Cyanidioschyzon merolae aurora kinase phosphorylates evolutionarily conserved sites on its target to regulate mitochondrial division

The mitochondrion is an organelle that was derived from an endosymbiosis. Although regulation of mitochondrial growth by the host cell is necessary for the maintenance of mitochondria, it is unclear how this regulatory mechanism was acquired. To address this, we studied the primitive unicellular red alga Cyanidioschyzon merolae, which has the simplest eukaryotic genome and a single mitochondrion. Here we show that the C. merolae Aurora kinase ortholog CmAUR regulates mitochondrial division through phosphorylation of mitochondrial division ring components. One of the components, the Drp1 ortholog CmDnm1, has at least four sites phosphorylated by CmAUR. Depletion of the phosphorylation site conserved among eukaryotes induced defects such as mitochondrial distribution on one side of the cell. Taken together with the observation that human Aurora kinase phosphorylates Drp1 in vitro, we suggest that the phosphoregulation is conserved from the simplest eukaryotes to mammals, and was acquired at the primitive stage of endosymbiosis.

M itochondria are organelles that play important roles in energy production and several metabolic pathways in eukaryotic cells, and are thought to be derived from a sole alphaproteobacterium that invaded the protoeukaryote 1,2 . The endosymbiosis of mitochondria is thought to be crucial for the evolution of eukaryotic cells 3 . Because the machinery regulating mitochondrial division, such as a mitochondrial division ring, is well conserved among eukaryotes, it can be predicted that the mechanism, which regulates mitochondrial division was acquired in the process of establishing this endosymbiosis [3][4][5] .
Some nuclear-encoded genes derived from the ancestor host cell, such as Drp1, Opa1, and Fis1, control mitochondrial morphology in animal cells 6 . Drp1 forms a mitochondrial division ring at the mitochondrial division site. Mitochondrial division is induced by contraction, which is caused by a structural change in Drp1. Drp1 recruitment to mitochondria depends on an integral membrane protein, Fis1, which is located in the mitochondrial outer membrane 6 . In addition, the mitochondrial outer membrane protein Mff recruits Drp1 to mitochondria independently of Fis1 6 . The dynamin-related protein Opa1, which is localized to the mitochondrial inner membrane, induces mitochondrial fusion 6 . In animal, yeast and plant cells, regulation of mitochondrial fragmentation in the G2/M phase by mitotic kinases is required for the proper distribution of mitochondria to daughter cells 6 . Previous studies using human cultured cells have revealed that mitochondrial proliferation occurs before mitosis, and that this phenomenon is controlled by phosphoregulation of Drp1 by cyclin-dependent kinase1 7 . The mitochondrial localization of Drp1 is regulated by Aurora A via phosphorylation of RalA, which belongs to the Ras family, part of the small GTPase superfamily 8 . Phosphorylated RalA forms a protein complex with Drp1, RalBP1, Cyclin B, and CDK1, which is recruited to the mitochondrial outer membrane in a Mff-dependent manner 8 .
Cyanidioschyzon merolae is an ultra-small (> 2 µm in diameter) unicellular red alga that inhabits hot and sulfate-rich springs. C. merolae has a simple cell structure, with a single mitochondrion and a single plastid 9,10 . The mitochondrion divides only once per cell cycle, during M phase [9][10][11] . These features would enable us to investigate the mitochondrial division precisely. In addition, sequencing of the C. merolae genome and the development of molecular genetic analysis methods have made it possible to investigate detailed molecular mechanisms in this organism [12][13][14][15][16] . Therefore, we assumed that C. merolae is the best model for investigating the molecular mechanisms of coordination between mitochondria and host cells.
Aurora kinases are members of a highly conserved mitotic kinase family, which controls mitotic spindle formation and chromosome segregation [17][18][19][20][21][22] . Although many other eukaryotes have two or three Aurora kinases, C. merolae has only one Aurora kinase gene (CmAUR) in its genome 23 . Previously, we revealed that a green fluorescent protein (GFP) fusion protein of CmAUR in C. merolae was localized to the spindle and the mitochondrion 23 . Localization to the mitochondrion was observed throughout the cycle, and CmAUR-GFP was especially accumulated at the mitochondrial division site in M phase, which is the only phase when mitochondrial division occurs in C. merolae 23 . Its localization to the mitochondrion was shown to be regulated by a kinesin-like, division-inducing protein (TOP) 24 . Given these findings, we presumed that CmAUR regulates mitochondrial division during M phase.
To test this hypothesis, here, we investigated whether regulation of mitochondrial division by a mitotic kinase occurs in primitive eukaryotic cells using C. merolae. Our results indicate that Aurora kinase regulates mitochondrial division in G2/M phase through phosphorylation of the Drp1 ortholog CmDnm1. Our results also show that human Aurora kinases directly phosphorylate Drp1 in vitro. These results suggest the conservation of this mechanism among eukaryotes. The mechanism may have been acquired in common ancestral cells and may have contributed to the establishment of endosymbiosis.

Results
Evaluation of the Aurora kinase ortholog CmAUR. Although some previous studies have focused on CmAUR localization in C. merolae 23,24 , little is known about its role as an Aurora kinase, such as in mitotic progression through histone H3 Ser10 phosphorylation in mitosis 25 . Thus, we considered that it was important to test the Aurora kinase function of CmAUR. First, we conducted immunostaining to determine the cellular localization of CmAUR. Previously, we reported a CmAUR antibody and confirmed its specificity 24 . The immunostaining showed that CmAUR localized to the mitotic spindle and mitochondrion (Fig. 1a, Supplementary Fig. 1). To investigate its function in cell cycle progression, we transiently expressed a kinase-dead version of CmAUR (CmAUR K208R ) in C. merolae. This mutant of CmAUR was designed by reference to previous reports of kinasedead human Aurora A 26 and B 27 , and we confirmed that recombinant CmAUR K208R protein had no autophosphorylation activity (Fig. 1b). Expression of CmAUR K208R in C. merolae delayed M phase, and this was most likely caused by a dominantnegative effect, analogous to a previous study in human 27 (Fig. 1c). These results suggested that CmAUR has an important role in M phase in C. merolae, like its orthologs in other eukaryotes. We also tested whether CmAUR has capability as a kinase of histone H3 Ser10. Similar to other species, C. merolae showed phosphorylation of histone H3 Ser10 during M phase, as reported previously 28 (Fig. 1d, Supplementary Fig. 2). Recombinant glutathione-S-transferase (GST)-CmAUR strongly phosphorylated H3 Ser10 in vitro (Fig. 1e). Although a weak histone H3 phospho Ser10 band was detected in the GST only-treated sample, the intensity was not different from the band in the non-GST-treated sample (Fig. 1e). Therefore, we consider that H3 Ser10 phosphorylation by GST-CmAUR depends on CmAUR. For further confirmation of whether CmAUR works as an Aurora kinase, we treated GST-CmAUR in vitro with hesperadin, an Aurora kinase inhibitor. Phosphorylation of H3 Ser10 by GST-CmAUR was inhibited by hesperadin (Fig. 1f). In summary, these results suggest that CmAUR functions as an Aurora kinase.
Relationship between mitochondrial division and CmAUR. To investigate the molecular processes regulated by CmAUR, we conducted immunostaining analyses of autophosphorylated CmAUR. Previous studies suggested that Aurora kinase activates itself through autophosphorylation 29,30 . Because the selfphosphorylation site of Aurora kinase is well conserved among Aurora kinase orthologs, a commercial human phospho-Aurora (phAUR) antibody was able to detect self-phosphorylated CmAUR ( Supplementary Fig. 3a-d). By using the phAUR antibody, we demonstrated that activated CmAUR partially localized to the mitochondrion (Fig. 2a, Supplementary Fig. 4). Coimmunostaining using phAUR and an antibody against the mitochondrial division ring component Mda1 revealed that phAUR was localized to the mitochondrial division ring (Fig. 2b). To investigate whether the localization changed depending on cell cycle progression, we analyzed phAUR and Mda1 in both G2 and M phase. The intensity of the phAUR signals on the mitochondrial division ring increased as mitochondrial division progressed (Student's t test, p < 0.01) but not Mda1 signals (Fig. 2b, c). Because Mda1 is a protein that accumulates on the mitochondrial division ring, this result suggests that the phAUR signals increased independently of the accumulation of other mitochondrial division-related proteins in M phase such as Mda1. After mitochondrial division, the phAUR signal was observed on structures other than the mitochondrion (Fig. 2a, white arrowheads). Considering the morphological features and a previous report on C. merolae mitosis 31 , we deduced that this structure was the mitotic spindle. However, localization on speckles, which are clearly different from the spindle, was also observed. This suggests that CmAUR does not co-localize exclusively with mitochondria, as already reported in human Aurora A 25 . We conclude that CmAUR activity is involved in both mitochondrial division and mitotic spindle formation.
To confirm the role of CmAUR in mitochondrial division during mitosis, we overexpressed both wild-type and kinasedead type CmAUR. Then we measured the frequency of each phase of the transformants (interphase, pre-mitochondrial division phase, and post-mitochondrial division phase) in the mitochondrial replication process (Fig. 2d). The transformants overexpressing both CmAUR K208R and wild-type CmAUR were arrested at the pre-and post-mitochondrial division phases (Fig. 2e, f; Supplementary Tables 1, 2). Because the proportion of wild-type and CmAUR K208R overexpressing cells in each phase did not seem to be different, it is not clear whether the kinase activity of CmAUR had an impact on the mitochondrial division. However, these data indicate that CmAUR kinase activity and/or another CmAUR function can regulate both nuclear and mitochondrial division. Division of the mitochondrion occurs before nuclear division in C. merolae 31 . Thus, the inhibition of mitochondrial division was most likely not caused by nuclear division. White arrowheads indicate signals localized to the spindle or spindle pole. Black arrowheads indicate plastid autofluorescence. Arrows indicate plastid DNA. The Pearson correlation coefficient (PCC) in each cell was calculated using areas without plastids. We observed 24 cells and present representative images in this figure. b Autophosphorylation assay of wild-type and mutant CmAUR. ATP-γS was used as a substrate for autophosphorylation, and the phosphorylation was detected by western blotting using thiophosphate ester antibody. c Mitotic inhibition in a dominant-negative mutant of CmAUR. CmAUR K208R is a kinase-dead mutant; GFP was used as a negative control. Relevant genes were transiently overexpressed in C. merolae and mitotic cells among transfectants were counted. The error bars indicate standard error of the mean. d Immunofluorescence image of histone H3 Ser10 phosphorylation in mitosis. More than 18 cells were observed. e In vitro phosphorylation assay of histone H3 Ser10 with recombinant CmAUR. Glutathione-S-transferase (GST) was used as a negative control. Distilled water was used as a blank sample. f In vitro phosphorylation assay of histone H3 Ser10 treated with hesperadin. The concentration of hesperadin was 10 µM. The concentration of dimethyl sulfoxide in the final solution was 0.1% (v/v). Bars: 1 µm a, 2 μm c.
To predict the localization of substrates phosphorylated by CmAUR in vivo, we conducted co-staining for CmAUR and with ProQ diamond. ProQ diamond binds total phosphorylated proteins and we previously demonstrated that ProQ diamond staining is effective in C. merolae 24 . In co-stained M-phase C.
merolae cells, CmAUR signals were partially co-localized with ProQ signals (Fig. 2g, h). Because CmAUR was localized to the mitochondrion (Fig. 1a, Supplementary Fig. 1 Identification of the substrates of CmAUR. Because localization analysis of phAUR suggested a relationship between mitochondrial division and CmAUR, we predicted that CmAUR is involved in mitochondrial division through phosphorylation of mitochondrial division ring components. To verify this presumption, we performed in vitro kinase assays using recombinant CmDnm1, Mda1, and TOP, which are known to be involved in mitochondrial division in C. merolae 11,24,32 . CmDnm1 is a highly conserved protein among eukaryotes and an orthologue of human Drp1 (DNM1L), which forms a ring around mitochondria and induces their division in eukaryotes such as mammalian, nematode, and yeast cells [33][34][35] . In these assays, CmAUR phosphorylated CmDnm1, Mda1, and TOP (Fig. 3a-c). These results suggested that CmAUR regulates mitochondrial division through phosphorylation of mitochondrial division ring components in C. merolae.
Determination of phosphorylation sites of CmDnm1 by CmAUR. On the basis of our findings that CmAUR is involved in mitochondrial division and in the phosphorylation of proteins related to mitochondrial division, we hypothesized that CmAUR regulates mitochondrial division via direct phosphorylation of mitochondrial division ring proteins. We predicted that phosphorylation of CmDnm1 by CmAUR would be important for mitochondrial division because CmDnm1, as a component of the mitochondrial division ring, has a crucial role in mitochondrial division in C. merolae 4,11 . To further analyze phosphoregulation of CmDnm1 by CmAUR, in vitro phosphorylation sites in CmDnm1 were identified by mass spectrometry ( Supplementary  Fig. 5). Taking evolutionary conservation among eukaryotes into consideration, nine residues were determined as candidate sites for phosphorylation. Through in vitro kinase assays of point mutants of these nine residues, four phosphorylation sites (T139, S570, S726, S732) were confirmed to be phosphorylated by CmAUR (Fig. 4a). In addition, to confirm each single phosphorylation site, we quantified the western blot signal reduction of phosphorylation site variants. As a result, signal densities of T139A and S726A variants were reproducibly reduced compared with wild-type CmDnm1 signals (Fig. 4b).
To reveal the function of these phosphorylation sites in mitochondrial division, we produced Dnm1 variants in which phosphorylation sites T139, S570, S726, and S732 were substituted to nonphosphorylatable alanine or phosphomimetic glutamic acid and overexpressed them in C. merolae. The effects of these variants were quantified by determining the frequency of each mitochondrial division phase in each transformant (Fig. 2d). The population in each mitochondrial division phase was not affected by overexpression of CmDnm1 S570A/E , CmDnm1 S726A , or CmDnm1 S732A/E , but the number of cells in pre-mitochondrial division phase was increased by overexpression of CmDnm1 T139A/E and CmDnm1 S726E (Fig. 4c; Supplementary Table 4). This result suggests that the expression of CmDnm1 T139A/E and CmDnm1 S726E causes suppression of mitochondrial division. Previously, it was reported that mitochondrial division is stalled by overexpression of dominant-negative Drp1 mutants (phosphodeficient or mimetic substitution within the GTPase domain) 33 . We observed that overexpression of CmDnm1 T139A/E and CmDnm1 S726E caused mitochondrial division arrest. These results suggest that T139A/E and S726E function, respectively, in a dominantnegative mode. Interestingly, overexpression of CmDnm1 with T139 variations caused abnormal phenotypes such as mitochondrial distribution on one side of the cell (Fig. 4c- 31 . Because aberrant phenotypes were not frequent among the transformants of CmDnm1 S726E , we predict that phosphoregulation of T139 has a more important role in mitochondrial division than does phosphoregulation of S726. To investigate whether the phosphorylation of T139 regulates the localization of CmDnm1, we constructed stably transformed lines that expressed superfolder GFP (sfGFP)-CmDnm1, sfGFP-CmDnm1 T139A , and sfGFP-CmDnm1 T139E under the control of a heat-inducible promoter 16 . The heat-inducible expression enables CmDnm1 variant localization to be observed without the effect of a transfection reagent containing polyethylene glycol. After heat induction, we immunostained cells using an Ef-Tu 36 antibody as a mitochondrial marker and a GFP antibody. We found no differences in the intracellular localization of each CmDnm1 variant (Supplementary Fig. 6). This result indicates that phosphorylation of T139 may regulate CmDnm1 independently from the intracellular localization control.
Finally, we examined if recombinant human Aurora kinase could directly phosphorylate recombinant Drp1, which is a CmDnm1 orthologue found in human. Aurora A and Aurora B could phosphorylate Drp1 in vitro (Fig. 4h). This result suggests that the direct phosphoregulation of CmDnm1 orthologs by Aurora kinase is conserved between C. merolae and human.

Discussion
The results of the present study suggest that CmAUR is involved in mitochondrial division through direct phosphorylation of CmDnm1, the Drp1 orthologue in C. merolae. Residues T139 and S726 of CmDnm1 were found to have functional impacts on mitochondrial division in C. merolae (Fig. 4c-g).
In this study, CmAUR was found to phosphorylate multiple components of the mitochondrial division machinery in vitro (Fig. 3a-c). These results suggested that CmAUR has multiple roles in regulating mitochondrial division through phosphorylation of multiple targets such as CmDnm1, Mda1, and TOP. In a previous report, overexpression and inhibition of Aurora A resulted in mitochondrial elongation in mammalian cells 37 ,  Table 3). b Quantification of in vitro kinase assays of CmDnm1 single amino-acid variants. Signals from different membranes were normalized against the luminosity of wild-type CmDnm1. The relative value of wild-type CmDnm1 was 1. n = 3 for T139A, n = 4 for S726A. The error bars indicate standard error of the mean. c Frequency of transformants expressing CmDnm1 variants. Alanine and phosphomimetic glutamic acid mutants of CmDnm1 T139 and CmDnm1 S726 were overexpressed in C. merolae cells. According to classification Fig. 2d,  suggesting that Aurora kinase has a complex role in controlling mitochondrial morphology. In summary, the complex role of Aurora kinase in mitochondrial division may be conserved in C. merolae and mammalian cells.
The nonphosphorylatable mutant CmDnm1 T139A and the phosphomimetic mutant CmDnm1 T139E exhibited deficient mitochondrial division phenotypes (Fig. 4c-g). This result suggested that residue T139 of CmDnm1 has an important role in mitochondrial division and that strict spatiotemporal control of the switching between phosphorylation and dephosphorylation is necessary for proper mitochondrial division. T139 is located in the GTPase domain ( Supplementary Fig. 7). This suggests that T139 phosphorylation may affect the recruitment of GTP to CmDnm1 and GTP hydrolysis.
However, the percentage of the mitochondrial dividing population of CmDnm1 T139A and CmDnm1 T139E -expressing cells in each phase did not seem to be different (Fig. 4c). From this result, we cannot exclude the possibility that the nonphosphorylated threonine residue at position 139 of CmDnm1 is important for the regulation of mitochondrial division and the phenotype on amino-acid substitution to alanine and aspartic acid was an artifact. In order to prove the importance of T139 phosphorylation of CmDnm1 in mitochondrial division, further investigations are needed. In the case of S726 in CmDnm1, a delayed mitochondrial division phenotype occurred only in the phosphomimetic mutant (Fig. 4c). Therefore, we considered that the only function of phosphorylation at S726 is to suppress mitochondrial division. SSDB (Sequence Similarity Data Base) motif research into CmDnm1 suggested that S726 in CmDnm1 is in the GTPase effector domain (GED) (Supplementary Fig. 7). The GED domain in the dynamin family is a component of the stalk which is involved in homodimerization of dynamin 38,39 . Interestingly, in the CmDnm1 orthologs of higher animals and plants, S726 is replaced with acidic amino acids ( Supplementary Fig. 8). Therefore, this site is thought to be responsible for a mitochondrial division control mechanism unique to C. merolae.
In this study, we discovered that the direct phosphoregulation of CmDnm1 by Aurora kinase is one of the important mechanisms in mitochondrial division in a primitive red alga. The equivalent of T139 in CmDnm1 is well conserved among eukaryotes ( Supplementary Fig. 8), and kinase assay using human recombinant Drp1 and human recombinant Aurora kinase indicated the direct phosphorylation of Drp1 by Aurora kinase (Fig. 4h). Moreover, recent study discovered the indirect regulation of mitochondrial division by Aurora kinase 8 and intramitochondrial localization of Aurora kinase in mammalian cells 37,40 . These facts demonstrate that the relationship between mitochondrial division and Aurora kinase is common to C. merolae and other organisms. Therefore, we conclude that the role of Aurora kinase in mitochondrial division is conserved in eukaryotes. These findings suggest that the regulation of mitochondrial division during the cell cycle by a mitotic kinase was acquired in common ancestral cells, and that this function may have contributed to the establishment of endosymbiosis.

Methods
Materials. We used C. merolae 10D and M4 provided by Professor Tsuneyoshi Kuroiwa.
Synchronization culture. Synchronized culture was conducted as previously described 41 .
Sample preparation for western blot of whole cell protein. Synchronized cultured mitotic cells and normally cultured interphase cells were resuspended in lysis buffer (25 mM pH7.5 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), 150 mM NaCl, 1 mM ethylenediaminetetraacetic acid (EDTA), 1% Triton X-100, 0.1% Na-deoxycholate, 0.1% sodium dodecyl sulfate) supplemented with complete protease inhibitor cocktail (Merck). After incubation for 90 minutes on ice, glass beads were added to the tube, and the tube was shaken using a bead beater (2600 rpm for 1 min). The supernatant of the extracts was used for western blotting.
Transformation of C. merolae. To produce expression vectors for C. merolae, the open reading frame (ORF) of CmAUR and the ORF-3′-UTR (140 bp) of Dnm1 were fused with the β-tubulin promoter of C. merolae (1955 bp) by polymerase chain reaction using PrimeSTAR HS (TaKaRa, Shiga, Japan). The PCR products were introduced into the multiple cloning sites of pTH2-PL and pMtGFP, respectively. Mutagenesis of CmAUR and Dnm1 was carried out by PCR with KOD plus-Neo polymerase (Toyobo, Osaka, Japan). Primers are shown in Supplementary Table 6. Each plasmid (10 μg) was used to transform C. merolae as previously described 15 .
In vitro kinase assay. The CmAUR kinase assay was performed as previously described 43 . When immunoblotting was used to detect phosphorylation, ATP was substituted by ATPγS to use the thiophosphate ester for detection. The substrates resolved in elution buffer were used at < 1/60th of the volume of the total reaction solution. The reaction volume was 25 µL. The reaction buffer contained 5 mM HEPES (pH7.5), 50 mM sucrose, 100 mM KCl, 0.1 mM CaCl 2 , 0.4 mM ethylene glycol tetraacetic acid, 1.2 mM ATPγS, and 20 mM MgCl 2 . After the reaction at 30°C for 30 min, 1.5 μL of 50 mM p-Nitrobenzyl mesylate and 2.6 μL of 500 mM EDTA were added to the mixture.
Mass spectrometry analysis of CmDnm1 phosphorylation. CmDnm1 phosphorylated by CmAUR in vitro as described above was separated on 7.5% TGX gels, and the gels were stained with Bio-Safe Coomassie G-250 Stain. An LTQ-Orbitrap XL (Thermo Fisher Scientific) coupled with an EASY-nLC 1000 (Thermo Fisher Scientific) was used for nano-LC-MS/MS analyses as described previously 44 . The result of this analysis is shown as "Experiment 1" in Supplementary Fig. 5. We also conducted LC-MS/MS analysis with another sample preparation method. The data obtained by the following method is shown in Supplementary Figure 5 as "Experiment 2". The sample solution from the in vitro kinase assay was dried by evaporation and the residue was dissolved in denaturing solution (8 M urea, 50 mM ammonium bicarbonate, 10 mM DTT). The mixture was incubated at 37°C for 30 min and then subjected to reductive alkylation treatment with iodoacetamide. Then, 50 mM ammonium bicarbonate was added and the urea concentration in the solution was diluted to 2 M. Finally, trypsin was added, and the mixture was incubated at 37°C for 16 h. After the reaction, the peptide solution was desalted using a C 18 Stage Tip. Phosphorylated peptides were fractionated and recovered from the peptide sample, and TiO 2 resin was used for specific adsorption. The procedure was based on that recommended for the Titansphere Phos-TiO Kit (GL Science, Tokyo, Japan). The eluted sample was dried under reduced pressure, and the residue was dissolved in a mixture of water, acetone, and trifluoroacetic acid Statistics and reproducibility. Student's t test and Welch's t test were used for testing the difference in population means. Data sets whose P value of f test were larger than 0.05 were used Student's t test, and one whose p value was < 0.05 were used Welch's t test. For statistical analysis of count data based on mitochondrial morphology, we performed Pearson's chi-squared test, and the residuals were calculated.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The mass spectrometry data were deposited in the jPOST repository 45