Differential regulation of actin microfilaments by human MICAL proteins

Actin plays a key role in maintaining cell shape, cell motility, chemotaxis and a multitude of crucial cellular functions including cytokinesis, signal transduction and endocytosis (Sutherland and Witke, 1999). Regulation of the axonal growth cone and the axonal guidance machinery is one such actin-regulated process that has been studied extensively over the past decade (Pak et al., 2008). Growth cones are highly motile sensory structures localized to the tip of the axons and essential for the guidance of neurons to their synaptic targets during embryonic development. Numerous soluble, matrix and cell-bound ligands mediate axonal guidance through binding to receptors at the tip of the axon or growth cone (Kalil and Dent, 2005). Upon binding to ligands, these receptors create localized changes in filamentous actin, which in turn modulate microtubule dynamics and thus induce changes in axonal direction.

Axons use both attractive and repulsive cues to accurately establish neuronal circuits. Semaphorins are one of the largest classes of guidance cues involved in axonal pathfinding, branching, fasciculation and neuronal polarity (He et al., 2002; Raper, 2000). Semaphorins bind to plexin receptors and modify the growth cone cytoskeleton to mediate repulsive signaling. Sema1A is a well-characterized transmembrane semaphorin involved in repulsive axonal guidance (Yu et al., 1998). When Sema1A binds to plexinA on neurons, this association promotes depolymerization of actin filaments. The mechanism of Sema1A-mediated axonal repulsion was clarified after the identification of MICAL (Molecules Interacting with CasL) in the fruit fly Drosophila melanogaster (Terman et al., 2002). Drosophila MICAL interacts with the cytoplasmic region of plexin and is required for pathfinding of motor axons.

Although MICAL was identified originally in mammals, its function has been studied primarily in Drosophila. Drosophila MICAL^−/− flies displayed abnormally shaped bristles, whereas overexpression of MICAL in wild-type Drosophila caused the bristles to branch (Hung et al., 2010). Similar to neuronal extension by axonal guidance, the process of bristle elongation is also dependent on actin dynamics (Sutherland and Witke, 1999). Bristles of MICAL^−/− flies had disorganized, intersecting and larger F-actin bundles, whereas in Drosophila overexpressing MICAL, bristles displayed a rearrangement of F-actin into a complex meshwork of short actin filaments (Hung et al., 2010). Moreover, MICAL directly induced actin depolymerization and significantly decreased the levels of actin filaments in vitro (Hung et al., 2010). These studies implicated MICAL as a direct effector of F-actin. Accordingly, Drosophila MICAL is likely to function downstream of semaphorin to cause actin destabilization and thus play an important role in repulsive axon guidance. Indeed, further evidence for this model is supplied by a study demonstrating that Sox14, a transcription factor necessary and sufficient to mediate dendrite severing, mediates dendrite pruning by directly regulating the expression of MICAL (Kirilly et al., 2009). Drosophila MICAL mutants also affect neuromuscular junctions, causing patterning and arrangement defects of synaptic boutons at the distal axonal termini (Beuchle et al., 2007).

Drosophila MICAL has been extensively studied; however, the roles of the mammalian MICAL proteins have not been well characterized. Human MICAL proteins have four conserved domains: an N-terminal flavin adenine dinucleotide (FAD) binding domain, a calponin homology (CH) domain, a Lin11, Isl-1 and Mec-3 (LIM) domain and a C-terminal coiled-coil (CC) domain (Fig. 1A) (reviewed by Hung and Terman, 2011; Zhou et al., 2011a). MICAL1 has the most closely related domain architecture to Drosophila MICAL, whereas MICAL3 displays the least homology (Fig. 1B). Both MICAL1 and Drosophila MICAL have proline-rich regions that are required for binding to SH3 domains. However, unlike MICAL1, the CH and LIM domains of MICAL2 are separated by approximately 380 residues, and MICAL2 lacks a recognizable C-terminal CC domain.

It was demonstrated that both MICAL1 and MICAL2 interact with plexins (Terman et al., 2002), and that MICAL1 also interacts with collapsin response mediator proteins (CRMPs), which function downstream of plexin signaling (Schmidt et al., 2008). MICAL1, MICAL2 and MICAL3 are expressed in the embryonic postnatal and adult nervous systems, suggesting a role in neuronal growth cone machinery (Pasterkamp et al., 2006). These studies imply that mammalian MICALs could potentially function in actin remodeling, like Drosophila MICAL. To date, it remains unknown whether mammalian MICALs have any role apart from their function as neurite out-growth regulators. In particular, little is known about the expression and potential role(s) of human MICAL proteins in non-neuronal cells. In this study, we address the function of the human MICAL proteins that are expressed in non-neuronal cells, and provide evidence for a mechanism describing their differential regulation of actin microfilaments.

To date, isolated studies have addressed the expression of MICAL proteins in neural cells, but even fewer studies have been performed in non-neural cells and tissues (Schmidt et al., 2008; Suzuki et al., 2002). We analyzed the expression of MICAL1 in a variety of both neural and non-neural cell lines, such as HeLa, retinal pigment epithelium (RPE), SKNMC neuroblastoma cells, squamous cell carcinoma (SCC), Caco-2 colon carcinoma cells, A431, LnCap prostate cancer cells and human foreskin fibroblast cells (Fig. 1C,D). As depicted, MICAL1 protein expression was detected in all of these cells. Moreover, the specificity of the MICAL1 antibodies could be verified by showing decreased MICAL1 immunoreactivity in MICAL1-depleted cells (Fig. 1D). On the other hand, utilizing both commercial antibodies and antisera generated in our laboratory, we were unable to detect MICAL2 or MICAL3 proteins in any of these cell lysates, although the antibodies we generated did recognize a band corresponding to overexpressed MICAL2 (our unpublished observations). To determine whether MICAL2 and MICAL3 mRNA is expressed in non-neural cells, we performed RT-PCR with several primers designed for MICAL2 and several isoforms of MICAL3. We found that MICAL2 mRNA is indeed expressed in HeLa cells, as are mRNAs encoding MICAL1 and actin (supplementary material Fig. S1A). In addition, we could detect mRNAs for both MICAL3 isoform1 and MICAL3 isoform3 (supplementary material Fig. S1B). Previously, it was demonstrated that splice variants of MICAL2 were expressed in prostate cancer cells (Ashida et al., 2006). By designing various primers against the different regions of MICAL2 we were able to verify the isoform that is expressed in HeLa cells, namely that corresponding to accession number NM_014632.2.

Because overexpression of MICAL in Drosophila caused the normal, parallel organization of bundled F-actin in the bristles into a complex meshwork of short actin filaments (Hung et al., 2010), we hypothesized that expression of exogenous MICALs in mammalian cells might influence actin organization. Accordingly, we began by transfecting hemagglutinin (HA)-tagged MICAL2 into HeLa cells (Fig. 2A–C). The exogenous MICAL2 localized to the plasma membrane, and its expression correlated with the loss of actin stress fibers as well as a concentration of actin in filopodia-like protrusions at the cell periphery (Fig. 2A–C and quantified in 2P; arrows denote actin at the cell periphery in actin-rich protrusions).

To delineate the domain(s) responsible for actin rearrangement, we performed a series of truncations (Fig. 2D–O). Removal of either the C-terminal LIM domain (Fig. 2D–F) or the LIM domain along with the linker region between the LIM and CH domains (Fig. 2G–I) did not prevent actin stress fiber loss upon transfection into cells. Moreover, we observed that the MICAL2 FAD domain is both required and sufficient for actin stress fiber loss (Fig. 2J–L). Instituting conserved glycine to tryptophan point mutations within the Drosophila MICAL FAD domain was previously shown to impair FAD binding and enzymatic activity (Terman et al., 2002). When a similar mutation was generated in the FAD domain of MICAL2, expression of this mutant FAD domain failed to induce the loss of actin stress fibers (Fig. 2M–O). The percentage of cells transfected with HA–MICAL2 that displayed loss of identifiable actin stress fibers (compared with 100% for untransfected control cells) was quantified for HA–MICAL2 and the MICAL2 mutants (Fig. 2P). In addition, quantification was carried out to calculate the percentage of F-actin in the transfected cells compared with untransfected cells (Fig. 2P). As observed, although several mutants displayed mild reductions in overall F-actin levels (for example, MICAL2 FAD + CH domains was ~80% compared with 100% in untransfected cells), these differences were relatively minor compared with the massive numbers of transfected cells displaying loss of stress fibers.

Unlike MICAL2, wild-type MICAL1 overexpression had no discernable effect on F-actin stress fiber loss or total F-actin compared with untransfected cells (Fig. 3A–C; quantified in Fig. 3S). Indeed, unlike MICAL2, the overexpressed MICAL1 was largely cytoplasmic in its distribution. Surprisingly, truncation of the C-terminal MICAL1 CC region led to a dramatic reduction in the percentage of cells exhibiting actin stress fibers as well as decreased levels of overall F-actin when overexpressed (Fig. 3D–L; quantified in Fig. 3S). Unexpectedly, in the absence of the last 268 amino acids (including the CC), MICAL1 exhibited a partial nuclear localization (Fig. 3D). Although the reason for this remains unclear, this mutant was nonetheless capable of inducing a considerable decrease in the percentage of cells displaying stress fibers along with a decrease in total F-actin (Fig. 3D–F; quantified in Fig. 3S). Expression of MICAL1 with a truncation prior to the LIM domain (MICAL1 FAD+CH) again resulted in a cytoplasmic distribution of the mutant protein, and similarly resulted in a radical decrease in the percentage of cells with actin stress fibers and the amount of total F-actin per cell (Fig. 3G–I; quantified in Fig. 3S). As demonstrated for MICAL2, expression of an active MICAL1 FAD domain was both required and sufficient for loss of stress fibers and decreased total actin, as indicated by the dramatic loss of microfilaments induced by MICAL1 FAD expression (Fig. 3J–L; quantified in Fig. 3S). Moreover MICAL1 FAD-transfected cells also displayed a decrease in the number of focal adhesion plaques (supplementary material Fig. S2). Similar to our results with MICAL2, a mutant predicted to display impaired FAD binding and reduced enzymatic activity prevented stress fiber loss and F-actin depletion (Fig. 3M–O; quantified in Fig. 3S).

Given that the presence of the MICAL1 CC domain appeared to be crucial for the prevention of stress fiber loss upon MICAL1 overexpression, we hypothesized that expression of a MICAL1 mutant with a mutated CC might lead to stress fiber loss, as observed with wild-type MICAL2 expression. As demonstrated by the Paircoil algorithm in supplementary material Fig. S3 (http://groups.csail.mit.edu/cb/paircoil/cgi-bin/paircoil.cgi), replacement of alanines 940 and 941 with proline residues was predicted to interfere with generation of the CC. Indeed, overexpression of this mutant revealed that in the absence of a functional CC, MICAL1 expression induced a loss of actin stress fibers and decreased the cellular F-actin levels (Fig. 3P–R; quantified in Fig. 3S). These data suggest that MICAL2 constitutively controls stress fibers via its FAD domain, whereas the MICAL1 FAD domain is normally inhibited via its CC domain.

Next, we directly addressed the hypothesis that the MICAL1 protein is functionally different from the constitutively active Drosophila MICAL and MICAL2 due to the presence of its inhibitory CC domain. For this, we constructed a pair of chimeric proteins: the N-terminal MICAL1 FAD and CH domains followed by the central region and LIM domain of MICAL2 (MICAL1/MICAL2), and the MICAL2 FAD and CH domains fused to the LIM and CH domains of MICAL1 (MICAL2/MICAL1) (Fig. 4A). In cells expressing HA–MICAL1/MICAL2, we observed a decrease in the number of stress fibers and the appearance of actin-rich protrusions (Fig. 4B–D; see outlined cells, insets and arrows). However, cells expressing HA–MICAL2/MICAL1 did not display decreased levels of actin stress fibers (Fig. 4E–G). At lower expression levels, HA–MICAL2/MICAL1 partially localized to filamentous structures that displayed minimal colocalization with F-actin (supplementary material Fig. S4). These results support the idea that the FAD domains of MICAL1 and MICAL2 are both potent actin stress fiber regulators, and that the MICAL1 CC region is responsible for an auto-inhibitory mechanism that prevents FAD function, the depletion of stress fibers and the formation of actin-rich structures at the cell periphery.

Given the dramatic effects of MICAL2 and MICAL1 FAD overexpression on actin stress fibers, we carried out a series of studies to determine whether expression of these proteins had more widespread effects on the cells. As demonstrated in supplementary material Fig. S5, despite the potent effects on stress fibers, these proteins had little effect on early endosomes (EEA1; supplementary material Fig. S5A), the Golgi (Giantin; supplementary material Fig. S5B), and lysosomes (our unpublished observations). We did observe some reorientation of α-tubulin to actin-rich structures upon high levels of HA–MICAL1 FAD overexpression (supplementary material Fig. S5C), as well as centralization of γ-tubulin in HA–MICAL1 FAD-transfected cells (supplementary material Fig. S5D). However, little change was observed for Arp-3, myosin Vb or myosin Vc (our unpublished observations) upon transfection of either HA–MICAL2 or the MICAL1 FAD domain, indicating that the effects are highly specific for actin stress fibers and that there is little impact on endocytic organelles, actin nucleation via the ARP2/3 complex, and actin motors.

Studies on Drosophila MICAL indicate that its depletion affects actin architecture and bristle morphology (Hung et al., 2010). Accordingly, we next assessed the effect of MICAL1 and MICAL2 depletion on cellular F-actin. As demonstrated, RNA interference (RNAi) was successful in depleting >90% of endogenous MICAL1 (Fig. 5D) and >70% of exogenously expressed MICAL2 (Fig. 5E). In contrast to mock-treated cells (Fig. 5A), those lacking either MICAL1 or MICAL2 displayed radically altered F-actin morphology, with the loss of prominent stress fibers, enhanced levels of cortical actin, an enrichment of actin puncta in the cytoplasm, and F-actin concentrated in actin-rich protrusions (Fig. 5B,C; quantified in Fig. 6E). In addition, depletion of MICAL3 (as reflected by loss of mRNA for the MICAL3 isoform3 and modestly decreased expression of the MICAL3 isoform1; supplementary material Fig. S1B) induced enhanced levels of stress fibers per cell, and more elongated cells were observed (supplementary material Fig. S6). These results indicate that mammalian MICALs play an important role in the organization of actin stress fibers.

To verify that the effects of MICAL protein knockdown are directly related to the loss of protein expression (and are not off-target effects), we instituted a ‘rescue’ system in which siRNA-resistant MICAL1 (with silent mutations in the cDNA) was reintroduced into depleted cells. As demonstrated (Fig. 6), the expression of HA–MICAL1 could be efficiently depleted with siRNA, and a siRNA-resistant HA–MICAL1 could be expressed in siRNA-treated cells (Fig. 6A,D). Moreover, cells expressing the siRNA-resistant HA–MICAL1 showed a decrease in actin-rich protrusions compared with the non-transfected (but MICAL1-depleted) cells (Fig. 6A–C; see non-transfected cells within dashed red lines and yellow arrows depicting protrusions). Because the generation of actin-rich protrusions was the most dramatic phenotype in MICAL1 knockdown cells, we used the presence of these structures to quantify the effects of knockdown of MICAL1 and MICAL2, as well as the ‘rescue’ of MICAL1 knockdown (Fig. 6E). As demonstrated, less than 15% of untreated cells normally displayed actin-rich protrusions, but knockdown of MICAL1 or MICAL2 increased the percentage of cells containing actin-rich protrusions to about 90% and 80%, respectively (Fig. 6E). Reintroduction of MICAL1 partially ‘rescued’ the knockdown phenotype, with only 50% of the cells exhibiting actin-rich protrusions (compared with >90% for MICAL1 knockdown). Because both MICAL2 knockdown and overexpression have complex effects on cellular actin, we were unable to carry out similar quantifiable rescue studies with this paralog.

Recent studies have shown that the MICAL1 FAD has the ability to produce H₂O₂ in vitro, and it has been proposed that generation of reactive oxygen species (ROS) might be an essential mechanism by which MICAL induces F-actin reorganization (Nadella et al., 2005; Schmidt et al., 2008). To determine whether the MICAL1 FAD domain is capable of inducing ROS generation, we used dihydrocalcein to detect ROS in HeLa cells expressing various MICAL constructs. Without transfection, very few cells contained detectable ROS levels (Fig. 7A). In cells expressing wild-type MICAL1, which does not induce loss of F-actin or actin-rich protrusions, ROS levels remained low (Fig. 7B,C and quantified in 7J). Cells transfected with the FAD domain (only) from MICAL1 displayed tenfold higher ROS levels (Fig. 7D,E and quantified in 7J). However, upon transfection of the enzymatically impaired MICAL1 FAD domain mutant (glycine to tryptophan point mutations), the levels of ROS were only slightly higher than the baseline levels observed upon MICAL1 expression (Fig. 7F,G and quantified in 7J). By contrast, MICAL2 expression induced three- to fourfold higher levels of ROS than did MICAL1 expression (Fig. 7H,I and quantified in 7J). To determine whether individual cells expressing the active MICAL1 FAD domain exhibited higher levels of ROS, we overexpressed Tomato-tagged MICAL1 FAD in HeLa cells treated with the fluorescent dye H₂DCFDA (supplementary material Fig. S7). First, we showed that expression of Tomato–MICAL1-FAD had similar effects to HA–MICAL1-FAD, causing loss of actin stress fibers and the generation of actin-rich protrusions (supplementary material Fig. S7A–C). To detect levels of ROS in individual transfected cells, we imaged fields of cells transfected with either Tomato alone (supplementary material Fig. S7D–F; control), or Tomato–MICAL1-FAD (supplementary material Fig. S7G–I). As demonstrated, expression of the Tomato protein had no effect on the levels of ROS produced in the cells (supplementary material Fig. S7D–F). On the other hand, cells expressing the Tomato-tagged FAD induced high levels of ROS in transfected cells (supplementary material Fig. S7G–I). Overall, these results lend credence to the notion that MICAL proteins exert their effects on actin microfilaments, at least in part, through the generation of ROS via their FAD domains.

Although Drosophila MICAL and its role in growth cone repulsion has been well studied over the years, relatively little is known about its mammalian counterparts, especially in the context of non-neural cells. Human MICAL1 was originally identified as an interaction partner for CasL and vimentin (Suzuki et al., 2002) and later as a Rab1 interaction partner (Weide et al., 2003). Other studies determined that MICAL1 functions in tandem with CRMP2 to regulate the semaphorin–plexin signaling pathway (Schmidt et al., 2008). MICAL1, whose monooxygenase domain-mediated NADPH oxidation is stimulated by F-actin (Zucchini et al., 2011), was also recently described as an interaction partner and negative regulator of nuclear Dbf2-related (NDR) kinases and thus impacts apoptosis (Zhou et al., 2011b). MICAL2, on the other hand, has a number of splice variants that are expressed in prostate cancer cells (Ashida et al., 2006), whereas MICAL3 has functions downstream of the semaphorin pathway in motor neurons (Bron et al., 2007). A very recent report shows that MICAL3 interacts with both Rab8 and ELKS, linking exocytic vesicles to the plasma membrane, and that its monooxygenase activity is required for exocytic activity (Grigoriev et al., 2011). Studies with rat models have led to the conclusion that MICAL proteins are expressed in embryonic, postnatal and adult nervous systems (Pasterkamp et al., 2006). Although most of these studies highlight the roles of mammalian MICAL proteins in development and in neural cells, the precise functions of these proteins in non-neural cells have remained largely unanswered.

Here, we have functionally compared mammalian MICALs using HeLa cells as a non-neural model system and have shown that MICAL1 and MICAL2 are both expressed at the RNA level. Various isoforms of MICAL3 are also expressed in HeLA cells, but we chose here to focus on MICAL1 and MICAL2, which are more closely related at the sequence level to Drosophila MICAL. Although the human MICAL-like proteins MICAL-L1 and MICAL-L2 play important roles in endocytic regulation (Sharma et al., 2009; Terai et al., 2006), we found that the human MICALs had no detectable effects on endocytic organelles and actin motor proteins. Consistent with Drosophila MICAL, human MICALs had only modest effects on α- and γ-tubulin localization. Instead, our various studies taken together suggest that the human MICALs play a key role in actin organization in non-neural cells. In particular, overexpression and domain-swapping studies point to the MICAL FAD domain as being crucial for actin-based structures. Specifically, overexpression of a MICAL2 construct, or of an intact FAD domain, led to a dramatic dismantling of actin stress fibers and a concomitant increase in actin-rich filopodia- or microspike-like protrusions. Because this was accompanied without an overall change in F-actin levels, we propose that enhanced MICAL2 expression could lead to a repurposing of cellular actin from stress fibers to protrusions.

By contrast, overexpression of wild-type MICAL1 had no effect on actin structures. However, domain experiments showed that this lack of phenotype was due to the presence of an autoinhibitory CC domain, because removal of this region led to a dramatic decrease in stress fibers, consistent with observations upon MICAL2 overexpression. As with MICAL2, an intact MICAL1 FAD domain was necessary and sufficient to recapitulate the uninhibited phenotype. Unlike MICAL2 overexpression, however, this shift in actin organization was accompanied by a decrease in overall F-actin levels, which highlights the fact that these two proteins, once activated, have subtly different functions. Indeed, domain-swapping experiments confirmed that the specific effect on actin could be mapped to the FAD domain, and that the autoinhibitory domain resides in the CC domain of MICAL1.

How do these results compare to our current understanding of Drosophila MICAL? Overexpression of Drosophila MICAL in bristles caused the F-actin to rearrange and form a complex meshwork of short filaments (Hung et al., 2010). Although these and other studies have also implicated Drosophila MICAL function in the disassembly of F-actin bundles and filaments, we have found important differences between Drosophila MICAL and human MICAL function. For example, unlike its human counterparts, Drosophila MICAL requires its CH as well as its FAD domain to be functionally active. Another key difference that we observed was in the subcellular localization of the proteins. For example, Drosophila MICAL is reported to be a cytoplasmic protein (Terman et al., 2002) and MICAL1 similarly displayed a cytoplasmic distribution, but MICAL2 was mostly localized proximal to the plasma membrane. This shows that the biology of mammalian MICALs, perhaps not surprisingly, might be more complicated than in the simpler Drosophila system. Because MICAL2, like Drosophila MICAL, is not autoinhibited, and because MICAL2 shows the highest protein sequence identity with Drosophila MICAL, MICAL2 might be functionally more similar to Drosophila MICAL than is MICAL1. Future studies to clarify the ‘true’ Drosophila MICAL ortholog might include rescue experiments using MICAL1 and MICAL2 in MICAL^−/− flies and testing whether the human MICALs are able to rescue the bristle morphology defects.

In this study, we found that stress fiber loss mediated by the MICAL FAD domain is accompanied by the production of ROS. Previous studies have demonstrated the ability of the MICAL1 FAD domain to generate ROS in vitro (Hung et al., 2010; Nadella et al., 2005; Schmidt et al., 2008). ROS have the ability to modify cysteine, methionine and tryptophan residues of actin, which might cause structural changes, the inability of actin to polymerize and its potential degradation (Dalle-Donne et al., 2002; Fedorova et al., 2010). Moreover, upon activation by its coenzyme NADPH, Drosophila MICAL causes a decrease in steady-state levels of actin polymerization (Hung et al., 2010). In addition, given that Drosophila MICAL does not destabilize actin by acting as a sequestering or capping protein, and that it decreases the rate and extent of actin polymerization (reviewed by Hung and Terman, 2011), the in vitro studies by Hung and co-workers suggest that Drosophila MICAL acts directly on actin through ROS production (Hung et al., 2010). However, until that study it had not been determined whether ROS are actually generated by MICAL proteins in cells (reviewed by Kolk and Pasterkamp, 2007).

Initially, we verified the ability of the MICAL1 FAD domain to produce ROS in vivo. It was previously determined that the anti-oxidant (-) epigallocatechin gallate (EGCG) can act as a pharmacological inhibitor of Drosophila MICAL monooxygenase activity in vivo (Terman et al., 2002). However, EGCG could potentially function either in monooxygenase inhibition, or more indirectly by suppressing MICAL-induced ROS. We found that a general antioxidant, N-acetyl cysteine (NAC), partially rescues stress fiber levels upon MICAL1 FAD domain overexpression (our unpublished observations). On the other hand, the MICAL2 FAD caused actin-rich protrusions, but not a loss of total F-actin (as observed for MICAL1). These studies are complicated, however, by the effects that NAC itself has on the actin cytoskeleton, and are thus difficult to interpret. However, the demonstration that individual cells transfected with an active FAD domain display both stress fiber loss as well as high levels of ROS is highly suggestive. It is likely that MICAL proteins, at least in part, affect actin microfilaments through their ability to produce ROS.

How might MICAL-induced ROS regulate F-actin structures? It is of interest that Rac1 has been implicated in depolymerization of actin microfilaments via ROS production (Moldovan et al., 1999). Also, Nimnual and colleagues reported that Rac suppresses Rho activity via this ability to generate ROS (Nimnual et al., 2003). Given that Rho dampening is associated with the dismantling of stress fibers (Ridley and Hall, 1992), it is possible that ROS generated by MICAL1 could feed into this same pathway and explain the similar phenotype we have observed. Regardless, because both Rac and MICAL proteins can function downstream of plexin, it remains to be determined whether Rac affects MICAL function or vice versa. It would also be interesting to see whether actin residues are modified by MICAL1 and/or MICAL2; potential differences in amino acid modifications might help explain why MICAL1 causes complete stress fiber depletion but MICAL2 primarily induces actin-rich protrusions. In addition, there is at least precedence for mammalian MICALs affecting actin structures other than stress fibers. For example, a previous study demonstrated that overexpression of the MICAL1 FAD domain induces a reduction in cell area via a retraction of the lamellipodia, leading to cell rounding (Schmidt et al., 2008). Moreover, these authors promoted the notion that MICAL1 has auto-inhibitive activity, and that binding to CRMP proteins releases this inhibition to allow the reduction in cell area to proceed (Schmidt et al., 2008). However, the mechanisms by which the MICAL1 FAD induces this reduction, and the mode by which the full-length protein is inhibitory, are not well understood. It is possible that the uninhibited MICAL1 FAD domain is so active that it dismantles the lamellipodia by completely destabilizing actin microfilaments. Similarly, the mechanism by which MICAL2 FAD overexpression, in our hands, led to an increase in actin-rich protrusions also remains to be delineated. The generation of filopodia is a complicated process that remains somewhat controversial (Faix et al., 2009), with a suite of upstream regulators including the Rho family GTPases. It might be that the dismantling of stress fibers increases the cellular pool of available actin, which is repurposed and shunted into pathways that create filopodial-like structures. Future experiments will be required to understand these interrelated processes.

In our experiments using RNA interference to deplete mammalian MICAL1 or MICAL2, we observed actin-rich protrusions of actin filaments, supporting our hypothesis that these proteins play a role in actin regulation in a non-neural context. It is interesting to note that overexpression of MICAL2 (and the FAD domains of either MICAL1 or MICAL2) induced phenotypic effects that bear similarity to those observed upon depletion of either protein, with notable loss of stress fibers and, in the case of MICAL2, generation of peripheral membrane actin-rich protrusions. Although the mechanism for this remains unknown, it is possible that knockdown of one MICAL induces compensatory action in another MICAL or one or more of its various isoforms, thus causing an overall level of enhanced ROS (as observed upon overexpression). Alternatively, it is possible that the absence of a particular MICAL alters the behavior of other actin-regulatory proteins also known to act at cytoplasmic plexin receptors, including Rho and Ras, or indeed of the GTPase-activating protein domain of plexin itself, all of which might be delicately balanced (reviewed by Hung and Terman, 2011).

Overall, our study indicates that the mammalian MICAL proteins not only play a role in semaphorin-mediated repulsive signaling in neural cells, but are also crucial regulators of actin stress fibers in non-neural cells. Actin is one of the most abundant proteins in eukaryotic cells, and plays an essential role in cell shape and survival, as well as providing the force for transport of intracellular molecules and cell movement (reviewed by Pollard and Cooper, 2009). By elucidating the specific functional roles of the mammalian MICAL family proteins in non-neural cells, our studies shed new light on the mechanisms regulating actin microfilaments.

Human MICAL1 and MICAL2 cDNA clones were purchased from Invitrogen (Carlsbad, CA) and cloned into a hemagglutinin fusion mammalian expression vector pHA-CMV (Clontech, Palo Alto, CA) using standard cloning procedures. MICAL1 and MICAL2 truncation mutants, FAD-dominant-negative and MICAL1 A940P, A941P mutants were generated using the QuickChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). siRNA-resistant constructs of HA–MICAL1 and HA–MICAL2 were generated similarly by creating silent cDNA mutations in the oligonucleotide-binding region. The MICAL1–MICAL2 chimera was created by cloning residues 1–634 of MICAL1 and residues 611–1124 of MICAL2 into the pHA-CMV vector. Similarly, the MICAL2–MICAL1 chimera was generated by cloning residues 1–635 of MICAL2 and residues 608–1067 of MICAL1 into the pHA-CMV vector. The MICAL1 FAD domain (residues 1–500) was cloned into ptd-Tomato-C3 (Clontech) using standard cloning procedures.

The following antibodies were used in this study: rabbit polyclonal antibody against MICAL1 (ProteinTech, Chicago, IL); mouse monoclonal antibodies directed against the HA epitope (Covance, Princeton, NJ); mouse anti-EEA1 (BD Biosciences, San Jose, CA); mouse anti-α-tubulin (Invitrogen); mouse anti-γ-tubulin (Sigma-Aldrich); mouse anti-actin and rabbit anti-Giantin (Abcam, Cambridge, MA); rabbit anti-HA (Bethyl, Montgomery, TX); rabbit antibody against phosphorylated paxillin (Biosource, Camerillo, CA); and rat anti-Hsc70 (Stressgen, Ann Arbor, MI). The Alexa-Fluor-488-conjugated phalloidin and secondary Alexa-Fluor-568-conjugated goat anti-mouse, Alexa-Fluor-488-conjugated goat anti-mouse, and Alexa-Fluor-568-conjugated goat anti-rabbit antibodies were purchased from Invitrogen. Goat anti-mouse horseradish peroxidase (HRP) was obtained from Jackson ImmunoResearch Laboratories (West Grove, PA). Donkey anti-rabbit HRP was obtained from GE Life Sciences (Piscataway, NJ). Dihydrocalcein and N-acetyl cysteine were purchased from Invitrogen and Sigma (St. Louis, MO), respectively. Dihydro-dichlorofluorescein diacetate (H₂DCFDA) was kindly provided by Surinder Batra (UNMC, Omaha, NE).

HeLa cells were transfected with plasmids using GeneExpresso DNA in vitro Transfection Reagent (Lab Supply Mall, Gaithersburg, MD) using the protocol specified by the company. On-Target SMART pool siRNA from Dharmacon (Lafayette, CO) was used to knock down MICAL1, MICAL2 and MICAL3 expression (using 300, 400 and 400 nM concentrations, respectively). siRNA knockdowns were performed as previously described (Sharma et al., 2009).

Immunofluorescence staining was carried out as previously described (Naslavsky et al., 2006). Image acquisition and processing were performed as previously described (Sharma et al., 2009), using a Zeiss LSM5 Pascal confocal microscope with a 63× objective (numerical aperture 1.4).

HeLa cells grown on coverslips or Lab-Tek chambered coverglass systems (Nunc, Rochester, NY) were either transfected or left untransfected. After 18 hours, cells were either fixed or washed and incubated with 10 μM dihydrocalcein or H₂DCFDA dissolved in DMEM (without Phenol Red, which interferes with the ROS measurements) for 30 minutes at 37°C. Dihydrocalcein- or H₂DCFDA-treated cells were then washed and incubated with DMEM containing 10% fetal bovine serum (but lacking Phenol Red). After 10 minutes, cells were imaged by confocal microscopy with a 10× objective. Manual counting was performed on dihydrocalcein-treated cells to calculate the number and ratio of fluorescent (ROS-containing) cells in each case. In parallel, fixed cells were washed, permeabilized and incubated with anti-HA followed by Alexa-Fluor-568-conjugated anti-mouse antibody. Cells were then mounted with solution containing DAPI. Images were obtained to quantify the transfection ratio. ROS levels were calculated from the ratio of cells fluorescing upon dihydrocalcein treatment divided by the overall transfection ratio. Experiments were performed three times and standard errors were determined.

Hela cells were grown on 35-mm dishes to ~90% confluency. mRNA from HeLa cells was extracted using RNeasy Mini Kit (Qiagen). cDNA was then prepared from HeLa mRNA using SuperScript III First-strand Synthesis System for RT-PCR (Invitrogen). PCR reactions were performed with pfuTurbo DNA polymerase (Agilent, Santa Clara, CA). Primers for RT-PCR that are specific for β-actin, MICAL1, MICAL2 and MICAL3 were designed using primer blast. The primers used for actin were 5′-CCTCGCCTTTGCCGATCC-3′ and 5′-GGATCTTCATGAGGTAGTCAGTC-3′. The primers used for MICAL1 were 5′-TGCTGGGACTGGTGGGGGAC-3′ and 5′-GCCGGTACACCAGGGCACAC-3′. The primer pairs that were used for MICAL2 were 5′-CACCACTGCCGCACACGACT-3′, 5′-AGTGCGCAAGCCACAGGGTC-3′; 5′-CCAAGACAACCTGCTATCCTATGC-3′, 5′-GTGCTCCAGCTCCTTAGTGATATAC-3′; 5′-GCCCGTGGATTCTTGGCGCA-3′, 5′-GTGCCCCGTCACGACCACAG-3′; and 5′-CGCAGGCTCTTTCCGGGGTG-3′, 5′-CACGTGTCGCTGCCTCCCAG-3′. The primer pairs 5′-ATCGGCATACGGAGACAGGGCTC-3′, 5′-ATCTCCAGAGGATCCGCCGCG->3′ and 5′-AGAGAACCCCAGGTTCAGGCGT-3′, 5′-TCTTCCCCTCACAACGCACTTGCT-3′ were used for MICAL3 isoform1 and isoform3, respectively.

The authors gratefully acknowledge the assistance of Subhankar Chakraborty and Srustidhar Das for their assistance with the reverse transcriptase PCR experiments.