Formin 3 directs dendritic architecture via microtubule regulation and is required for somatosensory nociceptive behavior

Dendrites function in the reception, integration and propagation of sensory or synaptic information and thereby mediate neural computation. Specification and maturation of diverse dendritic architectures is subject to complex intrinsic and extrinsic regulatory programs that modulate the formation and plasticity of a functional nervous system (Lefebvre et al., 2015). An important convergent regulatory target of extrinsic signaling and intrinsic factors is modulation of the neuronal cytoskeleton, which ultimately contributes to dendritic form and function (Das et al., 2017; Nanda et al., 2017). Thus, elucidating the molecular bases underlying dendritic cytoskeletal organization is crucial to uncovering principles governing neuronal diversity, as well as potential mechanistic links between architecture and behavior.

Regulation of the cytoskeleton is important for the establishment, maintenance and plasticity of neural morphology, as well as function (Franker and Hoogenraad, 2013). Upon achieving a mature shape, dendritic arbors retain a certain degree of plasticity that occurs via dynamic cytoskeletal modulation, thereby contributing to refinements in neural connectivity. Moreover, stabilization of mature dendrites is required for proper function, as morphological destabilization can lead to functional impairments in neurotransmission, result in neurodegeneration, disrupt organelle trafficking, and in the case of sensory neurons, disrupt responses to environmental stimuli, leading to behavioral defects such as peripheral insensitivity (Nanda et al., 2017; Honjo et al., 2016). Architectural organization and dynamic modulation of cytoskeletal fibers is controlled by a vast array of regulatory factors that impact assembly, disassembly, bundling, severing, stabilization and motor-based transport (Coles and Bradke, 2015; Kapitein and Hoogenraad, 2015). In addition, defects in cytoskeletal-mediated motor-based transport result in loss of dendritic identity, defects in neuronal polarity, and severe impairments in dendritic development (Rolls, 2011; Nanda et al., 2017).

Formins represent an important family of evolutionarily conserved, multi-functional cytoskeletal regulators (Goode and Eck, 2007). Formins are multidomain, dimeric functional molecules characterized by conserved Formin homology (FH) domains (Higgs, 2005). The functional role of formins in actin polymerization has been well documented, wherein the FH2 domain initiates actin nucleation and remains bound at the barbed end of an F-actin filament, moving along the growing filament and promoting elongation by preventing access of capping proteins (Chhabra and Higgs, 2007). F-actin elongation is further enhanced by association of Profilin with the formin FH1 domain (Kovar, 2006). Unlike other actin nucleators, such as Arp2/3, that form branched actin filaments, formins assemble linear actin filaments (Evangelista et al., 2002). Despite their basic function and properties, formins vary significantly; for instance, some bundle actin filaments, some sever or depolymerize actin filaments, and in recent studies formins have been implicated in regulation of microtubule (MT) organization and dynamics (Bartolini and Gundersen, 2010; Breitsprecher and Goode, 2013; Roth-Johnson et al., 2014). Formins appear to stabilize MTs both through their direct binding and/or by altering their post-translational state (Gaillard et al., 2011; Thurston et al., 2012). Recent studies have begun to elucidate the roles of formins in neural development, including dendrite morphogenesis (Galbraith and Kengaku, 2019); however, much remains unknown regarding the mechanistic roles of formins in dendritic arborization.

Here, we provide mechanistic links between formin-mediated dendritic cytoskeletal regulation and nociceptive behavioral sensitivity. Specifically, we demonstrate that Formin 3 (Form3) is required for dendritic arbor growth, with a major role in regulating a population of stable dendritic MTs. Moreover, form3 mutants exhibit defects in dendritic localization of mitochondria and satellite Golgi. Biochemical analyses further reveal that Form3 directly interacts with MTs via FH1-FH2 domains. Intriguingly, defects in INF2 have been causally linked to Charcot–Marie–Tooth (CMT) sensory neuropathies known to result in impaired peripheral sensitivity. Disruption of form3 in nociceptive neurons severely impairs noxious heat-evoked behaviors and leads to mislocalization of the TRPA channel Painless. Moreover, form3 mutant defects in nociceptive behavioral sensitivity and reductions in stable dendritic MTs can be partially recovered by expression of INF2 FH1-FH2 domains. Collectively, our findings suggest form3 and INF2 have conserved functions that may provide mechanistic insights into potential etiologies underlying CMT sensory neuropathies.

Molecular control of cytoskeletal organization and dynamics is essential for specification, stabilization and modulation of dendritic architecture. We previously conducted a systematic neurogenomic-driven genetic screen to identify putative cytoskeletal regulators of dendritic arborization using Drosophila Class IV (CIV) nociceptive sensory neurons as a model system (Das et al., 2017). We identified CIV expression of multiple members of the Formin gene family, including form3 and Frl (Das et al., 2017). Formins have been demonstrated to function by regulating the F-actin and MT cytoskeletons; however, their role(s) in neuronal development, and more specifically dendritogenesis, remain incompletely understood (Galbraith and Kengaku, 2019). Therefore, we performed a pilot RNAi screen to examine potential functional roles of the six Drosophila formins. Our screen revealed that form3 knockdown (form3-IR) produced severe defects in CIV dendritic arborization (Fig. 1B; Fig. S1A-H). Based on these results, we chose to conduct more in-depth phenotypic studies of Form3 function in dendritic morphogenesis.

CIV-specific form3-IR expression led to severe dendritic reductions, resulting in a highly rudimentary arbor with pronounced defects in distal higher order branching (Fig. 1A-B′,E-G,I). Dendritic branch order analyses showed significant reductions in higher branch orders (Strahler order 2) compared with control, with Strahler order 1 branches, representing terminals that account for the majority of control CIV branches, being undetectable in form3-IR neurons (Fig. 1G). To quantify effects on dendritic branch distribution, Sholl analyses were used to plot the density profiles of branches as a function of distance from the soma and to compare the peak of maximum branch density (critical value) and its corresponding radius. Both parameters showed significant reductions in CIV-specific form3-IR expression (Fig. 1I).

To independently validate phenotypic defects observed with form3-IR, we conducted mosaic analysis with a repressible cell marker (MARCM) clonal analyses using two previously published form3 point mutant alleles, namely form3^Em31 and form3^Em41 (Tanaka et al., 2004). Consistent with form3-IR analyses, MARCM mutant clones exhibited defects in CIV dendritic arborization characterized by short interstitial branches and reductions in distal terminal branching (Fig. 1C-D′). Morphometric analyses revealed significant reductions in terminal dendrites and concomitant reductions in total dendritic length (Fig. 1E,F). Moreover, Strahler analyses showed a similar result to that observed with form3-IR, where the reduction was observed in order 2 branches with a complete loss of the terminal order 1 branches for both form3 mutant alleles (Fig. 1H). Sholl analyses identified a significant reduction in the critical value for both form3^Em31 and form3^Em41 (Fig. 1J).

Collectively, our phenotypic analyses revealed relatively stronger phenotypic defects in form3-IR knockdowns compared with the form3^Em31 or form3^Em41 MARCM clones, which were comparatively similar in their phenotypic defects. This variability could be due to different effects of these genetic perturbations on Form3 protein levels. To address this question, we developed polyclonal antibodies against Form3 and investigated expression in control versus mutant backgrounds. Form3 is expressed in dendritic arborization (da) neuron subclasses, including CIV neurons, in control third instar larvae (Fig. S1I-I″). To assess antibody specificity and explore how form3-IR or form3 mutant alleles impact Form3 expression levels, we performed immunohistochemistry (IHC) analyses to compare Form3 protein levels in control, form3-IR and form3^Em41 MARCM clone CIV neurons. Form3 levels were reduced by 77% in form3-IR CIV neurons and by 44% in form3^Em41 mutant CIV neurons relative to control levels (Fig. 1K-O). These data indicate that form3 knockdown in CIV neurons results in a stronger reduction of Form3 levels relative to the form3^Em41 mutant neurons, which is consistent with the phenotypic variability observed in these genetic backgrounds.

To determine whether we could alleviate the loss-of-function dendritic defects observed with form3 knockdown, we generated a UAS-form3-IR resistant transgene (UAS-form3-IR^R), in which we introduced silent mutations in the form3-IR target sequence to render the transgene resistant to RNAi-mediated knockdown while limiting the use of rare codons that are recognized by low abundance tRNAs. We expressed this transgene in the form3-IR background of a CIV neuron and observed a partial, albeit significant, rescue with respect to number of dendritic branches and total dendritic length (Fig. S2). Although the RNAi-resistant transgene did not fully rescue the form3-IR-mediated dendritic defects, this could potentially be due to the relative strength of the form3-IR and/or insufficient levels of expression required to restore fully the dendritic abnormalities though the introduction of the silent mutations using less common codon usage. Alternatively, it is possible that the failure to rescue fully the form3-IR defects could be due to expression timing and levels that do not precisely mirror endogenous Form3 expression.

Collectively, these data indicate that Form3 promotes the appropriate number and position of branches along the proximal-distal axis of dendritic arbors, and is required to promote higher order branches. Given that form3-IR produced the strongest phenotypic defects in CIV dendritogenesis coupled to the strongest reduction in Form3 protein levels, and that the form3 homozygous mutant alleles are lethal prior to the third instar larval stage, we performed the remainder of our loss-of-function (LOF) studies using form3-IR.

Formins are well-known regulators of the cytoskeleton and thus we hypothesized that Form3 functions to modulate dendritic architecture. We utilized CIV-GAL4-driven expression of transgenic multi-fluorescent cytoskeletal reporters in combination with form3-IR in order to simultaneously visualize the F-actin (UAS-GMA) and MT (UAS-mCherry::Jupiter) cytoskeletons in vivo (Das et al., 2017). These studies revealed disruptions in F-actin cytoskeletal distribution and a severe reduction of dendritic stable MTs (Fig. 2A-B″). To confirm that the observed defects on dendritic stable MTs were not due to a non-specific effect of form3-IR on expression of the mCherry-tagged microtubule-associated protein (MAP) Jupiter, we performed independent validation experiments. IHC analyses of CIV neurons expressing form3-IR or form3^Em41 CIV MARCM clones were performed using antibodies against the Drosophila MAP1B molecule Futsch. Futsch is a known MAP that stabilizes MTs and has been shown to be a marker of the stable MT population (Halpain and Dehmelt, 2006; Hummel et al., 2000; Roos et al., 2000). Compared with genetic control CIV neurons, Futsch levels were reduced by 55% in form3-IR neurons at the cell body (Fig. 2C-D″,G). We also observed a 48% reduction in Futsch labeling at the cell body in form3^Em41 MARCM clones compared with control MARCM clones (Fig. 2E-F″,G). These analyses collectively suggest that Form3 has a major functional role in regulating a population of stable dendritic MTs.

To assess the effect of form3 disruption on the cytoskeleton across the dendritic arbor, we employed newly developed next-generation multichannel neuronal reconstructions (Das et al., 2017; Nanda et al., 2020). Multichannel reconstructions of cytoskeletal features revealed form3-mediated alterations in F-actin distribution and stable MT levels with form3-IR knockdown (Fig. 2H-I″). Alterations in F-actin organization upon form3 knockdown manifest as a reduction in the critical value (the peak of the curve) and a proximal shift of the critical radius (the radius at which the peak of the curve lies) of the relative F-actin quantity (Fig. 2J). In form3 knockdown neurons, we found a complete absence of Strahler order 1 branches, which make up the majority of the F-actin-rich terminal branches in control CIV neurons (Fig. 2K). As predicted, the effect on MTs was severe, as revealed by overall reductions in the relative MT quantity, as well as reductions at branch orders 2-5 and a complete absence of Strahler order 1 terminals (Fig. 2L,M).

To assess how form3 may contribute to dendritic arborization over developmental time, we conducted time-course studies examining CIV-specific form3 knockdown at first, second and third instar larval stages. These analyses revealed that form3-IR-mediated defects in dendritic hypotrophy increase as a function of developmental age (Fig. 3A-H). Phenotypic comparisons at the second instar stage revealed a modest qualitative increase in short interstitial branches emanating from the lower order dendrites in form3-IR (Fig. 3B,B′,F,F′), whereas third instar form3-IR neurons exhibited dramatic reductions in dendritic branches (Fig. 3C,C′,G,G′). However, quantitative analyses revealed no significant difference in the number of branches or total dendritic length at first or second instar stages compared with control, but a severe reduction in both length and branch number at the third instar stage (Fig. 3I,J). To address putative effects of maternal perdurance, we also investigated disruption of form3 function in the embryonic germline and CIV neurons using nanos-GAL4 to drive expression of form3-IR (Ye et al., 2004). As with CIV-specific form3 disruption (Fig. 3), morphometric analyses again revealed no significant difference in number of branches or total dendritic length at the first or second instar stages; however, third instar larvae displayed reductions in both morphometric parameters (Fig. S3). As control CIV dendrites continue to grow and ‘space-fill’ the receptive field over development (Fig. 3D), whereas form3-depleted CIV neurons exhibit a progressive loss of dendrites over development (Fig. 3H), these data suggest that the form3-IR phenotype exhibits a slowed growth, particularly at later stages of larval development.

A recent study demonstrated that dendritic arbor length of multiple da neuron subclasses at the third larval instar stage, including CIV neurons, is highly interrelated with local MT quantities (Nanda et al., 2020). Given the effect of form3-IR on dendritic growth and stable MTs at the third instar stage and the progressive loss of terminal branches observed in time-course analyses, we hypothesized that slowed dendritic growth and loss of dendritic branches correlates with progressive reduction in the population of stable MTs over development. Consistent with this hypothesis, we found that the stable MT signal is likewise progressively reduced over the developmental time in form3-depleted CIV neurons relative to controls (Fig. 3K-P). Quantitative analyses revealed significantly reduced relative MT quantities on the dendritic arbor over developmental time in form3-IR neurons (Fig. 3Q-S). Despite these reductions at first or second instar larval stages, we did not identify a significant change in either total dendritic length or terminal dendrites for form3-IR neurons, whereas third instar larval mutant neurons were reduced for both metrics. These findings may suggest that although stable MT levels are reduced at early larval stages, the remaining population of stable MTs is sufficient to support arborization, whereas at the third instar stage the level of stable MTs dips below a critical threshold required to sustain the growth of the complex, space-filling dendritic arborization profile of CIV neurons.

As form3 disruption produces severe defects in higher order dendritic branching, we sought to test a hypothesis that form3 overexpression may lead to excessive terminal branching. We overexpressed UAS-form3 in CIVs, which revealed a shift in branching distribution characterized by a reduction in branching immediately proximal to the cell body as well as exuberant terminal branching and elongated terminal dendrite extension (Fig. 4A-B′). Sholl analysis revealed a significant increase in the critical value (Fig. 4C), which occurs prominently around 105-200 μm from the soma (Fig. 4D). Moreover, Strahler analysis revealed a significant increase in the number of terminal branches, indicated by Strahler order 1 (Fig. 4E).

Another striking feature observed with form3 overexpression is a change in primary dendrite branch thickness. Relative to control, primary branch diameter was notably increased by form3 overexpression (Fig. 4A′,B′, arrowheads), suggesting that increased Form3 levels may alter underlying cytoskeletal organization. Therefore, we sought to assess the effect of form3 overexpression on the cytoskeleton. Multichannel cytoskeletal reconstructions (Fig. 4H-I″) revealed an altered F-actin signature in CIV neurons overexpressing Form3, which when quantified revealed a leftward shift of the critical value (peak of the curve) indicating an increase in relative F-actin quantity in the first few dendritic branch orders (Fig. 4H″,I″,J). Furthermore, cytoskeletal analyses revealed that Form3 overexpression leads to an increase in the MT intensity profile proximal to the cell body (Fig. 4H′,I′,K), suggestive of a role for Form3 in promoting MT density in primary dendritic branches. This raises the question of whether there may be changes in the cytoskeletal organization of dendritic terminals as well. Relative to control terminals (Fig. 4F,F′), Form3 overexpression qualitatively increased MT signal into the elongated terminal branches, potentially promoting their extension (Fig. 4G,G′).

In light of the partial rescue we observed with expression of the UAS-form3-IR^R transgene in the form3-IR mutant background, we examined the effects of overexpressing the RNAi-resistant transgene to determine whether it may phenocopy aspects of the form3 overexpression phenotype. These analyses revealed a qualitatively similar phenotype with thickened primary dendrites and clustered, elongated higher order branches (Fig. S4A-C). Sholl analyses revealed no statistical difference in the critical value or radius between UAS-form3 and UAS-form3-IR^R, suggesting that the RNAi-resistant version can recapitulate phenotypic features of the unmodified UAS-form3 transgene (Fig. S4D).

Based on the regulatory relationship between Form3 function and MT architecture, we sought to examine potential direct or indirect interactions between Form3 and MTs. To address this, we performed MT co-sedimentation (±MTs) and MT precipitation assays using GST-tagged Form3 constructs expressing either the FH1-FH2 domains or the FH2 domain alone compared with GST controls lacking Form3 sequences. Analyses of FH1-FH2 fusion proteins revealed specific co-sedimentation in pellet fractions only in the presence of MTs (Fig. 5A) indicative of a direct interaction. Analyses of the FH2 domain alone revealed low-level self-pelleting in the absence of MTs. Despite this, there was a notable increase in the amount of FH2 fusion protein that co-sedimented with MTs specifically, suggesting the FH2 domain is sufficient for Form3-MT direct interaction. To validate these results, we performed MT precipitation assays, revealing that tubulin precipitated only in the presence of the Form3 constructs, but not with GST alone (Fig. 5B).

Because the effects of form3 disruption were severe in MTs, we investigated the potential role of Form3 in regulating MT dynamics. To this end, we used a previously described approach, in which we expressed α-tubulin tagged with photoconvertible tdEOS in CIV neurons (Herzmann et al., 2017; Tao et al., 2016). When exposed to 405 nm light, tdEOS converts from green to red and the duration for which the red signal lasts can be used to assess MT turnover. We photoconverted a small segment of the dendrite close to the cell body in controls and form3-IR third instar larvae and imaged the animals at 30-min intervals for 1 h (Fig. 5C-D‴). Our results showed that there was no significant difference in the red fluorescent intensities (pseudo-colored magenta) between control and form3-IR animals over the course of our analyses, suggesting that the rate of MT turnover is similar in both these groups (Fig. 5E). However, it is important to note that form3-IR animals have significantly less MT signal to begin with compared with control (Fig. 5F). To assess whether a 1 h time interval may be too short to observe MT dynamics, we extended our time-lapse analyses to 2 h and our results remained the same, i.e. there was no statistical difference in the rate of turnover for the subpopulation of MTs that remain in form3-depleted neurons (Fig. S5). This raises the possibilities that Form3 may play a role in MT dynamics via de novo MT generation, or alternatively that Form3 may only function in stabilizing a subset of MTs that are normally bound by MAPs, including Futsch and Jupiter. We attempted to visualize de novo MT generation and dynamics using EB1::GFP +TIP labeling (Rolls et al., 2007; Ori-McKenney et al., 2012; Arthur et al., 2015) in form3-IR CIV neurons; however, the EB1::GFP signal in form3-depleted neurons was below the level of detection for reliable quantification (data not shown). Thus, we can neither rule in, nor rule out, the possibility of a role for Form3 in de novo MT generation/dynamics. The most parsimonious explanation of our data is that Form3 has a role in regulating at least a subpopulation of stable MTs and that a remaining MT subpopulation in form3-deficient neurons is not more unstable than controls.

Collective evidence implicates Form3 in MT stability, which, although clearly crucial for supporting dendritic complexity, may not fully explain why the arbor exhibits pronounced hypotrophy at the late larval stage. Therefore, we sought to examine the potential functional consequences of a destabilized MT cytoskeleton by examining the localization of organelles essential for supporting dendritic development. Studies have demonstrated that defects in mitochondrial localization can lead to dendritic degeneration/fragmentation in both invertebrates and vertebrates (Tsubouchi et al., 2009; López-Doménech et al., 2016). Moreover, INF2 (the human ortholog of Form3) has been shown to affect mitochondrial length and endoplasmic reticulum-mitochondrial interactions (Korobova et al., 2013). In vivo imaging revealed defects in dendritic mitochondria localization in form3-IR neurons, whereas proximal axonal localization appeared largely normal (Fig. 6A-B″). We also observed a significant decrease in dendritic mitochondria density (Fig. 6C). These data suggest that mitochondria localization is at least in part dependent on Form3, as well as a stable MT cytoskeleton.

Localization of satellite Golgi has been demonstrated to play key roles in regulating dendritic development and MT nucleation (Ye et al., 2007; Ori-McKenney et al., 2012). We hypothesized that form3 disruption impairs proper localization of satellite Golgi due to a disrupted MT cytoskeleton, which could, in part, contribute to the observed dendritic hypotrophy. We discovered aberrant localization in form3-IR neurons, with the majority of the satellite Golgi confined to the proximal branches, though we did observe some Golgi in the terminals (Fig. 6D-E′). The density of satellite Golgi was also significantly reduced (Fig. 6F). In controls, there was a distinct localization of satellite Golgi to dendritic branch points, which was often not the case in form3-IR. In addition, whereas controls display ‘islands’ of fused Golgi located along interstitial branches, such islands are not observed with form3-IR, but rather only small puncta (Fig. 6D,E). These data suggest that an intact MT cytoskeleton contributes to normal satellite Golgi translocation, and that defects in localization may contribute to the overall dendritic hypotrophy observed in form3-IR neurons.

To investigate the hypothesis that Drosophila Form3 and human INF2 share conserved functions, we generated transgenic fly strains to allow for inducible expression of either full-length INF2 or a truncated version in which autoinhibitory regulatory domains (DID/DAD) were deleted, leaving only the FH1 and FH2 domains (FH1-FH2) (Fig. 7A). Previous studies in Caenorhabditis elegans revealed that removal of the DID/DAD inhibitory domains was required for INF2 rescue of the worm ortholog exc-6 (Shaye and Greenwald, 2015), and Form3 protein does not contain the DID/DAD domains.

We initiated our INF2 analyses by first examining the functional consequences of overexpressing these transgenes in CIV neurons. Overexpression of the full-length INF2 transgene (Fig. S6A) modestly reduced the number of terminal branches, but had no significant impact on total dendritic length (Fig. S6B,C), whereas overexpression of the truncated INF2-FH1-FH2 transgene (Fig. 7E) did not significantly alter the number of terminal branches or total dendritic length relative to control CIV neurons (Fig. 7F,G). To determine whether INF2 expression can alleviate morphological defects observed in form3-depleted neurons, we introduced the INF2-FH1-FH2 variant into the form3-IR genetic background and out-crossed to a CIV-GAL4 driver. These analyses revealed a modest, but significant, recovery of CIV morphological defects (Fig. 7C,D) by mildly increasing the number of terminal dendrites and total dendritic length (Fig. 7F,G). Sholl analyses further revealed a significant increase in the critical radius with INF2-FH1-FH2 expression indicative of an increase in dendritic complexity (Fig. 7H).

As form3 disruption severely reduces the expression of markers for stable MTs (Fig. 2), we hypothesized that introduction of INF2-FH1-FH2 may be able to partially recover Futsch labeling on dendrites of form3-depleted neurons. Consistent with this hypothesis, analyses revealed an increase in Futsch labeling on the soma and dendrites of CIV neurons with INF2-FH1-FH2 expression (Fig. 7I-I″) relative to what we observed with form3-IR knockdowns (Fig. 2D-D″). These studies indicate that INF2-FH1-FH2 can partially alleviate form3-IR-mediated defects in dendritic complexity and stable MTs. Collectively, these analyses suggest a degree of functional conservation between Form3 and INF2, via FH1-FH2, in promoting dendritic complexity and MT stability.

CIV neurons function as polymodal nociceptors and are required to mediate a variety of nocifensive responses including an aversive body rolling (360°) in larvae upon exposure to potentially damaging thermal, chemical or mechanical stimuli (Tracey et al., 2003; Hwang et al., 2007; Lopez-Bellido et al., 2019; Himmel et al., 2019, 2017). Furthermore, recent evidence connects defects in dendritic architecture with noxious heat-evoked nocifensive behavioral sensitivity (Honjo et al., 2016). We hypothesized that form3 disruption in these neurons may lead to reduced peripheral sensitivity to noxious heat and thereby impair nocifensive behavior. Thus, we expressed form3-IR in CIV neurons and observed nocifensive behavioral responses when challenged with a noxious heat (45°C) stimulus. These analyses revealed a severe impairment in noxious heat-evoked behavioral responses (Fig. 8B,E,F; Movie 1), indicating a dramatic loss of peripheral sensitivity. This behavioral defect was not due to any general defect in locomotion, as both control and form3-IR larvae exhibit normal locomotor behavior. CIV-specific inhibition of form3 led to a dramatic increase in the latency to respond (among those few larvae which responded) (Fig. 8A,B,E). The majority of form3-IR larvae were classified as non-responders, as they failed to exhibit rolling or other documented nocifensive behaviors within the 20 s assay period (Fig. 8B,F).

We further sought to determine whether form3 or INF2-FH1-FH2 overexpression leads to changes in heat-induced nocifensive behavior. We found that neither form3 nor INF2-FH1-FH2 overexpression resulted in a significant change from controls with respect to latency (Fig. 8C-E); however, we did observe a reduction in the percentage of responders in both these conditions (Fig. 8F). Apart from nocifensive rolling, several other noxious heat-evoked behaviotypes have been described, including larval whipping, head thrashing, and seizures (Chattopadhyay et al., 2012). Relative to controls, and among those larvae that failed to execute rolling behavior, these behaviors were variably observed with greater incidence in form3 and INF2-FH1-FH2 overexpression, as well as to some modest extent in form-IR (Fig. 8F). This may suggest that the alterations in dendritic morphology observed with form3 or INF2-FH1-FH2 manipulations impair normal processing of thermal stimuli, resulting in altered behavioral repertoires.

Intriguingly, mutations in INF2 have been causally linked to CMT disease, yet the mechanisms of action in CMT pathology are incompletely understood (Boyer et al., 2011). Neurological features of CMT include peripheral motor and sensory neuropathies, and the primary phenotypes consist of progressive distal muscle weakness and atrophy, reduced tendon reflexes, foot and hand deformities, and peripheral insensitivity (Ekins et al., 2015). CMT sensory neuropathies lead to distal sensory loss resulting in a reduced ability to sense heat, cold and pain, yet the neural bases of these sensory defects have not yet been fully elucidated. Given the causative role of INF2 mutations in CMT disease and the impaired distal sensitivity to thermal stimuli in CMT patients, we next tested the hypothesis that introduction of the INF2-FH1-FH2 transgene into the form3-IR background would rescue the impaired behavioral responses to noxious thermal stimuli. We found that CIV expression of INF2-FH1-FH2 significantly rescued the behavioral latency defects observed in form3-IR larvae, leading to an increase in the percentage of behavioral responders (Fig. 8D-F; Movie 2). Although the introduction of INF2-FH1-FH2 only partially rescues the behavioral defects, it nonetheless suggests a conserved role for Form3 and INF2 in regulating peripheral sensitivity to noxious stimuli that has potential implications for the etiological bases of CMT peripheral sensory neuropathy.

Previous studies have linked CMT2C to aberrant ankyrin repeat (ARs) in the human TRPV4 channel (Landouré et al., 2010). Noxious heat detection in Drosophila larvae requires the function of two TRPA channels, Painless and TRPA1, both of which contain ARs (Tracey et al., 2003; Zhong et al., 2012). In the mechanosensing TRP channel NompC, ARs provide direct linkage of the channel to the MT cytoskeleton and are required for mechanical gating of the channel (Zhang et al., 2015). Moreover, studies of Painless have revealed that its AR domains are necessary for thermal, but not mechanical, nociception (Hwang et al., 2012). These findings led us to hypothesize that form3-mediated defects in the MT cytoskeleton may affect localization of AR-containing TRP channels such as Painless. To investigate this hypothesis, we conducted in vivo imaging of CIV neurons expressing UAS-Painless^p103::VFP while simultaneously expressing form3-IR. In contrast to controls, in which Painless^p103::VFP was strongly expressed on CIV cell bodies, as well as throughout dendrites and proximal axons (Fig. 8G-G″), in form3-IR neurons there was a notable reduction of Painless signal on proximal and distal CIV dendrites, despite comparable levels of cell body expression and, to a lesser extent, axonal expression (Fig. 8H-H″). These data suggest that form3-IR-induced defects in nocifensive behavior may result from a combination of factors, including severely reduced arbor complexity and aberrant localization of channels required for thermosensory nociceptive behavior.

Dendritic development is a complex phenomenon, which requires spatiotemporal regulation of local cytoskeletal interactors to direct specific morphological features of the neuron. Molecules involved in this process can have one of many roles, such as arbor specification, growth by enhancement, suppression by reduction, or simply maintenance of the dendritic arbor. Our results suggest Form3 is required in maintaining dendritic arbor growth, as form3 mutants exhibit progressive dendritic hypotrophy, ultimately leading to a highly rudimentary arbor.

The hypothesis that formins may regulate MTs has been proposed for some time, but only recently have biochemical studies begun to explore how formins interact with MTs and affect their dynamic properties. Studies have shown that multiple mammalian formins, as well as the fly formin Cappuccino, can interact directly with MTs (Breitsprecher and Goode, 2013; Roth-Johnson et al., 2014); however, whether this is a specific or more general property of formin molecules remains unclear. In cultured hippocampal neurons, the formin mDia1 (Diaph1) has been implicated in MT stabilization with knockdowns leading to increased MT catastrophe and reductions in tyrosinated MTs as well as increased spine density and dendritic complexity (Qu et al., 2017). Conversely, Hong et al. (2018), demonstrated that knockdown of mDia1 resulted in reductions in total dendritic length and number of branch points in cultured hippocampal neurons; thus, additional studies are needed to clarify the roles of mDia1 in dendritic development. At a mechanistic level, we present multiple converging lines of evidence implicating Form3 in positive regulation of a population of stable MTs. Consistent with this, previous studies demonstrate that formation of stabilized MTs requires the Form3 ortholog INF2 (Andrés-Delgado et al., 2012).

Proper mitochondrial function is required to support neuronal development and function, as mitochondria are fundamentally important for several cellular events, such as ATP production and Ca²⁺ regulation, as well as release and reuptake of neurotransmitters at synapses (Detmer and Chan, 2007). Several consequences of impaired mitochondrial dynamics have been studied, and it has been shown that mitochondria dysfunction is highly correlated with neurodegenerative diseases (Chan, 2006). Mutations in the mitochondrial GTPase mitofusin 2 cause autosomal dominant CMT type 2A disease (Züchner et al., 2004), and disruptions in OPA1, a protein that mediates mitochondrial fusion, lead to autosomal dominant optic atrophy, an inherited form of optic nerve degeneration (Alexander et al., 2000). Moreover, studies have demonstrated that defects in mitochondrial function, morphology or trafficking contribute to dendritic degeneration and loss of complexity in both invertebrates and vertebrates. Mutations in the Drosophila mitochondrial protein Preli-like (Prel), as well as its overexpression, cause mislocalization and fragmentation of mitochondria, leading to dendritic loss via a mechanism involving increased eIF2α phosphorylation and translational repression (Tsubouchi et al., 2009; Tsuyama et al., 2017). Disrupted mitochondrial distribution leads to a loss of dendritic complexity in mouse hippocampal neurons, which precedes neurodegeneration, supporting a crucial role of mitochondria in stabilizing complex dendritic architectures and maintaining neuronal viability (López-Doménech et al., 2016). These findings are also intriguing in terms of conserved functions between Form3 and INF2, the latter of which has been demonstrated to affect mitochondrial length and endoplasmic reticulum-mitochondrial interactions (Korobova et al., 2013).

In addition to mitochondrial localization defects, we also postulated that defects in MT cytoarchitecture may impair localization of satellite Golgi on the dendritic arbor. Satellite Golgi have been shown to play important functional roles in regulating dendritic growth and branching by serving as local sites for MT nucleation to support branch extension (Ye et al., 2007; Ori-McKenney et al., 2012). Furthermore, a recent study discovered a role for the formin mDia1 in proper positioning of Golgi to the base of major dendrites in cultured mouse hippocampal neurons (Hong et al., 2018). We discovered that form3-depleted neurons also exhibited defects in satellite Golgi localization and gross Golgi architecture. Interestingly, previous studies have implicated INF2 in the maintenance of Golgi architecture in cultured cells in which disruptions in INF2 function cause Golgi fragmentation (Ramabhadran et al., 2011). Combined, these analyses provide important insights into the putative mechanistic bases of form3-mediated defects in dendritogenesis, and identify an important role of a stable MT cytoskeleton in supporting the proper localization of dendritic organelles.

Numerous neurological disorders including lissencephaly, amyotrophic lateral sclerosis, spastic paraplegia, tauopathies, Alzheimer disease and CMT are linked to defects in the MT cytoskeleton and/or MT motor-based transport (Franker and Hoogenraad, 2013). We show here that form3 is required for dendritic development, but the potential role of INF2 is unknown with respect to neural development. Morphological defects caused by form3 disruption can be partially recovered by expression of the INF2 FH1-FH2 domains, with respect to dendritic growth and Futsch labeling of MTs. These analyses suggest potential evolutionarily conserved functions between form3 and INF2 in regulation of dendritic architecture and MTs.

Mutations in INF2 are known to be causative for CMT disease (Boyer et al., 2011), although the mechanistic functions of INF2 in disease pathogenesis are unclear. We observed that form3 disruption in CIV nociceptive neurons severely impairs noxious heat-evoked behavioral responses resulting in peripheral insensitivity, which is consistent with the sensory neuropathies observed in CMT patients. Intriguingly, we can partially revert this heat insensitivity by the introduction of INF2 FH1-FH2 in the form3 mutant background, revealing not only partial morphological recovery, but also behavioral rescue.

Previously, CMT diseases have been characterized by progressive defects in axonal development, myelination, protein translation, and intracellular traffic of vesicles and organelles (Bucci et al., 2012; Niehues et al., 2014). CMT disease has also been linked to various defects in mitochondrial dynamics with CMT causing mutations that alter energy production via a mitochondrial complex I deficiency (Cassereau et al., 2011). Our work, in addition to these mechanisms, suggests a possible additional mechanism whereby aberrant INF2 activity may lead to progressive sensory neuron dendritic hypotrophy that could dramatically reduce dendritic field coverage and contribute to peripheral insensitivity. CMT sensory neuropathies manifest as impaired peripheral sensitivity to heat, cold and pain, and mutations in thermosensitive TRP channels such as TRPV4 have been linked to CMT2C (Landouré et al., 2010). We discovered that form3 mutant CIV nociceptive sensory neurons exhibit defects in dendritic localization of the thermosensitive TRPA channel Painless. Interestingly, previous work demonstrates that the heat-sensitive TRPV1 channel is known to bind to MTs and disruption of the MT cytoskeleton leads to attenuation of TRPV1 currents impacting desensitization (Goswami et al., 2011; Laínez et al., 2010; Ferrandiz-Huertas et al., 2014). TRPV1 physically interacts with MTs and an intact MT cytoskeleton is required to preserve TRPV1 function as MT disassembly by treatment with colchicine impairs TRPV1 membrane insertion, revealing a role in ion channel trafficking to the membrane (Goswami et al., 2007; Goswami, 2012; Storti et al., 2012). Given that form3 mutant neurons exhibit dendritic MT defects and aberrant TRP channel localization, disruptions in these biological processes may contribute to the behavioral insensitivity to noxious heat observed in form3 mutants. Combined, our findings provide new insights into the roles of formins in dendritic development and may provide mechanistic insights into the potential etiological bases of INF2-mediated CMT sensory neuropathy.

Drosophila stocks were maintained at 25°C, and genetic crosses were performed at 29°C. Age-matched wandering third instar larvae of both sexes were used for all experiments. The following stocks were used in this study, where ‘B’ represents stocks from the Bloomington Drosophila Stock Center, ‘v’ represents those from the Vienna Drosophila Resource Center, and additional sources are provided parenthetically: GAL4⁴⁷⁷,UAS-mCD8::GFP/CyO,tubP-GAL80;GAL4^ppk.1.9,UAS-mCD8::GFP (CIV-GAL4); GAL4⁴⁷⁷, UASmCD8::GFP (CIV-GAL4); GAL4⁴⁷⁷;ppk::tdTomato; UAS-GMA::GFP;GAL4⁴⁷⁷,UAS-Jupiter::mCherry (Das et al., 2017); UAS-form3-B1 (Tanaka et al., 2004); form3^Em41,FRT^2A (missense: A376>V) (Tanaka et al., 2004; this study); form3^Em31,FRT^2A (nonsense: R428>Stop) (Tanaka et al., 2004; this study); GAL4^5-40,UAS-Venus,SOP-FLP⁴²; +; tubP-GAL80,FRT^2A (Drosophila Genomics Resource Center, stock 109-950); GAL4^Nanos;+;form3-IR/ppk::tdTomato (B7303; this study); UAS-mito-HA-GFP.AP (B8442); UAS-manII::EGFP; GAL4^ppk,UAS-myr::GFP; UAS-INF2-FH1-FH2; UAS-INF2-Full length; UAS-Painless^p103::VFP (Hwang et al., 2012); UAS-capu-IR (B2922); UAS-dia-IR (B33424, B35479); UAS-Fhos-IR (B31400, B51391); UAS-Frl-IR (B32447, v110438); UAS-DAAM-IR (B39058); UAS-form3-IR (B32398; focus of analyses in this study; v45594; v42302; v28437; v107473); and UASp-alphaTub84B.tdEOS (B51314). OregonR (ORR) or w¹¹¹⁸ were used as a control strain for genetic outcrosses. The UP-TORR Fly website (https://www.flyrnai.org/up-torr/) was used to computationally evaluate putative off-target effects for UAS-form3-IR transgenes used in this study and stock B32398 targets both form3 isoforms with zero predicted off-target effects in the fly genome.

Larval filets were processed and labeled as previously described (Sulkowski et al., 2011). Primary antibodies used in this study were: rabbit anti-Form3 (1:100; see below), mouse anti-Futsch (1:200; Developmental Studies Hybridoma Bank, 22C10), anti-HRP (1:200; Jackson ImmunoResearch, 123-475-021) and chicken anti-GFP (1:1000; Abcam, ab13970). Donkey anti-rabbit, anti-mouse and anti-chicken secondary antibodies (Jackson ImmunoResearch, 711-585-152, 703-545-155, 715-585-150, 123-475-021) were used at 1:200. IHC slides were then mounted in Fluoromount Aqueous Mounting Medium (Sigma-Aldrich), and imaged at room temperature on a Zeiss LSM780 confocal system with either a 20× Plan Apo, 0.8 NA (dry), 40× Plan Apo, 1.4 NA (oil) or 60× Plan Apo, 1.4 NA (oil) objective. Zen Blue software was used to quantify the mean intensity of the fluorescence. Live confocal imaging was performed as described (Iyer et al., 2013). Briefly, six to ten third instar larvae were analyzed for neurometric quantitation from segments A3-A5. For time-course analyses, larvae were live imaged at the indicated time point, returned to agar plates, aged at 25°C, and then re-imaged. Images were collected as z-stacks with a step size of 1.0-2.0 µm and 1024×1024 resolution.

Anti-Form3 polyclonal antibody was generated in rabbits using KLH-conjugated peptides (GenScript). Form3 epitope [RGSDASSPTRKPSQ (amino acids 321-334)] was predicted by the GenScript Optimum Antigen design tool. Anti-Form3 antibody was affinity purified against antigenic peptide and specificity confirmed by IHC analyses.

MARCM analyses were performed as described (Sulkowski et al., 2011). Briefly, form3^Em31, FRT^2A or form3^Em41, FRT^2A flies were crossed to GAL4^5-40,UAS-Venus,SOP-FLP⁴²; +; tubP-GAL80,FRT^2A flies (Drosophila Genomics Resource Center, stock 109-950). Third instar larvae with GFP-labeled form3 mutant neurons were subjected to live confocal microscopy.

Quantitative neurometric analyses were performed as previously described (Das et al., 2017). Maximum intensity projections of z-stacks were exported using Zen Blue software. Exported images were manually curated to eliminate non-specific auto-fluorescent spots, such as larval denticle belts, using a custom-designed program, Flyboys (available at https://github.com/dcox18/Morphometric-analysis). Images were processed, skeletonized, and neurometric data compiled as described (Iyer et al., 2013). Quantitative neurometric information was extracted and compiled using custom Python algorithms (available at https://github.com/dcox18/Morphometric-analysis). The custom Python scripts were used to compile the output data from the Analyze Skeleton ImageJ plugin, and the compiled output data were imported into Excel (Microsoft). Neurometric data were analyzed in Microsoft Excel and statistical tests were performed and plotted in GraphPad Prism 9. For Sholl analysis, we used ImageJ (Ferreira et al., 2014) to plot the number of intersections as a function of distance from the cell soma and to determine the peak of maximum branch density (critical value/number of intersections) and corresponding radius. For Strahler analysis, we used the centripetal labeling function of NeuronStudio (Wearne et al., 2005). Dendritic terminals are defined as Strahler order 1. Order increases when two or more branches of the same order intersect at a branching point.

Multi-signal cytoskeletal reconstructions and quantitative analyses were performed using a previously described protocol (Nanda et al., 2020), slightly modified from the protocol used by Das et al. (2017). Briefly, two-channel (GFP for F-actin and RFP for MT) image stacks (.czi file format) were first processed in Fiji (Schindelin et al., 2012) in which a third pseudo-channel was created by adding the signals from the two original channels. Multiple image tiles of a single neuron, if present, were stitched together in Vaa3D (Peng et al., 2014). These finalized image files were manually reconstructed using the reconstruction tool neuTube (Feng et al., 2015). The newly reconstructed SWC files (Cannon et al., 1998; Nanda et al., 2018) were topologically repaired in batch, using custom scripts built within the TREES toolbox (Cuntz et al., 2010) environment in MATLAB (MathWorks). These repaired SWC files are once again opened in neuTube, where additional editing and quality checks were conducted. The corrected reconstruction files and the image stacks were used as input in the Vaa3D multi-channel plugin (Nanda et al., 2018) to create enhanced reconstructions that represent the morphology along with the intensity and volume occupied by each channel. The neurons were then scaled (from voxel to physical dimension) and their cytoskeletal morphometrics (subcellular and overall structural features) were quantified using custom quantification scripts (quantification scripts can be downloaded from dx.doi.org/10.17632/wpzd2wxtgn.1).

Cytoskeletal quantity (MT or F-actin quantity) of a compartment is defined as the product of three parameters: (1) the portion of the volume inhabited by the cytoskeletal signal, (2) the mean signal intensity and (3) the compartment volume. All cytoskeletal quantities are represented relative to Class IV WT. MT and F-act quantities are normalized by an M-factor and an F-factor, where the two factors are the inverse of average total MT and average total F-actin of Class IV WT neurons, respectively. Dendrite length, microtubule and F-actin quantities (in the y-axis of Fig. 2J-M, Fig. 3Q-S, Fig. 5J,K) represent the total sum of length, microtubule and F-actin, respectively, at each Strahler order and within every 40-μm path distance bin. These values are computed for each neuron first, and then averaged across all neurons of a given neuron type.

For analyses of MT stability/turnover, α-tubulin 84B tagged with photoconvertible tdEOS (UAS-alphatub84B.tdEOS), and UAS-form3-IR were expressed under the control of GAL4⁴⁷⁷. For controls, GAL4⁴⁷⁷;UAS-alphatub84B.tdEOS flies were outcrossed to wild-type flies. Photoconversion experiments were carried out following previously described procedures (Tao et al., 2016). Briefly, an ∼30 µm² region of the dendrite near the cell body was photoconverted from green to red by exposing to 405 nm laser. The photoconverted neurons were live-imaged immediately after photoconversion (0 h) and at regular intervals of 30 min up to 2 h. Red fluorescent intensities were measured in this photoconverted region and a neighboring non-converted (green fluorescent) region in Zen Blue lite (Zeiss). The remaining fluorescence intensity (FI) was measured according to the following formula: (FI_converted−FI_neighboring)_timecourse/(FI_converted−FI_neighboring)_0h as previously described (Tao et al., 2016).

Coding sequences of the Form3 FH1-FH2 and FH2 domains were cloned in EcoRI-cut pGEX2T vector. The resulting plasmids were used to express FH1-FH2 and FH2 only fragments of Form3 as GST fusion proteins in BL21 Escherichia coli. Protein purification was carried out as previously described (Barkó et al., 2010) with minor modifications. For MT binding assays, MTs were assembled from tubulin protein (Cytoskeleton, Inc.) according to the manufacturer’s instructions. In MT co-sedimentation assays, GST::Form3-FH1-FH2 and GST::Form3-FH2 proteins were pre-cleared by ultracentrifugation (400,000 g for 30 min), then diluted in MT binding buffer (MBB; 10 mM Na-HEPES pH 7.0, 1 mM MgCl₂, 1 mM EGTA, 1 mM DTT, 20 µM taxol, 0.5 mM thesit, 10% glycerol) and mixed with pre-assembled MTs (0.5 µM). Control samples did not contain MTs. Protein mixtures were incubated for 30 min at room temperature then centrifuged at 100,000 g for 1 h at 25°C. Proteins in the supernatants and pellets were resolved by SDS-PAGE then stained with colloidal Coomassie Blue. In GST pull-down assays, purified Form3 proteins were immobilized on glutathione-S-sepharose beads, followed by incubation with taxol-stabilized MTs (0.5 µM) for 30 min at room temperature in MBB. The beads were washed thoroughly in MBB followed by protein elution in SDS-PAGE sample buffer and detection by anti-GST (Sigma-Aldrich) and anti-α-tubulin (Sigma-Aldrich) by western blotting.

A synthetic RNAi-resistant UAS-form3 transgene (UAS-form3-IR^R) was generated as previously described (Schulz et al., 2009; Jonchere and Bennett, 2013). Briefly, codon usage of the form3-RA transcript was modified by the introduction of silent mutations to the RNAi target sequence for the form3-IR transgene predominantly used in this study (B32398; TRiP reagent ID: HMS00393) in order to render the transgene resistant to RNAi-mediated knockdown. The RNAi target sequence is located in the N-terminal Form3 FH3 domain and the synthetic RNAi-resistant sequence covered amino acids 63-80 of the Form3-PA protein isoform, which is identical to the sequence in the other Form3 protein isoform, Form3-PB. The silent mutations were selected to minimize the use of rare codons recognized by low abundance tRNA species over a 54 bp segment of the coding sequence corresponding to amino acids 63-80, whereas the remainder of the form3 coding sequence was unaltered (see Supplementary Materials and Methods). Gene synthesis was performed by GenScript and synthetic DNA was subcloned in using EcoRI (5′) and KpnI (3′) into pUAST-attB. Transgenic UAS-form3-IR-R strains were generated by ΦC31-mediated integration with targeting to 2R (VK1) (GenetiVision).

For optimal expression, we synthesized D. melanogaster-codon-optimized INF2 cDNAs (GenScript). Two custom gene syntheses were performed to generate a full-length cDNA and a cDNA in which the DID and DAD autoinhibitory regulatory domains had been deleted, leaving only the FH1 and FH2 domains (FH1-FH2). Each synthesized gene was C-terminally FLAG-tagged (DYKDDDDK) and subcloned into pUAST-attB. Transgenic INF2 strains were generated by ΦC31-mediated integration with targeting to 2 L (attP40) (GenetiVision).

For noxious heat nociceptive assays, age-matched third instar larvae were recovered and briefly rinsed with water to remove any residual media. Larvae were then transferred to a black aluminum metal plate, which was pre-sprayed with water to generate a thin film facilitating larval movement. Larvae were allowed to acclimate to the plate and resume normal peristaltic locomotion before the plate was transferred to a temperature-controlled Peltier plate (TE Technology). The temperature was preset to 45°C to evoke nocifensive behaviors. Heat-evoked behaviors were recorded using a Nikon D5300 DSLR camera. Video files were processed using ImageJ and manually curated to evaluate response latency and characterize nocifensive behaviotypes (Chattopadhyay et al., 2012). The maximal latency period for behavioral response was set at 20 s after stimulus, and larvae that failed to exhibit a behavioral response in this interval were classified as non-responders.

Data are reported as mean, and error bars represent standard error of the mean (s.e.m.) or standard error of proportion (SEP) as indicated in figure legends. Data represent biological replicates, with the exception of the western blots for the GST pull-down assays, which represent technical replicates. Sample sizes were selected based upon accepted norms in the published literature and included both sexes in all analyses. Statistical analyses (reported in the figure legends) were performed using either one-way ANOVA with Bonferroni correction for multiple comparisons; Kruskal–Wallace test with Dunn correction for multiple comparisons; two-way ANOVA with Bonferroni correction; Mann–Whitney U-test; or unpaired t-test. All datasets were tested for normality (Shapiro–Wilk normality test) and homogeneity of variance (Bartlett's test or F test) before statistical analysis. All new genotypes presented here are available upon request.

We gratefully acknowledge Drs Akinao Nose, Yuh-Nung Jan, Daniel P. Kiehart, Chris Q. Doe and W. Daniel Tracey for reagents and fly strains. We thank the Bloomington Drosophila Stock Center (NIH P40ODO18537) and Vienna Drosophila Resource Center (VDRC) for fly strains used in this study.