DRP3 and ELM1 are required for mitochondrial fission in the liverwort Marchantia polymorpha

Mitochondria increase in number by the fission of existing mitochondria. Mitochondrial fission is needed to provide mitochondria to daughter cells during cell division. In Arabidopsis thaliana, four kinds of genes have been reported to be involved in mitochondrial fission. Two of them, DRP3 (dynamin-related protein3) and FIS1 (FISSION1), are well conserved in eukaryotes. The other two are plant-specific ELM1 (elongated mitochondria1) and PMD (peroxisomal and mitochondrial division). To better understand the commonality and diversity of mitochondrial fission factors in land plants, we examined mitochondrial fission-related genes in a liverwort, Marchantia polymorpha. As a bryophyte, M. polymorpha has features distinct from those of the other land plant lineages. We found that M. polymorpha has single copies of homologues for DRP3, FIS1 and ELM1, but does not appear to have a homologue of PMD. Citrine-fusion proteins with MpDRP3, MpFIS1 and MpELM1 were localized to mitochondria in M. polymorpha. MpDRP3- and MpELM1-defective mutants grew slowly and had networked mitochondria, indicating that mitochondrial fission was blocked in the mutants, as expected. However, knockout of MpFIS1 did not affect growth or mitochondrial morphology. These results suggest that MpDRP3 and MpELM1 but neither MpFIS1 nor PMD are needed for mitochondrial fission in M. polymorpha.

1 Graduate School of Agricultural and Life Sciences, The University of Tokyo, Tokyo, 113-8657, Japan. 2 Graduate School of Science, Kobe University, Kobe, 657-8501, Japan. 3 Graduate School of Biostudies, Kyoto University, Kyoto, 606-8502, Japan. 4 PRESTO, Japan Science and Technology Agency, Saitama, 332-0012, Japan. Nagisa Nagaoka and Akihiro Yamashita contributed equally to this work. Correspondence and requests for materials should be addressed to S.-i.A. (email: arimura@mail.ecc.u-tokyo.ac.jp) The Arabidopsis genome has two closely related homologues of Fis1 (FIS1A/BIGYIN and FIS1B). FIS1A and FIS1B are reported to target to mitochondria, peroxisomes and chloroplasts, and were shown to facilitate mitochondrial fission [29][30][31] . BLAST searches of the Arabidopsis genome (https://www.arabidopsis.org/Blast/index.jsp) did not reveal any homologues of other mitochondrial fission factors (Mdv1p/Caf4p, Mff or MiD49/51). The closest matches had e-values > 1e-4, indicating very weak matches. On the other hand, it has two plant-specific fission factors, elongated mitochondria1 (ELM1) and peroxisomal and mitochondrial division factors (PMDs). ELM1 was identified as the gene responsible for the elongated mitochondria mutant phenotype in Arabidopsis. In wild-type Arabidopsis, ELM1 surrounds mitochondria and interacts with both DRP3A and DRP3B 32 . These results suggest that ELM1 is required for relocalization of DRP3A from the cytosol to mitochondrial fission sites. Elongated mitochondria are also observed in the pmd1 mutant, but PMD1 and its paralog PMD2 do not physically interact with DRP3 or FIS1, suggesting that PMD1 facilitates mitochondrial proliferation in a DRP3/ FIS1-independent manner 33 .
Recent molecular phylogenetic analyses of land plants agreed the basal position bryophytes, encompassing liverworts, mosses and hornworts, to vascular plants, and liverworts are considered as one of the earliest diverging distant land plant lineages [34][35][36] . Therefore, liverworts are could be a key clade to our understanding of the commonalities and differences of mitochondrial fission factors among land plants. Here, we used reverse genetics to study candidate genes for mitochondrial fission in the liverwort Marchantia polymorpha. Marchantia is emerging as an experimental model organism, because it has little genetic redundancy in many cases and is well suited for various strategies for molecular genetics, such as introduction of reporter constructs, gene silencing and targeted gene modification [37][38][39][40] . Our results demonstrate that mitochondrial fission in a liverwort relies on some factors used in vascular plants (DRP3 and ELM1) but does not seem to rely on other factors (FIS1 and PMDs).

Results
The Marchantia genome has single copies of homologues of DRP3, FIS1 and ELM1. For searches of the M. polymorpha genome, we used the JGI M. polymorpha EST and genome databases ver. 3.1 (https://phytozome.jgi.doe.gov/pz/portal.html). BLASTx searches of the JGI M. polymorpha EST and genome databases for homologues of DRP3, FIS1, ELM1 and PMD revealed no PMD homologue (e-value of closest match > 1e-4), but did reveal single copies of homologues of DRP3 (0.0 or 6e-71), FIS1 (5e-35) and ELM1 (1e-138). We call these homologues MpDRP3 (Mapoly0069s0084), MpFIS1 (Mapoly0147s0019) and MpELM1 (Mapoly0038s0050), respectively. Two different mRNAs of MpDRP3 were found and appear to be transcribed from a single gene (Figs 1a-c and S1a). The predicted splicing variants, MpDRP3s (short) and MpDRP3l (long), differed by 27 bp in exon 13. RT-PCR analysis revealed that the mRNA expression levels of MpDRP3s and MpDRP3l in the thallus were similar (Fig. 1b). The amino acid sequences of MpDRP3s and MpDRP3l are highly similar to their Arabidopsis and Physcomitrella homologues (Fig. S1a). MpDRP3s had 62.4% identity to AtDRP3A and 79.7% identity to PpDRP3. The alternative splice site is not conserved in Arabidopsis and Physcomitrella (Fig. S1a, magenta). In addition, all five genes have predicted GTPase, dynamin middle and GTPase effector domains (Fig. 1c). Similarly, MpFIS1 (Fig. 1d) and MpELM1 (Fig. 1f ) had high similarities in length, amino-acid sequences ( Fig. S1b and c) and domain structures ( Fig. 1e and g) to Arabidopsis and Physcomitrella counterparts, respectively.
MpDRP3, MpFIS1 and MpELM1 were localized to mitochondria. To examine the subcellular localization of MpDRP3s and MpDRP3l, we observed a fluorescent marker Citrine-fused proteins (Citrine-MpDRP3s and Citrine-MpDRP3l) with a confocal laser scanning microscope (CLSM) (Fig. 2a). The fusion genes were given an N-terminal Citrine tag and placed under the control of the CaMV 35 S promoter. In transgenic plants expressing these proteins and stained with MitoTracker, a part of the Citrine-MpDRP3s and Citrine-MpDRP3l signals localized to the constriction sites of mitochondria (Fig. 2a, arrowheads). Other signals were mainly localized to the tips of mitochondria, while some remained in the cytosol. To better see the localizations of MpDRP3s or In addition, Marchantia thalli were transformed transiently with MpELM1-Citrine and mitochondrial-targeted RFP (MtRFP) by particle bombardment. Although the mitochondria appeared to be enlarged and aggregated, at least some of the MpELM1-Citrine surrounded the RFP fluorescence of mitochondria (Fig. 2c).

Mpelm1 and Mpdrp3 (but not Mpfis1) showed significant growth defects.
To clarify the role of each gene in mitochondrial fission and plant growth, we made MpFIS1, MpELM1 and MpDRP3 knockout mutants (named Mpfis1, Mpelm1 and Mpdrp3, respectively) by homologous recombination (HR)-mediated gene targeting (Fig. 3a) 38 . The knockouts were confirmed by RT-PCR (Fig. 3b). The wild-type plants grew to an area of about 200 mm 2 in 2-weeks from a gemma (Fig. 3c). Compared with the wild type, Mpfis1 did not have any noticeable developmental or growth defects (Fig. 3c). In contrast, Mpelm1 and Mpdrp3 showed similar growth retardation, i.e., their areas grew to less than 10 mm 2 in 2 weeks from a gemma. In addition, Mpelm1 and Mpdrp3 showed distorted thalli phenotypes (Fig. 3d), and they began to wither before reaching the wild-type size. However, these mutants were able to complete their asexual life cycle, i.e., they generated asexual propagules called gemmae, which develop into adult gametophytes.
Mpelm1 and Mpdrp3 (but not Mpfis1) showed severe morphological phenotypes of mitochondria. We then examined whether knockout of those genes affects mitochondrial morphology. In wild-type and  the wild-type or Mpfis1, most of the cells in Mpelm1 and Mpdrp3 were categorized as "Elongated" (Fig. 4b), indicating that mitochondrial fission was blocked in Mpelm1 and Mpdrp3, but not in Mpfis1.
As with the fluorescence microscopy, transmission electron microscopy showed that most of mitochondria in Mpelm1 and Mpdrp3 were much longer than those in the wild-type and Mpfis1 (Fig. 4c). In addition, enlarged structures of mitochondria included vesicle-like structures (black arrowhead) and vacuolated structures (black arrow). These structures might be similar to the dark areas inside mitochondrial red bodies (shown in the insets of Fig. 4c (yellow arrows)), which were not stained with MitoTracker. Although Mpelm1 and Mpdrp3 had enlarged mitochondria, their cristae were similar to those in the wild-type, and they maintained their electrochemical potential because the MitoTracker that we used (Orange CMTMRos) cannot accumulate in mitochondria that do not maintain a membrane potential. Transient expression of MpDRP3 and MpELM1 in the mutants complemented the morphological phenotypes of mitochondria. Because MpDRP3s and MpDRP3l had similar subcellular localizations, we examined whether both of them were functional. To test this, we examined the ability of the MpDRP3s and MpDRP3l to complement the defective mitochondrial networking phenotype of Marchantia lacking MpDRP3. When Mpdrp3 was transformed transiently with either Citrine-MpDRP3s or Citrine-MpDRP3l by particle bombardment, the elongated mitochondrial morphology reverted to a wild-type-like morphology. Similarly, Mpelm1 cells with MpELM1-Citrine have discrete complemented mitochondria (Fig. 5b).

Discussion
In this work, three of four candidate genes for mitochondrial fission (MpDRP3, MpFIS1 and MpELM1) were found in the genomic database of Marchantia polymorpha (Fig. 1). The finding that these three genes (but not PMD) are conserved in Arabidopsis and Marchantia, indicate that MpDRP3, MpFIS1 and MpELM1 have significant roles in land plants. The Arabidopsis genome has two copies of DRP3-, FIS1-and ELM1-type genes 13 . In contrast, Marchantia has single copies of these genes, which makes it a good model for revealing the mechanism of mitochondrial fission in land plants.
MpDRP3 exists in two alternative splicing isoforms (MpDRP3s and MpDRP3l) that differ by 27 bp in exon 13 (Fig. 1). The mRNA expression levels of MpDRP3s and MpDRP3l in the thallus were similar (Fig. 1b). Furthermore, the mitochondrial morphology in Mpdrp3 was rescued by the expression of either Citrine-MpDRP3s or Citrine-MpDRP3l transiently (Fig. 5a). Together, these findings suggest that MpDRP3s and MpDRP3l are functionally redundant, at least in the observed Marchantia thallus cells.
Because little or no difference was observed in plant growth (Fig. 3c) or mitochondrial morphology (Fig. 4b) between Mpfis1 and the wild-type, MpFIS1 does not appear to have a major role in mitochondrial fission. The mitochondrial localizations of Citrine-MpDRP3s and MpELM1-Citrine in Mpfis1 mutants in Fig. S4a and S4b suggest that MpFIS1, unlike yeast FIS1, is not required for the localization of these mitochondrial fission proteins from the cytosol. Citrine-MpFIS1, which was detected on the mitochondrial surface (Fig. 2b), probably binds by the C-terminal transmembrane domain of MpFIS1, as this domain is crucial for mitochondrial targeting in other eukaryotes 11,23,24 . Furthermore, Citrine-MpFIS1 was localized to ring-shaped bodies (Fig. 2b, left panel, arrowhead), which seems to be peroxisomes (Fig. S3). Although FIS1 also localizes to peroxisomes and is involved in peroxisomal fission in Arabidopsis 30,31,41,42 , the role of MpFIS1 in peroxisomal fission remains unclear.
Moreover, phenotypic analyses of plant growth (Fig. 3c) and mitochondrial morphology (Fig. 4b) suggested that not only MpDRP3 but also MpELM1 play crucial roles in mitochondrial fission. In Arabidopsis, drp3a drp3b double mutants show a severe growth-defect with network-shaped mitochondria 20 , but elm1 mutants do not show growth defects 32 . This is possibly because the latter mutants have residual mitochondrial fission activity, attributed to another ELM1 paralogue (At5g06180) in the Arabidopsis genome 32 . By contrast, the Marchantia genome has a single copy of ELM1, so that there would be no residual mitochondrial fission activity in Mpelm1 mutants.
The phenotypes in Mpelm1 strongly resemble those in Mpdrp3 (Figs 3c and 4a), which raises the possibility that MpELM1 is involved in an MpDRP3-dependent pathway. In other words, MpDRP3 may require MpELM1, in the same way that Dnm1p requires the cytosolic adaptor Mdv1p/Caf4p, which localizes Dnm1p to the mitochondrial fission sites in budding yeast 11,21 (Fig. 6, left). It is possible that MpDRP3 interacts with MpELM1, since DRP3A and DRP3B interact with ELM1 for their localization to mitochondria in Arabidopsis (two-way arrows in Fig. 6) 32 . Further biochemical studies in Marchantia are needed to test this model.
Dynamin-related proteins (green shapes in Fig. 6) are universally conserved in eukaryotes, including yeasts, humans and vascular plants 7,14,43,44 . The present results show that this is also the case in Marchantia. Indeed, MpDRP3 had high identities with DRP3 orthologues from A. thaliana and P. patens (Fig. 1c), suggesting that dynamin-related proteins are also crucial in bryophytes. Although the amino acid sequences of Fis1 homologues (blue bars in Fig. 6) are conserved across diverse eukaryotes and Fis1 plays an integral part for mitochondrial fission in yeast, it remains unclear whether Fis1 has a role in mitochondrial fission in mammals 26 . Considering that mammalian membrane-anchored proteins (Mff and MiD49/51) act as mitochondrial adaptors for Drp1 in place of (or with) Fis1 25-28 , it is possible that M. polymorpha has other mitochondrial fission factor(s) on the surface of mitochondria. Because MpFIS1 does not appear to play a role in mitochondrial fission (Fig. 4b), it probably has some other function. This raises the possibility that Fis1s in other eukaryotes might also have some other common function.
The Marchantia genome does not seem to have a PMD homologue. A search of the databases shows that homologues of PMDs are found only in Brassicaceae, a family of vascular plants, and related plants, suggesting that PMD is a mitochondrial fission factor specific to these plants. On the other hand, ELM1-like proteins are evolutionarily conserved across land plants 32 , but they are not present in the genomes of chlorophytes (Chlamydomonas reinhardtii, Volvox carteri and Chlorella vulgaris) or a rhodophyte (Cyanidioschyzon merolae). The present results show that the amino acid sequence and the function of ELM1 in the vascular plant Arabidopsis are conserved in a bryophyte, indicating that the ancestor of land plants acquired ELM1 as a mitochondrial fission factor.

Experimental procedures
Plant materials and growth conditions. Male, accession Takaragaike-1 (called Tak-  Plasmid constructions. Gateway cloning technology (Life Technologies) was used to construct plasmids. cDNA was amplified by PCR with the following primers (Table S1). The amplified DNA fragments were mixed with an entry vector, pENTR/D-TOPO, and the TOPO reaction were performed according to the manufacturer's instructions (Life Technologies). In addition, the stop codon of pENTR-MpELM1 was deleted by using Site-Directed Mutagenesis to add the fluorescent protein to their C-terminus. The resulting plasmids were mixed and reacted with an LR clonase with destination vectors, p2RGW7 46 , pMpGWB105 47 or pMpGWB106 47 , to construct binary vectors containing the CaMV 35 S promoter, fluorescent genes and a NOS terminator.
RT-PCR analysis. Total RNA was extracted from sections of thallus. RNA isolation was carried out using the RNeasy Plant Mini Kit (QIAGEN), according to manufacturer's instructions. First strand cDNA synthesis was carried out starting from 1 µg of total RNA by using SuperScript III Reverse Transcriptase (Invitrogen) and the oligo(dT) primers for reverse transcription. To detect the splicing variants and transcriptional products, RT-PCR analyses were performed using appropriate primers (Table S1). The expression levels of each gene were normalized to the expression level of the MpEF1.
Putative amino-acid sequence and domain structures. Sequence comparisons were performed by using the CLUSTAL W multiple alignment program and GeneDoc Program between sequences of Marchantia, Arabidopsis and Physcomitrella in NCBI database 48  Agrobacterium-mediated transformation. Transgenic plants were generated with AgarTrap (Agar-utilized Transformation with Pouring Solutions) as described previously 40 . Transformants were selected on plates containing hygromycin B (10 µg ml −1 ) and claforan (100 µg ml −1 ) and cultivated for approximately 1 month to obtain gemmae. Then, a gemma was picked to establish an isogenic line from the respective transgenic lines, as a gemma is developed from a single cell 39 .
Particle bombardment. Plasmids for visualizing mitochondrial fission factors were introduced via particle bombardment, by using a helium-driven particle accelerator (Bio-Rad, PDS/1000), with all basic adjustments set according to the manufacturer's recommendations. The bombardment parameters applied were as follows: bombardment pressure, 7.6 × 10 4 hPa; gold particles of 1.6 µm in diameter; 12 cm between the macrocarrier and plants; decompression vacuum, 915 hPa. After bombardment, plants were incubated for 2-days and observed by using CLSM.
MitoTracker Orange staining. Small sections (2-10 mm 2 ) were cut out of the Marchantia thallus with a sharp razor blade, stained with 0.5 µM MitoTracker Orange CMTMRos (Life Technologies) for about 20 minutes, and washed three times with distilled water.

Microscopic observations and image analysis.
Microscopic observations were performed by using a transmission electron microscope (JEOL, JEM-1010), a stereoscopic microscope (Leica, M125), a fluorescence microscope (Nikon, ECLIPSE Ti) and a confocal laser scanning microscope system (Nikon, C1Si). For the electron microscopic analysis, the pre-fixation was performed with phosphate buffer (50 mM) containing 4% paraformaldehyde and 2% glutaraldehyde. The pre-fixed samples were then post-fixed with 2% osmium tetroxide. The specimens were dehydrated in a graded ethanol series and embedded in Spurr resin (Spurr Low Viscosity Embedding Kit, Polysciences). Ultrathin sections (70 nm) were prepared and stained with 2% uranyl acetate and lead solution.
In fluorescence imaging, a 488-nm Ar/Kr laser was used for the excitation of Citrine. A 561-nm diode laser was also used for RFP, MitoTracker Orange and autofluorescence from chloroplasts. Emission signals were detected using a 515/30-nm filter for Citrine, a 590/70-nm filter for RFP and MitoTracker Orange and a 650-nm filter for autofluorescence. All of the data were processed using Adobe Photoshop CS4 Extended 11.0 (Adobe Systems).

Targeted gene knockout.
To generate the targeting vectors, pJHY-TMp1 was used 38 . The 5′ and 3′ homology arms were amplified with the appropriate primer pairs, and the PCR products were cloned into the AscI and PacI sites of pJHY-TMp1, respectively (Tables S1 and S2). Introduction of the targeting construct into M. polymorpha and selection of knockout lines were performed as previously described 37, 38 . Quantification of mitochondrial morphology. One experimenter stained the mutant and wild-type plants with MitoTracker Orange, and obtained images of 50 epidermal cells with CLSM in each experiment for three replicates. A second experimenter, blinded to the preparation of the images, categorized each cell into four groups (circular, tubular, elongated and elongated + enlarged structures) based on mitochondrial morphologies. All statistical testing was performed with one-way analysis of variance (ANOVA) followed by Tukey's multiple-comparison test. P < 0.05 was considered to be statistically significant; n.s., not significant.