Proteomics: Predicting Steroid Resistance in ICI Myocarditis
Immune checkpoint inhibitor-associated myocarditis (ICIAM) is rare, but pharmacovigilance data from Moslehi and colleagues (2018) put the fatality rate among reported cases near 50%. High-dose methylprednisolone comes first. Some patients never respond, and nothing measured at diagnosis tells clinicians who they are. Zhang and colleagues now describe a serum proteomics score, built in 33 patients and rechecked by ELISA in 14 more, that aims to flag steroid resistance before the first dose.
Key Takeaways
Baseline serum proteomics separated steroid-resistant from steroid-sensitive immune checkpoint inhibitor myocarditis better than routine labs such as LDH or troponin.
Three serum proteins, CXCL9, TRANCE and MMP1, may flag ICI myocarditis patients who need closer monitoring and earlier escalation planning.
In ICI myocarditis, steroid responders showed broader drops in serum inflammatory proteins, suggesting protein dynamics could become a readout of treatment effect.
The steroid-resistance model is preliminary, drawn from a single center with 47 patients and still lacking external validation.
Targeted Serum Proteomics Before and After Steroids
Of 63 patients diagnosed between 2023 and 2025, 47 had serum banked before steroids began. Resistance (srICIAM, versus steroid-sensitive ssICIAM) meant troponin falling less than 50% from peak, or persistent conduction or ventricular problems, after 72 hours of intravenous methylprednisolone. Protein profiling used the Olink Target 96 Inflammation panel, a proximity extension assay covering 92 immune proteins, with every sample in one run and no batch correction across plates required. Only proteins chosen by all three of LASSO, random forest and SVM-RFE entered the final model.
Key Findings
Steroids reshaped the serum proteome unevenly: 29 of 92 proteins fell nominally after treatment. Sensitive patients showed 22 decreases against 13 in resistant ones, and IL-22RA1 nominally rose in resistant patients while falling in sensitive ones.
Close coupling between CXCL9 and injury markers: CXCL9 correlated with LDH (rho 0.655) and troponin T (rho 0.538), both p < 0.001.
A three-protein score beat every single marker: Higher CXCL9 with lower TRANCE and MMP1 gave a development AUC of 0.91, against 0.81 for MMP1 alone and 0.76 for LDH.
Cross-platform hold-out check: Measured by ELISA in 14 held-out patients, the score reached an AUC of 0.88 after a new platform-specific cutoff was set.
Figure 1. Serum immune proteome dynamics during steroid therapy for ICI myocarditis. Panel A shows the cohort flow and training/validation split, and Panel B is a Venn diagram of proteins that changed in steroid-sensitive and steroid-resistant patients. Panels C through F are volcano plots of paired post- versus pre-treatment shifts, and Panel G tracks individual IL-22RA1 levels. Adapted from Zhang et al. (2026), Clinical and Translational Medicine.
Reading the CXCL9, TRANCE and MMP1 Signature
CXCL9 is the easy one to explain. Interferon-gamma induces it in macrophages, and the CXCL9/CXCL10 and CXCR3 axis has already been implicated in ICI myocarditis, which is why the authors float CXCL9 blockade as a therapy. The TRANCE (RANKL) story leans on glucocorticoid receptor signaling, and the MMP1 link runs through PD-L1. Without tissue or functional data, both remain hypotheses.
Where a Baseline Score Fits in Cardio-Oncology Care
Second-line drugs such as tofacitinib or intravenous immunoglobulin usually start only after 72 hours of steroids have visibly failed. A baseline score could move that planning to day one. No routine clinical variable, from age to number of ICI cycles, predicted resistance here, which strengthens the case for a protein readout. Our earlier summary of cathepsin C as a marker of cancer therapy heart risk covers a related cardiotoxicity question in breast cancer.
From Affinity Panel to Transferable Assay
For immuno-oncology sponsors, the lesson sits in the validation step: Olink reports NPX, a relative log2 unit, so the authors had to set a fresh ELISA cutoff. We hit the same wall in client work at Dalton when affinity-panel hits move toward a clinical assay, and we usually recommend confirming them by targeted MS-based proteomics with isotope-labeled peptide standards, which yields absolute amounts that travel between sites. That staged hand-off is the backbone of any multi-omics biomarker discovery program meant to survive a multicenter trial.
Frequently Asked Questions
What is proteomics used for in clinical biomarker research?
Proteomics measures hundreds to thousands of proteins at once in tissue or biofluids such as serum. Clinical teams use it to find proteins that separate patient groups or predict drug response, with either antibody panels like Olink or mass spectrometry.
Can a blood test predict steroid-resistant ICI myocarditis?
Not in routine care. A three-protein serum score reached an AUC of 0.91 in development and 0.88 by ELISA in a 14-patient hold-out, but the authors call it preliminary.
How many patients does a serum proteomics biomarker study need?
Pilot discovery often starts with a few dozen patients, like the 33 used for training here, but small cohorts overfit easily and need an independent validation set. Sample volumes are small, about 1 microliter of serum for affinity panels and typically under 50 microliters for MS-based workflows.
Conclusion
It's now believable that steroid-resistant ICI myocarditis leaves a serum signature before the first dose, with CXCL9 the most credible piece. Whether any cutoff holds outside one Shanghai center, or whether acting on it changes outcomes, isn't proven. Trial sponsors can help cheaply by banking pretreatment serum from every patient who develops myocarditis.
Related Reading
See how we run these analyses in one lab: Dalton's multi-omics CRO services.
Citation
Zhang, Y., Li, Y., Zhang, J., He, X., Shen, Y., Xu, Y., Zhang, H., Pan, J., Wang, X., Zhang, S., Wang, Y., & Cheng, L. (2026). A serum proteomics-based predictive model for steroid-resistant immune checkpoint inhibitor associated myocarditis. Clinical and Translational Medicine, 16(10), e70843. https://doi.org/10.1002/ctm2.70843
Note
This blog post summarizes findings from the above-cited research. Figures are adapted from the original publication. For full details, please refer to the source article.
By Seungjun Yeo, CEO at Dalton Bioanalytics. Specializing in multi-omics mass spectrometry for drug discovery and biomarker research.