Single-Cell Transcriptomics Reveals Immune Reconstitution in Patients with R/R T-ALL/LBL Treated with Donor-Derived CD7 CAR-T Therapy

Despite intensive multiagent cytotoxic chemotherapy regimens, fewer than 40% of adults and 75% of children with T-cell acute lymphoblastic leukemia (T-ALL) survive beyond 5 years (1, 2). T-cell lymphoblastic lymphoma (T-LBL) is considered the same disease as T-ALL, which differs in the degree of bone marrow (BM) infiltration (3). Allogeneic hematopoietic stem cell transplantation (allo-HSCT) can improve survival in first complete remission (CR), which is the only curative option for patients with relapsed/refractory (R/R) T-ALL (4). However, approximately 25% of patients with T-ALL experience relapse after allo-HSCT, indicating a poor prognosis and a substantial unmet therapeutic need (4, 5).

Chimeric antigen receptor T (CAR-T) cell therapy for ALL has been proven by the FDA for several years (6). CD7 is continuously expressed on the surface of T-ALL cells at higher levels than that on normal T cells, making it a feasible CAR-T therapeutic target (7). Efficient strategies have been developed to avoid CAR-T cell fratricide due to the expression of CD7 on the surface of CAR-T cells, which maintains viability and proliferation of CD7 CAR-T cells (7–10). Even though several phase I clinical trials claimed the manageable safety and potential efficacy of autologous (10, 11) or donor-derived (10, 12–15) CD7 CAR-T therapies for T-ALL/LBL, a couple of patients with T-ALL/LBL treated with CD7 CAR-T therapies encountered infection, relapse, and mortality, and the inducements remain to be addressed.

Here, we initiated a phase I clinical trial to investigate the safety and efficacy of CD7 CAR-T therapy, and found that CD7⁺ normal T cells were depleted and CD7⁻ T cells expanded in patients with T-ALL/LBL treated with CD7 CAR-T therapy. By other reports, Cd7 knockout mice develop normal lymphoid organs and immune responses, and Cd7 knockout T cells showed comparable antivirals function in vitro, suggesting CD7 is dispensable for T-cell development and functions (16–19). However, the function of CD7⁻ T cells repopulated after CD7 CAR-T treatment and the functional recovery of other immune cells in actual patients with T-ALL/LBL after serial therapies remain unknown. Accordingly, we evaluated the immune reconstitution via single-cell RNA sequencing (scRNA-seq) analysis of peripheral blood mononuclear cells (PBMC) of several canonical cases collected before versus after CD7 CAR-T infusion, which might provide clinicians with optimal treatment decisions for R/R T-ALL/LBL.

This open-label, phase I clinical trial (ChiCTR2200058969) was conducted to evaluate the safety and efficacy of donor-derived CD7 CAR-T cell therapy for patients with a pathologic confirmed CD7⁺ R/R T-ALL/LBL, as defined by the National Comprehensive Cancer Network (NCCN) Clinical Practice Guidelines 2021 for ALL (20). Eastern Cooperative Oncology Group (ECOG) status of participants was 0 to 2, without graft-versus-host disease or need for immunosuppressive agents, and without cancer other than T-ALL/LBL within 5 years, or severe hepatitis, cardiac and systemic diseases, or infections. The criterion of 5-year-other-cancer-free is used to distinguish cancer caused by CAR-T therapy from recurrency of other original cancers. Lymphodepletion (fludarabine at 30 mg/m²/day and cyclophosphamide at 250 mg/m²/day) was conducted on days −5, −4, and −3 before infusion. After 6-day manufacturing, the patients received CD7 CAR-T cells at the target dose of 1 × 10⁵/kg. The study protocol was approved by the Ethics Committee of Beijing Gobroad Boren Hospital compliant with the Declaration of Helsinki. Written informed consent was obtained from all study participants or their guardians.

The study's primary endpoint was to evaluate the safety of CD7 CAR-T cell therapy up to at least day 28 post-infusion. Cytokine release syndrome (CRS) and immune effector cell–associated neurotoxicity syndrome (ICANS) were graded according to the American Society for Transplantation and Cellular Therapy Series (ASTCT) Consensus (21). Other adverse events were graded with Common Terminology Criteria for Adverse Events (CTCAE-version 5.0; ref. 22). The secondary endpoints were efficacy and survival, and pharmacokinetic (PK) and pharmacodynamic (PD) parameters of CD7 CAR-T cells in patients. Efficacy assessment was evaluated on the basis of the NCCN Guidelines 2021. Minimal residual disease (MRD) was assessed by flow cytometry (FCM; sensitivity of threshold is 1 × 10⁻⁴). Response of extramedullary disease (EMD) was assessed by PET-CT and/or contrast CT and MRI. The evaluation of extramedullary and intramedullary disease was performed on day 28 after CAR-T cell infusion, and every 3 months thereafter. The proliferation and persistence of CD7 CAR-T cells were tested by qPCR and FCM (#564019, BD Biosciences). The variations in cytokines were detected by cytometric bead array (CBA)-based FCM (#551809, BD Biosciences and #740784, BioLegend).

According to the manufacturer's introduction, scRNA-seq libraries were constructed using Single Cell 3′ Library. The libraries were sequenced using an Illumina Novaseq6000 sequencer with a sequencing depth of at least 20,000 reads per cell with a pair-end 150 bp (PE150) reading strategy (performed by iomics)

Sequencing data alignment and gene expression matrix calculation were performed by cellranger 5.0.1 (10× Genomics). The reference genome was generated by combining human reference genome build 38 (hg38) with human immunodeficiency virus 1 (NC_001802.1). Human gene annotation version 32 (Ensembl 98) together with the annotation of HIV1gp1 were also combined for identifying expressed genes. Single-cell expression matrix of PBMCs from the healthy donor was downloaded from 10× Genomics website. Cells with a percentage of mitochondrial genes >20%, expressed genes <300 or >10,000, and estimated as doublets by scds package were removed from further analysis (23).

The Seurat package (version 4.1.1) was adopted for unsupervised cell clustering (24). The SCTransform module was used for batch effect correction (25). To get robust cluster results, we performed two rounds of clustering. For the first round, after removing mitochondrial and ribosomal genes, the 3,000 most variable genes were selected for integrating data from different samples and for principal component (PC) analysis. The first 30 PCs were used for identifying cell clusters with a resolution of 0.1. The UMAP was used for visualizing clustering results. The SingleR package (RRID:SCR_023120) was used for defining the cell type according to their similarity to Monaco immune data of celldex package (26). Cells annotated as granulocytes or as different major cell types from the majority of their belonging cluster were removed. Then, the second round of clustering was performed using the same parameters except for the resolution of 0.2. The final immune cell types were annotated by the expression of specific marker genes.

T cells and CD14⁺ monocytes were subjected to subclustering with a resolution of 0.4. T-cell subclusters were annotated by specific marker genes. The CAR-T cells were defined as cells with reads from the HIV1gp1 gene. The CD7⁻ T cells were defined as cells with no read from CD7.

Each subcluster of CD14⁺ monocytes was downsampled to 200 cells. Then, the cytoTRACE package was used for estimating the developmental potential of each subcluster (27). Genes were ranked by their correlation with the predicted developmental order of each cluster, and the top 100 positively correlated genes and the top 100 negatively correlated genes were functionally annotated by clusterProfiler (version 4.4; ref. 28).

The expression percentage of each gene sets calculated by the “PercentageFeatureSet” function of Seurat was defined as the gene set expression level for every cell. Student t test was used to test the significance between CD7⁻ and CD7⁺ T cells.

GSEA was performed by GSEA (version 4.2.3, RRID:SCR_003199). The pathways used for GSEA were downloaded from MsigDB catalog c2 KEGG gene sets. T-cell subclusters with more than 100 cells in all samples were tested by GSEA. Pathways with adjusted enrichment P-value <0.01 were defined as significant up-/downregulated. Enrichment heatmaps for pathways showed significance in at least three comparisons drawn by the pheatmap package. Pathway networks according to their overlapping genes were built by Preselected Functions (default parameters, Edge-weighted Force-directed. BioLayout for CluePedia) within ClueGO (version 2.5.9, RRID:SCR_005748) in Cytoscape (version 3.9.1, RRID:SCR_005748).

Clinical safety, efficacy, and post-infusion follow-up data are summarized and analyzed by using descriptive statistical methods. GraphPad Prism 9.0 (RRID:SCR_002798) was used to conduct the data analysis.

All data generated or analyzed during this study are included in this published article and its supplementary information files. The scRNA-seq data that support the findings of this study are available in BioProject (https://ngdc.cncb.ac.cn/bioproject/browse/PRJCA011733) and Genome Sequence Archive (https://ngdc.cncb.ac.cn/gsa-human/browse/HRA003023), and are also available from the corresponding author upon reasonable request.

Seven eligible patients were enrolled between November 2020 and December 2021, and followed up until May 1, 2022 (Supplementary Table S1). Donor-derived CD7 CAR-T cells were infused into 7 patients. The median age of the patients was 30 (range 28–38) years old, and all patients had a history of allo-HSCT before enrollment. Patients were heavily pretreated with a median of 4 (range: 3–8) lines of chemotherapies before allo-HSCT. The median duration from allo-HSCT to relapse was 12 months (range 5–24). Five patients had high-risk mutations or gene fusion, namely MLL break, SETD2, NOTCH1, BCOR, JAK1, RB1, KIT, KRAS, PHF6, and JAK3, at initial diagnosis (Supplementary Table S2). At enrollment, 4 patients were diagnosed with T-LBL, and 3 patients were diagnosed with T-ALL among which the median proportion of blasts was 94% (range 0.12%–98.2%) in BM. Two patients with T-ALL and 4 patients with T-LBL developed EMD, including diffuse involvement (n = 1), bulky mediastinal masses >7 cm in diameter (n = 1), and central nervous system (CNS) involvement (n = 3), and high-risk subtypes included early T-cell precursor (ETP)-ALL (n = 3, 42.9%) and primary refractory disease (n = 5, 71.4%). Among the patients with CNS involvement, 2 were T-ALL and 1 was T-LBL. All patients received at least one-line bridging chemotherapy before CAR-T cells infusion and 1 patient received radiotherapy on mediastinal lymph nodes.

Adverse events (AE) were summarized (Supplementary Tables S3 and S4). The rate of CRS and ICANS was 85.7% (n = 6) and 14.3% (n = 1), respectively. Only 1 patient (Case 2) had a grade 3 CRS and a serious grade 4 ICANS. The other 5 patients experienced grade 1 CRS, and all other patients had no ICANS. The median time to onset of CRS was day 5 (range 1–19). All patients' median duration time of CRS was 8 days (range 5–21).

Only 2 patients (28.6%) developed graft versus host disease (GVHD). Case 1 developed grade 1 intestinal chronic GVHD (cGVHD) resolved by prednisone acetate and ruxolitinib at month 6 after CAR-T cells infusion, which was evaluated unrelated to CAR-T treatment. Case 4 experienced grade 3 intestinal acute GVHD (aGVHD) at month 3 after CAR-T cells infusion, which was considered induced by intestinal infection, and finally died of intestinal aGVHD at month 4.

All patients had normal blood cell indices before enrollment. Hematologic grade 4 CAR-T cell–related AEs including leukopenia, thrombocytopenia, and neutropenia were observed in 6 patients (85.7%), among which 5 patients (5/6, 83.3%) had grade 4 neutropenia and 5 patients (5/5, 100%) had grade 4 leukopenia. Six patients (6/7, 85.7%) had thrombocytopenia, among which 4 patients (4/6, 66.7%) were grade 4, who had prolonged thrombocytopenia and did not receive platelet recovery until the cut-off time. The median (range) times to onset of grade 4 leukopenia, grade 4 neutropenia, and grade 4 thrombocytopenia were day 7 (day 0–day 21), day 4 (day 0–day 21), and day 7 (day 0–day 30), respectively. The median duration of leukopenia and neutropenia was 16 and 17 days, respectively. Among 6 patients with grade 4 thrombocytopenia, 4 patients did not recover from grade 4 thrombocytopenia till the end of the study visit, and the other 2 patients had 16 and 17 duration days.

Infections occurred in 5 patients (71.4%). Virus activations occurred in 4 patients (57.1%) and bacterium occurred in 3 patients (42.9%). No fungal infection was observed. Case 1 had pneumonia with legionella bacteria from month 10 to 12, considered related to immunosuppressant use for cGVHD and unrelated to CAR-T treatment, and received moxifloxacin and minocycline until chest CT testing appeared negative. Case 3 experienced repeatedly symptomatic Epstein–Barr virus (EBV) activation and posttransplantation lymphoproliferation disease (PTLD), resolved by rituximab and globulin. Case 4 had intestinal clostridium difficile and hemolytic staphylococcus infection at month 3, resolved by vancomycin and carbapenem antibiotic, and asymptomatic cytomegalovirus (CMV) activation at month 3.5, resolved by potassium and sodium phosphate. Case 5 had pneumonia with Enterobacter cloacae and respiratory syncytial virus at month 3, and died of pulmonary hemorrhage associated with bacterial and viral pneumonia at month 4. Case 6 had asymptomatic EBV activation from day 34 to 40.

On day 28, 4 patients with T-LBL without BM involvement at enrollment achieved CR, and 3 patients with T-ALL who had BM involvement achieved MRD-negative CR (Fig. 1A). Three patients who had CNS leukemia (Cases 2, 3, and 6) achieved CR on day 28, and 5 patients with their EMD (Cases 1, 3, 5, 6, and 7) encountered CR on day 90. The median followed-up time was 4 months (range 3–16). Three patients developed relapses during the trial: Case 1 relapsed with EMD at month 12, Case 3 relapsed with BM and EMD at month 5, and Case 6 relapsed with BM at month 3. Three patients died during the trial: Case 2 died of brain hemorrhage caused by low platelet number, Case 4 died of intestinal infection, and Case 5 died of lung infection at month 4. Four patients were alive until the cut-off time. The median progression-free survival (PFS) and overall survival (OS) were 4 and 4.6 months, respectively (Fig. 1B and C).

CAR-T cell expansion was detected by FCM and PCR testing CAR transgene in blood. CAR-T cells peaked between days 11 and 15 with a mean count of 51.1 cells/μL (range: 7.17–497) detected by FCM (Fig. 1,E). CAR-T cells of 2 patients (Cases 2 and 7) could not be detected by FCM at month 3, whereas CAR-T cells of the other 5 patients remained detectable until final assessments in blood by PCR testing (Fig. 1D and E). But a relatively low copy number of CAR-T cells exists in all patients after 3 months of CAR-T cells infusion (Fig. 1F). Serum biomarkers including IL6, IL10, TNFα, IFNγ, and soluble CD25 (sCD25) mildly increased in most patients in the course of CAR-T cells reaction, which indicated the activation of CAR-T cells after infusion (Fig. 1G–K). The absolute numbers of T/B/NK/NKT cells showed a transient decline and recovered by a low cell number on day 21, and all of these immune cells of Case 6 surged because of the relapse at month 3 (Fig. 1L; Supplementary Fig. S1). In addition, CD7⁺ T cells in the peripheral blood were eliminated in 7 patients on day 11 after CD7 CAR-T infusion, however, CD7⁻ T cells dramatically expanded in all patients (Fig. 1M and N). Further, the reduction of CD4⁺/CD8⁺ T-cell ratios, driven by the increase in the proportion of CD8⁺ T cells, might be caused by the cytotoxic effect (Fig. 1O).

To further profile immune status changes after CD7 CAR-T infusion in patients with T-ALL/LBL, PBMCs were collected from 3 patients (Cases 4, 5, and 6) before and after infusion for 3 weeks. On the basis of FCM (Fig. 2A–C) and PET-CT (Fig. 2D) analysis, Case 4 (T-ALL) and Case 5 (T-LBL) achieved CR after infusion, respectively, who then unfortunately died of severe infection after approximately 4 months. Case 6 (T-ALL) achieved short-term CR after infusion but relapsed at month 3 (Fig. 2E–G). One patient (Case 1, T-LBL) who achieved long-term CR until relapse at month 12 after CAR-T treatment (Fig. 2H), and 1 patient with no symptoms of relapse after allo-HSCT without CAR-T cell infusion (Control 1) were recruited as external controls. One PBMC scRNA-seq data of a healthy donor from 10× Genomics was adopted as health control (Control 2). In total, 61,757 cells from nine samples were clustered into 11 cell populations including four myeloid cell clusters, five T-cell clusters, one NK-cell cluster, and one B-cell cluster (Fig. 3A and B; Supplementary Table S5).

PBMC samples from short duration after CAR-T treatment showed a significant increase in the proportion of MKI67⁺ T cells (P = 0.049, K–S test; Fig. 3C), indicating the activation of T-cell proliferation right after CD7 CAR-T eliminates malignant and normal T cells. Moreover, we also observed a significant increase of CD16⁺ T cells (P = 0.049, K–S test; Fig. 3C), which have been reported to have an excessive cytotoxic function and are related to severe inflammation during COVID-19 infection (29). The phenomena may explain the system inflammation which persisted for nearly 2 weeks after CD7 CAR-T treatment.

To further evaluate the compositional changes among T cells before and after treatment, we reclustered T cells with increased resolution and identified 14 subclusters (Fig. 3D). The proportion of CAR-T cell residuals, which express lentiviral vector mRNA, decreased from 9.9% to 27.1% at week 3 to 0.2% at month 10. The major subtypes of CAR-T residuals belong to GZMK⁺ memorial T cells (11.9%–35.8%) and CD16⁺ α/β T cells (23.8%–32.9%), which indicated an enhanced memorial and cytotoxic function of CAR-T cells (Supplementary Table S5).

The CD7⁻ host T cells accounted for 34% to 52% of total T cells before treatment, and significantly expanded after treatment until accounted for 99.8% of total T cells at month 10, which is consistent with FCM results (Fig. 1M and N). By comparing the proportion of CD7⁺ host T cells with the proportion of GZMK⁺ and CD16⁺ α/β CAR-T cells in posttreatment samples, we found a significant negative correlation (R = −0.920, P = 0.0399, Pearson correlation; Fig. 3F), which indicated the T-cell–eliminating function of these CAR-T cells. The highest naïve CD8⁺ CAR-T cells observed in the sample with the highest CD7⁺ host T cells further indicated the ongoing T-cell–eliminating process in this patient.

GZMH⁺ T cells became the most abundant type (20.4%–48.7%) after treatment and only accounts for 7% of T cells of Control 2 (Fig. 3E). These cells were reported to exhibit a significant clonal expansion and played a pathogenic role in lupus patients (30). Moreover, we observed a more than three times decline in CD4⁺ naïve T cells posttreatment in each of the 3 patients (Fig. 3E). This phenomenon was also seen in lupus patients of Asian origin (30). These observations indicated the immune reconstruction process after CD7 CAR-T treatment may have similarities with the autoimmune response which is characterized by lymphopenia and “epitope spreading” (31).

We further compared the expression level of T-cell receptors (TCR), T-cell activation pathway, and TCR signaling pathway before and after infusion in four major T-cell subtypes, including GZKH⁺, GZMK⁺, naïve CD4⁺, and naïve CD8⁺. Nearly in all T-cell subtypes, the three gene sets showed equal or higher expression levels in posttreatment samples, except for the T-cell activation pathway in CD4⁺ naïve T cells of Case 5 (Fig. 4A).

Furthermore, GSEA was adopted to explore the functional differences between CD7⁻ and CD7⁺ cells from pretreatment samples on canonical pathways level. CD7⁻ T cells showed similar function to CD7⁺ T cells in either patients with T-ALL/LBL before CD7 CAR-T treatment or the health donor, as only limited enriched pathways are shared by different samples (Supplementary Fig. S2; Supplementary Table S6). We further built pathway networks to explore the hidden relationship among enriched pathways (Fig. 4B). The absence of T-cell function and immune system–related pathway in the enrichment results further supported that CD7 does not make a pivotal contribution to T-cell function (8).

We then compared the functions of CD7⁻ T cells before and after treatment. As shown in Fig. 4C, CD7⁻ T cells showed significant activation of lupus-related functions after treatment, which is consistent with the compositional changes of T cells described above. These pathways formed the largest module in the network (Fig. 4B). Interestingly, the activation of autoimmune-related function was also found by the comparison between Case 1 whose sample collected at the 10th month after infusion with non-recurrent control (Control 1) and healthy PBMC (Control 2), respectively (Fig. 4B; Supplementary Fig. S2; Supplementary Table S6). These data indicate that exposure to autoantigens may lead to a permanent imprint during immune reconstruction.

We also evaluated the long-term effects of CD7 CAR-T treatment by comparing CD7⁻ T cells of Case 1 at month 10 after infusion with those of Control 1 and Control 2. Besides the above-mentioned activation of lupus-related pathways, Case 1 also showed an inhibition of MTOR and ERBB signaling–related network modules (Fig. 4B; Supplementary Fig. S2; Supplementary Table S6). As reported previously, MTOR signaling can suppress memory CD8⁺ T-cell differentiation and function (32), and such inhibitions may lead to activated T-cell function in Case 1.

We then compared the expression level of four major cytokines including IL6, IL10, TNF, and INFG between post- and before-treatment samples and between CD7⁻ and CD7⁺ T cells. As shown in Supplementary Table S7 and Supplementary Fig. S3A, there was no significant expression difference between the comparisons in the majority of T-cell subtypes. Only INFG showed significant upregulation in Naive CD8 and Naive CD4 T cells in posttreatment samples and IL10 showed significant upregulation in Naive CD4 T cells in the CD7⁻ subgroup. Furthermore, the levels of multiple proinflammatory factors including IFNγ, G-CSF, GM-CSF, IL1α, IL1β, MIF, and TNFα were significantly higher in PBMCs collected from Case 7 under stimulation with CMV peptide for 24 hours than control treatment in vitro (Supplementary Fig. S3B–S3E). These results indicated the reconstructed CD7⁻ T-cell population after CD7 CAR-T treatment had competent immune functions.

Infection is a grievous AE after CD7 CAR-T cell infusion, which occurred in the other research as well (12). As shown in Fig. 4C, T cells of Case 5, who died from a severe infection after treatment, showed a significant cytokine function inhibition; in contrast, Case 6 with no symptom of infection showed enhanced chemokine signaling after treatment. Besides that, a tremendous reduction of monocyte numbers was found in Cases 4 and 5 (Fig. 5A), indicating the severe infection of these patients was caused by the failure of monocyte reconstitution.

To test the possible explanations of monocyte loss after CD7 CAR-T treatment, we further performed a detailed analysis of CD14⁺ monocytes. These cells can be divided into eight subclusters, among which cluster 4 showed the highest proportion in Case 6 before infusion who showed no infection without monocyte change after treatment (Fig. 5B and C; Supplementary Table S5). This cluster was predicted to be the least differentiated by cytoTRACE (Fig. 5D; ref. 27). The genes positively correlated with developmental potentials were overrepresented in antigen processing functions, and in the contrast, the negatively correlated genes were enriched in leukocyte differentiation and activation functions (Fig. 5E and F). The negative correlation between developmental potential with AP-1 (JUN and FOSB; Fig. 5E), a critical transcription factor during monocyte to macrophage differentiation (33), further supported the less differentiated state of cluster 4. Such data suggested that the lack of high developmental potential monocytes in pretreatment samples might lead to the impairment of monocyte reconstruction after CD7 CAR-T therapy, resulting in infectious AEs.

We compared the expression profiles of single T cells to reveal the possible relapse-related gene signatures in patients with T-ALL/LBL before and after CD7 CAR-T cell infusion. The relapse group contained pre-infusion samples of the three patients (Cases 4, 5, and 6), who were experiencing R/R T-ALL/LBL after allo-HSCT, and the post-infusion sample of Case 6, who experienced immediate relapse at month 4 after CD7 CAR-T treatment. Post-infusion samples of Cases 4 and 5, as well as Control 1, were recruited as non-relapsed cases. Accordingly, S100A8, S100A9, and LYZ were identified as potential universal relapse markers regardless of the therapy strategies, due to their significant upregulation in relapse samples in every subtype of T cells (Fig. 6A). Of note, S100A8 and S100A9, encoding calprotectin in a heterodimer form, are expressed mainly in myeloid cells and absent from lymphocytes in health conditions (34). Especially, such two genes were most upregulated in relapse ALL (35). In pre-CD7 CAR-T treatment samples of Cases 4, 5, and 6, the expression levels of S100A8 and S100A9 were significantly higher than the nonrelapse Control 1 and healthy Control 2 (P < 10⁻¹⁶, Wilcox-rank test; Fig. 6A). S100A8 and S100A9 expression reached the lowest level in post-CD7 CAR-T treatment samples of Cases 4 and 5, who showed no symptoms of relapse. In contrast, the highest expression of S100A8 and S100A9 were observed in posttreatment PBMCs of Case 6 (Fig. 6A). As shown in Fig. 6B, similar expression profiles of these three genes were also observed in B and NK cells, which indicates the upregulation of these genes may begin at lymphoid progenitors.

This open-label, phase I clinical trial involved 7 patients suffering R/R T-ALL/LBL after allo-HSCT; all patients achieved CR in their BM at month 1 and remained in CR for at least 3 months, which illustrated the efficacy of donor-derived CD7 CAR-T therapies for R/R T-ALL/LBL. However, several patients in our study encountered immunological insufficiency, infection, relapse, and mortality. Two patients died of infections in this trial: Case 4 died of intestinal aGVHD induced by intestinal infection at month 4, and Case 5 died of pulmonary hemorrhage associated with bacterial and viral pneumonia at month 4. Underlying the inducement of such severe infection, we compared the PBMCs of Case 4 and Case 5 with patients that did not suffer infection at sampling time by scRNA-seq. Without significant difference in T-cell function, the main reason for the severe infection seems to be the loss of monocytes after CAR-T therapy, because the recruitment of monocytes is essential for the clearance of bacterial, protozoal, fungal, and viral infections (36). Thus, to prevent posttreatment infection, the protection from loss of monocytes may be a challenge clinically.

Because CD7 is universally expressed in normal T and NK cells, whether CD7⁻ cells can fulfill the immune cell function after CD7 CAR-T treatment becomes the major question of this therapeutic strategy. In our study, when CD7⁺ T cells in the BM, including normal and abnormal T cells, were eliminated in all patients on day 11 after CD7 CAR-T infusion, CD7⁻ T cells dramatically increased in all patients, which was in keeping with the significant increase in the proportion of MKI67⁺ T cells, determined by scRNA-seq. Meanwhile, our observation indicates that in gene expression level, the reconstructed T-cell populations may have normal or even enhanced immune functions than pretreatment T cells, which was supported by the higher expression levels of TCRs, T-cell activation pathway, and TCR signaling pathway in posttreatment samples. In addition, the patient at his 10th month after CAR-T treatment showed inhibition of MTOR signaling-related pathways in comparison with healthy control and nonrelapsed T-ALL/LBL control, further supported the enhanced memory CD8⁺ T-cell function in that patient (32). Thus, our present work, for the first time, indicated that CD7⁻ T cells had a considerable proliferation ability and immunological function in patients with T-ALL/LBL after CD7 CAR-T treatment, by single cell level evaluation.

The main characteristics of the reconstructed T-cell populations are their similarity with T cells in autoimmune diseases, including the expansion of GZMH⁺ T cells, decline of Naive CD4⁺ T cells, and activation of autoimmune disease-related pathways, which indicated the immunologic competence of T cells after infusion as well. This character can even be seen in a patient in his 10th month after treatment, which can be explained by the exposure to autoantigen generated by the recognition and elimination of CD7 CAR-T cells to normal CD7⁺ T cells. This character also raises the possibility that although we observed a higher level of TCRs, their diversity may significantly decrease in posttreatment samples (37). Because a broad TCR repertoire and T-cell diversification are important for immune function, further investigations on measuring TCR diversity at the single cell level should be performed to get a comprehensive measurement of immune function of posttreatment samples (38).

We observed consistent upregulation of S100A8 and S100A9 in leukocyte lineage as a signature of the patients with relapsed T-ALL/LBL. S100A8 and S100A9 are proinflammatory factors mainly expressed by monocytes and neutrophils and upregulated during inflammatory processes, such as trauma, infections, and autoimmune diseases (39). Their expressions keep at low levels in T cells from the healthy donor according to data from the Human Protein Atlas. In various cancers, they are linked with progression, metastasis, and poor prognosis. In chronic lymphocytic leukemia, upregulated S100A8 is associated with rapid progression (40). High serum level of S100A8 and S100A9 is related to poor treatment response in T/NK lymphoma and Hodgkin lymphoma (41, 42). In breast cancer, S100A8 and S100A9 released by myeloid-derived suppressor cells are critical for the formation of premetastatic niches (43). In our relapsed samples, the nonmalignancy leukocyte lineage cells, which are indicated by the absence of genomic copy number alterations (Supplementary Fig. S4), showed high expression of S100A8 and S100A9 with significant correlation among different cell clusters (Fig. 6B). This observation suggests the upregulation of these two genes may occur at leukocyte progenitor cells in lymph nodes and form a microenvironment to promote relapse. Although the reasons for these upregulations need further investigation, their high expressions in leukocyte lineage can act as molecular markers of relapsed T-ALL. In addition, the fact that both R/R T-ALL/LBL after allo-HSCT and after CD7 CAR-T show high expression of these genes suggests their universal marker role no matter which therapy strategy is adopted.

In conclusion, our work provided the cellular level dynamics of immune homeostasis of CD7 CAR-T treatment in real patients, which is valuable for treatment decisions of R/R T-ALL/LBL. However, TCR diversity examination of the reconstructed CD7⁻ T cells after CD7 CAR-T infusion and monocyte loss mechanism deserve more investigations. To further evaluate the efficacy of CD7 CAR-T cell therapy, long-term clinical observations by recruiting more patients are undergoing.

No author disclosures were reported.

W. Chen: Data curation, software, formal analysis, funding acquisition, validation, visualization, methodology, writing–review and editing. H. Shi: Resources, data curation, formal analysis, validation, investigation, writing–original draft. Z. Liu: Data curation, formal analysis, validation, visualization, writing–original draft, writing–review and editing. F. Yang: Conceptualization, resources, data curation. J. Liu: Writing–review and editing. L. Zhang: Writing–review and editing. Y. Wu: Writing–review and editing. Y. Xia: Writing–review and editing. Y. Ou: Writing–review and editing. R. Li: Resources. T. Zhang: Resources. J. Zhang: Supervision. X. Ke: Supervision, project administration. K. Hu: Conceptualization, resources, supervision. J. Yu: Conceptualization, supervision, funding acquisition, writing–original draft, writing–review and editing.

We thank all the patients participating in the study for their courage and generosity. J. Yu reports grants from the National Natural Science Foundation of China (No. 81970186) and the Beijing Advanced Innovation Center Research Foundation (No. 10287) during the conduct of the study. W. Chen reports grants from the National Natural Science Foundation of China (No. 82141113) during the conduct of the study. No disclosures were reported by the other authors.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.