Fig 1.
Identification of AP-2 subunits among MEX3C-associated proteins.
A. Identifying AP-2 α and β2 subunits among MEX3C-associated proteins. MEX3C interacting proteins were pulled down by anti-Flag antibody from cell lysates expressing MEX3C-2-Flag. M: molecular weight marker; kDa: kilodalton. B. AP-2 pulldown was MEX3C-dependent. Since affinity purified proteins were loaded, a loading control was unavailable. The lysates were prepared from equal numbers of HEK293 cells transfected with MEX3C-2-Flag expressing DNA or vector DNA, the purification was done in parallel, and equal volumes of eluents from MEX3C-Flag positive (+) and negative (-) lysates were loaded. C. RNase treatment did not eliminate MEX3C/AP-2 association. RNase A treatment eliminated MEX3C/THOC4 association, demonstrating that the RNase treatment worked. D. Comparisons of relative amounts of protein pulled down by MEX3C-2 with and without RNase treatment. Densitometry of protein bands was analyzed with Image J software. Protein expression was normalized by the respective MEX3C-2-Flag.
Fig 2.
Mapping the MEX3C regions mediating MEX3C/AP-2 interaction.
A. MEX3C-1 and MEX3C-3 pull down endogenous AP-2 subunits. Transiently expressed MEX3C-1 and MEX3C-3 were Flag-tagged. C1 and C3 indicate MEX3C-1-Flag and MEX3C-3-Flag respectively. # indicates the IgG band. B. Deleting the central region or the C-terminal 249AA of MEX3C-1 abolished MEX3C/AP-2 interaction. MEX3C-1 (C1 in Figure), MEX3C-1-ΔKH (1ΔKH) and MEX3C-1-ΔC (1ΔC) were Flag-tagged and detected with anti-Flag antibody. MEX3C-1-ΔC appeared at the same position as the IgG heavy chain. # and * indicate the IgG and the AP-2 μ2 bands, respectively. C. Mutating the YXXΦ-like motifs (lane 1-mY) or the amino acids necessary for RNA binding (lane 1-mKH) did not affect MEX3C/AP-2 interaction. Transiently expressed MEX3C-1, MEX3C-1-mY, and MEX3C-1-mKH were Flag-tagged and immunoprecipitated by anti-Flag antibody to detect co-immunoprecipitated AP-2 subunits. The AP-2 α subunit pulled down by both mutants contained a smaller band in addition to the normally observed large band. D. The ring finger domain of MEX3C is necessary for MEX3C/AP-2 interaction. MEX3C-2 (C2 in Figure), MEX3C-2-KO (all lysine residues were mutated to arginine, lane “2KO”), and MEX3C-2-mRing (the C3HC4 motif in the MEX3C-2 ring finger domain was mutated to A3NC4, lane “2mRing”) were Flag-tagged. # indicates the IgG band and * indicates the AP-2 μ2 band. E. Deleting the ring finger domain (C-terminal 53 AA) of MEX3C-1 abolished MEX3C/AP-2 interaction. 1ΔRing: MEX3C-1-KO-ΔRing, all lysine residues of MEX3C-1 were mutated to arginine and the ring finger domain was deleted. 1KO: All lysine residues in MEX3C-1 were changed to arginine. Both mutants were HA-tagged and immunoprecipitated by anti-HA antibody to detect AP-2 subunits. F. Addition of ubiquitin chain(s) to MEX3C-1-KO-ΔRing restored MEX3C/AP-2 interaction. 1ΔR-ub: MEX3C-1-KO-ΔRing-UbKO, MEX3C-1-KO-ΔRing fused to one ubiquitin chain whose lysine residues were all mutated to arginine; 1ΔR-2xub: MEX3C-1-KO-ΔRing-2xUbKO, MEX3C-1-KO-ΔRing fused to two ubiquitin chains whose lysine residues were all mutated to arginine. All three mutants were HA-tagged. For A~F, “Mock” indicates vector-DNA transfected cells.
Fig 3.
Summary of interaction data of MEX3C variants and mutants with AP-2.
Black boxes indicate sequences that were identical among MEX3C-1, MEX3C-2, and MEX3C-3. Thin gray boxes indicate sequences shared by MEX3C-1 and MEX3C-2. Thick gray boxes indicate sequences unique to MEX3C-1. Unfilled boxes indicate motifs and domains studied. Boxes filled with vertical lines indicate ubiquitin chains with lysine changed to arginine. Dashed lines indicate deleted regions in deletion mutants. Dashed red boxes indicate regions essential for MEX3C/AP-2 interaction. Sequence numbering is based on human MEX3C-1. “+” indicates able to pull down AP-2 complex in immunoprecipitation experiments while “-” indicates the inability to do so. The tag (Flag or HA) for each variant and mutant is indicated. Detailed immunoprecipitation data are presented in Fig 2.
Fig 4.
Subcellular localization of MEX3C in the endolysosome compartment.
A. Co-localization of MEX3C-2 (GFP-tagged) with an endogenous AP-2 α subunit. Endogenous AP-2 α was immunostained by specific antibody. The boxed area was shown at high magnification in the following row. Co-localized signals were marked by *. B. Co-localization of transiently expressed MEX3C variants (Flag-tagged, stained by anti-Flag antibody) with transiently expressed AP-2 μ2 (mCherry-tagged). The boxed area was shown at high magnification in the following row. Co-localized signals are marked by *. C. GTP binding defective dynaminK44A mutant trapped wild-type MEX3C-1, but not MEX3C-1 mutants unable to interact with AP-2, at dynaminK44A enriched foci. K44A: DynaminK44A mutant; 1ΔC: MEX3C-1ΔC, MEX3C-1 C-terminal 247AA deleted; 1ΔKH: MEX3C-1ΔKH, MEX3C-1 KH domains (AA 206–407) deleted. Shown are representative images of multiple double-positive cells. Quantitative data are presented in Fig A in S1 File.
Fig 5.
Co-localization of MEX3C proteins with CD63 and AGO2.
A. MEX3C proteins showed partial co-localization with CD63. B. MEX3C proteins showed a high degree of co-localization with AGO2. The boxed image shows the ubiquitous localization of singly expressed MEX3C-3. In MEX3C-3 and AGO2 double-positive cells, all MEX3C-3 protein was cytoplasmic. For A-B, MEX3C proteins were Flag-tagged; CD63 and AGO2 were GFP-tagged and their similarity to endogenous proteins was validated by respective donating investigators (see Acknowledgments for list of investigators). Shown are representative images of multiple double-positive cells. Nuclei (stained by DAPI) were pseudocolored blue.
Fig 6.
MEX3C and AP-2 are involved in EV miR-451a secretion.
A. Western blotting confirmed efficiency of the MEX3C and AP-2 siRNAs. MEX3C was detected 24 hours after siRNA transfection; the loading control was β-actin. AP-2 α and AP-2 μ2 were examined 84 hours after siRNA transfection (after EV collection). The loading control was GAPDH. The panel on the right shows the relative expression after normalized to loading control determined by densitometry. Results are representative of two independent experiments. B. Real-time PCR analysis of EV RNA expression after MEX3C or AP-2 inhibition. Because inhibiting AP-2 α or AP-2 μ, or AP-2 α and AP-2 μ in HEK293T cells showed similar effects, these data were combined. * indicates p<0.001 by Tukey's multiple comparison test following ANOVA. For miR-451a and miR-320a, N = 3; for the others, N = 2. C. MEX3C or AP-2 inhibition did not affect cellular miR-451a expression (N = 3). D. EV miR-451a was protected from RNase by membrane (N = 5). ***, p<0.0001 between protease K and RNase treated samples with and without NP40 pre-treatment, analyzed by Tukey's multiple comparison test following ANOVA. For B, C and D, means ± standard error (s.e.m.) are shown.
Fig 7.
MEX3C and AP-2 are involved in exosomal miR-451a secretion.
A. siRNA-mediated MEX3C or AP-2 inhibition reduced miR-451a in exosome-enriched preparations. For AP-2 inhibition, siRNAs for AP2A1 and AP2M1 were transfected simultaneously. Shown are representative data from two independent transfections. B. DOX-induction inhibited MEX3C expression at the RNA level (N = 3). C. shRNA targeting the MEX3C coding region inhibited MEX3C expression from pFlag-MEX3C-1 after DOX-induction (N = 2). Co-transfected EGFP expression was used as the control for transfection efficiency and loading. D. DOX-induced MEX3C inhibition decreased exosomal miR-451a but not miR-320a expression (N = 3). For B and D, the mean ± s.e.m. are presented. *, ** and *** indicate p<0.05, 0.01 and 0.0001 when compared with control in Tukey’s multiple tests (B) and Bonferroni posttests (D).
Fig 8.
Role of AP-2 and MEX3C in exosomal miR-451a secretion.
A. HGS was efficiently inhibited by siRNA si-HGS-2 but not si-HGS-1. Right: Expression as measured by densitometry (normalized with β-actin expression). Shown are representative results from two experiments. B. Inhibiting HGS expression failed to decrease exosomal miR-451a expression (N = 2). C. GW4869 significantly inhibited exosomal miR-451a expression (N = 3). The mean ± s.e.m. are shown. * indicates p<0.05 compared with DMSO control in Bonferroni posttests following ANOVA. D. Proposed model for AP-2/MEX3C interaction. At least two regions in MEX3C are necessary for its interaction with AP-2 complex. The red question mark indicates uncertainty whether the MEX3C/AP-2 interaction is direct or indirect; the black question mark indicates that the identities of the substrates ubiquitinated by MEX3C ring finger domain, and the proteins that directly bind microRNA, are unknown. E. Proposed role of MEX3C in ceramide-mediated miRNA exosomal sorting. MEX3C increases the association of microRNA to the endosome. ILV: intraluminal vesicles; MVB: multivesicular body; *: ceramide. It is unknown whether MEX3C itself is sorted into the ILVs, so MEX3C was not shown in ILVs.