Lung cancer is one of the most common and deadliest cancers worldwide (1). Surgical resection is the main strategy for the treatment of non-small cell lung cancer (NSCLC). However, even with complete tumor removal, NSCLC patients have poor postoperative prognosis, with a 5-year survival rate of approximately 50% for stage II and 20% for stage III (2). This unfavorable outcome may be attributed to residual tumor cells, tumor micro-metastases, or circulating tumor cells (CTCs) and circulating tumor DNA (ctDNA) causing tumor metastasis or recurrence. Even neoadjuvant chemotherapy only improves the 5-year survival rate by 5%, which is relatively limited (3, 4). Squamous cell carcinoma, compared to other subtypes of NSCLC, lacks effective therapeutic targets and has a worse prognosis. In recent years, our understanding of the role of the immune system in regulating tumor development has significantly increased, leading to a revolution in the field of cancer treatment with the emergence of neoadjuvant immunochemotherapy (NICT). It can effectively activate immune responses and potentially eliminate residual lesions or small metastatic foci (5). NICT in NSCLC has shown significant advantages in terms of short-term outcomes, such as safety, tolerability, and major pathological response (6, 7). However, not all patients benefit from NICT, which also imposes a substantial economic burden. Therefore, identifying the potential beneficiaries of NICT, excluding low responders, reducing healthcare costs, and avoiding overtreatment are urgent issues that need to be addressed.

Pathological complete response (PCR) is a critical indicator for evaluating NICT efficacy in lung squamous cell carcinoma (8).. Previous studies have aimed to identify biomarkers associated with PCR. In the context of esophageal, breast, and colorectal solid tumors, biomarkers such as PD-L1 score, tumor mutation burden (TMB), tumor-infiltrating lymphocytes (TILs), and microsatellite instability (MSI) have been considered closely linked to the effectiveness of NICT. However, the findings from certain clinical trials exhibit inconsistency and even contradictory results (9–11). These discrepancies highlight significant variations in NICT response across different tumor types and individuals, possibly attributable to the intricate nature of the tumor immune microenvironment. Therefore, analyzing the immune microenvironment could serve as a reliable approach for predicting the efficacy of NICT in lung squamous cell carcinoma.

The efficacy of anti-tumor therapy is closely related to the complex tumor immune microenvironment (TME). The complexity of the TME is determined by factors including the quantity, spatial distribution, and function of immune cells. Previous studies have shown that the distance from immune cells to tumor nests is a critical factor affecting prognosis Moreover, the separation among CD20+ B cells, CD4+ T cells, and CD8+ T cells leads to distinguished spatial immune architectures affecting the functional state of immune cells (12). Furthermore, the density of functionally suppressed CD8+ T cells can accurately predict NICT’s treatment response. One possible mechanism is to reverse the immune suppression state of exhausted killer T cells (CD8+PD-L1+ T cells) by using PD-L1 inhibitors, thereby activating immune cell killing functions and achieving pathological remission (13). Hence, analyzing the quantity, functionality, and spatial distribution of immune cells is a viable option for predicting the attainment of PCR in NICT.

With the development of multiplex immunofluorescence technology (mIF), it becomes possible to directly observe the quantity, spatial distribution, and phenotype of immune cells in the microenvironment. By leveraging machine learning, the aforementioned image features can be quantitatively analyzed. This study aims to utilize these approaches to identify and characterize immune cell features within the tumor microenvironment (TME), select the most informative features, and establish a concise and efficient predictive model for assessing the efficacy of neoadjuvant therapy. The findings will serve as a valuable reference for clinical decision-making.

We utilized multiplex immunofluorescence technology and image quantitative analysis techniques to examine the pre-treatment immune microenvironment of lung squamous cell carcinoma patients. Our analysis revealed disparities in the immune microenvironment between the PCR and non-PCR groups, specifically in terms of immune cell quantity, distribution, and function. We identified the most valuable differences for predicting the effect of NICT treatment, namely the density of CD8+ T cells, PD-L1+ T cells, and PD-L1+CD8+T cells within the tumor nest (rather than the stroma). The prediction model established using these differences performed well in both the training and validation sets, providing an accurate and easy-to-implement plan for clinical prediction of NICT efficacy.

Initially, we analyzed the spatial distribution of immune cells and identified their predominant concentration within tumor nests and invasive margins. Likewise, the differences in markers, including CD8+, PD-L1+, Foxp3+, and CD163+, were primarily observed in these regions. Importantly, no statistically significant differences were observed in the stroma. Consequently, we hypothesized that immune cells responsible for immune function were primarily distributed within the tumor rather than the stromal area. In a study investigating neoadjuvant therapy for breast cancer, researchers discovered a notable association between the quantity of immune cells within tumor nests and pathological response, whereas no such correlation was observed within the stroma (16). Similarly, a separate study focusing on gastric cancer reported a significant relationship between the quantity of immune cells infiltrating tumor nests and invasive margins and patient overall survival (OS) (17). Collectively, these findings suggest that immune cells exert their anti-tumor effects primarily within the tumor nests and invasive margins.

Additionally, we performed an analysis of immune cell densities in this study. The PCR group exhibited higher quantities of CD20+ B lymphocytes, CD4+ helper T lymphocytes, and CD8+ killer T lymphocytes, while demonstrating lower levels of Foxp3+ regulatory T cells and CD163+ M2 macrophages. These findings suggest that both humoral immunity and cellular immunity play crucial roles in anti-tumor activity when there is increased immune cell infiltration. Conversely, the presence of regulatory T cells and M2 macrophages hindered the anti-tumor effect within the immune microenvironment. Notably, previous research indicated that a 10% rise in infiltrating immune cells (TILs) in breast cancer reduced the risk of death by 20% (18). Furthermore, a study focusing on T3N0 esophageal squamous cell carcinoma reaffirmed the close correlation between the quantity of TILs and overall survival (OS) within each subgroup (19). Another study illustrated the predictive value of various TILs subpopulations in melanoma for patient response to immunotherapy (20). Moreover, regulatory T cells have the capability to suppress bodily immunity (21), consequently impeding the body’s anti-tumor activity by inhibiting the activation and proliferation of reactive T cells within the body’s microenvironment. Considering the collective findings from prior studies and our own, we have compelling grounds to suspect that the quantity of immune cells and macrophages may represent crucial factors influencing the response to neoadjuvant therapy.

Lastly, we conducted an analysis of cell functions. Additionally, our study reveals higher levels of CD20+ cell density and TLS numbers in the PCR group, although this disparity did not reach statistical significance. Notably, CD20+ B lymphocytes not only contribute to humoral immunity but also form a critical component of tertiary lymphoid structures. Numerous studies underscore the significance of CD20+ B cells and TLS in tumor immunochemotherapy (22), thereby indirectly elucidating the elevated levels of CD20+ B cells and TLS in the PCR group. However, T cells appear to constitute the primary anti-tumor component. The densities of CD8+ cells in the PCR group are significantly greater than those in the other groups, suggesting that a higher quantity of killer T cells corresponds to a more pronounced pathological response.

Our study corroborates previous reports indicating that a higher quantity of effector T cells is associated with pathological complete remission of breast cancer. This association could be attributed to the direct involvement of CD8+ T cells in tumor eradication (23). Furthermore, the density of CD8+PD-L1 depleting killer cells was higher in the PCR group in our study, indicating a greater presence of these depleting T cells exerting an anti-tumor effect through the reactivation of immune checkpoint inhibitors (23). While previous studies suggested that CD8+ killer T cells were the primary anti-tumor cells, recent research has demonstrated that CD4+ T cells are also capable of secreting granase B to eliminate cancer cells and enhancing the killing function of CD8+ T cells through the secretion of interleukin-2 (IL-2) (24). While previous studies suggested that CD8+ killer T cells were the primary anti-tumor cells, recent research has demonstrated that CD4+ T cells are also capable of secreting granase B to eliminate cancer cells and enhancing the killing function of CD8+ T cells through the secretion of interleukin-2 (IL-2) (25). These findings indicate the significant roles of both CD4+ and CD8+ T cells in the anti-tumor process, highlighting their potential for cooperative interactions (26). According to the univariate analysis in this study, CD4+LAG-3+ appeared to be a significant influencing factor for PCR, providing support for this hypothesis. However, not all T cells contribute to immune killing. Immune killing by Foxp3+ regulatory T cells is regulated by negative feedback, and an excessive presence of regulatory T cells inhibits immune killing while promoting tumor escape (27). This study found a high density of Foxp3 and CD3+Foxp3+ in the non-PCR group, providing evidence for the aforementioned perspective. The function of M2 macrophages in restricting immune response and promoting angiogenesis (28) indirectly facilitates immune escape and tumor metastasis. The higher presence of CD163+ macrophages in the non-PCR group in this study aligns with this perspective. In summary, disparities are observed between the PCR group and the non-PCR group in the density, spatial distribution, and function of immune cells, potentially contributing to the occurrence of PCR. Building upon the aforementioned analysis, we employ optimal subset regression and LASSO regression to identify the three most effective variables: CD8+, CD8+PD-L1+, and PD-L1+ for improved prediction of PCR. Subsequently, a PCR prediction model is developed and internally validated. The training set exhibits excellent performance, while there is a noticeable decline in performance on the validation set. This may be due to issues such as overfitting and a limited number of samples. The study is conducted in a single center, and the lack of external validation is a limitation of this research. Nevertheless, compared to traditional TPS > 1 or TPS > 50 scores, the model demonstrates strong performance in improving sensitivity, specificity, and accuracy. This provides a reliable basis for selecting clinical treatment strategies.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving humans were approved by Sichuan Provincial People’s Hospital, University of Electronic Science and Technology of China. The studies were conducted in accordance with the local legislation and institutional requirements. The ethics committee/institutional review board waived the requirement of written informed consent for participation from the participants or the participants’ legal guardians/next of kin because Utilizing medical records and biological specimens obtained from previous clinical diagnoses and treatments.

MX: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Software, Validation, Writing – original draft, Writing – review & editing. LT: Data curation, Formal analysis, Software, Writing – review & editing. TZ: Data curation, Investigation, Methodology, Writing – review & editing. YH: Data curation, Investigation, Writing – review & editing. XL: Conceptualization, Supervision, Validation, Writing – review & editing. QZ: Conceptualization, Data curation, Investigation, Methodology, Project administration, Software, Supervision, Validation, Writing – original draft, Writing – review & editing.