S92 phosphorylation induces structural changes in the N-terminus domain of human mitochondrial calcium uniporter

The mitochondrial calcium uniporter (MCU) plays essential roles in mitochondrial calcium homeostasis and regulates cellular functions, such as energy synthesis, cell growth, and development. Thus, MCU activity is tightly controlled by its regulators as well as post-translational modification, including phosphorylation by protein kinases such as proline-rich tyrosine kinase 2 (Pyk2) and AMP-activated protein kinase (AMPK). In our in vitro kinase assay, the MCU N-terminal domain (NTD) was phosphorylated by protein kinase C isoforms (PKCβII, PKCδ, and PKCε) localized in the mitochondrial matrix. In addition, we found the conserved S92 was phosphorylated by the PKC isoforms. To reveal the structural effect of MCU S92 phosphorylation (S92p), we determined crystal structures of the MCU NTD of S92E and D119A mutants and analysed the molecular dynamics simulation of WT and S92p. We observed conformational changes of the conserved loop2-loop4 (L2-L4 loops) in MCU NTDS92E, NTDD119A, and NTDS92p due to the breakage of the S92-D119 hydrogen bond. The results suggest that the phosphorylation of S92 induces conformational changes as well as enhancements of the negative charges at the L2-L4 loops, which may affect the dimerization of two MCU-EMRE tetramers.

In details of conformational and electrostatic changes of MCU NTD by S92 phosphorylation. To reveal the structural effect of S92 phosphorylation in the MCU NTD, we generated the S92E mutant, an S92p mimic, of MCU NTD fused with the bacteriophage T4 lysozyme at the N-terminal end of MCU NTD (T4-MCU NTD S92E ) to improve protein solubility for crystallographic studies 6 . We determined the structure of T4-MCU NTD S92E at a resolution of 2.50 Å by molecular replacement using the MCU NTD WT (PDB ID: 4XSJ) and T4 lysozyme (PDB ID: 2LZM) structures as templates ( Fig. 2A; Table 1).
The overall structure of MCU NTD S92E was similar to the structure of MCU NTD WT (PDB ID: 4XSJ) with root-mean-square deviation (RMSD) of 0.57 Å for 87 C α atoms, and consisted of two helices (α1 and α2) and six β-strands ( Fig. 2A,B). The S92-D119 in the L2-L4 loops of MCU NTD WT formed a hydrogen bond at a distance of 2.5 Å; the R93 interacted with the E118 to form a salt bridge and stabilized the closed form of L2 loop (Fig. 2B,C). The mutation of S92 to E92 results in atomic clashes of the side chains between E92 and D119, broke the S92-D119 hydrogen bond, and induced conformational changes from the closed form of the L2 loop in MCU NTD WT to the open form (Fig. 2B,C). The peptide backbone of the L2 loop in the MCU NTD S92E moved away from L2 loop of MCU NTD WT (C α atom distance of 4.6 Å), and the side chain of R93 moved up to the position of S92 and formed a new hydrogen bond with E92 and D119. The MCU NTD S92E L90 in the hydrophobic interior, which also contained V88, L115, I122, V125, and I153 in MCU NTD S92E moved away from that of MCU NTD WT at a distance of 2.2 Å (Fig. 2B,C). Based on the MCU NTD S92E structure, we can suggest that the additional phosphate group by the S92 phosphorylation break the S92-D119 hydrogen bond due to atomic clashes between the phosphate group of S92p and the carboxyl group of D119, and induces a conformational change similar to that of the MCU NTD S92E .
In our previous studies, we unintentionally observed that the S92A mutation abolish the S92-D119 hydrogen bond in the structure of the MCU NTD S92A 6 ( Supplementary Fig. S3A). Intriguingly, the conformational changes of the L2-L4 loops in MCU NTD S92A   proteins that were incubated with protein kinase A (PKA), protein kinase C (PKC) isoforms (PKC mixture of α, β, and γ isoforms with lesser δ and ζ; PKC βII ; PKC δ ; PKC ε ), and [γ-32 P]ATP (P-32). We designed all Ser/Thr (T76, S87, S92, T100, S105, S107, S129, S138, T139, and T157) mutants of the MCU NTD except the S92 (MCU NTD AA S92 ). Full autoradiography results in Supplementary  To further investigate whether the S92-D119 hydrogen bond is important for maintaining the closed conformation of the L2-L4 loops, we prepared the D119A mutant to break the S92-D119 hydrogen bond. We determined the structure of the MCU NTD D119A mutant fused with N-terminus T4-lysozyme fusion (T4-MCU NTD D119A ) at 2.85 Å resolution. Overall, the structures of MCU NTD WT and the MCU NTD D119A mutant were similar, with an RMSD of 0.61 Å for 86 C α atoms (Fig. 2B). As expected, the MCU NTD D119A also broke the S92-D119 hydrogen bond from the L2-L4 loops of MCU NTD WT and caused structural changes in the L2-L4 loops, similar to that observed in the structure of MCU NTD S92E (Fig. 2B,D). The L2 loop conformation of the MCU NTD D119A moved away at a C α atom distance of 5.1 Å from that of MCU NTD WT , while the side chain of R93 residue, which moved up to the position of S92, did not form a new hydrogen bond because of lack of a hydrogen bonding counterpart by D119A mutation (Fig. 2B,D).
In addition, to understand whether the S92-D119 hydrogen bond disruption by S92p might contribute to flexibility of L2-L4 loops in the MCU NTD, we performed the ensemble refinement using PHENIX and calculated the root-mean-square fluctuation (RMSF) (Å) from the ensemble refinement results of the MCU NTD WT and the mutants (S92E and D119A) 41,42 . Overall structures of two mutants showed similar RMSF scores in dynamics to that of the MCU NTD WT , while dramatic RMSF changes were observed in the L2-L4 loops of the MCU NTD mutants compared to the MCU NTD WT (Fig. 3A,B).
To investigate whether phosphorylation of S92 in the MCU NTD affected electrostatic charges, we calculated and compared the side chain charges of the residues S92, S92p, S92E, and D119A, at the mitochondrial matrix pH of approximately 7.8 using the Henderson-Hasselbalch equation 43,44 . Negative charges in the mutant S92E (pKa Collectively, these findings suggest that the S92-D119 hydrogen bond formation or disruption, which depends on S92 phosphorylation, regulates the conformation of L2-L4 loops and additional negative charges in the phosphate group of S92p in the MCU complex.

Molecular dynamics simulation analysis of NTD WT and NTD S92p monomers. Molecular dynamics
(MD) simulations were performed on the NTD WT and NTD S92p monomer structures to identify the intra structural changes caused by phosphorylation of S92 in the NTD. The MD simulations clearly showed the flexibility change of the L2-L4 loop region (Fig. 4A,E,F). The fluctuations of all amino acid residues in NTD WT and NTD S92p monomer structures were measured by plotting of the RMSF. The RMSF values of the L2 and L4 loops of the NTD S92p structure were significantly higher than the values of the NTD WT (Fig. 4B). The average RMSF values of the L2 loop for the NTD WT and NTD S92p were 0.76 Å and 1.41 Å, respectively; the values of the L4 loop were 0.63 Å and 0.93 Å, respectively.
To investigate details of the atomic interaction between the residues near the L2 and L4 loops, the final MD trajectory structure was extracted. In NTD WT , the S92 and R93 in the L2 loop were hydrogen-bonded with D119 in the L4 loop (Fig. 4C). Conversely, in NTD S92p , only R93 participated in the hydrogen bond interaction, as the interaction of S92 with D119 was broken (Fig. 4D). Therefore, it can be inferred that phosphorylation on S92 can break the interaction between the S92 and D119.   (Fig. 5A,B,D). Additionally, the MCU-EMRE channel interactions formed a V-shaped tetrameric MCU-EMRE dimer. Moreover, a single mutation, D123R, in the L4 loop of the MCU NTD abolished the dimerization of the two MCU-EMRE channels, possibly by disrupting the electrostatic interactions with the neighboring arginine residues, R93 and R124 19 . To elucidate the effect of S92p on the dimerization of tetrameric MCU-EMRE, we compared the binding energy difference for the dimerization of the tetramer between WT and S92p NTDs using the PRODIGY web server 45 . The binding free energy of NTD S92p (−7.4 kcal/mol) was higher than that of WT (−10.5 kcal/mol), suggesting that the conformational changes and enhancement of negative charges by S92 phosphorylation may affect the dimerization of two MCU-EMRE channels (Fig. 5C-F).
To confirm this hypothesis, we performed MD simulations for the NTD WT and NTD S92p octamer structures. To compare the distance between the two tetramers (NTD-A and NTD-B), the three key monomer pair distances (M2-M8, M3-M7, and M4-M6) were monitored during 10 ns simulations times (Fig. 6). The snapshot structures at 10 ns (Fig. 6A,B) and the distance trajectory during the MD simulation (Fig. 6C-E) show that the distances between the paired monomers in NTD S92p were significantly increased compared to that of the NTD WT by approximately 1.5 to 5 Å. It suggests that the additional negative charges from the phosphate group might contribute to push each tetramers (NTD-A and NTD-B) away. Overall, our MD simulation studies suggest that the S92 phosphorylation can weaken dimerization of the MCU-EMRE tetramer.

Discussion
MCU activity is modulated by its regulatory proteins, including MICU1, MICU2, MCUb, EMRE, and MCUR1, as well as post-translational modifications such as phosphorylation 2,7,8,20,21 . In addition, the MCU NTD plays pivotal roles in MCUR1 interaction, MCUb NTD interaction, Mg 2+ binding selectivity, phosphorylation, redox sensor, oligomerization of MCU-EMRE channel complexes, and regulation of MCU Ca 2+ uptake activity 6,7,[17][18][19] . Thus, we believe functional roles of MCU NTD for its Ca 2+ uptake activity warrants further investigation, although recent studies of NTD deletion of MCU appears to be functionally dispensable in mitochondrial Ca 2+ uptake 19,46 Protein kinases can be localized in the sarcoplasmic reticulum (SR) and mitochondria, and modulate function of Ca 2+ channels by phosphorylation. Phosphorylation of Ca 2+ channels containing ryanodine receptor 2 and inositol 1,4,5-trisphosphate receptors regulates Ca 2+ release in the SR through PKA and CaMKII 20,47-51 . The MCU is directly phosphorylated by Pyk2 and AMPK and phosphorylated MCU facilitates Ca 2+ entry into the mitochondria 20,21 . The conserved S92 in the MCU is a putative recognition site (89-RLPS-92; RxxS motif) for phosphorylation by Ser/Thr kinases such as CaMKII, PKA, and PKC on the basis of prediction of KinasePhos 2.0  25 , although the regulatory functions of MCU activity by CaMKII still remain controversial [22][23][24][25][26] . Instead, our in vitro kinase assay results indicated MCU S92 was phosphorylated by PKC isoforms (PKC βII , PKC δ , and PKC ε ) localized in the mitochondrial matrix, but was not phosphorylated by PKA (Fig. 1D,F) 27,28 . Further studies are needed to understand the functional roles of MCU NTD phosphorylation by PKC isoforms.
Free radicals, such as ROS and reactive nitrogen species (RNS), generate in a well-modulated manner to maintain cellular homeostasis as signalling second messengers, and play critical roles in the activation of enzymes and alteration of lipids, protein, and DNA 52,53 . Under physiological conditions, the MCU uptakes Ca 2+ ions into the www.nature.com/scientificreports www.nature.com/scientificreports/ matrix; Ca 2+ ions play an essential role modulating ATP synthesis through TCA cycle and the electron transfer chain, and finally induce ROS production as by-products in the mitochondria 54 . However, continuous overload of mitochondrial Ca 2+ entry can produce large amounts of ROS and eventually lead to apoptotic or necrotic cell death 54 . Upon production of ROS, PKC βII , PKC δ , and PKC ε translocate to the mitochondria and modulate functions of enzymes and Ca 2+ channels by Ser/Thr phosphorylation, as well as ROS production [27][28][29] . In our studies, we observed that PKC βII , PKC δ , and PKC ε phosphorylated the MCU S92 in vitro. We speculate that regulation of the S92 phosphorylation by the PKC isoforms under physiological conditions play important roles in ROS homeostasis or programmed cell death by excessive ROS, driven by the MCU Ca 2+ uptake. Additional experimental evidence will be required to clarify the functional roles of PKC isoforms in the MCU [29][30][31] .
In conclusion, we identified that the PKC isoforms, PKC βII , PKC δ , and PKC ε , are capable of phosphorylating S92 in the MCU NTD. We also characterized local conformational changes in our structural determination of MCU NTD S92E and NTD D119A as well as in MD simulation analysis of the WT and S92p. The conformational changes and enhancement of negative charge of the L2-L4 loops in the MCU NTD by S92 phosphorylation may be essential for regulating MCU activity, despite there lacks of functional data for the MCU activity modulation by S92 phosphorylation. Further studies are required to reveal the functional effects of MCU S92 phosphorylation by the PKC. The results provide a framework for further studies investigating the functional and structural roles of MCU phosphorylation by PKC.  55,56 in the modified pET21a vector (Novagen), was constructed as previously described 6 . Point mutagenesis using polymerase chain reaction (PCR) was performed to construct the S92E or D119A mutants. For the in vitro kinase assay of the MCU NTD S92, we designed all Ser/Thr to Ala mutants (T76, S87, S92, T100, S105, S107, S129, S138, T139, and T157; MCU NTD AA ). PCR was used to synthesize the MCU NTD AA construct using 12 oligonucleotides; then, MCU NTD AA S92 was generated using the A92S mutation from the MCU NTD AA .
Data collection, structure determination, and refinement. Diffraction data of T4 lysozyme-MCU NTD S92E or T4 lysozyme-MCU NTD D119A crystals were collected at 100 K using synchrotron X-ray sources on beamlines 5 C at the Pohang Acceleratory Laboratory (PAL) (Pohang, South Korea). We finally collected diffraction data for T4 lysozyme-MCU NTD S92E at a resolution of 2.50 Å and for T4 lysozyme-MCU NTD D119A at 2.85 Å using a single wavelength, 0.9795 Å. The diffraction data were processed using the HKL2000 suite 57 . Molecular replacement was carried out using Phaser in the CCP4 suite 58 , using the structures of the bacteriophage T4 lysozyme (PDB ID: 2LZM) and MCU NTD (PDB ID: 4XTB) as templates. The obtained models were subjected to iterative rounds of model building and refinement using programs Coot 59 and REFMAC5 in CCP4 suite 58 . The details of data collection and refinement statistics are provided in Table 1.
The amino acid sequence and protein surface conservation of the MCU NTD were calculated using the ConSurf server 30 . The CCP4 program LSQKAB was used to superimpose the structures of MCU NTD WT (PDB ID: 4XTB), MCU NTD S92E , and MCU NTD D119A and to estimate RMSD (Å) scores of C α atoms 58 . The electrostatic surface charges of MCU NTDs (WT, S92E, S92p, and D119A) were analysed using the PDB2PQR server 60  Ensemble refinement. Ensemble refinement for T4 lysozyme-MCU NTD WT (PDB ID: 4XSJ), T4 lysozyme-MCU NTD S92E , and T4 lysozyme-MCU NTD D119A was performed using structures and structural www.nature.com/scientificreports www.nature.com/scientificreports/ factors by phenix.ensemble_refinement 41 . Default parameters were used in the phenix.ensemble_refinement, including pTLS = 0.8 and T bath = 5 K, and solvent updated every 25 cycles. The simulations have an equilibration phase (10τx) in which the temperature, X-ray weight and averaged structure factors stabilize, followed by an acquisition phase (10τx). The output structures by ensemble refinement were visualized using PyMOL version 1.5.0.4 (Schrödinger LLC) with a script 'ens_tool.py' . The RMSF difference histogram for the MCU NTD WT and mutants (S92E and D119A) was plotted using SigmaPlot 12.

Molecular dynamic simulations. Four molecular dynamics (MD) simulations of NTD WT and NTD S92p
monomer and octamer structures were carried out using GROMACS (GROningen Machine for Chemical Simulations) 2018.4 package 61-63 with amber99sb-star-ILDNP force field 64 . Molecular topologies for phosphorylated S92 were generated by AnteChamber Python Parser interface (ACPYPE) with generalized AMBER force field 2 (GAFF2) 65,66 . All four systems were solvated with TIP3P water molecules 67 in a dodecahedron box and Na + counter ions were added to neutralize the net changes of the systems by replacing water molecules. In all cases, bond lengths were constrained with LINCS 68 and long-range electrostatics were calculated using the smooth particle mesh Ewald (PME) method with a cut-off of 1.0 nm 69,70 . A cut-off of short-range non-bonded interactions, van der Waals (vdW), were truncated at 1.0 nm. All MD simulations were conducted energy minimization using the steepest descent method. Equilibration was then performed in two phases, during which position restraints applied to all heavy atoms of the protein. First, the simulations were run under NVT conditions at 300 K, using Berendsen's coupling algorithm 71 for 100 ps. The second phase of equilibration was carried out an NPT ensemble for 100 ps, using the Nose-Hoover thermostat 72,73 and the Parrinello-Rahman barostat 74,75 with coupling time constants of 2.0 ps and 5.0 ps to maintain 300 K and 1 bar, respectively. Production MD was then conducted for 10 ns without any restraint and under the same conditions as the NPT ensemble. All analyses of MD simulation results were performed using the analysis tools in the GROMACS package.
Accession numbers. Atomic coordinates and structure factors of T4 lysozyme-MCU NTD S92E and T4 lysozyme-MCU NTD D119A have been deposited in the PDB with the accession numbers, 6JG0 and 6KVX, respectively.