PMC

Summary and possible roles of Phldb2 for synaptic plasticity in the spine. Postsynaptic membrane area is shown in purple.

A PI3K inhibitor induces a decrease in the number of spines with Phldb2-positive heads. (A) Cultured hippocampal neurons were cotransfected with expression vectors for GFP-Phldb2 and tdTomato. GFP-Phldb2 localized in the dendritic spines. Neurons subjected to 10 μM Ly294002 treatment for 1 hr are shown in the right panels. Representative neurons are shown in the upper panels. Magnified images of the squares are shown in the panels below. (B) Head+ indicates a spine whose head is Phldb2 positive, whereas head- indicates a spine whose head is Phldb2 negative. Typical examples are shown in (A) Head+ spines are shown by arrows, whereas head- spines are indicated by arrowheads. (C) The densities of head+ and head- spines were measured in secondary and tertiary dendrites. GFP-Phldb2-transfected neurons were treated with Ly294002 (Ly294002, n = 322 spines) or did not receive Ly294002 treatment (vehicle, n = 686 spines) (Mean ± SEM. Student’s t-test, **P < 0.01). Scale bar represents 20 μm in (A).

Deletion of Phldb2 impairs performance on the T-maze left-right discrimination test. (A–C) Left-right discrimination in a T-maze test was examined. The graph shows the percentage of correct choices (A), duration (B) and total distance (C). The Phldb2^−/− mice (n = 11) made significantly fewer correct choices than the Phldb2^+/+ mice (n = 9). (Mean ± SEM. Two-way repeated measures ANOVA, F _(1,18) = 7.69, P = 0.013). The average duration and total distance were not significantly different between the Phldb2^−/− mice and Phldb2^+/+ mice [Mean ± SEM. Two-way repeated measures ANOVA, F _(1,17) = 1.625, P = 0.220 for (A) and F_(1,17) = 1.614, P = 0.221 for (B)]. (D) No obvious differences were observed in the locomotor activity of the two genotypes. The total distance moved in the 24-hr locomotor test did not differ between the Phldb2^−/− mice and the Phldb2^+/+ mice (Mean ± SEM. Student’s t-test, n = 14 for each mice, P = 0.660 for dark and P = 0.330 for light).

Phldb2 is necessary for LTP induction. (A) Electrophysiological analyses on the effects of Phldb2 on hippocampal LTP; time course of the normalized fEPSP slope recorded in slices from the Phldb2^+/+ mice and the Phldb2^−/− mice. LTP was induced by high-frequency stimulation (HFS) (100 Hz; 100 pulses, 1 sec) of the Schaffer collaterals. LTP was induced in the Phldb2^+/+ slices (n = 8) but not the in Phldb2^−/− slices (n = 6) (Mean ± SEM. Two-way repeated measures ANOVA, F _(1,12) = 1.09, P = 0.0035). One hundred percent corresponds to the pre-LFS baseline. (B) The input-output curve of fEPSP slope (mV/ms) versus presynaptic fibre volleys (FV; mV) at the Schaffer collateral pathway did not differ between Phldb2^+/+ slices (n = 10) and Phldb2^−/− slices (n = 13) (Mean ± SEM. One-way ANOVA analysis, F _(1,225) = 0.7596, P = 0.3844). (C) Paired-pulse facilitation, the short-term enhancement of synaptic efficacy following the delivery of two closely spaced stimuli, did not significantly differ between Phldb2^+/+ slices (n = 10) and Phldb2^−/− slices (n = 13) (Mean ± SEM. One-way ANOVA analysis, F _(1,171) = 1.431, P = 0.233).

Phldb2 gene deletion results in a significant decrease in synaptic AMPA receptor density. Replicas were prepared from the CA1 region of the hippocampus and labelled for AMPA receptors ((A–C): GluA1–3; (D–F): GluA1) in combination with the NR1 subunit of NMDA receptors as a marker for excitatory synapses. The synaptic identity of these IMP cluster areas (purple area) was further confirmed by immunolabelling for the NR1 subunit visualized with 10 nm immunogold (black arrowheads in A and D). Immunoreactivity for GluA1–3 or GluA1 was visualized with 5-nm immunogold particles (orange arrowheads) in (A,D), respectively. (B,E) The numbers of immunoparticles for GluA1–3 (B) or GluA1 (E) in individual IMP clusters were plotted against the IMP cluster areas. In both cases, a statistically significant positive correlation between the AMPAR labelling numbers and synaptic areas was found regardless of genotype (Pearson’s correlation test for GluA1–3: the Phldb2^+/+ mice, n = 43 synapses, r = 0.742, P < 0.001; the Phldb2^−/− mice, n = 36 synapses, r = 0.896, P < 0.001, and Spearman’s rank-order test for GluA1: the Phldb2^+/+ mice, n = 69 synapses, r = 0.344, P < 0.01; the Phldb2^−/− mice, n = 69 synapses, r = 0.558, P < 0.001). However, significant reductions in synaptic GluA1–3 (C) and GluA1 (F) labelling densities were detected in the Phldb2^−/− mice compared with the Phldb2^+/+ mice (Mean ± SEM. Student’s t-test for GluA1–3, *P < 0.05, Spearman’s rank-order test for GluA1, *P < 0.05).

Phldb2 interacts with PSD-95 and regulates PSD-95 turnover. (A) Expression vectors for HA-PSD-95 and GFP-Phldb2 or GFP-mock were cotransfected in COS-7 cells. Lysates were subjected to immunoprecipitation with anti-GFP antibody. PSD-95 was co-immunoprecipitated with Phldb2 (star). The grouping of blots was cropped from Supplementary Fig. . (B) Hippocampal neurons were transfected with tdTomato expression vectors and fixed at day 21 in vitro (DIV 21). High-magnification image of a mushroom spine. Endogenous PSD-95 was stained (green). A plot of the fluorescence intensity profile along the yellow line shows the distribution of endogenous PSD-95 (green line) and tdTomato (red line). The spine lengths were defined by the fluorescence intensity profile of tdTomato. (C) The spine head lengths along the yellow line were divided into 10 parts, shown on the X-axis. The fluorescence peak of PSD-95 was consistently slightly farther from the spine head in the Phldb2^−/− mice (n = 23 spines) than in the Phldb2^+/+ mice (n = 25 spines). The Y-axis represents normalized fluorescence intensity. For normalization, signal intensity in the dendritic spine head length was divided by peak intensity (intensity/peak intensity ratio) (Mean ± SEM. Student’s t-test, *P < 0.05). (D) Cultured hippocampal neurons were cotransfected with expression vectors for photoactivatable green fluorescent protein-tagged PSD-95 (PAGFP-PSD-95) and tdTomato in the Phldb2^+/+ mice and the Phldb2^−/− mice at DIV 20. PAGFP-PSD-95 was photoactivated by two-photon excitation with 730 nm laser light in the indicated areas (arrowheads) at time 0. (E) Semi-quantification of PAGFP-PSD-95 fluorescence in the spine. The rate of fluorescence of intensity of PAGFP-PSD-95 increased in the Phldb2^−/− mice (Phldb2^+/+ mice, n = 10 spines; Phldb2^−/− mice, n = 8 spines. Mean ± SEM. Two-way repeated measures ANOVA, F _(1,17) = 5.25, P = 0.035). Scale bars represent 5 μm.

Phldb2 gene deletion results in a significant decrease in synaptic NMDA receptor density without changes in the size of postsynaptic densities or the volume of spine heads. (A) Replicas were prepared from the CA1 region of hippocampus, and dendrites and dendritic spines in replicas were identified based on morphology under a transmission electron microscope. (B) Postsynaptic membrane specializations of excitatory synapses in replicas were identified in the exoplasmic (E)-face of the plasma membrane by clusters of intra-membrane particles (IMP clusters, purple) labelled for the NR1 subunit (arrowhead) and visualized with 10 nm immunogold. (C) The numbers of immunoparticles for NR1 in individual IMP clusters were plotted against the areas of IMP clusters. A statistically significant positive correlation between the NR1 labelling number and the synaptic area was found regardless of genotype (Pearson’s correlation test: the Phldb2^+/+ mice, n = 62 synapses, r = 0.680, P < 0.001; the Phldb2^−/− mice, n = 62 synapses, r = 0.763, P < 0.001). (D) The average labelling density for synaptic NR1 was significantly lower in the Phldb2^−/− mice than in the Phldb2^+/+ mice (Mean ± SEM. Student’s t-test, **P < 0.001). (E,F) Reconstruction of dendritic spines from serial FIB-SEM images clearly demonstrates a full view of a dendritic spine, its head portion, and a postsynaptic membrane specialization (green lines) defined by the postsynaptic density (PSD). (E) Examples of FIB-SEM images from Phldb2^+/+ mice and Phldb2^−/− mice (spine head in yellow, PSD in green and the rest of the spine in purple). (F) Examples of 3D-reconstructed spines from Phldb2^+/+ and Phldb2^−/− mice (spine in transparent purple, head in transparent yellow and PSD in green). (G) The PSD areas were plotted against the spine head volumes. A statistically significant positive correlation between the PSD area and the spine head volume was found in both genotypes (Spearman’s rank-order test: the Phldb2^+/+ mice, n = 30 synapses, r = 0.879, P < 0.001; the Phldb2^−/− mice, n = 30 synapses r = 0.835, P < 0.001). The average PSD areas (H) and head volumes (I) were not significantly different between the Phldb2^−/− mice and the Phldb2^+/+ mice [Spearman’s rank-order test, P = 0.598 for (H) and P = 0.330 for (I)].

Phldb2 deletion results in a decrease in the number of GluA2-positive spines and the surface accumulation of GluA2. (A) Expression vectors of myc-Phldb2 or myc-mock, and GFP-GluA1 or GFP-GluA2 were co-transfected in COS-7 cells. Lysates were immunoprecipitated with anti-myc antibody. Phldb2 was co-immunoprecipitated with GluA1 and GluA2, and an amount of Phldb2 was increased in the presence of GluA1 or GluA2. The grouping of blots was cropped from Supplementary Fig. . (B) At DIV 21, endogenous GluA2 subunits were sparsely distributed in the spines of the Phldb2^−/− mouse hippocampal neurons. (C) For rescue experiments, hippocampal neurons of the Phldb2^−/− mice were transfected with tdTomato expression vectors and myc-Phldb2 or myc-Phldb2 ΔPH expression vector. Neurons were fixed at DIV 21. Endogenous GluA2 subunits were stained (green). (D) The number of spine-associated GluA2 puncta was significantly reduced in the Phldb2^−/− mice and was rescued by Phldb2 expression. (Phldb2^+/+ mice, n = 348 spines; Phldb2^−/− mice, n = 187 spines; Phldb2^−/− mice with exogenous Phldb2, n = 283 spines; Phldb2^−/− mice with ΔPhldb2, n = 154 spines, Tukey-Kramer test, *P < 0.05, **P < 0.01). (E) Cultured hippocampal neurons were transfected with HA-GluA2 and tdTomato expression vectors at DIV 19. At DIV 21, HA-GluA2 at the membrane surface was visualized by staining for HA without Triton X-100 treatment (green). The neurons were then treated with Triton X-100, and total HA-GluA2 was observed (blue). Magnified images of the dendritic regions in white squares are shown in the right panels (arrowheads). (F) For Ly294002 treatment, 10 μM Ly294002 was added to the medium 60 min in advance of observation. (G) The fluorescence intensity of the surface GluA2 was divided by that of total GluA2 (surface/total ratio of GluA2) in the Phldb2^+/+ neurons and was defined as 1.0 for normalization. The normalized surface/total GluA2 ratio was lower in the Phldb2^−/− mice than in the Phldb2^+/+ mice. The surface/total ratio of GluA2 in the Phldb2^+/+ mice was decreased by Ly294002 treatment, whereas the same was not true in the Phldb2^−/− mice (Mean ± SEM. n = 10 neurons for each group, Student’s t-test, *P < 0.05). Scale bars = 10 µm (B,C) and 20 µm (E,F).