Abl signaling shapes the intrinsic fluctuations of actin to direct growth of a pioneer axon in Drosophila

The fundamental problem in axon growth and guidance is to understand how cytoplasmic signaling modulates the cytoskeleton to produce directed growth cone motility. We show here that the TSM1 pioneer axon of Drosophila extends by using Abl tyrosine kinase to shape the intrinsic fluctuations of a mass of accumulated actin in the distal axon. The actin mass fluctuates stochastically in length, but with a small, forward bias that drives the axon along its trajectory by promoting emergence of protrusions in leading intervals where actin accumulates, and collapse of protrusions in lagging intervals that actin has vacated. The actin mass is sculpted by Abl signaling, which probabilistically modulates its key parameters - its width and internal disorder - to drive its advance, while maintaining internal coherence. Comparison of TSM1 to other systems suggests that the mechanism we demonstrate here is apt to be common among pioneer axons in many organisms.

The process of axon guidance is central to patterning the nervous system during 49 development. Understanding how axons pathfind to their correct targets requires that we know the 50 mechanism by which guidance information in the environment controls the spatial organization 51 and dynamics of the growth cone cytoskeleton to produce directed extension of the axon. It is 52 widely accepted that in some cases broad, flat, growth cones extend the axon by harnessing the 53 mechanochemical properties of large adhesive lamellipodia 1,2 . There is reason to suspect, 54 however, that other axons pathfind and extend in an entirely different way. 55 Cells in culture can switch mechanisms of cell motility, depending on the developmental 56 context. For example, fibroblasts make large lamellipodia in 2-dimensional culture, yet assume 57 spindle shaped morphologies and move faster in 3D environments 3 . Mesenchymal cells use a slow 58 moving adhesive style of growth in 2D culture, but switch to a fast ameboid style of growth in 3D 59 conditions of low adhesion and strong confinement 4 . Dendritic cells rely on integrins for 2D cell 60 migration but 3D migration in vivo is protrusive and independent of integrin function 5 . These 61 examples demonstrate that the molecular components and mechanisms used by diverse cell types 62 to move in 2D versus 3D environments are consistently and fundamentally different. 63 Much as cell motility can be accomplished in multiple ways, there is also evidence for a 64 second mode of axon growth. It has been reported that filopodial protrusions alone can extend 65 some axons. In these cases, it seems that rather than applying adhesive traction to lamellipodia to 66 pull the growth cone forward, axon growth is instead accomplished by selective stabilization and 67 dilation of individual filopodia 6,7 . This filopodial style of growth has been proposed as a common 68 mechanism for axon growth 6,8 , but it is not clear how such a mechanism would work, or how axon 69 guidance signaling molecules would regulate it.
To understand how signaling controls axon growth and guidance, there has been a 71 sustained effort to find and characterize individual molecules with the goal of linking them to 72 fundamental steps in the mechanism of axon extension and guidance. One such protein, the non-73 We therefore set out to use live imaging of a single pioneer neuron in its native environment 88 to understand how Abl controls a pathfinding axon. We developed a method to image and 89 computationally quantify both the neuronal morphology and the cytoskeletal dynamics of the 90 TSM1 axon as it extends along its native trajectory in the intact Drosophila wing. We show here 91 that Abl regulates the organization and redistribution of an accumulated actin bolus that locally 92 controls filopodial morphogenesis to direct the extension and guidance of the TSM1 axon by We first verified that the Abl tyrosine kinase is required for growth and guidance of the 118 TSM1 axon, just as it is required for many other axons in the CNS and PNS of Drosophila, and of 119 vertebrates 9-12 . We found that perturbing Abl activity or expression caused aberrant TSM1 axon 120 growth and guidance phenotypes including instances where axonal growth stalled ( Figure 1C), no 121 axon extended from the cell body ( Figure 1D), the axon misrouted along its trajectory ( Figure 1E), 122 and/or the axon maintained aberrant collateral branches ( Figure 1C). Specifically, reducing Abl 123 expression by RNAi in the background of Dicer 2 overexpression (Abl KD) caused an increase in 124 aberrant TSM1 axon growth and guidance phenotypes (UAS-Abl RNAi, 51% vs. control 11%; p 125 < 0.0001, Figure 1F). The expressivity of TSM1 axon defects observed in Abl KD was enhanced 126 by heterozygosity for an Abl mutant allele (83% with UAS-Abl RNAi/Abl4 vs. UAS-Abl RNAi, 127 50%; p = 0.002, Figure 1F), confirming the fidelity of the RNAi phenotype. Expression of Dicer2 128 alone also increased the frequency of subtle defects in TSM1 morphology (32% vs 11% in control, 129 Figure 1F) but the spectrum of defects is significantly different from that in the presence of Abl 130 RNAi (p = 0.0035, chi-square, Supplemental Figure 2) and overexpression of Dicer 2 alone did 131 not affect cytoskeletal parameters measured in growing axons (see below). To discriminate the 132 catalytic function of Abl kinase from its scaffolding role, we treated wings with Imatinib, a specific 133 inhibitor of Abl Kinase activity, and found significant TSM1 axonal defects when compared to 134 control (61% in 10µM Imatinib vs. control 11%; p < 0.0001, Figure 1F). This result strongly 135 suggests that kinase activity is critical, though it does not exclude an important function for 136 scaffolding. We also observed axonal defects when Abl was overexpressed (Abl OE 35% vs. 137 control 11%; p = 0.0123, Figure 1F). Taken together, these results demonstrate that precise tuning 138 of Abl function is required for TSM1 axon extension and guidance. 139

Actin accumulates in the filopodial TSM1 growth cone 141
We next live -imaged TSM1 axon extension in explanted early-pupal wing imaginal discs 142 by spinning disc confocal microscopy, with simultaneous visualization of the axonal membrane, 143 using neuron-specific expression of CD4 Tandem-Tomato, and intracellular actin, using 144 LifeactGFP (which labels both F-and G-actin 22 ) in wild type, Abl KD and Abl OE backgrounds, 145 respectively. Multiplexed Z-stacks were collected every 3 minutes for 1.5 hours from fourteen 146 trajectories of each genotype. Axon morphology was traced stereoscopically in 3-dimensions, and 147 the intensity of the LifeactGFP signal was quantified along the axon shaft, from the base of the 148 axon to its distal tip. We focus on actin organization both because of its critical role in essentially 149 all forms of motility, and more specifically because the most extensively characterized output of 150 Abl signaling is its modulation of actin regulators. Control experiments showed that imaging of 151 developing transgenic wings in culture does not disturb the trajectory of the axon, or the ability of 152 these axons to reach the L1-L3 junction and fasciculate with the L3 nerve (data not shown). 153 Live-imaging revealed consistent, concerted advance of a zone of enhanced filopodial 154 density in the distal axon, together with an intra-axonal mass of accumulated actin, as the axon 155 extended. In all imaged timepoints, filopodia and transient axonal branches (referred to 156 collectively as protrusions) were the dominant morphological features of the axon. Large 157 lamellipodia were seen only very rarely (<1% of time points) (Figure 2A -C). While the length 158 and density of protrusions were highly variable along the axon, a region with enhanced density of 159 protrusions was almost always observed in the distal portion of the axon ( Figure 2B and E).  Figure 2G and 3E and Supplemental movie 1). We refer to the curve that reports the relative actin intensity as a function of its position along the axon as the "actin distribution" 164 throughout this manuscript (see, for example Fig 2D). Note that standard confocal imaging 165 methods can only record the dynamics of this bulk actin distribution, which incorporates actin 166 transport, diffusion, polymerization/de-polymerization, branching/de-branching, etc., it does not 167 resolve the motions of individual actin molecules. Actin accumulation in the distal axon was also 168 observed using other actin markers, including F-tractin Tandem-tomato, which selectively labels 169 Furthermore, the offset distance between the actin maximum and that of the protrusion density 207 appears to be maintained actively. We found a strong negative correlation between the rate of 208 change of the offset distance between these two positions in pairs of successive time-points (r=-0.50, p<0.0001) ( Figure 3C), which suggests that when the spacing between the actin and 210 protrusion maxima increases or decreases in any given time step, the gap size tends to be restored 211 during the next time point. Taken together, these data provide evidence that the actin accumulation 212 peak leads the advance of the axonal zone bearing the highest density of protrusions, with the 213 offset distance between these two axonal features being a regulated aspect of axon growth. 214 Based on these observations, we hypothesized that advance of the leading actin mass

TSM1 axons maintain parameter relationships in Abl KD and OE genetic backgrounds 228
The observations above suggest that the molecular mechanism that controls the structure 229 and dynamics of the actin distribution is the key to regulating the morphological development and 230 extension of the TSM1 axon. We therefore imaged TSM1 axon extension in Abl KD and Abl OE 231 genetic backgrounds, computationally measured a broad set of parameters that capture axonal morphology and intra-axonal actin organization and compared these data sets to those from 233 wildtype TSM1 trajectories. Among the parameters we investigated were morphological features, 234 including the number, length, and branch order of protrusions, and the 3-dimensional volume of 235 substratum explored by protrusions from the growth cone (here called growth cone volume), as 236 well as measures of the actin distribution, including the length, organization, and speed of advance 237 of the peak of actin accumulation (see Table 1 and Materials and Methods for a complete listing 238 of imaged features and their definitions). 239 To verify that the mechanism underlying axon extension in the Abl KD and Abl OE 240 conditions is fundamentally the same as that in wild type, we first queried the parameter set by 241 performing a Principal Components Analysis (PCA) using the wild type dataset of growth cone 242 measurements (Supplemental Figure 7). The first two principal components account for 47% of 243 the total variance in the data set. PC1, which accounts for 29% of the variance, is dominated by 244 contributions from morphological parameters including the number and length of protrusions 245 along the axon, while PC2 (18% of the variance) is dominated by parameters that measure how 246 the actin distribution is organized. We applied the two eigenvectors to the Abl KD and Abl OE 247 data sets to compare how similar the global relationships among parameters in the altered-Abl 248 conditions are to those in wild type ( Figure 4A). A substantial number of data points from the Abl 249 perturbed conditions occupied the same PC1 vs. PC2 parameter space as did the wild type data set. 250 Crucially, however, the data points from Abl KD and Abl OE that did not overlap wild type did 251 not just spread out isotropically, nor did they segregate into a discrete, separate domain of 252 parameter space. Instead, they formed a restricted distribution along a single vector emanating 253 from the cloud of wild type data. Stated otherwise, the ratio of the values of any pair of principal 254 components in the Abl-perturbed conditions varied along a simple, linear relationship relative to their values in wild type, even for the data points with the most severely altered absolute values. 256 This pattern was observed in all pairwise comparisons among the first 4 principal components, 257 which account for 71% of the variance in the data set, thus revealing consistent quantitative 258 relationships among the growth cone parameters that produce morphology and motility in 259 wildtype, Abl KD and Abl OE. This verifies that the fundamental relationships among 260 morphological and cytoskeletal properties of wild type growth cones are maintained in the Abl 261 KD and OE conditions, even for the most severely affected data points in our dataset, and therefore 262 validates the use of the altered Abl conditions to interrogate wild type growth cone dynamics. 263

Abl perturbation has large effects on the actin distribution and relatively smaller effects on 265 growth cone morphology 266
Comparison of parameters across all three genotypes revealed that the actin distribution 267 was far more sensitive to changes in Abl levels than were the morphological parameters of the 268 axon, consistent with the hypothesis that actin remodeling may be the more direct mechanistic 269 target of Abl-dependent signaling, upstream of changes in axonal morphology. Abl perturbation 270 affects many growth cone parameters, but to varying degrees. For example, the mean length of the 271 actin peak is decreased in the knockdown condition, and increased by Abl OE, yet is unaffected 272 by expression of Dicer 2 alone (mean ±SEM, 15.16 µm ± 0.19 in control vs. 14.0 µm ± 0.27 in 273 KD; p = 0.0051, 15.83 µm ± 0.24 in Dicer 2; N.S., and 16.72 µm ± 0.32 in OE; p = 0.0002; One -274 way ANOVA) ( Figure 4B). Similarly, the length of the protrusion peak in Abl KD is smaller than 275 wildtype (15.9µm ± 0.21 in control vs. 14.5µm ± 0.18 in KD; p<0.05) and is larger in Abl OE 276 (18.9µm ± 0.27; p< 0.001) ( Figure 4C). To determine which aspects of TSM1 architecture were 277 most affected by Abl, we calculated the Z-score for each parameter relative to the wildtype mean.
While disruption of Abl signaling does not shift the mean value of any morphological feature by 279 more than 0.7 standard deviations from its wildtype mean, actin parameter values are shifted by as 280 much as 2 -11 standard deviations away from their respective control means ( Figure 4D). As Abl 281 perturbation causes substantially larger effects on the distribution of actin in the axon as compared 282 to axonal morphology, we infer that the actin distribution is the more sensitive target of Abl 283 signaling. In light of the observation that the shape of the actin distribution is a sensitive target of 288 guidance signaling downstream of Abl, and that the pattern of actin redistribution predicts where 289 future axonal protrusions will grow, we next examined the evolution of the actin distribution itself 290 over time in each trajectory. We found that anterograde translocation of the actin peak was not 291 consistent. Rather, both the maximum of the actin peak, and its midpoint, displayed extensive, 292 seemingly stochastic, fluctuations proximo-distally along the axon, but with a small anterograde 293 bias that resulted, over time, in net forward motion of the actin peak ( Figure 5B This suggests that one effect of Abl is to regulate the coherence of the actin distribution in the 315 axon. 316 Next, we compared the actin distribution dynamics in the wildtype condition with the Abl 317 perturbed conditions and found that the actin distribution evolves in an orderly way from one-time 318 step to the next in wild type, but much less so when Abl is dysregulated ( Figure 6E  preferentially associated with the smallest growth cones in Abl KD (r = -0.14; p = 0.005), but it 342 was associated with the largest GCs in Abl OE (r = 0.25; p < 0.0001) ( Figure 6I). In the case of 343 Abl KD, this reflected a population of very short growth cones with hypercondensed actin, that is, 344 growth cones containing small, intensely-labelled foci of actin. Data presented above showed that 345 actin step size, ie the distance the actin maximum (or midpoint) advances in any single time step, 346 is correlated with the length of the actin peak ( Figure 5C). In Abl KD, however, the growth cone often contracts beyond the lower limit observed in wild type, and the step size falls to nearly zero 348 ( Figure 6J). This is a genotype in which we observe growth cone stalling, and failure to form an 349 axon. In contrast, in Abl OE, a genotype associated primarily with axon misrouting, maximal actin 350 fragmentation is associated preferentially with the largest actin distributions, and in particular, with 351 distributions broader than the upper limit typically observed in wild type ( Figure 6K Furthermore, our data demonstrate that the Abelson tyrosine kinase, a conserved regulator of 367 cytoskeletal dynamics that signals downstream of many guidance cue receptors, coordinates the 368 actin fluctuations that sum to produce the net forward motion of the distribution. Taken together, 369 our data suggest that the fundamental function of Abl during axon guidance and extension is to modulate, in a probabilistic way, the fluctuations and the coherence of an advancing actin wave 371 that directs the construction and consolidation of the growing axon in response to guidance cues 372 (Model Figure 7). 373 Growth cone advance is often discussed by invoking deterministic, clutch-like adhesive 374 mechanisms that harness the mechanochemical properties of leading lamellipodia 1,2,24 . For some 375 neurons, particularly those extending axons on relatively rigid, highly adherent substrata, these 376 models provide a plausible explanation for axon growth. However, TSM1 and many other axons 377 look non-lamellar, particularly in compliant, 3D environments like those often encountered by 378 pioneer axons in vivo 6,7,25,26 . In these contexts, growth cones are often dominated by filopodial 379 protrusions and seem to lack the large veil-like structures that we associate with the adhesive style 380 of growth. Moreover, the growth of these axons appears to be accomplished by protrusion and 381 selective stabilization of filopodia, rather than by traction forces applied to the lamellipodia-like We now find that the filopodial-dominated TSM1 growth cone extends its axon within the 388 intact Drosophila wing by regulated advance of an actin distribution, using a protrusive 389 mechanism of motility that is probabilistic rather than deterministic, and is based on the statistical 390 properties of disordered actin networks. Forward motion of the actin peak arises from a small 391 spatial bias that is applied to the fluctuating actin distribution; forward motion of the axon terminus 392 arises from preferential extension of filopodia from axon regions that now have high actin density and retraction of filopodia in regions of low actin density. In essence, the advancing actin peak 394 directs processive axon growth by locally promoting assembly of potential axonal tracks, while 395 the axon cannibalizes filopodia that lag behind the actin peak, and correspond to axonal paths that 396 were not taken. 397 The key to such a mechanism for axon extension is that forward expansion of the actin 398 distribution must be large enough to advance the actin bolus and produce net growth, yet 399 constrained enough that the actin bolus remains coherent and the growth cone behaves as a single, 400 unitary entity. We see here that balancing these two competing, yet related, requirements is the steps. Second, we also found that Abl regulates the width of the actin peak. This is seen most 409 clearly in the Abl perturbations, where reducing Abl promotes condensation of the actin mass 410 while increasing Abl promotes expansion of the peak. Specifically, Abl knockdown causes the 411 actin bolus to hyper-condense and fragment into small, tightly-packed foci that show reduced 412 spatial motion (Figure 4 and 6). Since expansion of the distribution provides the forward motion 413 required for 'inchworming', it is plausible that failure of that expansion in the Abl KD condition 414 is responsible for the growth cone stalling observed in this genotype. Conversely, overexpression 415 of Abl causes the actin distribution to expand, often to the degree that the peak exceeds the characteristic decay length over which the motions of individual actin molecules remain correlated 417 within disordered actin networks (~11-16µm) 23 . This likely contributes to the preferential 418 fragmentation we observe of the broadest actin peaks in the Abl OE condition, and plausibly could 419 predispose to axonal misrouting and inappropriate axon branching in part by allowing different 420 portions of the same growth cone to act independently. The two key functions of Abl, minimizing 421 disorder and controlling distribution width, are closely intertwined; in the Abl KD condition the 422 tendency is for the smallest growth cones to drive measures of disorder, while in Abl OE, the 423 degree of actin disorder increases with expanding length of the actin peak ( Fig 6I). The mechanism of axon growth we observe here is energetically expensive, with only ~5-474 15% of the total back-and-forth motion of the actin peak being captured in net advance of the actin 475 mass. Considered in context, however, this mechanism allows an extremely efficient use of 476 guidance information. The fluctuations of the actin peak cause it to repetitively sample leading and 477 trailing positions along the axon multiple times before moving irrevocably along its trajectory. We 478 propose that this is useful, and probably essential, for responding accurately to the shallow, often 479 noisy gradients of individual guidance cues, which are typically presented in complex 480 combinations in the substratum. In this sense, the mechanism we have found is most akin to the 481 mechanism of bacterial chemotaxis, in which external attractants and repellants provide only a 482 subtle spatial bias to essentially stochastic fluctuations of the motility machinery, relying on a 483 'random walk with a ratchet' to produce net guidance 34,35 . Additionally, we note that these actin 484 fluctuations survey all growth cone protrusions -off-axis lateral projections as well as on-axis extensions of the axon shaft-thereby allowing growth cone turning to derive from precisely the 486 same machinery as does linear extension of the axon. 487 The wave-like, anterograde propagation of an actin distribution, regulated by the conserved 488 Abelson tyrosine kinase, which probabilistically localizes the site of filopodial dynamics, 489 represents a novel mechanism of motility that drives the extension and guidance of the TSM1 axon  Table 1) 543 Custom scripts in Mathematica software were written to identify computationally the 544 position of highest filopodial density along the axon by using a sliding-window method to sum the 545 length and number of protrusions within a 5μm window that advanced 1um/per step along the 546 segmented axon. We varied window size from 1-10 μms and empirically found peak protrusion 547 density assignment was insensitive to the size of the window (data not shown). We then calculated 548 the square root of the second moment about the peak of protrusion density to determine the length 549 of the protrusive zone, both separately for the portions of the distribution leading and trailing the 550 peak position, and also globally to measure the length of the entire distribution. 551 SWC files from a complete trajectory of TSM1 growth and the corresponding 4D Z stacks, 554 (x, y, z, time) converted to Nikon image cytometry standard (ICS) format in Imaris, were loaded 555 into two MiPav plugins, PluginDrosophilaCreatesSWC and PlugIn3DSWCStats to extract the 556 actin distributions from each time point. Plugin code is available in MiPav, but its function is 557 described below. 558 In brief, image intensity is calculated by summing the actin intensity within sequential 559 frustums that encompass the axon.     Principal component analysis was performed using 10 measured variables that quantify aspects of 887 TSM1 axonal growth. The first four eigenvectors account for ~71% of the total variance in the 888 data set. 889