The Alzheimer's disease-linked protease BACE2 cleaves VEGFR3 and modulates its signaling

Abstract

The β-secretase β-site APP cleaving enzyme (BACE1) is a central drug target for Alzheimer's disease. Clinically tested, BACE1-directed inhibitors also block the homologous protease BACE2. Yet little is known about physiological BACE2 substrates and functions in vivo. Here, we identify BACE2 as the protease shedding the lymphangiogenic vascular endothelial growth factor receptor 3 (VEGFR3). Inactivation of BACE2, but not BACE1, inhibited shedding of VEGFR3 from primary human lymphatic endothelial cells (LECs) and reduced release of the shed, soluble VEGFR3 (sVEGFR3) ectodomain into the blood of mice, nonhuman primates, and humans. Functionally, BACE2 inactivation increased full-length VEGFR3 and enhanced VEGFR3 signaling in LECs and also in vivo in zebrafish, where enhanced migration of LECs was observed. Thus, this study identifies BACE2 as a modulator of lymphangiogenic VEGFR3 signaling and demonstrates the utility of sVEGFR3 as a pharmacodynamic plasma marker for BACE2 activity in vivo, a prerequisite for developing BACE1-selective inhibitors for safer prevention of Alzheimer's disease.

Keywords: Aging; Alzheimer disease; Drug therapy.

Figure 1. Identification of VEGFR3 as a BACE2 substrate candidate.

(A) Volcano plot of proteomic analysis of murine plasma from WT and B2KO mice (n = 6). VEGFR3 (FLT4) is highlighted in red. (B) Normalized VEGFR3 LFQ intensities extracted from A. (C) MSD-assay quantifications of sVEGFR3 in the same plasma samples. (D) Immunoblot detection of sVEGFR3 ectodomain in mouse plasma from A, using nonreducing and reducing conditions. (E) Volcano plot of proteomic analysis of murine plasma from an independent B2KO line (n = 9) compared with WT (n =9) and (F) the extracted normalized LFQ values. Volcano plots of the proteomic analyses of Bace1/Bace2 double-knockout (BDKO) mice (n =9) compared with the WT line (n = 9) (G) (corresponding extracted LFQ intensities of sVEGFR3 in F) and B1KO (n = 9) compared with an individual control WT line (n = 9) (H). (I) Normalized LFQ values extracted from H. (J) Localization of identified individual peptides (black dots) on the canonical VEGFR3 sequence. The signal peptide is shown in rose, the ectodomain is indicated in blue, the intracellular domain in green, and the transmembrane domain in yellow. Two sided Student’s t tests with a permutation-based FDR correction (FDR < 0.05; indicated by hyperbolic curves) were used for volcano plots (A, E, G, and H). Proteins with P < 0.05 are shown as red circles. Extracted LFQ quantifications (B, F, and I) of VEGFR3 with significance after FDR correction are labeled with plus signs. All dot plots were normalized on the WT mean and depict mean and SD. MSD-assay data (C) additionally depicts the P value calculated by unpaired t test. ****P < 0.0001.

Figure 2. Cleavage of VEGFR3 by BACE2.

(A) Schematic of VEGFR3 fragments. From left to right: The immature proVEGFR3 (200 kDa) can be cleaved by BACE2, releasing the immature, soluble ectodomain sol proVEGFR3 (130 kDa). The mature protein consists of 2 subunits linked through a disulfide bridge: VEGFR3α (75 kDa) and VEGFR3β (125 kDa). Upon BACE2 cleavage, the VEGFR3β-CTF (70 kDa) and sVEGFR3 (130 kDa) are generated, the latter of which consists of the VEGFR3α (75 kDa) and VEGFR3β-NTF (55 kDa) fragments. (B) Immunoblot detection of VEGFR3 in lysates and media of HEK293 cells transfected with empty control plasmids (Ctrl), Vegfr3, Vegfr3 + Bace2, and Bace2. B2, BACE2. Data show 3 independent experiments. sVEGFR3 is not detectable under reducing conditions. sol proVEGFR3 in the lysates appears at around 100 kDa and derives from BACE2 cleavage of immaturely glycosylated proVEGFR3 early in the secretory pathway upon BACE2 overexpression. (C) Localization and length of identified individual peptides (black dots) on the canonical VEGFR3 sequence. The ectodomain is indicated in blue, the intracellular domain in green, the signal peptide in orange, and the transmembrane domain in yellow. (D) N-terminal juxtamembrane region of VEGFR3 sequence. The identified semispecific peptide after LysN digestion is marked in yellow, the proposed cleavage site after amino acid alanine with 2 vertical lines, and the transmembrane region in gray. (E) Comparison of the fragment ion spectra of the identified C-terminal peptide of the LysN digestion KGC(cam)VN(+1)SSASVA (lower spectrum) to a synthetic peptide with the same sequence (upper spectrum). Identified y-ions are indicated in red, b-ions in blue, and fragment ions with neutral losses in green. Both spectra match with a dot product (73) of 0.93 for the fragment ion intensities.

Figure 3. Endogenous cleavage of VEGFR3 in LECs.

(A) Immunoblot detection after control treatment (–) or upon BACE1 and BACE2 knockdown (+). Lysates were blotted for VEGFR3, BACE1/2, and actin. Conditioned media were blotted for sVEGFR3. (B) Corresponding densitometric quantifications, deriving from VEGFR3β (lysate) and sVEGFR3 (medium). (C) Immunoblots of cells treated with DMSO (–) or 100 nM verubecestat (+). (D) Corresponding densitometric quantifications as in B. Dot plots were normalized on the control mean and depict mean and SD, alongside the calculated P values, calculated by unpaired t test. **P < 0.01; ****P < 0.0001. P values are only indicated where significance was observed. Data are derived from n = 6 biological replicates. Shown are representative data from 3 independent experiments.

Figure 4. BACE2-dependent changes in VEGFR3 signaling in LECs.

(A) Schematic for VEGFR3 signaling. Upon ligand binding, VEGFR3 dimerizes, resulting in intracellular autophosphorylation and activation of the downstream genes FOXC2 and DLL4. V, verubecestat, inhibitor of BACE2. Gene expression levels of (B) DLL4 and FOXC2 and (C) VEGFR3 after the application of DMSO, 100 nM verubecestat (V), and VEGF-C. (D and E) Gene expression levels of DLL4, FOXC2, VEGFR3, BACE1, and BACE2 after BACE knockdown (siB1/siB2) without or with (+V) subsequent verubecestat application. All dot plots were normalized on the control mean and depict mean and SD alongside the P values calculated by unpaired t tests against the DMSO control (B and C) or by 1-way ANOVA (D and E), in both cases followed by Bonferroni’s multiple-comparison test. *P < 0.05; **P < 0.01; ****P < 0.0001. P values are only indicated where significance was observed. In B, 1 data point was excluded from the DLL4 expression/VEGF-C156S data set, since it was identified as an outlier via the ROUT method. Data are derived from n = 6 (B and C) or n = 4 (D and E) biological replicates.

Figure 5. Inhibition of Bace2 in zebrafish embryos leads to enhanced facial lymphatic development.

(A) Laterally imaged zebrafish embryos and representative images comparing the facial lymphatic development between vehicle control– (Veh) and verubecestat-treated (100 μM) (Verub) zebrafish embryos at 3 dpf in a flt4:mCit transgenic background. Asterisks represent the starting points of measurements, while arrowheads depict the end points of the respective measurements. (B) Quantification of LFL vessel length (in μm) in vehicle control– and verubecestat-treated (100 μM) zebrafish embryos at 3 dpf (n = 18 for vehicle, n = 20 for verubecestat-treated group). (C) Representative images comparing the facial lymphatic development between vehicle control– and verubecestat-treated (100 μM) zebrafish embryos at 5 dpf in a fli:nucGFP and lyve1:dsRed transgenic background. Enlarged image of the LFL depicts nuclei of all endothelial cells (green, fli:nucGFP), the cytoplasm of the facial lymphatic vessel (red, lyve1:dsRed) and other lymphatic vessels, and the nuclei present in the facial lymphatic vessel (yellow, colocalization). (D) Corresponding quantification of the number of fli:nucGFP⁺ nuclei present in the LFL of 5 dpf embryos (n = 23 for control; n = 29 for verubecestat-treated group). All images with anterior to the left. All dot plots depict mean and SD, alongside P values calculated by unpaired t test. ***P < 0.001; ****P < 0.0001.

Figure 6. BACE2 inhibition reduces murine plasma sVEGFR3.

Volcano plots of proteomic analysis of murine plasma from (A) compound 89–treated (Cpd89) versus vehicle-treated (veh) mice and (B) LY2811376-treated versus vehicle-treated mice (n = 13, treated; n = 14, veh). VEGFR3 is highlighted in red. (C) Corresponding extracted LFQ intensities of sVEGFR3 and (D) MSD-assay quantifications of sVEGFR3. Plasma sVEGFR3 (E) and plasma sSEZ6L (F) levels in 8–10 B1KO, B2KO, and respective WT mice with (blue) or without (black) 3 days of 50 mg/kg per os twice a day verubecestat dosing. (G) Plasma levels of VEGFR3 and SEZ6L during 7 days of 0.1% dietary verubecestat (average drug intake, 97 mg/kg/d; n = 6 per group, all male, age: 7–10 weeks), respective to untreated control levels. Two-sided Student’s t tests with a permutation-based FDR correction (FDR < 0.05; indicated by hyperbolic curves) were used for volcano plots (A and B). Proteins with P < 0.05 are shown as red circles. (C) Significance after FDR correction is indicated with plus signs. All dot plots were normalized on the control mean and depict the SD alongside the calculated P values, calculated by 1-way (D and G) or 2-way (E and F) ANOVA with Bonferroni’s multiple-comparison test. *P < 0.05; ***P < 0.001; ****P < 0.0001. P values are only indicated where significance could be observed. Number of biological replicates in E and F was 9, except for Bace1-WT + verubecestat (n = 8) and for Bace2WT without verubecestat (n = 10).

Figure 7. Plasma sVEGFR3 is a superior marker for BACE2 activity.

(A) Plasma sVEGFR3 and (B) sSEZ6L levels in mice, fed with diet supplemented with 0 (black), 0.002%, and 0.1% of verubecestat (blue) (n = 4). (C) Photographs of the fur pigmentation of the corresponding mice. (D and E) Relative ELISA quantifications of sVEGFR3 in plasma and CSF Aβ40 as well as Aβ42 levels of (D) verubecestat-treated NHPs (n = 4) before and after treatment and (E) clinical trial participants treated with atabecestat (n = 9) or placebo (n = 4). Human Aβ1-40 (Aβ40) and Aβ1-42 (Aβ42) CSF data for the selected individuals were extracted from a previous publication (41). NHP data were normalized to the predose mean; human data are expressed as postdose/predose ratio for each individual. All dot plots were normalized on the control or predose mean, respectively, and (A and B) depict the SD alongside the calculated P values, calculated by 1-way ANOVA, followed by Bonferroni’s multiple-comparison test (A and B), paired t test (D), or unpaired t test (E). *P < 0.05; **P < 0.01; ****P < 0.0001.