A RIPK2 activity signature in prostate cancer: Modulation by RIPK2 inhibition and clinical association Authors: Ahmed M. Elgehama, Qian Yang, Jaceline Gislaine Pires Sanches, Zaoke He, Lauryn Ruegg, Sungyong You, Wei Yang Published: 2026 ABSTRACT Highlights•This study defines an 8-gene RIPK2 activity signature in prostate cancer.•The signature dynamically responds to RIPK2 inhibition in vitro and in vivo.•The signature shows stronger associations with metastasis and survival than RIPK2 mRNA.•RIPK2-dependent transcriptional regulation involves c-Myc and KDM5A. Introduction Receptor-interacting protein kinase 2 (RIPK2) is a Ser/Thr/Tyr kinase best known for its role in inflammation and innate immunity, primarily through the NOD/RIPK2/NF-κB signaling pathway [1]. More recently, RIPK2 has been implicated in cancer, where its overexpression across multiple tumor types correlates with disease progression and adverse clinical outcomes [2,3]. In response, many RIPK2 inhibitors have been developed, and several have advanced into early-stage clinical trials [1,4]. Prostate cancer (PC) remains a major health burden, particularly in the United States and other developed countries [5,6]. With an aging global population, PC-related mortality is projected to approach 700,000 deaths annually by 2040 [7]. Most PC-related deaths result from metastatic disease, which is associated with poor survival and substantial morbidity, underscoring the need for novel therapeutic strategies targeting metastatic progression. Our recent work identified RIPK2 as a key driver of PC progression and metastasis through a non-canonical RIPK2/MKK7/c-Myc phosphorylation pathway [2]. Genetic or pharmacologic targeting of RIPK2 suppressed PC metastasis in vitro and in vivo, supporting its potential as a therapeutic target in advanced PC [2]. However, clinical translation is limited by the lack of reliable biomarkers to assess RIPK2 signaling activity, complicating both patient selection and pharmacodynamic monitoring for anti-RIPK2 therapeutic strategies. Although RIPK2 protein expression has been proposed as a potential biomarker, protein abundance does not necessarily reflect kinase activity. RIPK2 signaling is regulated by post-translational modifications, protein-protein interactions, and subcellular localization, rather than solely by protein abundance. Reliance on protein levels alone may therefore misclassify tumors with active RIPK2 signaling. In contrast, a RIPK2-regulated gene signature would capture downstream transcriptional outputs of pathway activation and provide a functional readout of RIPK2 activity. Although RIPK2 potently activates c-Myc in PC cells [2], MYC signatures lack specificity, as c-Myc activity is modulated by diverse signaling pathways. Here, we address this critical gap by identifying a RIPK2-regulated gene signature through RNA sequencing of PC cells with genetic RIPK2 ablation, followed by validation with reverse transcription quantitative PCR (RT-qPCR). We further evaluate the signature’s responsiveness to RIPK2 inhibition in vitro and in vivo. We then examine its association with metastasis and survival outcomes in PC and compare its performance with RIPK2 mRNA expression alone. Finally, we explore the potential involvement of c-Myc and KDM5A as candidate mediators of RIPK2-dependent regulation of the signature genes. Collectively, these studies define a functionally validated RIPK2-regulated transcriptional output and provide a framework for translational evaluation of RIPK2-targeted therapeutic strategies. Material and methods Cell lines, culture, and treatment Human PC cell lines 22Rv1 (#CRL-2505), DU145 (#HTB-81), PC3 (#CRL-1435), C4–2B (#CRL-3315), and MDA-PCa-2b (#CRL-2422) were obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA). Control and RIPK2-knockout (RIPK2-KO) 22Rv1, DU145, and PC3 cells were generated using CRISPR/Cas9 technology [2]. 22Rv1 and C4–2B cells were cultured in RPMI-1640 medium (ATCC, #30–2001) supplemented with 10% fetal bovine serum (FBS; ATCC, #30–2021), Gentamicin (25 mg/mL), and Amphotericin B (125 µg/mL) (Sigma, St. Louis, MO, USA, #SBR00045–2ML). DU145 and PC3 cells were cultured in DMEM medium (ATCC, #30–2002) supplemented with 10% FBS and 1% penicillin/streptomycin. MDA-PCa-2b cells were cultured in BRFF-HPC1 medium (Fisher Scientific, Pittsburgh, PA, USA, #NC9970798) supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin. All cells were maintained under standard conditions (37 °C, 5% CO2). For pharmacological studies, cells were treated with ponatinib (MedChemExpress, Monmouth Junction, NJ, USA, #HY-12047), RIPK2-IN-4 (MedChemExpress, #HY-107978), AZD1480 (VWR International, Radnor, PA, USA, #ALA126326), MYCi975 (Cayman Chemical, #39622), CPI-455 (MedChemExpress, #HY-100421), or KDM5A-IN-1 (MedChemExpress, #HY-100014) at the indicated doses and time points. RNA sequencing and data analysis Total RNA was extracted from cells (three biological replicates per clone) using the RNeasy Mini Kit (Qiagen, Valencia, CA, USA, #74104). RNA integrity was verified before library preparation. Strand-specific mRNA libraries were prepared and sequenced by BGI (Shenzhen, China) on the DNBseq platform using 150-bp paired-end (PE150) reads, with a target depth of ∼25 million reads per sample. Adapter sequences, contaminants, and low-quality reads were removed to generate clean FASTQ files (Phred+33 encoding). Sequencing quality was assessed using FastQC (v0.11.7). High-quality reads were aligned to the human reference genome (hg38) with STAR (ref [8].) (-alignIntronMin 20 -alignIntronMax 1000,000 -alignSJoverhangMin 8 -quantMode GeneCounts). Gene-level quantification was performed using RSEM (ref [9].). Count data were normalized using the trimmed mean of M values (TMM) method implemented in edgeR [10]. Principal component analysis and sample correlation analyses were performed to assess data quality and reproducibility. Differentially expressed genes were identified from TMM-normalized counts, with false discovery rate control using the Benjamini-Hochberg method [11]. RNA extraction and RT-qPCR Total RNA was extracted by direct lysis using the SingleShot Cell Lysis Kit (Bio-Rad, Hercules, CA, USA, #1725080) according to the manufacturer’s instructions. Cells were washed once with phosphate-buffered saline and lysed in 50 µL of lysis buffer for 5 min at room temperature. Lysates (1–2 µL) were used for complementary DNA synthesis with the iScript Reverse Transcription SuperMix (Bio-Rad, #1708841). RT-qPCR was performed using the SsoAdvanced Universal SYBR Green Supermix (Bio-Rad, #1725272) with gene-specific primers (200 nM per reaction) on a CFX Opus 96 Real-Time PCR system (Bio-Rad, #12011319). Relative gene expression was calculated using the 2-ΔΔCt method and normalized to GAPDH or ACTB. Protein extraction and western blotting Cells were lysed in RIPA buffer (Boston Bioproducts, Milford, MA, USA, #BP-115) supplemented with Halt protease inhibitor cocktail (Thermo Fisher Scientific, Waltham, MA, USA, #87786). Lysates were mixed with Laemmli SDS sample buffer (Thermo, #J61337), reduced and denatured at 95 °C for 8 min, separated on Mini-PROTEAN TGX precast gels (Bio-Rad, #4568096), and transferred onto nitrocellulose membranes (Bio-Rad, #1704270) using the Trans-Blot Turbo transfer system (Bio-Rad, #1704150). Membranes were blocked with EveryBlot blocking buffer (Bio-Rad, #12010020) for 5 min and incubated with primary antibodies (1:1000 dilution) for 1 h at room temperature (RT) or overnight at 4 °C. Primary antibodies included RIPK2 (Cell Signaling Technology [CST], Danvers, MA, #4142), β-actin (CST, #4970S), JAK1 (CST, #3344T), JAK2 (CST, #3230T), pSTAT3 (CST, #9145T), STAT3 (CST, $12640T), c-Myc (Abcam, ab32072), KDM5A (CST, #3876T), H3K4me3 (CST, #9751T), H3 (CST, #9715S), and GAPDH (CST, 51332S). After washing, membranes were incubated for 1 h at RT with HRP-conjugated anti-rabbit IgG secondary antibody (1:5000; CST, #7074S), except for GAPDH, which was detected using an HRP-conjugated primary antibody. Signals were developed with SuperSignal West Pico PLUS substrate (Thermo Fisher, #34580) at RT and visualized using a ChemiDoc MP imaging system (Bio-Rad, #12003154). Subcutaneous xenograft studies All animal experiments were conducted under an approved Stony Brook University Institutional Animal Care and Use Committee (IACUC) protocol (IACUC2024–00005). Luciferase (Luc2)-labeled 22Rv1 cells were used to enable comparison with our prior in vivo findings [2]. Cells (1.5 × 107/mL) were resuspended in DPBS and mixed 1:1 (v/v) with Matrigel (BD, #356234). A total of 100 μL of the cell suspension was injected subcutaneously into one flank of 6-week-old male SCID/Beige mice (n = 12; Charles River, #CRL:250; CB17.Cg-PrkdcscidLystbg-J/Crl). When tumors reached approximately 200 mm3, mice were randomized to receive vehicle control or ponatinib (6 mg/kg) administered by oral gavage once daily. After 3 days of treatment, three mice per group were euthanized 3 h post-dose; the remaining mice were euthanized on day 7 (3 h post-dose). Xenograft tumors were harvested and weighed. Total RNA was isolated from flash-frozen tumors using a TRIzol/RNeasy workflow under RNase-free conditions [12]. Briefly, tissues were pulverized in liquid nitrogen and homogenized in TRIzol (Invitrogen, #15596026), followed by chloroform phase separation. The aqueous phase was mixed with ethanol and applied to RNeasy Mini spin columns (Qiagen, #74104) with on-column DNase I treatment according to the manufacturer’s instructions. Columns were washed, and RNA was eluted in RNase-free water. Samples with high tissue input were split across multiple columns to prevent clogging. For protein extraction, flash-frozen tumor tissues were similarly pulverized in liquid nitrogen and homogenized in RIPA buffer. Calculation of RIPK2 signature scores Signature scores were computed using R (v4.3.1). Because qPCR-based experimental data and clinical transcriptomic datasets differ in assay structure and analytical purpose, platform-appropriate scoring methods were used. These scores were used for within-platform analyses and should not be treated as directly interchangeable numerical values. For qPCR-based analyses, ΔCt values—which are inversely proportional to gene expression levels—were Z-score normalized across samples for each of the eight signature genes. For each biological sample, a composite signature score was generated by summing gene-specific Z-scores with directional weighing: Z-scores for RIPK2-repressed genes (LAMB3, PEAR1, PLXNA2, TRPV4) were summed directly, whereas Z-scores for RIPK2-dependent genes (ANGPTL4, ANKLE1, SGK2, and TRABD2A) were inverted (multiplied by −1) before summation. The total was divided by 8 (the number of genes) to yield an averaged signature score for ease of interpretation. Within this scoring approach, higher scores indicate higher inferred RIPK2 activity. For clinical transcriptomic analyses, datasets were retrieved from cBioPortal and the Gene Expression Omnibu [13,14]. Only datasets containing RIPK2 and all eight signature genes were included. After removing samples with missing values, log2-transformed expression values were center-normalized. In the resulting expression matrix, RIPK2-repressed signature genes (i.e., LAMB3, PEAR1, PLXNA2, and TRPV4) were assigned a weight of −1. Single-sample Gene Set Enrichment Analysis (ssGSEA) (ref [15].) was then performed to calculate signature scores using the 8-gene set. ssGSEA was implemented using the GSVA R package (v2.4.4) via the ssgseaParam() function, which applies a rank-based algorithm based on within-sample gene expression rankings. As with the qPCR-based score, higher ssGSEA scores indicate higher inferred RIP2K activity. Kaplan-Meier survival analysis Kaplan-Meier survival analysis was performed in R (v4.3.1) using the survival and survminer packages. Optimal cut points were determined using the surv_cutpoint() function. Survival curves were generated using the survfit() function and visualized using ggsurvplot(). Hazard ratios were estimated using Cox proportional-hazards models implemented with the coxph() function. Log-rank P values were calculated using the surv_pvalue() function. Likelihood-ratio χ2 statistics were obtained from Cox proportional-hazards models fitted with the coxph() function in the survival package. Time-dependent receiver operating characteristic (ROC) analysis was performed using the timeROC function from the timeROC package. Continuous RIPK2 signature scores or RIPK2 mRNA expression levels were used as markers, and area under the curve (AUC) values were estimated at 1, 3, and 5 years. Chromatin immunoprecipitation (ChIP)-qPCR analysis ChIP-qPCR analysis was performed using the Pierce Magnetic ChIP kit (Thermo Scientific, #26157) according to the manufacturer’s instructions. PC cells were crosslinked with 1% formaldehyde, and chromatin was enzymatically digested, followed by sonication (Branson Sonicator; 20-s pulses at 3 W) to generate DNA fragments of approximately 200–500 bp. Immunoprecipitation was carried out using an anti-H3K4me3 antibody (CST, #9751T; 2 μg per reaction), with normal IgG (Thermo Fisher, #1862244) as a negative control. After crosslink reversal and DNA purification, enrichment at promoter regions of the eight RIPK2 signature genes was quantified by quantitative PCR (qPCR). Enrichment was calculated as a percentage of input DNA and normalized to the IgG control. siRNA-mediated KDM5A knockdown PC cells were transiently transfected with two independent siRNAs targeting distinct regions of the KDM5A transcript (Life Technologies, #11836 and #11834) to enhance specificity. A non-targeting siRNA (Thermo Fisher, #4390843) served as a negative control. Transfections were performed using Lipofectamine RNAiMAX (Thermo Scientific, #13778075) according to the manufacturer’s instructions. Knockdown efficiency was assessed 72 h post-transfection by RT-qPCR and immunoblotting. Statistical analysis Statistical analyses were performed using Microsoft Excel and R (v4.3.1). For cell-based and xenograft experiments, three independent biological replicates were performed unless otherwise specified. Differences between two groups were assessed using unpaired two-tailed Student’s t-tests. Data are presented as mean ± standard error of the mean (SEM). All statistical tests were two-sided, and a P value < 0.05 was considered statistically significant (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001). Results Identification and validation of a RIPK2 activity gene signature in PC cells We previously generated CRISPR/Cas9-mediated RIPK2-knockout (RIPK2-KO) clones of 22Rv1, DU145, and PC3 cells using two independent guide RNAs (gRNAs) per line [2]. RIPK2 ablation was confirmed by immunoblotting (Fig. S1), and RNA-seq was performed to identify RIPK2-regulated genes. Initial quality assessment of raw sequencing data using FastQC confirmed high-fidelity reads across all libraries. All samples demonstrated optimal per-base sequence quality (mean Phred score > 35), stable GC content (49–52%), and negligible adapter contamination (<0.1%) (Supplementary File 1). Using an adjusted P-value <0.01 and absolute fold change >1.5, we identified thousands of differentially expressed genes in each RIPK2-KO clone compared to its matched control (Fig. 1A, Table S1). The DU145 RIPK2-gRNA2 (Rg2) clone showed fewer changes (464 upregulated and 945 downregulated genes), but consistent directionality was observed across clones. Overlap analysis revealed subsets of genes consistently up- or down-regulated in both RIPK2-KO clones within each cell line (Fig. 1B).Fig. 1Identification of candidate RIPK2 activity signature genes in PC cells. (A) Venn diagrams showing genes significantly upregulated (top) or downregulated (bottom) upon RIPK2 knockout (KO) across three PC cell lines. Cg1, control gRNA #1; Rg1, RIPK2 gRNA #1; Rg2, RIPK2 gRNA #2; Rg10, RIPK2 gRNA #10. (B) Heatmaps of genes commonly regulated by RIPK2-KO in different PC cell lines (n = 3). (C) Heatmap of the 13 candidate RIPK2 signature genes identified from the overlapping gene sets. Gene expression values were normalized to their respective Cg1 control cells and are presented as log2-transformed fold changes (Log2FC). Nominal P-values were calculated using an unpaired two-tailed Student’s t-test comparing RIPK2-KO with Cg1 control cells. P < 0.05 was considered statistically significant. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns, not significant.Fig 1 dummy alt text Identification of candidate RIPK2 activity signature genes in PC cells. (A) Venn diagrams showing genes significantly upregulated (top) or downregulated (bottom) upon RIPK2 knockout (KO) across three PC cell lines. Cg1, control gRNA #1; Rg1, RIPK2 gRNA #1; Rg2, RIPK2 gRNA #2; Rg10, RIPK2 gRNA #10. (B) Heatmaps of genes commonly regulated by RIPK2-KO in different PC cell lines (n = 3). (C) Heatmap of the 13 candidate RIPK2 signature genes identified from the overlapping gene sets. Gene expression values were normalized to their respective Cg1 control cells and are presented as log2-transformed fold changes (Log2FC). Nominal P-values were calculated using an unpaired two-tailed Student’s t-test comparing RIPK2-KO with Cg1 control cells. P < 0.05 was considered statistically significant. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns, not significant. To derive a robust RIPK2 activity signature, we focused on genes with consistent direction and magnitude (absolute fold change >1.5) across all RIPK2-KO clones, allowing at most one exception among the six clones. This analysis identified 13 candidate RIPK2-regulated genes (Fig. 1C). We next performed RT-qPCR validation in 22Rv1 and PC3 cells using primers listed in Table S2. In 22Rv1 cells, the Rg1 clone was replaced with Rg3 due to partial RIPK2 re-expression in Rg1 (Fig. 2A). Eight genes were consistently validated: four RIPK2-dependent genes (downregulated upon RIPK2-KO) and four RIPK2-repressed genes (upregulated upon RIPK2-KO) (Fig. 2B). Five additional candidate genes failed validation and were excluded (Fig. S2).Fig. 2RT-qPCR validation of candidate RIPK2 activity signature genes. (A) Representative immunoblot images of the indicated proteins in control and RIPK2-KO 22Rv1 cells. (B) Bar graphs showing relative mRNA expression levels of RIPK2-dependent genes (top) and RIPK2-repressed genes (bottom) in control and RIPK2-KO cells (n = 3). (C) Swarm plots of RIPK2 signature qPCR scores in control and RIPK2-KO cells (n = 3). Solid lines represent the mean signature scores. Nominal P-values were determined using an unpaired two-tailed Student’s t-test. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.Fig 2 dummy alt text RT-qPCR validation of candidate RIPK2 activity signature genes. (A) Representative immunoblot images of the indicated proteins in control and RIPK2-KO 22Rv1 cells. (B) Bar graphs showing relative mRNA expression levels of RIPK2-dependent genes (top) and RIPK2-repressed genes (bottom) in control and RIPK2-KO cells (n = 3). (C) Swarm plots of RIPK2 signature qPCR scores in control and RIPK2-KO cells (n = 3). Solid lines represent the mean signature scores. Nominal P-values were determined using an unpaired two-tailed Student’s t-test. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. Using these 8 genes, we calculated RIPK2 signature scores (see Methods). They were significantly reduced in RIPK2-KO clones relative to controls, paralleling RIPK2 protein levels and with the strongest reduction observed in the most efficient knockout clone (Rg10) (Fig. 2C). These data define an 8-gene RIPK2 activity signature in PC cells. Pharmacologic RIPK2 inhibition suppresses the signature in vitro To test whether the signature reflects acute RIPK2 activity, we treated PC cells with two structurally distinct RIPK2 inhibitors: ponatinib and RIPK2 inhibitor 4 (RIPK2-IN-4). Ponatinib is an FDA-approved multi-kinase inhibitor previously shown to suppress RIPK2/MKK7/c-Myc signaling and metastasis in vivo [2], whereas RIPK2-IN-4 is a selective and potent RIPK2 inhibitor [16]. Both inhibitors modulated expression of the eight signature genes and reduced signature scores in a dose-dependent manner in 22Rv1 and PC3 cells (Fig. 3A-B, Fig. S3). Time-course experiments revealed gene-specific dynamics (Fig. 3C, Fig. S4), but both inhibitors produced a significant time-dependent reduction in overall signature scores (Fig. 3D).Fig. 3RIPK2 inhibition modulates signature gene expression and reduces signature scores in a dose- and time-dependent manner. (A) Heatmaps showing relative mRNA expression of the indicated RIPK2 signature genes following treatment with increasing doses of ponatinib or RIPK2 inhibitor 4 (RIPK2-IN-4) for 24 h (n = 3). Expression levels were normalized to their respective vehicle controls. (B) Swarm plots of RIPK2 signature qPCR scores under the same dose-dependent conditions shown in (A) (n = 3). (C) Heatmaps showing relative mRNA expression of the indicated RIPK2 signature genes following treatment with 1 µM ponatinib or RIPK2-IN-4 at the indicated time points (n = 3). Expression levels were normalized to their respective vehicle controls. (D) Swarm plots of RIPK2 signature qPCR scores under the same time-course conditions shown in (C) (n = 3). (E) Swarm plots of RIPK2 signature qPCR scores in three additional PC cell lines treated with 1 µM ponatinib or RIPK2-IN-4 at the indicated time points (n = 3). (F) Swarm plot of RIPK2 signature qPCR scores in PC3 cells treated with 1 µM AZD1480 at the indicated time points (n = 3). Nominal P-values were calculated using an unpaired two-tailed Student’s t-test. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns, not significant.Fig 3 dummy alt text RIPK2 inhibition modulates signature gene expression and reduces signature scores in a dose- and time-dependent manner. (A) Heatmaps showing relative mRNA expression of the indicated RIPK2 signature genes following treatment with increasing doses of ponatinib or RIPK2 inhibitor 4 (RIPK2-IN-4) for 24 h (n = 3). Expression levels were normalized to their respective vehicle controls. (B) Swarm plots of RIPK2 signature qPCR scores under the same dose-dependent conditions shown in (A) (n = 3). (C) Heatmaps showing relative mRNA expression of the indicated RIPK2 signature genes following treatment with 1 µM ponatinib or RIPK2-IN-4 at the indicated time points (n = 3). Expression levels were normalized to their respective vehicle controls. (D) Swarm plots of RIPK2 signature qPCR scores under the same time-course conditions shown in (C) (n = 3). (E) Swarm plots of RIPK2 signature qPCR scores in three additional PC cell lines treated with 1 µM ponatinib or RIPK2-IN-4 at the indicated time points (n = 3). (F) Swarm plot of RIPK2 signature qPCR scores in PC3 cells treated with 1 µM AZD1480 at the indicated time points (n = 3). Nominal P-values were calculated using an unpaired two-tailed Student’s t-test. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns, not significant. Similar time-dependent suppression of signature genes and scores was observed in DU145, C4–2B, and MDA-PCa-2b cells (Fig. 3E, Fig. S5), although responses were more modest than in 22Rv1 and PC3 cells (Fig. S4 vs. Fig. S5), consistent with lower basal RIPK2 activity in these lines. Vehicle treatment (0.1% DMSO) did not affect the expression of signature genes over time (Fig. S6). To evaluate specificity, PC3 cells were treated with the well-characterized JAK inhibitor AZD1480. Although AZD1480 suppressed STAT3 phosphorylation in a time-dependent manner (Fig. S7A), it did not significantly alter RIPK2 signature gene expression or scores (Fig. 3F, Fig. S7B). Together, these data indicate that the 8-gene signature dynamically reports RIPK2 activity in vitro and is not broadly responsive to kinase inhibition. Ponatinib progressively suppresses the RIPK2 activity signature in vivo Although RIPK2-IN-4 is a highly selective RIPK2 inhibitor, its in vivo pharmacokinetic properties have not been fully characterized. We therefore selected ponatinib for in vivo studies because it has established oral bioavailability, demonstrated RIPK2 inhibition in mice, and strong translational relevance as an FDA-approved drug [2]. To determine whether the signature functions as a pharmacodynamic biomarker in vivo, male immunodeficient SCID/Beige mice bearing ∼200 mm3 subcutaneous 22Rv1 xenografts were treated daily with ponatinib (6 mg/kg) or vehicle control via oral gavage for 3 or 7 days (Fig. 4A).Fig. 4Ponatinib progressively suppresses the RIPK2 activity signature in vivo. (A) Schematic of the in vivo experimental design. Male immunodeficient SCID/Beige mice bearing ∼200 mm3 subcutaneous (s.c.) 22Rv1 xenograft tumors were treated daily with ponatinib (Pon, 6 mg/kg) or vehicle (Veh) control by oral gavage for 3 or 7 days. Created with BioRender.com. (B) Tumor weights of 22Rv1 xenografts harvested after 3 or 7 days of treatment (n = 3 mice per group). (C) Immunoblot images (left) and corresponding densitometric quantification (right) of the indicated proteins in xenograft tumor lysates (n = 3). (D) Grouped dot plot showing log2-transformed fold changes (Log2FC) of the indicated RIPK2 signature genes in xenograft tumors relative to vehicle controls (n = 3). (E) Swarm plot of RIPK2 qPCR signature scores in xenografts (n = 3). Solid lines represent the mean signature scores. Nominal P-values were calculated using an unpaired two-tailed Student’s t-test. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns, not significant.Fig 4 dummy alt text Ponatinib progressively suppresses the RIPK2 activity signature in vivo. (A) Schematic of the in vivo experimental design. Male immunodeficient SCID/Beige mice bearing ∼200 mm3 subcutaneous (s.c.) 22Rv1 xenograft tumors were treated daily with ponatinib (Pon, 6 mg/kg) or vehicle (Veh) control by oral gavage for 3 or 7 days. Created with BioRender.com. (B) Tumor weights of 22Rv1 xenografts harvested after 3 or 7 days of treatment (n = 3 mice per group). (C) Immunoblot images (left) and corresponding densitometric quantification (right) of the indicated proteins in xenograft tumor lysates (n = 3). (D) Grouped dot plot showing log2-transformed fold changes (Log2FC) of the indicated RIPK2 signature genes in xenograft tumors relative to vehicle controls (n = 3). (E) Swarm plot of RIPK2 qPCR signature scores in xenografts (n = 3). Solid lines represent the mean signature scores. Nominal P-values were calculated using an unpaired two-tailed Student’s t-test. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns, not significant. Ponatinib significantly reduced tumor size and weight at both time points, with stronger effects at Day 7 (Fig. 4B, Fig. S8). c-Myc protein levels decreased by 69% and 86% after 3 and 7 days, respectively (Fig. 4C). RIPK2 protein levels were significantly reduced at Day 7 but not Day 3 (Fig. 4C), consistent with reported lysosomal degradation of RIPK2 upon ponatinib treatment [17]. Importantly, ponatinib modulated signature genes in vivo in a manner consistent with RIPK2 inhibition: RIPK2-dependent genes decreased, RIPK2-repressed genes increased, and overall signature scores declined significantly; these changes were more pronounced at Day 7 than Day 3 (Fig. 4D–E). Collectively, these findings support the in vivo pharmacodynamic responsiveness of the RIPK2 activity signature, although future studies using selective RIPK2 inhibitors will be needed to confirm RIPK2 specificity. RIPK2 signature scores are more strongly associated with metastatic disease and adverse clinical outcomes than RIPK2 mRNA We next assessed the clinical relevance of the RIPK2 activity signature. Signature scores were significantly higher in metastatic PC than in primary tumors, whereas RIPK2 mRNA levels alone did not differ significantly between groups (Fig. 5A).Fig. 5RIPK2 activity signature scores correlate more strongly than RIPK2 mRNA levels with metastasis and survival outcomes in PC. (A) Violin plots comparing RIPK2 signature scores (left) and RIPK2 mRNA expression levels (right) between primary and metastatic PC samples in the Taylor cohort (n = 131 primary vs. 19 metastatic) [31]. (B) Kaplan-Meier curves of progression-free survival stratified by RIPK2 signature scores (left) or RIPK2 mRNA levels (right) in the TCGA Pan-Cancer Atlas PC cohort (n = 493). (C) Kaplan-Meier curves of disease-free survival in the Taylor cohort (n = 140). (D) Kaplan-Meier curves of overall survival in the TCGA Pan-Cancer Atlas PC cohort (n = 493). Optimal cut points were applied to stratify patients in all Kaplan-Meier curves. Hazard ratios (HRs) with 95% confidence interval (CIs) are shown. Nominal P-values were calculated using a two-sided log-rank test.Fig 5 dummy alt text RIPK2 activity signature scores correlate more strongly than RIPK2 mRNA levels with metastasis and survival outcomes in PC. (A) Violin plots comparing RIPK2 signature scores (left) and RIPK2 mRNA expression levels (right) between primary and metastatic PC samples in the Taylor cohort (n = 131 primary vs. 19 metastatic) [31]. (B) Kaplan-Meier curves of progression-free survival stratified by RIPK2 signature scores (left) or RIPK2 mRNA levels (right) in the TCGA Pan-Cancer Atlas PC cohort (n = 493). (C) Kaplan-Meier curves of disease-free survival in the Taylor cohort (n = 140). (D) Kaplan-Meier curves of overall survival in the TCGA Pan-Cancer Atlas PC cohort (n = 493). Optimal cut points were applied to stratify patients in all Kaplan-Meier curves. Hazard ratios (HRs) with 95% confidence interval (CIs) are shown. Nominal P-values were calculated using a two-sided log-rank test. Elevated signature scores were associated with shorter progression-free survival (PFS), disease-free survival (DFS), and overall survival (OS) using endpoint-specific optimized cut points (Fig. 5B–D). Similar trends were observed when patients were stratified using matched proportions (Fig. S9). In contrast, RIPK2 mRNA expression alone showed weaker and less consistent associations (Fig. 5B–D, Fig. S9). Because optimized Kaplan-Meier cut-points can overestimate significance, we performed time-dependent ROC analysis as a complementary, threshold-independent assessment of discriminatory performance. The RIPK2 signature showed modest but consistently higher AUC values than RIPK2 mRNA expression across the evaluated survival endpoints (Fig. S10). In Cox regression analyses, the signature yielded higher likelihood-ratio χ2 statistics than RIPK2 mRNA expression alone for all survival endpoints examined (PFS: 2.2 vs. 0.5; DFS: 3.3 vs. 0.5; OS: 3.1 vs. 0.9), indicating stronger associations with patient outcomes. The likelihood-ratio χ2 statistic reflects the strength of association between a variable and survival outcome, with higher values indicating better model fit and stronger prognostic relevance. Collectively, these retrospective analyses support that the RIPK2 activity signature is more strongly associated with clinically aggressive disease than RIPK2 transcript abundance alone. RIPK2-dependent regulation of the signature genes involves c-Myc and KDM5A Because c-Myc is a major downstream effector of RIPK2 signaling in PC cells [2], we tested whether it mediates RIPK2’s regulation of signature genes. Pharmacological inhibition of c-Myc phenocopied RIPK2 inhibition, and combined inhibition did not produce additive effects in 22Rv1 or PC3 cells (Fig. 6A and Fig. S11), supporting c-Myc as a key mediator. However, c-Myc inhibition suppressed overall signature scores less effectively than RIPK2 inhibition (Fig. 6B), suggesting the presence of additional regulatory factors.Fig. 6RIPK2-dependent regulation of the signature genes involves c-Myc and KDM5A. (A) Heatmap showing relative expression of RIPK2 signature genes under the indicated treatment conditions (1 μM inhibitor; 24 h; n = 3). (B) Swarm plot of RIPK2 signature scores in 22Rv1 and PC3 cells treated with 1 μM RIPK2 inhibitor 4 (RIN4), 1 μM MYCi975 (MYCi), or their combination (dual) for 24 h (n = 3). (C) Venn diagram showing transcription factors (TFs) significantly enriched from RIPK2-dependent or RIPK2-repressed signature gene sets. (D) Bar graphs showing relative mRNA expression levels of RIPK2 signature genes in control, RIPK2-KO, and RIPK2-rescue clones derived from 22Rv1 cells (n = 3). (E) Heatmap of RIPK2 signature gene expression in KDM5A knockdown (siKDM5A) cells relative to control siRNA (siCtrl) (n = 3). (F) Swarm plot of RIPK2 signature qPCR scores under the siRNA knockdown conditions shown in (E) (n = 3). si5A: siKDM5A. (G) Heatmap showing relative expression of RIPK2 signature genes following treatment with the indicated inhibitors (1 μM each; 24 h; n = 3). CPI, CPI-455; KDM, KDM5A-IN-1. (H) Swarm plot of RIPK2 signature qPCR scores under the inhibitor condition shown in (G) (n = 3). Nominal P-values were calculated using an unpaired two-tailed Student’s t-test. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns, not significant.Fig 6 dummy alt text RIPK2-dependent regulation of the signature genes involves c-Myc and KDM5A. (A) Heatmap showing relative expression of RIPK2 signature genes under the indicated treatment conditions (1 μM inhibitor; 24 h; n = 3). (B) Swarm plot of RIPK2 signature scores in 22Rv1 and PC3 cells treated with 1 μM RIPK2 inhibitor 4 (RIN4), 1 μM MYCi975 (MYCi), or their combination (dual) for 24 h (n = 3). (C) Venn diagram showing transcription factors (TFs) significantly enriched from RIPK2-dependent or RIPK2-repressed signature gene sets. (D) Bar graphs showing relative mRNA expression levels of RIPK2 signature genes in control, RIPK2-KO, and RIPK2-rescue clones derived from 22Rv1 cells (n = 3). (E) Heatmap of RIPK2 signature gene expression in KDM5A knockdown (siKDM5A) cells relative to control siRNA (siCtrl) (n = 3). (F) Swarm plot of RIPK2 signature qPCR scores under the siRNA knockdown conditions shown in (E) (n = 3). si5A: siKDM5A. (G) Heatmap showing relative expression of RIPK2 signature genes following treatment with the indicated inhibitors (1 μM each; 24 h; n = 3). CPI, CPI-455; KDM, KDM5A-IN-1. (H) Swarm plot of RIPK2 signature qPCR scores under the inhibitor condition shown in (G) (n = 3). Nominal P-values were calculated using an unpaired two-tailed Student’s t-test. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns, not significant. Interrogation of the KnockTF 2.0 database [18] identified KDM5A, ACO1, and EIF2AK2 as transcription factors enriched across both RIPK2-dependent and RIPK2-repressed signature gene sets (Fig. 6C). KDM5A was prioritized based on the number of signature genes regulated and its reported cooperation with c-Myc [19]. RIPK2-KO reduced c-Myc protein levels, and this effect was reversed by expression of a CRISPR-resistant RIPK2 rescue construct (Fig. S12). In contrast, RIPK2 did not alter KDM5A protein abundance or global levels of H3K4me3, the canonical substrate of KDM5A (Fig. S12) [20]. Nevertheless, ChIP-qPCR analyses revealed that RIPK2-KO significantly altered promoter-associated H3K4me3 levels at signature genes in both 22Rv1 (Fig. 6D, Table S3) and PC3 cells (Fig. S13). Specifically, H3K4me3 levels decreased at promoters of RIPK2-dependent genes and increased at promoters of RIPK2-repressed genes, consistent with the corresponding gene expression changes (Fig. 6D vs. Fig. 2B; Fig. S13 vs. Fig. 2B). These alterations were restored by RIPK2 rescue (Fig. 6D, Fig. S13), indicating that RIPK2 exerts locus-specific control of promoter-associated H3K4me3 without altering global H3K4me3 or KDM5A protein levels. To determine whether KDM5A functionally mediates RIPK2-dependent regulation of the signature genes, we next examined the effects of genetic and pharmacologic perturbations of KDM5A. KDM5A knockdown altered the expression of signature genes and signature scores in a pattern consistent with RIPK2 inhibition, without affecting RIPK2 protein levels (Fig. 6E–F, Fig. S14–S15). Likewise, pharmacologic KDM5A inhibition phenocopied RIPK2 inhibition, whereas combined RIPK2 and KDM5A inhibition did not yield additive effects (Fig. 6G–H, Fig. S16-S18). Together, these findings support a model in which RIPK2 regulates signature gene expression via c-Myc and through locus-specific modulation of promoter-associated H3K4me3, with KDM5A serving as a functional mediator. Discussion RIPK2 has recently emerged as a promising drug target in multiple malignancies, including PC. However, the lack of reliable biomarkers to measure RIPK2 activity has limited the development of precision strategies for patient selection and pharmacodynamic monitoring. In this study, we identified and validated an 8-gene signature whose expression levels are consistently modulated by both genetic and pharmacological inhibition of RIPK2. Notably, signature scores decreased in a dose- and/or time-dependent manner following RIPK2 inhibitor treatment in vitro and in vivo, supporting its potential utility as a dynamic readout of RIPK2 signaling output. Beyond its pharmacodynamic responsiveness, the clinical relevance of the RIPK2 activity signature merits consideration. We focused our clinical analyses on metastasis and survival rather than traditional clinicopathologic variables (e.g., TNM stage, PSA, Gleason score), which primarily reflect tumor burden or histologic differentiation rather than activation of specific oncogenic pathways. Unlike established or emerging multi-gene PC prognostic classifiers, including recent machine-learning and multi-omics signatures designed to capture broad determinants of disease aggressiveness, the RIPK2 activity signature was experimentally derived from RIPK2 perturbation and is intended to report RIPK2-regulated transcriptional output [21,22]. Thus, it should not be interpreted as a replacement for established prognostic classifiers. This distinction may also explain why its time-dependent AUC values were modest: the signature was designed to capture pathway-specific RIPK2 activity rather than the full spectrum of clinical, histologic, genomic, and treatment-related factors that determine patient outcomes. Nevertheless, the RIPK2 activity signature showed stronger clinical associations and consistently higher time-dependent AUC values than RIPK2 mRNA expression alone, supporting its potential utility as a pathway-specific biomarker associated with aggressive disease. Future studies should determine whether it provides complementary prognostic, predictive, or pharmacodynamic information in the context of existing clinical and transcriptomic risk models. The biological functions of several signature genes further underscore the potential relevance of this transcriptional output. Among RIPK2-dependent signature genes, ANGPTL4 and SGK2 have established roles in tumor progression and metastasis. ANGPTL4 promotes PC progression through PI3K/Akt signaling and mitochondrial reprogramming [23,24], while SGK2 has been implicated in PC metastasis through suppression of ferroptosis [25]. In contrast, the roles of ANKLE1 and TRABD2A in PC remain largely unexplored. Among RIPK2-repressed genes, PLXNA2 has been linked to ERG-driven migration and invasion [26]. In contrast, LAMB3, PEAR1, and TRPV4 are involved in cell adhesion, signaling, and mechanotransduction in other biological contexts [[27], [28], [29]], but their roles in PC remain poorly characterized. These observations suggest that RIPK2-dependent transcriptional remodeling may orchestrate multiple pro-metastatic signaling pathways, warranting further mechanistic investigation. Mechanistically, our data support a model in which RIPK2-dependent transcriptional regulation involves coordinated activity of c-Myc and KDM5A-associated chromatin remodeling. While c-Myc appears to contribute substantially to the regulation of the signature genes, the incomplete overlap between c-Myc and RIPK2 inhibition suggests that RIPK2 signaling engages additional transcriptional or epigenetic mechanisms. The identification of KDM5A as a convergent candidate mediator is notable, particularly given its reported cooperation with c-Myc in transcriptional control [19]. The observation that RIPK2 influences promoter-associated H3K4me3 without altering global H3K4me3 levels further suggests locus-specific chromatin regulation rather than broad epigenetic reprogramming. Together, these findings raise the possibility that RIPK2 modulates transcriptional output by context-dependently regulating chromatin-associated complexes, a hypothesis that warrants further mechanistic investigation. Because RIPK2 is a canonical mediator of innate immune signaling, future studies should determine how the RIPK2 activity signature relates to the immune contexture of PC [30]. The present signature was derived from RIPK2-perturbed PC cell lines and therefore most directly reflects tumor-cell-intrinsic RIPK2-regulated transcriptional output. However, tumor-cell RIPK2 activity may influence cytokine, chemokine, or stress-response programs, and bulk-tissue signature scores may also be influenced by immune or stromal compartments. Spatial transcriptomic, single-cell, and multiplex imaging studies will be needed to determine whether high RIPK2 signature activity is associated with specific immune infiltrates or immune-suppressive niches in PC. Several limitations merit consideration. First, although platform-appropriate scoring methods were used for pharmacodynamic experiments and clinical transcriptomic datasets, the current study does not define a numerically interchangeable or locked clinical scoring algorithm. Future clinical implementation will require a defined assay platform, assay-specific calibration, and prospective validation. Although two independent, non-overlapping public PC cohorts were analyzed, additional validation in clinically annotated and ideally prospectively collected cohorts will be required to establish generalizability, clinical utility, and integration with existing risk models. Second, the in vivo pharmacodynamic study used ponatinib, a multi-kinase inhibitor, and included only three mice per group at each terminal time point. Thus, the observed tumor suppression and signature-score reduction should be interpreted as supportive evidence that the signature can be modulated in tumors, rather than as definitive proof of RIPK2-specific in vivo efficacy or biomarker performance. Finally, additional studies are needed to establish direct mechanistic links among RIPK2, c-Myc, KDM5A, and promoter-associated H3K4me3 changes, and to define RIPK2-dependent chromatin regulation more comprehensively through tiled promoter analyses or genome-wide profiling. Conclusion In conclusion, we define a RIPK2-regulated gene signature that functions as a molecular readout of RIPK2 activity in PC. By integrating genetic, pharmacologic, chromatin, and clinical association analyses, this work provides a framework for translational evaluation of RIPK2 activity monitoring in RIPK2-targeted therapeutic strategies. Further validation in independent patient cohorts and exploration of minimally invasive platforms, such as liquid biopsy-based assays, will be important next steps toward clinical implementation and companion diagnostic development. Funding This work was supported by the Department of Defense (DoD) Prostate Cancer Research Program Translational Science Award W81XWH-22-1-1031, the National Cancer Institute (NCI) R01CA26694, as well as an OVPR seed grant and a startup fund from Stony Brook University (to WY). Data availability The RNA-seq data generated in this study have been deposited in the Gene Expression Omnibus (GEO) under accession number GSE292278 and are publicly available. Other data generated and/or analyzed during this study are available from the corresponding author upon reasonable request. CRediT authorship contribution statement Ahmed M. Elgehama: Writing – review & editing, Writing – original draft, Visualization, Validation, Methodology, Investigation, Formal analysis, Data curation. Qian Yang: Writing – review & editing, Software, Formal analysis. Jaceline Gislaine Pires Sanches: Writing – review & editing, Investigation, Data curation. Zaoke He: Writing – review & editing, Investigation. Lauryn Ruegg: Writing – review & editing, Investigation. Sungyong You: Writing – review & editing, Supervision, Software, Formal analysis. Wei Yang: Writing – review & editing, Writing – original draft, Visualization, Supervision, Resources, Project administration, Funding acquisition, Formal analysis, Conceptualization. Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.