Urine proteomics defines an immune checkpoint-associated nephritis signature

Abstract

Immune checkpoint inhibitor (ICI) therapy is a cornerstone treatment for many cancers, but it can induce severe immunotoxicity, including acute interstitial nephritis (AIN). Currently, kidney biopsy is required to differentiate ICI-AIN from other causes of acute kidney injury (AKI). However, this invasive approach can lead to morbidity, delayed glucocorticoid treatment for patients with AIN, and unnecessarily prolonged suspension of ICI therapy in non-AIN patients. Delayed or incorrect diagnosis of ICI-AIN is particularly detrimental, as over 50% of patients are at risk of permanent renal damage. Thus, there is an urgent need for non-invasive biomarkers that can rapidly and accurately distinguish ICI-AIN from other causes of AKI.The urine and plasma proteome contain actively secreted proteins that provide real-time insights into dynamic physiological processes. However, identification of effective biomarkers of disease using established technologies such as proximity ligation assays (PLA) and bead-based immunoassays is challenging due to their limited sensitivity and loss of precision in multiplex analysis.To address this, we employed cutting-edge NUcleic acid Linked Immuno-Sandwich Assay (NULISA) technology to measure protein expression in urine and plasma samples from AKI patients undergoing ICI therapy. NULISA offers 10,000-fold greater precision than PLA, enabling quantification of over 200 inflammatory proteins with unprecedented precision. Our analysis revealed that urine was more sensitive and specific than plasma in distinguishing ICI-AIN from non-AIN cases. Pathway analyses highlighted the involvement of JAK-STAT and tumor necrosis factor (TNF) signaling in ICI-AIN pathogenesis. We identified several novel urine biomarkers, including IL-5, Fas, TNFSF4, CD274, IL-20, TNFSF15, TSLP, TREM1 and CCL1 while confirming previously reported markers such as CXCL9 and TNF-α. Using statistical and machine learning methods, we constructed a novel urine biomarker signature-IL-5+Fas-that achieved an area under the curve of 0.94 for diagnosing ICI-AIN.By leveraging high-sensitivity proteomics, we developed a non-invasive strategy for diagnosing ICI-AIN. This approach will enable earlier intervention to mitigate immunotoxicity, preservation of antitumor efficacy of ICI therapy in non-AIN patients, and safe rechallenge of ICI therapy in patients previously treated for ICI-AIN.

Keywords: Biomarker; Immune Checkpoint Inhibitor; Immune related adverse event - irAE; Immunotherapy; Nephrotoxicity.

Figure 1. Analysis of protein expression pattern in urine and plasma samples. (A) Heatmap demonstrating distinct protein expression patterns in all urine and plasma samples. (B) Principal component analysis (PCA) shows a clear separation between urine and plasma specimens. (C) Scatterplot comparing fold changes of all proteins in ICI-AIN versus non-AIN samples in urine and plasma. Urine contains more differentially expressed proteins with greater fold changes compared with plasma. (D) Top pathways identified in urine specimens using the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis. Proteins with false discovery rate (FDR) <0.05 and absolute log 2-fold change >1 were analyzed using Enrichr. (E) Top pathways identified in plasma specimens under same criteria. AIN, acute interstitial nephritis; ICI, immune checkpoint inhibitor.

Figure 2. Identification of key markers in urine for differential diagnosis of ICI-AIN and non-AIN. (A) Volcano plot illustrating the differential expression of proteins in urine between the ICI-AIN (n=22) and non-AIN (n=27) groups. Each dot represents an individual protein. The x-axis shows the log2 fold change and the y-axis represents the −log10 FDR (adjusted p value). Blue dots represent proteins with an AUC > 0.8 (p-adj <0.01). P values were calculated using the Wilcoxon test and FDR corrected. (B) Dot plots of the top 10 urinary proteins distinguishing ICI-AIN from non-AIN cases. Statistical analysis was performed using Tukey’s multiple comparison test: *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001(p-adj). (C) Heatmap of top markers showing a clear separation between ICI-AIN and non-AIN cases. (D) PCA of the top markers demonstrating clear separation between the ICI-AIN and non-AIN groups. AIN, acute interstitial nephritis; FDR, false discovery rate; ICI, immune checkpoint inhibitor; NPQ, NULISA Protein Quantification.

Figure 3. Development and performance of the urine immune checkpoint-associated nephritis signature. (A) Bootstrap logistic regression analysis identified IL5 and Fas as top markers, selecting them in over 80% of models. (B) Classification and Regression Tree (CART) analysis identified Fas and IL5 as key markers for ICI-AIN. Of 49 cases (n=22 ICI-AIN, n=27 non-AIN), 14 ICI-AIN cases had a Fas greater or equal to 10.85 NPQ. IL5 greater or equal to 10.11 NPQ identified 6/8 remaining ICI-AIN cases and 25/27 non-AIN cases. Fas and IL5 levels are expressed in NPQ units. (C) Performance comparison of the urine ICI-AIN signature, using IL5 and Fas, against individual markers. The IL5+Fas signature achieved an AUC of 0.94, outperforming Fas, IL5, or CXCL9 alone. (D) Scatterplot of Urine Fas Expression versus Urine IL5 Expression. Each dot represents a patient sample. The black line depicts the threshold of equal probability for ICI-AIN and non-AIN, as predicted by the logistic regression model. AIN, acute interstitial nephritis; AUC, area under the curve; ICI, immune checkpoint inhibitor; NPQ, NULISA protein quantification; NULISA, Nucleic Acid Linked ImmunoSandwich Assay.