Prediction of immune-related adverse events in urological cancer during checkpoint inhibitor immunotherapy through immunohistochemical analysis of tumorous LCP1/ADPGK Authors: Liqing Xu, Xi Zhang, Zihao Tao, Yiceng Ying, Xueqin Zeng, Kaiwei Yang, Yang Wang, Ninghan Feng, Xuesong Li, Yaqian Li, Qi Tang Published: 2026 ABSTRACT Highlights•LCP1 positivity (AUC=0.8415) and ADPGK positivity (AUC=0.8759) individually predicted irAEs, and a bivariate model combining both markers achieved superior performance (AUC=0.9184 in the overall cohort).•LCP1 and ADPGK positivity significantly outperformed conventional hematologic parameters in predicting irAEs across urological cancer subtypes.•The model maintained high predictive efficacy for irAEs across cancer types and irAE onset timing, facilitating personalized ICI dosing to prevent treatment discontinuation.Keywords: Urological cancer, PD-1, PD-L1, irAE, Biomarker, LCP1, ADPGK Introduction In the United States, it is estimated that 170,000 people are diagnosed with urological cancers and >32,000 die from urological cancers each year [1]. In China, a similar number of new cases is reported annually; however, over 65,000 deaths from urological cancers are reported each year [2]. Immunotherapy, especially immune checkpoint inhibitors (ICIs), has revolutionised the treatment of urological cancers [3]. ICIs are now available as first-line treatment for patients who are ineligible for platinum-based chemotherapy. For example, atezolizumab and pembrolizumab have shown objective response rates (ORRs) of 23.5% and 28.6%, respectively, in urothelial carcinoma [[4], [5], [6]]. Furthermore, first-line combination therapies of ICIs and chemotherapy, such as nivolumab plus cisplatin-based chemotherapy, have demonstrated improved overall survival (OS), progression-free survival (PFS), ORR, and complete response rates (CRR) compared to chemotherapy alone [7]. In addition, ICIs, such as nivolumab and avelumab, have also shown efficacy in the post-platinum and maintenance settings [8,9]. Given the activation of the immune system by ICIs, there is a high risk of immune-related adverse events (irAEs) [3]. For instance, the incidence of grade ≥3 treatment-related adverse events (TRAEs) with pembrolizumab and atezolizumab monotherapy is 21.1% and 16%, respectively [3,5,6]. The addition of chemotherapy to ICIs results in a significant increase in the rate of grade ≥3 TRAEs, with 81% of patients treated with atezolizumab and 87% of patients treated with pembrolizumab [10,11]. So, predicting who is at risk of severe irAEs after ICI treatment is essential, especially as only a proportion of patients benefit significantly from these drugs. Many attempts have been made to discover biomarkers for predicting irAEs. Biomarkers such as autoantibodies, peripheral blood cell counts and ratios, serum proteins and cytokines have been suggested as potential predictors of ICI-induced irAEs [[12], [13], [14], [15], [16]]. But most of these markers have not been well validated or have shown inconsistent results in different cohorts [17,18], and some are still controversial even in single-center studies. A recent proteomic analysis of peripheral blood mononuclear cells (PBMCs) found that decreased expression of CXCR3 and CCR6 in PBMC subpopulations is linked to the occurrence of irAEs in patients treated with PD-1 inhibitors, suggesting a new avenue for the discovery of circulating biomarkers [19]. But the consistency and reproducibility of these circulating biomarkers across various tumor types and patient populations need to be confirmed. Furthermore, human leukocyte antigen (HLA) genotyping and the gut microbiome have also been suggested as potential irAE predictors [[20], [21], [22]]. Genetic variants and gene expression have also been explored, as they are known to play a role in autoimmune diseases [23]. Next-generation sequencing [24] and multi-omics analyses using public databases [25] have been used to detect mutations and differentially expressed genes in irAE development. In particular, multi-omics analyses using the U.S. Food and Drug Administration Adverse Event Reporting System (FAERS) and The Cancer Genome Atlas (TCGA) have proposed that the expression of LCP1 and ADPGK in tumors may be predictive of ICI-induced irAEs [25]. LCP1 and ADPGK have been shown to be involved in T-cell activation and immune regulation [[26], [27], [28], [29], [30]]. But whether LCP1 or ADPGK can be used to predict irAEs in patients with urological cancers receiving ICIs is unknown. In this study, we therefore conducted immunohistochemical staining of LCP1 and ADPGK in a number of urological cancer samples and found positive signals. Thus, we reviewed the characteristics of irAEs in a series of patients with urological cancers receiving PD-1 or PD-L1 antibodies and retrospectively assessed peripheral blood cell counts and ratios, as well as tumoral LCP1 and ADPGK expression, to determine the predictive value of these parameters. We found that tumoral LCP1 and ADPGK expression was superior to traditional hematological parameters in predicting the occurrence of irAEs in urological cancer patients treated with PD-1 or PD-L1 antibodies. Methods Patient cohort This study was approved by the Ethics Committee of Peking University First Hospital with the approval ID number of 2024– 490– 001. All patients were informed and signed on consents. Urological patients who underwent surgery and subsequently received anti-PD-1 or anti-PD-L1 therapy at Peking University First Hospital between August 2011 and December 2023 were included. The assessment window for irAEs is defined as from ICIs initiation to one year after the end of ICIs treatment. We used 3 months as a cutoff to define early irAE (≤3 months from ICIs initiation) and late irAE (>3 months from ICIs initiation). Demographics and clinical information were collected by chart review. Patients without surgical formalin-fixed and paraffin-embedded (FFPE) samples were excluded. irAEs were graded according to National Cancer Institute Common Terminology Criteria for Adverse Events version 5.0. Immunohistochemistry Assessment of tumoral LCP1 and ADPGK expression in FFPE tumor samples from before PD-(L)1 treatment were initially reviewed by two pathologists to (1) confirm the malignancy and histological subtype of the urological cancer, and (2) ensure that the tumor tissue constituted >50% of the sample, with <20% necrosis. Immunohistochemistry (IHC) was performed using standard protocols. Primary antibodies for LCP1 (1:500, LiangJue Technology, LJ-mAB-0002) and ADPGK (1:500, LiangJue Technology, LJ-mAB-0001) were applied to the tissue sections. Secondary antibodies and the DAB chromogenic detection system were purchased from Servicebio (G1301– 100ML, G1212– 200T). Hematoxylin was used for counterstaining. Whole-slide scanning was conducted using a PANNORAMIC 1000 scanner (3DHISTECH, Hungary) and CaseViewer 2.4 (3DHISTECH, Hungary) was used for image capture and analysis. Data analysis criteria Images were checked by pathologists XQ.Z. and SQ.W. and oncologists LQ.X. and X.Z. to ensure the staining is qualified for analysis. Images were taken at 600 dpi and were quantified using Aperio ImageScope software (v14.3) with the Positive Pixel Count v9 (PPCv9) algorithm as described previously [25,31]. At least seven representative regions of interest were selected for each sample and the average value was used as the final score for either LCP1 or ADPGK expression. Statistical analysis Data were analysed using IBM SPSS Statistics for Mac OS (version 27.0; IBM Corporation, Armonk, NY, USA). Logistic regression analysis was used to build the bivariate model of LCP1 and ADPGK positivity [25]. Logistic regression analyses were conducted for each individual dataset, with the results shown in Supplementary Table 1. Two-tailed Student’s t-tests were used for comparisons of positivity between two groups. P values < 0.05 were considered significant. Results General demographics and clinical characteristics This study included 112 patients with urological cancers, with an average age of 60.3 ± 12.5 years (Table 1). Of these, 29 patients had bladder cancer (BC), 23 had upper tract urothelial carcinoma (UTUC), and 60 had renal cell carcinoma (RCC). The rate of irAEs was higher in males (50.7%) than in females (35.9%). Overall, 51 of 112 patients developed irAEs of any grade. But there were no differences in the incidence of any-grade irAEs across cancer types. A total of 74 irAEs were observed in 51 (45.5%) patients, and 10 (8.9%) patients experienced grade 3 or grade 4 irAEs (Table 2). In the cohort, 23 patients (45.1%) had early-onset irAEs of any grade, and 21 patients (41.2%) had late-onset irAEs of any grade. Three patients (5.9%) had both early and late irAEs, and the onset of irAEs was unknown for four patients. In terms of organ distribution, endocrine and liver toxicities were relatively more common in both early- and late-onset irAEs. The organ-specific distribution of irAE onset is presented in Supplementary Fig. 3. The rate of severe (grade ≥3) events was 11.8% in the early irAE group and 3.9% in the late irAE group. Table 1Demographics and clinical characteristics.Table 1 dummy alt textNon-irAEirAETotalMaleNo. of patients36 (49.3%)37 (50.7%)73Ave. Age61.959.860.8FemaleNo. of patients25 (64.1%)14 (35.9%)39Ave. Age54.468.3559.4DiagnosisBladder cancer16 (55.2%)13 (44.8%)29Upper tract urothelial carcinoma12 (52.2%)11 (47.8%)23Renal cell carcinoma33 (55%)27 (45%)60ICINivolumab448Pembrolizumab191736Sintilimab121123Tislelizumab201434Toripalimab6511 Demographics and clinical characteristics. Table 2irAE characteristics.Table 2 dummy alt textDiagnosisICIirAEGrade(Highest)RCCTislelizumabPneumonitis; Myocarditis4RCCSintilimabHepatitis4BCPembrolizumabHepatitis3RCCPembrolizumabRash; Hepatitis; Hyperthyroidism; Colitis3RCCTislelizumabDiabetic ketoacidosis; Hypothyroidism3UTUCTislelizumabThrombocytopenia3UTUCTislelizumabPneumonitis; Rash3UTUCTislelizumabMyocarditis3BCToripalimabMyocarditis3RCCPembrolizumabMyocarditis3RCCPembrolizumabNephritis2RCCPembrolizumabHepatitis2RCCPembrolizumabHepatitis2BCTislelizumabHypothyroidism; Hepatitis; Rash2BCSintilimabHepatitis; Myocarditis; Hypothyroidism2RCCSintilimabHypoadrenocorticism2BCSintilimabHyperthyroidism; decreased ACTH2BCToripalimabHypothyroidism2RCCSintilimabHypothyroidism2RCCPembrolizumabHyperthyroidism2RCCPembrolizumabRash; Hepatitis; Diarrhea2RCCPembrolizumabNephritis1UTUCTislelizumabPneumonitis1RCCSintilimabAnemia1UTUCNivolumabHepatitis; Myocarditis1BCTislelizumabHepatitis1UTUCTislelizumabHepatitis1RCCToripalimabDiarrhea1BCToripalimabHepatitis1RCCSintilimabHyperthyroidism; Nephritis1RCCSintilimabHypothyroidism; Lumphopenia1RCCPembrolizumabThrombocytosis; Hyperthyroidism1RCCPembrolizumabThrombocytopenia; Hepatitis1UTUCTislelizumabHepatitis; Hypothyroidism1RCCPembrolizumabHepatitis; Hyperglycemia1UTUCTislelizumabHypothyroidism1RCCSintilimabHypoadrenocorticism1BCTislelizumabHypothyroidism1BCTislelizumabHypothyroidism; Hypoadrenocorticism1RCCSintilimabHyperthyroidism1RCCPembrolizumabHypothyroidism; Hypoadrenocorticism1BCPembrolizumabHypoadrenocorticism1BCTislelizumabHyperglycemia1UTUCToripalimabHypothyroidism1RCCPembrolizumabHyperglycemia1RCCSintilimabHyperthyroidism1BCNivolumabHyperglycemia1UTUCNivolumabHyperglycemia1RCCPembrolizumabHyperthyroidism1UTUCNivolumabHyperthyroidism1RCCPembrolizumabRash1 irAE characteristics. One patient who experienced pneumonitis and myocarditis, and another patient who experienced hepatitis, were considered to have grade 4 irAEs based on the National Cancer Institute CTCAE version 5.0. Two patients with RCC developed grade 4 irAEs and three RCC patients developed grade 3 irAEs, accounting for 50% of the ten patients who experienced grade 3 or grade 4 irAEs (Fig. 1A). All patients over the age of 80 years (4/112) experienced irAEs (Fig. 1A). Baseline routine blood tests did not differentiate between patients with and without irAE. A. Overview of patient age, irAE grades, cancer types and ICI utilized. B. Baseline (before ICI treatment) whole blood cell counts in patients without irAE, with any grade irAEs or with grade 3 and grade 4 irAEs. C. Baseline absolute lymphocyte counts in patients without irAE, with any grade irAEs or with grade 3 and grade 4 irAEs. D. Baseline absolute neutrophil counts in patients without irAE, with any grade irAEs or with grade 3 and grade 4 irAEs. E. Baseline absolute platelet counts in patients without irAE, with any grade irAEs or with grade 3 and grade 4 irAEs. F. Baseline neutrophil-lymphocyte ratio in patients without irAE, with any grade irAEs or with grade 3 and grade 4 irAEs. G. Baseline platelet-lymphocyte ratio in patients without irAE, with any grade irAEs or with grade 3 and grade 4 irAEs. H. Baseline T3 (3, 5, 3′-triiodothyronine) levels in patients without irAE, with any grade irAEs or with grade 3 and grade 4 irAEs. Baseline routine blood tests did not differentiate between patients with and without irAEs We retrospectively examined baseline routine blood parameters. But there were no differences in white blood cell count (WBC) (Fig. 1B), absolute lymphocyte count (Fig. 1C), absolute neutrophil count (Fig. 1D), or platelet count (Fig. 1E) between patients with and without irAEs. Likewise, the neutrophil-to-lymphocyte ratio (NLR) and platelet-to-lymphocyte ratio (PLR) were not predictive of irAEs (Fig. 1F-G), although some studies have reported associations [27]. Baseline triiodothyronine (T3) levels were significantly higher in patients who developed irAEs, but did not differentiate patients with grade 3–4 irAEs (Fig. 1H). Other blood parameters, such as red blood cell count (RBC), hemoglobin (HGB), liver and renal function tests, and lipid profiles, also did not distinguish patients with and without irAEs (Supplementary Figs. 1–2). Tumorous LCP1 and ADPGK serve as predictive biomarker of irAEs LCP1 was found to be both nuclear and cytoplasmic, while ADPGK was found to be mainly cytoplasmic (Fig. 2). LCP1 and ADPGK positivity were significantly increased in tumor samples from patients with irAEs (Figs. 2, 3A). Moreover, ADPGK positivity was significantly higher in patients with grade 3–4 irAEs than in patients with grade 1–2 irAEs (Fig. 3A). There were no significant differences in LCP1 and ADPGK expression across cancer types (Fig. 3B). Representative LCP1 and ADPGK staining in urological cancer sections. All slides were scanned using exactly the same parameters with the PANNORAMIC 1000 system equipped with a 20X objective lens. All magnifications were using the same pre-set parameters in the CaseViewer 2.4 software for image output. Tumorous LCP1 and ADPGK serve as predictive biomarkers of irAE. A. LCP1 positivity and ADPGK positivity in patients without irAE, with grade 1 and grade 2 irAEs or with grade 3 and grade 4 irAEs. B. LCP1 positivity and ADPGK positivity in renal cell carcinoma, bladder cancer and upper tract urothelial carcinoma patients. C. ROC curve of LCP1 positivity in predicting irAEs in all urological cancer patients. D. ROC curve of ADPGK positivity in predicting irAEs in all urological cancer patients. E. ROC curve of bivariate model established with a logistic regression of LCP1 and ADPGK positivity in all urological cancer patients, logit(P) = 4.898 - 7.013 × LCP1_Positivity - 8.108 × ADPGK_Positivity. F. LCP1 positivity in patients of different prognosis. G. ADPGK positivity in patients of different prognosis. Receiver operating characteristic (ROC) analysis showed that LCP1 positivity was predictive of irAEs with an area under the curve (AUC) of 0.8415 (P < 0.0001; 95% CI: 0.7679–0.9152) (Fig. 3C). ADPGK positivity showed slightly improved predictive performance, with an AUC of 0.8759 (P < 0.0001; 95% CI: 0.8112–0.9406) (Fig. 3D). Notably, the bivariate model combining LCP1 and ADPGK yielded superior predictive accuracy, with an AUC of 0.9184 (P < 0.0001; 95% CI: 0.8648–0.9719) (Fig. 3E). When the patients were divided into groups based on their clinical response (CR, partial response [PR], stable disease [SD], and progressive disease [PD]), no differences in the proportion of LCP1- or ADPGK-positive patients were found (Fig. 3F–G). But high levels of both markers were consistently associated with the presence of irAEs, regardless of response. Subgroup analysis by cancer type In BC, the rates of LCP1 and ADPGK positivity were significantly increased in patients with irAEs (Fig. 4A). The bivariate model demonstrated excellent predictive performance, with an AUC of 0.9856 (P < 0.0001; 95% CI: 0.9516–1.000) (Fig. 4C). In RCC, ADPGK was a better predictor than LCP1 (Fig. 4D-E); but the bivariate model showed better predictive ability (Fig. 4F). In UTUC, ADPGK showed relatively poor predictive performance (AUC = 0.7803), while LCP1 and the bivariate model had AUCs of 0.8485 and 0.8561, respectively (Fig. 4G–I). LCP1, ADPGK and bivariate model are valid in subtypes of urological cancers. A. LCP1 positivity and ADPGK positivity in bladder cancer patients with or without irAEs. B. ROC curves of LCP1 positivity and ADPGK positivity in predicting irAEs in bladder cancer patients. C. ROC curve of bivariate model in predicting irAEs in bladder cancer patients, logit(P) = 13.54 - 33.15 × LCP1_Positivity - 7.803 × ADPGK_Positivity. D. LCP1 positivity and ADPGK positivity in RCC patients with or without irAEs. E. ROC curves of LCP1 positivity and ADPGK positivity in predicting irAEs in RCC patients. F. ROC curve of bivariate model in predicting irAEs in RCC patients, logit(P) = 5.046 - 5.106 × LCP1_Positivity - 10.89 × ADPGK_Positivity. G. LCP1 positivity and ADPGK positivity in UTUC patients with or without irAEs. H. ROC curves of LCP1 positivity and ADPGK positivity in predicting irAEs in UTUC patients. I. ROC curve of bivariate model in predicting irAEs in UTUC patients, logit(P) = 4.112 - 9.989 × LCP1_Positivity - 3.136 × ADPGK_Positivity. Prediction efficacy varies among irAE subtypes, ICIs, and irAE timing We then divided the samples into groups based on the type of immune-related adverse event (irAE). LCP1 and ADPGK positivity were significantly increased in patients with pulmonary, gastrointestinal, endocrine, cutaneous, and renal irAEs (Fig. 5A, B). But LCP1 and ADPGK positivity did not significantly differentiate patients with hematologic irAEs from those without irAEs. Moreover, LCP1 positivity was not significantly associated with patients who developed cardiovascular irAEs (Fig. 5A). We also noticed that the contributions of LCP1 and ADPGK positivity differed among irAEs. Thus, we determined the ratio of LCP1 to ADPGK positivity and observed that it varied between irAE types (Fig. 5C). Patients with pulmonary irAEs had relatively higher ADPGK positivity, while those with renal irAEs had higher LCP1 positivity (Fig. 5C), implying that the LCP1/ADPGK ratio may be used to predict irAE phenotypes. Lastly, we examined LCP1 and ADPGK positivity in patients with different numbers of irAEs and found no significant differences in LCP1 or ADPGK positivity among patients with one, two, or multiple irAEs (Fig. 5D). Prediction efficacy varies among different adverse effects or ICIs. A. LCP1 positivity in samples from patients without irAE and samples from patients with different types of irAEs. B. ADPGK positivity in samples from patients without irAE and samples from patients with different types of irAEs. C. LCP1 positivity-ADPGK positivity ratios in samples from patients without irAE and samples from patients with different types of irAEs. D. LCP1 positivity and ADPGK positivity in samples from patients experienced different frequency of irAEs. E. LCP1 positivity in samples from patients treated with different ICIs. F. ADPGK positivity in samples from patients treated with different ICIs. To assess the generalizability of the predictive model, LCP1 and ADPGK positivity were also examined in the context of different ICI. LCP1 and ADPGK positivity were significantly increased in patients with irAEs in most ICI groups (Fig. 5F). The only exception was LCP1 in patients treated with sintilimab, where statistical significance was not reached despite a higher mean LCP1 positivity in the irAE group (Fig. 5E). Overall, these results indicate that tumoral LCP1 and ADPGK may be potential predictive markers for irAEs in various ICI treatments. We examined LCP1 and ADPGK positivity in the early- and late-onset irAE groups. As shown in Supplementary Fig. 4, in the early-irAE group, the irAE subgroup had significantly higher LCP1 and ADPGK positivity than the non-irAE subgroup (P < 0.0001 for both). Similarly, in the late-irAE group, the irAE group also exhibited significantly higher LCP1 and ADPGK positivity than the non-irAE group (P < 0.0001 for both). These results suggest that LCP1 and ADPGK positivity are consistently associated with irAE development, regardless of the time of onset. Impact of sample acquisition method Because the amount of evaluable tumor area varies among samples, especially when different methods are used to acquire the samples, the results of the quantification may be variable. We thus divided the samples into three groups based on the method of sample acquisition. The average positivity of LCP1 and ADPGK was greater in the irAE group, regardless of the method of sample acquisition (Fig. 6A). But in the samples obtained by puncture biopsy and transurethral resection of bladder tumor (TURBT), the positivity of LCP1 and ADPGK did not differ significantly between the non-irAE and irAE groups (Fig. 6A), which may indicate that false-positive or less accurate results may be obtained when the evaluable tumor area is small. However, when we restricted the analysis to surgical specimens, the predictive value of LCP1 positivity, ADPGK positivity, and the bivariate model for irAE prediction was significantly improved (Fig. 6B-D). Non-surgical samples impair prediction efficacy of LCP1, ADPGK and bivariate model. A. LCP1 positivity and ADPGK positivity in samples obtained from various methods. B. ROC curve of LCP1 positivity in predicting irAEs in surgically obtained urological cancer patients. C. ROC curve of ADPGK positivity in predicting irAEs in surgically obtained urological cancer patients. D. ROC curve of bivariate model in predicting irAEs in surgically obtained urological cancer patients, logit(P) = 4.809 - 7.308 × LCP1_Positivity - 7.993 × ADPGK_Positivity. Discussion Combination of tumorous LCP1 and ADPGK is a promising model in predicting irAEs in urological cancer and beyond In this cohort of 112 patients with urological cancers, 37 patients (33.0%) achieved CR, 20 (17.8%) achieved PR, and 11 (9.8%) experienced PD, suggesting that ICI-based therapies have a significant clinical benefit. However, 45.5% of patients experienced irAEs, 17 patients (15.1%) experienced multiple irAEs, and 10 patients (8.9%) experienced grade 3 or 4 irAEs. Eleven patients (10% of the cohort; data not shown) discontinued ICI treatment. Thus, predictive biomarkers for irAEs are needed to improve the benefit-to-risk ratio of ICI therapy. While many studies have attempted to predict ICI-induced irAEs, there is no consensus [32,33]. In our cohort, previously reported markers, including absolute lymphocyte count, NLR, and PLR [34] were not predictive. Our results are in line with multi-omics and experimental validations in lung cancer, where high tumoral LCP1 and ADPGK expression was linked to irAEs [25,35]. They also confirm these findings in urological cancers, consistent with our recent validation in bone and soft tissue sarcomas and other cancers [31,36]. In urological cancers, both markers successfully distinguished patients with and without irAEs, and ADPGK positivity was associated with irAE severity. The bivariate LCP1+ADPGK model had an AUC of 0.9184 in the overall cohort and an outstanding AUC of 0.9856 in BC, suggesting tumor-specific variations in predictive accuracy. Overall, these results suggest that LCP1 and ADPGK have the potential to be translated as reliable markers for predicting irAEs. We conducted bootstrap internal validation (1000 iterations) and LOESS calibration for LCP1, ADPGK and the bivariate model in five cohorts (Supplementary Figs. 5–9). The bootstrap-corrected estimates suggested a moderate level of optimism. In the overall cohort, the optimism-corrected AUC was 0.865 (∼0.05 lower than the apparent AUC), which was still >0.86 and suggested clinical utility. Likewise, in the surgical cohort (n = 96), the bivariate model exhibited a relatively modest difference between the apparent and corrected AUC (0.922 vs. 0.901). This finding suggests the model is more likely to perform well in larger datasets, further reinforcing the stability and validity of the bivariate model in the overall and surgical cohorts. Conversely, the largest gap between the apparent and bootstrap-corrected AUC was observed in the BC subgroup (0.986 vs. 0.771), which is likely due to the small sample size and the risk of overfitting, as also reflected by the broad confidence interval (0.722–0.988). This was also evident in the UTUC cohort, where the corrected AUC was 0.831 (95% CI: 0.636–0.981). These results highlight the need for external validation in larger cohorts, especially for cancer subtype-specific analyses. Elevated LCP1 and ADPGK activate T cells Biologically, LCP1 has been shown to regulate leukocyte adhesion via phosphorylation, thus facilitating T-cell migration and activation [28,37]. More recent research has also demonstrated that LCP1 deficiency results in hematologic and immune dysfunction [26]. At the molecular level, LCP1 has been shown to facilitate the formation of the NLRP3 inflammasome through the Pyk2-ASC pathway, leading to inflammatory tissue damage [29]. Thus, its involvement in cytoskeletal dynamics and immune regulation may promote hyperactive immune responses during ICI treatment, and LCP1 expression may serve as a predictor of irAEs. ADPGK is a key enzyme in metabolism that converts ADP to ATP and facilitates glucose metabolism [27,30]. Its role in immune cells, including T-cell activation and proliferation, implies that ADPGK may also regulate immune responses in cancer therapy [38,39]. Since LCP1 and ADPGK are both part of T-cell activation pathways, high tumoral expression of these proteins may reflect a pre-activation state of the immune system that increases the risk of immune-related toxicity. Crucially, this is a speculative interpretation. Our results are correlative. Functional experiments are needed to establish whether LCP1 and ADPGK derived from tumors directly affect T-cell function and lead to immune-related toxicity. Thus, the main interpretation of our findings should be in terms of their predictive ability, while the mechanistic interpretation should be used to guide future research. Stereo parameters are required for prediction of irAE types and grades While the sample size was small, we found that the ratio of LCP1-to-ADPGK positivity differed among irAEs. The three patients who experienced lung irAEs had relatively higher ADPGK positivity, while the three patients who experienced renal irAEs had higher LCP1 positivity. By contrast, the LCP1/ADPGK ratio in patients with endocrine irAEs was more variable without a clear trend. These findings suggest that in cohorts with a greater number of lung or renal irAEs, the LCP1-to-ADPGK ratio may be worth exploring as a second dimension of biomarker information that may help to discriminate between irAE phenotypes. In conclusion, LCP1 positivity, ADPGK positivity and the bivariate model all showed excellent predictive potential for irAEs, regardless of the immune checkpoint inhibitor used. Furthermore, our sub-analysis showed that sample collection method had a profound impact on the predictive performance. Biopsy samples acquired by TURBT and puncture biopsy were less accurate than surgical specimens, probably due to tumor heterogeneity and the small proportion of evaluable tissue. These results underline the need for harmonized pre-analytical procedures to achieve reliable biomarker evaluation. Notably, the predictive value of LCP1 and ADPGK was consistent across various ICI therapies, suggesting the robustness of these biomarkers. However, there are some limitations. First, the retrospective nature of the study may lead to selection bias. Second, the study was conducted at a single institution, which could affect the external validity of the results. Third, the small sample size requires careful interpretation of the predictive accuracy of the model. For instance, in the BC cohort, while the AUC was 0.9856, this was calculated from only 29 cases, which may impact the precision of the estimate. Thus, multi-center prospective studies with larger sample sizes and longer follow-up are required to further validate and improve the model. Fourth, while IHC is widely used in clinical practice, it is semi-quantitative; more standardized and quantitative methods will be needed for future clinical use. Fifth, tissue heterogeneity had a substantial impact on the predictive performance in non-surgical samples, highlighting the need for standardized sampling procedures. Pathway to clinical implementation and standardization The excellent predictive value of the LCP1/ADPGK bivariate model suggests the potential for its use in clinical practice. But to develop a clinically validated biomarker test, a systematic process is required. First, the clinical value of this biomarker needs to be validated in prospective clinical trials. An important part of this process is the reproducibility of the IHC assay. While staining and scoring in the current study were carried out in a consistent manner by trained pathologists in a single institution, these processes represent an internal standard and need to be validated externally. To be clinically relevant, LCP1/ADPGK testing requires a standardized and objective scoring system. This would involve the creation of a reliable and user-friendly algorithm to score staining. Second, the reproducibility of the assay must be confirmed through multi-institutional collaboration to develop standard operating procedures for: (1) pre-analytical variables, including tissue fixation time and processing; (2) analytical variables, including antibody clones, antigen retrieval, staining platforms and detection systems; and (3) post-analytical variables, especially training and certification of pathologists to ensure consistent interpretation. Thus, while our study offers a proof-of-concept, significant work needs to be done to develop and validate the LCP1/ADPGK IHC test for routine clinical use. Ethics approval and consent to participate This study was approved by the Ethics Committee of Peking University First Hospital. All patients were informed and signed on consents. Availability of data and material All data and material are available once published. Funding This study was supported by the National High Level Hospital Clinical Research Funding (Youth clinical research project of Peking University First Hospital, 2025YC-GD03 to QT), Immuno-Oncology Research Fund of the Cancer Foundation of China (CFCMYZLKYJJ011 to QT), National High Level Hospital Clinical Research Funding (High Quality Clinical Research Project of Peking University First Hospital) (2022CR72), Charitable Project of Beijing Guo Yinglu Urology Development Foundation, Wuxi "Taihu Talents Program" Medical and Health High-level Talents Project, and National Natural Science Foundation of China (82273350,U23A20460). Abbreviation LCP1: Lymphocyte Cytosolic Protein 1 ADPGK: ADP Dependent Glucokinase ICI: immune checkpoint inhibitor irAE: immune-related adverse events BC: bladder cancer RCC: renal cell carcinoma UTUC: upper tract urothelial carcinoma TURBT: transurethral resection of bladder tumor ROC: receiver operating characteristic AUC: area under curve CR: complete response PR: partial response SD: stable disease PD: progressive disease CRediT authorship contribution statement Liqing Xu: Writing – review & editing, Writing – original draft, Resources, Project administration, Methodology, Investigation, Formal analysis, Data curation. Xi Zhang: Resources, Data curation. Zihao Tao: Writing – original draft, Resources, Methodology, Investigation, Formal analysis, Data curation. Yiceng Ying: Resources, Data curation. Xueqin Zeng: Methodology. Kaiwei Yang: Methodology, Investigation, Data curation. Yang Wang: Methodology. Ninghan Feng: Writing – review & editing, Funding acquisition. Xuesong Li: Writing – review & editing, Project administration, Funding acquisition. Yaqian Li: Writing – review & editing, Writing – original draft, Project administration, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Qi Tang: Writing – review & editing, Project administration, Methodology, Investigation, Funding acquisition, Data curation, 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. Acknowledgments The authors are grateful to the urological pathology institute in Peking University First Hospital for helping in the preparation of the pathological specimens. We also thank Mr. Xin Liu and Mrs. Rui Liu for kindly providing testing antibodies against LCP1 and ADPGK. Footnotes Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.tranon.2026.102836. Contributor Information Ninghan Feng, Email: n.feng@njmu.edu.cn. Xuesong Li, Email: pineneedle@sina.com. Yaqian Li, Email: 18810980233@163.com. Qi Tang, Email: drtangq@bjmu.edu.cn. Supplementary materials References Associated Data Supplementary Materials Data Availability Statement All data and material are available once published.