Identification of h1218, a novel humanized chicken scFv recognizing a membrane-proximal epitope of human CD19

We first sought to identify an antibody against a CD19 epitope that was distinct from the one recognized by the FMC63 antibody that is used in all the currently FDA-approved CART19 (Fig. 1A). To this goal, we screened a proprietary chicken immune single-chain variable fragment (scFv) library against the human CD19 extracellular domain (M1-K291). Given the greater evolutionary distance between humans and chickens as compared to humans and mice, the use of a chicken library allows for the generation of antibodies recognizing unique epitopes compared to murine-developed antibodies. The initial screening included six hits, from which we selected clone 1218, which was characterized by the absence of competition with FMC63 for CD19 (M1-P278). Indeed, in epitope binning experiments using the Octet assay, we showed that FMC63 and h1218 could simultaneously bind to CD19 (Fig. 1B), confirming that they recognize distinct epitopes. The chicken 1218 scFv was humanized by CDR-grafting to human germline genes and backmutations, generating the humanized 1218 scFv (h1218) (Figure S1A). To define the epitope recognized by the 1218 scFv, we performed domain- and epitope-mapping mutagenesis. We divided the extracellular portion of CD19 into three domains and replaced each human (H) domain with cynomolgus (C) monkey domains obtaining the following chimeras: H–H-H-(cell), C-H–H-(cell), H-C-H-(cell), and H–H-C-(cell) (Table S1). The h1218 scFv was able to recognize H–H-C, but not C-H–H, suggesting that, unlike FMC63 scFv, h1218 scFv recognizes an epitope in the most membrane-proximal domain between amino acids 51 and 63 (Fig. 1C). To further characterize the specific epitope, we mutated single amino acid codons and found that the key amino acids required for the binding of h1218 scFv were L58, K59, and K63. Conversely, the membrane-distal amino acid H218/KSS [27] was recognized as a key residue for FMC63 binding (Fig. 1D). ELISA further confirmed this characterization by revealing a significant reduction in interferon-gamma production in h1218-CART19 cells upon mutation of the amino acids L58, K59, and K63. Mutations in amino acids T51, S53, and E55 also influenced this production, albeit to a lesser extent (Fig. 1E). Importantly, we established that the h1218 scFv had a similar affinity to CD19-ECD-Ck as compared to FMC63 scFv (mean KD, 1.85 × 10−7 M and 1.49 × 10−7 M, respectively), but the binding kinetics of the two antibodies differed significantly (Fig. 1F). Indeed, the h1218 scFv had faster binding to the target (on-rate) ability (mean Kon, FMC63 = 8.94 × 103(1/Ms) versus h1218 = 8.68 × 104(1/Ms)) and a significantly faster off-rate (mean Koff, FMC63 = 1.31 × 10–3(1/s) versus h1218 = 1.60 × 10–4(1/s)).

Fig. 1

The h1218 antibody is specific for CD19 and recognizes a non-FMC63 membrane proximal epitope. A Schematic of FMC63 and h1218 antibodies binding sites on CD19. B Binding of h1218 to FMC63 bound-human CD19 complex. Sensor chips were coated with FMC63 Fc and CD19-ECD-Ck from 1700 to 2300 s. Additional FMC63 or h1218 antibody was added to the FMC63-bound sensor chip at 2300 s and monitored for further binding activities. C h1218 antibody binding test on wild-type HEK293T (CD19 negative), HEK293T cell expressing human CD19 (huCD19, HHH), and HEK293T cells expressing one of the three chimeric CD19 forms that had cynomolgus residues replacement at different region of CD19, respectively (CHH, HCH, and HHC). D Mutagenesis study to identify key residues corresponding to the h1218 CD19 epitope. E Quantification of IFNγ release on HEK293T cells expressing WT or mutant CD19 to identify additional key residues corresponding to the h1218 CD19 epitope. F Binding affinity of FMC63 and h1218 scFv to recombinant human CD19-ECD-Ck (M1-P278). Each line represents affinity measured at a different scFv concentration. All the experiments were repeated at least twice

Lastly, to confirm the specificity of the h1218 scFv to CD19, we screened for binding interactions against HEK293 cells expressing 5,484 human plasma membrane proteins or cell surface-tethered human-secreted proteins (Retrogenix assay). We observed high-intensity binding to CD19, together with non-specific lower-intensity binding to three additional proteins (TMEM108, SUSD5, and GUCA2A) (Figure S1B). To test whether the h1218 scFv was able to functionally recognize only CD19 and not these three possible off-targets, we performed flow cytometry and ELISA to test the binding of h1218 to HEK293T cells overexpressing TMEM108, SUSD5, or GUCA2A. The results showed no activation of h1218-CART19 by these targets (Figure S1C-E). Therefore, we excluded any non-specific binding to TMEM108, SUSD5, or GUCU2A.

h1218-CART19 cells overcome CD19-FMC63 epitope loss resistance in preclinical models

Having developed a novel scFv targeting a membrane-proximal epitope of CD19, we sought to develop a second-generation CAR construct. We used lentiviral backbones (pTRPE or pLTG) and included a CD8α hinge and transmembrane domain, together with 4-1BB costimulatory and CD3ζ signaling domains. We used this construct to generate lentiviruses and transduce and expand human T cells as previously described [20, 28]. We first tested this new product (h1218-CART19) in FMC63-CART19-resistant models. Two point mutations in exon 3 of CD19 (R163L and L174V) have been described in post-FMC63-CART19 biopsies, which drive FMC63-CART19 resistance and lead to FMC63 CD19 epitope loss [8, 14]. To model CD19-mutant escape, we used CRISPR-Cas9 technology (Table S2) to knock out wild-type CD19 in the standard B-ALL cell line Nalm6 (Nalm6-CD19KO), and subsequently transduced these cells with lentiviral constructs harboring one of the two reported point mutations (Nalm6-CD19R163L or Nalm6-CD19L174V) (Fig. 2A). The surface expression of the CD19 mutants was confirmed by flow cytometry (Fig. 2B). We then evaluated the cytotoxic capacity of control untransduced (UTD) T cells, FMC63-CART19, and h1218-CART19 in vitro against the two engineered cell lines and found that h1218-CART19, but not FMC63-CART19, could recognize and target both Nalm6-CD19R163L and Nalm6-CD19L174V cells in short-term killing assays (Fig. 2C). Consistent with these findings, h1218-CART19 cells secreted IL-2 and TNF in the presence of both cell lines, whereas FMC63-CART19 cells did not (p < 0.0001 for both cytokines) (Fig. 2D). To confirm these results in vivo, we tested the efficacy of FMC63- and h1218-CART19 against the more frequently reported Nalm6-CD19L174V. We engrafted  immunodeficient NOD-SCID gamma chain deficient (NSG) mice with 1 × 106 Nalm6-CD19L174V cells on day -5 and randomized the mice to receive 0.75 × 106 UTD, FMC63-CART19, or h1218-CART19 T cells on day 0 (Fig. 2E). Notably, h1218-CART19-treated mice showed significant efficacy in terms of both tumor control and overall survival (median overall survival, FMC63-CART19 = 9 days versus h1218-CART19 = not reached, p = 0.0027) (Fig. 2E-F). Furthermore, the in vivo expansion of h1218-CART19 cells in the blood was also significantly higher than that in the controls (mean T cell count, FMC63 = 68 cells/100 µL blood versus h1218 = 300 cells/100 µL blood, p = 0.0096) (Fig. 2G). These results demonstrated the ability of h1218-CART19 to overcome FMC63-resistant CD19 mutations.

Fig. 2

h1218-CART19 recognize and kill malignant B cells carrying FMC63-resistant CD19 mutations. A Schematic representation of B cell leukemia cells with point mutations in the membrane-distal CD19 domain (CD19R163L or CD19L174V) as clinically identified post FMC63-CART19 treatment. B CD19 expression levels in Nalm6-CD19 KO, Nalm6 (wild type), Nalm6-CD19L174V (left) and Nalm6-CD19R163L (right) cell lines as measured by flow cytometry. C CART19 cytotoxicity against Nalm6-CD19L174V and Nalm6-CD19R163L at various effector to target (E:T) ratios (n = 3 independent donors) (luciferase assay). D IL-2 and TNF cytokine release measured by ELISA 24 h after CART19 and cancer cell co-culture at an E:T ratio of 5:1 (n = 2 donors). All values determined to be negative by comparison with the standard curve are shown as zero. E (Left) Schematic of the xenograft NSG mouse model: 1 × 106 luciferase + Nalm6-CD19L174V cells and 0.75 × 106 UTD, FMC63-CART19, or h1218-CART19 cells were intravenous injected with a 5-day interval. (Right) Tumor burden of engrafted mice treated with UTD (n = 5), FMC63-CART19 (n = 5), or h1218-CART19 (n = 5) as measured by bioluminescence imaging. The bold lines represent the median luminescence of each group. F Overall survival in each treatment group (p = 0.0027). G Absolute cell counts of huCD45+ huCD3+ T cells in 100μL mouse blood on day 9. All bar graphs and cytotoxicity curves are presented as the mean ± SEM. Survival curves were compared using the log-rank (Mantel-Cox) test, and one-way ANOVA was performed with Tukey’s correction for multiple comparisons; **** p < 0.0001, *** p < 0.001, ** p < 0.01, and * p < 0.05. All bar graphs are presented as the mean ± SEM. All experiments were repeated at least twice

Another unique, although rare, mechanism of FMC63-CD19 epitope loss is the accidental transduction of the FMC63-CAR19 gene into leukemic B-cell blasts, which causes CD19 epitope masking due to the in cis CAR19:CD19 interaction. We previously demonstrated that CAR19+ B-ALL blasts become resistant to FMC63-CART19 cells despite the fact that CD19 is still expressed on the surface [12]. We speculated that h1218-CART19 would be able to recognize these cells, given that they recognize a different CD19 epitope not "masked" by the FMC63-CAR19. Therefore, we used CD19+ Nalm6 cells transduced with a lentiviral vector carrying FMC63-CAR19 expressing both CD19 and FMC63-CAR19 (Fig. 3A-B). These cells were previously shown to be resistant to FMC63-CART19 [12]. In preclinical in vitro models, h1218-CART19 cells showed effective cytotoxicity in the short- and long-term and cytokine secretion (IL2) against Nalm6-FMC63 cells, whereas FMC63-CART19 cells demonstrated no activity (Fig. 3C-E). To confirm these results in vivo, we tested the efficacy of h1218-CART19 against FMC63-CD19-epitope-masked cancer cells by engrafting immunodeficient NSG mice with 1 × 106 Nalm6-FMC63 cells on day -5 and randomized the mice to receive 0.75 × 106 UTD, FMC63-CART19, or h1218-CART19 T cells on day 0 (Fig. 3F). Notably, all h1218-CART19-treated mice exhibited significantly enhanced tumor control and longer overall survival than FMC63-CART19-treated mice (median overall survival, FMC63-CART19 = 7 days versus h1218-CART19 = not reached, p = 0.0016) (Fig. 3F-G). These results suggest that the novel h1218-CART19 product, targeting a membrane-proximal epitope of CD19, can recognize and respond to cancer cells that have relapsed after CART19 with mutations in CD19 or epitope masking.

Fig. 3

h1218-CART19 recognize and kill relapsed FMC63-CAR19+ Nalm6. A Schematic representation of a B cell lymphoma/leukemia cell accidentally transduced with the FMC63-CAR19 lentivirus during manufacturing that leads to epitope masking and resistance to FMC63-based CART19. B FMC63 expression level on Nalm6 cells measured by flow cytometry. C IL-2 cytokine release quantification by ELISA 24 h after CART and cancer cell co-culture at a E:T ratio of 5:1 (n = 2 donors). All values determined as negative by comparison to the standard curve are shown as zero. D 48-h CART cytotoxicity against Nalm6-FMC63 at various E:T ratios (n = 2 donors) by flow cytometry. E Tumor growth in the presence of CART cells over 10 days (n = 3 donors) by flow cytometry. (Left) Fold change over time of Nalm6-FMC63 cells compared to day 0 counts. (Right) Nalm6-FMC63 fold change on day 6. F (Left) Schematic of the xenograft mouse model: 1 × 106 luciferase+ Nalm6-FMC63 cancer cells were engrafted 5 days before intravenous injection of 0.75 × 106 UTD or CART cells. (Right) Tumor burden over time in mice bearing Nalm6-FMC63 with UTD (n = 5), FMC63-CART19 (n = 5), or h1218-CART19 (n = 5). Bolded line represents the median of each group. G. Overall survival in mice bearing Nalm6-FMC63 (p = 0.0016). All bar graphs and cytotoxicity curves are represented as mean ± SEM. One-way ANOVA was performed with Tukey correction for multiple comparisons; survival curves were compared using the log-rank (Mantel-Cox) test; **** p < 0.0001, *** p < 0.001, ** p < 0.01, and * p < 0.05. All bar graphs are represented as mean ± SEM. All the experiments were repeated at least twice

h1218-CART19 cells demonstrate greater efficacy than FMC63-CART19 cells in human preclinical models of B-cell neoplasms.

As previously discussed, the h1218 scFv is characterized by faster on- and off-rates than those of the FMC63 scFv. We hypothesized that the h1218 CAR could lead to improved outcomes compared to standard FMC63-CART19 by reducing activation-induced cell death (AICD) upon antigen engagement (Fig. 4A). To test this hypothesis, we assessed h1218-CART19 and FMC63-CART19 activities in clinically relevant preclinical models of NHL/ALL in vitro and in vivo. In the short-term (48 h) in vitro killing assays, FMC63- and h1218-CART19 cells demonstrated similar cytotoxic effects against Nalm6 and other hematological malignancies (Fig. 4B; Figure S2A-B). This result was consistent with the fact that in the short-term, there was no significant difference in IFNγ cytokine release levels between FMC63- and h1218-CART19 cells against B-cell lymphoma/leukemia lines, including Raji (Burkitt lymphoma), Pfeiffer (DLBCL), Toledo (DLBCL), and Nalm6 (B-ALL) (Fig. 4C). In in vivo experiments, h1218-CART19 exhibited dose-dependent activity against Raji cells (Figure S3A-B), and when we used high doses of CAR T cells (1.5 × 106), both FMC63-CART19 and h1218-CART19 had similar anti-tumor cytotoxicity to Raji and Nalm6 cells in vivo (Fig. 4D).

Fig. 4

h1218-CART19 demonstrates enhanced efficacy than FMC63-CART19. A Schematic of h1218-CART19 targeting the membrane-proximal domain of CD19 in B cell lymphoma/leukemia as compared to standard FMC63-CART19. B Nalm6 tumor killing by UTD, h1218-CART19, or FMC63-CART19 cells at various E:T ratios (n = 3 donors) by flow cytometry. C IFNγ release as measured by ELISA upon stimulation of CD19-expressing tumor cells (Raji, Pfeiffer, Toledo, and Nalm6) at an E:T ratio of 3:1 (n = 2 donors). D (Left) Schematic of the in vivo experiment: either 1.5 × 106 UTD or CART19 cells were infused 7 days after intravenous injection of  luciferase + Raji or lNalm6 cells. Tumor progression in mice bearing Raji (middle) or Nalm6 (right) cells is shown. E Quantification of Nalm6 fold change over 14 days in the presence of UTD, FMC63-CART19, or h1218-CART19 at low-E:T ratio model (n = 2 donors) using flow cytometry. F (Left) Schematic of the in vivo xenograft model: 0.75 × 106 UTD or CAR T cells were infused intravenously 5 days after luciferase+ Nalm6 engraftment. (Middle) Tumor progression over time in mice bearing Nalm6 cells treated with UTD (n = 3), FMC63-CART19 (n = 7), or h1218-CART19 (n = 7) measured by luminescence. Bolded lines represent the median tumor burden in the corresponding group. (Right) Tumor burden on day 42 after CART19 injection. G Overall survival in mice bearing Nalm6. H (Left) CAR T cell expansion kinetics in the peripheral blood after CART19 injection. Bolded lines represent the median CAR T expansion. (Right) Quantification of CAR T cells in the blood 14 days after CART19 injection using flow cytometry. All bar graphs and cytotoxicity curves are presented as the mean ± SEM. Student's t-test was used to compare two groups; one-way ANOVA was performed with Tukey’s correction for multiple comparisons; survival curves were compared using the log-rank (Mantel-Cox) test; **** p < 0.0001, *** p < 0.001, ** p < 0.01, and * p < 0.05. All bar graphs are presented as the mean ± SEM. All experiments were repeated at least twice

However, in "stress-test" experiments where we administered lower and sub-optimal CART19 doses (0.75 × 106), we observed enhanced efficacy by h1218-CART19 cells compared to FMC63-CART19 both in vitro and in vivo. In particular, in a long-term in vitro killing assay with a low effector: target (E:T) ratio of 0.0625:1, h1218-CART19 showed improved tumor control against the B-ALL cell line Nalm6 compared to FMC63-CART19 (Fig. 4E). Furthermore, in an in vivo stress-test model of Nalm6 where only 0.75 × 106 CAR+ T cells were injected into NSG mice, h1218-CART19 demonstrated significantly better tumor control than FMC63-CART19 (p = 0.042) (Fig. 4F; Figure S3C). This enhanced tumor control was also associated with longer overall survival (h1218-CART19 = not reached versus FMC63-CART19 = 28 days, p = 0.0017) (Fig. 4G). Lastly, we measured CART expansion and persistence in the blood of mice and observed higher expansion and persistence of h1218-CART19 than that of FMC63-CART19 (Fig. 4H). These results demonstrate that the novel h1218-CART19 has stronger preclinical anti-tumor activity than that of FMC63-CART19 and is associated with higher CART19 expansion and persistence.

Faster on- and off-rates of h1218-CART19 cells are associated with decreased activation-induced cell death compared to FMC63-CART19

To better understand the functional characteristics and mechanism that support the increased anti-tumor efficacy of h1218-CART19 versus FMC63-CART19, we next studied the cellular interaction between CART19 and tumor cells. Given the hypothesis that faster on- and off-rates of h1218 scFv are associated with enhanced CAR T function, we first quantified the cellular avidity of FMC63-CART19 and h1218-CART19 using the z-Movi platform (LUMICKS). In line with the previous affinity kinetics results of the scFv, h1218-CART19 cells showed reduced cellular avidity compared to FMC63-CART19 cells (mean avidity, FMC63 = 55.57% versus h1218 = 39.23%, p < 0.0001) (Fig. 5A). We then studied the immune synapse characteristics of different CAR T cells as previously described [29]. We quantified F-actin to measure the stability of the synapse and the clustering of CD19 to measure the physical synapse formation. In line with the lower avidity of h1218- as that of FMC63-CART19, we found that synaptic F-actin clustering and accumulation of CD19 protein were reduced in h1218-CART19 cells. We then studied early CAR signaling by measuring perforin and phosphorylated-CD3ζ (pCD3ζ) chain polarization to the synapse. Similarly, we observed less perforin and pCD3ζ at the immunological synapse at a 2-h stimulation time-point in h1218-CART19 cells than in FMC63-CART19 cells (Fig. 5B-C). These results are in line with the faster off-rate observed for h1218. Given the increased anti-tumor efficacy and in vivo proliferation of h1218-CART19 observed in our previous experiments, we hypothesized that h1218-CART19 was more efficacious than FMC63-CART19 owing to reduced early CAR T activation and AICD and subsequent higher survival of CAR T cells after tumor encounter. To prove this hypothesis, we measured AICD upon target encounters. Notably, h1218-CART19 showed lower caspase 3/7 cleavage at baseline and reduced AICD as measured by caspase 3/7 cleavage upon stimulation compared to FMC63-CART19 (Fig. 5D), suggesting higher levels of AICD in FMC63-CART19 cells, which may lead to decreased T cell numbers in the long-term.

Fig. 5

h1218-CART19 demonstrates lower avidity and less activation-induced cell death than FMC63-CART19. A Quantification of UTD, FMC63-CART19, and h1218-CART19 binding avidity to Nalm6 after 15-min co-culture (n = 2 donors) by Lumicks analysis. B Representative confocal microscopy images of F-actin, CD19, perforin, and phosphorylated-CD3ζ (pCD3ζ) (CAR) expressed in FMC63-CART19 or h1218-CART19 cells when engaged with biotinylated CD19 protein. C Quantification of F-actin, perforin polarization, and pCD3ζ in FMC63-CART19 or h1219-CART19 cells engaged with biotinylated CD19 protein (n = 2 donors). In total, 200 events were recorded for each group. D (Left) Representative flow cytometric analysis of Caspase3/7 + FMC63-CART19 and h1218-CART19 cells with and without 4-h stimulation by Nalm6. Caspase3/7 + population of CART19 cells is boxed in red. (Middle) Baseline level of Caspase3/7 + population in FMC63- or h1218-CART19 cells at 0 h. (Right) Differential increase in the Caspase3/7 + population after 4-h stimulation. All graphs are represented as the mean ± SEM. Student's t-test was used to compare two groups; one-way ANOVA was performed with Tukey’s correction for multiple comparisons; **** p < 0.0001, *** p < 0.001, ** p < 0.01, and * p < 0.05. All bar graphs are presented as the mean ± SEM. All experiments were repeated at least twice

A first-in-human clinical trial of h1218-CART19 in relapsed and refractory NHL: study design and patients

Given the promising preclinical results of h1218-CART19, we sought to translate this preclinical product into a clinical good manufacturing product (GMP) product termed AT101 (AbClon Inc.). We started a phase I/II multi-center first-in-human clinical trial to establish the tolerability and the recommended phase II dose (RP2D) of AT101 in patients with B-NHL (NCT05338931). The study utilized a standard 3 + 3 dose escalation design (see “Methods” section) (Figure S4A). Here, we report the initial results of the Phase I portion of the trial. Key inclusion criteria comprised an established pathological diagnosis of B-cell non-Hodgkin lymphoma relapsed or refractory to at least one line of standard therapy; an ECOG performance status of 0–1, and adequate hematological, kidney, liver, lung, heart, and bone marrow functions. Patients previously treated with cellular therapies, autologous stem cell transplantation (ASCT), or CD3/CD20 bispecific antibodies were eligible for enrollment. Patients were initially screened for eligibility and signed an informed consent document before enrollment in the study. Autologous T cells were collected through standard apheresis and manufactured as AT101 using closed and automated systems. Patients were assigned a dose level upon the completion of apheresis. Upon receipt of the final AT101 product, the patients received lymphodepleting chemotherapy (LD) with intravenous fludarabine (25 mg/m2) and cyclophosphamide (250 mg/m2) on days -4, -3, and -2 prior to AT101 infusion. AT101 was administered as a single intravenous infusion at one of three dose levels (DL): DL-1:0.2 × 106 cells/kg, DL-2:1.0 × 106 cells/kg, and DL-3:5.0 × 106 cells/kg. Patients were allowed to receive bridging therapy at the discretion of the principal investigator during AT101 manufacturing and up to 2 weeks before AT101 infusion.

As shown in the CONSORT diagram (Figure S4B), of 14 patients registered, 14 patients were enrolled, and 12 (85.7%) were infused. Two patients were not infused because of manufacturing failure due to insufficient CAR T expansion (n = 1; CAR T fold change on day 6 of expansion was 0.70) and bacterial contamination of the apheresis product due to asymptomatic bacteremia likely caused by bacterial translocation originating from an intestinal lesion (n = 1). AT101 CAR T products reached a median of 5.6 (5.2–6.0) population doublings at the end of manufacturing. The final median CD4/CD8 ratio was 0.8 (0.1–1.7), and 92% (86–95%) of the cells were CD45 + CD3 + T cells. The median frequency of CAR + cells among AT101 cells was 46% (25–57%) (Table S3). The median vein-to-vein time was 56 (48–97) days (Table S3), given that the current release and quality control (QC) tests (listed in the Methods) are performed using time-consuming culture methods.

Patient demographics for the 12 infused patients are summarized in Table 1 and detailed in Table S4. The median age was 62.5 (39–84) years, and 58.3% (7/12) of patients were female. The median number of previous lines of treatment was 3 (2–8). The non-Hodgkin lymphoma subtypes included diffuse large B-cell lymphoma (DLBCL, n = 7/12, 58.3%), follicular lymphoma (FL, n = 3/12, 25.0%), mantle cell lymphoma (MCL, n = 1/12, 8.3%), and marginal zone lymphoma (MZL, n = 1/12, 8.3%). Three patients had an ECOG score of 0, and nine patients of 1. Five patients (41.7%) had received a previous autologous stem cell transplantation (ASCT), one patient (8.3%) had received a non-gene edited NK cellular therapy, and three patients (25%) had received CD20/CD3 bispecific antibodies. Two patients (16.7%) had refractory disease. The median sum of the product of diameters (SPD) by CT scan was 1,041 (190–11,836) mm2. Two patients (6.7%, patient 9 and 12) received bridging therapy, in particular CHOP (cyclophosphamide, doxorubicin, vincristine, and prednisone) or ESHAP (etoposide, cisplatin, cytarabine, and methylprednisolone), between apheresis and CAR T infusion. Patient 9 did not respond to bridging therapy whereas patient 12 achieved a partial response. Five patients (41.7%) had elevated lactate dehydrogenase (LDH) levels. One patient had bulky disease (> 7 cm) (Table S4). Overall, this study enrolled patients who were heavily pretreated with, including novel T cell-based immunotherapies, but not CART19.

Table 1 Patient baseline characteristicsSafety of h1218-CART19 in patients with relapsed or refractory B cell non-Hodgkin lymphoma

There were no infusion-related acute adverse events. Overall, 58.3% of the patients had at least one grade ≥ 3 adverse event (Table 2). As expected, neutropenia was the most common severe adverse event affecting seven patients and was likely related to LD chemotherapy. Among the 12 patients, six had a low B cell count before lymphodepletion (< 50 cells/μL). All patients experienced B cell aplasia following lymphodepletion and AT101 infusion. Patient 8, who had progressive disease (PD) at 3 months, recovered B cells at the same time as the disease progressed (Figure S5A). Grade 3 anemia was observed in four patients (33.3%), and grade ≥ 3 thrombocytopenia in two patients (16.7%). The levels of hemoglobin and platelets decreased early after lymphodepletion but promptly recovered in the responding patients (Figure S5B-C). Ferritin levels peaked during the first two weeks after AT101 infusion (Figure S5D). One patient (patient 12) experienced grade 3 sepsis on day 19 and promptly recovered with antimicrobial treatment; however, 27 days after the first infection, the patient developed new septic shock due to candidemia and died of multi-organ failure (Table 2, Table S4).

Cytokine-release syndrome (CRS) was observed in 33.3% of patients (n = 4), and it was severe (grade 3) in one patient (8.3%). No patients at DL-1 developed CRS, but three patients at DL-2/3 had grade 1 CRS, and one additional patient at DL-3 had grade 3 CRS. The median time of CRS onset was 6.5 (2–11) days after AT101 infusion and resolution happened within one day. None of the patients required steroids for the CRS management. Immune cell-related neurotoxicity syndrome (ICANS) occurred in 25% (n = 3) of patients, and it was severe (grade 4) in one patient (8.3%) at DL-1, which was the only dose-limiting toxicity (DLT) in this trial. The patient developed encephalopathy on day 12, which required intubation for airway protection. However, the patient recovered completely without neurological sequelae after six days of intravenous dexamethasone and intrathecal hydrocortisone treatment. Due to this DLT, 3 additional patients were enrolled in the DL-1 cohort without any further ICANS. One patient from the DL-2 cohort and one patient from the DL-3 cohort had grade 1–2 neurotoxicity, characterized by tremors and delirium, respectively. The median time of ICANS onset was 12 (8–12) days, with complete resolution within 3 (2–6) days with the use of dexamethasone (Table 2; Table S5).

Table 2 Summary of adverse events incidence, grade and duration by dose levelEfficacy of h1218-CART19 in patients with relapsed or refractory B cell non-Hodgkin lymphoma

Of the 12 patients treated, one patient (8.3%) did not respond to AT101. All other patients responded with an overall response rate (ORR) of 91.7% (95%CI, 56.2–97.0), and a CR was observed in 10 patients (75%) (Fig. 6A-D). At month 1, the ORR was 83.3% (95%CI, 51.6–97.9), and CR was observed in 8 patients (66.7%). At month 3, of the 11 patients still alive, the 8 responding patients were all in CR (72.7%) (95%CI, 43.6–92.1) (Figure S6A). Of note, considering only the patients (n = 6) receiving higher doses of AT101 (DL-2 and DL-3), the CR was 100% (Fig. 6D). In the overall population, one patient (patient 8) with a partial response (PR) lost the response before month 3; all other responding patients (10/11) maintained their response, with one patient (patient 13) switching from PR to CR at month 3. With a median follow-up of 9.3 months (1.5–16.5 months), the progression-free survival (PFS) was 75.0%, and the overall survival (OS) was 82.5%. The median PFS and OS were not reached (Fig. 6E-F). One patient (patient 12) in CR died of septic shock. Median PFS and OS of patients achieving CR were not reached (Figure S6B-C). Based on these results, DL-3 was selected as the recommended phase-II dose.

Fig. 6

h1218-CART19 phase I clinical trial patient characteristics and response. A Swimmers' plot of individual patients with their responses to AT101 over time. Specific NHL subtype and baseline tumor burden is indicated (sum of the product of the diameters (SPD), mm2). See box legends for detail. Note: patients 7 had non-measurable disease involvement of sigmoid colon at time of AT101 infusion and patient 12 did not have a measurable target lesion after bridging therapy. B Waterfall plot depicting the change in tumor burden from baseline to the best response post-treatment for each patient. C Patient 10's PET/CT-scan before AT101 infusion and 1 month after AT101 infusion, showing complete metabolic response. D (Top) Best overall responses across all 12 patients. (Bottom) Best overall response by dose level (DL-1, DL-2, and DL-3). E Progression-free survival of patients treated with AT101. Patients at risk listed below. F Overall survival of patients treated with AT101. Patients at risk listed below

h1218-CART19 AT101 expansion, persistence, and serum cytokine levels

To study the function of AT101 in patients and define possible correlates of response and toxicity, we measured CAR T kinetics in peripheral blood using flow cytometry and qPCR, and serum cytokine levels using the Luminex assay [30]. The AT101 transgene in the blood peaked on day 11 (11–14) after CAR T infusion. There was a correlation between peak expansion and infused dose (DL-1:2.5 ± 1.3, DL-2:5.0 ± 3.4, DL-3:113.6 ± 6.50 × 104 CAR gene copies/µg DNA) (Fig. 7A). Furthermore, we measured cytokine levels in the serum at baseline and following AT101 infusion. Interestingly, in all patients, we observed an increase in key inflammatory cytokines after CAR T infusion, particularly sFas ligand, granzyme A, and perforin. The peaks of these cytokines were observed on day 11 (Fig. 7B; Figure S7).

Fig. 7

h1219-CART19 phase I clinical trial CAR T expansion, blood cell counts, and serum cytokine levels. A AT101 expansion and persistence in the blood detected by qPCR in DL-1, DL-2 and DL-3. B Mean serum levels of sFas ligand, granzyme A, or perforin across 12 patients over time by Luminex. C AT101 expansion in complete response patients (n = 9) and not complete response patients (n = 3) by qPCR. D B cell count in complete response patients (n = 9) and not complete response patients (n = 3). E (Left) sFas ligand and (right) serum amyloid A levels on day 1 pre AT101 infusion and days 2–168 post AT101 infusion in complete response patients and not-complete response patients by Luminex. F IP-10 serum level change over time in patients without CRS (n = 8) and patients with CRS (n = 4) by Luminex. G Platelet level change over time in patients without CRS (n = 8) and patients with CRS (n = 4) by Luminex. H Ferritin level change over time in patients without CRS (n = 8) and patients with CRS (n = 4)

We then investigated whether patients achieving CR displayed different proportions of CAR T transgenes and cytokine secretion in the peripheral blood. Indeed, the AT101 transgene peak was 6.0 × 104 copies/μg DNA in patients who developed CR compared with 2.0 × 104 copies/μg DNA in patients who did not develop CR (Fig. 7C). Similarly, reduced B cell levels after treatment and during follow-up were associated with CR (Fig. 7D). Additionally, sFas ligand levels were higher in CR patients than in non-CR patients. In contrast, serum amyloid A levels were higher in the non-CR group than in the CR group (Fig. 7E). Furthermore, patients with CRS had higher cytokine levels, particularly IP-10 (Fig. 7F), and lower platelet counts (Fig. 7G). Lastly, we observed higher peaks of ferritin in patients experiencing CRS as compared to patients with no CRS (Fig. 7H).

In summary, we developed a new CART19 product from bench to bedside and demonstrated that targeting a novel membrane-proximal epitope of CD19 with fast on- and off-rate CAR leads to a potent anti-tumor effect in patients with manageable toxicity.