Immunotherapy for Cancer Kara L. Davis, D.O. Anne T. and Robert M. Bass Endowed Faculty Scholar in Pediatric Cancer and Blood Diseases Assistant Professor of Pediatrics Bass Center for Childhood Cancer and Blood Disorders Stanford University Objectives

• Childhood leukemia leads the way for targeted T- cell therapies • What are chimeric antigen T cells • Use in heme malignancies and solid tumors • Limitations in Efficacy • Toxicity concerns • Immune Checkpoint Inhibition for anti-cancer effect • Successes and Challenges Childhood Leukemia was universally fatal 100 years ago

Gittins, Arch. Dis Child, 1933 Now childhood leukemia is a poster child for success in cancer treatment

Pui et al, NEJM 2006 Survival after relapse from childhood ALL K Nguyen et al 2147 Table 4 Univariate analyses of risk factors associated with survival after relapse

Univariate analyses of survival post 5-year survival rates P-value relapse risk factors (N 1961) post relapse±s.e. (%) ¼ Time to relapse Early 21.0±1.8 o0.0001 Intermediate 37.9±2.7 Late 53.1±3.9

Relapse site Isolated marrow 24.1±2.1 o0.0001 Concurrent marrow 39.4±5.0 Isolated CNS 58.7±3.2 Isolated testicular 58.0±8.2 Other extramed±CNS 55.9±9.9

Figure 2 Kaplan–Meier estimates of survival after relapse for patients Age group stratified by NCI risk group at diagnosis and timing of relapse. There o1 year 19.8±4.4 o0.0001 were significant differences in outcome for those SR vs HR patients 1–9 years 45.0±2.1 experiencing either an early (33.1±3.6 vs 14.9±2.1%, Po0.0001), 10+ years 18.2±2.8 intermediate (52.2±3.7 vs 22.0±3.9%, Po0.0001) or late relapse (59.6±4.6 vs 39.5±7.2%, Po0.0001). Trial era But relapsed leukemia is still a bigEarly problem 36.2±1.9 0.66 Late 36.6±3.1

Sex Male 35.4±2.2 0.51 • OverFemale time, from 38.0±2.5 Lineage 1988-2002B studies 37.2±2.1 o0.0001 addressingT relapsed23.0 ±ALL4.0 CNS status at diagnosis haveCNS-3 not improved 14.5±4.5 o0.0001 CNS-2 and CNS-1 37.7±1.7 outcomes WBC o50 k 41.6±2.0 o0.0001 X50 k 24.3±2.6

Race White 38.5±2.0 0.02 Hispanic 30.5±3.8 Figure 3 Kaplan–Meier estimates of survival after relapse for patients Black 30.1±5.9 stratified by treatment era and timing of relapse. There was no Other Nguyen et al, Leukemia,38.1±6.9 2008 difference in survival between early and late trials among patients who ± ± experienced an early relapse (19.3 2.1 vs 23.4 3.4%, P 0.10), or NCI risk intermediate relapse (39.3±3.3 vs 36.0±4.8%, P 0.49)¼ or late Standard 50.4±2.4 o0.0001 relapses (53.3±4.3 vs 54.4±9.2%, P 0.66). ¼ ¼ High 22.6±2.1

CCG-1952, CCG-1953, CCG-1961 and CCG-1962. A total of Abbreviations: CNS, central nervous system; NCI, National Cancer 1961 children were included in this analysis (Figure 3). The Institute; WBC, white blood cells. results demonstrated that patients who relapsed early had dismal survival regardless of treatment era (5-year post-relapse survival: 19.3±2.1 vs 23.4±3.4%, P 0.10). Similarly, there were no diagnosis were excluded from the multivariate analyses, as they ¼ statistically significant differences in survival between those cannot be classified by NCI risk group. Age at diagnosis, CNS treated on early and late trials, for patients experiencing either disease, sex, lineage and NCI risk group were significant an intermediate (39.3±3.3 vs 36.0±4.8%, P 0.49) or late predictors of survival post relapse. Trial era, WBC at diagnosis ¼ relapse (53.3±4.3 vs 54.4±9.2%, P 0.66). and race were not significant in predicting survival in the ¼ Univariate analyses (n 1961) were conducted to study the multivariate analysis. Multivariate analyses stratified for time ¼ association of presenting clinical and laboratory features at to relapse, the site of relapse and age at diagnosis resulted in initial diagnosis, time to relapse and the site of relapse with the same significant predictors as before. Trial era, WBC at survival post relapse (Table 4). Significant associations were diagnosis and race were not significant factors. found with site and timing of relapse, age at diagnosis, WBC at diagnosis, lineage (B vs T), CNS status at diagnosis, race and NCI risk; exceptions were trial era and sex. Multivariate analyses Discussion stratified by timing of relapse and the site of relapse were conducted on the subset of patients (n 1391) with complete Despite improved treatments and increasing dose intensity in ¼ data for all the variables (Table 5). Infants o1 year of age at primary therapy for newly diagnosed patients with ALL, patients

Leukemia The search for surface targets in B cell cancers

Piccaluga et al., Leuk & Lymphoma 2010 What is a CAR-T cell?

VH V V VL scFv L H α β

ε δ ϒ ε

+ TM

. CD28/41BB ζ ζ

costim CD3ζ

Monoclonal antibody T-cell Receptor Chimeric Antigen Receptor How CAR-T cells work…

T cell Tumor cell

CAR enables T cell to Expression of recognize tumor cell CAR antigen Viral DNA Insertion

Antigen

Tumor cell CAR T cells apoptosis multiply and release cytokines

Courtesy of David Miklos Timeline of Development CAR T cells: CAR design impacts function Therapeutic Advances in Hematology 6(5)

First generation CAR Second generation CAR Third generation CAR

ScFv antigen-binding domain VH VH VH V V V L L L hinge linker Costim 1 C Costim 1 C CD3

‘first signal’ ‘first signal’ ‘first signal’ CD28 CD137 (4-1BB) Costim 2 C CD3 CD27 CD28 ‘second signal’ ‘second signal’ CD134 (OX40) CD137 (4-1BB) CD3 ‘third signal’ Tasian et al., Ther Adv Hema, 2015

Figure 1. Generations of chimeric antigen receptors (CARs) utilized in clinical testing. Constructs encoding synthetic CARs targeting tumor-associated antigens (such as CD19) can be stably transduced into human T cells for infusion into patients with relapsed/refractory cancer. CARs are comprised of (1) an extracellular MHC- independent antigen-binding domain usually derived from a monoclonal antibody single chain variable fragment (ScFv), (2) an extracellular spacer domain or ‘hinge’ (in some CARs), (3) a transmembrane linking domain and (4) an intracellular co-stimulatory T cell signaling domain or multiple domains.

With increased understanding of the critical Given the potential for clinically significant CAR importance of the intracellular ‘second signal’ T-cell induced toxicities (discussed below), opti - activation for CAR T-cell efficacy, subsequent mal CAR design thus must carefully balance generations of CARs have optimized ScFv and desired antitumor potency with minimization of linker component design and, more importantly, hazardous side effects [Gardner and Jensen, 2014]. have incorporated additional intracellular costim- ulatory signaling domains [e.g. CD27, CD28, Various groups have focused intensively in the past CD134 (OX40), CD137 (4-1BB)] in efforts to decade on targeting the B-lymphocyte antigen, increase the expansion, persistence and potency CD19, a phosphoglycoprotein ubiquitously of CAR T cells, as well as to prevent cellular expressed on malignant and nonmalignant B-cells. exhaustion in vivo (Figure 2). Based upon the clinical efficacy and apparent toler- ability of targeting CD20 (another commonly Most CARs used in current clinical trials of engi- expressed B cell antigen) with the anti-CD20 mon- neered T cells for patients with B-ALL are derived oclonal antibody rituximab in patients with B-cell from second generation constructs with CD3ζ hematologic malignancies [Cramer and Hallek, and another signaling endodomain. Delivery of a 2012], it was hypothesized that the CD19 receptor second costimulatory signal has indeed appeared could be similarly targeted with engineered T cells to induce significantly greater T-cell expansion expressing a CD19-redirected CAR and that treat- and longer-term persistence in vivo to date in ment with these CD19 CAR T cells would also be treated patients. ultimately tolerable in patients [Cooper et al. 2004; Kochenderfer and Rosenberg, 2013]. Third generation CAR constructs comprised of CD3ζ and two additional co-stimulatory endodo- In preclinical studies, several research teams mains transduced into T cells are also under clini- observed that co-incubation of CD19 CAR T cells cal evaluation, but have not thus far resulted in with B-cell leukemia or lymphoma cell lines greater efficacy than second generation CARs induced potent T-cell degranulation, cytokine [Davila et al. 2012; Sadelain et al. 2013; Kenderian production and tumor cytotoxicity in vitro. et al. 2014; Mackall et al. 2014]. Furthermore, treatment of human ALL cell

230 http://tah.sagepub.com Design affects Function

CD19-CD28-Zeta CD19-41BB-Zeta

V V H Anti-CD19 H Anti-CD19 VL VL

NCI TM vs. TM UPENN KITE NOVARTIS MSKCC CD28 4-1BB SEATTLE JUNO (JCAR015) JUNO (JCAR017) CD3ζ CD3ζ

Persist less than 3 months Persist 6 months or more in most patients in most patients Increased Persistence of CD19.BB.z vs CD19.28.z CAR in Clinical TrialsCD19-CAR

100000 CD19.28.z-Bethesda CD19.BB.z-Penn 10000 1000 100 10

1 0 3 6 9 12 Days Since Infusion 0.1 transgene copies/100 ng of DNA 0 7 14 21 28 35 42 49 56 100 200 300 Lee, Lancet, 2014 Days Since Infusion Months Since Infusion Maude, NEJM, 2014 Efficacy of CAR-T cells Current strategies for treatment after relapse are suboptimal Clo/CTX/VP-16 for Advanced Childhood ALL

Clofarabine, Etoposide, CyclophosphamideBlinatumomab

Locatelli et al, Br J Haematol, 2009 Fig 3. Overall survival from start of treatment by response to Von Stackelberg et al., Blood, 2016 Fig 1. Overall survival from start of treatment. treatment.

donor. This combination of agents might also be considered for patients who had previously received allogeneic HSCT. The rationale for combining these three cytotoxic agents lay with the observation that cyclophosphamide induces DNA interstrand cross-links. These cross-links can be rapidly repaired by leukaemia cells, as demonstrated in chronic lymphocytic leukaemia (CLL) lymphocytes after in vitro exposure to activated cyclophosphamide (Yamauchi et al, 2001). Pre-treatment with clofarabine, leading to inhibition of both ribonucleotide reductase and DNA synthesis and repair, may significantly prevent or even impede completion of DNA strand break repair, resulting in an increase in apoptosis of leukaemia blasts. Support to this hypothesis is provided by a recent report on the combined use of clofarabine and Fig 2. Overall survival from start of treatment by immunophenotype. cyclophosphamide in adults with acute leukaemia, showing that clofarabine followed by cyclophosphamide was associated with increased DNA damage and apoptosis in both myeloid survival for patients who did or did not achieve a CR or CRp and lymphoid blasts obtained both from peripheral blood and was 39% (95%CI, 5–73%) and 0%, respectively (P <0Æ001) BM (Karp et al, 2007). Etoposide binds to topoisomerase II (Fig 3). Among the 14 patients who responded to treatment, and inhibits its function in ligating cleaved DNA molecules leukaemia recurrence was observed in six (43%) patients. The resulting in the accumulation of single- or double-strand DNA median duration of remission for the 14 patients who achieved breaks. This hampers DNA replication and transcription, and either CR or CRp, censoring patients at time of transplanta- promotes apoptotic cell death. tion, was 6 months (range, 3–8Æ5). The median survival The drug dosages chosen for this study were based on the duration of the 11 patients who did not respond to treatment preliminary data presented by Hijiya et al (2007) with some or had PR was 3 months (range, 1–7). modifications, namely a decrease in cyclophosphamide dosage and an increase in etoposide dosage. This scheme proved to be safe and tolerable, as none of the 25 patients enrolled in our Discussion study died due to treatment-related complications. The only This multicentre study showed that, for children with multiple relevant toxicity that we observed was that related to relapsed or resistant ALL, a single combination course of liver dysfunction, although this was never fatal and always clofarabine, cyclophosphamide and etoposide was associated reversible. Hepatic toxicity has been frequently reported when with acceptable organ toxicity and showed promising activity clofarabine was used as single agent, as well as in combination with an OR rate >50%. In addition, this regimen was with other drugs, including cyclophosphamide and cytarabine particularly effective in BCP ALL and induced remission of (Kantarjian et al, 2003; Jeha et al, 2004; Faderl et al, 2005; Jeha sufficient duration to allow many patients with a suitable et al, 2006; Karp et al, 2007; Faderl et al, 2008). In a phase II donor to proceed to HSCT, even from an HLA-disparate extension of the initial escalating dose, phase I study reported

ª 2009 Blackwell Publishing Ltd, British Journal of Haematology, 147, 371–378 375 Articles

A B 150 CR-MRD negative 200 p=0·0039 CNS2 CR CNS1 100 SD 150 PD 50

. . No CRS 100 ...... * . . 0 ...... Grade 1 or 2 CRS

..... (thousands) .... Grade 3 CRS 50 –50 .... Grade 4 CRS The new england journal of medicine Absolute CAR T-cells in CSF in T-cells CAR Absolute ...... Change in marrow blasts (%) ...... –100 ...... Across0 centers, excellent outcomes with CAR Individual ALL patients (n=20) No neurotoxicity Neurotoxicity per microgram of genomic DNA. One patient A C (Patient 17) received infusions again at 3 months D 1.0 10 Patients achieving MRD-negative remission

10 14 Number of responding patients and 6 months because of early loss of CTL019 with haematogones present 100 0.9 9 cells with B-cell recovery, and this patient sub- 10 12 0.8 80 sequently had persistence of CTL019. In the 108 10 60 0.7 0.6 patient with the longest remission (2 years), B-cell 107 8 40 0.5 aplasia (absence of CD19-positive cells) (Fig. 3) survival (%) 6 20 NCI, CD19-28z CAR; n=20 10 6 Leukaemia-free 0.4 continued for a year after the loss of CTL019 0 105 4 0 50 100 150 200 250 300 350 400 450 500 550 600 0.3 cells detectable by flow cytometry, suggesting 0.2 functional persistence of CTL019 cells below Absolute circulating B cells 2 Number at risk 12 11 10 7 6 6 5 2 2 2 1 Survival rate at 6 mo, 4 0.1 <10 Probability of Event-free Survival 67% (95% CI, 51–88) the limits of detection by flow cytometry, where- 0 0.0 All patients enrolled as CTL019 remained detectable by means of 0 3 7 14 28 42 68 0 3 6 9 12 15 18 21 24 Days after CAR T-cell infusion 100 quantitative PCR. The probability of relapse- Months since Infusion 80 free B-cell aplasia at 6 months was 73% (95% E No. of 30 19 14 5 1 1 1 1 1 60 Lee et al. Lancet 2015 CI, 57 to 94). 150 Patient 8 Patients Maude et al. NEJM 2015 Blasts 40 CAR T-cells 20 CTL019 for Relapse after Allogeneic Stem-Cell 100 B Overall survival (%) 0 Transplantation 0 50 100 150 200 250 300 350 400 450 500 550 600 1.0 50 Days after CAR infusion 0.9 In the 18 patients who were treated for relapse (thousands) Total CSF cells Total Number at risk 21 20 17 12 10 9 6 3 2 2 1 0.8 of disease after allogeneic stem-cell transplanta- 0 0.7 tion, the median donor chimerism at the time of Pre 10 20 30 40 50 F 0.6 leukapheresis was 100% (range, 68 to 100). No 150 Patient 11 100 0.5 graft-versus-host disease was observed after in- 80 0.4 fusion of CTL019. Event-free survival and overall 100 60 0.3 survival did not differ significantly between the 40 0.2 Survival rate at 6 mo, 78% (95% CI, 65–95) patients who had previously undergone stem-cell 20 Probability of Overall Survival 0.1 50 1·0

(thousands) transplantation and those who had not under- Total CSF cells Total 1·0 0.0 0 3 6 9 12 15 18 21 24 gone stem-cell transplantation (P = 0.21 for event- 0 0·8 0·6 Months since Infusion free survival and P = 0.24 for overall survival).

Pre 10 20 30 40 CAR (%) expressing Days after CAR T-cell infusion 0·4 No. of 30 26 19 10 4 2 1 1 1 Number of circulating T cells of circulating Number 0·2 Patients Therapy after Administration of CTL019 0 G Five patients withdrew from the study after the 10 3 7 14 28 42 68 10 Figure 1. Probability of Event-free and Overall Survival 109 administration of CTL019 to receive other thera- 109 at 6 Months. 8 py; three of these patients underwent allogeneic 10 8 10 Panel A shows the time to an event after infusion of stem-cell transplantation while their disease was 107 CTL019. Events were relapse (in seven patients), no 7 in remission, and the disease remained in remis- 106 10 response (in three patients), and the myelodysplastic 5 syndrome (in one patient). Tick marks indicate the sion 7 to 12 months after the infusion of CTL019. 10 6 10 time of data censoring at the last follow-up or on the Patient 12, who had undergone a previous stem- 4

Absolute circulating blasts date of initiation of alternative therapy (in four pa- <10 105 cell transplantation, had a post-transplantation tients). The curve in Panel B shows overall survival. relapse of T-cell ALL that aberrantly expressed –1 3 7 14 28 Data were censored at the time of the last follow-up. Days after CAR T-cell infusion 4 Absolute circulating CAR T cells CAR Absolute circulating <10 In both panels, dashed lines represent 95% confidence CD19, was refractory to two intensive reinduc- intervals. tion regimens, and entered a morphologic remis- H 3 7 14 28 42 68 Day –1 Day 10 Day 53 sion after the infusion of CTL019, but the patient 6 105 79% 10 had minimal residual disease (0.09%). She sub- 104 105 sequences remained detectable by means of quan- sequently received bortezomib and an infusion 103 titative polymerase-chain-reaction (PCR) assay in of donor lymphocytes, and the disease remained CD3+ cells 4 102 10

CAR patients with sustained remissions until 2 years in remission without minimal residual disease at CD3 103 (Fig. 2C and data not shown). This assay showed 11 months. In Patient 11, the myelodysplastic 105 12% 32% 0% 0% 44% 0% very high levels of proliferation of CTL019 cells; syndrome developed and led to overt acute my- 2 104 10 all patients had peak levels greater than 5000 cop- eloid leukemia with a monosomy 8 clone that 103 Copies CAR per 100 ng DNA Copies ies per microgram of genomic DNA, and 26 pa- also shared cytogenetic features with the original 1 102 <10

CD19 tients had peak levels greater than 15,000 copies B-cell ALL. Mononuclear cells 102 103 104 105 102 103 104 105 102 103 104 105 1 71428 42 68 CD34 Days after CAR T-cell infusion

1510 n engl j med 371;16 nejm.org October 16, 2014 The New England Journal of Medicine 520 www.thelancet.com Vol 385 February 7, 2015 Downloaded from nejm.org on November 7, 2015. For personal use only. No other uses without permission. Copyright © 2014 Massachusetts Medical Society. All rights reserved. Tisagenlecleucel in B-Cell Lymphoblastic Leukemia Tisagenlecleucel in B-Cell Lymphoblastic Leukemia

the null hypothesis of an overall remission rate 9.4 (SAS Institute). Additional details regarding Figure 2. Duration of Remission, Event-free Survival, A Duration of Remissionof 20% against the alternative hypothesis of an the statistical analysis are provided in the Sup- overall remission rate of 45% or higher at an plementary Appendix. and Overall Survival. 1.0 overall one-sided significance level of 2.5%. Panel A shows the duration of remission, defined as 0.9 An interim analysis was planned after the Results the time to relapse after the onset of remission, in the first 50 patients who received a tisagenlecleucel 61 patients who had a best overall response of either 0.8 infusion had completed 3 months of follow-up Patients complete remission or complete remission with incom- or discontinued participation in the study. The Between April 8, 2015, and the data cutoff on plete hematologic recovery. Panel B shows event-free 0.7 results with regard to the primary end point April 25, 2017, a total of 107 patients were survival among the 75 patients who received an infusion, were considered to be significant in the interim screened, and 92 were enrolled (Fig. 1). A total 0.6 defined as the time from tisagenlecleucel infusion to analysis if the one-sided P value was lower than of 75 patients received an infusion of tisagenle- the earliest of the following events: no response (8 pa- 0.0057. Key secondary end points were tested cleucel, with a median time from enrollment 0.5 sequentially (after the primary end point was to infusion of 45 days (range, 30 to 105). The tients), relapse before response was maintained for at significant) to control the overall alpha. median duration of follow-up among patients 0.4 least 28 days (2), or relapse after having complete re- The results with regard to overall remission who received a tisagenlecleucel infusion was mission or complete remission with incomplete hema- 0.3 rate, response duration, event-free survival, over- 13.1 months. At enrollment, patients who received tologic recovery (17). A total of 32 patients had still not all survival, cellular kinetics, and safety that are tisagenlecleucel had a median age of 11 years had an event at the time of data cutoff. Data for 16 more 0.2 presented in this report are from an updated (range, 3 to 23), a median of 3 previous therapies Probability of Continued Remission No. of patients, 61 patients were censored for event-free survival — 8 pa- analysis that included 75 patients who received (range, 1 to 8), and a median marrow blast per- 0.1 No.tisagenlecleucel of events, 17 and had completed 3 months of centage of 74% (range, 5 to 99); 46 patients tients for allogeneic stem-cell transplantation during Median duration of remission, not reached follow-up or discontinued the study at an earlier (61%) had undergone previous allogeneic hemato- remission, 7 patients for new cancer therapy other than 0.0 point. For the time-to-event analyses, Kaplan– poietic stem-cell transplantation (Table S1 in the stem-cell transplantation during remission (4 received 0 Meier2 curves4 6 were8 used10 to estimate12 14 survival16 18dis- 20Supplementary22 Appendix). ELIANA:humanized anti-CD19 Pivotal CAR T cells, 1 received Multicenter ponatinib, Novartistributions TrialMonths after infusion. since Onset All statistical of Remission tests were Before tisagenlecleucel infusion, 72 of 75 pa- 1 receivedThe new vincristineengland journalsulfate and of blinatumomab,medicine and performed with the use of SAS software, version tients (96%) received lymphodepleting chemo- No. at Risk 61 54 43 33 23 18 8 7 3 1 0 1 received antithymocyte globulin), and 1 patient for lack of adequate assessment. Ten patients were followed Figure 1. Screening, Enrollment, Treatment, and Follow-up. for relapse afterOriginal new therapy, Article 4 of whom had a relapse B Event-free and Overall Survival The first patient’s first visit occurred on April 8, 2015. The median time from 1.0 107 Patients were screened or died. Panel B also shows overall survival among the tisagenlecleucel infusion to data cutoff was 13.1 months. The reasons for 75 patients who received an infusion from the date of patients not enrolling in the study after screening included not meeting the 0.9 inclusion criteria or meeting the exclusion criteria (11 patients, including tisagenlecleucel infusion to the date of death from any <5% blasts in the bone marrow in 8 patients), death before acceptance of Tisagenlecleucel in Children and Young 92 Were enrolled cause. Nineteen patients died after tisagenlecleucel in- 0.8 the apheresis sample at the manufacturing facility (2 patients; 1 who died Adults withfusion, andB-Cell 56 patients Lymphoblastic had their data censored Leukemia at the from pulmonary hemorrhage and 1 who died from multiorgan failure), phy- 0.7 Overall survivalsician decision (1), and apheresis-related issue (1). All patients who com- time of the last follow-up. Tick marks indicate the time 17 Were excluded S.L. Maude, T.W. Laetsch, J. Buechner, S. Rives, M. Boyer, H. Bittencourt, 7 Had tisagenlecleucel pleted screening and whose apheresis product was received and accepted P. Bader, M.R.of Verneris, censoring. H.E. Stefanski, G.D. Myers, M. Qayed, B. De Moerloose, 0.6 product-related issues by the manufacturing facility were enrolled in the study. Of the 75 patients H. Hiramatsu, K. Schlis, K.L. Davis, P.L. Martin, E.R. Nemecek, G.A. Yanik, Event-free survival 7 Died who received an infusion, 65 (87%) received bridging chemotherapy be- 3 Had adverse events tween enrollment and infusion, and 72 (96%) received lymphodepleting 0.5 C. Peters, A. Baruchel, N. Boissel, F. Mechinaud, A. Balduzzi, J. Krueger, chemotherapy (fludarabine–cyclophosphamide [71 patients] or cytarabine– C.H. June,B-Cell B.L. Levine, Aplasia P. Wood, T. Taran, M. Leung, K.T. Mueller, Y. Zhang, etoposide [1]). Seventeen enrolled patients did not receive a tisagenlecleucel K. Sen, D. Lebwohl, M.A. Pulsipher, and S.A. Grupp Probability 0.4 75 Underwent infusion infusion because of product-related issues (7 patients), death (7 patients; All patients with a response to treatment had 4 from disease progression and 1 each from sepsis, respiratory failure, and B-cell aplasia, and most patients in the study 0.3 No. of No. of Median fungemia), and adverse events (3 patients; 1 each from graft-versus-host ABSTRACT Patients27 DiscontinuedEvents Survival Rate at 6disease, Mo systemic mycosis, and fungal pneumonia). Tisagenlecleucel product- received immunoglobulin replacement in accor- 0.2 11 Died related issues included an inability to manufacture as a result of poor cell 9 Had lack of efficacymo % (95% CI) growth for 6 patients and a technical issue unrelated to cell growth for 1 pa- 5 Underwent new therapy BACKGROUNDdance with local practice. The median time to Overall Survival 75 19 19.1 90 (81–tient.95) Patients who received the infusion but discontinued follow-up were 0.1 Event-free 75 for27 ALL while innot complete 73 (60–82) In a single-centerB-cell phase recovery 1–2a study, was the anti-CD19not reached chimeric (Fig. antigen S4 receptor in the (CAR) The authors’ full names, academic de- remission followed for survival. At the time of data cutoff, 27 patients had discontin- T-cell therapy tisagenlecleucel produced high rates of complete remission and was grees, and affiliations are listedSurvival in the 2 Withdrew orreached were withdrawn ued follow-up owing to death (11 patients; 7 from disease progression and Appendix. Address 0.0 reprint requests to by guardian associated withSupplementary serious but mainly Appendix). reversible toxic Theeffects probabilityin children and ofyoung 1 each from encephalitis, cerebral hemorrhage, systemic mycosis, and hepa- Dr. Maude at 3012 Colket0 Translational2 4 6 8 10 12 14 16 18 20tobiliary22 disorders related to allogeneic hematopoietic stem-cell transplan- adults with relapsedmaintenance or refractory of B-cellB-cell acute aplasia lymphoblastic at 6 monthsleukemia (ALL).after Research Bldg., 3501 Civic Center Blvd., tation), lack of efficacy (9 patients; nonresponse or relapse), new therapy Philadelphia, PA 19104, or at maude@ Months since Tisagenlecleucel Infusion 48 Remained in follow-up while in complete remission (5), and patient or guardian decision (2); 48 METHODS infusion was 83% (95% CI, 69 to 91). email . chop . edu. No. at Risk patients remained in follow-up. ALL denotes acute lymphoblastic leukemia. We conducted a phase 2, single-cohort, 25-center, global study of tisagenlecleucel Drs.Overall Pulsipher survival and Grupp75 contributed72 64 58 55 40 30 20 12 8 2 0 equally to this article. in pediatric andCytokine young adult Response patients withMaude CD19+ etrelapsed al. NEJM or refractory 2018 B-cell ALL. Event-free survival 75 64 51 37 33 19 13 8 3 3 1 0 The primary end point was the overall remission rate (the rate of complete remission N Engl J Med 2018;378:439-48. or complete remissionAmong withthe incomplete75 patients hematologic who received recovery) within tisagenlec 3 months.- DOI: 10.1056/NEJMoa1709866 n engl j med 378;5 nejm.org February 1, 2018 441 Copyright © 2018 Massachusetts Medical Society. RESULTS leucel, transient increases in serum interleu- The New England Journal of Medicine Downloaded from nejm.org at Lane Medical Library, Stanford University Med Center on March 12, 2018. For personal use only. No other uses without permission. For this plannedkin-6, analysis, interferon 75 patients gamma, received an and infusion ferritin of tisagenlecleucel levels oc- andprotein level was observed in most patients, Copyright but© 2018 Massachusetts Medical Society. All rights reserved. could be evaluatedcurred for duringefficacy. The the overall cytokine remission release rate within syndrome 3 months was with large variability. 81%, with all patients who had a response to treatment found to be negative for minimal residualafter disease, infusion; as assessed these by increases means of flow tended cytometry. to be The more rates of event-free survivalpronounced and overall in survival patients were 73%with (95% grade confidence 4 cytokine interval [CI], Safety 60 to 82) andrelease 90% (95% syndrome CI, 81 to 95), thanrespectively, in patientsat 6 months withand 50% lower (95% CI,The safety analysis set included all 75 patients 35 to 64) and 76% (95% CI, 63 to 86) at 12 months. The median duration of remis- sion was not gradesreached. Persistence(Fig. S5 ofin tisagenlecleucel the Supplementary in the blood Appendix). was observed forwho received an infusion of tisagenlecleucel; the as long as 20Similar months. Grade trends 3 or 4 wereadverse observedevents that were in suspected the levels to be related of median time from infusion to data cutoff was to tisagenlecleucelother occurred cytokines, in 73% including of patients. interleukin-10,The cytokine release inter syndrome- 13.1 months (range, 2.1 to 23.5). Eighteen pa- occurred in 77% of patients, 48% of whom received tocilizumab. Neurologic events occurredleukin-12p70, in 40% of patients interleukin-1 and were managedβ, interleukin-2, with supportive inter care,- andtients (24%) received their infusions in an outpa- no cerebral edemaleukin-4, was reported. interleukin-8, and tumor necrosis fac- tient setting. All patients had at least one adverse CONCLUSIONStor α. A transient increase in the C-reactive event during the study; 71 of 75 patients (95%) In this global study of CAR T-cell therapy, a single infusion of tisagenlecleucel provided durable remission with long-term persistence in pediatric and young adult patients with relapsed or refractory B-cell ALL, with transient high-grade toxic effects. (Funded by Novartis Pharmaceuticals;n ClinicalTrials.gov engl j med 378;5 number,nejm.org February 1, 2018 443 NCT02435849.)

n engl j med 378;5 nejm.org February 1, 2018 439 The New England Journal of Medicine Downloaded from nejm.org at Lane Medical Library, Stanford University Med Center on March 12, 2018. For personal use only. No other uses without permission. Copyright © 2018 Massachusetts Medical Society. All rights reserved.

CD19 CAR in NHL Refractory LBCL patients do poorly with standard therapies

• SCHOLAR-1, a retrospective, international, patient-level, multi-institution study and the largest reported analysis of outcomes in patients with refractory large B cell lymphoma, demonstrated that these patients have a very poor prognosis1 – N = 636 (post-rituximab era, 2000-2017) – ORR = 26% – CR rate = 7% – Median OS = 6.3 mo – These results provided a benchmark for evaluation of new approaches • Previous analyses of SCHOLAR-1 standardized to ZUMA-1 with ≥ 6 mo follow-up suggested the benefit of axi-cel in refractory large cell lymphoma2 1. Crump M, et al. Blood. 2017 ZUMA1: Study of Kite’s KTE-c19 in NHL

• Axicabtagene ciloleucel (Yescarta) is an autologous CAR T therapy with a CD3ζ/CD28- based signaling that recognizes and eliminates CD19-expressing cells

• NHL is the most common hematologic malignancy in the US • Outcomes in refractory aggressive NHL are poor • SCHOLAR-1 patient level meta-analysis of refractory NHL • ORR of 26% (CR of 7%) • Median OS of 6.3 months ZUMA-1 at Median f/u 15.4 Months: 42% Progression-Free and 56% Alive Progression-Free Survival Overall Survival

Landmark PFS Landmark OS 6-month 49 6-month 78 12-month 44 12-month 59 18-month 41 18-month 52 Neelapu et al ASH 2017 578

Limitations in Efficacy Antigen Escape and T cell Exhaustion Experience in B-ALL Has Illuminated Mutually Exclusive Causes for CAR T Cell Failures

Pre-CD19 Relapse Post-CD19 CAR Therapy CAR Therapy CD22 Antigen Antigen Disease Disease CD22

High negPos/lo ControlControl RelapseRelapse CD19 CD19

AntigenAntigen • Appears to be the most common cause of No No pos Response Neg/lo relapse Low Response RelapseRelapse

Target Antigen Expression • Incidence not truly known Low High • Emerging data suggests a similar rate in DLBCL CAR T Cell Expansion/Persistence • Addressing this problem is important for B-ALL but also identifies a vulnerability for CAR T cells which will be amplified in AML and solid tumors Tisagenlecleucel in B-Cell Lymphoblastic Leukemia

Figure 2. Duration of Remission, Event-free Survival, A Duration of Remission and Overall Survival. 1.0 Panel A shows the duration of remission, defined as the time to relapse after the onset of remission, in the 0.9 61 patients who had a best overall response of either 0.8 complete remission or complete remission with incom- plete hematologic recovery. Panel B shows event-free 0.7 survival among the 75 patients who received an infusion, 0.6 defined as the time from tisagenlecleucel infusion to the earliest of the following events: no response (8 pa- 0.5 tients), relapse before response was maintained for at least 28 days (2), or relapse after having complete re- 0.4 mission or complete remission with incomplete hema- 0.3 tologic recovery (17). A total of 32 patients had still not had an event at the time of data cutoff. Data for 16 more 0.2 Probability of Continued Remission No. of patients, 61 patients were censored for event-free survival — 8 pa- 0.1 No. of events, 17 tients for allogeneic stem-cell transplantation during Median duration of remission, not reached remission, 7 patients for new cancer therapy other than 0.0 stem-cell transplantation during remission (4 received 0 2 4 6 8 10 12 14 16 18 20 22 humanized anti-CD19 CAR T cells, 1 received ponatinib, Months since Onset of Remission 1 received vincristine sulfate and blinatumomab, and No. at Risk 61 54 43 33 23 18 8 7 3 1 0 1 received antithymocyte globulin), and 1 patient Relapsefor with CD19- ALL was majority of cases lack of adequate assessment. Ten patients were followed for relapse after new therapy, 4 of whom had a relapse B Event-free and Overall Survival or died. Panel B also shows overall survival among the 1.0 75 patients who received an infusion from the date of 0.9 tisagenlecleucel infusion to the date of death from any cause. Nineteen patients died after tisagenlecleucel in- 0.8 fusion, and 56 patients had their data censored at the Overall survival time of the last follow-up. Tick marks indicate the time 0.7 of censoring. 0.6 Event-free survival 0.5 B-Cell Aplasia Probability 0.4 All patients with a response to treatment had 0.3 No. of No. of Median B-cell aplasia, and most patients in the study Patients Events Survival Rate at 6 Mo received immunoglobulin replacement in accor- 0.2 mo % (95% CI) dance with local practice. The median time to Overall Survival 75 19 19.1 90 (81–95) 0.1 Event-free 75 27 not 73 (60–82) B-cell recovery was not reached (Fig. S4 in the Survival reached Supplementary Appendix). The probability of 0.0 0 2 4 6 8 10 12 14 16 18 20 22 maintenance of B-cell aplasia at 6 months after Months since Tisagenlecleucel Infusion Maude et al, NEJM, 2018 infusion was 83% (95% CI, 69 to 91). No. at Risk Overall survival 75 72 64 58 55 40 30 20 12 8 2 0 Cytokine Response Event-free survival 75 64 51 37 33 19 13 8 3 3 1 0 Among the 75 patients who received tisagenlec- leucel, transient increases in serum interleu- kin-6, interferon gamma, and ferritin levels oc- protein level was observed in most patients, but curred during the cytokine release syndrome with large variability. after infusion; these increases tended to be more pronounced in patients with grade 4 cytokine Safety release syndrome than in patients with lower The safety analysis set included all 75 patients grades (Fig. S5 in the Supplementary Appendix). who received an infusion of tisagenlecleucel; the Similar trends were observed in the levels of median time from infusion to data cutoff was other cytokines, including interleukin-10, inter- 13.1 months (range, 2.1 to 23.5). Eighteen pa- leukin-12p70, interleukin-1β, interleukin-2, inter- tients (24%) received their infusions in an outpa- leukin-4, interleukin-8, and tumor necrosis fac- tient setting. All patients had at least one adverse tor α. A transient increase in the C-reactive event during the study; 71 of 75 patients (95%)

n engl j med 378;5 nejm.org February 1, 2018 443 The New England Journal of Medicine Downloaded from nejm.org at Lane Medical Library, Stanford University Med Center on March 12, 2018. For personal use only. No other uses without permission. Copyright © 2018 Massachusetts Medical Society. All rights reserved. NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12320 ARTICLE ab 100 42 400 Pre-CAR 100 100 Pre-CAR CRP (mg l 80 Post-CAR 80 80 Post-CAR 40 300 60 Progressive disease 60 60 40 38 200 20 40 40

–1 CAR not detected

36 100 % positive cells 20 20

) 0 0 0 Temperature (°C) 34 0 CD19CD22CD10CD34 CD33 CD14 2 3 4 5 2 3 4 5 0 10 20 30 CD11b CD117 0 0 HLA-DR 10 10 10 10 10 10 10 10 c d 100 42 400 100 100 CRP (mg l 80 40 300 60 80 80 Complete response 60 60 38 200 40 20 40 40 36 100 –1

CAR 0.4% d28 % positive cells ) 0 20 20 relapse @6 m

Temperature (°C) 34 0 0 0 0 10 20 30 CD19CD22CD10CD34 CD33 CD14 2 3 4 5 2 3 4 5 CD11b CD117 0 0 HLA-DR 10 10 10 10 10 10 10 10 ef 100 42 400 100 100 CRP (mg l 80 80 80 40 300 ARTICLE 60 Mixed response Relapse with Lineage switch Pediatr Blood Cancer 2016;63:1113–1115 Received 16 May 201640| Accepted 21 Jun 2016 | Published 27 Jul 2016 DOI: 10.1038/ncomms12320 OPEN 60 60 38 200 BRIEF REPORT CD19 CAR immune pressure induces B-precursor40 Lineage Switch in MLL-Rearranged Infant40 Leukemia Following NATURE COMMUNICATIONS20 | DOI: 10.1038/ncomms12320 –1 acute lymphoblastic leukaemia lineage switch CD19-Directed Therapy ARTICLE

36 100 CAR 0.2% d28 % positive cells ) 0 20 1 202 1 exposing inherent leukaemic plasticity Ahmad Rayes, MD, ∗ Richard L. McMasters, MD, and Maureen M. O’Brien, MD

Temperature (°C) 34 0 Rearrangements0 of the mixed lineage leukemia (MLL) gene oc- refractory0 to cytotoxic chemotherapy who was treated with bli- Elad Jacoby1,w, Sang M.ab Nguyen1, Thomas J. Fountaine1, Kathryn Welp1, Berkley Gryder2, Haiying Qin1,cur frequently in infants with both acutePost-CAR lymphoblastic leukemiaPre-CARnatumomab. Following rapid initial clearance of peripheral lym- (ALL) and acute myeloid leukemia2 (AML).3 Conversions4 of leukemia5 phoblasts, bone marrow evaluation2 demonstrated3 4 a leukemic5 0 10 20 30 Yinmeng Yang1, Christopher D. Chien1, Alix E. Seif3, Haiyan Lei1, Young K. Song2, Javed Khan2, Daniel W. Lee1, CD19CD22CD10CD34 CD33 CD14 cell lineage are rare,0 but occur most105 commonly in the set- lineage switch105 to CD19-negative0 monoblastic AML. Complete re- 1 4 4 CD11b1 CD1171 ting of MLL-rearrangement. Blinatumomab0 is a bidirectional an- mission was achieved with myeloid-directed chemotherapy. Pediatr Crystal L. Mackall , Rebecca A. Gardner , Michael C. Jensen , Jack F. Shern & Terry J. Fry tibody targeting CD19 with10 significant1 activity10 in relapsed10 B- Blood Cancer 2016;63:1113–1115.10 10C 2016 Wiley10 Periodicals,10 Inc. HLA-DR CD19 ⃝ Percentprecursor positive ALL. We report an infant with104 ALL with t(4;11)(q21;q23) 104 CD22 Key words: blinatumomab; CD19; infant; leukemia; lineage; MLL 80 103 103 Adoptive immunotherapy using chimeric antigen receptor (CAR) expressing T cells targeting B220 60 ghthe CD19 B lineage receptor has demonstrated marked success in relapsed pre-B-cell acute 40 2 2 INTRODUCTION 10 methotrexate,10 etoposide, and cyclophosphamide. Bone marrow lymphoblastic leukaemia (ALL). Persisting CAR-T cells generate sustained pressure against CD43 100 100 100 20 0 evaluation0 after this therapy showed persistent ALL with 13– Acute leukemias usually adhere to normal lineage diferentia- CD19 that may drive unique mechanisms of resistance. Pre-B ALL originates from a 15% residual lymphoblasts. 42 100 CD127 tion programs, although they may express markers2 of both3 lym-4 5 2 3 4 5 CRP (mg l committed pre-B cell or an earlier progenitor, with potential to reprogram into other Due to refractory disease, the patient was taken of study 80 phoid and80 myeloid lineages. Rarely, lineage switching010 may10 result10 10 80 010 10 10 10 and enrolled on the COG phase 1/2 trial of blinatumomab hematopoietic lineages. Here we report changes in lineage markers including myeloid from alteration of leukemic cell diferentiation programs follow- BP1 CD19 (AALL1121, NCT01471782)CD22 for children with relapsed or re- ing therapy or at relapse.[1–3] InfantB220 leukemias frequently have 80 conversion in patients following CD19 CAR therapy. Using murine ALL models we study the fractory B-precursor ALL. Bone marrow evaluation imme- 40 Complete response 60 rearrangements60 of the mixed lineage leukemia (MLL; KMT2A) 60 long-term effects of CD19 CAR-T cells and demonstrate partial or complete lineage switch as CD11b 105 diately prior105 to blinatumomab therapy showed 82% CD19- gene on chromosome 11q23. MLL gene rearrangement positive lymphoblasts (Figs. 1A and 2) with t(4;11)(q21;23) and 60 a consistent mechanism of CAR resistance depending on the underlying genetic oncogenic (MLL-R) is associated with a propensity for lineage switch- 40 Gr1 4 new acquisition4 of add(19)(p13) since the original diagnosis. driver. Deletion of Pax5 or Ebf1 recapitulates lineage reprogramming occurring during CD19 ing, typically40 from B-precursor acute10 lymphoblastic leukemia 4010 38 WBC count was 12 103 /mcl with 25–30% circulating lym- (ALL) to acute myeloid leukemia (AML) following cytotoxic × CAR pressure. Our findings establish lineage switch as a mechanism of CAR resistance phoblasts. Blinatumomab was initiated at 5 mcg/m2/day and the 40 KIT chemotherapy or at relapse.[4] 3 3 exposing inherent plasticity20 in genetic subtypes of pre-B-cell ALL. 20 10 treatment10 was complicated by grade 1 cytokine release syndrome –1 CAR detected, Blinatumomab is a bispecifc T-cell engager antibody con- 20 that was managed with dexamethasone without interruption of 36 struct that simultaneously binds CD3-positive cytotoxic T cells % positive cells 2 2

) the blinatumomab infusion. By day 6 of the infusion, WBC 20 relapse @ 30d 0 and CD19-positive B cells, resulting10 in T-cell-mediated lysis of 10 count was 2.9 103/mcl without circulating blasts. Per proto- normal and0 malignant B cells.[5,6] Blinatumomab has demon- 0 × E2a_F6 E2a_F5 E2a_B2 E2a_E9 E2a_B4 E2a_B3 0 0 2 E2a_D5 E2a_D3 E2a_C2 E2a_G2 E2a_G4 col, the blinatumomab dose was escalated to 15 mcg/m /day on E2a_F12 E2a_E10 E2a_B11 E2a_B21 E2a.PBX E2a_C10 E2a_D31 E2a_C12 strated efcacy in the treatment2 of relapsed3 and4 refractory5 2 3 4 5 Temperature (°C) 34 0 day 8 without adverse events. CD19-positive B-precursor ALL.[7] This report details2 the3 case4 5 2 3 4 5 0 010 10 10 10On day 15 of0 the100 infusion,10 protocol-mandated10 10 bone mar- of an infant with refractory MLL-R ALL who developed lin- –5 0 5 10 15 20 10 10 10 10 row evaluation revealed10 83% blasts10 with morphologic10 10 and CD19CD22CD10CD34 CD33 CD14 eage switch from CD19-positive B-precursor ALLCD11b to CD19- Gr1 immunophenotypic switch to acute monoblastic leukemia CD11b CD117negative AML under the selective pressure of CD19-targeted HLA-DR (Figs. 1B andJacoby 2). Cytogenetics et al, demonstrated Nat Comm an identical, 2015 kary- therapy with blinatumomab. CD19 otype to that seenCD11b in the ALL immediately prior to bli- CD19 CAR Off-pressure Off-pressure natumomab [t(4;11)(q21;23), add(19)(p13)]. Figure 2 shows CASE REPORT representative dot plots of the blast immunophenotype be- A male 3-month-old infant presented with pallor, tachyp- fore and after blinatumomab therapy; CD19 was unde- nea, and hepatosplenomegaly. Laboratory evaluation showed tectable following blinatumomab. Targeted gene sequencing Figure 1 | Phenotypic alterations in clinical ALL samples following CD19 CAR. (a,c,e,g) Fever curves,hyperleukocytosis CRP values with white blood cell and (WBC) count clinical 821 (Foundation response Medicine, Cambridge, to MA) CD19 of the monoblastic CAR × 103/mcl, anemia, and thrombocytopenia. The peripheral blood 1 2 Pediatric Oncology Branch, Center for Cancer Research, National Cancer Institute, National Institutes of Health, Bethesda, Maryland 20892, USA.smearGenetics showed abundant lymphoid blasts and fow cytome- treatment. (b,d,f,h) Bar graph demonstrating percent of cellsBranch, expressing National Cancer Institute, National cell Institutes of surface Health, Bethesda, Maryland markers 20892, USA. 3 Division of by Oncology, flow The Children’s cytometry Hospital of in the bone marrow, gated on leukaemic try con4frmed the diagnosis of B-precursor ALL.100 Cytoge- Philadelphia, and Department of Pediatrics, Perelman School of Medicine of the University of Pennsylvania, Philadelphia, Pennsylvania 19104, USA. Ben 1Division of Oncology, Cancer and BloodE2a:PBX Disease Institute, Cincin- Towne Center for Childhood Cancer Research, Seattle Children’s Hospital, and Department of Pediatrics, University of Washington, Seattle,netic Washin studiesgton showed t(4;11)(q21;q23) (MLL-AFF1 rearrange- 2 ment). Cerebrospinal fuid was positive for rare blasts. After the nati Children’s Hospital Medical Center, Cincinnati,2nd Ohio Division 98105, USA. w Present address: Departmentc of Pediatric Hematology and Oncology, The Edmond and Lily Safra’s Children’s Hospital, Sheba Medical Center, of Pathology and Laboratory Medicine, Cincinnati Children’s Hos- blasts, with representative flow cytometry histograms of CD19 and CD11b on the right (grey, control; blue,emergence pre-CAR; of double-volume exchange red, transfusion, post the75 patient CAR). (a,b) Pre-CAR sample and Sackler School of Medicine, Tel Aviv University, Israel. Correspondence and requests for materials should be addressed to T.J.F. (email: [email protected]). pital Medical Center, Cincinnati, Ohio 3rd was enrolled on Children’s Oncology Group (COG) protocol Confict of interest: Nothing to declare NATURE COMMUNICATIONS | 7:12320 | DOI: 10.1038/ncomms12320 | www.nature.com/naturecommunications AALL06311 (NCT00557193). Following induction chemother-50 and day 30 post-CAR sample from a patient who did not experience CRS following CD19 CAR. (c,d) Pre-CARapy (vincristine, daunorubicin, and asparaginase,180 corticosteroids, days∗Correspondence post-CAR to: Ahmad Rayes, samples Division of Oncology, Cancerfrom a and cytarabine), bone marrow evaluation showed 30–40% resid- and Blood Disease Institute, Cincinnati Children’s Hospital Medical þ CD19 þ ual lymphoblasts by morphology and fow cytometry.25 The pa- Center, 3333 Burnet Avenue, MLC 7015, Cincinnati, OH 45229. patient with normal karyotype multiple relapsed ALL, who had a severe CRSCD22 followed by an MRD-negativetient was randomized complete to the experimental treatment response arm includ- E-mail: with [email protected] CD19 CAR, with no ing lestaurtinib, a FLT3 tyrosine kinase inhibitor, withPercent survival high-dose Received 8 January 2016; Accepted 25 January 2016

+ C 0 CD19 (%) 1/1 (100%) 0/1 (0%) 3/9 (33%)⃝ 2016 Wiley Periodicals,16/21 Inc. (76%) CAR þ cells persisting beyond 60 days. (e,f) Pre-CAR and day 30 post-CAR samples from a patient withDOI 10.1002/pbc.25953 normal karyotype0 ALL who10 experienced 20 30 a mild Published online 23 February 2016 in Wiley Online Library þ (wileyonlinelibrary.com). Days CRS and CAR expansion, but had persistent disease. (g,h) Pre-CAR andd post-CAR samples from an infant with MLL-rearranged ALL, treated with CD19- 41BB-zeta CAR, who relapsed with myeloid blasts. 100 E2a:PBX 2nd

CD19 75 3rd 38–40 CD22 malignant B cells and pre-B ALL . Similar to the patient model, CD19-CAR treatment induced50 prolonged remissions in presented in Fig. 1g,h, extended phenotyping using multi- E -RET leukaemia with occasional late25 mortality (Supplementary m Percent survival GR1 0 parameter flow cytometry and identified that, in addition toCD11b Fig. 6). However, at necropsy bone marrow0 10 showed 20 30 normal- Days loss of B-cell markers, late relapses acquired phenotypic markers appearing hematopoiesisNo CD19+ population on propagation with no evidence of leukaemia by flow of alternative lineages, including CD11b, Gr1 and cKIT (Fig.e 2d cytometry, PCR for the E -RET transgene and failed to generate m 100 E2a:PBX and Supplementary Table 1). Unsupervised hierarchical leukaemia following injection into75 secondary2nd recipients CD19 3rd clustering of cell-surface markers further illustrated that earlyCD22 (Supplementary Fig. 6). Persistence50 of CAR T cells was relapses were similar to parental pre-B ALL with the exception of confirmed by PCR. Thus, in contrast25 to E2a:PBX leukaemia, Percent survival CD19 loss, whereas later relapses were characterized by multiple Em-RET pre-B cell ALL did not relapse0 following CD19 CAR via GR1 0 10 20 30 phenotypic alterations (Fig. 2d). Unsupervised clustering of RNACD11b lineage switch suggesting that the occurrenceDays of this resistance sequencing data further confirmed distinct featuresFigure 3 of | Lineage later switch phenotypemechanism is not detectable may in parental depend leukaemia on and is the stable genetic when CAR pressure driver. is removed. (a) Primary E2a:PBX ALL was compared with earlier relapses (Fig. 2f and Supplementarysingle-cell Fig. cloned by5), limiting dilution, expanded and analysed by flow cytometry. Heatmap of cell surface marker expression evaluated by flow cytometry. (b) CD45.2 þ B220 þ leukaemic blasts from sample 59–61 obtained at day 58 following CAR demonstrating intermediate B/myeloid phenotype expressing with loss of B-cell-associated transcripts and gain of expressionfor CD22, CD11b and of Gr1, but negative for CD19. (c–e) In vivo passage of post-CAR CD19 À relapsed leukaemia, with E2a:PBX control, in absence of additional CAR treatment. (c) B/Myeloid biphenotypic sample 25–2 passaged twice off pressure, with re-emergence of a CD19 þ population in 16/21 mice myeloid or T-cell genes consistent with lineagein 3rd switching. passage (E2a:PBX controlMyeloidn 16, 2nd n clones10, 3rd n 12). not (d) Lineage detectable switch samples prior 24–6 (n 5 to per group)CD19 and (e CAR) A001 (E2a:PBX exposure control n .3, ¼ ¼ ¼ ¼ ¼ 2nd n 3, 3rd n 4) passaged twice. Respective survival curves for 1 of 2 experiments shown on the right. Interestingly, principal component analysis revealed¼ greater¼ Lineage switch in patients following cytotoxic therapy results diversity among lineage switch relapses (Fig. 2g), indicatingability for pre-B that ALL tofrom switch lineages selection under CAR or immune geneticthe reprogramming H3K27ac histone mark20 in. lymphoid, The diverse mixed B/myeloid gene and pressure can result in emergence of a stable, alternatively myeloid relapses to evaluate the changes in activated chromatin persistent CD19 CAR can induce multiple genedifferentiated expression CAR-resistantexpression phenotype. profile of the post-CARsites in lineage-specific mixed transcriptional phenotype as a and potential myeloid mechanism of post-CAR lineage reprogramming. In correlation with the profiles that allow pre-B ALL escape. relapses derived from E2a:PBX1transcriptome ALL profile following(Fig. 2f), the post-CAR CD19 CD19 CAR-T negative We next used an Em-RET murine pre-B ALLGenomic to accessibility test associatessuggested with lineage emergence of relapse. We oflymphoid leukaemia relapse had with H3K27ac distinctsignals in B-cell-associated lineage the hypothesis that CAR resistance mechanismnext may performed differ chromatincharacteristics immunoprecipitation resulted sequencing of fromtranscription reprogramming. factor promoter regions To including evaluateCd19, Pax5 forand depending on the genetic basis41,42. As with theNATURE E2a:PBX1 COMMUNICATIONS | 7:12320the | DOI: possibility 10.1038/ncomms12320 of | www.nature.com/naturecommunications selection of a pre-existent alternative lineage cell5

NATURE COMMUNICATIONS | 7:12320 | DOI: 10.1038/ncomms12320 | www.nature.com/naturecommunications 3 Immune Escape via Isoform Switch or CD19 mutation

Sotillo et al, Cancer Discovery, 2015 Aligning leukemia cells to developing B cells identifies expanded subpopulations

Good et al., Nat Med 2018 Focus on Failure with CAR-T cells: Who will have CD19- relapse?

Assess MRD ~60% ~40% Relapse CAR T and B cell Relapse Remission aplasia CD19-

Months ? Different mechanisms for CD19 loss could be detected before CAR

Multi-Dimensional Scaling (MDS) plot CNA analysis by DNA-seq VAF in DNA WES (Chr16) CD19 ex1-14 Healthy Hemizygous del(16) DELETED N/A CHOP105R (p13.11p11.1) W111delinsWPLR 100% (22/22)

CNA analysis by DNA-seq CNA analysis by DNA-seq VAF in DNA VAF in DNA WES (Chr16) CD19 ex1-14 WES (Chr16) CD19 ex1-14 WT >50% LOH N/A CHOP133R Hemizygous del(16) CHOP101R INTACT G67fs 28% (13/47) G100fs 100% (34/34) P204fs 15% (10/66) Pablo Domizi, David Barre Hypothesis Generation: Focus on MOST Relevant Cell Populations

101 105 135 136

100 100 100 100

80 80 80 80

60 60 60 60 Pre CAR T 40 40 40 40 Percentage Percentage Percentage Percentage

20 20 20 20

0 0 0 0

HSC HSC HSC HSC Pro-BIPro-BIIPre-BIPre-BII Pro-BIPro-BIIPre-BIPre-BII Pro-BIPro-BIIPre-BIPre-BII Pro-BIPro-BIIPre-BIPre-BII Pre-Pro-B MatureMature BI BII Pre-Pro-B MatureMature BI BII Pre-Pro-B MatureMature BI BII Pre-Pro-B MatureMature BI BII ProgenitorProgenitor I II Immature BI Early non-BI Unclassified ProgenitorProgenitor I II Immature BI Early non-BI Unclassified ProgenitorProgenitor I II Immature BI Early non-BI Unclassified ProgenitorProgenitor I II Immature BI Early non-BI Unclassified Progenitor III ImmatureImmature BII BIII Early Maturenon-BII Non-B Progenitor III ImmatureImmature BII BIII Early Maturenon-BII Non-B Progenitor III ImmatureImmature BII BIII Early Maturenon-BII Non-B Progenitor III ImmatureImmature BII BIII Early Maturenon-BII Non-B Population Population Population Population

100 100 100 100

80 80 80 80

60 60 60 60

Relapse 40 40 40 40 Percentage Percentage Percentage Percentage

20 20 20 20

0 0 0 0

HSC HSC HSC HSC Pro-BIPro-BIIPre-BIPre-BII Pro-BIPro-BIIPre-BIPre-BII Pro-BIPro-BIIPre-BIPre-BII Pro-BIPro-BIIPre-BIPre-BII Pre-Pro-B MatureMature BI BII Pre-Pro-B MatureMature BI BII Pre-Pro-B MatureMature BI BII Pre-Pro-B MatureMature BI BII ProgenitorProgenitor I II ImmatureImmature BI BII Early non-BI Unclassified ProgenitorProgenitor I II ImmatureImmature BI BII Early non-BI Unclassified ProgenitorProgenitor I II ImmatureImmature BI BII Early non-BI Unclassified ProgenitorProgenitor I II Immature BI Early non-BI Unclassified Progenitor III Immature BIII Early Maturenon-BII Non-B Progenitor III Immature BIII Early Maturenon-BII Non-B Progenitor III Immature BIII Early Maturenon-BII Non-B Progenitor III ImmatureImmature BII BIII Early Maturenon-BII Non-B Population Population Population Population RESEARCH LETTER

D32–D70 a 1 × 106 CAR T cells 0.4 × 106 CAR T cells 0.2 × 106 CAR T cells b 8 103 108 108 108 10 19-28 19-BB 19-28ζ 106 102 106 19-BBζ 106 106 NT 104 NA 1 4 4 4 10 10 10 10 102 CAR T cells NA 0 0 Inhibitory receptor expression 2 2 2 EOMES/T-bet ratio Average radiance 10 10 10 10 10 PD-1 LAG-3 TIM-3 4 4 4 –1 –1 –1 11 18 25 32 39 46 53 60 67 11 18 25 32 39 46 53 60 67 11 18 25 32 39 46 53 60 67 Number of inhibitory receptors Days Days Days 19-28ζ19-BBζ 19-28ζ19-BBζ 031 2 c D14 D32–D70 d 0.009 e 0.1 NALM6 0.78 0.007 CD19–mCherry Gated on NALM6 Gated on CAR T cells 0.009 30,000 0.01 15,000 15,000 0.001 105 105 105 20,000 UT 10,000 10,000 4 4 4 10 10 10 19-del 10,000 19-28ζ 5,000 5,000 103 103 103 MFI CD19 ALLLETTER cells lose CD19 when it ends up onCAR thehttps://doi.org/10.1038/s41586-019-1054-1 CAR-TMFI CD19 cells 19-BBζ

NA MFI mCherry 0 0 0 102 102 102 0124 0 124 0124

CD19 molecules per cell CD19 molecules per cell D0 D6 NT NT Time (h) Time (h) Time (h) 19-28ζ19-BBζ 19-28CAR19-BBζ ζT cell trogocytosis NALM6and cooperative killing regulate tumour antigen escape f + – g NALM6 Mohamad Hamieh1, Anton Dobrin1, Annalisa Cabriolu1, Sjoukje J. C. van der Stegen1, Theodoros Giavridis1, Jorge Mansilla-Soto1, Blasts CAR T cells Justin Eyquem1, Zeguo Zhao1, Benjamin M. Whitlock2, Matthew M. Miele3, Zhuoningtrog Li3, Kristentrog M. CunananNALM64, Morgan Huse2, CD19–mCherry 1 h co-culture 3,5 1,5 1,5 1,2 0.02 Ronald C. Hendrickson , Xiuyan WangQuantitative, Isabelle Rivière mass & Michel Sadelain * 4 spectrometry D1 D2 D1 D2 D1 D2 10 0.02 2,000 0.02 ChimericAs antigen CAR-trogT-cells receptors (CARs)gain+ are CD19 synthetic antigen or receptors CARCD19 (Extended Data Fig. 3d). The transfer of CD19 protein from ALL-1 Heavy aa that reprogram T cell specificity, function and persistence1. Patient- NALM6 cells to T cells thus displayed the hallmarks of CAR-mediated 0.02 1,500 ALL-2 (7 days) derived CARCD22, T cells have they demonstrated demonstrate remarkable efficacy against trogocytosis,CD81 as further compounded by inhibition with blockers3 of a range of B-cell malignancies1–3, and the results of early clinical actin polymerization17 (Extended Data Fig. 3e). Co-culture with CD19-10 4 CLL-1 trials suggestincreased activity in multiple features myeloma . Despite of high complete knockoutCD22 NALM6 cells expressing a CD19–mCherry fusion molecule 1,000 response rates, relapses occur in a large fraction of patients; some resulted in the detection of both mCherry and CD19 in TMFI CD19 cells, CLL-2 FACS sorting 1,2,4–9 CAR of these areexhaustion antigen-negative and others are antigen-low . demonstrating whole-proteinIntensity CD19 membrane extraction (Fig. 1e). Normal aa Unlike the mechanisms that result inQuantitative complete and permanent mass Loading of CD19–mCherry-expressing NALM6 cells with heavy amino2 6,8,9 10 500 (7 days) antigen loss , those that lead to escape ofspectrometry antigen-low tumours acids and 19-28ζ cells with light amino acids and then the sorting of + − 024 024 remain unclear. Here, using mouse models of leukaemia, we show mCherry-positive trogocytosis-positive (trog ) and -negative (trog ) Hamieh et al. Nature 2019 02.5 5.07.5 10.0 that CARs provoke reversible –antigen loss through trogocytosis, an singlet T cells after brief co-culture unequivocally demonstrated CD19 Time (h) Time (h) active process inCAR-trog which the target antigen is transferred to T cells, peptides in the trog+ but not the trog− fraction (Fig. 1f, Extended Data thereby decreasing target density on tumour cells and abating T cell Fig. 3f). CD81, which forms a complex with CD19, was also detected + − h T cell/T cell co-cultureactivity by promoting fratricide T cell killing and T cell exhaustion. i in trog but not trog T cells (Fig. 1f), while concomitantly lost in the j These mechanisms affect both CD28- and 4-1BB-based CARs, albeit co-cultivated NALM6 cells (Extended Data Fig. 3g). By contrast, CD22 NALM6 D6 after sorting NALM6 differentially, depending on antigen density. These dynamic features remained CD19unchanged in CTL NALM6 cells and was not detected by mass 0.34 0.05 CD19–mCherry can be offset by cooperative killing and combinatorial targeting to spectrometrytransd. in the T cells (Fig. 1f, Extended Data Fig. 3h, i). CD19–mCherry 0.02 augment tumour0.02 responses to immunotherapy.0.02 CD19 trogocytosis occurred similarly after co-culture with NALM6, 0.02 Tracer + 25 40 We15 modelled0.02 CAR therapy relapse80 by infusing limiting doses of SUP-B15, Raji and CD19T cellSK-OV-3 cells (Extended Data Fig. 4a) and, + dye CD19 CAR T cells in the well-established NALM60.02 acute lymphoblastic importantly, after the co-incubation of autologous20 19-28ζ cells with 10–16 30 leukaemia (ALL) model in immunocompromised60 mice (Extended primary samples from patients with ALL15 or chronic lymphocytic Data10 Fig. 1a). CARs encompassing CD28 or 4-1BB0.68 co-stimulatory Tleukaemia cell (CLL) (Fig. 1g, Extended Data Fig. 4b). CAR-induced domains (referred to as 19-28ζ or 19-BBζ, respectively) effectively trogocytosis was observed +with all other tested10 CAR targets, including 20 0.20 40 6 6 CAR Lysis (%) controlled NALM6 cells0.05 at the dose of 0.4 × 10 –1.0 × 10 CAR CD22, B-cell maturation antigen (BCMA)5 and mesothelin (Extended T cells,5 but allowed for frequent leukaemia relapse at the dose of Data Fig. 4c–e).transd. The same antigen escape profile was achieved after 10 0.11 IFN γ (%) 6 0.2 × 10 cells (Fig. 1a, ExtendedGzmB (%) 20 Data Fig. 1b–e). Although both targeting CD22 in vivo (Extended Data Fig.0 5). NALM6 cells, which CAR CAR FACS sorting FACS sorting CAR types of CAR T cell showed limited evidence of exhaustion two weeks express approximately 2,000 CD22 molecules per cell at baseline levels, Tracer 0 + after infusion0 (Extended Data Fig. 1f,0 g), 19-BBζ cells were markedly relapsed with expression of 750 CD22 molecules40:1 per cell20: 1after10: 1treatment5:1 exhausted by the time of relapse, whereas 19-28ζ cells were no longer with 0.2 × 106 22-BBζ cells (Extended Data Fig. 5c). Trogocytic target dye E:T PD-1+LAG-3+TIM-3 (%) detected (Fig. 1b, Extended Data Fig. 1h), consistent with clinical expe- acquisition is thus a general feature of CAR T cells, probably applying rience+ 13–16 and CAR stress test models12. CD19 expression+ was reduced to many if not all antigens. + + 19-28ζ-trog+ + 19-28ζ/19-28ζ-trog 19-BBζ/19-BBζ-trog 19-28ζ/T-CD19+ 19-BB− ζ/T-CD19 19-BBζ-trog in progressing 19-BBζ-treated NALM6 cells, averaging 4,500 mole- Co-culture of sorted trog but not trog CAR T cells with fresh – – 19-28 /19-28 -trogcules– per cell, down19-BB from the/19-BB starting 11,000,-trog which– remain unchanged 19-2819-28ζ cells elicited/T-CD19 the production– of IFN19-BBγ and GzmBζ/T-CD19 by 19-28ζ cells– 19-28ζ-trog 19-BBζ-trog ζ ζ in 19-28ζ relapses and inζ untreated miceζ (Fig. 1c). The loss of CD19 (Fig. 1h). Whenζ stably expressing CD19 at approximately equivalent occurred early on, as it was already present by day 14 and thus occurred levels to those detected after trogocytosis (Extended Data Fig. 6a), both in the presence of abounding CAR T cells (Fig. 1b, c and Extended Data 19-28ζ and 19-BBζ cells engaged in CD19+ T cell killing, more so in the Fig. 1 | Trogocytic antigen extractionFig. 1f). The same promotes patterns were found tumour with CD19 CARs escape. comprising former, consistent withof their CD19 greater effector in NALM6 function12,18 (Fig. CD19–mCherry 1i). The cells (middle) and CAR T cells single chain Fv (scFv) antibody fragments SJ25C116 or FMC6313–15 culture of sorted trog+ and trog− T cells for 6 days revealed increasing a, Tumour burden was monitored using bioluminescence image (right; representative+ of three donors, n = 3 independent samples). UT, −1(Extended− Data2 Fig.− 12). Concurrent with decreased CD19 expression in expression of PD-1, LAG-3 and TIM-3 in trog cells, more so in 19-28ζ (average radiance, photons s tumour cm cells, sra large fraction) in miceof CAR T cellsbearing stained positive NALM6 for CD19 cellsthan 19-BB ζ cells (Fig.untransduced. 1j, Extended Data Fig. 6b).19-del In T cells denotes stably CAR T cells that lacks co-stimulatory and (Extended Data Fig. 3a). Notably, CD19 expression in the retrieved co-expressing CAR and CD19, those that did not succumb to T cell after treatment with CD19 CARNALM6 T cellscells was reversible(n = after6–7 short-term mice culture per (Fig. group; 1d). Because two fratricide killing tendedζ-chain to acquire signalling exhaustion markers domains. (Extended f, Heat map of protein expression (CD19, there was little variation in the expression of6 Cd19 mRNA (Extended Data Fig. 6c–e). CD19 trogocytosis was associated with diminished+ − independent experiments are pooledData Fig. 3b), thesefor findings the 0.2indicated × that 10 a reversible, CAR post-transcrip T cell -dose).cell-surface CAR expression,CD81 both and in vitro CD22) and in vivo, and in co-localized trog and trog 19-28ζ cells and CD19–mCherry- tional loss of CD19 occurred in the presence of CAR T cells. CD19 intracellular CAR and CD19 (Extended Data Fig. 7). NT, non-treated mice. b, Left, expressionCAR didT notcell vary counts when fresh NALM6 (n = cells 3–7 were segregated mice fromper The infusion of ‘fresh’expressing 19-BBζ T cells 10 NALM6 days after the initial cells. 19-BB ζaa, amino acids. g, MFI of CD19 on blasts (left) CAR T cells in transwells, but promptly decreased when T cells were treatment failed to rescue the relapse-prone+ mice (Fig. 2a), suggesting group). Middle, EOMES/T-betco-cultured ratio (Extended (n = Data3–5 Fig. mice 3c, d). CD19 per was notgroup). lost in co-cultures Right, that CD19 density hadand already CAR decreased T to cells a level eluding (right) 19-BB ζderived from patients with ALL (ALL-1 and expression of PD-1, LAG-3 andwith TIM-3 untransduced in T cells CAR or T cells T expressing cells (a non-signallingn = 3–5 CD19 mice threshold per efficacy. 19-28ALL-2)ζ CAR T or cells, with however, CLL did rescue (CLL-1 these and CLL-2) (n = 4 patient samples). 1Center for Cell Engineering, Memorial Sloan Kettering Cancer Center, New York, NY, USA. 2Immunology Program, Sloan Kettering Institute, New York, NY, USA. 3Microchemistry and Proteomics group). D32–D70, days 32–70.Core c Laboratory,, CD19 Sloan Kettering expression Institute, New York, NY, in USA. 4NALM6Quantitative Sciences Unit,cells Stanford from University School of Medicine,h Palo, Percentage Alto, CA, USA. 5Molecular Pharmacology of CAR Program, T cells stained with cell tracer dye and positive Sloan Kettering Institute, New York, NY, USA. *e-mail: [email protected] ζ ζ = for IFNγ or granzyme B (GzmB) (n = 4 donors). i, Cytotoxic chromium mice treated with 19-28 or 19-BB CAR T cells (day 14; n 3–7 mice N A T U R E | www.nature.com/nature per group; days 32–70; n = 3–9 mice per group). d, CD19 expression (51Cr) release assay (representative of 6 donors, n = 3 independent in ex vivo cultured NALM6 cells retrieved from 19-BBζ-treated mice at samples). j, Percentage of CAR T cells co-expressing PD-1, LAG-3 day 0 (D0; time of retrieval) or day 6 (D6) (n = 9 independent samples), and TIM-3 (n = 4 donors). NA, not available (no detectable cells). or NALM6 cell in vitro (n = 4 independent samples). In b–d, cells were P values were determined by two-sided Mann–Whitney U-test collected from bone marrow or extramedullary tumour sites of mice (c, g, h, j) or two-sided Wilcoxon matched-pairs signed-rank test (d). treated with 0.2 × 106 CAR T cells. e, Mean fluorescence intensity (MFI) Data are mean ± s.e.m. relapses (Fig. 2a). Differential antigen sensitivity between the two CARs effector:target ratios may overcome clonal sensitivity thresholds. In was confirmed in models using stable, graded CD19 levels (Fig. 2b, microwells containing one CAR T cell and one NALM6 cell (1:1 E:T; Extended Data Fig. 8). Mice bearing NALM6med or NALM6low tumours Fig. 3a), 19-28ζ cells were more likely to lyse NALM6 cells than 19-BBζ (obtained after monoalleleic or biallelic disruption of the Cd19 gene, cells (57% versus 39% over 24 h; Fig. 3b). The time to tumour cell respectively) showed diminished responses to low-dose CAR therapy, death after the formation of a T cell–NALM6 cell conjugate showed that with 19-28ζ cells consistently inducing longer survival than 19-BBζ 19-28ζ cells killed NALM6 cells after 261 ± 18 min (mean ± s.e.m.), counterparts (Fig. 2b). These studies thus confirmed that a reduc- whereas 19-BBζ cells did so in 569 ± 35 min (mean ± s.e.m.) (Fig. 3b). tion of target antigen density alone can foster CAR T cell resistance. The lesser killing ability of 19-BBζ cells did not result from poorer Together with the promotion of T cell exhaustion (Fig. 1j, Extended antigen recognition, as the frequency of non-lytic stable conjugate for- Data Fig. 6b–e) and fratricide T cell killing (Fig. 1h, i and Extended mation was not less but higher than measured with 19-28ζ cells, and the Data Fig. 6d), CAR target extraction is thus poised to promote time spent in a non-lytic conjugate was greater (Fig. 3c). antigen-low tumour relapse. In microwells containing two CAR T cells and one tumour cell (2:1 E:T; Notably, 19-BBζ cells were able to control wild-type NALM6 when Fig. 3d), the frequency of tumour lysis increased to about 75% for administered at a higher dose (Fig. 1a). We hypothesized that higher both 19-28ζ and 19-BBζ cells (Fig. 3e). Although this increase could

N A T U R E | www.nature.com/nature Patient #4: 10 yo, refractory B-ALL, High burden disease, Complete response followed by CD19+CD22+ relapse

Day 6 Day 14 Day 28 2 months 3 months Blood Blood Blood Bone Marrow Blood Ki-67 CD27 CD38 T-bet CD279 (PD1) CAR-T Product

CD8 Day 12 Day 28 tSNE2 tSNE1 Low High Gated on CAR+ T cells AP-1 Transcription Factors are part of Exhaustion

ü c-Jun OE increases IFNg and IL-2 expression in exhausted HA-28z T cells, but has no effect in non- exhausted T cells

Lynn et al, In Revision Alternative Targets Have Emerged: CD22

71% MRD Negative Response Rate at Dose level 2+

Fry et al., Nature Med, 2017 Antigen density has emerged as a salient question in the CD22 CAR

Fry et al., Nature Med, 2017 Bispecific Targeting in ALL: CD19 and CD22

• Two trials of CD19/22-bispecific CAR at Stanford have launched: Pediatric and Adult • First-in-human testing of a bispecific CAR bi-specific chimeric antigen receptor • Relapsed refractory B cell malignancies • Primary objective: safety and CAR-T cell target 2 feasibility Target Bearing Cancer Cell • Secondary objectives: Response rate, Incidence of target 1 CD19 negative escape • Heavy correlative science 6 patients infused, 4 in remission at Day 27

Disease Age Burden Dose Toxicity Day 27 response Indication

17 yo F Isolated BM MRD+ by NGS at DL1: 1x106 /kg Gr 1 CRS/ MRD- at Day 27 relapse infusion CRES

2 yo M MLLr CNS/BM MRD+ by NGS at DL1: 1x106 /kg Gr 1 CRS MRD- at Day 27, s/p relapse infusion; CNS2 HSCT 16 yo F Primary MRD+ 0.15 by flow DL1: 1x106 /kg Gr 1 CRES MRD- at Day 27 refractory (HA)

8 yo M BM relapse 6% disease by DL1: 1x106 /kg Gr 2 CRS Morphologic flow remission at Day 27

2 yo F BM relapse 15% disease DL1: 1x106 /kg Gr 2 CRS Non-responder

12 yo F 2nd BM relapse 7% disease DL1: 1x106 /kg Gr 1 CRS TBD So many questions remain…. How much disease is best? When to use CAR T cells? How to use CAR T cells? Only in Patients After Minimal Disease? A lot with Residual Only at relapse? Chemotherapy for Diagnosis? Disease? All? Bridge to Transplant? CAR T cells for Solid Tumors DIPG and beyond

Mount and Majzner, Nat Med, 2018 DIPG: in need of fresh thinking

Diffuse Intrinsic Pontine Glioma – Devastating brainstem tumor of childhood, 300 cases/yr in US – Standard of care: median life expectancy <12mo. – No improvements in survival since introduction of targeted

radiotherapy Mackay et al, Cancer Cell 2017 Cell surface screen identifies GD2 as immunotherapy target in DIPG GD2.4-1BBz CAR R. Majzner; C. Mackall xenografts Single-dose intravenousGD2-CAR TcelltherapyclearsDIPG

SU-DIPG6 Mount, Majzner etal.Nature Med.2018 SU-DIPG13P* 100 CD19-CAR n=22 GD2-CAR n=23

p <0.0001 50 Percent surviva l

0 0 10 20 30 40 50 DPT GD2 CAR T cells widely infiltrate the brain after intravenous administration GD2-CAR Trial in Pediatric Solid Tumors Slated to Open later in 2019

• Neuroblastoma, osteosarcoma, DIPG • GD2-CAR Incorporates a suicide domain that can be “drug activated” in the event of untoward toxicity • Children with DIPG will be monitored closely by a team of immunotherapists, neurointensivists and neurosurgeons • We will seek to identify the most appropriate patients in terms of disease burden/life expectancy Other CAR T cell therapies in the clinics Malignancy Target Multiple Myeloma BCMA, CD38, CD138 Lung cancer HER2 Pancreatic cancer CEA Brain tumors EGFR, HER2, GD2, IL13ra2 AML CD33, CD123 Hodgkin Lymphoma CD30, CD20 Pancreatic Mesothelin Neuroblastoma CD171, GD2 T cell ALL, T LL CD5 Solid tumors Glypican 3, mesothelin, ROR1, CD70 Immune Checkpoint Inhibitors in Cancer

Taking the breaks off the immune system

NYT Oct 1, 2018 Immune Checkpoint Inhibitors have made a huge impact in oncology

Nature Rev Cancer Wei et al, Cancer Disc 2018 Pediatric tumors have low mutational burden

Lawrence et al, Nature, 2013 Checkpoint inhibitors work best when the target malignancy is highly mutated

Le et al, NEJM, 2015 Nivolumab in Relapsed or Refractory Hodgkin’s Lymphoma of 87% and a rate of progression-free survival of ASCT failure and No ASCT and No brentuximab 86% at 24 weeks. Adverse events were mainly of brentuximab failure brentuximab failure grade 1 or 2. The rate of adverse events was sim- A Response Characteristics ilar to that in trials of nivolumab in patients with 4 Patient No. solid tumors. Given the limited therapeutic op- 20 First complete response tions for patients with Hodgkin’s lymphoma 19 First partial response 18 Therapy duration whose disease progresses after autologous stem- Transplantation 17 21-23 Ongoing response cell transplantation and the relatively short- 16 lived responses to brentuximab after relapse,24 15 nivolumab-mediated PD-1 blockade may repre- 14 13 sent a promising targeted treatment for these 12 patients. 11 10 The new england journal of medicine The frequent chromosome 9p24.1 amplifica- 9 tion and associated PD-1 ligand overexpression 8 in Hodgkin’s lymphoma and the pronounced but 7 A 6 Figure 2. Genetic and Immunohistochemical Analyses ineffective inflammatory response seen in in- 5 volved lymph nodes provided a compelling ratio- 4 100 kb of PDL1 and PDL2 Loci, PD-L1 and PD-L2 Protein nale for evaluating the efficacy of PD-1 blockade 3 Expression, and Epstein–Barr Virus Status in Patients 2 with Hodgkin’s Lymphoma. in patients with relapsed or refractory disease.chr9:5,270,000 1 chr9:5,700,000 PDL1 PDL2 In this study, all the patients with available tu- 0 8 16 24 32 40 48 56 64 72 80 88 Shown are the results of analyses of the PDL1 and PDL2 loci and PD-L1 and PD-L2 protein expression in mor specimens had concurrent gain of theOr PDL1 in some cancersRP11-599H20 that upregulateWeeks RP11-635N21 PD-L1 or PD-L2 and PDL2 loci, increased expression of the PD-1 Reed–Sternberg cells. Panel A shows the location and ligands, and evidence of active JAK-STAT signal- B Change in Tumor Burden color labeling of bacterial artificial chromosome B clones used for the fluorescence in situ hybridization ing. In this group of patients, the incidence of a 10 Stable Complete DiseasePDL1/2 Gain Partial ResponsePDL1/2 AmplificationResponse (FISH) assay for PDL1 (red) and PDL2 (green) on copy-number gain in PDL1 and PDL2 was higher 0 than in previously reported series of patients −10 CEP9 chromosome 9p24.1. Representative images obtained with newly diagnosed Hodgkin’s lymphoma,14,25 −20 PD-L1 from patients show a copy-number gain in PDL1 and suggesting that this disease-specific genetic al- −30 PD-L2 PDL2 (Panel B, left), with six green–red (yellow over- teration may have adverse prognostic signifi- −40 lap) signals (signifying a fusion signal), as compared −50 with three centromeric signals (aqua), or PDL1 and cance. The low rate of EBV positivity (1 of 10 Change (%) −60 PDL2 amplification (Panel B, right), with more than patients) that was observed is consistent with −70 three times as many green–red (yellow overlap) sig- the low predominance of EBV in Hodgkin’s lym- −80 nals as centromeric signals (aqua). In Panel C, the ex- phoma of the nodular-sclerosis type.26 −90 pression of PD-L1 (upper row) and PD-L2 (lower row) In available biopsy samples obtained from the −100 is indicated by brown staining in Reed–Sternberg cells patients, tumor-infiltrating T cells largely ex- Individual Patient Data (N=23) obtained from the same patients as in Panel B. Arrows pressed low levels of PD-1 on standard immunoC - Ansell et al NEJM 2015 Figure 1. Response Characteristics and Changes in Tumor Burden in Patients indicate malignant cells. PD-L1 is evaluated in con- histochemical analysis. Previous studies have with Hodgkin’s Lymphoma Receiving Nivolumab. junction with PAX5 to identify PAX5-positive cells suggested that PD-1 blockade selectively en- Panel A shows the response onset and duration for the 20 study patients (shown in red in the upper row). PD-L2 is assessed in hances the function of CD8+ T cells that have who had a response to treatment with nivolumab. The color of each bar in- association with phosphorylated signal transducer and low or intermediate, rather than high, levels of dicates whether previous autologous stem-cell transplantation (ASCT) or AX5 activator of transcription 3 (pSTAT3), which reflects PD-1 expression.27 The levels of PD-1 on tumor- brentuximab therapy had failed in that patient. The length of the bar shows the time until the patient had a complete response or a partial response, Janus kinase–STAT activation (shown in red in the infiltrating T cells were significantly less predic- along with the duration of the response. Six patients elected to discontinue lower row). The scale bars represent 50 µm. Panel D tive of response to nivolumab therapy than was PD-L1/ P the study in order to undergo stem-cell transplantation after having a re- shows PD-L1 and PD-L2 protein expression and status PD-L1 expression on solid tumors in recent sponse to nivolumab. Eleven patients continued to have a response at the with respect to Epstein–Barr virus–encoded messen- 6 time of this writing (indicated by an arrowhead). Panel B shows the per- clinical trials, findings that are consistent with ger RNA (EBER) in study patients who could be evalu- our results. centage reduction in tumor burden from baseline in all 23 study patients. Two patients met the criteria for a complete response without having a 100% ated. All the patients who were included in the analy- The frequent clinical responses to nivolumab decrease in tumor burden. One patient with a partial response had a 99% sis had structural bases for increased copy numbers in therapy in heavily pretreated patients with re- decrease in tumor burden but had positive results on positron-emission PDL1 and PDL2 on chromosome 9p24.1, including ex- lapsed or refractory Hodgkin’s lymphoma and tomography. tra copies of 9p (polysomy 9p), copy gain in PDL1 or genetic alterations of the PD-1 ligand loci high- PDL2, or amplification in PDL1 or PDL2 (Table S2 and Fig. S2 in the Supplementary Appendix). Epstein–Barr PD-L2/pS TA T3 virus status was evaluated by means of a FISH assay. HRS denotes Hodgkin’s Reed–Sternberg, and IHC im- munohistochemical. n engl j med 372;4 nejm.org january 22, 2015 317 The NewD England Journal of Medicine Downloaded from nejm.org at STANFORD UNIVERSITY Patienton January No. 4, 2017. For personalCytogenetic use only. No other IHC-positiveuses without permission.Nuclear EBER light the importance of the PD-1 immune eva- Copyright © 2015 Massachusetts Medical Society. All rights reserved. pSTAT3 Alterations HRS cells sion pathway and the genetically defined sensi- Polysomy PDL1/2 PDL1/2 PD-L1 PD-L2 9p Gain Amplification tivity to PD-1 blockade in this disease. 1 + – – + + + – Supported by Bristol-Myers Squibb, by grants from the Na- 2 + – – + + + – tional Institutes of Health (U54CA163125 and P01AI056299, to 3 + – – + + + – Dr. Freeman; and R01CA161026, to Dr. Shipp), and by a grant 4 + + – + + + – from the Miller Fund (to Dr. Shipp). 5 + + – + + + – Dr. Ansell reports receiving grant support from Seattle Genet- 6 + + – + + + + ics, Celldex Therapeutics, Millennium, Idera, and Regeneron; 7 + + + + + + – Dr. Lesokhin, receiving consulting fees and grant support from 8 + + + + + + – Bristol-Myers Squibb; Dr. Halwani, receiving grant support 9 – + + + + + – from Seattle Genetics, Millennium, Kyowa Hakko Kirin, AbbVie, Genentech, and Bristol-Myers Squibb; Dr. Freeman, receiving 10 – – + + + + – royalties from patents related to PD-1, PD-L1, and PD-L2 pathways (US 6808710, US 6936704, US 7038013, US7101550, US 7432059,

318 n engl j med 372;4 nejm.org january 22, 2015 The New England Journal of Medicine Downloaded from nejm.org at STANFORD UNIVERSITY on January 4, 2017. For personal use only. No other uses without permission. Copyright © 2015 Massachusetts Medical Society. All rights reserved. Bouffet et al recurrence with a 6 3 7 mm nodule in the initial tumor bed and a second bMMRD tumors (n = 5) did not exhibit hypermutation (mean, 16 3 17 mm lesion in the left temporal lobe. 40 6 SD, 18). Importantly, bMMRD GBMs had a significantly Six months earlier, her brother, age 3.5 years, was diagnosed with a higher mutational load than sporadic pediatric and adult gliomas right frontoparietal GBM, which was treated with surgery, focal irradi- and all other brain tumors (P ,.001; Fig 2A). To test the extent to ation, and temozolomide. Ten months after the initial diagnosis, sur- veillance MRI revealed an asymptomatic diffuse multinodular GBM which hypermutation translates to a strong neoantigen signature, a recurrence. Both tumors harbored TP53 mutations and lack of MGMT current predictor of response to immune checkpoint inhibition, we promoter hypermethylation, which are poor prognostic markers in performed genome-wide somatic neoepitope analysis using similar childhood GBM.16-18 algorithms previously used for melanoma, lung, and colon Importantly, both children exhibited cafe-au-lait´ spots, a shared feature cancers.9,14,15 For each study, we compared our cohort of tumors fi of neuro bromatosis type 1 and bMMRD, prompting clarifying molecular with other tumors that were reported to respond to immune investigations. These revealed germline homozygous c.2117delA, p. checkpoint inhibitors (Fig 2B). Strikingly, bMMRD GBMs had a Lys706SerfsX19 mutation in PMS2, loss of PMS2 staining in the tumor and fi normal tissue (Fig 1B), and normal NF1 germline testing. Together, these signi cantly higher number of predicted neoantigens, whereas clinical and molecular tests confirmed a diagnosis of bMMRD.3 other tumors responded with a fraction of the neoantigens found in our patients (P , .001; Fig 2B). The mean neoantigen load was seven to 16 times higher than those of immunoresponsive mel- RESULTS anomas, lung cancers, and microsatellite-unstable GI cancers. Because of this information, the index patient and her To examine whether immune checkpoint inhibitors would be brother’s tumors were sequenced using the rapid SickKids Cancer applicable for bMMRD cancers, we surveyed the extent of Sequencing Program (Toronto, Ontario, Canada) protocol, KiCS,5 hypermutation across bMMRD tumors form various tissues. as a part of the international bMMRD consortium effort Exome sequencing of 37 cancers collected from the bMMRD (Appendix Table A1, online only). Both tumors harbored driver consortium revealed that all malignant tumors (n = 32) were mutations in POLE (Appendix Fig A1, online only) and were found hypermutant. Although bMMRD brain tumors had the highest to harbor 24,680 and 21,919 mutations per exome, respectively. mutational load resulting from secondary polymerase mutations Given the rationale that tumors with molecular signatures similar (mean, 17,740 6 standard deviation [SD], 7,703), all other high- to those developing in children with bMMRD are responsive to grade tumors were hypermutant, harboring more than 100 exonic immunomodulators and the lack of effective therapies for children mutationsBi-allelic (mean, 1,589 6 SD,Mismatch 1,043; Fig 2A). Lower-grade Repairwith recurrent disseminated Deficiency GBM, the index patient’s parents were • ABHomozygous germline mutations in one mismatch repair gene (PMS2, MLH1, MSH2, MSH6) 2.0 bMMRD Child brain Adult brain • Most penetrant cancer predisposition syndrome described P < .001 P < .001 to date (100% cancer incidence within1.5 first two decades of life)

• Not1000 uncommon in populations with high consanguinity 1.0 100• Estimated 40% of pediatric GBM in Jordan is associated

10 with bMMRD 0.5 Total Neoantigen Proportion Exonic Mutations (log) • High grade bMMRD tumors are “hypermutant”

• high levels of single nucleotide variants0.0 (e.g. neoantigens) bMMRD Melanoma Lung Colon L/L PA MB PA MB LGG GBM ATRT DIPG GBM GBM • >>>PNET/MBGI cancers 100 mutations/exome Tumor Type

Other malignant Low-grade tumors • Estimated acquisitionCancer Type of 600 new mutations/cell division

Fig 2. Tumor mutation and neoantigen analysis. (A) Boxplot comparing the number of mutations per tumor exome in several biallelic mismatch repair deficiency (bMMRD) cancer types with pediatric and adult brain tumors. (B) Ratio of the number of neoantigens found in immunoresponsive tumors from melanoma (n = 27), lung cancer (n = 14), and colon cancer (n = 7) data sets compared with median number of neoantigens in bMMRD glioblastoma multiforme (GBM; n = 13). ATRT, atypical teratoid rhabdoid tumor;Bouffet DIPG, et al. JCO 2016 diffuse intrinsic pontine glioma; L/L, leukemia/lymphoma; LGG, low-grade glioma; MB, medulloblastoma; PA, pilocytic astrocytoma; PNET, primitive neuroectodermal tumor.

2208 ©2016byAmericanSocietyofClinicalOncology JOURNAL OF CLINICAL ONCOLOGY Information downloaded from jco.ascopubs.org and provided by at STANFORD UNIV MEDICAL CENTER on September Copyright © 2016 American18, 2016 Society from of 171.65.65.46Clinical Oncology. All rights reserved. Response of Recurrent GBMs to Nivolumab in Siblings with bMMRD

Sibling #1 ~24,000 mutations/ exome

Sibling #2 ~22,000 mutations/ exome Bouffet et al. JCO 2016 PD-1 inhibition may have limited applicability to pediatric cancers PD-L1 Expression by Pediatric Tumor Type

100 PD-L1 + (%) PD-L1 - (%)

50 Percentage

0

Ewing sarcomaOsteosarcoma (25) (20) MedulloblastomaNeuroblastoma (40) (118) Burkitt lymphoma (10) Rhabdomysarcoma (53)

Glioblastoma multiforme (14) Majzner et al., Cancer, 2017 In children, responses seen in Hodgkin’s Disease 1

Intervention start Did not start treatment t Progressive Disease e c Stable Disease u b j

S Partial Response Complete Response Continued treatment

Disease Cohort Neuroblastoma Osteosarcoma Rhabdomyosarcoma Ewing sarcoma Hodgkin lymphoma Non-Hodgkin lymphoma Melanoma Neuroblastoma (MIBG w/o RECIST)

0 5 10 15 20 25 30 Time since enrollment (months) Each bar represents one subject in the study. There is more to learn about how the immune system interacts with cancer

Keren et al, Cell, 2018 Objectives

• Childhood leukemia leads the way for targeted T- cell therapies • What are chimeric antigen T cells • Use in heme malignancies and solid tumors • Limitations in Efficacy • Toxicity concerns • Immune Checkpoint Inhibition for anti-cancer effect • Successes and Challenges Thank you! • Cancer Cellular Therapy • Correlative Science Unit • Crystal Mackall • Sean Bendall • David Miklos • Bita Schaff • Liora Schultz • Rohit Gupta • Lori Muffly • Holden Maecker • Tina Baggott • Davis Lab: Pablo Domizi, Astraea Jager • Sharon Mavroukakis • David Barrett • Katie Kong • Cell Therapy Lab • Monje Lab: Michelle Monje • Chris Mount • LCGM • Mackall Lab • Apheresis • Patients and Families • Administration