NADPH-dependent oxidation of CRMP2 through a MICAL1-Prx1 redox relay controls neurite outgrowth

CRMP2/DPYL2 is an effector protein in the semaphorin signaling pathway that controls cytoskeletal dynamics, linking extracellular signals to the formation of axonal networks. CRMP2 is regulated by post-translational modifications including a dithiol-disulfide redox switch. The mechanisms of reduction of this switch were established, the signal-induced oxidation, however, remained unclear. Here, we show that CRMP2 is oxidized through a redox relay involving the flavin-mooxygenase MICAL1 and the peroxidase Prx1 as specific signal transducers. Using molecular oxygen and electrons provided by NADPH, MICAL produces hydrogen peroxide and specifically oxidizes Prx1 through direct interactions between the proteins. Subsequently, Prx1 oxidizes CRMP2. The lack of any components of this redox relay dysregulates neurite outgrowth. Consequently, both oxidation and reduction of CRMP2 require reducing equivalents in the form of NADPH. 16 17 18 19 20 21 22 23 24 25 26


Introduction
The semaphorin (Sem) signaling pathway contributes to the regulation of axonal outgrowth and neuronal connectivity. Activation of the Sem3A pathway induces growth cone collapse and retraction of the outgrowths by controlling the de-polymerization of the cytoskeleton that subsequently allows the axons to change their direction. Sem3A acts via a hetero-dimeric transmembrane receptor composed of neuropilin 1 (NP1) and a plexin A (PlexA) family member 1,2 .
Effector proteins include the kinases CDK5 and GSK3β that sequentially phosphorylate collapsin response mediator protein 2 (CRMP2, gene: DPYSL2) 3 . CRMP2 was reported to regulate microtubuli dynamics by binding to α-β-tubulin dimers. MICAL (molecule interacting with CasL) proteins bind to the cytoplasmic domain of PlexAs 4 . The human genome encodes three MICAL proteins. All contain an N-terminal FAD-dependent monooxygenase (MO) domain, followed by a calponin homology (CH) and a LIM domain. Human MICAL 1 and 3 contain an additional Cterminal Rab-binding domain (RBD) 5 . MICALs can produce H 2 O 2 upon activation of the Sem3A pathway 6 . Furthermore, MICALs were characterized as specific oxidases of actin methionyl residues 7,8 , thus regulating microfilament dynamics.
In general, redox-mediated signal transduction occurs by both reversible oxidation and reduction of key molecules. The side chains most vulnerable to redox modifications are cysteinyl and methionyl residues. Redox modifications occur -often rapidly -under physiological conditions and they are highly specific with respect to both, the molecules involved and the nature of the redox modification; reversible oxidation and reduction are key mechanisms in cell signaling 9,10 . While the 49 50 ≈10 6 -fold higher rate constants and are generally high abundant proteins in mammalian cells. It was thus suggested that these peroxidases may function as receptors and transducers of redox signaling, e.g. 12 . CRMP2 has been reported to be regulated by a dithiol-disulfide redox switch of two cysteinyl 504 residues in the homo-tetrameric quaternary complex, that is essential for neurite outgrowth and axonal connectivity 6,13 . In vivo, the reduction of this disulfide appears to be specifically regulated by the cytosolic isoform of glutaredoxin 2 (Grx2c), a process linked to both formation of a neuronal network and the progression of cancer cells [13][14][15] . MICAL-produced H 2 O 2 was suggested as potential source of CRMP2 oxidation 14 , however, the rate constant of the direct reaction of CRMP2 with H 2 O 2 is also too low to be of significance in vivo 16 . In fact, two studies provided evidence for the involvement of Prx1 in the oxidation of CRMP2 17,18 . We thus hypothesized a signal-induced redox relay in the semaphorin signaling pathway that involves both MICAL1 and Prx1 as signal transducers and CRMP2 as effector protein.

Results
We hypothesized that Sem3A signaling induces a redox relay involving a FAD-dependent monooxygenase of the MICAL family, one of the cytosolic Prxs, and CRMP2. The latter should then affect axonal outgrowth by binding to tubulin and actin, thus controlling cytoskeletal dynamics. We have used purified recombinant human CRMP2 to perform affinity capture assays in cell extracts. Using Western blotting, we confirmed the interaction between CRMP2 and both actin and tubulin (Fig. 1a). Moreover, also MICAL1 and Prx1 were captured in this assay; their release from the complex required reducing equivalents, indicating redox-dependent associations.
We proposed the production of hydrogen peroxide by a MICAL protein that leads to the oxidation of a Prx. Subsequently, this Prx may oxidize the Cys504 redox switch in tetrameric CRMP2. We cloned, recombinantly expressed, and purified the two cytosolic Prxs (Prx1-2) and the monooxygenase domains of MICAL1-3, henceforth named MICAL(1-3)-MO. All three recombinant MICAL-MO domains contained oxidized FAD, indicated by additional absorption bands at approx. 370 and 450 nm (Fig. 1b). For the comparison of spectra of both liberated and MICAL bound FAD, see Fig. S1. These proteins oxidized NADPH in the presence of O 2 , a reaction that yields H 2 O 2 , proving their activity as monooxygenases (Fig. 1c). Next, we incubated both Prx1 and Prx2 with the three MICAL-MO domains in a one-to-one stoichiometry (one Prx dimer to one MICAL-MO) in the presence of NADPH. In this setting, all MICAL-MOs led to the oxidation of the two Prxs, i.e. the formation of the catalytic inter-molecular disulfide between the peroxidatic and resolving cysteinyl residues of two Prx monomers (Fig. 1d). When a mutant of the resolving cysteinyl residue of the Prxs was used, dimer formation could not be observed. As a control, we have also confirmed that MICALs (in a 1:1 stoichiometry in the presence of oxygen and NADPH) do not directly oxidize CRMP2 (Fig. S2).

/ 39 64
To analyze the mechanisms of the MICAL reaction and the oxidation of the Prxs, we have turned to stopped-flow kinetics. As seen before, MICAL1-MO oxidized NADPH in a concentrationdependent manner in the presence of O 2 at pH 7.4 and 298 K ( Fig. 2a- We have also performed the oxidation of MICAL-MO in the presence of Prxs. The remaining spectra, however, did not suggest any other intermediates, e.g. a charge transfer complex between the FAD and the Prx. The reaction remained bi-phasic and the rate constants were in the range of those determined for the reaction in the absence of a Prx ( Fig. 2k-l, Table 1). We thus conclude that  Table 1, Fig. S6). We also and dimeric Prx1. Using a start structure obtained by molecular docking, we have analyzed the dynamics of the complex over the time course of 250 ns (Fig. 4). Within 50 ns, the simulation reached an equilibrium ( Fig. 4d-e). In this complex, one of the peroxidatic cysteinyl residues of the Prx-dimer is placed at the exit of a channel in MICAL1 that connects to the FAD/FADH 2 active site of the enzyme ( Fig. 4f-g), the side chains of both the peroxidatic and resolving cysteinyl were at the low end of the spectrum (Fig. 4h), suggesting a rather stable conformation.
Next, we aimed to confirm the proposed redox relay in an established human cellular model of neuronal differentiation. SH-SY5Y neuroblastoma cells can be differentiated into mature neurons through a variety of different mechanisms including treatment with retinoic acid (RA) 19 . At first, we have treated the cells in the neuroblast state with 1µg·ml -1 Sem3A and analyzed the oxidation of Prx1 and CRMP2. Both proteins form a homodimeric inter-molecular disulfide that can be identified after non-reducing Western blotting. We found an increase in both oxidized Prx1 and CRMP2 between 5 to 10 minutes after addition of the ligand (Fig. 5a). Since we proposed a Sem3A-induced redox-relay from MICAL1 to Prx1 to CRMP2, we repeated the experiment with  Table 1).
The crystal structures available for MICAL-MO domains imply a conformational change of the FAD from an 'out' conformation in the oxidized state to an 'in' conformation in the reduced state 37, 44 . This may be the explanation why the reduction of the FAD by NADPH is the rate limiting step in the MICAL monooxygenase reaction mechanism (Fig. S4). It may also explain why the rate constant of this fully reversible reaction step appeared to be independent of temperature or pH ( 2a-f). The rate was likely not limited by the electron transfer reaction itself, but rather by the following conformational changes. Reduced FAD immediately reacted with molecular oxygen. The apparently bi-phasic reaction likely resulted from the formation of an unstable peroxy-flavin, which decays to H 2 O 2 and the oxidized flavin (Fig. S4) 36 . The rate constant of the first step was too high for us to spectroscopically verify the peroxy-flavin intermediate (Fig. 2g- Fig. 4a-c, Fig. S8). Such a binding event brings the peroxidatic cysteinyl residue of one monomer of the dimeric Prx in close proximity to the active site opening in the MICAL1-MO domain where the FAD is bound (Fig. 4f-g), the H 2 O 2 produced may then be directly channeled to the Prx, which is supported by the finding that catalase does not inhibit Prx oxidation.
( Fig. 3). The oxidation of CRMP2 by Prx1 was demonstrated before. However, it seems that this reaction requires at least one additional factor that may act as scaffold bringing Prx1 and CRMP2 into close enough proximity for the reaction to take place 17,31 .
Different studies suggested an involvement of the oxidized form of CRMP2 in the semaphorin signaling pathway before although without clarifying the mechanism of oxidation nor a sufficient explanation for the specific reduction. The use of a mutant thioredoxin in this enzymatic reaction leads to the formation of an artificial disulfide 6,36 .
CRMP2 is subject to a variety of post-translational modifications. Five phosphorylation sites are located at the C-terminus 3,46-49 , following the redox-sensitive cysteinyl residue 504 13 Table S1 13,54 .

Recombinant expression and purification
With the following substitutions, the expression rate law becomes: Integration: Integration using the method of partial fractions: Rearrangement and re-substitution yields: Upon re-arrangement, we obtain the linearized equation of the integrated second order kinetics: In practice, [NADPH] 0 was given and controlled based on its molar absorptivity at 340 nm;  and Prx1 models were prepared by homology modeling using pdb entries 6ici and 5hqp, respectively. The simulations were performed in GROMACS 2016.3 56 , with AMBER-99ff-ILDN force field 57 . The parametrization of FAD was done with ACPYPE 58 script. The protein structure was solvated with TIP3P water in a cubic box under periodic boundary conditions and at least 1 nm away from the edge of the box. Na + and Clions were added to neutralize the charge of the system. An initial energy minimization was performed using steepest descent algorithm until the system converged to 1000 kJ·mol -1 ·nm -1 . System equilibration was performed for 100 ps at a constant number of molecules, volume, and temperature 300 K (NVT) and for duration of 100 ps with production simulation was 250 ns (125,000,000 time steps, 2 fs each). The bonded interactions of hydrogens were constrained with LINCS algorithm 59 . The Parrinello-Rahman 60 method was used for pressure coupling and the modified Berendsen thermostat -velocity rescale 61 Table S2).      39 / 39