WNT Activates the AAK1 Kinase to Promote Clathrin-Mediated Endocytosis of LRP6 and Establish a Negative Feedback Loop

Abstract

β-Catenin-dependent WNT signal transduction governs development, tissue homeostasis, and a vast array of human diseases. Signal propagation through a WNT-Frizzled/LRP receptor complex requires proteins necessary for clathrin-mediated endocytosis (CME). Paradoxically, CME also negatively regulates WNT signaling through internalization and degradation of the receptor complex. Here, using a gain-of-function screen of the human kinome, we report that the AP2 associated kinase 1 (AAK1), a known CME enhancer, inhibits WNT signaling. Reciprocally, AAK1 genetic silencing or its pharmacological inhibition using a potent and selective inhibitor activates WNT signaling. Mechanistically, we show that AAK1 promotes clearance of LRP6 from the plasma membrane to suppress the WNT pathway. Time-course experiments support a transcription-uncoupled, WNT-driven negative feedback loop; prolonged WNT treatment drives AAK1-dependent phosphorylation of AP2M1, clathrin-coated pit maturation, and endocytosis of LRP6. We propose that, following WNT receptor activation, increased AAK1 function and CME limits WNT signaling longevity.

Keywords: AAK1; AP2M1; LRP6; WNT signaling; clathrin; endocytosis; feedback loop; gain-of-function screen; kinase.

Graphical abstract

Figure 1

Gain-of-Function Kinome Screen Reveals AAK1 as a Repressor of WNT Signaling (A) HEK293T-B/R cells were transfected with the indicated construct for 24 hr. Cells were then treated for 16 hr with WNT3A or Lcell CM. Bars represent average Firefly/Ren relative fluorescence units (RFU) from three technical replicates. (B and C) Luciferase assay of HT1080 (B) or RKO (C) stable B/R cells transfected with either control or AAK1 siRNA for 56 hr. Cells were then treated with either Lcell or WNT3A CM for 16 hr. Bars represent average Firefly/Ren RFU from three technical replicates. Western blot analysis illustrates knockdown efficiency of two independent AAK1 siRNAs. (D) IncuCyte imaging of HT1080 cells stably expressing a BAR-mCherry fluorescent reporter transiently transfected with indicated siRNA construct. WNT3A CM was added at 18 hr, then cells were imaged for 50 hr post-transfection. Graph represents data points averaged across four technical replicates. (E) Live-cell imaging of HT1080 cells stably expressing a BAR-mCherry fluorescent reporter transiently transfected with the indicated expression construct, AAK1, or FLAG control. WNT3A CM was added at 8 hr, and cells were monitored for an additional 56 hr. Data represent the average of four technical replicates. (F and G) qPCR analysis of AXIN2 and NKD1 in HEK293T (F) or HT1080 (G) cells 72 hr after transfection with the indicated siRNA. Cells were treated with WNT3A CM for 6 hr prior to harvest. Bars represent average glyceraldehyde-3-phosphate dehydrogenase (GAPDH)-normalized gene expression across three technical replicates. (H) qPCR analysis of AXIN2 (left) and NKD1 (right) in HEK293T cells transfected with overexpression construct for 24 hr, then treated with WNT3A CM for 6 hr prior to harvest. Bars represent average GAPDH-normalized gene expression across three technical replicates. (I) Luciferase assay of HEK293T cells transfected with indicated pathway-specific Firefly luciferase reporter constructs and expression constructs prior to a 16 hr treatment with recombinant human (rh) WNT3A (200ng/mL), rhTNFα (200ng/mL), or rhTGFβ1 (200ng/mL). Bars represent average Firefly/Ren RFU from three technical replicates. ^∗p < 0.05, ^∗∗p < 0.005, and ^∗∗∗p < 0.0005. All data are representative of biological triplicates, unless otherwise noted. Error bars represent SE. For complete statistics, see STAR Methods. See also Table S1.

Figure 2

SGC-AAK1-1 Is a Potent and Specific Inhibitor of AAK1 and BMP2K (A) Chemical structures of the AAK1/BMP2K chemical probe SGC-AAK-1, the related AAK1/BMP2K inhibitor 25A, and the negative control compound 34A that has a similar chemical structure but does not inhibit AAK1 or BMP2K. (B) SGC-AAK-1 selectively bound AAK1 and BMP2K with a more than 30-fold difference in K_i in a TR-FRET binding displacement assay over the related kinases GAK and STK16. Sixteen-point dose-response curves were measured in duplicate. (C) SGC-AAK1-1 has good selectivity over the human kinome. SGC-AAK1-1 was used at 1 μM concentration and tested against over 400 wild-type human kinases. K_D determination was completed for all kinases that exhibited >85% inhibition and/or were observed as common off-targets for this structural class. Kinases for which SGC-AAK1-1 has a K_D < 100 nM are marked with a large red circle and 100 nM < K_D < 1.0 μM with a smaller red circle. (D) ITC confirmed a K_D of 120 nM for binding of SGC-AAK1-1 to AAK1. (E) A co-crystal structure of BMP2K (2.41 Å) bound to 25A revealed the binding mode of the inhibitor in the ATP site. This structure revealed five key hydrogen bonds and revealed the orientation of the central hydrophobic portion of the inhibitor with respect to the binding pocket. (F and G) NanoBRET cellular target engagement assays in HEK293T cells showed that 25A (F) and SGC-AAK1-1 (G) both entered cells and directly bound to AAK1 and displaced a fluorescent tracer molecule from the ATP site. Cells were treated with serially diluted inhibitors in the presence of four different concentrations of a NanoBRET tracer molecule, and IC₅₀ values were calculated from a linear interpolation of the IC₅₀ values at each tracer concentration to obtain the predicted IC₅₀ in the absence of tracer. Measurements were made with three technical replicates. (H) Binding of SGC-AAK1-1 to BMP2K in cells was weaker than to AAK1 with IC₅₀ > 1 μM. Displacement of the same fluorescent tracer molecule required increased concentrations of SGC-AAK1-1, revealing an elevated IC₅₀ value for binding to BMP2K versus AAK1 in the cellular target engagement assay. (I) SGC-AAK1-1 inhibited the phosphorylation of the AP2M1 subunit at Thr156 in HEK293T cells in a concentration-dependent manner or with 10 μM 34A. HEK293T cells were treated with inhibitors for 2 hr before western blot analysis. Error bars represent SE. See also Figure S1 and Tables S2, S3, S4, and S5.

Figure 3

AAK1 Inhibition Potentiates WNT Signaling and Stabilizes β-Catenin (A) Representative western blot of HT1080 cells treated with indicated compounds (2.5 μM) for 1 hr (left). Box-and-whisker plot represents quantification of four biological replicate experiments using LiCOR software (right). (B) BAR luciferase assay of HT1080-B/R stable cells treated with WNT3A CM and the indicated dose of compound for 16 hr. Data are averaged over three technical replicates. (C) qPCR analysis of AXIN2 or NKD1 from HT1080 cells treated with indicated compound (1.25 μM) for 12 hr in the presence of Lcell or WNT CM and normalized to GAPDH. Graph represents analysis averaged across three technical replicates. (D and E) Western blot analysis of RKO cells following a 15 min WNT3A pulse, then treated with 1.25 μM of either 34A or SGC-AAK1-1 for the indicated time. Dashed line (D) represents cropped blot for removal of redundant, untreated sample. ^∗p < 0.05, ^∗∗p < 0.005, and ^∗∗∗p < 0.0005. All data are representative of biological triplicates, unless otherwise noted. Error bars represent SE. For complete statistics, see STAR Methods.

Figure 4

AAK1 Represses β-Catenin-Mediated WNT Signaling through Regulation of Endocytosis (A) HT1080-B/R cells were transfected with indicated siRNA for 56 hr and then treated with WNT3A or CHIR99021 (1 μM) for 16 hr (left). HT1080-B/R cells were transfected with the indicated overexpression DNA construct and allowed to recover for 12 hr. Cells were then treated with WNT3A CM or CHIR99021 (1 μM) for 16 hr (right). Data are averaged over three technical replicates. (B) BAR luciferase assay from HT1080 cells transiently transfected with BAR-firefly luciferase, TK-Ren, and the indicated expression constructs. Twelve hours post-transfection, cells were treated for 16 hr with Lcell or WNT3A CM and then assayed for BAR activity. Western blot of FLAG-AAK1 and FLAG-GFP is shown to illustrate expression of AAK1 constructs. Presented data are averaged across three technical replicates. (C) Luciferase assay in HT1080-B/R cells transfected with indicated siRNAs for 50 hr, then co-transfected with indicated expression constructs for 12 hr. Cells were then treated for 16 hr with either Lcell or WNT3A CM. Data are averaged over three technical replicates. Right: western blot of clathrin, indicating knockdown efficiency. (D) Luciferase assay of HT1080-B/R cells transfected with indicated siRNA for 56 hr, then treated as indicated for 16 hr. Data are averaged over three technical replicates. (E) Luciferase assay of HEK293T-B/R cells transfected with indicated constructs for 12 hr, then treated with either Lcell or WNT3A CM for 18 hr. β-Catenin pm (point mutant) is a construct with the GSK3β target residues (S33/S37/T41) on β-catenin mutated. Western blot analysis of FLAG and GFP shown to confirm expression of overexpression constructs. Data are averaged over three technical replicates. ^∗p < 0.05, ^∗∗p < 0.005, and ^∗∗∗p < 0.0005. All data are representative of biological triplicates, unless otherwise noted. Error bars represent SE. For complete statistics, see STAR Methods. See also Figure S2.

Figure 5

AAK1 Downregulates LRP6 Cell Surface Levels (A and B) Western blot analysis of HEK293T cells transfected with an AAK1 overexpression construct for 24 hr (A) or AAK1 siRNA for 72 hr (B) prior to membrane fractionation using a surface biotinylation assay. Data represent average surface staining of four biological replicate experiments quantified by LiCOR software. (C) LiCOR quantification of western blot analysis of LRP6 levels with AAK1 knockdown. The box-and-whisker plot represents total LRP6 expression normalized to β-tubulin (loading control) from six biological replicates quantified by LiCOR software. (D and E) LRP6 surface staining as detected by flow cytometry analysis in HEK293T cells transfected with an AAK1 overexpression construct for 24 hr (D) or an AAK1 siRNA construct for 72 hr (E). All experiments were quantified using FlowJo software. Experiment was performed in four biological replicates. (F) LRP6 surface levels as detected by flow cytometry analysis in HEK293T cells treated as indicated for 8 hr. The box-and-whisker plot represents LRP6 surface levels normalized to BSA control for biological triplicates. Raw values for each experiment are shown in Table S6. (G) HEK293T cells were transfected with FLAG-AAK1 or FLAG-GFP overexpression constructs. Twenty-four hours post-transfection, cells were treated for 6 hr with the autophagy inhibitor bafilomycin A1 (100 nm) or DMSO control prior to cell harvest, then analyzed using western blot for endogenous LRP6 and AAK1 levels. The box-and-whisker plot represents LRP6 expression normalized to β-tubulin from four biological replicates quantified by LiCOR software. ^∗p < 0.05 and ^∗∗p < 0.005. All data are representative of biological triplicates, unless otherwise noted. For complete statistics, see STAR Methods. See also Figure S3 and Table S6.

Figure 6

WNT Treatment Induces AAK1-, PIP2-Dependent Phosphorylation of AP2M1 (A) Western blot analysis of HT1080 cells treated with WNT3A CM for indicated times. (B) Western blot analysis of HT1080 cells transfected with either AAK1 or control siRNA for 72 hr. Cells were then treated with rhWNT3A (200 ng/mL) for 15 min, then the medium was changed for fresh, complete DMEM, and cells were incubated for 9 hr. (C) Western blot analysis of pAP2M1 levels in HEK293T cells treated with either rhWNT3A (3A) or rhWNT5A (5A) (200 ng/mL) for 9 hr. The box-and-whisker plot represents pAP2M1/AP2M1 expression from four biological replicates quantified by LiCOR software. (D) Western blot analysis of HT1080 cells treated with either CHIR99021 compound (1 μM) or rhWNT3A (200 ng/mL) for the indicated time. (E) HEK293T cells were transfected with FLAG-dnTCF7L2 for 24 hr. Cells were then treated with WNT3A for 15 min, then the medium was changed for fresh, complete DMEM, incubated for 9 hr, and analyzed using western blot. (F) Western blot analysis of HEK293T WT, DVL TKO clone #4, DVL TKO clone #6, and LRP5/6 DKO cells treated with WNT3A CM for 9 hr. (G) Western blot analysis of HT1080 cells treated with WNT3A CM for 15 mins, then the medium was changed for complete DMEM containing either carbachol (50 μM) or neomycin (100 μM). Cells were incubated for 9 hr prior to cell harvest. (H) Western blot analysis of HT1080 cells transfected with siRNA against PI4K2α for 72 hr and pulsed with WNT3A CM for 15 mins and then incubated for 9 hr. All panels are representative of biological triplicates, unless otherwise noted. For complete statistics, see STAR Methods. See also Figure S4.

Figure 7

Proposed Model for AAK1-Dependent and CME of LRP6 (A) Within minutes of WNT ligand exposure, FZD and LRP5/6 co-complex, allowing formation of the LRP6 signalosome in nascent clathrin-coated pits. Signalosome formation requires DVL domain swapping and polymerization, as well as PIP2 and AP2 recruitment. This promotes accumulation of β-catenin and activation of the WNT signaling pathway. (B) Prolonged exposure to WNT ligand drives AAK1-dependent AP2M1 phosphorylation, CME, and ultimately removal of the WNT receptors from the plasma membrane.