MICAL2 acts through Arp3B isoform-specific Arp2/3 complexes to destabilize branched actin networks

The Arp2/3 complex (Arp2, Arp3 and ARPC1-5) is essential to generate branched actin filament networks for many cellular processes. Human Arp3, ARPC1 and ARPC5 exist as two isoforms but the functional properties of Arp2/3 iso-complexes is largely unexplored. Here we show that Arp3B, but not Arp3 is subject to regulation by the methionine monooxygenase MICAL2, which is recruited to branched actin networks by coronin-1C. Although Arp3 and Arp3B iso-complexes promote actin assembly equally efficiently in vitro, they have different cellular properties. Arp3B turns over significantly faster than Arp3 within the network and upon its depletion actin turnover decreases. Substitution of Arp3B Met293 by Thr, the corresponding residue in Arp3 increases actin network stability, and conversely, replacing Arp3 Thr293 with Gln to mimic Met oxidation promotes network disassembly. Thus, MICAL2 regulates a subset of Arp2/3 complexes to control branched actin network disassembly.


Abstract
The Arp2/3 complex (Arp2, Arp3 and ARPC1-5) is essential to generate branched actin filament networks for many cellular processes. Human Arp3, ARPC1 and ARPC5 exist as two isoforms but the functional properties of Arp2/3 iso-complexes is largely unexplored. Here we show that Arp3B, but not Arp3 is subject to regulation by the methionine monooxygenase MICAL2, which is recruited to branched actin networks by coronin-1C. Although Arp3 and Arp3B iso-complexes promote actin assembly equally efficiently in vitro, they have different cellular properties. Arp3B turns over significantly faster than Arp3 within the network and upon its depletion actin turnover decreases. Substitution of Arp3B Met293 by Thr, the corresponding residue in Arp3 increases actin network stability, and conversely, replacing Arp3 Thr293 with Gln to mimic Met oxidation promotes network disassembly. Thus, MICAL2 regulates a subset of Arp2/3 complexes to control branched actin network disassembly.

Main text
The Arp2/3 complex, consisting of seven subunits (Arp2, Arp3, ARPC1-5) is conserved in all eukaryotes, with the exception of some algae, protists and microsporidia (1-4). In mammals, the Arp3, ARPC1 and ARPC5 subunits exist as two isoforms (Arp3/3B, ARPC1A/B, and ARPC5/5L), which in humans are 91, 67 and 67% identical respectively (5)(6)(7). Several observations indicate that different Arp2/3 complexes have distinct regulatory or functional properties. For example, mutations that lead to a severe reduction or loss of ARPC1B expression in humans result in immunodeficiency and inflammation due to defects in cytotoxic T lymphocyte maintenance and activity (8)(9)(10)(11)(12)(13). ARPC5L, together with g-actin, is required for peripheral nuclei positioning, while ArpC5 and b-actin are involved in transverse triad formation in in vitro assembled myofibers (14). The ARPC1 and ARPC5 isoforms also differentially affect the actin nucleating properties of the Arp2/3 complex as well as the stability of the branched filament networks they generate (15). In contrast, to ARPC1 and ARPC5 isoforms, the impact of Arp3B on the activity and function of the Arp2/3 complex remains to be established.
To investigate whether Arp3B impacts on the activity of the Arp2/3 complex we used actinbased motility of vaccinia virus as a model system (15)(16)(17)(18). GFP-tagged Arp3 isoforms are both recruited to vaccinia-induced actin tails (Fig. 1A), consistent with their ability to incorporate into Arp2/3 complexes (Fig. S1A). Strikingly, GFP-tagged Arp3 and Arp3B induced the formation of longer and shorter actin tails respectively (Fig. 1A), suggesting that Arp3B may not be as effective as Arp3 in promoting actin polymerization. To investigate if this is the case, we examined the impact of RNAi-mediated depletion of Arp3 isoforms on actin-based motility of vaccinia (Fig. S1B). As described previously (19,20), loss of Arp3 results in a significant reduction of all Arp2/3 complex subunits (Fig. 1B). Concomitant with this, there was a dramatic reduction in the length of vaccinia-induced actin tails (Fig.   1C, S1C), that is rescued by expression of GFP tagged Arp3 but not Arp3B (Fig. 1D, S1D, E).
Knockdown of Arp3B had no appreciable impact on the level of Arp2/3 complex subunits (Fig. 1B). Nevertheless, there was a significant increase in actin tail length (Fig. 1C).
Moreover, expression of GFP-Arp3B in cells treated with Arp3 siRNA results in short actin tails (Fig. 1D). These observations clearly demonstrate that Arp3 and Arp3B containing complexes have different ability to promote vaccinia-induced actin polymerization.
To examine whether the differences in actin tail lengths are due to Arp3B complexes being less efficient than those with Arp3 in promoting actin polymerization, we performed in vitro pyrene actin assembly assays using defined recombinant human Arp2/3 complexes ( Fig. 2A,   S2A). We found that Arp2/3 complexes containing Arp3 and Arp3B were equally efficient at stimulating actin polymerization, with ARPC1B/ARPC5L containing complexes being better than those with ARPC1A/ARPC5 as observed previously (15) (Fig. 2A). We also performed TIRF microscopy to directly visualize Arp2/3 mediated actin assembly and branching (Fig.   2B, Movie S1). Automated quantification with the AnaMorf ImageJ plugin (21), reveals no differences in the rate of actin branch assembly (Fig. 2C). We therefore wondered whether short actin tails induced by Arp3B are due to increased actin disassembly. Using photoactivatable Cherry-GFP PA -actin we measured the rate of actin tail disassembly in cells treated with ARP3B siRNA (Fig. S2B). Loss of Arp3B increases the half-life of actin disassembly from 8.00 ± 0.16 to 9.82 ± 0.19 sec (Fig. 2D Human Arp3 and Arp3B are 91% identical with amino acid differences spread throughout their length (Fig. S3A). To narrow down the region/amino acids responsible for the difference in actin network stability, we examined the impact of expressing a series of RNAi resistant Arp3/3B hybrids on actin tail length in cells lacking endogenous Arp3 and Arp3B ( Fig. 3A, S3B). We found that residues 281-418 of Arp3B are sufficient to confer the short actin tail phenotype (Fig. 3B). This region contains 10 conservative and 3 non-conservative substitutions between the two proteins (Fig. S3A). Analysis of the structure of Arp2/3 in its active state reveals that two non-conserved substitutions (T293M and P295S) are on the surface of Arp3 in a hydrophobic loop which is positioned close to Arp2 (22-24) (Fig. 3C). To investigate their possible role in mediating isoform differences we exchanged residues 293 and 295 between Arp3 and Arp3B to generate GFP tagged Arp3 MS and Arp3B TP (Fig. 3D). GFPtagged Arp3 MS and ARP3B TP switched phenotypes inducing the formation of short and long actin tails respectively in cells depleted of endogenous Arp3 and Arp3B (Fig. 3E, S3C, S3D).
Actin tails remained long when single Arp3B substitutions were introduced into Arp3 (Fig.   3E, S3C, S3D). In contrast, actin tails remained short when Ser295 in Arp3B was changed to proline, but became long when Met293 was substituted for threonine. This demonstrates that Met293 is essential for Arp3B to induce short actin tails.
Oxidation of Met44 and Met47 to methionine sulfoxide (Met-SO) in actin by MICAL proteins promotes actin filament disassembly (25,26). Given this, we wondered whether the Arp3Bdependent actin tail phenotype depends on oxidation of Met293. To investigate this possibility, we replaced Thr293 in Arp3 and Met293 in Arp3B with glutamine to mimic the Met-SO state (27, 28). In both cases, the glutamine mutants induced the formation of shorter actin tails (Fig 4A, S4A). Consistent with the notion that Met293 in Arp3B is oxidized we found that GFP-tagged MICAL2 but not MICAL1 is recruited to actin tails (Fig. 4B, S4B Movies S6 & S7). Moreover, knockdown of MICAL2 also reduces the half-life of actin disassembly in tails by ~ 2 sec, while loss of MICAL1 had no impact (Fig. 4C, S4C). Depletion of MICAL2 or Arp3B alone or together resulted in similar long actin tails (Fig. 4D, S4D).
Lastly, loss of MICAL2 suppressed the short actin phenotype induced by GFP-Arp3B over expression (Fig. 4E, S4E). Our data suggest that MICAL2 mediated oxidation of Met293 in Arp3B promotes faster actin network disassembly, resulting in shorter actin tails.
It is striking that GFP-MICAL2 is not recruited immediately behind the virus but further down the actin tail (Fig. 4B), as we previously observed for coronin (15). In light of this, we investigated if there is a connection between the two proteins given the role of coronin in promoting disassembly of branched actin networks (29, 30). Depletion of coronin-1C resulted in a dramatic loss of GFP-MICAL2 recruitment to actin tails (Fig. 5A, S5A, Movies S8 & S9) and suppression of the short actin tail phenotype induced by GFP-Arp3B (Fig. 5B, S5B).
In agreement with this, GFP-Trap pulldowns demonstrated that coronin-1C associates with GFP-tagged MICAL2 but not MICAL1 (Fig. 5C). We previously demonstrated that coronininduced actin tail disassembly depends on cortactin (15). Consistent with this, loss of cortactin also suppresses the short actin tail phenotype induced by GFP-Arp3B but had no impact on Arp3 induced actin tails (Fig. 5D, S5C). Furthermore, photoactivation experiments demonstrate that depletion of cortactin increases the half-life of Arp3B from 6.29 ± 0.17 to 9.48 ± 0.22 seconds but had no impact on Arp3 dynamics (Fig. 5E, S5D). In addition, depletion of coronin-1C or MICAL2 also results in a similar ~3 sec increase in Arp3B half-life, while having minimal impact on Arp3 (Fig. 5F, S5E). Concomitant with this, there was also a similar increase in the half-life of actin disassembly when Arp3B, coronin-1C or MICAL2 were depleted alone or in pairwise fashion (Fig. 5G, S5F).
Actin assembly drives many cellular processes including cell migration, however, the disassembly of actin networks is equally as important for the actin cytoskeleton to perform its numerous functions (15,29,30). Our observations demonstrate that the recruitment of MICAL2 by coronin-1C enhances the disassembly of branched actin networks in an Arp3B but not Arp3 dependent fashion. Proteomic data indicates that Arp3B is widely expressed albeit at significantly lower levels than Arp3 (http://pax-db.org); for example, in HeLa cells Arp3B is 61 fold less abundant than Arp3 (31). Our data suggest that the relative levels of Arp3 and Arp3B will influence the stability of Arp2/3 induced branch networks. The regulation of a subset of Arp2/3 complexes by MICAL2, together with the impact of ARPC1 and ARPC5 isoforms, allows for further fine tuning of branched actin network dynamics in different cellular contexts and processes.

Vaccinia virus infection, antibodies, immunoblot and immunofluorescence analysis
HeLa were infected with the Western Reserve strain of Vaccinia virus 72 hours after siRNA transfection and processed for immunofluorescence analysis as previously described (15).
Actin tails were labelled with Alexa488 or Texas red phalloidin (Invitrogen). Samples from each siRNA condition were kept for immunoblot analysis. All secondary (Alexa conjugated) antibodies were purchased from Molecular Probes, Invitrogen.
The other antibodies used in this study are listed below: ARPC2/p34-Arc polyclonal
The siRES GFP PA -Arp3 or GFP PA -Arp3B lentiviral constructs were used to infect Lifeact-RFP stable HeLa cells (15) and cells expressing both proteins were selected using a combination of hygromycin and puromicin (100 µg/ml and 1 µg/ml respectively).

Live cell imaging and photoactivation.
Photoactivation experiments involving GFP PA -tags were performed as previously described (15). To visualise actin structures in live imaging, GFP-MICAL1 and GFP-MICAL2 stable cells were transfected with pE/L-LifeAct-iRFP670 using FUGENE (Promega) and infected with vaccinia virus as previously described (35
Lysates were normalised and incubated with GFP-Trap agarose beads (Chromotek) and pulldown was performed according to manufacturer's instructions.

Recombinant expression and purification of Arp2/3 complexes.
Human Arp3B was amplified by PCR using pLVX-GFP-Arp3B as a template and cloned into the BamHI/NotI sites of the pFL vector which also contains ArpC2 (15). The other baculovirus expression vectors and the Arp2/3 complex purification protocol were previously described (15).

Pyrene actin polymerisation, TIRF branching assays and analysis
Pyrene labelled and unlabelled actin monomers were mixed in G-buffer (

Reverse Transcription Quantitative PCR (RT-qPCR).
RNA extraction, cDNA synthesis and qPCR were performed as previously described (35)

Statistical Analysis
For actin tail length experiments, 10 cells per condition were acquired and 10 tails per cell were measured in each independent experiment (100 tails per condition in total). Data from three independent experiments were combined and the error bars represent the SEM for n = 300 tails per condition. When only 2 conditions were compared, Student t test was performed to determine statistical significance. When more than two samples were compared, Tukey's multiple comparisons test was used to determine statistical significances. All data concerning was analysed using Graphpad Prism 8.