Crystal structure and mutation analysis revealed that DREP2 CIDE forms a filament-like structure with features differing from those of DREP4 CIDE

Cell death-inducing DFF45-like effect (CIDE) domain-containing proteins, DFF40, DFF45, CIDE-A, CIDE-B, and FSP27, play important roles in apoptotic DNA fragmentation and lipid homeostasis. The function of DFF40/45 in apoptotic DNA fragmentation is mediated by CIDE domain filament formation. Although our recent structural study of DREP4 CIDE revealed the first filament-like structure of the CIDE domain and its functional importance, the filament structure of DREP2 CIDE is unclear because this structure was not helical in the asymmetric unit. In this study, we present the crystal structure and mutagenesis analysis of the DREP2 CIDE mutant, which confirmed that DREP2 CIDE also forms a filament-like structure with features differing from those of DREP4 CIDE.

Scientific RepoRts | (2018) 8:17810 | DOI: 10.1038/s41598-018-36253-y proteins assemble into highly oligomeric forms in solution. Based on the structural study of DREP4 CIDE, which showed a helical filament-like structure even in the crystallographic asymmetric unit, we found that the CIDE domains of both DREP4 and DREP2 form filament-like structures in solution. In this study, the details of the helical assembly of the CIDE domain were determined and the filament-like helical oligomeric complex of DREP2 CIDE was confirmed by structural analysis and mutagenesis studies. Additionally, the differences in the characteristics of filament-like structures between DREP2 and DREP4 were determined.

CIDE domain of DREP2 forms a highly oligomeric state in solution.
Apoptotic DNA fragmentation, which is mainly mediated by the DFF40/DFF45 heterocomplex, is the hallmark of apoptotic cell death and conserved in fly. Four CIDE domain-containing proteins have been identified in fly: DREP1, DREP2, DREP3, and DREP4. Unlike DFF45, which contains only conserved acidic residues, the DREP2, DREP4, DFF40, and FSP27 CIDE domains contain two patches, acidic and basic, in one CIDE domain (Fig. 1A). Although the pattern of charge distribution of each CIDE domain is similar in that they contain two oppositely charged patches, their behaviours differed in solution by forming various oligomeric states based on the size-exclusion chromatography results (Fig. 1B). The DREP4 CIDE domain showed the greatest oligomerization in solution (Fig. 1B). In size-exclusion chromatography, the DREP2 CIDE domain was eluted at approximately 11-14 mL, corresponding to a molecular weight of 100-300 kDa, indicating formation of a highly oligomeric complex (Fig. 1B). We also performed native-PAGE with purified DREP2, DREP4, DFF40, and FSP27 proteins to check the behaviour of the CIDE domain in native state. As shown by the Fig. 1C, we found that DREP2 and DFF40 were stuck on the loading wall and did not migrate well on the native-gel, which might be due to large size of the homo-oligomeric structure. DREP4 migrated a little on the gel. This is also due to the large size of the DREP4 CIDE as indicated by size-exclusion chromatography. Interestingly, however, majority of the FSP27 CIDE, which was eluted as 18 mL on the size-exclusion chromatography, did not migrated well on the gel, although some low molecular weight Scientific RepoRts | (2018) 8:17810 | DOI:10.1038/s41598-018-36253-y particle was shown on the gel. This might be because FSP27 also formed higher oligomeric filament structure during the concentration process. Oligomerization of the CIDE domain of DREP4 was highly dependent on the concentration of salt 30 . To analyse the salt dependency of complex formation of DREP2, we conducted multi-angle light scattering (MALS), which reveals the absolute molecular weight of a particle. The molecular weight of DREP2 CIDE in 50 mM NaCl and 1 M NaCl were calculated by MALS. The theoretical molecular weight of DREP2 CIDE was 10.99 kDa, while the experimental weights determined by MALS were 281.6 kDa (2.04% fitting error) in 50 mM NaCl and 215.3 kDa (0.77% fitting error) in 1 M NaCl, indicating that complex formation of DREP2 CIDE is not dependent on the salt concentration (Fig. 1D,E). Oligomerization of the CIDE domain was expected, as several studies showed that CIDE domain-containing proteins formed highly oligomeric complexes via this domain 31 .
Structure of DREP2 CIDE. The 2.3 Å high resolution crystal structure of the DREP2 CIDE domain was solved using the molecular replacement (MR) method followed by refining to an R work of 22.0% and R free of 25.6%. The structure of the DREP2 CIDE domain exhibited an atypical CIDE domain fold, which is composed of an α/β roll fold with two α helices and five β strands, containing the α2 helix, but not β4 strand (Fig. 1F). This case is similar with the structure of ICAD CIDE and CIDE-A. The two helices comprised of residues 30-41 and 64-68 are indicated as α1 and α2, respectively. The four strands comprised of residues 10-15, 22-27, 50-53, and 74-78 are indicated as β1, β2, β3, and β5, respectively (Fig. 1F). Unlike DREP4 CIDE, in which one turn of the helical complex is formed by 10 molecules of DREP4 CIDE in the crystallographic asymmetric unit, there were four molecules in the asymmetric unit of DREP2 CIDE, which were referred to as chains A-D (Fig. 1G). A model of chains A and D was built from residues 7-84, while that of chains B and C was built from residues 8-84. There were extra residues of Leu, Glu, and His at the C-terminus, which were part of the vector construct. Based on the Ramachandran plot, 95% of the residues were in the most favourable region, whereas 5% were in the allowed regions. The collection and processing of data and refinement statistics are summarized in Table 1. Interestingly, there was no apparent symmetry between the four chains (Fig. 1G). Each monomer was nearly identical, as indicated by superimposition with a root mean square deviation (R.M.S.D.) of approximately 0.7-0.9 Å (Fig. 1H).
Four interfaces are formed in the oligomeric CIDE domain of DREP2. The four CIDE domains in the asymmetric unit are arranged to form a square-shaped complex through non-symmetric interactions between molecules, resulting in mediation of the tetrameric arrangement by four unique interfaces involving different parts of the surface and residues of the DREP2 CIDE domain. The interface between chains C and D (hereafter, interface C) showed the most extensive interactions burying the 500-Å 2 accessible surface area. Inspection of

Data collection
Wild-type R36E interface C revealed that the main interactions are mediated by polar residues involving K9, K13, W15, R22, K23, and N72 from chain D and D56, T58, Q59, E61, E64, and Y65 from chain C. These residues form three salts bridges and four hydrogen bonds, as well as contribute to van der Waals interactions ( Fig. 2A). The second largest interface, which is formed by chains A and B (hereafter, interface A), buries the 495-Å 2 accessible surface area and is mediated by the same residues observed in interface C, except for residues K9 (chain D) and E64 (chain C) (Fig. 2B). Interestingly, these two interfaces (A and C) are highly similar to those observed in the homo-dimeric and hetero-dimeric CIDE domain complex 23,26,28,29 . These two CIDE homo-dimers further assembled into a tetrameric complex by forming two unique interfaces between chains B and C (hereafter, interface B) and between chains A and C (hereafter, interface D). Interface B buries the 368-Å 2 accessible surface area, and the interaction is mediated by residues E31, T29, and G61 from chain B and E31, K38, and A45 from chain C (Fig. 2C). The last interface, D, buries the 134-Å 2 accessible surface area via an interaction mediated by residues E61, D62, and R67 from chain C and residues E36, D39, and K40 from chain A (Fig. 2D). In the tetrameric assembly, chain C forms three unique interfaces with chains A, B, and D involving a surface area of 1002 Å 2 which accounts for 20% of the total surface area of 4951 Å 2 of chain C, while chain D only forms an interface with chain C. Therefore, the chain C molecule is important in the tetrameric arrangement of the DREP2 CIDE domain in the asymmetric unit.
Putative oligomeric structure of CIDE domain of DREP2. The DREP2 CIDE domain exists as a highly oligomeric complex in solution, containing 16-18 molecules in the complex as determined by size-exclusion chromatography and MALS. Because the current tetrameric structure of the DREP2 CIDE domain is smaller than the calculated size, we examined the entire molecular packing of the CIDE domains in the crystal (Fig. 3A). Interestingly, each tetrameric complex of the DREP2 CIDE domain, which was detected in the asymmetric unit, was linked to the next tetrameric unit using interfaces previously detected in the hetero-dimeric and homo-dimeric complexes of CIDE domains ( Fig. 3B) 26,29 . The solution structure of the hetero-dimeric complex between DFF40 CIDE and DFF45 CIDE showed that the interaction is mediated by a basic patch (K9, K18, K32, and R36) on DFF40 CIDE and acidic patch (D66, D71, D72, and D74) on DFF45 CIDE 27 (Fig. 3C). Another structural study of the homo-dimer of the FSP27 CIDE domain revealed that homo-dimerization of the CIDE domain is mediated by basic patch formed by R46, R55, and K56 of one FSP27 CIDE and acidic patch formed by E87, D88, and E93 of the second CIDE (Fig. 3C). This interaction strategy may be the same as that used by neighbouring molecules in the structure of DREP2 CIDE. Superposition of the structures showed that the hetero-dimeric   (Fig. 3D). These findings indicate that the orientation of the interface formed by crystallographic packing was similar to that formed by the hetero-dimeric complex and homo-dimeric complex of the CIDE domains. This structural analysis and the newly solved filament-like helical structure of DREP4 CIDE suggest that DREP2 CIDE forms a similar filament-like helical structure in solution. Therefore, we modelled the helical complex of DREP2 by identifying the symmetrical molecules responsible for forming the helical complex in the crystal lattice (Fig. 3A,E). Our previous structural study of the filament-like CIDE domain showed that 10 molecules of DREP4 CIDE form one turn of the helical assembly in the crystallographic asymmetric unit 30 . In the crystal lattice, the helical structure is continuous and stacks along the a-axis of the unit cell with a 56.5 Å rise/turn and ~105 Å diameter 30 . The modelled helical structure of DREP2 based on the crystal contains four molecules in the crystallographic asymmetric unit arranged into a filament assembly with eight subunits per turn, rise of 50.3 Å, and diameter of ∼90 Å (Fig. 3E).

Confirmation of helical complex of DREP2 by mutagenesis.
To determine the importance of the interface formed by crystallographic packing in the oligomeric complex followed by formation of further filament-like structures, we conducted a mutagenesis study. Based on interface analysis of the tetrameric DREP2 and previous analysis of the interface formed by the hetero-dimeric DFF40/DFF45 CIDE complex showing that the main forces generated in the interface were salt bridges formed by K9, K18, K32, and R36 on DFF40 and D66, D71, D72, and D74 on DFF45, we introduced a mutation at K13 on the DREP2 CIDE domain to D (hereafter, K13D), which is aligned with K9 in DFF40, to disrupt the interfaces A and B in the tetrameric DREP2 structure. We also introduced a mutation at R36 in the DREP2 CIDE domain to E (hereafter, R36E), which is critical for the formation of interface D in the tetrameric DREP2 structure (Fig. 2D). While wild-type DREP2 CIDE and R36E mutants formed a highly oligomeric homo-complex in solution that was eluted from the gel-filtration column at approximately 11 mL, the K13D mutant produced a monomeric peak in solution that was eluted from the gel-filtration column at approximately 18 mL, indicating that only K13D (disrupting interfaces A and B) disrupted complex formation (Fig. 4A). The molecular weight of the disrupted K13D mutant was confirmed by MALS. The theoretical molecular weight of the K13D mutant was 10.85 kDa and the experimental molecular weight determined by MALS was 12.14 kDa (8.59% fitting error), with a polydispersity of 1.000 (Fig. 4B). MALS showed that R36E still formed a ~217.4 kDa complex, with a similar size as the wild-type (Fig. 4C). To exclude the possibility that dissociation of the self-complex resulted from structural distortion caused by mutations, we conducted far UV circular dichroism (CD) analysis. As shown in Fig. 4D, the wild-type and two mutants showed similar CD spectrum patterns, with two pronounced minima at 208 and 222 nm and a maximum at 215 nm. These findings indicate that the mutations did not affect the DREP2 CIDE domain structure. Accordingly, the real form of DREP2 CIDE in solution is a helical filament-like complex rather than a tetrameric complex which is packed and detected in the asymmetric unit. The R36E mutant was crystallized under similar conditions as those used to produce the wild-type crystal. The structure of R36E was solved and refined to an R work of 18.0% and R free of 23.4%. Four molecules in the asymmetric unit were the same as those in the wild-type (Fig. 5). Mutated E36 formed a salt bridge with R67 from neighbouring molecule chain C (Fig. 5). This interface was formed as a crystallographic artefact, which was confirmed by mutagenesis analysis.

Characterization of DREP2 CIDE and comparison with DREP4 CIDE. Previous structural analysis
showed that the CIDE domain forms helical oligomers through repetitive head-to-tail polymerization via charged interfaces, which are disrupted by a high concentration of NaCl 30 . As the salt concentration was increased, the DREP4 CIDE particle sizes decreased. Because the particle size of the CIDE domain varied depending on the buffer condition, protein concentration-dependent oligomerization, which is another important factor affecting filament formation, was also examined by size-exclusion chromatography and electron microscopy. The size of the DREP2 CIDE particle did not change as protein concentration was decreased from 30 to 1 mg/mL (Fig. 6A). Electron microscopy of negatively stained samples also showed that both DREP2 and the R36E mutant formed similar-sized rings and filaments in solution and that the filament size was not affected by protein concentration (Fig. 6B). In contrast, filament formation of DREP4 CIDE was dependent on the protein concentration based  on the results of size-exclusion chromatography (Fig. 6C) and electron microscopy (Fig. 6D). Smaller particles eluted at 15 mL in the size-exclusion chromatography (Fig. 6C) and smaller filaments were detected by electron microscopy (Fig. 6D) when low concentrations of DREP4 CIDE were used, strongly indicating that the filament formation of DREP4 CIDE depends on the protein concentration, whereas the DREP2 CIDE filament is not sensitive to protein concentration. This indicates that each CIDE domain exhibits unique features in filament formation, although the structure and surface features are identical.

Discussion
Apoptotic DNA fragmentation, which is a hallmark of apoptosis, is mediated by apoptotic nuclease DFF40. DFF40 contains a CIDE domain, which is a protein interaction domain. The CIDE domain-mediated interaction of DFF40 with DFF45, an inhibitor of DFF40 that contains a CIDE domain, is critical for controlling the activity of DFF40. Filament-like assembly of DFF40 via CIDE domain after removing DFF45 upon activation of apoptotic signalling is necessary for DNA fragmentation. In addition to its function in DNA fragmentation, the CIDE domain-containing proteins CIDEA, CIDEB, and FSP27 play important roles in lipid homeostasis. It has been reported that CIDEA, CIDEB, and FSP27 localize at lipid droplet contact sites, promoting lipid transfer and lipid droplet fusion in adipocytes and hepatocytes. Because of this involvement, CIDE-containing proteins were emerged targets for therapeutic intervention of metabolic disorders. The high resolution structure of the DREP2 CIDE showed that the structure exhibited an atypical CIDE structure that contains two helices, α1 and α2, and four strands, β1, β2, β3, and β5, by replacing β4 with a loop. A structural homology search conducted using DALI 32 revealed that the DREP2 CIDE domain is highly similar to ubiquitin-like domains and other CIDE domains (Table 2). DALI server picked six structures, including CIDE-A (PDBid: 2EEL), CAD (PDBid: 1F2R-I), ICAD (PDBid: 1F2R-C), ubiquitin (PDBid: 4NQK-E), SUMO-3 (PDBid: 1U4A-A), and HUB-1 (PDBid: 3PLU-A), as top matches. Pair-wise structural alignments of the DREP2 CIDE domain and structural homologues showed that the length and orientation of the α2 helices in the DREP2 CIDE domain differed slightly from those of in the other structures (Fig. 7A-F). A district β4 was only detected in the structure of CAD (Fig. 7B). No ubiquitin-related proteins contained β4 (Fig. 7D-F). One of the distinct features of ubiquitin-related proteins, including ubiquitin and Sumo, is that β2 is longer than the CIDE domain (Fig. 7A-F).   Based on the structural similarity between the CIDE domain and ubiquitin, it would be interesting to functionally characterize and compare these domains. Although the structure of DREP2 CIDE is similar to those of other CIDEs in that it contains a typical α/β roll fold with two α helices and five (or four) β strands, the high-resolution structure revealed a possible biologically important higher oligomerization mechanism of the CIDE domain that functions through several novel dimeric interfaces formed between homo-dimers. Before solving helical filament-like oligomeric structure of DREP4, determining the DREP2 structure, which formed an unusual tetramer without any symmetry in the crystallographic asymmetry unit, was difficult. The current study showed that DREP2 CIDE also formed a filament-like structure with features differing from those of DREP4 CIDE in solution. The filament-like oligomeric structure of DREP2 was confirmed by mutagenesis analysis.

Proteins and accession numbers Z-score RMSD (Å) Identity (%) References
Although the function of CIDE domain-mediated filament-like structure in apoptotic DNA fragmentation was established in the current structural study, CIDE domain-mediated lipid metabolism and learning/memory in brain synapses require further investigation. The characterization and elucidation of the oligomeric forms of the CIDE domain and CIDE domain-containing proteins will provide important information regarding the function of various CIDE domain-containing proteins in apoptosis, lipid metabolism, and particularly lipid droplet (LD) growth and learning/memory in the brain synapses.

Methods
Sequence alignment. Clustal W has been used for analysing the amino acid sequences of CIDEs (http:// www.ebi.ac.kr/Tools/clustalw2/index.html).
Protein expression and purification. The expression and purification methods used in this study have been described in detail elsewhere 21,29,30 . Briefly, the DREP2 CIDE (amino acids 1-84), the DREP4 CIDE (amino acids , and FSP27 (amino acids 38-119) were expressed in Escherichia coli BL21 (DE 3) under overnight induction at 20 °C. The protein contained a carboxyl terminal His-tag and was purified by nickel affinity and size-exclusion chromatography with a S200 gel filtration column 10/30 (GE Healthcare, Little Chalfont, UK) that had been pre-equilibrated with 20 mM Tris-HCl at pH 8.0 and 150 mM or 500 mM NaCl. The protein was then concentrated for further use.
Native PAGE shift assay. Oligomeric forms of each CIDE domains were monitored by native (non-denaturing) PAGE on a PhastSystem (GE Healthcare) with pre-made 8-25% acrylamide gradient gels (GE Healthcare). Separately purified proteins were directly loaded onto the gel. Coomassie Brilliant Blue was used for staining and detection of the shifted bands. The uncropped scan of gel is provided at Supplementary Fig. 1. Crystallization and data collection. The crystallization conditions were initially screened at 20 °C by the hanging drop vapor-diffusion method using various screening kits. Initial crystals were grown on the plates by equilibrating a mixture containing 1 μL of protein solution (7-8 mg/mL protein in 20 mM Tris-HCl at pH 8.0, 500 mM NaCl) and 1 μL of a reservoir solution containing 300 mM magnesium formate dihydrate and 100 mM Bis-Tris pH 6.2 against 0.4 mL of reservoir solution. The best crystal was obtained by further optimization searching over a range of concentrations of protein and precipitant and pH ranges. The diffraction data set was collected at beamline BL-4A of the Pohang Accelerator Laboratory, Republic of Korea. Data processing and scaling were carried out using the HKL2000 package 33 . The mutant crystal was obtained in the similar crystallization condition. 2.3 Å and 3.3 Å data were collected for wildtype and mutant, respectively.
Structure determination and analysis. The initial molecular-replacement (MR) method was carried out using Phaser 34 with the solution structure of the CIDE-N domain of human cell death activator CIDE-A (PDB code 2EEL), which has 37% amino-acid sequence identity, as a search model. The MR solution gave rotation-function and translation function Z-scores of 5.4 and 18.9, respectively. Model building and refinement were performed using COOT 35 and Refmac5 36 , respectively. Model quality was evaluated using PROCHECK 37 . Pymol was used to generate all the cartoon figures 38 . Mutagenesis. Site-directed mutagenesis was conducted using a Quick-change kit (Stratagene, La Jolla, CA, USA) according to the manufacturer's protocols. Mutagenesis was then confirmed by sequencing. Mutant proteins were prepared as described above.
Oligomerization assay by size-exclusion chromatography. For gel filtration analysis to detect oligomerization formation, the target protein was applied to a gel-filtration column (Superdex 200 HR 10/30, GE Healthcare) that had been pre-equilibrated with 20 mM Tris-HCl 8.0 and 500 mM NaCl. The peak fractions were collected and subjected to SDS-PAGE.

Multi-angle light scattering (MALS).
The molar mass of the highly oligomerized CIDE domain of DREP2 was determined by MALS. The target protein was injected onto a Superdex 200 HR 10/30 gel filtration column (GE Healthcare). The chromatography system was coupled to a three-angle light scattering detector (mini-DAWN EOS) and refractive index detector (Optilab DSP) (Wyatt Technology, Santa Barbara, CA, USA). Data were collected every 0.5 s at a flow rate of 0.2 mL/min and analysed using the ASTRA program, which gave the molar mass and mass distribution (polydispersity) of the sample. cuvette at a bandwidth of 1.0 nm, rate of 50 mm/min, and 5-s response time. The protein samples in buffer containing 20 mM Tris-HCl at pH 8.0 and 150 mM NaCl were diluted to 0.1 mg/mL prior to use. Four scans were accumulated and averaged, after which the α-helical content was calculated from the molar ellipticity at 222 nm 39 .
Electron microscopy. Wild-type DREP2 CIDE and R36E mutant samples after affinity chromatography purification were diluted to 0.8 mg/mL to prepare the high concentration sample and 0.1 mg/mL to prepare the low concentration sample. DREP4 CIDE samples were also diluted to 1 mg/mL as the high concentration sample and 0.1 mg/mL as the low concentration sample. For negative staining, 10 μL of each protein sample was placed onto a glow discharged copper grid and stained with 1% uranyl formate at pH 4.5 for 30 s and air-dried. The grids were imaged using a Tecnai G² Spirit BioTWIN Transmission Electron Microscope (FEI Company, Hillsboro, OR, USA) and recorded with an AMT 2k CCD camera (Thermo Fisher Scientific, Waltham, MA, USA).

Protein Data Bank accession codes.
Coordinates and structural factors have been deposited in the Protein Data Bank.