High expression of CCL3/CCL4/CCL5/CCR5 promotes exhausted CD8 + T cells terminal differentiation and is associated with poor prognosis in pediatric B-ALL patients

Abstract

This study aims to identify differentially upregulated ligand-receptor interactions between B-ALL cells and exhausted CD8⁺ T cells and to develop a multivariate Cox regression model for predicting the overall survival of pediatric B-ALL patients based on CCL3/CCL4/CCL5 expression levels. Pediatric B cell-acute lymphoblastic leukemia (B-ALL) is a hematopoietic malignancy. T cell exhaustion has an important impact on the prognosis of leukemia. The interaction between tumor cells and T cells can influence the degree of T cell exhaustion. However, the effects of B-ALL cells on exhausted T cell subpopulations and how the interaction influences the prognosis of B-ALL patients remain unclear. Single-cell RNA sequencing (scRNA-Seq) data from pediatric B-ALL patients were downloaded from GEO. Cell interaction analysis identified ligand-receptor pairs between B-ALL cells and exhausted CD8⁺ T cell. To confirm the function of CCL3/CCL4/CCL5/CCR5 in prognosis prediction, quantitative real-time polymerase chain reaction (qRT-PCR) was employed. We further developed an innovative stratified model that integrates CCL3, CCL4, and CCL5 through multi-Cox regression. Clustering of scRNA-Seq data revealed an increased proportion of exhausted CD8⁺ T cells in relapsed B-ALL, especially terminal exhausted CD8⁺ T cells (CD8_Ex), with increased exhaustion and decreased proliferation scores. Moreover, the CCL3/CCL4/CCL5-CCR5 axis was upregulated in interactions between B-ALL cells and terminal CD8_Ex. Transcriptome data from 221 pediatric B-ALL samples revealed that high CCL3/CCL4/CCL5/CCR5 levels correlate with low overall survival (OS). A multivariate Cox regression model incorporating CCL3/CCL4/CCL5 predicted prognoses. Finally, a model based on the adult B-ALL patients from our center also accurately predicted prognoses. We report for the first time the crucial role of the CCL3/CCL4/CCL5-CCR5 axis in the differentiation of terminal exhausted CD8⁺ T cells in B-ALL. High expression of CCL3, CCL4, CCL5, and CCR5 correlates with poor prognosis in B-ALL, suggesting potential biomarkers and therapeutic targets.

Introduction

B cell-acute lymphoblastic leukemia (B-ALL) is a hematologic malignancy resulting from uncontrolled proliferation of B-lymphoid progenitor cells marked by rapid expansion of immature B cells in the bone marrow (BM) and peripheral blood. ¹ Despite significant advances in the understanding of its molecular biology and improvements in treatments strategies, the disease remains challenging due to its high relapse of over 60% and drug resistance. ² How the properties of the host immune system affect outcome in leukemia has garnered attention in recent years, particularly the involvement of CD8⁺ cytotoxic T lymphocytes (CTLs).^3,4 CTLs are crucial for recognizing and eliminating tumor cells, but their function can be significantly impaired in the tumor microenvironment, leading to exhaustion.^4,5 This exhaustion negatively impacts leukemia prognosis and the efficacy of immunotherapies.^6–8 Exhausted CD8⁺ T cells exhibit heterogeneity, comprising various subpopulations, such as progenitor and terminal exhausted CD8⁺ T cells, which may influence patient outcomes.^9,10 The interaction between tumor cells and exhausted T cells can influence the state of T cell exhaustion. For instance, in non-small cell lung cancer, the proportion of tumor cells correlates with exhausted T cells, and the ligand-receptor interactions between these two cell states showed that T cells were recruited to tumor cells through the production of chemokines. ¹¹ However, the role of ligand-receptor interactions between B-ALL cells and exhausted T cells remain underexplored in B-ALL patients. C-C chemokine ligands (CCLs) including CCL3, CCL4, and CCL5 together with their receptor C-C chemokine receptor 5 (CCR5) have been shown to play a critical role in the progression of multiple tumors by promoting tumor cell growth and creating an immunosuppressive environment.^12,13 Regulatory T cells are recruited to tumor sites via a CCR5-dependent pathway where they modulate the effector functions of lymphocytes, including suppressing T cell-mediated cytotoxicity.^14,15 However, the role of the CCL3/CCL4/CCL5-CCR5 axis in the exhaustion of T cells remains poorly understood. In this study, a single-cell RNA sequencing dataset including seven newly diagnosed and relapsed pediatric B-ALL patients from GSE130116 were used to explore the impact of B-ALL cells on the T cell exhaustion state and prognosis, aiming to explore new targets to assess B-ALL treatment and prognosis.

Materials and methods

Data sources

Single-cell RNA sequencing (scRNA-seq) data from pediatric B-ALL patients were obtained from the GSE130116 dataset, which includes seven pairs of samples from pediatric patients at initial diagnosis or relapse. Transcriptome data were sourced from the TARGET-P2-BALL database, containing a gene-expression matrix of bone marrow from 221 newly diagnosis child patients. A total of 23 samples were collected from the newly diagnosed B-ALL patients at our clinical center (JNU database). This study adhered to the Declaration of Helsinki and was approved by the Ethics Committee of the First Affiliated Hospital of Jinan University with all participants providing written informed consent.

Single-cell RNA sequencing data processing

The “Seurat” package (version 4.1) was used to perform quality control on the seven pairs of B-ALL samples from GSE130116, applying the criteria of nFeature_RNA >500 and <4000 and percent.mt <15. Following the data quality control, variance analysis was conducted on the filtered genes, identifying the top 2000 genes with significant variation among cells for subsequent cell type identification. Principal components for these highly variable genes were calculated, and principal components with P values less than 0.05 were used to identify clusters with the UMAP2 algorithm. Finally, cell types clusters were manually annotated based on the signature gene expression profiles. To investigate intercellular communication, the “CellChat” package was used in conjunction with the CellChatDB database, which integrates various ligand-receptor interactions. This analysis aimed to explore the mechanisms of intercellular communication at the single-cell level. Marker genes of terminal exhausted CD8⁺ T cells were identified by FindAllMarkers from “Seurat” R package. “GSVA” v1.42.0 package was used to perform single-sample gene set enrichment analysis (ssGSEA) score based on top 50 marker genes of terminal exhausted CD8⁺ T cells within each individual B-ALL patient sample from the TARGET database.

Validation of biomarker expression by quantitative real-time polymerase chain reaction

Mononuclear cells were separated from B-ALL patient samples by Ficoll density centrifugation (Sigma Aldrich, German). Then, total RNA was extracted from the PBMCs using TRIzol reagent (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s instructions and reverse transcribed into complementary DNA (cDNA) using the PrimeScript™ RT reagent Kit (Takara, Japan) according to the experimental instructions. The relative expression levels of CCL3, CCL4, CCL5, and CCR5 were measured by quantitative real-time polymerase chain reaction with SYBR Master Mix (TIANGEN, Beijing, China), and B2M was selected as an internal control. The expression levels of CCL3, CCL4, CCL5, and CCR5 are presented as 2^−ΔCT. The primer sequences in this study are listed as follows:

• CCL3 Forward: 5′-AGTTCTCTGCATCACTTGCTG-3′

• CCL3 Reverse: 5′-CGGCTTCGCTTGGTTAGGAA-3′

• CCL4 Forward: 5′-GCTTCCTCGCAACTTTGTGG-3′

• CCL4 Reverse: 5′-GGATTCACTGGGATCAGCACA-3′

• CCL5 Forward: 5′-CTGCTTTGCCTACATTGCCC-3′

• CCL5 Reverse: 5′-CACACTTGGCGGTTCTTTCG-3′

• CCR5 Forward: 5′-TTCTGGGCTCCCTACAACATT-3′

• CCR5 Reverse: 5′-TTGGTCCAACCTGTTAGAGCTA-3′

Statistical analysis

Statistical analyses were performed in R (version 4.4.1). A t-test was used for comparing two groups, while the log-rank test from the “survminer” package was applied to compare Kaplan–Meier (KM) curves. Receiver operating characteristic (ROC) curve analysis was conducted using the “timeROC” package. A two-tailed P value of less than 0.05 was considered statistically significant.

Results

After integrating, reducing dimensionality, and clustering the scRNA-Seq data from newly diagnosed and relapsed pediatric B-ALL patients, seven types of cells were identified including B-ALL cells (DNTT, MME, ERG, BCL2, CD19, CD79A), B cells (CD19, CD79A), T cells (CD3D, CD3E, IL7R), NK cells (NKG7, GZMB, GNLY), myeloid cells (CD14, S100A8, S100A9), HSPC (AVP, SPINK2, CRHBP), erythrocytes (HBA1, HBA2, HBB) (Supplemental Figures S1A and B). Then, the T cells were grouped into nine distinct subtypes: CD4⁺ T cells (CD4⁺ CD8⁻; Naïve, Exhausted CD4⁺ T cells, Th, Treg), CD8⁺ T cells (CD4⁻ CD8A⁺ CD8B⁺; Naïve, Effector, Memory, Exhausted CD8⁺ T cells), γδ T cells (TRDC) (Online Materials and Methods: Figure 1a and Supplemental Figure S1C). The exhausted CD8⁺ T cell subtype, defined by high expression of immunosuppressive markers PDCD1 and HAVCR2, demonstrated an increased trend in relapsed samples compared with de novo samples, suggesting that a higher proportion of exhausted CD8⁺ T cells may be correlated with poor prognosis for relapsed B-ALL (P = 0.031, Figure 1b and c). Next, exhausted CD8⁺ T cells were grouped into three subclusters to illuminate their heterogeneity in B-ALL (Figure 1d). Among exhausted CD8⁺ T cells, progenitor exhausted CD8⁺ T cells (progenitor CD8_Ex) expressed higher levels of naïve/memory T cell-related genes (TCF7, LEF1, IL7R, CCR7) and lower levels of inhibitory genes (PDCD1, TIGIT, HAVCR2, LAG3) (Figure 1e). Intermediate exhausted CD8⁺ T cells (intermediate CD8_Ex) exhibited elevated levels of effector genes such as GZMB, GZMG, and GNLY. Lastly, terminal exhausted CD8⁺ T cells (terminal CD8_Ex) demonstrated the highest expression of in-hibitory genes. Interestingly, progenitor CD8_Ex presented a lower proportion while terminal CD8_Ex were higher in relapsed samples compared with de novo samples (Figures 1f and g). Furthermore, terminal CD8_Ex in relapsed samples had a lower proliferative score and higher exhausted score than de novo samples (Figure 1h). These results indicated that a higher proportion of exhausted CD8⁺ T cells in relapsed B-ALL patients differentiate into a terminal state, and their function are impaired. To explore the potential mechanism of how B-ALL cells interact with terminal CD8_Ex, the CellChat package was applied to simulate intercellular communication. In relapsed cases, some ligand receptors that B-ALL cells express on exhausted T cell subpopulations were upregulated compared to newly diagnosed cases. Specifically, there was a selective upregulation of the CCL3/CCL4/CCL5-CCR5 axis in interactions involving terminal CD8_Ex as opposed to progenitor CD8_Ex (Figure 1i). Moreover, the expression levels of the CCL3, CCL4, CCL5, and CCR5 genes were upregulated in relapsed B-ALL samples (Figure 1j). Therefore, the CCL3/CCL4/CCL5-CCR5 pathway may promote exhausted CD8⁺ T cell terminal differentiation.

Figure 1.

Interactions between B-ALL cells and exhausted CD8⁺ T cell subpopulations in newly diagnosed and relapsed pediatric B-ALL patients based on scRNA-Seq. (a) UMAP projection of nine T cell clusters of newly diagnosed and relapsed pediatric B-ALL samples. Each dot corresponds to one single cell colored according to cell cluster. (b) Stacked bar chart showing the constitution of the nine T cell clusters in the de novo and relapsed groups. (c) Comparison of the frequency of exhausted CD8⁺ T cells in all T cells from the de novo (n = 7) and relapsed (n = 7) samples. Each dot represents one BM sample, while the center line indicates the median value. (d) UMAP plot showing three exhausted CD8⁺ T cells. (e) Heatmap showing the expression of marker genes of subclusters in exhausted CD8⁺ T cells. (f) Stacked bar chart showing the constitution of subclusters in exhausted CD8+ T cells in the de novo and relapsed groups. (g) Comparison of the frequency of progenitor and terminal exhausted CD8⁺ T cells in exhausted CD8⁺T cells from de novo (n = 7), relapsed (n = 7) samples. (h) Comparison of the proliferation score and exhaustion of terminal CD8_Ex between de novo and relapsed samples. (i) Bubble plots showing the intercellular communication between B-ALL cells and progenitor or terminal exhausted CD8⁺ T cells. (j) Boxplots showing the expression levels of the CCL3, CCL4, CCL5, and CCR5 genes in de novo and relapsed samples.

To further explore the impact of CCL3/CCL4/CCL5/CCR5 on B-ALL prognoses, we conducted a prognostic analysis using transcriptomic and clinical data from pediatric B-ALL patients sourced from the TARGET database. Based on the optimal cutoff for the expression of each gene, the patients were divided into high and low expression groups. A KM plot indicated that high levels of CCL3, CCL4, CCL5, and CCR5 were correlated with low overall survival (OS, P = 0.001, P = 0.001, P = 0.008, and P = 0.033) (Figures 2a–d). Considering an additive effect on outcome with multiple genes involved in the CCL3/CCL4/CCL5-CCR5 pathway, we further characterized the predictive value of co-expression patterns in B-ALL. B-ALL patients with high expression of any two among CCL3, CCL4, CCL5, and CCR5 exhibited the poorest OS (CCL3/CCL4, P < 0.001; CCL3/CCL5, P = 0.001; CCL4/CCL5, P < 0.001, CCL3/CCR5, P = 0.001; CCL4/CCR5, P < 0.001; CCL5/CCR5, P = 0.004) (Figures 2e–g; Supplemental Figure S2A–C). Additionally, the area under the ROC Curve (AUC) of three two-gene models was significant for predicting 3- and 5-year survival (Figures 2h–j; Supplemental Figure S2D–F). The four genes were then subjected to multivariate Cox regression analysis to establish a prognostic model (Figure 2k). Furthermore, a CCL3/CCL4/CCL5 three-gene model was established, and the formula for calculating the prognostic risk model was as follows: h(t) = h₀(t) × exp (CCL3high × 0.507 + CCL4high × 0.432 + CCL5high × 0.333). The risk score of each patient is the expression of three candidate genes multiplied by the coefficient of their multi-Cox regression (Figure 2m). According to the optimal risk score cutoff, pediatric B-ALL patients were divided into a high-risk group and a low-risk group. A KM plot of the two risk groups indicated that the high-risk group had a significantly worse prognosis than the low-risk group (P < 0.001, Figure 2l). Finally, we used ssGSEA to assess the terminal CD8_Ex scores of BALL patients from the TARGET database by scoring them based on the marker genes for terminal CD8_Ex from de novo scRNA-Seq dataset. B-ALL patients with higher terminal CD8_Ex scores had worse OS (P = 0.024) and higher risk score derived from the multi-Cox regression model (P < 0.001) (Supplemental Figure S2G and H). The correlation analysis also revealed a positive relationship between terminal CD8_Ex scores and risk scores (Supplemental Figure S2I).

Figure 2.

Construction of a prognostic model based on the CCL3/CCL4/CCL5/CCR5 genes in the TARGET-P2-BALL dataset. (a–d) KM curves showing the OS of high and low expression of each gene in the TARGET-P2-BALL dataset. (a) CCL3. (b) CCL4. (c) CCL5. (d) CCR5. (e–g) KM curves showing the OS of two-gene co-expression of CCL3, CCL4, and CCL5. (e) CCL3/CCL4. (f) CCL3/CCL5. (g) CCL4/CCL5. (h–j) ROC curves showing the prognostic value of the two-gene co-expression of CCL3, CCL4, and CCL5. (h) CCL3/CCL4. (i) CCL3/CCL5. (j) CCL4/CCL5. (k) Forest plot showing multi-Cox regression analysis of B-ALL patients. (l) KM curve survival analysis of the low- and high-risk groups based on the multi-Cox regression model. (m) Radar plot showing the contribution of the three genes to OS, which was determined by the coefficients of the three genes in the multivariate COX regression model.

To further validate the prognostic value of the CCL3/CCL4/CCL5/CCR5 genes for adult patients, 23 adult B-ALL patient samples were collected from our center (JNU database). However, the expressions of the four genes were not significantly associated with OS, possibly due to the small sample size (Supplemental Figures S3A–D). We next used the CCL3/CCL4/CCL5/CCR5 genes to construct a prognostic model, and a CCL3/CCL4/CCL5 three-gene model was established with the following formula: h(t) = h₀(t) × exp (CCL3high × 1.818 + CCL4high × (−3.595) + CCL5high × 1.833) (Supplemental Figure S3E). The risk score of each patient is the expression of the three candidate genes multiplied by the coefficient of their multi-Cox regression. According to the optimal risk score cutoff, the JNU dataset was divided into a high-risk group and a low-risk group, and a KM plot of the two risk groups indicated that the high-risk group had a significantly worse prognosis than the low-risk group (P = 0.004; Supplemental Figure S3F and G).

Discussion

The quantity and function of T cells largely determine the effectiveness of anti-tumor responses and influence the prognosis of patients with malignancy. Under the continuous stimulation of tumor antigens and the influence of inflammatory factors, T cells gradually exhibit reduced proliferative capacity and diminished effector function, a phenomenon known as T cell exhaustion. In addition to the decline in effector function, exhausted T cells also upregulate or co-express various inhibitory receptor molecules such as PD-1, CTLA-4, LAG-3, and TIM-3. ¹⁶ Other characteristics include changes in the expression and function of key transcription factors, metabolic alterations, and epigenetic modifications. T cell exhaustion is a progressive process; therefore, exhausted CD8⁺ T cells are heterogeneous and consist of different exhausted subpopulations. ¹⁷ Previous studies have revealed that heterogeneous exhausted T cells can transition from progenitor exhausted T cells to a terminal subset, and this is related to the prognosis of tumor patients.^9,16 Moreover, tumor cells can remodel the tumor microenvironment and induce the immunosuppression of T cells through multiple mechanisms including metabolism, immune checkpoints, and the recruitment of T cells. However, whether B-ALL cells are involved in the exhausted T cell transition and how they interact with different states of exhausted T cells remain unclear.

In this study, scRNA-seq data including seven pairs of pediatric B-ALL samples (de novo and relapsed) were employed to understand the interaction between tumor cells and exhausted CD8⁺ T cell subpopulations. Progenitor CD8_Ex possess relatively higher expression of naïve/memory T cell-related genes and lower expression of inhibitory genes compared to terminal CD8_Ex, consistent with previous reports. ¹⁶ Furthermore, relapsed samples exhibited a higher proportion of the terminal CD8_Ex subset with a lower proliferation and higher score than de novo samples, which indicates that the balance of the exhausted T cell proportions shifts from progenitor towards terminal CD8_Ex in relapsed patients, and this bias may contribute to the poorer prognosis in relapsed B-ALL. Next, we found that relapsed B-ALL cells specifically interact with terminal CD8_Ex through the CCL3/CCL4/CCL5-CCR5 axis. Moreover, the expression levels of the CCL3, CCL4, CCL5, and CCR5 genes were upregulated in relapsed B-ALL samples. Therefore, we propose the hypothesis that B-ALL cells promote exhausted CD8⁺ T cell terminal differentiation through the CCL3/CCL4/CCL5-CCR5 axis and lead to poor prognosis.

Chemokines are key molecules in cancer development and disease progression. The infiltration of immune cells into the tumor microenvironment is a decisive factor in cancer prognoses. ¹⁸ Chemokine signaling is crucial for recruiting immune cells with anti-tumor effects, such as CD8⁺ T cells and natural killer cells. Therefore, chemokines are also essential for assessing the prognosis of tumor patients. In this study, a pediatric B-ALL dataset sourced from the TARGET database was analyzed, and we found that the OS of patients with high expression levels of CCL3, CCL4, CCL5, and CCR5 was significantly worse than that of patients with low expression levels. Results of a three-gene CCL3/CCL4/CCL5 model also showed that the OS of high-risk patients was poorer than that of low-risk patients. Lastly, data from 23 B-ALL patients from our center were used to further validate the prognostic value for adult B-ALL patients. The CCL3/CCL4/CCL5 three-gene model can also predict the OS of adult B-ALL patients.

In previous studies, the chemokines CCL3, CCL4, CCL5, and their receptor CCR5 were overexpressed in various tumors, including both hematologic malignancies and solid tumors, and were associated with tumor progression, which is consistent with the findings of the current study.^12,13,19,20 However, other studies have also demonstrated that CCL5 and CCL4 enhance the response to immunotherapy.^12,13 Therefore, the CCL3/CCL4/CCL5 pathway acts as a double-edged sword in cancer, depending on the type of immune cells expressing CCR5, which determines their recruitment to tumor sites by these chemokines. If anti-tumor T cells and dendritic cells are recruited to the tumor microenvironment through CCR5, this pathway facilitates the elimination of tumor cells. However, this pathway can be used by cancer cells such as the B-ALL cells in our study, which secreted CCL3, CCL4, and CCL5 to recruit CCR5⁺ regulatory T cells, exhausted T cells, and other immunosuppressive cells, thereby modulating the immunosuppressive microenvironment. In our study, the correlation between ssGSEA scores based on the terminal CD8+ T cell marker genes and risk scores from the multivariable Cox regression model based on the expression of CCL3, CCL4, and CCL5, suggests CCL3/CCL4/CCL5-CCR5 pathway may play a essential role in recruiting and inducing terminal exhausted CD8⁺ T cells, which could contribute to poor prognosis in B-ALL patients. Nonetheless, the mechanism of CCR5 expression regulation remains unclear.

There are also limitations to this study. The sample size is limited, which may affect the reliability of the results. A small sample size can increase variability and reduce the statistical power, potentially impacting the accuracy and generalizability of the findings. These findings needed further validation by a larger sample cohort. Moreover, the effects of CCL3/CCL4/CCL5-CCR5 axis on the differentiation of exhausted CD8⁺ T cells needs to be verified by experiments in future research.

Conclusion

CCL3/CCL4/CCL5-CCR5 may promote the terminal differentiation of exhausted CD8⁺ T cells in pediatric B-ALL patients. High expression of CCL3, CCL4, CCL5, and CCR5 is associated with poor prognosis in B-ALL. We developed a predictive model with the CCL3/CCL4/CCL5 genes that could serve as a biomarker for the prediction of OS in pediatric B-ALL patients. This study provides potential mechanisms for the terminal differentiation of exhausted CD8⁺ T cells, offering valuable insights into the regulation of leukemia immunity.

Supplemental Material

sj-docx-1-iji-10.1177_03946320251346823 – Supplemental material for High expression of CCL3/CCL4/CCL5/CCR5 promotes exhausted CD8+ T cells terminal differentiation and is associated with poor prognosis in pediatric B-ALL patients

Supplemental material, sj-docx-1-iji-10.1177_03946320251346823 for High expression of CCL3/CCL4/CCL5/CCR5 promotes exhausted CD8+ T cells terminal differentiation and is associated with poor prognosis in pediatric B-ALL patients by Jiamian Zheng, Yupei Zhang, Xueting Peng, Cunte Chen, Jie Chen, Liye Zhong, Yangqiu Li and Songnan Sui in International Journal of Immunopathology and Pharmacology