To identify CAR-T cells, we used recombinant PE-conjugated CD19 protein for the first eight patients due to the unavailability of anti-FMC63 antibody, which was then used for the remaining patients. Example of staining is presented in Supplementary Figure S1 . CAR-T cells were further stained with a multicolor antibody panel to determine their differentiation immunophenotype using antibodies against antigens CD3, CD4, CD8, CD14, CD45RA, CD62L, CD27, CD28, PD-1, and TIGIT (11). Additionally, samples were analyzed with a second antibody panel to determine the composition of all significant leukocyte subsets using antibodies against antigens CD3, CD4, CD8, CD14, CD16, CD19, CD45, CD56, and TCRgd. Figure 1 shows an example of typical CAR detection. Gating strategy of CAR-T memory subsets is similarly presented in Figure 2. Pre-gating on CD3⁺ cells is shown in Supplementary Figure S2. Controls for CAR staining (healthy donor) and immunophenotype staining (FMOs, healthy donor) are presented in Supplementary Figures S3, S4, respectively. Threshold for the detection of CAR⁺ cells was set at 0.1% out of CD3 T cells, and the percentage of CAR-T cells was further converted to the absolute counts per microliter using the total white blood cell count values obtained from hematology analyzer. The antibody panels are presented in Supplementary Table S1 .

The first goal was to characterize patients’ apheresis and to determine whether differences in their T cell memory subsets are associated with stronger in vivo expansion of CAR-T cells and improved clinical outcome. For each patient, the apheretic product samples were cryopreserved, as well as from five healthy donors functioning as control samples. Prior to flow cytometry analysis, the cells were thawed and let to rest in media overnight. We have found out that 27 out of 32 subjects had undetectable B-cells in apheresis due to preceding treatment by rituximab (Figure 3A, Supplementary Figure S2). To characterize the T cells, we determined their immunophenotype with the differentiation antibody panel (omitting the anti-FMC63 antibody). The patients’ T cells were characterized by highly variable numbers of Tem/Tcm/Temra CD8⁺ and CD4⁺ memory subsets, similarly to what was detected in healthy donors (Figure 3B, Supplementary Figure S6). Nevertheless, the patients’ T cells contained more exhausted T cells expressing PD-1 receptor compared to healthy donors—50% vs. 12% CD4⁺ (p = <0.001) and 35% vs. 14% CD8⁺ (p = 0.046). In addition, the non-responders were characterized by significantly reduced numbers of CD8⁺ CD45RA⁺ CD27⁺ T cells (Figure 3C) in the apheresis, in agreement with report by Fraietta et al. (12).

Subsequently, the manufactured CAR-T cells were analyzed using aforementioned antibody panel with added anti-FMC63 antibody. The samples for measurements were obtained from discarded infusion bags after administration to the patients. The infusion bags and their filters were thoroughly washed with PBS to acquire the remaining cells. In some cases, we were unable to measure administered products for technical reasons. We observed that the products contained a highly variable percentage of CAR⁺ cells (range 1%–40%) and that the majority of CAR-T cells were CD4 positive (Figure 4A). There was no significant difference of percentage of CD4⁺ or CD8⁺ CAR-T cells in product between responders and non-responders (Figure 4B). The immunophenotype of CAR-T cells was likewise highly variable between individual patients in regards to the proportion of Tscm/Tcm/Tem/Temra, and CD27/CD28/PD-1/TIGIT positive cells (Figure 4C, Supplementary Figure S7). There was a slightly higher PD1+TIGIT− percentage of CD4⁺CAR⁺ T cells in responders (p = 0.039, Figure 4C). CAR⁺ T cells in the product had almost identical immunophenotype as the CAR⁻ population (data not shown), suggesting that their differentiation during manufacturing was driven more by polyclonal anti-CD3/CD28 activation than by CAR signaling. One product (P31) did not meet the entry specifications since it contained only 1% of CAR-T cells. Despite such a low percentage of CAR-T cells, this patient responded to the therapy and achieved PR by PET/CT (nonetheless relapsed later). Conversely to apheresis, we did not detect a significant difference in the numbers of CD8⁺ CD45RA⁺ CD27⁺ T cells in the product (Figure 4C).

Expansion and immunophenotype of CAR-T cells were determined in the patients’ blood samples after administration at three timepoints—between days 2–4 (T1), days 10–14 (T2), and days 30–60 (T3). Since the study was performed in a non-clinical trial setting, the samples could be obtained only during regular medical examinations including signing of the informed consent form by the patient. To determine the kinetics of CAR-T cells’ expansion, we measured percentage of CD4⁺/CD8⁺ CAR-T out of CD3⁺ T cells and calculated their absolute counts in blood. The CAR-T cell percentages were compared between T1 and T2 to assess CAR-T cells’ expansion kinetics. If the percentage of either CD4⁺ or CD8⁺ CAR⁺ at T2 was higher than their percentage at T1, the patient was categorized as showing a detectable expansion. Figure 5A shows the kinetics of CAR-T cell expansion for each patient. In addition, we calculated the absolute number of CAR-T cells per µl of blood in the responders and the non-responders at T1 and T2 (Figure 5B). Table 1 shows the measured values of CAR-T cells in blood for each patient and demonstrates that only 27% (3/11) of non-responders had a detectable expansion of CAR-T cells and that 90% (19/21) of responders had detectable expansion (Table 2). Out of 25 patients with DLBCL, 10 subjects had no expansion of CAR-T cells (all patients with B-ALL were responders with detectable CAR-T cell expansion). 80% of these “non-expanders” were also non-responders, but only 40% of them had primary refractory disease. Following this, we determined the immunophenotype of detected CAR-T cells in the blood similarly as was done for the product (Supplementary Figure S8). Threshold for the detection of CAR-T cells was set at 0.1% of CD3⁺ T cells, samples with CAR⁺ percentage below this value were not analyzed. Significant differences between responders and non-responders are presented in Figure 6A. At time point 2, responders had higher numbers of CD8⁺CAR⁺ Tcm (p = 0.030) and PD1−TIGIT− than non-responders (p = 0.048). At time point 3, CD28⁺CD27⁻ percentage was higher in non-responders (p = 0.025), although only 3 samples were used for the analysis. Clustering by EmbedSOM algorithm (13) shows relative distribution of memory subsets of CAR-T cells for time point 1 (Figure 6B).

In addition to blood samples, we analyzed samples obtained from bone marrow and in three subjects with lung infiltration also from the malignant pleural effusion. Figure 7A shows percentages of CAR-T cells detected at a single time point in blood with corresponding bone marrow or pleural effusion (lung) sample. Interestingly, we observed effective migration of CAR-T cells into the tumor sites, demonstrated by high CAR-T percentage in the pleural effusion. We also detected long-term persisting CAR-T cells in the bone marrow of several subjects.

Immunophenotype of CAR-T cells in bone marrow was compared to paired blood samples (Figure 7B). CD8⁺ CAR-T cells in blood had elevated Tcm (p = 0.038), Temra (p = 0.046), and CD45RA⁺CD27⁻ (p = 0.032) phenotypes compared to bone marrow. CD8⁺ CAR-T cells in bone marrow had slightly higher percentage of PD1+TIGIT+ (p = 0.003), PD1+TIGIT− (p = 0.028) populations, and lower PD1−TIGIT+ (p = 0.004) percentage. Phenotype of CAR-T cells in pleural effusion was not statistically analyzed due to insufficient number of paired blood samples with detectable CAR-T cells (Supplementary Figure S9).

Four patients from the selected study group were administered a second dose of CAR-T cells after they relapsed; the source of the CAR-T cells was a remaining cryopreserved bag of the original product. Subject P09 received the second dose simultaneously with anti-PD-1 antibody pembrolizumab. Subject P16 received the second dose concurrently with lenalidomide treatment. Subjects P18 and P19 were pediatric B-ALL patients in clinical remission, although B-cell presence in the peripheral blood was established, suggesting the therapy’s failure. Therefore, the second CAR-T cells’ dose was used successfully as a bridging therapy before allo-HSCT. Furthermore, we observed that the subjects P16, P18, and P19 had a non-detectable expansion of the second CAR-T cells’ dose. In contrast, in the subject P09, we observed high levels of CAR-T cells in the pleural effusion even though they were almost undetectable in the blood sample. Figure 7C shows percentage of CAR-T cells at D+14 for first and second administration. The immunophenotype of these CAR-T cells was very similar to the immunophenotype of CAR-T cells detected in the patient’s blood sample following the administration of the first dose. Unfortunately, despite such a very efficient expansion of CAR-T cells, subject P09 did not achieve complete remission (CR) and the disease continually progressed.

Samples were acquired from patients diagnosed with DLBCL and B-ALL who were treated with tisagenlecleucel at the Institute of Hematology and Blood Transfusion, General University Hospital in Prague, and University Hospital in Motol. All patients (or their parents/guardians) signed informed consent form. Following samples were collected from each patient: 1) apheresis used for CAR-T cell manufacturing, 2) CAR-T-cell product from infusion bag, 3) whole blood sample from patients following CAR-T infusion at three time points—first early after administration (days 2–4), second at the time of expected maximal CAR-T-cell expansion (days 10–14), and third after the retraction of the immune response (days 30–60). The control samples were obtained from age-matched healthy volunteers. In addition to FACS analysis, a total blood cell count was determined as part of a regular medical examination. Peripheral blood mononuclear cells (PBMCs) were isolated from whole blood samples by gradient centrifugation using Ficoll-Paque premium (GE Healthcare). After isolation, PBMCs were resuspended in PBS and were either stained immediately or frozen in CryoStor CS10 (StemCell Technologies) for further processing. For unfreezing, cells were thawed and cultivated in cell culture media overnight. For all experiments, CellGro media (CellGenix, Germany) supplemented with 10% heat-inactivated fetal calf serum (Gibco, United States) was used with addition of antibiotics penicillin and streptomycin (Gibco, United States).

The study group includes patients with pediatric B-ALL (n = 5, age 3–10), adult B-ALL (n = 2, age 25–27), or with DLBCL (n = 25, age 34–77) who were eligible for treatment with tisa-cel (Table 1). In total, 32 patients received the treatment under non-clinical trial settings according to recommended clinical practice. All patients with B-ALL were in complete remission before infusion of CAR-T cells. In contrast, 24 out of 25 (96%) of the remaining patients with DLBCL had active disease with a significant tumor burden. After administration of CAR-T cells, the patients were monitored, and when necessary, treated for CRS/neurotoxicity symptoms with corticosteroids or tocilizumab according to recommended CAR-T cells’ medication protocols. A grade 2 CRS was observed in 5 patients, which required hemodynamic support. However, no patient required mechanical ventilation, all rapidly recovered from CRS symptoms. The effects of CAR-T cell therapy were determined at three and six-months’ time points via PET/CT (18) and by flow cytometry. Those patients who either progressed or failed to achieve at least a PR at 3 months were classified as non-responders. The remaining individuals were classified as responders and were further followed. All relevant pretreatment and post-treatment parameters are presented in Table 1. Patient P28 had a significant CAR-T expansion, so we included them as a responder; however, this patient died within a month due to complications connected with primary disease. In summary, the patients’ responses to CAR-T cell treatment and associated complications were similar to the currently known clinical experience with tisa-cel (19).

Approximately five million freshly isolated PBMCs per sample were used for staining. For the first eight patients, we used PE-Labeled Human CD19 (20–291) Protein (Acro Biosystems). Later, anti-FMC63-FITC antibody by Acro Biosystems was used. Comparison of staining with protein CD19 and anti-FMC63 antibody is shown in Supplementary Figure S1. Besides detection of CAR-T cells the samples were stained with antibodies against antigens CD3, CD4, CD8, CD45RA, CD62L, CD27, CD28, CD57, PD-1, TIM-3, and TIGIT. A second detection panel against antigens CD3, CD4, CD8, CD14, CD16, CD19, CD45, CD56, and TCRgd was used to determine significant leukocyte subsets in the samples. Antibody panels used in this study were used previously (20) and can be found in Supplementary Tables S1, S2. All antibodies were titrated before use, and fluorescence-minus-one controls for selected antibodies were measured. Firstly, PBMCs were stained using a fixable blue dead cell stain kit (Thermo Fisher Scientific, United States), washed with FACS buffer, and then stained with antibody mix in Brilliant Stain Buffer (BD) at room temperature for 30 min. The samples were washed with FACS buffer prior to measurement.

Sample data were acquired on a five-laser BD LSRFortessa instrument (BD Biosciences). All measurements were standardized using 8-peak Rainbow beads (Spherotech, Lake Forest, IL). BD CompBeads (BD, anti-rat #552844, and anti-mouse #552843) were used for compensation.

Manual analysis of cytometry data was performed using FlowJo software (TreeStar).

Statistical analysis was carried out using GraphPad Prism version 9 (GraphPad Software) and MATLAB (R2022b, The MathWorks Inc.).

Non-responders to therapy with CD19-specific CAR-T cells (tisagenlecleucel) are characterized by undetectable expansion of CAR-T cells and in the majority of cases are diagnosed with primary refractory aggressive B-cell lymphomas..