Automated tracking of dendritic protrusions in live neurons. 11 DIV hippocampal neurons were transfected with a vector expressing Lifeact-ruby, and 24 hrs later were imaged at 5 min intervals for 1 hr. (a) Individual protrusions were tracked across the time series manually and with our automated method. This histogram shows the percentages of stable, new, and pruned protrusions for both methods. (b) Images depict 12 DIV neurons expressing Lifeact-ruby (white; top) overlayed with automated 3D reconstructions (bottom). The dendrite shaft is red and each tracked protrusion is labeled 1 through 4 (shades of blue and green). Images from t = 0 and 60 min are shown, and the scale bar is 5 μm. (c) Head width and (d) protrusion length were plotted versus time for each protrusion; the labels 1 through 4 in the legend correspond to the labels 1 through 4 in panel (b).

Acute BDNF induces specific types of spine remodeling. 12 DIV hippocampal neurons expressing Lifeact-ruby were treated with vehicle or 100 ng/ml BDNF and imaged at 5 min intervals for 1 hr. Each protrusion was classified as stubby, mushroom, or thin at t = 0 and 60 min. (a-d) At left, each each time series (0 - 60 min) depicts a representative type of dendritic protrusion remodeling observed during our analysis. At right, the automated 3D reconstructions illustrate the classification of each protrusion at t = 0 and 60 min (stubby: yellow, mushroom: green, thin: blue). (e) The diagram shows the percentages of pre-existing stubby, mushroom, and thin protrusions that were remodeled (to stubby, to mushroom, to thin) or pruned under control and BDNF-treated conditions. (f) Using the same dataset as in (e), we calculated the total incidence for each type of remodeling under control and BDNF-treated conditions. The histogram depicts the percentages of total protrusions that were initially stubby, mushroom, or thin and were either remodeled (to stubby, to mushroom, or to thin) or pruned. (g) The initial (t = 0) mean protrusion head width, neck width and length as well as the head width/neck width ratio were determined for mushrooms spines that were either stable (maintained mushroom morphology), remodeled into thin protrusions, or pruned within the 60 min imaging period following BDNF treatment (N = 24 - 32 spines; ANOVA, post-hoc Tukey's test; head width: ^‡P = 0.032, *P = 0.017; neck width: *P = 0.020, ^‡P = 0.037; length: ^‡P = 0.032, *P = 0.017; head/neck ratio: *P = 0.002, ^‡P = 0.001). (h) The group means listed above were determined for thin protrusions (t = 0) that were stable (maintained thin morphology), remodeled into mushroom spines, or pruned during 30 min. BDNF treatment (N = 36 - 47 spines; ANOVA, post-hoc Tukey's test; length: *P = 0.002; head/neck ratio: *P = 0.037).

Automated detection of dendritic spines using images of Lifeact-expressing neurons. (a) Images depict a representative 3D reconstruction of a 17 DIV rat hippocampal neuron expressing Lifeact-ruby (white) and the automated trace generated by Filament Tracer (blue). Scale bar is 10 μm. (b) Automated and manual spine counts were performed within the same dendritic regions. Top: A dendritic segment of a Lifeact-ruby-expressing neuron (white) and the corresponding automated trace (dendrite: red, spines: blue) as well as manually marked spines (yellow spheres) are shown. Middle: Lifeact-ruby signal (white) is overlayed by the automated trace (shown alone at right). Bottom: Lifeact-ruby signal (white) is overlayed by manual spine marks (yellow; shown alone at right). (c) For each dendritic segment, the manual spine count was plotted against the automated count. Linear regression analysis showed that manual and automated spine detection were significantly correlated (n = 28; β = 0.824; p < 0.001).

Acute BDNF treatment induces maturation of the dendritic spine population. 11 DIV hippocampal neurons were transfected with a vector expressing Lifeact-ruby, and 24 hrs later, the neurons were treated with vehicle or 100 ng/ml BDNF followed by time-lapse imaging every 5 minutes for 1 hr. Each protrusion was tracked across time and measured using the automated method. (a) Head volume was plotted as the percent change from the initial time point (T₀). Statistical analyses were performed to compare protrusion head volume at T₀ and T₆₀ (N = 105 - 135 spines; Kruskal Wallis test with repeated measures; Control: P = 0.548; BDNF: P = 0.001). (b) Neck width was plotted and analyzed as above (Control: P =0.91; BDNF: P = 0.0002). (c) Dendritic protrusion length was plotted and analyzed as above (Control: P = 0.648; BDNF: P = 0.017). (d) Dendritic protrusions were classified as stubby, mushroom, or thin at each time point. The percentages of total protrusions within each class are presented for T₀ and T₆₀ (*P = 0.038, ^‡P = 0.044, ^#P = 0.015). (e) The number of protrusions within a dendritic region was determined using Imaris Filament Tracer and compared between T₀ and T₆₀ (N = 25 - 30 neurons; repeated measures ANOVA, post-hoc Tukey's test [Control: P = 0.219; BDNF: P = 0.014]).

A PI3 kinase inhibitor rescues spine morphology in neurons from Fmr1 knockout mice. (a) Hippocampal neurons cultured from wild type (WT) or Fmr1 knockout (KO) mice were treated with vehicle or a PI3 kinase inhibitor (LY294, 10 μM) for 72 hours starting at 15 DIV. Neurons were transfected with a plasmid encoding Lifeact-ruby at 17 DIV and fixed 24 hours later. The images depict representative dendritic regions from deconvolved z-series images (scale bar is 5 μm). (b) Spine density was measured using our automated approach for WT and KO neurons treated with either vehicle or LY294 (n = 55-60 neurons; ANOVA [F = 3.996, P = 0.009]; post-hoc Fisher's LSD: *P = 0.017, ^#P = 0.028). Cumulative distributions of (c) spine head width, (d) length, and (e) volume were plotted for each group (Kolmogorov-Smirnov test: head width [WT vs. KO: P = 0.002, WT vs. WT LY294: P = 0.235, KO vs. KO LY294: P = 0.009], spine length [WT vs. KO: P = 0.009, WT vs. WT LY294: P = 0.537, KO vs. KO LY294: P = 0.014], spine volume [WT vs. KO: P < 0.001, WT vs. WT LY294: P = 0.158, KO vs. KO LY294: P < 0.001]). (f) Spines were classified as stubby, mushroom, and thin based on the automated geometric measurement, and the values were plotted as percentage of total spines per treatment group (n = 55 - 60 neurons; ANOVA with post-hoc Fisher's LSD: *P = 0.043, #P = 0.006, ^‡P = 0.043).