A small TAT-TrkB peptide prevents BDNF receptor cleavage and restores synaptic physiology in Alzheimer's disease

Abstract

In Alzheimer's disease (AD), amyloid β (Aβ)-triggered cleavage of TrkB-FL impairs brain-derived neurotrophic factor (BDNF) signaling, thereby compromising neuronal survival, differentiation, and synaptic transmission and plasticity. Using cerebrospinal fluid and postmortem human brain samples, we show that TrkB-FL cleavage occurs from the early stages of the disease and increases as a function of pathology severity. To explore the therapeutic potential of this disease mechanism, we designed small TAT-fused peptides and screened their ability to prevent TrkB-FL receptor cleavage. Among these, a TAT-TrkB peptide with a lysine-lysine linker prevented TrkB-FL cleavage both in vitro and in vivo and rescued synaptic deficits induced by oligomeric Aβ in hippocampal slices. Furthermore, this TAT-TrkB peptide improved the cognitive performance, ameliorated synaptic plasticity deficits and prevented Tau pathology progression in vivo in the 5XFAD mouse model of AD. No evidence of liver or kidney toxicity was found. We provide proof-of-concept evidence for the efficacy and safety of this therapeutic strategy and anticipate that this TAT-TrkB peptide has the potential to be a disease-modifying drug that can prevent and/or reverse cognitive deficits in patients with AD.

Keywords: Alzheimer’s disease; BDNF; TAT peptide; TAT-TrkB; TrkB receptor; amyloid β; drug screening; hippocampal plasticity; learning; memory; protein cleavage.

Graphical abstract

Figure 1

Cleavage of TrkB-FL is a hallmark of AD progression, and it is correlated with the levels of Aβ_1-42, even in early stages of the disease (A) Heatmaps of pairwise correlation p values between parameters measured in the CSF of patients with MCI due to AD (n = 47) and MCI controls (n = 23). Correlations were evaluated between parameters related to BDNF signaling (BDNF, TrkB-FL, TrkB-ICD, and the TrkB-ICD:TrkB-FL ratio), molecular biomarkers of AD (Aβ_1-42, p-Tau, and t-Tau), and age. Only the correlation between the levels of TrkB-ICD and Aβ_1-42 in patients with MCI due to AD was found to be statistically significant (R = −0.46, p = 1.25 × 10⁻³, Pearson’s correlation coefficient; a Bonferroni-adjusted α = 3.13 × 10⁻³ was considered to account for multiple comparisons). Univariate analyses of each parameter are shown in Figures S1A–S1G. (B and C) Scatterplots depicting values of individual patients for two pairs of parameters: Aβ_1-42 versus TrkB-ICD levels (B) and p-Tau versus TrkB-ICD levels (C). Linear regressions with 95% confidence ribbons are overlaid. (D) Classification of AD pathology according to Braak staging (top). Representative western blot of human temporal cortical samples probed with anti-TrkB C-terminal and anti-β-actin antibodies (bottom). (E and F) Quantification of TrkB-FL (E) and TrkB-ICD (F) levels in human patients with AD, according to the severity of AD-related pathology. Data are normalized for group Braak stages 0–II (Braak stages 0–II: n = 11; III–IV: n = 7; V–VI: n = 8). Individual replicates are shown with group medians (solid lines) overlaid (E: H = 9.26, p = 9.78 × 10⁻³; F: H = 11.03, p = 4.03 × 10⁻³; Kruskal-Wallis test, followed by Dunn’s test for multiple comparisons). (G) Microarray quantification of TrkB-FL transcript (NTRK2 gene) levels normalized to the SV2B gene, according to disease stage. Data are normalized for group Braak stages 0–II (Braak stage 0–II: n = 27; III–IV: n = 11; V–VI: n = 17). Individual replicates are shown with group medians (solid lines) overlaid (H = 5.57, p = 0.06; Kruskal-Wallis test, followed by Dunn’s test for multiple comparisons). See also Figure S1 and Tables S1 and S2.

Figure 2

The cleavage product TrkB-ICD causes dendritic spine loss and neuronal hyperexcitability and changes the expression levels of synapse-related genes (A) Schematic diagrams of the LV payload. (B–E) Representative microscopy images of primary cultured neurons (DIV14) transduced with either LV-CaMKIIα-EGFP (LV-GFP) (B) or LV-CaMKIIα-TrkB-ICD-IRES-ZsGreen1 (LV-ICD) (D). Yellow arrowheads label cells co-expressing MAP2 (a neuronal marker) and the LV payload, while blue arrowheads indicate untransduced neurons. Quantifications of transduction efficiency and neuronal selectively are shown in (C) and (E) for LV-GFP and LV-ICD, respectively (n = 3 independent cultures per condition). (F) Representative western blot of transduced and untransduced primary cultured neurons (DIV10) probed with anti-TrkB C-terminal and anti-GAPDH antibodies. (G) Quantification of TrkB-FL (top) and TrkB-ICD (bottom) levels in transduced and untransduced primary cultured neurons (DIV10). Individual replicates (circles) are shown with mean ± SEM (bars and error bars) overlaid (n = 3–4 replicates from n = 3 independent cultures per condition; top: F(2,8) = 1.23, p = 0.34; bottom: F(2,8) = 39.51, p = 3.85 × 10⁻³; one-way ANOVA, followed by Tukey’s test for multiple comparisons). (H–J) Representative microscopy images of primary cultured neurons (DIV14) transduced with either LV-GFP (H) or LV-ICD (I) and stained with phalloidin (F-actin, dendritic spines) and a MAP2 antibody (a-MAP2, neuronal marker). Both transduced (yellow arrowheads) and neighboring untransduced neurons (blue arrowheads) were quantified. Higher-magnification insets of dendritic spines are shown in (J). (K) Quantification of dendritic protrusions in primary culture neurons (DIV14), both transduced and untransduced. Individual replicates (circles) are shown with boxplots overlaid (n = 13–15 cells, from n = 3 independent cultures per condition; two-way ANOVA followed by Šídák’s test for multiple comparisons; main effects, interaction, and pairwise comparisons are shown). (L–N) Resting membrane potential (L), frequency (M), and amplitude (N) of mEPSC events recorded from primary cultured neurons transduced with either LV-GFP or LV-ICD. Individual replicates (circles) are shown with mean ± SEM (bars and error bars) overlaid (n = 7–8 cells, from n = 3 independent cultures per condition; L: t = 2.93, df = 12, p = 1.26 × 10⁻²; M: t = 2.60, df = 11, p = 2.46 × 10⁻²; N: t = 0.74, df = 11, p = 0.47; unpaired t tests). (O) Representative mEPSC tracings for each condition, as indicated. (P) For each GO term cataloged, the differences in the NESs for the two statistical comparisons (“TrkB-ICD overexpression” and “LV transduction”) are plotted against the respective multiple-correction-adjusted p values for the TrkB-ICD overexpression contrast. Solid lines are hyperbolic thresholds (with asymptotes at an NES differential of 0.5 and an FDR of 0.05). (Q) The GO terms shortlisted in (P) are shown, with indication of their catalog ID, NES, and Benjamini-Hochberg corrected p value for the TrkB-ICD overexpression contrast. (R) Differential gene expression results represented as a volcano plot for the TrkB-ICD overexpression contrast. The horizontal dashed line represents the significance threshold of p < 0.05 after Benjamini-Hochberg correction for multiple testing. Each dot represents a gene; significantly upregulated genes due to TrkB-ICD overexpression are labeled in red. See also Figure S2.

Figure 3

A TAT-TrkB peptide with a lysine-lysine linker attenuates TrkB-FL cleavage, has high affinity for cell membranes, and is readily translocated to the cytosol and across an in vitro model of the BEB (A) Schematic depicting the TrkB-FL aminoacid sequence flanking calpains’ cleavage site, and the design of the TAT-TrkB peptide. The predicted 3D structure of the peptide is shown (bottom right); pink triangles indicate the position of the cleavage site. (B and C) Western blots of homogenates from adult rat brain cortex (B) and primary neuronal culture (DIV8) (C). Calpain-mediated cleavage of TrkB-FL was prevented by the TAT-TrkB peptide (20 and 50 μM). (D and E) Representative experiments of TAT-TrkB partition coefficient determination in simpler DMPC:Chol (85:15) (D) and more complex DMPC:Chol:SM (75:15:10) (E) LUVs (n = 3 independent experiments). The fitted curves were obtained by applying Equation 1 to the second derivative values of the absorbance spectra at 322 nm (see materials and methods and Figure S4 for more details). (F and G) Representative experiments of T_m determination. Normalized count rate of DMPC:Chol (F) and DMPC:Chol:SM (G) LUVs (4,000 μM) with and without TAT-TrkB (4 mg/mL) are plotted as a function of temperature at pH 7.4 (n = 3 independent experiments, F: T_m = 18.4°C ± 0.5°C and T_m = 17.4°C ± 0.7°C in the absence and presence of TAT-TrkB, respectively; U = 1, p = 0.20; G: T_m = 17.7°C ± 0.4°C and T_m = 17.4°C ± 0.7°C in the absence and presence of TAT-TrkB, respectively; U = 3, p = 0.70; Mann-Whitney U tests). Fitted lines were plotted according to Equation 2 (see materials and methods). (H) Representative epifluorescence image of live primary cultured neurons (DIV14) after a 2-h incubation with TAT-TrkB-6-FAM. (I) Western blot of primary cultured neurons (DIV8). Aβ-driven cleavage of TrkB-FL was attenuated by the TAT-TrkB peptide (5 and 20 μM) incubated either 1 h or 4 h before the start of the 24 h incubation with Aβ_25-35 (25 μM). (J) Translocation kinetics of TAT-TrkB peptide through an in vitro model of the BEB (n = 3 independent experiments for each time point; each data point is the average of triplicates). Individual data points (black circles) are plotted, with averages (pink bars) and a hyperbolic regression with a 95% confidence interval ribbon overlaid (for schematic and BEB integrity assays, see Figure S5). See also Figures S3–S5.

Figure 4

The TAT-TrkB peptide prevents synaptotoxic effects of Aβ_1-42 oligomers (A) Time course of averaged normalized changes in fEPSP slope after delivery (0, 60, and 120 min) of three weak θ-burst (4 × 4), separated by 1-h intervals, to hippocampal slices exposed to vehicle, Aβ oligomers (200 nM), or both Aβ oligomers (200 nM) and TAT-TrkB (20 μM). Data are expressed as mean ± SEM (n = 7–8 slices, from n = 7–8 rats per condition). (B and C) Quantifications of the LTP1 magnitude (B) and SI (C) from the recordings shown in (A), and from hippocampal slices exposed to TAT-TrkB (20 μM) alone (n = 8 slices, from n = 8 rats). Individual replicates (circles) are shown with mean ± SEM (bars and error bars) overlaid (B: H = 11.52, p = 9.22 × 10⁻³; C: H = 15.77, p = 1.26 × 10⁻³; Kruskal-Wallis test, followed by Dunn’s test for multiple comparisons). (D–F) Time courses of averaged normalized changes in fEPSP slope after delivery (0 min) of a weak θ-burst (4 × 4) in the absence (colored open circles) or presence (colored filled circles) of BDNF (30 ng/mL) to hippocampal slices exposed to vehicle (D), Aβ oligomers (200 nM, E) or Aβ oligomers (200 nM) and TAT (20 μM, F). Data are expressed as mean ± SEM (D: n = 13 [no BDNF], n = 6 [with BDNF]; E: n = 8 [no BDNF], n = 7 [with BDNF]; F: n = 15 [no BDNF], n = 13 [with BDNF] slices, from different rats). (G) Quantification of LTP magnitudes from the recordings shown in (D)–(F) and in Figure S10F. BDNF significantly increased LTP magnitude after incubation with vehicle or TAT-TrkB (20 μM) alone (D: U = 7, p = 3.23 × 10⁻³; E: U = 21, p = 0.44; F: U = 94, p = 0.89; Mann-Whitney U tests). (H) Top: time course of averaged normalized changes in fractional release of [³H]glutamate from hippocampal synaptosomes, evoked by two 15-mM K⁺ stimuli (at 5–7 min and 29–31 min; shaded areas), in the absence (colored open circles) or presence (colored filled circles) of BDNF (30 ng/mL). Bottom: same time course as above, where the internal control curves were subtracted from BDNF curves. This procedure makes the incremental effect of BDNF on glutamate release during the second K⁺-induced depolarization more evident. (I) Time courses of averaged control-subtracted changes in fractional release of [³H]glutamate from hippocampal synaptosomes induced by BDNF, after incubation with vehicle (n = 12 experiments, 24 rats), Aβ oligomers (200 nM) (n = 5 experiments, 10 rats), or both Aβ oligomers (200 nM) and TAT-TrkB (20 μM) (n = 3 experiments, 6 rats). (J) Area of control-subtracted S2 peaks from the recordings in (I) and after incubation with TAT-TrkB (20 μM) alone (n = 5 experiments, 10 rats; see Figure S10L). Individual replicates (circles) are shown with mean ± SEM (bars and error bars) overlaid (U = 9.25, p = 2.61 × 10⁻²; Kruskal-Wallis test, followed by Dunn’s test for multiple comparisons). See also Figures S6–S10.

Figure 5

In vivo treatment of 5XFAD mice with TAT-TrkB attenuates TrkB-FL cleavage and prevents hippocampal-dependent memory deficits, LTP impairment, and loss of dendritic spines (A) Schematic representation of the experimental groups and of the time line for treatments and experiments. (B) Representative western blot of hippocampal homogenates probed with anti-TrkB and anti-GAPDH antibodies. (C and D) Quantification of TrkB-FL (C) and TrkB-ICD (D) levels in hippocampal homogenates of 5XFAD and WT mice, treated with either vehicle (Veh; 0.9% NaCl) or TAT-TrkB (TAT; 25 mg/kg). Data were normalized for the vehicle-treated WT group (n = 9–15 mice per condition). Individual replicates are shown with mean ± SEM (bars and error bars) overlaid (two-way ANOVA, followed by Šídák’s test for multiple comparisons; main effects, interaction, and pairwise comparisons are shown). (E–H) Hippocampal-dependent memory performance was assessed by the MWM test, in which learning/acquisition (E and F) and memory retention (G and H) were evaluated. The latency to reach the platform is plotted as a function of the training day (F) during the learning phase (n = 18–23 mice per condition; data are expressed as mean ± SEM; repeated-measures three-way ANOVA; main effects and interactions are shown). Representative traces of the mice traveling during the 4th training day are shown in (E). The time spent on the target quadrant during the probe test is shown for each experimental group in (G) (n = 18–23 mice, per condition; individual replicates are shown with mean ± SEM [bars and error bars] overlaid; two-way ANOVA, followed by Šídák’s test for multiple comparisons; main effects, interaction, and pairwise comparisons are shown). Representative traces of the mice traveling during the probe test are shown in (H). (I and J) Quantification of the time spent in the open arms (I) and of the total number of entries in both open and closed arms (J) of the EPM (n = 18–23 mice, per condition). Individual replicates are shown with mean ± SEM (bars and error bars) overlaid (two-way ANOVA, followed by Šídák’s test for multiple comparisons; main effects, interaction, and pairwise comparisons are shown). (K) Time course of averaged normalized changes in fEPSP slope after delivery (0 min) of a weak θ-burst (4 × 4) to hippocampal slices from mice of each experimental group (n = 5–6 mice, per condition). (L) Quantification of LTP magnitudes from the recordings shown in (K). Individual replicates are shown with mean ± SEM (bars and error bars) overlaid (two-way ANOVA, followed by Šídák’s test for multiple comparisons; main effects, interaction, and pairwise comparisons are shown). (M–P) Quantification of PSD-95 (N) and Syn-1 (P) levels in hippocampal homogenates of 5XFAD and WT mice, treated with either vehicle (0.9% NaCl) or TAT-TrkB (25 mg/kg). Data were normalized for the vehicle-treated WT group (N: n = 7–11 mice per condition; P: n = 5–6 mice per condition). Individual replicates are shown with mean ± SEM (bars and error bars) overlaid (two-way ANOVA, followed by Šídák’s test for multiple comparisons; main effects, interaction, and pairwise comparisons are shown). Representative immunoblots are shown in (M) and (O), respectively. (Q–U) Quantification of F-actin (dendritic) protrusion density in the stratum radiatum of CA1 hippocampal area (U: n = 18 imaging fields, from n = 3 mice per condition). Individual quantifications are shown with mean ± SEM (bars and error bars) overlaid (mixed-effect linear regression, followed by Tukey’s test for multiple comparisons; main effects, interaction, and pairwise comparisons are shown). Representative microscopy images are shown in (Q)–(T). See also Figure S11.

Figure 6

In vivo treatment of 5XFAD mice with TAT-TrkB reduced Aβ aggregation and prevented Tau hyperphosphorylation (A–D) Representative microscopy images of Aβ plaques in the hippocampus and cortex of 6-month-old 5XFAD mice (C and D). No plaques were observed in WT littermates (A and B). Insets show the dentate gyrus in higher magnification. (E and F) Quantification of Aβ plaque density (E) and area (F) in the hippocampus of 5XFAD mice treated with either vehicle (Veh; 0.9% NaCl) or TAT-TrkB (TAT; 25 mg/kg). In (E), individual quantifications are shown with boxplots overlaid (n = 23–24 imaging fields, n = 3 mice per condition; U = 268, p = 0.87, Mann-Whitney U test). In (F), individual plaques are shown with violin plots and boxplots overlaid (n = 289–309 plaques, n = 3 mice per condition; U = 40,790, p = 2.35 × 10⁻⁴, Mann-Whitney U test). (G and H) Quantification of Aβ plaque density (G) and area (H) in the cortex of 5XFAD mice treated with either vehicle (0.9% NaCl) or TAT-TrkB (25 mg/kg). In (G), individual quantifications are shown with boxplots overlaid (n = 23–24 imaging fields, n = 3 mice per condition; U = 314, p = 0.42, Mann-Whitney U test). In (H), individual plaques are shown with violin plots and boxplots overlaid (n = 558–657 plaques, n = 3 mice per condition; U = 178,169, p = 0.11, Mann-Whitney U test). Representative microscopy images are shown in Figure S12. (I and J) Quantification of (RIPA-)soluble Aβ levels in hippocampal homogenates of 5XFAD and WT mice, treated with either vehicle (0.9% NaCl) or TAT-TrkB (25 mg/kg) (J). Data were normalized for the vehicle-treated WT group (n = 13–15 mice per condition). Individual replicates are shown with mean ± SEM (bars and error bars) overlaid (two-way ANOVA, followed by Šídák’s test for multiple comparisons; main effects, interaction, and pairwise comparisons are shown). A representative immunoblot is shown in (I). (K–N) Representative microscopy images of p-Tau⁺ cells in the dentate gyrus of the hippocampus of 6-month-old 5XFAD mice and WT littermates. (O and P) Quantification of average normalized p-Tau intensity (O) and of the number of pTau⁺ cells (P) in the hippocampus of 5XFAD and WT mice, treated with either vehicle (0.9% NaCl) or TAT-TrkB (25 mg/kg). In (O), average normalized p-Tau intensities of individual p-Tau⁺ cells are shown with boxplots overlaid (n = 70–88 cells, n = 3 mice per condition). In (P), numbers of p-Tau⁺ cells per imaging field are depicted with boxplots overlaid (n = 22–26 imaging fields, n = 3 mice per condition). Data were analyzed with mixed-effects linear regressions, followed by Tukey’s test for multiple comparisons; main effects, interaction, and pairwise comparisons are shown. (Q and R) Quantification of p-Tau levels in hippocampal homogenates of 5XFAD and WT mice, treated with either vehicle (0.9% NaCl) or TAT-TrkB (25 mg/kg) (R). Data were normalized for the vehicle-treated WT group (n = 5–6 mice per condition). Individual replicates are shown with mean ± SEM (bars and error bars) overlaid (two-way ANOVA, followed by Šídák’s test for multiple comparisons; main effects, interaction, and pairwise comparisons are shown). A representative immunoblot is shown in (Q). See also Figure S12.

Figure 7

In vivo treatment with TAT-TrkB does not cause muscle, kidney, or liver damage (A and B) Body weight is plotted as a function of experiment duration (A) (data are expressed as mean ± SEM). Changes in body weight from the start to the end of treatment are depicted in (B) (n = 6–7 mice per condition; individual replicates are shown with boxplots overlaid). (C–E) Quantification of creatine kinase (C), creatinine (D), and urea (E) in the serum of 5XFAD and WT mice, treated with either vehicle (Veh; 0.9% NaCl) or TAT-TrkB (TAT; 25 mg/kg) (n = 3–4 mice per condition; individual replicates are shown with mean ± SEM [bars and error bars] overlaid). Horizontal dashed lines indicate the range of reference values. (F–I) Representative microphotographs of H&E-stained sections of kidney. No histological changes were detected. (J and K) Quantification of AST (J) and ALT (K) in the serum of 5XFAD and WT mice, treated with either vehicle (0.9% NaCl) or TAT-TrkB (25 mg/kg) (n = 3–4 mice per condition; individual replicates are shown with mean ± SEM [bars and error bars] overlaid). Horizontal dashed lines indicate the range of reference values. (L–O) Representative microphotographs of H&E-stained sections of liver. No histological changes were detected. Analyses in (B)–(E) and (J) and (K) are two-way ANOVAs; main effects and interaction are shown.