Flavin monooxygenases regulate Caenorhabditis elegans axon guidance and growth cone protrusion with UNC-6/Netrin signaling and Rac GTPases

The guidance cue UNC-6/Netrin regulates both attractive and repulsive axon guidance. Our previous work showed that in C. elegans, the attractive UNC-6/Netrin receptor UNC-40/DCC stimulates growth cone protrusion, and that the repulsive receptor, an UNC-5:UNC-40 heterodimer, inhibits growth cone protrusion. We have also shown that inhibition of growth cone protrusion downstream of the UNC-5:UNC-40 repulsive receptor involves Rac GTPases, the Rac GTP exchange factor UNC-73/Trio, and the cytoskeletal regulator UNC-33/CRMP, which mediates Semaphorin-induced growth cone collapse in other systems. The multidomain flavoprotein monooxygenase (FMO) MICAL (Molecule Interacting with CasL) also mediates growth cone collapse in response to Semaphorin by directly oxidizing F-actin, resulting in depolymerization. The C. elegans genome does not encode a multidomain MICAL-like molecule, but does encode five flavin monooxygenases (FMO-1, -2, -3, -4, and 5) and another molecule, EHBP-1, similar to the non-FMO portion of MICAL. Here we show that FMO-1, FMO-4, FMO-5, and EHBP-1 may play a role in UNC-6/Netrin directed repulsive guidance mediated through UNC-40 and UNC-5 receptors. Mutations in fmo-1, fmo-4, fmo-5, and ehbp-1 showed VD/DD axon guidance and branching defects, and variably enhanced unc-40 and unc-5 VD/DD axon guidance defects. Developing growth cones in vivo of fmo-1, fmo-4, fmo-5, and ehbp-1 mutants displayed excessive filopodial protrusion, and transgenic expression of FMO-5 inhibited growth cone protrusion. Mutations suppressed growth cone inhibition caused by activated UNC-40 and UNC-5 signaling, and activated Rac GTPase CED-10 and MIG-2, suggesting that these molecules are required downstream of UNC-6/Netrin receptors and Rac GTPases. From these studies we conclude that FMO-1, FMO-4, FMO-5, and EHBP-1 represent new players downstream of UNC-6/Netrin receptors and Rac GTPases that inhibit growth cone filopodial protrusion in repulsive axon guidance.


Introduction
The formation of neural circuits during development depends on the guidance of growing axons to their proper synaptic targets. This process relies on the growth cone, a dynamic actin based structure present at the tip of a growing axon. Growth cones contain a dynamic lamellipodial body ringed by filopodial protrusions, both important in guiding the axon to its target destination [1][2][3][4]. Guidance receptors present on the leading edge of the growth cone sense and respond to various extracellular guidance cues, which attract or repel axons enabling them to reach their proper target destination [5,6].
In C. elegans, UNC-6/Netrin is secreted by the ventral cells and along with its receptors UNC-40 and UNC-5 is required for the dorsal ventral guidance of circumferential neurons and axons [8,13,14]. Previous studies of repelled VD growth cones in Netrin signaling mutants revealed a correlation between attractive axon guidance and stimulation of growth cone protrusion, and repulsive axon guidance and inhibition of growth cone protrusion [15]. For example, in unc-5 mutants, growth cones were larger and more protrusive, and often displayed little or no directed movement. This is consistent with observation that increased growth cone size was associated with decreased neurite growth length [16]. Conversely, constitutive activation of UNC-5:UNC-40 signaling in repelled VD growth cones led to smaller growth cones with severely reduced filopodial protrusion [15,17]. Thus, directed growth cone repulsion away from UNC-6/Netrin requires a balance of pro-and anti-protrusive activities of the receptors UNC-40 and UNC-40:UNC-5, respectively, in the same growth cone [15]. Genetic analysis has identified a cytoskeletal signaling pathway involved in stimulation of growth cone protrusion in response to the attractive UNC-40 signaling that includes CDC-42, the Rac-specific guanine nucleotide exchange factor TIAM-1, the Rac-like GTPases CED-10 and MIG-2, as well as the cytoskeletal regulators Arp2/3 and activators WAVE-1 and WASP-1, UNC-34/Enabled, and UNC-115/abLIM [18][19][20][21][22][23], consistent with findings in other systems [7]. Mechanisms downstream of UNC-5 in axon repulsion are less well described, but the PH/ MyTH4/FERM molecule MAX-1 and the SRC-1 tyrosine kinase have been implicated [24,25]. We delineated a new pathway downstream of UNC-5 required for its inhibitory effects on growth cone protrusion, involving the Rac GEF UNC-73/Trio, the Rac GTPases CED-10 and MIG-2, and the cytoskeletal-interacting molecule UNC-33/CRMP [17].
Collapsin response mediating proteins (CRMPs) were first identified as mediators of growth cone collapse in response to the Semaphorin family of guidance cues [26], and we have shown that UNC-33/CRMP inhibits growth cone protrusion in response to Netrin signaling [17]. This motivated us to consider other mediators of Semaphorin-induced growth cone collapse in Netrin signaling. In Drosophila, the large multidomain cytosolic protein MICAL (Molecule Interacting with CasL) is required for the repulsive motor axon guidance mediated by interaction of Semaphorin 1a and Plexin A [27,28]. MICAL proteins are a class of flavoprotein monooxygenase enzymes that bind flavin adenine dinucleotide (FAD) and use the cofactor nicotinamide dinucleotide phosphate (NADPH) to facilitate oxidation-reduction (Redox) reactions [27]. MICAL regulates actin disassembly and growth cone collapse in response to semaphorin via direct redox interaction with F-actin [29,30]. MICAL molecules from Drosophila to vertebrates have a conserved domain organization: an N-terminal flavin-adenine dinucleotide (FAD)-binding monooxygenase domain, followed by a calponin homology (CH) domain, a LIM domain, a proline-rich domain, and a coiled-coil ERM α-like motif [27,31].
The C. elegans genome does not encode a MICAL-like molecule with the conserved domain organization described above. However, it does encode five flavin monooxygenase (fmo) genes similar to the Flavin monooxygenase domain of MICAL: fmo-1, fmo-2, fmo-3, fmo-4 and fmo-5 [32]. Like MICAL, the C. elegans FMO molecules contain an N-terminal FAD binding domain and a C-terminal NADP or NADPH binding domain [27,32]. The C. elegans gene most similar to the non-FMO portion of MICAL is the Eps-15 homology domain binding protein EHBP-1 [33], which contains a CH domain as does MICAL.
In this work, we test the roles of the C. elegans FMOs and EHBP-1 in Netrin-mediated axon guidance and growth cone protrusion. We find that fmo-1, fmo-4, fmo-5 and ehbp-1 mutants display pathfinding defects of the dorsally-directed VD/DD motor neuron axons that are repelled by UNC-6/Netrin, and that they interact genetically with unc-40 and unc-5. We also find that VD growth cones in these mutants display increased filopodial protrusion, similar to mutants in repulsive UNC-6/Netrin signaling (e.g. unc-5 mutants), and that transgenic expression of FMO-5 inhibits growth cone protrusion, similar to constitutively-activated UNC-40 and UNC-5. We also show that FMO-1, FMO-4, FMO-5 and EHBP-1 are required for the growth cone inhibitory effects of activated UNC-5, UNC-40, and the Rac GTPases CED-10 and MIG-2. Together, these genetic analyses suggest that FMO-1, FMO-4, FMO-5, and EHBP-1 normally restrict growth cone protrusion, and that they might do so in UNC-6/ Netrin-mediated growth cone repulsion.

Genetic methods
Experiments were performed at 20˚C using standard C. elegans techniques [34]. Mutations used were LGI: unc- 40 -6]. Multiple (!3) extrachromosomal transgenic lines of Pfmo-5::fmo-5 for overexpression data of fmo-5 were analyzed with similar effect, and one was chosen for integration and further analysis. Genotypes containing M+ indicate that homozygous animals from a heterozygous mother were scored. The ehbp-1(ok2140M+) strain was balanced with the nT1 balancer, since ehbp-1(ok2140) homozygous animals are sterile. RNAi was performed by feeding as previously described [36].

Analysis of axon guidance defects
VD neurons were visualized with a Punc-25::gfp transgene, juIs76 [37], which is expressed in GABAergic neurons including the six DDs and 13 VDs, 18 of which extend commissures on the right side of the animal. The commissure on the left side (VD1) was not scored. In wildtype, an average of 16 of these 18 VD/DD commissures are apparent on the right side, due to fasciculation of some of the commissural processes. In some mutant backgrounds, fewer than 16 commissures were observed (e.g. unc-5). In these cases, only observable axons emanating from the ventral nerve cord were scored for axon guidance defects. VD/DD axon defects scored include axon guidance (termination before reaching the dorsal nerve cord or wandering at an angle greater than 45˚before reaching the dorsal nerve cord) and ectopic branching (ectopic neurite branches present on the commissural processes). In the case of double mutant analysis with unc-40 and unc-5 only lateral midline crossing (axons that fail to extend dorsally past the lateral midline) were considered. Fisher's exact test was used to determine statistical significance between proportions of defective axons. In double mutant comparisons, the predicted additive effect of the mutants was calculated by the formula P1+P2-(P1 Ã P2), where P1and P2 are the phenotypic proportions of the single mutants. The predicted additive effect of single mutants was used in statistical comparison to the observed double mutant effect.

Growth cone imaging
VD growth cones were imaged as previously described [15,22]. Briefly, animals at 16 h posthatching at 20˚C were placed on a 2% agarose pad and paralyzed with 5mM sodium azide in M9 buffer, which was allowed to evaporate for 4 min before placing a coverslip over the sample. Some genotypes were slower to develop than others, so the 16 h time point was adjusted for each genotype. Growth cones were imaged with a Qimaging Rolera mGi camera on a Leica DM5500 microscope. Projections less than 0.5 µm in width emanating from the growth cone were scored as filopodia. Filopodia length and growth cone area were measured using ImageJ software. Filopodia length was determined by drawing a line from the base where the filopodium originates on the edge of the peripheral membrane to the tip of the filopodium. Growth cone area was determined by tracing the periphery of the growth cone, not including filopodial projections. Significance of difference was determined a two-sided t-test with unequal variance.
The 19 D-class motor neurons cell bodies reside in the ventral nerve cord. They extend axons anteriorly and then dorsally to form a commissure, which normally extend straight dorsally to the dorsal nerve cord (Fig 2 and Fig 3B). On the right side of wild-type animals, an average of 16 commissures were observed, due to the fasciculation of some processes as a single commissure (Fig 2C and Materials and Methods). fmo-1, 4, and 5 and ehbp-1 mutants showed significant defects in VD/DD axon pathfinding, including ectopic axon branching and wandering (~3-5%; see Materials and Methods and Fig 3A, 3C and 3D), although most crossed the lateral midline despite wandering. fmo-2 and fmo-3 mutations showed no significant defects compared to wild-type ( Fig 3A). We used RNAi directed against ehbp-1, which has the potential to eliminate any maternally-supplied mRNAs, but not translated proteins. ehbp-1 (RNAi) resembled ehbp-1 M+ mutants (Fig 3A), suggesting that maternal mRNAs are not involved in VD/DD axon guidance.
Most double mutants showed no strong synergistic defects compared to the predicted additive effects of the single mutants ( Fig 3A). However, the fmo-2; fmo-3 and the fmo-2; fmo-4 double mutants showed significantly more defects compared to the predicted additive effects of the single mutants. The fmo-4; ehbp-1 double mutant displayed significantly reduced defects than either mutation alone. The fmo-1; fmo-4 fmo-5 triple mutant also showed no synergistic defects as compared to single mutants alone ( Fig 3A). Lack of extensive phenotypic synergy suggests that the FMOs do not act redundantly, but rather that they might have discrete and complex roles in axon guidance, as evidenced by fmo-4; ehbp-1 mutual suppression.

fmo-1, fmo-4 and fmo-5 act cell-autonomously in the VD/DD neurons
Expression of the fmo-1, fmo-4, and fmo-5 coding regions were driven in VD/DD motor neurons using the unc-25 promoter. Punc-25::fmo transgenes significantly rescued lateral midline In the early L2 larval stage, the VD neurons (green) extend axons anteriorly in the ventral nerve cord after which the axons turn dorsally and migrate to the dorsal nerve cord to form commissures. Only two of the 13 VD neurons are shown. While migrating towards the dorsal nerve cord, VD growth cones display an extended, protrusive morphology with highly dynamic filopodial protrusions (VD8). VD7 shows the final structure of the VD neurite. (B) Fluorescent micrograph of an early L2 larval wild-type commissure indicated by an arrow, and a VD growth cone indicated by an arrowhead. CB, cell body; DNC, dorsal nerve cord; and VNC, ventral nerve cord. Scale bar represents 5μm. (C) Diagram of an L4 hermaphrodite after all the VD axon outgrowth is complete. The 18 commissures on the right side of the animal are shown (black lines), and axon guidance defects of these commissures were scored. One commissure (VD1) extends on the right side and was not scored. Of the 18 commissures on the right side, two (DD1 and VD2) extend as a single fascicle. Others pairs occasionally extended as single fascicles as well, resulting in an average of 16 observable commissures per wild-type animal.
Previous studies showed that fmo-1 and fmo-5 promoter regions were active in intestinal cells and the excretory gland cell, whereas the fmo-4 promoter was active in hypodermal cells, duct and pore cells [32,40]. ehbp-1 is expressed in all somatic cells including neurons [33]. Furthermore, cell-specific transcriptome profiling indicated that fmo-1, fmo-4 and fmo-5 were expressed in embryonic and adult neurons, including motor neurons [41][42][43]. We fused the upstream promoter regions of fmo-1, fmo-4, and fmo-5 to gfp. We could observe no fmo-1::gfp expression in transgenic animals, in contrast to previous studies using a LacZ reporter [32]. However, transcriptome profiling indicates neuronal expression of fmo-1 [43]. Our fmo-1::gfp transgene might be missing regulatory regions required for expression. fmo-4::gfp was expressed strongly in hypodermal cells, excluding the seam cells and vulval cells, consistent with previous studies [32] (Fig 6D). We also observed fmo-4::gfp expression in cells in the ventral nerve cord (Fig 6D and 6D'). Pfmo-5::fmo-5::gfp was expressed strongly in the intestine as previously reported [32] (Fig 6E). We also observed expression along the ventral nerve cord ( Fig  6E and 6E'). In sum, previous expression studies combined with those described here suggest that fmo-1,4,5 and ehbp-1 are expressed in neurons, and that fmo-1,4, and 5 can act cell-autonomously in the VD/DD motor neurons in axon guidance. The growth cones of dorsally-directed VD commissural axons are apparent in early L2 larvae (Fig 2B). We imaged VD growth cones at 16 hours post-hatching, when the VD growth cones have begun their dorsal migrations, as described previously [15]. fmo-1, fmo-4, and fmo-5 mutant growth cones displayed longer filopodial protrusions compared to wild type (e.g. 0.96 µm in wild type compared with 1.55 µm in fmo-5(tm2438); p < 0.001) (Fig 7). This effect was not significant in ehbp-1(ok2140) (Fig 7). Growth cone area was not significantly different in any mutant. These results suggest that fmo-1, fmo-4 and fmo-5 normally limit growth cone filopodial protrusion length. This is consistent with ectopic axon branches observed in post-development VD/DD neurons in these mutants (Fig 3), as other mutants with increased growth cone filopodial protrusions (e.g. unc-5, unc-73, unc-33) also display ectopic branches, likely due to failure of filopodial retraction and subsequent consolidation into a neurite [15,17].
Expression of the fmo-5 coding region driven in VD/DD motor neurons using the unc-25 promoter also significantly rescued axon guidance defects as well as the long filopodial protrusions seen in fmo-5(tm2438) (Fig 8). We expressed the fmo-5 coding region in the hypodermis using the dpy-7 promoter [44] and observed no significant rescue of axon guidance defects or filopodial protrusions (Fig 8). These data confirm that fmo-5 can act cell-autonomously in the VD/DD neurons in axon guidance and growth cone filopodial protrusion.

FMO-5 can inhibit growth cone protrusion
fmo-5 loss-of-function mutant growth cones displayed excessively-protrusive filopodia (Fig 7) and suppressed activated UNC-5:UNC-40 and Rac signaling (Figs 9-11). Transgenic expression of wild-type FMO-5 driven by its endogenous promoter rescued the axon guidance defects and long filopodial protrusions seen in fmo-5(tm2438) mutant VD growth cones ( Fig  12A-12E). In a wild-type background, fmo-5 transgenic expression resulted in growth cones with smaller area and shortened filopodia (Fig 13A, 13B and 13E), indicating that wild-type FMO-5 activity can inhibit growth cone protrusion. This inhibition was not observed in the fmo-5(tm2438) background, possibly due to the decreased levels of FMO-5 compared to the wild-type background.
In contrast to unc-5 mutants, unc-40 single mutants display shortened filopodial protrusions and a relatively normal growth cone size (Fig 13A and 13B) [15]. This is likely due to the dual role of UNC-40 in both stimulating protrusion as a homodimer and inhibiting protrusion as a heterodimer with UNC-5. fmo-5 transgenic expression had no effect on filopodial protrusions in unc-40, but did reduce growth cone size, consistent with a role of FMO-5 in inhibiting protrusion.

Discussion
Results here implicate the C. elegans flavoprotein monooxygenase molecules FMO-1, FMO-4 and FMO-5 in inhibition of growth cone protrusion via UNC-6/Netrin receptor signaling in repulsive axon guidance (Fig 14). The MICAL molecule found in vertebrates and Drosophila is a flavoprotein monooxygenase required for semaphorin-plexin mediated repulsive motor axon guidance [27,45]. Here we focus on UNC-6/Netrin signaling and have not analyzed the role of the FMOs in semaphorin signaling in C. elegans. MICAL is a multi-domain molecule that also includes a calponin homology (CH) domain, a LIM domain and multiple CC domains. No molecule encoded in the C. elegans genome has a similar multi-domain organization. However, the Eps-15 homology (EH) domain binding protein EHBP-1 is similar to the non-FMO portion of MICAL and contains a CH domain [33]. We show here that EHBP-1 also is also involved in inhibition of growth cone protrusion and axon guidance. Thus, while C. elegans does not have a multidomain MICAL-like molecule, it is possible that the functional equivalents are the FMOs and EHBP-1. FMO-1, FMO-4, FMO-5 and EHBP-1 regulate axon guidance and growth cone filopodial protrusion fmo-1, fmo-4, fmo-5, and ehbp-1 mutants display defects in dorsal guidance of the VD/DD motor axons that are repelled from UNC-6/Netrin (Fig 3). Double and triple mutant analysis did not uncover significant redundancy, suggesting that these molecules might have discrete and complex roles in axon guidance. Consistent with this idea, fmo-4 and ehbp-1 mutually suppress VD/DD axon guidance defects. Furthermore, transgenic expression of FMO-5 rescued excess growth cone and filopodial protrusions of fmo-1 mutants. This suggests that FMO-5 can partially compensate for loss of FMO-1, and that the function of FMO-5 does not depend on FMO-1. Combined with lack of phenotypic synergy, these data suggest that the FMOs act in a common pathway, where loss of one abolishes pathway function, and that FMO-5 might act downstream of FMO-1 in this pathway. fmo-2 and fmo-3 mutations displayed no significant defects alone, suggesting that they are not involved in axon guidance. fmo-2 did significantly enhance fmo-4. Possibly, fmo-2 and fmo-3 might have roles in axon guidance that were not revealed by the mutations used.
Drosophila and vertebrate MICAL regulate actin cytoskeletal dynamics in both neuronal and non-neuronal processes through direct redox activity of the monooxygenase domain [27,30,[46][47][48][49][50]. In Drosophila, loss of MICAL showed abnormally shaped bristles with disorganized and larger F-actin bundles, whereas, overexpression of MICAL caused a rearrangement of Factin into a complex meshwork of short actin filaments [29]. Here we show that loss of fmo-1, fmo-4, and fmo-5 resulted in longer filopodial protrusions in the VD motor neurons (Fig 7), suggesting that their normal role is to limit growth cone filopodial protrusion. Indeed, transgenic expression of wild-type FMO-5 resulted in VD growth cones with a marked decrease in growth cone filopodial protrusion (Fig 13). Growth cone size was not affected in any loss-offunction mutation, but growth cone size was reduced by transgenic expression of wild-type FMO-5 (Fig 13), suggesting a role of the FMO-5 in both filopodial protrusion and growth cone lamellipodial protrusion. Previous studies have shown that Drosophila MICAL may require both its FMO and CH domain to induce cell morphological changes; however, mammalian MICAL in non-neuronal cell lines requires only its FAD domain suggesting a difference in the mechanism of action in these MICALs [29,51]. These data suggest that in some cases, the FMO domain is sufficient for the function of MICAL. Thus, single domain FMOs as in C. elegans could function despite lacking the multi-domain structure of MICAL. Loss of the C. elegans MICAL-like molecule EHBP-1, which contains a CH domain and is similar to the non-FMO portion of MICAL (Fig  1), also resulted in VD/DD axon guidance defects, but did not significantly affect growth cone filopodial protrusion. EHBP-1 might act with the FMOs in axon guidance. Phenotypic differences could be due to EHBP-1-dependent and independent events, or to the wild-type maternal contribution in ehbp-1 homozygous mutants derived from a heterozygous mother. It is also possible that EHBP-1 affects axon guidance independently of the FMOs. EHBP-1 is involved in Rab-dependent endosomal vesicle trafficking by bridging interaction of endosomal Rabs with the actin cytoskeleton [33,52]. MICAL has also been implicated in Rab-dependent endosomal biogenesis and trafficking [53][54][55], suggesting that FMO/EHBP-1 and MICALs might share common functions, although it remains to be determined if FMOs in C. elegans regulate endosomal trafficking.
MICAL has been shown to directly oxidize cysteine residues in F-actin, leading to actin depolymerization and growth cone collapse [29,30,56,57]. We speculate that FMO-1, FMO-4, and FMO-5 might act by a similar mechanism to inhibit growth cone filopodial protrusion. Previous studies have shown that the single calponin homology (CH) domain containing protein CHDP-1 promotes the formation of cell protrusions in C. elegans by directly binding to Rac1/CED-10 through its CH domain [58]. The role of EHBP-1 however, is less clear, but previous studies have shown that Drosophila MICAL might require both its FMO and CH domain to induce cell morphological changes [29]. Thus, in axon guidance, FMO-1, FMO-4, and FMO-5 might require the CH domain provided by EHBP-1 in some instances. Mammalian MICAL requires only the FMO domain [51], suggesting that in some cases the CH domain is not required and the FMO domain can act alone. Future studies will be directed at answering these questions.

FMOs can act autonomously in the VD/DD neurons
Expression of full length fmo-1, fmo-4 and fmo-5 coding regions under the control of the unc-25 promoter specific for GABA-ergic neuron expression (including the VD/DD neurons) rescued VD/DD axon guidance defects (Figs 6 and 8). Furthermore, the promoters of fmo-4 and fmo-5 were active in ventral nerve cord cells (Fig 6). Expression of full length fmo-5 coding region under the control of the unc-25 promoter rescued axon guidance defects as well as the long filopodial protrusions seen in fmo-5(tm2438), whereas expression from the hypodermal dpy-7 promoter did not (Fig 8). Cell-specific transcriptome profiling indicated that fmo-1, fmo-4 and fmo-5 were expressed in embryonic and adult neurons, including motor neurons [41][42][43]. Together, these results suggest that the FMOs can act cell-autonomously in the VD/ DD neurons in axon guidance and growth cone filopodial protrusion.

FMO-1, FMO-4 and FMO-5 mediate UNC-6/Netrin receptor signaling in growth cone inhibition of protrusion
Our findings suggest that the FMOs act with the UNC-40 and UNC-5 receptors to mediate UNC-6/netrin repulsive axon guidance and inhibition of growth cone protrusion. fmo-1, fmo-4, and fmo-5 mutations enhanced axon pathfinding defects in unc-40 and hypomorphic unc-5 mutants (Figs 4 and 5). The axon guidance defects of the fmos were weaker than those of unc-40 and unc-5 mutants (e.g. the fmos displayed few lateral midline crossing defects despite axon wandering). We speculate that the FMOs are but one of several pathways mediating the effects of UNC-40 and UNC-5 in axon pathfinding. ehbp-1 did not enhance unc-40 or unc-5, suggesting discrete roles of these molecules or wild-type maternal ehbp-1 contribution. fmo-1, fmo-4, fmo-5, and ehbp-1 mutations each suppressed the effects of activated MYR::UNC-40 and MYR::UNC-5 on inhibition of growth cone protrusion (Fig 10). In this case, both filopodial protrusion and growth cone area was restored, consistent with a role of these molecules in inhibiting both growth cone filopodial and lamellipodial protrusion. We also find the fmo-5 transgenic expression suppressed unc-5(e152) growth cone area and filopodial protrusions (Fig 13). That the FMOs and EHBP-1 were required for the effects of the constitutively active MYR::UNC-40 and MYR::UNC-5 suggest that they act downstream of these molecules in growth cone inhibition of protrusion. While the loss-of-function and gain-of-function data are consistent with acting downstream of UNC-40 and UNC-5, it is possible that the FMOs define a parallel pathway in growth cone protrusion.
Interestingly, mutations in these genes have very distinct penetrances (e.g. the axon guidance and protrusion defects of unc-5 are much stronger than those of the fmo mutants). One explanation for this is that these molecules act in networks rather than simple linear pathways. The FMOs might be one of many mechanisms acting downstream of UNC-5, and multiple pathways might converge on UNC-33 (e.g. UNC-5 and UNC-33 are major "nodes" in this network). Further loss-and gain-of-function studies will be required to understand this signaling network.

FMO-5 interacts with UNC-33/CRMP
Previous studies have shown that the C. elegans CRMP-like molecule UNC-33 is required in a pathway downstream of Rac GTPases for inhibition of growth cone protrusion in response to UNC-6/Netrin [17]. unc-33 loss-of-function mutants with FMO-5 transgenic expression displayed a mutually-suppressed phenotype. The excessively-long filopodial protrusions of unc-33 mutants were reduced to wild-type levels, but were significantly longer than in animals with FMO-5 transgenic expression, and the growth cone area was reduced to resemble FMO-5 transgenic expression alone (Fig 13). This hybrid phenotype makes it difficult to determine if FMO-5 and UNC-33 act in the same pathway, in parallel pathways, or both.
One proposed mechanism of cytoskeletal regulation by MICAL is the production of the reactive oxygen species (ROS) H 2 O 2 by the FAD domain in the presence of NADPH [59].
Upon activation by Sema3A, MICALs generate H 2 O 2 , which can, via thioredoxin, promote phosphorylation of CRMP2 via glycogen synthase kinase-3, leading to growth cone collapse [60]. Thus, the FMOs have the potential to inhibit growth cone protrusion through direct oxidation of F-actin resulting in depolymerization, and through redox regulation of the activity of UNC-33/CRMP (i.e. to act both in the UNC-33 pathway and in parallel to it).

Conclusion
In summary, we present evidence of a novel role of the C. elegans flavin-containing monooxygenase molecules (FMOs) in inhibition of growth cone protrusion downstream of UNC-6/ Netrin signaling. The FMOs acted downstream of the UNC-6/Netrin receptors UNC-5 and UNC-40, and downstream of the Rac GTPases CED-10 and MIG-2 (Fig 14). Future studies will determine if the FMOs regulate UNC-33/CRMP, if they cause actin depolymerization, or both, to inhibit growth cone protrusion.