Click here to close
Hello! We notice that you are using Internet Explorer, which is not supported by Xenbase and may cause the site to display incorrectly.
We suggest using a current version of Chrome,
FireFox, or Safari.
???displayArticle.abstract???
Cells such as astrocytes and radial glia with many densely ramified, fine processes pose particular challenges for the quantification of structural motility. Here we report the development of a method to calculate a motility index for individual cells with complex, dynamic morphologies. This motility index relies on boxcar averaging of the difference images generated by subtraction of images collected at consecutive time points. An image preprocessing step involving 2D projection, edge detection, and dilation of the raw images is first applied in order to binarize the images. The boxcar averaging of difference images diminishes the impact of artifactual pixel fluctuations while accentuating the group-wise changes in pixel values which are more likely to represent real biological movement. Importantly, this provides a value that correlates with mean process elongation and retraction rates without requiring detailed reconstructions of very complex cells. We also demonstrate that additional increases in the sensitivity of the method can be obtained by denoising images using the temporal frequency power spectra, based on the fact that rapid intensity fluctuations over time are mainly due to imaging artifact. The MATLAB programs implementing these motility analysis methods, complete with user-friendly graphical interfaces, have been made publicly available for download.
Figure 1. Radial glia process movement in vivo is apparent over a timescale of minutes. (a) Maximum intensity projections of two-photon microscope z-series images of a radial glia cell expressing farnesylated EGFP in the intact tadpolebrain. (b) RGB overlay of the time-lapse images of a radial glia cell where red denotes the position of the cell at 0 minutes, green after 5 minutes, and blue after 10 minutes of imaging. (c) Magnified images of several highly dynamic sites from the radial glial cell shown in (b). Scale bar = 10 μm.
Figure 2. Image preprocessing for analysis. (a) Maximum intensity projections of two-photon images of a radial glia cell expressing farnesylated EGFP in vivo collected at 0, 5, and 10 minutes. (b) Same images after Sobel edge detection and binarization. (c) Binarized images underwent dilation with a 6-pixel radius. Pixel dilation helps eliminate artifactual discontinuities in the images (arrowhead in insets) resulting from the binarization step. Scale bar = 10âμm.
Figure 3. Analysis of pixel redistribution using boxcar averaging. (a) Preprocessed images from 0, 5, and 10 minutes time points. (b) Subtracting consecutive images from one another creates a map of the redistributed pixels between time points, shown in red. (c) Boxcar averaging of the pixel redistribution images from (b) using 9 à 9 pixel window. (d) Result of multiplication of the images in (b) and (c). The average of the nonzero pixel values in this image is used to calculate the cell motility index. (e) Plot of motility index values obtained for the datasets in Figures 5(c), 5(d), 6(c), and 6(d) using a range of different boxcar sizes. Motility values in the different conditions are most clearly separated when a 9 à 9 pixel boxcar size is used. Scale bar = 10âμm.
Figure 4. Temporal frequency filtering helps exclude artifactual pixel redistributions. (a) Time series of two-photon images of a portion of a radial glia cell. Scale bar = 2âμm. (b) Integrated intensity of a healthy cell over the course of 18 minutes, preprocessed for analysis. Red pixels have a value of 1 during most time points and blue pixels during few time points. Black pixels are negative background sites with value 0 at all time points. Time-dependent pixel intensity functions (i) for a pixel located within the main process, (ii) for a pixel into which a filopodium extended, and (iii) for a noisy pixel located along the edge of the cell. (c) Frequency map showing noisy pixels that were discarded by temporal frequency filtering. The colors represent discrete high frequencies measured in the excluded range. (d) Frequency map of the same cell, showing pixels that were accepted by the frequency filter. Each color shows a discrete frequency within the accepted range. In (c) and (d), black represents sites where pixel intensity did not change over the 18 minutes.
Figure 5. Comparing the relative sensitivity of the two motility measurement methods using images of normal radial glia cells (N = 10) before and after injection of 100 μM MK-801 into the ventricle of the tadpole to reduce motility. ((a), (b)) Example of glial motility before (a) and after (b) drug treatment. (c) Motility index changes measured using Method 1 from [15]. (d) Results using the boxcar-based Method 2. Each individual cell is depicted with the same color in (c) and (d). Note that Method 2 gives more consistent results. All values are normalized to the mean predrug motility index to facilitate comparison. Scale bar = 10 μm in upper panel and 2 μm in lower panel of (a).
Figure 6. Use of paraformaldehyde- (PFA-) fixed cells to reveal nonbiological artifacts that contribute to the motility index scores. (a) The motility of control living cells (N = 4) was compared with (b) dead cells fixed in paraformaldehyde (N = 5). ((c), (d)) Bars indicate the motility values of the two groups normalized to the mean of control values using Method 1 in (c) and Method 2 in (d). Error bars are SEM. (e) Applying a low-pass temporal frequency filter with a threshold of 0.0026âHz (6.4âmin period) reduces the motility index scores (measured using Method 2) of the PFA-fixed cell group by substantially more than the control group, indicating that the filter is effective at eliminating nonbiological noise. Dashed lines indicate full width at half maximum. Scale bars = 10âμm.
Bosgraaf,
Quimp3, an automated pseudopod-tracking algorithm.
2010, Pubmed
Bosgraaf,
Quimp3, an automated pseudopod-tracking algorithm.
2010,
Pubmed
Costantino,
Semi-automated quantification of filopodial dynamics.
2008,
Pubmed
Dormann,
Simultaneous quantification of cell motility and protein-membrane-association using active contours.
2002,
Pubmed
Dunaevsky,
Developmental regulation of spine motility in the mammalian central nervous system.
1999,
Pubmed
Fischer,
Glutamate receptors regulate actin-based plasticity in dendritic spines.
2000,
Pubmed
Guberman,
PSICIC: noise and asymmetry in bacterial division revealed by computational image analysis at sub-pixel resolution.
2008,
Pubmed
Holtmaat,
Experience-dependent structural synaptic plasticity in the mammalian brain.
2009,
Pubmed
Hua,
Neural activity and the dynamics of central nervous system development.
2004,
Pubmed
Huisken,
Optical sectioning deep inside live embryos by selective plane illumination microscopy.
2004,
Pubmed
Iino,
Glia-synapse interaction through Ca2+-permeable AMPA receptors in Bergmann glia.
2001,
Pubmed
Lavialle,
Structural plasticity of perisynaptic astrocyte processes involves ezrin and metabotropic glutamate receptors.
2011,
Pubmed
Lichtman,
The neuronal naturalist: watching neurons in their native habitat.
2001,
Pubmed
Lippman,
Dendritic spine morphogenesis and plasticity.
2005,
Pubmed
Machacek,
Morphodynamic profiling of protrusion phenotypes.
2006,
Pubmed
Niell,
Live optical imaging of nervous system development.
2004,
Pubmed
Nishida,
Direct astrocytic contacts regulate local maturation of dendritic spines.
2007,
Pubmed
Planchon,
Rapid three-dimensional isotropic imaging of living cells using Bessel beam plane illumination.
2011,
Pubmed
Ruthazer,
Bulk electroporation of retinal ganglion cells in live Xenopus tadpoles.
2013,
Pubmed
,
Xenbase
Schneider,
NIH Image to ImageJ: 25 years of image analysis.
2012,
Pubmed
Thévenaz,
A pyramid approach to subpixel registration based on intensity.
1998,
Pubmed
Tremblay,
Regulation of radial glial motility by visual experience.
2009,
Pubmed
,
Xenbase
