Circulating low-density neutrophils as biomarkers of resistance to first-line anti-PD-1/PD-L1 immunotherapy in non-small cell lung cancer.

Introduction
Immune checkpoint inhibitors (ICI) have significantly enhanced the prognosis of patients with advanced NSCLC. Anti-PD-1/PD-L1 antibodies have become the standard frontline treatment, as single agent when PD-L1 tumour expression is high (≥50%), or combined with chemotherapy otherwise[[1], [2], [3], [4], [5], [6], [7], [8], [9], [10]]. However most patients do not respond to treatment, and no reliable predictive biomarkers have been identified yet[[11], [12], [13]].
The relevance of myeloid cells in the immune response against cancer has been highlighted in the recent years and have been identified as one of the dominant cell populations in the tumour microenvironment (TME)[14]. Neutrophils exhibit intrinsic plasticity, enabling them to adopt dual roles by presenting both pro-tumour (N1) and anti-tumour (N2) phenotypes[[15], [16], [17], [18]]. A population of circulating neutrophils named LDNs, which share mainly morphological and immunosuppressive characteristics with cells described as polymorphonuclear myeloid derived suppressor cells (PMN-MDSCs), have been detected in patients with cancer as well as in autoimmune diseases[19]. This population, absent in healthy volunteers, has immunosuppressive properties and heterogeneous maturation state in contrast to conventional neutrophils or high-density neutrophils (HDNs)[[20], [21], [22]]. Although an active role of these cells in resistance to ICI has been reported, it remains poorly understood[[23], [24], [25]].
Our group demonstrated an association between high baseline levels of LDNs and resistance to first-line ICI monotherapy in a cohort of 31 patients with advanced NSCLC. Surprisingly, this association was not observed in those treated with chemoimmunotherapy, where a depletion of LDN was observed in responders[26]. In this article, we provide further evidence in this line of research by including outcomes from a validation cohort. We have also explored the characteristics of this myeloid subpopulation using quantitative proteomics and multiparametric flow cytometry, and explored additional potential mechanisms associated with ICI resistance in plasma samples as well as in an in vivo model.

Methods

Study design and patient enrolment
A total of 120 patients recently diagnosed with advanced NSCLC were prospectively recruited at the Medical Oncology Department of the Hospital Universitario de Navarra (HUN) between April 2018 and January 2025. This cohort included 60 patients with PD-L1 expression ≥50% who received first-line ICI monotherapy, and 60 patients with PD-L1 expression <50% who were treated with first-line chemo-immunotherapy (CT+ICI). Data from a cohort of healthy donors (n = 11) were also evaluated (Figure S1). In the monotherapy cohort, LDN quantification was not assessable in four cases due to sample processing limitations, and clinical outcome data were unavailable for three patients because of incomplete follow-up. Written informed consent was obtained from all participants. Exclusion criteria were previous treatment for advanced disease, as well as mixed histology. Epidemiological and clinical variables were recorded. None of the patients presented with sepsis or received G-CSF at the start of treatment. Corticosteroid use was defined as >10mg/day of prednisone or equivalent at treatment start. Tumour responses were evaluated according to RECIST 1.1[27] and Immune-Related Response Criteria[28]. Progressive disease was confirmed by investigator assessment based on at least one sequential tumour assessment, except in the case of clear clinical deterioration. Somatic genetic variants were assessed by Next-Generation Sequencing (NGS) with the Oncomine Focus Assay (Thermo Fisher) targeting 52 genes in patients with non-squamous and non-smoking-associated squamous cell carcinoma. Some cases were not evaluated due to sample unavailability.

Flow cytometry
Blood samples (10mL) were obtained from each patient immediately before the first immunotherapy cycle and after the first radiological assessment. Samples were collected in EDTA tubes and processed within two hours from extraction (Fig. S1). PBMCs were isolated by Ficoll® gradient and HDNs by Mono-Poly™ resolving medium following the manufacturer’s indications[29]. After isolation, cells were washed, lysed using ACK lysing buffer, and stained with the indicated antibodies in a final volume of 50μL for 10min in ice. The gating strategy used to identify LDNs based on CD66b⁺ and CD116⁺ expression is shown in Fig. S1 and fluorochrome-conjugated antibodies (1:50 dilutions) are detailed in the supplementary materials.

Circulating inflammatory profile
A panel of 43 key cytokines involved in human inflammatory processes (HCYTA-60K-43C, Merck Millipore) was assessed by MILLIPLEX® technology in the plasma of 38 patients (19 from ICI cohort and 19 from CT+ICI cohort). Plasma from baseline blood samples was separated by centrifugation within 2 hours from extraction and stored at -80°C for further analysis. Samples were analysed following manufacturer indications. Each sample was measured in duplicate in a xMAP INTELLIFLEX equipment (Merck Millipore). The cytokines quantified are detailed in the supplementary materials.

Cell isolation and proteomics
For proteomic analysis, HDNs and LDNs from 9 patients were compared. HDNs and LDNs were purified as described before and then isolated with magnetic beads coupled to anti-CD15 antibodies (Invitrogen, ThermoFisher Scientific). Lysates were centrifuged at 100,000 g (1 h, 15°C), and the resulting supernatant was quantified with the Bradford assay kit (BioRad, Barcelona, Spain). Proteins were reduced with DTT (final concentration of 20 mM; room temperature, 30min), alkylated with iodoacetamide (final concentration of 30 mM; room temperature, 30 min in the dark), diluted to 0.9 M with ABC and digested with trypsin (Promega, Madison, WI, USA; 1:20 w/w enzyme protein ratio, 18 h, 37°C). Protein digestion was interrupted by acidification (pH < 6, acetic acid), and the resulting peptides were cleaned-up using Pierce™ Peptide Desalting Spin Columns (ThermoFisher Sci., Waltham MA, USA). Data independent acquisition (DIA)-mass spectrometry is detailed in the supplementary materials.

Tumour associated neutrophils
The evaluation of TANs was performed with the collaboration of the Pathology Department of the HUN based on morphological analysis of diagnostic tumour tissue samples stained with Haematoxylin and Eosin (H&E). Both intraepithelial and stromal TANs were evaluated within fibrovascular cores of papillary structures, covering a 0-100% range. Only primary or metastatic tumour samples were included, excluding lymph nodes, cytological samples, superinfection, ulceration and intravascular neutrophils. TANs were analysed in the enriched cell zone and categorized into four groups following a semiquantitative classification: Group 0 (<1%, 40x), Group 1 (1-4%, 40x, 1+), Group 2 (5-19%, 20x, 2+), and Group 3 (>20%, 10-20x, 3+).

In vivo experiment
A total of forty-eight, six-week-old female C57BL/6J (Envigo) mice were subcutaneously injected in the flank with 0.5 × 10⁶ Lewis Lung Carcinoma (LLC) cells. Tumour dimensions were measured every two days throughout the experiment, and tumour volume was calculated using the formula: (width × length²)/2. When tumours reached approximately 50mm³, mice were randomized into six treatment groups (n = 8 per group), with treatments administered intraperitoneally. Targeting of MDSCs was achieved using an anti-Gr1 antibody, which binds to Ly6 receptors expressed on the surface of PMN-MDSCs[30]. Treatment groups were: a) control (vehicle-treated); b) anti-PD1 (100 μg/mouse; a total of 3 injections); c) anti-GR1 (200 μg/mouse; a total of 3 injections); d) anti-PD1 combined with anti-GR1 (same dose and injections described before); e) chemotherapy consisting of cisplatin [3 mg/kg] + pemetrexed [100 mg/kg] + anti-PD1 (3 injections each) (CT+ICI); f) CT+ICI plus anti-GR1 (see supplementary materials, Fig. 5A for more details about the therapeutic scheme). At the end of the treatment period, mice were euthanized by CO2 administration. For the murine experiments, LLC cells were obtained from ATCC (#CRL-1642); anti-mouse Gr1 antibody (#BE0075, clone RB6-8C5) and anti-PD-1 antibody (#BE0273, clone RMP1-14) were purchased from BioXCell; pemetrexed (1000 mg/40 mL, 25 mg/mL) was obtained from TEVA; and cisplatin (#232120) was purchased from Sigma-Aldrich. The housing conditions of the mice are described in the supplementary materials.

Data collection. bioinformatics and statistical analysis
Flow cytometry data was analysed using FlowJo (BD Biosciences, Franklin Lakes, NJ, USA). Statistical analysis was performed using GraphPad Prism v8.0 (GraphPad Software, San Diego, CA, USA) and SPSS packages (IBM, Armonk, NY, USA). Percentages of LDNs did not follow a normal distribution; therefore, group comparisons were performed using nonparametric tests. The nonparametric tests used are detailed in the supplementary materials.
Fast progressive disease (fast-PD) was defined as death within 12 weeks after the initiation of treatment. Clinical Benefit Rate (CBR6) longer than 6 months was defined as absence of progression or death within 6 months after the initiation of treatment. Prognostic scores as Neutrophil-to-Lymphocyte ratio (NLR), Derived Neutrophil-to-Lymphocyte ratio (dNLR), Gustave Roussy Immune Score (GRIm-S) and Lung Immune Prognostic Index (LIPI) score were calculated as described elsewhere[31,32]. The definitions of some variables are detailed in the supplementary materials.
For proteomic analysis, mass spectrometry data files were analyzed using Spectronaut (Biognosys) by direct DIA analysis (dDIA[33], detailed in the supplementary materials). Mass-spectrometry data and search results files were deposited in the Proteome Xchange Consortium via the JPOST partner repository (https://repository.jpostdb.org)(34) with the identifier PXD066804 for ProteomeXchange and JPST003967 for jPOST. (for reviewers: https://repository.jpostdb.org/preview/659964475688b7ac725729; Access key: 1669). Results from proteomics were further analysed by STRING (Search Tool for the Retrieval of Interacting Genes) software and represented by Cystoscape stringApp (v 3.10.3). The pathway enrichment analysis was performed in the Ingenuity Pathway Analysis (IPA, QIAGEN Bioinformatics) based in the QIAGEN Knowledge Base (QKB). Selected pathways correspond to the most differentially regulated z-scores, all statistically significant.

Results

Study population
In the overall population, the cohort receiving first-line ICI monotherapy (n=60) had PD-L1 expression ≥50%. Baseline characteristics are summarized in Table S1. Median PFS was 6.7 months, and median OS was 21.8 months. CBR at 6 months was 53.7% and the incidence of fast-PD disease was 24.1%. In this cohort, patients with dNLR > 3 experienced more frequently a CBR < 6 months (75% vs 25%, p = 0.05). In contrast, other unfavourable prognostic markers, including LIPI, GRIm, and NLR, were not significantly associated with CBR < 6 months. In the cohort receiving first-line chemo-immunotherapy (n = 60), 47% of patients had PD-L1 tumour expression <1%. Baseline characteristics are presented in Table S2. In this cohort, median PFS was 6.4 months, and median OS was 12.3 months. CBR at 6 months was 56.7% and the incidence of fast-PD was 16.7%. Moreover, unfavourable prognostic markers such as, LIPI, GRIm, dNLR and NLR, were not significantly associated with outcomes.
Baseline LDN levels showed a weak correlation with blood neutrophil counts levels (r = 0.24, p = 0.008) but no association with NLR or dNLR (Fig. S2). A multiple linear regression analysis showed no significant overall association between baseline LDNs levels and clinical variables (tumour burden, body mass index, diabetes mellitus, liver metastases, corticosteroid use, antibiotic therapy before treatment) (Table S3).

High baseline circulating LDNs associate with resistance to first line ICI monotherapy in NSCLC
Previous studies by our group[26] reported an association between LDNs levels above 7.09% in peripheral blood and the resistance to first-line ICI monotherapy in patients with advanced NSCLC. In the present study, these findings are confirmed in a validation cohort of 60 patients. Higher baseline LDN levels were observed in those with non-CBR compared to those with CBR>6 months (27.2% vs 10.1%; p = 0.005). Compared to cancer patients, LDNs were not detectable in blood samples from healthy donors (mean 20.6 vs 0.6; p = <0,001) (Fig. 1A). The threshold of 7.09% for baseline LDN levels, established in the initial cohort, was confirmed as a potential predictive biomarker for response to ICI monotherapy, with an AUC value of 0.76 (p = 0.002) in the validation cohort. This threshold identified patients with a CBR < 6 months with a sensitivity of 83.8% and specificity of 73.1% (Fig. 1B). Patients who experienced disease progression as best response to ICIs had significantly higher baseline LDN levels (32.1% vs 13.6%, p<0.0001) (Fig. 1C). Notably, patients with baseline LDN levels above 7.09% experienced progression as best response more frequently (64.5% vs 9%, p < 0.0001) and showed a lower ORR compared to those below the threshold (17% vs 50%, p = 0.004). Furthermore, the incidence of fast-PD (FPD) was significantly higher in this group (92.8% vs 7.1%; p = 0.003) (Fig. 1D). Although baseline LDN levels differed between responders and non-responders, LDN levels remained stable during treatment in both groups (responders: 1.6% vs. 4.4%, p = 0.21; non-responders: 18.1% vs. 11.3%, p = 0.43) (Figure S3). Patients with baseline LDN levels above 7.09% exhibited a median of PFS of 2.37 months (95% CI: 1.37–3.36) compared with 21.85 months (95% CI: 0.00–43.93) for patients with LDN ≤7.09% (p < 0.001) (Fig. 1E). Similarly, median mOS was 11 months (95% CI: 0.16–21.86) versus 28.2 months (95% CI: 14.97–41.41) in the lower LDN group (p = 0.002), indicating a strong association between elevated baseline LDN levels and poorer survival outcomes (Fig. 1F). In the multivariate Cox model, baseline LDNs emerged as an independent predictor of response (HR = 20.38, 95% CI 1.87–221.79, p = 0.013), whereas other established clinical prognostic variables (tumor burden, liver metastases, LDH levels, histology, NLR) were not significant (Table S4).

Circulating LDNs levels do not associate with responses to first line CT+ICI combination therapy in NSCLC
To validate our previous findings[26] suggesting that the association between baseline LDN levels and response to ICI was treatment-dependent, we analysed a validation cohort of patients with advanced NSCLC treated with first-line CT+ICI. Consistent with our earlier results, higher baseline LDN levels were not associated with resistance to treatment, as no differences were found between non-responders and responders (mean 17.9% vs 30.1%; p=0.09), even responders presented a trend towards higher LDN levels (Fig. 2A). A progressive decline in LDN levels was detected in the patients who responded to the treatment (n=29), with significantly lower LDN levels observed at the time of radiological assessment (mean baseline LDNs 30.1% vs 3rd cycle LDNs 6.4%; p=0.0006) (Fig. 2B). However, this effect was not observed in non-responders (Fig. 2C). Patients with high LDN levels (71.7% of the cohort) did not show a higher incidence of progressive disease as the best response (21.9% vs 41.1%; p=0.14) nor were there differences in PFS (mPFS; 8.1 months [95% CI: 4.47–11.89] vs 5.2 months [95% CI: 0.56–10.02]; p =0.18) (Fig. 2D) or OS (mOS; 15.2 months [95% CI: 9.96–20.60] vs 9.8 months [95% CI: 0.64–19.01]; p = 0.87) (Fig. 2E). In the multivariate Cox model, none of the evaluated variables, including LDN levels, tumour PD-L1 expression, LDH, NLR, and other clinical factors, were independently associated with survival (Table S5).

LDNs show a differential phenotype compared to HDNs
To assess the differences in the phenotype of LDNs and HDNs, we analysed the expression of maturation and activation surface markers[21] in LDNs and HDNs from 24 patients (14 from ICI cohort and 10 from CT+ICI cohort). Compared to HDNs, LDNs exhibited more frequently an aged phenotype (CXCR4+ CXCR2−; LDNs: 22.3% vs HDNs: 2.5%; p < 0.001) while the immature phenotype was less frequent (CXCR4+ CXCR2+; LDNs: 67.2% vs HDNs: 84%; p=0.03). Furthermore, we evaluated phenotypic differences in HDNs between NSCLC patients and healthy donors (n=5), and no significant differences were found (Fig. 3A). Additionally, to assess phenotype changes during treatment, LDNs were analysed at baseline and after three treatment cycles in 11 patients (3 from ICI cohort and 8 from CT+ICI cohort), revealing increased population heterogeneity characterized by a rise in aged LDNs (18.2% vs 51.1%; p=0.014) and a reduction in mature LDNs (74.2% vs 38.1%; p = 0.007) (Fig. 3B). No significant differences in phenotypes were observed between responders and non-responders at baseline. However, in non-responders, there was a marked tendency toward a decrease in immature LDNs and an increase in aged LDNs over the course of treatment (Fig. S4). Regarding other markers commonly used to evaluate neutrophil maturity (CXCR2, CXCR4, CD10, CD101, CD62L), a greater heterogeneity was observed in LDNs, whereas HDNs were more homogeneous (Fig. S5A). Compared with HDNs, LDNs tended to exhibit a higher expression of PD-L1 (PD-L1⁺ LDNs: 2.3% vs HDNs: 1%). However, no significant differences were observed in other neutrophil activation markers, LOX-1 (LOX-1⁺ LDNs: 77.6% vs HDNs: 75%) or HLA-DR expression (HLA-DR⁻LDNs: 97% vs HDNs: 98.6%) (Fig. S5B).
To study the characteristics of LDNs and to identify proteins differentially expressed related with HDNs, the proteomic profiles of these cells were compared by quantitative proteomics. The proteomic analysis of LDNs and HDNs identified 365 differentially expressed proteins, with 178 upregulated and 187 downregulated proteins in LDNs (Fig. 4A-B). Functional interactome analysis of upregulated proteins in LDNs showed an increase in chemotaxis and neutrophil migration (Platelet-factor 4 (PF4), PPBP), including key proteins involved in integrin signalling (Vinculin), cell adhesion and transepithelial migration (NHERF1, PDIA3, TALIN1), cytoskeletal modulation (Ezrin-radixin-moesin family), suggesting an autocrine loop inducing LDN mobilization. Furthermore, LDNs exhibited higher levels of proteins linked to the formation of neutrophil extracellular traps (NETs) (PF4, Lamin-B1, Moesin), and a reduced capacity for neutrophil degranulation (ADRM1, CAMP, EIF4A3, PHB), suggesting immune dysfunction. Several upregulated proteins were associated with increased cell survival and proliferation, including those related to the inhibition of apoptosis (Cathepsin B, CSTF2, EIF3D, Histone 1.4, Lamin-B1), enhanced haematopoiesis and cellular differentiation (Multimerin-1, PAK2). LDNs also presented a greater dependence on mitochondrial metabolism and increased oxidative phosphorylation (ATP5F1B, STP5ME, ATP5PB, ATP5PO), and ATP synthesis (ACAT 1, COC5A, COX5B, COX6B1). However, in contrast to HDNs, LDNs showed a downregulation of proteins involved in reactive oxygen species (ROS) detoxification. Additionally, a wide range of upregulated proteins in LDNs were associated with protein synthesis, including ribosomal proteins (RPL, RPS), transcription, post-transcriptional modifications and translation related proteins (HNRNPA, HNRNPF, HNRNPM, H1-10, TPR). LDNs presented downregulation of proteins involved in protein degradation, particularly through ubiquitination, compared with HDNs (Fig. 4C).
Pathway enrichment analysis of proteins supported the previous results. Compared with HDNs, LDNs were enriched for pathways related with DNA replication and protein synthesis (EIF2 signalling, initiation, elongation, and termination of translation). However, HDNs were enriched for pathways related to neutrophil degranulation, immune response (antigen processing and presentation via MHC class I), apoptosis and response to hypoxia or oxidative stress (Fig. 4D). Significantly enriched pathways with corresponding z-scores, p-values, and FDR-adjusted q-values (Benjamini-Hochberg correction) are shown in Table S6.

Relationship between circulating LDNs and TANs
To investigate whether the expansion of circulating LDNs correlated with higher myeloid infiltration in the tumour microenvironment, samples from diagnostic biopsies of 40 patients with high and low LDN baseline levels were evaluated. TANs were detected in 17 of these samples and no significant association with baseline LDNs was observed (p=0.74) (Fig. S6). No association was found between TAN infiltration and other variables as histological subtype, age, tumour burden, NLR, or best response to treatment (Table S7).

Immunosuppressive cytokines predominate plasma inflammatory profile in patients with poor prognosis
To analyse potential circulating cytokines mediating the expansion and phenotypic differentiation of LDNs, we compared a panel of 43 inflammatory cytokines in plasma samples from selected patients with high versus low baseline levels of LDNs (n=38). Levels for all the proteins in patients with low vs high LDNs as well as p values are shown in Table S8. Even though statistical significance was not reached, probably due to the limited sample size, selected cytokines showing most prominent changes are shown in Fig. S7. Interestingly, several cytokines known to support the expansion and immunosuppressive activity of LDNs and other MDSCs, including M-CSF, IL-6, IL-1β, and TGF-β were elevated in patients with high levels of LDNs. Notably, IL-1β and M-CSF showed particularly high levels, suggesting a prominent role in driving the observed myeloid expansion. Additionally, other cytokines such as CXCL9 and IFN-γ were elevated, which are involved not only in MDSCs expansion but also their activation in the TME. IL-22 and IL-25, known to contribute to regulatory immune processes, were in slightly higher concentration. These patients showed elevated levels of EGF, known to stimulate cell proliferation and differentiation through EGFR activation.

Depletion of the Gr1+ population in a NSCLC model with high LDNs and anti-PD-1 resistance, increases response to ICI and chemotherapy-ICI therapy
To investigate the role of LDNs in NSCLC resistance to ICI, we used the aggressive LLC mouse model, reported as refractory to anti-PD-1[34]. Tumour progression in LLC-bearing mice coincides with LDN accumulation in tumours, spleen, and bone marrow, making it suitable for studying LDNs in therapy resistance[35]. As a broader approach to target LDNs, we used an anti-Gr1+ blocking antibody[30], establishing several treatment groups: control (vehicle-treated), anti-PD-1, anti-Gr1, CT+ICI (cisplatin, pemetrexed and anti-PD-1), and anti-Gr1 + CT+ICI (Fig. 5A).
As expected, anti-PD-1 monotherapy had no effect on LLC tumours (Fig. 5B). Anti-Gr1+ monotherapy resulted in a significant reduction in tumour volume (p<0.05, Fig. 5C). Surprisingly, the combination of anti-PD-1 and anti-Gr1+ showed a weaker therapeutic effect than anti-Gr1+ alone, with no significant difference compared to controls (Figure S8). CT+ICI treatment caused a highly significant reduction in tumour volume (p<0.001; 33.7% decrease) compared to controls (Fig. 5D) and the combination of CT+ICI with anti-Gr1+ produced the most pronounced antitumour effect (p<0.001; 49.4% decrease, Fig. 5E). A graphic scheme including all study groups is shown in supplementary materials (Fig. 5A). These results suggest that targeting the LDNs population may enhance the therapeutic efficacy of CT+ICI. The detrimental effect observed with the combination of anti-Gr1+ and anti-PD-1 warrants further investigation.

Discussion
ICIs have transformed the treatment of advanced NSCLC, with PD-L1 expression as the main recommended biomarker[36,37]. However many patients with high PD-L1 expression will progress within the first weeks of treatment[2,6,7]. Understanding resistance and identifying reliable biomarkers is crucial. Circulating immune cells, especially neutrophils and MDSCs, have been linked to prognosis and ICI response. The prognostic significance of circulating neutrophils, often presented as NLR or dNLR, predicts resistance to ICI therapy in NSCLC[32,38,39] and other tumour types[40]. We also observed higher dNLR in poor responders, but baseline LDNs did not correlate with NLR/dNLR, suggesting that LDN expansion is independent. In patients with NSCLC, elevated circulating LDNs are associated with poor prognosis and ICI resistance[23,24,26]. Our data confirms that baseline LDNs predict resistance to first-line ICI monotherapy but not CT+ICI, highlighting its predictive and not prognostic, value. Responders to CT+ICI showed LDN decline, suggesting chemotherapy depletes LDNs and enhances immunity.
LDNs have been characterised by their immunosuppressive role in cancer and inflammatory conditions like sepsis. Our quantitative proteomic analysis comparing LDNs to HDNs revealed a distinct profile, with enrichment in cell survival, immunosuppression, protein synthesis, and mitochondrial metabolism pathways. These findings are consistent with previous research in MDSCs, such as Zhang Y et al.[41], showing upregulated DNA replication, protein translation and anabolic metabolism, while pathways related to immune response and degranulation were enriched in HDNs. Similarly, Leite et al.[42], who analysed the proteome of LDNs in sepsis, reported the upregulation of pathways associated with T-cell suppression, proliferation and NET formation, consistent with our findings.
Some studies have reported phenotypic variability in LDNs compared to HDNs[20,43]. Consistently, our study found LDNs to be more heterogeneous and aged (CXCR4+CXCR2-), while HDNs were mature and uniform. LDNs also showed low, albeit increased levels of PDL1+ phenotype compared to HDNs.
Several studies have explored the prognostic value of circulating cytokines in NSCLC and their role in immunotherapy response. IL-8 is linked to immunosuppression and recruitment of MDSCs, while IL-6 promotes tumour survival[[42], [43], [44]]. In our study, most plasma cytokines showed no significant differences between groups. However, patients with high LDNs had increased levels of cytokines like IFN-γ, TNF-β, and IL-25, as well as the myeloid growth factors M-CSF, IL-6 and IL-22, which influence neutrophil chemotaxis. Consistent with Barrera et al.[25,46], we observed higher IL-1β in patients with elevated LDNs, reflecting their pro-inflammatory and immunosuppressive roles.
To clarify the role of LDNs population in the differential response of NSCLC patients, we used the LLC mouse model to evaluate the impact of a LDN depleting approach (anti-Gr1+ therapy) on tumour response. Anti-Gr1+ monotherapy targeting this population produced a significant though modest antitumour activity, although this effect was abrogated when anti-Gr1+ was combined with anti-PD-1 therapy. However, a significant synergy was observed when anti-Gr1+ treatment was combined with chemotherapy and anti-PD-1. These results suggest that, although LDNs appear to play a key role in resistance to PD-1 blockade in NSCLC, additional mechanisms or immune cell populations likely contribute, thereby substantially increasing its complexity. PD-1 inhibition may promote compensatory immunoregulatory pathways, including the expansion and functional reprogramming of immunosuppressive myeloid populations and the induction of immunosuppressive cytokines, thereby rendering anti-Gr1⁺ mediated cell depletion insufficient to sustain an antitumour response and indicating that the addition of chemotherapy may be required to achieve effective tumour control. In mesothelioma mouse models, the combination of cisplatin, pemetrexed and anti-PD-1 has also shown antitumour effect by inhibiting MDSC tumour infiltration[45]. Our results show that targeting Gr1+ enhances CT+ICI efficacy, likely depleting LDNs, reducing systemic immunosuppression and boosting antitumour activity.
We observed a specific association between circulating LDNs and resistance to ICI monotherapy in frontline NSCLC treatment. These findings suggest that myeloid cell expansion and plasticity influence the response to immunotherapy. Moreover, the lack of correlation between circulating LDNs and TAN infiltration suggests that this mechanism is likely systemic rather than local, potentially driven by tumour-derived soluble factors. In our previous work[26], we showed plasma from NSCLC patients with high LDNs suppresses T cell cytotoxicity and promotes tumour growth, involving LDN-related cytokines. However, baseline cytokine levels in this study were not differentially higher in patients with elevated LDNs, potentially because of liberation to the systemic circulation upon PD-1 blockade. This would be in agreement of the findings of the study by Glodde N et al, describing in a murine model a c-MET-driven myeloid expansion in peripheral blood upon immunotherapy treatment[18].
Further studies focusing on TANs and their characterization should be conducted in larger tissue samples to account for tumour heterogeneity. While myeloid expansion correlates with ICI monotherapy resistance, it does not impair chemo-immunotherapy response, possibly due to chemotherapy-induced LDN depletion. Our a murine model with high Gr1⁺ LDNs, Gr1⁺ depletion enhanced ICI response, especially combined with chemotherapy (CT+ICI), which would provide a rationale for the advantage for CT+ICI compared with ICI monotherapy in patients with NSCLC PD-L1 ≥ 50% and myeloid peripheral expansion reported by previous works[46,47]. IL-6, which is secreted by PD-L1 overexpressing cells, has proven to be a main regulator of the immunosuppressive activities of myeloid cells, and was elevated in the patients with high LDNs from our cohort[48]. Accordingly, the combination of PD-1 immunotherapy and IL-6 blockade exhibited synergy in a preclinical study, reinforcing the main role that myeloid cells play for the immunotherapy resistance in NSCLC[49].
However, several limitations of our work should be addressed. A larger prospective cohort would be required to confirm the findings and to explore the effect in patients treated with other ICI currently approved for the treatment of NSCLC, including not only anti-PD1 or anti-PDL1 antibodies but also anti-CTLA4.
These findings highlight the need for new strategies to overcome myeloid-driven resistance. Although specific therapies for MDSCs have not yet been identified, several promising approaches are under investigation such as blocking IL-8/CXCR2 pathway, NOX or Arg1 inhibitors or cytokine-neutralizing antibodies[19,50]. Combination regimens with chemotherapy or dual immune checkpoint blockade with CTLA-4 inhibitors may further remodel the myeloid compartment[51].

Conclusion
High baseline LDN levels are associated with resistance to first line ICI monotherapy in patients with NSCLC. The combination with chemotherapy could overcome this resistance by depleting these cells and thereby preventing their immunosuppressive role. LDNs exhibit a distinct proteomic profile, with a shift towards mitochondrial metabolism and immunosuppression. Plasma from patients with high LDNs levels is enriched in cytokines involved in inflammatory immune processes and myeloid expansion. Direct targeting of LDNs could further enhance the efficacy of immunotherapy and should be explored in additional studies.

Funding
The authors are supported and funded by the Spanish Association against Cancer (AECC, PROYE16001ESCO), SOSCLC (AECC 70% Survivorship Challenge), Instituto de Salud Carlos III (CP12/03114, FIS. PI17/02119 and PI22/01253) and Spanish Society of Medical Oncology (SEOM) (Beca SEOM para Proyectos de Investigación Traslacional en Inmuno-Oncología 2023).

Ethics approval and consent to participate
The Ethics Committee for Investigation with Medicinal Products from Navarre approved this study with the following record ID, PI_2023/139. Informed consent was obtained from all subjects. The animal experiment was approved by the Ethical Committee for Animal Experimentation of the University of Navarra (ref. 082-23). All experiments were performed according to the principles established in the Declaration of Helsinki and the Department of Health and Human Services Belmont Report.

Declaration of generative AI in scientific writing
The authors used ChatGPT-5.2 during the preparation of the manuscript to improve the readability and language of the manuscript. After using this tool/service, the authors have reviewed and edited the content as needed and take full responsibility for the content of the published article.

Consent for publication
All authors have given their consent for publication of this manuscript. Any identifiable patient data have been anonymized and permission for publication has been obtained where applicable.

Availability of data and material
The datasets generated and analysed during the current study are available from the corresponding author on reasonable request.

CRediT authorship contribution statement
N Castro: Writing – review & editing, Writing – original draft, Software, Project administration, Methodology, Investigation, Formal analysis, Data curation. I Labiano: Writing – original draft, Supervision, Methodology, Investigation, Formal analysis. M Martínez-Aguillo: Writing – review & editing, Supervision, Conceptualization. AE Huerta: Writing – original draft, Visualization, Methodology, Investigation, Formal analysis. I Morilla: Writing – review & editing, Supervision, Conceptualization. L Teijeira: Resources. D Serrano: Writing – original draft, Investigation. I Caseda: Writing – original draft, Investigation. A Lecumberri: Writing – original draft, Software, Formal analysis. I Amat: Investigation, Data curation. M Zuazo: Investigation. L Chocarro: Investigation. E Blanco: Investigation. D Escors: Methodology, Funding acquisition, Conceptualization. G Kochan: Investigation, Conceptualization. J Fernández Irigoyen: Writing – original draft, Investigation. R Vera: Supervision. A Calvo: Writing – original draft, Supervision, Investigation, Funding acquisition. M Alsina: Writing – review & editing, Writing – original draft, Investigation. H Arasanz: Writing – review & editing, Writing – original draft, Validation, Supervision, Project administration, Methodology, Investigation, Funding acquisition, Conceptualization.

Declaration of competing interest
The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Hugo Arasanz reports financial support was provided by Spanish Society of Medical Oncology. Hugo Arasanz reports financial support was provided by Spanish Association Against Cancer. Alfonso Calvo reports was provided by Spanish Association Against Cancer. David Escors reports financial support was provided by Carlos III Health Institute. Natalia Castro reports a relationship with Roche that includes: speaking and lecture fees. Natalia Castro reports a relationship with Merck Sharp & Dohme UK Ltd that includes: travel reimbursement. Natalia Castro reports a relationship with Merck KGaA that includes: speaking and lecture fees and travel reimbursement. Arturo Lecumberri reports a relationship with Laboratoires Pierre Fabre that includes: speaking and lecture fees and travel reimbursement. Arturo Lecumberri reports a relationship with Merck KGaA that includes: travel reimbursement. Arturo Lecumberri reports a relationship with Novartis Pharmaceuticals that includes: travel reimbursement. Arturo Lecumberri reports a relationship with Merck Sharp & Dohme UK Ltd that includes: travel reimbursement. Ruth Vera reports a relationship with Servier Monde that includes: consulting or advisory. Ruth Vera reports a relationship with Roche that includes: consulting or advisory and speaking and lecture fees. Ruth Vera reports a relationship with Merck Sharp & Dohme UK Ltd that includes: consulting or advisory and speaking and lecture fees. Ruth Vera reports a relationship with Amgen Europe GmbH that includes: speaking and lecture fees. Ruth Vera reports a relationship with AstraZeneca R&D Reims that includes: speaking and lecture fees. Maria Alsina reports a relationship with Servier Monde that includes: consulting or advisory. Maria Alsina reports a relationship with Merck Sharp & Dohme UK Ltd that includes: consulting or advisory. Maria Alsina reports a relationship with Bristol-Myers Squibb Company that includes: consulting or advisory. Maria Alsina reports a relationship with Eli Lilly and Company that includes: consulting or advisory. Maria Alsina reports a relationship with Amgen Europe GmbH that includes: consulting or advisory. Hugo Arasanz reports a relationship with AstraZeneca Pharmaceuticals LP that includes: consulting or advisory. Hugo Arasanz reports a relationship with Roche that includes: consulting or advisory. Hugo Arasanz reports a relationship with Takeda Oncology that includes: speaking and lecture fees. Hugo Arasanz reports a relationship with Bristol-Myers Squibb Company that includes: travel reimbursement. Hugo Arasanz reports a relationship with Merck Sharp & Dohme UK Ltd that includes: travel reimbursement. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.