolved 2D classes were pooled and subjected to heterogenous refinement in cryoSPARC with C1 symmetry using the aforementioned initial model. Then particles from one good class out of 6 classes were subjected to non-uniform refinement in cryoSPARC and yielded a 3 Å map. The resulting C1 map was locally sharpened in PHENIX for model building in Coot^49,50. ModelAngelo was used for initial model building for high confidence region⁷¹. Then the amino acid sequences of MCU and EMRE were manually rebuilt in Coot. Lipids and ions were modeled according to density⁷². The final model was refined in real space using PHENIX and assessed in Molprobity⁷³.

Fourier shell correlation (FSC) curves were then calculated between the refined model vs summed half maps generated in cryoSPARC, and resolution was reported as FSC = 0.5. The directional resolutions were calculated using a remote 3DFSC processing server (https://3dfsc.salk.edu/)⁷⁴. UCSF Chimera was used to visualize and segment density maps, and figures were generated using Chimera and Pymol (Version 3.0, Schrödinger)⁵¹.

Molecular Docking

The GOLD docking software (CCDC, version 2024.3)⁵³ was used to assess the binding of spermine to MCU-EMRE. The CHEMPLP scoring function was employed for docking⁵⁴. The search space covered the full matrix-exposed surface of the MCU–EMRE complex, divided into three overlapping regions (selectivity filter, matrix gate, and matrix surface). We performed exhaustive sampling for each region, with 200,000 genetic algorithm runs, and 200 poses retained per region. Resulting poses were clustered based on RMSD, and representative binding modes were identified. Clusters corresponding to equivalent binding modes across the four MCU subunits were merged due to the four-fold symmetry of the channel.

Molecular Dynamics Simulations

All simulations were performed using the GROMACS software package v.2022.3 and the CHARMM36m force field^55,56. Spermine was parameterized using the CGenFF web server⁵⁷. Initial spermine coordinates were guided by the putative non-protein density observed below the selectivity filter in the cryo-EM map. Spermine was placed on the matrix side of the pore, and no positional restraints were applied during production runs, allowing spontaneous diffusion. System configurations were prepared using CHARMM-GUI, embedding the MCU-EMRE cryo-EM structure into lipid bilayer patches containing 347 POPC lipids⁷⁵. Parameters for cardiolipin (TLCL2) were obtained from the lipid21 force field⁷⁶. Cardiolipin was included at the four sites resolved in the cryo-EM map, providing an EM-guided minimal model of specific protein–lipid contacts. The simulation boxes were solvated with CHARMM TIP3P water molecules⁷⁷, and 75 mM CaCl₂ and 75 mM KCl were added to ensure charge neutrality and achieve an ionic strength of 150 mM. The Ca²⁺ concentration and parameters were chosen according to the multisite Ca²⁺ model guidelines³³.

The negative mitochondrial membrane potential was mimicked by applying an electric field of −290 mV along the z-axis. All systems were subjected to energy minimization using the steepest descents algorithm. The temperature was maintained at 300 K using the Parrinello–Donadio–Bussi velocity rescaling method with a coupling time constant of 1.0 ps, and the pressure was maintained at 1 bar using the C-rescale barostat with a coupling time constant of 5.0 ps^78,79. The simulation time step was set to 2.0 fs. Long-range electrostatic interactions were calculated using the particle–mesh Ewald (PME) method with a Fourier spacing of 0.12 nm and a real-space cutoff of 1.0 nm⁸⁰. Lennard–Jones interactions were force-switched off between 1.0 nm and 1.2 nm. Bond lengths were constrained using the LINCS algorithm⁵⁸. The standard stepwise minimization and equilibration protocol generated by CHARMM-GUI was used to optimize the system. Production simulations ranged between 100 ns and 850 ns. Analyses were performed using GROMACS tools, MDAnalysis, and Python⁵⁹. All data and scripts are available from the author upon reasonable request.

QUANTIFICATION AND STATISTICAL ANALYSIS

Cryo-EM data collection and refinement statistics are summarized in Table 2. Statistical comparison was via Mann-Whitney U-tests or 1-way ANOVA with post hoc Bonferroni-corrected t-test, as indicated in corresponding figure legends, using OriginPro 2025. Bars represent mean ± SEM. Statistically significant results were those where the p-value was <0.05. Dose-response curves were generated by fitting a modified Hill function in OriginPro.

Supplementary Material

Document S1. Tables S1-S2. Figures S1-S14, and supplemental references.

KEY RESOURCES TABLE

REAGENT OR RESOURCE	SOURCE	IDENTIFIER
Antibodies
GAPDH	Cell Signaling Technology	2118L, RRID:AB_561053
FLAG-HRP	Sigma	A8592, RRID:AB_439702
MCU	Cell Signaling Technology	14997S, RRID:AB_2721812
Anti-Rabbit IgG HRP conjugated goat polyclonal	VWR	102645-194
Bacterial and virus strains
Biological samples
Chemicals, peptides, and recombinant proteins
Oregon Green BAPTA 6F	Thermo	O23990
CGP-37157	Cayman Chemicals	15611
Cyclosporin A	Sigma	30024
Carbonyl cyanide-p-trifluoromethoxyphenylhydrazone (FCCP)	Cayman Chemicals	15218
Antimycin A	Sigma	A8674
Ru360	Sigma	557440
Lauryl maltose neopentyl glycol (LMNG-3)	Anatrace	NG310
Cholesteryl hemisuccinate tris salt	Anatrace	CH210
Soy extract polar	Avanti	541602C
Sodium butyrate	Alfa Aesar	A11079
FreeStyle 293 expression medium	Invitrogen	12338018
Doxycycline hyclate	Millipore Sigma	D9891
1, 7-Diaminoheptane	Sigma	D17408
N8-Acetylspermidine dihydrochloride	Sigma	A3658
N1-Acetylspermidine dihydrochloride	Cayman Chemical	9001535
Putrescine dihydrochloride	Sigma	P7505, P5780
N1-Acetylspermine trihydrochloride	Fluka	01467
Spermidine trihydrochloride	Sigma	S2501
Spermidine	Cayman Chemical	14918
Spermine tetrahydrochloride	Sigma	S2876
Spermine	Cayman Chemical	18041
Di(ethyl)dihydroxyhomospermine	Panbela Therapeutics	SBP101
Critical commercial assays
Sep-Pak C18 cartridges	Waters	WAT023590
Superose 6, 10/300 GL	GE Healthcare Life Sciences	17-5172-01
Strep-Tactin®XT 4Flow® high capacity resin	IBA-Lifesciences	2-5030-002
Deposited data
CryoEM map of spermine bound MCU-EMRE complex	This study	EMDB: EMD-77084
The coordinates of spermine bound human MCU-EMRE complex	This study	PDB: 13IO
Experimental models: Cell lines
HEK293T/17 SF cell line	ATCC	ACS-4500, RRID:CVCL_4V93
MCU KO HEK-293T cell line	Vamsi Mootha	N/A
MICU1 KO HEK-293T cell line	Vamsi Mootha	N/A
Experimental models: Organisms/strains
Mouse C57BL/6NJ	Jackson Laboratory	JAX: 005304, RRID:IMSR_JAX:005304
Oligonucleotides
Forward primer, cloning human MCU into MCU-mGFP-EMRE-pTRE-Bidirectional (IDT) CCTCTAGACATATGCTGCAGACGCGTTTATTTCTCAAATTGAGGGTGGGAC	This study	N/A
Reverse primer, cloning human MCU into MCU-mGFP-EMRE-pTRE-Bidirectional (IDT) TGAACCGTCAGATCGCTCTGGAGAATTCGCCACCATGGCGGCCGCCGCAG	This study	N/A
Forward primer, cloning human EMRE into MCU-mGFP-EMRE-pTRE-Bidirectional (IDT) CGGGGCCCAGATCTCCGCGGGGATCCGCCACCATGGCGTCCGGAGCGGCTC	This study	N/A
Reverse primer, cloning human EMRE into MCU-mGFP-EMRE-pTRE-Bidirectional (IDT) ATCCTCTGGAGATATCGCCGGCGGCCGCGACCGGTTTATTTCTCAAATTGAGGGTGGGA	This study	N/A
Forward primer, for creating of MCU-EEAA mutation (IDT) CGCCAGGCATATGTTTATCCAGCAGCCAGAGACAGACAATACTTAC	This study	N/A
Reverse primer, for creating of MCU-EEAA mutation (IDT) TGGCTGCTGGATAAACATATGCCTGGCGTGTCATTACAAAATAT	This study	N/A
Recombinant DNA
Plasmid: MCU-Flag-pLYS1	Vamsi Mootha	N/A
Plasmid: MCU-S259A-pLYS1	Vamsi Mootha	N/A
Plasmid: MCU-EEAA-pLYS1	This study	Addgene 256373
Plasmid: MCU-mGFP-EMRE-pTRE-Bidirectional	This study	Addgene 256374
Software and algorithms
RELION 5.0	Scheres47	https://relion.readthedocs.io/en/release-5.0/
CryoSPARC 4.6.2	Punjani et al.48	https://guide.cryoSPARC.com
Coot 0.9.8.4	Emsley et al.49	https://www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot/
Phenix 2.0-5848	Adams et. al.50	https://phenix-online.org
UCSF ChimeraX 1.9	Pettersen et al.51	https://www.rbvi.ucsf.edu/chimerax/
CTFFIND4	Rohou and Grigorieff52	http://grigoriefflab.janelia.org/ctf
GOLD docking software (CCDC)	Jones et. al.53	https://www.ccdc.cam.ac.uk/solutions/software/gold/
CHEMPLP	Korb et. al.54	N/A
GROMACS	Abraham et. al.55	https://www.gromacs.org/
CHARMM36m	Huang et. al.56	https://mackerell.umaryland.edu/charmm_ff.shtml
CGenFF	Vanommeslaeghe et. al.57	https://cgenff.com/
LINCS	Hess et. al.58	N/A
MDAnalysis	Gowers et. al.59	https://www.mdanalysis.org/
Other
Au 1.2/1.3 holey, 300-mesh gold grids	Quantifoil	R1.2/R1.3

HIGHLIGHTS.

• Mitochondrial matrix polyamine levels vary greatly across organs

• CryoEM, docking, and simulations reveal spermine binds within the MCU pore

• Matrix polyamines inhibit bidirectional Ca²⁺ flux by electrostatic pore occlusion

• CryoEM reveals phospholipids contribute to the MCU conduction pathway