CD49a+ NK cells promote M2 polarization and are associated with poor pathological response in NSCLC Authors: Xin Li, Zelin Zhao, Weihong Zhang, Meng Wang, Siyuan Zhang, Nan Dong, Wenwen Yu, Yang An, Xiubao Ren, Hua Zhao, Yu Zhang Published: 2026 ABSTRACT Natural killer (NK) cells are key for tumor immune defense. Tissue-resident NK subsets often differ from classical CD16+ NKs and can be immunosuppressive. The exact traits and mechanisms of these tissue-resident NKs in the tumor microenvironment (TME) of non-small cell lung cancer (NSCLC) are still unclear. This study aimed to comprehensively analyze NK cells in the NSCLC tumor microenvironment to identify distinct tissue-resident subsets, assess their clinical relevance, and investigate their functional properties and underlying mechanisms in shaping immunosuppression. We identified CD49a+ NK cells as a distinct tissue-resident subset in non-small cell lung cancer surgical specimens. Single-cell RNA sequencing revealed that this subset is associated with poor pathological response to neoadjuvant therapy. The abundance of CD49a+ NK cells negatively correlates with the formation of tertiary lymphoid structures (TLS). These cells exhibit an altered phenotype, characterized by upregulated immune checkpoint genes, downregulated cytotoxicity-related genes, and uniquely elevated expression of CSF-1. This CD49a+ NK cell subset is functionally linked to driving M2 macrophage polarization, and M2 polarization tends to inversely correlate with TLS density. Our findings indicate that CD49a+ NK cells contribute to the induction of M2 macrophage differentiation within the tumor stroma and are unfavorable for TLS formation. Our findings identify a specific CD49a+ tissue-resident NK cell subset that fosters an immunosuppressive microenvironment in NSCLC by impairing TLS formation and driving M2 macrophage polarization, a mechanism that likely contributes to adverse responses to neoadjuvant therapy. Introduction NSCLC constitutes about 85% of lung cancer cases and stands as the leading cancer-related death cause globally1,2. Despite significant improvements in overall NSCLC survival due to therapeutic advances, the clinical prognosis for patients diagnosed with stage IIIA (N2) NSCLC continues to be poor3. Cancer immunotherapy through immune checkpoint blockade (ICB) has become a frontline therapeutic option for advanced NSCLC lacking actionable driver mutations4,5. Giving immunotherapy before surgery has emerged as a beacon of hope for individuals with early-stage lung cancer that can be surgically removed. The advent of immune checkpoint inhibitors has revolutionized the way we treat cancer. Unfortunately, only a small subset of lung cancer patients achieves durable clinical benefit from these therapies4,6–8. The tumor microenvironment (TME) is critically involved in key oncogenic processes, including tumor initiation, advancement, metastatic spread, and the development of therapy resistance9. Among its components, NK cells play a pivotal role in antitumor immunity due to their ability to directly mediate cytotoxicity and secrete immunostimulatory cytokines without the need for prior antigen sensitization10. Evidence has shown that neoadjuvant therapy can promote the expansion and activation of CD16+ NK cells, which positively correlates with pathological response11. However, studies have also demonstrated that tissue-resident NK (trNK) cells exhibit functional impairments, including reduced cytotoxicity, diminished cytokine production, and weakened chemotaxis, displaying features of dysfunction12. Therefore, investigating whether trNK cells are key factors contributing to adverse outcomes in NSCLC neoadjuvant therapy is of great importance. Prior studies indicate that trNK cells in human lung tissue are commonly characterized by the concurrent presence of four specific proteins: CD69, CD49a, CD103 and CXCR613–15. CD69 interacts with S1PR1 (sphingosine-1-phosphate receptor 1) and promotes its internalization/inactivation, rendering lymphocytes insensitive to the S1P gradient. This reduces their ability to egress from tissues or lymphoid organ16.CD49a pairs with β1/CD29 to form VLA-1, a receptor for collagen and laminin that mediates cell adhesion to and localization within the basement membrane and interstitium. In tissues formed following infection—such as in lung-resident memory CD8+ T cells—VLA-1/CD49a contributes to the retention and maintenance of protective memory cells within the tissue and influences local motility and immune surveillance17–19.CD103 is a key adhesion molecule for epithelial localization and tissue residency. It pairs with β7 to form αEβ7, whose canonical ligand is E-cadherin (CDH1), enabling CD103-expressing lymphocytes to anchor near epithelial cells. CD103 is therefore commonly used as a marker of TRM cells20,21. CXCR6 is the receptor for the chemokine CXCL16 and participates in immune cell recruitment, adhesion, and tissue localization. By regulating tissue residency and intra-tissue compartmentalization, CXCR6 influences the distribution and localization of TRM cells across different microenvironments in tissues such as the respiratory tract, thereby affecting local defense capacity22,23. TrNK cells generally exhibit a CD16− phenotype, though significant heterogeneity exists across different tissues. A subset of CD69+CXCR6+ trNK cells, displaying immunoregulatory and exhausted-like phenotypes, CD69+CXCR6+ trNK cells specifically accumulate within the TME of NSCLC. The functional impairment of these cells can be partially restored by PD-1 and CTLA-4 blockade24. In human bronchoalveolar lavage samples, CD49a+CD69+CD103+/− CD56brightCD16− cells are identified as trNK cells with enhanced responsiveness in the lung25. These findings highlight that trNK cells exhibit diverse phenotypes across different tissue types, making it crucial to clarify the specific phenotype of trNK cells in NSCLC for in-depth research. TLSs refer to organized aggregates of immune cells that form within nonlymphoid tissues26. Accumulating evidence indicates that the presence and maturation level of TLS correlate with clinical outcomes in tumors, highlighting their potential as novel biomarkers for prognosis27. However, whether trNK cells are associated with TLS formation remains unclear. In our investigation, we sought to ascertain characteristic trNK cells in NSCLC, evaluated their association with pathological response to neoadjuvant therapy, and investigated whether trNK cells affected the efficacy of neoadjuvant treatment by influencing TLS formation. Furthermore, we sought to explore the potential of trNK cells as predictive biomarkers for neoadjuvant therapy in NSCLC. Methods Tumor tissue, lymph node, normal lung epithelial tissue, and peripheral blood mononuclear cell collection Fresh tumor tissue samples were obtained via surgical resection immediately following the procedure. Tumor tissues, lymph nodes, normal lung epithelial tissues, and peripheral blood mononuclear cells (PBMCs) were collected from 21 patients for flow cytometric analysis. Peripheral blood was collected in tubes containing EDTA as an anticoagulant. The serum was subsequently separated by centrifugation and stored at − 80 °C refrigerator until flow cytometric analysis was performed (Detailed patient information is provided in Table 1). Prior to study commencement, informed consent was obtained from all participants, and the study was approved by the Ethics Committee of Tianjin Medical University Cancer Institute and Hospital. Flow cytometry Human peripheral blood mononuclear cells (PBMCs) and single-cell suspensions derived from fresh tumor specimens and matched adjacent tissues were subjected to flow cytometric analysis. Cells were harvested after a 4-hour stimulation washed with FACS buffer, and subsequently incubated at room temperature for 20 min with the following antibody cocktail for surface staining: anti-human CD45PerCP/Cy5.5Biolegend, Cat#368504anti-human CD3APCBiolegend, Cat#300312anti-human CD3PerCP/Cy5.5Biolegend, Cat #317336anti-human CD16APC-Cy7Biolegend, Cat #302018anti-human CD56PE-Cy7Biolegend, Cat #362510anti-human CD49aPEBiolegend, Cat #328304anti-human CXCR6FITCBiolegend, Cat #356020anti-human CD69PE-CF594BD Biosciences, Cat #562617anti-human CD107APC-Cy7Biolegend, Cat#328630anti-human CD39PerCP/Cy5.5Biolegend, Cat #328218anti-human CD103BV421Biolegend, Cat #350214anti-human IFN-γAPC-Cy7Biolegend, Cat #502530anti-human GZMBPerCP/Cy5.5Biolegend, Cat#396416anti-human PerforinPerCP/Cy5.5Biolegend, Cat #353314anti-human M-CSFAPCBiolegend, Cat #IC2161Aanti-human CD163PEBiolegend, Cat #326506 anti-human CD45PerCP/Cy5.5Biolegend, Cat#368504anti-human CD3APCBiolegend, Cat#300312anti-human CD3PerCP/Cy5.5Biolegend, Cat #317336anti-human CD16APC-Cy7Biolegend, Cat #302018anti-human CD56PE-Cy7Biolegend, Cat #362510anti-human CD49aPEBiolegend, Cat #328304anti-human CXCR6FITCBiolegend, Cat #356020anti-human CD69PE-CF594BD Biosciences, Cat #562617anti-human CD107APC-Cy7Biolegend, Cat#328630anti-human CD39PerCP/Cy5.5Biolegend, Cat #328218anti-human CD103BV421Biolegend, Cat #350214anti-human IFN-γAPC-Cy7Biolegend, Cat #502530anti-human GZMBPerCP/Cy5.5Biolegend, Cat#396416anti-human PerforinPerCP/Cy5.5Biolegend, Cat #353314anti-human M-CSFAPCBiolegend, Cat #IC2161Aanti-human CD163PEBiolegend, Cat #326506 Following two washes with wash buffer, we conducted cellular analysis using a BD Aria II flow cytometer manufactured by BD Biosciences in San Diego, USA. The data we collected were then processed and examined with FlowJo software, version 10.1, which was developed by FlowJo LLC in Ashland, OR, USA. Patients and tumor specimens in the validation cohort Our study enrolled a total of 48 patients with primary lung cancer from the Tianjin Medical University Cancer Hospital between June 2022 and September 2024, including 22 cases who underwent surgical resection without neoadjuvant therapy and 26 cases who received neoadjuvant therapy prior to surgery. Inclusion criteria comprised complete clinical records, standardized postoperative management, confirmed pathological diagnosis. An R0 resection was achieved in all enrolled patients, with neoadjuvant treatment predominantly based on anti-PD-1 therapy(Pembrolizumab, 2 mg per kilogram of body weight, 21 days each cycle, for two cycles before surgical resection, and then followed by two cycles after surgical resection). For multiplex Immunohistochemistry staining, we utilized formalin-fixed, paraffin-embedded (FFPE) tumor tissue samples collected from the cohort of 48 patients (Table 2). This study received approval from the Ethics Committee of Tianjin Medical University Cancer Hospital, with written informed consent obtained from all participants prior to their involvement. Table 1Clinicopathological characteristics of the 21 lung cancer patients n (%).NSCLC patientsNumber(%)Sex Male11(52%) Female10(48%)Age < 655(23.8%) 65–759(42.8%) > 757(33.3%)Race Asian21(100%)Tumor stage I16(76.2%) II2(9.5%) III2(9.5%) IV1(4.7%)Histological type Adenocarcinoma20(95.2%) Squamous cell carcinoma1(4.7%) Table 1Clinicopathological characteristics of the 21 lung cancer patients n (%).NSCLC patientsNumber(%)Sex Male11(52%) Female10(48%)Age < 655(23.8%) 65–759(42.8%) > 757(33.3%)Race Asian21(100%)Tumor stage I16(76.2%) II2(9.5%) III2(9.5%) IV1(4.7%)Histological type Adenocarcinoma20(95.2%) Squamous cell carcinoma1(4.7%) Clinicopathological characteristics of the 21 lung cancer patients n (%). Table 2Baseline characteristics of the 48 lung cancer patients.CharacteristicNeoadjuvant therapy (n = 24)P valuePathological responseMPR(n = 12)Non-MPR(n = 12)Gender, n>0.05 Male78 Female54Median age, years (range)62(44–69)61(53–67)>0.05Histological subtype33>0.05Squamous carcinomaAdenocarcinoma99Clinical stage, n>0.05 I32 II45 III55 Table 2Baseline characteristics of the 48 lung cancer patients.CharacteristicNeoadjuvant therapy (n = 24)P valuePathological responseMPR(n = 12)Non-MPR(n = 12)Gender, n>0.05 Male78 Female54Median age, years (range)62(44–69)61(53–67)>0.05Histological subtype33>0.05Squamous carcinomaAdenocarcinoma99Clinical stage, n>0.05 I32 II45 III55 Baseline characteristics of the 48 lung cancer patients. Single-cell sequencing A total of 12 patients diagnosed with stage IIIA non-small cell lung cancer (NSCLC) were enrolled in the single-cell sequencing analysis, including 4 treatment-naïve patients and 8 patients who received neoadjuvant pembrolizumab plus chemotherapy. Among them, there were 5 patients with adenocarcinoma and 7 with squamous cell carcinoma. The baseline clinicopathological characteristics of these patients are summarized in Table 2. This study was approved by the Ethics Committee of Tianjin Medical University Cancer Institute and Hospital (approval number: bc2020060) and complied with the Declaration of Helsinki and Good Clinical Practice guidelines. Written informed consent was obtained from all patients. For patients enrolled in the neoadjuvant pembrolizumab plus chemotherapy group, none had autoimmune diseases, interstitial lung disease, or a prior history of cancer. No glucocorticoids or immunosuppressive agents were used. None had previously received checkpoint inhibitors or other drugs targeting T-cell costimulatory or immune checkpoint pathways. Patients with acute or chronic hepatitis virus infection or active tuberculosis were also excluded. Enrolled patients received neoadjuvant therapy before surgical resection, consisting of intravenous pembrolizumab (2 mg/kg body weight) on day 1, together with paclitaxel (175 mg/m²) for squamous cell carcinoma or pemetrexed (500 mg/m²) for adenocarcinoma, plus carboplatin (area under the curve 5; 5 mg/mL/min) on day 1, repeated every 21 days for two cycles, followed by two additional cycles after surgical resection. The sequencing data were aligned to the GRCh38 human reference genome and quantified using Cell Ranger 3.1.0. Subsequently, a gene expression matrix was constructed, and Seurat 3.2.1 was employed to filter cell barcodes, removing dead cells and artifacts based on the following criteria: <200 or > 6000 genes detected per cell, UMI count < 1000, or mitochondrial gene content > 10%.TCR-negative cells extracted from NK/NKT populations were subsequently analyzed as NK cells. For detailed information please refer to our team’s previous article28. Co-culture of NK cells and macrophages in vitro NK cells and monocytes were isolated from PBMCs using fluorescence-activated cell sorting (FACS). A549 cells were cultured and expanded in vitro, and the conditioned medium was collected. NK cells were then treated with this A549-conditioned medium, while the control group received NK cells cultured in 1640 medium supplemented with 10% fetal bovine serum. Subsequently, NK cells from both groups were collected and co-cultured with monocytes at a 5:1 ratio. After incubation at 37 °C for 2 days, cells were harvested for flow cytometric analysis. A549-stimulated NK cells were sorted by flow cytometry into CD49a+ NK and CD49a− NK subsets, which were then co-cultured with monocytes according to the above protocol. Subsequently, the cells were collected for flow cytometry analysis. Multiplex immunohistochemistry for tertiary lymphoid structure quantification The Opal 7-Color Manual IHC Kit (PerkinElmer, Cat# NEL811001KT) was employed for multiplex fluorescence staining according to the manufacturer’s protocol. Primary antibodies were used at the following dilutions: mouse anti-human CD56 monoclonal antibody (Cell Signaling Technology, Cat #3576S; 1:400), mouse anti-human CD49a monoclonal antibody (NOVUS, Cat # NBP2-76478; 1:1000), mouse anti-human CD20 monoclonal antibody (Abcam, Cat # ab9475; 1:800), mouse anti-human Bcl-6 monoclonal antibody (NOVUS, Cat # NBP3-07540; 1:200), and recombinant rabbit anti-human CD163 monoclonal antibody (HUABIO, Cat# ET1704-43; 1:2000).Following each round of tyramide signal amplification, slide sections underwent microwave heat treatment. Nuclear counterstaining was performed with DAPI after completion of all antigen labeling procedures. For multispectral imaging, the stained slides were scanned following the manufacturer’s protocol. The Mantra System (PerkinElmer) was utilized to produce im3 image cubes for subsequent processing. Spectral data for all incorporated fluorophores examined via the inForm image analysis tool (v2.4.4 PerkinElmer). From each slide, ten fields showing enriched immune cell populations were selected for image acquisition. Corresponding patient characteristics are detailed in Supplemental Table 1. Whole-slide scanning for TLS quantification was conducted on a TissueFAXSi-plus panoramic scanner (TissueGnostics, Vienna, Austria), equipped with the Pixelink digital color camera and controlled by TissueFAXS acquisition software (v7.0.6245). Multispectral image analysis was subsequently carried out using the StrataQuest software (v7.0.1.165, Tissue Gnostics). TLSs were defined as lymphoid aggregates histologically resembling organized lymphoid structures, specifically containing CD20⁺ B cell zones within the tumor area. TLS density was determined by counting all such aggregates situated within the tumor parenchyma or in direct contact with tumor cells at the invasive margin, and was expressed as the number of TLS per square millimeter (mm²). Statistical analysis Statistical analyses were conducted with GraphPad Prism software (version 9). Group comparisons made via unpaired t-test, paired t-tests were used to analyze flow cytometry data, statistical significance across multiple study groups was evaluated using a one-way ANOVA, with subsequent pairwise comparisons conducted via Tukey’s post hoc analysis. Any P-value exceeding 0.05 was regarded as lacking statistical significance (NS). Results Identification of distinct resident NK cell subsets in NSCLC To identify potential markers of trNK cells in NSCLC patients, we analyzed the proportions of NK cells expressing CD69, CD103, CXCR6, and CD49a in tumor tissues (Tumor), draining lymph nodes (LNs), adjacent normal lung epithelial tissues (N), and peripheral blood mononuclear cells (PBMCs) obtained from 21 patients with NSCLC (Fig. 1A), The patient information is detailed in Table 1. Notably, the CD49a⁺ NK cell subset was present at a significantly higher frequency within the tumor than in PBMC, LNs, and normal lung epithelium (Fig. 1B). The proportion of CD69⁺ NK cells was significantly elevated in tumor tissues relative to PBMC and normal lung epithelium, but showed no significant difference compared to LNs (Fig. 1C). Our results showed that the frequencies of both the CD103⁺ and CXCR6⁺ NK exhibited markedly greater presence in tumor samples compared to those in PBMC and LNs, while no significant difference was observed compared to normal lung epithelium (Fig. 1D,E). Based on these findings, we concluded that CD49a⁺ NK cells likely represented the tissue-resident NK (trNK) cell population in NSCLC tumors. Fig. 1Characterizing iunique tumor-resdent NK cell subpopulations in NSCLC.(A) Schematic of the overall study design. Flow cytometry was performed on CD45+ immune cells isolated from tumor tissue (Tumor), matched adjacent non-cancerous tissue (N), regional draining lymph nodes (LNs), peripheral blood mononuclear cells (PBMC). Figure created with https://biogdp.com/. (B–E) Profiling tissue-residency markers. Expression of CD49a, CD69, CD103, and CXCR6 was compared by flow cytometry among immune cells from tumor (Tumor), draining lymph node (LN), adjacent normal tissue (N), and peripheral blood (PBMC). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 (one-way ANOVA or Student’s t test). Characterizing iunique tumor-resdent NK cell subpopulations in NSCLC. (A) Schematic of the overall study design. Flow cytometry was performed on CD45+ immune cells isolated from tumor tissue (Tumor), matched adjacent non-cancerous tissue (N), regional draining lymph nodes (LNs), peripheral blood mononuclear cells (PBMC). Figure created with https://biogdp.com/. (B–E) Profiling tissue-residency markers. Expression of CD49a, CD69, CD103, and CXCR6 was compared by flow cytometry among immune cells from tumor (Tumor), draining lymph node (LN), adjacent normal tissue (N), and peripheral blood (PBMC). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 (one-way ANOVA or Student’s t test). The abundance of CD49a+ NK cells was negatively correlated with TLS formation To investigate whether trNK cells are associate with TLS formation, we performed multiplex immunohistochemistry staining on tumor tissue sections from 26 patients who received neoadjuvant therapy(non-MPR = 13, MPR = 13) and 22 treatment-naïve patients, the patient information is detailed in Table 2, the first row shows TLS in the three groups, the second row shows the intratumoral stromal CD49a⁺ NK cells, and the third row shows CD49a⁺ NK cells in the peritumoral TLS region (Fig. 2A). First, the relationship between TLS density and different treatment groups was analyzed. Compared with the naive group, the neoadjuvant therapy group exhibited a higher TLS density, which is consistent with previous research findings (Fig. 2B). Furthermore, we compared TLS density between non-MPR and MPR groups among patients receiving neoadjuvant therapy, with the MPR group showing a higher density than the non-MPR group (Fig. 2C). Additionally, in the naive group, non-MPR group, and MPR group, the number of CD49a⁺ NK cells was significantly higher in the non-MPR group than in the other two groups (Fig. 2D). In the neoadjuvant therapy group, the MPR group had a higher number of grade 3 TLSs than the non-MPR group (Fig. 2E). We performed a correlation analysis between CD49a⁺ NK cells and TLS density in 48 patients and found a significant negative correlation between the two groups (Fig. 2F). Fig. 2Multicolor immunohistochemistry reveals a negative association between CD49a+ NK cells and TLS formation. (A) Multiplex immunohistochemistry analysis comparing naïve, non-MPR and MPR groups reveals differences in TLS density, the stromal abundance of CD49a+ NK cells, and the density of CD49a⁺ NK cells in the peritumoral region surrounding TLSs. (B) Comparison of TLS density between naive and Neoadjuvant therapy Groups. (C) Comparison of TLS density between non-MPR and MPR Groups. (D) Comparison of CD49a⁺ NK cell proportion among the naive group, non-MPR group, and MPR group. (E) TLS grading in the non-MPR and MPR groups. (F) Correlation between CD49a+ NK cells and TLS density in all patient samples. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 (one-way ANOVA or Student’s t test). Multicolor immunohistochemistry reveals a negative association between CD49a+ NK cells and TLS formation. (A) Multiplex immunohistochemistry analysis comparing naïve, non-MPR and MPR groups reveals differences in TLS density, the stromal abundance of CD49a+ NK cells, and the density of CD49a⁺ NK cells in the peritumoral region surrounding TLSs. (B) Comparison of TLS density between naive and Neoadjuvant therapy Groups. (C) Comparison of TLS density between non-MPR and MPR Groups. (D) Comparison of CD49a⁺ NK cell proportion among the naive group, non-MPR group, and MPR group. (E) TLS grading in the non-MPR and MPR groups. (F) Correlation between CD49a+ NK cells and TLS density in all patient samples. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 (one-way ANOVA or Student’s t test). CD49a+ NK cells mediate resistance to the pathological effects of neoadjuvant therapy To validate the above findings and clarify the role of CD49a+NK cells in the immune microenvironment, we conducted single-cell sequencing on tumor tissue samples from 8 NSCLC patients who underwent neoadjuvant therapy and 4 treatment-naïve patients (Table 3). Surprisingly, no significant differences were detected among the proportions of NK cells in PBMCs, normal lung epithelium, lymph nodes (LNs), or tumor tissues among the naïve group, the non- Major Pathological Response group (non-MPR), and the Major Pathological Response group (MPR) (Fig. 3A). The proportion of different NK cell subsets was analyzed using a pie chart. In the non-MPR group, the proportion of CD49a⁺ NK cells was significantly elevated. Furthermore, within dual-positive CXCR6+CD49a+ or CD103+CD49a+ (CXCR6⁺ITGA1⁺ or ITGAE⁺ITGA1⁺) or triple-positive CXCR6+CD103+CD49a+ (CXCR6⁺ITGAE⁺ITGA1⁺) subpopulations, any NK subset expressing ITGA1 consistently showed higher abundance in non-MPR group and lower abundance in the MPR group. These findings suggest that the enrichment of the trNK cell population we identified may be indicative of an unfavorable outcome following neoadjuvant therapy. (Fig. 3B). Table 3The basic clinicopathological characteristics of patients in single-cell sequencing cohort.PatientAgeSexSmoking historyPD-L1 (TPS)PathologyRVT%Pathologic responseNaiveP0175MaleYesN/AADN/AN/AP0265FemaleNoN/AADN/AN/AP0373MaleYesN/ASCCN/AN/AP0458MaleNoN/AADN/AN/ANeoadjuvantP0559MaleYes50%SCC10%MPRP0669MaleYes5%SCC95%non-MPRP0751FemaleYes0%SCC0%MPRP0866FemaleNo90%AD80%non-MPRP0967MaleNo3%SCC0%MPRP1062MaleYes30%SCC0%MPRP1146MaleNo60%AD60%non-MPRP1268MaleYes15%SCC95%non-MPRPD-L1, programmed cell death-ligand 1; TPS, tumor proportion score; AD, adenocarcinoma; SCC, squamous cell carcinoma; RVT, residual viable tumor cells; N/A, not applicable; MPR, major pathologic response. Table 3The basic clinicopathological characteristics of patients in single-cell sequencing cohort.PatientAgeSexSmoking historyPD-L1 (TPS)PathologyRVT%Pathologic responseNaiveP0175MaleYesN/AADN/AN/AP0265FemaleNoN/AADN/AN/AP0373MaleYesN/ASCCN/AN/AP0458MaleNoN/AADN/AN/ANeoadjuvantP0559MaleYes50%SCC10%MPRP0669MaleYes5%SCC95%non-MPRP0751FemaleYes0%SCC0%MPRP0866FemaleNo90%AD80%non-MPRP0967MaleNo3%SCC0%MPRP1062MaleYes30%SCC0%MPRP1146MaleNo60%AD60%non-MPRP1268MaleYes15%SCC95%non-MPRPD-L1, programmed cell death-ligand 1; TPS, tumor proportion score; AD, adenocarcinoma; SCC, squamous cell carcinoma; RVT, residual viable tumor cells; N/A, not applicable; MPR, major pathologic response. The basic clinicopathological characteristics of patients in single-cell sequencing cohort. PD-L1, programmed cell death-ligand 1; TPS, tumor proportion score; AD, adenocarcinoma; SCC, squamous cell carcinoma; RVT, residual viable tumor cells; N/A, not applicable; MPR, major pathologic response. Fig. 3Single-cell sequencing reveals a correlation between CD49a+ NK cells and poor pathological response in neoadjuvant therapy. (A) Comparison of NK cell proportions across different sites among the Naive, non-MPR, and MPR groups via single-cell sequencing. (B) Pie chart showing the proportion of each resident subpopulation in the naive, MPR, and non-MPR groups. Single-cell sequencing reveals a correlation between CD49a+ NK cells and poor pathological response in neoadjuvant therapy. (A) Comparison of NK cell proportions across different sites among the Naive, non-MPR, and MPR groups via single-cell sequencing. (B) Pie chart showing the proportion of each resident subpopulation in the naive, MPR, and non-MPR groups. CD49a+ NK cells exhibit elevated immune checkpoint genes, reduced cytotoxicity, and high CSF-1 expression Currently, the mechanism by which CD49a⁺NK cells exert immunosuppressive effects in NSCLC tissues remains unclear. Through analysis of cytotoxic profiles, cytokine expression, and immune checkpoint markers in NK cell subsets from single cell RNA-seq data( 8 NSCLC patients who underwent neoadjuvant therapy and 4 treatment-naïve patients, the patient cohort and results are the same as in 3.3; see Table 3 for details), we found that compared to other subsets, CD49a⁺ NK cells showed diminished cytotoxic gene (PRF1) levels and elevated immune checkpoint expression CTLA4, TIGIT, LAYN, PDCD1(PD-1), TGFB1,HAVCR2(Tim-3), ENTPD1,CSF-1,ZNF683(Hobit), TNFRSF18,LAG3—a wide range of genes have been linked to NK cell exhaustion29,30. These features likely contribute to the immunosuppressive phenotype of the CD49a⁺ NK cells. Notably, CD49a⁺ cells also significantly expressed high levels of CSF-1 and the recognized T cell exhaustion marker ENTPD1 (Fig. 4A)31. To validate our findings, we randomly selected four patients with NSCLC and analyzed their tumor tissues by flow cytometry. Through flow cytometry analysis, we confirmed that the CD49a+ NK cell exhibited significantly downregulated expression of cytotoxic molecules (GZMB, Perforin, IFN-γ) compared to the CD49a− NK cell subset. However, the CD49a⁺ NK cells upregulate M-CSF and CD39, which is consistent with the findings from single-cell RNA sequencing. (Fig. 4B). Fig. 4Characterizing the function of CD49a+ NK cells. (A) Functional gene analysis of CD49a+ NK cells from single cell RNA-seq data. (B) Flow cytometric validation of functional gene expression in CD49a+ NK cells. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 (one-way ANOVA or Student’s t test). Characterizing the function of CD49a+ NK cells. (A) Functional gene analysis of CD49a+ NK cells from single cell RNA-seq data. (B) Flow cytometric validation of functional gene expression in CD49a+ NK cells. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 (one-way ANOVA or Student’s t test). In vitro validation of CD49a+ NK cells in promoting M2 macrophage polarization To investigate the mechanism by which CD49a⁺ NK cells influence macrophages, we established an in vitro co-culture system. NK cells were isolated from peripheral blood of three healthy donors and first incubated with A549 cell-conditioned medium for 24 h, after which they were collected. These preconditioned NK cells were then co-cultured with autologous monocytes at a 5:1 ratio. Control NK cells were prepared by incubation in standard culture medium, followed by the same co-culture procedure with monocytes. After two days of co-culture with monocytes, we observed that NK cells post-stimulated with A549-conditioned medium showed upregulated CD49a expression (Fig. 5A). Further flow cytometry analysis revealed that, upon A549-conditioned medium stimulation, CD49a⁺ NK cells secreted significantly higher levels of CD39 than CD49a⁻ NK cells, and a similar trend was observed for M-CSF. In contrast, CD49a⁻ NK cells exhibited markedly higher expression of cytotoxic markers, including CD107a, GZMB, Perforin, and IFN-γ (Fig. 5B). Moreover, the proportion of M2 macrophages was significantly increased in the co-culture system (Fig. 5C). All of the above results were obtained from three technical replicates. To further determine the specificity of tumor-induced CD49a+ NK cells in promoting M2 polarization, we sorted CD49a+ NK cells stimulated with A549 by flow cytometry, with the remaining cells designated as CD49a− NK cells. These two populations were then co-incubated with M0 macrophages. We compared the levels of M-CSF secreted by NK cells and the proportion of M2 macrophages in the co-culture system among three groups: the unstimulated A549 control group (CONTROL), the post-stimulated CD49a+ NK group, and the post-stimulated CD49a− NK group. The post-stimulated CD49a+ NK group exhibited a stronger ability to secrete M-CSF and to induce the polarization of M0 macrophages into the M2 phenotype (Fig. 5D). All the above results were obtained from three technical replicates. These findings further validate our sequencing data. These findings further validate our sequencing data. Fig. 5CD49a+ NK cells promote M2 macrophage polarization in vitro. (A) Flow cytometric analysis of the proportion of CD49a+ NK cells. (B) Flow cytometry was used to validate the secretion of cytotoxic factors and functional proteins by CD49a-NK and CD49a+NK cells. (C) Flow cytometric analysis of the M2-macrophages proportion in the two groups. (D) Flow cytometric analysis of M-CSF secretion by NK cells and the proportion of M2 macrophages across the three groups. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 (one-way ANOVA or Student’s t test). CD49a+ NK cells promote M2 macrophage polarization in vitro. (A) Flow cytometric analysis of the proportion of CD49a+ NK cells. (B) Flow cytometry was used to validate the secretion of cytotoxic factors and functional proteins by CD49a-NK and CD49a+NK cells. (C) Flow cytometric analysis of the M2-macrophages proportion in the two groups. (D) Flow cytometric analysis of M-CSF secretion by NK cells and the proportion of M2 macrophages across the three groups. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 (one-way ANOVA or Student’s t test). Multicolor immunohistochemistry validated that CD49a+ NK cells facilitate the generation of M2 macrophages M-CSF is known to be a key mediator in the M0 to M2 macrophage polarization process. Therefore, we examined the correlation between the quantities of CD49a⁺ NK cells and M2 macrophages. We performed further multiplex immunohistochemical staining on tumor tissues from the naive, non-MPR, and MPR groups after neoadjuvant therapy (The patient cohort is the same as that in Result 3.2, detailed patient cohort information is provided in Method 2. 3 and Table 2). The TLS size in the MPR group was significantly greater than that in the other two groups. (Fig. 6A). Analysis revealed that in the non-MPR group, which exhibits higher CD49a⁺ NK cell infiltration, the tumor microenvironment (TME) contained a greater number of M2 macrophages (Fig. 6B). Moreover, these CD49a⁺ NK cells and M2 macrophages were often spatially adjacent, and their quantities showed a trend toward positive correlation (Fig. 6C). Furthermore, in the non-MPR group, the data showed a negative association between the abundance of M2 macrophages and the density of TLS (Fig. 6D). Fig. 6Multicolor immunohistochemistry revealed that CD49a+ NK cells drive M2 macrophage polarization. (A) Multiplex immunohistochemistry analysis comparing naïve, non-MPR and MPR groups reveals differential densities of CD49a+ NK cells and CD163+ M2 macrophages. (B) Comparison of CD163+ M2 macrophage density among the naïve, non-MPR, and MPR groups. (C) Correlation between the proportion of CD49a+ NK cells and M2 macrophages. (D) Correlation between TLS density and M2 macrophages proportion. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 (one-way ANOVA or Student’s t test). Correlations were evaluated using Spearman’s rank correlation analysis. Multicolor immunohistochemistry revealed that CD49a+ NK cells drive M2 macrophage polarization. (A) Multiplex immunohistochemistry analysis comparing naïve, non-MPR and MPR groups reveals differential densities of CD49a+ NK cells and CD163+ M2 macrophages. (B) Comparison of CD163+ M2 macrophage density among the naïve, non-MPR, and MPR groups. (C) Correlation between the proportion of CD49a+ NK cells and M2 macrophages. (D) Correlation between TLS density and M2 macrophages proportion. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 (one-way ANOVA or Student’s t test). Correlations were evaluated using Spearman’s rank correlation analysis. Discussion Over the last twenty years, the introduction of molecularly targeted agents and immunotherapeutic approaches has substantially enhanced the clinical prognosis of patients with NSCLC. Nevertheless, most late-stage NSCLC patients acquire treatment resistance and subsequently encounter disease advancement32,33. The emergence of drug resistance is intimately connected to immunosuppressive elements thriving in the TME, including but not limited to myeloid-derived suppressor cells (MDSCs), M2 macrophages, regulatory T cells (Tregs), regulatory B cells (Bregs), cancer-associated fibroblasts (CAFs), and exhausted T cells34–39. Within the TME, NK cells undergo a significant receptor reprogramming characterized by heightened inhibitory signaling coupled with diminished activating receptor expression. This transformation gives rise to dysfunctional NK cell subsets, notably exhausted TIM-3⁺ populations and tissue-resident CD49a⁺NK cells40. We found that all NK cell populations expressing ITGA1 (regardless of co-expression with CXCR6 or ITGAE) consistently exhibited higher abundance in the non-MPR group and lower abundance in the MPR group. These results indicate that the enrichment of this identified trNK cell subset may serve as a potential indicator of poor response to neoadjuvant therapy. CD49a⁺ NK cells in certain TME commonly demonstrate diminished cytotoxic capacity and inconsistent inflammatory cytokine secretion, and are associated with poorer clinical prognosis41–43. Across multiple murine tumor models, CD49a is highly expressed on tumor-infiltrating NK cells. These CD49a⁺ NK cells display a more exhausted phenotype than CD49a⁻ cells. Importantly, genetic deletion or antibody-mediated blockade of CD49a on NK cells delays tumor growth and extends survival, an effect underpinned by the essential role of NK cells in antitumor immunity. Furthermore, monoclonal antibody inhibition of CD49a alone curbs tumor progression, and its combination with anti–PD-L1 therapy synergistically enhances antitumor efficacy44. Neoadjuvant chemoimmunotherapy promotes the formation of TLSs within resected NSCLC tissues. Increased TLS density is associated with favorable treatment responses, consistent with studies demonstrating that the presence of TLSs and B cells was significantly associated with extended survival and improved therapeutic outcomes28,45,46. In our study, multiplex immunohistochemistry analysis revealed that the non-MPR group harbored a greater abundance of CD49a⁺ NK cells than the MPR and naïve groups. In contrast, the MPR group exhibited a higher TLS abundance compared to the naïve and non-MPR groups. A significant negative correlation was observed between these two parameters, suggesting that CD49a⁺ NK cells may act as a negative factor in TLS formation. To elucidate the mechanism by which CD49a⁺ cells exert immunosuppressive effects, we dove deep into single-cell RNA sequencing data harvested from non-small cell lung cancer patient tumor samples, specifically comparing profiles of patients who had undergone neoadjuvant therapy. Our investigation zeroed in on contrasting characteristics between CD49a⁻ and CD49a⁺ natural killer cell populations. Notably, in addition to elevated expression of exhausted molecules, the CD49a⁺ cell subset showed significantly increased expression of CSF‑1 and ENTPD1. CSF‑1 is produced by fibroblasts, tumor cells, and other sources. It exists in multiple forms, including membrane‑bound, secreted glycoprotein, and proteoglycan variants47. CSF-1R is currently recognized as the exclusive receptor for CSF-1, and the signaling pathway mediated by CSF1/CSF1R plays a vital role in the differentiation and recruitment of mononuclear phagocytes, notably M2 macrophages.M2 macrophages/tumor‑associated macrophages facilitate neoplastic progression, vascularization, invasion, dissemination, and therapeutic resistance48,49. M2 macrophages promote tumor cell proliferation and survival through the secretion of multiple cytokines, such as Epidermal growth factor, PDGF, TGF-β1, HGF, and EGFR ligands, and bFGF. Among these, EGFR family ligands are particularly significant in tumor development, especially in breast and lung cancers50,51. Subsequently, we validated the difference in M2 macrophage counts among the naïve, non-MPR and MPR groups using multiplex fluorescence staining, revealing a significantly higher number of M2 macrophages in the non-MPR group compared to the naïve group and MPR group. The number of M2 macrophages was low in the naive group, which may be associated with the lower inflammatory response in this group .Further investigation into the relationship between CD49a⁺ NK cells and M2 macrophages demonstrated a significant positive correlation. The data indicate that in the NSCLC TME, CD49a⁺ NK cells drive immunosuppression through M2 polarization mediated by CSF-1. In summary, this study identified a unique population of CD49a⁺ trNK cells in NSCLC, which was associated with poor response to neoadjuvant therapy. These cells exhibited an immunosuppressive phenotype, characterized by reduced cytotoxicity, elevated expression of immune checkpoint molecules and CSF-1, and functionally promoted M2-like macrophage polarization while being correlated with diminished TLS formation. Collectively, CD49a⁺ trNK cells foster an immunosuppressive tumor microenvironment that is associated with a poor prognosis to neoadjuvant therapy in NSCLC.