Novel Structure of the N Terminus in Yeast Fis1 Correlates with a Specialized Function in Mitochondrial Fission*

Mitochondrial fission is facilitated by a multiprotein complex assembled at the division site. The required components of the fission machinery in Saccharomyces cerevisiae include Dnm1, Fis1, and Mdv1. In the present study, we determined the protein structure of yeast Fis1 using NMR spectroscopy. Although the six α-helices, as well as their folding, in the yeast Fis1 structure are similar to those of the tetratricopeptide repeat (TPR) domains of the human Fis1 structure, the two structures differ in their N termini. The N-terminal tail of human Fis1 is flexible and unstructured, whereas a major segment of the longer N terminus of yeast Fis1 is fixed to the concave face formed by the six α-helices in the TPR domains. To investigate the role of the fixed N terminus, exogenous Fis1 was expressed in yeast lacking the endogenous protein. Expression of yeast Fis1 protein rescued mitochondrial fission in Δfis1 yeast only when the N-terminal TPR binding segment was left intact. The presence of this segment is also correlated to the recruitment of Mdv1 to mitochondria. The conformation of the N-terminal segment embedded in the TPR pocket indicates an intra-molecular regulation of Fis1 bioactivity. Although the TPR-like helix bundle of Fis1 mediates the interaction with Dnm1 and Mdv1, the N terminus of Fis1 is a prerequisite to recruit Mdv1 to facilitate mitochondrial fission.

Mitochondria are dynamic organelles that change their morphology by fusion and fission. Such processes, apparently counteracting each other, are facilitated by two independent molecular machineries. Mitochondrial outer membrane fusion is regulated by the integral membrane proteins Mfn1 and Mfn2 in the case of mammals (1) or Fzo1 in the case of yeast (2,3). These proteins span the mitochondrial outer membranes twice, exposing an N-terminal GTPase domain and a C-terminal coiled-coil domain to the cytosol. A recent study of mouse Mfn1 revealed that the C-terminal coiled-coil domain points outward from mitochondrial membranes to form homodimers in an antiparallel fashion, which is implicated in tethering of two mitochondria together at a distance of about 10 nm (4). Subsequent inner membrane fusion is mediated by a protein called OPA1 in the case of mammals (5,6) or Mgm1 in yeast (7,8), which localizes to the inner membrane with the GTPase domain facing the intermembrane space.
Mitochondrial fission is mediated by a dynamin-related protein, Drp1, identified as Dnm1 in yeast (9 -11). Drp1 and Dnm1 localize in the cytosol as well as in foci at the dividing site of the mitochondrial outer membrane surface during mitochondrial fission (9,12). In contrast to proteins that are involved in fusion, Drp1 and Dnm1 do not span membrane bilayers.
Drp1-and Dnm1-mediated mitochondrial fission is achieved with other accompanying proteins. Genetic studies of Saccharomyces cerevisiae identified Mdv1 and Fis1 as proteins involved in the Dnm1-dependent fission process (13,14). Mdv1 is a peripheral membrane protein that localizes with Dnm1 at punctate structures along the mitochondrial outer membranes to regulate Dnm1. Fis1 is an integral membrane protein that is uniformly distributed along the mitochondrial outer membranes that is required for immobilization of Dnm1 and Mdv1. A high molecular weight complex of Dnm1, Mdv1, and Fis1 constructs the punctate structures at the dividing site and facilitates constriction and division of mitochondrial membrane (13,14). Mdv1 orthologs have not been identified in nematode, fruit fly, or vertebrates, and it is not known whether another protein takes its place or whether the molecular machinery of mitochondrial fission in the various species is different. In contrast, human and mouse orthologs of Fis1 have been identified, suggesting that the role of Fis1 in mitochondrial fission is conserved among lower and higher eukaryotes. Although increased levels of human Fis1 in mammalian cells may (15)(16)(17)(18) or may not (19) accelerate mitochondrial fission, a reduction in human Fis1 level results in notable extensions in the length of mitochondria (17,18,20), indicating that, as in yeast, Fis1 is required for mitochondrial fission in mammals. However, the regulation of Drp1-and Fis1-mediated mitochondrial fission remains unclear in both yeast and mammals.
Comparing orthologs is one way to help understand the general scheme of protein function as well as their molecular evolution. In previous studies (19,21), it was reported that human Fis1 assembles into a novel tetratricopeptide repeat (TPR) 1 -like helix bundle, and it was suggested, based on similarity to other TPR domain proteins, that its concave surface may provide a means to recruit other proteins such as Drp1. * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The Here we present the three-dimensional protein structure of yeast Fis1 and show that, in contrast to human Fis1, an extended N-terminal domain binds to the concave surface of the TPR motif. We also show that this self-interacting region of the N terminus that is absent in human Fis1 is required for yeast Fis1 bioactivity.

MATERIALS AND METHODS
Recombinant Protein-The protein corresponding to residues 1-138 of yeast Fis1 was prepared for NMR studies. The residues 139 -152, half of the membrane-spanning domain and residues that face the mitochondrial intermembrane space, were excluded to solubilize the recombinant protein. The cDNA of Fis1 was cloned from S. cerevisiae genomic DNA library (American Type Culture Collection). NdeI and XhoI sites were introduced next to the initiation and stop codons of the cDNA, respectively, and inserted into pET-17b (Novagen Inc.). Then, the fragment corresponding to nucleotides 1-413 was excised using NdeI and RsaI. pET21b was treated with XhoI followed by Mung bean nuclease and then with NdeI. The excised DNA and the digested pET21b vector were ligated. The resulting plasmid encodes residues Met 1 -Val 138 of Fis1 linked to the C-terminal hexa histidine tag and thus was termed pET21-yFis1-His 6 . Escherichia coli BL21(DE3) (Novagen Inc.) harboring the plasmid was cultured in Martek-9N or Martek-9CN (Spectra Stable Isotope) to produce uniformly 15 N-labeled and uniformly 15 N-, 13 C-labeled protein, respectively. The recombinant protein was isolated from the cytosol by metal chelate affinity chromatography on a Ni 2ϩ -resin column (Novagen Inc.) using 20 mM Tris-HCl, pH 7.9, 500 mM NaCl, 0 or 100 mM EDTA and then further purified by collecting the flow-through from an ion-exchange chromatography on a mono-Q column (Amersham Biosciences) using 20 mM Tris-HCl, pH 8.9, without NaCl. No detergents were used in any step of the protein purification. All NMR samples contained 0.5-1.0 mM protein in 10 mM Tris acetate, pH 5.5, in 90% H 2 O/10% D 2 O or 100% D 2 O. The recombinant protein of human Fis1 for the study of backbone dynamics was prepared as described previously (19).
NMR Spectroscopy-All NMR spectra were acquired at 32°C on Bruker 600 or 800 MHz NMR spectrometers. The spectra were processed using the NMRPipe (22) and analyzed with PIPP (23). The following experiments were used for assignments of 1 H, 13 (27), and three-dimensional HCCH-TOCSY (28). For stereospecific assignment for methyls of leucines and valines, a 1 H-13 C CT-HSQC experiment (29) was carried out using the protein obtained from a bacterial culture using a minimal medium containing 10% [U-13 C]glucose-90% [U-12 C]glucose (30). Proton homonuclear nuclear Overhauser effects (NOEs) were obtained using three-dimensional 15 N-edited NOESY (26), four-dimensional 15 N/ 13 C-edited NOESY (31), and fourdimensional 13 C/ 13 C-edited NOESY (32) experiments. Residual dipolar couplings for N-H and C ␣ -H ␣ were calculated from the difference in corresponding scalar couplings measured in the presence and absence of Pf1 phage (11 mg/ml in 150 mM NaCl) (33). A two-dimensional IPAP 15 N-1 H HSQC experiment (34) was used to obtain the one-bond N-H scalar couplings. A CT-(H)CA(CO)NH experiment (35) was used to obtain the one-bond C ␣ -H ␣ scalar couplings. Relaxation values ( 15 N T 1 and 15 N T 1 ) for backbone amides were calculated from the peak intensities measured using conventional pulse programs (36). Steady state heteronuclear NOE for the backbone amides was derived from the ratio of peak intensities of experiments performed with and without 1 Hpresaturation using a reported pulse program and corrected to compensate for errors caused by incomplete 1 H magnetization recovery (37). The measurements were repeated twice.
Structure Calculation-Peak intensities from NOESY experiments were translated into a continuous distribution of proton-proton distances. Generic hydrogen bond distance restraints were employed for ␣-helical regions that were determined based on secondary 13 C ␣ chemical shifts and medium range NOE patterns. The TALOS program (38) predicted the backbone dihedral angles (, ) from 13 C ␣ , 13 C ␤ , 13 CЈ, 1 H ␣ , and 15 N H chemical shifts. Statistically significant angles were used as structural restraints with at least 20°margins. Residual dipolar coupling restraints were separated into mobile and rigid regions determined based on the backbone dynamic data. Structures of yeast Fis1 were calculated by a distance geometry and simulated annealing protocol (39) with the incorporation of dipolar coupling restraints (40) using the program XPLOR-NIH (41). Structure calculations employed 2235 inter-residue and 803 intra-residue proton-proton distance restraints, 124 hydrogen bond distance restraints, 92 and 92 angle restraints, and 90 N-H and 41 C ␣ -H ␣ dipolar couplings.
Fis1 Function in Yeast-Yeast vector pH62 was kindly provided by Dr. Reed B. Wickner (NIDDK, National Institutes of Health). pH62 is a derivative of pRS315 and contains CEN replicon, LEU2 marker, and ADH1 promoter. The DNA fragment containing Fis1 cDNA was excised from the pET17-yFis1 plasmid using XbaI and XhoI and then inserted into pH62 at XbaI and XhoI sites (pH62-FIS1). Plasmids for deletion mutants were constructed in the same way. The deletion mutant DM1 lacks the N-terminal 14 residues, DM2 lacks the N-terminal 5 residues, and DM3 lacks Phe 6 -Tyr 14 (see Fig. 7A). All yeast strains are derived from strain yAN001 (met15 leu2 his3 ura3). Cells were transformed using the lithium acetate method (42) and were grown in synthetic glucose medium missing the appropriate nutrients to select for the plasmids. Mitochondria were visualized using a mitochondria-targeted GFP (pVT100U-mitoGFP; kindly provided by Dr. Benedikt Westermann, Universitä t Bayreuth, Germany). Exponentially growing cells of strains yAN008, yAN009, yAN010, yAN011, yAN012, and yAN013 (Table I) were embedded in 0.2% agarose, and confocal pictures were taken using a LSM 510 Meta (Zeiss). Mitochondria were classified in two different phenotypes. More than 200 cells were counted to obtain the percentage of the phenotypes.
Detection of Exogenous Fis1 Expression in ⌬fis1 Yeast-Cells of strains yAN003, yAN004, yAN005, yAN006, and yAN007 (Table I) were grown overnight in synthetic medium (SD-Leu) to logarithmic phase and harvested by centrifugation. After washing once with ice-cold water, the cells were resuspended in ice-cold lysis buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 50 mM NaF, 5 mM EDTA, 0.1% IGEPAL CA-630, 1% Triton X-100) supplemented with protein inhibitor mixture (Sigma) and mixed with acid-washed glass beads (Sigma). The cells were broken by shaking in a mixer mill (Retsch) for 5 min, and then debris was removed by 3 min of centrifugation at 4°C. Protein lysates were analyzed by Western blot using rabbit anti-Fis1 antibody (kindly provided by Dr. Jodi Nunnari, University of California, Davis), mouse anti-porin antibody (Molecular Probes), and horseradish peroxidase-coupled secondary antibodies (Amersham Biosciences).
Subcellular Localization of Mdv1-The ⌬fis1 yeast harboring pH62-FIS1 constructs was transformed with pGALL-GFP-MDV1 (kindly provided by Dr. Jodi Nunnari, University of California, Davis). The strains yAN015, yAN016, yAN017, yAN018, and yAN019 (Table I) were grown overnight in synthetic medium (SRaf-Leu-URA) containing 2% raffinose to allow galactose induction and lacking leucine and uracil to select for plasmids. Production of GFP-Mdv1 was induced by treatment with 2% galactose for 2 h. Cells were fixed and analyzed by confocal microscopy. Location of GFP-Mdv1 was classified in two categories: diffuse in cytosol or localized to mitochondria. More than 200 cells were analyzed to score the localization of GFP-Mdv1.

RESULTS
Structure Description of Yeast Fis1-The NMR derived structures of yeast Fis1 are presented in Fig. 1. The core domain of Fis1 consists of six ␣-helices. The six ␣-helices are determined based on a combination of NMR data (Fig. 2) ␣6 for Lys 111 -Glu 127 . The first 6 residues of the N-terminal tail do not adopt an ordered conformation. Their 13 C ␣ and 13 C ␤ chemical shifts are close to their random coil values, and no 1 H-1 H NOE of midrange or long range can be observed. The region corresponding to residues Leu 10 -Tyr 14 shows characteristics of an ␣-helix, such as positive 13 C ␣ secondary chemical shifts and NOEs between H ␣ (i) and H N (i ϩ 3) and between H ␣ (i) and H ␤ (i ϩ 3). However, as NOEs between H ␣ (i) and H N (i ϩ 4) are absent, this region is not defined to be an ␣-helix. The C-terminal 17 residues of the recombinant protein, which include the hydrophobic region and the artificial His tag, show a disordered conformation in solution. Their secondary chemical shifts for 13 C ␣ and 13 C ␤ are close to zero, and no long range 1 H-1 H NOE are observed. As seen in the ensemble of the 20 lowest energy structures (Fig. 1A), the region corresponding to residues Thr 9 -Glu 127 converged into a well defined conformation. The atomic root mean square deviation about the mean of coordinates for six ␣-helices of the 20 structures was 0.4 Ϯ 0.1 Å for backbone heavy atoms and 1.0 Ϯ 0.1 Å for all heavy atoms.
A data base search performed using the DALI program for structural similarities (43) revealed, at the top of the list, 1IYG, the structure of mouse Fis1 as a representative structure of Fis1 orthologs. Currently, there are three PDB depositions for Fis1 orthologs: 1PC2 is the solution structure of human Fis1 (19), 1NZN is the crystal structure of human Fis1 (21), and 1IYG is the solution structure of mouse Fis1. The core domain of human, mouse, and yeast Fis1 contains six ␣-helices with the same folds. Due to the high degree of structural similarity between yeast and mouse Fis1, the rest of the list from the DALI search is the same as the list that we obtained previously using the human Fis1 structure, 1PC2, as a query (19). Namely, the overall fold of the six ␣-helices in yeast Fis1 is similar to the fold of helices composed of the TPR motif, although, like other Fis1 sequences, no significant sequence similarity to the typical TPR motif is found within the TPR-like core domain of yeast Fis1. The typical TPR motif contains degenerate 34-amino-acid sequences with 8 loosely conserved consensus residues (X 3 (WLF)X 2 (LIM)(GAS)X 2 (YLF)X 8 (ASE)X-3 (FYL)X 2 (ASL)X 4 (PKE)X 2 ) and usually presents in a tandem array of multiple copies (44). The TPR motif is found in a number of functionally different proteins. The TPR-containing domains facilitate specific protein-protein interactions at the concave surfaces, although the common features of the interaction partners have not been defined. The structural analogy of yeast Fis1 to the typical TPR-containing proteins, although the protein sequences are discrete, suggests that yeast Fis1 may bind to other proteins at its concave protein surface.
Structural Comparison between Yeast and Human Fis1-To evaluate the structural conservation between yeast and human Fis1, we compared these two structures (Fig. 3). The core domain of both yeast and human Fis1 consists of six ␣-helices. The length of the ␣1-helix in yeast Fis1 is shorter than that in human Fis1 (Fig. 3A). For the other five ␣-helices, the lengths of corresponding helices in yeast and human Fis1 are the same. Moreover, the folds of six ␣-helices are similar (Fig. 3, B and C), as indicated by the data base search described in the previous section. The pairwise root mean square difference, calculated using the backbone atoms (N, C ␣ , CЈ, and O) of the corresponding six helices for yeast and human Fis1, is 1.8 Å. Among the six ␣-helices from the two structures, the ␣1-helix shows a major difference in its location; it is shifted along its axis and results in the largest deviation in the pairwise comparison. We previously found that human Fis1 contains two TPR-like motifs, where the ␣2-loop-␣3 hairpin forms one TPR-like motif and the ␣4-loop-␣5 hairpin forms the other. The two loops in the yeast Fis1 structure (one between ␣2 and ␣3 and the other between ␣4 and ␣5) show the same conformation as those in the human Fis1 structure. The pairwise comparison of overall folds between yeast and human Fis1 using DALI results in the Z-score of 12.
A major difference between the yeast and human Fis1 structures is the N terminus prior to the ␣1-helix. The N-terminal tail of yeast Fis1 is located at the concave side of the helix bundle (Fig. 1), whereas that of human Fis1 is not. A difference can also be seen in the backbone dynamics of the region (Fig. 4). In the human Fis1 structure, the entire N-terminal tail is flexible (Fig. 4A). In contrast, the N terminus of the yeast Fis1, which is longer than that of human Fis1 by 8 residues (Fig. 3A), is made of two parts with different characteristics (Fig. 4B). The first portion from the N-terminal end (Met 1 -Phe 6 ) is flexible, and the second portion (Trp 7 -Tyr 18 ) is rigid. It is not clear where the boundary between the flexible and rigid regions is. The backbone dynamics of Trp 7 cannot be measured because the cross-peak for the amide 15 N and 1 H chemical shifts of Trp 7 overlaps with those of Val 133 and Val 134 . However, the aromatic side chain of Trp 7 appears to be stabilized by hydrophobic interaction as 1 H-1 H NOEs to Tyr 81 within ␣4, Leu 103 within ␣5, and Val 113 within ␣6 are observed. The rigid portion (residues Trp 7 -Tyr 18 ) is bound to the concave side that is composed of the six helices. Hydrophobic interactions between Trp 7 , Pro 8 , Leu 10 , Ala 13 within this loop and hydrophobic residues at the concave surface of the TPR-like domain stabilize the loop (Fig. 5). In addition, two negatively charged residues, Asp 12 and Glu 15 , are proximate of positively charged Lys 51 (Fig. 5), indicating that electrostatic interactions assist in the stabilization of the loop. Lack of these interactions in human Fis1 relegates its N-terminal tail out of the concave surface of the TPR-like domain (Fig. 6A).
The rigid region of the N terminus of yeast Fis1 (Fig. 6B) is located in the hydrophobic pocket, similar to that of the N terminus of the mouse Fis1 structure (1IYG) (Fig. 6C). However, in the construction of the recombinant mouse Fis1 protein, a cloning artifact resulted in the extraneous amino acids, Gly Ϫ7 -Ser Ϫ6 -Ser Ϫ5 -Gly Ϫ4 -Ser Ϫ3 -Ser Ϫ2 -Gly Ϫ1 , becoming attached to the N-terminal end. The native sequence of mouse Fis1 is identical to that of human Fis1 except for only five residues within ␣1 and ␣3 helices, and therefore, the N-terminal loop of mouse Fis1 that binds into the pocket is not physiological due to the 7-residue extension, and the native mouse Fis1 N terminus would be expected to be flexible as that in the human Fis1 structure (19). Nevertheless, the structure of the N-terminal extended recombinant mouse Fis1 is similar to that of the yeast Fis1 structure, corroborating the clear potential for TPR pocket binding of N-terminal extensions as in yeast Fis1. The first 6 residues (Gly Ϫ7 -Ser Ϫ2 ) of recombinant mouse Fis1 do not converge into one conformation, as shown in the 20 ensemble structures, and appears to be flexible, whereas the FIG. 2. Data obtained using NMR spectroscopy to establish the secondary structure of yeast Fis1. Sequential and medium range NOE connectivities characteristic of ␣-helices as well as 13 C ␣ and 13 C ␤ secondary shifts are presented along with the protein sequence. The height of bars for the NOE connectivities presentation indicates the NOE intensity. The absence of typical NOEs within the ␣-helical regions mostly means that data were not included due to the difficulty of picking overlapped peaks. The height of the bars for the secondary shifts presentation are the difference from the random coil values. The location of the ␣-helices defined from the NMR data is indicated at the bottom. rest of the N terminus prior to the ␣1 helix appears to be rigid and positions in the hydrophobic pocket made by six ␣-helices. The residues from 9 to 14 in the recombinant mouse protein (corresponding to Glu 2 -Asn 6 in the native sequence) form a helix, similar to the region corresponding to the residues Leu 10 -Tyr 14 in yeast Fis1. This ironically explains that the N-terminal behavior of yeast Fis1 in solution is physically appropriate. The hydrophobic surface composed by six ␣-helices is a general binding pocket in the yeast, mouse, and human orthologs. Specific binding can be achieved by electrostatic  interactions and steric fit, in addition to hydrophobic interactions. To date, the identity of proteins predicted to bind to this region in human and mouse Fis1 is not known, although the current work shows that yeast Fis1 entails at least a selfassociation with its extended N terminus.
The biological Role of the N Terminus-The finding of the N terminus embedded in the binding pocket led us to investigate the importance of the Fis1 N terminus in mitochondrial fission. We constructed plasmids carrying FIS1 (Fig. 7A) and introduced them into a S. cerevisiae strain that lacks the fis1 gene. GFP localized to the mitochondrial matrix by import sequence of cytochrome c oxidase subunit IV was used to observe the mitochondrial morphology in the transformed yeast (Fig. 7B). The ⌬fis1 yeast showed net-like or condensed mitochondria, indicating that the fission machinery is not working properly as initially reported (13,14). Yeast harboring the empty plasmid, pH62, showed the same aberrant mitochondrial morphology. Ectopic expression of wild-type Fis1 in the ⌬fis1 yeast resulted in fragmentation of mitochondria, consistent with the conclusion that it properly rescues the fission machinery. Upon expression of the Fis1 deletion mutant DM1, which lacks the majority of the Nterminal loop (the first 14 residues) in the ⌬fis1 yeast, the mitochondria appeared mostly fused and condensed, indicating that mitochondrial fission could not be rescued in the absence of the first 14-residue segment. Two other N-terminal deletion mutants were used to explore the role of N-terminal binding to the TPR domain. Expression of the deletion mutant DM2, which lacks the flexible portion of the N-terminal loop (the first 5 residues), reconstituted mitochondrial fission, whereas the deletion mutant DM3, which lacks Phe 6 -Tyr 14 corresponding to most of the rigid region of the N-terminal loop, was inert. To rule out the possibility that the Fis1 deletion mutants DM1 and DM3 are unstable or not expressed, resulting in the failure to reconstitute mitochondrial fission, we confirmed that all of the deletion mutants were present at levels comparable with that of the wild-type protein (Fig. 7C). The loss of function for the deletion mutants DM1 and DM3 is due to the lack of the N-terminal rigid region.
In the multistep process of mitochondrial fission, Fis1 is involved in mitochondrial localization of Mdv1 (13,14,45). To investigate whether the N-terminal rigid region of the yeast Fis1 protein is required for binding to Mdv1, we tested the recruitment of GFP-Mdv1 to mitochondria upon expression of the different Fis1 constructs (Fig. 8). GFP-Mdv1 showed a cytoplasmic localization in ⌬fis1 yeast, and expression of either Fis1 or Fis1-DM2 led to the recruitment of Mdv1 to the mitochondria. However, neither Fis1-DM1 nor Fis1-DM3, both lacking the N-terminal rigid region, was able to relocalize the cytosolic Mdv1 to the mitochondria. Upon co-expression with Fis1 or Fis1-DM2, GFP-Mdv1, through its uniform distribution on the mitochondria, revealed net-like mitochondria, typical of the ⌬fis1 phenotype (Fig. 8). In contrast to endogenous Mdv1 that produced fragmented mitochondria (Fig. 7B), overexpression of GFP-Mdv1 inhibited mitochondrial fission, as described previously in the literature (46). GFP-Mdv1 is evenly dispersed along the mitochondria, instead of forming punctate structures at discrete foci on the mitochondria, apparently due to the FIG. 6. N termini of the recombinant proteins. A, the N terminus of the recombinant human Fis1 (PDB accession number ϭ 1PC2) has a native sequence, which shows a disordered conformation. B, the N terminus of the recombinant yeast Fis1 has a native sequence, a part of which is located in the pocket made by the six-helix bundle. C, the N terminus of the recombinant mouse Fis1 (PDB accession number ϭ 1IYG) contains the additional sequence (indicated as in yellow: Gly Ϫ7 -Ser Ϫ6 -Ser Ϫ5 -Gly Ϫ4 -Ser Ϫ3 -Ser Ϫ2 -Gly Ϫ1 ) fused to the native sequence. Note that the wild-type sequence of the N terminus of mouse Fis1 is identical to that of human Fis1. A part of the wild-type sequence constructs a short helix that is located in the pocket made by the six-helix bundle. These data indicate that the N-terminal loop, and specifically the region that binds into the concave face of the TPR domain, is required for yeast Fis1 bioactivity. We propose that the TPR-like helix bundle formulates a module for proteinprotein interaction and that the N-terminal loop fitting into the concave surface is a prerequisite for appropriate interactions through the TPR-like helix bundle. DISCUSSION Among the many Fis1 orthologs, human and yeast Fis1 are the best characterized. Fis1 exists at the mitochondrial outer membrane and is thought to recruit other proteins that are involved in mitochondrial fission, such as Dnm1 and Mdv1 in the case of yeast and Drp1 in the case of mammals. Fis1 possesses a transmembrane region at its C terminus that anchors it into the mitochondrial outer membrane. The N-terminal core domain is exposed to the cytosol and is proposed to act as a binding module. We previously determined the structure of human Fis1 and characterized the potential TPR-mediated molecular mechanisms by which Fis1 interacts with its binding partners (19). However, it is still not clear how the interaction is regulated. In the present study, we determined the structure of yeast Fis1 and compared it with the human counterpart. It is helpful to consider the differences and similarities between yeast and human Fis1 to investigate the general scheme of their molecular function and their specific functions in the different species.
Examination of the sequence alignment (Fig. 3A) between human and yeast Fis1 reveals that the region corresponding to residues Met 1 -Gln 40 of yeast Fis1 is the least similar to that of the human ortholog. Nevertheless, yeast Fis1 forms an ␣-helix (residues Pro 19 -Ser 31 ) that is unexpectedly similar to the ␣1helix of human Fis1. The rest of the human and yeast Fis1 is very similar in both sequence as well as structure. Based on the analysis using DALI against the data base of protein structures, overall folds of the two proteins are statistically identical. Both human and yeast Fis1 contain a TPR-like helix bundle that forms a hydrophobic pocket on the concave side. A helix FIG. 7. Mitochondrial morphology of ⌬fis1 yeast with or without ectopic production of Fis1 protein. A, a schematic representation of the Fis1 constructs. In addition to wild-type FIS1, three different deletion mutants, DM1, DM2, and DM3, were designed. The N termini of proteins to be expressed using the pH62 vector are shown. The gray regions correspond to the ␣1 helix in the determined structure. The numbers correspond to the residue numbers of the wildtype protein. In B, bars indicate the ratio of two different phenotypes of mitochondria. Mitochondria of more than 200 cells were counted for each strain and classified as either fragmented (black bar) or net-like (open bar) phenotype. The percentages of fragmented mitochondria are listed. A representative cell for each strain is shown below. Mitochondria were visualized using pVT100U-mitoGFP (upper panels), and the corresponding cells are shown in the bright field images (lower panels). Strains used here are yAN008, yAN009, yAN010, yAN011, yAN012, and yAN013. C, ectopic Fis1 proteins in ⌬fis1 yeast were analyzed by Western blot. The outer mitochondrial membrane protein porin served as a loading control. Strains used here are yAN003, yAN004, yAN005, yAN006, and yAN007.
FIG. 8. Mitochondrial recruitment of GFP-Mdv1 by FIS1 expression in ⌬fis1 yeast. Subcellular localization of GFP-Mdv1 was analyzed by confocal microscopy. More than 200 cells were counted for each strain to determine the percentage of mitochondrially localized GFP-Mdv1 (black bar). GFP-Mdv1 (upper panels) and a corresponding bright field image (lower panels) for a representative cell for each strain are shown below. Strains used here are yAN015, yAN016, yAN017, yAN018, and yAN019. bundle from a typical TPR motif constructs a hydrophobic pocket that mediates protein-protein interactions (47,48), and due to the structural similarity between Fis1 and other TPRcontaining proteins, we propose that Fis1 mediates proteinprotein interactions at this concave surface. It has been reported that a proline substitution within the TPR-like domain of yeast Fis1 results in an inhibition of mitochondrial fission, due to the lack of binding between Fis1 and Mdv1 (45). Proline is known to be unfavorable for the formation of ␣-helices. It is probable that the proline mutation at Leu 80 within the ␣4 helix of yeast Fis1 prevents the formation of a proper ␣-helix, thus inhibiting the functional TPR-like structure. Interestingly, many of the hydrophobic residues in yeast and human Fis1 are conserved, suggesting that they may be critical for the stabilization and/or the assembly of the helices into a TPR-like helix bundle.
In contrast to the highly conserved structure of the TPR-like helix bundle, the N-terminal tail shows major structural differences between human and yeast Fis1. The sequence alignment (Fig. 3A) and the details on interactions between the N-terminal tail and the concave surface formed by the TPR-like domain in yeast Fis1 (Fig. 5) provide reasonable explanations for why the N-terminal tail of human Fis1 does not adopt a fixed conformation. The residues involved in hydrophobic interactions, Trp 7 , Pro 8 , and Leu 10 in yeast Fis1, are missing in human Fis1. The residue involved in electrostatic interaction, Asp 12 in yeast Fis1, is also missing in human Fis1. The length of N-terminal tail of human Fis1 is so short that the hydrophobic residues Val 3 and Leu 4 cannot reach deep enough into the hydrophobic pocket to establish stable hydrophobic interactions. Given that the N terminus of yeast Fis1 regulates protein function (Fig. 7), the distinct N termini of these proteins suggest a different regulation of human and yeast Fis1.
Comparison of the human and yeast Fis1 sequences with those from Arabidopsis, Drosophila, Caenorhabditis elegans, and pufferfish (Fig. 9) shows that the length and sequence of the N terminus differs markedly and more than any other part of the molecule. The truncated N terminus of Drosophila Fis1 would lack both helices ␣1 and ␣2 in the yeast structure, and Arabidopsis has an 11-residue N-terminal extension beyond that of even the long S. cerevisiae N terminus. The residues in the N terminus of yeast Fis1 that have contact with the TPR bundle are not well conserved in other sequences. The fact that the artificial extension of mouse Fis1 forms a similar conformation and interacts with residues in the homologous pocket in mouse Fis1 indicates that the binding specificity of the concave pocket in the TPR domain for N-terminal extensions is not highly specific.
Mdv1, a protein that binds to Fis1 and is required for mitochondrial fission in yeast, is not found in mammals. Interest-ingly, we find that the N-terminal region of yeast Fis1 protein that docks to the TPR-like domains and is absent in mammalian Fis1 is required for appropriate interactions between Mdv1 and Fis1 in yeast (Fig. 8). The Fis1 function in binding to Mdv1 is directly related to its function in mitochondrial fission. Deletion of the rigid Phe 6 -Tyr 14 region of yeast Fis1, essential for recruiting Mdv1, abolished its bioactivity. This explains why human Fis1 cannot reconstitute the defect of mitochondrial fission in ⌬fis1 yeast (17), consistent with our finding that the unique N-terminal region of yeast Fis1 is required for mitochondrial fission. The required portion is embedded in the pocket formed by the TPR-like domain, suggesting that this intra-molecular association is a regulatory mechanism for the function of Fis1 in yeast. Since this pocket in other TPR domain proteins is involved in protein-protein interactions, the occupation of the yeast Fis1 hydrophobic pocket by its N-terminal residues may provide a way to compete for and thus regulate interaction of Fis1 between different proteins.