The Actin Nucleator Cobl Is Controlled by Calcium and Calmodulin

Actin nucleation triggers the formation of new actin filaments and has the power to shape cells but requires tight control in order to bring about proper morphologies. The regulation of the members of the novel class of WASP Homology 2 (WH2) domain-based actin nucleators, however, thus far has largely remained elusive. Our study reveals signal cascades and mechanisms regulating Cordon-Bleu (Cobl). Cobl plays some, albeit not fully understood, role in early arborization of neurons and nucleates actin by a mechanism that requires a combination of all three of its actin monomer–binding WH2 domains. Our experiments reveal that Cobl is regulated by Ca2+ and multiple, direct associations of the Ca2+ sensor Calmodulin (CaM). Overexpression analyses and rescue experiments of Cobl loss-of-function phenotypes with Cobl mutants in primary neurons and in tissue slices demonstrated the importance of CaM binding for Cobl’s functions. Cobl-induced dendritic branch initiation was preceded by Ca2+ signals and coincided with local F-actin and CaM accumulations. CaM inhibitor studies showed that Cobl-mediated branching is strictly dependent on CaM activity. Mechanistic studies revealed that Ca2+/CaM modulates Cobl’s actin binding properties and furthermore promotes Cobl’s previously identified interactions with the membrane-shaping F-BAR protein syndapin I, which accumulated with Cobl at nascent dendritic protrusion sites. The findings of our study demonstrate a direct regulation of an actin nucleator by Ca2+/CaM and reveal that the Ca2+/CaM-controlled molecular mechanisms we discovered are crucial for Cobl’s cellular functions. By unveiling the means of Cobl regulation and the mechanisms, by which Ca2+/CaM signals directly converge on a cellular effector promoting actin filament formation, our work furthermore sheds light on how local Ca2+ signals steer and power branch initiation during early arborization of nerve cells—a key process in neuronal network formation.


Introduction
promotes Cobl's interactions with syndapin I and that Cobl, CaM, F-actin, and syndapin I accumulate at branch initiation sites.
Taken together, our study addresses the means of Cobl regulation and reveals several molecular mechanisms that enable the Ca 2+ /CaM signaling pathway to steer Cobl. With Cobl, we identified an actin filament-promoting effector by which local Ca 2+ /CaM-signaling directly powers cellular morphogenesis during early neuronal network formation.

Dendritic Protrusions Initiate from Cobl and F-actin Accumulations
The extremely arborized morphologies that neurons develop during their early morphogenesis are a prerequisite for the formation of all neuronal networks. By some, yet unknown means, the underlying reorganization of cell shape involves Cobl, a protein that has the ability to promote the formation of new actin filaments [12,14]. Dendritic arborization has furthermore been suggested to be controlled by local Ca 2+ influx. Indeed, 3-D-time-lapse calcium imaging in developing hippocampal neurons showed that protrusions, which formed in areas marked by transient Ca 2+ influx, were originating from sites enriched for F-actin. These F-actin-rich sites either already existed prior to the calcium signal ( Fig 1A) or F-actin accumulated after a calcium pulse (Fig 1B).
Analyzing the distribution and dynamics of the actin nucleator Cobl in developing neurons, we observed that the formation of protrusions from dendritic structures coincided with Cobl enrichment. Branching was a dynamic process with branches being initiated, shrinking back to the mother dendrite and being reinitiated until they were firmly established and grew out. Cobl accumulated in spatially restricted dendritic sites prior to the induction of almost all branching events (Fig 1C, arrows; S1 Fig; S1 Movie). Quantitative analyses showed that Cobl was highly enriched at branch initiation sites 30 s before initiation of protrusion. The intensity of GFP-Cobl at such sites was more than twice as high as at adjacent control region of interest (ROI), whereas GFP showed no intensity differences when control ROIs were compared to branch initiation sites ( Fig 1D).
Dual imaging of GFP-Cobl and LifeAct-RFP revealed that Cobl accumulations and dendritic branch inductions were accompanied by local F-actin formation. Interestingly, the signals of both Cobl and F-actin were particularly high at the base of initiated branches. Furthermore, we observed that the maximum of Cobl accumulation hereby usually preceded that of F-actin accumulation (Fig 1E; S2 Movie).
In line with the live imaging data, immunostainings of dendrites of developing neurons showed that sites marked by increased F-actin were often marked by protrusive morphology and displayed accumulations of endogenous Cobl (Fig 1F). Initiation of dynamic, dendritic protrusions (marked by green *) often is preceded by Cobl accumulation (white arrows). Retraction events are marked by red°a nd static phases with yellow I. Dendrite branch induction is a dynamic process with often several protrusive attempts until a dendritic branch is firmly established and strongly elongated. Three consecutive initiations of a protrusion from the same site are shown. Data (see S1 Fig for original data of the GFP channel) are shown in a color-coded manner (heat map purple to white, see legend for color coding). Bar, 5 μm. Please also see S1 Movie. (D) Quantitative analyses of GFP-Cobl enrichment at dendritic protrusion initiation sites 30 s prior to protrusion initiation in comparison to adjacent control regions of interest (ROIs). Errors represent standard error of the mean (SEM). Statistical significances, one-way ANOVA with Tukey's post test. GFP-Cobl, n = 25; GFP, n = 18. ***p < 0.001. For underlying data, see S1 Data. (E) Simultaneous recordings of GFP-Cobl and LifeAct-RFP show that initiation of dendritic protrusions coincides with Cobl accumulation followed by F-actin buildup at the dendritic base (arrows). Labeling as in C. Bar, 2 μm. Please also see S2 Movie. (F) Anti-Cobl and anti-MAP2-stained hippocampal neuron at DIV6 transfected with LifeAct-GFP to mark putative sites of dendritic branch induction show colocalization of Cobl with F-actin at actin-rich sites protruding from the dendrites (arrow). Bar  Immunohistological examinations showed that Cobl and CaM display overlapping expression. Cells with pronounced Cobl expression in the hippocampus were also marked by significant CaM expression (Fig 2E, arrows). The same was true for the cerebellum. In particular, Purkinje cell dendrites relying on Cobl for branching [15] showed marked anti-CaM immunosignals (Fig 2F, arrows).
These findings suggested Cobl to be a Ca 2+ /CaM-regulated cytoskeletal effector. Indeed, the Cobl/CaM interaction was also observable in coprecipitation analyses with endogenous Cobl from rat brain lysates. The Cobl/CaM interaction was strongly Ca 2+ dependent and occurred in a wide range of Ca 2+ concentrations (tested were physiological concentrations down to 2 μM) ( Fig 2G).
We next aimed at further corroborating the Cobl/CaM interaction by coimmunoprecipitations of the endogenous proteins with suitable anti-CaM antibodies. A subpool of endogenous Cobl indeed was specifically coimmunoprecipitated with CaM from rat brain lysate (Fig 2H). Calcium imaging using GCaMP5G showed that also Ca 2+ signals correlated with both Cobl accumulation and protrusion initiation (Fig 3B, arrow).

Cobl-Mediated Dendritic Arborization Requires Ca 2+ /CaM Signals
To experimentally test the effects of Ca 2+ /CaM signals on Cobl's functions in dendritogenesis, we next employed two different CaM inhibitors: W7 (N-(6-Aminohexyl)-5-chlor-1-naphthalinsulfonamide) [17] and CGS9343B (1,3- [4,l]-benzoxazepin-4-y1-methy1]-4-piperidinyl]-2H-benzimidazol-2-one(1:1) maleate) [18] (Fig 4). W7 and CGS9343B both completely suppressed Cobl-induced dendritic arborization in developing neurons (Fig 4A-4H). Also in control cells, W7 and CGS9343B incubation caused some reduction of dendritic branch points. These effects were smaller than the strong suppression of the Cobl overexpression phenotype indicating that the CaM inhibitors indeed suppress Cobl functions and do not act in a parallel pathway. The effect of W7 and CGS9343B in control cells may include an inhibition of the functions of endogenous Cobl that has been demonstrated to be a crucial factor in dendritic arborization [12] and is expressed throughout the development of the brain (S5 Fig). Similar to dissociated neurons, significant reductions of dendritic branch points by W7 and CGS9343B were also observed in Purkinje cells in developing cerebellar slice cultures (S6 Fig). Interestingly, the observed impairments mirror quite well the Coblloss-of-function phenotype in cerebellar slices [15].
Highlighting the neuronal morphology of primary rat hippocampal neurons at days in vitro (DIV)6 with PM-mCherry in 3-D-time-lapse studies showed many dynamic protrusions originating from Cobl-enriched sites in untreated neurons. After addition of CGS9343B, this dynamic behavior rapidly came to a standstill. Neuronal structures became static, and Coblenriched sites appeared less frequently (Fig 4I; S4 Movie). Quantitative analyses showed that the frequencies of protrusion initiation decreased to about 10% of the values of the respective cells before addition of CaM inhibitor (Fig 4J).
Ca 2+ Promotes the G-actin Binding of Cobl, While Ca 2+ /CaM Attenuates the Actin Binding of the First WH2 Domain To unravel the mechanisms of CaM's crucial role in Cobl-mediated dendritogenesis, we next mapped the CaM binding sites. Surprisingly, we identified multiple CaM interactions with Cobl. The N-terminus contains at least three independent CaM-binding areas (aa48-112, 147-176 and 175-229) (Fig 5A and 5B). Coprecipitation and corecruitment studies furthermore showed that CaM binding additionally involved regions in the middle of Cobl (aa750-1005) (Fig 5C) as well as in front of the . Bar, 2 μm. Please also see S3 Movie. (B) MIPs from 3-D-time-lapse recordings of the calcium sensor GCaMP5G and mCherry-Cobl reveal that local rises in calcium levels coincide with subsequent induction of Cobl accumulation and dendritic protrusions arising from Cobl-enriched sites (arrows). Bar, 5 μm. Inset is a high-intensity image of the protrusion. Cobl and CaM accumulations, respectively, at sites giving rise to protrusion are marked by arrows. Green *, protrusion that was initiated or grew when compared to previous image; (green *), minor, aborted protrusion; red°, protrusion that shrunk compared to previous image; yellow I, static protrusion.  [19][20][21].
Our observations raised the exciting possibility that Ca 2+ /CaM modulates Cobl's cytoskeletal functions. As neither full-length Cobl nor extended C-terminal parts of Cobl can be purified [12,14], and in vitro reconstitutions of actin polymerization are hampered by Ca 2+ -containing physiological buffers, we focused on actin coimmunoprecipitations and in vitro reconstitutions of actin binding to explore putative Ca 2+ /CaM-induced modulations of Cobl functions.
Coimmunoprecipitation of endogenous actin with GFP-Cobl 1001-1337 , GFP-Cobl 1176-1337 , and GFP-Cobl 1206-1337 unveiled that Ca 2+ promotes actin binding. This positive effect was independent of CaM binding and also independent of the first WH2 domain neighboring the CaM binding interface but prominently involved the second and third WH2 domain of Cobl (Fig 5E-5G). Interestingly, the Ca 2+ -mediated increase of actin binding was not reversed upon subsequent lowering of Ca 2+ levels ( Fig 5F).
Since the increased actin association was more striking for GFP-Cobl 1176-1337 lacking the CaM binding site than for GFP-Cobl 1001-1337 containing an interface for direct CaM binding (S9 Fig), we next addressed the effects of Ca 2+ /CaM specifically on the first WH2 domain. Interestingly, the first WH2 domain (Cobl 1176-1224 ) required the CaM-binding region (Cobl 1001-1176 ) for coimmunoprecipitation of actin (S10 Fig). This specific coimmunoprecipitation of actin by Cobl 1001-1224 was strongly suppressed by 500 μM Ca 2+ , which may only be reached in direct vicinity of Ca 2+ channels, as well as by lower Ca 2+ levels (2 μM), which are more commonly and more widely reached in neurons ( Fig 5H).
Importantly, in vitro reconstitutions with purified proteins demonstrated that this suppression of actin binding of Cobl's first WH2 domain solely involved actin, Cobl and Ca 2+ /CaM. In the presence of CaM, we observed a Ca 2+ -specific CaM/Cobl complex formation and a statistically significant reduction of actin binding. In contrast, in the absence of CaM, no such difference between Ca 2+ and Ca 2+ -free conditions was observed (Fig 5I; S11 Fig). Thus, the suppression of the actin binding of the first WH2 domain involves direct Ca 2+ /CaM association and relies on the Ca 2+ sensor CaM.
The first WH2 domain is crucial for Cobl's actin filament promoting function [12,14], and actin filament formation as well as Cobl accumulation was observed at the initiation point of Ca 2+ -triggered, newly forming dendritic protrusions (Fig 1). Both observations strongly suggested that actin filament formation at branch initiation sites plays a supporting role during dendritic arborization. It was therefore interesting that Ca 2+ was overall promoting actin association of the Cobl C-terminus, despite a simultaneously occurring Ca 2+ /CaM and dendritic branch points (H) normalized to the GFP+DMSO controls of each assay. Note that Cobl-mediated dendrite formation is suppressed by the CaM inhibitors W7 and CGS9343B. Data represent mean ± SEM. GFP +DMSO, n = 197; GFP+W7, n = 120; GFP+CGS9343B, n = 169; GFP-Cobl+DMSO, n = 101; GFP-Cobl+W7, n = 53; GFP-Cobl+CGS93943B, n = 80. Statistical significances (to control, marked above column; between other conditions, indicated by lines) were determined using one-way ANOVA with Tukey's post-test. *p < 0.05; **p < 0.01; ***p < 0.001. (I) 3-D-time-lapse imaging of GFP-Cobl dynamics and neuronal morphogenesis visualized by PM-mCherry before and after incubation with the CaM inhibitor CGS9343B. Shown is a selection (times as indicated) of MIPs recorded by spinning disc microscopy. Green *, protrusion that was initiated or grew when compared to previous image (examples numbered to allow for tracking); red°, protrusion that shrunk compared to previous image; yellow I, static protrusion. Note that whereas most neuronal structures were dynamic, CGS9343B addition (red line; 11:20 min:s) impeded this morphological dynamics and neuronal structures largely were static until the end of recording. Bar, 5 μm. Please also see S4 Movie. (J) Quantitation of the frequencies of protrusion initiation before and after addition of DMSO and DMSO+inhibitor (CGS9343B), respectively. DMSO, n = 12; DMSO+inhibitor, n = 10 dendrite sections. For data underlying G, H, and J see S1 Data. Statistical significances, one-way ANOVA with Tukey's posttest. ***p < 0.001. The Actin Nucleator Cobl Is Controlled by Calcium and Calmodulin binding-mediated suppression of the actin association of the first WH2 domain. This suggested that actin binding may be transiently increased even further once Ca 2+ levels drop again, as under such conditions the dissociation of CaM may release the suppression. In line with a transient effect, the Ca 2+ /CaM-mediated block of actin binding to the first WH2 domain indeed was fully reversible upon reduction of Ca 2+ levels ( Fig 5J).

Ca 2+ /CaM Regulation of Cobl Homology Domain-Mediated Membrane Association
The observation that CaM inhibition impaired Cobl accumulation at branch initiation sites (Fig 4I and 4J) prompted us to study the influence of Ca 2+ /CaM on Cobl's association with the cell cortex. This process involves Cobl Homology domain interactions with syndapin I [16]. Specific coimmunoprecipitation as well as specific reconstitutions of Cobl-CaM interactions in intact cells confirmed that also the CaM interactions with the Cobl Homology domain are of relevance in vivo (Fig 6A-6C).
We thus addressed the exciting hypothesis that Ca 2+ /CaM signaling may not only control Cobl's actin cytoskeletal functions but may also orchestrate Cobl's membrane association. Whereas we were unable to purify the full Cobl Homology domain, we succeeded in purifying an alternative protein (TrxHis-Cobl 54-450 ). In vitro reconstitutions with liposomes revealed that the N-terminus of Cobl itself has membrane-binding activity and therefore floated with liposomes irrespective of calcium presence (Fig 6D and 6E).
Interestingly, Ca 2+ /CaM addition effectively suppressed the direct membrane-binding ability of the Cobl Homology domain (Fig 6F; upper middle panel).

Ca 2+ /CaM Promotes Cobl/Syndapin I Complex Formation
Cortical targeting of cytoskeletal effectors is a key aspect in shaping cells. The Ca 2+ /CaM-mediated suppression of Cobl's direct lipid association thus was puzzling. We therefore analyzed a putative regulation of Cobl/syndapin I interactions. In vitro reconstitutions demonstrated that direct and simultaneously occurring interactions of Cobl with CaM and syndapin I give rise to complexes containing all three components (Fig 7A and 7B).
Strikingly, the anchoring of Cobl to membranes via the F-BAR domain protein syndapin I was not suppressed by Ca 2+ /CaM addition. Complexes containing all three components, i.e., Cobl, CaM, and syndapin I, floated with liposomes ( Fig 7C-7E).
Together, the formation of complexes composed of all three components and their ability to associate with membranes suggested that Cobl's intrinsic lipid association constitutively ensures some Cobl presence at the cell cortex. Upon association of Ca 2+ /CaM, this ability of Cobl is switched off. As a consequence, F-BAR domain-mediated membrane associations by The Actin Nucleator Cobl Is Controlled by Calcium and Calmodulin syndapin I [22,23] start to dominate the spatial control of Cobl at the cell cortex. Consistent with such a scenario, we observed that addition of the Cobl-binding SH3 domain of syndapin I did not suffice for restoring Cobl's membrane association in the presence of Ca 2+ /CaM ( Fig 7F).
Thus, upon Ca 2+ /CaM association, Cobl localization becomes fully dependent on SH3 domain interactions and on F-BAR-mediated membrane association of syndapin I. This regulatory mechanism would be even more effective if syndapin I associations were promoted upon Ca 2+ /CaM. In order to address this directly in vivo, we conducted quantitative coimmunoprecipitation studies. We observed a Ca 2+ -mediated increase of Cobl 1-408 /syndapin  The Actin Nucleator Cobl Is Controlled by Calcium and Calmodulin I complex formation at both 500 μM (not shown) and 2 μM calcium (+55.3 ± 17.0%; p < 0.05) when compared to conditions without calcium (Fig 8A and 8B; S12 Fig).
This increase was, to a large extent, reversible. At least upon prolonged incubation with ethylenglycol-bis(aminoethylether)-N,N,N 0 ,N 0 -tetraacetic acid (EGTA) after Ca 2+ stimulation, ). The syndapin I interaction with Cobl is promoted upon activation of Ca 2+ signaling (2 μM Ca 2+ ) (indicated by the green upright arrowhead). For further confirmation of the Ca 2+ /CaM association-dependent increase of the syndapin I interaction, also see coimmunoprecipitations under alternative conditions shown in S12 Fig. (B,C) Quantitative evaluations of coimmunoprecipitated syndapin I normalized to immunoprecipitated Cobl and displayed as mean percental difference ± SEM from the respective Ca 2+ -free conditions show that the syndapin I interaction is reversibly promoted by increasing calcium. Note that syndapin I coimmunoprecipitation by the CaM bindingdeficient Cobl mutant Cobl 1-408ΔCaM is insensitive to changes of calcium levels. The observed regulation of Cobl/syndapin I complex formation thus requires the CaM binding interface of Cobl. GFP-Cobl 1-408 /Flag-Sdp I +/− Ca 2+ , n = 3 each; with Ca 2+ /EGTA, n = 2. GFP-Cobl 1-408ΔCaM /Flag-Sdp I +/− Ca 2+ , n = 3 each; with Ca 2+ /EGTA, n = 2. Statistical significances were calculated by one-way ANOVA with Tukey's post-test. *p < 0.05. (D,E) Coimmunoprecipitations from rat brain lysates (D) demonstrate the promotion of endogenous Cobl/syndapin I complexes upon calcium addition (highlighted by green upward arrowhead; E, quantitation; n = 2. For data underlying B, C, and E, see S1 Data. syndapin I coimmunoprecipitation intensities only remained moderately increased and were not significantly different from control anymore (+21.0 ± 11.4%; Fig 8A and 8B).
The striking promotion of Cobl's association with syndapin I was dependent on Cobl's ability to associate with the calcium sensor CaM because a Cobl mutant incapable of binding to CaM (Cobl 1-408ΔCaM ) did not respond to changes of Ca 2+ levels. Instead, Cobl 1-408ΔCaM coimmunoprecipitated constant amounts of syndapin I (Fig 8A and 8C).
Endogenous coimmunoprecipitations from rat brain lysates showed a corresponding calcium-mediated increase of Cobl/syndapin I complex formation that was consistent with thẽ 60% increase of syndapin I association in the heterologous coimmunoprecipitations using the Cobl Homology domain described before (Fig 8D and 8E).

Syndapin I Coincides with Cobl Accumulations at Sites of Dendritic Membrane Protrusion
In primary neurons undergoing dendritogenesis (DIV7), GFP-Cobl and Flag-mCherry-syndapin I colocalized at discrete sites within the dendritic arbor ( Fig 9A). 3-D-time-lapse recordings showed that protrusions emanated from sites that were enriched for both syndapin I and Cobl. Both proteins showed very good spatial overlap at nascent dendritic branch points. Similar to the dynamic behavior of Cobl during dendritic branch induction, also syndapin I was found to accumulate at sites of branch initiation shortly before branch induction started. After protrusions had been established and grew, syndapin I and Cobl both redistributed to a more disperse localization in the mother dendrite and in the formed branch ( Fig 9B; S13 Fig).
Immunostainings of endogenous Cobl and syndapin I in developing neurons confirmed the presence of syndapin I-enriched sites in dendrites that also showed accumulations of anti-Cobl signals. Often such sites were not symmetric but protruded from one side of the dendrite and may thus represent initiation sites for dendritic branching (Fig 9C).

CaM Association Is Critical for Cobl-Mediated Dendrite and Dendritic Branch Formation
To address whether CaM association is crucial for orchestrating Cobl functions during dendritogenesis, we next decided to employ GFP-Cobl mutants lacking N-and C-terminal CaM- binding interfaces or combinations thereof (Fig 10A). The importance of the most C-terminal CaM-binding area identified (Fig 5) hereby was addressed in form of two separate mutants, as further biochemical experiments revealed that the interface Cobl 1001-1176 contained at least two areas that independently interact with CaM (Cobl 1001-1101 and Cobl 1100-1176 ). This increased the number of independent CaM interface on Cobl to at least six (Fig 10A; S14 Fig).
Cobl ΔCaM-N,C lacked all CaM-binding areas that we had identified and consistently did not associate with CaM anymore (Fig 10B). The Cobl N-terminus lacking the CaM interfaces (Cobl ΔCaM-N ) still associated with the two other components that are known to bind to the Cobl Homology domain and are critical for the functions of Cobl in vivo, Abp1 and syndapin I [15,16] (S15 Fig). Thus, this mutant should allow for dissecting the known requirements of syndapin I and Abp1 association from a putative importance of CaM associations in functional studies. Importantly, despite preserved Abp1 and syndapin I interactions, Cobl ΔCaM-N,C overexpression did not result in the extensive dendritic arborization observed upon overexpression of wild-type Cobl. Thus, the N-terminal and the more C-terminal CaM-binding interfaces are crucial for inducing Cobl overexpression phenotypes (Fig 10C-10L).
We next tested the functional importance of the identified CaM interfaces individually. Similar to Cobl ΔCaM-N,C , also the individual deletions Cobl ΔCaM-N and Cobl ΔCaM-C were unable to trigger dendritic arborization (Fig 10F, 10G, 10K and 10L). Thus, Cobl functions seem to require both CaM binding sites of the N-terminal part of Cobl, which we showed to be involved in modulating the syndapin I interactions (Fig 8), as well as C-terminal CaM binding sites, which we unraveled to modulate actin associations (Fig 5).
Cobl with even smaller deletions (Cobl ΔCaM-C ', Cobl ΔCaM-C ", and Cobl ΔCaM-C "') also failed to give rise to the Cobl gain-of-function phenotype. Instead, the morphology of transfected neurons remained indistinguishable from control cells (Fig 10H-10L). Therefore, the CaM binding area Cobl 1001-1176 addressed in our mechanistic studies of Ca 2+ /CaM-mediated actin association was critically required for Cobl-mediated dendritic branching.
To address whether these data reflect a requirement of both N-and C-terminal CaM binding areas for physiological Cobl functions, we next conducted Cobl loss-of-function experiments and corresponding rescue experiments (Fig 10M-10V). We subjected cerebellar slices to gene gun transfections with GFP-reported plasmids. As described previously [15], Cobl RNAiimpaired dendritic branching of Purkinje cells. This Cobl loss-of-function phenotype was rescued by resupplying the cells with RNAi-insensitive, wild-type Cobl (GFP-Cobl Ã ) (Fig 10M-10O and 10V).
In contrast, substitution of Cobl with any of the four CaM-binding-deficient mutants Cobl ΔCaM-N,C Cobl ΔCaM-N , Cobl ΔCaM-C , and Cobl ΔCaM-C ' did not only fail to rescue the Cobl loss-of-function phenotype, but impaired dendritogenesis further (Fig 10P-10S and 10V). Cobl ΔCaM-C " and Cobl ΔCaM-C "' also were unable to rescue the Cobl loss-of-function defects in dendritic branching in the developing slice cultures (Fig 10T-10V).
Thus, Cobl's crucial role in dendritic branch induction critically relies on both the N-and the C-terminal CaM association sites, for which we have revealed the molecular mechanisms of Cobl regulation.

Discussion
Shaping neurons demands that Ca 2+ signals are converted into locally restricted and transient activity of force-generating effectors [24]. The activity of such effectors must be targeted to the dendritic plasma membrane and must cease to exist once a branch is induced successfully. Whereas other actin cytoskeletal effectors are controlled by Rho-type GTPases, we here describe a direct interaction of the calcium sensor protein CaM with a powerful cytoskeletal effector remodeling neuronal morphology, the actin nucleator Cobl (Fig 11).
To our knowledge, only few direct links of Ca 2+ /CaM to the actin cytoskeleton have been discovered in neurons thus far. However, none of them explains how Ca 2+ signals may trigger dendritogenesis. Neither the described CaM associations with brain-enriched spectrin isoforms [25] nor competitive binding of the F-actin bundling protein α-actinin and CaM to NMDA receptors and L-type Ca 2+ channels [26,27] offer obvious mechanisms bringing about dendritic arborization.  The Actin Nucleator Cobl Is Controlled by Calcium and Calmodulin With Cobl, we have identified a CaM-associating component that effectively promotes the local formation of actin filaments. Actin cytoskeletal forces have the power to shape cells and cellular compartments. Whether other WH2 domain-containing actin nucleators [10,11] or actin filament formation via the Arp2/3 complex and/or Formins also are directly controlled by Ca 2+ /CaM remains to be addressed.
We found by yeast-2-hybrid, coprecipitations, coimmunoprecipitations, and corecruitment studies in intact cells that CaM associates with Cobl. Reconstitutions with purified proteins demonstrated that Cobl's interactions with CaM are direct. The Ca 2+ dependency of the interactions strongly suggested that Cobl/CaM complexes are involved in translating Ca 2+ signals sensed by CaM into cellular answers. Colocalizations in different parts of the brain and a dynamic coappearance of Cobl and CaM at induction sites of dendritic protrusion support this hypothesis.
Several lines of evidence from functional studies underscore the importance of Cobl/CaM interactions during dendritogenesis: CaM inhibitors fully suppressed Cobl-mediated dendrite formation and branching in quantitative end-point analyses of fixed neurons and in live imaging studies. Our data strongly argue that a direct association and not merely CaM signaling is required, as overexpression of Cobl mutants incapable of associating with CaM failed to mimic Cobl-mediated effects on dendrite formation. Furthermore, re-expression of such Cobl mutants failed to rescue the Cobl loss-of-function phenotype in dendritic branching of cerebellar Purkinje cells. These findings clearly point to a critical function of CaM associations with Cobl.
Our analyses revealed multiple independent CaM binding areas. At least three of them reside in the Cobl Homology domain. In addition, at least three further binding motifs are located in Cobl's C-terminal part. Usually, the CaM lobes wrap around target segments, burying them in a hydrophobic channel, and thereby enforce alter target conformations [28]. Consistent with such putative conformational changes, we observed dramatically altered properties of both the N-and C-terminus of Cobl upon Ca 2+ /CaM signaling. Overall, calcium strongly promoted the actin association with the C-terminus of Cobl in a manner that was independent of CaM association and independent of the first WH2 domain of Cobl. In principle, such Ca 2+mediated effects could be due to different mechanisms, i) Ca 2+ modulates the properties of actin (G-actin or G-actin/F-actin balance), and this is reflected by changed association with Cobl WH2 domains, ii) Ca 2+ modulates the properties of the Cobl WH2 domains number 2 and 3, and this leads to changed actin affinities or iii) Ca 2+ signals to further cellular components, and these either modulate actin properties, Cobl properties, or both.
The observed Ca 2+ -mediated increase in Cobl's overall actin binding occurred despite a simultaneous suppression of the actin binding of the first WH2 domain upon CaM association with neighboring sites, as demonstrated by quantitative coimmunoprecipitation studies and by in vitro reconstitutions with purified components. In line with CaM acting as calcium sensor protein, this suppression was released with decreasing Ca 2+ levels, whereas the Ca 2+ -promoted actin binding of the Cobl C-terminus persisted. Since all three WH2 domains need to work together and the first WH2 domain is crucial for actin filament formation [12,14], the release of the suppression of the first WH2 domain after a transient Ca 2+ signal would elegantly ensure that Cobl responds to calcium transients. In contrast to simple on/off mechanisms, transient activations would also exclude Cobl activity during longer lasting NMDA and Ca 2+ /CaM signaling, such as during excitotoxicity conditions. Indeed, such conditions do not promote filament formation but are marked by filament loss [29]. Related to high Ca 2+ levels during excitotoxicity-reducing actin dynamics and filament formation, high calcium was reported to inhibit the F-actin-driven formation of filopodia-like dendritic spines and stabilized them, whereas lower calcium signals were reported to promote the formation of these F-actin-rich structures. It therefore is conceivable that, besides the crucial role in dendritogenesis, Coblmediated functions and Ca 2+ /CaM-mediated control of Cobl may in addition play some role in synapse formation and/or plasticity processes, as these involve both Ca 2+ signaling and actin cytoskeletal reorganizations [30].
We observed local F-actin accumulations at the base of dendritic protrusions. We have demonstrated earlier that Cobl overexpression promotes dendritic arborization, and Cobl lossof-function results in a reduction of branches [12,15]. Correlative 3-D-time-lapse studies, inhibitor studies, and mutational analyses demonstrated that dendritic branching is associated with Cobl, F-actin and syndapin I at branch initiation sites and is controlled by changing calcium levels, by Ca 2+ /CaM signaling and by direct Cobl association of CaM.
We have identified a direct plasma membrane association of Cobl that also was CaM-regulated. This ability of the Cobl Homology domain was suppressed upon CaM association. At the same time, Ca 2+ /CaM promoted Cobl's association with syndapin I [16,31] and thereby promoted indirect membrane associations of Cobl. Direct membrane association of Cobl may ensure its general availability at the plasma membrane and may also explain why syndapin I loss-of-function only partially suppressed Cobl's cortical localization [16]. Upon CaM binding, syndapin I interactions will increasingly influence Cobl's localization. Indeed, our studies revealed that dendritic branching events correlated with accumulations of both syndapin I and Cobl.
In line with this conclusion, syndapin I has been demonstrated to be involved in dendritogenesis [16]. Interestingly, we observed accumulations of syndapin I specifically at nascent dendritic branch sites. It thus seems that F-BAR domain-mediated membrane curvature sensing and/or induction by syndapin I [22,32] spatially steers the actin nucleator Cobl at the cell cortex. Consistently, mutational overexpression analyses in neuronal cultures as well as rescue experiments of cerebellar Cobl loss-of-function revealed that CaM binding to the actin-binding C-terminal part and to the syndapin I-binding N-terminal part of Cobl were crucial for Cobl functions in dendritogenesis.
Together, our work unveils that Cobl is regulated by the calcium sensor protein CaM and reveals the Ca 2+ /CaM-controlled molecular mechanisms that are crucial for Cobl's cellular functions. The regulation by Ca 2+ /CaM seems to distinguish Cobl from established actin nucleators, such as the Arp2/3 complex and Formins, which are directly and indirectly regulated by Rho-type GTPases.
Our examinations of the Ca 2+ /CaM-mediated mechanisms that control Cobl's activity in neurons furthermore provided deep insights into how local Ca 2+ signals steer and power branch initiation during early arborization of nerve cells.
Correct cloning by PCR was verified by sequencing in all cases.
Rabbit skeletal muscle actin was from Cytoskeleton. GST-and TrxHis-tagged fusion proteins were purified from E. coli as described previously [16,39].
Analyses addressing direct interactions of GST-Cobl 1001-1176 with CaM were done in lysis buffer with 150 mM NaCl containing 0 and 1 μM Ca 2+ , respectively (set according to [41]). Eluted proteins were analyzed by SDS-PAGE and subsequent anti-TrxHis and anti-GST immunoblotting.

Preparation of HEK293 Cell Lysates
24-48 h after transfection, HEK293 cells were washed with PBS, harvested and subjected to sonification for 10 s and/or lysed by incubation in lysis buffer containing EDTA-free protease inhibitor Complete (Roche) and 120-150 mM NaCl for 20 to 30 min at 4°C. Cell lysates were obtained as supernatants from centrifugations at 16,000 xg (20 min at 4°C).

Coprecipitation of Proteins from HEK293 Cell Lysates
For coprecipitation experiments, extracts from HEK293 cells expressing different GFP fusion proteins were incubated for 3 h at 4°C with purified GST-fusion proteins immobilized on glutathione sepharose beads (GenScript) as described [33]. Bound protein complexes were eluted with 20 mM-reduced glutathione, 120 mM NaCl, 50 mM Tris/HCl pH 8.0.
For coprecipitations with CaM, HEK293 cell lysates were prepared in EGTA-free lysis buffer containing 150 mM NaCl, and EDTA-free protease inhibitor cocktail and 200 μM calpain inhibitor I. Cell lysates were supplemented with either 1 mM EGTA or to be tested Ca 2+ concentrations.
For binding curves, Ca 2+ concentrations ranging from 0 to 500 μM were set according to [41]. After incubation with 25 μl CaM-sepharose 4B for 3 h at 4°C and washing, bound proteins were isolated by boiling in SDS sample buffer.
Lysates, supernatants, and eluates were analyzed by immunoblotting using anti-GST and anti-GFP antibodies, respectively.
Coimmunoprecipitations of endogenous actin together with GFP and GFP-tagged Cobl deletion mutants were done according to [12] with slight modifications. Lysates of HEK293 cell were incubated in EGTA-free lysis buffer containing 100 mM NaCl, 5 mM DTT, 200 μM calpain inhibitor I and varying amounts (2 μM, 500 μM) of Ca 2+ or no free Ca 2+ (EGTA addition) with 5 μg rabbit anti-GFP antibody/well (6-well plate) for 3 h at 4°C.
Reversibility of Ca 2+ -induced suppression of actin binding of Cobl 1001-1224 by CaM association was tested by incubating 2 μM Ca 2+ -treated samples subsequently with 1 mM EGTA for 2 h. The CaM-independent Ca 2+ -induced (500 μM Ca 2+ ) increase of actin binding to Cobl's Cterminal part (Cobl 1176-1337 ) was tested for reversibility by subsequent EGTA addition (5 mM EGTA final). Antibody-associated protein complexes were isolated with protein A/G-agarose (2 h, 4°C), washed with a buffer with the respective Ca 2+ concentrations and eluted by boiling in a mix of 8 M urea and 4x-SDS-sample buffer. The eluates were immunoblotted with anti-GFP and anti-actin antibodies and analyzed quantitatively using fluorescently labeled secondary antibodies and a LI-COR Odyssey System.
Quantitative coimmunoprecipitation analyses of Flag-syndapin I with GFP-Cobl proteins were done similarly except that the lysis buffer lacked DTT and contained 75 mM NaCl and protein A-agarose (Santa Cruz Biotechnology) was used. Confirmatory experiments were additionally done at 100 mM NaCl for consistency with previously published conditions for heterologous syndapin I/Cobl coimmunoprecipitations [16].
Coimmunoprecipitations of endogenous Cobl and CaM were performed using rat brain lysates in lysis buffer with 30 mM NaCl, 500 μM Ca 2+ and 200 μM calpain inhibitor I using mouse anti-CaM (G-3) antibodies according to coimmunoprecipitation procedures described [16]. Coimmunoprecipitations of endogenous Cobl and syndapin I were performed using rat brain lysates in lysis buffer with 30 mM NaCl and 400 μM calpain inhibitor I with and without 2 μM Ca 2+ using guinea pig anti-Cobl DBY antibodies as described [16], except that antibodies were added to brain lysates and then isolated with protein A-agarose.
The amounts of coimmunoprecipitated proteins under different conditions were normalized to the amount of immunoprecipitated GFP-Cobl and Cobl, respectively, and expressed as percent difference from Ca 2+ -free conditions. Statistical analyses were performed using oneway ANOVA with Tukey's post-test. Ã p < 0.05, ÃÃ p < 0.01, ÃÃÃ p < 0.001.

Cell culture, Transfection, and Immunostaining
Culturing of HEK293 and COS-7 cells and immunolabeling were essentially as described [42]. HEK293 and COS-7 cells were transfected using TurboFect (Thermo Scientific). Mitochondria of COS-7 cells were stained with 0.2 μM MitoTracker Deep Red FM (Molecular Probes) in medium at 37°C for 1 h and cells were subsequently fixed with 4% PFA for 7 min.
Phalloidin stainings and antibody incubations were done in the same buffer without Triton X-100 according to [42,44].

Immunolabeling of Mouse Brain Sections
Brain sections of adult (7 weeks) male mice were prepared and immunolabeled as described [15].

RT-PCR from Murine Brain
Brain preparations at different developmental stages, mRNA and cDNA preparation, as well as RT-PCR, were done according to [15] using the primers GCTCCGGAAGACTGCAGAACA (forward-WH2; positioned at exon border 12/13) and CGAGCAAGGGAACCTTTCTTAGTC (reverse-WH2; positioned at exon border 14/15) for Cobl detection and the primers ATT-GACCTCAACTACATGGTCTACA (forward) and CCAGTAGACTCCACGACATACTC (reverse) for GAPDH as control.

Spinning Disc Live Microscopy of Developing Neurons
Primary hippocampal neurons undergoing dendritic arbor formation were transiently transfected with Lipofectamine 2000 at DIV6. For imaging, the culture medium was replaced by 20 mM HEPES pH 7.4, 0.8 mM MgCl 2 , 1.8 mM CaCl 2 , 5 mM KCl, 140 mM NaCl, 5 mM D-glucose (live imaging buffer) adjusted to isoosmolarity using a freezing point osmometer (Osmomat 3000; Gonotec). For inhibitor studies, 10 μM (final) CaM inhibitor CGS9343B (Tocris) was used in DMSO and accompanied with the respective solvent control (0.1% DMSO final). Live imaging was conducted 16-28 h after transfection in an open coverslip holder placed into a temperature-and CO 2 -controlled incubator built around a spinning disc microscope. The microscope was a motorized Axio Observer combined with a spinning disc unit CSU-X1A 5000 and equipped with a 488 nm/100 mW OPSL laser, a 561 nm/40 mW diode laser and a QuantEM 512SC EMCCD camera (Zeiss). Images were taken as Z-stacks (stacks of 7-17 images at Z-intervals of 0.31 μm depending on cellular morphology) at time intervals of 10 s and 3 s (Ca 2+ imaging) with exposure times of 50-200 ms and 3%-12% laser power using a C-Apochromat objective (63x/1.20W Korr M27; Zeiss).
Image processing was done using ZEN, Imaris software, and Adobe Photoshop. Quantitative comparisons of signal intensities at dendritic branch initiation sites with signal intensities at dendritic sites that were not branching were done by placing a circular ROI at the branch initiation site (covering the dendrite diameter) and on an adjacent dendrite section (distance 2 ROI diameters). The fluorescence signal intensities of GFP and GFP-Cobl were measured at both sites 30 s prior to initiation of protrusion and expressed as relative enrichments relative to control ROI.
Frequencies of protrusion initiation from dendrite sections of neurons incubated with DMSO control and CaM inhibitor CGS9343B in DMSO were compared to those prior to treatment and expressed as protrusions per 10 μm dendrite section and min.

Quantitative Analyses of Dendrites of Hippocampal Neurons in Culture and of Purkinje Cells in Cerebellar Slice Cultures
Dendrite analyses of transiently transfected (DIV4) hippocampal neurons immunostained for MAP2 (NM_013066.1; GI:6981181) were performed with !2 independent neuronal preparations on 2-6 independent coverslips per condition in each assay at DIV6. Neurons were sampled systematically on each coverslip. Morphometric measurements were based on anti-MAP2 immunolabeling and performed with ImageJ according to [16,44]. Preparation of cerebellar slices, gene gun transfection and morphometric analyses of Purkinje cell dendrites in the Molecular Layer of the cerebellum were done as described [15].
Supporting Information S1 Data. Numerical information underlying all quantitative analyses presented in the main and the supplementary figures in the Supporting Information. S1 Data is an Excel file containing all data underlying Fig 1D; Fig 4G,  Note that both Cobl fusion proteins bind to CaM in a specific manner (A-C) and that quantitative Western blotting analyses (D,E) show that half-maximal binding is already reached at 0.68 μM and 0.95 μM Ca 2+ , respectively. Please also note that 2 μM Ca 2+ used in some biochemical assays of this study corresponds to about 80%-90% of maximal binding observed and that 500 μM Ca 2+ ensures plateau levels of CaM association. GFP-Cobl 106-324 , n = 4; GFP-Cobl 1001-1337 , n = 7; GFP, n = 3. For data underlying D and E see S1 Data.