Safety and Efficacy of Autologous RNA Chimeric Antigen Receptor T-cell (rCAR-T) Therapy in Myasthenia Gravis: a prospective, multicenter, open-label, non-randomised phase 1b/2a study

Abstract

Background:

Chimeric antigen receptor (CAR) T-cells are highly effective in treating hematological malignancies, but associated toxicities and necessity of lymphodepletion limit their use in autoimmune disease. To explore the use of CAR T-cells in treating autoimmune disease and to improve their safety, we engineered them with RNA (rCAR-T)—rather than the conventional DNA approach—to target the B-cell maturation antigen (BCMA) expressed on plasma cells (PC). To test the suitability of our approach, we used rCAR-T to treat patients with myasthenia gravis (MG), a prototypical autoantibody disease mediated in part by pathogenic PCs.

Methods:

MG-001 was a prospective, multicenter, open-label, phase 1b/2a study of Descartes-08, an autologous anti-BCMA rCAR-T, in adults (age ≥18 years) with generalized MG (gMG) (NCT04146051). Lymphodepletion chemotherapy was not used. In Part-1 (Phase 1b), patients with MGFA Disease Class III–IV gMG would receive three ascending doses of Descartes-08 to determine a Maximum Tolerated Dose (MTD). In Part-2 (Phase 2a), gMG patients with MGFA Disease Class II-IV would receive six doses at the MTD. Part-2 was conducted in the outpatient setting. The primary objective was to establish safety and tolerability; secondary objectives were to assess MG disease severity and biomarkers.

Results:

14 participants were enrolled (n=3 in Part-1, n=11 in Part-2). Ten participants were women and four were men. Median follow-up in Part-2 was 5 months (range: 3–9 months). There was no dose-limiting toxicity, cytokine release syndrome (CRS), or neurotoxicity. Common adverse events were headache, nausea/vomiting, and fever, which resolved within 24 hours of infusion. Fevers were not associated with elevated markers of CRS (interleukin-6, interleukin-2, and tumor necrosis factor-α). Mean [95% confidence interval] improvements from baseline to Week 12 were −5·9 [−9, −2·8] for MG-ADL, −7 [−11, −3] for QMG, −14 [−19, −9] for MGC, and −9 [−15, −3] for MG-QoL-15r.

Conclusion:

In this first study of an rCAR-T therapy in an autoimmune disease, Descartes-08 appeared safe and was well-tolerated. Descartes-08 infusions were followed by numerical decreases on MG severity scales equal to or above what is considered clinically meaningful and persisting on most recent follow-up of up to 9 months. rCAR-T therapy warrants further investigation as a potential new modality to treat autoimmune disease.

Funding:

Cartesian Therapeutics. Research reported in this publication was also supported by the National Institute of Neurological Disorders and Stroke of the National Institutes of Health under Awards Number R25NS088248 and NS115426-01A1.

Introduction

Chimeric antigen receptor (CAR) T-cells have been hailed as a versatile new class of effective, molecularly precise therapy. The CAR molecule combines the extracellular target binding domain of an antibody directed toward the desired target with the intracellular T-cell activation protein domains.¹ This combination enables T-cell activation upon contact with the target cell antigen, bypassing antigen presenting cells and multiple regulatory checkpoints.² However, due to their dependence on preconditioning lymphodepletion chemotherapy and association with severe toxicities, conventional CAR T-cells have been reserved primarily for the treatment of advanced cancers.³

Conventional CAR T-cell engineering relies on DNA to express the CAR, and gene transfer underlies much of these cells’ observed toxicities.⁴ The DNA is integrated permanently into the T-cell genome and replicates with each cell division.⁵ Lymphodepletion, usually with fludarabine and cyclophosphamide, is necessary prior to administration to create the appropriate cytokine environment for the infused DNA CAR T-cells to proliferate in vivo and reach their therapeutic concentration.⁴ However, as the activated cells proliferate, the CAR signal is also amplified. This leads to unpredictable pharmacokinetics (PK) and characteristic severe adverse events, such as cytokine release syndrome (CRS) and immune effector cell-associated neurotoxicity syndrome (ICANS) that prolong hospitalization after treatment.⁶ These aspects of DNA CAR T-cells limit their suitability for use beyond advanced cancers. To date, within autoimmune disease indications, only patients with severe forms of systemic lupus erythematosus (SLE) and neuromyelitis optica (NMO) have received CAR T-cell therapies. These have been restricted to DNA-based approaches and only in the context of expanded use and under prolonged hospital monitoring.^7,8

To expand the range of conditions treatable with CAR T-cells beyond cancer, we engineered these cells with RNA (rCAR-T), rather than DNA, on the premise that the temporary, non-replicable influence of mRNA would confer predictable PK and consequently a more favorable safety profile. rCAR-T uses the same advances in RNA engineering which enabled the widespread use of mRNA vaccines — optimal 3’ and 5’ untranslated regions, poly(A) tail length, and 5’ capping — to increase mRNA stability while enhancing translational efficiency.⁹ Since the CAR-encoding mRNA does not replicate together with the activated and proliferating rCAR T-cells, the load of CAR+ cells is determined and limited by the administered dose, and declines over time, potentially enabling more precise PK control over the therapy. Because our approach uses ex vivo T-cell proliferation, it does not require the specific cytokine environment induced by lymphodepletion. Although the therapeutic effects achieved with these approaches may be lasting, reaching the full therapeutic effect requires repeat dosing; thus, a robust manufacturing platform is required to generate a sufficient number of autologous CAR T-cells.

To test the suitability of our approach in autoimmunity, we engineered an rCAR-T to treat patients with generalized myasthenia gravis (gMG), a prototypical autoimmune disease in which autoantibodies target the neuromuscular junction, causing chronic, fluctuating, potentially debilitating weakness and muscle fatigue. There is a substantial unmet medical need for MG patients who do not respond to current (typically immunosuppressive) therapies or who experience serious side effects.^10–14 Autoantibody-producing plasma cells (PCs) are a key cellular component of MG pathophysiology.^15,16 Existing MG therapies do not adequately or specifically target PCs.¹⁷ The specific expression of B-cell maturation antigen (BCMA) on the surface of mature PCs provides an opportunity to do so.

Here we report the results of a prospective, multi-center, open-label clinical trial to test the safety and preliminary clinical efficacy of Descartes-08, anti-BCMA rCAR-T¹⁸, in patients with gMG.

Methods

Trial Design

MG-001 was a prospective, open-label, multi-center, non-randomised, Phase 1b/2a trial evaluating the safety and clinical activity of Descartes-08 in adult gMG patients requiring immunosuppression. Eligible participants underwent leukapheresis to obtain peripheral blood mononuclear cells (PBMCs), from which Descartes-08 was prepared following Good Manufacturing Practices. Immunosuppression was not withheld before the collection of PBMCs. In Part-1 of the study (n=3), each participant received three ascending weekly doses of Descartes-08 as a 15–30-minute intravenous infusion of 3·5 × 10⁶ CAR+ cells/kg (Dose Level 1, DL1) 17·5 × 10⁶ CAR+ cells/kg (DL2), and 52·5 × 10⁶ CAR+ cells/kg (DL3) to determine the maximum tolerated dose (MTD) at the safety interim analysis, prior to proceeding to Part-2 (see Clinical Protocol in the Supplementary material for details). The allowed dose margin was ±45% for all dose levels. Part-2 (n=11) tested 3 dosing schedules at the MTD: twice weekly for 3 weeks (Arm-1), weekly for 6 weeks (Arm-2), and monthly for 6 months (Arm-3). Participants were evaluated at screening, infusion visits, Weeks 8, 12, 16, and 20, and Months 6, 9, and 12. Enrollment into Part-2 would stop when sufficient data on safety, feasibility, and clinical activity had been gathered to permit the selection of the dosing schedule to be tested in later-stage trials.

Permitted MG concomitant medications were corticosteroids (≤40 mg prednisone/day equivalent), azathioprine, mycophenolate mofetil, pyridostigmine and complement inhibitors, provided the dose was stable for at least 8 weeks before the first infusion. No dosing change was allowed for concomitant MG-specific medications during the study, other than corticosteroids. The dose of corticosteroids was not allowed to be increased but it could be tapered at the site investigator’s discretion after Week 4. Intravenous immunoglobulin (IVIg) and plasma exchange (PLEX) were prohibited within 4 weeks of baseline and during the study. Other biologics, including rituximab and efgartigimod, were prohibited within 8 weeks of baseline and during the study.

The study was conducted in accordance with the principles of the Declaration of Helsinki, the Good Clinical Practice guidelines, and applicable United States regulatory standards. Independent institutional review boards provided written approval of the protocol and amendments. All patients provided written informed consent. Safety monitoring was performed by the site investigator, the sponsor (Cartesian Therapeutics) medical monitor, and an external monitoring committee. Data were entered by research staff at each site into a 21 CFR Part 11 compliant electronic database, which was analyzed by the study sponsor. In Part 1, safety was evaluated after each infusion by the site investigator and sponsor medical monitor. Throughout the study, the clinical data were reviewed periodically by a sponsor-funded safety monitoring committee containing experts who have no other relationship with the sponsor or trial conduct. Drafting of the manuscript was made by the authors listed, which include employees from the sponsor.

Participants

Eight study sites in the United States recruited subjects between January 7, 2020 and August 3, 2022. Key inclusion criteria were: age ≥18 years; diagnosis of MG (MGFA Disease Class III–IV in Part-1, Class II–IV in Part-2) with presence of an MG-associated autoantibody (anti-acetylcholine receptor [AChR], muscle specific kinase [MuSK], or low-density lipoprotein receptor-related protein 4 [LRP4]). If seronegative, unequivocal response to cholinesterase inhibitors, and abnormal repetitive nerve stimulation, or increased jitter on single-fiber EMG was required. Patients had to have MG Activities of Daily Living (MG-ADL) score ≥6 at both screening and baseline; and requiring immunosuppression. Key exclusion criteria were the presence of a major chronic illness that was not well managed; IVIg or plasma exchange within 4 weeks of baseline (first infusion) visit; and the use of non-permitted immune modulators. See Clinical Protocol in the Supplementary material for full eligibility criteria.

Intervention

Descartes-08 is an autologous CD8+ T-cell-only product that is transfected with RNA to express anti-BCMA targeting CAR protein over the course of a week. Only CD8+ T-cells are used for manufacturing since CD4+ T-cells are known for their memory function, rather than direct killing function, which is not relevant to RNA-transfected CAR T-cells that express the CAR molecule over days. Following leukapheresis, CD8 selection and manufacturing of autologous Descartes-08 cells including their ex vivo proliferation and mRNA transfection was performed. The autologous product was divided into multiple aliquots, frozen, and tested for sterility, cell quality, CAR expression, and potency. Aliquot(s) of cells were thawed at each infusion visit and infused by a peripheral intravenous line. For Part-1, participants were admitted to the hospital for their first infusion and observed as inpatient for three days and thereafter as outpatient daily until Day 7. Participants in Part-2 were assigned to a treatment Arm at investigator’s discretion, taking into consideration each participant’s preferences (predominantly determined by the time burden of each infusion schedule) and enrolment to date. Participants who received Descartes-08 as outpatients were observed for an hour post-infusion. Any participant with a fever of ≥38°C (≥100·4°F) within 7 days of infusion was admitted for 24-hour observation and evaluated for an infectious etiology with blood and urine cultures. These admissions were not considered serious adverse events (SAEs).

End Points and Assessments

The primary objective of Part-1 was to determine tolerability. The primary endpoint of Part-2 was safety (frequency and severity of AEs) to final follow-up. Secondary endpoints were mean changes from baseline at each follow-up visit for up to 12 months (a primary efficacy time point was not pre-specified) in four validated scales of MG severity: Myasthenia Gravis Activities of Daily Living (MG-ADL), Quantitative MG (QMG), MG Composite (MGC), and MG Quality of Life 15-revised (MG-QoL-15r) scores, as well as the MG Post-Intervention Status (MG PIS). MG-ADL is an 8-item, 24-point, patient-reported scale that assesses the impact of myasthenic symptoms on daily functioning. By convention, a 2-point change is considered clinically meaningful.¹⁹ QMG is a standardized, quantitative, 39-point scoring system consisting of 13 provider-assessed items, which include hand grip strength and forced vital capacity. MGC is a 10-item, 60-point weighted instrument composed of selected components of the MG-ADL and QMG scores. A 3-point change in QMG and MGC is considered clinically meaningful.^20,21 MG-QoL-15r is a 15-item, 30-point quality-of-life, patient-reported instrument.²² Currently, there is no consensus as to what is considered a clinically meaningful change in MG-QoL-15r. Anti-acetylcholine receptor (AChR) antibody titers were measured by radioimmunoassay in a CLIA-certified laboratory (Quest Diagnostics). Anti-muscle specific kinase (MuSK) and anti-lipoprotein receptor-related protein 4 (LRP4) antibodies were measured by semi-quantitative cell-based cluster assay (Oxford University).²³

Exploratory Analyses

We evaluated PK using quantitative RT-PCR for CAR mRNA. Enzyme-linked immunosorbent assay (ELISA) was used to measure soluble levels of BCMA (sBCMA, R&D Systems) in a pre-specified analysis, and for B-cell Activating Factor (BAFF, R&D Systems) and A Proliferation-Inducing Ligand (APRIL, Invitrogen) in a post-hoc analysis. sBCMA is a surrogate measure of total PCs; BAFF and APRIL are markers of B-cell survival. Tetanus, diphtheria, pertussis, meningococcus, and SARS-CoV-2 antibody titers and immunoglobulin levels were measured in a CLIA-certified laboratory (Quest Diagnostics) to evaluate plasma cell function and humoral immunity after Descartes-08. Immunophenotyping of B cells, T cells, and dendritic cells was performed by flow cytometry. Serum cytokine levels were measured using a multiplex bead-based assay (BioLegend). Post-hoc, high-throughput next-generation sequencing of Complementary Determining Region 3 (CDR3) from cDNA was conducted to follow distinct T-cell receptor (TCR) clonotypes (Adaptive Biotechnologies).

Statistical Analyses

As this was an early-phase open-label study with a dose-escalation regimen, no formal power analysis was performed. Baseline demographics and primary endpoints (type and frequency of AEs) were analyzed using descriptive statistics. Categorical variables were expressed as percentages (or fractions when n<10), while continuous variables were expressed as mean and standard deviation, or median and range for variables with skewed distribution (including those with mean/SD ratio <2). Secondary and exploratory endpoints were expressed as mean change with a 95% Confidence Interval (CI), or proportion with Standard Error (SE). SE of proportion was calculated as $\surd \widehat{\text{p}}\left(1-\widehat{\text{p}}\right)/\text{n}$, where $\widehat{\text{p}}$ is the sample proportion and n is the sample size. We calculated critical values for the 95% CI based on the t distribution using a two-tailed test with significance level 0.05 and n-1 degrees of freedom since the sample size was <30 in all analyses. To assess normality, we first used a Q-Q plot to compare quantiles of our data to quantiles of normal distribution. When n≥4 we then used Shapiro-Wilk test, with p>0·05 considered normal distribution,^24,25 and presented individual data points when n ≤3. When absolute change in exploratory biomarkers from baseline showed skewed distribution, relative change was used instead. All analyses were performed using Mathematica version 13·1·0·0 (Wolfram Research, Champaign, IL).

Role of the funding source

The Sponsor (Cartesian Therapeutics) played a part in the design of the study, data collection, interpretation, analysis, writing of the manuscript, and the decision to submit.

Results

Patient Characteristics

Sixteen patients were screened. Two did not qualify due to low baseline MG-ADL score (n=1) or lack of generalized disease (n=1) (Figure 1). Fourteen patients, ages 18–83 years, meeting all eligibility criteria, received at least one dose of Descartes-08, and were included in the safety analysis (Table 1). Most participants (10 of 14, 71%) had MGFA class III disease, with a median MG-ADL score of 11 (range 6–15), median QMG score of 16 (range 8–24), and median MGC score of 23 (range 12–30) at screening. Eleven patients had a history of anti-AChR antibodies; two had anti-MuSK antibodies; and one was seronegative for AChR, MuSK and LRP4 antibodies. Most participants (11 of 14, 79%) continued pyridostigmine, and 10 of 14 (71%) received corticosteroids throughout the study (mean dose 19.5 mg/day, range 5–40 mg/day). All participants had previously received at least one of the following: IVIg, corticosteroids, non-steroidal immune suppressants, or PLEX (Table 1).

Figure 1. CONSORT diagram of MG-001 participants.

SAE: serious adverse event.

Table 1.

Demographics and baseline characteristics¹

	Part-1 Participants	Part-2 Participants			All Participants
		Arm 1	Arm 2	Arm 3
Characteristic	(n=3)	(n=3)	(n=7)	(n=1)	(n=14)
Age, mean (SD), years	57 (16)	43 (26)	52 (16·8)	70	52 (18)
Female sex, no. (%)	2/3	2/3	5/7	1/1	10 (71%)
Weight, mean (SD), kg	83 (26)	80 (23)	88 (20)	71	84·4 (21·2)
BMI, mean (SD), kg/m2	31 (5·8)	26 (0·9)	34·5 (8·4)	28·5	31·6 (8·1)
Race/ethnicity
White, non-Hispanic	1/3	2/3	7/7	1/1	11 (79%)
White, Hispanic	1/3	0	0	0	1 (7%)
Asian	1/3	1/3	0	0	2 (14%)
MGFA class at screening
II	0	0	3/7	0	3 (21%)
III	3/3	2/3	4/7	1/1	10 (72%)
IV	0	1/3	0	0	1 (7%)
Age at disease onset, median (range), years	44 (25–58)	25 (15–54)	26 (14–79)	44	40 (14–79)
Duration of disease, median (range), years	13 (5–25)	14 (4–15)	16 (3–27)	26	14 (3–27)
Myasthenia gravis antibody status
Anti-AChR antibody	3/3	3/3	4/7	1/1	11 (79%)
Anti-MuSK antibody	0	0	2/7	0	2 (14%)
Seronegative	0	0	1/7	0	1 (7%)
Baseline score, mean (SD)
QMG	14·3 (3·5)	17·3 (1·5)	15·9 (4·5)	8	15·3 (4·1)
MG ADL	8·7 (3·8)	10·7 (3·5)	10·7 (3)	6	10 (3·2)
MGC	19·3 (6·7)	24·7 (4·7)	23 (5·2)	13	21·9 (5·7)
MG QoL15r	23·3 (3·1)	19·1 (9)	19·7 (5·1)	12	19·9 (5·8)
Prior MG therapies (standard of care)
Pyridostigmine	3/3	3/3	7/7	1/1	14 (100%)
Prednisone	3/3	3/3	7/7	1/1	14 (100%)
Other immunosuppressants	3/3	3/3	7/7	1/1	14 (100%)
Eculizumab	0	0	2/7	0	2 (14%)
Rituximab	0	0	2/7	0	2 (14%)
Prior IVIg	1/3	3/3	7/7	1/1	12 (86%)
Prior plasma exchange	0	3/3	5/7	0	5 (36%)
Diagnosis of thymoma	0	3/3	0	0	0
Prior thymectomy	1/3	2/3	3/7	0	6 (43%)
Prior MG crisis requiring intubation	1/3	1/3	2/7	0	4 (29%)
MG therapy at baseline
Pyridostigmine	2/3	3/3	6/7	0	11 (79%)
Prednisone	3/3	2/3	5/7	0	10 (71%)
Azathioprine	0	0	1/7	0	1 (7%)
Mycophenolate mofetil	1/3	0	0	0	1 (7%)

Part-1: intra-patient dose escalation

Part-2 Arm 1: twice weekly dosing; Arm 2: weekly dosing; Arm 3: monthly dosing

¹Baseline characteristics of individual participants are provided in Table s1.

Descartes-08 Manufacturing and Infusion

We successfully produced Descartes-08 from all patients despite their use of immunosuppressants, with a potency comparable to a healthy volunteer (Figure s1). Overall, final cell products were median 99·1% CD8+ (range 99·1–99·7%), 77·0% CD3+CD56− (95%CI 69·2–76·7%), and 92·0% perforin+ (95%CI 89·1–95%), with low exhaustion markers CD57 and PD1, and downregulation of Helios compared to original PBMCs (Table s2). Three Part-1 participants received a median of 6·4 × 10⁹ CAR+ cells (range 6·3–7·2) over three infusions. In Part-2, 11 participants received a median of 17·3 × 10⁹ CAR+ cells (range 9·65–33·12) divided across a median of six infusions (range 3–6). Two of the 11 participants (18%) in Part-2 withdrew and did not complete all planned infusions: one due to urticaria (Arm 1, see below) and another for personal reasons unrelated to safety (Arm 3).

Safety and tolerability

There were no dose-limiting toxicities (DLTs), treatment-related SAEs, or Grade ≥3 AEs in Part-1 (Table 2), making DL3 the MTD. One SAE unrelated to Descartes-08 (Grade 2 influenza requiring hospitalization, which occurred after apheresis but before treatment initiation) was reported in Part-1. Two SAEs were reported in Part-2. The first was Grade 3 urticaria 24 hours after the third infusion in a participant with a prior history of drug-induced urticaria; skin biopsy was consistent with a typical drug reaction, and serum tryptase and cytokine levels were normal (Table s3). The urticaria was deemed possibly related to Descartes-08 and resolved completely after the administration of intravenous steroids. Per protocol, the participant was taken off-study due to steroid administration, and no post-treatment MG severity assessments were performed. The second SAE was a non-ST segment elevation myocardial infarction occurring 72 hours after the sixth infusion in an 83-year-old man with a history of hypertension and hyperlipidemia. Coronary angiography revealed multivessel disease requiring revascularization, whereupon the patient fully recovered. The investigator deemed the event unrelated to Descartes-08.

Table 2.

Adverse events (AEs)¹ Related² to Descartes-08 per Investigators’ Assessment

		No. (%)
Term	Grade3	Part-1 (n=3)	Part-2, all Arms (n=11)	Part-2, Arm-1 (n=3)	Part-2, Arm-2 (n=7)	Part-2, Arm-3 (n=1)
Hand numbness	2	1/3
Headache	1	1/3	5 (45%)	1/3	3/7	1/1
Muscle soreness	1	1/3	1 (9%)		1/7
Nausea	1	1/3	4 (36%)	2/3	2/7
Rash	3		1 (9%)	1/3
Itchy throat	1		2 (18%)		1/7	1/1
Vomiting	1		3 (27%)	2/3	1/7
Weakness	1		2 (18%)	2/3
Line infiltration	1		1 (9%)	1/3
Fever	1		4 (36%)	1/3	3/7
Shortness of breath4	1		2 (18%)	1/3	1/7
Chills	1		2 (18%)	1/3	1/7
Diarrhea	1		1 (9%)	1/3	1/7
Gum inflammation	1		1 (9%)		1/7
Teeth sensitivity	1		1 (9%)		1/7
Night sweats	1		1 (9%)		1/7
Restless leg	1		1 (9%)		1/7
Lightheadedness	1		1 (9%)		1/7

¹Each AE was counted once per patient, at the highest grade reported.

²Relatedness includes possibly, probably, or likely

³There were no Grade ≥3 events reported in Part-1, and no Grade 2 or Grade 4 events reported in Part-2.

⁴Not associated with hypoxia

No patient showed evidence of functional immunosuppression (i.e. new hypogammaglobulinemia or disappearance of previously therapeutic levels of vaccine titers) or opportunistic infection. Headache, fever, and nausea were the most common treatment-related AEs in Part-2, reported in 5 (45%), 4 (36%), and 4 (36%) of 14 patients, respectively. All fevers occurred 4–6 hours after infusion and resolved within 24 hours. Associated symptoms were body aches (1 of 14, 9%), night sweats (1 of 14, 9%), and lethargy (2 of 14, 18%). None of these patients had positive urine or blood cultures or received empiric antibiotics. Serum cytokine levels from febrile patients showed increases in interferon-γ (IFN- γ) and its downstream chemokines CXCL10 and CCL2, but not in interleukin-2, interleukin-6, or tumor necrosis factor-α (Table s3). There were no episodes of hypotension, hypoxia, or requirement for the use of tocilizumab or steroids for treatment of or prevention of CRS.

Clinical activity

Part-1 participants showed variable reductions in mean MG-ADL, QMG, MGC, and QoL-15r scores at Weeks 3–24 (Figure s2). One participant tapered their dose of prednisone from 40 mg daily before treatment to 25 mg daily at final (12-month) follow-up.

Part-2 participants enrolled in Arms 1 and 2 experienced decreases in all disease severity scores by Week 5; in Arm-2, these reductions deepened further by Week 8 and subsequently plateaued (Figure 2). Mean [95% CI] improvements at Week 12 — the last time point when data for all Arm-1 and Arm-2 participants were available — were −5·9 [−9, −2·8] (MG-ADL), −7 [−11, −3] (QMG), −14 [−19, −9] (MGC), and −9 [−15, −3] (MG-QoL-15r) (Table 3). Repeatable numerical improvement in multiple MG scales pointing towards clinical symptom improvement was observed in all 9 participants who reached Week 8, and 8 of 9 (89%) at Week 12. By Week 16, all Arm-1 participants were off-study due to AE (n=1) or worsening symptoms requiring additional treatment (n=2); all 7 Arm-2 participants continue having MG scale scores below baseline during 6 months median follow-up (range 4–9 months). Three Arm-2 participants (43%) achieved minimal symptom expression (MSE, defined as MG-ADL of 0 or 1)²⁶ in at least one post-treatment follow-up; two of the three have maintained MSE through the most recent (6-month) follow-up. Two other Arm-2 participants who, before enrollment, required weekly or biweekly IVIg infusions have not needed further IVIg during follow-ups at 4 and 6 months, respectively. There were no changes in concomitant MG-specific medications reported in any Part-2 participants during the study period.

Figure 2. Mean change from baseline in disease severity scores for participants enrolled in Part-2.

Arm 1: twice weekly infusions (n=2, first post-treatment follow-up at Week 5); Arm 2: once weekly infusions (n=7 up to Week 16, 6 up to Week 20, and 5 up to Month 6, first post-treatment follow-up at Week 8); Arm 3: monthly infusions (n=1, first post-treatment follow-up at M9 [not reached due to withdrawal from the study]). MG-ADL: Myasthenia Gravis Activities of Daily Living; QMG: Quantitative Myasthenia Gravis; MGC: Myasthenia Gravis Composite; MG-QoL-15r: Myasthenia Gravis Quality-of-Life Revised Scale. Note that one of the three Arm-1 participants went off the study before the first post-treatment assessment (not presented in the figure). The single Arm-3 participant withdrew from study after receiving 3 of 6 planned monthly infusions. The shaded bands represent 95% CI. Arm-1 and Arm-3 lines represent individual participants.

Table 3.

Measures of Disease Severity at Week 12¹

		By treatment arm		By MG type
Variable	All, n=9 (mean change, 95% CI)	Arm 1, n=2 (individual values)	Arm 2, n=7 (mean change, 95% CI)	AChR+, n=6 (mean change, 95% CI)	MuSK+, n=2 (individual values)	Seronegative, n=1 (individual values)
MG-ADL	−5·9 [−9, −2·8]	−6, −8	−6 [−15, +3]	−6 [−11, −1]	−3, −4	−8
QMG	−7 [−11, −3]	−5, −3	−8 [−20, +4]	−5 [−10, 0]	−9, −5	−17
MGC	−14 [−19, −9]	−7, −11	−15 [−29, −1]	−14 [−21, −7]	−14, −7	−22
MG-QoL15r	−9 [−15, −3]	−8, +4	−11 [−23, +1]	−8 [−17, +1]	−10, −6	−14
MG-ADL decrease ≥2 at Week 12, No. (%)	8/9	2/2	6/7	5/6	2/2	1/1
MGC decrease ≥3 at Week 12, No. (%)	9/9	2/2	7/7	6/6	2/2	1/1
QMG decrease ≥32 at Week 12, No. (%)	8/9	2/2	6/7	5/6	2/2	1/1
MG-ADL decrease ≥6 at Week 12, No. (%)3	5/9	2/2	3/7	4/6	0/2	1/1

¹Includes data from Part-2 (Arm-1 and Arm-2) participants who completed 12-week follow-up. One Arm 1 participant was off-study prior to the first post-treatment assessment. Clinical efficacy outcomes for the single Arm 3 participant are shown in Figure 1

²All participants who had the pre-specified ≥2-point improvement in QMG also had a ≥3-point improvement.

³Post-hoc analysis of depth of response.

Biomarker analysis

CAR RNA was detected in peripheral blood 1–2 hours post-infusion and at no other time point (Figure s3A). Bone marrow biopsies are not customarily done on MG patients, so bone marrow was not assessed for the presence of CARs or CAR RNA. Comparison of TCR clonotypes using TCR sequencing on PBMCs before and 57–85 days after infusion revealed newly expanded clones dominating the overall T cell repertoire (Figure 3B). Clonotypes that expanded from Screening to Day 1 did not show similar dominance, suggesting that the expansion was an effect of Descartes-08 (Figure s3B–C). Other than transient elevations of IFN-γ, there were no consistent changes during or after treatment in any of the 13 cytokines measured (Table s2).

Figure 3. Exploratory biomarkers of participants enrolled to Part-2.

A: Mean change from baseline in anti-AChR antibody levels of participants with detectable antibodies on Day 1 (n=5). B: Relative frequency of TCR clonotypes that expanded from Screening to Day 1 (Apheresis) and from Day 1 to Day 57–85 (Descartes-08 Treatment), shown as a proportion (0–1) of all clonotypes detected at that time point. Data were available for 3 participants, one each with AchR+, MuSK+ and seronegative MG. Individual clonotypes are shown in Figure s3. C–I: Mean change from baseline serum sBCMA (C), BAFF (D), APRIL (E), anti-meningococcal antibodies (all serogroups, F), anti-tetanus antibodies (G), total IgG (H), and total IgM levels (I) in Part-2 Arm-1 (n=2) and Arm-2 (n=7) participants. All error bars represent 95% CI.

sBCMA, a surrogate marker of total PCs, was measured in all Arm-1 and Arm-2 participants and showed a highly skewed between-participants distribution at all assessed time points (median [range] at Day 1 42·08 [11·90–154·59] ng/mL, Week 5 27.92 [17·69–110·722] ng/mL, Week 8 26·27 [16·06–104·40] ng/mL, and Week 12 30·49 [20·97–112·88] ng/mL) with no appreciable relative change (Figure 3C). Circulating levels of B-cell survival factor BAFF and APRIL were also assessed (Figures 3D–E). The mean [95% CI] level of BAFF was 1050·3 [770·8–1329·8] pg/mL at baseline and decreased after treatment by up to 148 [18–278] pg/mL at Week 12. APRIL, which had a highly skewed distribution (median [range] at Day 1 50·90 [0·86–571·57] pg/mL) also decreased, by up to 46% [95%CI 23–85%] at Week 8, and 40% [0–90%] at Week 12.

Eight of the 11 participants in Part-2 (73%) had anti-AChR antibody titers by history. In 5 of 8 (63%) these autoantibodies were present at baseline (median [range] 50·90 [0·86–571·57] nmol/L) and decreased by 22% [95% CI 1–43%] at Week 8 (Figure 3A). One of two patients with documented anti-MuSK disease had antibodies at screening; no change was observed during treatment (Table s3).

To evaluate plasma cell function and humoral immunity after Descartes-08, immunoglobulin levels and titers of anti-meningococcal and anti-tetanus antibodies were measured in all Part-2 participants at Screening and follow-up. Anti-meningococcal IgG antibody titers were measurable in 6 participants (mean 4·4 [2·1–6·7] μg/mL) and decreased by 2·5 [1·2–3·8] μg/mL at Week 5, 1·5 [0–3] μg/mL at Week 8, and 1·9 [0·7–3·1] μg/mL at Week 12 (Figure 3F). Anti-tetanus IgG antibodies were detected in all 9 participants (median [range] at Day 1 93·7 [8·5–850·9] Units/mL) and did not change (Figure 3G). The median total IgG among the 9 participants was 1325 [range 383–2862] mg/dL and decreased up to 18% [95%CI −6–42%] by Week 12, while IgA and IgM levels remained unchanged during treatment (Figures 3H–I). IgE was undetectable in all participants.

Discussion

In our prospective, open-label, multi-center trial of 14 patients with gMG, Descartes-08 appeared safe and was associated with numerical changes on a range of MG outcome measures that directionally suggest clinical improvement; most notably, there was resolution of IVIg dependence in 2 participants and induction of MSE in 3 more participants. These improvements have been maintained to date in all participants who received weekly infusions for 6 weeks and currently at 6–12 months of follow-up.

This study demonstrated the feasibility of preparing autologous RNA CAR T-cells (rCAR-T) for patients on immunosuppressive therapy, of using the cells without lymphodepletion chemotherapy, and of administering rCAR-T in the outpatient setting with minimal post-infusion monitoring due to the notable safety profile of the product. While DNA-based CAR T-cell therapies are also moving to the outpatient setting, they still require close post-infusion monitoring with daily clinic visits as well as reservation of hospital beds in case severe toxicities develop. In accordance with time-restricted expression of RNA-based CAR molecules in vitro¹⁸, mRNA detection in our study was transient.

Consistent with the hypothesized mechanism of targeting PCs, we observed decreases in BAFF and APRIL, B-cell survival factors and ligands of BCMA that have previously been shown to correlate with MG severity.²⁷ There were only small numerical decreases in vaccine antibody titers and IgG with negligible decrease in sBCMA and no evidence of immunosuppression (i.e. increased occurrence of infections or complete depletion of protective vaccine titers). These observations suggest that, as expected, only a fraction of PCs was impacted by rCAR-T.

In theory, Descartes-08 could have inhibited all humoral immunity, and fear of immune suppression has prompted preclinical work on targeting specific PC subsets.²⁸ In practice, however, we did not observe hypogammaglobulinemia, susceptibility to infection, or other evidence of broader PC destruction. The measurable effect of Descartes-08 on the PC niche was, therefore, modest compared to the magnitude of numerical MG scale improvement observed. A possible explanation for this discordance is the propensity for pathogenic PC clones to reside in primary and secondary lymphoid organs such as the thymus and bone marrow,^15,29 a compartment more accessible to CAR T-cells than the loose connective tissue of the gastrointestinal tract in which the majority of non-pathogenic plasma cells reside.^30,31 BCMA is also expressed on plasmacytoid dendritic cells (pDCs) when they are activated through Toll-like receptors³² and may be an additional target for Descartes-08 cells. Chronic innate activation of pDCs drives their secretion of type I interferons promoting autoimmunity.³³ Other mechanisms, such as suppression of autoreactive T-helper cell clones by KIR+ CD8+ T cells, cannot be excluded.³⁴ Notably, we observed large and persistent changes in the TCR clonotype repertoire, the mechanism of which we have yet to elucidate.

Two recent studies of conventional CAR-T treatments in patients with autoimmune disorders highlight how toxicity may limit the broader use of DNA-engineered cells despite their therapeutic potential. In one study, five patients with refractory SLE received a single infusion of 1×10⁶ DNA-modified anti-CD19 CAR T-cells per kilogram under an expanded access protocol.⁷ While they maintained drug-free remissions for a median of 8 months of follow-up, all required inpatient admission and preconditioning chemotherapy; and all developed hematologic toxicity and CRS. In the other study, 12 patients with relapsed/refractory NMO received 0·5–1×10⁶ DNA-modified anti-BCMA CAR T-cells per kilogram in a Phase 1 trial.⁸ Most (92%) achieved remission. However, all had grade 1–2 CRS and grade 3 or higher adverse events, including neutropenia (100%), anemia (50%), and thrombocytopenia (25%). More than half (58%) developed infections, including 25% with serious cytomegalovirus infections and 8% with serious pneumonia. Other plasma cell-targeting therapies are also associated with severe toxicities. Administration of daratumumab, an anti-CD38 monoclonal antibody, in 7 patients with autoantibody-associated neurologic disorders resulted in five grade 3 or higher toxicities, including one treatment-related death.³⁵ Consistent with our initial hypothesis that rCAR-T therapy has less potential for toxicity as it obviates the need for lymphodepletion chemotherapy and has predictable PK, we observed no CRS, neurotoxicity, or hematologic toxicities in any of the participants treated with Descartes-08. While some CRS grading systems do classify any occurrence of fever as grade 1 CRS, unchanged IL-2, IL-6, and TNF-α levels during the fever episodes (Table s3), the timing of the fevers, and the lack of associated symptoms typical for CRS^36,37 contravene this classification. The safety profile of Descartes-08 allowed for repeat dosing and outpatient infusions for all 11 Part-2 participants. Only one patient required hospitalization due to urticarial rash that resolved within 24 hours of steroid treatment. Of note, allergic reactions have been reported with repeated infusions of rCAR T-cells.³⁸

Currently, MG is treated with broad immunosuppression, IVIG, and PLEX, or with recently approved disease-modifying drugs that target specific elements of the pathogenic pathway. With the exception of non-selective B-Cell inhibitor rituximab, all of these treatment modalities target mechanisms that are downstream of antibody production, and even rituximab was not associated with a meaningful decrease in antibody titers.³⁹ Complement inhibitors work at the level of the neuromuscular junction, require chronic administration (eculizumab every 2 weeks or ravulizumab every 8 weeks), and carry a black box warning for meningococcal infections and sepsis.^12,13 Efgartigimod, a ligand of the neonatal Fc receptor, is not associated with meningococcal infections and is administered as 4 weekly infusions; however, it is intended for cyclic administration with median duration of the first cycle of 10 weeks.¹¹ In contrast, the numerical decreases in MG severity scales associated with Descartes-08 appear to persist many months after all infusions were completed at Week 6. Therefore, Descartes-08 may find use as an infrequent, as-needed treatment, avoiding side effects from continuous exposure to immunomodulators and/or immunosuppressants.

Our study has the inherent limitations of a small, open-label trial, including the potential for placebo effect, and limited ability to draw inferences from secondary, exploratory and subgroup analyses. First, sBCMA assay used in the study was developed for evaluation of patients with multiple myeloma, who have significantly higher serum sBCMA values than the ones we observed in MG.⁴⁰ It therefore may not have been suitable for detecting small changes from already low baseline values. Second, three participants had MGFA Class II MG, which may improve to MSE even with conventional treatments. However, our inclusion criteria required that patients have significant disease symptoms (MG-ADL≥6) despite being on longstanding conventional regimens which were stable for at least 8 weeks prior to Descartes-08 treatment. Although it is unlikely that the observed magnitude of symptom improvement could be attributed solely to this unchanged maintenance dose, or to the placebo effect, in the absence of a control arm such a possibility cannot be excluded. The consistent temporal relationship between Descartes-08 initiation and measured MG scale improvements, with a latency of about 5–8 weeks and persistence for up to 12 months, also supports the notion that those improvements were not spontaneous, or due solely to concomitant medications. Third, there were only 2 participants with MuSK, and only 1 with seronegative MG. Analyses of PBMCs obtained from patients with MuSK MG suggest that plasmablasts contribute to the production of MuSK-specific autoantibodies in patients experiencing relapse, and that BCMA density on plasmablasts may be lower than on plasma cells.^41,42 Although it appears that the two participants with MuSK MG did have different response kinetics than others (Table 3), a larger sample size is required to confirm this hypothesis.

In addition to Descartes-08 appearing to be safe and well-tolerated, the magnitude of measured responses is promising, as the proportion of participants in our study who experienced numerical decrease in MG scales equal to or greater than what is considered clinically meaningful appeared greater than the reported placebo effects in other MG trials.^11–14 Comparison to historical controls is not conclusive, and therefore a more complete assessment of efficacy is underway in a randomized, placebo-controlled study using six weekly doses of Descartes-08 for MG (NCT04146051).

Evidence before this study

We searched MEDLINE, Embase, and PubMed databases up to February 5, 2023, for relevant clinical studies in myasthenia gravis (MG) on the use of cell therapy, with no date or language restrictions. The search terms used were “cell therapy” (or “chimeric antigen receptor”), and “myasthenia gravis.” We did not identify any reports on the use of cell therapy, whether autologous or allogeneic, unmodified or engineered to express chimeric antigen receptors, in clinical studies of any phase. There was a single report on the use of engineered T cells in an experimental autoimmune MG mouse model.

Added value of this study

MG-001 demonstrated the feasibility of producing autologous RNA chimeric antigen receptor T-cells (rCAR-T) from patients with generalized MG (gMG) receiving a background therapy of prednisone and/or steroid-sparing immunosuppressants. The method of CAR expression through RNA engineering obviated the need for lymphodepletion chemotherapy, which is required for conventional, DNA-engineered CAR T-cells. Repeated rCAR-T infusions were not associated with cytokine release syndrome, neurotoxicity, or hematologic adverse events typical of DNA CAR-Ts. Treatment of participants with weekly or twice weekly infusions for six doses was associated with clinically meaningful decreases in all measures of MG severity, including induction of Minimal Symptom Expression and elimination of dependence on intravenous immunoglobulin infusions in some participants. These effects were persistent at the most recent assessment (9 months of follow-up).

Implications of all the available evidence

This study demonstrated the feasibility of rCAR-T as a novel treatment modality for gMG. RNA-engineered CAR-Ts may offer an improved safety profile compared to other forms of CAR-T therapy. Further, this therapeutic approach may result in a numerical decrease of MG severity scales equal to or above what is considered clinically meaningful for months after treatment. Our findings require corroboration in an ongoing randomized double-masked placebo-controlled trial. More broadly, these results may support a new strategy that employs rCAR-Ts to combat autoimmunity beyond MG.