Open Access

André Ferreira Castro

• Even one century after Santiago Ramón y Cajal’s groundbreaking contribu- tions to neuroscience, one of the most fundamental questions in the field is still largely open, namely understanding how the shape of a dendrite is adapted to its specific biological function. A systematic investigation of this problem is challenging both technically and conceptually because neurons have diverse genetic, molecular, morphological, connectional and functional properties. In the light of the preceding, dendritic arborisation (da) neurons of the Drosophila melanogaster larva PNS have proven to be an excellent model system for the study of such growth and patterning processes. Structure and function in these cell classes are intimately intertwined, as class type-specific dendritic arbour differentiation processes are required to satisfy a given phys- iological need. Also, there is a remarkable genetic toolkit that enables one to selectively and reproducibly label, image and manipulate each one of these sensory neuron classes. In this thesis, I address the aforementioned open problem by linking single-cell patterning, information processing and wiring optimisation in sensory da neurons to behaviour in Drosophila larva. In particular, I study Class I ventral peripherical dendritic arborisation (c1vpda) neurons. These are a class of proprioceptive neurons that relay information on the position of the larva’s body back to the CNS during crawling behaviour to assure proper locomotion. Their stereotypical comb- like shaped dendritic branches spread along the body-wall, and they get noticeably deformed during crawling behaviour. The bending of the den- dritic branches is hypothesised to be a possible mechanism to transduce the mechanosensory inputs arising from cuticle folding. Interestingly, c1vpda neurons do not necessarily satisfy optimal wiring constraints since they are required to pattern into a specific shape to fulfil their function. Therefore, I considered the da system to study how the specific functional requirements may be combined with optimal wiring constraints during development. Although the molecular machinery of dendrite patterning in c1vpda neurons is well studied, the precise elaboration of the comb-like shaped dendrites of these cells remains elusive. Moreover, even though a lot of work has been put into the description and quantification of growth processes of the nervous system, there are still few solid and standardised models of arbour staging and patterning. Importantly, the defining parameters that determine the dendrite elaboration program that in turn is responsible for creating the final arbour morphology are still unknown. As a result, unraveling possible universal stages of dendrite elaboration shared between different model systems and cell types is challenging. Thus, in order to understand the development of the fine regulation of branch outgrowth that leads to the observed terminal arbour morphology in the mature cell, I collected in vivo, long-term, non-invasive high temporal res- olution time-lapse recordings of dendritic trees during the differentiation process in the embryo and its maturation phase in the larva. For further analysis, I developed new algorithms that quantified the structural changes in dendrite morphology in the time-lapse videos. My approach provides a framework to analyse such developmental data, or any dataset comprising continuous morphological dynamical processes in an unbiased way. Using these newly developed methods, I examined the development of a sample of c1vpda cells and identified five stages of differentiation in these data: initial stem polarization, extension, pruning, stabilization, and isometric stretching during larval stages. The beginning of the growth process is marked by the polarisation of the main stem. Subsequently, during the extension phase, branches emerge interstitially from the existing main stem. Later, higher-order branches sprout from pre-existing lateral branches, increasing arbour complexity. This is followed by a pruning stage where developmental intermediate dendritic branches are removed. This step leads to a spatial rearrangement of the dendritic tree. The end of the pruning step is followed by a stabilisation period where arbour morphology remains virtually unaltered in the embryo. After hatching, c1vpda dendrites experience an isometric scaling, with their branching complexity and pattern being invariant across all larval stages. After dissecting the c1vpda dendrites spatiotemporal differentiation process, I established a link between dendritic shape and behaviour. I measured intra- cellular Ca++ activity in the dendrite branches of l1 larvae during forward locomotion, while simultaneously recording branch deformation using a dual genetic line. I reported that post-embryonic c1vpda dendrites Ca++ responses increased in freely crawling larvae. Furthermore, I showed strong correlations between Ca++ signal and deformation of the comb-like dendritic ranches during body-wall contractions. Then, using a geometrical model, I provided evidence that the pruning stage could reorganise the dendrite morphology to maximise mechanosensory re- sponses during body wall contraction. I showed that the angle orientation of each side branch correlates with the bending curvature and thus with the me- chanical displacement of the cell membrane during locomotion. During the pruning phase, I observed a preferential reduction of less efficient branches with low bending curvature, influencing the mechanisms of dendritic sig- nal integration of c1vpda sensory neurons. I proceeded to quantify branch dynamics at single tip resolution during pruning, providing evidence that a simple random pruning mechanism is sufficient to remodel the tree structure compatible with the observed way. I used these time-lapse data to constrain a new computational noisy growth model with random pruning based on optimal wiring principles. This model is able to generate highly realistic synthetic c1vpda morphologies. The model furthermore requires few parameters to generate highly accurate temporal development trajectories and morphologies at single-cell level. Utilising this data and model enabled me to investigate upon the hypothesis that a noisy dendrite growth and random pruning mechanism synergise to achieve den- dritic trees efficient in terms of both wiring and function. My findings show how single neurons can create functionally specialised dendrites while min- imising wiring costs, elucidating how general principles of self-organisation may be involved in the generation of these structures.