Idecabtagene vicleucel and endogenous T‐cell phenotypes linked to progression‐free survival in relapsed multiple myeloma

Abstract

Improved understanding of how chimeric antigen receptor (CAR) and endogenous T‐cell phenotypes associate with progression‐free survival (PFS) is needed to individualize and improve immunotherapy outcomes in multiple myeloma (MM). We characterized and investigated the prognostic value of immune dynamics defined with multidimensional and computational flow cytometry in 258 bone marrow aspirates from 107 MM patients treated with idecabtagene vicleucel in the KarMMa trial. At screening, there was an association between longer PFS and increased percentages of PD1⁺ cytotoxic T cells and ICOS⁻TIGIT⁺ Tregs. At Month 1 after infusion, the presence of >1% CAR‐T cells, a CD4/CD8 CAR‐T‐cell ratio >0.09, and greater abundance of non‐activated non‐cytolytic CAR‐T cells identified patients with longer PFS. Notably, 47 of the 115 endogenous T‐cell clusters were prognostic at the latest assessment performed after CAR‐T infusion. Lower percentages of ICOS⁺TIGIT⁻ Tregs showed the strongest association with longer PFS. Altogether, this large dataset of dynamic immune profiling throughout treatment with anti‐B‐cell maturation antigen (anti‐BCMA) CAR‐T cells uncovered that the T‐cell composition prior to apheresis as well as CAR‐T and endogenous T‐cell phenotypes after infusion associates with PFS. These findings have potential for translation into routine laboratory practice for improved understanding of response and resistance to CAR‐T cells in relapsed/refractory MM.

INTRODUCTION

The development of chimeric antigen receptor (CAR) T‐cell constructs targeting the B‐cell maturation antigen (BCMA) has changed the treatment landscape for patients with relapsed and refractory multiple myeloma (RRMM). ¹ , ² , ³ , ⁴ , ⁵ This was because of the high response rates on intent‐to‐treat, as well as the prolonged progression‐free survival (PFS) observed in some patients. These results were attained virtually without data about the determinants of response and resistance to CAR‐T‐cell therapy in MM, and it is plausible that greater knowledge could lead to treatment individualization and improved outcomes.

Resistance to anti‐BCMA CAR‐T therapy can be primary, as defined by an initial lack of response, ⁶ , ⁷ or secondary, as defined by relapse after an initial transient response. ⁸ There are emerging biomarkers of primary and secondary resistance such as patient' and disease characteristics, ⁹ , ¹⁰ tumor burden, ¹¹ , ¹² and antigen escape. ¹³ , ¹⁴ , ¹⁵ However, current understanding of how endogenous and CAR‐T‐cell phenotypes impact patients' PFS is limited to relatively small series studied with different methods. ¹² , ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ Accordingly, immune profiling is not part of routine laboratory practice in patients treated with CAR‐T cells.

The identification of immune biomarkers could facilitate a more individualized approach at the time of treatment initiation, but also after infusion to enhance CAR and endogenous T‐cell immune surveillance. ²¹ Owing to the limited understanding of how patient outcomes are shaped by T‐cell fitness prior to apheresis, as well as by CAR and endogenous T‐cell phenotypes post‐infusion, we aimed at characterizing immune dynamics and investigating its prognostic value in RRMM patients treated with idecabtagene vicleucel (ide‐cel) in the KarMMa Phase 2 trial (NCT03361748). ¹

RESULTS

Study design

In total, 107 RRMM patients were included in this study according to the availability of immune profiling data generated at screening, or at Months 1, 3, 6, and 12 after ide‐cel infusion (Figure 1A). All bone marrow (BM) aspirates were collected prior relapse (Figure S1). Baseline characteristics of patients with immune profiling performed at screening (n = 63) are described in Table 1. When compared to patients without immune profiling at screening because of limited BM volumes or cellularity (n = 65), there were no differences in demographics and major disease characteristics, except for a slightly inferior median number of prior lines of therapy in patients enrolled in this study (5 vs. 6, Table 1). The median PFS of patients with immune profiling performed at screening was 9 (interquartile range [IQR], 3.96–17.3) months.

Figure 1.

Immune profiling in the KarMMa clinical trial. (A) A total of 107 patients were included in this study according to the availability of immune profiling data generated at screening, or at Months 1, 3, 6, and 12 after infusion. (B) Immune profiling was performed in a total of 258 bone marrow aspirates using 8‐color monoclonal antibody combinations designed to characterize the T‐cell compartment. All combinations included a customized reagent for the quantification and phenotyping of idecabtagene vicleucel (ide‐cel). (C) Data contained in 1604 flow cytometry standard (FCS) files from all patients and time points were merged and analyzed using the computational flow cytometry workspace FlowCT, which resulted in the identification of 158 clusters including 115 endogenous and 43 chimeric antigen receptor (CAR) T‐cell clusters.

Table 1.

Baseline characteristics of the patients with immune profiling performed at screening (n = 63) versus those without immune profiling data at screening (n = 65).

	Patients included (n = 63)	Patients excluded (n = 65)	P value
Age (years), median (range)	60 (33–75)	61 (38–78)	0.81
Male, n (%)	41 (65)	35 (54)	0.27
Time since diagnosis (years), median (range)	5.8 (1–18)	6.3 (2–18)	0.29
International Staging System at screening, n (%)			0.63
I	26 (41)	22 (34)
II	24 (38)	26 (40)
III	13 (21)	17 (26)
del(17p), t(4;14) and/or t(14;16) at diagnosis, n (%)			0.12
Yes	28 (48)	17 (32)
No	30 (52)	36 (68)
Not available	5 (8)	12 (18)
del(17p), t(4;14) and/or t(14;16) at diagnosis at screening, n (%)			0.7
Yes	17 (55)	14 (47)
No	14 (45)	16 (53)
Not available	32 (51)	35 (54)
≥50% plasma cells in bone marrow at screening, n (%)			0.10
Yes	27 (45)	38 (61)
No	33 (55)	24 (39)
Not available	3 (5)	3 (5)
Number of prior regimens, median (range)	5 (3–16)	6 (3–14)	0.015
Prior stem cell transplant, n (%)	60 (95)	60 (92)	0.72
Refractory status, n (%)
Immunomodulatory drugs	61 (97)	65 (100)	0.24
Proteasome inhibitors	58 (92)	58 (89)	0.81
Anti‐CD38 monoclonal antibodies	60 (95)	60 (92)	0.72
Double‐refractory	56 (89)	58 (89)	1.0
Triple‐refractory	54 (86)	54 (83)	0.87
Penta‐refractory	13 (21)	20 (31)	0.27

Immune profiling was performed in a total of 258 BM aspirates using 8‐color monoclonal antibody combinations designed to characterize T cells (Figure 1B and Table S1). All combinations included a customized reagent for the detection of CAR‐T cells. Data from all patients and time points were merged and analyzed using computational flow cytometry, ²² which resulted in the identification of 158 clusters including 115 endogenous and 43 CAR‐T‐cell clusters (Figure 1C and Table S2). This approach enabled a holistic analysis of phenotypes associated with PFS (Figure S2).

Endogenous T‐cell phenotyping prior apheresis

We first investigated the association between T‐cell phenotypes at screening and PFS. Of the 115 endogenous T‐cell clusters, 41 were significantly associated with PFS (Table S3). Of these, nearly two‐thirds (25/41, 61%) were CD8⁺ clusters. As expected, higher percentages of terminal effector memory RA (TEMRA) CD8⁺ T cells identified patients with inferior PFS compared to those with lower percentages (medians of 6 vs. 16 months, P = 0.047, Table S3). Of further interest was the observation of a positive correlation between increased percentages of clusters expressing one or more inhibitory checkpoints (e.g., PD1, TIM3, or TIGIT) and longer PFS (Table S3). Accordingly, there were two clusters in which lower percentages were associated with a ≥5‐fold increased risk of progression and/or death (i.e., hazard ratio [HR] ≥ 5.0). One cluster was phenotypically defined as CD3⁺CD4⁻CD8⁺CD57⁻ICOS⁺PD1⁺TIM3⁻ (i.e., potentially reactive cytotoxic T cells) and the other as CD3⁺CD4⁺CD8⁻CD25⁺CD127lo⁻ICOS⁻TIGIT⁺ (i.e., potentially tissue‐protective regulatory T cells [Tregs]). Patients having higher percentages of these two clusters displayed longer PFS (Figure 2A,B). Of note, there were no significant differences observed in T‐cell phenotypes at screening according to the number of years between diagnosis and treatment, International Staging System, and tumor burden (i.e., ≥50% CD138⁺ plasma cells in BM) at screening, number of prior lines of therapy, as well as prior stem cell transplantation (data not shown). Conversely, there were differences in T‐cell phenotypes between patients with versus without extramedullary disease (Table S4). Furthermore, there were significant correlations between T‐cell phenotypes at screening and sBCMA levels at baseline, as well as with the difference in sBCMA levels from baseline to Month 1 after CAR‐T‐cell infusion (Figure S3).

Figure 2.

Prognostic value of endogenous T‐cell phenotyping prior apheresis. (A, B) Progression‐free survival of patients stratified according to low versus high percentages of CD3⁺CD4⁻CD8⁺CD57⁻ICOS⁺PD1⁺TIM3⁻ (0.04% cutoff) and CD3⁺CD4⁺CD8⁻CD25⁺CD127lo⁻ICOS⁻TIGIT⁺ (1.10% cutoff) T cells at screening. HR, hazard ratio.

Quantification and characterization of CAR‐T cells after infusion

After identifying endogenous T‐cell clusters at screening associated with clinical outcome, we next investigated if the percentage and phenotype of CAR‐T cells after infusion were also prognostic. Of the 76 patients profiled at Month 1 after infusion, 55 had detectable CAR‐T cells in BM (median 0.5%, IQR 0.12%−1.21%). Furthermore, in 16 of the 55 cases, the percentage was >1%. These patients showed significantly longer PFS when compared to those with <1% CAR‐T cells at Month 1 after infusion (Figure 3A).

Figure 3.

Prognostic value of chimeric antigen receptor (CAR)‐T‐cell phenotyping at Month 1 after infusion. (A, B) Progression‐free survival of patients stratified according to >1% versus <1% CAR‐T cells and >0.09 versus <0.09 CD4/CD8 CAR‐T‐cell ratio in bone marrow at Month 1 after ide‐cel infusion. (C, D) Progression‐free survival of patients stratified according to low versus high percentages of CD3⁺CD4⁻CD8⁺CD16⁻CD56⁻GranzymB⁺HLADR⁺ (8.37% cutoff) and CD3⁺CD4⁺CD8⁻CD16⁻CD56⁻GrymB⁻HLADR⁻ (3.74% cutoff) CAR‐T cells in bone marrow at Month 1 after ide‐cel infusion. HR, hazard ratio.

Beyond the percentage of intratumor CAR‐T cells, we observed that a CD4:CD8 CAR‐T‐cell ratio > 0.09 in BM also identified patients with longer PFS (Figure 3B). Of note, these cases with a favorable prognosis were characterized by increased CD4⁺ rather than decreased CD8⁺ CAR‐T cells. Indeed, similar to other studies, ²³ we found that the median percentage of CD8⁺ CAR‐T cells at Month 1 after ide‐cel infusion was higher than that of CD4⁺ CAR‐T cells (medians of 0.28% vs. 0.08%, respectively; P = 0.003).

The inherent migration of CAR‐T cells to the BM naturally leads to higher percentages than that observed in peripheral blood (PB) and facilitates their phenotypic characterization. We leveraged this to identify which phenotypic traits besides the CD4:CD8 ratio were prognostic. Of the 43 CAR‐T‐cell clusters identified using the multidimensional and computational flow cytometry approach described above, 12 were significantly associated with PFS (Table S5). Interestingly, these included CD4⁻/CD8⁻, CD4⁺/CD8⁺, CD4⁺, and CD8⁺ CAR‐T cells, which were generally defined according to the expression of GrzmB, checkpoint markers, and antigen‐dependent differentiation markers (Table S5). The two clusters associated with the highest risk of progression and/or death (i.e., HR ≥ 6.0) were either CD8⁺ or CD4⁺, and both defined by the expression of GrzmB and HLADR (Table S5). Namely, patients having higher percentages of CD3⁺CD4⁻CD8⁺CD16⁻CD56⁻GranzymB⁺HLADR⁺ CAR‐T cells (HR: 6.40, 95% confidence interval [CI] 2.01−20.34; P = 0.002) and lower percentages of CD3⁺CD4⁺CD8⁻CD16⁻CD56⁻GranzymB⁻HLADR⁻ CAR‐T cells (HR: 6.43, 95% CI 2.58−16.04; P < 0.001) showed significantly inferior PFS (Figure 3C,D). These results suggest that a greater abundance of potentially non‐activated non‐cytolytic CAR‐T cells after peak expansion may translate into a more durable anti‐tumor response. Accordingly, we observed that hyper‐expanders (i.e., patients with CAR‐T‐cell expansions at Month 1 above the upper quartile) showed significant differences in the abundancies of five CAR‐T‐cell types when compared to the remaining cases. Hyper‐expanders showed increased percentages of CD3⁺CD4⁺CD8⁻CD16⁻CD56⁻GranzymeB⁻HLADR⁻ CAR‐T cells, together with decreased percentages of CD3⁺CD4⁻CD8⁺CD16⁻CD56⁻GranzymeB⁺HLADR⁻, CD3⁺CD4⁻CD8⁺CD27⁺CD45RAloCD95⁻CD197⁻, CD3⁺CD4⁻CD8⁺CD16⁻CD56⁻GranzymeB⁺HLADR⁺, and CD3⁺CD4⁻CD8⁺CD27⁺CD45RA⁺⁺ CD95‐CD197⁻ CAR‐T cells (Figure S4).

T‐cell dynamics throughout CAR‐T‐cell therapy

There is limited understanding of how the phenotype of endogenous T cells is modified after CAR‐T‐cell infusion and if there are prognostic phenotypes. Thus, we investigated intratumor T‐cell dynamics between screening and Months 1, 3, 6, and 12 after CAR‐T infusion. Paired samples were respectively available in 44, 27, 21, and 21 patients.

Of the 115 endogenous T‐cell clusters analyzed, 25 showed different distribution between screening and Month 1 after CAR‐T infusion (Figure 4A). It was interesting to note that most subsets that became expanded were CD4⁺ and Tregs expressing inhibitory checkpoints (e.g., ICOS, PD1, TIM3, or TIGIT). Conversely, the subsets showing the most significant contraction were either CD4⁺, CD8⁺, and CD4⁻CD8⁻ T cells expressing the interleukin‐7 receptor alpha chain (CD127) that is typically associated with memory and effector functions. Similar results were observed in the paired comparison between screening and Month 3 after CAR‐T‐cell infusion, though, overall, there were fewer clusters showing different distribution (n = 10, Figure 4B). Furthermore, there were no significant differences observed in the T‐cell distribution between paired samples analyzed at screening and Months 6 and 12 after CAR‐T‐cell infusion. These results suggest that the greatest impact on the composition of the endogenous T‐cell compartment occurs soon after CAR‐T‐cell expansion and progressively fades thereafter. Another interesting observation was that the percentage of 15 CAR‐T‐cell clusters was significantly correlated with the endogenous counterpart, particularly at Month 1 after infusion (Table S6).

Figure 4.

Dynamics and prognostic value of endogenous T‐cell phenotyping after chimeric antigen receptor (CAR)‐T‐cell infusion. (A) Volcano plot showing clusters with significantly different distribution between paired samples collected at screening and Month 1 after CAR‐T‐cell infusion (n = 44). (B) Volcano plot showing clusters with significantly different distribution between paired samples collected at screening and Month 3 after CAR‐T‐cell infusion (n = 27). (C) Progression‐free survival of patients stratified according to low versus high percentages of CD3⁺CD4⁺CD8⁻CD25⁺CD127lo⁺ICOS⁺TIGIT⁻ T regulatory cells in bone marrow at the latest assessment after CAR‐T‐cell infusion (0.16% cutoff). HR, hazard ratio.

In contrast to the findings at screening, there were nearly as many CD4⁺ (n = 22) and CD8⁺ (n = 23) T‐cell clusters associated with clinical outcomes after CAR‐T‐cell infusion (Table S7). Another difference with the findings at screening was that increased percentages of clusters expressing one or more inhibitory checkpoints (e.g., PD1, TIM3, or TIGIT) at post‐infusion timepoints were associated with inferior PFS (Table S7). Furthermore, the cluster showing the highest risk of progression and/or death was defined by CD3⁺CD4⁺CD8⁻CD25⁺CD127lo⁺ICOS⁺TIGIT⁻ phenotype (HR: 5.25, 95% CI 1.60−17.28; P = 0.006). Patients having higher percentages of ICOS⁺ Tregs had a median PFS of 2.4 months, whereas those having lower percentages had a median PFS not reached (Figure 4C).

DISCUSSION

An exquisite understanding of the mechanisms leading to high CAR‐T‐cell efficacy and of those promoting resistance is paramount to optimize clinical outcomes with this innovative treatment modality. Loss of the target antigen and the impact of suboptimal responses have been studied in more detail. ¹³ , ¹⁴ , ¹⁵ By contrast, knowledge remains scarce about inefficient CAR‐T‐cell expansion or persistence in vivo, rapid exhaustion of CAR‐T cells, and impaired endogenous immune surveillance. Here, we report one of the largest studies of dynamic immune profiling in RRMM patients treated with anti‐BCMA CAR‐T cells. We uncovered that T‐cell phenotypes prior to apheresis are associated with PFS. Furthermore, CAR‐T‐cell phenotypes after infusion may be as prognostic as their percentage. We also showed how CAR‐T cells may modulate endogenous T‐cell phenotypes and how they associate with PFS. Altogether, this study supports immune profiling for improved understanding of response and resistance to CAR‐T cells in RRMM.

There is emerging evidence suggesting that the first genetic events preceding the diagnosis of MM could take place between the second and third decades of life. ²⁴ , ²⁵ This finding together with the years that may take from the progression from pre‐malignant conditions onto newly diagnosed active MM, implies that by this stage, patients' T cells have been exposed to tumor antigens for decades. In RRMM patients such as those enrolled in the KarMMa trial, the multiple lines of therapy and their immunosuppressive effects have an unmeasured impact on T cells. Thus, the finding that there were no significant differences observed in T‐cell phenotypes at screening according to clinical characteristics and prior treatment suggests that the individual effect of host, tumor, and therapy in patients' immune status cannot be fully dissected at the stage of RRMM. In such a case, immune profiling could hypothetically act as a surrogate of the cumulative effect of these factors in T‐cell fitness in this heavily treated population. Our results showing an association between longer PFS and higher percentages of cytotoxic T cells that, based on their PD1⁺TIM3⁻ phenotype, could be in a progenitor or precursor exhausted state, may be due to potential self‐renew in response to persisting antigen and differentiation into more effector‐like exhausted and terminally differentiated exhausted CD8⁺ T cells. ²⁶ We also showed an association between higher percentages of ICOS⁻TIGIT⁺ Tregs and longer PFS. This finding is unexpected since previous evidence suggested that non‐responders to Ide‐cel had significantly more Tregs in PB at day of leukapheresis. ²³ Interestingly, though, it has been reported that beyond its immunosuppressive role, TIGIT⁺ Tregs may have tissue‐protective effects. ²⁷ Validation of these results in ongoing and future studies is needed to clarify if immune profiling at screening could potentially be used to guide, for example, the type of bridging therapy and improve clinical outcomes in patients in whom CAR‐T cells may be less potent. ²⁸ External validation is also required because some patient numbers in some subgroups are very low owing to the maxstat methodology to identify prognostic cutoffs.

Lin et al. showed how the concentration of ide‐cel at Day 14 in the PB of RRMM patients enrolled in the CRB‐401 Phase 1 study was associated with different duration of response. ¹⁷ Recently, the number of CD3⁺ CAR‐T cells in PB samples collected at Days 7, 14, and 30, from 27 patients treated with ide‐cel, was shown to be significantly different between patients with stable versus progressive disease. ²³ Even more recently, the percentage of CAR‐T cells within total CD3⁺ T cells in PB was found to be significantly higher in RRMM patients achieving long‐ versus short‐term responses. ²⁰ Although monitoring of CAR‐T cells in PB is more convenient for the patient and therefore can be performed more frequently, our results suggest that BM aspirates collected soon after infusion, for example, to assess measurable residual disease, ⁶ , ⁷ can be leveraged to determine the persistence of intratumor CAR‐T cells and obtain additional prognostic information. It was interesting to observe that with the exception of CD3⁺CD4⁺CD8⁺CD27⁺CD45RA⁻CD95⁻CD197lo CAR‐T cells, no other subset was associated with patients' MRD status at Month 1 after infusion (Figure S5). The fact that this cluster was expanded in MRD‐negative versus ‐positive patients suggests that greater tumor debulking may result in higher representation of central memory instead of effector memory or terminally differentiated effector CAR‐T cells.

We observed that higher CD4:CD8 CAR‐T‐cell ratios and increased percentages of potentially non‐activated CD4⁺ CAR‐T cells associated with longer PFS. Overall, these results are consistent with recent observations from CARTITUDE‐1, where RRMM patients achieving long‐term remission and survival after treatment with ciltacabtagene autoleucel (cilta‐cel) displayed higher CAR⁺CD4⁺ central memory T cells. ²⁹ Indeed, we observed that patients displaying increased percentages of CD4⁺ central memory CAR‐T cells (i.e., cluster CD3⁺CD4⁺CD8⁻CD27⁺CD45RA⁻CD95⁻CD197lo CAR‐T) had median PFS not reached (Figure S6). These results, together with the remarkable observation that decade‐long remission in patients with B‐cell chronic lymphocytic leukemia was driven by long‐persisting CD4⁺ CAR‐T cells, ³⁰ urge a better understanding of their role in durable responses.

Knowledge about immune correlates of clinical outcomes after CAR‐T‐cell therapy is increasing with recent studies. Rade et al. ³¹ presented a longitudinal single‐cell atlas of anti‐BCMA CAR‐T‐cell therapy in MM, revealing that cilta‐cel drives potent CD4⁺ cytotoxic responses and superior outcomes, but also severe CRS and prolonged cytopenia. Interestingly, RRMM patients unable to achieve a PFS longer than 5 years after cilta‐cel infusion exhibited strong negative correlations between myeloid‐derived suppressor cells and both endogenous and CAR‐derived CD4‐memory T‐cell subsets. ³² These results are in contrast to other recent studies suggesting that long‐term PFS is primarily influenced by the patient's immune landscape and the tumor burden rather than anti‐BCMA CAR‐T‐cell properties. ³³ These discordances may be associated with the CAR‐T‐cell product used in the different studies, since both Rade ³¹ and Atanackovic ³⁴ showed, amongst others, different dynamics and phenotype at peak expansion, transcriptional modulation after expansion, and long‐term persistence. Our study, which analyzed the largest cohort of RRMM patients treated with ide‐cel and specifically analyzed BM instead of PB samples, confirmed the association between CD4⁺ CAR‐T cells and longer PFS, and uncovered that the endogenous T‐cell composition prior to apheresis as well as after infusion may help predicting resistance.

As data on resistance to CAR‐T‐cell therapy increase and new therapeutic approaches emerge, there might be opportunities to tailor interventions based on the evaluation of genomic antigen escape, depth of response achieved after CAR‐T‐cell infusion, and endogenous immune surveillance versus escape after CAR‐T‐cell clearance. It was interesting to observe that the greatest impact on T‐cell phenotypes occurred at Month 1 and progressively faded over time. Notwithstanding the limitation of how the potential effects that lymphodepletion and CAR‐T‐cell infusion cannot be dissected, our findings suggest that patients displaying traits of a chronic and exhausted immune response, as suggested by increased percentages of endogenous ICOS⁺TIGIT⁻ Tregs, which have been described as the most immunosuppressive Treg subset, ³⁵ are at greater risk of progression and may benefit from earlier exposure to drugs empowered to revert immune escape instead of treatment at the time of clinical relapse. ²¹ This as well as the previous observations requires validation in future studies performing longitudinal immune profiling prior, during, and after CAR‐T‐cell therapy.

Until such data becomes available, we herein report findings that have potential for translation into routine laboratory practice. Of note, none of the prognostic clusters were significantly correlated with the leukocyte count at the respective time points (Figure S7), and therefore immune profiling may provide complementary information. Namely, the 1% cutoff of BM CAR‐T cells and the 0.09 cutoff for the ratio of CD4:CD8 CAR‐T cells at Month 1 after infusion, as well as the percentage of endogenous ICOS⁺TIGIT⁻ Tregs at later time points. While the holistic nature of our analysis resulted in the unbiased identification of clusters with complex phenotypes that are not easy to monitor in routine practice, we observed that more conventional endogenous and CAR‐T‐cell types defined by conventional nomenclature and phenotype (e.g., central or effector memory) associated with PFS (Table S8). These immune biomarkers may be even more useful towards individualized approaches if incorporated into prognostic scores that capture other patient‐ and disease‐related characteristics prior apheresis ⁹ , ¹⁰ as well as depth of response after CAR‐T infusion. ⁶ , ⁷ Future studies should explore the utility of immune profiling to predict other clinically relevant outcomes such as cytokine release syndrome, immune effector cell‐associated neurotoxicity, infections, and hematological adverse events.

PATIENTS AND METHODS

Patients and treatment

The design, primary endpoint (i.e., objective response), and patient demographic information for the ongoing KarMMa clinical trial (NCT03361748) were reported previously. ¹ In brief, eligible patients were 18 years of age or older; had received at least three previous regimens including an immunomodulatory agent, a proteasome inhibitor, and an anti‐CD38 antibody; had disease that was refractory to their last regimen; had measurable disease; and had adequate organ function. ² Subjects' median age (range) was 61 (33–78) with slightly more than half of the subjects being male (59%). The majority of subjects were white (80.5%), whereas 2.3% were Asian and 4.7% were Black or African American. The majority of subjects were not Hispanic or Latino (80.5%).

Ide‐cel was manufactured and administered as described previously. ¹ Briefly, it was manufactured from autologous PB mononuclear cells, stimulated with antibodies to CD3 and CD28, transduced with a lentiviral vector containing the anti‐BCMA CAR, and expanded over a period of 10 days. The actual doses delivered were within 20% of the target doses. Patient distribution was 2%, 51%, and 47% for the respective ide‐cel dosages of 150 × 10⁶, 300 × 10⁶, or 450 × 10⁶ CAR⁺ T cells per kg body weight. Bridging therapy was allowed during manufacturing but was stopped at least 14 days before lymphodepletion and was restricted to certain drug classes and to drugs previously received.

The study protocol was approved by local or independent institutional review boards or ethics committees at participating sites. All the patients provided written informed consent.

Immune profiling

Fresh BM aspirates were processed following the EuroFlow lyse‐wash‐and‐stain protocol. ³⁶ Eight‐color monoclonal antibody combinations were developed to dissect T cells into subsets defined by the expression of activation, antigen‐dependent differentiation, cytolytic, and exhaustion markers. All combinations included a customized reagent for the quantification and phenotyping of ide‐cel. Samples were measured in a FACSCanto II flow cytometer (BD Biosciences [BDB], San Jose, CA) using the FACSDiva software v8.0.1 (BDB). Cytometer linearity control and calibration/compensation were performed according to EuroFlow SOP to ensure longitudinal instrument standardization. ³⁶

Data were analyzed using FlowCT, a semiautomated workflow developed for the deconvolution of immunophenotypic data and objective reporting on large datasets. ²² Briefly, FlowCT starts by creating a matrix with expression data generating a SingleCellExperiment object and correcting possible discrepancies in markers' nomenclature. Subsequently, it performs internal data quality control and normalization, using Gaussian and Warp alignment of marker intensities to correct for technical variability across samples. Automated clustering followed by dimensionality reduction to visualize clusters' identity was performed before manual annotation using the Infinicyt software (BDB). Subclustering of endogenous and CAR‐T cells was performed in merged data from all assessments (Figure S8). Each monoclonal antibody combination was analyzed separately (Figures S9 and S10). A total of 158 clusters (including 115 endogenous T‐cell and 43 CAR‐T subsets) were identified, 33, 45, 39, and 41 in the first, second, third, and fourth 8‐color monoclonal antibody combination. A minimum of 20 events was required to define clusters, and in this dataset, the smaller cluster had 236 events.

Statistical analysis

Statistical differences between groups were assessed using the Mann–Whitney U and chi‐squared tests with Bonferroni correction. Prognostic cutoffs were identified using maximally selected rank statistics with the maxstat v0.7 R package. PFS was landmarked to the time point of immune profiling and measured until progression and/or death. Briefly, differences were tested for statistical significance with the (two‐sided) log‐rank test, and HR (with two‐sided 95% CIs) were estimated with a Cox regression model. Statistical analyses were performed using R version 4.0. P values <0.05 were considered statistically significant.

AUTHOR CONTRIBUTIONS

Bruno Paiva: Conceptualization; methodology; data curation; supervision; formal analysis; validation; investigation; writing—review and editing; writing—original draft. Irene Manrique: Methodology; investigation; validation; formal analysis; data curation; writing—review and editing. Ethan Thompson: Investigation; writing—review and editing. Timothy B. Campbell: Investigation; writing—review and editing. Camila Guerrero: Investigation; formal analysis; methodology; writing—review and editing. Nathan Martin: Investigation; writing—review and editing. Larry D. Anderson Jr.: Investigation; writing—review and editing. Jesús G. Berdeja: Investigation; writing—review and editing. Sagar Lonial: Investigation; writing—review and editing. Noopur S. Raje: Investigation; writing—review and editing. Yi Lin: Investigation; writing—review and editing. Philippe Moreau: Investigation; writing—review and editing. Jesús F. San‐Miguel: Investigation; writing—review and editing. Nikhil C. Munshi: Investigation; writing—review and editing. Shari Kaiser: Conceptualization; methodology; data curation; investigation; validation; formal analysis; funding acquisition; visualization; project administration; resources; writing—original draft; writing—review and editing; supervision.

CONFLICT OF INTEREST STATEMENT

B. Paiva reports other support from BMS‐Celgene during the conduct of the study; grants and personal fees from BMS, GSK, Sanofi, and Roche, grants from BeiGene, personal fees from Janssen, Adaptive, and Becton Dickinson outside the submitted work. T. B. Campbell reports other support from Bristol Myers Squibb during the conduct of the study; other support from Bristol Myers Squibb outside the submitted work. N. Martin reports personal fees from BMS during the conduct of the study; personal fees from BMS outside the submitted work. L. D. Anderson Jr. reports personal fees from Janssen, Celgene, BMS, Amgen, Pfizer, Karyopharm, GSK, AbbVie, Beigene, Cellectar, Sanofi, and Prothena outside the submitted work. J. G. Berdeja reports grants from BMS during the conduct of the study; grants and personal fees from CRISPR Therapeutics, grants from 2 Seventy Bio, AbbVie, Amgen, C4 Therapeutics, CARsgen, Cartesian, Celularity, Fate Therapeutics, Genentech, GSK, Ichnos Sciences, Incyte, Karyopharm, Sanofi, Lilly, Novartis, Poseida, Teva, and Zentalis, personal fees from Kite Pharma, Legend Biotech, Roche, and Secura Bio, grants and personal fees from Janssen, Takeda outside the submitted work. S. Lonial reports personal fees from Amgen, BMS, Takeda, Celgene, Novartis, Janssen, Genentech, AbbVie, Pfizer, Regeneron, and GSK outside the submitted work; and trial support from Takeda, Novartis, Janssen, and BMS; member of the board of directors with stock in TG Therapeutics (no cancer agents currently). N. S. Raje reports personal fees from BMS during the conduct of the study; grants and personal fees from Pfizer, grants from 2seventy bio, personal fees from Janssen, Immuneel, Caribou Biosciences, Sanofi, GSK, and AbbVie outside the submitted work. Y. Lin reports other support from BMS during the conduct of the study; other support from Janssen, Kite/Gilead, Caribou Biosciences, Pfizer, Takeda, Chimeric Therapeutics, NekTar, NexImmune, and Sanofi outside the submitted work. P. Moreau reports personal fees from Celgene, Janssen, AbbVie, Sanofi, Amgen, and Pfizer outside the submitted work. J. F. San‐Miguel reports personal fees from AbbVie, Amgen, BMS‐Celgene, Roche, GSK, Janssen, Karyopharm, Merck, Novartis, Sanofi, and Takeda outside the submitted work. N. C. Munshi reports personal fees from BMS during the conduct of the study; personal fees from Pfizer, Novartis, DCT, Janssen, and GSK outside the submitted work; in addition, N. C. Munshi has a patent for Oncopep issued and licensed. S. Kaiser reports personal fees from Bristol Myers Squibb outside the submitted work; in addition, S. Kaiser has a patent for uses of anti‐BCMA chimeric antigen receptors pending to Celgene/Bristol Myers Squibb. No disclosures were reported by the other authors.

ETHICS STATEMENT

The study protocol was approved by local or independent institutional review boards or ethics committees at participating sites. All the patients provided written informed consent.

FUNDING

This study was supported by 2seventy bio (formerly bluebird bio) and Celgene, a Bristol Myers Squibb company.

Supporting information

Supporting Information.

DATA AVAILABILITY STATEMENT

Deidentified patient‐level data will be made available through the global clinical research data‐sharing platform, in compliance with applicable privacy laws, data protection, and requirements for consent and anonymization. A request for data sharing can be made through the portal link at https://www.bms.com/researchers-and-partners/independent-research/data-sharing-request-process.html. Requests are reviewed by an independent review committee (Duke Clinical Research Institute) in accordance with PhRMA and EFPIA Principles for Clinical Trial Data Sharing.