RNAi-mediated p38δ silencing mitigates anthracycline cardiotoxicity in female mice

Abstract

The anthracycline antibiotic doxorubicin (DOX) is a potent chemotherapy drug, but its use is limited by dose-dependent cardiotoxicity. We previously reported that genetic deletion of p38δ protects female mice from DOX-induced cardiotoxicity (DIC), suggesting that inhibiting this kinase could be an effective treatment. Here, we developed a fully chemically stabilized small interfering RNA (siRNA) that effectively silences p38δ. In an acute DIC model, silencing p38δ reduced mortality and morbidity, preserved heart structure and function, and decreased fibrosis in female mice. It also alleviated DOX-induced electrophysiological remodeling and decreased cardiac inflammation and senescence-associated secretory phenotype. Transcriptomic analysis of DOX-treated p38δ-deficient hearts revealed the downregulation of genes associated with inflammation, ion transport, and impulse generation, and the upregulation of genes involved in oxidative stress management, autophagy, and immune signaling. These findings support silencing p38δ as a promising approach to protect against DIC, highlighting the potential of siRNA-based therapies to mitigate anthracycline cardiotoxicity.

NEW AND NOTEWORTHY

This study presents the first isoform-specific inhibition of p38δ using a DCA-conjugated, fully chemically stabilized siRNA. Our lead compound, si644, achieves potent, durable, and well-tolerated p38δ silencing in vivo in mouse hearts and ex vivo in human cardiac organotypic slices. si644-mediated p38δ silencing protects female mice from DIC by mitigating DOX-induced structural, electrophysiological, fibrotic, and inflammatory remodeling. These findings establish p38δ inhibition as a promising therapeutic strategy for preventing anthracycline cardiotoxicity.

INTRODUCTION

The global cancer burden has been increasing, with about 20 million new cancer cases diagnosed in 2022 and projections surpassing 35 million annually by 2050 (1). Recently, the number of cancer survivors has grown due to advances in diagnostics and modern treatments (2). As a result, there were approximately 18.1 million cancer survivors in the US in 2022, and this number is expected to reach 26 million by 2040 (3). Among this group, cardiovascular disease, rather than cancer, remains the leading cause of death (4).

DOX, a common anthracycline chemotherapy drug, is well-known for causing cancer therapy-related cardiac dysfunction. While DOX is highly effective in treating various cancers, its use is limited by cumulative dose-dependent cardiotoxicity, affecting up to 9% of patients receiving DOX (2,5). DOX-induced cardiac dysfunction can lead to cardiomyopathy and heart failure in cancer patients and survivors. Currently, there are no effective treatments to prevent DIC. This creates a dilemma where treating one world’s major cause of death (cancer) can unintentionally contribute to another leading cause (cardiovascular disease) (3,4).

The mammalian p38 mitogen-activated protein kinase (MAPK) family comprises four isoforms: p38α, p38β, p38γ, and p38δ, which are encoded by the MAPK14, MAPK11, MAPK12, and MAPK13 genes, respectively (6). All four p38 MAPK isoforms are expressed in the heart (7). p38 kinases regulate DIC because they are activated by DOX-induced stresses, such as systemic inflammation, excessive reactive oxygen species (ROS) production, DNA damage, and mitochondrial dysfunction, which trigger adaptive responses (2,8–10). However, the isoform-specific functions of p38 kinases in DIC are incompletely understood. Of these, p38δ is the least characterized, and its potential as a therapeutic target in heart disease and cancer remains under investigation. Recently, we demonstrated that p38δ systemic genetic deletion significantly protected female but not male mice from DIC through a mechanism involving increased cardiac autophagy (8). Murine and human studies identify p38δ as pro-oncogenic (11–16), suggesting that targeting it in DIC would not impair anthracycline efficacy. Currently, there are no available p38δ-selective conventional small-molecule inhibitors (17,18).

RNA interference (RNAi) enables highly specific gene silencing using small interfering RNA (siRNA), with target specificity determined by the sequence of the antisense strand. This makes the method highly programmable and capable of effectively silencing nearly any target gene (19). Improvements in conjugate-mediated delivery of fully chemically stabilized siRNAs have enabled their successful therapeutic use, leading to the recent FDA approval of multiple siRNA drugs for hepatic targeting to treat both rare and common diseases (19). Recently, the effectiveness of hydrophobic conjugate docosanoic acid (DCA)-mediated delivery of chemically engineered siRNAs has been demonstrated, achieving efficient, non-toxic, and long-lasting gene silencing—which can last weeks to months—in the murine heart (20). The versatility and precision of siRNAs in targeting various disease pathways, along with their safe and effective delivery to the heart in vivo, make this approach a promising strategy for selectively inhibiting p38δ in the heart. Notably, the temporary nature of siRNA-mediated gene silencing could be especially beneficial for clinical applications, as patients would not need to have p38δ silenced after completing chemotherapy.

In this study, we identified and validated an siRNA for the selective silencing of the MAPK13 gene, which encodes p38δ, a kinase previously suggested to be resistant to functional inhibition by conventional small-molecule protein kinase inhibitors (17,18). We screened a panel of siRNA sequences that were computationally predicted to silence p38δ. We identified the lead compound called si644, which is cross-reactive with human and mouse MAPK13 mRNA transcripts. In mice, si644 achieved selective, efficient, durable, and non-toxic silencing of p38δ in the heart. The therapeutic efficacy of si644-mediated p38δ silencing was validated in a mouse model of acute DIC. We also provide mechanistic insights into the cardioprotective effects of p38δ targeting in the context of DIC. Furthermore, si644 enabled effective p38δ silencing in ex vivo human left ventricular (LV) organotypic cardiac slices. The siRNA drug approach presented in this work offers new opportunities to modulate other disease targets in the heart.

MATERIALS AND METHODS

Design of p38δ/MAPK13 siRNAs.

The design process for siRNAs was based on a 20-nucleotide target sequence derived from human MAPK13 (accession number: NM_002754.5) and mouse Mapk13 (accession number: NM_011950.2) transcripts. Target sequences, including 10 nucleotides upstream and 15 downstream, were analyzed using a weighted scoring matrix, with the highest-scoring sequences selected for each species. To ensure specificity and efficacy, siRNA sequences that exhibited (1) more than 56% GC content, (2) single-nucleotide runs of four or more, or (3) complete homology to human miRNA seed sequences at positions 2–7 of the antisense strand were excluded. Cross-species targeting was determined by identifying perfect homology within the 16-nucleotide region of the target sequence (positions 2–17) between the two species. Each siRNA was named according to its position in the target transcript (e.g., human siRNA 644 corresponds to positions 644–663 of NM_002754.5), with names for cross-species targeting based on the human transcript.

Oligonucleotide Synthesis.

Oligonucleotides were synthesized using standard phosphoramidite solid phase synthesis on Dr Oligo 48 or MerMade 12 synthesizers, employing 2′-fluoro RNA or 2′-O-methyl RNA phosphoramidites from HonGene Biotech or ChemGenes. Antisense strands were synthesized on UnyLinker 500 Å support at 1-μmol (in vitro) or 5-μmol (in vivo) scales, while cholesterol-conjugated sense strands used tetraethylene glycol cholesterol support at 1-μmol scale. For in vivo applications, a custom CPG was utilized, which allowed conjugation of docosanoic acid to the 3’end of the sense strand. Phosphoramidites were generally dissolved in 0.1 M anhydrous acetonitrile, with 2′-O-Methyluridine phosphoramidite requiring 15% anhydrous DMF addition. The synthesis process involved washing with anhydrous DNA sequencing grade ACN, phosphoramidite activation using 5-(benzylthio)-1H-tetrazole (BTT), a four-minute coupling time, capping with specific solutions, oxidation using iodine solution, sulfurization with 3-[(dimethylaminomethylene)amino]-3H-1,2,4-dithiazole-5-thione (DDTT), and detritylation using 3% trichloroacetic acid in dichloromethane. All reagents were purchased from ChemGenes or Hongene. Post-synthesis treatments varied for in vitro and in vivo applications, with the former being washed with 10% diethylamine in ACN. In vitro oligonucleotides were deprotected with ammonia gas at 65°C for 1.5 hours, then precipitated on-column using 0.1 M NaOAc in 85% ethanol. Afterward, they were washed with 85% ethanol and finally eluted with water.

Oligonucleotide Deprotection for in vivo Applications.

The sense strands were deprotected and cleaved from the CPG using a solution of 40% aqueous methylamine and 30% ammonium hydroxide at room temperature for 2 hours. The antisense strands were cleaved with a mixture of 30% ammonium hydroxide and 3% diethylamine at 35°C for 20 hours. After incubation, the solutions were filtered and evaporated using a SpeedVac. The dry pellets were reconstituted in 5% acetonitrile for purification.

HPLC Purification.

Oligonucleotide purification was conducted using an Agilent 1290 Infinity II HPLC system. Sense strands were purified using a reverse-phase preparative column (Hamilton PRP-C18) with the following conditions: buffer A was 50 mM sodium acetate in 5% acetonitrile, and buffer B was 100% acetonitrile. The elution gradient ranged from 0–20% over 3 minutes, followed by 20–70% over 23 minutes, and concluded with a cleaning and recalibration step lasting 9 minutes. The column operated at 60°C with a flow rate of 40 mL/min. Antisense strands were purified with an anion-exchange column (SOURCETM 15Q) using buffer A of 10 mM sodium acetate in 20% acetonitrile and buffer B of 1 M sodium perchlorate in 20% acetonitrile, with the same gradient and flow rate as for sense strands, but at a column temperature of 55°C. Oligonucleotides were detected by UV absorbance at 260 nm, and peak fractions were automatically collected for identity confirmation. Quality control was performed using liquid chromatography-mass spectrometry (LC/MS). Pure fractions were pooled, frozen, dried overnight in a SpeedVac, and reconstituted in water before desalting with Sephadex G-25 Fine.

LC-MS Analysis.

The purity and identity of all oligonucleotides were assessed using an Agilent 6530 accurate mass Q-TOF LC-MS system with ion-pair reverse phase chromatography. Chromatographic separation was performed on an Agilent 2.1 × 50 mm AdvanceBio C18 oligonucleotide column under the following conditions: buffer A comprised 9 mM triethylamine and 100 mM hexafluoroisopropanol in water, and buffer B contained 9 mM triethylamine and 100 mM hexafluoroisopropanol in methanol. The column was maintained at 60°C with a flow rate of 0.5 mL/min. Peaks were detected by UV absorbance at 260 nm. Mass spectrometry parameters included the use of electrospray ionization in negative mode, a mass scan range of 100–3200 m/z, and a scan rate of 2 spectra per second. The capillary voltage was set at 4000 V, and the fragmentor voltage at 180 V.

Cell Culture.

The murine oral cancer (MOC) squamous cell carcinoma cell line MOC1, generously provided by Dr. Clint T. Allen (National Institute of Deafness and Other Communication Disorders, National Institutes of Health), was generated, characterized, and cultured as described (21), validated, and pathogen-tested as reported (22,23). MOC1 cells were cultured in complete MOC media consisting of Iscove’s Modified Dulbecco’s Media (IMDM)/Ham’s F12 Nutrient Mixture (Thermo Fisher Scientific) (2:1) (22). HaCaT cells, a spontaneously immortalized human keratinocyte cell line (24), kindly provided by Dr. Norbert E. Fusenig (German Cancer Research Center), were cultured in Dulbecco’s Modified Eagle’s Media (Corning Cellgro; #10–013CV). All cell lines were grown in complete growth media, which consisted of the specified basal media supplemented with 10% fetal bovine serum (Gibco; #26140). The cells were incubated at 37 °C in a humidified atmosphere with 5% CO2, and no antibiotics were used during the experiments. Cells were passaged every 2–4 days and discarded after 15 passages to reduce the risk of genetic drift.

In Vitro Passive Uptake Screening.

HaCaT and MOC1 cells were treated with cholesterol-conjugated siRNAs at a concentration of 1.5 μM, representing the maximum dose for dose-response assays. Cells were cultured in 3% fetal bovine serum (FBS) media, prepared by mixing 6% FBS media with Opti-MEM media (Gibco, #31985–079) in a 1:1 ratio. After a 72-hour incubation, cells were collected in a lysis mixture consisting of 1 part lysis mixture (Invitrogen, #13228) to 2 parts water, supplemented with 0.2 mg/mL proteinase K (Invitrogen, #25530–049) and incubated at 55 °C for 30 minutes. Target mRNA levels were quantified using Quantigene 2.0 assays (Affymetrix). QuantiGene detection probesets were obtained from ThermoFisher: human MAPK13 (#SA-12678), mouse Mapk13 (#SB-3032493), human HPRT (#SA-10030), and mouse Hprt (#SA-15463). MAPK13 mRNA expression was normalized to the housekeeping genes HPRT or Hprt, depending on the cell line used.

Mouse Studies.

All animal procedures were approved by the Institutional Animal Care and Use Committee at Northwestern University and followed the National Institutes of Health Guide for the Care and Use of Laboratory Animals. Mice were on a C57BL/6J background; WT females were obtained from The Jackson Laboratory (Stock No. 000664). Generation and genotyping of germline p38δ KO mice have been previously documented (25). p38δ KO mice are viable, fertile, show no abnormalities, and demonstrate normal baseline cardiac function (25,26). Female p38δ KO mice were bred in-house and used for experiments. All mice were 13–15 weeks old at the start of experiments, housed in pathogen-free facilities under a 12-h light/dark cycle at 70–74°F and 30–70% humidity, with food and water provided ad libitum.

Experimental Groups.

Evaluation of the efficiency and safety of p38δ KD using DCA-conjugated siRNAs:

WT female mice received subcutaneous injections of 20 mg/kg of siNTC or si644 twice, 10 hours apart, for either 72 hours or 14 days (n = 5 per group). p38δ KD and DOX administration: WT female mice were subcutaneously injected with 20 mg/kg of siNTC (n = 20), si644 (n = 20), or sterile saline (n = 10) twice, 10 hours apart. Seven days after siRNA treatment, all mice received an intraperitoneal injection of 30 mg/kg DOX (Thermo Fisher Scientific, J64000-MA) to induce acute DIC. Survival and health of the mice were monitored for 11 days. p38δ KO mice and DOX treatment: WT and p38δ^−/− female mice (n = 5 per group) received 30 mg/kg of DOX, and heart tissue samples were collected after three days for RNA analysis.

Survival and Morbidity Assessment.

Survival and health status were monitored in mice treated with DCA-conjugated siRNAs with or without DOX throughout the study period. A health status score was assigned daily, ranging from 0 to 1, using a previously described evaluation method (8). At the end of the survival period, mice were anesthetized with isoflurane and euthanized via cervical dislocation. Hearts were excised, the aorta was cannulated, and the hearts were flushed with PBS. The extracted hearts were either fixed in 4% paraformaldehyde for histological analysis, preserved in RNALater^™ (Millipore Sigma, MFCD03453003) for mRNA quantification, or flash-frozen in liquid nitrogen for Western blot analysis.

Electrocardiography.

In conscious mice, ECGs were recorded using the emka Technologies ecgTUNNEL device (emka Technologies, Sterling, VA). Mice were guided into the tunnel and restrained with each paw in contact with a silver ECG electrode. After a 5-minute acclimation, ECGs were recorded for at least 2 minutes in lead I and II configurations. For each animal, five representative cardiac cycles were selected and averaged. ECG analysis was performed with a custom MATLAB program to measure the following parameters: P-wave duration (P, ms), PR interval (PR, ms), QRS duration (QRS, ms), QT interval (QT, ms), RR interval (RR, ms), and corrected QT (QTc) interval.

Echocardiography.

Mice were anesthetized with a 2% isoflurane/oxygen mixture. They were weighed to monitor body weight during treatment periods. Mice were transferred to an imaging stage heated to 37°C, where anesthesia depth was maintained with isoflurane delivered through a nose cone. Chest hair was removed using depilatory cream, and ultrasound gel (Aquasonics, Clear) was applied to the sternum. Transthoracic M-mode echocardiography of the LV in the short-axis view was performed using a Vevo 3100 (VisualSonics) ultrasound machine or a 40-MHz probe from the high-frequency ultrasound system (Prospect, S-Sharp, New Taipei City, Taiwan). The recordings were analyzed with Vevo LAB 5.7.0 software or Prospect software.

Picrosirius red staining.

At study end, extracted hearts were flushed with PBS, fixed in 4% paraformaldehyde (48 hours), stored in 70% ethanol, paraffin-embedded, sectioned transversely, and stained with Picrosirius red. LV, RV, and septal images were acquired at 5× and 20× magnification (Zeiss). Fibrosis was quantified using a custom MATLAB program with color deconvolution, expressed as the percentage of Picrosirius red–positive pixels per total area. For each region, three images were obtained (LV: anterior, posterior, free wall; RV: anterior, posterior, free wall; septum: three sites) and averaged per sample. These averages were reported as a percentage of fibrosis.

Langendorff Perfusion.

At the end of the survival period following siRNA and DOX treatment, mice in the saline + DOX (n = 9), siNTC + DOX (n = 12), and si644 + DOX (n = 14) groups were deeply anesthetized with isoflurane. The heart was removed after cervical dislocation and thoracotomy. The aorta was cannulated and connected to a Langendorff system, where it was suspended vertically in a bath and perfused retrogradely with warmed (37°C) and oxygenated (95% oxygen and 5% carbon dioxide) modified Tyrode’s solution (in mM, NaCl 130, NaHCO3 24, NaH2PO4 1.2, MgCl2 1, glucose 5.6, KCl 4, and CaCl2 1.8) with a pH of 7.4. The heart was maintained at a constant flow rate (1.0 – 2.0 ml/min), adjusted as needed to sustain a coronary pressure of approximately 80 mmHg. ECG electrodes were placed in the bath to record pseudo-ECGs.

Optical Mapping.

Langendorff-perfused hearts were electromechanically uncoupled with 15 μM blebbistatin (Cayman Chemicals 13186) and equilibrated for 20 minutes. Hearts were then loaded with the voltage-sensitive dye RH237 (30 μL of 1 mg/mL RH237 stock solution plus 970 μL Tyrode’s solution, Biotium 61018) and calcium fluorescent dye Rhod2-AM 30 μL of 1 mg/mL Rhod2-AM stock solution plus 30 μL of Pluronic F-127 plus 940 μL Tyrode’s solution, Thermo Fisher Scientific R 1244 and Biotium 59055, respectively). Illumination (520 ± 5 nm, Prizmatix UHP-Mic-LED-520) excited both dyes, and emission was recorded using CMOS cameras (MiCAM, SciMedia). Baseline recordings were acquired, then hearts were paced at 1.5× threshold via a platinum epicardial point electrode. Data were analyzed in Rhythm 3.3.0 for V_m (V_m RT, APD₅₀, APD₈₀, CV_L, CV_T, and AR_CV) and intracellular calcium (Ca²⁺ RT, CaTD₈₀, and Ca²⁺ τ). Due to severe disease progression, some mice died from DIC before the study concluded. For optical mapping analysis, saline+DOX mice and siNTC+DOX mice exhibiting a DIC phenotype were grouped together as CNTLs. The DIC phenotype in DOX-treated mice was defined as the development of LV hypertrophy by Day 4 after DOX treatment. Specifically, DOX-treated mice with an LVpWd greater than the Day 4 average (0.829 mm) were classified as having DOX-induced LV hypertrophy, indicating DIC.

Mouse Cytokine/Chemokine Array.

Flash-frozen LV and liver tissues were lysed in RIPA buffer (Sigma 89901) with protease and phosphatase inhibitors (Thermo Fisher A32955). Protein concentration was measured in duplicate using the Pierce BCA kit (Thermo Fisher 23227). Cytokine and chemokine levels were quantified using the Mouse 36-Plex Array (MD36) by Eve Technologies (Calgary, AB).

Bulk RNA Sequencing.

Female WT (n = 5) and p38δ KO (n = 5) murine hearts were perfused with PBS and preserved in RNALater^™ (Millipore Sigma, MFCD03453003). Bulk RNA sequencing was performed by Genewiz (Azenta Life Sciences). Reads were aligned to the mouse genome with STAR and quantified using RSEM. Differential expression was assessed with DESeq2 using apeglm shrinkage; significance was set at padj < 0.1. Principle component analysis was performed on regularized log-transformed data. DEGs (padj < 0.1, |log₂FC| > 1) were visualized by volcano plots and heatmaps (z-score normalized counts). GO Biological Process and KEGG enrichment were performed with enrichGO/enrichKEGG (padj < 0.1). For GSEA, genes were ranked by log₂FC and analyzed against Hallmark gene sets (MSigDB, Category) using clusterProfiler, with FDR q < 0.05 considered significant. Data is available at the NIH Gene Expression Omnibus repository (accession number GSE308615).

Ex vivo Human LV Organotypic Slice Preparation and Culture.

Human LV slices were obtained from healthy donor hearts unsuitable for transplant through Gift of Hope (Chicago, IL) and Novabiosis (Indianapolis, IN). The protocol was deemed IRB exempt by Northwestern University. Donor demographics and comorbidities are summarized in Supplemental Table 11 (https://doi.org/10.6084/m9.figshare.31306834); none showed heart failure or dysfunction. Human LV organotypic slices were prepared as previously described (27). Slices were placed in 6-well plates (2 mL prewarmed Medium 199 with 2% penicillin-streptomycin, 1× insulin-transferrin-selenium, 10 mM BDM) and cultured in triplicate per heart (n = 5 hearts per group). Treatments included control medium, 1 μM siNTC, or 1 μM si644 for 72 hours at 37°C, 30% O₂, 5% CO₂, 20 rpm. The culture medium was replaced after 48 hours with 1 mL of the respective culture condition, and slices were cultured for an additional 24 hours. After 72 hours, slices were halved and preserved in RNALater^™ (Millipore Sigma MFCD03453003) or flash-frozen at −80°C.

Western Blot.

Flash-frozen human LV organotypic slices were lysed in RIPA buffer (Sigma, 89901) supplemented with protease and phosphatase inhibitors (Thermo Fisher, A32955). Total protein concentration was measured using a Pierce BCA protein assay kit following the manufacturer’s instructions (Thermo Fisher, 23227), with samples processed in duplicates. Proteins were separated by molecular weight through electrophoresis using a 4–15% Criterion TGX gel (Bio-Rad, 5671084) and then transferred to a PVDF membrane. The membrane was blocked with bovine serum albumin and horizontally cut into two pieces according to the molecular weight of the protein ladder (Bio-Rad, 1610375) for separate detection of housekeeping and target proteins. Both membranes were incubated overnight with primary antibodies. Afterward, they were washed, incubated with secondary antibodies for 2 hours at room temperature, and washed again. Protein bands were detected through chemiluminescence using an Azure300 imager. Densitometry analysis was performed with ImageJ to quantify the relative expression levels of the proteins of interest, with all values normalized to the housekeeping protein. The primary antibodies included p38δ antibody S526A, obtained from the Division of Signal Transduction Therapy (Dundee, UK) (25), and α-actinin antibody 12413S, purchased from Cell Signaling Technology, Inc. (Danvers, MA, USA). The secondary antibodies used were: anti-rabbit IgG, HRP-linked antibody (7074P2, Cell Signaling Technology, Inc.); rabbit anti-sheep IgG (H+L), HRP-linked antibody (31480, Thermo Fisher).

TCGA Data Analysis.

To evaluate the expression of p38α/MAPK14 and p38δ/MAPK13 and their impact on clinical outcomes in breast cancer, we used the Gene Expression Profiling Interactive Analysis (GEPIA) 2 web server (28) to query The Cancer Genome Atlas Breast Cancer (TCGA BRCA) dataset of primary breast cancer samples (29). For expression analysis, we chose the Box Plots module to compare MAPK13 and MAPK14 levels in BRCA tumor samples (n = 1085) versus normal breast tissue (n = 291), utilizing matched BRCA tumor and normal tissue data from TCGA as well as data from normal breast tissue in the Genotype-Tissue Expression (GTEx) dataset. The parameters set were: |Log2 FC| cutoff = 1 and p-value cutoff = 0.01. For survival analysis, we used the Survival Plots module to assess overall survival in BRCA patients based on MAPK13 or MAPK14 expression levels. Patients were divided into high- and low-expression groups (n = 535 patients each) according to the median expression level (50% high vs. 50% low). Kaplan–Meier curves were generated, and statistical significance was determined using the log-rank test. Hazard ratios and 95% confidence intervals were calculated with the Cox proportional hazards model.

Statistical Analysis.

All data were analyzed using GraphPad Prism 9 software (GraphPad Software, Inc.) or RStudio. Data are presented as mean ± SD. Sample sizes were determined based on an expected 30% difference in measured parameters, considered biologically significant, with an estimated sigma of 10–20% of the expected mean. Power calculations used standard α = 0.05 and β = 0.8. No animals or samples were excluded from analysis. When applicable, animals were randomized into treatment and control groups. For comparisons involving more than two groups, one-way ANOVA was used with treatment as the independent variable. Post hoc comparisons were performed using unpaired Student’s t-tests with Tukey correction to evaluate differences in MAPK13 mRNA knockdown efficiency, survival rates on individual days, and regional fibrosis percentages. Overall survival across treatment groups was assessed with the log-rank test. Two-way ANOVA with repeated measures was applied to longitudinal data, including health scores, ECG parameters, echocardiographic measurements, and the left ventricular weight-to-body weight ratio, with treatment group and time point as factors. Nonlinear regression analysis was performed using the least squares regression method. For parameters showing restitution properties (V_m RT, APD₅₀, APD₈₀, CV_L, CV_T, Ca²⁺ RT, CaTD₈₀, and Ca²⁺ τ), an exponential plateau model was used. The best-fit parameters were compared between groups using the sum-of-squares F test. P<0.05 indicates statistical significance.

RESULTS

Identification of a lead siRNA that effectively silences p38δ/MAPK13 expression.

Using a proprietary bioinformatics algorithm, we generated a panel of 48 siRNA sequences targeting the 5’ untranslated region (UTR), open reading frame, or 3’ UTR regions of the p38δ/MAPK13 human and mouse mRNA transcripts, as previously described (30). The 3’ ends of the sense strands of the siRNAs were conjugated to cholesterol to enhance cellular internalization through passive uptake in vitro (30). The panel was used for in vitro screening in human HaCaT and mouse MOC1 cells (Fig. 1A, B). We identified siRNA sequence 644, which achieved approximately 80% and 65% silencing of human and mouse p38δ/MAPK13 mRNAs, respectively. The sequence and chemical modifications of this compound are listed in Supplemental Table 1. Conversely, treatment with the non-targeting control (NTC) siRNA did not affect p38δ/MAPK13 expression compared to untreated (UNT) cells in both screens (Fig. 1A, B). We then determined the median inhibitory concentration (IC₅₀) for 644 using seven-point dose-response experiments, establishing this siRNA as potent (Fig. 1C, D). Fig. 1E shows the accession number and the mRNA targeting site of the 644 siRNA.

Fig. 1.

Identification of lead siRNA-644 that effectively silences p38δ/MAPK13 expression in vitro. A, B: In vitro screening of siRNAs in human HaCaT (a) and mouse MOC1 (b) cells. Cells were treated with fully chemically modified, cholesterol-conjugated siRNAs at 1.5 μM for 72 hours. p38δ/MAPK13 mRNA levels were measured using the QuantiGene 2.0 assay. The siRNA number indicates the 5’ position of the mRNA target site. 644 siRNA is shown in red. UNT, untreated control; NTC, non-targeting control siRNA. Data are shown as a percentage of UNT (n = 3 biologically independent samples, mean ± s.d.). C, D: Seven-point dose-response curves of lead siRNA 644 in human HaCaT (c) and mouse MOC1 cells (d). M, molar concentration of siRNA (n = 3 biologically independent samples, mean ± s.d., ****P < 0.0001; one-way ANOVA). E: Accession number and mRNA targeting site of lead compound 644; ORF, open reading frame.

si644 effectively silenced p38δ in mouse hearts over 14 days and was well tolerated in vivo.

We first tested whether 644 siRNA, covalently linked to the hydrophobic conjugate docosanoic acid (DCA) to improve retention in heart tissue and cellular uptake, could silence p38δ in vivo. Female wild-type (WT) C57BL/6 mice were subcutaneously injected with 20 mg/kg of DCA-conjugated, fully chemically modified 644 or NTC siRNAs (Supplemental Table 1; hereafter called si644 and siNTC, respectively) twice at a 10-hour interval (Fig. 2A). We then measured p38δ/MAPK13 mRNA levels in the heart 72 hours and 14 days after injections. As shown in Figs. 2B and 2C, si644 treatment achieved 53% and 58% silencing of p38δ at 72 hours and 14 days post-injection, respectively, compared to siNTC controls. Mouse survival and overall health were monitored for 14 days after si644 administration, with no significant differences observed versus siNTC-treated mice (Fig. 2D, E). Consistent with previously reported normal cardiac function in unstimulated p38δ knockout (KO) mice (26), our data showed that si644-mediated p38δ silencing in mouse hearts did not significantly affect electrocardiogram (ECG) or echocardiogram parameters in female mice 14 days after treatment, compared to siNTC (Fig. 2F, G; Supplemental Figs. 1, 2, and Supplemental Table 2). These findings indicate that si644 can effectively, durably, and safely silence p38δ in vivo for at least two weeks.

Fig. 2.

si644 effectively silences p38δ in the heart without causing cardiotoxic effects in female mice. A: Schematic of the experimental protocol, involving two subcutaneous injections of female mice with 20 mg/kg of siNTC or si644, administered 10 hours apart, with mice monitored for up to 14 days after siRNA treatment. B, C: p38δ/MAPK13 mRNA levels were measured using the QuantiGene 2.0 assay at 72 hours (b) and 14 days (c) post-treatment (mean ± s.d., unpaired Student’s t-test). D: Survival rates in female mice treated with siNTC and si644 (n=5 per group). A log-rank test was used to assess statistical significance over the survival period. E: The health status was monitored in the specified groups over 14 days after siRNA treatment, as described in Methods. A two-way ANOVA with repeated measures was used to compare health status on the indicated days. F: Representative ECG traces recorded before (black traces) and 14 days after siNTC (blue trace) or si644 treatment (magenta trace). G: Representative echocardiograms recorded before (Day 1) and 14 days after siNTC or si644 treatment. Created in BioRender. Amaral, P. (2025) https://BioRender.com/vuxu2rg

si644-mediated silencing of p38δ in the heart reduced DOX-induced mortality and morbidity in female mice.

Next, we assessed the therapeutic effect of si644 in vivo using a mouse model of acute DIC. Female mice were treated with either siNTC or si644 on Day −6, as shown in Fig. 3A, followed by an intraperitoneal injection of 30 mg/kg of DOX on Day 0. We then monitored the survival and health of the mice for 11 days after DOX treatment. si644-treated mice showed a significantly higher survival rate, with 85% (17/20) surviving the 11 days post-DOX, compared to 55% (11/20) of the siNTC-treated mice (Fig. 3B). Consistent with the improved survival, we recorded significantly better health scores from Days 5 through 11 after DOX in this group compared to their siNTC-treated counterparts (Fig. 3C). These findings demonstrate that si644-mediated p38δ silencing significantly alleviates acute DOX-induced mortality and morbidity in female mice.

Fig. 3.

si644-mediated p38δ silencing protects against DIC-induced morbidity, mortality, and cardiac structural remodeling in female mice. A: Schematic of the experimental protocol involving two subcutaneous injections in female mice on Day −6 of 20 mg/kg of siNTC or si644, administered twice 10 hours apart, followed by an intraperitoneal injection of 30 mg/kg of DOX on Day 0, with mice monitored for up to 11 days after the DOX injection. Created in BioRender. Amaral, P. (2026) https://BioRender.com/19ooile. B: Survival rates of female mice treated with siNTC + DOX or si644 + DOX are shown. The sample size for each group is indicated beside the respective Kaplan-Meier curve. A log-rank test was used to assess the statistical significance of differences during the survival period. A two-way ANOVA with repeated measures was used to compare survival on each day, with p-values displayed below the curves. C: The health status of the groups was monitored using a health scoring method described in the Methods. A two-way ANOVA with repeated measures was used to compare health status on each day, with p-values shown above the curves. D: Representative echocardiograms from the groups on Day 0 and Day 4 after DOX treatment. Yellow bars indicate LV posterior wall thickness during diastole. E: LV heart weight, measured via echocardiography, was normalized to total body weight on Day 0 and Day 4 post-DOX. Differences between groups were analyzed using two-way ANOVA with repeated measures. F – Q: Quantification of structural and mechanical parameters from echocardiograms on Day 0 and Day 4 after DOX. Differences between groups were assessed via two-way ANOVA with repeated measures. Parameters include LVESD/LVEDD, LVaWs/LVaWd, LVpWs/LVpWd, LVESV/LVEDV, stroke volume (SV), ejection fraction (EF), fractional shortening (FS), and cardiac output (CO). Sample sizes: n=20 mice per group for survival and health status assessments; n=15 mice per group for LV heart weight and echocardiography measurements.

si644-mediated p38δ silencing in the heart preserved cardiac structure and function in DOX-treated female mice.

DIC is characterized by LV dysfunction and remodeling (2,31). To assess the structural and functional changes in the LV, we performed M-mode echocardiography on Day 0 (before DOX administration) and Day 4 after DOX treatment (Fig. 3D). M-mode echocardiograms were analyzed to measure LV posterior and anterior wall thickness, LV diameter, and LV volume during systole and diastole. From these measurements, LV mass, ejection fraction (EF), fractional shortening (FS), stroke volume (SV), and cardiac output (CO) were calculated (Fig. 3E–Q, Supplemental Table 3).

Notably, DOX treatment increased LV weight in siNTC-treated but not in si644-treated mice on Day 4 after DOX administration (Fig. 3E). Additionally, DOX treatment reduced CO, an indicator of DIC, in siNTC-treated but not si644-treated mice on Day 4 post-DOX (Fig. 3I). In contrast, DOX-induced increases in EF and FS were observed in both siNTC- and si644-treated mice (Fig. 3F, G).

Furthermore, we observed that LV anterior and posterior wall thickness parameters during systole and diastole—LVaWs, LVaWd, LVpWs, and LVpWd (Fig. 3J–M)—which are related to DOX-induced cardiac hypertrophy, were increased with DOX in siNTC-treated but not in si644-treated mice on Day 4 after DOX administration. Conversely, LV end-systolic and end-diastolic diameters (LVESD and LVEDD, respectively) and volumes (LVESV and LVEDV, respectively) were reduced in both siNTC- and si644-treated mice four days after DOX administration (Fig. 3N–Q).

In summary, the echocardiographic analysis indicated that DOX-induced disruptions in cardiac mechanical function and LV wall thickness were alleviated in si644-treated mice, which is linked to decreased DOX-induced morbidity and mortality in this group.

The effects of si644-mediated p38δ silencing in the heart on ECG parameters in DOX-treated female mice.

We then evaluated cardiac electrical function by recording ECGs in conscious female mice treated with either siNTC or si644 four days after DOX administration. Representative ECG traces are shown in Fig. 4A, and ECG parameters are listed in Fig. 4B–E and Supplemental Table 3. P wave and QRS interval durations, reflecting the speed of electrical excitation spreading across the atria and ventricles, respectively, remained unchanged by DOX treatment in both the siNTC- and si644-treated groups (Fig. 4B, D), possibly because of the relatively short period after acute DOX exposure, as previously reported (32). Conversely, DOX treatment significantly prolonged the PR interval in siNTC-treated but not in si644-treated mice (Fig. 4C), while the heart rate-corrected QT (QTc) interval was prolonged by DOX in both groups (Fig. 4E).

Fig. 4.

si644-mediated p38δ silencing does not reduce DOX-induced QT toxicity. A: Representative ECG traces before (black) and after DOX administration in siNTC- (blue) and si644-treated (magenta) mice. B – E: Summary of electrical function parameters measured for the indicated groups. Two-way ANOVA with repeated measures was used to evaluate differences among groups. Sample size: n=15 mice per group.

si644-mediated p38δ silencing in the heart reduced DOX-induced septal fibrosis in female mice.

Myocardial fibrosis and extracellular matrix remodeling contribute to heart failure and DIC (8,33). We measured collagen deposition in the LV, right ventricle (RV), and the interventricular septa of DOX-treated female mice at the end of the 11-day survival period using Picrosirius Red staining (Fig. 5). Notably, DOX-induced septal fibrosis was significantly reduced, as indicated by decreased Picrosirius Red staining, in si644-treated mice compared to their siNTC-treated counterparts (Fig. 5C, D). In contrast, DOX-induced fibrosis in the LVs and RVs did not differ significantly between the two groups (Fig. 5A, B). The preservation of cardiac structure in DOX-treated female mice with si644-mediated p38δ silencing aligns with the improved survival, overall health, and cardiac function observed in this group.

Fig. 5.

si644-mediated p38δ silencing reduces DOX-induced cardiac septal fibrosis in female mice. A – C: Quantification of the fibrosis percentage area in the LV (a), RV (b), and septum (c) for the specified groups. Unpaired Student’s t-tests were used to assess the statistical significance of differences between groups in each region. LV: left ventricle, RV: right ventricle. Sample size: n=5 mice per group. D: Representative images of septal tissue sections stained with Picrosirius Red (collagen) from the specified groups.

si644-mediated p38δ silencing in the heart protected against DOX-induced electrical cardiac remodeling in female mice.

Next, we performed ex vivo optical mapping of transmembrane potential (V_m) and intracellular calcium transients (CaT) in hearts from female mice treated with saline (untreated control, UNT), siNTC, or si644 on Day −6. On Day 0, all groups were treated with DOX and harvested 11 days after DOX exposure or upon reaching a clinical endpoint (two consecutive health scores of 0.25). Due to lower survival among DOX-treated UNT and siNTC groups compared with DOX-treated si644, and since mortality and morbidity between UNT and siNTC groups did not differ significantly (data not shown), for subsequent optical mapping analysis, we combined UNT and siNTC mice in one group, which we refer to as the Control group (CNTL), to have comparable experimental cohort sizes for CNTL- and si644-treated animals for this analysis.

Representative optical action potential (OAP) traces recorded from CNTL+DOX- and si644+DOX-treated hearts illustrate differences in OAP morphology between the two experimental groups (Fig. 6A). Activation maps generated from OAP traces revealed similar activation time patterns across groups (Fig. 6B). To quantify electrophysiological remodeling, we analyzed OAP restitution properties, including OAP rise time (V_m RT), action potential duration at 50% (APD₅₀) and 80% (APD₈₀) repolarization, conduction velocity in the longitudinal (CV_L) and transverse (CV_T) directions, and the anisotropic ratio of conduction velocity (AR_CV) at different basic cycle lengths (BCL) (Fig. 6C–H). Notably, si644-mediated p38δ silencing protected against DOX-induced APD prolongation, significantly reducing APD₅₀ and APD₈₀ (Fig. 6D, E); however, it did not prevent DOX-induced CV_L slowing (Fig. 6F).

Fig. 6.

si644-mediated silencing of p38δ reduces DOX-induced electrical cardiac remodeling in female mice. A, B: Representative optical action potentials and activation maps recorded from CNTL+DOX (blue trace) and si644+DOX (magenta trace) treated female mice, respectively. C – H: Restitution curves are shown for transmembrane potential (V_m RT), action potential duration at 50% and 80% repolarization (APD₅₀, APD₈₀), longitudinal and transverse conduction velocities (CV_L and CV_T), and anisotropy (AR_CV). I: Representative optical CaT traces recorded from CNTL+DOX (blue trace) and si644+DOX (magenta) treated mice. J – L: Restitution curves are displayed for calcium rise time (Ca²⁺ RT), calcium transient duration at 80% reuptake (CaTD₈₀), and calcium decay constant (Ca²⁺ τ). Nonlinear regression analysis with an exponential plateau model identified differences among groups. BCL, basic cycle length. Sample sizes: n = 16 in CNTL+DOX group; n = 15 in si644+DOX group.

Besides remodeling OAP morphology, DOX promotes a proarrhythmic substrate in the heart through calcium mishandling (32). Representative optical CaT traces from the hearts of the CNTL+DOX and si644+DOX-treated groups highlight differences in CaT morphology (Fig. 6I). Hearts from the si644+DOX-treated group showed protection against DOX-induced prolongation of calcium transient duration at 80% reuptake (CaTD₈₀) (Fig. 6K) and exhibited a trend toward protection from prolonged calcium decay (Ca²⁺ τ, p = 0.0963) (Fig. 6L). Additionally, hearts from this group showed a trend toward improved calcium rise time (Ca²⁺ RT, p = 0.085) (Fig. 6J). These findings suggest that si644-mediated p38δ silencing reduced DOX-induced electrophysiological remodeling and calcium mishandling, contributing to its cardioprotective effects against DIC.

si644-mediated p38δ silencing reduced DOX-induced inflammation and the senescence-associated secretory phenotype (SASP) in the hearts of female mice.

Cardiac inflammation plays a key role in the development of DIC (2,33). To assess the effect of si644-mediated p38δ silencing on DOX-induced cardiac inflammation in female mice, we conducted comprehensive profiling of cytokines and chemokines in hearts from (siNTC+DOX)- and (si644+DOX)-treated female mice collected 11 days after DOX administration or at two consecutive health scores of 0.25. We observed a significant reduction in several pro-inflammatory mediators in (si644+DOX)-treated hearts, including G-CSF, IFN-α, IL-3, IL-4, IL-5, IL-6, CXCL1, M-CSF, and TNF-α, compared to (siNTC+DOX)-treated controls (Fig. 7 and Supplemental Table 4). Furthermore, there were trends indicating lower levels of GM-CSF (Fig. 7E), IL-13 (Fig. 7K), IL-15 (Fig. 7L), as well as IL-17A, IL-18, IL-28B, and CCL4 (Supplemental Table 4).

Fig. 7.

si644-mediated p38δ silencing reduces DOX-induced heart inflammation and the senescence-associated secretory phenotype (SASP) in female mice. A – C: Heatmap showing chemokine and cytokine protein expression profiles in the hearts of siNTC+DOX-treated and si644+DOX-treated female mice 11 days after DOX administration or when the animals reached two consecutive health scores of 0.25. D – O: Quantification of the indicated chemokines and cytokines with a p < 0.07 is shown. Unpaired two-tailed Student’s t tests were used to compare differences in chemokine and cytokine protein levels. Sample size: n = 5 mice per group.

DOX is known to induce cardiac cell senescence and the release of pro-inflammatory SASP components, which have been linked to DIC (2,34). Interestingly, several prominent SASP components, including IL-6, CXCL1, G-CSF, M-CSF, GM-CSF, and IL-13, were among the pro-inflammatory factors whose levels were decreased in (si644+DOX)-treated female mice hearts.

DOX also induces hepatotoxicity characterized by pro-inflammatory cytokine production and immune cell recruitment (35). We next assessed whether p38δ silencing modulates DOX-induced hepatic inflammation. Comprehensive cytokine and chemokine profiling of liver tissue collected under the same conditions revealed that (si644+DOX)-treated livers exhibited lower CCL2 and IL-1β levels than (siNTC+DOX)-treated controls, indicating attenuation of DOX-induced proinflammatory signaling in the liver (Supplemental Fig. 3 and Supplemental Table 5). No significant differences in inflammatory mediators were observed in skeletal muscle under the same conditions (data not shown).

These findings demonstrate that si644-mediated silencing of p38δ significantly modifies the heart’s inflammatory environment by decreasing DOX-induced inflammation and SASP while also reducing hepatic inflammatory responses to DOX. These changes correspond with the reduced DIC observed in the (si644+DOX) group compared to the (siNTC+DOX) group.

The systemic deletion of p38δ in female mice caused widespread transcriptional changes in DOX-treated hearts, promoting protection against DIC.

To investigate the molecular mechanisms behind cardioprotection against DIC caused by targeting p38δ in female mice, we performed bulk RNA sequencing on cardiac samples from p38δ KO (p38δ^−/−) and WT female mice three days after DOX administration at 30 mg/kg (Fig. 8A). Principal component analysis showed a clear separation between p38δ^−/− and WT samples, although one WT sample behaved differently (Fig. 8B).

Fig. 8.

Global p38δ deletion modulates cardiac gene expression and related pathways during DIC in female mice. A: Study protocol diagram showing 30 mg/kg of DOX administered to WT C57BL/6 or p38δ^−/− female mice (n = 5 per group), with heart tissues collected three days later for bulk RNA sequencing. Created in BioRender. Amaral, P. (2026) https://BioRender.com/i8oz7yq. B: Principal component analysis shows clustering of gene expression profiles in DOX-treated WT (blue dots) and p38δ^−/− (magenta dots) samples. C: Volcano plot highlights downregulated (blue) and upregulated (magenta) DEGs in DOX-treated p38δ^−/− samples compared to WT controls. DEGs were identified using the DESeq2 package in R, with a significance cutoff of padj < 0.1 and a log2 fold change >1 for upregulated and <−1 for downregulated genes. D: Heatmap displays normalized gene expression of all DEGs across each mouse in each group. E, F: Top 10 downregulated (e) and upregulated (f) GO BP in DOX-treated p38δ^−/− samples compared to WT controls. G, H: Top 10 downregulated (g) and upregulated (h) KEGG pathways. I: Top 5 suppressed Hallmark gene sets from GSEA. J: Network analysis of the Hallmark Myogenesis suppressed gene set. The EnrichGO package in R was used to identify enriched GO BP. The EnrichKEGG package in R was used to identify enriched KEGG pathways. The ClusterProfiler package in R and Hallmark Molecular Signatures Database (MSigDB, category H) were used for GSEA. An FDR q-value < 0.05 was considered significant for enrichment. Sample size: n = 5 mice per group. DEG: differential gene expression, padj: p-adjusted value.

The analysis of differentially expressed genes (DEGs) identified 55 DEGs in DOX-treated p38δ KO female hearts compared to their WT counterparts, including 19 downregulated and 36 upregulated genes (Figs. 8C, D). Among these, the downregulated DEGs in DOX-treated p38δ KO hearts included Cacna1h, a gene encoding a subunit of a voltage-dependent Ca²⁺ channel involved in cardiac electric impulse generation, and the Slc39a8 gene, which encodes a metal-ion transporter ZIP8 involved in mediating lipid peroxidation and ferroptosis (Fig. 8D) (36). p38δ depletion in DOX-treated female hearts also caused a reduction in the expression of several inflammation-related genes, including Tagap1, Chrdl1, C2, Chst4, Ifit3, and Ptgds, indicating decreased inflammatory signaling (Fig. 8D) (37–39).

In contrast, several upregulated DEGs were linked to the dynein motor complex, including Dnah8, Dynlt1b, Dynlt1a, and Dynlt2a2 (Fig. 8D), suggesting possible changes in intracellular transport functions. Additionally, genes involved in counteracting oxidative stress, such as Glo1 and Nnt, were upregulated (Fig. 8D), indicating an adaptive antioxidant response (40,41). An autophagy regulator, Wdfy1, which is a part of the E3 ubiquitin ligase complex (42), was also upregulated in DOX-treated p38δ KO female hearts, along with Hmga1b, a gene involved in chromatin remodeling and gene regulation (Fig. 8D) (43).

Gene Ontology (GO) biological process (BP) analysis revealed that p38δ deficiency affected multiple pathways in response to DOX (Fig. 8E, F). The downregulated GO BPs included pathways related to cation transport, leukocyte adhesion, and excitatory synaptic transmission (Fig. 8E), which aligns with the observed anti-inflammatory and electrophysiological effects of p38δ deletion (8). The upregulated GO BPs involved pathways related to small GTPase activity, Rac protein signaling, and the regulation of mitochondrial membrane permeability (Fig. 8F), suggesting potential mechanisms for controlling cytoskeletal organization and mitochondrial protection. KEGG pathway analysis further supported these findings, highlighting upregulation of motor protein-related pathways (Fig. 8H). Importantly, the downregulated KEGG pathways included those linked to ferroptosis and sex hormone signaling, such as GnRH secretion and progesterone-mediated oocyte maturation (Fig. 8G).

Moreover, Gene Set Enrichment Analysis (GSEA) revealed a significant suppression of gene sets associated with early and late estrogen response, Kras signaling downregulation, the apical junction, and myogenesis in DOX-treated p38δ KO female hearts compared to DOX-treated WT controls (Fig. 8I, J, and Supplemental Tables 6–10). The suppression of the Hallmark Myogenesis gene set supports the electrophysiological remodeling observed through optical mapping, gene expression analysis, GO BP, and KEGG. Notably, the suppressed Hallmark Myogenesis genes include Cacna1h and Sln (sarcolipin), an inhibitor of the calcium-pumping activity of the cardiac sarcoplasmic/endoplasmic reticulum Ca2+ ATPase (SERCA2a) (Fig. 8J). Overall, the transcriptome analyses suggest that deleting p38δ influences DOX-induced transcriptional responses, leading to reduced DIC in female hearts.

si644 effectively silences p38δ in ex vivo human organotypic cardiac slices.

Human ex vivo organotypic cardiac slices bridge the gap between isolated cells and whole-heart preparations, preserving native cell diversity and maturity, intercellular connections, and extracellular matrix (44,45). Derived from donor hearts with mature cardiomyocytes, they provide a clinically relevant model for studying the adult human heart ex vivo. Using this model, we evaluated whether si644 can effectively silence p38δ in ex vivo human heart tissue. Human LV organotypic cardiac slices were cultured with 1 μM si644 or siNTC for 72 hours (Fig. 9A, B). After treatment, we observed 61% mRNA silencing and 77% protein silencing of p38δ in si644-treated slices compared to siNTC-treated or untreated controls (CNTL), confirming the efficacy and potency of si644 in silencing p38δ in human LV organotypic cardiac slices (Fig. 9C–E). Furthermore, these results demonstrate the feasibility of the siRNA drug strategy outlined in this work for silencing other disease targets in ex vivo human heart tissue.

Fig. 9.

si644 effectively silences p38δ in ex vivo cultured human LV organotypic cardiac slices. A: Human LV organotypic cardiac slices from donor hearts unsuitable for transplant were prepared as described in the Methods section (n = 5 per group). Slices were treated in triplicate per heart with either culture media alone (untreated control, CNTL) or media containing 1 μM of siNTC or si644 for 72 hours, with media changed and fresh siRNA added at 48 hours (indicated by the arrow). B: LV slices. C: p38δ/MAPK13 mRNA levels in slices were measured using the QuantiGene 2.0 assay. D, E: Representative immunoblot (d) and quantification (e) of p38δ protein levels normalized to α-actinin in the specified samples. A one-way ANOVA was used to assess the statistical significance between the treatment groups.

DISCUSSION

In this study, we developed a DCA-conjugated, fully chemically stabilized siRNA that targets p38δ, achieving efficient, durable, and non-toxic silencing of the p38δ/MAPK13 gene in the heart. Our lead compound, si644, effectively silenced p38δ expression both in vivo in mouse hearts and ex vivo in organotypic human heart tissue. Importantly, si644 treatment was well tolerated in vivo, consistent with previous reports demonstrating the safety of DCA-siRNA compounds in animal models (20,30). Furthermore, si644-mediated p38δ silencing protected female mice against acute DIC (Fig. 10), aligning with our earlier findings of a similar protective effect from systemic genetic p38δ ablation in female mice treated acutely with DOX (8). Given p38δ’s broad expression and its involvement in cardiac and non-cardiac diseases, p38δ-selective siRNA offers potential for both mechanistic studies and targeted therapies beyond DIC.

Figure 10. siRNA-mediated p38δ silencing mitigates DIC.

si644-mediated p38δ silencing improves survival and health, reduces wall thickening and interventricular septal fibrosis, prevents APD prolongation, restores calcium handling, and alleviates inflammation/SASP. Created in BioRender. Amaral, P. (2026) https://BioRender.com/urpnzs8.

Current treatments for DIC include pharmacotherapies, medical devices, and heart transplants in severe cases (46). While these approaches manage symptoms, they do not prevent DIC. Therefore, there is an unmet need for DIC therapies that do not compromise the effectiveness of DOX in cancer treatment. Preclinical studies have linked p38 kinases to DIC (2,8–10). Our previous work showed that p38δ systemic genetic deletion protected female mice from DIC by increasing cardiac autophagy (8). The protective effects of p38δ deletion have also been observed in models of non-alcoholic fatty liver disease (47), bacterial endotoxin-induced hepatitis (48), and various cancers (11–15). Despite the potential benefits of targeting p38δ, this kinase remains understudied, and currently, no p38δ-specific small-molecule kinase inhibitors are commercially available (17,18). This underscores the strong rationale for using RNAi technology to develop targeted p38δ-specific siRNA that effectively silences human and mouse p38δ/MAPK13 mRNA, shows therapeutic promise in a mouse model of acute DIC, and reduces DOX-induced cardiac damage in female mice.

DIC typically appears as dilated cardiomyopathy with reduced EF (5). However, transient increases in EF have been observed during acute DOX treatment, likely due to a hypercontractile stress response (49). Our previous research shows that EF remains stable or temporarily increases after acute DOX treatment (8,9). The rise in EF in our acute DIC model probably results from decreases in LV chamber size, which mechanically elevates EF despite overall reduced cardiac function. In this study, DOX-treated female mice with intact p38δ expression (i.e., siNTC-treated) displayed decreased CO and a higher ratio of LV weight to body weight, along with thicker LV anterior and posterior walls during systole and diastole (LVaWs, LVaWd, LVpWs, and LVpWd), indicating pathological cardiac remodeling. Silencing p38δ with si644 preserved these parameters, suggesting protection against DOX-induced structural and functional decline. Notably, previous research has demonstrated that systemic genetic deletion of p38δ reduces LV mass and cardiomyocyte size without affecting EF or FS under unstimulated conditions, and it prevents angiotensin II-induced hypertrophy (26).

DOX is also known to cause electrical remodeling, with QT prolongation being a key symptom (32,50). At the cellular level, DOX prolongs APD by disrupting cardiac ion channel expression and function, including reduced rapidly and slowly activating delayed rectifier potassium currents (I_Kr and I_Ks) (32). Although less studied, DOX’s effect on the transient outward potassium current (I_to) is likely important, especially in rodent models where I_to is the main repolarizing current in ventricular myocytes (51). In disease states such as oxidative stress, hypertrophy, ischemia, and diabetic cardiomyopathy, downregulation of I_to has been linked to APD prolongation (52,53). In our study, silencing p38δ effectively prevented DOX-induced APD prolongation, indicating a regulatory role for p38δ in cardiac ion channel expression and reducing electrical remodeling.

Besides affecting V_m, DOX disrupts excitation-contraction coupling (ECC) by promoting RyR2 leak and impairing SERCA2a activity (54), both of which contribute to contractile dysfunction. We observed that silencing p38δ reduced these calcium-handling abnormalities. Although few studies directly link p38δ with ECC or cardiac ion channel regulation, recent evidence shows that p38δ, along with p38γ, contributes to arrhythmia susceptibility by modulating calcium handling and repolarizing currents, including RyR2 activity and I_to, in aging mouse hearts (55). These findings align with our results and support a role for p38δ silencing in maintaining electrophysiological and calcium balance during DOX stress.

DIC involves a heightened inflammatory environment driven by activation of the innate immune system (2,33,56). This inflammatory state promotes immune cell recruitment, fibroblast activation, extracellular matrix remodeling, and cellular stress, ultimately leading to senescence and cardiac dysfunction (2,33). After DOX-induced myocardial injury, increased numbers of senescent cardiomyocytes, endothelial cells, cardiac fibroblasts, and cardiac progenitor cells have been observed (2,57–59). These senescent cells develop a proinflammatory phenotype, releasing SASP factors that amplify immune cell recruitment and exacerbate tissue injury, thereby further promoting cardiac damage (60). In this study, we found that silencing p38δ with si644 reduces the secretion of SASP factors, including IL-6, IL-13, IL-15, G-CSF, GM-CSF, CXCL1, along with other proinflammatory cytokines and chemokines in the heart during DIC. These findings suggest that p38δ silencing reduces DOX-induced cardiac inflammation by suppressing SASP, consistent with the protection observed in p38δ KO and knockdown female mice ((8) and this study). Given inflammation’s key role in many types of cardiac injury, the anti-inflammatory effects of p38δ inhibition could have broader therapeutic benefits beyond DIC.

Beyond the heart, DOX is well known to elicit a pro-inflammatory response in the liver, contributing to hepatotoxicity characterized by hepatocyte stress, immune cell infiltration, and fibrotic remodeling. Although cardiotoxicity remains the predominant and dose-limiting adverse effect of doxorubicin therapy, hepatic inflammation is a recognized secondary toxicity that may contribute to the systemic inflammatory burden during treatment (61). DOX exposure increases hepatic expression of inflammatory mediators that promote the recruitment and activation of immune cells, thereby exacerbating liver injury (35). Our findings demonstrate that p38δ silencing attenuated DOX-induced increases in hepatic CCL2, a well-established marker of liver injury that amplifies monocyte and macrophage recruitment (62). In addition, p38δ silencing reduced hepatic IL-1β expression, a pro-inflammatory cytokine and downstream effector of NLRP3 inflammasome signaling (63). Consistent with our findings, prior work has demonstrated that p38δ regulates hepatic inflammatory signaling and immune cell recruitment, and that genetic loss of p38δ attenuates liver inflammation and neutrophil infiltration (47). Together, these findings suggest that p38δ modulates DOX-induced hepatic inflammatory pathways by regulating immune cell recruitment and NLRP3 inflammasome-associated cytokine production. Suppression of hepatic inflammation may reduce the systemic inflammatory burden during DIC and potentially contribute to the cardioprotective effects observed with si644 treatment.

Our findings showed that p38δ systemic genetic deletion caused widespread transcriptomic changes in the hearts of DOX-treated female mice three days after treatment. Bulk RNA sequencing of DOX-treated female WT and p38δ KO hearts revealed significant changes in the expression of several genes previously linked to the regulation of DOX-induced cardiac injury, which may contribute to the observed protection against DIC when targeting p38δ. Collectively, these transcriptional changes suggest that p38δ deletion may confer cardioprotection by enhancing oxidative stress management (Glo1 and Nnt) (40,41), improving intracellular transport (Dnah8, Dynlt1b, Dynlt1a, Dynlt2a2) (64), reducing inflammation (Tagap1, Chrdl1, C2, Chst4, Ifit3, and Ptgds) (65,66), inhibiting pro-arrhythmic and hypertrophic pathways (Cacna1h) (67), promoting epigenetic remodeling (Hmga1b) (43), and decreasing apoptosis and ferroptosis (Slc39a8) (68). This underscores the complex and multifaceted regulatory role of p38δ in DIC.

GO BP and KEGG pathway enrichment analyses further supported these DEG changes, revealing upregulation of pathways related to cytoskeletal organization and intracellular trafficking, along with downregulation of pathways associated with inflammation, ferroptosis (69), and electrical remodeling. Together, these transcriptomic changes suggest that p38δ deletion promotes an adaptive stress response while limiting maladaptive remodeling after acute DOX exposure. Interestingly, KEGG and GSEA analyses also identified suppression of hormone-related pathways, including GnRH secretion, progesterone-mediated oocyte maturation, and early and late estrogen response gene sets, suggesting a potential endocrine role in the female-specific cardioprotection observed in p38δ KO mice. These findings indicate that p38δ deletion may reprogram hormone-sensitive pathways to enhance cardioprotection in females and potentially improve cancer treatment in hormone-responsive (HR)-positive breast cancer.

In fact, the anticancer effects of p38δ gene deletion have been demonstrated in several mouse cancer models, including breast, colon, skin, and endometrial cancers (11–15). Additionally, the p38δ gene has been identified as a pro-oncogenic, tumor-essential gene in human genomic studies (16), indicating that DOX’s anticancer effectiveness is unlikely to decrease and may even improve by targeting p38δ for DIC protection.

Consistent with the tumor-promoting role of p38δ, analysis of The Cancer Genome Atlas Breast Cancer (TCGA BRCA) dataset from primary breast cancer patient samples (29) using the GEPIA2 bioinformatics tool (28) showed that p38δ/MAPK13 expression was significantly higher in tumor samples (n = 1085) than in normal breast tissues (n = 291) (Supplemental Fig. 4A). Furthermore, BRCA patients with low p38δ/MAPK13 levels had significantly longer overall survival than those with high p38δ/MAPK13 levels (Supplemental Fig. 4C). In contrast, p38α/MAPK14 levels showed no significant difference between BRCA tumors and normal tissues, and the overall survival of BRCA patients was not significantly affected by the expression levels of this isoform (Supplemental Fig. 4B, D). These human data highlight the translational relevance of targeting p38δ for DIC protection and anticancer responses.

In this study, we employed a well-established acute DIC mouse model characterized by rapid, severe cardiac injury after a single, near-lethal dose of DOX. This model is valuable for investigating mechanisms of acute cardiac stress and for establishing proof-of-concept for interventions that reduce severe acute cardiac damage and mortality caused by DOX. However, because DOX is administered in multiple cycles in clinical settings, a chronic model with repeated smaller doses would better evaluate the effect of p38δ targeting on progressive cardiac dysfunction. Although p38δ targeting has been linked to anti-oncogenic effects in several cancer models (11–15), its influence on DOX-mediated tumor suppression remains unexplored. Therefore, investigating the impact of p38δ targeting in tumor-bearing mice treated with chronic DOX regimens would provide important insights into cardioprotective effects while preserving anti-tumor efficacy. Here, we evaluated the effectiveness of siRNA-mediated p38δ silencing as a cardioprotective strategy in female mice, based on prior evidence of sex-specific protection from DIC in p38δ knockout mice (8). Future work should evaluate the effects of p38δ silencing on DIC in male mice and incorporate both acute and chronic DIC models to more fully define the translational potential of p38δ silencing in DIC.

In summary, we report the development of a p38δ-selective siRNA inhibitor that effectively silences the MAPK13 gene in the heart, shows cross-reactivity between human and mouse p38δ, and is well-tolerated in vivo. Importantly, silencing p38δ provides therapeutic protection against DIC in female mice (Fig. 10). Mechanistic studies show that p38δ silencing influences multiple disease-related pathways, including ECC, fibrosis, inflammation, oxidative stress, intracellular trafficking, ferroptosis, and hormone-responsive signaling. These findings not only reveal a new regulatory role for p38δ in cardiac stress responses but also suggest that this inhibitor could benefit cancer patients by mitigating anthracycline-induced heart damage without compromising the anti-cancer effects of DOX. The inhibitor’s cross-species reactivity further strengthens its translational potential and supports ongoing efforts toward clinical application.

Supplementary Material

Supplemental Tables 1–11, Supplemental Figures 1–4: https://doi.org/10.6084/m9.figshare.31306834

NONSTANDARD ABBREVIATIONS AND ACRONYMS

APD₅₀

Action Potential Duration at 50% Repolarization

APD₈₀

Action Potential Duration at 80% Repolarization

BP

Biological Pathway

BPM

Beats Per Minute

CaT

Calcium Transient

CaTD₈₀

Calcium Transient Duration at 80% Reuptake

Ca²⁺ RT

Calcium Rise Time

Ca²⁺ τ

Calcium Decay

CNTL

Control

CO

Cardiac Output

CV_L

Longitudinal Conduction Velocity

CV_T

Transverse Conduction Velocity

DIC

Doxorubicin-Induced Cardiotoxicity

DCA

Docosanoic Acid

DEG

Differential Gene Expression

DOX

Doxorubicin

ECG

Electrocardiogram

EF

Ejection Fraction

FDR

False Discovery Rate

FS

Fractional Shortening

GO

Gene Ontology

GSEA

Gene Set Enrichment Analysis

KEGG

Kyoto Encyclopedia of Genes and Genomes

KD

Knockdown

KO

Knockout

LFQ-MS

Label-Free Quantification Mass Spectrometry

LFC

Log Fold Change

LV

Left Ventricle

LVaWd

Left Ventricular Anterior Wall during Diastole

LVaWs

Left Ventricular Anterior Wall during Systole

LVEDD

Left Ventricular End-Diastolic Diameter

LVESD

Left Ventricular End-Systolic Diameter

LVEDV

Left Ventricular End-Diastolic Volume

LVESV

Left Ventricular End-Systolic Volume

LVpWd

Left Ventricular Posterior Wall during Diastole

LVpWs

Left Ventricular Posterior Wall during Systole

MAPK

Mitogen-Activated Protein Kinase

NTC

Non-Targeting Control siRNA

OAP

Optical Action Potential

padj

P-Adjusted Value

PBS

Phosphate-Buffered Saline

RNAi

RNA interference

ROS

Reactive Oxygen Species

RV

Right Ventricle

SASP

Senescence-Associated Secretory Phenotype

siRNA

Small Interfering RNA

SV

Stroke Volume

TCGA

The Cancer Genome Atlas

UTR

Untranslated Region

UNT

Untreated

V_m RT

Transmembrane Potential Rise Time

WT

Wild-type

DATA AVAILABILITY

Data will be made available upon reasonable request.