Poster Z10 - Network Topology Analysis for the Localization of Fluorescent Lifetime Fluctuations in Different Orders of Dendrites

Introduction: cAMP-dependent protein kinase, PKA, plays an important role in shaping developing dendrites in multiple organisms, including Drosophila, mice, and rats [1, 2]. However, whether there is spatial control of PKA activation during dendritogenesis is not known. To address this gap in our knowledge, we used the PKA activation sensor tAKAR4α and performed Fluorescence Lifetime Imaging Microscopy (FLIM) in developing cultured rat hippocampal neurons. We compared activation of PKA in different orders of dendrites in baseline conditions of development and with application of brain-derived neurotrophic factor (BDNF), the most well-studied regulator of dendrite branching. We developed an analysis pipeline that combines computer vision methods and network topology to automatically localize fluctuations in lifetime of fluorescence. We chose graph theory and network topology to preserve the physical structure of the neuron while allowing the analysis of the changes to probe lifetime at discrete points, or ‘nodes’, in the cell. We based our choice of methods as they have recently been applied to create novel automated neuronal tracing algorithms. Our final goal is to analyze temporal and spatial fluctuations in time-series data to create a method to automatically identify points of high PKA activation based on changing donor lifetime values from the tAKAR4α probe.

Materials and

Methods: Cell Culture: All work was performed in accordance with the guidelines set by the IACUC at Rutgers, The State University of New Jersey. The hippocampi of Sprague Dawley rat embryos were dissected at 18 days of gestation. Hippocampus cultures were prepared in 35 mm glass bottom dishes pre-coated with 100 μg/mL poly-D-lysine. Cultures were maintained in Neurobasal medium with 2% B27 supplement and 1% GlutaMAX.

Transfection: On day in vitro (DIV) 4, cultures were transfected using Lipofectamine 2000 and one of the following 3 plasmids: mCerulean, tAKAR4α, and tAKARαT391A, a mutant which does not change conformation following PKA activation.

Imaging: On DIV7, cultures were imaged using the Leica TCS SP8 confocal microscope to obtain intensity and donor lifetime data from the 3 probes. Following baseline imaging, either vehicle (0.1% DMSO) or 25 ng/mL BDNF was added, and thirty-nine additional frames were taken.

Algorithm development: Intensity data were segmented twice: first, using the SCIRD-TS algorithm to segment dendrites [3], and second, using a Gabor filter bank to isolate the soma. The combined segmentation mask was skeletonized and converted into a network. Nodes were classified as ‘dendrites’, ‘terminals’, ‘branchpoints’, or ‘soma’ based on their degree and location. Calculation of the shortest path between the soma and each terminal was used to determine branchpoint order. Optical flow vectors were calculated on the lifetime matrices using the Horn-Schunck algorithm to quantify fluctuations between frames. Pixels with a high variance of optical flow magnitudes were extracted as points with fluctuating lifetime.

Results, Conclusions, and Discussions: The segmented dendrites can successfully be ordered using the shortest path algorithm (Fig. 1). The script orders nodes into one of the following groups: ‘soma’, ‘primary dendrite’, ‘secondary dendrite’, ‘tertiary dendrite’, or ‘higher-order dendrite’. The main hurdle is misidentification of a subset of nodes as part of the soma rather than as dendrites, which then affects dendrite order classification. More restrictive soma segmentation algorithms will be tested to create a more accurate soma segmentation mask and reduce dendrite misordering. Our optical flow-based algorithm identifies timepoints of high optical flow variance and the location of these fluctuations in different areas of the neuron (Fig. 2). The next step is to isolate these hotspots and determine whether the lifetime increases or decreases at those pixels to differentiate whether the sensor is activated or returns to its baseline state, and this will be key to answer biological questions about the spatio-temporal activation of PKA during neuronal maturation in vitro and in response to BDNF.

References:
1. Zhong, H., et al., Subcellular dynamics of type II PKA in neurons. Neuron, 2009. 62(3): p. 363-374.
2. Copf, T., Developmental shaping of dendritic arbors in Drosophila relies on tightly regulated intra-neuronal activity of protein kinase A (PKA). Developmental biology, 2014. 393(2): p. 282-297.
3. Annunziata, R. and E. Trucco, Accelerating convolutional sparse coding for curvilinear structures segmentation by refining SCIRD-TS filter banks. IEEE transactions on medical imaging, 2016. 35(11): p. 2381-2392.

Acknowledgements (Optional): This work was supported by grant # 1R21NS122159-01 from NINDS.