A small-molecule TNIK inhibitor targets fibrosis in preclinical and clinical models

The study addresses the lack of effective anti-fibrotic therapies for idiopathic pulmonary fibrosis (IPF) and other fibrotic diseases such as chronic kidney disease (CKD). IPF is a progressive, lethal disease with median survival of 2–3 years and limited approved treatments (nintedanib and pirfenidone), which only slow disease progression and have adverse effects. Fibrosis is driven by dysregulated chronic inflammation and TGF-β–mediated myofibroblast differentiation. The research question is whether artificial intelligence-driven target discovery and molecular design can rapidly identify a novel, druggable anti-fibrotic target and generate a selective small-molecule inhibitor with efficacy across organs. The purpose is to leverage multiomics- and text-based AI (PandaOmics) to prioritize targets and AI-driven structure-based design (Chemistry42) to deliver a clinical-stage TNIK inhibitor, addressing an urgent clinical need in fibrosis while accelerating the discovery timeline.

Background highlights include: (1) IPF epidemiology: prevalent in patients over 60, with rising incidence and substantial healthcare burden; untreated IPF shows variable course with poor survival. (2) Current therapies: nintedanib (inhibits FGFR, PDGFR, VEGFR receptor tyrosine kinases) and pirfenidone (suppresses TGF-β expression) are guideline-recommended but not curative. (3) Fibrosis biology: excessive ECM deposition by myofibroblasts, with TGF-β central to EMT/FMT and fibrosis, and additional roles for WNT, Hippo/YAP-TAZ, JNK, and NF-κB pathways. (4) Renal fibrosis: manifests as tubulointerstitial fibrosis, driven by ischemia/inflammation and elevated TGF-β signaling, with no approved anti-fibrotic drugs for CKD. (5) Prior AI applications have successfully identified targets/biomarkers in aging and cancer. The role of serine–threonine kinases like TNIK in IPF has been largely unexplored compared to tyrosine kinases, motivating investigation of TNIK as a novel target.

• AI target discovery: Used PandaOmics to integrate multiomics datasets from IPF patient lung tissues, biological network analysis, and literature/text data. Applied scoring and filters (protein/receptor kinase scenario, novelty, small-molecule druggability, protein class) to generate a ranked target list, with TNIK scoring number 1. Transparency analyses linked TNIK to focal adhesion, myofibroblast differentiation, mesenchymal migration, and to IPF-associated genes via causal inference. Single-cell RNA-seq (GSE136831) showed elevated TNIK in cytotoxic T cells, myofibroblasts, and club cells in IPF; virtual TNIK knockout (scTenifoldKnk with MCODE enrichment) suggested activation of Hippo signaling and downregulation of YAP-TAZ.
• AI drug design: Leveraged TNIK kinase-domain crystal structures and Chemistry42 structure-based workflow. Targeted ATP-binding site with two-point pharmacophore (hinge Cys108-NH hydrogen bond; hydrophobic interaction near gatekeeper Met105/back pocket). Generated, scored, and synthesized compounds optimized for physicochemical properties, novelty, and synthetic accessibility. Lead optimization for ADME properties resulted in INS018_055 (WO2022179528A1) predicted to deeply occupy the TNIK back pocket while forming a hinge hydrogen bond.
• Biophysics and selectivity: Surface plasmon resonance measured binding kinetics (INS018_055 versus NCB-0846, KY-05009). KinaseProfiler screened 430 kinases at 10 µM, with IC50 follow-up on 42 hits to determine selectivity and potency, focusing on fibrosis-relevant kinases.
• In vitro functional assays: EMT/FMT assays in MRC-5, primary human IPF and healthy lung fibroblasts, and primary bronchial epithelial cells. Western blots in A549 for E-cadherin, N-cadherin, p-SMAD2/3, β-catenin subcellular localization, NF-κB p65 phosphorylation. shRNA knockdown of TNIK in A549 to compare phenocopy. Bulk RNA-seq with GO/KEGG enrichment to assess pathway modulation.
• In vivo lung fibrosis (mouse): Bleomycin-induced lung fibrosis in C57BL/6 males; oral INS018_055 dosing (3, 10, 30 mg/kg BID) versus nintedanib (60 mg/kg QD). Lung function (Penh) at day 21; histopathology (Masson’s trichrome, modified Ashcroft score), IHC for collagen I and α-SMA; BALF cytokines/cell counts; inflammation by H&E. Combination studies with pirfenidone at suboptimal and maximum therapeutic doses.
• Inhalation lung fibrosis (rat): Bleomycin-induced lung fibrosis in Sprague Dawley rats; aerosolized INS018_055 (0.1, 0.3, 1, 6 mg/ml; 30 min/day × 21 d) with delivered dose quantification; lung and plasma exposure; lung function (FVC, compliance, resistance); histopathology.
• Kidney fibrosis: HK-2 cell α-SMA assay and cytotoxicity; unilateral ureteral obstruction (UUO) model in mice with INS018_055 (3, 10, 30 mg/kg BID) versus ALK5 inhibitor SB525334 (100 mg/kg QD). Endpoints: Sirius Red area, kidney hydroxyproline, collagen I IHC score.
• Skin fibrosis: NHDF assays for α-SMA, fibronectin, procollagen I; rat bleomycin-induced skin thickening with topical INS018_055 (0.05%, 0.15%, 0.45%) versus vehicle/rapamycin; hydroxyproline and collagen assays.
• Clinical PK and safety: Phase 0 microdose IV PK in healthy participants (Australia). Phase I randomized, double-blind, placebo-controlled trials in healthy volunteers: New Zealand SAD (10–120 mg) including fed/fasted at 90 mg; MAD QD 7 days (30, 60, 120 mg). China SAD (30, 60, 120 mg) and MAD BID 7 days (30, 60, 90 mg). PK endpoints (Cmax, Tmax, AUC, t1/2, accumulation) and safety/tolerability; dietary effect at 90 mg cohort.
• Statistics: Appropriate ANOVA with post hoc tests (Šídák’s, Dunn’s), Fisher’s LSD, Bonferroni, Welch’s t-tests for blots; enrichment analysis via gseapy.enrichr with Benjamini–Hochberg correction.

• Target identification: TNIK ranked top anti-fibrotic target by PandaOmics; linked to WNT, TGF-β/SMAD, Hippo/YAP-TAZ, JNK, and NF-κB pathways. scRNA-seq showed higher TNIK expression in IPF myofibroblasts, cytotoxic T cells, and club cells.
• Binding and selectivity: INS018_055 exhibited strong TNIK binding (SPR KD = 4.32 nM) versus NCB-0846 (21.6 nM) and KY-05009 (137 nM). Kinase profiling showed TNIK as the most inhibited target (IC50 = 31 nM) with selectivity improvements over NCB-0846 among reported kinases; additional affinity for fibrosis-related kinases (e.g., ALK4, TGFBR1, DDR1) at higher concentrations.
• In vitro anti-fibrotic effects: In MRC-5 cells, INS018_055 reduced TGF-β-induced α-SMA with IC50 ≈ 27.14 nM and CC50 = 84.3 µM (low cytotoxicity). In primary IPF lung fibroblasts, IC50 for α-SMA inhibition were 50, 79, 63 nM; in healthy fibroblasts, 79, 200, 320 nM. In primary bronchial epithelial cells, INS018_055 inhibited FN1 with IC50 of 250–400 nM (IPF donors), outperforming nintedanib (1,600–7,900 nM), and 63–500 nM in healthy donors. In A549 EMT models, INS018_055 dose-dependently increased E-cadherin, decreased N-cadherin and p-SMAD2/3, reduced chromatin-associated β-catenin, and suppressed p-p65 induced by TGF-β/TNF-α. TNIK knockdown phenocopied these effects. RNA-seq showed reversal of TGF-β-induced ECM, focal adhesion, and collagen pathways; Hippo signaling modulation with downregulation of fibrosis-related targets (PAI-1, SMAD7, CTGF, GLI2, PUMA).
• In vivo lung fibrosis (mouse): Oral INS018_055 for 2 weeks significantly improved lung function (reduced Penh) to levels comparable to nintedanib. Histology showed >50% reduction in fibrotic area at 3 mg/kg BID and >75% at 10 and 30 mg/kg BID; decreased modified Ashcroft scores; reduced α-SMA and collagen I IHC. Inflammation and BALF cytokines (IL-6, IL-1β) and myeloid cell counts were reduced.
• Combination therapy: Suboptimal-dose INS018_055 plus pirfenidone improved Penh and reduced fibrosis markers more than either agent alone. With therapeutic-dose pirfenidone (200 mg/kg BID), combining INS018_055 (up to 10 mg/kg BID) achieved complete prevention of clinical symptoms.
• Inhalation (rat): Aerosolized INS018_055 achieved ~50-fold higher lung versus plasma exposure. All inhaled doses significantly improved FVC, increased compliance, and reduced airway resistance; higher doses (1 and 6 mg/ml) markedly reduced fibrosis and inflammation histologically.
• Kidney fibrosis: In HK-2 cells, α-SMA inhibition IC50 = 0.104 µM with CC50 = 37.37 µM. In UUO mice, INS018_055 (3–30 mg/kg BID) significantly reduced Sirius Red-positive area, kidney hydroxyproline, and collagen I IHC score, comparable to ALK5 inhibitor SB525334.
• Skin fibrosis: In NHDFs, INS018_055 inhibited α-SMA (IC50 ~25 nM), fibronectin (~135 nM), and procollagen I (~232 nM). Topical INS018_055 (0.05–0.45%) reduced hydroxyproline and collagen in bleomycin-induced dermal fibrosis.
• Clinical PK and safety: Phase I (New Zealand) SAD showed rapid absorption (median Tmax 1.00–1.53 h), multiphasic decline with t1/2 7.42–9.74 h, dose-proportional Cmax and AUC except Cmax plateau at 90 mg; food reduced Cmax (270 to 169 ng/ml) and AUC modestly and delayed Tmax (1.01 to 3.00 h). MAD (30–120 mg QD, 7 d) showed Tmax 1–2 h, t1/2 9.36–11.9 h, no accumulation (day 7/day 1 ratios: Cmax 0.739–0.905; AUC0-τ 0.996–1.05). China Phase I showed comparable SAD/MAD PK; slight supra-proportional exposure from 60 to 120 mg in SAD. Across studies, adverse events were mild and resolved; overall safe and well tolerated with good oral bioavailability.

The AI-enabled pipeline rapidly identified TNIK as a previously underexplored serine–threonine kinase target in fibrosis and delivered a selective, potent small-molecule inhibitor, INS018_055, within approximately 18 months from target discovery to preclinical candidate. Mechanistically, TNIK intersects key profibrotic and inflammatory pathways (TGF-β/SMAD, WNT/β-catenin, Hippo/YAP-TAZ, NF-κB), aligning with observed reversal of EMT/FMT signatures, reduced β-catenin chromatin association, and suppression of p-SMAD2/3 and p-p65. The compound demonstrated robust anti-fibrotic and anti-inflammatory efficacy across lung, kidney, and skin models, improved lung function in bleomycin-induced fibrosis, and exhibited synergy with pirfenidone. Compared to earlier TNIK inhibitors (e.g., NCB-0846), INS018_055 shows improved selectivity, reduced cytotoxicity at therapeutic exposures, and favorable PK/safety in healthy volunteers. These findings support TNIK inhibition as a cross-organ anti-fibrotic strategy and illustrate the promise of AI-driven target discovery and medicinal chemistry to accelerate development. Nevertheless, translating preclinical efficacy and Phase I safety to clinical benefit in IPF requires validation in patient populations.

This study demonstrates that generative AI platforms can uncover novel anti-fibrotic targets and efficiently design drug candidates. TNIK was prioritized by PandaOmics and validated functionally; INS018_055, a selective TNIK inhibitor, showed potent anti-fibrotic and anti-inflammatory effects in vitro and in vivo across lung, kidney, and skin models, improved lung function in mice, and was safe with favorable PK in two Phase I trials. The work underscores the potential of TNIK inhibition as a therapeutic approach for IPF and other fibrotic conditions. Ongoing Phase II trials in IPF (NCT05975983, NCT05938920) will determine clinical efficacy. Future research should assess long-term safety, efficacy across fibrotic organs, optimal delivery routes (including inhalation), and potential combination strategies with existing antifibrotics.

• Clinical validation is pending: efficacy has not yet been demonstrated in IPF patients; Phase II/III trials are needed to confirm therapeutic benefit.
• Preclinical model generalizability: Bleomycin-induced fibrosis and UUO models may not fully recapitulate human disease heterogeneity and chronicity.
• Comparative selectivity data: Cross-study comparisons (e.g., versus NCB-0846) are constrained by methodological differences; broader kinome profiling and off-target assessments in human tissues are warranted.
• AI target discovery constraints: The authors note limitations in multiomics-based discovery, including data complexity, batch effects, and reliance on available datasets and literature trends.
• Phase I populations: Safety and PK were characterized in healthy volunteers; safety in patients with comorbidities and long-term dosing remain to be established.