Peripheral blood samples from 50 patients with CLL, including 32 newly diagnosed patients and 18 refractory/relapsed (R/R) patients, and 8 age-matched healthy individuals (HIs) were analyzed after obtaining informed consent according to the hospital Medical Ethical Committee. Peripheral blood mononuclear cells (PBMCs) were isolated using Human Mononuclear Cell Separation Medium 1.077 (Bio-Processing System, 25610) density-gradient centrifugation, and they were cryopreserved until analysis. Clinical characteristics of 50 patients with CLL in the HNCH were listed in Table S1 .

Cells were washed in phosphate-buffered saline containing 2% FBS and incubated at 4°C for 30 min with combinations of the following antibodies: CD3-APC-A750 (A94680, Beckman Coulter), CD4-KrO (A96417, Beckman Coulter), CD8-APC-A700 (B49181, Beckman Coulter), and TCF1-PE (655208, Biolegend). Relevant isotype control mAbs were purchased from BD Biosciences. After two washes in phosphate-buffered saline containing 2% FBS, cells were analyzed by FACS Calibur (Becton Dickinson, San Jose, CA). Data were processed with Navios Flow Cytometer software (Beckman Coulter, Brea, CA, USA).

The gene expression profiles of CLL cells and HIs were queried in the Gene Expression Omnibus (GEO) database, and the GSE19147 (31), GSE66425 (32), and GSE50006 datasets (33) were obtained. The GSE19147 dataset included 25 CLL patients and 8 HIs. GSE66425 included 30 PBMC samples from CLL patients and 5 samples of PBMCs from HIs, and GSE50006 included 32 B cell samples from HIs and 188 CLL samples. Differences in expression between CLL patients and HIs were compared using GEO2R.

To verify the data from the public datasets, total RNA was prepared from CLL patients and HIs using TRIzol Reagent (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s instructions. cDNA was synthesized from equal amounts of total RNA (1 μg) using HiScript^® III RT SuperMix for real-time quantitative PCR (qPCR) (+gDNA wiper) (Vazyme, Q711-02, China) and analyzed by ChamQ Universal SYBR qPCR Master Mix. Table S2 lists the primers used for PCR, which was performed using the ABI 7500 FAST real-time PCR system. The levels of transcripts were quantified by the 2^-ΔΔCT (cycle threshold) method.

The TTFT of CLL patients was queried in the GSE39671 dataset (34), which included 130 CLL patients and their TTFT information. In addition, we collected samples from 43 CLL patients from HNCH to confirm the TTFT results. The TTFT and OS for CLL patients is shown in Kaplan-Meier (KM) survival curves and analyzed by the log-rank Kaplan-Meier method using the “survival” package in R software (version 4.1.0). CLL prognoses were queried in the GSE22762 dataset (35), which included 107 CLL patients and OS information. CLL patients were segregated into 2 distinct categories according to TCF1 expression. The cutpoint value was calculated by the “survminer” package in R, and the RMST was obtained by the “survRM2” package in R. All cutpoint values are shown in Figure S1 .

The genes differentially expressed in correlation with TCF7 in the GSE39671 and GSE22762 datasets were analyzed by the Spearman correlation coefficient. Additionally, the Venny 2.1.0 program was used to screen for common genes in the two databases (R > 0.3 and P < 0.05), and the DAVID version 6.8 (36) was used to conduct corresponding biological process and KEGG pathway enrichment analysis, and a bubble plot was drawn with the “ggplot2” package in R.

A PPI network for co-expressing genes was constructed using the Search Tool for the Retrieval of Interacting Genes (STRING) database (https://string-db.org/cgi/) (37), and the results were visualized using Cytoscape software (version 3.4.0) (38). The cutoff criterion for the confidence score was the default setting (> 0.7). In addition, the closely connected protein-interactive regions were obtained by the Molecular Complex Detection (MCODE) plug-in, and the correlation of hub genes in GSE39671 and GSE22762 was plotted with the Circos online tool, and the biological process results were also shown in a PPI network by the STRING database. Further, the hub genes from MCODE were filtered again using the Markov Clustering (MCL) algorithm.

Data were analyzed using GraphPad Prism 8.2.1. Results are presented as the mean ± SD. Student’s unpaired t-test was used for differential expression analysis, the log-rank test was used to indicate the statistical significance of survival or TTFT correlation between groups, and the correlation of two genes was tested by the Spearman correlation coefficient. Survival curves were analyzed by the log-rank Kaplan-Meier method. Cox regression analysis was constructed to determine the hazard ratio (HR). All statistically significant variables (P < 0.05), as found in the univariate analyses, were included in multivariate analysis based on a Cox proportional hazards model. P < 0.05 was considered statistically significant.

To investigate the effects of TCF1 in CLL, we developed the research plan shown in the study flow chart ( Figure 1 ). We first analyzed the expression of the TCF1, in PBMCs, CLL cells, T cells, and normal B cells in CLL patients from the GEO database and HNCH, respectively. The expression of TCF1 in PBMCs was mainly concentrated in T cells, it was quite low in CLL cells, and it was lower in normal B cells ( Figures S2A , S2B ). Moreover, TCF1 expression in PBMCs from CLL patients was lower than that in HIs in the GSE66425 dataset (P = 0.012; Figure 2A ). Although there was no significant difference compared with HIs in the GSE19147 dataset, the proportion of TCF1+CD3+ T cells in CLL patients displayed a clear downward trend (P = 0.098; Figure 2B ). We verified the above results in 50 patients with CLL from HNCH. Amazingly, TCF1 had low expression not only in PBMCs as determined by qPCR (P = 0.011; Figure 2C ) but also in CD3+ T cells as determined by FCM (P = 0.004; Figure 2D ) in CLL patients compared with HIs, which is highly consistent with the above findings from the GEO databases. Therefore, compared with HIs, TCF1 was significantly downregulated in the PBMCs of CLL patients, particularly in CD3+ T cells.

Most patients with CLL are diagnosed at the early stage of asymptomatic disease and are not treated until treatment indicators appear. TTFT is an important index that evaluates disease stability in patients with CLL like lymphocyte doubling time (39). Thus, we explored the predictive effects of TCF1 on the TTFT and OS of patients with CLL using the GSE39671 and GSE22762 datasets, respectively. The results demonstrated that CLL patients with low TCF1 expression have a shorter TTFT (5-year TTFT rate: 36% vs. 57%, P = 0.009; Figure 2E ) and OS (5-year OS rate: 28% vs. 98%, P < 0.001; Figure 2F ) than those with high TCF1 expression. Next, we confirmed the above findings in 43 CLL patients from HNCH and found that low TCF1 expression appeared to be correlated with short TTFT (5-year TTFT rate: 11% vs. 57%, P < 0.001; Figure 2G ). Furthermore, we employed RMST to confirm the TTFT and OS data and found that patients with low TCF1 expression have a shorter RMST than those with high TCF1 expression ( Figures 2E–G , right panel). Collectively, there was a clear trend where patients with low TCF1 expression have rapid disease progression and a short survival time. Thus, TCF1 can be used as a predictive biomarker of inferior prognosis for CLL patients.

The 114 genes commonly co-expressed with TCF7 in the GSE39671 (592 genes) and GSE22762 datasets (903 genes) ( Figure 3A , left panel) were filtered by Venny and then built into a protein-protein network using the STRING database. Cytoscape (MCODE plug-in) was used to establish the most important module, which is highlighted in yellow ( Figure 3A , middle panel). Based on the degree score, we selected the module with the highest score that included 14 genes: TCF7, BCL11B, RUNX3, LCK, CD3E, LAT, AKT3, PIK3R1, CD86, IL10, FLT3LG, SLAMF1, CD7, and CD244. These genes were identified as potential hub genes, and the expression levels of the 14 hub genes were plotted by the Circos webtool for the GSE22762 and GSE39671 datasets ( Figure 3A , right panel).

The results of KEGG pathway analysis, which was performed using the DAVID database and plotted by R, indicated that the 14 hub genes mainly participate in T cell receptor signaling and T cell differentiation ( Figure 3B , right panel). Notably, we found that, of the 14 genes in the PPI network, BCL11B, RUNX3, CD3E, and LCK are directly related to TCF7 gene (highlighted in green) ( Figure 3B , left panel). We next reconstructed the PPI network of these five genes using the STRING database ( Figure 3C , left panel). The biological process category of each gene is displayed in different colors and mainly include regulation of T cell activation, T cell differentiation, and T cell receptor recombination ( Figure 3C , right panel), which is highly consistent with the above KEGG analysis results. These data indicate that TCF7 coordinates with hub genes in regulating T cell immunity.

To determine genes closely related to TCF7, we clustered the 5 genes using the MCL method. Remarkably, the 5 genes divided into 2 clusters, cluster 1 (including TCF7, RUNX3, and BCL11B) and cluster 2 (including LCK and CD3E). It was evident that the genes in cluster 1 were closely related to TCF7 ( Figure 3D ). We further verified the expression of BCL11B and RUNX3 in PBMCs from CLL patients (HNCH) and HIs by qPCR. Intriguingly, BCL11B expression was significantly decreased in CLL patients compared with that in HIs (P = 0.008; Figure 3E , left panel), but there was no significant difference in RUNX3 between CLL patients and HIs (P = 0.898, Figure 3E , right panel). Likewise, we found that BCL11B expression in CD3+ T cells was significantly decreased in CLL-CD3+ T cells compared to HI-CD3+ T cells in the GSE19147 dataset (P = 0.044 Figure 3F , left panel), but RUNX3 did not demonstrate a difference between CLL and HIs (P = 0.550, Figure 3F , right panel). In addition, there was a high positive correlation between TCF7 and BCL11B expression in the GSE22762 (R = 0.71, P < 0.001) and GSE39671 (R = 0.41 P < 0.001) datasets ( Figure 3G ). Therefore, among the genes co-expressed with TCF7 in CLL patients, BCL11B may be the closest partner gene for TCF7 involved in T cell immune regulation.

To investigate the effects of BCL11B alone or in combination with TCF1 on the prognosis of patients with CLL, we further analyzed the effects of BCL11B on TTFT and OS for CLL patients in the GSE39671 and GSE22762 datasets, respectively ( Figures S3A , S3B ). Low expression of BCL11B was significantly associated with shorter TTFT and poorer OS compared with high expression (5-year TTFT: 28% vs. 77%; 5-year OS: 33% vs. 98%) ( Figures 4A, B , left panel). Similarly, patients with low BCL11B expression had shorter RMST than those with high expression ( Figures 4A, B , right panel). As expected, the above findings were confirmed with 50 patients with CLL from HNCH (5-year TTFT: 0% vs. 28%) ( Figure 4C ). Next, we analyzed the effects of BCL11B in combination with TCF1 on prognosis, compared with those who were TCF1^highBCL11B^high, CLL patients who were TCF1^lowBCL11B^low had a poorer TTFT and OS ( Figures 4D, E , left panel) as well as shorter RMST ( Figures 4D, E , right panel). Specifically, the 5-year TTFT was 26% and 74%, and the 5-year OS was ≤ 17% and 100%, respectively. Similarly, those results also were validated in HNCH, the 5-year TTFT and for TCF1^lowBCL11B^low and TCF1^highBCL11B^high was 0% and 37%, respectively ( Figure 4F ). Therefore, BCL11B can be used as a predictive biomarker for inferior prognosis for CLL patients. More importantly, the combination of TCF1 and BCL11B expression could more accurately assess the prognosis of CLL patients compared with either alone.

To explore the correlation between TCF1 expression and the characteristics of patients with CLL, we next integrated a series of clinical characteristics, including disease state, Rai stage, β2 microglobulin (β2M) level, lactate dehydrogenase (LDH) level, gender, age, lymphocyte percentage, IGHV mutation status, cytogenetic abnormalities such as del(17p) or P53 mutation (TP53 aberration), del(11q), del(13q) and trisomy12, absolute lymphocyte count (ALC), and bulky disease (≥ 5 cm) ( Figures 5A–N ). Notably, decreased TCF1 expression was significantly associated with relapsed and refractory disease (P = 0.001; Figure 5A ), Rai stage 3-4 (P = 0.012; Figure 5B ), and β2M ≥ 3.5 (mg/L) (P = 0.009; Figure 5C ), which mainly serve as clinical indexes for disease progression and adverse prognosis. In relapsed and refractory CLL patients, we found there was no significant difference in the effect of first-line treatment regimens on TCF1 expression except for a slight upward trend in the ibrutinib group, which may be related to the small number of cases in different groups ( Table S1 ).

Cox regression analysis demonstrated that low TCF1 and BCL11B expression are independent predictive factors for short TTFT of CLL patients ( Table 1 ). Specifically, in univariate analysis, ≥ 65 years old, female, high β2M level, high LDH level, unmutated IGHV, TP53 aberration, del(11q), trisomy12, Rai stage 3-4, R/R, low TCF1, and low BCL11B were significant risk factors for short TTFT. Statistically significant factors for TTFT (P < 0.05) were included in the multivariate analysis, revealing that unmutated IGHV, TP53 aberration, trisomy12, Rai stage 3-4, R/R, low TCF1, and low BCL11B were independent risk factors for shortened TTFT. Thus, TCF1 is a potential clinical biomarker for predicting disease progression for CLL patients.

To explore the reduced expression of TCF1 in various T cell subgroups, we subsequently detected the percentage of TCF1+ cells in the CD3+, CD3+CD4+, and CD3+CD8+ T cell subgroups of 33 CLL patients and compared these with HIs. We restricted all analyses to CD3+ T cells ( Figure 6A ). The results demonstrated that the percentages of TCF1+ cells in the CD3+, CD3+CD4+, and CD3+CD8+ T cell populations from CLL patients were significantly lower than that in corresponding T cell subgroups from HIs (P values are 0.004, < 0.001 and < 0.001, respectively), particularly for CD3+CD8+ T cells ( Figures 6B, C ). We further compared the percentages of TCF1+ cells among the different T subgroups and found that TCF1+ cells had a higher percentage in CD3+CD4+ T cells irrespective of CLL patients or HIs ( Figure 6C ). Notably, the percentage of TCF1+ cells in the CD3+CD8+ T cell population was significantly lower than that in the CD3+CD4+ T cell population, particularly in CLL patients. It has been reported that TCF1+CD8+ T cells promote effective antitumor immunity (40). Based on previous studies and our current findings, decreased TCF1+CD8+ T cells in CLL patients indicates T cell immune dysfunction. Therefore, TCF1 has the potential to be a biomarker for T cell immune status and a therapeutic target in CLL patients.