Single-cell transcriptional profiling uncovers the association between EOMES⁺CD8⁺ T cells and acquired EGFR-TKI resistance

Abstract

Acquired resistance to tyrosine kinase inhibitors (TKIs) is reportedly inevitable in lung cancers harboring epidermal growth factor receptor (EGFR) mutations, emphasizing the need for novel approaches to predict EGFR-TKI resistance for clinical monitoring and patient management. This study identified a significant increase in eomesodermin (EOMES)⁺CD8⁺ T cells in the TKI-resistant patients, which was correlated with poor survival. The increase in EOMES⁺CD8⁺ T cells was further confirmed in both tissue samples and peripheral blood of patients with TKIs resistance. The integrated analysis of pseudotime and Gene set variation showed that the increase in EOMES⁺CD8⁺ T cells may be attributed to TRM T cell conversion and metabolic reprogramming. Overall, this work suggested an association between the increased number of EOMES⁺CD8⁺ T cells and acquired TKI drug resistance, supporting the utility of EOMES⁺CD8⁺ T cells as a biomarker for TKI treatment response.

Background

Tyrosine kinase inhibitors (TKIs) are effective in the treatment of 50%–60% of non-small cell lung cancer (NSCLC) patients harboring activating epidermal growth factor receptor (EGFR) mutations in Asia [1–3]. However, most patients are at a high risk of developing acquired drug resistance after treatment [4, 5]. Therefore, identifying a biomarker to predict EGFR-TKI resistance could be valuable for clinical monitoring and patient management.

The mechanism underlying EGFR-TKI resistance remains unknown in 15%-20% of patients [6, 7]. The current studies on EGFR-TKI resistance are predominantly retrospective, with heterogeneity in sample size, patient race, disease stage, and treatment strategy, resulting in inconsistent results [8]. Recently, it was reported that the EGFR-TKI rechallenge was effective following anti-programmed cell death receptor 1 (PD-1) monotherapy [9–11], implying the critical role of immune cells in tumor microenvironment (TME) in TKI resistance. Typically, TME is highly complex and dynamic during EGFR-TKI treatment. EGFR mutations can trigger low-level immune responses, and TKI treatment can lead to TME reprogramming by recruiting cytotoxic CD8⁺ T cells and dendritic cells (DCs) while depleting Foxp3⁺ regulatory T cells (Tregs) and suppressing M2-like polarization. Consequently, local immunosuppression can be quickly but temporarily relieved after TKI treatment. Nonetheless, with long-term treatment, antitumor effector cells may changes or decrease [12].

In this work, we attempt to identify a biomarker associated with EGFR-TKI resistance by incorporating three key phases during TKI treatment: TKI naïve (TN, before TKI treatment), residual disease (RD, TKI-sensitive group), and progressive disease (PD, TKI-resistant group) [13, 14].

Transformation of Tissue-resident memory T cells induces an increase in EOMES⁺CD8⁺ T cells under TKI resistance

An integrative analysis was performed using scRNA-seq data derived from the normal (n=4), TN (n=7), RD (n=10), and PD (n=8) samples (Figure S1, Table S2, and supplemental material). We analyzed the profile of CD8⁺ T cells, given their potential role in TKI resistance and antitumor immunity [9–11, 15]. Two subsets of CD8⁺ T cells were identified based on canonical markers, including tissue-resident memory (TRM) T cells (CD103⁺, CD69⁺, and SPR1⁻) and exhausted T cells (CD8⁺, ICOS⁺, TIGIT⁺, and PDCD1⁺). EOMES is commonly believed to be linked to T cell exhaustion and development. A distinct new subset of cells termed EOMES⁺CD8⁺ T cells was defined (Figures 1A; Table S6) [16, 17]. A significant increase in EOMES⁺CD8⁺ T cells and reductions in TRM and exhausted T cells in the PD group compared to the RD group was found (Figure 1A). The CellChat analysis imply that with the administrarion of TKIs and development of drug resistance, the intercellular interaction between immune cells and epithelial cells undergoes dynamic changes, and the damage of signaling pathways between multiple cells may be the reason for resistance to TKIs (Figure S2, and supplemental material).

Figure 1. Identifying major CD8⁺ T cell subtypes and evolutionary profiles.

(A) UMAP projection of 3735 T cells and the proportions of the main T cell subtypes in Normal, TN, RD, and PD groups. Each dot refers to a single cell and represents a different color based on their belongings. (B) Canonical cell markers of the three main T cell subtypes (exhausted, EOMES⁺CD8⁺, and TRM). (C) The development trajectory of CD8⁺ T cells inferred by diffusion map, colored by cell subtype and sample group. (D) A five-plex staining panel illustrating the components of EOMES⁺CD8⁺ T cells and TRM T cells. (E) Comparison of cellular infiltration between PD (n=7) and RD (n=10) patients (P-values: TRM T cells <0.0001, EOMES⁺CD8⁺ T cell <0.0001). Scale bar = 100 μm.

To further identify the dynamic changes in the three CD8⁺ T cell subsets, pseudotime analysis was utilized to depict their evolutional characteristics. A similar evolutionary process involving transforming TRM cells to exhausted T cells was observed in TN and RD groups. In the PD group, TRM cells were predominantly converted into EOMES⁺CD8⁺ T cells, possibly explaining the increase in EOMES⁺CD8⁺ T cells in PD patients (Figure 1C), similar to the evolutionary trend in the control samples. There were no significant differences in the evolutionary characteristics of exhausted T cells across TN, PD, and RD groups. Interestingly, EOMES⁺CD8⁺ T cells were significantly increased in the PD group compared to the other three groups (Figure 1C). We performed five-color multiplex fluorescent immunohistochemistry (mfIHC) on formalin-fixed paraffin-embedded tissues comprising 10 RD and 7 PD samples. Representative images from individual samples are displayed in Figure 1D. Consistent with our scRNA-seq analysis, mfIHC demonstrated a higher EOMES⁺ T cell component in PD samples than in RD samples. Conversely, TRM T cell infiltration was more abundant in RD tissues (Figure 1D–E).

Molecular changes in CD8⁺ T cells and their association with clinical survival

Next, we analyzed the molecular characteristics of the three subsets of CD8⁺ T cells. TRM T cells are the dominant memory CD8⁺ T cells in TME, capable of maintaining their “tissue-resident” property as a specific subtype of tumor-infiltrating lymphocytes (TIL) by mediating CD103⁺, CD69⁺, and SPR1⁻ expressions [18]. In addition, we found that TRM T cells expressing TIM-3 (HAVCR2)⁺ and PD-1⁺, that were active in respons to PD-1 inhibitors and absent in the PD group [18]. Meanwhile, the PD group exhibited a significant decrease in TRM proportions (Figures 2A–B). Such findings may explain the poor efficacy of PD-1 monotherapy in patients with acquired resistance to EGFR-TKI in clinical practice.

Figure 2. Dynamic molecular characteristics of EOMES⁺CD8⁺ T cells and TRM T cells.

(A) Proportions of the three T cell subtypes (exhausted, EOMES⁺CD8⁺, and TRM) in Normal, TN, RD, and PD groups. Each cell subtype is represented by a different color. (B) The heatmap visualizes the differences in the four groups of cancer hallmarks colored per cell by GSVA. (C) Dynamic changes in the expression of canonical cell markers of the three T cell types in the four groups. (D) The heatmap visualizes the differences in metabolic patterns colored per cell by GSVA in the four groups. (E) Kaplan-Meier curves display the prognostic value of EOMES⁺CD8⁺ T cells for PFS (left) and OS (right). (F) Associations of EOMES⁺CD8⁺ T cells (left) and TRM T cells (right) with immune response estimated by TIDE scores.

Moreover, we found a sustained increase in CCR7 level (one of the homing signals) from the beginning of TKI treatment to aquirsition of TKI resistance (Figure 2B), suggesting an increased tendency of TRM T cells to “migrate” out of TME during disease progression. In contrast, EOMES⁺CD8⁺ T cells were observed only after the occurance of TKI resistance. In the PD group, the homing markers CCR7 and SELL significantly increased, but the tissue-resident marker ITGAE (combining E-cadherin) decreased (Figure 2B). Overall, our findings imply that the upsurge in EOMES⁺CD8⁺ T cells may come from the TRM T cell conversion, consistent with our pseudotime analysis results. SVA enrichment analysis of 21 molecular pathways was then conducted to detect subtle pathway activity changes in TRM CD8⁺ T cells and EOMES⁺CD8⁺ T cells during the progression from RD to PD. Intriguingly, the two cell types exhibited reverse tendenciy in PI3K-AKT-mTOR, TGFβ, and DNA replication pathways, associated with developing CD8⁺ T cells (Figure 2C) [19–21], implying the potential involvement of these pathways in TKI resistance and the diverse functions of the two cell types.

TRM T cells were reported to compete for nutrients with tumor cells in TME via fatty acid-oxidative phosphorylation [22]. Therefore, the metabolic patterns of EOMES⁺CD8⁺ T cells and TRM T cells were further analyzed. Significant differences in the metabolic pattern of TRM T cells were demonstrated between PD and RD/TN groups, with decreased fatty acid metabolism (fatty acid biosynthesis and degradation) in the PD group (Figure 2D). Generally, the metabolic burden limits the activity of tissue-resident memory T cells. However, such a metabolic burden might increase with decreased fatty acid metabolism in TRM cells. Moreover, higher levels of active fatty acid metabolism were observed in EOMES⁺CD8⁺ T cells than in TRM T cells in the PD group (Figure 2D). These findings indicate that metabolic reprogramming loss in TRM T cells may be linked to their transformation to EOMES⁺CD8⁺ T cells.

Further analysis was focused on the relationship between the two subsets of cells and the clinical survival outcome in a pooled cohort of lung adenocarcinoma patients with EGFR mutations (Table S7). We stated that higher expression levels of signature genes of EOMES⁺CD8⁺ T cells were correlated with poorer progression-free survival (PFS) (p=0.028; HR, 1.697; 95% CI: 1.052–2.74) and overall survival (OS) (p=0.0098; HR, 2.218; 95% CI: 1.193–4.123) (Figure 2E). Additionally, the TIDE score exhibited a weak positive correlation with EOMES⁺CD8⁺ T cell levels (p=0.107, r = 0.11) and a negative correlation with TRM T cell expression (p=0.028, r = −0.14) (Figure 2F) [23]. These results imply that EOMES⁺CD8⁺ T cells might play a substantial role in acquired TKI resistance by reshaping the immune microenvironment.

Peripheral EOMES⁺CD8⁺ T cells had a positive relationship with acquired TKI resistance

As accessibility makes circulating biomarkers from peripheral blood ideal for monitoring disease activity, we evaluate the association of peripheral blood cells and TKI resistance. For this purpose, we first examined an external single-cell data set containing 14 lung cancer samples, each contained paired tissue and peripheral blood data [24]. As a result, 4439 CD8⁺ T cells were obtained, including EOMES⁺ T cells, TRM T cells, and other killer cells (Cyto) (Figure 3A). The correlation analysis displayed that the proportion of EOMES⁺ T cells in peripheral blood and tissue had a strong positive linear relationship with TKI resistance(P<0.001, R=0.92) (Figure 3B). Next, 24 peripheral blood samples were obtained from eight patients with RD (n=4) and PD (n=4) groups (Figure 3C). Consistent with scRNA-seq analysis and previous studies, CD8⁺ T cell count increased in the PD group [25]. More importantly, the peripheral EOMES⁺CD8⁺ T cell count in lymphocytes in the PD group was approximately three times higher than that in the RD group (17.38% vs. 6.19%, p<0.0001) (Figure 3D, 3E). A cutoff value of 13.3 was then determined using ROC analysis, yielding an area under the ROC curve of 1.0 (Figure 3F). While further validation with larger samples is warranted, these findings reveal that EOMES⁺CD8⁺ T cells in peripheral blood have a potential to be a promising biomarker for EGFR-TKIs resistance.

Figure 3. Different EOMES expression in CD8⁺ T cells in the peripheral blood samples from lung cancer patients.

(A) UMAP projections of 14 samples containing paired tissues and peripheral blood and the distribution of CD8⁺ T cell subsets contained in each sample. (PTC: CD8⁺ cytotoxic T cells from peripheral blood, TTC: CD8⁺ cytotoxic T cells from tumor tissue, NTC: CD8⁺ cytotoxic T cells from adjacent normal tissue). (B) The proportion of EOMES⁺T cells in peripheral blood and corresponding tissues has a positive association (P<0.001, R=0.92). (C) Procedure for performing a liquid biopsy in lung cancer patients. (D) Flow cytometric analysis of the ratio of apoptotic CD8⁺ T and EOMES⁺CD8⁺ T cells to lymphocytes in RD and PD groups. All samples were detected in triplicates. (E) tSNE map of total CD8⁺ T cells (green dots) and CD8⁺EOMES⁺ T cells (red dots) in lymphocytes (gray dots). (F) ROC curve analysis.

Discussion

Prioor studies substantiate that EOMES expression is the characteristics of exhausted CD8⁺ T cells [26–30]. EOMES was identified as a marker of thymic precursors of self-specific memory-phenotype CD8⁺ T cells (CD8-MP) in humans [16] and is critical in maintaining immune homeostasis [31]. Our study disclosed that EOMES⁺CD8⁺ T cells were significantly increased in lung cancers after acquired resistance to TKIs and correlated with poor survival outcomes.

The mechanism underlying increased EOMES⁺CD8⁺ T cells in response to TKI resistance remains poorly understood. By analyzing intercellular interactions, we provide clues regarding the potential role of numerous intercellular signaling pathways in immune and epithelial cells during acquisition of EGFR-TKI resistance, involving cytokines, extracellular matrix, cell adhesion, and neural-related pathways. Pseudotime analysis suggested that increased EOMES⁺CD8⁺ T cells in the PD group may be derived from TRM T cells, while alterations of fatty acid metabolism (decrease in TRM cells and increase in EOMES⁺CD8⁺ T cells) may drive the conversion. Most importantly, given the vital function of TRM T cells in anti-PD-1 immunotherapy, transformation of TRM T cells to EOMES⁺CD8⁺ T cells may account for TKI resistance development and poor response to immune checkpoint blockade (ICB) immunotherapy after acquiring resistance. Besides, these findings indicated that promoting fatty acid metabolism in TRM T cells may improve the efficacy of TKI therapy and immunotherapy. In addition, PD-1 expression progressively decreases after acquired resistance in TRM T cells, raising concerns about whether anti-PD-1 immunotherapy should be administered early in this patient population.

In summary, this work demonstrated a correlation between EOMES+CD8+ T cells and acquired EGFR-TKI resistance, supporting the utility of EOMES as a biomarker for therapy response. However, additional studies with a larger sample size are required to validate our findings, and the time window between molecular prediction and radiographic progression of TKI resistance warrants further investigation.