Sixteen archived LUAD samples (formalin-fixed, paraffin-embedded (FFPE) blocks) were kindly provided by the Annunziata Hospital (Cosenza, Italy). Samples were collected during the surgical tumor resection between 9 January 2020, and 10 November 2020. Informed consent was obtained from all subjects. The main clinical information associated with each sample were not correlated with any clinical studies or immune checkpoint inhibitor therapy.

The immunohistochemical (IHC) staining for PD-L1 was performed by using the FDA-approved Dako Agilent PD-L1 IHC 22C3 pharmDx kit on the Dako Autostainer Link 48 platform. PD-L1 positive control material for protocol establishment included FFPE specimens from normal tonsil. The analysis was limited to cell blocks with >100 tumor cell. The analysis of stained tissue samples was performed by 2 expert pathologists involved in the routine evaluation of clinical samples for diagnostic purposes. The estimation of PD-L1 expression below or beyond the 2 cut-off points (1% or 50%) were done according to current clinical practice for the determination of patients eligible for anti-PD-1 treatments. According to the Tumor Proportion Score (TPS) patients were classified as low TPS (1%–49%) and high (TPS ≥50%) expressors, respectively (De Marchi et al., 2021).

To determine the iron content in both intra-tumoral and peritumoral TAMs, each tumor sample was stained by using Perls Van-Gieson Kit (Bio-Optica). Briefly, paraffin-embedded tissue samples were rehydrated, stained with Perls solution for 20 min and then rinsed in distilled water. Sections were then counterstained with Van-Gieson solution for 10 min and, rinsed in water. Finally, sections were dehydrated in 70%, 95%, 100% ethanol and embedded using xylene-based mounting media. Producer guidelines have been used to quantify the positivity or negativity to Perls staining. Patients were defined as iron positive or negative according to the positivity (despite of the threshold) or negativity to Perls staining.

LUAD cell lines A549 and H460 were purchased from the American Type Culture Collection (ATCC, Rockville, MD, United States) and grown in RPMI-1640 (Sigma-Aldrich, St. Louis, MO, United States) supplemented with (v/v) fetal bovine serum (FBS) (Invitrogen, San Diego, CA), 100 U/mL of penicillin, and 0.1 mg/mL of streptomycin (Sigma-Aldrich, St. Louis, MO, United States). Cells were maintained in a 5% CO₂ humidified atmosphere at 37°C and periodically tested for the presence of mycoplasma. For each experiment, cells were seeded to obtain 70%–80% confluence.

Ferlixit (62.5 mg/5 mL, SANOFI) was obtained from the outpatient pharmacy at Unit of Cardiology, Magna Graecia University, Germaneto, while the antioxidant (±)-6-hydroxy- 2,5,7,8-tetra-methylchromane-2-carboxylic acid (trolox) was ordered from Cayman Chemical (Cayman Chemical Company, Ann Arbor, United States). Cells were seeded in a 6-well plate in serum-free medium. Ferlixit was added into the medium at the final concentration of 250 µM for 24 h while trolox was used at 200 μM for 6 h.

To obtain total protein extracts, protein extraction was performed using RIPA buffer as previously described (Biamonte et al., 2015; Zolea et al., 2015; 2016; Di Sanzo et al., 2022), supplemented with cOmplete™ Protease Inhibitor Cocktail provided in EASYpacks (Roche Diagnostics, Mannheim, Germany). Otherwise, nuclear, and cytoplasmic protein extracts were obtained as previously described (Chirillo et al., 2020; Scaramuzzino et al., 2021). Equal amounts of protein (50 μg) from each sample were separated by 8%–12% SDS-PAGE. The migration was performed at 200 V for 1 h and 30’. Then, proteins were transferred to nitrocellulose membranes (Sigma-Aldrich, St. Louis, MO, United States) at 50 V for 2 h. The membranes were blocked in 5% milk or 5% BSA for 1 h at room temperature and incubated overnight at 4°C with primary antibodies against PD-L1 (1:1,000, PA5-28115, ThermoFisher Scientific), c-Myc (1:500, sc-42, Santa Cruz Biotechnology), Nrf2 (1:200, sc-365949, Santa Cruz Biotechnology), NF-κB (1:500, sc-8008, Santa Cruz Biotechnology), HIF-1α (1:500, sc-10790, Santa Cruz Biotechnology) p-STAT3 (1:500, 4,113s, Cell Signaling Technology), STAT3 (1:500, 9,139s, Cell Signaling Technology), p-mTOR (1:500, 5,536s, Cell Signaling Technology), mTOR (1:500, 2,972s, Cell Signaling Technology), p-ERK (1:500, 9,106s, Cell Signaling Technology), ERK (1:500, 9,107s, Cell Signaling Technology), c-JUN (1:500, sc-1694, Santa Cruz Biotechnology), p70 S6 Kinase (S6, 1:500, 2,708, Cell Signaling Technology) and Phospho-p70 S6 Kinase (p-S6, 1:500, 9,206, Cell Signaling Technology). The membranes were washed for 30 min and then incubated for 1 h at room temperature with peroxidase-conjugated secondary antibodies (Peroxidase AffiniPure Sheep Anti-Mouse IgG, 1:10,000; Peroxidase AffiniPure Donkey Anti-Rabbit IgG, 1:10,000; Peroxidase AffiniPure Donkey Anti-Goat IgG, 1:10,000; Jackson ImmunoResearch Europe Ltd). Signals were detected using chemiluminescence reagents (ECL Western blotting detection system, Santa Cruz Biotechnology, Dallas, Texas) and acquired by Uvitec Alliance Mini HD9 (Uvitec Cambridge, United Kingdom). To calculate the relative expression of specific protein, a goat polyclonal anti-γ-Tubulin antibody (γ-TUB, 1:3,000; sc-17787; Santa Cruz Biotechnology) serves as a reference for citosolic sample loading, while Lamin A (LAMIN, 1:2000, sc-20680, Santa Cruz Biotechnology) was used as loading control for nuclear samples. The protein band intensity on western blots was quantified and normalized to that of γ-TUB or LAMIN by using ImageJ software (http://rsb.info.nih.gov/ij/).

Intracellular labile iron concentration was determined by flow cytometry using the fluorescent iron sensor calcein acetoxymethyl ester (CA-AM). Briefly, cells were incubated with 0.25 μM CA-AM (Aldrich, Missouri, United States) for 30 min at 37°C in the dark. Then, cells were washed twice with PBS (1X) to remove the excess of CA-AM, and thus treated with 200 μM L1 (3-Hydroxy-1,2-dimethyl-4(1H)-pyridone, Sigma-Aldrich, Missouri, United States) or left untreated. The analysis was performed by FACS BD LSRFortessaTM X-20 cytofluorometer (BD Biosciences). The difference in cellular fluorescence after and before incubation with L1 reflected the labile iron pool: Δ M e a n   F l u o r e s c e n c e   I n t e n s i t y , Δ M F I = Δ M F I a f t e r − Δ M F I b e f o r e

Generation of mitochondrial ROS was measured by flow cytometry with the use of MitoSOX Red Mitochondrial Superoxide Indicator (Thermo Fisher Scientific Inc.). After treatments, cells were incubated with 5 µM MitoSOX Red for 10 min at 37°C, washed in PBS (1X), and then analyzed by flow cytometry using a FACS BD LSRFortessaTM X-20 cytofluorometer (BD Biosciences). A minimum of 20.000 cells was analyzed per condition. Fluorescence was measured using FlowJo software program (Tree Star, Inc.). Each experiment was performed in triplicate.

For the flow cytometry analysis of surface PD-L1, cells were incubated with PD-L1 antibody (anti-human CD274, APC, BioLegend, San Diego, California, United States) for 30 min in the dark. After washing twice with PBS (1X), cells were acquired in a FACS BD LSRFortessaTM X-20 cytofluorometer (BD Biosciences). Data were analyzed using FlowJo software (Tree Star, Inc.). Three independent experiments were conducted.

For identifying cells actively undergoing apoptosis, a double staining with Annexin V and PI was performed using Alexa Fluor^®488 Annexin V/Dead Cell Apoptosis Kit (Thermo Fisher Scientific, Waltham, Massachusetts, United States) as previously described (Scicchitano et al., 2023). After staining, cells were incubated at room temperature for 15 min in the dark. Each tube was diluted with 400 μL of Annexin Binding Buffer and then cells were analyzed by flow cytometry using the FACS BD LSRFortessaTM X-20 cytofluorometer (BD Biosciences). Data were analyzed using FlowJo software (Tree Star, Inc.). Three independent experiments were conducted.

Total RNA was extracted using the Trizol method (Life Technologies, Carlsbad, CA, United States) as previously described (Biamonte et al., 2017; De Vitis et al., 2023). Then, 1 µg of total RNA were retrotranscribed using High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific, Waltham, Massachusetts, United States). qRT-PCR was performed using the SYBR Green qPCR Master Mix (Thermo Fisher Scientific, Waltham, Massachusetts, United States) (Guglielmelli et al., 2010; Biamonte et al., 2021). Analysis was performed on QuantStudio 3 Applied Biosystems by Thermo Fisher Scientific. The relative mRNA expression level was calculated by the 2^−ΔΔCT method and glyceraldehyde 3-phosphate dehydrogenase (GAPDH) and actin beta (ACTB) were used as the housekeeping genes (Di Sanzo et al., 2016; 2018).

A549 and H460 cells were transfected using Lipofectamine 3,000 transfection reagent (Thermo Fisher Scientific, Waltham, Massachusetts, United States) according to the manufacturers’ protocol. c-Myc transient knockdown was performed by using a specific c-Myc siRNA (s912, Thermo Fisher Scientific, Waltham, Massachusetts, United States). To ensure an optimal control, the two cell lines were further transfected with Silencer™ Select Negative Control siRNA (Thermo Fisher Scientific, Waltham, Massachusetts, United States). The transfection efficiency was evaluated by Western blot at 48 h.

Human Peripheral blood mononuclear cells (PBMCs) were isolated from sodium heparin anticoagulated whole blood drawn from healthy donors. PBMCs isolation was performed through a density gradient centrifugation method using Ficoll Histopaque (Sigma-Aldrich, St. Louis, MO, United States). Then, PBMCs were cultured in RPMI 1640 medium and stimulated with Dynabeads Human T-Activator CD3/CD28 (11131D, Thermo Fisher Scientific) for 1 h at 37°C. Stimulated and not-stimulated PBMCs were co-colture for 3 h with H460 and A549 cells, previously treated with 250 μM ferlixit (24 h). T cell responses were assessed by ELISPOT assay using the Human IFN-ɣ ELISpot Kit (856.051.005S, Diaclone) according to the manufacturer’s protocol. Briefly, the assay plate was re-hydrated by washing with 1X PBS and blocked with capture antibody overnight at 37°C. The day after, culture medium collected from the above-mentioned co-cultures was added to each well and incubated for 2 h at room temperature followed by 1 h-incubation after adding both the detection antibody and the streptavidin conjugated with alkaline phosphatase. Finally, BCIP/NBT substrate solution was added, kept for 15–25 min until a color change was noted. Plate development was stopped with a water wash and the plate was air-dried at room temperature, avoiding exposure to light. Spots were enumerated using an automated spot counter (BIOSYS Bioreader 3,000 Auto Macroscope ELIspot Plate Reader) and data were expressed as mean values of triplicate determinations (Garofalo et al., 2023).

Statistical analysis was performed using R environment (R: a language and environment for statistical computing, n.d.). The expression data and relative clinical information under the project TCGA Lung Adenocarcinoma (LUAD) were downloaded from the GDC Data Portal using R package TCGAbiolinks (Mounir et al., 2019). TIMER2.0 (Li et al., 2020), a comprehensive resource platform tool, was used for the systematic analysis of immune infiltrates in TCGA-LUAD dataset. Non-parametric test was used for statistical analyses, as PDL1 and CD71 expression values did not follow a normal distribution (Shapiro-Wilk test).

Correlation of PD-L1 expression with iron positivity in tissue samples was tested by Fisher’s exact test analysis. For all the analyses, the unpaired, two-tailed Student’s t-test was used to test for significant differences between two experimental groups. A p-value <0.05 was considered statistically significant. Each experiment was performed at least three times; results are, then, presented as mean ± SD.

First, we assessed whether and how PD-L1 expression in tumor cells correlates with iron content within the TME on FFPE tissue samples derived from 16 LUAD patients, whose clinicopathological features are reported in Table 1. Upon staining with the 22C3 clone anti-PD-L1 antibody, according to the TPS, 5 out of 16 samples (31%) were classified as “high PD-L1 expressors” (TPS ≥50%) while 11 out of 16 (69%) were classified as “low PD-L1 expressors” (TPS = 1–49%). By using the Perls Van-Gieson staining method, samples were also classified as “low iron” when the presence of iron inclusions within the TME were not detectable (see representative image in Figure 1A, top left) while samples with clearly recognizable iron inclusions were considered as “high iron” (Figure 1A, bottom left). Notably, 7 out of the 11 patients (64%) with low PD-L1 expression presented a low iron density within the TME, while 4 out of the 5 patients (80%) expressing high levels of PD-L1 presented a high iron density (Figure 1B, Fisher Exact Test: p < 0.001). These results suggest that a higher iron density within the TME might correlate with higher PD-L1 expression levels in LUAD cells.

In light of the direct correlation between intratumoral iron density and PD-L1 levels, we hypothesized that iron levels could causally regulate PD-L1 expression. To this end, we investigated the effects of iron addition within culture media on PD-L1 expression in LUAD cell lines. To this, we cultured A549 and H460 for 24 h in their relative culture media with or without 250 µM ferlixit, a Fe³⁺ compound normally used to treat patients with anemia. As shown in Figure 2A, when growth in the iron-rich culture condition, H460 and A549 cells show an intracellular accumulation of free-iron (A549^untreated ΔMFI: 1948; A549^ferlixit ΔMFI: 6,944; H460^untreated ΔMFI: 10,559; H460^ferlixit ΔMFI: 17,067), and an overproduction of mitochondrial ROS (A549^untreated MFI: 103; A549^ferlixit MFI: 324; H460^untreated MFI: 267; H460^ferlixit MFI: 474, Figure 2B). This is accompanied by a significant upregulation of PD-L1 surface levels in both th cell lines (Figure 2C). The strength of the upregulation inversely correlates with PD-L1 steady state amounts of the two LUAD cell lines: in H460 cells, showing very high levels of baseline PD-L1, iron causes a 2-fold increase, while in A549, with very low baseline PD-L1 expression, iron determines more than 20-fold increase. Notably, in both cell lines, PD-L1 overexpression is consistently counteracted by treatment with the antioxidant and ROS quencher trolox (200 µM for 6 h) (A549^untreated MFI: 148; A549^ferlixit MFI: 4,672; A549^{ferlixit+trolox} MFI: 2,639; H460^untreated MFI: 6,762; H460^ferlixit MFI: 10,723, H460^{ferlixit+ferlixit} MFI:8,893, Figure 2C), thus strongly suggesting that iron promotes PD-L1 overexpression in a ROS-dependent manner. No signs of apoptosis or ferroptosis are detectable by annexin/PI flow cytometry analysis, thus excluding the possibility that PD-L1 overexpression in A549 and H460 is determined by potential cytotoxic effects triggered by the iron-mediated oxidative stress (Supplementary Figure S1).

In cancer cells, the expression of PD-L1 is intricately regulated either at transcriptional, post-transcriptional, and post-translational levels (Ribas, 2015; Ju et al., 2020). As shown by qPCR analysis in Figure 3A (see also Supplementary Figure S2A), we demonstrated that iron promotes PD-L1 overexpression essentially at mRNA levels and that this is accompanied by the enhanced nuclear translocation of c-Myc, a well-defined trancription factor of PD-L1 (Figure 3B). Once again, The administration of trolox reduces the translocation of c-Myc, which correlated with the lack of PD-L1 upregulation in both the cell lines (Figure 3B). No significant variations were observed, instead, for other PD-L1 transcription factors c-JUN, p-mTOR, HIF-1α, p-ERK, p-s6, NF-kB, p-STAT3, and Nrf2 (Supplementary Figure S2B). To better dissect the role of c-Myc on iron-mediated PD-L1 regulation, we performed the transient knockdown of c-Myc (48 h) in A549 and H460 cultured either in iron-rich or non iron-rich culture media. As shown in Figures 3C, D (see also Supplementary Figure S2C), c-Myc silencing alone causes a reduction of both cytoplasmic and nuclear c-Myc and a parallel downregulation of PD-L1 mRNA and protein levels. Besides, c-Myc knockdown in A549 and H460 grown in iron-rich culture conditons significantly attenuates its translocation and, as a consequence, the overexpression of PD-L1 induced by iron. The effects of c-Myc on iron-mediated PD-L1 regulation was further confirmed at surface levels (Figure 3E). Together, these results suggest that, in A549 and H460 LUAD cells, the iron-dependent PD-L1 upregulation is mediated by the transcription factor c-Myc.

To assess whether the overexpression of PD-L1 induced by iron in LUAD cells can affect the immune function of T cells, we measured the production and the release of IFN-γ by T cells in a co-culture system. PBMCs isolated from peripheral blood donated by healthy volunteers were stimulated with antiCD3/CD28 beads to activate T cells. Then, stimulated and not-stimulated (ns) PBMCs were co-cultured for 3 h with H460 and A549 cells in a culture medium supplemented with 250 μM ferlixit. As shown in Figure 4A, iron supplementation strongly enhances the overexpression of PD-L1 in A549 co-cultured with antiCD3/CD28 stimulated PBMCs (A549^untreated PD-L1 MFI: 236; A549^{untreated+PBMCs ns} PD-L1 MFI: 259; A549^{untreated+PBMCs antiCD3/CD28} PD-L1 MFI: 499; A549^{ferlixit+PBMCs ns} PD-L1 MFI: 1738; A549^{ferlixit+PBMCs antiCD3/CD28} PD-L1 MFI: 4,219). In agreement, the IFN-γ ELISPOT assay shows a remarkable reduction of IFN-γ release from stimulated PBMCs co-cultured with A549 in the iron-rich culture medium compared to those co-cultured with A549 without iron supplementation (Figure 4B). This result is partially replicated when stimulated PBMCs were co-cultured with H460 cells. Indeed, PD-L1 is only slightly overexpressed in H460 co-cultured with stimulated PBMCs in the iron-rich culture condition (H460^untreated PD-L1 MFI: 2,712; H460^{untreated+PBMCs ns} PD-L1 MFI: 2,563; H460^{untreated+PBMCs antiCD3/CD28} PD-L1 MFI: 2,697; H460^{ferlixit+PBMCs ns} PD-L1 MFI: 5,018; H460^{ferlixit+PBMCs antiCD3/CD28} PD-L1 MFI: 5,787) and IFN-γ release undergoes a small reduction (Figures 4A, B). Overall, these data show that an iron rich-environment promotes the overexpression of a functional PD-L1, which can significantly suppress the immune function of T cells in A549 cells. The less efficient iron-mediated induction of a functional PD-L1 in H460 cells requires further investigation; however, it can be attributed to the already very high levels of PD-L1 at baseline.

Cancer cells equipped with hyper functional iron uptake, often exerted by the overexpression of proteins devoted on iron intake (i.e., CD71), deprive the TME of iron to boost their protumoral functions or to suppress the anticancer activities of innate immune cells (Liang and Ferrara, 2021). Here, we finally explored whether an “iron addiction” phenotype, characterized by higher levels of CD71, may correlate with PD-L1 baseline expression in LUAD. To this, we analyzed PD-L1 and CD71 mRNA expression by using The Cancer Genome Atlas (TCGA) for LUAD dataset. We observed that in a cohort of primary LUAD patients (n = 433), higher levels of PD-L1 significantly correlates with higher CD71 expression levels (p = 2.15e-07, rho = 0.3) (Figure 5A). The correlation becomes more significant in LUAD patients with a higher T cell immune infiltration (n = 24) identified by using TIMER 2.0 (p < 0.004, rho = 0.6; Figure 5B).