Toxicity related to immunotherapy in urothelial cancer: Tumour genetic risk variants

Abstract

The treatment of urothelial cancer (UC) has been revolutionized by the introduction of immune checkpoint inhibitors (ICIs). However, while ICIs have achieved better therapeutic results than traditional chemotherapeutic agents, immune-related adverse events (irAEs) can lead to poor treatment outcomes. This exploratory study was designed to seek tumour single-nucleotide polymorphisms (SNPs) related to irAEs in patients with UC treated with ICIs. Tumour specimens were collected at several hospitals in Madrid (Spain) from 102 patients treated with nivolumab, atezolizumab, avelumab or pembrolizumab. DNA was extracted from formalin-fixed paraffin-embedded tumour biopsies, and 49 SNPs genotyped using a MassARRAY platform. Among the 49 SNPs analyzed, rs1738074 was associated both with overall toxicity (p < 0.001), and with the appearance of specific toxicities such as dermatitis and pneumonitis (p = 0.005 and p = 0.036, respectively). In effect, 100% of patients with pneumonitis were homozygous for the variant allele (p = 0.020). Other SNPs were also linked to a higher risk of irAEs: rs2301756 to pneumonitis (p = 0.005), rs55733913 and rs2117997 to dermatitis (p = 0.026, p = 0.021, respectively) and rs4988956 to hepatitis (p = 0.027). Contrarily, two SNPs, rs11571302 and rs66502444, showed a possible protective role against asthenia (p = 0.012, p = 0.030, respectively). This exploratory study highlights the possible useful role of tumour SNPs as biomarkers to predict the risk of UC patients developing irAEs. Further studies are needed to validate the utility of SNPs as biomarkers of ICI-induced toxicity.

Introduction

Urothelial carcinoma (UC) originates in the urothelial cells lining the urinary tract and accounts for over 90% of all cases of bladder cancer. Besides the bladder, UC can also develop in the upper urinary tract (renal pelvis, ureters), these tumours sharing similar histological features and treatment strategies [1]. UC is closely linked to an advanced age, and the median age of bladder cancer diagnosis is 73 years and, in consequence, it is also associated with a higher likelihood of comorbidities, reduced physical fitness, and with an augmented risk of experiencing treatment-related toxicity [2].

Platinum-based regimens have been the standard treatment in UC since 2000, providing an acceptable response rate of up to 70%, but with low survival rates. Only 15% of patients were alive at 5 years, and the median overall survival was 15 months [3]. Patients ineligible for cisplatin, who may account for up to 50% of cases, did not have an optimal treatment regimen; most received the carboplatin-gemcitabine doublet, with response rates of 36% and a median overall survival of approximately 8.5 months [3]. Recently, the limited therapeutic options for UC have been transformed by the introduction of immunotherapy targeting immune checkpoint molecules that help tumour cells evade the host immune system. In tumour cells PD-L1 (programmed cell death ligand-1) is overexpressed and interacts with its receptor PD1 (programmed cell death-1) on activated T cells, triggering the inactivation and apoptosis of T cell proliferation [4]. Anti PD1 drugs, such as pembrolizumab and nivolumab, and anti-PDL1 agents, such as atezolizumab and avelumab, are immune checkpoint inhibitors (ICIs) that disrupt this interaction and are standard of care agents recommended in European guidelines for the treatment of urothelial tumours [5]. Nivolumab was approved by the FDA for this use based on the results of the CheckMate 275 phase II trial, which showed an objective response rate of 19.6% in patients with advanced or metastatic UC who showed resistance to platinum treatment [6]. Atezolizumab has also shown great efficacy in UC treatment. A meta-analysis of clinical studies to evaluate its predictive accuracy involving 4148 individuals indicated that atezolizumab led to significantly improved overall- and progression free survival in UC patients [7]. The most recent therapeutic strategy for first-line treatment of metastatic urothelial carcinoma includes, since 2024, the combination of antibody-drug conjugates with immunotherapy. In this setting, the combination of enfortumab vedotin and pembrolizumab has represented a therapeutic breakthrough, changing the treatment standard by doubling median overall survival. The median overall survival currently achieved is nearly 34 months [8].

ICIs promote immune system hyperactivation possibly causing immune-related adverse events (irAEs) that can lead to different types of complications, for instance, hepatitis, dermatitis, pneumonitis or asthenia, as considered in the present study [9]. A recent review of the side effects of immunotherapy including 87 trials revealed that immune-related deaths occur in 0.26% of patients, and that the primary cause is pneumonitis, followed by hepatic failure and hepatitis accounting for 35.7%, 10.7% and 7.1% of total deaths, respectively [9]. Among the any-grade irAEs most frequently reported were rash (13.8%), hypothyroidism (11%), and diarrhoea (10.4%). According to the results of a recent meta-analysis, the incidence of irAEs leading to treatment discontinuation was 7.6% of patients receiving ICI monotherapy and 24.3% of patients under dual ICI therapy [10]. Severe irAEs can lead to poor treatment outcomes for cancer patients. To address this issue effectively, clinicians need to gain a thorough understanding of the risk factors involved in irAEs and of potential genetic biomarkers of this toxicity.

Different ICI treatments, therapy combinations, types of tumour and clinical settings have been associated with several irAE profiles [10], but scarce data have been reported for molecular biomarkers. Only a few studies have examined the possible role of some germline SNPs as biomarkers [11–13]. In addition, it seems that circulating tumour DNA could be a promising biomarker to help identify patients who are more likely to benefit from adjuvant atezolizumab [5]. In the present work, we hypothesized that SNPs identified in tumour biopsy specimens could predict irAEs, as occurs with germline or circulating tumour DNA. The aim of this study was to establish possible links between 49 tumour SNPs and the risk of developing ICI-induced irAEs in treated UC patients.

Materials and methods

Patients and samples

In this exploratory observational study, we focused on patients with urothelial carcinoma who were eligible for treatment with ICIs (nivolumab, atezolizumab, avelumab or pembrolizumab). Patients were treated over the period 2017 to 2024 at the oncology units of the University Hospital of Móstoles, University Hospital of the Alcorcón Foundation, University Hospital of Getafe and University Hospital of Fuenlabrada, in Madrid, Spain.

Inclusion criteria were age ≥18 years, urothelial carcinoma treated with monotherapy ICI and signed informed consent by the patient. Exclusion criteria were having a different type of cancer, receiving another experimental treatment in a clinical trial within the previous four weeks, and having insufficient paraffin-embedded tumour tissue available. No patient received chemotherapy or any other drug at the same time.

Medical oncologists from the different hospitals recorded the following clinical-pathological data in accordance with existing data protection laws: biopsy cellularity percentage, tumour location, sex, age, treatment, immunotherapy line, and the type and grade of toxicity: colitis, dermatitis, hepatitis, pneumonitis, thyroid dysfunction, adrenal dysfunction, diabetes mellitus, hypophysitis, arthritis, ocular afectation, nephritis, asthenia, anaemia, lymphopenia, myocarditis and myositis. Adverse events were graded according to the Common Terminology Criteria for Adverse Events (CTCAE), version 5.0, from grade 0 (no toxicity) to grade 5 (death due to irAEs) including grades from 1 to 4 that are specifically defined for different types of toxicity [14]. For instance, ICI-induced pneumonitis is an inflammatory reaction of the pulmonary parenchyma characterized by cough, dyspnoea, fever, and/or hypoxemia, often accompanied by radiological findings. According to CTCAE v5.0, grade 1 corresponds to asymptomatic or minimal radiologic findings; grade 2 involves moderate symptoms that interfere with daily activities; grade 3 denotes severe symptoms or need for hospitalization; grade 4 is life-threatening, with respiratory failure or need for mechanical ventilation [14, 15]. ICI-induced dermatitis is one of the most common immune-mediated toxicities and manifests clinically as maculopapular rash, pruritus, erythema, or desquamation, with various possible patterns such as morbilliform, lichenoid, or bullous eruptions. According to CTCAE v5.0 criteria, grade 1 corresponds to mild skin eruptions affecting less than 10% of body surface area; grade 2 is defined as moderate involvement (10–30% of body surface area); grade 3 indicates extensive (>30%) or severe symptoms such as intense pruritus, pain, or significant desquamation, requiring hospitalization; grade 4 is life-threatening, with epidermal necrosis or severe mucocutaneous involvement [14, 15]. Asthenia is a disorder characterized by a state of generalized weakness and a pronounced inability to summon sufficient energy to accomplish daily activities. According to CTCAE v5.0, toxicity is classified as grade 1 (mild; fatigue relieved by rest), grade 2 (moderate, fatigue not relieved by rest; limiting instrumental activities of daily living (preparing meals, shopping for groceries or clothes, using the telephone, managing money etc.), grade 3 (severe, fatigue not relieved by rest, limiting self-care activities of daily living (bathing, dressing and undressing, feeding self, using the toilet, taking medications, and not bedridden); there is no asthenia grade 4 or 5 [14, 15]. ICI-induced hepatitis is characterized by elevated blood levels of AST (aspartate aminotransferase), ALT (alanine transaminase), and/or bilirubin compared to the ULN (upper limit of normal). The severity of toxicity was classified from grade 1 to grade 4 depending on the range of increase of AST, ALT and bilirubin and symptomatic liver dysfunction according guidelines [14, 15].

Pathologists were responsible for selecting and obtaining sample slices. They first selected the optimal area of haematoxylin-eosin-stained formalin-fixed paraffin-embedded tumour biopsies and then processed only those whose cellularity exceeded 50%. Next, the first two slices were discarded, and 5 slices of 5 µm each obtained (total 2.5 mm³). DNA was extracted using the QIAamp DNA FFPE Tissue Kit (Qiagen, Germantown, MD, USA), as indicated in the manufacturer’s instructions. The purified DNA was eluted in 30 µL of DNAse-free water, and quality and concentration assessed in a Nanodrop 2000 system (Thermo Fisher Scientific, Wilmington, DE, USA).

SNP selection and genotyping

Forty-nine SNPs were chosen for examination based on a literature search in PubMed (www.pubmed.com) using the keywords SNP, immunotherapy, and cancer, and also based on our previous results [16]. The polymorphisms selected because of their potential role as biomarkers of ICI-induced irAEs were mainly associated with DNA metabolism, immune response, RNA metabolic processes, reproductive processes, stimulus response, and establishment of localization [11–13, 17, 18] (Table 1). The 49 SNPs chosen affected different genic regions: ADAD1 (rs17388568), ANO5 (rs7481951), BAZ2B (rs56328422), CFAP36 (rs3762513), CTLA4 (rs11571302, rs16840252, rs1863800, rs231775, rs3087243, rs4553808, rs5742909), FAR2 (rs7954686), GLIS3 (rs10814859), IFNG (rs2069705), IFNL4 (rs12979860), IFNW1 (rs10964859), IL1RL1 (rs4988956), IL22RA1 (rs75824728), IL23R (rs2295359), IL7 (rs16906115), PACRG (rs55733913, rs66502444), PD-L1 (rs4143815), PNPT1 (rs782637), PREX2 (rs6993547), PTPN11 (rs2301756), RAC1 (rs9374), RGMA (rs4778080), ROBO1 (rs162263), SEMA5A (rs3026321), TAGAP (rs1738074), TYK2 (rs12720250) and UNG/ACACB (rs246079); 9 RNA coding regions: LINC00973/RNU6-1263P (rs2062059), LOC102724355 (rs239731), LOC105374140 (rs6440251), LOC105377125 (rs11711517), LOC124902023 (rs12548560), LOC124902908 (rs2117997), LOC100129175 (rs359975), MAMSTR/FUT2/LOC105447645 (rs504963), and MIR146A/MIR3142HG (rs2910164); and 7 intergenic regions: rs113861051, rs1276216, rs35807769, rs470753, rs4800887, rs5908080, and rs9268543.

Table 1.

The 49 SNPs analysed, and classification of the genomic regions affected by them according to their gene ontology.

Gene Ontology	Genomic region	SNP wt > v
Immunologic response	CTLA4 or near CTLA4 (cytotoxic T-lymphocyte associated protein 4)	rs11571302 G > T
		rs1863800 C > T
		rs3087243 G > A
		rs4553808 A > G
		rs16840252 C > T
		rs231775 A > G
		rs5742909 C > T
	MIR146A/MIR3142HG	rs2910164 C > G
	IFNW1 (interferon omega 1)	rs10964859 C > G
	IL7 (interleukin 7)	rs16906115 G > A
	IFNG (interferon gamma)	rs2069705 G > A
	IL1RL1 (interleukin 1 receptor like 1)	rs4988956 G > A
	IL23R (interleukin 23 receptor)	rs2295359 C > T
	PDL-1 (ligand for PD-1 receptor)	rs4143815 C > G
	IL22RA1 (interleukin 22 receptor subunit alpha 1)	rs75824728 G > A
	TYK2 (tyrosine kinase 2)	rs12720250 G > T
	IFNL4 (interferon lambda 4)	rs12979860 C > T
Central nervous system development	SEMA5A (semaphorin 5A)	rs3026321 A > T
	RGMA (repulsive guidance molecule BMP co-receptor A)	rs4778080 G > T
	ROBO1 (roundabout, axon guidance receptor, homolog 1)	rs162263 T > G
Regulation of gene expression	TAGAP (T cell activation RhoGTPase activating protein)	rs1738074 T > C
	PNPT1 (polyribonucleotide nucleotidyltransferase 1)	rs782637 T > C
	LINC00973/RNU6-1263P (long intergenic non-protein coding RNA 973, U6 small nuclear 1263, pseudogene)	rs2062059 A > G
	GLIS3 (GLIS family Zinc Finger 3)	rs10814859 T > C
	MAMSTR/FUT2/LOC105447645 (MEF2 activating motif and SAP domain containing transcriptional regulator)	rs504963 G > A
	BAZ2B (bromodomain adjacent to Zinc Finger domain 2B)	rs56328422 T > C
	LOC105377125 (ncRNA)	rs11711517 C > T
	LOC124902023 (ncRNA)	rs12548560 A > G
	LOC124902908 (ncRNA)	rs2117997 A > G
	LOC102724355 (ncRNA)	rs239731 T > C
	LOC105374140 (ncRNA)	rs6440251 T > G
Signalling pathways	PTPN11 (protein tyrosine phosphatase non-receptor type 11)	rs2301756 A > G
	PREX2 (phosphatidylinositol-3,4,5-trisphosphate dependent Rac exchange Factor 2)	rs6993547 G > C
	RAC1 (Rac family small GTPase 1)	rs9374 G > A
Cell motility	CFAP65/ LOC100129175 (cilia and flagella associated protein 65, lncRNA)	rs359975 C > G
	CFAP36 (cilia and flagella associated protein)	rs3762513 C > T
Chaperons and chaperonins	PACRG (parkin co-regulated gene protein)	rs55733913 T > C
		rs66502444 A > T
Nucleic acid metabolism	UNG/ACACB (uracil N-glucosidase)	rs246079 A > G
	ADAD1 (adenosine deaminase domain containing 1)	rs17388568 G > A
Membrane channel	ANO5 (anoctamin 5)	rs7481951 A > T
Lipid metabolism	FAR2 (fatty acyl-CoA reductase 2)	rs7954686 A > G
Unknown	Intergenic region	rs113861051 A > G
		rs1276216 A > T
		rs35807769 A > G
		rs470753 G > A
		rs4800887 G > A
		rs5908080 T > C
		rs9268543 A > T

wt wild-type allele, v variant allele.

Genotyping of the 49 SNPs was carried out by the Spanish National Genotyping group (FPGMX-CeGen) using iPlexGold chemistry on the MassARRAY platform, following the manufacturer’s instructions (Agena Bioscience, San Diego, CA, USA) and using Assay Design Suite v3.0 software (Agena Bioscience, San Diego, CA, USA), GRCh38 version. The SNPs were genotyped twice in independent assays. The PCR conditions were: total volume 5 µL, 20 ng of template DNA, 1× PCR buffer, 2 mM MgCl₂, 500 µM dNTPs, primers 100 nM, and 1U/reaction of DNA polymerase (Agena Bioscience, San Diego, CA, USA). The PCR protocol started with a 2-min denaturation at 94 °C, followed by 45 cycles of 94 °C for 30 s, 56 °C for 30 s, and 72 °C for 1 min, finishing with a 1-min extension at 72 °C. To eliminate free dNTPs, the PCR products were incubated with 0.6 U shrimp alkaline phosphatase (SAP) for 40 min at 37 °C, and the enzymes were deactivated by heating at 85 °C for 5 min.

iPLEX Gold assays were conducted in a total volume of 9 µL, which included 0.222× iPLEX buffer Plus, 0.5× iPLEX termination mix and 0.5× iPLEX enzyme. A mixture of extension primers was prepared to achieve a final concentration ranging from 0.73 µM to 1.46 µM for each primer. The thermal cycling procedure commenced with an initial denaturation at 94 °C for 30 s. This was followed by 40 cycles at 94 °C for 5 s, incorporating an internal loop of 5 cycles at 52 °C for 5 s and 80 °C for 5 s, and concluded with a final extension at 72 °C for 3 min. Subsequently, the iPLEX Gold reaction products underwent desalting using Clean Resin (Agena Bioscience, San Diego, CA, USA) as in the manufacturer’s guidelines. The desalted products were then transferred onto a 384 Spectrochip II with the help of the Chip Prep Module 384 (Agena Bioscience, San Diego, CA, USA). Spectra were obtained by MA4 (Agena Bioscience, San Diego, CA, USA) mass spectrometry, and the spectra were manually revised by qualified personnel utilizing MassARRAY Typer software v5.0.9 (Agena Bioscience, San Diego, CA, USA). All experiments were performed in 384-well plates, which included negative controls and a set of Coriell samples (Na10861, Na11994, and Na11995) for quality assurance purposes. SNPs with call rates below 95% were omitted to maintain the accuracy and precision of subsequent data analyses. Internal controls showed 100% reproducibility and successful genotyping.

Statistical analysis

For quantitative data, mean and standard deviation or median and interquartile rank (IQR) are provided based on their distribution as determined by the Kolmogorov-Smirnov normality test. Qualitative data are expressed in both absolute numbers and percentages. The relationship between different genotypes of each SNP and the occurrence of an irAE was evaluated using contingency tables, with either the chi-squared test or Fisher’s exact test applied as appropriate. Genotypes were grouped for comparisons in two ways considering carriers of 1 or 2 variant alleles (v): v/v versus wt/wt + wt/v and v/v + wt/v versus wt/wt. Significant relationships were further explored by calculating odds ratios and 95% confidence intervals. To verify genetic stability within the group, Hardy-Weinberg equilibrium was checked for each gene. All statistical tests were performed using IBM SPSS Statistics for Windows, Version 29.0.2.0 (Armonk, NY). Significance was set at p ≤ 0.05.

Results

Participants were 102 patients diagnosed with UC who met the inclusion criteria. The demographic and baseline characteristics of the patients are detailed in Table 2. Median age at time of diagnosis was 71 years (IQR = 64–76). The majority of patients were male 80.4% (Table 2). The most frequent type of tumour analyzed among patients with primary tumours (92 of 102) was urinary bladder (75) following by renal pelvis (12), ureter (3), prostate (1) and urethra (1). Ten patients suffered metastases to different sites: 2 cervical, 2 peritoneum, 2 ganglions, 1 adrenal gland, 1 small intestine, 1 prostate and 1 kidney. All the patients received ICI monotherapy: most of them (75.5%) were under treatment with atezolizumab, following by avelumab (13.7%), pembrolizumab (7.8%) and nivolumab (3%) (Table 2). ICI-induced adverse events were observed in 36.3% of our cohort of UC patients; 25.5% showing moderate symptoms; 9.8% severe symptoms and one patient dying because of the adverse effect of treatment (Table 2). Asthenia was recorded as the irAE showing the higher frequency among the UC patients (18.6%) following by dermatitis (8.8%), hepatitis (3.9%), pneumonitis (3.9%), colitis (2.9%), anaemia (2.9%), thyroid dysfunction (2.9%) and other minor irAEs (<2%) including arthritis, lymphopenia, ocular affectation, adrenal dysfunction, diabetes and myositis (Table 2).

Table 2.

Patient demographic and baseline characteristics (n = 102).

Patient characteristics		No. (%)
Sex	Male	82 (80.4%)
	Female	20 (19.6%)
Biopsy specimen	Primary tumour	92 (90.2%)
	Metastasis	10 (9.8%)
Immunotherapy regimen	Atezolizumab	77 (75.5%)
	Avelumab	14 (13.7%)
	Pembrolizumab	8 (7.8%)
	Nivolumab	3 (3%)
Overall toxicity	No toxicity (grade 0)	65 (63.7%)
	Moderate toxicity (grade 1 + 2)	26 (25.5%)
	Severe toxicity (grade 3 + 4)	10 (9.8%)
	Death (grade 5)	1 (<1%)
Immune-related adverse event	Asthenia	19 (18.6%)
	Dermatitis	9 (8.8%)
	Hepatitis	4 (3.9%)
	Pneumonitis	4 (3.9%)
	Colitis	3 (2.9%)
	Anaemia	3 (2.9%)
	Thyroid dysfunction	3 (2.9%)
	Arthritis	2 (1.9%)
	Lymphopenia	1 (<1%)
	Ocular affectation	1 (<1%)
	Adrenal dysfunction	1 (<1%)
	Diabetes	1 (<1%)
	Myositis	1 (<1%)

Genotype frequencies for the 49 SNPs analysed in tumour biopsies are presented in Table 3. According to the 1 000 Genomes Project Consortium, the allelic frequencies observed here are in line with those of European populations described at this website [19]. In our cohort, some SNPs displayed the variant allele (v) at frequencies over 80%: rs12762162 and rs5908080 located in the intergenic region and rs239731 and rs6440251 in non-coding RNA genes (Table 3). Moreover, good agreement with Hardy-Weinberg equilibrium was observed in 96% of the SNPs analysed, indicating good stability in the population without the influence of selective evolutionary pressures.

Table 3.

Genotype and allele frequencies of the 49 SNPs examined. European frequencies from the 1000 Genomes database are indicated [19].

SNP wt > v	Genotype frequency	Allele frequency	1000 Genomes European allele frequency
rs11571302 G > T	GG = 16 (16.0%) GT = 48 (48.0%) TT = 36 (36.0%)	G = 40.0% T = 60.0%	G = 51.1% T = 48.9%
rs3087243 G > A	GG = 24 (24.0%) GA = 46 (46.0%) AA = 30 (30.0%)	G = 47.0% A = 53.0%	G = 53% A = 47%
rs4553808 A > G	AA = 67 (66.3%) AG = 29 (28.7%) GG = 5 (5.0%)	A = 80.7% G = 19.3%	A = 83.1% G = 16.9%
rs16840252 C > T	CC = 70 (68.6%) CT = 28 (27.5%) TT = 4 (3.9%)	C = 82.4% T = 17.6%	C = 83.1% T = 16.9%
rs231775 A > G	AA = 54 (54.0%) AG = 38 (38.0%) GG = 8 (8.0%)	A = 73.0% G = 27.0%	A = 64.1% G = 35.9%
rs5742909 C > T	CC = 84 (83.2%) CT = 16 (15.8%) TT = 1 (1.0%)	C = 91.1% T = 8.9%	C = 91.6% T = 8.4%
rs2910164 C > G	CC = 4 (4.0%) CG = 35 (35.0%) GG = 61 (61.0%)	C = 21.5% G = 78.5%	C = 26% G = 74%
rs3026321 A > T	AA = 82 (81.2%) AT = 17 (16.8%) TT = 2 (2.0%)	A = 89.6% T = 10.4%	A = 91.4% T = 8.6%
rs4778080 G > T	GG = 13 (12.9%) GT = 51 (50.5%) TT = 37 (36.6%)	G = 38.1% T = 61.9%	G = 36.5% T = 63.5%
rs1738074 T > C	TT = 17 (16.8%) TC = 45 (44.6%) CC = 39 (38.6%)	T = 39.1% C = 60.9%	T = 43.3% C = 56.7%
rs782637 T > C	TT = 20 (19.8%) TC = 40 (39.6%) CC = 41 (40.6%)	T = 39.6% C = 60.4%	T = 46.7% C = 53.3%
rs2062059 A > G	AA = 38 (37.6%) AG = 52 (51.5%) GG = 11 (10.9%)	A = 63.4% G = 36.6%	A = 62.1% G = 37.9%
rs2301756 A > G	AA = 89 (88.1%) AG = 11 (10.9%) GG = 1 (1.0%)	A = 93.6% G = 6.4%	A = 90.3% G = 9.7%
rs359975 C > G	CC = 74 (73.3%) CG = 25 (24.8%) GG = 2 (2.0%)	C = 85.6% G = 14.4%	C = 87.7% G = 12.3%
rs3762513 C > T	CC = 22 (22.7%) CT = 37 (38.1%) TT = 38 (39.2%)	C = 41.8% T = 58.2%	C = 47.8% T = 52.2%
rs55733913 T > C	TT = 61 (61.0%) TC = 34 (34.0%) CC = 5 (5.0%)	T = 78.0% C = 22.0%	T = 76.7% C = 23.3%
rs66502444 A > T	AA = 69 (68.3%) AT = 29 (28.7%) TT = 3 (3.0%)	A = 82.7% T = 17.3%	A = 81.7% T = 18.3%
rs10964859 C > G	CC = 43 (44.3%) CG = 40 (41.2%) GG = 14 (14.5%)	C = 64.9% G = 35.1%	C = 63.3% G = 36.7%
rs12979860 C > T	CC = 51 (54.3%) CT = 30 (31.9%) TT = 13 (13.8%)	C = 70.2% G = 29.8%	C = 69.1% T = 30.9%
rs16906115 G > A	GG = 82 (82.8%) GA = 17 (17.2%) AA = 0	G = 91.4% A = 8.6%	G = 92.5% A = 7.5%
rs2069705 G > A	GG = 10 (10.0%) GA = 41 (41.0%) AA = 49 (49.0%)	G = 30.5% A = 69.5%	G = 32.7% A = 67.3%
rs4988956 G > A	GG = 42 (42.9%) GA = 46 (46.9%) AA = 10 (10.2%)	G = 66.3% A = 33.7%	G = 59.5% A = 40.5%
rs75824728 G > A	GG = 84 (85.7%) GA = 12 (12.3%) AA = 2 (2.0%)	G = 91.8% A = 8.2%	G = 89.7% A = 10.3%
rs10814859 T > C	TT = 54 (55.1%) TC = 33 (34.7%) CC = 10 (10.2%)	T = 72.4% C = 27.6%	T = 69.3% C = 30.7%
rs56328422 T > C	TT = 49 (50.5%) TC = 37 (38.2%) CC = 11 (11.3%)	T = 69.6% C = 30.4%	T = 71.7% C = 28.3%
rs11711517 C > T	CC = 7 (6.9%) CT = 43 (42.6%) TT = 51 (50.5%)	C = 28.2% T = 71.8%	C = 22.4% T = 77.6%
rs12548560 A > G	AA = 46 (45.5%) AG = 44 (43.6%) GG = 11 (10.9%)	A = 67.3% G = 32.7%	A = 70.7% G = 29.3%
rs1276216 A > T	AA = 2 (2.0%) AT = 32 (31.7%) TT = 67 (66.3%)	A = 17.8% T = 82.2%	A = 15.6% T = 84.4%
rs2117997 A > G	AA = 43 (43.0%) AG = 43 (43.0%) GG = 14 (14.0%)	A = 64.5% G = 35.5%	A = 68.8% G = 31.2%
rs239731 T > C	TT = 2 (2.0%) TC = 30 (29.7%) CC = 69 (68.3%)	T = 16.8% C = 83.2%	T = 15.6% C = 84.4%
rs35807769 A > G	AA = 63 (62.4%) AG = 30 (29.7%) GG = 8 (7.9%)	A = 77.2% G = 22.8%	A = 74.8% G = 25.2%
rs470753 G > A	GG = 8 (8.3%) GA = 33 (34.0%) AA = 56 (57.7%)	G = 25.3% A = 74.7%	G = 21.7% A = 78.3%
rs6440251 T > G	TT = 5 (5.1%) TG = 28 (28.9%) GG = 64 (66.0%)	T = 19.6% G = 80.4%	T = 13.9% G = 86.1%
rs246079 A > G	AA = 28 (28.6%) AG = 49 (50.0%) GG = 21 (21.4%)	A = 53.6% G = 46.4%	A = 58.8% G = 41.2%
rs162263 T > G	TT = 45 (44.1%) TG = 45 (44.1%) GG = 12 (11.8%)	T = 66.2% G = 33.8%	T = 66.3% G = 33.7%
rs4800887 G > A	GG = 74 (74.0%) GA = 22 (22.0%) AA = 4 (4.0%)	G = 85.0% A = 15.0%	G = 73.5% A = 26.5%
rs7481951 A > T	AA = 19 (19.0%) AT = 40 (40.0%) TT = 41 (41.0%)	A = 39.0% T = 69.0%	A = 45% T = 55%
rs7954686 A > G	AA = 65 (64.3%) AG = 33 (32.7%) GG = 3 (3.0%)	A = 80.7% G = 19.3%	A = 80.9% G = 19.1%
rs9374 G > A	GG = 71 (71.0%) GA = 25 (25.0%) AA = 4 (4.0%)	G = 83.5% A = 16.5%	G = 76.7% A = 23.3%
rs12720250 G > T	GG = 91 (91.9%) GT = 8 (8.1%) TT = 0	G = 96.0% T = 4.0%	G = 92.1% T = 7.9%
rs6993547 G > C	GG = 68 (69.4%) GC = 26 (26.5%) CC = 4 (4.1%)	G = 82.7% C = 17.3%	G = 79.4% C = 20.6%
rs9268543 A > T	AA = 91 (90.1%) AT = 9 (8.9%) TT = 1 (1.0%)	A = 94.6% T = 5.4%	A = 84.6% T = 15.4%
rs113861051 A > G	AA = 87 (85.3%) AG = 15 (14.7%) GG = 0	A = 92.6% G = 7.4%	A = 91.8% G = 8.2%
rs504963 G > A	GG = 18 (18.4%) GA = 58 (59.2%) AA = 22 (22.4%)	G = 48.0% A = 52.0%	G = 53.4% A = 46.6%
rs5908080 T > C	TT = 16 (16.2%) TC = 3 (3.0%) CC = 80 (80.8%)	T = 17.7% C = 82.3%	T = 25.7% C = 74.3%
rs17388568 G > A	GG = 56 (56.0%) GA = 37 (37.0%) AA = 7 (7.0%)	G = 74.5% A = 25.5%	G = 70.8% A = 29.2%
rs1863800 C > T	CC = 27 (27.6%) CT = 42 (42.8%) TT = 29 (29.6%)	C = 49.0% T = 51.0%	C = 55.0% T = 45.0%
rs2295359 G > A	GG = 39 (39.0%) GA = 47 (47.0%) AA = 14 (14.0%)	G = 62.5% A = 37.5%	G = 68.6% A = 31.4%
rs4143815 G > C	GG = 42 (44.2%) GC = 42 (44.2%) CC = 11 (11.6%)	G = 66.3% C = 33.7%	G = 67.0% C = 33.0%

wt wild-type allele, v variant allele.

Genotype frequencies of the SNPs analysed in our cohort, categorized by the presence or absence of overall ICI-induced toxicity and, separately, by ICI-induced pneumonitis, dermatitis, asthenia or hepatitis are provided in Supplementary Table 1 (Table S1). Initially, comparisons across the three genotypes (wt/wt, wt/v, v/v) were performed and SNPs showing significant associations are shown in Table S1 both according to overall toxicity and to specific toxicities experienced by ≥3 patients. Only the SNP rs1738074 was associated with overall toxicity (p = 0.01, Table S1). To determine whether the presence of the SNP gave rise to higher or lower overall toxicity, genotypes were grouped by the presence or absence of the variant allele. Significant genotype groups comparisons are summarized in Supplementary Table 2 (Table S2). This analysis for rs1738074 revealed that the presence of the variant allele could be associated with higher ICI-induced toxicity of any kind, as the group of v/v patients versus the rest of genotypes showed an almost 5 times greater significant risk (95% CI: 2.06–11.79, p < 0.001, Fig. 1). The SNP rs11711517 did not show significant results when the three genotypes were first compared but did so when genotypes were grouped: all patients (100%) showing any type of irAE were carriers of 1 or 2 variant alleles of rs11711517 indicating the presence of the mutant allele was associated with a higher risk of toxicity of any type (p = 0.045, Fig. 1). Uncertain results were obtained for the rs2117997 polymorphism which seemed associated with the development of overall toxicity (p = 0.046, Table S1) yet genotype group comparisons revealed no significance of the variant allele.

Fig. 1. Significant SNPs obtained in genotype groups comparisons, according to the presence or absence of any type of ICI-induced toxicity (overall toxicity) or specific toxicity.

Odds ratios and 95% confidence intervals are shown.

Genotype frequencies of the 49 SNPs analysed categorized by specific toxicities are also detailed in Table S1. Significant comparisons among three genotypes categorized by the presence or absence of ICI-induced pneumonitis, dermatitis, asthenia and hepatitis highlighted the SNPs: rs11571302 associated with asthenia (p = 0.018); rs1738074, rs2301756, rs66502444, rs12548560, rs9374 and rs5908080 with pneumonitis (p = 0.036, p = 0.000; p = 0.030, p = 0.038, p = 0.032, p = 0.026, respectively); rs1738074, rs55733913 and rs2117997 with dermatitis (p = 0.005, p = 0.043 and p = 0.021, respectively); and rs4988956 with hepatitis (p = 0.027) (Table S1). Among the 6 significant SNPs linked to pneumonitis, 2 also showed significant differences when genotype groups were compared, indicating a greater risk of pneumonitis associated with the variant alleles: rs2301756 associated with a 29.33-fold greater risk of pneumonitis (95% CI: 2.76–312.19; p = 0.005) and rs1738074 with 100% of patients with pneumonitis featuring the genotype v/v (p = 0.020) (Fig. 1). This polymorphism, rs1738074, was also linked to a significantly increased risk of dermatitis (15.74-fold, 95% CI: 1.88–131.59; p = 0.002) as well as the SNPs rs55733913 and rs2117997, both associated with a risk increased more than 6-fold of dermatitis (p = 0.026, p = 0.021, respectively, Fig. 1). In contrast, two SNPs seemed to have a protective effect against asthenia, rs11571302 and rs66502444 (p = 0.012, p = 0.030, respectively, Fig. 1). For hepatitis, only the SNP rs4988956 showed significant differences between genotypes (p = 0.027, Table S1) but comparison among genotype groups did not reveal significant results.

Discussion

The management of urothelial carcinoma presents difficulties due to its molecular diversity and the variable effectiveness of existing treatments. Immune-checkpoint inhibitors have meant a significant advance, strengthening immune responses against tumours adept at evading detection. Improving our knowledge of biomarkers to forecast prognosis and treatment is much needed. In a recent review, it was proposed that FGFR3 and HER2 alterations could serve as biomarkers guiding targeted therapies for UC as PD-L1 expression is unreliable for predicting the ICI response because of inconsistent results [20]. Circulating tumour DNA shows prognostic promise but needs further validation for predictive use, and emerging biomarkers indicate a shift toward comprehensive tumour profiling [20]. Our exploratory study aims to contribute to existing knowledge about tumour SNPs as biomarkers of toxicity induced by ICIs to help choose the best treatment for each patient with UC.

In response to ICI treatment, the SNP rs1738074 in gene TAGAP was associated both to overall toxicity (p = 0.01, Table S1), and to pneumonitis and dermatitis when these specific irAEs were analysed separately (p = 0.036, 0.005, respectively, Table S1). For a more detailed analysis, when genotypes were grouped, the v/v genotype of this SNP showed a greater association with toxicity than the wt/wt + wt/v group, with both overall toxicity (4.93-fold, 95% CI: 2.06–11.79, p < 0.001, Fig. 1) and pneumonitis and dermatitis (p = 0.020 and p = 0.002, Table S1). The TAGAP gene codes for a protein that may function as a Rho GTPase-activating protein and mutations in this gene are thought to be associated with several autoimmune diseases such as, for instance, multiple sclerosis and rheumatoid arthritis [21]. Other authors have also found an association between this SNP and an increased risk of severe irAEs in melanoma patients in a germline whole-exome sequencing study [22]. Our results also associated another SNP, rs11711517, with a higher risk of developing general toxicity (p = 0.045, Table S1). This SNP was observed in high frequency not only in our cohort (71.8%, Table 3) but also in healthy European populations (77.6%, Table 3) and is located in the uncharacterized LOC105377125 genic region that expresses a ncRNA with a possible role in gene expression regulation (Table 1). In contrast, in melanoma patients, this SNP has been linked to reduced odds of developing irAEs [11].

The immune-related cutaneous adverse events dermatitis typically involves pruritus, maculopapular rash, eczema, lichenoid and psoriasiform dermatitis, vitiligo, and bullous dermatoses, and higher frequencies of such reactions have been detected in response to anti-PD-1 agents compared to anti-CTLA-4 inhibitors [23]. In our exploratory study, 8.8% of patients developed dermatitis and our analysis revealed that 3 SNPs were significantly associated with a higher risk of developing this complication. Among these, rs1738074 mentioned above was found in association with overall toxicity as well as with a 15.7-fold increased risk of dermatitis (95% CI: 1.88–131.59, p = 0.002, Table S1). The SNP rs55733913, located in the PACRG gene and associated with autoimmune thyroid diseases, was found to increase the risk of dermatitis in our cohort of UC patients by 6.4-fold (95% CI: 1.27–32.91, p = 0.026, Fig. 1). The SNP rs2117997, belonging to genic region LOC124902908 was related to a non-coding RNA, increasing by 6.5-fold the risk of dermatitis (95% CI: 1.49–28.18, p = 0.021, Fig. 1). Similar results were observed in previous work of our team, where tumoral SNP rs55733913 was also linked with 6.77-fold higher risk of hepatitis in melanoma patients treated with ICIs [16]. The SNPs rs55733913 and rs2117997 have been examined in the genetic germline of melanoma patients treated with ICIs, but with contradictory results: the variant alleles were associated with a reduced likelihood of developing ICI-induced general toxicity [11]. Despite these opposing results, it could be of interest to clarify the role of these biomarkers in ICI-induced toxicity for use in cancer patients under treatment with ICIs.

Immune checkpoint inhibitor-induced pneumonitis is a relatively rare but clinically significant irAE affecting the lung parenchyma. Its overall incidence is estimated at between 5 and 19% for any grade and 6% for high-grade events, with a higher frequency observed in patients receiving anti-PD-1 inhibitors versus anti-CTLA-4 agents, and it can be severe and potentially life-threatening [24]. General clinical manifestations include cough, chest discomfort, dyspnoea, and occasionally fever or hypoxia. Differentiating ICI-induced pneumonitis from other causes such as chronic obstructive pulmonary disease, infection, or chemotherapy- or targeted therapy-related pneumonitis is challenging due to overlapping radiologic and symptomatic profiles [25]. The SNP rs1738074 was again associated with a significantly higher risk of ICI-induced pneumonitis in patients homozygous for the variant allele compared to wt/wt + wt/v patients (p = 0.020, Fig. 1). Another SNP, rs2301756 (in the PTPN11 gene) similarly gave rise to a 29.3-fold risk of ICI-induced pneumonitis in our group of patients with at least one variant allele (95% CI: 2.76–312.19, p = 0.005, Fig. 1). The protein encoded by this gene is a type of protein tyrosine phosphatase known for its role in regulating cell processes including cell growth, differentiation and oncogenic transformation (Table 1). Our analysis of ICI-induced hepatitis in our patient cohort revealed a significant difference among genotypes of SNP rs4988956 (p = 0.027, Table S1). Although comparison among genotype groups did not reveal significant results, it was observed a tentative risky effect of the variant allele because 91.5% of patients without hepatitis were wt/wt or wt/v (Table S1).

Asthenia was the most common irAE detected in our cohort of patients (18.6%, Table 2). In patients with metastatic melanoma treated with ICIs, asthenia has been correlated with improved disease control [26]. Further, a lower risk of ICI-induced asthenia has been described for PD-1/PD-L1 inhibitors than CTLA4 inhibitors [27]. Although our patients were not treated with the anti-CTLA4 drug, we found significant correlation between the presence of at least one mutant allele near of the CTLA4 gene SNP rs11571302 and protection against ICI-induced asthenia (p = 0.012, Fig. 1). Xiang et al. described a relationship between the presence of this SNP and a higher incidence of allergic rhinitis in response to environmental allergens [28]. Similarly, the SNP rs66502444, was also associated with a 5-fold lower risk of ICI-induced asthenia in patients with at least one variant allele (0.20-fold, 95% CI: 0.04–0.94, p = 0.030, Fig. 1). These results were confirmed by Abdel-Wahab et al. who also found that the presence of the variant allele of this SNP had a protective effect against irAEs in patients with melanoma receiving ICIs [11]. Nevertheless, in a previous study, we correlated the presence of this allele with a higher risk of ICI-induced hepatitis in patients with melanoma treated with ICIs [16].

Although immunotherapy has improved the treatment of UC, the best management strategy for toxicity remains unresolved, particularly in terms of patient selection, long-term toxicity and the use of predictive biomarkers. Despite its limited sample size, this exploratory study offers interesting preliminary data about SNPs associated with overall toxicity and with specific adverse effects such as pneumonitis, asthenia, dermatitis or hepatitis. As other authors have proposed, i.e. Membribes el al., there is growing consensus that no single biomarker is likely sufficient to address the heterogeneity linked to UC and further work is needed to identify useful biomarkers through the integration of genomic, proteomic and immune signatures with clinical parameters across different platforms including IA [20].

Supplementary information

Table S1 ^{(63.1KB, docx)}

Table S2 ^{(30KB, docx)}

Author contributions

AFS wrote the main manuscript; AFS, ARL, MG, AN and BAP designed the research; ARL, MG and JMR processed samples and performed genotyping; AN revised and selected bibliography for prepared project; FIC, IA, RK, MB selected the paraffin-embedded tumour sections with >50% cellularity; MC, AMM, SH, DM selected patients, collected informed consents and clinical-pathological variables; MR and JMR performed statistical analysis of data; All authors reviewed the manuscript.

Funding

This work was supported by the grant Dr. Luis Álvarez 2023 (IdiPAZ).

Data availability

All data were available in the tables, figure and the supplementary information.

Competing interests

Authors declare no competing financial interests.

Ethics approval and consent to participate

All participants provided written informed consent for study participation. The study was approved by the Ethics Committee of the University Hospital of Fuenlabrada (identification code: APR 18/20, September 2018) and all methods were conducted in accordance with the Declaration of Helsinki.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41397-026-00423-3.