rs10204525 binding to miR-4717-3p modulates PD-1 expression and predicts the development of immune-related adverse events in patients with advanced cancer treated with anti-PD-1/PD-L1 therapy

Abstract

Background

Predictive biomarkers for anti-programmed cell death-1 (PD-1) and anti-programmed cell death ligand 1 (PD-L1) therapy are needed. Here, we validated the role of PD-1 single nucleotide polymorphisms (SNPs) in predicting the development of immune-related adverse events (irAEs) in patients with advanced cancer treated with anti-PD-1/PD-L1-based immunotherapy and defined the molecular mechanisms underlying the role of identified SNP candidate.

Methods

Blood samples, clinical-pathological characteristics, survival outcomes and irAEs were collected from two cohorts of patients: (1) patients with advanced cancer treated with anti-PD-1/PD-L1 alone and (2) patients with advanced non-small cell lung cancer (NSCLC) treated with anti-PD-1 in combination with platinum-based chemotherapy with or without anti-cytotoxic T-lymphocyte antigen 4 (CTLA-4) therapy. PD-1 SNPs including rs2227981, rs7421861, rs11568821, rs36084323, rs2227982 and rs10204525 were genotyped and correlated with clinical-pathological characteristics and irAEs. Putative miRNAs binding to PD-1 SNP candidate were identified by in silico analysis. Validation of miRNA binding to PD-1 SNP allele specificity as well as evaluation of the induced PD-1 modulation was performed in vitro using patient-derived peripheral blood mononuclear cells (PBMCs). Susceptibility of non-cancer cells to immune cells incubated with anti-PD-1 based on PD-1 SNP allele specificity and miRNA modulation was performed by co-culturing non-cancer human epidermal keratinocyte (HaCaT) cells and bronchial epithelial BEAS-2B cells with human leukocyte antigen (HLA)-matched PBMCs, obtained from patients with cancer treated with anti-PD-1/PD-L1-based immunotherapy and carrying a different PD-1 SNP.

Results

Most of the analyzed PD-1 SNPs were not associated with the development of irAEs. In contrast, rs10204525 exhibited a significant association with the occurrence of both grade 1–2 and 3–4 irAEs in both cohorts of patients. Specifically, patients carrying C/C had a higher rate of irAEs as compared with those carrying C/T. rs10204525 mapped on 3′ untranslated region (3’-UTR) region of PD-1. miR-4717-3p bound to rs10204525 based on its allele specificity. Modulation of miR-4717-3p expression as well as of miR-4717-3p binding to rs10204525 differentially regulated PD-1 expression and induction in PMBCs harboring C/C or C/T genotypes as well as their ability to recognize and destroy HLA-matched HaCaT cells, even more in the presence of anti-PD-1 therapy. Specifically, PBMCs carrying a C/T genotype displayed a significantly lower ability to recognize and destroy non-cancer cells as compared to those carrying C/C. These results were further validated by co-culturing of both BEAS-2B and HaCaT non-cancer cells with PBMCs carrying differential rs10204525 genotypes, isolated from additional patients with cancer, incubated with anti-PD-1 or anti-PD-1 in combination with anti-CTLA-4 therapy.

Conclusions

These findings have high clinical relevance since they define rs10204525 binding to miR-4717-3p-mediated PD-1 expression and induction as a mechanism modulating the reactivity of immune cells to non-cancer cells as well as a novel biomarker for predicting irAEs in patients with advanced cancer treated with anti-PD-1/PD-L1-based immunotherapy.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Predictive biomarkers for immune‑related adverse events (irAEs) in patients with cancer treated with anti‑programmed cell death-1 (PD‑1)/programmed cell death ligand 1 (PD‑L1)-based immunotherapy are limited.

WHAT THIS STUDY ADDS

• This study demonstrates that rs10204525 single nucleotide polymorphisms (SNPs) in PD-1 gene are significantly associated with the occurrence of both grade 1–2 and grade 3–4 irAEs in patients with advanced cancer treated with anti‑PD‑1/PD‑L1 therapy alone or in combinations.

• It reveals a molecular mechanism: allele‑specific binding of miR‑4717‑3p to rs10204525 differentially modulates PD‑1 expression in peripheral blood mononuclear cells and influences immune cell reactivity toward non‑cancer cells in the presence of immune checkpoint inhibitors.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• rs10204525 SNP could serve as a clinically useful biomarker to stratify patients by risk of irAEs prior to immunotherapy, allowing for personalized monitoring or potential implementation of more intensified/de-intensified combinatorial strategies of immune checkpoint inhibitor-based immunotherapies.

Background

Immune checkpoint inhibitor (ICI)-based immunotherapy, employing monoclonal antibodies (mAbs), targeting cytotoxic T-lymphocyte antigen 4 (CTLA-4), programmed cell death-1 (PD-1) and its ligand 1 (PD-L1) has dramatically changed the management of several types of advanced solid tumors such as head and neck squamous cell carcinoma (HNSCC), melanoma, non-small cell lung cancer (NSCLC) and renal cell carcinoma (RCC).1,14 Nonetheless, the efficacy of ICI-based immunotherapy is limited to a subset of treated patients, with only a subgroup of patients achieving long-term clinical benefit. In addition, the development of severe immune-related adverse events (irAEs) can lead to permanent or even fatal consequences. As a result, there is an urgent need to identify efficient predictive biomarkers for both tumor response and development of severe irAEs. Based on ICI-based immunotherapy and cancer type, severe irAEs are reported in about 10–60% of treated patients.1,11 irAEs manifest as autoimmune conditions at different times of onset, capable of virtually affecting any organ or tissue, often resembling “de novo” autoimmune diseases.^{15 16} Severe irAEs require a withdrawal from ICI treatment in most cases.1,1115 16 While several biomarkers have been investigated for ICI clinical benefit, no biomarker has been identified for predicting the development of irAEs.17,24 Genetic variants can influence multiple aspects of different types of disease including cancer. Over 70,000 genetic variants (mainly, single nucleotide polymorphisms (SNPs)) are associated with specific cancer features, survival outcomes, treatment response and drug-related toxicity. PD-1 SNPs have been clearly linked to the development of autoimmune conditions such as Crohn’s disease, systemic lupus erythematosus, type I diabetes, rheumatoid arthritis and multiple sclerosis,²⁵ as well as to a differential risk of cancer26,31 and to viral control.³² On the other hand, contrasting results have been reported between PD-1 SNP and development of ICI-mediated irAEs. In the present study, we validated the role of PD-1 SNPs in predicting the development of irAEs in a cohort of patients with advanced cancer treated with anti-PD-1/PD-L1 therapy as well as to define the molecular mechanisms underlying the role of the identified SNP candidate.

Methods

Study population

The study was performed without interfering with clinical practice. Two cohorts of patients were included. In the first, Caucasian patients with confirmed advanced tumors, treated with single-agent anti-PD-1/PD-L1 as first or later lines of treatment in the specific disease, were recruited from July 2017 to June 2022 at “San Giovanni di Dio e Ruggi D’Aragona” University Hospital. In the second cohort, Caucasian patients with confirmed diagnosis of advanced NSCLC, treated as first line with platinum-based chemotherapy (PBCT) and pembrolizumab (anti-PD-1) or PBCT and nivolumab (anti-PD-1) plus ipilimumab (anti-CTLA-4), were recruited from April 2021 to May 2024 at the same institution. Selection of patients to be included in the studies was performed based on: (1) signed informed consent for clinical-pathological data acquisition; (2) age >18 years; (3) no previous treatment with anti-PD-1/PD-L1 mAbs; (4) treatment with baseline prednisone equivalent dose ≤10 mg/day; (5) absence of symptomatic brain metastases; (6) absence of active autoimmune disease; and (7) signed informed consent for blood sample collection and analysis. Based on patient availability, the first cohort included patients with advanced HNSCC, melanoma, NSCLC and RCC. Patients with NSCLC with epidermal growth factor receptor (EGFR), anaplastic lymphoma kinase (ALK), c-ros oncogene 1 (ROS1), V-raf murine sarcoma viral oncogene homolog B (BRAF), mesenchymal-epithelial transition factor (MET), REarranged during Transfection (RET), human epidermal growth factor receptor 2 (HER2) and neurotrophic tropomyosin receptor kinases (NTRK) tumor alterations were excluded from both the cohorts. Evaluation of ALK, BRAF, EGFR, HER2, MET, NTRK, RET and ROS1 alterations was performed on tumor samples (when available) or liquid biopsy according to national pathology guidelines. Clinical-pathological characteristics including age, sex, Eastern Cooperative Oncology Group (ECOG) performance status (PS), smoking status, comorbidities, presence/absence of asymptomatic brain metastases, previous chemotherapy and/or targeted therapy were collected. Then, the response rate, median progression-free survival (PFS), median overall survival (OS) and rates of irAEs were calculated. Patient privacy and personal data were preserved by assigning a progressive anonymous identification number. Patients received anti-PD-1 (nivolumab or pembrolizumab), anti-PD-L1 (atezolizumab) and anti-CTLA-4 (ipilimumab) mAbs until progressive disease (PD), unacceptable toxicity or clinical indications according to European Society for Medical Oncology guidelines. irAEs were defined as adverse events displaying a certain, likely, or possible correlation with ICIs according to Common Terminology Criteria for Adverse Events (CTCAE) V.4.0.³³ irAEs were graded according to CTCAE V.4.0³³ and prospectively collected. Safety assessments were made continuously during the ICI treatment. For patients who discontinued ICI, safety assessments were continued up to 100 days after the last dose of ICI. Radiographic imaging was performed every 8 weeks. Response rate was determined according to Response Evaluation Criteria in Solid Tumors V.1.1³⁴ and reported as complete response (CR), partial response (PR), stable disease (SD) and PD. Objective response rate (ORR) was defined as the proportion of patients with a CR or PR. PFS was defined as the time from the start of treatment to the first documented PD or death by any cause. OS was defined as the time from the start of treatment to death by any cause or last follow-up date. Patients dead from COVID-19 were excluded from the study. The study was approved by the local ethics committee (prot./SCCE n.85275), in accordance with the Declaration of Helsinki and its amendments.

PD-1 SNP genotyping

Six previously reported PD-1 SNPs (rs10204525, rs2227981, rs7421861, rs11568821, rs36084323 and rs2227982) related to autoimmunity, cancer predisposition and cancer prognosis were selected to be genotyped.26,31 Peripheral blood mononuclear cells (PBMCs) were obtained from recruited patients before starting treatment with anti-PD-1/PD-L1-based immunotherapy and isolated as described.³⁵ Isolated PBMCs were stored at −80°C. PD-1 SNPs were genotyped from DNA extracted from isolated PBMCs using the Maxwell 16 blood DNA purification kit (Promega, Madison, Wisconsin, USA) according to the manufacturer’s protocol. Quantity and quality of purified DNA were assessed in every sample using a NanoDrop spectrophotometer (Thermo Scientific, Wilmington, Delaware, USA). Genotyping of PD-1 SNPs was performed using TaqMan genotyping assay (PD-1: C_172862_10 for rs10204525, C_57931286_20 for rs2227981, C_26891639_10 for rs7421861, C_57931290_10 for 11568821, C_57931321_10 for rs36084323 and C_57931287_10 for rs2227982) as described.³⁶

In silico analysis

PD-1 SNPs were mapped using National Center for Biotechnology Information single nucleotide polymorphism database (NCBI dbSNP) (http://www.ncbi.nlm.nih.gov/SNP) and ENSEMBL V.58 (http://www.ensembl.org/) databases. Identification of putative microRNAs (miRNAs) targeting the binding sites within the 3′-UTR of PD-1, encompassing rs10204525, was based on miRNASNP-v4 (https://gong_lab.hzau.edu.cn/miRNASNP/snpdetail). Minimum-free energy of hybridization (MFE) score, based on the binding affinity of identified miRNAs, was evaluated using the miRTarbase web server (https://awi.cuhk.edu.cn/~miRTarBase/miRTarBase_2025/php/index.php).

Cell cultures

PBMCs were isolated from blood sample collection of patients included in the study. PBMCs were cultured in Roswell Park Memorial Institute (RPMI) medium (Euroclone, Milan, Italy) supplemented with 10% fetal bovine serum (FBS) (Euroclone) and 1% penicillin-streptomycin (Euroclone), at 37°C in a humidified incubator containing 5% CO₂. Human epidermal keratinocyte (HaCaT) and bronchial epithelial (BEAS-2B) cell lines were obtained from the American Type Culture Collection (ATCC; Manassas, Virginia, USA). HaCaT cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM) (Euroclone) supplemented with 10% FBS (Euroclone) and 1% penicillin-streptomycin (Euroclone); BEAS-2B cells were cultured in airway epithelial cell basal medium (Lonza; Basel, Switzerland) supplemented with the bronchial epithelial cell growth kit (Lonza) according to the manufacturer’s recommendations (ATCC). All cells were regularly tested for Mycoplasma contamination with the use of a MycoAlert Mycoplasma Detection Kit (Lonza) following the manufacturer’s protocol.

Cell transfection experiments

PBMCs were seeded into 6-well plates at a density of 2×10⁶ cells per well. All transfection experiments were carried out by reverse transfection with HiPerFect reagent (QIAGEN, Hilden, Germany) according to the manufacturer’s instructions. The following oligos at the indicated concentration were used: 5 nM of miR-4717–3 p mimic (catalog no. 339173, QIAGEN); 60 nM of inhibitors (catalog no. 232724697, IDT, Coralville, Iowa, USA); 7.5 nM of PD-1-specific miR-4717–3 p target site blocker (TSB) (catalog no. 339194, sequence: AGGGTGGGCACATGGGG, QIAGEN) or recommended negative miRNA mimic (catalog no. 339173, QIAGEN), NC1 negative inhibitor (IDT), and negative TSB (sequence: ACGTCTATACGCCCA, QIAGEN). Following a 48 hours transfection period, PBMCs were incubated with 100 ng/mL of interferon (IFN)-γ (PeproTech EC, London, UK) for 48 hours. Functional assays relative to RNA and protein analyses were performed at 24 hours and 48 hours post-transfection, respectively.

RNA isolation and quantitative real-time

Following a 24 hours post-transfection period, miRNAs were extracted from PBMCs using the miRNeasy Mini Kit (catalog no. 217004, QIAGEN) according to the manufacturer’s instructions. The quantity and quality of RNA samples were assessed using a NanoDrop spectrophotometer (Thermo Scientific). For miRNA detection, 10 ng of RNA were reverse transcribed using the miRNA LNA PCR starter kit (catalog no. 3616517, QIAGEN) including U6 as endogenous control. Expression levels of miR-4717–3 p were normalized to miR103a-3p and relative expression levels were calculated within each independent experiment using the formula 2−ΔΔCT. Total RNA was extracted using TRIzol (Thermo Fisher, Waltham, Massachusetts, USA) according to the manufacturer’s protocol. Complementary DNA (cDNA) was transcribed using the SensiFAST cDNA Synthesis Kit (Bioline) according to the manufacturer’s protocol. Reverse transcription reactions were performed in a T100 Thermal Cycler (Bio-Rad Laboratories, Hercules, California, USA) according to the manufacturer’s protocol. For evaluation of PD-1 messenger RNA (mRNA) expression levels, quantitative real-time PCR (qRT-PCR) was used. GAPDH was used as an internal reference gene. qRT-PCR was performed in duplicate using SensiFAST SYBR No-ROX Kit (Bioline, Memphis, Tennessee, USA) according to the manufacturer’s protocol. PCR reactions were carried out in LightCycler 480 II (Roche, Basel, Switzerland). Primer sets were PD-1 fw: CGTGGCCTATCCACTCCTCA; PD-1 rv: ATCCCTTGTCCCAGCCACTC; GAPDH fw: CTGACTTCAACAGCGACACC, GAPDH rv: TAGCCAAATTCGTTGTCATACC. Gene expression profile analysis was performed using the 2−ΔΔCq method.³⁷ Data are representative of the results obtained in three independent experiments.

Western blot analysis

Following a 48 hours post-transfection period, PBMCs were harvested and lysed as described.³⁸ Cell lysates were analyzed by western blot with the following antibodies: PD-1 (catalog no. 86163), GAPDH- (catalog no. 5174) specific Abs, horseradish peroxidase anti-rabbit Ab (catalog no. 7074) (Cell Signaling, Danvers, Massachusetts, USA). Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as a loading control. Data are representative of the results obtained in three independent experiments.

Flow cytometry analysis

PBMCs were seeded into 6-well plates at a density of 2×10⁶ cells per well and incubated with IFN-ɣ (100 ng/mL). Untreated cells were used as a control. Following a 48 hours incubation, PBMCs were harvested, and cell surfaces were stained with phycoerythrin (PE) anti-PD-1 IgG1k Ab (catalog no. 329905, BioLegend, San Diego, California, USA). PE anti-mouse IgG1k (catalog no. 406607, BioLegend) was used as a specificity control for PD-1 staining. Staining was performed as described.³⁵ Stained cells were analyzed using a FACSVerse flow cytometer (BD Biosciences, Swindon, UK). Data expressed as mean fluorescence intensity are representative of the results obtained in three independent experiments.

Human leukocyte antigen (HLA) class I genotyping

HLA class I genotyping was performed at the Transplant Hematological Unit of “San Giovanni di Dio e Ruggi D’Aragona” University Hospital. HLA class I genotypes were characterized as described.³⁹ HLA class I genotyping was validated using the TRON Cell Line Portal (https://www.cellosaurus.org/CVCL_0038).

Co-culture of non-cancer cells and HLA-matched PBMCs incubated with anti-PD-1 mAb alone or in combination with anti-CTLA-4 mAb

PBMCs isolated from six recruited patients were selected and cultured based on (1) tumor type, (2) achievement of clinical benefit from anti-PD-1/PD-L1±anti-CTLA-4 therapy, (3) presence of different rs10204525 genotype (C/C or C/T), and (4) HLA class I antigen matching with the HaCaT or BEAS-2B cell lines. Transfected PBMCs (following a 48 hours post-transfection period) and non-transfected PBMCs were added to HaCaT or BEAS-2B cells seeded into 24-well plates at a density of 2×10⁵ cells per well (5:1 ratio) and incubated with 10 µg/mL of anti-PD-1 nivolumab (Thermo Fisher) ±3.3 µg/mL of anti-CTLA-4 ipilimumab (Thermo Fisher) for 24 hours and 48 hours, respectively. A mass ratio of approximately 3:1 of nivolumab/ipilimumab incubation was used to emulate clinical regimens in CheckMate 012 and CheckMate 9LA trials, where patients received nivolumab at 3 mg/kg combined with ipilimumab at 1 mg/kg (≈3:1 by mass).^{14 40} Purified human IgG4 and IgG (BioLegend) were used as isotype controls for nivolumab and/or ipilimumab. Monocultured HaCaT and BEAS-2B cells were used as controls in all co-culture experiments.

Cell viability assay

Following a 24 hours or 48 hours of co-culture, HaCaT or BEAS-2B cells were washed with phosphate-buffered saline (PBS) and their viability was determined by cell counting kit-8 assay (Dojindo Laboratories, Rockville, Maryland, USA) according to manufacturers’ protocol. The absorbance at 450 nm was determined using the Sunrise microplate reader (TECAN, Männedorf, Switzerland). All experiments were performed in triplicates and repeated three times.

IFN-γ ELISA

Following a 24 hours or 48 hours of co-culture, the culture medium was harvested and IFN-γ levels were measured using ELISA Max Deluxe Set Human IFN-γ (BioLegend) assay according to the manufacturers’ protocol. The absorbance at 450 nm was determined using the Sunrise microplate reader (TECAN). All experiments were performed in triplicates and repeated three times.

Annexin V-FITC/PI assay

Following a 24 hours or 48 hours of co-culture, HaCaT and BEAS-2B cells were washed with PBS and their apoptotic rate was determined by Annexin V-FITC/PI assay using an Annexin V-FITC Early Apoptosis Detection Kit (Cell Signaling Technology) to distinguish viable, apoptotic, and necrotic cells according to the manufacturer’s protocol. Stained cells were analyzed using a FACSVerse flow cytometer. Data are representative of the results obtained in three independent experiments.

Statistical analysis

All data was collected using Microsoft Excel. Statistical analyses were performed using Stata V.13 software released by StataCorp LP (College Station, Texas, USA) or GraphPad Prism V.6.0 released by GraphPad Software (La Jolla, California, USA). Continuous data were expressed as medians and ranges, whereas categorical data were expressed as frequencies and percentages. PFS and OS were calculated using the Kaplan-Meier method. Correlations between clinical-pathological characteristics, ORR and development of irAEs with PD-1 SNPs were performed using the Fisher’s exact test, Mann-Whitney U test and the Kruskal-Wallis method, as appropriate. Correlations between clinical-pathological characteristics, PD-1 SNPs and development of irAEs with survival outcomes (PFS and OS) were performed using the log-rank test. The difference between groups was calculated using the two-sided, unpaired t-test or one-way analysis of variance. The difference between groups was considered significant when the p value was <0.05.

Results

Clinical-pathological characteristics of patients with cancer treated with anti-PD-1/PD-L1 therapy

A total of 72 Caucasian patients with a confirmed diagnosis of advanced cancer from “San Giovanni di Dio e Ruggi D’Aragona University Hospital” were included in the first cohort. 49 (68.06%), 9 (12.50%), 8 (11.11%) and 6 (8.33%) patients were affected by advanced NSCLC, RCC, HNSCC and melanoma, respectively. Baseline medical record information including clinical-pathological characteristics of patients is summarized in table 1.

Table 1. Baseline clinical-pathological characteristics of patients included in the first cohort.

Median age:	66 years (range, 43–84)
Sex
Male	59 (81.94%)
Female	13 (18.06%)
ECOG PS
0	26 (36.11%)
1	28 (38.89%)
2	18 (25.00%)
Smoking status
People who have never smoked	13 (18.06%)
People who have quit smoking	44 (61.11%)
People who smoke	15 (20.83%)
Comorbidities
Hypertension	43 (59.72%)
Dyslipidemia	19 (26.39%)
Diabetes	14 (19.44%)
COPD	8 (11.11%)
Heart failure	6 (8.33%)
Chronic renal failure	0 (0.00%)
Immune disorder	0 (0.00%)
Type of cancer
HNSCC	8 (11.11%)
NSCLC	49 (68.06%)
Melanoma	6 (8.33%)
RCC	9 (12.50%)
Asymptomatic brain metastases
Yes	14 (19.44%)
No	58 (80.56%)
Line of treatment
First line	11 (15.28%)
Second line	61 (84.72%)
Previous chemotherapy
Yes	51 (70.83%)
No	21 (29.17%)
Previous targeted therapy
Yes	10 (13.89%)
No	62 (86.11%)
Type of mAb
Nivolumab	55 (76.39%)
Pembrolizumab	9 (12.50%)
Atezolizumab	8 (11.11%)
Median number of anti-PD-1/PD-L1 therapy cycle	8 (1-83)
Type of response
CR	4 (5.56%)
PR	16 (22.21%)
SD	12 (16.67%)
PD	40 (55.56%)
ORR	20 (29.41%)
Survival outcomes
Median follow-up	21.93 months (range, 5.53–46.67)
Median PFS	4.15 months (range, 0.46–46.67)
Median OS	10.70 months (range, 0.46–46.67)

COPD, chronic obstructive pulmonary disease; CR, complete response; ECOG PS, Eastern Cooperative Oncology Group performance status; HNSCC, head and neck squamous cell carcinoma; mAb, monoclonal antibody; NSCLC, non-small cell lung cancer; ORR, objective response rate; OS, overall survival; PD, progressive disease; PD-1, programmed cell death-1; PD-L1, programmed cell death ligand 1; PFS, progression-free survival; PR, partial response; RCC, renal cell carcinoma; SD, stable disease.

The median age was 66 years (range, 43–84 years). 59 patients (81.94%) were male. 54 (75.00%) and 18 (25.00%) had ECOG PS of 0–1 and 2, respectively. 13 patients (18.06%) were people who have never smoked, while 44 (61.11%) and 15 (20.83%) were people who have quit smoking and people who smoke, respectively. Relevant comorbidities included hypertension (59.72%), dyslipidemia (26.39%), diabetes (19.44%), chronic obstructive pulmonary disease (11.11%) and heart failure (8.33%). Asymptomatic brain metastases were present in 14 (19.44%) patients. Most of the patients (84.72%) were treated with anti-PD-1/PD-L1 therapy as second line of treatment. 51 (70.83%) and 10 (13.89%) patients had previously received chemotherapy and targeted therapy, respectively. Anti-PD-1 and anti-PD-L1 mAbs were administered in 88.89% and 11.11% of patients, respectively. 55 (76.39%) and 9 (12.50%) patients were treated with anti-PD-1 nivolumab and pembrolizumab, respectively; 8 (11.11%) patients were treated with anti-PD-L1 atezolizumab. ORR was 27.78%. CRs, PRs, SDs and PDs were reported in 4 (5.56%), 16 (22.21%), 12 (16.67%) and 40 (55.56%) patients, respectively. At a median follow-up of 21.93 months (5.53–46.67 months), 21 out of 72 patients (29.17%) were still alive. Median PFS and OS were 4.15 (0.46–46.67 months) and 10.70 months (0.46–46.67 months), respectively (figure 1).

Figure 1. PFS and OS of patients with advanced cancer treated with anti-PD-1/PD-L1 mAbs, atezolizumab, nivolumab and pembrolizumab. At a median follow-up of 21.93 months (5.53–46.67 months), median PFS and OS were 4.15 months (A) and 10.70 months (B), respectively. PFS and OS analysis were performed using the Kaplan-Meier method. mAbs, monoclonal antibodies; PD-1, programmed cell death-1; PD-L1, programmed cell death ligand 1; PFS, progression-free survival; OS, overall survival.

Grade 1–2 and grade 3–4 irAEs were reported in 45 (62.50%) and 6 (8.33%) of treated patients, respectively (table 2).

Table 2. Rates of irAEs reported in the first cohort.

	Grade 1–2	Grade 3–4
Any event	45 (62.50%)	6 (8.33%)
Led to discontinuation of treatment	0 (0.00%)	3 (4.17%)
Adrenal insufficiency	8 (11.11%)	0 (0.00%)
Amylase increase	2 (2.78%)	0 (0.00%)
Arthritis	3 (4.17%)	0 (0.00%)
Asthenia	23 (31.94%)	0 (0.00%)
Alanine aminotransferase increase	1 (1.39%)	1 (1.39%)
Aspartate aminotransferase increase	3 (4.17 %)	1 (1.39%)
Colitis	2 (2.78%)	0 (0.00%)
Creatinine increase	7 (9.72%)	0 (0.00%)
Decreased appetite	5 (6.94%)	2 (2.78%)
Diarrhea	7 (9.72%)	0 (0.00%)
Fever	8 (11.11%)	0 (0.00%)
Gynecomastia	1 (1.39%)	0 (0.00%)
Hypophysitis	1 (1.39%)	0 (0.00%)
Lipase increase	3 (4.17 %)	1 (1.39%)
Nausea	4 (5.56%)	3 (4.17 %)
Oral mucositis	3 (4.17%)	0 (0.00%)
Pancreatitis	1 (1.39%)	2 (2.78%)
Pneumonitis	2 (2.78%)	1 (1.39%)
Pruritus	9 (12.50%)	0 (0.00%)
Rash	4 (5.56%)	1 (1.39%)
Thyroiditis	18 (25.00%)	0 (0.00%)
Vitiligo	1 (1.39%)	0 (0.00%)
Vomiting	5 (6.94%)	0 (0.00%)

irAE, immune-related adverse event.

The most frequently reported irAEs of grade 1–2 and grade 3–4 were asthenia (31.94%) and nausea (4.17 %), respectively. Three (4.17%) patients discontinued anti-PD-1/PD-L1 therapy because of the development of irAEs. No treatment-related death was reported.

Genotyping and mapping of PD-1 SNPs in patients with cancer treated with anti-PD-1/PD-L1 therapy

rs2227982, rs36084323, rs11568821, rs7421861, rs2227981 and rs10204525 PD-1 SNPs were genotyped in the study population. The frequencies of analyzed PD-1 SNPs are described in table 3.

Table 3. Frequencies of PD-1 SNP genotypes.

Genotype	Frequency	Location
rs2227982:
G/G	69 (95.83%)	Exon
G/A	3 (4.17%)
A/A	0 (0.00%)
rs36084323:		Promoter region
C/C	69 (95.83%)
C/T	3 (4.17%)
T/T	0 (0.00%)
rs11568821:
C/C	58 (80.56%)	Intron
C/T	13 (18.06%)
T/T	1 (1.39%)
rs7421861:
A/A	41 (56.94%)	Intron
A/G	26 (36.11%)
G/G	5 (6.94%)
rs2227981:
G/G	11 (25.00%)	Exon
A/G	21 (47.73%)
A/A	12 (27.27%)
rs10204525:
C/C	60 (83.33%)	3’-UTR
C/T	12 (16.67%)
T/T	0 (0.00%)

PD-1, programmed cell death-1; SNPs, single nucleotide polymorphisms; 3'-UTR', 3′ untranslated region.

Mapping of PD-1 SNP analyzed showed that rs2227982 and rs2227981 localized on an exonic region (chr2:241 851 281 and chr2:241851121, respectively); rs11568821 and rs7421861 on intronic regions (chr2:241 851 760 and chr2:241853198, respectively); rs36084323 on promoter region (chr2:241859444) and rs10204525 on 3’-UTR region (chr2:241850169).

Association between clinical-pathological characteristics, clinical outcomes, irAEs and PD-1 SNPs in patients with cancer treated with anti-PD-1/PD-L1 therapy

No significant associations between clinical-pathological characteristics, clinical outcomes and development of irAEs were found. In addition, no significant associations between clinical-pathological characteristics, clinical outcomes and PD-1 SNPs were detected. Lastly, no significant associations between PD-1 SNPs including rs2227981, rs7421861, rs11568821, rs36084323, rs2227982 and development of irAEs were found. However, the presence of different genotypes in rs10204525 significantly correlated with the development of irAEs. Specifically, patients carrying the C/C genotype in rs10204525 developed more grade 1–2 and 3–4 irAE (p=0.0053 and p<0.0001, respectively) as compared with patients carrying the C/T genotype. Worth noting, in patients carrying a C/T genotype in rs10204525, no grade 3–4 irAEs were reported, as well as no grade 1–2 irAEs of special interest including arthritis, alanine/aspartate aminotransferase increase, pancreatitis, nausea, and adrenal insufficiency.

To further validate these results, the predictive role of rs10204525 was investigated in patients with advanced non-oncogene addicted NSCLC who were treated in first-line therapy with the combination of PBCT and pembrolizumab or the combination of PBCT and nivolumab plus ipilimumab. 27 patients were enrolled from April 2021 to May 2024. Baseline medical record information, including clinical-pathological characteristics of patients, is summarized in online supplemental table 1. Grade 1–2 and grade 3–4 irAEs were reported in 18 (66.67%) and 9 (33.33%) of treated patients, respectively (online supplemental table 2). The most frequently reported irAEs of grade 1–2 and grade 3–4 were asthenia (33.33%) and pneumonitis (14.81%), respectively. Five (18.52%) patients discontinued combination therapy because of the development of irAEs. No treatment-related death was reported. rs10204525 PD-1 SNP was genotyped as C/C, C/T, and T/T in 18 (66.67%), 9 (33.33%) and 0 (0.00%) patients, respectively. Even in this cohort, patients carrying the C/C genotype in rs10204525 developed more grade 1–2 and 3–4 irAE (p=0.0012 and p=0.0116, respectively) as compared with patients carrying the C/T genotype. No grade 3–4 irAEs in patients carrying a C/T genotype in rs10204525 were detected.

Identification of putative miRNAs binding to rs10204525 in 3’-UTR of PD-1 based on its allele specificity

rs10204525 mapped on 3’-UTR region of PD-1. SNPs localized on 3’-UTR regions can affect the binding affinity of specific miRNAs that in turn modify target gene expression. As a result, we investigated whether miRNAs can affect PD-1 expression by rs10204525 allele specificity. To identify the putative miRNAs, miRNASNP-v4 databases were used. miR-3945942-3p, miR-4717-3p, miR-5589-3p, and miR-4802-3p were identified to bind rs10204525 based on its allele specificity (table 4).

Table 4. Identification of miRNAs targeting rs10204525 in the 3′ UTR of PD-1 by ΔMFE between allele C and T.

Putative miRNAs	Length	Seed region	ΔMFE (kcal/mol)
miR-942–3p	21	7	−12.8
miR-4717–3p	20	8	−26.70
miR-3115–5589 –3p	22	8	-20.60
miR-4802 –3p	22	7	-13.70

miRNA, microRNA; PD-1, programmed cell death-1; 3'-UTR', 3′ untranslated region; ΔMFE, Differential minimum-free energy of hybridization.

All four identified miRNAs recognized a seedregion of 7–8 nucleotides encompassing the critical locus rs10204525. However, miR-4717–3p demonstratedthe most favorable thermodynamic profile asindicated by the lowest ΔMFE. The latter suggestedthe strongest miRNA:mRNA affinity. As a result, miR-4717–3p was candidate for further investigation on its potential role in the modulation of PD-1 expression by rs10204525 allele specificity binding. As a result, miR-4717–3p was candidate for further investigation on its potential role in the modulation of PD-1 expression by rs10204525 allele specificity binding.

Modulation of PD-1 expression and induction by allele-specificity binding of miR-4717-3p to rs10204525 in human PBMCs

PBMCs carrying C/C (PBMCs^C/C) and C/T (PBMCs^C/T) in rs10204525 were isolated from blood sample collection, cultured, and transfected with (1) specific miR-4717–3 p mimic, (2) specific miR-4717–3 p inhibitor, and (3) TSB. Evaluation of miR-4717–3 p in PBMCs transfected with negative miRNA mimic, negative miRNA mimic plus negative TSB and negative inhibitor did not change the levels of miR-4717–3 p (online supplemental figure 1). As a result, negative miRNA mimic was selected as a negative control for further experiments. The validity of miRNA transfection was verified by quantifying miR-4717–3 p levels under basal conditions and following IFN-ɣ incubation (figure 2A). Specifically, miR-4717–3 p levels were increased by miR-4717–3 p mimic with or without TSB, while they were decreased by specific miR-4717–3 p inhibitor, regardless of rs10204525 genotype. IFN-γ incubation, used to mimic PBMC activation and to induce PD-1 upregulation,⁴¹ did not affect miR-4717–3 p levels in any of the transfected experimental conditions in both PBMCs^C/C and PBMCs^C/T. On the other hand, analysis of PD-1 expression, under basal conditions and following IFN-γ incubation, demonstrated that miR-4717–3 p modulation differentially affected PD-1 mRNA and protein levels in PBMCs^C/C and PBMCs^C/T (figure 2B–D, online supplemental figure 2). Specifically, under basal conditions, PD-1 expression was significantly higher in PBMCs^C/T transfected with negative miRNA mimic as compared with those of PBMCs^C/C. In addition, modulation of mir-4717–3 p by miR-4717–3 p mimic and inhibitor significantly reduced and increased, respectively, PD-1 levels in PBMCs^C/C, but not in PBMCs^C/T. Similarly, transfection with the TSB alone increased PD-1 expression in PBMCs^C/C, but not in PBMCs^C/T. Lastly, as expected, IFN-γ incubation increased PD-1 levels, regardless of rs10204525 genotype and/or miR-4717–3 p modulation. However, even in this case, IFN-γ-mediated PD-1 upregulation was increased to a greater extent in PBMCs^C/T as compared with that of PBMCs^C/C. These results validated the cause-effect relationship between the allele specificity of rs10204525 with miR-4717–3 p binding. The latter, in turn, negatively regulated expression and induction of PD-1 in immune cells only in the presence of C/C rs10204525 genotype, but not in C/T.

Figure 2. Modulation of PD-1 expression by miR-4717–3 p based on rs10204525 allele specificity in human PBMCs under basal conditions or following IFN-ɣ treatment. Both PBMCs^C/C and PBMCs^C/T were transfected with specific miR-4717–3 p mimic, inhibitor, TSB and a combination of miR-4717–3 p mimic with TSB for 48 hours. Negative miRNA mimic was used as a control. Both transfected PBMCs^C/C and PBMCs^C/T were seeded into 24-well plates at a density of 2×10⁵ cells per well and incubated with IFN-ɣ (100 ng/mL). Following a 24 hours incubation at 37°C in a 5% CO₂ atmosphere, expression levels of (A) miR-4717–3 p normalized to miR-103–3 p and (B) PD-1 mRNA normalized to GAPDH were evaluated by RT-PCR. The performances of each miR-4717–3 p mimic, inhibitor or TSB were compared with negative miRNA mimic and expressed as means±SD of the results obtained in three independent experiments (***p≤0.001). (C) Following a 48 hours incubation at 37°C in a 5% CO₂ atmosphere, PBMCs were harvested, lysed and analyzed by western blot with PD-1-specific Ab. GAPDH was used as a loading control. The performances of each miR-4717–3 p mimic, inhibitor or TSB were compared with negative miRNA mimic and plotted below their representative images as means±SD of the results obtained in three independent experiments (***p≤0.001). (D) Following a 48 hours incubation at 37°C in a 5% CO₂ atmosphere, PBMCs were harvested and cell surface stained with PE anti-PD-1 IgG1k Ab. PE anti-mouse IgG1k was used as a specificity control. Data, expressed as representative histogram profiles of MFI-PD-1, are the results obtained in three independent experiments. Ab, antibody; IFN, interferon; MFI, mean fluorescence intensity; mRNA, messenger RNA; PBMC, peripheral blood mononuclear cell; PD-1, programmed cell death-1; PE, phycoerythrin; RT-PCR, real-time PCR; TSB, target site blocker.

Modulation by allele-specificity binding of miR-4717–3 p-mediated PD-1 expression and induction on the in vitro reactivity of HLA-matched PBMCs to recognize and destroy non-cancer cells incubated with anti-PD-1 nivolumab

To assess the clinical relevance of the association between the development of irAEs and the modulation of expression/induction of PD-1 on immune cells mediated by allele-specificity binding of miR-4717–3 p to rs10204525, we co-cultured non-cancer cells (HaCaT cells) with miR-4717–3 p modulated HLA class I-matched PBMCs^C/C and PBMCs^C/T, under basal conditions and following incubation with the anti-PD-1 nivolumab (figure 3). Analysis of HLA class I profile revealed that HaCaT cells were homozygous for HLA-A*31:01 and heterozygous for HLA-B*40:01:02, 51:01:01 as well as for HLA-C*03:04:01, 15:02:01. PBMCs were isolated from blood sample collection of two recruited patients (Patient #1 and #2) and were selected based on their similar clinical-pathological characteristics, differential rs10204525 genotype, and HLA class I antigen matching to HaCaT cells (HLA-A*31:01 and HLA-C*03:04:01 for PBMC^C/C; and HLA-A*02:01 for PBMC^C/T). Noteworthy, both patients were affected by advanced NSCLC, were long responders to second-line anti-PD-1 therapy, one developed irAEs in presence of C/C genotype for rs10204525, while the other one, in presence of C/T genotype, did not (table 5).

Figure 3. Modulation by rs10204525 allele-specificity of in vitro activity of anti-PD-1 nivolumab on transfected PBMCs co-cultured with HLA-matched HaCaT cells. HaCaT cells were seeded into 24-well plates at a density of 2×10⁵ cells per well. Following a 48 hours incubation at 37°C in a 5% CO₂ atmosphere, cells were co-cultured with HLA-matched PBMCs^C/C and PBMCs^C/T isolated from Patient #1 and #2 transfected with miRNA specific constructs (miR-4717–3 p mimic, inhibitor and a combination of the mimic and TSB) in a 1:5 ratio and treated with 10 µg/mL of anti-PD-1 nivolumab and its human isotype IgG4 for 24 hours. Negative miRNA mimic transfected PBMCs^C/C and PBMCs^C/T were used as controls. (A) Following a 48 hours incubation after transfection, PBMCs were harvested and cell surface stained with PE anti-PD-1 IgG1k Ab. PE anti-mouse IgG1k was used as a specificity control. Data, expressed as MFI, are representative of the results obtained in three independent experiments. (B) HaCaT cell viability was determined by CCK-8 assay. Data are expressed as mean percentages of the viability of co-cultured HaCaT cells±SD as compared with HaCaT cells alone. The mean percentage of cell viability and SD were calculated from three independent experiments; each of them was performed in triplicate. (C) IFN-γ levels in the medium harvested from co-cultured cells were measured by an ELISA Max Deluxe Set Human IFN-γ kit. Data are expressed as a means of IFN-γ levels±SD of the results obtained in three independent experiments; each of them performed in triplicate. (D) The induction of apoptosis of HaCaT cells was determined by flow cytometry analysis of annexin V and PI staining. The levels of apoptosis are plotted and expressed as a mean fraction of annexin V+cells±SD of the results obtained in three independent experiments. *p<0.05. **p<0.01. ***p<0.001. ns indicates not statistically significant. Ab, antibody; CCK-8, cell counting kit-8; HaCaT, human epidermal keratinocyte; IFN, interferon; MFI, mean fluorescence intensity; PBMC, peripheral blood mononuclear cells; PD-1, programmed cell death-1; PE, phycoerythrin; PI, propidium iodide; TSB, target site blocker.

Table 5. Clinical-pathological characteristics of patients with NSCLC selected for PBMC-co-culturing experiments.

	Patient #1	Patient #2	Patient #3	Patient #4	Patient #5	Patient #6
rs10204525	C/C	C/T	C/C	C/T	C/C	C/T
HLA-A	A02:01/A31:01	A02:01/A11:01	A24:02/A31:01	A02:01/A31:01	A24:01/A03:01	A01:01/A02:01
HLA-B	B07:02/B08:01	B15:01/B40:01	B08:01/B51:01	B15:01/B40:01	B07:02/B51:01	B08:01/B15:01
HLA-C	C03:04/C07:02	C03:04/C15:02	C07:02/C15:02	C03:04/C07:01	C07:01/C15:02	C03:04/C07:01
ECOG PS	0	0	1	0	0	0
Smoking status	People who have quit smoking	People who have quit smoking	People who smoke	People who have quit smoking	People who have quit smoking	People who smoke
Comorbidities	Hypertension	Hypertension	None	Hypertension	Hypertension Dyslipidemia	None
Histology	Adenocarcinoma	Adenocarcinoma	Adenocarcinoma	Adenocarcinoma	Adenocarcinoma	Adenocarcinoma
Asymptomatic brain metastases	No	No	No	No	No	No
Previous PBCT	Yes	Yes	Yes	Yes	No	No
Previous targeted therapy	No	No	No	No	No	No
Type of ICI-based immunotherapy	Nivolumab	Nivolumab	Nivolumab	Nivolumab	PBCT and nivolumab plus ipilimumab	PBCT and nivolumab plus ipilimumab
Number of ICI-based immunotherapy cycles	45	40	50	38	3	14
Best response from ICI-based immunotherapy	PR	PR	PR	PR	PR	CR
Grade 1–2 irAEs	Asthenia Creatinine increase Lipase increase	None	Asthenia Nausea Thyroiditis	Asthenia	Alanine aminotransferase increase Aspartate aminotransferase increase Amylase increase Lipase increase Oral mucositis	Thyroiditis
Grade 3–4 irAEs	None	None	Rash	None	*Alanine aminotransferase increase *Aspartate aminotransferase increase	None

^*Led to discontinuation of ICI-based immunotherapy.

CR, complete response; ECOG PS, Eastern Cooperative Oncology Group performance status; HLA, human leukocyte antigen; ICI, immune checkpoint inhibitor; irAE, immune-related adverse event; NSCLC, non-small cell lung cancer; PBCT, platinum-based chemotherapy; PBMC, peripheral blood mononuclear cell; PR, partial response.

Quantification of PD-1 surface protein expression on PBMCs across all experimental conditions was used to validate the efficiency of miR-4717–3 p modulation by transfection (figure 3A). Consistent with previous findings, modulation of miR-4717–3 p affected PD-1 expression on PBMCs^C/C. Indeed, a significant decrease and increase were detected in PD-1 surface expression on transfection with miR-4717–3 p mimic and miR-4717–3 p inhibitor/TSB, respectively. In contrast, in PBMCs^C/T, transfection with miR-4717–3 p mimic, miR-4717–3 p inhibitor or the combination of miR-4717–3 p mimic and TSB did not affect PD-1 expression. Cell growth inhibition and apoptosis induction of HaCaT cells as well as IFN-γ release by PBMCs were increased to a greater extent when HaCaT cells were co-cultured with PBMCs^C/C as compared with HaCaT cells co-cultured with PBMCs^C/T (figure 3B–D). In addition, cell growth inhibition and apoptosis induction of HaCaT cells as well as IFN-γ release by PBMCs^C/C were dramatically decreased and not significantly affected, respectively, by miR-4717–3 p mimic and miR-4717–3 p inhibitor transfection of PBMCs^C/C co-cultured with HaCaT cells as compared with those by negative miRNA mimic transfected PBMCs^C/C (figure 3B–D). Similar results to miR-4717–3 p inhibitor transfection were obtained when PBMCs^C/C were transfected with miR-4717–3 p mimic and TSB. In contrast, cell growth inhibition and apoptosis induction of HaCaT cells as well as IFN-γ release by PBMCs^C/T were not changed when HaCaT cells were co-cultured with modulated miR-4717–3 p PBMCs^C/T as compared with that of HaCaT cells co-cultured with negative miRNA mimic transfected PBMCs^C/T (figure 3B–D). Finally, treatment with anti-PD-1 nivolumab significantly and differentially affected PBMC-mediated recognition and destruction of HaCaT cells based on rs10204525 genotype. Specifically, nivolumab treatment significantly increased cell growth inhibition and apoptosis induction of HaCaT cells as well as IFN-γ release by negative miRNA mimic-transfected PBMC^C/C as compared with cells incubated with isotype control. Additionally, nivolumab treatment synergistically increased the significant increase of cell growth inhibition and apoptosis induction of HaCaT cells as well as IFN-γ release by miR-4717–3 p mimic transfected PBMC^C/C as compared with control groups (figure 3B–D). In contrast, in all experimental conditions involving the modulated transfection of miR-4717–3 p in PBMCs^C/T co-cultured with HaCaT cells, no significant changes, even following nivolumab incubation, were detected. Nivolumab treatment slightly decreased the viability of HaCaT cells (figure 3B) as well as slightly increased PBMC-mediated IFN-γ release (figure 3C) and apoptosis induction of HaCaT cells (figure 3D) as compared with HaCaT cells treated with IgG4 isotype control, co-cultured with all types of transfected PBMCs^C/T.

Validation of rs10204525 genotype-mediated reactivity of HLA-matched PBMCs to recognize and destroy non-cancer cells incubated with anti-PD-1 mAb alone or in combination with anti-CTLA-4 mAb

Lastly, to confirm the obtained clinical results as well as the in vitro results, we co-cultured BEAS-2B cells as well as HaCaT cells with additional HLA class I-matched PBMCs^C/C and PBMCs^C/T, under basal conditions and following incubation with nivolumab (figure 4) or with nivolumab in combination with ipilimumab (figure 5). BEAS-2B were homozygous for HLA-A02:01 and HLA-C07:01, heterozygous for HLA-B*07:02, 15:01. PBMCs were isolated from blood sample collection of four additional recruited patients (Patient #3, #4, #5 and #6; table 5). HLA were at least one allele matched. All patients were affected by advanced NSCLC and achieved clinical benefit from anti-PD-1-based immunotherapy. Patient #3 and #4 received nivolumab as second line, while patients #5 and #6 received the combination of PBCT and nivolumab plus ipilimumab as first line. All patients carrying a C/C genotype developed grade 3–4 irAEs, while those carrying a C/T did not. In addition, those carrying a C/C genotype developed more grade 1–2 irAEs as compared with those carrying a C/T genotype. Noteworthy, grade 3–4 irAEs led to discontinuation of ICI-based immunotherapy in patient #5 (table 5). Analysis of cell growth inhibition and apoptosis induction in non-cancer cells co-cultured with PBMCs carrying different rs10204525 as well as IFN-γ release by PBMCs demonstrated that PBMCs^C/C exhibited significantly enhanced reactivity to non-cancer cells in response to anti-PD-1 alone (figure 4) or in combination with anti-CTLA-4 (figure 5) as compared with PBMCs^C/T. Treatment with anti-CTLA-4 increased the anti-PD-1 mediated enhanced susceptibility of non-cancer cells to PBMCs, regardless of rs10204525 genotype. However, this effect was more marked in the presence of C/C genotype as compared with C/T genotype.

Figure 4. Differential immune response by rs10204525 genotype in PBMCs from patients with NSCLC co-cultured with non-cancer cells on anti-PD-1 treatment. HaCaT and BEAS-2B cells were seeded into 24-well plates at a density of 2×10⁵ cells per well. Following a 24 hours incubation at 37°C in a 5% CO₂ atmosphere, cells were co-cultured with HLA-matched PBMCs^C/C and PBMCs^C/T isolated from Patients #1–4 in a 1:5 ratio and treated with 10 µg/mL of anti-PD-1 nivolumab and its human isotype IgG4 for 48 hours. Monocultured HaCaT and BEAS-2B cells were used as controls. (A) Following a 48 hours incubation, non-cancer viability was determined by CCK-8 assay. Data are expressed as mean percentages of the viability of co-cultured HaCaT and BEAS-2B cells±SD as compared with HaCaT or BEAS-2B cells alone. The mean percentage of cell viability and SD were calculated from three independent experiments; each of them was performed in triplicate. (B) IFN-γ levels in the medium harvested from co-cultured cells were measured by an ELISA Max Deluxe Set Human IFN-γ kit. Data are expressed as means of IFN-γ levels±SD of the results obtained in three independent experiments; each of them performed in triplicate. (C) The induction of apoptosis of HaCaT and BEAS-2B cells was determined by flow cytometry analysis of annexin V and PI staining. The levels of apoptosis are plotted and expressed as a mean fraction of annexin V+cells±SD of the results obtained in three independent experiments. *p<0.05. **p<0.01. ***p<0.001. ns indicates not statistically significant. BEAS-2B, bronchial epithelial; CCK-8, cell counting kit-8; HaCaT, human epidermal keratinocyte; IFN, interferon; NSCLC, non-small cell lung cancer; PBMC, peripheral blood mononuclear cells; PD-1, programmed cell death-1; PI, propidium iodide.

Figure 5. Enhanced immune response in rs10204525 C/C PBMCs isolated from patients with NSCLC on anti-PD-1 or combined anti-PD-1/CTLA-4 treatment in co-culture with non-cancer cells. HaCaT and BEAS-2B cells were seeded into 24-well plates at a density of 2×10⁵ cells per well. Following a 24 hours incubation at 37°C in a 5% CO₂ atmosphere, cells were co-cultured with HLA-matched PBMCs^C/C and PBMCs^C/T isolated from Patients #5 and #6 in a 1:5 ratio and treated with 10 µg/mL of anti-PD-1 nivolumab alone or in combination with 3.3 µg/mL of anti-CTLA-4 ipilimumab for 48 hours. Human IgG was used as an isotype control. Monocultured HaCaT and BEAS-2B cells were used as controls. (A) Following a 48 hours incubation, non-cancer cell viability was determined by CCK-8 assay. Data are expressed as mean percentages of the viability of co-cultured HaCaT and BEAS-2B cells±SD as compared with HaCaT or BEAS-2B cells alone. The mean percentage of cell viability and SD were calculated from three independent experiments; each of them was performed in triplicate. (B) IFN-γ levels in the medium harvested from co-cultured cells were measured by an ELISA Max Deluxe Set Human IFN-γ kit. Data are expressed as means of IFN-γ levels±SD of the results obtained in three independent experiments; each of them performed in triplicate. (C) The induction of apoptosis of HaCaT and BEAS-2B cells was determined by flow cytometry analysis of annexin V and PI staining. The levels of apoptosis are plotted and expressed as a mean fraction of annexin V+cells±SD of the results obtained in three independent experiments. *p<0.05. **p<0.01. ***p<0.001. ns indicates not statistically significant. CCK-8, cell counting kit-8; CTLA-4, cytotoxic T-lymphocyte antigen 4; HaCaT, human epidermal keratinocyte; IFN, interferon; NSCLC, non-small cell lung cancer; PBMC, peripheral blood mononuclear cells; PD-1, programmed cell death-1; PI, propidium iodide.

Discussion

Anti-PD-1/PD-L1-based immunotherapy has revolutionized the therapeutic landscape of many types of cancers, improving the survival and quality of life of treated patients as compared with conventional therapies.1,11 Although administration of anti-PD-1/PD-L1 mAbs is generally well-tolerated, some of the treated patients develop severe irAEs, potentially causing permanent sequelae or even fatal consequences.1,11 Incidence of severe irAEs depends on type of ICI-based immunotherapy, ranging from 10% to 70%. Our study included two cohorts of patients: first, patients with advanced cancer were treated with anti-PD-1/PD-L1 as a single agent; second, patients with advanced NSCLC were treated with anti-PD-1 in combination with platinum-based chemotherapy with or without ipilimumab. The representativeness and validity of the patient populations included are corroborated by the rate of irAEs and treatment discontinuation obtained as well as by the obtained clinical outcomes in the different tumor types we analyzed (ORR and survival outcomes), being in line with data reported in the literature for anti-PD-1/PD-L1-based immunotherapy.1,14 To date, several biomarkers have been investigated to predict clinical benefit from anti-PD-1/PD-L1 mAbs. However, few studies have focused on the identification of biomarkers to identify patients who are more likely to develop anti-PD-1/PD-L1-mediated irAEs. So far, none of the investigated biomarkers demonstrated to be enough efficient for this type of prediction. Here, we validated the role of a PD-1 SNP that efficiently predicts both grade 1–2 and 3–4 irAEs from anti-PD-1/PD-L1-based immunotherapy in patients with advanced cancer. In the past few years, investigation on the predictive role to irAEs of PD-1 SNPs has produced contrasting results.42,46 Our data demonstrated that not all PD-1 SNPs have clinical significance for predicting irAEs since only rs10204525 correlated with the occurrence of irAEs. Specifically, both grade 1–2 and grade 3–4 irAE rates were significantly higher in patients carrying C/C genotype in rs10204525 as compared with those carrying C/T genotype in both patients with cancer treated with anti-PD-1/PD-L1 alone and anti-PD-1/PD-L1 in combination with chemotherapy with or without anti-CTLA-4 therapy. In line with our results, Kobayashi et al demonstrated that the allele G in rs10204525 is associated with a higher risk to develop severe and multiple anti-PD-1-mediated irAEs.⁴⁶ In contrast, Bins et al as well as Refae et al showed no correlation between rs10204525 genotype and development of anti-PD-1-mediated irAEs.^{42 44 45} In the patient populations we analyzed, based on patient availability, we included in the first cohort Caucasian patients with advanced cancer, including those affected by HNSCC, NSCLC, RCC and melanoma, treated with anti-PD-1/PD-L1 as a single agent, in both first and second line of treatment. Differential types of tumors analyzed as well as the differential ethnicity and clinical-pathological characteristics of patient with cancer population included might justify the reported contrasting results. However, in our case, the results obtained from a general cancer population were validated by a second cohort of patients with advanced NSCLC who received anti-PD-1 therapy in combination with chemotherapy with or without anti-CTLA-4 therapy, all as first line, according to clinical practice. To the best of our knowledge, no study has investigated the association between rs10204525 and development of irAEs in patients with cancer treated with anti-PD1/PD-L1 in combination with other oncological treatments. Noteworthy, in the patient populations we analyzed, three (in the first cohort) and five (in the second cohort) out of the patients discontinued anti-PD-1/PD-L1 therapy because of the development of severe irAEs. All these patients carried C/C genotype in rs10204525. Lastly, the predictive role of rs10204525 to irAEs is further supported by its functional significance in autoimmunity, cancer predisposition and cancer prognosis.26,31

Contrasting results have also been reported on the association between the development of irAEs and clinical response to ICIs in patients with solid tumors.47,49 In our case, development of irAEs did not correlate with clinical survival outcomes. Similarly, none of the PD-1 SNPs analyzed correlated with clinical survival outcomes. In this case, different types of tumors, lines of treatments and low number of patients analyzed limit our results. Further studies are needed to clarify whether PD-1 SNPs and irAEs might correlate with OS of tumor-specific patients with cancer treated with ICIs.

The clinical relevance of the association between rs10204525 and development of irAEs is supported by the molecular mechanisms we show for the first time to underline the predictive value of rs10204525. To the best of our knowledge, no study has demonstrated that rs10204525 predicts the development of irAEs by mediating a differential binding of miR-4717–3 p to PD-1, regulating its expression and induction. The cause-effect relationship between miR-4717–3 p and rs10204525 is demonstrated by the differential modulation of miR-4717–3 p in PBMCs obtained from patients carrying differential rs10204525 genotypes. As a result, in PBMCs carrying a C/C genotype, miR-4717–3 p binds to 3’-UTR of PD-1 and causes PD-1 downregulation, both under basal conditions and following IFN-γ incubation. The latter was used to induce PD-1 expression on PBMCs.⁴¹ In contrast, in PBMCs carrying a C/T genotype, lack of binding of miR-4717–3 p to PD-1 does not affect PD-1 expression and induction. As a result, PBMCs^C/T display higher levels of PD-1 expression as a mechanism of regulation of the host immune response.⁵⁰

Clinical translation of the differential development of irAEs as well as of the modulation of PD-1 expression and induction by rs10204525 allele-specificity binding of miR-4717–3 p is obtained by co-culturing non-cancer cells with patient-derived HLA-matched PBMCs carrying differential rs10204525 genotypes. Normal keratinocytes and bronchial epithelial cells were used as non-cancer cell models. These experimental conditions, by testing the ability of anti-PD-1 therapy as well as anti-PD-1 and anti-CTLA-4 in combination to unleash the reactivity of PBMCs carrying differential rs10204525 genotype to recognize and destroy non-cancer cells, are expected to mimic the development of an irAE. PBMCs carrying a C/C genotype in rs10204525 display a decreased miR-4717–3 p-mediated PD-1 expression and induction, that in turn is associated with an increased recognition and destruction of non-cancer cells, even more in the presence of anti-PD-1-based immunotherapy. In contrast, the increased expression of PD-1 in PBMCs carrying a C/T genotype in rs10204525 not binding miR-4717–3 p is associated with a reduced reactivity of PBMCs to recognize and destroy non-cancer cells. As a result, rs10204525 has an impact on PD-1 mediated PBMC immunoreactivity (figure 6).

Figure 6. Graphical representation of the influence of rs10204525 genotype on PD-1 expression and increased sensitivity of HaCaT cells to anti-PD-1 therapy. 3’-UTR, 3′ untranslated region; GzmB, Granzyme B; HaCaT, human epidermal keratinocyte; IFN, interferon; PBMC, peripheral blood mononuclear cells; PD-1, programmed cell death-1; TNF, tumor necrosis factor.

Worth noting, PBMCs carrying differential rs10204525 genotypes used in co-culture experiments were isolated from patients with advanced NSCLC, all of them achieving clinical benefit from anti-PD-1 therapy alone or in combination with chemotherapy and anti-CTLA-4. Those carrying the C/C genotype developed more and severe irAEs as compared with those carrying the C/T genotype. Selection of patients with similar clinical-pathological characteristics and their HLA class I antigen matching to used non-cancer cells ensures the physiological relevance and robustness of the co-culture models, allowing for the replication of immune interactions in a clinically relevant HLA context. These results have high clinical relevance since they underscore the importance of genetic factors in determining the safety of anti-PD-1/PD-L1-based immunotherapy, paving the way for a more personalized, risk-based approach to monitor and managing the expected development of irAEs. In addition, a reduced sensitivity to develop irAEs observed in patients carrying the C/T genotype is likely to suggest a more safe potential implementation of more intensified or novel combinatorial strategies of ICI-based immunotherapies within this subgroup.

Although we provided a validation cohort, our study presents some limitations, including the small sample size of study populations and the heterogeneity of types of tumors in the first cohort. As a result, the impact of rs10204525 genotype on the development of irAEs in anti-PD-1/PD-L1-based immunotherapy should be further validated in prospective larger studies. Moreover, in our study, most patients received either anti-PD-1 or anti-PD-L1 therapy, while only a small proportion of patients (in the second cohort) received the combination of platinum-based chemotherapy and anti-PD-1 plus ipilimumab (anti-CTLA-4). Some lines of evidence suggested a different irAE profile for anti-PD-1 and anti-PD-L1 therapy.^{51 52} Nevertheless, in our case, no significant difference was detected in anti-PD-1-mediated and anti-PD-L1-mediated irAEs. On the other hand, in line with data from the literature,812,14 a higher incidence and grade of irAEs were reported in combinatorial treatments of anti-PD-1 therapy as compared with anti-PD-1/PD-L1 alone. We did not have the power to detect differences in the development of irAEs between patients treated with anti-PD-1 plus chemotherapy and anti-PD-1 plus chemotherapy and anti-CTLA-4. Although anti-CTLA-4 increased the susceptibility of non-cancer cells to the anti-PD-1-mediated PBMC activity, regardless of rs10204525 genotypes, we show that this effect is more marked in the presence of C/C genotype as compared with C/T genotype. Further studies are needed to evaluate whether rs10204525 genotype can have a differential impact on development of irAEs based on the combination of anti-PD-1/PD-L1 with other types of oncological treatments including ICIs, chemotherapy and targeted therapy. Lastly, further investigation is needed to investigate the potential molecular mechanisms that might underline clinical significance of other PD-1 SNPs.

Overall, our study has high clinical relevance since it identifies rs10204525 and miR-4717–3 p-mediated modulation of PD-1 expression and induction as a novel biomarker to predict patients with cancer at a higher likelihood to develop irAEs as well as a mechanism of immune cell reactivity. This novel biomarker may be useful to improve the management of patients with cancer treated with anti-PD-1/PD-L1 therapy.

Data availability statement

Data are available upon reasonable request.