Structural and kinetic insights into flavin-containing monooxygenase and calponin-homology domains in human MICAL3

The structure of human MICAL3 provides information about its mechanism.


Introduction
Flavin-dependent monooxygenases catalyze a variety of oxygenation reactions, including regioselective, chemoselective and stereoselective oxidation reactions, which can be accomplished by single oxygen transfer to target substrates (Hung et al., 2010(Hung et al., , 2011. These enzymes are involved in several metabolic processes, including the biosynthesis of hormones and vitamins, the inactivation of signaling molecules, the excretion of xenobiotic substrates and the guidance of axons (Kaya et al., 2015;Drazic & Winter, 2014;Lee et al., 2013;Nadella et al., 2005). With regard to the regulation of axon guidance, the molecule interacting with CasL (MICAL) protein, which belongs to the flavin-dependent monooxygenase family, has recently received attention for its unique function in controlling actin assembly in conjunction with MsrB1 (Lee et al., 2013). MICAL is a multi-domain enzyme that is highly conserved across most animal kingdoms and participates in actin cytoskeletal reorganization (Kolk & Pasterkamp, 2007). It has been functionally characterized to possess stereoselective methionine-oxidation activity for actin disassembly in mice and fruit flies. Its flavin-containing monooxygenase (FMO) domain oxidizes the two conserved methionine residues of actin to methionine-R-sulfoxides, while the calponin-homology (CH) domain is involved in actin binding, thereby enhancing the catalytic activity of the FMO domain (Gimona et al., 2002).
Three MICALs, MICAL1, MICAL2 and MICAL3, have been identified in mammals. They contain FMO and CH domains and share a common function of F-actin disassembly. In addition, MICAL2 and MICAL3 have been reported to localize to the nucleus, but MICAL1 is cytosolic (Lundquist et al., 2014;Fischer et al., 2005). Furthermore, MICAL1 is autoinhibited by its C-terminal coiled-coil domain, but MICAL2 is constitutively active for the regulation of actin stress fibers (Giridharan et al., 2012). Following the discovery of MICAL in Drosophila, its biological role in actin cytoskeleton reorganization has been shown in various organisms. MICALs oxidize two conserved methionine residues of actin to methionine-Rsulfoxide in a stereospecific manner in the presence of NADPH, thus leading to the disassembly of F-actin into actin monomers (Giridharan & Caplan, 2014). MsrB then reduces these methionine-R-sulfoxides back to methionines. After this redox change, the actin monomer can be reassembled into an F-actin polymer, thereby conferring cytoskeletal reorganization (Lee et al., 2013). These cytoskeletal alterations regulate multiple cellular events including cell morphology and axon growth, and exocytosis in various tissues. However, the underlying mechanism of this regulation by MICAL proteins is poorly understood (Wu et al., 2018). The structure of mouse MICAL1 has been determined, which has helped to provide an understanding of how it functions at the molecular level (Siebold et al., 2005;Nadella et al., 2005;Sun et al., 2006). However, the structures of MICAL2 and MICAL3 have yet to be determined, and thus the lack of structural comparison among mammalian MICALs has limited our understanding of both the shared and unique characteristics of these proteins. Here, we report the first crystal structure of human MICAL3, which contains the FMO and CH domains, at 1.9 Å resolution. Based on a comparison of human MICAL3 with mouse MICAL1 (PDB entry 4txi; Alqassim et al., 2016), we present their structural similarities and differences, as well as their kinetics, which depend on the interactions of the FMO domain with the CH domain.

Cloning of human MICAL1 and MICAL3
The DNA regions encoding the FMO and CH domains of MICAL3 (hMICAL3 FMOCH ; amino-acid residues 1-700), as well as the FMO domain of MICAL3 (hMICAL3 FMO ; aminoacid residues 1-494), were amplified from human cDNA. Truncated forms of MICAL1 comprising the FMO and CH domains (hMICAL1 FMOCH ; amino-acid residues 1-616) and the FMO domain (hMICAL1 FMO ; amino-acid residues 1-489) were amplified from human MICAL1 cDNA purchased from the Korean Human Gene Bank. The oligonucleotide primers used for amplification can be found in Supplementary Table  S1. After PCR amplification, hMICAL3 FMOCH was cloned into the pET-28b plasmid (Novagen) using the restriction enzymes BamHI and XhoI. Other MICAL forms were digested with the same enzymes and inserted into the pET-28a vector. E213G and R530G substitutions were introduced into hMICAL3 FMOCH in pET-28b by site-directed mutagenesis (hMICAL3 FMOCH Á213,530; Haque et al., 2018). To enhance the solubility of the recombinant MICAL forms, they were transformed and overexpressed in the Escherichia coli Rosetta 2 pLysS strain (Novagen). The cells were grown at 37 C to an optical density at 600 nm (OD 600 ) of approximately 0.6 in LB medium and were induced at 18 C with 0.5 mM isopropyl -d-1-thiogalactopyranoside (IPTG). After 18 h, the cells were harvested by centrifugation at 2300g for 1 h and stored at À20 C.

Protein purification
For each MICAL form, the harvested cells were lysed by sonication in a buffer consisting of 50 mM Tris pH 8.5, 150 mM NaCl, 2 mM -mercaptoethanol . After centrifugation at 13 000 rev min À1 for 1 h at 4 C, the supernatant was loaded onto an Ni-NTA column (GE Healthcare) equilibrated with binding buffer (50 mM Tris pH 8.5, 150 mM NaCl, 2 mM -mercaptoethanol) and eluted with elution buffer (50 mM Tris, 150 mM NaCl, 2 mM -mercaptoethanol, 1 M imidazole). The purified MICAL proteins (except for hMICAL3 FMOCH ) were concentrated using Amicon ultracentrifugal filters (Ultracel-30, 30 kDa cutoff; Millipore) and stored at À80 C. For the subsequent crystallization experiments, hMICAL3 FMOCH was concentrated and further purified by gel-filtration chromatography on a Superdex S200 column (GE Healthcare) using elution buffer consisting of 50 mM Tris pH 8.5, 100 mM NaCl, 1 mM 1,4-dithiothreitol, 1% glycerol. The predicted molecular weight of hMICAL3 FMOCH is $79.4 kDa. The monomeric form identified in some fractions during gel-filtration chromatography was collected, concentrated to 25 mg ml À1 using Amicon ultracentrifugal filters and stored at À80 C.

Structure determination
A diffraction data set was collected on beamline 5A at Pohang Accelerator Laboratory, Republic of Korea at a wavelength of 0.9794 Å and the data were processed and scaled using SCALEPACK and DENZO from the HKL-2000 program package (Otwinowski & Minor, 1997). The crystal belonged to space group P2 1 , with unit-cell parameters a = 65.233, b = 94.363, c = 71.825 Å . The structure was solved by molecular replacement with mouse MICAL1 (PDB entry 4txi; Alqassim et al., 2016) and the NMR structure of the CH domain of human MICAL3 (PDB entry 2d88; RIKEN Structural Genomics/Proteomics Initiative, unpublished work) as the initial models. Molecular replacement was performed with Phenix, model building was performed with Coot and final refinement was conducted with REFMAC5 in the CCP4 suite (Liebschner et al., 2019;Emsley et al., 2010;Winn et al., 2011).

Electron microscopy
The purified sample of human MICAL3 FMOCH was diluted with 50 mM Tris pH 8.5, 100 mM NaCl, 1 mM 1,4-dithiothreitol, 1% glycerol to a final concentration of 80 nM. The treated samples (5 ml) were immediately applied onto a carbon-coated grid which had previously been glowdischarged (Harrick Plasma, Ithaca, New York, USA) for 3 min in air. The grids were negatively stained using 1% uranyl acetate. The prepared grids were examined on a Tecnai 10 transmission electron microscope (FEI, USA) equipped with a lanthanum hexaboride (LaB 6 ) cathode operating at 100 kV (the instrumentation used at Kangwon Overall structure and characterization of human MICAL3 FMOCH . (a) Sequence alignment of human MICAL3, human MICAL2, human MICAL1 and mouse MICAL1. Strictly conserved residues are boxed in red, while similar residues are shown as red letters. The sequence-alignment tools used were ClustalW and ESPript. 3 10 -Helices are represented by , strict -turns are represented by TT and strict -turns by TTT.
Center for Systems Imaging). Images were collected using a US1000 CCD camera (Gatan, USA) at a magnification of 0.32 nm per pixel (Kim et al., 2012). Single-particle 3D reconstruction from the 1180 negatively stained individuals was performed using the EMAN2 package approach with C1 symmetry (Ludtke, 2016). UCSF Chimera was used for the visualization and analysis of 3D volumes (Pettersen et al., 2004).

Steady-state kinetic measurements
NADPH consumption was monitored in 96-well microplates at a wavelength of 340 nm using a SpectraMax i3 spectrophotometer (Molecular Devices). All reactions were executed in an F-buffer-based assay mixture. The F-actin was prepared according to the supplier's instructions. 1 mg of actin from rabbit skeletal muscle (Cytoskeleton Inc.) was reconstituted in 100 ml distilled water. It was diluted to 50 mM in G-buffer (5 mM Tris-HCl, 0.2 mM CaCl 2 , 0.2 mM ATP, 1 mM DTT pH 8.0) and incubated for 1 h on ice. Following ultracentrifugation at 100 000g for 1 h at 4 C, the supernatant that contained G-actin was separated carefully. 10Â polymerization buffer (50 mM Tris-HCl, 500 mM KCl, 20 mM MgCl 2 , 10 mM ATP pH 7.5) was then added to the supernatant and incubated at 25 C for 1 h with shaking. After polymerization, the F-actin was diluted to 8 mM with F-buffer (a mixture of G-buffer and polymerization buffer in a 9:1 ratio) and stored on ice until necessary. NADPH was dissolved in 20 mM Tris-HCl to make a 10 mM stock that was serially diluted as needed. NADPH and MICAL proteins were prepared at ten times the final concentration and 10 ml of each was added to 80 ml F-buffer (without F-actin) or F-actin. For the NADPH standard, the binding buffer for MICAL purification was added instead of the MICAL protein. All experiments were performed in triplicate, and the reaction was recorded every 10 s for 15 min. Kinetic parameters were calculated by fitting a nonlinear regression to the Michaelis-Menten equation using GraphPad Prism 5.

Overall structure of human MICAL3 FMOCH
To gain structural insights into the role of human MICAL3 (hMICAL3) in actin sulfoxidation, recombinant hMICAL3 containing the FMO and CH domains (hMICAL3 FMOCH ) was expressed and purified for crystallization. The structure of hMICAL3 FMOCH was determined at 1.9 Å resolution. hMICAL3 contains multiple domains: FMO, CH, Lin11, Isl-1, Mec-3 (LIM) and C-terminal coiled-coil domains. In particular, the FMO and CH domains of MICAL3 are homologous to the FMO and CH domains of MICAL1 and MICAL2 [ Fig. 1(a)] and are known to play a role in methionine sulfoxidation and actin binding, respectively (Hung et al., 2011;Giridharan et al., 2012). The FMO domain of hMICAL3 FMOCH has the conserved GXGXXG motif that forms the central part of the consensus hMICAL3 sequence connecting the first -strand and -helix in the Rossmann fold. hMICAL3 is an oxidoreductase enzyme that depends on FAD and NAD(P) for its catalytic activity (Vanoni et al., 2013). In the structure, Cys97, Val126, Tyr298, Asp398 and Trp405 interact with one FAD molecule. In particular, the indole ring of Trp405 stabilizes the flavin isoalloxazine ring of the FAD via a coaxial stacking interaction, while the other residues interact via hydrogen bonds [Figs. 1(b) and 1(c)]. The crystallographic R work and R free factors are 20.4% and 22.4%, respectively, and the Ramachandran favored region score is $98.1% (Table 1) not observed in the structure. The secondary structure in this region could not be predicted using JPred (Cole et al., 2008), suggesting that this region is flexible.

CH domain of hMICAL3 FMOCH
The CH domain of hMICAL3 FMOCH is composed of five -helices and contains a conserved actin-binding motif (522-SKLLGWCQR-530). The hMICAL3 FMOCH structure shows that the CH domain interacts with the FMO domain, as is the case in mouse MICAL1 (mMICAL1; Alqassim et al., 2016). In particular, five residues, Lys523, Arg530, Gln531, Tyr620 and Leu627, in the CH domain interact with Lys207, Thr208, Pro210, Glu213 and Glu215 in the FMO domain (Fig. 2). The main chain of Glu213 forms hydrogen bonds to Lys523, and the side chain of Gly213 forms a hydrogen bond and a salt bridge to Tyr620 and Arg530, respectively. In addition, the main chains of Lys207 and Thr208, which correspond to Asn201 and Pro202 of mMICAL1, participate in hydrogen bonding to the side chains of Arg530 and Gln531. Interestingly, the other four binding residues of the FMO domain are not conserved in mMICAL1, except for Glu215 [ Fig. 1(a)].

Interactions between the CH and FMO domains enhance catalytic efficiency
Generally, the type 2 CH domain does not bind to F-actin directly. Instead, it facilitates the binding of F-actin to other parts of the protein (Sun et al., 2006;Gimona & Mital, 1998;Gimona & Winder, 1998). To investigate the effects of the CH domain on MICAL enzyme activity, truncated MICAL proteins with or without the CH domain were produced (hMICAL1 FMO , hMICAL1 FMOCH , hMICAL3 FMO and hMICAL3 FMOCH ). It is known that MICAL has NADPH oxidase activity that underlies F-actin disassembly simultaneously with the oxidation of NADPH (Zucchini et al., 2011). Using this property, kinetic parameters for each MICAL form were determined from the initial velocity of the NADPH oxidase reaction, which depends on the NADPH concentration. The initial velocity of the reaction for hMICAL1 FMOCH and hMICAL3 FMOCH at each NADPH concentration was enhanced by adding F-actin, whereas there was no change in the initial velocity for hMICAL1 FMO and hMICAL3 FMO when F-actin was added [Figs. 3(a) and 3(b)]. When F-actin was not added, the k cat /K m value for hMICAL3 FMO ($2.4794 s À1 M À1 ) was similar to that for hMICAL3 FMOCH ($2.0776 s À1 M À1 ). However, the k cat /K m value for hMICAL3 FMOCH ($30.9857 s À1 M À1 ) increased dramatically on adding F-actin, while the k cat /K m value for hMICAL3 FMO ($2.8793 s À1 M À1 ) changed little even after adding F-actin (  Binding site between the CH domain and the FMO domain in human MICAL3. Residues that participate in the interaction between the FMO and CH domains are labeled. The FMO domain is shown in yellow and the CH domain is shown in green. The binding residues between the FMO domain and the CH domain in human MICAL3 are shown as stick models. The dotted lines indicate the interaction distances between residues. on adding F-actin when the CH domain is present was also observed in hMICAL1. Nevertheless, the ratio of the increase was much higher in hMICAL3 FMOCH ($15 times) compared with hMICAL1 FMOCH ($6 times). Consequently, these results reveal that the CH domain of hMICAL3 participates in increasing the F-actin substrate specificity, leading to enhanced catalytic efficiency. Moreover, it appears that the CH domain of hMICAL3 might make a more efficient interaction with the FMO domain for F-actin substrate specificity relative to the CH domain of hMICAL1. We chose two residues, Glu213 in the FMO domain and Arg530 in the CH domain, which were predicted to be key residues associated with the interaction between the FMO and CH domains and mutated them to examine the effect of disrupting the FMO-CH interaction on the catalytic efficiency (hMICAL3 FMOCH Á213,530; E213G and R530G). Notably, the kinetic parameters of hMICAL3 FMOCH Á213,530, including the initial velocity and catalytic efficiency, are more similar to those of hMICAL3 FMO than hMICAL3 FMOCH [Figs. 3(b) and 3(c) and Table 2]. Moreover, F-actin does not increase the catalytic efficiency of hMICAL3 FMOCH Á213,530. Therefore, this result suggests that the FMO-CH interaction in hMICAL3 is required to increase the catalytic efficiency by conferring specific binding to F-actin.

Structural comparison of CH domains between mouse MICAL1 and human MICAL3
The sequence similarity between the FMO domains of hMICAL3 (residues 1-494) and mMICAL1 (residues 1-489) is $58%. The FMO domains showed structural similarity, with a root-meansquare (r.m.s.) deviation of 0.56 Å when 397 C atoms were aligned in PyMOL ( Fig. 4; Janson et al., 2017). Likewise, the CH domains showed structural similarity, with an r.m.s. deviation of 0.95 Å , and were considered to be type 2 CH domains ( Fig. 4; Janson et al., 2017). However, the relative position of the CH domain of hMICAL3 (hMICAL3 CH ) in the crystal structure was different from that in mMICAL1. The CH domain can potentially occupy three positions (named positions A, B and C) in the context of an asymmetric unit [ Fig. 5(a)]. It was reported that the CH domain of mMICAL1 was located at position B (Alqassim et al., 2016), whereas our structure showed that the CH domain of hMICAL3 was located at position A, where it was rotated approximately 90 from position B [ Fig. 5(b)]. There is an invisible region (residues 495-517) between the Cterminus of the FMO domain and the N-terminus of the CH domain in the hMICAL3 FMOCH structure, with a distance of $25.3 Å . Essentially, the calculated distances between the FMO and CH domains are $25.3 Å for position A, $48.8 Å for position B and 68.9 Å for position C. Therefore, considering the distance of the invisible region, the CH domain of hMICAL3 fits position A. Furthermore, we performed electron microscopy followed by singleparticle analysis to identify the relative position of the CH domain in hMICAL3 (Fig. 6). The visual similarity seen in 3D reconstruction and 2D class averages with corresponding views of raw images Steady-state kinetic analysis of MICAL forms. Initial velocity (v) was measured in an F-bufferbased mixture at various NADPH concentrations. The initial velocity of each reaction was divided by the total enzyme concentration (E, 400 nM).
[ Fig. 6(a)] and the fitting of the crystal structure of MICAL3 to the 3D volume of negatively stained MICAL3 molecules supports the idea that the CH domain is located at position A [c.f. the 3D reconstruction in Fig. 6(b) and the models in Fig.  5(a)]. We then built an initial model of the MICAL3 FMOCH -Factin interaction [ Fig. 6(c)] using position A; the active site seems to be close to the D-loop where the target methionine residues are found.

Discussion
MICALs are involved in actin cytoskeleton reorganization through methionine oxidation. However, our understanding of MICALs is limited to genetic and cell biology results, which have been presented in recent papers (Vanoni et al., 2013;Giridharan & Caplan, 2014;Lim et al., 2019). To broaden the scope of these findings, further structural research is required to understand the exact mechanism of how MICALs oxidize methionine. The full-length MICAL protein contains FMO, CH and LIM domains, has a C-terminal domain with unknown function and is highly insoluble, which makes it difficult to obtain a crystal structure. However, its truncated form containing the FMO and CH domains is much more soluble than the full-length form but still retains catalytic activity for F-actin disassembly. Thus, the truncated form of MICAL has been used in various biochemical and biological experiments to gain structural insight into the catalytic mechanism. Of the three mammalian MICALs, only the structure of mouse MICAL1 has been reported and the structure is of a truncated    form containing the FMO and CH domains. Interestingly, mMICAL1 FMOCH and hMICAL3 FMOCH are structurally highly similar and share the same catalytic function of depolymerizing F-actin via oxidation of the two conserved methionine residues (Fig. 4). Nevertheless, the crystal structure and electron-microscopy data of hMICAL3 FMOCH show that the spatial arrangement of the FMO and CH domains differs from that in mMICAL1 (Figs. 5 and 6). In the crystal structure of hMICAL3 FMOCH , three possible positions of the CH domain could arrange in an asymmetric unit (Fig. 5) and one of them matches with the observations from electron microscopy.
These data also suggest that the crystal structure of mMICAL1 differs from the solution shape from SAXS data (Alqassim et al., 2016). Both crystal structures clearly show differences, which include the relative location of the CH domain and the length of the invisible region between the FMO and CH domains. The invisible region between the FMO and CH domains is 23 amino acids (495-517) in hMICAL3 and 18 amino acids (490-507) in mMICAL1. Therefore, this could suggest that different binding conformations are adopted by the FMO and CH domains in hMICAL3 and mMICAL1. Although hMICAL3 and mMICAL1 have highly conserved FMO and CH domains (Fig. 3), there are several reasons why MICAL3 and MICAL1 may have a different mechanism. Firstly, in the FMO domain of hMICAL3 the loop is longer than that in mMICAL1 and hMICAL1 (Fig. 7). The sequence of this loop is conserved in hMICAL3 and hMICAL2, but is not conserved in MICAL1. However, the sequence of this loop in MICAL1 is conserved in the human and mouse enzymes (Fig. 7). Therefore, MICAL2 and MICAL3 may have a similar binding mode by the type 2 CH domain, but MICAL1 does not. hMICAL3 and mMICAL1 are biologically similar in structure, but it is difficult to determine whether they have the same mechanism (Wu     Electron-microscopic analysis of human MICAL3 FMOCH and a model of MICAL3 FMOCH -F-actin interaction. (a) Structural comparison taken from 3D analysis: representative surface views of the reconstructed 3D structure (top row) and the corresponding views of 2D class averages (middle row) and raw particles (bottom row). The 10 nm scale bar applies to all of the panels in (a). (b) Superimposition of an equivalent view of the crystal structure (yellow) on the 3D envelope of negatively stained human MICAL3. (c) The initial model was built by manually docking human MICAL3 FMOCH to F-actin (PDB entry 3lue; Galkin et al., 2010). The CH domain (dark blue cartoon) was oriented first and the FMO domain (purple cartoon) was arranged so that the active site was close to actin. Each actin monomer is represented in a different color (tinted surfaces). et al., 2018). In the middle and bottom panels of Fig. 7, the superimposition of hMICAL3, hMICAL2 and mMICAL1 shows that the -helices of hMICAL2 and hMICAL3 completely superimpose but the -helix of mMICAL1 is only shifted slightly. This sequence also shows that hMICAL3 and hMICAL2 are completely conserved and mMICAL1 is not conserved at all. Therefore, we suggest that hMICAL3 and hMICAL2 may have a similar mechanism and can be grouped into the same class.
The FMO domain that exhibits monooxygenase activity is localized at the N-terminus of MICAL and is highly conserved among species. The CH domain that is usually found in actinbinding proteins is adjacent to the FMO domain and is also highly conserved. CH domains are classified into three types: types 1, 2 and 3. Whereas type 3 CH domains are mainly found in regulatory proteins associated with muscle contraction and signaling proteins, type 1 and 2 CH domains are usually found in cytoskeletal proteins (Zhou et al., 2011). MICALs have a typical type 2 CH domain. The results of kinetic experiments with hMICAL1 and hMICAL3 reveal that when F-actin is present as a substrate, MICAL FMOCH shows a much higher catalytic efficiency than MICAL FMO . In the case of MICAL FMO , there was no significant change in the activity depending on the presence of F-actin as a substrate. These kinetic data demonstrate that the FMO domain has catalytic activity but that the CH domain must be present for substrate specificity. Moreover, it was shown that the activity of MICAL FMO is slightly higher than that of MICAL FMOCH without F-actin. These two results suggest that the FMO domain performs the catalytic function regardless of the CH domain but that the CH domain is essential for the F-actin substrate specificity of MICALs. In addition, type 2 CH domains do not possess the ability to bind directly to F-actin, as described previously (Vanoni et al., 2013;Giridharan & Caplan, 2014;Zhou et al., 2011). Therefore, we conclude that the interaction between the FMO and CH domains might generate the substrate specificity, particularly for F-actin.
We also examined the change in catalytic efficiency upon mutating the MICAL3 FMOCH protein by replacing Glu213 and Arg530, two residues that are important for maintaining the interaction between the FMO and CH domains. Upon mutation, the catalytic efficiency is reduced to levels consistent with hMICAL FMO . Thus, these findings consistently support the idea that interaction between the FMO and CH domains increases the F-actin substrate specificity and the subsequent catalytic efficiency. Finally, although MICAL has catalytic activity for methionine oxidation, only actin is currently known to be its substrate. The mechanism of sulfoxidation remains disputed and is considered to be owing to direct oxidation by the transfer of a single oxygen molecule or via an indirect oxidation by the production of reactive oxygen species such as hydrogen peroxide. However, it was found that consumption of NADPH without F-actin substrate produced hydrogen peroxide in MICAL1 and MICAL3, but this was not proportional to the activity level of F-actin disassembly (Wu et al., 2018). Consequently, these findings suggest that the CH domain is crucial for conferring F-actin disassembly via methionine oxidation. In this context, the CH domain of MICAL3 interacts with the FMO domain in more efficient ways than the CH domain of MICAL1.1