The X-linked intellectual disability protein TSPAN7 regulates excitatory synapse development and AMPAR trafficking

Abstract

Mutations in TSPAN7--a member of the tetraspanin protein superfamily--are implicated in some forms of X-linked intellectual disability. Here we show that TSPAN7 overexpression promotes the formation of filopodia and dendritic spines in cultured hippocampal neurons from embryonic rats, whereas TSPAN7 silencing reduces head size and stability of spines and AMPA receptor currents. Via its C terminus, TSPAN7 interacts with the PDZ domain of protein interacting with C kinase 1 (PICK1), to regulate PICK1 and GluR2/3 association and AMPA receptor trafficking. These findings indicate that, in hippocampal neurons, TSPAN7 regulates AMPA receptor trafficking by limiting PICK1 accessibility to AMPA receptors and suggest an additional mechanism for the functional maturation of glutamatergic synapses, whose impairment is implicated in intellectual disability.

Figure 1

TSPAN7 Overexpression Induces Filopodia and Spines in Cultured Neurons (A and B) Primary rat hippocampal neurons were transfected at DIV1 with pIRES2-EGFP, pIRES2-EGFP-TSPAN7 or pIRES2-EGFP-TSPAN7ΔC and fixed at DIV5 (A) or DIV7 (B). Scale bar represents 10 μm, higher magnification inset. Mean (±SEM) filopodia density in axons (A, DIV5) and dendrites (B, DIV7) is significantly greater in cells expressing TSPAN7 construct than GFP controls and neurons expressing TSPAN7ΔC (DIV5: ^∗∗∗p < 0.001; DIV7: ^∗∗p = 0.002, Tukey after ANOVA; at least nine neurons examined for each construct). (C) Neurons were transfected at DIV11 with EGFP, EGFP plus HA-TSPAN7, or EGFP plus HA-TSPAN7ΔC, fixed at DIV21 and stained with anti-HA antibody. EGFP was used to outline dendrites and spines. HA-TSPAN7 (middle) is present in patches along dendrites and dendritic spines, whereas TSPAN7ΔC does not localize to spines. Bottom panels show boxed sections in top panels at higher magnification. Scale bar represents 10 μm. The histograms compare mean (±SEM) spine density (number of spines/10 μm of dendrite), length and head width (at least nine neurons and 180 dendrite sections examined for each construct). Spine density was significantly greater in HA-TSPAN7 neurons than controls and HA-TSPAN7ΔC neurons (^∗∗p = 0.009, ^∗p = 0.02, Tukey after ANOVA), whereas spine width was significantly greater in HA-TSPAN7 and EGFP controls than HA-TSPAN7ΔC neurons (^∗p = 0.013, ^∗∗p = 0.007; Tukey test ANOVA).

Figure 2

Effects of TSPAN7 Overexpression in Mature Hippocampal Neurons (A) Neurons at DIV11 were transfected with EGFP, pIRES2-EGFP-TSPAN7, or pIRES2-EGFP-TSPAN7ΔC, and stained at DIV18. Scale bar represents 10 μm. The histograms show that TSPAN7 overexpression increases staining intensity and cluster density for GluA2 and PSD-95 whereas TSPAN7ΔC overexpression decreases them. (B) Double labeling of endogenous TSPAN7 with either Bassoon, GluA2, surface β1 integrin, PSD-95, or GluN1, showing that TSPAN7 localizes with these markers to varying extents. Scale bar represents 10 μm. The histograms show percentages of TSPAN7-positive puncta colocalizing with the indicated protein. (15 cells per condition).

Figure 3

TSPAN7 Knockdown Reduces Dendritic Spine Width and Enhances Spine Motility and Turnover (A) Hippocampal neurons were transfected at DIV11 with either scrambled siRNA14 (scrambled), siRNA14, siRNA14 plus a wild-type variant of TSPAN7 resistant to siRNA14 (rescue WT), or siRNA14 plus TSPAN7ΔC resistant to siRNA14 (rescue ΔC), and immunostained at DIV18 for TSPAN7. TSPAN7 appears much reduced in TSPAN7-knockdown dendrites, and TSPAN7ΔC does not localize in spines. Scale bar represents 10 μm. (B) Histograms comparing mean spine density (number of spines/10 μm of dendrite), spine length and spine width in neurons transfected as in (A). Mean dendritic spine width (error bar represents SEM) in neurons transfected with siRNA14 was reduced relative to neurons transfected with scrambled siRNA14 (at least 10 neurons and 1,000 spines examined for each construct; ^∗∗∗p < 0.001). The effect of TSPAN7 silencing on spine width was rescued by coexpressing rescue WT but not rescue ΔC. (C) Time-lapse imaging comparing spine dynamics in neurons transfected with either scrambled siRNA14, siRNA14, siRNA14 plus rescue WT (rescue WT), or TSPAN7. The number of spines that disappeared (arrows at 0 min) and that appeared after 40 min (arrows at 40 min) were expressed as percentage of total spines. The histogram shows that mean number of lost (^∗∗∗p < 0.001) and new spines (^∗∗∗p < 0.001) was higher in neurons transfected with siRNA14 than scrambled siRNA14 (Tukey after ANOVA, at least 10 neurons per condition; scale bar represents 10 μm). The effects on spine turnover were fully rescued by wild-type TSPAN7 (rescue WT). The overexpression of TSPAN7 increases the number of ex novo spines (^∗∗∗p < 0.001). Error bars represent SEM. (D) Neurons were transfected at DIV11 with scrambled siRNA14 or siRNA14, and chemical LTP was induced. siRNA14 prevents spine enlargement and spine number increase due to LTP. Scale bar represents 10 μm.

Figure 4

Immunostaining for Endogenous Synaptic Markers in Neurons Expressing Either siRNA14, Scrambled, Rescue WT, or Rescue ΔC (A) In each panel, EGFP is on the left, immunostaining for GluA1, GluA2/3, PSD-95, GluN1, or surface β1 integrin is in the middle and merge on the right. Individual channels are shown in black and white. Scale bar represents 10 μm. (B) Histograms show mean (error bars represent SEM) staining intensity (left) and number of clusters per 10 μm of dendrite (right) for the specified markers. Values are normalized to the level in scrambled siRNA14 cells (at least 9 neurons and 100 synapses/neuron analyzed per condition). Neurons expressing siRNA14 have significantly lower staining intensity for GluA1 (^∗p = 0.019), GluA2/3 (^∗∗∗p < 0.001), and PSD-95 (^∗∗p = 0.001) but not GluN1 or surface β1 integrin; the density of clusters was also reduced for GluA1, GluA2/3, and PSD-95 (^∗∗p = 0.003, ^∗∗p = 0.004, ^∗∗∗p < 0.001, respectively; Tukey after ANOVA). Rescue WT but not rescue ΔC rescued the effects of siRNA14 on GluA1, GluA2/3, and PSD-95.

Figure 5

TSPAN7 Knockdown Reduces Excitatory Synaptic Function (A) Overlaid miniature excitatory postsynaptic currents (mEPSCs) are shown recorded from primary hippocampal pyramidal neurons transfected with either empty vector pLL3.7 (Control), siRNA14, siRNA47, or siRNA14 + wild-type TSPAN7 rescue (Rescue). Gray traces are aligned mEPSCs and colored traces are population averages for representative cells. (B) Summary of mEPSC amplitudes, shown as cumulative distributions and bar graphs (inset). Numbers of recorded cells are indicated in the bar graphs (^∗p < 0.05; ^∗∗p < 0.01). Knockdown of TSPAN7 significantly reduces mEPSC amplitude relative to control and rescued neurons. (C) Representative mEPSC recordings are shown from neurons transfected with either the empty vector pLL3.7 (Control), siRNA14, siRNA47, or siRNA14 + wild-type TSPAN7 rescue (Rescue). (D) Summary of mEPSC frequency data, shown as cumulative distributions and bar graphs (inset). Numbers of recorded cells are indicated in the bar graphs (^∗p < 0.05). TSPAN7 knockdown reduces mEPSC frequency relative to control and rescued neurons.

Figure 6

TSPAN7 Directly Interacts with PICK1, Associates with β1 Integrin and GluA2/3, and Regulates PICK1-GluA2/3 Interaction (A) Representation of TSPAN7 showing transmembrane domains (red) and TSPAN7 C-terminal tail (aa 234–249). The latter was used as bait in two-hybrid screen of a human fetal brain cDNA library. Four prey cDNA clones were isolated (clones 40, 48, 15, and 28)—all encoded PICK1. (B) Yeast two-hybrid test with, as bait, the full length C terminus of TSPAN7 (wt) or a truncated version of it (Δ4, aa 234–245) without the last four aas (putative PDZ binding motif). As potential prey, full length PICK1 (aa 1–415), N terminus + PDZ (aa 1–108), PDZ domain (aa 14–108), N terminus + PDZ + linker region (aa 1–147), BAR domain (aa 142–367), N terminus + PDZ + linker + BAR (aa 1–367), and two constructs mutated at sites expected to abolish PDZ-dependent binding: PDZ (KD → AA) (aa 14–108) and full length PICK1 (KD → AA) were probed. Full length PICK1, and N terminus + PDZ + linker + BAR (aa 1–367) constructs interacted with the TSPAN7 C-terminal tail (wt) (3+). (C) Coimmunoprecipitation and pull-down experiments with TSPAN7 and PICK1. In COS7 cells HA-TSPAN7 coimmunoprecipitated with myc-PICK1 (top). In GST pull-downs in COS7 cells (bottom), the C terminus of TSPAN7 (ct wt, 234–249 aa) pulled down PICK1 (myc-PICK1) and PICK1's PDZ domain (myc-PDZ) but not the mutated PDZ domain (KD → AA; myc-PDZ^∗) or the PICK1 fragment lacking the first 121 aas (flag-Δ121). (D) PICK1 and TSPAN7 interaction in neurons. Immunofluorescence labeling of hippocampal neurons (top) shows that endogenous TSPAN7 and PICK1 colocalize to some extent in dendrites and spines. Coimmunoprecipitation experiments on brain extracts using monoclonal PICK1 antibody (bottom) show that TSPAN7 and PICK1 are associated. (E) The C-terminal tail of TSPAN7 (ct wt; 234–249 aa) pulls down PICK1, GluA2/3, and β1 integrin (left) from primary hippocampal neuron extracts. Coimmunopurification of AMPAR complexes using anti-GluA2/3 antibodies (middle and right panels) shows that PICK1 and GluA2/3 associate more in siRNA14-infected neurons than scrambled siRNA14-infected neurons (1.22 ± 0.05 versus 1.01 ± 0.01, ^∗∗p = 0.004; PICK1/GluA2/3 ratio in siRNA14-expressing neurons relative to that in scrambled siRNA14-expressing neurons). β1 integrin associates with GluA2/3 only in the presence of TSPAN7 (middle).

Figure 7

TSPAN7 Knockdown Results in Increased GluA2 Internalization (A) Representative images of GluA2 internalization experiments 0, 5, and 10 min after a 10 min incubation with anti-GluA2 antibody, in TSPAN7 knockdown and scrambled siRNA14 neurons, untreated (left) or treated with dynasore (right). (B) Extent of GluA2 internalization was quantified as the ratio of intracellular to total fluorescence (internalization index) in untreated and dynasore-treated neurons. The internalization index is significantly higher in TSPAN7 knockdown neurons than scrambled siRNA14 neurons, at all times, in untreated conditions (left panels) (0 min: 1.26 ± 0.05 versus 1.00 ± 0.09, ^∗p = 0.04; 5 min 1.74 ± 0.04 versus 1.46 ± 0.04, ^∗∗∗p < 0.001; 10 min: 1.27 ± 0.07 versus 1.08 ± 0.05, ^∗p = 0.04; Tukey after ANOVA; values normalized to scrambled siRNA14 neuron levels at time 0). Dynasore treatment (80 μM, 30 min; right panels) abolishes all differences in the internalization index between TSPAN7 knockdown and scrambled siRNA14 neurons (0 min: 1.00 ± 0.17 versus 1.06 ± 0.17, p = 0.81; 5 min: 1.04 ± 0.28 versus 1.07 ± 0.18, p = 0.93; 10 min: 1.04 ± 0.09 versus 0.97 ± 0.09, p = 0.64; Tukey after ANOVA). AMPARs are internalized more efficiently in the absence of TSPAN7.

Figure 8

Effects of TSPAN7 Knockdown and Overexpression on GluA2 Trafficking Are Dependent on PICK1 (A) Representative images of GluA2 internalization experiments on transfected neurons 5 min after completion of 10 min incubation with anti-GluA2. Neurons were transfected with EGFP (control), siRNA14, siRNA14 plus wild-type TSPAN7 resistant to siRNA14 (rescue WT), siRNA14 plus TSPAN7ΔC resistant to siRNA14 (rescue ΔC), TSPAN7, or TSPAN7ΔC. The extent of GluA2 internalization is shown in the histograms on the right. TSPAN7 knockdown (siRNA14) increases the internalization index, rescue WT rescues the effect of TSPAN7 knockdown by reducing the internalization index close to control levels, whereas rescue ΔC does not. Overexpression of TSPAN7, but not of TSPAN7ΔC, reduces the GluA2 internalization index compared to control. (B–D) Representative images of GluA2 internalization experiments performed as in (A). Neurons were transfected with EGFP (control), siPICK1, siPICK1 plus siRNA14 (B), PICK1 or PICK1 plus TSPAN7 (C). Extent of GluA2 internalization (internalization index) is shown in histograms on the right (D). PICK1 knockdown (siPICK1) decreases the GluA2 internalization index and, when cotransfected with siRNA14, prevents the siRNA14-dependent increase in GluA2 internalization. PICK1 overexpression increases the GluA2 internalization index and, when cotransfected with TSPAN7, prevents TSPAN7-induced reduction in GluA2 internalization. Scale bar represents 10 μm.