1. Introduction
Lung cancer is a significant cause of cancer-related deaths worldwide, and non-small-cell lung cancer (NSCLC) accounts for approximately 85% of all lung cancer cases [
1]. Nowadays, programmed cell death protein 1 (PD-1)/programmed death-ligand 1 (PD-L1) immune checkpoint inhibitors (ICI), which unleash T lymphocyte-mediated immune responses, have become a breakthrough therapy for lung cancer. The CheckMate 057 trial is a clinical trial that evaluated the efficacy and safety of nivolumab, a PD-1 programmed death-ligand inhibitor, compared to docetaxel in patients with previously treated advanced NSCLC. The trial showed that nivolumab improved overall survival and had a better safety profile than docetaxel, leading to its approval by the FDA as a treatment option for patients with advanced NSCLC who had previously received platinum-based chemotherapy.
Currently, immunohistochemical-based ((IHC)-based) PD-L1 expression is the only predictive marker approved for ICIs, reflecting the immune characteristics of the tumor to escape immune surveillance. According to research on the early survival of patients in the CheckMate 057 trial, patients with poor prognostic factors combined with no or low PD-L1 expression were at a higher risk of death within the first 3 months of nivolumab treatment than with docetaxel [
2]. In the same context, in Korea, reimbursement for nivolumab is limited to patients with PD-L1 expression levels ≥ 10% and who have progressed to platinum-based chemotherapy due to its better hazard ratio among subgroups with PD-L1 expression levels ≥ 10%. However, PD-L1 expression alone may not comprehensively reflect the complexity of the tumor microenvironment (TME), and other immunosuppressive mechanisms within the TME may attenuate ICI response. Collectively, these mechanisms contribute to the low mutational burden, poor intrinsic antigenicity of tumor cells, absence of priming, defective antigen presentation during the priming phase, and functional exhaustion of tumor-infiltrating lymphocytes by suppressive immune regulatory cells [
3].
Different strategies have been explored to counteract immune evasion by shifting the balance in favor of antitumor immune activation, and combination cytotoxic chemotherapies have been used as potent immune modulators [
4]. So far, several published studies have demonstrated the beneficial effect of low-dose cyclophosphamide in amplifying the immune response against tumors by reducing regulatory T cells (Tregs) [
5]. Doxorubicin is considered a good immunomodulator as it induces CD8
+ IFN-γ
+ T cell responses and significantly upregulates T cell activation markers on the surface of CD4
+ T cells. In addition, it promotes cytokine secretion of interleukin (IL)-1, IL-2, and tumor necrosis factor α (TNF-α), enhancing the immune shift from Th2 to Th1 [
6,
7]. Furthermore, doxorubicin selectively impairs myeloid-derived-suppressor-cell (MDSC)-induced immunosuppression [
8]. These two drugs are conventional cytotoxic drugs with good accessibility and cost benefits. Herein, we comprehensively assessed the impact of cyclophosphamide and Adriamycin (CA) induction therapy on the antitumor effects of nivolumab in advanced NSCLC with PD-L1 < 10%.
2. Materials and Methods
2.1. Study Design
This was a single-center, single-arm, phase II study. The patients received 4 cycles of cyclophosphamide (500 mg/m2) and adriamycin (50 mg/m2) induction therapy every 3 weeks. Nivolumab 360 mg was started from the second cycle of the induction phase. After induction therapy, nivolumab 480 mg was administered every four weeks until disease progression, unacceptable toxicity, or for a maximum of two years. The study was approved by the research ethics board (IRB No. NCC2018-0267) and conducted in accordance with the ethical principles of the Declaration of Helsinki. This study was registered at clinicaltrials.gov (NCT03808480), and written informed consent was obtained from all the patients.
2.2. Participants
To be eligible for the study, all patients aged 18 years or over were required to have histological or cytological proof of advanced non-squamous NSCLC. Patients with measurable lesions were eligible for inclusion. Each patient was required to have wild-type epidermal growth factor receptor (EGFR), anaplastic lymphoma kinase (ALK), a PD-L1 expression less than 10%, and a performance status of 1 or lower on the Eastern Cooperation Oncology Group (ECOG) scale with adequate organ function. All the patients previously received platinum-based chemotherapy. These patients should have received fewer than three chemotherapy regimens. The main exclusion criteria were a history of malignancies other than lung cancer in the past three years, previous ICI therapy, a history or current diagnosis of interstitial lung disease, and any unstable systemic disease.
2.3. Assessment of Response
The patients underwent chest and abdominal computed tomography (CT), brain magnetic resonance imaging (preferred), or CT and positron emission tomography at baseline. Assessments were performed every 2 cycles of therapy and every 2 months thereafter during the long-term follow-up. Using these imaging techniques, the clinical response to nivolumab was assessed in accordance with the Response evaluation criteria in solid tumors version 1.1. The following terms were used to describe the manner in which a tumor responded to treatment: complete response (CR; disappearance of all target lesions), partial response (PR; ≥30.0% reduction in the sum of the diameters of the target lesions), progressive disease (PD; ≥20.0% increase in the sum of the diameters of the target lesions), and stable disease (SD; insufficient to qualify as PR or PD). The objective response rate (ORR) was the proportion of patients who achieved CR or PR, and the disease control rate (DCR) was the proportion of patients who achieved CR, PR, or SD. Progression-free survival (PFS) was defined as the time from assignment to the study until the date of disease progression or death from any cause, whichever occurred first. Overall survival (OS) was defined as the time from the start of the study to the date of death or last follow-up for patients who were alive at the end of the study. The data of patients alive at the last follow-up visit were censored. Adverse events (AEs) were summarized according to the National Cancer Institute Common Terminology Criteria for Adverse Events (version 5.0).
2.4. Immunohistochemistry of PD-L1 Expression
PD-L1 expression was determined using a Ventana PD-L1 SP263 antibody (Ventana Medical Systems, Tucson, AZ, USA). PD-L1 expression levels in tumor cells were determined by the percentage of stained cells in each slide, which was estimated in increments of 5%, except for a 1% positivity value. Patients in whom at least 1% of the tumor cells stained positive for PD-L1 were determined positive.
2.5. Blood Next-Generation Sequencing Analysis (NGS) by GuardantOMNI
DNA extraction and next-generation sequencing were subsequently performed at Guardant Health using the Guardant OMNI (Redwood City, CA, USA) 2.15 Mb, 500 gene panel) to identify SNVs, indels, gene fusions, copy-number variants, microsatellite status, and tumor mutation burden (TMB) [
9]. The GuardantOMNI algorithm was used to report plasma TMB. This algorithm includes all somatic synonymous and nonsynonymous single-nucleotide variants (SNVs) and indels, excluding germline, clonal hematopoiesis of indeterminate potential, driver, and resistance variants, with statistical adjustment for sample-specific tumor shedding of ctDNA and coverage [
10]. pTMB-unevaluable samples are those with limited tumor shedding, including all somatic mutations <0.3% of the maximum somatic allele fraction, or low unique molecule coverage.
2.6. Proteomics
2.6.1. Sample Preparation for Mass Spectrometry
Plasma samples were depleted using High SelectTM top14 abundant protein depletion spin columns (Thermo Fisher Scientific, Rockford, IL, USA) and digested using S-Trap™ spin columns (Protifi, Huntington, NY, USA) according to the manufacturer’s instructions. Desalted peptides were labeled using the TMT reagent (Thermo Fisher Scientific), and 24 noncontinuous peptide fractions were separated using an Agilent 1260 Infinity HPLC system (Agilent, Palo Alto, CA, USA). Five percent of each fraction was aliquoted for global proteome analysis, and the remaining 95% was combined into 12 fractions for phosphoproteome analysis. Enriched phosphopeptides were obtained by combining 12 fractions with Ni-NTA agarose beads (Qiagen, Valencia, CA, USA) and converting them to Fe3+-NTA beads.
2.6.2. LC-MS/MS Analysis
The peptides prepared for global/phosphoproteome analysis were resuspended in 0.1% formic acid in water and analyzed using a FAIMS Pro interface (Thermo Fisher Scientific) mounted on an Orbitrap Eclipse Tribrid mass spectrometer (Thermo Scientific) equipped with an Ultimate 3000 RSLCnano system (Thermo Scientific). Solvents A and B contained 0.1% formic acid (FA) in water and 0.1% FA in acetonitrile, respectively. Peptides were loaded onto a trap column (Acclaum PepMapTM 100, 75 mm × 2 cm) and separated on an analytical column (EASY-Spray column, 75 mm × 50 cm; Thermo Fisher Scientific). Three CVs (−40/−60/−80) were used within a total MS cycle time of 6 s. Full MS scans were acquired over the range m/z of 350–2000 with a mass resolution of 120,000 (at m/z 200). Tandem mass spectra were acquired using an Orbitrap mass analyzer with a mass resolution of 30,000 at m/z 200 using the TurboTMT feature.
2.6.3. Protein Identification and Quantitation
A database search of all raw data files was performed using the Proteome Discoverer 2.5 software (Thermo Fisher Scientific). SEQUEST-HT was used to search the SwissProt Human database. The database search parameters included a precursor ion mass tolerance of 10 ppm, fragment ion mass tolerance of 0.02 Da, static modifications for carbamidomethylation (+57.021 Da/C) and TMT tags (+229.163 Da/K and N-terminal), and variable modifications for oxidation (+15.995 Da/M) and phosphorylation (+79.966 Da/S, T, Y). We obtained an FDR of less than 1% at the peptide level and filtered with high peptide confidence. The reporter abundance in each sample was calculated based on the S/N values or intensities of the reporter ion quantifier node.
2.6.4. Protein Expression Analysis
UniProtKB IDs were mapped to gene names using the UniProt ID mapping tool [
11]. Proteins with high dropout rates (>0.5) and high standard deviations in scaled abundance (>100 or >5 according to their distribution) were excluded from further analysis. To identify differentially expressed proteins (DEPs) between the responder and non-responder groups, we used the limma R package paired test to compare pre-and post-treatment conditions, and DEPs were selected at
p < 0.25 [
12]. Differences in gene expression were tested under paired conditions. The log-scaled fold-change (log2(post)–log2(pre)) expression for each sample was calculated, and limma was performed between responders and non-responders. Next, gene set enrichment analysis (GSEA) was applied to identify pathways using the KEGG gene set of DAVID (
p < 0.05) [
13,
14] (
Supplementary Table S1). Because the GSEA results uncovered ‘phagosome’, and ‘ECM-interaction receptor’ pathways involved in cellular interaction, we investigated whether the protein expression of ligand receptors exhibited the difference. The ligand receptor pairs were obtained from a knowledge-based database and a single-cell lung cancer study [
15]. We compared our DEPs with the ligand receptor collections using Venn analysis.
2.7. Flow Cytometry Analysis
The total numbers of various immune cells, including myeloid-derived suppressor cells (MDSC), cytotoxic CD8+T-cells, and regulatory T cells, were counted in the peripheral blood of the patients using flow cytometry. Fresh PBMCs were separated using HetasepTM (Stem Cell, Vancouver, BC, Canada) as described in the manufacturer’s protocol and stained for fluorescence-activated cell sorting (FACS) analysis. The antibody for each immune cell is listed in the following: APC-Cy7 labeled anti-human CD11b (BD Biosciences Clone: ICRF44), BV510 labeled anti-human CD33 (BD Biosciences Clone: WM53), BV786 labeled anti-human HLA-DR (BD Biosciences Clone: G46-6), BUV395 labeled Anti-human CD14 (BD Biosciences Clone:MP9), FITC labeled anti-human Lineage (BD Sciences), BUV labeled anti-human CD3 (BD Science Clone: UCHT1), PerCP-Cy5.5 labeled anti-human CD8 (BD Science Clone: RPA-T8), BV650 labeled anti-human CD4 (BD Science Clone SK3), BY711 labeled anti-human CD25 (BD Science Clone: 2A3), AF7647 labeled anti-FVS (BD Science Clone: Live/Dead), BUY496 labeled anti-human CD16 (BD Science Clone:3G8), AF647 labeled anti-human CD127 (BD Science Clone: RDR5), and PE-CF594 labeled anti-human CD15 (BD Biosciences Clone: W6D3). Stained cells were analyzed using Fortessa (BD Bioscience, Franklin Lakes, NJ, USA), and the data were analyzed using FlowJo (v10) software (FlowJo LLC, Vancouver, BC, Canada). MDSC subpopulation phenotypes were defined as described by Passaro et al. as follows: granulocytic MDSC, Lin-CD11b+CD33+CD11b+HLA-DR−/lowCD14−CD15+; monocytic MDSC, Lin-CD11b+CD33+CD11b+HLA-DR−/lowCD14+CD15−.
2.8. Validation Cohort Selection and Enzyme-Linked Immunosorbent Assay (ELISA) for TFRC
The validation cohort was composed of patients who had been treated with anti-PD1 or-PD-L1 antibodies between April 2016 and December 2022. Clinical data, including patient age, gender, smoking status, ECOG PS, line of therapy, PD-L1 expression level, PD-1 inhibitor type, and response to immunotherapy, were retrospectively analyzed. To assess the TFRC level of each patient, plasma was obtained by placing whole blood into a tube containing 0.5 m EDTA. The samples were centrifuged at 3000 rpm, 4 °C for 15 min. Plasma was collected and stored at −80 °C until further processing. TFRC concentrations were detected using ELISA kits (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer’s protocols.
2.9. Statistical Methods
A sample size (n = 18) was calculated to provide 80% power to demonstrate that the best ORR exceeded 10% at a one-sided type I error rate of 10% when the expected ORR in the treatment group was 30%. A total of 22 patients were planned for enrollment, considering a drop-out rate of 15%. The safety analysis population included all patients who received at least one dose of the study drug, and efficacy analyses were performed on the intention-to-treat population. Standard statistical methods were used, including descriptive statistics, the ꭓ2 test, logistic regression, the Kaplan–Meier method, two-sided 95% confidence interval (CI), and the stratified log-rank test. Data analyses were performed using SPSS (version 25.0; IBM Corporation, Armonk, NY, USA) and GraphPad Prism (version 9.5; GraphPad Software, San Diego, CA, USA).
4. Discussion
Currently, IHC-based PD-L1 expression is the only approved predictive marker for anti–PD-1 therapy, and its accuracy is insufficient to discriminate responders from non-responders completely. Among the immunosuppressive mechanisms that weaken the ICI response, suppressive immune regulatory cells such as Tregs and MDSCs play a key role in promoting tumor progression and inhibiting adaptive and innate immunity [
19,
20]. Therefore, targeting these suppressive mechanisms is essential for maximizing the efficacy of ICIs. Our study showed that two patients with a durable response maintained the lowest ratio (<1) of MDSC to CD8+T-cell throughout the treatment, consistent with previous features. However, our results also indicated that most patients had no effect on CA induction patients, which is different from the expectations of previous in vivo studies. Our results may be explained by other baseline characteristics. In an early survival analysis of patients with CM 057, Peters et al. showed that patients with poor prognostic characteristics (<3 months since the last treatment, progressive disease as the best response to prior treatment, ECOG performance status of 1), together with low or no PD-L1 expression, were at a higher risk of death within the first 3 months of treatment with nivolumab than with docetaxel [
2]. Therefore, the fact that most of the patients in our study had an ECOG status of 1 and less than 3 months after the last treatment may have led to a poor prognosis.
Innate immune checkpoints, which prevent malignant cells from being detected and removed through phagocytosis and suppress innate immune sensing, are also crucial for tumor-mediated immune escape and might be potential targets for cancer immunotherapy. Our study revealed that the expression levels of some proteins associated with phagocytosis were related to this response. Based on a previous study, TGFB1, known to interact with TGF-beta receptor 2, was involved in the interaction between metastatic lung cancer cells with mo-Macs (monocyte-macrophage). CCL5 interacts with exhausted CD8+ T cells and mo-Mac cells across primary lung tumors and metastases. THBS1 was involved in alveolar macrophage, interacting with the a3b1 complex of cancer cells. Meanwhile, TFRC expressed in malignant cells participates in the interaction between mo-Macs and malignant cells, including precancerous cells. Four proteins implied that three proteins of Mo-Mac cells were activated post-treatment, but the TFRC of the tumor cells decreased. Therefore, the results indicated that the treatment effectively inhibited tumor cells in responder patients.
TFRC can be used to predict immune phenotypes and immune cell infiltration in pancreatic cancer [
21]. TFRC is positively associated with many immunomodulators and is co-expressed with several significant immune checkpoints. Ferroptosis is a recently discovered form of regulated cell death characterized by the accumulation of lipid peroxides and subsequent destruction of cellular membranes [
22,
23]. The role of the TFRC gene and its protein product, the transferrin receptor, is responsible for iron uptake by cells, and iron is a key regulator of ferroptosis. Iron is necessary for ROS generation of reactive oxygen species, which can promote lipid peroxidation, ultimately leading to ferroptosis. TFRC plays a role in regulating iron homeostasis, and the overexpression of this gene can lead to increased iron uptake and susceptibility to ferroptosis [
24]. Studies have shown that ferroptotic cells can release damage-associated molecular patterns recognized by immune cells and trigger an immune response against cancer cells. These results indicate that the expression level of TFRC can affect the tumor-immune microenvironment, providing a new reference for the prognosis of ICI treatment [
25]. To the best of our knowledge, this is the first study to prove TFRC level as a predictive marker for anti-PD-1 therapy response in NSCLC patients with low PD-L1 expression. However, caution is needed for this mechanism to be a potential target for cancer immunotherapy because it has been shown that in vitro treatment of macrophages with the transferrin receptor ligand, transferrin, can promote the M2 polarization state; in other words, it can also contribute to cancer progression and therapeutic resistance [
26].