Identification and Validation of a Novel WNK2 Inhibitor: A New Genetically Informed Target for Osteoarthritis Drug Development

Abstract

Osteoarthritis (OA) is a disease characterized by loss of joint space, degeneration of cartilage at articular surfaces, remodeling of bone and other joint tissues, low-grade inflammation, and pain. It is a leading cause of disability in the aging population. We do not have effective disease-modifying drugs because we lack a comprehensive understanding of the genetic pathways underlying OA development. Our recent work demonstrated that dominant hypermorphic mutations in the osmotic stress sensing kinase, WNK2, are associated with OA susceptibility. To explore WNK2 inhibition as a potential disease-modifying osteoarthritis drug, we identified a WNK2 inhibitor by using a combination of computational and experimental approaches. We used AlphaFold2 to model the WNK2 kinase domain with high confidence (pLDDT score >90), enabling virtual screening of 4.8 million compounds in the ZINC database. Structure-based filtering prioritized 53 candidates interacting with WNK2-specific residues Glu245 and Glu249, with binding energies from −4 to −8 kcal/mol. M04, (1-(6-((3,5-bis­(trifluoromethyl)­benzyl)­(methyl)­amino)­pyrimidin-4-yl)­pyrrolidin-2-yl)­methanol, emerged as a partially selective WNK2 inhibitor with minimal cytotoxicity in human T/C-28a2 chondrocyte cells (cell viability IC₅₀ = 416 μM). M04 inhibited IL1β-mediated proinflammatory gene expression in primary human chondrocytes. RNaseq analysis comparing M04 vs a pan-WNK inhibitor in primary human chondrocytes indicated that M04 was a more effective at reducing expression of IL1β-induced OA-associated pathways. Further studies indicated the R-enantiomer is the active form of the molecule. These findings position M04 as a promising lead for WNK2-targeted OA therapy, supporting further in vivo studies for optimization and validation.

Introduction

Osteoarthritis (OA) is a disease characterized by loss of joint space, degeneration of cartilage at articular surfaces, remodeling of bone and other joint tissues, low-grade inflammation, and pain. , At present, there are no disease-modifying drugs available for OA, and treatment options are limited to pain management and joint replacement. A lack of targeted therapies highlights the need to identify molecular pathways underlying OA progression and develop effective treatments. − To address this issue, we have taken a combined genetic and functional approach that identifies molecular pathways that are critical for joint homeostasis. We identified families in which age-associated OA occurs as a fully penetrant, dominantly inherited trait and performed whole exome sequencing to discover rare coding alleles segregating with the OA phenotype. − By focusing on familial forms of OA, we avoid the confounding effects of studying populations in which disease occurs for heterogeneous reasons. The affected genes regulate pathways that are necessary for maintenance of the joint under normal behavioral conditions.

We recently identified 4 independent OA families with rare coding mutations in With-No-Lysine (K) 2 (WNK2). These coding variants alter evolutionarily conserved residues that are necessary for sensing and responding to hyperosmotic stress. The WNK protein kinases (WNK1–4) are intracellular sensors that respond to hyperosmotic stress by regulating ion channel activity and signaling pathways. WNKs directly phosphorylate the downstream kinases STE20/SPS1-related proline-alanine-rich kinase (SPAK) and oxidative stress-responsive kinase 1 (OSR1), which subsequently phosphorylate and regulate sodium- and potassium-chloride cotransporters to modulate osmoregulation. Our functional studies demonstrated that the WNK2 function mediates a significant portion of the response to acute and chronic hyperosmotic stress in chondrocytes. We further showed that the WNK2 OA-associated variants are hypermorphic and their expression is sufficient to promote expression of pathways associated with OA, a response that is further amplified under conditions of hyperosmotic stress. Our data indicate that reduction of WNK2 activity may be a promising therapeutic target for the treatment of OA.

Mutations in WNK1 and WNK4 are associated with hypertension and renal salt absorption and may be novel drug targets. WNK463 is a compound that inhibits all four WNK proteins with IC50 values of 5 nM (WNK1), 1 nM (WNK2), 6 nM (WNK3), and 9 nM (WNK4). This compound has been shown to produce significant, dose-dependent effects on blood pressure and heart rate in animal models. Specifically, WNK463 treatment reduces blood pressure and increases heart rate in spontaneously hypertensive rats (SHRs) when administered orally (1–10 mg/kg; 4 h). Additionally, it causes an increase in urine output and sodium and potassium excretion from the urine. Although WNK463 is a potent pan-WNK inhibitor it has a short half-life of 2.1 h in Sprague–Dawley rats. Furthermore, WNKs have other roles beyond the renal system, and systemic inhibition of all WNKs may not be a viable therapeutic option. − Overall, WNK463 has a number of limitations, making it imperative to develop more effective and targeted inhibitors for use in OA.

We propose that developing new WNK inhibitors could offer a novel therapeutic approach to prevent or slow the progression of OA. Having genetically supported therapeutic targets increases the success rate of FDA drug approval more than 2.5-fold. − Here, our genetic analysis of families with OA and integrated workflow led to the identification of a partially selective WNK2 inhibitor with minimal cytotoxicity. These results highlight M04 as a promising candidate for osteoarthritis therapy through its inhibition of WNK2, supporting further optimization of M04 and related compounds as potential treatments for OA.

Results

3D Structure of the WNK2 Protein and the PKD

WNK2 is a large protein (2122 amino acids) that has several conserved domains, including an N-terminal serine/threonine-protein kinase domain (PKD), a chloride-binding/regulatory region, a C-terminal autoinhibitory domain, coiled-coil domains, a SPAK/OSR1 binding motif, and an intrinsically disordered C-terminus that is necessary for phase separation and volume recovery. We used AlphaFold2 to generate a 3D structure of the full-length WNK2 protein and the WNK2 PKD. The accuracy of AlphaFold2 prediction was confirmed by the Predicted Local Distance Difference Test (pLDDT) scores, which provide valuable insights into the reliability of these predicted features. The average pLDDT score for the full-length WNK2 protein was 44.43, indicating a low level of confidence and a disordered visual structure (Figure A). In contrast, the average pLDDT score for the protein kinase domain was significantly higher at 89.79 (Figure B,C). Most amino acids within this domain had pLDDT scores exceeding 90, reflecting a more confidently predicted structure (Figure C). Therefore, we concentrated solely on the WNK2–PKD domain for drug discovery. A direct comparison and TM-score measurement of WNK2 were not performed because no crystal structure of the WNK2 PKD is available.

1.

3D structure of the full-length WNK2 protein and PKD domain. (A) The predicted 3D structure of the full-length WNK2 protein using AlphaFold2. (B) The predicted protein kinase domain (WNK2–PKD) model using AlphaFold2 shows secondary structure elements with red representing helices, yellow representing sheets (beta strands), and blue representing loops or turns. The N-terminal (N-term) and C-terminal (C-term) regions are labeled. (C) Confidence levels of the WNK2–PKD prediction based on pLDDT scores. Blue indicates very high confidence (pLDDT > 90), cyan represents confident regions (70 < pLDDT ≤ 90), yellow denotes low confidence (50 < pLDDT ≤ 70), and orange represents very low confidence (pLDDT ≤ 50).

3D Structure of a Selected Binding Pocket in the WNK2 PKD

As the protein kinase domains of WNK1–WNK4 are highly conserved, sharing approximately 85–90% sequence identity, we sought to identify a binding pocket that contains amino acids unique to WNK2 (Figure S1). Among several open binding pockets found, we selected Binding Pocket-4 (BP4) due to two amino acids, Glu245 and Glu249, that are exclusively present in WNK2 and not in other WNKs (Figures and S1). These residues are considered targetable sites, making BP4 a promising candidate for inhibitor design and molecular docking studies. The amino acid residues in BP4 include Leu116, His119, Thr120, Asp180, Glu181, Ser182, Val185, Lys244, Glu245, Arg247, Tyr248, Glu249, Ile250, and Lys251. The XYZ spatial coordinates of BP4 are (−5.17, 5.97, −10.70), and it has an approximate diameter of 54 Å.

2.

3D structure of the selected binding pocket WNK2 PKD domain. Left panel: The predicted structure of WNK2–PKD highlighting the binding pocket (BP-4) in yellow and blue. The N-terminal (N-term) and C-terminal (C-term) regions are labeled, along with the secondary structure elements depicted in red (alpha-helices), yellow (beta-sheets), and blue (loops and turns). Right panel: A closer view of the binding pocket (BP-4), showing the amino acid residues contributing to the binding interface. Key residues include Leu116, Glu245, Glu249, Arg247, Tyr218, Ser214, and others, which are critical for ligand interactions and selectivity.

Identification of Candidate WNK2 Interacting Molecules

After selecting BP4 as the screening site, we carried out a virtual screen of 4.8 million commercially available compounds from the ZINC database that met Lipinski’s Rule of Five criteria. Using the Molecular Operating Environment (MOE), a pharmacophore-based search was first applied to enrich for compounds with features compatible with the BP4 pocket, reducing the library to approximately 100,000–150,000 candidates. These compounds were then docked to BP4 using induced-fit refinement in MOE, with at least 100 poses generated for each compound. Candidate molecules were ranked based on docking score, binding orientation, and predicted interactions within the pocket, and 53 compounds were selected for further analysis (Table ).

1. Binding Energies (kcal/mol) of the Top 53 WNK2–PKD Interacting Molecules (M01–M53) Calculated Using MOE and AutoDock Vina .

molecule ID	binding energy (kcal/mol)		molecule ID	binding energy (kcal/mol)		molecule ID	binding energy (kcal/mol)
	MOE	Autodock Vina		MOE	Autodock Vina		MOE	Autodock Vina
M01	–4.79	–9.19	M21	–4.72	–8.05	M41	–4.60	–9.01
M02	–5.53	–8.11	M22	–8.29	–8.94	M42	–4.57	–8.01
M03	–5.85	–8.67	M23	–7.77	–7.08	M43	–4.53	–8.62
M04	–4.75	–9.19	M24	–6.81	–8.34	M44	–4.49	–7.98
M05	–4.76	–8.04	M25	–5.92	–8.74	M45	–4.49	–6.82
M06	–4.82	–8.02	M26	–5.71	–7.30	M46	–4.42	–7.37
M07	–5.47	–8.85	M27	–5.70	–8.68	M47	–4.39	–8.68
M08	–5.45	–9.22	M28	–5.68	–7.39	M48	–4.30	–8.51
M09	–5.37	–8.06	M29	–5.67	–9.46	M49	–4.21	–8.20
M10	–5.36	–8.47	M30	–5.64	–7.79	M50	–4.11	–9.10
M11	–5.35	–8.34	M31	–5.56	–9.15	M51	–3.95	–6.74
M12	–5.32	–9.25	M32	–5.55	–8.90	M52	–3.89	–8.36
M13	–5.23	–9.08	M33	–5.52	–8.58	M53	–3.27	–8.29
M14	–5.11	–9.40	M34	–5.45	–8.40
M15	–5.11	–7.78	M35	–5.44	–8.57
M16	–4.94	–9.91	M36	–5.43	–8.32
M17	–4.90	–8.59	M37	–5.29	–7.59
M18	–4.80	–8.61	M38	–5.08	–8.21
M19	–4.80	–9.25	M39	–5.07	–9.17
M20	–4.76	–8.05	M40	–4.62	–7.85

^aThe table highlights the comparative results from both docking tools, aiding in the prioritization of candidates for further evaluation.

Top 53 Candidate WNK2 Interacting Molecules

Among ∼150 K candidate molecules, 53 were selected for their favorable binding energies, docking poses, and interactions with key residues (Glu245 and Glu249) within the WNK2–PKD binding pocket. The structures of these molecules are displayed in Figure . The binding affinities of these top candidates, determined using both MOE and AutoDock Vina, ranged from approximately −3 to −10 kcal/mol (Table ).

3.

Selected WNK2–PKD interacting molecules (M1 to M53) and their corresponding binding energies. The 3D structures of the 53 molecules (M01 to M53) were identified through virtual screening against the WNK2 protein kinase domain (WNK2–PKD). Each inhibitor is labeled accordingly and is shown alongside its calculated binding energy (S-binding energy) in kcal/mol, as determined using MOE.

Identification of the Top 6 Protein–Molecule Interactions

The selected 53 molecules were visually inspected to validate their binding modes and key interactions. Among these, six inhibitor (M01–M06) interactions with the WNK2–PKD binding pocket are shown in Figure : M01 forms hydrogen bonds with Glu245 and Ile250, stabilized by polar contacts from Glu181, Lys244, Arg247, Tyr248, and Glu249, along with hydrophobic support from Leu116 and Val185 (Figure A). M02 engages hydrogen bonds with Ser182 and Glu249, further supported by polar and charged residues (Glu181, Lys244, Arg247, Tyr248, and Glu245) and hydrophobic contacts (Leu116, Val185, and Ile250) (Figure B). M03 anchors via hydrogen bonds to Asp180, Glu245, and Glu249, supplemented by polar interactions with Ile250 and other surrounding residues and stabilized by Leu116 and Val185 (Figure C). M04 exhibits specificity through an ionic interaction with the WNK2-unique residue Glu249, complemented by hydrophobic fluorinated rings and polar contacts from Lys244 and Lys251 (Figure D). M05 secured by hydrogen bonds to Glu181, Glu245, and Asp180, with additional polar interactions from Glu249, Lys244, and hydrophobic contacts from Leu116 and Val185 (Figure E). M06 establishes a key hydrogen bond with Glu245 and gains stability from polar residues (Asp180, Glu181, Glu249, Lys244, and Ser182) and hydrophobic surfaces (Leu116 and Val185) (Figure F). Collectively, these six molecules demonstrate favorable binding within the WNK2–PKD pocket, making them promising candidates for further optimization and development.

4.

Interaction profiles of selected molecules (M01 to M06) with the binding pocket BP-4 of the WNK2–PKD. (A–F) 2D interaction diagrams of the top six molecules (M01 to M06) with the binding pocket BP-4 of the WNK2 protein kinase domain (WNK2–PKD), generated using MOE. Each panel shows the interactions between the protein and a bound ligand (stick representation). Key interactions are highlighted: (1) hydrogen bonds as dashed lines between donor and acceptor atoms, (2) hydrophobic contacts as blue spheres around interacting residues, (3) aromatic interactions as pink spheres between aromatic rings. The ligand is shown in stick representation with different colors representing different chemical groups: carbon (gray), oxygen (red), nitrogen (blue), and fluorine (green).

Cell Toxicity of Selected Inhibitors

To determine if any of the top WNK2–PKD interacting molecules had cytotoxic effects, we determined cell viability in human T/C-28a2 chondrocytes using the XTT viability assay over a range of concentrations (0.01 μM to 400 μM). To evaluate tolerability, IC₅₀ valuesrepresenting the concentration needed to decrease cell viability by 50%were calculated for each compound. M04 displayed the highest IC₅₀ (416 μM), suggesting minimal cytotoxicity and favorable tolerability in chondrocytes. In comparison, M05 and M01 exhibited lower IC₅₀ values of 12.77 μM and 12.80 μM, respectively, indicating higher levels of cytotoxicity. The remaining compounds showed moderate profiles, with IC₅₀ values of 39 μM for M03, 329.5 μM for M02, and 388.8 μM for M34 (Figure ). These findings underscore M04 as the most suitable candidate for continued investigation based on its low cytotoxicity in chondrocytes.

5.

Cell viability assay of WNK2–PKD interaction molecules using the XTT viability assay. (A–F) Dose–response curves for the six molecules (M01, M02, M03, M04, M05, and M06) tested using the XTT assay. The X-axis represents the inhibitor concentration (μM) on a logarithmic scale, while the Y-axis represents cell viability (%). IC₅₀ values (the concentration at which 50% cell viability is inhibited) are indicated for each inhibitor in the inset boxes of their respective panels.

M04 Is a WNK2 Inhibitor

WNKs directly phosphorylate and activate the kinases SPAK and OSR1 to modulate osmoregulation through direct modulation of ion channels and altering gene expression (Figure A). ,, To determine if selected WNK2–PKD interacting molecules can inhibit the kinase activity of WNK2, we examined the accumulation of phosphorylated SPAK (pSPAK) in T/C-28a2 chondrocytes under isotonic conditions and acute osmotic stress. Our previous studies indicated the majority of SPAK phosphorylation in T/C-28a2 chondrocytes is due to WNK2 activity. T/C-28a2 chondrocytes were treated with isotonic media (Iso/300mOsm) or media supplemented with sorbitol (200 mM/500mOsm) for 5 min in the presence of DMSO control, candidate inhibitors M01 (20 μM), M02 (100 μM), M03 (20 μM), M04 (100 μM), M05 (20 μM), and M06 (20 μM), or the pan-WNK1–4 inhibitor WNK463 (100 μM), and protein extracts were collected for immunoblot analyses (Figure B). Among the tested compounds, M04 demonstrated a significant reduction in pSPAK levels under isotonic conditions and acute osmotic stress (Figure C). Treatment of chondrocytes with 200–400 μM M04 reduced pSPAK comparable to WNK463 (Figure B,C). M01 and M03 also reduced pSPAK levels but failed to inhibit pSPAK to the same degree as M04 and WNK463 (Figures D,E and S2). These results confirm that M04 effectively inhibits WNK2 protein kinase activity, supporting its potential as a targeted therapeutic candidate.

6.

M01, M03, and M04 inhibition of pSPAK levels under isotonic and osmotic stress conditions. (A) Graphical representation of WNK2 activity. WNK2 phosphorylates SPAK and OSR1 in response to osmotic stress, initiating downstream signaling pathways and altering gene expression. This process regulates ion transporters NKCC1 and KCC2 activity to maintain osmotic homeostasis in cells. (B) Immunoblot analysis of pSPAK levels under isotonic (top; Iso, 300 mOsm) and sorbitol-induced hyperosmotic stress conditions (bottom; Sor, 200 mM sorbitol, 500 mOsm). Cells were treated with candidate inhibitors M01 (20 μM), M02 (100 μM), M03 (20 μM), M04 (100 μM), M05 (20 μM), and M06 (20 μM) or the pan-WNK1–4 inhibitor WNK463 (100 μM). GAPDH was used as a loading control. For both panels, lane assignments are as follows: 1, M01; 2, M02; 3, M03; 4, M04; 5, M05; 6, M06; +C, WNK463 positive control with sorbitol; −C, WNK463 positive control without sorbitol; and D, DMSO vehicle control. Dose-dependent effects of M04, M01, and M03 are shown in panels (C), (D), and (E), respectively, on pSPAK levels under isotonic (top) (ISO; 300 mOsm) and sorbitol-induced hyperosmotic (bottom) (Sor; 200 mM/500 mOsm) stress conditions. Lane labels: –C (WNK463, positive control, without sorbitol), +C (WNK463, positive control, with sorbitol), D: DMSO, B: Blank, and 1 μM, 10 μM, 50 μM, 100 μM, 200 μM, and 400 μM inhibitors M04, M01, and M03. GAPDH was used as a loading control.

To better characterize the druggability of M04, we evaluated its predicted ADMET profile using SwissADME and ADMETlab 3.0 (Table S1). M04 exhibited physicochemical features consistent with a drug-like small molecule, including a molecular weight of ∼434 g/mol, a TPSA of 52.49 Ų, seven rotatable bonds, one hydrogen-bond donor, and no Lipinski rule-of-five violations. M04 also showed a favorable predicted QED value and no PAINS alerts, suggesting limited structural liability for nonspecific assay interference. ADMETlab 3.0 further predicted several properties that may require consideration during future optimization, including high plasma protein binding, potential P-gp and CYP3A4 inhibition, BBB permeability, and possible hERG-related liability. Overall, these in silico analyses indicate that M04 has drug-like characteristics suitable for an early stage hit compound, while also identifying pharmacokinetic and safety-related features that should be assessed experimentally in future lead-optimization studies.

Molecular Modeling of Ligand Interactions with the WNK1–4 Kinase Domains

To determine the subtype preference of M04 with the WNK1–4 kinase domains, we performed molecular docking and molecular dynamics (MD) analyses. We first used WNK463 as a control ligand. WNK463 had favorable binding across all 4 WNK kinase domains with ICM docking scores of −9.449 (WNK1), −12.5 (WNK2), −9.06 (WNK3), and −19.31 (WNK4). During the 10 ns MD trajectory, the RMSD/RMSF, contact persistence, and GB/SA results supported stable predicted binding of WNK463 to the WNK1–4 kinase domains (Supporting Information).

We next analyzed the interactions of M04 with the WNK1–4 kinase domains. The combined molecular docking and 10 ns MD analyses suggest that M04 shows a predicted subtype preference for WNK3, with WNK2 as the secondary binding subtype, whereas binding to WNK1 and WNK4 appears weaker and/or less dynamically stable. Although the initial docking scores favored WNK2 and WNK4, the MD-refined GB/SA binding-energy profiles provided a clearer selectivity trend. After excluding isolated numerical GB/SA outliers, M04 showed the most favorable mean ΔGbind with WNK3 (∼−11.6 kcal/mol), followed by WNK2 (∼−9.0 kcal/mol), while WNK4 (∼−6.7 kcal/mol) and WNK1 (∼−5.9 kcal/mol) were less favorable.

This corresponds to an approximate ΔΔG preference of ∼2.6 kcal/mol for WNK3 over WNK2 and ∼5–6 kcal/mol for WNK3 over WNK1/WNK4, supporting a predicted WNK3-biased binding profile for M04. This energetic trend is also supported by the MD stability metrics. WNK3–M04 showed the lowest ligand/protein RMSD profile among the M04 complexes and maintained the highest ligand-contact count, whereas WNK1–M04 and WNK4–M04 showed larger RMSD drift during the 10 ns simulations (Supporting Information). In sum, the docking, RMSD, ligand-contact, and GB/SA analyses indicate that M04 has predicted preferential binding toward WNK3, with moderate binding to WNK2 and weaker/stability-limited binding to WNK1 and WNK4.

In Vitro WNK1–4 Inhibition Assays

M04 was originally developed with the aim of selectively targeting WNK2 by utilizing amino acids that are unique to WNK2. To test the specificity of M04, we performed luminescence-based kinase assays using the recombinant kinase domains of WNK1–4 and SPAK as the substrate. We validated the assay by performing the kinase assay in the presence of 200 μM WNK463. As expected, WNK463 inhibited WNK1–4 (Figure ). We observed a dose-dependent reduction in WNK2 activity in the presence of MO4 (Figure B). MO4 did not affect WNK1 activity (Figure A). We also observed dose-dependent reduction in WNK3 and WNK4 activity, although the extent of inhibition was not as complete as WNK2 and did not reach the same level as WNK463 (Figure C,D). This indicates that M04 is able to interact with multiple WNK proteins and inhibit their kinase activity. Together, these results suggest that, although M04 inhibits WNK2, it also partially inhibits WNK3 and WNK4, suggesting the binding sites of these kinases overlap. The structure of M04 needs to be refined further to improve its ability to discriminate between closely related WNK isoforms and increase its specificity for WNK2.

7.

In vitro inhibition of WNK kinase activity by M04. Kinase activity of WNK1 (A), WNK2 (B), WNK3 (C), and WNK4 (D) was measured using the ADP-Glo Kinase Assay. Luminescence (relative light units, RLU) was measured after subtraction of the background signal (black). Kinase reactions were performed in the presence of DMSO (vehicle control), increasing concentrations of the inhibitor M04 (1, 10, 50, 100, and 200 μM) or the pan-WNK inhibitor WNK463 (200 μM) (positive control). Dose-dependent decreases in luminescence indicate inhibition of kinase activity by M04. Data represent the mean ± SD from [n = 4] independent experiments. Statistically significant differences (drug treatment vs DMSO) of p ≤ 0.05 were determined by ordinary one-way ANOVA, followed by Dunnett’s multiple-comparisons test.

Enantioselective Synthesis and Inhibition of pSPAK with the M04 R-Enantiomer

We determined which enantiomer of M04 was the active form of the molecule. We synthesized racemic, (R)-, and (S)-1-(6-((3,5-bis­(trifluoromethyl)­benzyl)­(methyl)­amino)­pyrimidin-4-yl)­pyrrolidin-2-yl)­methanol (5, (R-)-M04), and (S)-1-(6-((3,5-bis­(trifluoromethyl)­benzyl)­(methyl)­amino)­pyrimidin-4-yl)­pyrrolidin-2-yl)­methanol M04 compounds (Scheme 1, Supporting Information). The structural identity of M04 was confirmed by ¹H NMR spectroscopy (Supporting Information).

We next tested whether the racemic, (R)-, or (S)-M04 enantiomer was able to inhibit pSPAK accumulation in T/C-28a2 chondrocytes under acute osmotic stress. T/C-28a2 chondrocytes were treated with media supplemented with sorbitol (200 mM/500mOsm) for 5 min in the presence of DMSO control, 400 μM racemic M04, 400 μM (R)-M04, 400 μM (S)-M04, or 100 μM WNK463, and protein extracts were collected for immunoblot analyses (Figure ). Treatment of chondrocytes with either the racemic or (R-)-M04 demonstrated a reduction in pSPAK levels comparable to WNK463­(Figure ). The (S)-M04 enantiomer had a minor effect on the pSPAK levels (Figure ). These results confirm that the (R)-M04 enantiomer is the active form of M04.

8.

The (R)-enantiomer of M04 inhibits WNK2 activity and pSPAK in response to osmotic stress. Immunoblot analysis of pSPAK levels under sorbitol-induced hyperosmotic stress conditions (Sor, 200 mM sorbitol, 500 mOsm). Cells were treated for 5 min in the presence of DMSO control, 400 μM racemic M04, 400 μM (R)-M04, 400 μM (S)-M04, or 100 μM WNK463. GAPDH was used as a loading control.

Stability Evaluation of the Protein–Ligand Complex

To determine the stability of the WNK2–PKD–M04 interaction, we performed root mean square fluctuation (RMSF) analysis. We determined that Glu249, Lys244, and Lys251, whose RMSF values are 0.086, 0.094, and 0.0921 nm respectively, demonstrate strong structural stability and integrity without a ligand (Figure A). In the presence of M04, these values decreased slightly to 0.089, 0.126, and 0.087 nm, respectively, indicating further stabilization of these residues by the protein–ligand interaction without compromising the protein’s structural integrity (Figure A). A reduced flexibility of Glu249, Lys244, and Lys251 upon ligand binding suggests that these residues contribute directly to protein–ligand interaction, thereby promoting structural stability. RMSD analysis performed over a 100 ns simulation revealed minimal deviations from the initial conformation, confirming the sustained stability of the protein–ligand complex throughout the simulation (Figure B). Furthermore, the radius of gyration (Rg) analysis showed no significant changes in protein compactness, indicating that M04 binding did not disrupt the overall structural architecture of the protein (Figure C). In sum, these findings highlight M04’s potential as a strong and stable inhibitor with significant therapeutic potential, emphasizing its use as a highly reliable candidate for further functional studies.

9.

Analysis of WNK2–PKD protein dynamics in the presence and absence of M04. (A) RMSF plots showing the fluctuation of residues in WNK2–PKD and WNK2–PKD + M04. The X-axis represents amino acids number, and the Y-axis represents RMSF in nanometers (nm). (B) RMSD plots showing the time evolution of the root-mean-square deviation of the protein backbone atoms from the initial structure. The X-axis represents simulation time, and the Y-axis represents RMSD in nm. (C) Radius of gyration plots showing the overall size and compactness of the protein over time. The X-axis represents time, and the Y-axis represents radius of gyration in nm.

M04 Inhibition of WNK2 Reduces IL1β-Mediated Proinflammatory Gene Expression in Chondrocytes

IL1β expression in the synovial joint is associated with the OA phenotype and is used to induce a catabolic, injury-associated OA transcriptional response in cultured chondrocytes. − Our previous data indicated that the OA-associated WNK2 variants are hypermorphic and increased pro-OA gene expression in chondrocytes, including proinflammatory pathways. We tested if M04 inhibition of WNK2 can reduce IL1β-induced proinflammatory gene expression in T/C-28a2 chondrocytes and primary human chondrocytes. We cultured T/C-28a2 chondrocytes in hyperosmotic media (100 mM sorbitol), treated with IL1β for 24 h (10 ng/mL) in the presence of DMSO (control), 100 μM M04, or 10 μM WNK463, and we quantified changes in gene expression by RT-qPCR analysis. Treatment of T/C-28a2 chondrocytes with IL1β resulted in the upregulation of IL6 and CCL2 and downregulation of the main proteoglycan of cartilage, ACAN (Figure A). Treatment of IL1β-stimulated chondrocytes with M04 resulted in a significant reduction in IL6 and CCL2 expression back to control levels compared with IL1β-treated control chondrocytes. In addition to the reduction in proinflammatory gene expression, M04 treatment resulted in increased expression of the main collagen in cartilage, COL2A1, and prevented the IL1β-induced downregulation of ACAN (Figure A). We also treated chondrocytes with the pan-WNK1–4 inhibitor, WNK463. Similar to M04, WNK463 did not alter gene expression in control chondrocytes. In IL1β-stimulated chondrocytes, WNK463 treatment led to a significant upregulation of IL1β, IL6, and ACAN and reduced the expression of CCL2.

10.

M04 inhibition of WNK2 alters the transcriptional response of chondrocytes in response to IL1β treatment. RT-qPCR analysis of IL1β, IL6, CCL2, COL2A1, ACAN, FGF1, and MMP13 gene expression in (A) T/C-28a2 chondrocytes and (B) primary human chondrocytes cultured in hyperosmotic media (100 mM sorbitol) and treated with IL1β (10 ng/mL) for 24 h in the presence of DMSO, 100 μM M04, or 10 μM WNK463. Gene expression was normalized to ACTB, and relative expression was calculated using the ΔΔCq method. mRNA levels are reported as fold-change (mean ± SEM) relative to DMSO-treated Ctrl samples. Statistically significant differences of p ≤ 0.05 (*), p ≤ 0.01 (**), p ≤ 0.001 (***), and p ≤ 0.0001 (****) were determined by two-way ANOVA with Tukey’s multiple comparisons test; n = 3 or 4 biological replicates as indicated in the scatter plots.

We next wanted to determine if M04 could reduce IL1β-induced proinflammatory gene expression in primary human chondrocytes. Primary chondrocytes were cultured in hyperosmotic media (100 mM sorbitol), treated with IL1β for 24 h (10 ng/mL) in the presence of DMSO (control), 100 μM M04, or 100 μM WNK463, and we quantified changes in gene expression by RT-qPCR analysis. Similar to IL1β-stimulated T/C-28a2 chondrocytes, primary chondrocytes exposed to IL1β-upregulated proinflammatory gene expression compared with controls and M04 treatment led to a significant reduction in CCL2 and FGF1 expression (Figure B). M04 and WNK463 had no significant effect on control chondrocytes. WNK463 treatment of IL1β stimulated chondrocytes resulted in upregulation of many catabolic, proinflammatory factors including, IL1β, IL6, CCL2, MMP13, FGF1, and RUNX2 compared with IL1β-stimulated chondrocytes or control chondrocytes (Figure B). These data indicate that M04 inhibition of WNK2 reduces IL1β-mediated catabolic proinflammatory gene expression in chondrocytes while also upregulating anabolic gene expression. Our data also suggests that some WNK activity is needed for this response as inhibition of all WNK activity in IL1β stimulated chondrocytes amplifies the catabolic proinflammatory response.

M04 Inhibits Proinflammatory Pathways and Promotes Growth and Repair Pathways Compared with a Pan-WNK Inhibitor in IL1β-Treated Primary Human Chondrocytes

Our RT-qPCR analysis suggested that M04 and WNK463 elicit different transcriptional responses to IL1β stimulation, and that some WNK activity is needed for the anabolic response (Figure ). This is consistent with our previous work examining the transcriptional response to osmotic stress in WNK2 null T/C-28a2 chondrocytes. We hypothesized that partial WNK pathway inhibition using M04 would be a more effective treatment strategy than complete pathway inhibition with WNK463. To determine the pathways that differ between M04 and WNK463, we treated primary human chondrocytes with and without IL1β in the presence of M04 (100 μM) or WNK463 (10 μM) and performed RNaseq and pathway enrichment analysis. We first examined the transcriptional response to M04 vs WNK463 in non-IL1β-stimulated primary human chondrocytes (Figure A,B). M04 treatment led to the upregulation of genes involved in the remodeling response (MMP2 and MMP13) and downregulation of genes regulating BMP signaling (GREM2 and BMP4) (Figure A). Compared with WNK463, M04 led to the activation of pathways involved in cellular growth, inhibition of Hedgehog signaling, and the immune response while downregulating those involved in autophagy, the response to starvation, and chemical stress (Figure B). We next wanted to determine the genes and pathways regulated by M04 in the presence of IL1 β. M04 led to a significant reduction of pro OA-associated gene expression (BMP2, BMP4, NFKB1, MMP13, and FN1) and upregulation of genes involved in maintaining joint homeostasis (SLC1A5 and GREM1) (Figure C). Compared with WNK463, even in the presence of IL1β MO4 led to activation of pathways involved in growth, TGFβ signaling, the circadian rhythm, and ion (potassium) transport. M04 downregulated several major proinflammatory signaling pathways, including TLR, interferon α/β/γ, interleukin, and RIPK-mediated activation of NFκB (Figure D).

11.

M04 inhibits proinflammatory pathways and promotes growth and repair pathways compared with WNK463 in IL1β-treated primary human chondrocytes. (A, C) Comparative analysis of RNaseq performed on RNA isolated from primary human chondrocytes with and without IL1β (10 ng/mL) treatment in the presence of M04 (100 μM) or WNK463 (10 μM) for 24 h. The volcano plots indicate genes significantly upregulated (red) or downregulated (blue) in M04 vs WNK463-treated primary human chondrocytes in the absence (A) or presence (C) of IL1β. (B, D) Bubble plots illustrate the top pathways identified from the differentially expressed genes in A and C. In the bubble plots, the x-axis represents the gene ratio (number of genes in a pathway significantly altered by each treatment relative to the total number of genes in that pathway), while the y-axis displays the enriched pathway. Each bubble represents a specific pathway, with the color of the bubble indicating the positive (increased pathway gene expression, red) or negative (decrease pathway gene expression, blue) Z score. Darker colors indicate higher significance. The Z score is indicated next to each bubble. The size of the bubble indicates the –Log (p value) and corresponds to the level of significant enrichment. n = 4 for each condition.

Together, our molecular analyses (Figures and ) indicate that M04 and WNK463 produce different biological responses under control and IL1β stimulation conditions in primary human chondrocytes. In comparison with complete WNK pathway inhibition (using WNK463) M04, which reduces WNK pathway activity, is associated with stronger cellular growth pathways and activation of anabolic pathways while reducing several major OA-associated proinflammatory pathways. This data supports our hypothesis that a reduction rather than complete inhibition may be an effective therapeutic strategy for the treatment of OA.

Discussion and Conclusions

There are no disease-modifying drugs available for OA. A lack of targeted therapies highlights the need to identify genetic pathways underlying OA progression and develop effective treatments. − Identification of disease causing genes provides rational targets for effective drug development. Drugs that target disease causing proteins more than doubles the success rate in clinical trials independent of the source of the mutation (e.g., rare vs common alleles). − The genetic analysis of families with dominant forms of OA allows us to identify highly penetrant coding alleles that have a determinate effect on promoting OA, independent of prior biases on how protein activity may affect the OA phenotype or the tissue specific requirement of the mutant alleles. ,,− ,−

Our previous work identified 4 independent OA families with rare coding mutations in WNK2. We demonstrated that WNK2 is a key mediator of the chondrocyte response to hyperosmotic stress, and the OA-associated alleles are hypermorphic, which regulate the gene expression of key OA-associated pathways. These data suggest that WNK2 inhibition may be a therapeutic target in OA. Currently, there are no chemical inhibitors that specifically target WNK2.

Our computational approach effectively combined pharmacophore modeling, virtual screening, molecular dynamics, and experimental validation to identify potentially selective WNK2 inhibitors. In this study, we identified M04, (1-(6-((3,5-bis­(trifluoromethyl)­benzyl)­(methyl)­amino)­pyrimidin-4-yl)­pyrrolidin-2-yl)­methanol, as an inhibitor of the WNK2 kinase domain. M04’s interaction with WNK2 was confirmed through its ability to reduce SPAK phosphorylation in cell-based assays, indicating effective target engagement (Figure C). Notably, M04 exhibited an IC₅₀ of 416 μM in chondrocyte viability assays, suggesting low cytotoxicity. In vitro kinase assays indicated that M04 preferentially inhibited WNK2, but it also affected WNK3 and WNK4 activity (Figure ). We are cautious not to overinterpret this experiment as we were unable to use full-length WNK proteins in this assay. We used recombinant WNK proteins consisting of the kinase domain, which fails to accurately represent the complex 3D structure of WNKs.

Our previous work using a pan-WNK inhibitor (WNK463) and WNK2 null chondrocytes indicated that WNK2 is largely responsible for SPAK phosphorylation at physiological levels of hyperosmotic stress. Consistent with this data, inhibition of WNK2 with M04 was effective in suppressing OA-associated IL1β-mediated proinflammatory gene expression in cultured and primary human chondrocytes M04. Based on these results, M04 inhibition of WNK2 could be a novel therapeutic approach for limiting the onset or progression of OA.

Our study had several limitations. We restricted our drug screen to the WNK2 kinase domain since we were unable to generate a high-confidence model of full-length WNK2 (average pLDDT 44), thereby potentially missing other druggable sites. For example, the human OA-associated mutations are in the disordered C-terminus, a region of the WNK proteins that is necessary for phase separation and restoring cell volume. , Targeting this region of the protein may allow for novel strategies to modulate WNK2 activity independent of its kinase domain. Further optimization of the M04 structure may be needed for in vivo efficacy. Our studies were limited to in vitro analyses, and we did not test M04 in an animal model of OA. We are currently generating mice harboring loss-of-function (null) and gain-of-function (using the human mutations) Wnk2 alleles. These models will allow us to precisely define the role of WNK2 in vivo and test whether M04 inhibition of WNK2 protects against injury or age-associated OA.

In conclusion, the present study shows the potential for targeting WNK2 for OA therapy and the effectiveness of integrating human genetics with computational and experimental approaches to drug discovery.

Experimental Section

Generation of the WNK2 3D Structure

AlphaFold2 was used to predict the WNK2 (NP_001269323) 3D structure and its protein kinase domain (WNK2–PKD) (amino acids 195–454) (version 2.3.2). To assess the reliability of the predicted structures, we analyzed the Predicted Local Distance Difference Test (pLDDT) scores generated by AlphaFold2. Protein preparation was conducted using the Molecular Operating Environment (MOE) (version 2024.06) to ensure that no residues or atoms were missing, and necessary adjustments were made. As part of this comprehensive preparation process, hydrogen atoms were added, water molecules were removed, 3D protonation was performed, and gas-phase charges were assigned using the MMFF94 force field, followed by systematic minimization of the structure.

Preparation of Ligands

Ligands were sourced from the ZINC database (https://zinc.docking.org), focusing on commercially available compounds that met Christopher Lipinski’s Rule of Five (RO5). A total of 4.8 million compounds were downloaded and prepared for docking. The ligand preparation was done with Open Babel (version 3.1.0), a chemical toolbox that allows manipulation of molecules. For each ligand, hydrogen atoms were added in order to ensure correct protonation states, followed by the assignment of atomic charges to illustrate electrostatic interactions, and energy minimization to simplify geometries and minimize steric clashes. Additionally, solvation was performed to simulate a realistic environment for molecular interactions. The compounds were saved to SDF format and converted to appropriate formats for the docking program, such as PDBQT for autodock and .mol2 for molecular dynamic simulations. In preparing the ligands, standard criteria were followed, ensuring reproducibility and accuracy of the prepared structures.

WNK2 Binding Pocket Identification

The potential binding pockets of WNK2–PKD were identified using MOE’s site finder program. Protein surfaces are scanned by this algorithm to identify potential cavities based on their size, shape, residue composition, and chemical properties. This allowed us to understand which sites are potentially more druggable by calculating the “Protein–Ligand Binding” score (PLB), the number of hydrophobic alpha spheres (Hyd), and total number of alpha spheres (Side). Each candidate pocket was visually inspected and computationally evaluated to ensure it was open and accessible for small-molecule binding. Pymol (Version 3.0.3) was used to confirm that the cavities identified were not occluded and could accommodate ligands by calculating solvent-accessible surface area (SASA). An open and accessible binding pocket (BP-4) was discovered in WNK2, distinguished by two unique residues, Glu245 and Glu249, not found in other WNK family members, and was utilized for the screening process.

Molecular Docking Protocol

Molecular docking was performed in two sequential stages. In the first stage, a pharmacophore search was conducted using the MOE tool. The pharmacophore model was designed with specific interaction criteria: the δ-carbon (Cδ1 and Cδ2) and β-carbon of Leu116 were selected for likely hydrophobic interactions. In Glu245 and Glu249 residues, the Cδ carboxylate and hydroxyl groups were targeted for both anionic and hydrogen bond interactions, while the β-carbon and γ-carbon atoms of these residues were included for additional hydrogen bonding potential. A minimum interaction distance of 3 to 5 Å was defined, along with volume exclusion criteria restricting inhibitors to bind exclusively within BP4, with a 5 Å exclusion zone (Table S2). In total, 100 K–150 K compounds determined to match the pharmacophore model met at least five criteria. As a second step, the 100 K–150 K compounds identified from the pharmacophore search were docked into BP4 using the same selection criteria previously defined using induced fit refinement. For each compound, a minimum of 100 docking poses was generated to ensure accurate binding predictions. The compounds with the highest binding energies (ΔG), ranging from −3 to −10 kcal/mol, were visualized using MOE. A total of 53 compounds were selected for further analysis (Table S2). The selected 53 compounds were redocked using AutoDock Vina (version 4.2.6). In AutoDock Vina, the binding free energy results aligned well with MOE, confirming their accuracy.

T/C-28a2 Chondrocyte Cell Culture

To test the inhibitors of WNK2, we utilized the human chondrocyte cell line T/C-28a2 (Millipore Sigma). Previous studies in our lab have shown that T/C-28a2 cells express WNK2. The T/C-28a2 was cultured in Dulbecco’s modified Eagle medium (DMEM) (Corning Life Sciences) supplemented with 10% fetal bovine serum (FBS; Avantor), 4.5 g/L glucose, 1 mM sodium pyruvate, and 100 U/mL penicillin/streptomycin in a 5% CO₂ humidified atmosphere incubator at 37 °C. Cell passage was performed on 70% confluent cells by trypsinization with 0.25% trypsin–EDTA (Thermo Fisher).

Primary Human Chondrocyte Cell Culture

Primary human chondrocytes (CELLvo HC-a (passage 1), HCA-2–100–000, Lot # HCA-20-13.1) were obtained from StemBioSys, Inc. CELLvo chondrocytes were isolated from cadaveric knee articular cartilage of a healthy 28-year-old female African American donor. CELLvo HC-a chondrocytes were cultured on CELLvo ChondroMatrix-coated plasticware (StemBioSys, Inc.) in low-glucose Dulbecco’s modified Eagle medium (DMEM) (Corning) supplemented with 1 g/L glucose, 1 mM sodium pyruvate, 4 mM l-glutamine, 15% fetal bovine serum, and 100 U/mL penicillin/streptomycin in a 5% CO₂ humidified atmosphere incubator at 37 °C.

Treatment of Primary Human Chondrocytes for RNaseq Analysis

Primary human chondrocytes were cultured as described above. Cells were treated with DMSO (control), IL1β (10 ng/mL), M04 (100 μM), or WNK463 (10 μM) for 24 h, and the cells were collected for RNA isolation (as described below) and RNaseq analysis.

Cell Viability Measurement (XTT Assay)

To evaluate cell viability at various inhibitor concentrations, an XTT assay (Thermo Fisher) was performed according to the manufacturer’s protocol. Human chondrocyte cells (T/C-28a2) were cultured as described above and exposed to different concentrations of inhibitors (ranging from 0.1 μM to 100 μM in DMSO) and for 24 h at 37 °C. DMSO-treated cells were treated as controls. After incubation, XTT reagent was prepared by mixing 6 mL of XTT reagent with 1 mL of Electron Coupling Reagent. The working solution (70 μL) was added to each well and incubated for an additional 4 h at 37 °C. Absorbance was measured at both 450 and 660 nm. The specific absorbance, which reflects cell viability, was then calculated using the following formula. Specific absorbance = [A450 nm (Test) – A450 nm­(Blank)] – A660 nm­(Test)]. The calculated specific absorbance values for each inhibitor concentration were normalized to the untreated control to obtain percent viability. These data were then plotted against the inhibitor concentration. Nonlinear regression analysis was performed in GraphPad Prism using a four-parameter variable-slope logistic model of inhibitor concentration versus normalized response to determine the half-maximal inhibitory concentration (IC₅₀).

Preparation of Cell Lysates and Immunoblot Analysis

To determine whether the selected inhibitors affect WNK2 phosphorylation of SPAK, the immediate target of WNK2, we measured SPAK phosphorylation by performing immunoblot analysis. Chondrocytes were cultured as described above, and the cells were subjected to osmotic stress as previously described. In brief, once the chondrocytes reached 70% confluence, the culture medium was replaced with fresh complete medium containing various concentrations of inhibitors (ranging from 0.1 μM to 100 μM in DMSO) for 15 min. After this, the media were replaced with the same concentrations of inhibitors, either in isotonic medium (300 mOsm) or with sorbitol (200 mM/500 mOsm). The chondrocytes were incubated for an additional 10 min in either isotonic or stress-inducing medium, then harvested, and pelleted at 4 °C. The total exposure to osmotic stress was 15 min. Proteins were isolated using RIPA buffer supplemented with protease and phosphatase inhibitors. The isolated proteins were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) using a 10% acrylamide gel and transferred to a PVDF (0.45 μm) membrane. The membrane was blocked with 3% BSA in Tris-buffered saline with 0.1% Tween-20 (TBST) overnight at 4 °C. The membranes were incubated overnight at 4 °C with anti-pSPAK (EMD Millipore 07-2273). Blots were washed 3X with TBST and then incubated with secondary antibodies conjugated with horseradish peroxidase (anti-mouse/rabbit IgG HRP, Thermo Fisher Scientific). Signal detection was achieved through enhanced chemiluminescence (SuperSignal West Pico Chemiluminescent Substrate 34080, Thermo Fisher Scientific). GAPDH (Santa Cruz Biotechnology 365062) served as the loading control.

Protein Kinase Assay

Kinase activity was assessed using the ADP-Glo Kinase Assay Kit (Promega), following the manufacturer’s instructions with minor optimizations. Recombinant kinase domains of human WNK1–4 proteins were obtained from Abcam. The reactions were performed with 100 μM ATP, 10 μM SPAK peptide substrate, and recombinant proteins: 50 ng of WNK1, 100 ng of WNK2, 50 ng of WNK3, or 100 ng of WNK4. Test compounds or DMSO (vehicle control) were added at indicated M04 inhibitor concentrations (1, 10, 50, 100, and 200 μM) or the pan-WNK1–4 inhibitor WNK463 (200 μM) as a positive control (n = 4 for each reaction). Luminescence (RLURelative Light Units) was measured using a plate luminometer, and background signals were subtracted from all readings. Statistical analysis was performed using ordinary one-way ANOVA with DMSO as the control, followed by Dunnett’s multiple-comparisons test.

Molecular Dynamics Simulation

The stability and interactions of the selected molecule M04 were investigated through molecular dynamics (MD) simulations using GROMACS (version 2022.3) with the charmm36-jul2022 force field. We started with a cleaned WNK2–PKD protein structure, which was augmented with hydrogen atoms and underwent energy minimization. The protein topology was prepared using the CHARMM36 force field. The cleaned ligands were then formatted for topology using the CGenFF web server using the CHARMM36 force field, which generated the necessary topology files. Then, protein–ligand complexes were constructed and placed in a dodecahedron box filled with water, resulting in a charged protein-solvated system. Ions were introduced to neutralize the system. Energy minimization was conducted to alleviate steric clashes, followed by a two-phase equilibration: NVT (constant temperature and volume) for 20 ns and NPT (constant pressure and temperature) for an additional 20 ns. After equilibration, 100 ns of production MD simulations was performed on the equilibrated systems. All evaluations, including stability and interaction assessments, were performed with the GROMACS tools. Root-mean-square fluctuation (RMSF), a measure of protein–ligand interaction stability, was analyzed using GROMACS both in the presence and absence of selected inhibitors. The fluctuations in residue indicate flexibility, with a higher RMSF indicating increased interaction potential and a lower RMSF indicating decreased interaction potential. Root-mean-square deviation (RMSD), which measures the deviation of the protein or complex from its initial conformation, and radius of gyration (Rg), which measure the compactness of the protein structure by measuring the distribution of atomic mass around the center of mass, were calculated during the molecular dynamics simulation to evaluate the structural stability and binding affinity of the system over time. All simulations were performed in triplicate to ensure the reproducibility and reliability of the results.

Quantitative Reverse Transcription PCR (qRT-PCR)

T/C-28a2 and primary human chondrocytes were lysed using Trizol (Thermo Fisher Scientific), and total RNA was collected using the Direct-zol RNA Miniprep Kit (Zymo Research). RNA quantity and purity was assessed using the Nanodrop 2000 (Thermo Fisher). A total of 50 ng of RNA was combined with PrimeTime Master Mix (IDT) and gene-specific primers (IDT; see below), and qRT-PCR was performed according to the manufacture’s protocol. Gene expression was normalized to ACTB, and relative expression was calculated using the ΔΔCq method. mRNA levels are reported as fold-change (mean with 95% confidence interval) relative to each control condition.

	gene name	assay ID
1	ACTB	Hs.PT.39a.22214847
2	ACAN	Hs.PT.56a.742783
3	IL1β	Hs.PT.58.1518186
4	CCL2	Hs.PT.58.45467977
5	IL6	Hs.PT.58.40226675
6	COL2A1	Hs.PT.58.4107778
7	MMP13	Hs.PT.58.40735012
8	FGF1	Hs.PT.58.24337365
9	RUNX2	Hs.PT.56a.19568141

RNA Sequencing (RNaseq) and Data Analysis

The RNaseq workflow and data analysis were performed as previously described. Briefly, following drug treatment, total RNA was isolated from primary human chondrocytes and processed for bulk RNA sequencing. Sequencing reads were assessed for quality, aligned to the human reference genome, and quantified at the gene level using DESeq2. Differential gene expression analysis was performed between treatment groups using fold-change and adjusted p-value thresholds. Pathway enrichment and canonical pathway analyses were performed to identify the biological processes altered by treatment. Differential gene expression and pathway-level results were visualized by using volcano plots and pathway enrichment plots.

Statistical Analysis

Nonlinear regression was used to calculate the half-maximal inhibitory concentration (IC₅₀) values, fitting inhibitor concentration against normalized cellular response using a four-parameter logistic model in GraphPad Prism (version 10.3.1). Each experiment included at least three independent replicates, with results reported as mean ± standard deviation (SD).

qPCR analysis of gene expression in T/C-28a2 chondrocytes and primary human chondrocytes: Gene expression was normalized to ACTB and relative expression was calculated using the ΔΔCq method. mRNA levels are reported as fold-change (mean ± SEM) relative to DMSO-treated Ctrl samples. Statistically significant differences of p ≤ 0.05 (*), p ≤ 0.01 (**), p ≤ 0.001 (***), and p ≤ 0.0001 (****) were determined by two-way ANOVA with Tukey’s multiple comparisons test, n = 3 or 4 biological replicates as indicated in the scatter plots.

Glossary

Abbreviations

WNK

With-No-Lysine (K) protein kinase

OA

osteoarthritis

SPAK

STE20/SPS1-related proline-alanine-rich kinase

OSR1

oxidative stress-responsive kinase 1

PKD

protein kinase domain

RMSF

root mean square fluctuation

RNaseq

RNA sequencing

Data is available upon request. Restrictions apply to the availability of the genomic and medical data analyzed in this study to preserve patient confidentiality. Access to UPDB data is controlled through the Resource for Genetic and Epidemiologic Research.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsomega.6c04569.

• Sequence analysis of WNK family proteins and identification of target residues for WNK2-specific inhibitor design; unmodified immunoblots used in Figure ; enantioselective synthesis of M04; in silico ADMET and drug-likeness prediction for M04; details of the top 53 inhibitors (M01–M53) including their compound IDs (lab assigned), Mcule or zinc IDs, SMILES notation, and rule of five violations; NMR data for M04; and molecular docking and molecular dynamics analyses for WNK463 and M04 (PDF)

M.J.J. and S.R.V. conceived and designed the study, conducted experiments, analyzed data, and wrote the manuscript. S.B. and A.S. performed the synthesis of M04, analyzed data, and provided feedback on the manuscript. Y.M. conducted experiments and analyzed data.

The authors declare the following competing financial interest(s): SRV and MJJ have filed a U.S. Patent Application (No. 19/672,326) titled Compounds That Inhibit WNK2 Activity And Methods For Treating Osteoarthritis.