Introduction

Gastric cancer (GC) ranks the fifth most prevalent cancer and the fifth leading cause of cancer death worldwide in 2022 [1]. The therapeutic landscape in advanced GC remains challenging. Typical chemotherapy regimens, including platinum and fluoropyrimidine-based combinations, offer only modest improvements in overall survival and come with a range of potential adverse events that limit their clinical utility [2]. In addition, targeted agents such as human epidermal growth factor receptor-2 (HER-2) inhibitors (e.g. trastuzumab) and immune checkpoint inhibitors can only show promise in certain subsets of patients because of the molecular and phenotypical heterogeneity of GC, and response rates and durability remain suboptimal for many individuals with advanced GC [3,4,5,6]. Therefore, there is an unmet clinical need for more effective novel therapeutic approaches and better predictive biomarkers.

Claudin18.2 (CLDN18.2), a tight-junction molecule highly expressed in gastrointestinal cancers, is an appealing target for cancer therapy in GC [7]. Two phase III trials have demonstrated that a monoclonal antibody (zolbetuximab) that targets CLDN18.2 can significantly improve survival of patients with advanced GC [8, 9]. We also found CLDN18.2-specific chimeric antigen receptor (CAR)-T cells (CT041) has achieved promising results with a remarkable overall response rate of 54.9% in advanced CLDN18.2-positive GC [10, 11]. However, the applicability of CLDN18.2-specific CAR-T cell treatment remains limited due to the variable efficacy. Furthermore, the production costs for typical CAR-T cells exceed $500,000 per patient [12]. Therefore, it is imperative to identify reliable biomarkers that can discern which patients are more likely to derive therapeutic benefit from CAR-T cell treatment.

Previous studies have demonstrated the prognostic value of the neutrophil-to-lymphocyte ratio (NLR) in multiple cancer types [13]. However, the prognostic value of NLR has not been defined for CAR-T cell treatment in solid tumors. Moreover, these studies mainly focused on correlation analysis and the fundamental molecular mechanism by which increased circulating neutrophils contribute to cancer progression after CAR-T cell treatment  remains unclear.

The objective of this study was to evaluate the prognostic value of the NLR in advanced GC patients treated with CLDN18.2-specific CAR-T cells and to delineate the potential molecular mechanisms of CAR-T cell treatment resistance contributed by circulating neutrophils through single-cell RNA sequencing (scRNA-seq) analysis.

Methods

Patients

We retrospectively analyzed clinical and laboratory data from the phase I trial (NCT03874897) evaluating the safety and efficacy of CLDN18.2-specific CAR-T cell treatment (CT041) in patients with previously treated, CLDN18.2-positive gastrointestinal cancers. The detailed study design and clinical outcomes of this trial have been previously reported [10, 11]. The main inclusion criteria included: histologically confirmed GC who have failed at least one line of treatment; tumor tissue is positive for CLDN18.2 (expression intensity ≥2+ and positive tumor cell rate ≥40%). The informed consent was obtained from all patients. The study was approved by the ethics committee of Peking University Cancer Hospital (2018YJZ75).

Clinical data

Demographic, clinical, pathological information and peripheral blood cell counts (e.g. white blood cells, neutrophils, and lymphocytes) were obtained from the electronic data capture system of each patient at baseline (within five days before the preconditioning chemotherapy). NLR was calculated as the absolute count of neutrophils divided by the absolute count of lymphocytes. The cutoff points were set at three, in accordance with previous published studies, for the purposes of simplicity and ease of clinical use [14].

Outcomes

The efficacy outcomes included the objective response rate (ORR), disease control rate (DCR), progression-free survival (PFS), and overall survival (OS). PFS was defined as the time from CAR-T cell infusion to tumor progression or death. OS was defined as the time from CAR-T cell infusion to death. ORR was defined as the proportion of patients achieving complete response (CR) or partial response (PR), and DCR was defined as the proportion of patients achieving CR, PR, or stable disease (SD). The anti-tumor efficacy was assessed according to Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1.

Single cell sequencing

We collected peripheral blood samples from five patients in this clinical trial at baseline and at three and seven days after the infusion of CT041, respectively [15]. The scRNA-seq libraries were constructed in accordance with the protocol outlined in the GEXSCOPE Library Kit [16]. Individual libraries were diluted, pooled, and sequenced on an Illumina NovaSeq 6000 with 150 bp paired-end reads. The raw reads from the scRNA-seq were processed using the CeleScope (v1.9.0) pipeline. We used STAR (v2.6.1) to map the reads to the reference genome GRCh38, and used featureCounts (v2.0.1) to generate gene expression matrix files.

We used Seurat (v3.1.2) to quality control, dimensionality reduction, and clustering. The gene expression matrix was subjected to normalization and scaling using the NormalizeData and ScaleData functions. The top 2000 variable genes were selected for principal component analysis (PCA) using the FindVariableFeatures function. Cell clusters were visualized using t-distributed stochastic neighbor embedding (t-SNE) with the Seurat functions RunTSNE.

Differentially expressed gene analysis

We used the Seurat FindMarkers function in reference to the Wilcoxon rank sum test with default parameters to obtain significantly differentially expressed genes (DEGs). We selected the genes expressed at least 10% of the cells in both the compared groups and with an average log (fold change) value exceeding 0.25 as DEGs. The Bonferroni correction was employed to compute the adjusted P value. We conducted Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses by using clusterProfiler v4.0 to investigate the potential functions of subclusters [17].

Cell-type recognition and cell-cell interaction analysis

We identity different cell types of each cluster with the expression of canonical markers found in the DEGs using the SynEcoSys database that contained all cell-type features from PanglaoDB, CellMakerDB, and published literature (Table S1). For cluster annotation, the frequency of each cell type was calculated for each cluster, and the cell type with the highest frequency was selected as the cluster identity [18].

Cellphone DB (v2.1.0) was employed to predict cell-cell interactions between different cell types. Pairs of predicted interactions were deemed significant and depicted in a heatmap plot when the P value was less than 0.05 and the average log expression was greater than 0.1.

Statistical analysis

Continuous variables were assessed using t test, and categorical variables were analyzed with Chi-squared test. The Kaplan-Meier method was employed to estimate the OS and PFS survival curves, and the log-rank test was utilized to compare differences. The Cox proportional hazards regression model was utilized to identify the independent prognostic factors for survival, and variables (P < 0.1) from the univariate analysis entered into the multivariate analysis. A P value of less than 0.05 was considered statistically significant. GraphPad Prism and R software were used for statistical analysis.

Results

Elevated peripheral blood NLR was associated with worse prognosis

The baseline characteristics of the 72 advanced GC patients included in the study are presented in Table 1. The mean age of patients was 47 years, and a higher proportion of male patients (66% vs. 32%, P = 0.005) were observed among those with peripheral NLR ≥ 3. The majority of patients (90%) received a dose of 2.5 × 108 CAR-T cells. All patients had advanced and pretreated diseases, and 47% had more than three metastatic sites. The ORR for all patients with advanced GC was 44.4%, and the DCR was 94.4%. The median follow-up period was 31.7 months. The median PFS of all patients was 5.3 months (95% confidence interval [CI], 3.9–6.7), and the median OS was 8.1 months (95% CI, 5.9–10.3).

Table 1 Patient demographics and baseline characteristics.
Full size table

Patients with NLR ≥ 3 had a significantly lower ORR than those with NLR < 3 (34.2% vs.55.9%, P < 0.001) (Fig. 1a). No significant difference was observed in the DCR between the NLR ≥ 3 and NLR < 3 groups (94.7% vs. 94.1%, P = 0.909). Patients with NLR ≥ 3 exhibited a significantly shorter median PFS (3.6 vs. 8.0 months, P < 0.001, Fig. 1b) and OS (5.6 vs. 13.8 months, P < 0.001, Fig. 1c) compared to those with NLR < 3. The univariate analysis demonstrated that only NLR ≥ 3 had statistically significant impaired impact on both PFS (hazard ratio [HR] 2.59, 95%CI 1.56–4.29) and OS (HR 2.44, 95%CI 1.47–4.05) (Fig. 2). The multivariate analysis demonstrated that both NLR ≥ 3 and the presence of a greater number of previous lines of therapy were independent factors of poor prognosis for both PFS (HR 2.36, 95%CI 1.31–4.26) and OS (HR 3.64, 95%CI 1.93–6.85) (Fig. 2). In addition, patients with NLR ≥ 3 had a significantly higher prevalence of grade ≥ 2 cytokine release syndrome (CRS) than those with NLR < 3 (66.7% vs. 33.3%, P = 0.008).

Fig. 1: Association between neutrophil to lymphocyte ratio (NLR) and the clinical outcomes of advanced gastric cancer patients receiving CLDN18.2-specific CAR-T cell therapy.
figure 1

a Objective response rate; b survival curves of progression-free survival; and c overall survival.

Full size image
Fig. 2: Forest plot of univariate and multivariate analysis of clinicopathological characteristics associated with survival.
figure 2

a Progression-free survival; and b overall survival.

Full size image

Circulating neutrophil subclusters identified by single-cell RNA sequencing were related to CAR-T cell treatment resistance

A total of 173,495 cells were identified from 15 peripheral blood samples obtained from five patients. A total of nine distinct cell types were identified using established marker genes, including T cells, B cells, plasma cells, basophils, platelets, erythrocytes, macrophages (MPs), plasmacytoid dendritic cells (pDCs), and neutrophils (Fig. 3a, Table S1).

Fig. 3: Cellular profiling in peripheral blood samples.
figure 3

a Cellular constitution of peripheral blood samples in t-distributed stochastic neighbor embedding (t-SNE); b peripheral blood different cellular proportion; and c cellular constitution change at baseline and at three and seven days after CAR-T cell infusion in progression disease (PD) and partial response (PR) group.

Full size image

According to the RECIST 1.1 criteria, one patient achieved PR, and four patients demonstrated PD after CAR-T cell treatment. The proportion of neutrophils (91.0% vs. 90.7%, P = 0.277) and T cells (7.3% vs. 7.0%, P = 0.283) was comparable between the PD and PR groups at baseline (Fig. 3b). However, the proportions of neutrophils were significantly higher in the PD group compared with the PR group following CAR-T infusion at day 3 (96.8% vs. 62.0%, P < 0.001) and day 7 (40.1% vs. 3.5%, P < 0.001) after CAR-T cell infusion (Fig. 3b, c).

We further subdivided neutrophils into five subclusters using unsupervised clustering analysis (Fig. 4a). The proportion of neutrophil subcluster-3 (NE-3) was significantly higher in the PD group than in the PR group at baseline (6.7% vs. 1.9%, P < 0.001), day 3 (18.4% vs. 7.0%, P < 0.001), and day 7 (20.6% vs. 12.4%, P < 0.001) after CAR-T cell infusion, indicating NE-3 may be related to CAR-T cell treatment resistance (Fig. 4b). In contrast, the proportions of the other four subclusters were similar at all time points between the PD and PR groups and remained relatively unchanged after CAR-T cell treatment. High expression genes including MMP9, CST7, CD177, PADI4, LCN2, LTF, RETN, MMP8, and CAMP were identified as DEGs in NE-3 (Figs. 4c and S1).

Fig. 4: Cellular subcluster analysis of neutrophils in peripheral blood samples.
figure 4

a Cellular subcluster of neutrophils in peripheral blood samples; b peripheral blood different neutrophils subclusters proportion change at baseline and at three and seven days after CAR-T cell infusion in progression disease (PD) and partial response (PR) group; c heatmap of differentially expressed genes (DEGs) of neutrophils subclusters; d enrichment analysis of high-expression differentially expressed genes (DEGs) of neutrophils cluster 3 (NE-3) in biological processes, molecular functions (MF), cellular components (CC), and signaling pathways in GO and KEGG analysis.

Full size image

The biological functions of the high-expression DEGs of NE-3 were analyzed by enrichment analysis (Fig. 4d). The GO analysis revealed that the biological processes (BP) of NE-3 were associated with the response to foreign entities such as bacteria, and fungus. For molecular functions (MF), the genes were found to be enriched in the binding of cadherin, serine-type peptidase, and lipopolysaccharide. For cellular components (CC), NE-3 was found to be enriched in granule and vesicle lumen. The KEGG enrichment analysis revealed that the interleukin-17 (IL-17) signaling pathway was enriched in NE-3.

To further investigate the characteristics of neutrophils, we estimated the pseudotime trajectory of different neutrophil subclusters. NE-3 exhibited an intermediate state, whereas NE-1/2 and NE-4/5 were positioned at the two ends of the trajectory, respectively (Fig. S2A, B).

Circulating neutrophils suppressed naive T cells attributed to treatment resistance

The maximum concentration of CAR expansion copies after the initial CAR-T infusion was numerically lower in patients with an NLR ≥ 3 than those with an NLR < 3 (median: 4354 vs. 5382 copies per microgram of genomic DNA, P = 0.504). We further analyzed cellular interactions between different cell types in both the PD and PR groups. The number of predicted cellular interactions between neutrophils and T cells was higher in the PD group than in the PR group (Fig. 5a, b).

Fig. 5: Cellular interaction analysis among different cell types in peripheral blood.
figure 5

a Heatmap of cellular interactions in progression disease (PD) group and b partial response (PR) group; c different T cell subclusters constitution in t-distributed stochastic neighbor embedding (t-SNE); and d bubble plot of ligand-receptor pairs between neutrophil subclusters (ligand) and T cell subclusters (receptor).

Full size image

To investigate the detailed cellular interactions between neutrophil and T cell subclusters, we subdivided all T cells in peripheral blood into five subclusters, including naive T cells, CD8+ effector T cells, regulatory T cells, T helper cells, and natural killer cells (Fig. 5c). The proportion of naive T cells in the peripheral blood decreased after CAR-T cell treatment. Furthermore, this proportion was lower in the PD group than in the PR group at all time points (Fig. S3). A further comparison of the cellular interactions of the different subclusters of neutrophils and T cells was conducted between the PD and PR groups. The PD group demonstrated a greater number of cellular interactions between all neutrophil subclusters (particularly NE-3) and naive T cells, regulatory T cells, and T helper cells in comparison to the PR group (Fig. S4A, B).

Based on the results, we assumed that CAR-T treatment resistance was related to the suppression of naive T cells caused by circulating neutrophils. Consequently, we identified potential ligand-receptor pairs between different subclusters of neutrophils (ligand) and T cells (receptor). Elevated levels of immunosuppressive signaling between naive T cells and neutrophils were observed, including the SELL-SELPLG and CD58-CD2 pairs (Fig. 5d).

Discussion

The study demonstrated that elevated NLR in the peripheral blood was correlated with worse survival outcomes in advanced GC patients receiving CLDN18.2-specific CAR-T cell treatment. In addition, through scRNA-seq, we revealed the presence of a specific neutrophil subcluster (NE-3) was related to disease progression after CAR-T cell infusion. The NE-3 subcluster is characterized by the expression of pro-tumoral factors like MMP-9 and the enrichment in the IL-17 signaling pathway. We also found that these neutrophils could suppress the function of naive T cells via immunosuppressive pathways, which ultimately contributed to the development of CAR-T cell treatment resistance.

Our previous phase I trial (NCT03874897) yielded encouraging efficacy outcomes for CLDN18.2-specific CAR-T cell treatment in patients with advanced GC. However, it is crucial to recognize that not all patients can benefit equally from this treatment [11]. NLR has emerged as a robust and easily accessible biomarker reflecting systemic inflammation (e.g. infections, steroid treatment, stress), which is increasingly recognized for its prognostic value in various cancers [19, 20]. It has been established that NLR is a readily available prognostic biomarker in advanced GC and elevated NLR is generally associated with a poorer response to chemotherapy and/or immunotherapy [13, 19, 21]. However, regarding CAR-T cell treatment, the prognostic significance of NLR in advanced GC is currently being explored. In the present study, we observed that baseline NLR in peripheral blood plays a significant role as a prognostic indicator among advanced GC patients treated with CLDN18.2-specific CAR-T cells. Furthermore, through scRNA-seq analysis, we found that proportion of circulating neutrophils, particularly a subcluster of neutrophils (NE-3), was significantly higher in the PD group than in the PR group before and after CAR-T treatment, indicating that elevated circulating neutrophils could diminish the odds of a positive response to cell therapy. Prior research has demonstrated that patients with elevated NLRs tend to exhibit greater infiltration of tumor-associated neutrophils (TANs) within their tumor tissues [22, 23]. TANs have been shown to facilitate an environment conducive to cancer progression, including through the promotion of angiogenesis, the suppression of immune surveillance, and the remodeling of the extracellular matrix through the release of vascular endothelial growth factors, matrix metalloproteinases, and other immunosuppressive cytokines [22, 24, 25]. In addition, we discovered that NE-3 exhibited high expression of pro-tumor factors (e.g., MMP-9) and was enriched in the IL-17 signaling pathway, indicating a potential correlation between neutrophils and the establishment of a pro-tumor microenvironment. MMP9, secreted by neutrophils, has been demonstrated to play a crucial role in establishing a pro-tumorigenic milieu conducive to tumor initiation, progression, and metastasis [26, 27]. Yazawa et al. observed that patients with elevated NLR exhibited elevated IL-17 production in gastrointestinal cancers, which could further lead to lymphocyte proliferation suppression [28]. Furthermore, IL-17 has been demonstrated to induce neutrophil extracellular traps, which effectively suppress T-cell immunity and mediate resistance to immunotherapy in pancreatic cancer [29]. Therefore, NLR can serve as a measurable biomarker in clinical practice to stratify patients who may benefit from CAR-T cell treatment, underscoring the necessity for further investigation into specific neutrophil-targeted interventions to enhance the efficacy of CAR-T cell treatment in advanced GC.

In this study, we observed close interactions among subclusters of circulating neutrophils and T cells, which could potentially contribute to disease progression after CAR-T cell treatment. The cell-cell network analysis and T cell subclusters division indicated that neutrophils play a prominent role in exacerbating this suppression of T cells in circulation, especially naive T cells. Previous studies have indicated that neutrophils can suppress the immune activity of lymphocytes by producing chemokines and cytokines [30]. For instance, in liver cancer, increased circulating neutrophils have been shown to dampen other immune cells with potential antitumor effects such as NK cells and lymphocytes [31]. Kargl et al. reported that there was an inverse association between neutrophils and cytotoxic T cells in lung cancer tissues, and the ratio of these two cell types was also associated with anti-PD-1 immunotherapy response [32]. Concurrently, suppressed peripheral blood T cells could mirror a reduction in the infiltration of T cells into tumors, which would compromise the anti-tumor response. In addition, although there is no direct evidence that neutrophils can inhibit CAR-T cell expansion, a retrospective study also found that increased neutrophil counts were associated with lower CAR-T cell expansion [33]. These findings not only refine our understanding of the complex immunological dynamics following CAR-T cell infusion but also reveal potential promises for exploring combinatory approaches that integrate naive T cell enhancement with neutrophil suppression strategies.

There are several limitations in our study. First, the sample size for peripheral blood scRNA-seq analysis, particularly the PR group, is limited, which raises concerns about the potential for selection bias and the generalizability of the findings. Second, the analysis of TANs in GC tissues was not included, which limited the comprehensive understanding of how neutrophils within the tumor microenvironment contribute to CAR-T cell treatment resistance. Third, the duration of follow-up for the analysis of neutrophil dynamics following CAR-T cell infusion is not extensive, which hinders the ability to ascertain the durability of neutrophil-related effects on CAR-T cell treatment.

In conclusion, this study emphasizes the pivotal role of neutrophils in CLDN18.2-specific CAR-T treatment resistance in patients with advanced GC. By demonstrating that a high baseline NLR and the presence of a specific neutrophil subcluster are associated with poor treatment response, we propose that neutrophil-related biomarkers could serve as prognostic factors. The immunosuppressive impact of neutrophils on the function of T cells, particularly naive T cells, suggests potential therapeutic avenues to overcome CAR-T cell treatment resistance, such as neutrophil-targeted interventions or strategies to enhance T cell resilience. These insights not only deepen our understanding of the intricate dynamics of the tumor microenvironment but also facilitate the development of precision medicine strategies to enhance the efficacy of CLDN18.2-specific CAR-T cell treatment in advanced GC. Future studies should focus on validating these biomarkers in larger cohorts and exploring the potential of neutrophil-mediated immunosuppression treatments to enhance the efficacy of cell therapy.