Materials and methods

We follow the REMARK guideline for the reporting of tumor marker prognostic studies [13].

Patients

This prospective biomarker study was conducted at the Department of Oncology, Lillebaelt Hospital, University Hospital of Southern Denmark, Vejle, Denmark.

All patients were prospectively included and gave oral and written consent to participate. The study conforms to the Declaration of Helsinki and Danish data protection legislation. The study protocol was approved by the Regional Health Research Ethics Committee of Southern Denmark (ID-number S-20180123).

Patients with primary NSCLC with or without locoregional lymph-node involvement planned for curative RT were offered enrollment in the study (corresponding to TNM 8th edition stage Ia–IIIc) [14].

Inclusion criteria

Stereotactic or conventional RT for primary NSCLC with curative intent and concurrent chemotherapy if indicated, age ≥ 18 years, WHO performance status 0–2, written and orally informed consent, consent to translational research and biobank.

Exclusion criteria

Distant metastases, severe comorbidity making the patient unlikely to complete the planned 2-year follow-up period, other malignant disease within 5 years prior to study enrollment, except basocellular or squamous skin cancer and carcinoma in situ cervicis uteri.

To increase the number of patients in the analyses, we included ten comparable patients from another cohort, which had blood sampled twice during their diagnostic work-up and curative RT treatment. These blood samples were collected between June 2022 and December 2023. These ten patients were all diagnosed with NSCLC, and six of these patients received SBRT, while four patients received long-course RT.

Permission to use the blood sampled from these patients for this study was obtained as an amendment to the original protocol accepted by the Research Ethics Committee of Southern Denmark (S-20180123). This other cohort did not include NKVue blood sampling, so the added ten patients could not contribute data to NKA analyses.

Treatment

All patients received standard treatment according to Danish guidelines [2, 15].

The RT was delivered either as SBRT with 3–5 fractions (n = 33) or as long-course chemoradiotherapy with 24–33 fractions (66 Gy in 33 fractions (n = 26), 66–95 Gy in 33 fractions inhomogeneously escalated according to the NARLAL2 trial (n = 3) [16], or 50–66 Gy in 24 fractions inhomogeneously escalated according to the HERAN trial (n = 6) [17]).

Follow-up

The first follow-up visit was made five weeks after the completion of RT. Subsequent follow-up visits were planned with 3-month intervals for the first two years and then 6-month intervals for the next three years for a total of 5-year follow-up, as recommended by Danish guidelines [2]. A CT scan of the thorax and abdomen preceded each follow-up visit. In this paper, we only provide information on 12 months of follow-up, as many patients have not yet completed 2-year follow-up.

Blood sampling and storage

Blood sampling was performed before the start of radiation treatment, at each cycle of chemotherapy, at the end of radiation treatment, and at each follow-up visit for up to 2 years after treatment, or until relapse. At each blood sampling, three 9 mL venous blood samples were collected in EDTA tubes for plasma, and a 9 mL sample was collected in a clot activator tube for serum. The blood collection tubes were centrifuged at 2000 g for 10 min within 4 h and stored at − 80 °C until analysis. Storage time ranged from a few months to 4 years. One milliliter of venous blood was also sampled into a Promoca™-containing NK Vue® tube for analysis of NKA and placed in an incubator at 37 °C within 15 min of the blood sampling. After 20–24 h of incubation, the plasma was collected and stored at − 80 °C for a maximum of 54 days before analysis.

Analysis of methylated ctDNA

Four milliliters of plasma was thawed at room temperature and centrifuged at 10,000 g for 10 min. ~ 9000 copies/mL of an exogenous spike-in DNA fragment (CPP1) were added to the plasma to evaluate DNA extraction efficiency [18]. DNA was extracted using the DSP Circulating DNA kit (Qiagen, Hilden, Germany) on the QiaSymphony SP instrument (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. Extraction efficiency was determined with a ddPCR assay targeting CPP1 and calculating the percentage of recovered CPP1 copies/mL. Total cfDNA concentration and contamination with high-molecular-weight DNA were determined with ddPCR assays amplifying a 65 bp and 250 bp region of the EMC7 gene [19]. Potential contamination with lymphocyte DNA was evaluated with an immunoglobulin-specific ddPCR assay [18]. The extracted DNA was concentrated to 20 µL with an Amicon Ultra-0.5 centrifugal filter unit (Merck, Darmstadt, Germany) and bisulfite converted using the EZ DNA methylation lightning kit (Zymo Research, Irvine, CA, USA) according to the manufacturer’s instructions. Water was used as a non-template control, leukocyte DNA from healthy donors (20 µL corresponding to ~ 20 ng) as a non-cancer control, and Universal Methylated DNA Standard (Zymo Research, Irvine, CA, USA) as a positive control. All the controls were analyzed in parallel with the patient samples, and all analyses were blinded to the clinical endpoints.

The analyses of ctDNA methylation were performed in duplicate using the Bio-Rad QX600 Droplet Digital PCR Systems (Bio-Rad, Hercules, CA, USA). Droplets were generated using an Automated Droplet Generator (Bio-Rad). The PCR was performed on a Veriti™ Thermal Cycler (Applied Biosystems, Waltham, MA, USA) with the following PCR conditions: 95 °C for 10 min, 44 cycles of 95 °C for 15 s and 56 °C for 1 min, and 98 °C for 10 min using 2 °C/s ramp rate. All data were analyzed with QXManager Software 2.0 (Bio-Rad). Four methylated loci near the genes HOXA9, MCIDAS, TFAP2B, and SP9 were included in the multiplex. These methylated loci were selected based on previous studies (HOXA9) and data from several databases on candidate tumor-agnostic lung cancer methylated areas. These databases include The Cancer Genome Atlas Program (TCGA) and Gene Expression Omnibus (GEO) [20]. The selection process and validation of candidate targets is described in a recent paper from our group [21]. The analyses were developed and validated in-house in our lab.

The four methylated ctDNA markers were previously tested in a cohort of 40 patients from the lung clinic at Vejle Hospital, where patients with benign lung disease, such as chronic obstructive pulmonary disease or asthma, are treated. These 40 patients were not known to have any malignant disease at the time of blood sampling. The limit of blank (LoB) for each marker was established following the measurements from these 40 non-cancerous subjects, and using a set ≥ 95% specificity, the LoB was set at two positive droplets for the HOXA9, MCIDAS, and TFAP2B markers and at three positive droplets for the SP9 marker. In addition, a sample must have at least two markers above LoB to be classified as positive.

Analysis of soluble PD-L1

sPD-L1 was analyzed in all samples using the Human/Cynomolgus Monkey PD-L1/B7-H1 Quantikine ELISA kit in a 1:2 dilution (R&D Systems, Minneapolis, Minnesota, USA) according to the manufacturer’s recommendations. As controls, four serum samples were analyzed in duplicates. The in-house intra-assay variation was < 13%, and the inter-assay variation was < 11%. The analyses were blinded to the clinical endpoints.

Analysis of NK-cell activity

NK-cell activity was measured utilizing the level of IFN-γ as a surrogate with the NK Vue® Kit (NKMAX, Seongnam-si, South Korea). This technique has previously been reported by our group [11, 12]. Frozen plasma samples were thawed at room temperature and centrifuged at 11,500 g for 1 min before being analyzed using the NK Vue® enzyme-linked immunosorbent assay (ELISA) (NKMAX, Seongnam-si, South Korea). The manufacturer’s instructions were followed in all steps of the procedure. In-house measurements of the ELISA showed an intra-assay coefficient of variation < 10% and an inter-assay coefficient of variation < 14%. A level of < 250 pg/mL IFN-γ was considered abnormal as recommended by the manufacturer. The analyses were blinded to the clinical endpoints.

Statistics

Continuous variables were reported as means with standard deviation, if normally distributed (tested with visual inspection of QQ-plots), and as median with range and/or interquartile range (IQR) if non-normally distributed.

Associations between biomarker values (positive/negative) and 12-month relapse were calculated with Chi-square statistics, or with Fisher’s exact test if any observed frequency was below N = 5. Relapse was established with a biopsy if possible or otherwise defined as radiological progression as per RECIST 1.1 [22]. In addition to using the established LoB for the combined four ctDNA markers, we used Receiver-Operating Characteristic (ROC) curves to evaluate the discriminative capabilities of the different biomarkers measured at baseline in detecting patients who had a relapse within 12 months. Area under the curve (AUC) values were reported for the individual markers and their combinations. Logistic regression models were used to estimate the probability of relapse as investigated in the multivariate ROC analyses.

Final data analysis was performed using STATA ver. 18 (StataCorp, College Station, TX, USA).