3D Particle Tracking for Noninvasive In Vivo Analysis of Synaptic Microtubule Dynamics in Dendrites and Neuromuscular Junctions of Drosophila

Cells organize to form functional structures through the coordination of intra- and intercellular changes via morphogenesis. A remarkable example of morphogenesis is the development of the highly specialized neuronal structure. Neurons display remarkable polarization, in which they extend two structurally and functionally distinct types of processes, dendrites and axons¹, which can achieve immense lengths. The complexity of neuronal development arises not only from the sheer size of dendrites and axons but also from the difficulty in forming their intricately branched geometries^2,3. Neuronal morphogenesis and its consequences in learning and memory⁴ motivate the ongoing investigation of both its genetic control and the underlying cell biological mechanisms. Such mechanisms include, but are not limited to, intracellular membrane transport and the many cytoskeletal rearrangements needed for changes in neuronal morphology^1,2,3.

Studies of neuronal morphogenesis have produced a variety of advanced visualization techniques. Static methods, such as electron microscopy or fluorescence microscopy of fixed probes, are widely used to perform high-resolution morphological and structural analysis. However, besides the artifacts that are inevitable to any preservation method, static visualization cannot capture the dynamic changes that underpin morphogenesis. Thus, many pivotal insights originated from time-lapse fluorescence microscopy of living tissues. Early work by Lichtman and colleagues^5,6,7 utilized in vivo imaging of the mammalian nervous system to investigate axon regeneration/degeneration, organization of synaptic components, and long-range axonal transport. Furthermore, seminal studies in primary neuronal explants were critical to establishing the importance of microtubule (MT) dynamics to axonal elongation and motility^8,9. Crucially, early neuronal explant studies established the use of fluorescently-tagged end-binding family proteins (EBs) to gain invaluable insights into MT plus-end dynamics in developing neurons at the level of individual MT polymers¹⁰. These studies arose from observations that the EB family member EB1 preferentially localizes to MT plus ends¹¹ in S. cerevisiae¹² and in cultured cells¹³. Since then, EB1 and other plus tip tracking proteins (+TIPs)^14,15 have been widely used in in vivo studies of MT dynamic instability¹⁶, including in the context of neuronal development¹⁷.

Drosophila is a powerful model for in vivo imaging studies of MT dynamics during neuronal development due to the vast genetic and imaging tools available for fly studies^18,19 as well as the similarities in structure and function between Drosophila and vertebrate neurons¹. A key early study of the neuromuscular junction (NMJ) of Drosophila larvae performed repeated noninvasive imaging of a fluorescent membrane marker through the translucent cuticle of intact animals to document presynaptic terminal morphogenesis²⁰. Using a similar method to image whole, live Drosophila larvae, an initial demonstration of subcellular, particle-level analysis of processive movement of motor cargos in the axons was provided²¹. More recently, meticulous studies by Rolls and colleagues in the sensory dendrites of intact Drosophila larvae^{22,23,24,25,26,27} characterized postsynaptic MT plus-end dynamics by performing particle tracking and analysis of green fluorescent protein (GFP)-tagged EB1. Such studies in Drosophila^{22,23,24,25,26,27} and other systems^{28,29,30,31,32} have significantly advanced understanding of single-polymer behavior of MT plus ends in the dendrites of developing neurons³³.

Despite the impressive in vivo studies of postsynaptic MT dynamics^{22,23,24,25,26,27,28,29,30,31}, there have been far fewer comparable studies of presynaptic MT dynamics at the developing axon terminal. MT dynamics at the Drosophila larval NMJ has been studied using fluorescent speckle microscopy (FSM) and fluorescence recovery after photobleaching (FRAP)³⁴. These techniques evaluate the overall tubulin kinetics but not the behavior of individual MT plus ends. As of this writing, there has been one sole investigation of individual MT plus ends at the Drosophila NMJ: This study combined live time-lapse imaging with manual analysis of kymographs to characterize a population of dynamic, EB1-GFP labeled "pioneering MTs" that appeared distinct from a broader population of stabilized MTs³⁵. This lack of research on presynaptic MT dynamics may be due at least in part to anatomy: While it is relatively straightforward to obtain images of dendrites due to their proximity to the larval cuticle, NMJs are obstructed by other tissues, making it challenging to acquire images with sufficient signal-to-noise ratio for particle-level analysis. Nonetheless, given the well-established importance of the presynaptic MTs to synaptic morphogenesis and stabilization³⁶, as well as their links to neurodevelopmental and neurodegenerative disorders³⁷, bridging this gap between understanding of pre- and postsynaptic MTs is likely to yield invaluable insights. 

An additional challenge to the analysis of in vivo MT dynamics in general, in contrast to in vitro or ex vivo analysis, is the limited automated software tools that can extract dynamics parameters from in vivo data. Presently, one of the most popular and powerful techniques for analysis of +TIP-labeled MT plus ends is plusTipTracker^38,39, a MATLAB-based software that allows automated tracking and analysis of multiple dynamics parameters. Notably, plusTipTracker measures not only MT growth but also shrinkage and rescues: while +TIP labels such as EB1-GFP only associate with growing plus ends, plusTipTracker can algorithmically infer shrinkage rates and rescue events. However, while plusTripTracker has been very successfully applied to many contexts, including previous multiparametric analysis of ex vivo MT dynamics in Drosophila S2 cells⁴⁰, plusTipTracker is not optimal for analysis of in vivo data given their lower signal-to-noise ratio. As a result, in vivo studies of plus-end dynamics at dendrites^{22,23,24,25,26,27} and at the NMJ³⁵ of Drosophila have relied on manual generation and analysis of kymographs using software such as ImageJ⁴¹, or on semiautomated strategies that involve numerous human-in-the-loop components.

This study presents an experimental and analytical workflow that reduces the experimental and analytical overhead required to perform noninvasive polymer-level analysis of presynaptic MT dynamics in both sensory dendrites and the motor axon terminal of Drosophila third-instar larvae. The protocol utilizes immobilized, intact larvae and therefore avoids injuries known to trigger stress responses as well as other nonphysiological conditions that might perturb in vivo MT dynamics. To label dynamic MT plus-ends, EB1-GFP is pan-neuronally expressed using the Gal4/UAS system⁴², allowing visualization of MTs at both dendrites and NMJ with a single driver. While some early steps are inevitably subject to human decision-making, such as the selection of animal specimens and identification of regions to image, the steps following data acquisition are largely automated. Crucially, optimization of a new software enabled automated, unbiased analysis requiring minimal human input. While other particle tracking methods are available^43,44,45, this study utilizes a proprietary software because it was algorithmically well-suited to address the particular challenges of this particular dataset. The software is now available to users for a variety of applications. Specifically, the use of coherence-enhancing diffusion filtering⁴⁶ is integral to automated segmentation and background removal, and custom algorithms are implemented specifically to automate particle detection and tracking. This strategy could effectively handle the low signal-to-noise ratio inherent to the data in this study, as well as other challenges, such as movement of EB1-GFP comets through different focal planes. While it is not feasible to exhaustively test the performance of this software against all other particle analysis software, the performance of the present strategy equaled or approached the standard human performance. Furthermore, to the authors’ knowledge, there has been no other software specifically trained on in vivo data from sensory dendrites and the presynaptic terminal. Given that the performance of image analysis algorithms is often highly specific to the data they were designed for and that generalized computer vision is not yet possible, it is expected that training the described software to the specific in vivo data of interest is the most algorithmically sound approach.

Given the extensive work on dendritic MTs^{22,23,24,25,26,27} as well as the consistent quality of data that can be acquired from this system, the image acquisition and software analysis strategy was first validated in Drosophila sensory dendrites. Importantly, it was found in dendrites that the use of different neuronal Gal4 drivers, even in otherwise identical wild type backgrounds, results in significant differences in EB1-GFP dynamics due to differences in genetic background, emphasizing the importance of using a single Gal4 driver for consistent results. This strategy was next used for multiparametric analysis of EB1-GFP dynamics at the presynaptic terminal of the NMJ. To further illustrate the investigative value of this method, this imaging and software strategy was used to assess both pre- and postsynaptic EB1-GFP dynamics following knockdown of dTACC, the Drosophila homolog of the highly conserved TACC (transforming acidic coiled coil) family^47,48. Prior work in Drosophila S2 cells⁴⁰, as well as work by Lowery and colleagues in the Xenopus growth cone^49,50,51, has shown that TACC family members regulates MT plus-end dynamics. Furthermore, recently reported evidence from confocal and super-resolution immunofluorescence imaging showed that dTACC is a key regulator of presynaptic MTs during neuronal morphogenesis⁵², raising the question of whether dTACC regulates live MT dynamics. This report demonstrates a method that can indeed detect differences in live MT behaviors upon dTACC knockdown. Thus, this study presents an in vivo method that can effectively identify and characterize key regulators of MT dynamics within the developing neuron, particularly in the presynaptic compartment.