Kinase interaction analysis predicts actions of the WNK-OSR1/SPAK pathway

Abstract

The WNK-OSR1/SPAK protein kinase pathway regulates ion homeostasis and cell volume, but its other functions are not well understood. To discover undefined signaling functions, we utilized experimentally-derived binding specificity to predict interactions and relative affinities with the conserved C-terminal (CCT) domains of OSR1 and SPAK, which bind short linear motifs. The upstream kinases WNKs 1-4 and their relatives, the pseudokinases NRBP1/2, also contain CCT-like domains which have conserved folds and motif binding pockets. Motifs were scored using peptide arrays, conservation, cytosolic localization, and solvent accessibility. Out of nearly 3700 motifs in the human proteome, 90% of previously published motifs ranked in the top 2% of those predicted. Interactions with selected candidates, including TSC22D1 and CAVIN1, were validated, and their localization and/or modifications were coupled to changes in WNK1 signaling. We also identified additional motif variants and confirmed binding to the NRBP1 CCT-like domain. Our results stress the diverse functionality of CCT/CCT-like domains and implicates unexpected interactions driving WNK biology.

Fig. 1. Contributions to OSR1 and SPAK CCT domain binding by residues within and around core RFxV/I and RxFxV/I motifs.

A Domains of SPAK (STK39) and OSR1 (OXSR1). Conserved C-terminal (CCT) domains interact with R-F-x-V/I and R-x-F-x-V/I motifs. B Crystal structure of the complex of OSR1 CCT and a WNK1-derived hexapeptide, GRFQVT (PDB ID: 2V3S). The peptide interacts partially as a terminal β-strand (β-strand addition), causing sidechains to alternate direction, suggesting adjacent residues do not strongly influence each other’s binding. C Motif numbering scheme. D Peptide array setup. Peptides are covalently linked to the membrane. Each spot represents a single position within the base sequence that was varied to every amino acid. Primary anti-His₆ antibody detects bound CCT. Fluorophore-labeled secondary antibody detects primary. E Quantification of peptide array spot intensities. Original array images in Supp. Fig. 1. Asterisk (*) above red residue indicates position varied for every amino acid. Caret (^) below residue indicates site in base peptide that deviates from wild-type sequence. The number to the right indicates the motif position mutated. Top (blue) row: hWNK4 (human WNK4) 1012–1024 G1215A. Middle (orange) rows: wild-type hWNK4 1012–1024. Bottom (purple) rows: hWNK1 1253–1265 R1257A (R-x-F-x-V/I motif; wild-type hWNK1 sequence contains both motifs, R-R-F-x-V/I). See main text for detailed explanations. n = 2 arrays per CCT domain, normalized by the mean intensity of the array. Ranges reported in Supp. Fig. 2. F Affinity determined by fluorescence anisotropy peptide competition with titrated, unlabeled peptides displacing 25 nM labeled peptide (NH₃⁺-NLVGRF-[DAP-FAM]-VSPVPE-COO⁻] [diaminopropionic acid (DAP)]. SPAK/OSR1 CCTs are constant at 1.5/3.0 μM. Red letters mutated relative to wild-type sequence, and blue letter is position 0. The top group is hWNK4 1012–1024, the bottom is hWNK1 1253–1265. Affinity reported as inhibition dissociation constant (K_i, µM). Upper/lower limits of the 95% confidence interval are indicated. Fold-change decrease in affinity is a fold-change increase in K_i relative to wild type. Goodness of fit reported as R-squared (R²). n = 3. Curves and fluorescent probe binding in Supp. Fig. 3. GraphPad Prism 10, one-site fit to the K_i model used for data analysis. Asterisk (*) indicates previously published.

Fig. 2. Summary of protein interaction prediction process.

A Total motifs in the initial search depending on motif type (R-F-x-V/I or R-x-F-x-V/I). B Multiple sequence alignment of all experimentally identified motifs identified from the literature^{,–,,,,,,,–}. Literature references for each motif are in the Supp. Fig. 4A. Alignments generated with Clustal Omega (clustal.org). Sequence logo generated with WebLogo (weblogo.berkeley.edu). C Flowcharts for process predicting CCT interactions and with scoring information in Supp. Fig. 4B.

Fig. 3. OSR1 and SPAK motif interaction prediction summary.

A Plots comparing scores of all identified motifs within motif classes found using the Scansite3 web server (scansite3.mit.edu). A lower score indicates a higher predicted relative binding affinity. Orange rectangles indicate previous experimentally validated interactions (direct binding or inferred from experimental results (Fig. 2B and Supp. Fig. 4A)). Dashed lines indicate cut-offs used for further analysis (false-positive filtering and subsequent scoring). B Similar to (A) but without scores included on the y-axis. Previously validated motifs cluster near the top of the rankings. C Summary of motifs retained at each step. D Overall scores of all retained motifs subjected to the final round of scoring. Blue line indicates all 501 scored motifs. Red triangles indicate positions of previously validated motifs. Asterisks (*) indicate motifs initially identified in our Scansite3 analysis that led to publication of validated interactions prior to publication of this study^,. The inset graph depicts motifs ranked 1 to 80 for clarity. E Histogram depicting distribution of scores. Each bin represents the total number of motifs (blue) and previously validated motifs (red) within the bin. Bin size is 5 points. Total motifs per bin listed below. F Gene ontology (GO) analysis of the top 50 motifs.

Fig. 4. Multiple identified motifs interact with OSR1 and SPAK in vitro.

A AlphaFold structural predictions of motifs in the proteins^,. Images generated using AlphaFold 3D Viewer. Green sidechains with red lettering indicate core motifs. Structure coloring indicates model confidence. Dark blue: very high, light blue: high, yellow: low, orange: very low. Low model confidence can indicate the region is unstructured in isolation. Model confidence was not taken into account during structural analysis. B Affinity determined by fluorescence anisotropy peptide competition assays (see Fig. 1F and Supp. Fig. 3A). OSR1 CCT WT peptide: solid red line; OSR1 CCT V/I→A: dashed red line; SPAK CCT WT peptide: solid blue line; SPAK CCT V/I→A: dashed blue line. Flanking arginines are added to some peptides to enhance solubility. Error bars represent a 95% confidence interval. Individual data points not included for clarity. Data points can be found in the Supp. Data 3 (n = 3). C Quantitation of results in (B). n = 3 (see Fig. 1F).

Fig. 5. Validation of interactions via co-IP in HEK293T cells.

A Overexpressed FLAG-hSPAK 50-545 bait protein pre-loaded on beads pulled down overexpressed myc-tagged prey proteins (HA tag for ATG9A) from HEK293T cell lysates. 100 µM CCT blocking peptide (WNK1 1253–1265; NH₃⁺-SAGRRFIVSPVPE-COO⁻) as control. Supp. Fig. 5 presents the prey input and co-IP bands at identical detection intensities for comparison of the amount of protein remaining after wash steps. B Quantification of (A). n ≥ 3. Error bars indicate 95% confidence intervals. Replicates shown as dots. Green is statistically significant; red is not.

Fig. 6. Cavin1 phosphorylation is influenced by WNK signaling.

A–D siRNA-mediated knockdown of WNK1 (72 h) or inhibition with the pan-WNK inhibitor, WNK463 (18 h, 1 µM), in human dermal microvascular endothelial cells (HDMEC). Mean ± SD (n = 3; *P < 0.05, **P < 0.01; unpaired two-tailed t-test). E Cavin1 motif. Blue: basic/positively charged amino acids; red: acidic/negative; underlined: R-x-F-x-V; Y156 is the site of phosphorylation. F AlphaFold3 models analyzing Cavin1 pTyr156 effects on intermolecular interactions of SPAK CCT (human 446–545) with Cavin1 13-mer used in experiments and intramolecular interactions of the 13-mer alone. Model statistics in Supp. Fig. 9A. G Diagram indicating potential intramolecular interactions (dashed lines) between acidic phospho-Tyr156 at +5 position of Cavin1 motif and multiple basic residues at the -4, -3, -2, and +1 positions. Such interactions might reduce CCT domain interactions with the motif. H Fluorescence anisotropy peptide competition (see Fig. 1F). The higher line indicates weaker binding. Quantification unavailable due to Cavin1 p-Tyr peptide instability at higher concentrations. Error bars represent a 95% confidence interval. Individual data points not included for clarity. Data points can be found in the Supp. Data 3 (n = 3).

Fig. 7. TSC22D1 interacts with OSR1/SPAK and impacts WNK signaling.

A Multiple sequence alignment of TSC22D proteins around the motif. TSC22D3 lacks the region containing the motif. B co-IP of endogenous TSC22D1 with 3xFLAG-OSR1 in HEK293T cells. 50 µM CCT blocking peptide (WNK1 1253–1265; NH₃⁺-SAGRRFIVSPVPE-COO⁻) used as negative control (n = 3). C Similar to (B) except bacterially expressed GST-OSR1 CCT used for pull-down (n = 3). D Colocalization of endogenous OSR1 and stably-expressed TSC22D1-mNeonGreen in A549 cells ±1 µM pan-WNK inhibitor, WNK463, treated 24 h. Scale bar: 20 µm. (n = 3). E, F Western blot of partial siRNA-mediated knockdown of TSC22D1 in HDMEC after 48 h siRNA treatment. Mean ± SEM (n = 3; *P < 0.05, **P < 0.01, ***P < 0.001; unpaired two-tailed t-test). pSPAK/pOSR1 blot shown twice at varying intensities.

Fig. 8. CCT-like domains and alternative binding motifs expand complexity.

A Region of the human kinome tree containing WNK1-4, OSR1, SPAK, and NRBP1/2; all proteins contain CCT or CCT-like (CCTL) domains. B Diagram depicting locations of CCT and CCTL domains in proteins. C Multiple sequence alignment of regions of CCT/CCTL domains that comprise motif binding pockets consisting of β-strand–loop–α-helix. Conserved binding residues indicated. D Comparison of crystal structure of OSR1 CCT:GRFQVT hexapeptide complex (left, PDB:2V3S) with AlphaFold3-derived models of OSR1 CCT (430–527), WNK1 CCTL1 (human, 482–597), and NRBP1 CCTL (436–535) bound to TSC22D1 RFxV motif (395–419)^,,,. For NRBP1:TSC22D1, when using full-length proteins, differing models were obtained depending on the AlphaFold modeling server used. Both results can be found in the Supp. Fig. 9B. Identical results were obtained using truncated vs full-length proteins for OSR1 and WNK1. E, F Structural alignment of AlphaFold3 models in (D). Binding residues depicted as sticks. G Model of SPAK CCT:GRFQVT complex made by overlaying SPAK CCT structure (PDB: 7O86) onto the OSR1 CCT:GRFQVT complex structure. D477 (green) was mutated in the present study to inhibit binding while L491 (green) was mutated in previous study. H, I 3xFLAG-tagged SPAK 50-545 ± D477Q or NRBP1 ± E488Q used to co-IP myc-NKCC2 1-170 and myc-TSC22D1, respectively, in HEK293T cells. Mean ± SEM of mutants relative to WT (n = 3, 95% confidence intervals shown). J Affinity determined by fluorescence anisotropy peptide competition assays (n = 3, see Fig. 1F and Supp. Fig. 3A. Asterisk (*) indicates results previously published.