Phosphorylation of human aquaporin 2 (AQP2) allosterically controls its interaction with the lysosomal trafficking protein LIP5

Abstract

The interaction between the renal water channel aquaporin-2 (AQP2) and the lysosomal trafficking regulator-interacting protein LIP5 targets AQP2 to multivesicular bodies and facilitates lysosomal degradation. This interaction is part of a process that controls AQP2 apical membrane abundance in a vasopressin-dependent manner, allowing for urine volume adjustment. Vasopressin regulates phosphorylation at four sites within the AQP2 C terminus (Ser²⁵⁶, Ser²⁶¹, Ser²⁶⁴, and Thr²⁶⁹), of which Ser²⁵⁶ is crucial and sufficient for AQP2 translocation from storage vesicles to the apical membrane. However, whether AQP2 phosphorylation modulates AQP2-LIP5 complex affinity is unknown. Here we used far-Western blot analysis and microscale thermophoresis to show that the AQP2 binds LIP5 in a phosphorylation-dependent manner. We constructed five phospho-mimicking mutants (S256E, S261E, S264E, T269E, and S256E/T269E) and a C-terminal truncation mutant (ΔP242) that lacked all phosphorylation sites but retained a previously suggested LIP5-binding site. CD spectroscopy indicated that wild-type AQP2 and the phospho-mimicking mutants had similar overall structure but displayed differences in melting temperatures possibly arising from C-terminal conformational changes. Non-phosphorylated AQP2 bound LIP5 with the highest affinity, whereas AQP2-ΔP242 had 20-fold lower affinity as determined by microscale thermophoresis. AQP2-S256E, S261E, T269E, and S256E/T269E all had reduced affinity. This effect was most prominent for AQP2-S256E, which fits well with its role in apical membrane targeting. AQP2-S264E had affinity similar to non-phosphorylated AQP2, possibly indicating a role in exosome excretion. Our data suggest that AQP2 phosphorylation allosterically controls its interaction with LIP5, illustrating how altered affinities to interacting proteins form the basis for regulation of AQP2 trafficking by post-translational modifications.

Keywords: aquaporin; membrane protein; membrane trafficking; phosphorylation; protein-protein interaction.

Figure 1.

Crystal structure of human AQP2. Cartoon representation of the 2.75 Å human AQP2 crystal structure (Protein Data Bank code 4NEF) with residues that could not be resolved displayed as beads. Phosphorylation and ubiquitination sites are highlighted in red and gray, respectively. The proximal part of the C terminus that has been shown to bind LIP5 (20) forms a short helix (green).

Figure 2.

Phosphorylation of AQP2 in P. pastoris. A, purified AQP2 produced in P. pastoris was transferred to a PVDF membrane and probed for phosphorylation using Phos-tag BTL-111. A clear phosphorylation signal is seen for non-treated AQP2 (lane 1) with bands corresponding to monomeric (30 kDa), dimeric (60 kDa), and tetrameric (120 kDa) forms of the protein. Following treatment with alkaline phosphatase for 2, 4, and 20 h (lane 2-4), no bands can be seen, illustrating complete removal of phosphorylation. B, Western blot stained with antibody against the His tag shows that equal amount of AQP2 is present in all samples. Protein bands corresponding to those seen in A are marked with arrows. Protein masses are indicated in kDa.

Figure 3.

CD spectroscopy and thermal stability of AQP2 and AQP2 mutants. A, far-UV CD spectra for non-treated wild-type AQP2 (nt AQP2), dephosphorylated wild-type AQP2 (dep AQP2), and dephosphorylated AQP2 phosphorylation mutants (S256E, S261E, S264E, T269E, and S256E/T269E). B, melting curves for dephosphorylated AQP2 and AQP2 phosphorylation mutants. The curves are obtained by plotting the MRE at 222 nm from CD spectra obtained at different temperatures. C, melting curves as in B for non-treated wild-type AQP2, dephosphorylated wild-type AQP2, and AQP2-S256E. The melting curve for non-treated AQP2 is more similar to AQP2-S256E than to dephosphorylated AQP2. D, bar chart displaying melting temperatures for the different AQP2 constructs. All phosphorylation mutants have a significantly lower T_M than dephosphorylated AQP2 (*** indicates p < 0.001, and ** indicates p < 0.01; NS indicates not significant).

Figure 4.

Far-Western blot analysis of the interaction between AQP2 and LIP5. Equal amounts of dephosphorylated wild-type AQP2, AQP2 mutants, and AQP4 were transferred to a nitrocellulose membrane and incubated with purified LIP5. The interaction was detected using an antibody directed against LIP5. A, typical far-Western blot showing that wild-type AQP2, but not AQP4, interacts with LIP5. AQP2-S256E, S261E, S264E, T269E, S256E/T269E, and ΔP242 all interact with LIP5 but show significant variation in signal strength. B and C, staining the membrane with antibodies against the His tag (B) and Ponceau reagent (C) confirmed that protein equivalents had been loaded in each lane. AQP2-ΔP242 does not contain a His tag and is consequently not stained with anti-His antibodies (B). Marker protein masses are indicated in kDa. D, relative intensities of far Western blot signal for AQP2 mutant constructs and AQP4 compared with wild-type AQP2 (set as 100%). The values were calculated from three independent blots. The AQP2 variants S256E, S261E, S264E, and S256E/T269E show significantly less binding than wild-type AQP2 (p < 0.001, indicated by ***) but were not significantly different from each other. The binding was further reduced for ΔP242 (p < 0.001), whereas the signal intensity of AQP2 T269E was more similar to wild-type (p < 0.01, indicated by **). AQP4 did not interact with LIP5 (p < 0.001).

Figure 5.

Phosphorylation modulates the affinity between AQP2 and LIP5. The interaction between LIP5 and AQP2 was measured with MST, and the data were fitted according to a one-to-one binding model. A, dephosphorylated WT AQP2 interacts with LIP5 (K_d = 191 ± 43.2 nm). The data for AQP4 (negative control) had a low amplitude, could not be fitted to the equation, and did not reach saturation; therefore binding was interpreted as non-significant. B, C-terminal truncation of AQP2 (ΔP242) reduces the affinity 20-fold (K_d = 3.63 ± 0.44 μm). C, comparison of MST curves for WT AQP2 and phospho-mimicking mutants. There is a significant difference between WT and S256E (K_d = 1.00 ± 0.25 μm), S261E (K_d = 745 ± 141 nm), T269E (K_d = 721 ± 55.0 nm), and S256E/T269E (652.7 ± 62.2 nm) (*, p < 0.05; **, p < 0.01; ***, p < 0.001). There is no significant difference between WT and S264E (K_d = 278 ± 49.1 nm). D, bar chart summarizing K_d values determined in A–C.

Figure 6.

LIP5 affinity correlates with AQP2 stability. Plotting the dissociation constant (log K_d) for the interaction between AQP2 variants and LIP5 against the inverted AQP2 melting temperature (1/T_M) reveals a linear relationship (R² = 0.4369). The correlation is weaker for the more distal mutation S264E and T269E, whereas the double S256E/T269E mutant fits well to the linearity.

Figure 7.

Model for the role of AQP2 phosphorylation in regulating LIP5-mediated targeting to MVBs. Red and green boxes indicate phosphorylation sites with decreased and increased affinity to LIP5, respectively. During unstimulated conditions, AQP2 resides in storage vesicles and is mainly phosphorylated on Ser²⁶¹. Dephosphorylation of Ser²⁶¹ and phosphorylation of Ser²⁵⁶ trigger targeting of AQP2 to the apical membrane, whereas additional phosphorylation at Thr²⁶⁹ further increases the AQP2 apical membrane residence time. Phosphorylation at these three sites reduces LIP5 affinity. Dephosphorylation, as well as ubiquitination of AQP2, in the apical membrane induces endocytosis. Dephosphorylated AQP2 has the highest affinity for LIP5, stimulating its sorting into the inner vesicles of MVBs by coordinating the actions of ESCRT-III complex and VPS4. AQP2 phosphorylated at Ser²⁶⁴ also has higher affinity for LIP5 and may be involved in sorting AQP2 into MVBs for secretion in exosomes.