Multiple Myeloma: The Next Frontier

Article https://doi.org/10.
1038/s41467-023-44648-3
Neutrophil activation and clonal CAR-T

re-expansion underpinning cytokine release
syndrome during ciltacabtagene autoleucel
therapy in multiple myeloma
Received: 21 February 2023 Shuangshuang Yang 1,4, Jie Xu 1,4, Yuting Dai 1,4, Shiwei Jin 1, Yan Sun 1
,
Jianfeng Li 1, Chenglin Liu 1, Xiaolin Ma 1, Zhu Chen1, Lijuan Chen2,
Accepted: 21 December 2023
Jian Hou 3, Jian-Qing Mi1 & Sai-Juan Chen 1
1234567890():,;
1234567890():,;
Check for updates Cytokine release syndrome (CRS) is the most common complication of chi-
meric antigen receptor redirected T cells (CAR-T) therapy. CAR-T toxicity
management has been greatly improved, but CRS remains a prime safety
concern. Here we follow serum cytokine levels and circulating immune cell
transcriptomes longitudinally in 26 relapsed/refractory multiple myeloma
patients receiving the CAR-T product, ciltacabtagene autoleucel, to under-
stand the immunological kinetics of CRS. We find that although T lymphocytes
and monocytes/macrophages are the major overall cytokine source in man-
ifest CRS, neutrophil activation peaks earlier, before the onset of severe
symptoms. Intracellularly, signaling activation dominated by JAK/STAT path-
way occurred prior to cytokine cascade and displayed regular kinetic changes.
CRS severity is accurately described and potentially predicted by temporal
cytokine secretion signatures. Notably, CAR-T re-expansion is found in three
patients, including a fatal case characterized by somatic TET2-mutation, clo-
nal expanded cytotoxic CAR-T, broadened cytokine profiles and irreversible
hepatic toxicity. Together, our findings show that a latent phase with distinct
immunological changes precedes manifest CRS, providing an optimal window
and potential targets for CRS therapeutic intervention and that CAR-T re-
expansion warrants close clinical attention and laboratory investigation to
mitigate the lethal risk.
Over a decade, chimeric antigen receptor modified T cells (CAR-T) toxicity varies individually from clinically manageable to life-
therapy emerges to be an attractive approach against hematological threatening. Cytokine release syndrome (CRS) is the most common
malignancy, marking a cancer treatment paradigm shift. The overall adverse effect after CAR-T infusion. Severe CRS (Grade ≥ 3) is reported
therapeutic effectiveness of CAR-T is quite encouraging, but its in nearly 46% of B-cell acute lymphoblastic leukemia (ALL)1 and 13% of
1
Shanghai Institute of Hematology, State Key Laboratory of Medical Genomics, National Research Center for Translational Medicine at Shanghai, Ruijin
Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai 200025, China. 2Department of Hematology, First affiliated Hospital of
Nanjing Medical University, Jiangsu Province Hospital, Nanjing 210029, China. 3Department of Hematology, Ren Ji Hospital affiliated to Shanghai Jiao Tong
University School of Medicine, Shanghai 200127, China. 4These authors contributed equally: Shuangshuang Yang, Jie Xu, Yuting Dai.
e-mail: jianqingmi@shsmu.edu.cn; sjchen@stn.sh.cn
Nature Communications | (2024)15:360 1

Article https://doi.org/10.1038/s41467-023-44648-3
B-cell lymphoma2 patients receiving anti-CD19 CAR-T therapy, and 41% the European Medicines Agency, and the regulatory agency of Japan11
of multiple myeloma (MM)3 patients infused with anti-BCMA CAR-T. recently.
CRS is theoretically an indispensable inflammatory response In the present work, peripheral blood (PB) samples are collected
during CAR-T therapy. Appropriate CRS is thought to be helpful to from r/r MM patients who received Cilta-cel. Longitudinal cytokine
mitigate cancerous cells. However, excessive CRS can result in vital profiling and gene transcriptomes are performed to characterize the
organ injury, and often put patients at risk. Hence, an in-depth dynamics of acute inflammatory events in response to CAR-T. We
understanding of the CRS mechanism of action is warranted to prop- reveal a kinetic change in the axis of signaling activation-cytokine
erly manage severe immune response. Serological tests show that CRS release-clinical manifestation, showing a stepwise evolution of CAR-T-
occurs with a robust elevation of diverse cytokines4, of which, IL-6, IL- related acute inflammation. We also point out that multiple factors
1β and GM-CSF, mediated by the monocyte/macrophage signalings, may be able to collectively trigger CAR-T re-expansion, and subse-
play crucial roles in cytokine storm development5–7. However, the quently prompt a toxic reaction that is probably severer than tumor-
comprehensive chronological pathophysiologic landscape of CRS alone-driven CRS.
upon CAR-T therapy hasn’t yet been fully elucidated. This study provides a solid rationale for a timely and precise
Ciltacabtagene autoleucel (Cilta-cel, previously known as LCAR- application of targeted drugs, and proposes a notion that CAR-T re-
B38M) is a biepitopic CAR-T product targeting BCMA antigen, and has proliferation requires more close monitoring and effective medical
achieved remarkable success in the treatment of relapsed/refractory management to alleviate the lethal severity.
(r/r) MM patients in China3,8,9 and United States of America (US)10. In
light of it, Cilta-cel successively obtained approval from the US FDA, Results
Patient characteristics, clinical response and study design
The samples investigated were obtained from 26 Cilta-cel-exposed r/r
Table 1 | Baseline characteristics MM patients, of whom, 16 cases were from phase I Legend-2 trial as we
Total Phase I Phase II p value
reported previously3, and 10 from phase II CARTIFAN-1 trial9. Patients’
Case, n 26 16 10
demographic data are summarized in Table 1. Statistically, there were
no significant differences of age, gender, MM subtype and other
Median age (range) 56.5 (35~73) 55.5 (35~73) 60 (52~64) 0.133†
baseline characteristics between the two subpopulations from the
Sex, n (%) 1.000‡
above-mentioned separate trials (Table 1).
Male 17 (65.4) 10 (62.5) 7 (70.0) As of the manuscript preparation, the overall response rate of the
Female 9 (34.6) 6 (37.5) 3 (30.0) 26 cases was 88.5%, with 80.8% obtaining complete responses (CR).
Type of multiple mye- 1.000‡ The progression-free survival (PFS) rates were 39.2% at 3 year and
loma, n (%) 26.1% at 5 year; the overall survival (OS) rates were 53.1% at 3 year and
IgG 13 (50.0) 8 (50.0) 5 (50.0) 45.5% at 5 year (Supplementary Fig. 1a~b).
IgA 8 (30.8) 5 (31.3) 3 (30.0) PB specimens were collected at the given time points. Serum was
IgD 1 (3.8) 1 (6.3) 0 (0.0) separated from PB for cytokine profiling. Purified mononuclear cells of
Light chain 4 (15.4) 2 (12.5) 2 (20.0) PB (PBMCs) were assigned to three experimental cohorts as shown in
Fig. 1a and Supplementary Data 1.
Kappa 0 (0.0) 0 (0.0) 0 (0.0)
Lambda 4 (15.4) 2 (12.5) 2 (20.0)
Characterization of CRS
Nonsecretory multi- 0 (0.0) 0 (0.0) 0 (0.0) Within 30 days post Cilta-cel administration, CRS was observed in all
ple myeloma
patients at a median onset time of 6 days (range, 1~10 days). Fever
Extramedullary lesion, 4 (15.4) 4 (25.0) 0 (0.0) 0.136‡
signaled as an earliest event in 65.4% cases, appearing as early as 6 days
n (%)
at average. In the other 34.6% patients, hypotension, hypoxemia or
Median number of prior 4 (3~11) 4 (3~11) 4 (3~5) 0.358†
therapy Lines (range) organ malfunction presented earlier than fever in CRS outbreak. 34.6%
patients had Grade 1~2 CRS and 65.4% suffered Grade 3 or worse
Autologous stem cell 11 (42.3) 8 (50.0) 3 (30.0) 0.428‡
transplantation, n (%) (Fig. 1b). In accordance with clinical symptom, CRS grading was
Bone lesions, n (%) 21 (80.8) 12 (75.0) 9 (90.0) 0.617‡
adjusted and the change of CRS grading aligned with the kinetic
amount of circulating Cilta-cel (Fig. 1b~c). CD8:CD4 ratio of CAR-T
High-risk cytogenetic 10 (40.0) 6 (40.0) 4 (40.0) 1.000‡
abnormalitiesa, n (%) increasingly elevated as CRS progressed, reaching the peak at day
16~20 (Fig. 1d). Based on the two typical T cell immunophenotying
Prior therapies, n (%)
markers CCR7 and CD45RA12, cytotoxic CAR-T in CRS was enriched for
Proteasome inhibitor 24 (92.3) 0.749‡
effector memory (CCR7- CD45RA−) and effector memory T cells re-
Bortezomib 23 (88.5) 13 (81.3) 10 (100.0)
expressing CD45RA (CCR7- CD45RA+), particularly accumulated at day
Carfilzomib 2 (7.7) 2 (12.5) 0 (0.0) 10~12 (Fig. 1e).
Ixazomib 2 (7.7) 1 (6.3) 1 (10.0) To examine whether the baseline clinical and laboratory char-
Immunomodula- 23 (88.5) 1.000‡ acteristics had an impact on CRS, we compared several key parameters
tory drugs between severe (grade > 2) and mild CRS (grade ≤ 2) groups. The
Lenalidomide 19 (73.1) 10 (62.5) 9 (90.0) patients subjected to combination lymphodepletion (Cyclopho-
Thalidomid 12 (46.2) 6 (37.5) 6 (60.0) sphamide + Fludarabine) experienced more severe CRS than those
Pomalidomide 1 (3.8) 1 (6.3) 0 (0.0) receiving Cyclophosphamide alone (15/18 versus 2/8, p = 0.008),
Median CAR-T infused 0.595 0.575 0.63 0.336†
which was consistent with the observation previously stated13. At
dose (×106/kg) (range) (0.28~1.52) (0.28~1.52) (0.42~0.79) baseline, endogenous lymphocyte count was significantly lower in the
†
p value is calculated by t-test.
severe CRS group than that in the mild group (p = 0.016). However,
‡
p value is calculated by Fisher’s exact test. All tests are two-sided. there was no statistical difference in the neutrophils (p = 0.545) and
a
High-risk cytogenetic abnormalities are defined by the presence of the following abnormalities: monocyte (p = 0.199) count between mild and severe CRS groups.
del(17p), t(4;14), or t(14;16). One patient’s cytogenetic information is not available.
With respect to the CAR-T, the mean frequencies of CD8+ and CD4+
CAR-T in the mild CRS group were respectively 59% (0.46 × 106/kg) and

Fig. 1 | Overview of the study design and dynamic changes of CAR-T kinetics. related to CRS. AA1/AA2 denote two time points at acute aggravation phase. c The
a Overview of the study design including sample and clinical data collections, proliferation of Cilta-cel in the peripheral blood (n = 12) shown by the absolute
cytokine profiling, CAR-T immunophenotyping, transcriptome sequencing and number (mean ± SEM). d The chart shows the CD8:CD4 ratio of CAR-T (n = 9). Data
TCR repertoire analyses. b The graph (top panel) shows the mean value of CRS is presented as mean ± SEM. e The chart presents the immnunophenotypic sig-
grades in all the 26 patients on days after Cilta-cel infusion (mean ± SEM). The nature of CAR-T in the peripheral blood (n = 7). Data is shown as mean ± SEM. Tcm:
swimmer plot (bottom panel) shows the kinetics of the CRS severity in each patient central memory T cell (CCR7+ CD45RA−); Tn: naive T cells (CCR7+ CD45RA+);
within one month post infusions. One patient suffering lethal toxicity is shown a TemRA: effector memory RA+ T cells (CCR7− CD45RA+); Tem: effector memory
prolonged observation. Each row represents one patient. Colors on the bars indi- (CCR7− CD45RA−) T cells. Source data are provided as a Source Data file.
cate the grades of CRS on different days. The sign of ‘flame’ marks the onset of fever
41% (0.25 × 106/kg), and those in the severe group were respectively on these time slots, serological data were accordingly assessed to draw
62% (0.43 × 106/kg) and 38% (0.32 × 106/kg). Neither CD8+ nor CD4+ an overall landscape of cytokine profile.
reached proportionally and numerically significant differences As compared with the baseline, the inflammatory molecules
between the two CRS groups. The killing capacity of CAR-T was eval- substantially up-regulated at the late fever period, reaching a high level
uated by the apoptotic proportion of BCMA-expressing tumor in co- at the acute aggravation period (Fig. 2a). Several well-studied cyto-
culture. Though phase I and phase II adopted different effector-to- kines, such as IL-6, Granzyme B, IL-10, G-CSF and CXCL10, rocketed up
target (E:T) ratios in vitro, there was no evident tumor apoptotic dis- by 50~100 folds at the peak, indicating a highly active inflammatory
parity between severe and mild CRS cohorts. response at the periods spanning from day 6 to 15 (Fig. 2b). By the
In addition, the individual therapeutic response to CRS in terms of means of unsupervised clustering, we identified five cytokine clusters
CRS symptoms, grades, management, and outcome was summarized (CCs), among which, CC3, CC4, and CC5 displayed a positive correla-
in Supplementary Data 2. We found no significant difference in clinical tion with mean CRS grade, whereas CC2 exhibited a negative correla-
outcome between the patients with severe and mild CRS, suggesting tion (Fig. 2c). Notably, among the cytokines positively correlated with
clinical prognosis did not correlate with CRS grade. CRS severity, sIL-2Rα, sTNF RII (both belonging to CC4), and Granzyme
B (belonging to CC5) exhibited the strongest correlations (Fig.2a, d
Dynamic changes of cytokine profile Supplementary Data 3). To interpret the correlation of cytokine
To gain insight into the development of CRS, patients’ clinical pre- expression and CRS level, we adopted the computational methodol-
sentation separated the entire 21-day inflammatory reaction into ogy CONNECTOR14 and partitioned the 26 patients into two groups:
5 slots, including baseline (before infusion), latent (day 3~5), fever (day Group1 (n = 8) and Group2 (n = 18) (Fig. 3a~b). Further analysis of the
6~9), acute aggravation (day 10~15, AA), and resolving (day 20~21) mean log2FC of cytokines revealed a distinct cytokine kinetic mode of
periods. Since a robust cytokine secretion at the AA period, we col- Group1 from that of Group2 (Fig. 3c~d). The patients in Group1 dis-
lected individual samples two times (Day 10~12 and Day 13~15), deno- played a higher peak level of CC3, CC4, and CC5 cytokines, as well as
ted as AA1 and AA2, respectively to gain adequate information. Based the whole, as compared with Group2. Based on this grouping, we can

a Day 3~5 Day 6~9 Day 10~12 (AA1) Day 13~15 (AA2) Day 20~21
Day
CRS grade Pearson’s r Day
Patient r −0.2 Day 3~5
0.0 Day 6~9
CCL20 0.2 Day 10~12
CC1 CX3CL1 Day 13~15
CCL19 0.4 Day 20~21
CCL2
CXCL1 -log10() CRS grade
CXCL2 0
IL−15 2.5
sBCMA 1
sCD40 Ligand 2
CC2
5.0 3
EGF
CCL5 7.5
PDGF−AB/BB Patient
PDGF−AA 10.0
CZ−01
Angiopoietin−1 CZ−02
FGF−basic CZ−03
IL−33
sVEGF R3 JS−02
sFlt−3 Ligand JS−03
sVEGF R1 JS−04
Angiopoietin−2 JS−05
sIL−1 RII JS−06
sTNF RI JS−07
sPD−L1 JS−08
TRAIL JS−09
IL−7 RJ−01
sIL-6 Rα RJ−02
CCL11 RJ−03
CC3
CCL22 RJ−04
TGF−α RJ−05
IL−5 RJ−22
IL−12p70 RJ−23
IL−13
RAGE RJ−24
CCL4 RJ−25
VEGF RJ−26
sVEGF R2 RJ−27
gp130 RJ−29
IL−17A RJ−30
IL−1 α RJ−31
IFN−β RJ−32
IL−17E
IL−4 log2FC
IFN−α
sIL-2 Rα 4
sTNF RII
IL−18 2
IFN−γ
CXCL9 0
CC4
IL−3
GM−CSF −2
IL−1β
IL−2 −4
CCL3 Cytokine
TNF−α cluster (CC)
G−CSF
IL−8 CC1
IL−6 CC2
CC5
Granzyme B CC3
IL−10 CC4
CXCL10 CC5
IL−1ra
b
10,000
Peak FoldChange Over Baseline
R=0.56, p=8.8E−11 R=0.56, p=9.5E−11 R=0.52, p=2.8E−09

100
logFC (Granzyme B)
4
7.5
logFC (sTNF RII)
logFC (sIL-2 Rα)
3 2
5.0
2
1
1
2.5
1
0
0.0
0
0.5 1.0 1.5 0.5 1.0 1.5 0.5 1.0 1.5
IL B
Mean CRS grade Mean CRS grade Mean CRS grade

C SF
N α
C RII
IL 10
sP N−α
1
IL γ
XC F
TN L9
C 1β
IL a
M α
IF L3
IF L2
G −10
IL 8
IL 6
−8
L- 3
sT F−
e
−L
C CS
G 2R
−1
−
sI L−
−1
m
N
−C
XC
−
C
C
F
D
zy
I
n
ra
G
c Cytokine cluster (CC1) Cytokine cluster (CC2) Cytokine cluster (CC3) Cytokine cluster (CC4) Cytokine cluster (CC5)
4 6
4 1 R=-0.26, p =0.0054 R=0.19, p =0.04 R=0.31, p =0.00081 R=0.44, p =6E−07
R=0.12, p =0.21
1.0
3
Mean logFC
Mean logFC
Mean logFC
Mean logFC
0 4
Mean logFC
2 0.5
2
−1 2
0.0 1
0
−2
−0.5 0 0
−2 −3
0.5 1.0 1.5 0.5 1.0 1.5 0.5 1.0 1.5 0.5 1.0 1.5 0.5 1.0 1.5
Mean CRS grade Mean CRS grade Mean CRS grade Mean CRS grade Mean CRS grade
Fig. 2 | Dynamic changes of cytokines. a The heat map exhibits dynamic changes are shown in details in (d). b The median peak FC of each cytokine ranked at the top
of 61 cytokines (Y axis) in 26 patients (color bars on X axis) at different time slots (X 20 is shown in the plot. The peak FC for each evaluable case is shown as a grey dot,
axis) after CAR-T therapy. The cytokine level is defined as the fold change (FC) of a and the median FC value is shown in black. c~d, Scattered diagrams show the
certain cytokine value relative to its baseline, and transformed in the format of correlation between the mean CRS grade and the mean log2FC changes of cyto-
log2FC. The vertically arranged dots at the right side illustrate the correlation kines (c), as well as the correlation between the mean CRS grade and the cytokine
between the mean CRS grade and the cytokine level, of which the top three cyto- level (d). p values in a, c and d are calculated using two-sided Pearson’s r correla-
kines (sIL-2Rα, sTNF RII and Granzyme B) with the highest correlation coefficients tion. Source data are provided as a Source Data file.

a b
Stability Matrix, G= 2 Stability Matrix, G= 3 Stability Matrix, G= 4
Same fDB Tightness
Black line for the most Black line for the most Black line for the most
probable clustering: 0.92 probable clustering: 1 probable clustering: 0.9825 cluster 14
counting 1.1
1.00
Indexes Values
1.0 12
A: 1 A: 1 0.75
0.50 0.9
A: 0.84 0.25 10
0.00 0.8
C: 1 D: 1
8
0.7
B: 1
0.6 6
B: 1 B: 1 C: 0.93 2 3 4 2 3 4
Number of Clusters
G: 2 ; h: 1 G: 3 ; h: 2 G: 4 ; h: 2 Most probable
c
Cytokine cluster (CC1) Cytokine cluster (CC2) Cytokine cluster (CC3) Cytokine cluster (CC4) Cytokine cluster (CC5)
0.0
4
1 0.50
Mean log FC
Mean log FC
Mean log FC
Mean log FC
Mean log FC
2
0.00076
−0.5 3
0.00045
0 0.25 2
0.0098
−1.0 1
ay 15 0.00045
0.00 1
2 0.0033
ay 15 0.0074
ay 12 0.019
1 0.035
−1 −1.5
0 0
1 2
20 5
1
ay 12
ay ~15
1
ay 5
ay 9
D ~5
ay 9
D ~5
ay 9
ay 9
ay 9
0
0
D 0~1
D 3~1
~2
~2
~2
D 0~1
D 3~1
~2
D 0~1
~2
3~
D 6~
D 6~
D y6~
3~
D 6~
3~
D 6~
ay
ay
ay
ay
ay
D 0~
D 0~
D 3~
~
3
3
13
20
20
20
13
20
ay
ay
ay
ay
ay
ay
ay
ay
D
D
1
1
1
1
a
ay
ay
ay
ay
ay
D
D
D
D
Patient group Group1 Group2
d e
Total cytokines
1.0
CRS grade
Patient group 2 Patient group

0.0068
Mean log FC
Group1 Group1
0.0019
Group2 Group2
0.5 1
0.0267
0.0420
0.0097
0.0167
0.0 0
12
15
18
21
24
27
30
2
1
5
0
3
6
9
0
~1
~1
~2
3~
6~
ay
Days after infusion

10
13
20
ay
ay
D
ay
ay
ay
D
Fig. 3 | Dynamic correlation between cytokine expression and CRS level by log2FC changes of the cytokines in the separate clusters (c) and the total cytokines
CONNECTOR. a Stability matrix for parameter G (the number of clusters) = 2, 3, 4 (d) at different observation time. e, The graph shows the dynamic changes of CRS
separately. b Violin plots of the fDB (functional Davies-Bouldin) calculated on each grade in two groups after Cilta-cel infusion. Data is represented as mean ± SEM. The
run and for different number of clusters G (left panel). Violin plots of the total p values are determined using Mann-Whitney U test, and those with statistical
tightness (T) calculated on each run and for different number of clusters G (right differences (p < 0.05) are labeled above the X axis. Source data are provided as a
panel). G = 2 corresponds to the optimal scale for the fDB indexes minimization Source Data file.
while taking the sample size into account. Dynamic changes between the mean
clearly distinguish the temporal changes in CRS grading during the Meanwhile, a small set of molecules, such as Angiopoietin-1 (Ang-
treatment course. The patients in Group1 experienced severe CRS 1), PDGF-AA, PDGF-AB/BB and EGF, were noticeable to be down-
which peaked around day 12, whereas the patients in Group2 having a regulated as CRS progressed (Fig. 2a) and inversely associated with the
relatively mild CRS with a low peak around day 9 (Fig. 3e). The tem- severity of coagulation dysfunction (Supplementary Fig. 2b). These
poral dynamics of cytokine secretion was highly consistent with CRS cytokines play pivotal roles in stabilizing endothelial integrity16,17 which
grade regardless of individual discrepancy. was indispensable for preventing patients from coagulation system
The above results demonstrate that the timing of cytokine pro- disruption18,19. By contrast, Angiopoietin-2 (Ang-2), an angiogenic fac-
duction aligned with the clinical symptoms, revealing a highly regular tor, showed positive correlation with CRS grading (Fig. 2a). The ratio of
systemic inflammatory reaction upon CAR-T-tumor contact. Ang-2 and Ang-1, an indicator of endothelial activation13,20,21, was
responsively elevated as coagulation dysfunction turned worse (Sup-
Dysregulated cytokines and their clinical relevance plementary Fig. 2b). It implied that cytokines could pose destructive
In the Cilta-cel treatment, the top major adverse effects were hepatic effect to the endothelial steady-state condition, representing a
dysfunction and coagulation dysfunction, both accounting for 92.3% pathophysiological feature of CAR-T induced cytokine storm.
with the majority being asymptomatic. Correlation analysis of inflam-
matory factors and end organ damage identified a constellation of Immune cells-mediated signaling regulation of cytokine release
cytokines was positively related to aspartate aminotransferase (AST) To explore the upstream signaling responsible for cytokine regulation,
and alanine aminotransferase (ALT) elevation. Granzyme B and IFN-γ, transcriptome sequencing was performed in the PBMCs at different
known to cause liver impairment15, showed the strongest correlation periods. Three algorithms collectively identified 10,917 differentially
(Pearson’s r > 0.8) (Supplementary Fig. 2a). Overall, CRS induced expressed genes (DEGs) (Supplementary Fig. 3a), which were assigned
hepatic damage was transient and reversible since supportive care was into eight clusters with consistent expression trends using soft clus-
effective in most patients. tering method (Fig. 4a). We noticed that clusters 2 and 7, which

Fig. 4 | Signaling involving CRS by RNA sequencing. a Soft clustering of DEGs (GSEA) of RNA sequencing at indicated time slots. The dot size represents the
generated by transcriptome sequencing. b Gene ontology (GO) analyses of the magnitude of signaling activation (GSEA-computed p.adjust values). Pathway
biological processes indicated by core genes in the eight clusters. The dot size and direction is the median log2FC of significant transcripts relative to the baseline in
darkness represent the significance of pathways in the relevant cluster (GO-com- each pathway (blue, repression; red, activation). EC indicates endothelial cell.
puted p.adjust values). Pre-condi: pre-conditioning. c Gene set enrichment analysis Source data are provided as a Source Data file.
represented the signaling activated at the earlier phase (day 3~9), were cycle, were positively correlated with CRS severity, and cluster 7 was
significantly enriched for a variety of biological processes involved in negatively correlated with CRS severity (Supplementary Fig. 3c).
neutrophil, T lymphocyte, platelet, monocyte/macrophage activation, Considering CRS is a dynamic process, these data suggested that
cytokine signaling up-regulation, as well as excessive oxidative stress neutrophils and platelets were activated in the early stage of CRS, and
(Fig. 4a~b, Supplementary Fig. 3b). In addition, B cells affiliated genes so was oxidative stress. Distinctly, at the CRS aggravation stage, T-cell
were significantly enriched in the cluster 6, accompanied by the gene activation and proliferation were augmented.
transcription being decreasingly down-regulated since lymphodeple- To better understand the timing of immune cell activation cas-
tion and Cilta-cel therapies (Fig. 4a~b). Moreover, correlation analysis cade, gene set enrichment analysis (GSEA) was performed to compare
showed that clusters 1 and 5, characterized by T cell activation and cell each time point to the baseline. Interestingly, the neutrophils

Fig. 5 | Fatal cytokine storm in RJ-31 following the first CRS. a The grading of alone or in combination with 500 mg methylprednisolone) were intravenously
adverse events (AEs) in terms of hepatic impairment and coagulation dysfunction administered starting from day 26 as advanced supporting care couldn’t effectively
throughout the course of double CRS in RJ-31. b Kinetics of circulating Cilta-cel in control the hepatic failure worsening mainly presented by the hyperbilirubinemia
RJ-31. Absolute number of CAR-T is shown in green, and the ratio of CD8+ to CD4+ that rapidly progressed to grade 4. The dot size represents the magnitude of sig-
CAR-T is shown in black. c Cytokine profile as demonstrated by log2 FC of cytokine naling activation or repression (GSEA-computed p.adjust values). Pathway direc-
level relative to the baseline. Ang2:Ang1 represents the ratio of angiopoietin-2 to tion is the median log2FC of significant transcripts relative to the baseline in each
angiopoietin-1. Cytokines with a FC value in the top three on day 8 are shown in pathway (blue, repression; red, activation). EC indicates endothelial cell. e The DNA
blue. And cytokines that are significantly elevated only in the second CRS are copy numbers of the exogenous transgene in the liver and peripheral blood (PB) are
highlighted in red. d GSEA of the RNA sequencing at indicated time slots. To detected by qPCR on day 38. Source data are provided as a Source Data file.
clinically interfere with CRS, high-dose corticosteroids (20 mg dexamethasone

responded strongest as early as day 3~5 (latent period), while immune massive cytotoxic CD8+ T cell infiltration containing toxic granzyme B
response from macrophages and T cells started at the latent period, and TIA-1, high KP-1 (CD68) expression (Supplementary Fig. 6b~g),
and turned strong at the fever period of day 6~9 (Fig. 4c), suggesting reflective of a T cell mediated immune response in the liver, which
that neutrophils acted as a prominent pioneer in initiating inflamma- should be stronger than that in the PB owing to a higher accumulation
tory reaction, whereas macrophages and T lymphocytes were more of CAR-T. Since CAR-T were normally activated by the targeted anti-
engaged in the centre of subsequent cytokine cascade. As the immune gen, we hypothesized the presence of BCMA-expressing cells may act
cells responded, the key inflammatory signaling pathways TNF, IL-6/ as a stimulus. Albeit no imaging evidence to support tumor infiltration
JAK-STAT3, JAK/STAT, IL-2/STAT5 and complement/coagulation cas- in the liver, immunohistochemistry indeed revealed a scattered dis-
cades were significantly upregulated accordingly (Fig. 4c), which were tribution of plasma cells expressing CD138 and BCMA (Supplementary
well-known for their central roles in cytokine storm19,22–25. Similar Fig. 6h~n), indicating that the 2nd_CRS was probably initiated in the liver
conclusions were drawn by gene set scores (Supplementary Fig. 4). with cytotoxic Cilta-cel accumulation, which was most likely relevant
These data were predictive of potential targets and appropriate time to the exposure of localized BCMA antigen.
for the use of signaling blockers.
A dominant clone of Cilta-cel was identified in RJ-31 at the
CAR-T re-expansion in three patients 2nd_CRS
The majority of patients experienced CAR-T expansion once, but three Unexpectedly, RNA sequencing of PBMCs in RJ-31 uncovered a dra-
cases (RJ-25, 27 and 31) presented re-expansion of Cilta-cel following matic skewing of TCR repertoire to a dominance possessing TRAV13-
the first CAR-T peak (Supplementary Fig. 5a, b, Fig. 5). Re-expansion 1~J20/TRBV12-5~J2-1, which took up 89% of the entire TCR repertoire on
onset time in RJ-25 and RJ-27 was half a year after CAR-T infusion day 28 (Fig. 6a). TCR diversity (inverse Simpson index) and clonality
whereas that in RJ-31 was only 22 days. In parallel, RJ-25 and RJ-27 had (clonality index) indicated that RJ-31 uniquely showed a deep reduc-
very limited cytokines produced (Supplementary Fig. 5c, d). By con- tion of TCR diversity and a remarkable increase of T cell clonality since
trast, RJ-31 suffered life threatening CRS. Such discrepancies prompted day 22 (Fig. 6b). The other cases all showed a transient, light change of
our interest to figure out the potential causes. We found, at the half of TCR diversity and clonality, followed by a quick return to the original
year, RJ-25 was diagnosed with herpes zoster eye’s disease, and RJ-27 status (Fig. 6b), reminiscent of the common findings previously
was tested positive for human rhinovirus with on-going complete described26,27. Lentiviral vector integration site analysis further
responses (Supplementary Fig. 5e). Viral infection in both cases was revealed a high abundance of a clone with monoallelic integration at
symptomatic and most likely owing to persistent hypoimmunoglo- the downstream (nearly 14 kbp away) of ETS1 gene without affecting its
bulinemia as plasma cells remained deficient (Supplementary transcription confirmed by quantitative PCR (Fig. 6c, Supplementary
Fig. 5f~h). Since T lymphocytes are the key guardians of adaptive Fig. 7), ruling out the association of Cilta-cel dominance establishment
immunity, it was speculated that CAR-T might took part in immune and lentiviral insertion.
response to pathogen invasion and increasingly proliferated again. To functionally evaluate the property of the dominant clone,
With respect to RJ-31, this patient consequently died of hepatic failure purified CAR-positive cells at the 2nd_CRS (day 28), as well as the mat-
due to fatal CRS during the second CAR-T amplification. To deeply ched control in the 1st_CRS (day 11) were harvested for single-cell RNA
comprehend the origin of the excessive CRS in RJ-31, this special case sequencing. Overall, CAR-T in the 2nd_CRS was enriched for the cyto-
was investigated in details. toxic CD8+ effectors (Fig. 6d, Supplementary Fig. 8a~b). As a compar-
able control, a small dominant CAR-T clone, TRAV19~J34/TRBV4-1~J2-7
Case study of a fatal cytokine storm following the primary CRS (referred to as Major_A) in the 1st_CRS was selected for comparison with
RJ-31, a 61-year-old male, presented the first severe CRS (1st_CRS) on day the dominant TRAV13-1~J20/TRBV12-5~J2-1 (referred to as Major_B)
9, and relieved soon after Tocilizumab administration. The adverse identified 73% in the 2nd_CRS (Fig. 7e). This investigation showed that
events (AEs) of hepatic impairment (Grade 3) and coagulation dys- the Major_B possessed a stronger cytotoxic capacity respective of lytic
function (Grade 2) were transient during the primary immune reaction cell composition (Supplementary Fig. 8c) and higher toxicity score
and relieved on day 12 (Fig. 5a). A second CRS (2nd_CRS) appeared on (Fig. 6f~g, Supplementary Fig. 8d). Additionally, the significant DEGs in
day 20 with progressing jaundice (Fig. 5a), which lasted 18 days until he Major_B were involved in Toll-like receptor signaling pathway, apop-
died of liver failure and hepatic encephalopathy in spite of the absence totic mitochondrial changes and calcium iron related processes
of fever, hypotension at the initial stage. The Cilta-cel in the blood (Fig. 6h), which was in line with the finding generated by PBMCs
stream presented two peaks on day 11 and day 26~28, respectively, and transcriptome sequencing. Meanwhile, the entire CAR-T population
the ratio of CD8+ to CD4+ circulating CAR-T was significantly higher at could be representative of the clone when the dominance took up 73%
the second episode compared with that at the 1st_CRS (Fig. 5b, Sup- ~78% during day 28~38 (Figs. 5b, 6c, e).
plementary Fig. 6a). Cytokine profile confirmed that severe inflam-
matory storm consecutively occurred twice in RJ-31. The 1st_CRS was HHV7 activation and TET2 mutation might be collectively
manifested by a transient increase mainly in Granzyme B, IL-6, IL-10, responsible for clonal amplification of Cilta-cel in RJ-31
while 2nd_CRS exhibited a durable and robust rise of more cytokines, Given CAR-T clonotype rarely skewing to a single dominance
such as IL-8, IFN-γ, CX3CL1, sIL-1 RII, sVEGF R1, Ang-2 (Fig. 5c). Com- upon tumor stimulation, we speculated that the dominant clone
pared with the common signaling activation on day 8, GSEA pathway was contributed by another factor other than BCMA-expressing
enrichment showed a more fierce T cell immune activity, platelet cells. To mine the unknown peptide, the TCR clone was shown
degranulation and a stronger potency of various inflammatory sig- highly matched with herpesvirus5 (HHV5) in the ImmuQuad Bio-
naling involving Erk, IKK/NF-κB, Toll-like receptor pathways and oxi- tech database. Nevertheless, transcriptome sequencing of RJ-31
dative stress, mitochondrial damage on day 22, concomitant with PBMCs surprisingly found herpesvirus7 (HHV7) but not HHV5 was
highly active glycolysis, pyruvate metabolic and fatty acid metabolic activated starting from the resolving period of the first CRS. Back
processes (Fig. 5d). A widespread down-regulation of cytokine tran- to the aforementioned database, HHV7 was not included as a
scriptional signaling since day 26 might result from corticosteroids potential target for TCR alignment indeed. Though HHV5 and
intervention (Fig. 5d). HHV7 only shared 38.6% homology at peptide level, HHV7 was
Hepatic biopsy was obtained on day 38. The transgene copy hypothesized as a stimulus of the CAR-T dominance formation.
number of Cilta-cel in the liver was 9.5 times higher than that in the This hypothesis was verified by a transient positivity of the HHV7
matched PB of day 38 (Fig. 5e). Immunohistological staining showed a fragment between day 20 and 26 by RT-qPCR, which was exactly

a b
Day -5 4 8 11 20 22 26 28 33 TCR alpha
TRAJ47
TRAJ22 TRAJ8TRAJ29
TRAJ54 TRAJ22 TRAJ9
TRAJ29 TRAJ37TRAJ9
TRAJ11 TRAJ8 TRAJ48
TRAJ39
TRAJ23
TRAJ8
TRAJ48 TRAJ41 TRAJ43
TRAJ49 TRAJ42TRAJ32 0 TRAJ22
TRAJ41 TRAJ45
TRAJ36
TRAJ16
TRAJ13 TRAJ23TRAJ44TRAJ54
TRAJ56 TRAJ58
TRAJ26 TRAJ42
TRAJ36 TRAJ52
TRAJ54 TRAJ27
TRAJ39
TRAJ23
TRAJ21
TRAJ9 TRAJ8 TRAJ22
TRAJ41
TRAJ40
TRAJ36 0
TRAJ30 TRAJ4 TRAJ3 TRAJ15 TRAJ300 0 TRAJ39 TRAJ43 0 TRAJ22 TRAJ540 0 TRAJ4
TRAJ49 72
0.8
0 0 0 63
0 0 0 TRAJ37 0 0 0 0
TRAJ3 0 0 0 0 TRAJ6 0 0 0 0 0 0 TRAJ3 TRAJ4 0 0 0 0 0 0 0 64 64 72 70
4 TRAJ28 0 0 TRAJ36
0 0
0 TRAJ4
TRAJ4 0
0 TRAJ5 3
TRAJ30 0 0 TRAJ4 0
TRAJ5 8 0 0 TRAJ4 8
TRAJ42 0 TRAJ4
TRAJ3 TRAJ3
TRAJ4
56
0
TRAJ T8RAJ48 0 0
0 TRDJ 0 0 TRAJ210 0 0 3TRAJ TRAJ41 0
0 0 TRAJ2 0 0 0
TRAJ 56
80
56 80 TRAJ 77 TRAJ3
0
TRAJ TRAJ 10 TRAJ TRAJ
9 TRAJ 3 TRDJ 0 4TRAJ
TRA
71 TRAJ 0
9 TRAJ2 0 0
TRAJ5 24TRAJ4
0
0 0 4 40
TRAJ
0
0 0 0 TRAJ4
45 0 0
0 TRAJ TRAJ2
0 0
4 38 0 7TRAJ
57TRAJ TRAJ
8TRAJ TRAJ
34TRAJ TRAJ
TRAJ2 0 30 TRAJ 0 TRAJ
TRA
8 0 0 31TRA 48 TRA 0 0 9 TRAJ 0
49 4TRAJ
0 29
TRA 20 0 TRAJ 40 0 0 240
TRAJ 0 0
0
45 16 0 1TRA
13
0
0 28
TRA 20 48 9TRA 20
0 TRA
4
0 0 TRA
11TRA 0 TRA
12 TRAJ 0 TRA
18
J26 J53 0 5 TRAJ
0 0 TRA
J50 0
TRA
48
0 0 42 TRA
0 TRA
13
0
39
J42 0 J58 0 TRA
11J13 TRA TRAJ0 TRA
J49 J9 0 J34
0
52TRA
J47 J20 10
TRA 0 TRA
J39 TRAJ 42
0 J48
TRA
7
0 TRA
J12 TRAJ 0
TRA TRAJ5
J4 0 J53
TRA TRAJ1
0 0 J30
TRA 0
0 0 J54
TRA
J13 TRAJ 0 TRA
J28
TRA TRA
TRA 0 J15
TRA
0
J31 0 0 J44 TRA
J16 TRA 0 TRA 0 J12
TRA TRA TRA J48 0
J49
TRA
J24
TRA 0 TRA J22 0 0
TRD 0
TRA
J3 TRA 8 0 J32
TRA 0 J43
TRA
0
0 TRA
J15
TRA 0 TRA
J9
J48TRA
0
0
J53
0 J49
TRA 0
0
0 TRAJ37 TRA
J52 J21 0
TRA
J40 0
TRA 40 0 J37TRA
J4
40 0 J54TRA 0 TRA
0
J52 0
J39 J41 0
TRA 0 J30 0 0
J53 J6J13
0 0 J32 J56 J38 J23 J41 0
TRA
0 0 0
TRA 0
J17 J12
TRA 0
0 0 35
TRA
0 TRA J35 0
TRA0
TRA 0
TRA
0
TRA J27 0 J33 J10 0 J37 J45
J20
0 0
TRA00
TRA 0 0
TRA 0
TRA 0
TRA
J8
0
TRA 0
TRA0
TRA0
00
J28
TRA
0 0 0 0 0
TRA
TRA
0
J31
J39
J43
0
TRA
J53
TRA
TRA
J49
TRA
TRA
TRA
TRA
TRA
TRA
J10
J52
TRA
TRA
0
J42 TRA
TRA
TRA
J12 0 0
TRA
TRA 0
J38 0
J23 00
TRA
J37
J6
J380 0
J17 0
TRA0 0
TRA
32
32
J350 0
TRA
0
J21
J31
TRA
0
TRA0
100 1,000
TRA
TRA
J40
J47
TRA
J16
TRA
TRA
J29
TRA
J33
J24
J10
TRD
J43
TRD
0
TRA
J42
TRA
J32
J15 0
28
0
J20
TRA
J18
J47
J41
J44
J18
TRA
TRA
0
TRA
J6 0 0
J24J5
TRA
TRA
TRA
0
0
TRA
0 0
TRA
TRA
J52
J33 0
J44
TRA
J38
0
J45
J13
TRA 0
TRA
TRA
TRA
TRA
00
J38
0
TRAJ
J1TRA
0
J1
TRA
0 0
TRAJ
J32
J3TRAJ0
TRA
TRAJ
TRA
J44
TRAJ
J23
J23
J11
TRAJ0 0
TRAJ
TRAJ
J17J210 0
TRAJ
J23
J420 0 0
0
TRAJ
0
J9 0
7
TRAJ
0
0 0
J32
TRAJ0 0
TRAJ
4TRAJ
TRAJ
J5
TRAJ
TRAJ
TRAJ
TRAJ
J32
40
0
J21 0 0
24
0 0
J33
TRAJ
24
TRAJ
00
TRAJ
0
TRAJ
J36 0
TRAJ0
24
0
J5TRAJ
45
12
40
TRAJ
TRAJ
TRAJ
TRAJ
TRAJ
TRAJ
TRAJ 0
22
TRAJ
35TRAJ0
21TRAJ
21
0
21
53
TRAJ
37
00
47TRAJ1
TRAJ
TRAJ
0
35 0
1858
TRAJ
TRAJ
28TRAJ2
11TRAJ3
0 0
58
27TRAJ5
32 0 0
27TRAJ3
26
TRAJ
00
TRAJ2
TRAJ3
0 0
00
TRAJ6
32TRAJ7
5TRAJ1
29
47 0
TRAJ3
TRAJ3
0
24
TRAJ7
45
45TRAJ1
12
0 0
52
56
13
7TRAJ2
TRAJ1
0
0
TRAJ3
TRAJ4
TRAJ5
12
TRAJ7
TRAJ3
TRAJ6
0
0
0 0
0
57TRAJ1
40TRAJ8
0
TRAJ4
0 0
33TRAJ24
15
10TRAJ26
15
TRAJ2
00
TRAJ36
0 0
TRAJ2
TRAJ1
TRAJ34
0
TRAJ44
0
0 TRAJ53
Inverse simpson index

0 0
TRAJ50
TRAJ6
TRAJ43
7TRAJ47
6TRAJ8
16
16
6TRAJ44
1TRAJ47
0 0
TRAJ53
TRAJ28
0
0 0
3TRAJ3
00
16
0
3TRAJ52
8TRAJ11
14
14
14
TRAJ27
TRAJ29
TRAJ31
14
TRAJ31
1TRAJ26
5TRAJ41
2TRAJ10
0 0
3TRAJ50
7TRAJ33
00 0
00
TRAJ43
TRAJ29
00
00
TRAJ5
0
TRAJ34
TRAJ37
TRAJ17
0TRAJ29
2 TRAJ44
0
TRAJ49
0
0
TRDJ1
TRAJ45
0
TRAJ42
TRAJ20
TRAJ36
TRAJ30
0 0 0
TRAJ26
TRAJ34
0 0 00
0 0
TRAJ50
00
00
TRAJ27
TRAJ57
TRAJ57
0 0 0
0 0
0 00
0
TRAJ17
TRDJ1
TRAJ5
TRAJ57
TRAJ18
TRAJ27
TRAJ50
TRAJ9
TRAJ56
TRAJ22
TRAJ24
TRAJ7
TRAJ26
TRAJ18
TRAJ46
0 00 0
TRAJ16
TRAJ35
0 0
TRAJ50
TRAJ38
TRAJ13
0
TRAJ39
TRAJ31
TRAJ34
TRAJ7
TRAJ56
TRAJ15
00 0 0
TRAJ35
000 0
0
8
0
TRAJ22
0 0 00
TRAJ16
7
TRAJ26
TRAJ36
TRAJ40
TRAJ16
0 0 0
7
7
7
7
000 0 0
TRAJ58
TRAJ30
0 00
TRAJ38
0 0 0
TRAJ57
TRAJ31
TRAJ59
TRAJ7
0
TRAJ34
0 0 00
0
TRAJ11
TRAJ17
0 0 0 0 00
TRAJ39
TRAJ57
TRAJ21
TRAJ5
TRAJ6
TRAJ11
TRAJ21
TRAJ41
TRAJ16
TRAJ46
TRDJ3
TRAJ50
TRAJ27
TRAJ18
TRAJ57
TRDJ2
00
TRAJ56
TRAJ33
TRAJ15
TRAJ18
0 0 0
00 0
TRAJ17
0 0 0 0 0
TRAJ7
TRAJ38
TRAJ61
TR AJ 1 5
0
0
TR AJ 54
0 0 0 00
TRAJ61
TT
T R A J3 5
TRAJ7
00 0
TRAJ5
T RA J58
TRAJ56
T R AJ 1 4
0 00
RR
0 0
AA
0
0
0
0
0 0
0 0
0
0
J1
0
J3
60
0.6
TRAV13−2
TRAV36DV7
Clonality index
TRAV8−7
TRAV8−2
TRAV40
TRAV8−6
TRAV34
TRAV40
TRAV40
TRAV9−1
TRAV9−1
TRAV6
TRAV25
TRAV23DV6
TRAV39
TRAV6
TRAV34
TRAV8−6
TRAV38−1
TRAV2
TRAV22
0 0
0 0 00 0
TRAV9−1
0
TRAV22
00
TRAV8−6
0 0 0
TRAV8−2
0 0
0 0 0
00
TRAV24
0 0 0
0
TRAV6
0
0
TRAV26−1
TRAV39
TRAV26−2
0
0 00
TRAV1−1
TRAV9−1
TRAV1−1
TRAV24
TRAV10
0 0
TRAV3
TRAV8−1
TRAV26−1
TRAV34
TRAV8−6
0 0 0
TRAV8−1
TRAV8−6
TRAV38−1
TRAV10
TRAV19
TRAV8−2
0
TRAV4
TRAV39
0 0 0 00
0
TRAV35
0 00
TRAV26−
0 0
0 0
TRAV25
TRAV4
TRAV30
TRAV25
TRAV30
TRAV38
TRAV8−5
00 0 0
TRAV27
0 00
TRAV27
0
TRAV1−2
TRAV25
TRAV8−6
TRAV5
0 0
TRAV8−1
TRAV8−2
TRAV27
TRAV13
TRAV35
TRAV19
TRAV26
TRAV22
0
TRAV16
0 00
000 0
TRAV8−
TRAV12
TRAV26
0
TRAV30
TRAV16
0 0
TRAV39
7
TRAV2
TRAV1−
TRAV19 7
0 0
00
7
0 0
7
TRAV8−
TRAV1
7
TRAV8−
0
TRAV3
TRAV9−
TRAV1−
TRAV22
TRAV27
8
TRAV2
8
TRAV2
0
0
0 0
TRAV5
TRAV8
TRAV2
00
TRAV2
TRAV2
9
TRAV8
36DV7 0
−2DV8
2 TRAV3
TRAV3
TRAV9
0 0
00
0 0
00 0
TRAV8
0
TRAV1
TRAV6
TRAV1
TRAV2
0
TRAV3
−2TRAV3
TRAV2
TRAV3
TRAV1
0 0
0 0
TRAV1
0
TRAV8
−2 7
0 0
−3
TRAV2
2 TRAV
TRAV2
TRAV
TRAV
−2
0
00
TRAV
4DV4
8−2DV
00
TRAV
TRAV
0 0
0
1 −3
02
TRAV
3TRAV
1−3
0
TRAV
1 TRAV
TRAV
0 0
TRAV
2TRAV
13−1
19
TRAV
0 0
0
0 6−2
0
−3
0
8−1
TRAV
−2
14
TRAV
TRAV
0
14
0
TRAV
0
TRAV
TRAV14
TRAV
14
98−3
94
6−2
TRAV
TRAV
0
−1
TRAV
2−1
0
0 0
TRAV
1−1
TRAV
0 0
−6
TRAV
1−1
16
TRAV
TRAV
TRAV
8−3
38−1
58
7
0
TRAV
0
8−4
0
0
16
8−4
0 0
25
TRAV
V41 0
TRAV
TRAV
22
TRAV
0 0
8−2
8−1
16
TRAV
5 6TRAV
26−1
29DV
0
0
TRAV
TRAV
TRAV
0 0
TRAV
38−3
12−1
2 12−2
0
2TRAV
0 0
00
4TRAV
TRAV
TRAV
24
TRAV
23DV
TRAV
24
0
26−1
38−2
TRAV
0
TRA
22
TRA
29DV
9−2
TRA
0
TRA
8−2
00
23DV
TRA
0
−1 0
TRA
0
29DV
0
24
TRA
TRA
TRA
1−2
5 TRA
0
TRA
510
0
21
26−1
8−1
8−6
21
TRA
TRA
0
TRA
TRA
TRA
TRA
V17
TRA
V12−
TRA
0
0
35
V12−
DV8
0
00
0
0 0
V9−2
10
6
V1−2
8−2
1−2
TRA
V25
V12−
TRA
V20
0
3TRA
V13 0
0
22
0
V12−
V4
TRA
5V27
TRA
6V38
00
0
24
24
TRA
TRA
5V5
V8
TRA
TRA
0
TRA
V12−
−2TRA
V6
0
V10
TRA
0
V26−
V2
TRA
V21
0
TRA
24
TRA
0
0
−1
V23
TRA
0
V12
0
−2D
1TRA
V2
1TRA
0
TRA
V1−
TRA
TRA
V16
−2D
V21
3TRA
TRA
0
3
V20
0
V4
TRA
00
V30
TRA
V13
TRA
0
V23
0
V8−
TRA
0
TRA 7
TRA
TRA
0
V29
0
0
V21
V6
V23
TRA TRA 0
DV6
TRA 0
−2
0
TRA
V38
0
0
TRA
28
V38
1V14
TRA
V2
0
TRA
2V8−
V13
V9−
28 TRA 0 −2 TRA
V1−
0
V8
0
TRA 0
V23
V12
TRA TRA 0 TRA0 0
V12
DV6
1V5
V12
0 7
V26
TRA V14
DV5
TRA
DV6
TRA TRA
TRA TRA V1− 0 V39 0 TRA V12
−1
V10 0 TRA 0
TRA
V21 0
V10
0
TRA
0
V5
0
V12 0
DV4
6
2
V4 V41
1
0
TRA 0 V1− 0
TRA 0
TRA TRA
DV6
00
−2
V4 0 V8− 32 −1
V36
−3
0 32 TRA
DV4
−1
V29 0 TRA
V14
V3 0 DV7 TRA0
V26 −2 1TRA V24
V29 0 V13 TRA 0 −1
V17
0
0 TRA 0
TRA
TRAV 0 V39
TRA V13 TRA 2
V24 4V36 0
TRA V17
0 DV7
TRADV5 0 V17 V23D0
DV4 V5 V36 TRA0
V22 0
−1 0 0 DV5
TRAV0
−2 V8−4 0 V41 0 35
TRAV 0 V12−
TRAV
TRAV 0 V26− 0 TRAV
DV7 0
V16 TRA V16 0
V5
TRAV
38−2 0
0
TRA 7
V29D V1−2 0 0
V41
TRA V25 12−3
TRA V17 TRAV 0 TRAV
TRAV
V6 0 TRAV 0 TRAV 126−2 0 8 19 TRAV 0 V16
TRAV 0 40
DV8 0 0
12−3 21 0 TRA 0 0
20 38−2 0
TRA 19−2
TRAV
12−3 0 0 41
40
TRAV
8−4 0
TRAV
TRAV 0 13−2 12−1
TRAV TRA 29DV 13−1 TRAV
1−2 0 0 TRAV TRAV 0 0 10 TRAV TRAV
17 19 42
13−1
TRAV1 0 17 TRAV 0 0 0 DV8
TRAV30 6 TRAV
21
0
80 13−1 13−1 19
26−2 5 TRAV
9−2 0 0
2038−1 0
TRAV 13−1
TRAV2
0 0
5TRAV5
9−2 TRAV TRAV1 0 0
0
21 0 TRAV TRAV2
0
35TRAV12 0 19 48
80 48 77
TRAV
TRAV13
0 0 TRAV
DV7 TRAV120 0 0
4DV4 TRAV TRAV3
5 0 1
TRAV1 TRAV TRAV
−2
Dominant
0
TRAV35 0 0 TRAV35
0
0
0
0
0 4DV4 0
TRAV21 0 0
0
TRAV3 0 0
DV7 0 0 0 72 70
49
−1TRAV3 0 0 0
TRAV12−2 0 0
TRAV36 TRAV38−2
−2 TRAV210 0 0 0 0 0
TRAV1 TRAV8−3
0
5 TRAV12−2 0
0 0 TRAV4 TRAV41
−2 0 0 56 72 56
63 56
0
TRAV8−4
TRAV41
0 TRAV12 TRAV13−2
0
TRAV8−3 TRAV27 0 0 0 TRAV3 TRAV23DV6 0 0
TRAV36 TRAV16 0 0 TRAV17 64 64
TRAV14DV4 TRAV36DV7
TRAV9−2 TRAV25 TRAV41TRAV8−4
TRAV9−2 TRAV16
DV8 TRAV8−2
TRAV6 TRAV8−1
TRAV24 TRAV24
TRAV1−1 TRAV29DV
TRAV12−1
TRAV13−2
TRAV12−3 TRAV14DV4 TRAV14DV4 TRAV8−4
TRAV36DV7
TRAV26−2 TRAV38−2DV8 TRAV20
TRAV8−4 TRAV13−2
0.4
clone V19-J34 V13-1~J20
10
0.2
TRBJ2−4 TRBJ2−4 TRBJ1−4 TRBJ1−4 TRBJ2−4 TRBJ2−7
TRBJ1−4 6
TRBJ1−4 0
TRBJ1−1 TRBJ2−3 TRBJ2−1 TRBJ2−5
0 6 0 TRBJ2 0 0 TRBJ2−1 10 0 2 0 0 5 6 12
TRBJ2−1 49 56 TRBJ2−1 49 56 TRBJ2−1 56
36 −5 5 −1 2 15
0 0 0
TRBJ2 63 63
49
63
0 −6 TRBJ2
12
0 5
TRBJ
0
TRBJ −1 TRBJ 36 0 42 42
TRBJ 2 5 −4 42
30 TRBJ 6
0 TRBJ1 0 2−5 TRBJ2 10
0
1−1 70 70
0 2−1 0
2−3 30
6 TRB 35
70
35
2−1 6
TRB TRB 35
J2−4 5 5 0
5
5
0
0 J2−5
0 TRB TRB J2−4 2 0 J1−5 J2−2 TRB 77 77 77
24
7 6
24 28
TRB TRB 0
J1−2 TRB 1 0
J1−4
28
28
TRB
J2− 0 0
0
J1− 0 0
0 J2−
TRB
TRB
TRB J2− 0
18 1 TRB 84
84
TRB
18
84
3
TRB
J2−
J2−
0
18
21
21
15
TRB0
TRB0
TRB
21
J2−5
TRB
J2−3
15
5
0
0
J2−7
J2−7
J1−1 0
J2−3
TRBJ
TRB0
TRBJ
J2−3
TRBJ
TRBJ
TRBJ
TRBJ
0
4
TRBJ
6 0
12
0
2−7
2−7
12
12
TRB
0
2−7
J2−40
TRBJ
TRBJ
1−4 0
14
TRBJ
0
1−2
10
TRBJ
14
2−30 0
TRBJ
TRBJ
TRBJ
14
2−2 0
TRBJ
10
1−20
1−1 0
TRBJ
2−2
0
TRBJ
TRBJ
TRBJ2
TRBJ2
TRBJ1
2−4
TRBJ1
1−6
TRBJ1
2−4
0
2−3
TRBJ1
TRBJ2
TRBJ2
1−5 0
TRBJ1
0
1−2
TRBJ2
TRBJ1
0
TRBJ1
TRBJ1
0
2
2−6 0 0
2−6
TRBJ1
0
6
TRBJ2−
TRBJ1−
TRBJ1−
TRBJ2−
6
TRBJ2−
−7 0 0 0
−7
−4TRBJ2−
TRBJ2−
TRBJ1−
6
TRBJ2−
−5TRBJ1−3
TRBJ1−
0
TRBJ1−
0
5
−2 0
TRBJ1−
−2
7
−2TRBJ1−6
5
TRBJ1−3
−2
−2
TRBJ1−1
−2
TRBJ1−3
−2
TRBJ1−5
TRBJ2−5
0
−1 0 0 0 0
−4
TRBJ1−4
0
TRBJ2−6
00 0
TRBJ1−5
−6
0
TRBJ1−3
TRBJ1−3
TRBJ1−3
TRBJ1−3
TRBJ2−6
0
TRBJ1−5
0 0
66
5TRBJ1−2
TRBJ2−2
TRBJ1−6
TRBJ1−6
6
1TRBJ1−3
0
2
TRBJ1−6
0 0
5
TRBJ2−6
TRBJ2−6
5 0 0
5 0
0 0
0 0
0 00 0
TRBJ1−6
3 0
5 0
5
0
TRBJ1−6
0
0
0
0
0
0
TRBTRBV6−7
VTRBV1
TRBV26
BTRBV5−8
TRBV6−9
TRBV7−1
0.0
BV6−182−2
TRTRBV5−7
TTRTRBV10−2
TRBV11−3
V−752−8
TRTRBV6−6
RBBVV1101−−13
B7V−71−5
TRBV5−8
0 0 0 00
V1
V1 2 −5
TRTBRV
TTRBV7−7
0
B V 12−4
00 0 0 0
1−
0
B V 6 −4
TRTB
T R B V1 4
TRTRBV7−3
TR
0 0 00
TRBTVR1B
TR
TRT
TR
0 0 0
73−
000
0 0
TRTRBV11−3
TTR
0 0
TTRBV10−1
TRBV7−4
TB
0
0
00
VV
TRBV24−1
BTRBV3−1
TBR
B
RV
0
R
0
RV
RVB
B
B
TRBV7−1
RV
TTRBV4−3
TRBV6−9
0
R
TRBV4−1
TBB
1TRBV11−2
VTRBV3−2
BVTRBV13
BBTRBV10−3
R
0 0 00 0
B
TRBV6−8
TB
0 00 0 0 0 0
R
VTRBV3−1
T
VTRBV6−4
TRBV7−7
TRBV23−1
0
B3
0 00 0 0
0 0 0 0 0
0 0 00 0 0 0 0
0 0 0 00 0 0 0 0
V
0B1
1V
TTRR
0
TRBV15
0 0 0 0 0 0 00 0
1TRBV23−1
VVTRBV23−1
TR
1VTRBV15
R
B
TR
RTRBV3−1
7
1
TRBV11−3
TRBV12−4
B−
TRBV5−4
V−62
2−
−0
0 0 0
1−TRBV4−3
VB
10TRBV5−4
TRBV24−1
47−TRBV7−2
7−−44TRBV7−3
2TRBV10−1
V
6
TRBV7−2
VTRBV4−2
1
TRBV6−1
00 0 0 0
1
6−V−
−TRBV6−1
TRBV10−1
7TRBV10−1
−TRBV12−4
91TRBV21−1
−
44TRBV21−1
−4 TRBV21−1
1 TRBV6−2
−2TRBV7−6
TRBV4−1
26 TRBV6−2
TRBV18
2TRBV11−1
2
TRBV12−3
2
TRBV4−1
3 TRBV10−2
16
TRBV29−1
TRBV23−1
TRBV25−1
3TRBV10−3
TRBV21−1
0 0 00 0
0
6
TRBV11−1
TRBV7−3
5
TRBV12−5
7
5
TRBV25−1
TRBV7−6
TRBV12−1
TRBV30
TRBV7−3
00 0
TRBV4−1
TRBV6−4
TRBV3−2
TRBV6−6
0 0
0 0 00
0 0 00
7
TRBV21−1
TRBV15
TRBV6−6
TRBV10−2
TRBV24−1
0 0
TRBV6−2
0 0 0 0
TRBV15
TRBV24−1
TRBV4−2
0
TRBV6−5
TRBV27
0 00 0
TRBV21−
TRBV6−2
TRBV7−1
TRBV25−
TRBV5−5
TRBV11−
0 00 0
TRBV3−1
0
TRBV20−
TRBV10−
TRBV5−8
TRBV30
TRBV21−
TRBV18
TRBV7−8
TRBV12
TRBV6−4
TRBV7−
0 0
0 0
0 0
0 0
00 0 0
TRBV18
TRBV6−
00
TRBV15
TRBV7−
TRBV10
9
TRBV7−
TRBV5−
TRBV12
12
TRBV7−
TRBV25
TRBV25
TRBV7−
10
TRBV4−
TRBV7−
0
0 0
TRBV1
10
TRBV5−
0 0
TRBV7
TRBV1
0 0
TRBV7
12
0
TRBV6
TRBV2
TRBV7
0
14
TRBV7
TRBV2
0 0 0
12
0
0
0 0
0 00
14
TRBV9
TRBV1
TRBV6
TRBV5
TRBV2
14
TRBV7
1 1TRBV2
TRBV1
TRBV2
11
2−5
0
3−3
00 0
0 5 10 15 20 25 30 35 0 5 10 15 20 25 30 35
TRBV6
0
TRBV3
0
−1
TRBV1
3 TRBV
0
8TRBV5
0
2
TRBV
1TRBV7
−3
TRBV6
TRBV
61−3
−3
42−3
TRBV
TRBV1
TRBV9
TRBV
6TRBV
−1
0 0
0 0
−1
TRBV4
0 0
0
TRBV1
−3
0
TRBV
2 TRBV
3−2
00
−81−1
0 0
0
−2
9−1
TRBV
TRBV
7−9
53−1TRBV
TRBV
TRBV
TRBV
−7
0
14
TRBV
TRBV 0
TRBV
0 0
−5TRBV
−6
18
3−1
0
TRBV
−7TRBV
0
TRBV
0−3
9−1
TRBV
−5
12−5
00
0
11−3
−1TRBV
TRBV
TRBV
00
TRBV
TRBV
321−1
4−1
−5
12−3
TRBV
TRBV
11−2
TRBV
11−2
0 0
00
0
−4
−4TRBV
0 0
21
18
7−2
6−3
0
4TRBV
TRBV
11−1
0
7−2
TRBV
21
9 TRBV
24−1
6−5
5−6
5
4−3
0
TRBV
5−1
0 0
TRBV
TRBV
0
21
TRBV
TRBV
1512−5
197−8
0
0
6−2
0
TRBV5
0
6−1
TRB
12−3
00
TRB
TRB
0
13
12−3
TRB
11−2
TRB
0
TRBV
0
6−1
24−1
0
TRBV
7−8
TRB
TRB
27
TRB
0
25−1
0
TRB
6−5
0
TRB
0
10−3
0
TRB
0 0
TRB
6−6
TRB
24
0
V28
14
5−6
V6−5
TRB
6−1
5−5
20−1
0
TRB
TRB
0
V21−
15
V25−
TRB
TRB
V5−6
TRB
V27
TRB
0
0
TRB
V23−
V6−1
TRB
0
V2
TRB
TRB
2TRB
V27
0
V4−3
TRB
0
0
V11−
TRB
0
V9
V2
TRB
TRB
TRB
6
V12−
28
0
0
0
0
0
V19
0
TRB
TRB
V19
V2
TRB
V7−6
V19
TRB
V5−4
24
TRB
0
V5−
0
V20
11
TRB
TRB
0
2
V4−
28
V3−
TRB
TRB
V12
28
TRB
TRB
V28
0
0
0
V30
TRB
V5−
V6−
TRB
V6−
0
V13
TRB
2TRB
V28
0
0
0
TRB
V29
TRB
0
V27
V5−
TRB
0
TRB
TRB
4
TRB
V4−
52
V9
TRB
0
−1
TRB
0
TRB 0
V5−
TRB
2V25
TRB
1V7−
−4
TRB
V4−
0
0
0
0
−1 TRB 0
V28
TRB 0
TRB
4V30
0
30
V6−
9
V19
TRB TRB
TRB
0
−1
TRB TRB 0 V110 5 TRB
4V7−
V4−
V12 V25 0
V11
V5−
2V6−
TRB TRB
V28
0 0
Days after infusion Days after infusion
4
V14 V2 0
V19
V6− 0 V29
−1
TRB V7−
1
3
V9 12 0 TRB V7− V6− 0
3
0 0
TRB
V4−
0
V12 −3 −2 V3−
−1 0 V5− 35 V5− 35
1
TRB
−2
30 V19 5TRB TRB TRB
9
6
0 TRB
5TRB 0 0 TRB V12
TRB
0
V13 9 35
TRB
V5−
3
0
−4V18 0 0
V11− −3V10−0 0
TRB
V18 1V5−
TRB 0 TRB 6TRB0
V5− 0
0
1V7−0 1 V5− 0
0
V3−1 V24− 0
TRBV TRB TRBV 0 TRB1 1V6−6
TRBV TRBV TRB 84
1
TRBV 0
TRBV 0 0 V28 0
0 V10−
1TRBV 0 9 V4−1
0 84
1
TRBV 0 17−6 0 TRBV
V24−
2 TRBV
3 TRBV
0
V20− 0 84 42
18TRBV 0 7−9 TRBV
0 0 TRB 15 0 0 10−2
TRBV 0 TRBV
2184−30 0 6 12−5 42 42
6
27 0 TRBV
129−1 0 13
0 4−2
TRB TRBV 4−1 12−5
5−6TRBV2 TRBV 0 20−1 TRBV7
12−5 0 TRBV 0 5
29−1 5−6 77
27 0 0 10−2
TRBV10 6 −9 TRBV6
5−8 0 0 0
27
TRBV 18
TRBV1 0
0 TRBV20 TRBV 77 49 TRBV 77 12−5
TRBV10
0 TRBV7−0 TRBV3 0 TRBV TRBV6
−2 0 4−2 6−1
TRBV140 0 12 TRBV
0 6 9 TRBV7−2 0
−5TRBV4− 0 −2 TRBV4−0
0 0 0 −5
TRBV TRBV5−0 0
0−3 0 −1 TRBV 0−1 0
49
TRBV
49
TRBV14
−2 TRBV20−1
0 0 8 0 0 0 0 TRBV7 0 0 0
0
TRBV7 −6TRBV7−3 0 0 0 0 TRBV13 0 0
TRBV5−4 0
6 70 56 70 70
0
TRBV29−1 0 TRBV5−1
3 0 0 0 TRBV5−1
3 0
0 TRBV6 6TRBV28 0 0 0 TRBV20 TRBV2 0 0 63 63 56 63 56
TRBV28 TRBV5−6 TRBV18 TRBV7−8 TRBV7−6
TRBV3−1 TRBV14
TRBV7−2 TRBV4−2 TRBV20−1TRBV9
TRBV14
TRBV10−1
TRBV5−6
TRBV7−8 TRBV10−3
TRBV27 TRBV9
TRBV6−1 TRBV19
TRBV2 TRBV5−4
TRBV2 TRBV6−1
TRBV11−2 TRBV28
TRBV29−1 TRBV7−9
Dominant
V4-1~J2-7 V12-5~J2-1 TCR beta
clone
0.8
1,000
Inverse simpson index
c
100 Gene
Clonality index
Relative abundance (%)
0.2 0.4 0.6

LowAbund PIK3R4 UBE3C CIAO3
100
ETS1 ARHGAP35 RSPRY1 NELL2
75 TBC1D5 TLE4 PPP3R1 NAV2
PEX14 ZZEF1 CYTIP LINC01450
10
50 CDK17 KLRF1 CCSER1 PHF14
PGM1 MALAT1 ATAD5 LOC101929019
NASP TAF4 RBBP8 LOC101928851
0.0
25 NUB1 ZNF608 G3BP2 C12orf42
1
EHF MSH4 ZMYND11 MIR8054 0 5 10 15 20 25 30 35 0 5 10 15 20 25 30 35
Days after infusion Days after infusion
0
Patient RJ-22 RJ-24 RJ-26 RJ-31
Day11 Day22 Day28 Day38 Day38
Liver RJ-23 RJ-25 RJ-27 RJ-32
Blood
d f g
100 6 P < 2.22E−16 P < 2.22E−16
P < 2.22E−16 6 P < 2.22E−16
P < 2.22E−16
P < 2.22E−16
Gzmb_Gnly _related P=2.4E−12 P=0.018
CellType
75
Pecentage %
Cytotoxic score
cytotoxic score
CD4+ GZMK+ Exhausted 4
CD4+ Central Memory
50 CD8+ NKT 2 2
CD8+ GZMK+ Exhausted
CD8+ TIGIT+ Adhesion
25 CD8+ TRBV12−5+ Cytotoxic 0 0
CD8+ Cell cycle
_A
_B
_A
_A
_B
_A
er
er
0
er
er
or
or
or
th
th
or
or
or
th
th
aj
aj
aj
O
aj
aj
aj
O
O
RS RS
M
M
st _C nd _
C
1 2 1st_CRS 2nd_CRS 1st_CRS 2nd_CRS
st nd
e 1st_CRS h 1 _CRS 2 _CRS
Major_A
Major_A
Major_B
Other
Other
TRA:V19~J34 (Major_A)
TRB:V4−1~J2−7
(Major_A) Clone Rank Leukocyte adhesion to vascular endothelial cell
Top 1 Acute inflammatory response
Response to tumor necrosis factor
Top 2
Toll−like receptor signaling pathway
Top 3 Regulation of metal ion transport
Top 4 Response to metal ion
Top 5 Release of sequestered calcium ion into cytosol by sarcoplasmic reticulum
Rest Release of sequestered calcium ion into cytosol by endoplasmic reticulum
Apoptotic mitochondrial changes
-log10(p.adjust)
2nd_CRS Response to calcium ion
Humoral immune response 2
TRA: V13−1~J20 (Major_A) I−B kinase/NF−B signaling 4
TRB:V12−5~J2−1 Regulation of apoptotic signaling pathway
6
(Major_B) Clone Rank Type I interferon signaling pathway
Intrinsic apoptotic signaling pathway in response to DNA damage
Top 1
Response to IFN-r
Top 2 GeneRatio
T cell activation
Top 3 Natural killer cell activation 0.15
Top 4 Natural killer cell mediated cytotoxicity 0.10
Top 5 Phosphatidylinositol 3−kinase signaling
Prostanoid metabolic process 0.05
Rest (Major_B)
Unsaturated fatty acid biosynthetic process
i j
45
Protein length 30
VAF of TET2mut (%)
15
38.07% 31.21% 34.20%
TET2
NM_001127208 23.63%
Catalytic core domain 4.42%
Cys-rich DBSH Low complexity DBSH 0.6
P=0.014
0.4
/ / 1000 12
1200
200 1400 1600 1800 2000 0.40%
0.2 0.40%
0.27%
p.Q1834X 0.0 0.27%
0.10%
(RJ-31 mutation)
- + + + - +
3 MC D3 D3 C R -
1 R R 2
H2
O
HC HC PB C C CA PBM CA BMC CD3 + CA + CAR
HC
a y11 y22 y28 P
CD3 D 3
C
Healthy Pre-treatment D Da D a
Contorl in RJ-31 Post-treatment Post-treatment in
in RJ-31 RJ-31 on Day 33
consistent with the course of 2nd_CRS (Supplementary Fig. 9a~b). RJ-27, both of whom also had a high CD8+ CAR-T re-expansion. At the
Hence, HHV7 infection was probably the key stimulus for the same time point, the endogenous T cells exhibited a higher CD8+
corresponding CAR-T clone re-expansion. fraction compared to that at the early treatment phase (Supplemen-
Along with the engineered cells predominately with the CD8+ tary Fig. 9c). Taken together, the three cases shared a common feature,
subset outgrew (Fig. 5b), CD8+ endogenous (CAR-negative) T cells which was, a high CD8 CAR-T and endogenous T cell subpopulation
preferably proliferated as RJ-31 suffered from viral infection (Supple- level at virus infection. This phenomenon indicated a competent
mentary Fig. 9c). It was reminiscent of the viral infection of RJ-25 and immunity against the exogenous microbe. However, we noticed that

Fig. 6 | Identification of a dominant CAR-T clone in RJ-31. a Circle graphs show charts show the cytotoxic immunophenotypes of two dominant clones respectively
the type of genetically VDJ rearrangement of TCR alpha chain (TRA) and beta chain derived from CAR-T in the two CRSs. The annotations for the cytotoxic T subset are
(TRB). The type of dominant clone detected at certain period is shown below the from our database (f) and those reported previously62 (g) (n = 317, 5250, 4832, 19,
graphs. b TCR diversity (inverse Simpson index) and T cell clonal expansion 1756 cells following the order of the groups on the X axis). Box plots show the first
(clonality index) in eight patients after Cilta-cel treatment. c The abundance of quartile (lower end of the box), the third quartile (upper other end of the box) and
integration site marks CAR-T clone in the blood and liver of RJ-31. Top ten major the median value (centre line). The p values are determined using two-sided
integration sites are shown in different colors, and the other sites with relatively low Mann–Whitney U test. h GO enrichment of the top DEGs in the two dominant
abundances are in grey. d Proportions of CD4 or CD8 subtypes in the purified CAR- clones (GO-computed p.adjust values). i, Diagram of the TET2 catalytic domain. The
T population at the first and second CRS strikes, respectively. e The pie chart shows heterozygous point mutation (arrow) identified in RJ-31 is at the C-terminal of TET2
the frequencies of the top 5 ranked clones in the purified CAR-T population. The protein, where the catalytic domain locates. j, Statistical plot of the variant allele
clone with the biggest proportion is defined as a dominant clone. The clone type is fractions (VAFs) of TET2mut in RJ31 (n = 3 independent experiments). HCs represent
characterized with a pair of specific TRA and TRB. The UMAP plots show the three healthy controls. Data is presented as mean ± SEM. The p value is determined
dominant clone distribution at the 1st_CRS (blue) or 2nd_CRS (red). f~g The violin using two-sided t-test. Source data are provided as a Source Data file.
the T clonotype of RJ-25 and RJ-27 kept the polyclonal state (Supple- CRS is a result of an activation of a multicellular network. Other
mentary Fig. 5i). This triggered us to explore additional cause poten- than T lymphocyte and monocyte/macrophage, the primary cytokine
tially contributing to the accumulation of a single CAR-T clone. producers recognized5–7, neutrophils were found the earliest activated
Interestingly, PBMCs transcriptome sequencing showed a het- after Cilta-cel infusion. As innate immune effectors, neutrophils work
erozygous mutation in TET2 gene from day 20 on. This newly at the first line of response for the acute inflammation31 and take part in
identified abnormality was validated by targeted DNA sequencing, the pathophysiology of CRS in COVID-1932,33 and acute GVHD;34
which demonstrated the variation (C > T) of TET2 localized at the whereas, the role of neutrophils in CAR-T related CRS remains unclear.
genomic site of Chr4:106197167, resulting in an early translational Our finding provides compelling evidence that neutrophilic immu-
termination of the catalytic structural domain of TET2 protein nostimulatory reaction also drives a major cellular response at the
(referred to as TET2mut, Fig. 6i, Supplementary Fig. 10a~b). TET2 early stage of CRS outbreak in Cilta-cel therapy.
mutation is commonly considered as an initiating event of clonal The stepwise procedure of Cilta-cel-induced CRS depicted a
hematopoiesis28,29, and dysfunction of TET2 protein causes an kinetic change in the axis of signaling-cytokine-presentation. On one
exacerbation of immune inflammation, particularly characterized hand, inflammatory signaling pathways are shown to be activated in
by an IL-6 elevation30. To precisely determine if a low abundance of 3~5 days prior to cytokine storm, illuminating that close monitoring
TET2mut existed before the 2nd_CRS, ddPCR was performed and and essential supporting care are needed as early as day 3~5 to reduce
revealed a small TET2mut clone already existed in PBMCs at pre- the risk of unmanageable toxicity caused by delayed intervention. On
treatment indeed and mainly concentrated in CD3 positive host the other hand, the hyper-activation of IL-2/STAT5, IL-6/JAK-STAT3,
T cells (Fig. 6j, Supplementary Fig. 10c). This variation was also TNFα signaling indicates the potential targets for drug administration.
detected in the CAR-T on day 11 with low variant allele fraction (VAF) Tocilizumab is the most common agent to help control CRS25. TNFR
of 0.40%, and on day 28 with higher VAF of 38.07% by ddPCR blocker, Etanercept, has been applied in some cases3. Other agents
(Fig. 6j). Moreover, in the liver biopsy on day 38 when the case RJ-31 approved for controlling GVHD, such as JAK2 inhibitor (Ruxolitinib), IL-
died of the lethal toxicity, TET2mut was found to have similar VAF 2 receptor antagonist (Basiliximab), can also be considered used
levels to that of day 28 peripheral CAR-T (Supplementary Fig. 10d). against the over-reacted inflammation in the CAR-T therapy22,23.
It was worth noting that TET2mut VAF of the CD3-negative subset was Meanwhile, the optimal time point of targeted agent usage is expected
merely 0.4% at a late time point (day 33), which was basically close not to be later than day 10, at which, both cytokines and activated
to the VAF (0.1%) at pre-treatment (Fig. 6j, Supplementary Fig. 10c). signaling are reaching peak levels.
It suggested that the somatic TET2mut present at leukapheresis was Although re-expansion of CAR-T could occur under certain
more carried by T lymphocytes, consequently potentiating the circumstances35–37, multi-factor-induced CAR-T re-expansion hasn’t
engineered T cells clonal expansion, and TET2 mutation was unable been elucidated yet. In this study, we report a rare case who suffered
to drive non-T hematopoietic cells clonal expansion under the cir- two sequential CRS hits within one month post CAR-T infusion. To our
cumstance of microbial infection. knowledge, among 219 r/r MM cases receiving Cilta-cel therapy so
In sum, virus activation, a rare somatic mutation of TET2, along far3,9,10,38, this was the only one who developed a lethal CRS with an
with plasma antigen stimulation in the RJ-31 liver might collectively evidenced clonal expansion of CAR-T. We assume that several factors
contributed to the clonal amplification of Cilta-cel, prompting a sec- may result in this unexpected issue: first, late immune reaction could
ond severe CRS and conferring an excessive inflammation, and fatal be induced by CAR-T and antigen-exposing cells interaction again at
damage to the liver function. localized lesion39; second, virus reactivation may occur in the setting of
immune incompetency after CAR-T treatment40,41; third, the somatic
Discussion TET2 mutation is most likely the key intrinsic cause of CAR-T’s clonal
This study uncovered an orchestrated chronological process of expansion and subsequent fatal inflammatory storm in RJ-31, which
inflammation in response to CAR-T. The CRS pathophysiological pro- was not relevant to Cilta-cel manufacturing.
cesses investigated in this work are summarized in Fig. 7. We pointed With respect to the somatic TET2mut found in this study, RJ-31 aged
out that CAR-T could be activated by either targeted tumor antigen or 61 was speculated to have the clonal hematopoiesis of indeterminate
TCR-specific pathogen. In the common setting, CAR bound tumor cells potential (CHIP), which occurs in apparently healthy elderly people in
to produce pro-inflammatory factors that initially lead to neutrophils particular, with DNMT3A, TET2 and ASXL1 loss-of-function mutations
activation, followed by participation of other immune cellular com- being the most commonly detectable ones42,43. A small clone with TET2
ponents. In a viral invasion setting, the viral peptide presented by mutation could conceal in RJ-31’s T lymphocytes before their collec-
antigen presenting cells could allow CAR-T to be activated in a classical tion for CAR-T preparation, constituting a risk of wild outgrowth of
TCR/MHC manner. Other factors, such as target antigen exposure in CAR-T cells derived from the autologous lymphocytes under certain
the malignant or normal cells might reinforce CAR-T expansion, and circumstance, such as viral infection. Nevertheless, the other two
genetic abnormality (such as somatic mutation of TET2) in CAR-T patients (RJ-25 and RJ-27) were not detected with somatic TET2 muta-
might confer a proliferating advantage as well. tion, as well as other gene mutations associated with CHIP including

Hospitalization Discharge
Common CRS (n=26) Rare CRS in RJ-31 Undetectable

CRS in RJ-25
Immunity and RJ-27
CRS Grade5
grade Grade1 to Grade3
......
Days 0 3 9 12 15 18 21 24 27 30 33 36 39 180
Tumor-driven Multi-factors-driven Pathogen-driven
Homeostasis
Blood
Antigen Presentation
Antigen Presentation
Cell-cell interaction
Cell activation Granzyme B
sIL-2Rɑ Activated cells put into
Cytokine release/
the battle immediately
Interaction processes of the main cellular components
Degranulation sTNF RII

IL-6
IL-1ra
pMHC
ROS IL-10 ROS
IL-8
CXCL10
TET2
Polyclonal proliferation IFN-γ
Clonal re-proliferation Polyclonal
TNF-ɑ
Angiopoietin-2 re-proliferation
......
TET2
TET2
TET2
Organ
TET2
TET2 TET2
TET2
ROS
Initial
Antigen presenting cell Plasma cell CAR
CAR-T
BCMA
Cytokines
Highly-effecive Neutrophil Monocyte Diverse TCR
CAR-T Virus
Platelet
Persistent ROS Reactive Oxygen Species
Macropage Mast cell
CAR-T
Fig. 7 | Development of CRS after Cilta-cel therapy. Schematic illustration shows boosted up in a TCR/MHC manner, even are able to expand clonally. The immune
the CRS development upon Cilta-cel therapy. Within one month after CAR-T infu- reaction in both blood stream and tissues can be accordingly enhanced and might
sion, the common CRS is driven by the contact of CAR-T with tumor antigen, cause irreversible organ damage (middle). Beyond six months after infusion, low
followed by participation of immune cellular components (left). Under the cir- level persisting CAR-T skew to a memory-like status and are less likely to induce a
cumstance of pathogen invasion and/or genetic mutation, activated CAR-T can be robust cytokine storm against microbial infection (right).
DNMT3A and ASXL1, so the risk of TCR clonality formation was largely CAR-T case probably lies in the following reasons: first, the distinction
minimized. Therefore, for the RJ-31 case, the genetic defect was con- in CAR-T immunophenotypes. The cytotoxic effector phenotype in RJ-
sidered the overriding determinant to drive the clonal proliferation of 31 was of higher cytokine-producing capacity, compared with the
CAR-T, and the viral infection was probably an antigen stimulus to central memory phenotype found in the CD19 CAR-T case; second, the
trigger adaptive immunity with CAR-T cells being a part. difference in collaborative effects. Albeit TET2mut should play a pro-
The phenomenon in RJ-31 was reminiscent of a case (Patient-10) minent role in CAR-T’s aggressive growth in RJ-31, the influence from
reported in the anti-CD19 CAR-T therapy (referred to as the CD19 CAR- other two factors, viral activation and antigen stimulation, could not
T case, hereafter), in which, the CAR-T infused for the second time bore be neglected. Relatively, the CD19 CAR-T case was not reported to have
TET2 sequence abnormalities on both alleles, enabling clonal pro- other triggers besides TET2 abnormalities; third, the difference in the
liferation of CD8+ central memory CAR-T, and rapidly achieving tumor CRS occurring time. The next CRS struck consecutively when RJ-31
clearance accompanied by high-grade CRS44. Though the similarities in hadn’t completely recovered from the first storm, whereas the CD19
TET2 mutagenesis, CD8+ graft clonal outgrowth and severe toxicity, CAR-T case didn’t experience sequential CRS; fourth, the distinction in
the disparity of the consequent outcomes between RJ-31 and the CD19 cytokine profiles. The CD19 CAR-T case was observed having common

inflammatory molecules elevation; however, RJ-31’s cytokine signature coagulation dysfunction and liver damage depended on any of the
was considerably broadened and more toxic. Despite the objective above-mentioned AEs that met the end of the highest AE stage.
observations above, such genetic events which happened to either RJ-
31 or the CD19 CAR-T case are occasional. Screening of CHIP- CAR-T functionality assessment
associated genes, particularly TET2, DNMT3A and ASXL1, can be an The killing capacity of CAR-T was evaluated at the quality control step
optional measure in CAR-T manufacturing in some elderly populations of manufacturing. CAR-T were co-incubated for 24 h with luciferase
presenting CHIP bias, such as immune cytopenia, marrow dysplasia, labeled-RPMI-8226 cells at an E:T ratio of 20:1 (phase I) or 5:1 (phase II).
and nutritional deficiency. Controls were the untransfected T cells from the same donor. After
Additionally, the life-threatening CRS in RJ-31 evoked our aware- 24 h incubation in 37°C, 5% CO2 cell culture incubator, the cells in each
ness of proper and timely administration of immunoglobulins or well were added with 25 μl ONE-Glo Firefly luciferase assay reagent mix
specific anti-microbe agent, minimizing the possibility of pathogenic (Promega, E6120) and incubated at room temperature for 1 min.
infection at the period when immunoglobulin declines and circulating Relative light unit (RLU) of luciferase activities from the remaining
CAR-T remain in the effector status. Our work also conveyed an living target cells were read in a microplate reader (Tecan Spark 10 M)
implication for the use of a bi-specific CAR-T simultaneously targeting and were used for cell viability calculation.
BCMA antigen and a commonly affected virus, such as CMV. Such a In phase I with an E:T ratio of 20:1, the mean apoptotic cell fraction
combination targeted approach may help empower immunity and of the mild CRS group was 98% (range: 87–100%), and that of the
lower the risk of severe virus infection. Furthermore, CAR-T con- severe group was 95% (range: 83–100%), without significant difference.
stitutively carrying suicide genes are in great demand to cope with life- For the 10 investigated patients in phase II with an E:T ratio of 5:1, the
threatening side effects. average killing capacity of CAR-T was 61% and 56% in mild (one patient)
In conclusion, our study intensively evaluated the physiological and severe (9 patients) groups, respectively, and wasn’t reflective of an
process and the timing of CAR-T therapy-related CRS development, evident apoptotic disparity.
deepening our understanding of systemic toxicities in cellular immu-
notherapy. Although this study was carried out in the r/r MM patients Flow cytometry
with Cilta-cel treatment, we believe that the key findings is not Fresh PBMCs were harvested, and then incubated with antibodies for
restricted to this typical disease type and engineered product, instead, 30 min at 4 °C for staining. CAR-T immunophenotyping was detected
the conclusions can be shared by some of other CAR-T treatments. Of using fluorochrome-conjugated antibodies against human CD45 (BD
course, CAR-T-related CRS studies with a larger sample size or distinct Biosciences, Cat: 555483; Clone: HI30; Lot: 9337233), CD3 (BD Bios-
tumors are highly encouraged to consolidate our discoveries in the ciences, Cat:555335; Clone: UCHT1; Lot: 7200698), CD4 (BD Bios-
future. ciences, Cat:557871; Clone: RPA-T4; Lot: 7348612), CD8 (BD
Biosciences, Cat:564116; Clone: SK1; Lot: 8137958), CD45RA (BD Bios-
Methods ciences, Cat:563733; Clone: HI100; Lot: 8073653), CCR7 (BD Bios-
Patients and clinical assessment ciences, Cat:562555; Clone: 150503; Lot: 7355865) and FITC-labeled
The Legend-2 (phase I) and CARTIFAN-1 (phase II) studies were two human BCMA protein (ACRO Biosystems, Cat: BCA-HF254; Lot: FL894-
pivotal trials of Cilta-cel in China. This study included all participants 20BHF1-VK). All the data were collected using an LSRII flow cytometer
with available specimens in our clinical center, Ruijin Hospital affiliated (Becton Dickinson, Franklin Lakes, NJ, USA) and were analyzed with
with Shanghai Jiao Tong University School of Medicine (referred to as FlowJo software V10 (TreeStar, Ashland, OR, USA). Gating strategy can
RJ hereafter). The participating sites of the phase I trial comprised four be found in Supplementary Fig. 11.
centers with one in the West China and the other three in the East
(NCT03090659, ChiCTR-ONH-17012285). The East area (RJ, First Serum cytokine profiling
Affiliated Hospital of Nanjing Medical University and Changzheng Magnetic Luminex Assay kits were purchased from R&D Systems
Hospital) with RJ as the leading site administrated Cilta-cel treatment (catalog number: FCSTM18-45 and LXSAHM-16). 61 interleukins,
in 17 patients, of whom, 16 patients’ sequential serum samples were interferons, chemokines, TNFs and soluble receptors/ligands were
obtained while in one case the sample was unavailable. The phase II detected and collectively called "cytokines" throughout the context. In
trial was conducted in eight centers (NCT03758417). Each was in detail, they were CCL2, CCL3, CCL4, CCL5, CCL11, CCL19, CCL20,
charge of its own patient samples collection. As of manuscript pre- CX3CL1, CXCL1, CXCL2, CXCL10, IL-1α, IL-1β, IL-1ra, IL-2, IL-3, IL-4, IL-5,
paration, RJ had treated 10 phase II patients whose serum samples were IL-6, IL-7, IL-8, IL-10, IL-12 p70, IL-13, IL-15, IL-17A, IL-17E, IL-33, EGF, FGF
entirely investigated. Therefore, 26 r/r MM patients were included in basic, VEGF, PDGF-AA, PDGF-AB/BB, TGF-α, G-CSF, GM-CSF, sPD-L1,
this study without any selection bias. All the evaluated patients sCD40 Ligand, sFLT-3Ligand, IFN-α, IFN-β, IFN-γ, TNF-α, TRAIL, Gran-
underwent cyclophosphamide-based (with or without fludarabine) zyme B, sIL-2Rα, CXCL9, gp130, sIL-1R II, sIL-6Rα, sTNF RI, sTNF RII,
lymphodepletion therapy followed by Cilta-cel infusion at a median RAGE, angiopoietin-1, angiopoietin-2, IL-18, CCL22, sBCMA, sVEGF R1,
dose of 0.595×106 CAR-T cells/kg. The written informed consent to sVEGF R2, sVEGF R3. Samples cryopreserved at −80 °C were thawed
sample collection and data mining was obtained from all the patients. and processed according to the manufacturers’ protocols. Assay plates
The study was approved by the RJ ethics committee. containing patient serum were placed in the Luminex X-200 instru-
Patient clinical response was assessed by the criteria of treatment ment which provided a platform for the tests. Data processing was
response according to International Myeloma Working Group (IMWG) carried out by Milliplex analyst software based on five-parameter
consensus recommendations45. AEs were graded according to National logistic curve-fit. All data were subject to strict quality control adhering
Cancer Institute Common Terminology Criteria for Adverse Events to the following criteria: the standard curve for each analysis had a R2
(NCI-CTCAE v. 4.03). Determination of CRS was based on the criteria value > 0.99, and the coefficient of variation of standard curve and
proposed by Neelapu et al.46. Five grades were set for CRS stratifica- sample replicates were below 20%.
tion, in which, Grade 1–2 was defined as mild toxicity and Grade 3 to 5
was referred to as severe event. The term of coagulation dysfunction CONNECTOR analysis
used in this study was a collective description of the AEs relevant to Since the CC4 and CC5 exhibit the most robust correlations with CRS
platelet count, activated partial thromboplastin time or fibrogen. The severity, we adopted the computational methodology CONNECTOR to
term of liver damage used in this study was a collective description of stratify patients based on their affiliation with CC4 and CC5. Utilizing
the AEs relevant to AST, ALT, fibrogen and Bilirubin. The grading of the mean log2FC as the observation, we have performed 1,000

iterations using the ClusterAnalysis function within CONNECTOR. This T cell receptor repertoire analysis
analysis effectively partitions the 26 patients into two distinct groups: The inverse Simpson index (1/D)50 and clonality index51 were calculated
Group1 (comprising 8 patients: CZ-02, CZ-03, JS-06, RJ-01, RJ-02, RJ-04, as reported to quantitatively define the diversity of T cell clones.
RJ-22, RJ-30) and Group2 (comprising 18 patients: CZ-01, JS-02, JS-03, The amino acid sequences of CDR3 localized on TRB and TRA
JS-04, JS-05, JS-07, JS-08, JS-09, RJ-03, RJ-05, RJ-23, RJ-24, RJ-25, RJ-26, were in a perfect match with the TCR-antigen database developed by
RJ-27, RJ-29, RJ-31, RJ-32). Immuquad@ biotech, respectively. The screening method was carried
out based on the algorithm of Levenshtein distance52 ≤ 3, along with V
Nucleic acid extraction and RNA sequencing (RNA-seq) and J genes exactly matched with the ones in the database.
Total RNA and genomic DNA were isolated from PBMCs, whole blood
or liver tissue using QIAampR DNA Mini Kit (Qiagen) or RNeasy Micro Virus sequences alignment and preprocessing
Kit (Qiagen). The quality of RNA/DNA was firstly verified by Agilent We have download 13,778 virus genomes from National Center for
Bioanalyzer 2100. Biotechnology Information (NCBI, https://ftp.ncbi.nlm.nih.gov/) and
RNA-seq libraries were prepared according to protocol and built an in-house viral database. To reduce the impact of genomic
then sequenced on NovaSeq 6000 platform (Illumina). Raw FASTQ homology between human and virus, a human-virus genome was built
files were aligned to human reference genome GRCh38 (release 37). and we used two methods STAR (v2.7.9a)53 and salmon (v1.3.1)47 to
The human reference genome and its annotation file were down- calculate the virus transcripts from RNA-seq data.
loaded from GENCODE database (https://www.gencodegenes.org/).
We used salmon (v1.3.1)47 to generate the count and transcripts per Mutations calling and screening
kilobase of exon model per million mapped reads (TPM) matrix. The To identify mutations from RNA-seq data, raw reads were aligned to
transcript counts were then merged using DESeq248 and trans- human reference using STAR (v2.7.9a)53, which was downloaded from
formed as fragments per kilobase million (FPKM) to evaluate the UCSC Genome Browser (http://hgdownload.soe.ucsc.edu/). SAMtools
gene expression level by normalizing the length of genes using the (v1.8-47) was used to generate name-sorted and indexed BAM files and
TPM matrix. To ensure the robustness of our results and identify the PICARD tool (https://broadinstitute.github.io/picard/) was used to
DEGs, three well-established and recommended methods limma mark duplicate reads in BAM files. The mutation calling steps followed
(v3.44.3), edgeR (v3.30.3), DESeq2 (v1.28.1) were applied. The ana- the Genome Analysis Toolkit (GATK, v4.1.8)54 forum recommended
lysis workflow for DESeq2, limma and edgeR, which were con- best-practice pipeline. The steps were described as followed: (i) two
structed based on the linear model, followed the guidelines software, GATK HaplotypeCaller and VarScan2 (v2.4.4), were used to
provided in the respective software packages. Significant DEGs were generate mutations including single-nucleotide variations (SNVs) and
identified using adjusted p < 0.05. And significant DEGs between small insertions/deletions (INDELs); (ii) mutations were merged toge-
any two time points were included for further soft clustering ana- ther and annotated by ANNOVAR55 using specific databases such as
lysis. As demonstrated in Supplementary Fig. 3a, the majority of snp138, avsnp147, avsnp150, COSMIC, SIFT and PolyPhen. (iii) RNAmut
DEGs were consistently identified by all three methods, indicating a was used for verification mutation sites from RNA-seq data; (iv) the
high level of confidence in our DEGs selection process. Details of filtration conditions were ≥ 10x depth in the variant sites, ≥ 3% variant
analysis workflow were described in Supplementary Fig. 3a legend allele frequency (VAF) and > 3 individual mutant reads, not a SNP and
and “Code availability”. annotated by COSMIC database or predicted as damaging by
ANNOVAR.
Soft clustering method
The approach of soft clustering was implemented using the fuzzy Lentiviral vector integration site analysis
c-means algorithm by the Mfuzz R software package (v2.48.0), aiming Integration site analysis was implemented based on standard ligation
to assign the genes into different clusters in a gradual manner target amplification PCR and next-generation sequencing. Briefly,
according to their expression values49. After standardization, genes 500 ng of genomic DNA was sheared to an intermediate length of
were assigned a unique cluster by membership score (α). The gene 500 ± 50 bp using the Covaris M220 Gene Fragmentation Instrument.
with α score greater than 0.6 was considered as a core gene and was After purification, the sheared DNA was extended by biotinylated 3′-
used for subsequent functional enrichment analysis. Details of analysis long terminal repeat (3′-LTR)-specific primer (Primer 1) and captured
workflow were described in Supplementary Fig. 3b legend and “Code by streptavidin-coated beads (DynabeadsTMM-280; invitrogen). The
availability”. captured genome was seeded to a ligation cassette including a
sequencing barcode (UMI) according to the kit (Fast-Link Ligation Kit;
Gene ontology (GO) analysis and gene set enrichment Epicentre). Amplification of integration site was conducted via a pro-
analysis (GSEA) cedure of two nested PCR. The first exponential PCR primers were
The functional enrichment analysis of GO was performed with referred to as Primer 2 and Primer 3. After magnetic capturing, the
clusterprofiler R package (v3.16.1). And GSEA was conducted product underwent the second exponential PCR using Primer 4 (PE 2.0
based on molecular signatures database v7.4 and GSEABase R UMI short). Primer sequences are summarized in the
package (v1.50.1). Gene sets used in GSEA were downloaded from SupplemetaryTable 1.
the Molecular Signatures Database (MSigDB, v7.4) of the Broad The purified product proceeded to the subsequent Illumina Miseq
Institute. next-generation sequencing platform for paired-end reads with para-
The genes used for GSEA were selected from the FPKM profile. meter settings referenced in the literature56,57. Raw sequencing data
Genes with low expression (expressed in less than 30% of the samples) were analyzed with Genome Integration Site Analysis Pipeline58 based
were excluded. During the GSEA analysis at various observation time on human reference genome GRCh38.
points, genes were pre-ranked based on their log2FC, which was cal-
culated by comparing their expression levels with the baseline. Path- Single-cell RNA sequencing (scRNA-seq)
way direction is the median log2FC of significant transcripts (core CAR-positive cell populations from RJ-31’s PBMCs on day 11 and day 28
enrichment genes) relative to the baseline. In this visualization, an were obtained by flow cytometry cell sorting (FACSAriaTM III BD Bios-
adjusted p < 0.05 with pathway direction > 0 indicated activation, ciences) with FITC-labeled human BCMA protein (ACRO Biosystems).
while an adjusted p < 0.05 with pathway direction <0 indicated The target cells underwent single cell isolation to form Gel Bead in
repression. Emulsion by 10x Genomics ChrominumTM. And then scRNA-seq

libraries were performed according to the Chromium Single Cell 5′ Preparation Kit (for Illumina) and then sequenced on NovaSeq 6000
Library Construction Kit and Chromium Single Cell Human TCR platform (Illumina). Variant calling from targeted DNA sequencing was
Amplification Kit (10x Genomics). After quality verification, the performed as reported previously63.
libraries were sequenced on NovaSeq 6000 platform (Illumina).
The raw sequencing reads in FASTQ format of single-cell RNA Droplet digital PCR (ddPCR)
sequencing (scRNA-seq) data were aligned to the human GRCh38 To identify the low-frequency mutation in TET2, ddPCR was performed
reference (2020-A version) and the gene expression matrices were in RJ-31’s genome DNA at different time points. To make the assay
generated using CellRanger (10X Genomics, default settings, Ver- more convinced, water and three healthy controls were included as
sion 6.0.2). Both CellRanger software and the reference were blank and negative controls, respectively. ddPCR was accomplished on
downloaded from 10X Genomics website (https://www. a Bio-Rad QX200 system using ddPCR supermix for Probe (No dUTP)
10xgenomics.com/). And for V(D)J T-cell analysis, the single-cell T (Bio-Rad) and analyzed on QuantaSoft (Version 1.7.4). The primers and
cell receptor (TCR) sequencing (scTCR-seq) data were aligned to probe sequences targeting TET2 were forward, 5′-GTCCAAG-
human GRCh38 (ensembl-5.0.0) using the vdj module in CellRanger. GAGGCTTACACAAAT-3′; reverse, 5′-ACCTCATCGTTGTCCTCTGC-3′;
The gene expression matrices generated by CellRanger were then wild probe, 5′VIC-CACACCCTGGACTAG-3′MGB; mutate probe 5′FAM-
imported and merged using the Seurat R package59 for downstream CACACCCTAGACTAGT-3′MGB. To ensure the quality of the experi-
data analysis. Cells that expressed less than 800 genes or over 10% ment, three replicates were completed for each sample, and each
mitochondrial RNA were filtered out. Cells which matched with TCR reaction was guaranteed to have a droplet count of more than
Concentrate of mutate
were identified as T cells and for further study. Top 3,000 highly 10,000. VAF = Concentrate of mutate and wild
variable genes were identified to perform principal component
analysis (PCA) and ComBat method in R package SVA60 was used to Statistical analysis
correct batch effects. UMAP (uniform manifold approximation and Data analyses were performed using R version 4.0.3 or software
projection) was chosen to further cell visualization. Marker genes of GraphPad Prism version 8.0.2. Statistical tests were described in the
each cluster were calculated using ‘FindAllMarkers’ function in figure legends.
Seurat. Cell types were identified and annotated by the following
criteria: (i) annotation generated in SingleR61 was used as reference, Reporting summary
(ii) manually annotated using the top expressed genes in each Further information on research design is available in the Nature
cluster, (iii) clone type identified by scTCR-seq. Portfolio Reporting Summary linked to this article.
To compare the toxicity capacity of cells, we calculated the
average expression level of toxicity-related genes for each cell to Data availability
quantify cytotoxicity, called cytotoxic score. Based on our own scRNA- Raw sequencing data generated during this study have been deposited
seq data, we computed the Pearson correlation coefficients of all genes in the Genome Sequence Archive in National Genomics Data Center,
with GZMB and GNLY separately. And then intersection of the top 20 China National Center for Bioinformation/Beijing Institute of Geno-
ranked genes significantly associated with GZMB and GNLY were mics, Chinese Academy of Science (https://ngdc.cncb.ac.cn/gsa-
defined as Gzmb_Gnly_related gene set. Meanwhile, we extracted the human), with accession number ‘GSA-Human: HRA005381’ [https://
reported toxicity-related gene set62 as a supplement. ngdc.cncb.ac.cn/gsa-human/browse/HRA005381]. These data are
under controlled use conditions set by human privacy regulations,
Quantitative real-time polymerase chain reaction (qPCR) only available upon reasonable request. Access can be obtained by
To measure the copy number of the transgene in the Cilta-cel, qPCR of approval via the Data Access Committee of the GSA-human database
targeted fragment was subsequently performed as reported (https://ngdc.cncb.ac.cn/gsa-human/document/GSA-Human_Request_
previously3. Guide_for_Users_us.pdf). Original data for graphs is provided in the
And the transcriptional levels of the gene encoding herpesvirus7 Source Data file. Source data are provided with this paper.
(HHV-7) and ETS1 were measured by RT-qPCR. 500 ng of RNA was
reverse-transcribed into cDNA using HiScript III RT SuperMix kit Code availability
(Vazyme), and ChamQ SYBR qPCR Master Mix kit (Vazyme) was Code used in the study is available on GituHub (https://nrctm-bioinfo.
applied for qPCR of the genetic segment of interest. The procedure github.io/MM_CART_project/index.html) and has been assigned a
was run and analyzed on ViiATM 7 system (Life technologies). Gene (https://doi.org/10.5281/zenodo.10200875).
transcriptional level was assessed using cycle threshold and normal-
ized to a housekeeping gene GAPDH. The sequences of the primers References
were as follows: GAPDH forward 5′-GTCTCCTCTGACTTCAACAGCG-3′ 1. Maude, S. L. et al. Tisagenlecleucel in children and young adults
and reverse 5′-ACCACCCTGTTGCTGTAGCCA-3′, HHV-7 encoded gene with B-cell lymphoblastic leukemia. N. Engl. J. Med. 378,
forward 5′-AGAGGTTGCTTCTTCGAGTCA-3′ and reverse 5′- 439–448 (2018).
GCGCTTGTCAAAATCCTCAAAT-3′, ETS1 encoded gene forward 5′- 2. Neelapu, S. S. et al. Axicabtagene ciloleucel CAR T-cell therapy in
GGAATGTGCAGATGTCCCAC-3′ and reverse 5′-CATTCACAGCCCA- refractory large B-cell lymphoma. N. Engl. J. Med. 377,
CATCACC-3′. 2531–2544 (2017).
3. Xu, J. et al. Exploratory trial of a biepitopic CAR T-targeting B cell
Sanger sequencing maturation antigen in relapsed/refractory multiple myeloma. Proc.
The PCR was performed using Phanta Max Super-Fidelity DNA Poly- Natl Acad. Sci. USA 116, 9543–9955 (2019).
merase kit (Vazyme, P505-d1). The sequences of the primers of TET2 4. Teachey, D. T. et al. Identification of predictive biomarkers for
were forward 5′-GTCTCAGCCGATGGATCTGTA-3′ and reverse 5′- cytokine release syndrome after chimeric antigen receptor T-cell
ATCTGTTGTAAGGCCCTGTGA-3′. therapy for acute lymphoblastic leukemia. Cancer Discov. 6,
664–679 (2016).
Targeted DNA sequencing 5. Norelli, M. et al. Monocyte-derived IL-1 and IL-6 are differentially
Targeted DNA sequencing was performed on the gene coding required for cytokine-release syndrome and neurotoxicity due to
sequence. Library enrichment was carried using NadPrep DNA CAR T cells. Nat. Med. 24, 739–748 (2018).

6. Giavridis, T. et al. CAR T cell-induced cytokine release syndrome is 27. Turtle, C. J. et al. CD19 CAR-T cells of defined CD4+:CD8+ compo-
mediated by macrophages and abated by IL-1 blockade. Nat. Med. sition in adult B cell ALL patients. J. Clin. Investig. 126,
24, 731–738 (2018). 2123–2138 (2016).
7. Sterner, R. M. et al. GM-CSF inhibition reduces cytokine release 28. Busque, L. et al. Recurrent somatic TET2 mutations in normal elderly
syndrome and neuroinflammation but enhances CAR-T cell func- individuals with clonal hematopoiesis. Nat. Genet. 44,
tion in xenografts. Blood 133, 697–709 (2019). 1179–1181 (2012).
8. Zhao, W. H. et al. Four-year follow-up of LCAR-B38M in relapsed or 29. Jaiswal, S. et al. Age-related clonal hematopoiesis associated with
refractory multiple myeloma: a phase 1, single-arm, open-label, adverse outcomes. N. Engl. J. Med. 371, 2488–2498 (2014).
multicenter study in China (LEGEND-2). J. Hematol. Oncol. 15, 30. Cong, B., Zhang, Q. & Cao, X. The function and regulation of TET2 in
86 (2022). innate immunity and inflammation. Protein cell 12, 165–173 (2021).
9. Mi, J. Q., et al. Phase II, open-label study of ciltacabtagene auto- 31. Siraki, A. G. The many roles of myeloperoxidase: from inflammation
leucel, an anti-B-cell maturation antigen chimeric antigen receptor- and immunity to biomarkers, drug metabolism and drug discovery.
T-cell therapy, in Chinese patients with relapsed/refractory multiple Redox Biol. 46, 102109 (2021).
myeloma (CARTIFAN-1). J. Clin. Oncol. 41,Jco2200690 (2022). 32. Barnes, B. J. et al. Targeting potential drivers of COVID-19: Neu-
10. Berdeja, J. G. et al. Ciltacabtagene autoleucel, a B-cell maturation trophil extracellular traps. J. Exp. Med. 217, e20200652 (2020).
antigen-directed chimeric antigen receptor T-cell therapy in 33. Middleton, E. A. et al. Neutrophil extracellular traps contribute to
patients with relapsed or refractory multiple myeloma (CARTITUDE- immunothrombosis in COVID-19 acute respiratory distress syn-
1): a phase 1b/2 open-label study. Lancet 398, 314–324 (2021). drome. Blood 136, 1169–1179 (2020).
11. Ri, M. et al. Ciltacabtagene autoleucel in patients with relapsed/ 34. Socié, G. et al. Prognostic value of apoptotic cells and infiltrating
refractory multiple myeloma: CARTITUDE-1 (phase 2) Japanese neutrophils in graft-versus-host disease of the gastrointestinal tract
cohort. Cancer Sci. 113, 4267–4276 (2022). in humans: TNF and Fas expression. Blood 103, 50–57 (2004).
12. Maecker, H. T., McCoy, J. P. & Nussenblatt, R. Standardizing 35. Wang, D. et al. A phase 1 study of a novel fully human BCMA-
immunophenotyping for the Human Immunology Project. Nat. Rev. targeting CAR (CT103A) in patients with relapsed/refractory multi-
Immunol. 12, 191–200 (2012). ple myeloma. Blood 137, 2890–2901 (2021).
13. Hay, K. A. et al. Kinetics and biomarkers of severe cytokine release 36. Bernabei, L. et al. B-cell maturation antigen chimeric antigen
syndrome after CD19 chimeric antigen receptor-modified T-cell receptor T-cell re-expansion in a patient with myeloma following
therapy. Blood 130, 2295–2306 (2017). salvage programmed cell death protein 1 inhibitor-based combi-
14. Pernice, S. et al. CONNECTOR, fitting and clustering of longitudinal nation therapy. Br. J. Haematol. 193, 851–855 (2021).
data to reveal a new risk stratification system. Bioinformatics 39, 37. Sun, Z., Xie, C., Liu, H. & Yuan, X. CD19 CAR-T cell therapy induced
btad201 (2023). immunotherapy associated interstitial pneumonitis: a case report.
15. Casulleras, M., Zhang, I. W., López-Vicario, C. & Clària, J. Leuko- Front. Immunol. 13, 778192 (2022).
cytes, systemic inflammation and immunopathology in acute-on- 38. Zhao, W. H. et al. A phase 1, open-label study of LCAR-B38M, a
chronic liver failure. Cells 9, 2632 (2020). chimeric antigen receptor T cell therapy directed against B cell
16. Tillie, R. et al. A switch from cell-associated to soluble PDGF-B maturation antigen, in patients with relapsed or refractory multiple
protects against atherosclerosis, despite driving extramedullary myeloma. J. Hematol. Oncol. 11, 141 (2018).
hematopoiesis. Cells 10, 1746 (2021). 39. Tanyi, J. L. et al. Possible compartmental cytokine release syndrome
17. Thomas, M. & Augustin, H. G. The role of the Angiopoietins in vas- in a patient with recurrent ovarian cancer after treatment with
cular morphogenesis. Angiogenesis 12, 125–137 (2009). mesothelin-targeted CAR-T cells. J. Immunother. 40,
18. Boisramé-Helms, J., Kremer, H., Schini-Kerth, V. & Meziani, F. 104–107 (2017).
Endothelial dysfunction in sepsis. Curr. Vasc. Pharmacol. 11, 40. Handley, G. et al. Human herpesvirus 6 myelitis after chimeric
150–160 (2013). antigen receptor T-cell therapy. Int. J. Infect. Dis. 112,
19. Gavriilaki, E., Sakellari, I., Gavriilaki, M. & Anagnostopoulos, A. A new 327–329 (2021).
era in endothelial injury syndromes: toxicity of CAR-T cells and the 41. Wang, D. et al. Viral infection/reactivation during long-term follow-
role of immunity. Int. J. Mol. Sci. 21, 3886 (2020). up in multiple myeloma patients with anti-BCMA CAR therapy.
20. Hay, K. A. et al. Factors associated with durable EFS in adult B-cell Blood Cancer J. 11, 168 (2021).
ALL patients achieving MRD-negative CR after CD19 CAR T-cell 42. Jaiswal, S. Clonal hematopoiesis and nonhematologic disorders.
therapy. Blood 133, 1652–1663 (2019). Blood 136, 1606–1614 (2020).
21. Page, A. V. & Liles, W. C. Biomarkers of endothelial activation/dys- 43. Arends, C. M. et al. Hematopoietic lineage distribution and evolu-
function in infectious diseases. Virulence 4, 507–516 (2013). tionary dynamics of clonal hematopoiesis. Leukemia 32,
22. Pan, J. et al. Ruxolitinib mitigates steroid-refractory CRS during CAR 1908–1919 (2018).
T therapy. J. Cell. Mol. Med. 25, 1089–1099 (2021). 44. Fraietta, J. A. et al. Disruption of TET2 promotes the therapeutic
23. Wei, S. et al. Adjuvant ruxolitinib therapy relieves steroid-refractory efficacy of CD19-targeted T cells. Nature 558, 307–312 (2018).
cytokine-release syndrome without impairing chimeric antigen 45. Kumar, S. et al. International Myeloma Working Group consensus
receptor-modified T-cell function. Immunotherapy 12, criteria for response and minimal residual disease assessment in
1047–1052 (2020). multiple myeloma. Lancet Oncol. 17, e328–e346 (2016).
24. Chen, Y. et al. Therapeutic potential of TNFα and IL1β blockade for 46. Neelapu, S. S. et al. Chimeric antigen receptor T-cell therapy -
CRS/ICANS in CAR-T therapy via ameliorating endothelial activa- assessment and management of toxicities. Nat. Rev. Clin. Oncol. 15,
tion. Front. Immunol. 12, 623610 (2021). 47–62 (2018).
25. Fajgenbaum, D. C. & June, C. H. Cytokine storm. N. Engl. J. Med. 47. Patro, R., Duggal, G., Love, M. I., Irizarry, R. A. & Kingsford, C. Sal-
383, 2255–2273 (2020). mon provides fast and bias-aware quantification of transcript
26. Wang, X. et al. Quantitative characterization of T-cell repertoire expression. Nat. Methods 14, 417–419 (2017).
alteration in Chinese patients with B-cell acute lymphocyte leuke- 48. Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold
mia after CAR-T therapy. Bone Marrow Transplant. 54, change and dispersion for RNA-seq data with DESeq2. Genome Biol.
2072–2080 (2019). 15, 550 (2014).

49. Kumar, L. & M, E. F. Mfuzz: a software package for soft clustering of 82230006, 81970189, 82070227); the Shanghai Clinical Research Cen-
microarray data. Bioinformation 2, 5–7 (2007). ter for Hematological disease (19MC1910700); the Shanghai Shenkang
50. Yew, P. Y. et al. Quantitative characterization of T-cell repertoire in Hospital Development Center (SHDC2020CR5002,
allogeneic hematopoietic stem cell transplant recipients. Bone SHDC2020CR2066B); the Shanghai Major Project for Clinical Medicine
Marrow Transplant. 50, 1227–1234 (2015). (2017ZZ01002); the Innovative Research Team of High-level Local Uni-
51. Zhang, L. et al. 3D: diversity, dynamics, differential testing - a pro- versities in Shanghai; the Shanghai Pujiang Program (No. 22PJ1409800).
posed pipeline for analysis of next-generation sequencing T cell
repertoire data. BMC Bioinform. 18, 129 (2017). Author contributions
52. Young, B., Faris, T. & Armogida, L. Levenshtein distance as a mea- S.-J.C., Z.C., J.-Q.M. and S.Y. designed the project. J.-Q.M., J.X., S.J., L.C.,
sure of accuracy and precision in forensic PCR-MPS methods. For- J.H. performed the clinical trial. S.Y., C.L., X.M., Y.S. performed the
ensic Sci. Int. Genet. 55, 102594 (2021). experiments. S.Y., Y.D., J.L. provided the bioinformatics analysis. S.-J.C.,
53. Dobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Bioinfor- Z.C., J.-Q.M., S.Y., J.X. supervised the study and wrote the manuscript.
matics 29, 15–21 (2013). All authors have read and approved the final version of this manuscript.
54. McKenna, A. et al. The Genome Analysis Toolkit: a MapReduce
framework for analyzing next-generation DNA sequencing data. Competing interests
Genome Res 20, 1297–1303 (2010). The authors declare no competing interests.
55. Wang, K., Li, M. & Hakonarson, H. ANNOVAR: functional annotation
of genetic variants from high-throughput sequencing data. Nucleic Additional information
Acids Res. 38, e164 (2010). Supplementary information The online version contains
56. Gabriel, R. et al. Comprehensive genomic access to vector inte- supplementary material available at
gration in clinical gene therapy. Nat. Med. 15, 1431–1436 (2009). https://doi.org/10.1038/s41467-023-44648-3.
57. Paruzynski, A. et al. Genome-wide high-throughput integrome
analyses by nrLAM-PCR and next-generation sequencing. Nat. Correspondence and requests for materials should be addressed to
Protoc. 5, 1379–1395 (2010). Jian-Qing Mi or Sai-Juan Chen.
58. Afzal, S., Wilkening, S., von Kalle, C., Schmidt, M. & Fronza, R. GENE-
IS: time-efficient and accurate analysis of viral integration events in Peer review information Nature Communications thanks Francesca
large-scale gene therapy data. Mol. Ther. Nucleic Acids 6, Cordero, Nicolaus Kroeger and the other, anonymous, reviewer(s) for
133–139 (2017). their contribution to the peer review of this work. A peer review file is
59. Butler, A., Hoffman, P., Smibert, P., Papalexi, E. & Satija, R. Inte- available.
grating single-cell transcriptomic data across different conditions,
technologies, and species. Nat. Biotechnol. 36, 411–420 (2018). Reprints and permissions information is available at
60. Leek, J. T., Johnson, W. E., Parker, H. S., Jaffe, A. E. & Storey, J. D. The http://www.nature.com/reprints
sva package for removing batch effects and other unwanted var-
iation in high-throughput experiments. Bioinformatics 28, Publisher’s note Springer Nature remains neutral with regard to jur-
882–883 (2012). isdictional claims in published maps and institutional affiliations.
61. Aran, D. et al. Reference-based analysis of lung single-cell
sequencing reveals a transitional profibrotic macrophage. Nat. Open Access This article is licensed under a Creative Commons
Immunol. 20, 163–172 (2019). Attribution 4.0 International License, which permits use, sharing,
62. Tirosh, I. et al. Dissecting the multicellular ecosystem of metastatic adaptation, distribution and reproduction in any medium or format, as
melanoma by single-cell RNA-seq. Science 352, 189–196 (2016). long as you give appropriate credit to the original author(s) and the
63. Cheng, W. Y. et al. Transcriptome-based molecular subtypes and source, provide a link to the Creative Commons license, and indicate if
differentiation hierarchies improve the classification framework of changes were made. The images or other third party material in this
acute myeloid leukemia. Proc. Natl Acad. Sci. USA 119, article are included in the article’s Creative Commons license, unless
e2211429119 (2022). indicated otherwise in a credit line to the material. If material is not
included in the article’s Creative Commons license and your intended
Acknowledgements use is not permitted by statutory regulation or exceeds the permitted
This work was supported by the State Key Laboratory of Medical use, you will need to obtain permission directly from the copyright
Genomics, the Double First-Class Project (WF510162602) from the holder. To view a copy of this license, visit http://creativecommons.org/
Ministry of Education; the Shanghai Collaborative Innovation Program licenses/by/4.0/.
on Regenerative Medicine and Stem Cell Research (2019CXJQ01); the
Overseas Expertise Introduction Project for Discipline Innovation (111 © The Author(s) 2024
Project;B17029); the National Natural Science Foundation of China (Nos.

Multiple Myeloma: The Next Frontier

Uploaded by

Copyright:

Available Formats

You might also like

Multiple Myeloma: The Next Frontier

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Multiple Myeloma: The Next Frontier

Uploaded by

Copyright:

Available Formats

Article https://doi.org/10.

Neutrophil activation and clonal CAR-T

Nature Communications | (2024)15:360 1

Nature Communications | (2024)15:360 2

Nature Communications | (2024)15:360 3

R=0.56, p=8.8E−11 R=0.56, p=9.5E−11 R=0.52, p=2.8E−09

Mean CRS grade Mean CRS grade Mean CRS grade

Nature Communications | (2024)15:360 4

Patient group 2 Patient group

Days after infusion

Nature Communications | (2024)15:360 5

Nature Communications | (2024)15:360 6

Nature Communications | (2024)15:360 7

Nature Communications | (2024)15:360 8

Inverse simpson index

TRB TRB 0 V110 5 TRB

Days after infusion Days after infusion

0.2 0.4 0.6

Nature Communications | (2024)15:360 9

Nature Communications | (2024)15:360 10

Common CRS (n=26) Rare CRS in RJ-31 Undetectable

Degranulation sTNF RII

Nature Communications | (2024)15:360 11

Nature Communications | (2024)15:360 12

Nature Communications | (2024)15:360 13

Nature Communications | (2024)15:360 14

Nature Communications | (2024)15:360 15

Nature Communications | (2024)15:360 16

You might also like