Redox modification of nuclear actin by MICAL-2 regulates SRF signaling

MICAL-2 is enriched in the nucleus

Because of the important roles of MICAL-1 in depolymerizing actin in axonal growth cones (Hung et al., 2010), we sought to understand if MICAL-2 and -3 have related functions. In order to understand the roles of MICAL-2 and -3, we examined the subcellular localization of each MICAL isoform. Cherry-tagged MICAL-1 was cytoplasmic in HEK293T cells, however, Cherry-MICAL-2 and Cherry-MICAL-3 were nuclear (Figure 1A). To confirm that these localizations are not due to the presence of the Cherry tag, we monitored the localization of endogenous MICALs by subcellular fractionation using isoform-specific antibodies (Figures S1A and S1B). MICAL-1 was cytoplasmic, while MICAL-2 and -3 were enriched in the nuclear fraction in HEK 293T cells (Figure 1B). The nuclear localization of MICAL-2 was also confirmed by immunofluorescence in HEK293T, COS7, and HeLa cells (Figures S1C and S1D).

The PredictProtein algorithm (Rost et al., 2004) revealed a putative bipartite nuclear localization sequence (NLS) in both MICAL-2 and -3, which was not found in MICAL-1 (Figure 1C). Substitutions in lysine and arginine residues in the NLS abolished the nuclear localization of MICAL-2 and -3 (Figure 1A), confirming that these sequences constitute a NLS domain in these proteins.

MICAL-2 redox activity induces actin depolymerization

Because MICAL-2 and -3 share 60% and 63% identity, respectively, with MICAL-1 in the catalytic flavoprotein monooxygenase domain (Terman et al., 2002), we asked if they also depolymerize F-actin. To test this, we purified the bacterially expressed catalytic domain and the adjacent actin-binding calponin homology (CH) domain (designated ‘redoxCH’) for each MICAL isoform (Hung et al., 2010). To measure actin depolymerization, we used a previously described assay for MICAL-1-dependent actin depolymerization using pyrene-labeled actin (Hung et al., 2010). As expected, MICAL-1^redoxCH induced actin depolymerization in an NADPH-dependent manner (Hung et al., 2010) (Figure 1D). Similarly, MICAL-2^redoxCH and -3^redoxCH induced actin depolymerization in an NADPH-dependent manner. The rate of NADPH consumption by MICAL-2 and -3 was markedly enhanced by F-actin, suggesting that F-actin is a specific substrate of these enzymes (Figure S1E).

MICAL-2 regulates nuclear actin polymerization

The ability of MICAL-2 and -3 to depolymerize F-actin in vitro, combined with their nuclear localization, suggests that these isoforms may regulate actin polymerization within the nucleus. We therefore examined the effect of inhibiting MICAL-2 activity on F-actin levels within the nucleus in various cell lines. Expression of either of two dominant-negative MICAL-2 constructs, MICAL-2CT, which contains a deletion of the N-terminal catalytic domain, and MICAL-2GV, which contains a glycine to valine mutation at amino acid 95 that blocks FAD-binding in flavoprotein monooxygenases (Eppink et al., 1997), resulted in marked reorganization of actin into linear filaments that span large portions of the nucleus (Figures 1E and S1F). Similar mutations in MICAL-3 expression constructs did not appear to alter nuclear actin (data not shown).

MICAL-2 knockdown (Figure S1B) resulted in the appearance of nuclear F-actin as detected by phalloidin staining in fixed cells (Figure S1G). Similarly, expression of GFP-tagged LifeAct, which binds to F-actin (Riedl et al., 2008), revealed nuclear F-actin in live cells expressing MICAL-2-specific shRNA (Figure S1H). The MICAL-2 knockdown constructs did not affect cellular viability (Figures S1I and S1J).

To validate that these actin filaments are localized within the nucleus, we co-stained cells with an antibody to lamin A/C, a marker of the inner nuclear membrane (Figure S1K). Actin filaments were indeed within the nucleus, rather than surrounding the nuclear membrane (Figures S1L–O). Collectively, these data indicate that MICAL-2 enzymatic activity inhibits polymerization of nuclear actin in cells.

MICAL-2 expression induces nuclear localization of MRTF-A

Since MICAL-2 regulates F-actin within the nucleus, and since nuclear G-actin mediates nuclear export of MRTF-A (Vartiainen et al., 2007), we considered the possibility that MICAL-2 reduces nuclear MRTF-A by promoting its export. To test this, HEK293T cells were infected with a virus expressing either GFP or GFP-MICAL-2 and serum-starved for 18 h under low-serum conditions (0.3% fetal bovine serum [FBS]) (Miralles et al., 2003) (Figures 2A and 2B). Contrary to our expectation, overexpression of MICAL-2 under these conditions caused a marked increase in nuclear localization of MRTF-A (Figures 2A and 2B).

We next asked if the nuclear localization of MRTF-A induced by serum is mediated by MICAL-2. Treatment of serum-starved HEK293T cells with 10% FBS for 10 min altered MICAL-2 phosphorylation levels, suggesting that serum may regulate the function of MICAL-2 (Figures S2A and S2B). To determine if MICAL-2 is required for serum-mediated accumulation of MRTF-A in the nucleus, we monitored MRTF-A localization following knockdown of MICAL-2. In HEK293T cells expressing either of two MICAL-2-specific shRNAs, but not control shRNA, serum-induced localization of MRTF-A to the nucleus was markedly reduced (Figures S2C and S2D). Thus, MICAL-2 induces nuclear localization of MRTF-A in the absence of serum stimulation and is required for the increase in nuclear MRTF-A levels after serum stimulation.

We considered the possibility that MICAL-2 could increase nuclear MRTF-A by increasing expression levels of MRTF-A. However, MICAL-2 expression has no discernable effect on global levels of MRTF-A (Figure S2E). Additionally, the nuclear accumulation of MRTF-A does not appear to reflect binding of MRTF-A to MICAL-2 (Figures S2F and S2G). Lastly, expression of MICAL-2 did not induce stress fibers or increase cytosolic F-actin levels in HEK293T cells (Figures S2I–L). Thus, MICAL-2 does not increase nuclear MRTF-A by increasing cytosolic F-actin.

MICAL-2 mediates NGF-dependent nuclear localization of MRTF-A

We next examined whether MICAL-2 is required for nuclear localization of MRTF-A after growth factor stimulation. MRTF-A was readily detectable in the cytosol in untreated PC12 cells, and exhibited increased nuclear localization in response to NGF treatment (Figures 2C and 2D). However, MRTF-A nuclear localization in response to NGF was markedly reduced by expression of either of two MICAL-2-specific shRNAs, but not by a control shRNA (Figures 2C and 2D).

We next asked whether MICAL-2 promotes the nuclear localization of MRTF-A in primary neurons. For these experiments, we used rat embryonic day (E) 14 dorsal root ganglia (DRG) sensory neurons, which are typically cultured in NGF. Because these neurons require NGF for survival (Chun, 1977), we did not test culture conditions lacking NGF. In these neurons, MRTF-A was readily detectable in both the nucleus and cytoplasm (Figures 2E and 2F). Infection of these cells with lentivirus expressing either of two MICAL-2-specific shRNAs resulted in a marked decrease in nuclear MRTF-A (Figures 2E and 2F). Taken together, these data indicate that MICAL-2 is required for efficient nuclear localization of MRTF-A.

MICAL-2 expression induces the SRF/MRTF-A-dependent reporter

Because MICAL-2 induces nuclear localization of MRTF-A, we asked if MICAL-2 activates SRF/MRTF-A-dependent gene transcription. To test this, we determined the effect of MICAL-2 expression on a transcriptional reporter containing luciferase under the control of a CArG [CC(A/T)₆GG] element. In serum-starved HEK293T cells, minimal levels of luciferase were detected (Figure 3A). However, expression of MICAL-2 resulted in a 15-fold increase in luciferase expression (Figure 3A). This effect was not seen following expression of MICAL-2 mutants that were catalytically inactive, or which contain NLS mutations (Figure 3A). MICAL-2 did not induce a transcriptional reporter containing the highly related TCF/ETS SRE element, which requires SRF but not MRTF-A (Buchwalter et al., 2004) (Figure 3B). Collectively, these data indicate that MICAL-2 induces SRF/MRTF-A-dependent reporter gene expression.

MICAL-1 and MICAL-3 overexpression did not activate the reporter, indicating that activation of SRF/MRTF-A-dependent gene expression is limited to MICAL-2 in HEK293T cells (Figure 3C, S3A and S3B).

To confirm that MICAL-2-mediated activation of SRF/MRTF-A gene transcription requires MRTF-A, we expressed MRTF-A-ΔTAD, a dominant-negative MRTF-A mutant. This protein lacks the transcription activation domain (TAD) that is necessary for an active SRF/MRTF-A complex (Wang et al., 2003). Expression of MRTF-A-ΔTAD blocked the effect of MICAL-2 on the SRF/MRTF-A reporter (Figure 3D). Thus, MICAL-2 induces the reporter in an MRTF-A-dependent manner.

MICAL-2 expression induces endogenous SRF/MRTF-A target gene expression

We next asked if MICAL-2 induces endogenous SRF/MRTF-A target genes. SRF/MRTF-A target genes have been validated in numerous cell lines (Medjkane et al., 2009; Selvaraj and Prywes, 2004) and neuronal cells (Knöll and Nordheim, 2009). Expression of MICAL-2 in serum-starved HEK293T cells resulted in increased levels of several known SRF/MRTF-A target genes in these cells by 30-50%, including Acta2, Cyr61, SRF, and VCL (Figure S4A). Expression of MICAL-2 had a negligible effect on MRTF-A-independent SRF-dependent genes that contain the TCF/ETS SRE element in their promoter TSP1, PTGS1, and Egr2.

Expression of MICAL-2 in PC12 cells similarly induced SRF/MRTF-A target genes Acta2, Cyr61, and SRF by 100-200% (Figure S4C). MICAL-2 expression did not induce SRF-dependent, MRTF-A-independent genes, such as TSP1, PTGS1, and Egr2 (Figure S4C). These data indicate that MICAL-2 expression induces endogenous SRF/MRTF-A-dependent gene expression.

MICAL-2 is required for serum and NGF-induced SRF/MRTF-A target gene expression

We next examined whether MICAL-2 mediates the induction of SRF/MRTF-A target genes in response to serum and NGF. In fibroblasts, serum induces the expression of Acta2, Cyr61, VCL, and SRF in an SRF- and MRTF-A-dependent manner (Descot et al., 2009; Lee et al., 2010) (Figure S4B). However, expression of either of two MICAL-2-specific shRNAs blocked this effect (Figure S4B). MICAL-2-specific shRNAs did not affect the induction of MRTF-A-independent SRF-dependent genes TSP1, PTGS1, and Egr2. In PC12 cells, NGF induces the expression of Acta2, Cyr61, VCL, and SRF in an SRF- and MRTF-A-dependent manner (Figure S4D). MICAL-2 shRNAs impaired NGF-induced expression of these SRF/MRTF-A target genes (Figure S4D), but did not affect NGF-mediated induction of MRTF-A-independent SRF-dependent genes TSP1 and PTGS1. Similarly, rat E14-15 DRG neurons cultured in NGF exhibited a reduction in Acta2, Cyr61 and VCL expression upon infection with lentivirus expressing either of two MICAL-2 shRNAs, while MRTF-A-independent SRF-dependent genes TSP1, PTGS1 and Egr2 were unchanged (Figure S4E). Thus, MICAL-2 promotes serum- and NGF-dependent increases in SRF/MRTF-A-dependent gene expression.

MICAL-2 is required for NGF-dependent neurite outgrowth

To further address whether MICAL-2 mediates NGF-induced SRF/MRTF-A-dependent transcription, we monitored the effect of MICAL-2 knockdown on NGF-dependent neurite outgrowth in PC12 cells and embryonic DRG sensory neurons. In PC12 cells expressing MICAL-2-specific shRNA, neurite length 4 d after NGF treatment was reduced by ~40% compared to LacZ shRNA-infected PC12 cells (Figures 4A and 4B). Knockdown of MICAL-2 in differentiated PC12 cells also decreases neurite formation after NGF treatment (Figure 4C). Similar to PC12 cells, DRG neurons expressing MICAL-2-specific shRNA exhibited a ~45% decrease in axon growth rate at DIV5 compared to LacZ-shRNA expressing neurons (Figures 4D and 4E). Taken together, these data indicate that MICAL-2 mediates NGF-induced SRF/MRTF-A signaling in both PC12 cells and primary embryonic sensory neurons.

Embryonic MICAL-2 knockdown impairs SRF/MRTF-A-dependent gene expression

We next asked if MICAL-2 mediates SRF/MRTF-A-dependent gene transcription induced by physiologic signals. SRF-dependent gene transcription is required for heart development in zebrafish (Chong et al., 2012). MICAL isoforms have been characterized in zebrafish (Xue et al., 2010), with the cardiac-enriched mical2b exhibiting highest sequence identity to mouse MICAL-2.

We therefore knocked down mical2b during embryonic development using antisense morpholinos (Figures S5A and S5C). We used either a splice-blocking morpholino or a translation-blocking morpholino, and injected them into developing embryos at the 1-cell stage. Depletion of mical2b transcripts by the splice-blocking morpholino was validated by qRT-PCR (Figure S5B). Knockdown of mical2b resulted in small hearts that failed to undergo normal looping at 24 hpf (Figures 5A and 5B), with thin, linear morphology compared to wild-type hearts at 48 hpf (Figures 5C and 5D). Additionally, cardiomyocytes in morphant hearts were spatially disorganized, rather than displaying the relatively even distribution seen in wild-type hearts, as shown at 52 hpf in the atria (Figures 5E–G). Taken together, these data confirm that mical2b expression was effectively knocked down in zebrafish.

We next asked if mical2b-deficient animals exhibited selective defects in SRF/MRTF-A–dependent gene transcription. We found that many cardiac muscle-specific SRF/MRTF-A target genes, such as cardiac alpha actin, smooth muscle alpha actin, cardiac actin smooth muscle, skeletal alpha actin, smooth muscle 22a, srf, and calponin (Chong et al., 2012; Davis et al., 2008; Descot et al., 2009) were markedly reduced in mical2b knockdown animals (Figure 5H). The control genes vascular endothelial growth factor (vegf), fibroblast growth factor 4 (fgf4), and ETS translocation variant 4 (pea3) which regulate MRTF-A-independent cardiac development in zebrafish, were not affected by mical2b knockdown (Figure 5H) (Liang et al., 2001; Znosko et al., 2010). Co-injection of the morpholinos with RNA encoding full-length human MICAL-2, but not MICAL-2GV, normalized expression of the SRF/MRTF-A-dependent genes (Figure 5H). Thus, mical2b regulates SRF/MRTF-A-dependent gene transcription in response to physiological stimuli during zebrafish development.

MICAL-2 does not activate SRF/MRTF-A through RhoA

We considered the possibility that MICAL-2 activates SRF/MRTF-A signaling by increasing RhoA activity. However, application of cell-permeable C3-transferase, an inhibitor of RhoA, or Y-27632, a ROCK inhibitor, did not block MICAL-2-induced activation of the SRF/MRTF-A reporter in serum-starved HEK293T cells (Figure 6A) or MICAL-2-dependent MRTF-A nuclear accumulation (Figure 6B). These data, along with our finding that MICAL-2 does not affect cytosolic F-actin levels (Figure S2I–L), indicate that MICAL-2 does not activate SRF/MRTF-A reporter expression through the RhoA pathway.

MICAL-2 induces depletion of nuclear actin

We next asked if MICAL-2 activates SRF/MRTF-A-dependent gene transcription by depleting nuclear G-actin. To test this, we measured nuclear actin levels in HEK293T cells with Alexa 594-DNase I, which specifically binds G-actin (Suck et al., 1981). Consistent with previous studies (Mouilleron et al., 2011; Vartiainen et al., 2007), we found that serum-starvation was associated with higher G-actin levels in the nucleus than in the cytoplasm (Figures 6C and 6D). MICAL-2 overexpression caused a significant reduction in nuclear G-actin, even in the absence of serum (Figures 6C and 6D). The reduction in nuclear G-actin induced by MICAL-2 was similar to the reduction induced by serum stimulation (Figures 6C and 6D). These data suggest that MICAL-2 induces nuclear localization of MRTF-A by reducing nuclear G-actin levels.

We next asked if MICAL-2 promotes the reduction in nuclear G-actin seen after NGF stimulation. In differentiated PC12 cells, NGF induced a significant decrease in nuclear G-actin (Figures S6A and S6B). Expression of GFP-MICAL-2 in unstimulated PC12 cells also induced a significant decrease in the ratio of nuclear:cytosolic G-actin (Figures S6A and S6B). Furthermore, in PC12 cells expressing either of two MICAL-2-specific shRNAs, NGF-dependent reduction in nuclear G-actin levels was significantly impaired (Figures S6C and S6D). Consistent with these findings, DRG neurons expressing either of two MICAL-2-specific shRNAs exhibited significantly increased nuclear G-actin levels (Figures S6E and S6F). These data suggest that MICAL-2 mediates NGF-induced reductions in nuclear G-actin.

We next asked if the effects of MICAL-2 were due to a direct effect on actin. MICAL-1 oxidizes methionine 44 in actin to form methionine sulfoxide (Hung et al., 2011). We therefore asked if a M44Q substitution, which mimics features of the oxidized form of actin, alters actin localization in cells. Wild-type GFP-actin was largely cytosolic though a small portion was found in the nucleus (Figures S6G and S6H). However, GFP-actin M44Q was significantly excluded from the nucleus. On the other hand, GFP-actin M44L, which is not oxidizable, exhibited higher nuclear levels than wild-type actin (Figures S6G and S6H). These data suggest that oxidation of Met44 promotes exclusion of actin from the nucleus. GFP-actin and GFP-actin M44L colocalized with phalloidin staining at similar levels, however GFP-actin M44Q colocalization with phalloidin was significantly reduced (Figure S6I). The lack of incorporation of GFP-actin M44Q in F-actin is consistent with the known depolymerizing effect of MICAL-2 on actin. We next asked if oxidation of actin methionine 44 is required for MICAL-2-dependent reduction in nuclear G-actin. To test this, we monitored the effect of MICAL-2 on nuclear localization of GFP-actin and GFP-actin M44L. In these experiments we found that MICAL-2 coexpression decreased nuclear GFP-actin but not GFP-actin M44L (Figures 6E and 6F). These data suggest that MICAL-2-dependent oxidation of methionine 44 leads to a reduction in nuclear G-actin.

The SRF/MRTF-A pathway inhibitor CCG-1423 is a MICAL-2 inhibitor

CCG-1423 is a small molecule inhibitor of SRF/MRTF-A-dependent transcription that was identified in a cell-based screen using a SRF/MRTF-A reporter (Evelyn et al., 2007). While CCG-1423 has shown utility in preclinical disease models, its specific molecular target is not known.

CCG-1423 treatment reduces nuclear MRTF-A levels (Jin et al., 2011), which is similar to the effect of MICAL-2 knockdown (Figures 2C–F and Figures S3A–B). Therefore, we wondered if the effects of CCG-1423 on SRF/MRTF-A signaling are mediated by inhibition of MICAL-2. As a first test, we asked if CCG-1423 blocks MICAL-2 signaling in cells. In this experiment, we monitored the effect of CCG-1423 on MICAL-2-dependent induction of the SRF/MRTF-A reporter in serum-starved HEK293T cells. Expression of MICAL-2 induced luciferase expression, while treatment with 5 μM CCG-1423 for 4 hr reduced it by 75% (p<0.001, Figure 7A). These data indicate that CCG-1423 blocks MICAL-2-mediated induction of the SRF/MRTF-A reporter.

We next asked if CCG-1423 binds MICAL-2. To detect CCG-1423 binding to MICAL-2, we used a thermal denaturation assay. In this assay, the melting temperature (T_m) of a protein is measured by increasing the temperature of a sample in the presence of a fluorescent dye, such as 1-anilinonaphthalene-8-sulfonic acid (1,8-ANS) (Pantoliano et al., 2001). As the protein unfolds, 1,8-ANS binds the exposed hydrophobic domains, increasing the dye’s fluorescence. Ligand binding is detected by observing an increase or decrease in the T_m. Incubation of MICAL-2 with CCG-1423 decreased the T_m by −0.3°C with an EC₅₀ of 3.8 μM. A structurally similar compound with markedly reduced potency at inhibiting SRF/MRTF-A signaling (CCG-100594) (Evelyn et al., 2010) had no significant effect on T_m (Figure 7B and Figures S7A–C), indicating that these effects are specific to CCG-1423. Taken together, these data indicate that CCG-1423 directly binds MICAL-2.

We next asked if CCG-1423 binding affects MICAL-2 catalytic activity. To test MICAL-2 activity, we used an NADPH depletion assay. Recombinant MICAL-2, comprising the enzymatic domain and the CH domain (MICAL-2^redoxCH), was incubated with NADPH in the presence of F-actin, and the consumption of NADPH was monitored by UV absorbance. Addition of CCG-1423 (10 μM) markedly reduced NADPH consumption by MICAL-2 compared to CCG-100594 (Figure 7C). CCG-1423 increases the K_m of NADPH and exhibited a K_i of 1.57 μM (Figure 7D). Taken together, these data indicate that CCG-1423 binds MICAL-2 and inhibits its enzymatic activity.

Cell Culture and reagents

Primers, shRNA, plasmids, cell lines, and antibodies used in this study are listed in Tables S1 and S2 and the Extended Experimental Procedures.

qRT-PCR

RNA was prepared using TRIzol (Invitrogen). We synthesized cDNA using Superscript III (Invitrogen) and performed qRT-PCR utilizing iQ SYBR Green Supermix (Bio-rad), and the Realplex Mastercycler ep s (Eppendorf). For a list of primers used see Supplementary Table S1 and the Extended Experimental Procedures.

Luciferase and MICAL Assays

Activation of SRF/TCF and SRF/MRTF-A-dependent gene transcription was measured using the pGL4.33 and pGL4.34 reporter plasmids, respectively (Promega). Luciferase levels were measured using ONE-glo (Promega) and with a Spectramax L luminometer (Molecular Devices).

Recombinant MICAL-1, -2 and -3-induced actin depolymerization was measured by monitoring the loss of fluorescence of pyrene-labeled actin (Cytoskeleton) as described previously (Hung et al., 2010). MICAL-1, -2 and -3 activity assays were measured by in an NADPH depletion assay as previously described (Hung et al., 2010). Further details can be found in the Extended Experimental Procedures.

Zebrafish

Wildtype zebrafish (AB/Tu hybrid strain) were maintained as described (Westerfield, 1993). Transgenic reporter strain tg(myl7:egfp) was obtained from ZIRC, and tg(myl7:actn3b-egfp) was kindly provided by D. Yelon (UCSD). Morpholinos were purchased from Genetools, Inc. Sequences and experimental details can be found in the Extended Experimental Procedures and Spuuplementary Table S2.