Figure 1.
Postsynaptic structural and functional changes induced by NMDA-evoked LTD.
Neuronal cultures were prepared and assayed on DIV 21-23, and NMDA-evoked LTD (nLTD) was induced as described in Methods. (A1) nLTD is associated with long lasting changes on spine density and morphology. Data are expressed as mean ± SEM; ***p<0.0001, Mann-Whitney test (number of dendrites: ctrl = 135, nLTD = 121; number of spines: ctrl = 3753, nLTD = 1490). (A2) Selected dendritic regions of neurons expressing eGFP illustrate typical spine density and morphology in the absence (control) versus presence of NMDA for 4 min (nLTD). Image width = 27.3 µm. (B1) Loss of surface AMPA receptors accompanies the spine shrinkage induced by nLTD. Changes in SEP-GluA2 fluorescence (F/F0) and spine area (A/A0) were quantified using time-lapse imaging during the nLTD stimulus for the indicated times. Data are expressed as mean ± SEM (n = 110 spines). (B2) Time montage of an individual dendritic spine expressing mcherry and SEP-GluA2 at selected time points after nLTD induction (time ‘0’ = start of NMDA addition). Image width = 4.4 µm. (C) The calcineurin inhibitor FK506 (1 µM, 1 hr prior to addition of NMDA) reduces nLTD-associated spine loss in eGFP-expressing neurons. Data are expressed as mean ± SEM; ***p<0.001 compared to control alone; ###p<0.001 compared to nLTD alone; two-way ANOVA, followed by Bonferroni post hoc test (number of dendrites: ctrl = 56, nLTD4 min = 46, FK506 = 39, nLTD + FK506 = 70). (D1) nLTD is associated with a decrease in F-actin in live neurons observed using Lifeact. Data are expressed as mean ± SEM; (n = 44 spines). (D2) Time montage of an individual spine expressing Lifeact-RFP and eGFP at indicated time points following nLTD induction (time ‘0’ = start of NMDA addition). Image width: 3 µm. (E) Actin filament concentration is reduced during nLTD. Relative F-actin concentration within spines was quantified as the ratio of the integrated intensity of Alexa568-phalloidin divided by spine area. Data are expressed as mean ± SEM; ***p<0.001, unpaired t-test; a.u. = arbitrary units; (number of spines: ctrl = 304, nLTD = 259). (F) Free barbed ends of F-actin decrease in spines during nLTD. FBE concentration was quantified as described in Methods. Data are expressed as mean ± SEM; ***p<0.001, unpaired t-test; a.u. = arbitrary units; (number of spines: ctrl = 236, nLTD = 217). Experiments in A, C, E and F and their corresponding supplemental material were analyzed in a blind fashion.
Figure 2.
nLTD decreases the molecular proximity between endogenous spine cofilin and F-actin
. (A) Two examples from control cultures illustrate that endogenous cofilin is present both in neurons (MAP2-positive) and astrocytes (GFAP-positive) Yellow arrows point to astrocytic GFAP-positive processes, which often express levels of endogenous cofilin greater than or equal to that in neurons. Image width = 127 µm. (B) FRET was carried out as described in Methods. In brief, fixed neurons expressing membrane-targeted CFP were used to select spines for monitoring FRET between fluorescently tagged secondary antibodies against cofilin and actin. (B1) Example from a typical experiment to illustrate the acceptor photobleaching method. Shown is immunostaining in a control sample for endogenous F-actin and cofilin in a field containing a dendrite expressing membrane targeted CFP (CFP image shown at upper left). Image width = 11 µm. The white circle indicates the area that was targeted for acceptor photobleaching. Note that bleaching was restricted to a small region within the actin image, and this same region underwent a corresponding increase in intensity in the cofilin image when the actin is photobleached. For the assay FRET is quantified only within the spine area, not throughout the circular bleached region. (B2) Quantification of FRET between cofilin and F-actin at dendritic spines of control (ctrl) neurons versus neurons expressing chronophin (CIN) and incubated in the absence or presence of NMDA, as indicated. Data are expressed as mean ± SEM of the “% FRET” values determined using the Zeiss LSM 510 software module for FRET acceptor photobleaching (see Methods); *p<0.05, **p<0.01 compared to control alone; #p<0.05 compared to nLTD alone; XXp<0.01 compared to CIN alone, two-way ANOVA, followed by Bonferroni post hoc test., (number of spines: ctrl = 43, nLTD = 29, CIN = 49, CIN+nLTD = 77). Note that CIN overexpression acted as predicted to significantly increase FRET between endogenous cofilin and F-actin. (B3) Ratio of cofilin-actin FRET signal with or without the induction of LTD in control neurons versus neurons overexpressing CIN. p<0.001, unpaired t-test with Welch's correction. Data calculated from that in Fig. 2B2. (C1) The in situ Duolink proximity ligation assay (PLA) technology was carried out as described in Methods. White dots represent the detection of cofilin-actin interaction complexes. Image width = 127 µm. (C2) Cofilin-actin interaction in control versus nLTD was quantified as the number of neuron-associated Duolink puncta per dendritic length (see Methods). Two different combinations of cofilin and actin primary antibodies were used — one each raised in either rabbit (Rb) or mouse (M), and paired with the appropriate anti-rabbit or anti-mouse secondary antibody. Regardless of antibody combination we found that nLTD decreased the detection of cofilin-actin interaction (data shown in either blue or yellow bars). Note the low levels for interaction between phospho-cofilin (p-cof) and actin (red bar), as expected. Data are expressed as mean ± SEM; ****p<0.0001, unpaired t-test, number of fields: [α-cofM/α-actinRb]: 44; ***p<0.001, Mann-Whitney test, number of fields: α-cof Rb/α-actin M = 29; number of fields: α-pcof Rb/α-actin M = 20.
Figure 3.
Overexpression of either chronophin or cofilin prevents nLTD-induced loss and shrinkage of dendritic spines.
Neurons were transfected with either the cofilin phosphatase chronophin (CIN) or wildtype cofilin itself, as indicated, together with the cell filler eGFP prior to induction of nLTD. Image width = 27.3 µm; all images show eGFP). Morphological data was quantified using eGFP images and are expressed as mean ± SEM; *p<0.05, **p<0.01, ***p<0.001, compared to control alone; ## p<0.01, ###p<0.001 compared to nLTD alone; XXp<0.01 compared to CIN or cofilin (cof) alone, two-way ANOVA, followed by Bonferroni post hoc test. (A) CIN overexpression prevents nLTD-associated spine loss and shrinkage. Selected dendritic regions of neurons co-expressing CIN and the cell filler eGFP at 4 min post nLTD. For spine density, number of dendrites: ctrl = 38, nLTD = 40, CIN = 32, CIN+nLTD = 59); for spine length and width, number of spines: ctrl = 659, nLTD = 551, CIN = 635, CIN+nLTD = 1109). (B) Cofilin overexpression prevents nLTD-associated spine loss and shrinkage. Selected dendritic regions of neurons co-expressing ectopic cofilin and the cell filler eGFP at 4 min post-nLTD. For spine density, number of dendrites: ctrl = 27, nLTD = 21, cof-wt = 54, nLTD+cof-wt = 58); For spine length and width, number of spines: ctrl = 542, nLTD = 348, cof-wt = 1063, nLTD+cof-wt = 1095). (C) Chronophin overexpression prevents the nLTD-induced reduction in F-actin concentration in dendritic spines. F-actin concentration in spines was quantified as the ratio of the integrated intensity of Alexa 568-phalloidin divided by spine area on neurons incubated in the absence or presence of NMDA for 4 min (nLTD). number of spines: ctrl = 304, nLTD = 259, CIN = 231, CIN+nLTD = 253; (a.u. = arbitrary units). (D) Chronophin overexpression prevents the LTD-induced reduction in actin filament free barbed ends (FBEs) in dendritic spines. FBE concentration was quantified as described in Methods on neurons incubated in the absence or presence of NMDA for 4 min (LTD). Number of spines: ctrl = 236, nLTD = 217, CIN = 96, nLTD+CIN = 167 (a.u. = arbitrary units). Experiments in this figure and its corresponding supplemental material were analyzed in a blind fashion.
Figure 4.
Cofilin concentration selectively decreases in spines following nLTD and involves the capthepsin B/L class of proteases.
(A) Analysis of total cofilin concentration versus phospho-cofilin concentration in the absence and presence of nLTD. (A1) Representative example of an eGFP-expressing dendrite costained for phosphorylated (p-cof) and total cofilin (cof). As described in Methods, a digital processing procedure was used to quantify immunoreactivity only in spine heads (i.e. not the neck or parent dendrite) and also to exclude spines that overlapped with other adjacent structures like astrocytes. At left are shown the eGFP image (top) and the merged image for eGFP, p-cofilin and total cofilin (bottom). At right are shown the p-cofilin and total cofilin (bottom) images, with the perimeter of the dendrite (green outline) overlayed onto each image. Image width = 20 µm. (A2) Quantification of average phosphorylated (left) and total cofilin (right) concentrations in dendritic spine heads in the presence or absence of NMDA; integrated immunostaining intensity within each spine was divided by spine area. Data are expressed as mean ± SEM; ***p<0.001; unpaired t-test; number of spines: ctrl = 1608, nLTD = 714; a.u. = arbitrary units. (B) Analysis of total cofilin concentration in dendritic spines versus dendritic shafts in the presence and absence of nLTD. (B1) Cultures were transfected with eGFP prior to being triple-immunostained for MAP2, cofilin/ADF, and GFAP (not shown in the image). The boxed region at lower right indicates the region magnified in the two panels in b2 at right. Image width = 70 µm. (B2) Following digital removal of regions stained for GFAP, digital outlines were created based on the eGFP image (green) and the MAP2 image (red – which, as shown here, appears yellow in the merged image where eGFP and MAP2 are colocalized within the dendritic shaft –). These outlines were then overlaid onto the cofilin image in order to quantify the concentration of cofilin within each subcellular region (see Methods for details. Image width = 11 µm. (B3) nLTD has no effect on average cofilin concentration in the dendritic shaft. Data are expressed as mean ± SEM; p = 0.9529, Mann-Whitney test number of neurons: ctrl = 26, nLTD = 29. a.u. = arbitrary units. (B4) nLTD induces a significant decrease in cofilin concentration in dendritic spines that is prevented by incubation with the cathepsin B/L inhibitor CA074Me (4 µM, 30 min). Data are expressed as mean ± SEM; *p<0.05, one-way ANOVA, followed by Tukey post hoc test (number of dendrites: ctrl = 10, nLTD = 25, CA074Me+nLTD = 27). (B5) The cathepsin inhibitor prevents spine loss during nLTD. **p<0.01 one-way ANOVA followed by Tukey post hoc test (number of neurons: ctrl = 26, nLTD = 29, CA074Me+nLTD = 27). a.u. = arbitrary units. Experiments in this figure and its corresponding supplemental material were analyzed in a blind fashion.
Figure 5.
ADF/Cofilin gene silencing induces spine loss and shrinkage.
Upper Panel: Images from selected dendritic regions of neurons expressing the eGFP-tagged pSuper empty vector, the scrambled control, or the cofilin/ADF-RNAi (cof/ADF-RNAi) for 4 days prior to fixation. Image width = 27.3 µm. Lower Panel: Quantification of dendritic spine density, length and width. Data are expressed as mean ± SEM; ***p<0.001, Kruskal-Wallis test followed by Dunn's multiple comparisons test, number of dendrites: empty v. = 34, scramble = 13, cof/ADF-RNAi = 61; number of spines: empty v. = 642, scramble = 251, cof/ADF-RNAi = 712. Experiments in this figure and its corresponding supplemental material were analyzed in a blind fashion.
Figure 6.
LIMK overexpression induces PLD-dependent spine expansion and prevents nLTD-induced spine shrinkage.
(A) Quantification of the levels of phospho-cofilin immunoreactivity within dendritic spines in the absence versus presence of overexpressed LIMK for 24 hr. Data are expressed as mean ± SEM; ***p<0.001, unpaired t-test with Welch's correction; number of spines: ctrl = 20 nLTD = 24; a.u. = arbitrary units. (B) LIMK overexpression reduces FBEs. Concentration of labeled FBEs in spines of control neurons versus those overexpressing LIMK. Data are expressed as mean ± SEM; **p<0.01, Mann-Whitney test, number of spines: ctrl = 40, nLTD = 153. (C) Spine expansion induced by overexpression of LIMK is blocked by the PLD inhibitor FIPI. Left: Selected dendritic regions of neurons expressing either eGFP together with mcherry or eGFP together with LIMK for 24 hours while incubated in the presence or absence of FIPI. Only the eGFP channel is shown here. Image width = 27.3 µm. Quantification of spine morphology (length and width) is shown at right. Data are expressed as mean ± SEM; number of spines: ctrl = 646, FIPI = 694, LIMK = 830, LIMK+FIPI = 885; * p<0.05, ***p<0.001 compared to control alone; ###p<0.001 compared to LIMK alone; XXp<0.01 compared to FIPI alone, two-way ANOVA, followed by Bonferroni post hoc test. Experiments were analyzed in a blind fashion. (D) PLD inhibition blocks the prevention of LTD-induced spine shrinkage by LIMK. Changes in spine area (A/A0) were quantified using time-lapse imaging of neurons expressing LIMK and eGFP during incubation with NMDA for the indicated times. Data are expressed as mean ± SEM; differences in spine area for the two treatment groups were statistically significant (p = 0.0428) by repeated-measures two-way ANOVA, number of spines: 18. At right selected dendritic regions from a time-lapse series during nLTD; yellow arrows point to a subset of spines. Image width = 10 µm.
Figure 7.
Pharmacological inhibition of PLD prevents gLTP-induced spine enlargement and F-actin polymerization.
Time montages of dendritic segments from neurons expressing eGFP and Lifeact incubated in the absence (A) or presence (B) of the PLD inhibitor FIPI prior to induction of gLTP, as described in Methods. Note the substantial enlargement in dendritic spine size that occurs within 20 min post-gLTP; this expansion is almost completely suppressed by FIPI. Two examples are shown for each condition to illustrate the fact that both small spines and larger spines undergo expansion during gLTP. Cyan and magenta arrows indicate, respectively, spines that are small and large before any treatment. Image width = 13 µm. (C) Left: Quantification of changes in spine size during LTP in the absence versus presence of FIPI, as evaluated using time-lapse images of the cell filler eGFP. Differences in spine area and F-actin intensity for the two treatment groups were statistically significant at 10 and 30 min. Data are expressed as mean ± SEM, **p<0.01, two-way ANOVA, followed by Bonferroni post hoc test, number of spines: ctrl = 39, FIPI = 59. Right: Quantification of changes in spine F-actin concentration during LTP in the absence versus presence of FIPI, as evaluated using the Lifeact probe to label endogenous F-actin. Data are expressed as mean ± SEM, *p<0.05, **p<0.01, two-way ANOVA, followed by Bonferroni post hoc test, number of spines: ctrl = 39, FIPI = 59. (D) Quantification of average phospho-cofilin concentration in dendritic spines of neurons exposed to control or gLTP conditions for 10 min and then either fixed (10 min) or exchanged back to control buffer solution for an additional 20 minutes before fixation (30 min). Data are expressed as mean ± SEM, ***p<0.001, unpaired t-test, number of neurons (10 min): ctrl = 18, gLTP = 11; ***p<0.001 Mann-Whitney test, number of neurons (30 min): ctrl = 35, gLTP = 42.
Figure 8.
Blocking interaction between phospho-cofilin and PLD1 prevents gLTP-induced spine head enlargement.
(A) Neuronal cultures were transfected with either empty vector (HA-empty) or HA-tagged F3 fragment of PLD1 (HA-F3), which prevents activation of endogenous PLD1 by endogenous phospho-cofilin, prior to induction of gLTP, as described in Methods. Example images of transfected control neurons are shown in (A). Image width = 22.7 µm. Quantification of results (B) showed that gLTP induced a significant increase in spine area in the presence of the control construct, but gLTP induced a small but significant decrease in spine area in the presence of the blocking fragment. The fragment by itself had no significant effect on spine area. Data are expressed as mean ± SEM; number of spines (HA-empty vector): ctrl = 3331, gLTP = 3464; number of spines (HA-F3): ctrl = 3608, gLTP = 5377; ***p<0.001 compared to control alone; ###p<0.001 compared to gLTP alone; XXXp<0.001 compared to HA-F3 alone, two-way ANOVA, followed by Bonferroni post hoc test. (C) Diagram summarizing key findings from this study. The left panel illustrates the observation that dendritic spine shrinkage during nLTD is associated with a decrease in the concentration of cofilin within dendritic spines, while spine growth during gLTP is associated with an increase in the phosphorylation of cofilin on Ser-3, in agreement with previous reports. The upper right panel illustrates our proposed model that under basal conditions cofilin activity is centrally important in maintaining spine volume due to its promotion of actin turnover. Cofilin contributes fresh barbed ends as well as a critical recycling of actin monomer. The mass action of many polymerizing actin filaments provides a steady force against the plasma membrane that helps maintain spine volume. An abrupt loss of cofilin, as during nLTD, leads to a reduction in this internal force. The bottom right panel outlines the signaling pathways that mediate cofilin's role in nLTD-induced spine shrinkage and gLTP-induced spine growth. Experiments in this figure were analyzed in a blind fashion.