Inclusion of an Igg1-Fc Spacer Abrogates Efficacy of CD19 CAR T

ORIGINAL ARTICLE Inclusion of an IgG1-Fc spacer abrogates efﬁcacy of CD19 CAR T cells in a xenograft mouse model

H Almåsbak1, E Walseng2,3,8, A Kristian4,5,8, MR Myhre1, EM Suso1, LA Munthe6,7, JT Andersen6,MYWang1, G Kvalheim1, G Gaudernack2 and JA Kyte1

Cancer therapy with T cells expressing chimeric antigen receptors (CARs) has produced remarkable clinical responses in recent trials, but also severe side effects. Whereas most protocols use permanently reprogrammed T cells, we have developed a platform for transient CAR expression by mRNA electroporation. This approach may be useful for safe clinical testing of novel receptors, or when a temporary treatment period is desirable. Herein, we investigated therapy with transiently redirected T cells in vitro and in a xenograft mouse model. We constructed a series of CD19-specific CARs with different spacers and co-stimulatory domains (CD28, OX40 or CD28-OX40). The CAR constructs all conferred T cells with potent CD19-specific activity in vitro. Unexpectedly, the constructs incorporating a commonly used IgG1-CH2CH3 spacer showed lack of anti-leukemia activity in vivo and induced severe, partly CD19-independent toxicity. By contrast, identical CAR constructs without the CH2-domain eradicated leukemia in vivo, without notable toxicity. Follow-up studies demonstrated that the CH2CH3-spacer bound soluble mouse Fcγ-receptor I and mediated off-target T-cell activation towards murine macrophages. Our findings highlight the importance of non-signalling CAR elements and of in vivo studies. Finally, the results show that transiently redirected T cells control leukemia in mice and support the rationale for developing an mRNA-CAR platform.

Gene Therapy (2015) 22, 391–403; doi:10.1038/gt.2015.4; published online 5 February 2015

INTRODUCTION electroporation.21–27 The mRNA strategy represents a safer Adoptive T-cell therapy holds considerable promise as a novel alternative for first-in-man studies with novel receptors, and may concept for cancer treatment. Recent studies with T cells also be useful in patients where a limited treatment period is retargeted with chimeric antigen receptors (CARs) against CD19 desirable. We have previously reported a GMP platform for have induced remarkable clinical responses in heavily pre-treated transient redirection of T cells with mRNA encoding the CD19- patients with acute lymphocytic leukemia, chronic lymphocytic specific CAR fmc63-IgG1Fc-CD28-OX40-CD3ζ (19-IgFc-28OXζ).21 1–6 leukemia or B-cell lymphoma. CARs are recombinant receptors We demonstrated that this method resulted in efficient transfec- comprising an extracellular antigen binding domain from a tion of 490% of the cell population, highly uniform CAR monoclonal antibody and signalling domains from the T-cell expression and potent killing of CD19+ cell lines, primary leukemia 7 receptor (TCR) complex. The second generation CAR constructs cells and lymphoma cells. include a signalling domain from a co-stimulatory molecule (for Here, we report in vitro and in vivo studies of a series of fmc63- example, CD28, 41BB or OX40), and third generation CARs 8–11 based CARs, using a xenograft model of NOD-SCID gamma-c comprise two co-stimulatory domains. knockout (NSG) mice inoculated with the human leukemia cell line The striking clinical responses achieved with CD19 CAR T cells NALM-6. We unexpectedly observed that T cells transfected with have provoked a surge of interest in developing CARs and TCRs 21 our originally preferred 19-IgFc-28OXζ CAR construct exerted against new targets. Yet, serious toxicity observed in several trials only marginal effect on the tumor growth in vivo, in spite of the has highlighted the need to develop safer strategies for clinical 4,6,12–17 strong in vitro activity. Moreover, this construct caused serious testing. Most investigators have used viral vectors perma- fi nently integrating the receptor sequence into the T-cell genome. toxicity. On the basis of this surprising nding, we conducted a fl This approach may give durable anti-cancer activity after a single series of studies evaluating the in uence of alternative CAR T-cell infusion, but also represents a safety risk. As the gene- signalling domains and spacer variants. The results attribute the modified T cells persist and expand in the patient, serious adverse striking discrepancy between in vitro and in vivo activity to the events may develop.13–15,17 Insertional mutagenesis also remains a inclusion of a commonly used IgG1-based CAR spacer. Further, we possible concern, though not materializing in clinical trials so demonstrate the ability of T cells transiently expressing optimized far.18–20 To circumvent these issues, we and others are developing CD19 CARs to effectively eradicate leukemia in a mouse xenograft methods for transient CAR/TCR expression based on mRNA model, without detectable toxicity.

1Section for Cell Therapy, Department of Oncology, Oslo University Hospital Radiumhospitalet, Oslo, Norway; 2Department of Immunology, Oslo University Hospital Radiumhospitalet, Oslo, Norway; 3Department of Cancer Biology, Scripps Research Institute, Jupiter, FL, USA; 4Department of Tumor Biology, Oslo University Hospital Radiumhospitalet, Oslo, Norway; 5KG Jebsen Center for Breast Cancer Research, Institute for Clinical Medicine, University of Oslo, Oslo, Norway; 6Centre for Immune Regulation and Department of Immunology, Oslo University Hospital Rikshospitalet, Oslo, Norway and 7Institute of Clinical Medicine, University of Oslo, Oslo, Norway. Correspondence: Dr JA Kyte, Section for Cell Therapy, Department of Oncology, Oslo University Hospital Radiumhospitalet, Mail Box 4950 Nydalen, Oslo, 0424, Norway. E-mail: [email protected] 8These authors contributed equally to this work. Received 28 June 2014; revised 3 November 2014; accepted 6 January 2015; published online 5 February 2015 IgG1 spacer abrogates CAR efficacy in vivo H Almåsbak et al 392 RESULTS differences between the IL-2 +/ − groups (P = 0.43, t-test). Three days CAR19-IgFc-28OXζ T cells failed to control leukemia in vivo and after T-cell transfer, the study animals started to regain weight. The caused serious toxicity results suggested that there was a considerable component of off- On the basis of our promising in vitro studies of a third generation target toxicity and that this toxicity could not be avoided by 19-IgFc-28OXζ CAR,21 we conducted an in vivo study evaluating omitting adjuvant IL-2. this CAR construct in NSG mice challenged with the NALM-6 cell line. Transiently redirected T cells (mRNA electroporated) were Superior in vivo functionality of a second generation CAR with compared with constitutively redirected T cells (retroviral trans- short spacer duction). Separate series with high or low tumor load were set up We next asked whether T cells expressing lower densities of the to model advanced disease and minimal residual disease (MRD), 19-IgFc-28OXζ CAR could retain killing capacity while causing less respectively (Figures 1a and b). Control groups of mice were toxicity. The mRNA technology allows for controlling CAR treated with mock-transfected T cells (mock T cells) and IL-2. Flow expression levels by adjusting the concentration of mRNA at cytometry analysis showed CAR expression in 68% of the stably electroporation.21 We compared T cells expressing high (CARhigh; transduced T cells and 495% of the mRNA-transfected T cells 100 μg mRNA ml − 1) and low (CARlow;40μg mRNA ml − 1) CAR (data not shown). surface levels. Further, we compared with a second generation In the advanced disease model, bioluminescence imaging on CD19 CAR without OX40 and with a minimal spacer replacing the day 4 (before T-cell injection) indicated similar leukemia progres- IgG1Fc domain (19-short-28ζ) (Supplementary Figure 1a). The sion in the CAR T-cell-treated groups (transient or stable CAR surface CAR expression correlated well with the mRNA concentra- expression) (Figure 1c). Strikingly, the bioluminescence signals one tions at electroporation (Supplementary Figure 1b). In vitro day after T cell injection were significantly lower in control mice functionality assays showed that T cells expressing the different receiving mock T cells, compared with the animals treated with CAR constructs (19-IgFc-28OXζ or 19-short-28ζ) responded with CAR-expressing cells (P = 0.002, analysis of variance/Student New- similar levels of interferon (IFN)-γ production and degranulation man Keuls (ANOVA/SNK)). Three days after T cell injection, we (Supplementary Figure 1c). This applied both to CD4+ and CD8+ could readily detect T cells in mice receiving mock T cells, whereas T cells. Further, we observed that a fraction of the mock- no remaining T cells could be detected in mice treated with CAR transfected T cells displayed activity against the ff-Luc-NALM-6 T cells (data not shown). All CAR T-cell-treated mice developed (Supplementary Figure 1c). ⩾ 15% decrease in body weight and signs of distress, including We injected 24 NSG mice with 1 × 106 ff-Luc-NALM-6 cells at day ruffled fur, shivering and lethargy. This required euthanasia 0 (Figure 2a). Subsequently, three mouse groups were treated with according to internal guidelines. By contrast, the mice treated CAR T cells expressing either (i) 19-IgFc-28OXζ (H), (ii) 19-IgFc-28OXζ with mock T cells appeared unaffected. (L) or (iii) 19-short-28ζ (H). Leukemic control mice were treated with The transfer of CAR T cells was also accompanied by severe mock-transfected T cells or no T cells. Further, healthy mice received toxicity in the MRD model. However, after an initial weight loss, 19-IgFc-28OXζ (H) or 19-short-28ζ (H) CAR T cells. The biolumines- four out of five mice treated with CAR mRNA-redirected T cells cence data showed that mock T cells exerted a moderate anti- recovered and survived (Figure 1d). In contrast, all mice treated leukemia activity, compared with the leukemic controls not treated with virally transduced T cells had to be killed within 1 week with T cells (Figures 2b and c; Po0.05, ANOVA/SNK). Strikingly, only owing to weight loss and lethargy. Imaging at day 15 indicated the 19-short-28ζ CAR provided the T cells with significant additional that the CAR-mRNA T cells exerted increased antitumor activity anti-leukemia activity (Po0.05, ANOVA/SNK). The 19-IgFc-28OXζ compared with mock T cells (P = 0.002, t-test) (Figure 1e). The variants (H and L) were less efficient than mock-transfected T cells mock T cells also appeared to reduce leukemia burden in in controlling leukemia (Figures 2b and c; Po0.05, ANOVA/SNK). comparison with a control that had not received T cells. Furthermore, we found a greater weight loss in mice treated with 19-IgFc-28OXζ (H) T cells compared with the mock T cells (Po0.05, The toxicity caused by CAR 19-IgFc-28OXζ T cells was partly ANOVA/SNK), suggesting considerable toxicity (Figure 2d). Even off-target healthy control mice receiving these CAR T cells lost more weight than the other groups, in line with our previous experiment The adverse effects caused by the CAR 19-IgFc-28OXζ T cells were pointing to off-target toxicity. most dramatic in the advanced disease model, suggesting that the toxicity was partly target dependent, but was also considerable in the MRD model. To investigate whether the toxicity was in part CAR T cell therapy induced high IFN-γ serum levels dependent + off-target, and whether the toxicity could be avoided by omitting on CD19 targets adjuvant IL-2, we conducted a follow-up study injecting CAR We analyzed cytokine levels in serum drawn from mice at day 23, mRNA T cells +/ − IL-2 into leukemia-free mice (Figure 1f). The that is, 2 days after the transfer of 20 × 106 T cells (Figure 2e). IFN-γ mice developed approximately 10% weight loss with nadir 2 days was the dominant cytokine of those measured. The IFN-γ serum after T-cell transfer, together with lethargy. There were no detectable levels of CAR 19-IgFc-28OXζ- and CAR 19-short-28ζ-treated mice

Figure 1. Severe toxicity and low anti-leukemia activity of CD19-IgFc-28OXζ CAR T cells in murine xenograft model. Donor T cells were transiently (mRNA transfection) or permanently (retroviral transduction) redirected with a CD19-IgFc-28OXζ CAR. Mock-transfected donor T cells were used as controls. (a) Advanced disease model: 14 NSG mice (mRNA CAR: n = 5; retroviral CAR: n = 5; mock: n = 4) were inoculated with 20 × 106 ffLuc-NALM-6 cells 4 days before injection of 20 × 106 T cells intravenously and 1000 U IL-2 intraperitoneally. (b) Minimal residual disease model: 15 NSG mice (mRNA CAR: n = 5; retroviral CAR: n = 5; mock: n = 4; no T cells: n = 1) were inoculated with 0.25 × 106 ffLUc- NALM-6 cells 4 h before transfer of 20 × 106 T cells and 1000 U IL-2. (c) Advanced disease model, efficacy: Bioluminescence imaging was performed prior to (day 4) and 1 day after (day 5) injection of T cells. Left: Photon-density heat maps of three representative animals from each group. Right (bar chart): Mean bioluminescence signals (counts per second per cm2) for each group. Error bars indicate standard deviation (s.d.). (d) MRD model, toxicity: Mean weight curves (± s.d.) for surviving mice in the three treatment groups. (e) MRD model, efficacy: Bioluminescence imaging at day 15 after T-cell transfer on all surviving mice (mRNA: n = 4; mock: n = 4; no T cells: n = 1). Mean bioluminescence (counts per second per cm2)(± s.d.) for each group is shown. All five mice in the retroviral CAR group and one of the five mRNA CAR mice had been killed within day 7 owing to toxicity. (f) Follow-up study to investigate the observed toxicity: Eight healthy, non leukemic mice were treated with 20 ×106 mRNA CAR-T cells with (n = 4) or without (n = 4) adjuvant IL-2 intraperitoneally. The graph shows relative weight loss (weight loss/initial body weight) over time.

Gene Therapy (2015) 391 – 403 © 2015 Macmillan Publishers Limited IgG1 spacer abrogates CAR efficacy in vivo H Almåsbak et al 395 with extensive tumor burden was high and comparable results showed that redirected T cells acquired anti-leukemia (251 ± 94 ng ml − 1 and 334 ± 87 ng ml − 1, respectively). The IFN-γ activity and did not reveal any difference between the included levels were 100-fold lower in mice with only residual leukemia, CARs. We have previously observed that fmc63 CARs without suggesting that cytokine secretion was dependent on tumor co-stimulatory domains (first generation CARs) are considerably burden. In the CAR 19-IgFc-28OXζ (L) group, cytokine levels were less potent (data not shown). Among the second and third low, consistent with a lack of activity in T cells with low CAR generation constructs compared here, high fractions of both CD4+ expression. and CD8+ cells produced IFN-γ (CD4: range 64–97% and CD8: range 73–98%). The CD107a degranulation percentage was higher + – + – Toxicity and lack of CAR activity in vivo were related to the IgG1-Fc among CD8 (range 79 98%) than CD4 T cells (range 37 51%). spacer Co-cultured T cells were also stained with annexin-V to estimate the apoptotic fraction. Indeed, a high fraction of the CD8+ and We then asked whether the observed toxicity and in vivo + + – – fi CD4 cells were annexin-V (range 47 76% and 19 33% inef ciency of the third generation CAR were caused by the respectively, Figure 4g). We have repeatedly observed that a potent signaling domains (CD28-OX40-ζ) or by the IgG1-based ζ considerable fraction of CAR T cells undergo apoptosis, as spacer. To address this question, we compared the 19-short-28 measured by annexin-V staining, upon encounter with CD19+ CAR with a CAR with the same signaling units, but including the target cells. However, the apoptotic fraction varied somewhat with IgG1-based spacer (CD19-IgGFc-28ζ). Leukemic mice (n = 3 in each 6 target cell line and was generally slightly lower than that seen in group) were treated with a single dose of 10 × 10 CAR T cells Figure 4g (data not shown). Apoptosis induction was linked to the ζ (Figure 3). Mice treated with CAR 19-IgFc-28 T cells lost on degree of T-cell activation; we observed a higher apoptotic average 9% of their body weight 3 days after T-cell transfer and fraction in T cells with stronger functional activity (data not shown ζ appeared lethargic, whereas mice treated with CAR 19-short-28 and 21). cells lost only 2% body weight and seemed healthy. Biolumines- cence imaging of the mice 19 days post treatment again T-cell therapy of leukemic NSG mice suggests superiority of a CAR demonstrated that the 19-short-28ζ CAR T cells efficiently with short spacer eradicated leukemia (Figure 3). By contrast, the 19-IgFc-28ζ CAR T cells were inefficient. We then compared the in vivo potency of T cells expressing short (hinge only) or half (hinge-CH3) spacers combined with either CD28-ζ or CD28-OX40 − ζ endodomains. Twenty-six NSG mice Similar in vitro functionality of CAR constructs independent of were inoculated with NALM-6 cells and treated with CAR T cells spacer and endodomains (short-28, half-28OX, half-OX or half-28), or mock-transfected We next investigated the influence of different spacer variants. T cells (Figure 5a). The T cells prepared for injection were The original IgG1-based spacer comprised the CH3CH2 domains characterized in vitro and exhibited similar CAR expression levels and a 12aa hinge. We constructed the nine possible combinations independent of endodomain configuration (Supplementary Figure 2a). when pairing three signaling domain variants (CD28OX40ζ, OX40ζ When the T cells were co-cultured with NALM-6 cells, all CAR ζ or CD28 ) with three spacer designs: (i) full (hinge-CH2CH3) (ii) half constructs provided T cells with similar cytotoxic activity (hinge-CH3) and (iii) short (12aa hinge only) (Figure 4a). We (Supplementary Figure 2b). This also applied to the full-spacer transfected the T cells with equimolar amounts of mRNA encoding 28OX CAR that was included as a control. The CAR-expressing the different constructs and found that all CARs were well T cells, but not mock T controls, secreted IFN-γ, IL-2, TFN-α and expressed (Figure 4b). The three full-spacer CARs showed higher IL-10 (Supplementary Figure 2c). expression levels than the half/short spacer constructs. The CAR In the NALM-6-bearing mice, the T cells with short or half spacer expression was independent of endodomain configuration. We CARs all exerted considerable and statistically significant anti- also found that the CAR decay rates indicated similar receptor leukemia activity, compared with mock-transfected T cells (Figures stability, independent of spacer configuration and signalling 5b and c; Po0.001; one way ANOVA/SNK). No signs of toxicity domains (Figure 4c). The mRNA concentration at electroporation were detected through regular weight measurements and correlated positively with CAR expression levels (not shown) and monitoring. Bioluminescence imaging suggested that the short- with the in vitro potency of the CAR T cells (Figure 4d). spacer CAR was more potent than its half-spacer counterpart To compare potency between the nine different constructs, CAR (Po0.05 at three out of four time points measured after day 3; T cells were co-cultured with NALM-6 cells (Figures 4e,f and g). The one way ANOVA/SNK). T cells expressing the three different

Figure 2. Contrasting in vivo efficacy and toxicity of CD19-CARs configured with different spacers and signalling domains. (a) T cells were electroporated with high (H) or low (L) concentrations of CAR mRNA. Equimolar amounts of mRNA were used to adjust for differences in molar mass between the CAR constructs. Twenty-four NSG mice received 1 × 106 ffLuc-NALM-6 cells intravenously at day 0. The leukemic mice were randomized into three CAR T groups (n = 6) and two control groups (n = 3). The CAR T groups received T cells electroporated with (i) 19-IgFc- 28OXζ (H), (ii) 19-IgFc-28OXζ (L) or (iii) 19–28ζ (H) constructs. The leukemic control mice were treated with mock-transfected T cells (mock T cells) or left without T cells. Healthy mice (not injected with leukemia cells) were treated with 19-IgFc-28OXζ (H) or 19–28ζ (H) CAR T cells. All treatment groups received five injections of T cells from day 1 to day 21, as indicated. (b) Bioluminescence imaging was performed at the indicated time points. Photon-density heat maps are shown for all groups inoculated with ffLuc-NALM-6 cells. (c) The graph shows mean bioluminescence signal (counts per second per cm2) ± s.d. for each mouse group. All mice groups were inoculated with NALM-6 cells at day 0, except the two groups indicated as ‘no leukemia’. Statistical analysis (one-way ANOVA/SNK) demonstrated that mock T cells exerted a moderate but significant anti-leukemia activity, compared with the ‘no T cells’ control (Po0.05). Only the 19–28ζ CAR provided the T cells with significant additional anti-leukemia activity (Po0.05). (d) The weight loss during the first 3 days of study is presented for each group as the mean relative weight (actual weight divided by the weight at day 0). The mice treated with 19-IgFc-28OXζ (H) T cells recorded a significantly higher weight loss than the group treated with mock T cells (Po0.05, ANOVA/SNK). (e) Serum cytokine levels in mice with low or high tumor burden 2 days after injection of CAR T cells. The different mice groups had contrasting tumor burdens at the time of the last T-cell injection (day 21). The mice that had previously received CAR T cells received the same type of CAR T cells as at previous injections (19-IgFc- 28OXζ (H), 19-IgFc-28OXζ (L) or 19-short-28ζ (H)). The previously non-treated mice, at this time point carrying disseminated leukemia, received 19-short-28ζ CAR T cells at day 21. Blood was drawn from 16 mice at day 23 and the cytokine levels were measured by ELISA. The figure shows the cytokine values for the individual animals in each group (n = 3–5 per group). Bioluminescence heat-maps below the graphs indicate the tumor level of each animal at day 22.

© 2015 Macmillan Publishers Limited Gene Therapy (2015) 391 – 403 IgG1 spacer abrogates CAR efficacy in vivo H Almåsbak et al 396 half-spacer CARs (28OX, OX or 28) were equally efficient in survival of 61 versus 29 days (Figure 5d). The survival data for mice controlling leukemia, regardless of endodomain configuration. treated with different CAR T cells did not reveal substantial For Kaplan-Meier survival analysis, the four CAR T-cell groups survival differences between the CAR variants (short-28, half-28OX, were pooled and compared with both control groups taken half-OX, half-28; data not shown). together. The result demonstrated a significant survival advantage from CAR T-cell therapy (log rank test; Po0.001), with a median The IgG1-spacer CAR binds FcγR and mediates off-target activation from macrophages NSG mouse macrophages are dysfunctional, but express Fcγ receptors (FcγRs), including the high-affinity CD64 (FcγRI). We hypothesized that the detrimental effect of the CH2CH3-spacer in vivo may have been due to the engagement of FcγR on mouse macrophages with FcγR-binding motifs in the CH2 domain. This hypothesis was supported by our finding that half-spacer (hinge- CH3) CAR variants showed improved in vivo efficacy. To pursue this hypothesis, we stained CAR T cells with three soluble FcγRs from humans and mice (mouse CD64, human CD64, human CD32A; see Methods). The results indeed demonstrated strong binding of mouse FcγRI (CD64) to T cells expressing the full-spacer CAR (Figure 6a). By contrast, the short-spacer CAR T cells showed no FcγR binding. T cells with the full-spacer CAR, but not the other variants, had a weak binding to soluble human FcγRI (data not shown). Human CD32A (low affinity FcγRIIa) showed no binding to any CAR T cells (not shown). Next, we assessed the ability of NSG mouse macrophages to confer off-target activation of CAR T cells. We cultured CAR T cells for 20 h with macrophages obtained from NSG mice. The assay (Figures 6b and c) demonstrated substantial upregulation of activation markers CD25 and CD69 in T cells expressing the full- spacer CAR, consistent with stimulation through the FcγR-binding motifs in the IgG1 spacer. Compared with mock-transfected controls, a slight upregulation of CD25 and CD69 were observed in all CAR T cells (short-spacer, full-spacer). The latter observation is consistent with previous observations from many groups on ‘leaky’ CARs. Of note, the short-spacer CAR T cells retained similar CD25 and CD69 levels whether cultured with or without macrophages.

DISCUSSION Several phase I trials have showed impressive clinical responses from CD19 CAR T-cell therapy against B-cell malignancies, whereas others have reported only modest efﬁcacy.1–3,5,6,28–33 As many variables differ between clinical protocols, it is challenging to Figure 3. Lack of anti-leukemia activity in vivo is related to the IgG1- identify the factors that determine clinical efﬁcacy. Optimization of Fc spacer. Donor T cells were expanded and transfected with mRNA CAR design has in particular concentrated on modifying the encoding two different CARs that were identical apart from the cytoplasmic signalling moiety.7,26,34 Other structural units have spacer region: (i) CAR with a short 12aa spacer (19-short-28ζ) (ii) CAR received less attention, but some studies have indicated that the ζ – with a IgG1-based spacer (19-IgGFc-28 ). Seven NSG mice were spacer may be of importance.35 38 The IgG hinge-C 2C 3 domains inoculated with 1 × 106 ffLuc-NALM-6 cells intravenously at day 0 H H 6 are frequently used in spacers owing to its stabilizing effects on and treated once with 10 × 10 T cells, at day 1. The mice received 7 either 19-short-28ζ CAR T cells (n = 3), 19-IgGFc-28ζ CAR T cells CAR expression. Here, we report in vivo studies demonstrating (n = 3) or mock-transfected T cells (n = 1). The animals were imaged that this spacer has a detrimental effect in a xenograft at day 19 post treatment. Photon-density heat maps of the leukemia model. individual animals are depicted. The bar chart shows mean We compared CARs with and without the CH2CH3-spacer and bioluminescence signals (± s.d.). found that inclusion of this spacer abrogated functionality and

Figure 4. Comparison of CARs with different spacers and signalling domains. (a) Nine CD19-specific CAR constructs were designed, combining three different IgG1-based spacers (full, half, short) with three different signalling variants (28OXζ,28ζ,OXζ). The short spacer variant contained only a 12aa hinge. (b) T cells were electroporated with equimolar amounts of mRNA encoding the nine CAR constructs. After overnight culture, T cells were analyzed for CAR expression and functional activity against NALM-6 cells. CAR expression was detected with biotinylated protein-L/streptavidin APC recognizing the kappa light chain of the scFv. Overlay histograms of constructs with the same spacer (full, half or short) and different signalling domains are shown, with mock-transfected T cells as a control. (c) The graph (left) shows CAR expression kinetics over 96 h, calculated as relative fluorescence intensity compared with mock controls. The bar chart (right) shows slopes ( ± s.d.) for the protein L decay curves between 24 and 48 h, as an indicator for receptor stability. (d) CAR T cells electroporated with mRNA at different concentrations were co-cultured with NALM-6 cells (E: T ratio 1:3) for 4 h and analyzed by flow cytometry for degranulation (CD107a). (e, f) CAR T cells expressing nine different constructs were co-incubated with NALM-6 cells (E:T ratio 1:3) for 8 h and analyzed by flow cytometry for intracellular IFNγ production and degranulation. After 14 h, the cells were analyzed for annexin-V expression (g). Data are representative for two to four independent assays.

Gene Therapy (2015) 391 – 403 © 2015 Macmillan Publishers Limited IgG1 spacer abrogates CAR efﬁcacy in vivo H Almåsbak et al 397 yielded toxicity. Others have reported that the optimal spacer system were not detectable in vitro and were therefore unlikely to length for some CARs depends on the proximity of the target be caused by spacer size, sterical hindrance or ﬂexibility. We 35–37 epitope from the cell surface. However, the differences in our further found that CARs with a hinge-CH3 spacer (half spacer)

Gene Therapy (2015) 391 – 403 © 2015 Macmillan Publishers Limited IgG1 spacer abrogates CAR efficacy in vivo H Almåsbak et al 399 lacking the CH2 domain controlled leukemia progression, without determine the true importance of the findings in our detectable toxicity. As the CH2 domain includes the FcγR-binding xenograft model. motif, this result was consistent with the hypothesis that FcγR The mock-transfected T cells exerted a moderate anti-leukemia binding caused a detrimental effect.39 effect, both in vitro and in vivo. This effect may be due to donor In NSG mice, the macrophages are functionally impaired, but allo-reactivity against NALM-6 antigens or recognition of peptides FcγR-mediated off-target activation of the CAR T cells may still from the EGFP-ffLuc reporter.44 Consistent with the latter occur. Murine FcγRs display promiscuous specificities towards possibility, we observed that more T cells degranulated in co- human IgGs.40 In the present study, we found that soluble mouse cultures with EGFP-ffLuc-transduced NALM-6 cells than with the FcγRI (CD64) bound the full-spacer CAR, and not the short-spacer unmodified counterpart (data not shown). It is also possible that CAR. This observation confirmed that FcγRI, known to be the small number of NK/NKT cells remaining after T-cell expansion expressed by macrophages in NSG mice, was capable of binding conferred some anti-tumor activity. the full-spacer CAR construct. The finding implied inter-species We found that adverse effects from T cells with CARs harboring cross-reactivity and that the FcγRI binds to IgG1-Fc motifs even the CH2CH3 spacer were considerable even in non-leukemic mice, when these are embedded into a CAR. The soluble human FcγRI indicating that the toxicity had a CD19-independent component. bound only weakly to the full-spacer CAR. Whether this means This observation is consistent with our in vitro demonstration that that engagement of FcγRI would be less of a problem in humans, macrophages from NSG mice mediated activation of CAR T cells. is not possible to know, as FcγRI on cellular membranes may bind The possible implications of a cytokine storm in the NSG mice are with higher avidity than the soluble counterpart. Of note, unclear. The ability of human cytokines to bind murine receptors Hombach et al.41 have demonstrated that human immune cells is limited, and the NSG mice are deficient for the common IL-2 expressing various FcγRs may activate and become activated gamma chain. However, some cytokines are not dependent on the in vitro upon binding to CAR T cells expressing an IgG1-CH2CH3 IL-2 gamma chain for signalling, including IL-6 that has been spacer. reported as one of the critical components in CAR-mediated We further found that macrophages from NSG mice induced toxicity.1,3,6 off-target activation of T cells expressing the IgG1-spacer CAR, but Our unexpected finding of serious toxicity demonstrates the not the short-spacer construct. In vivo, this may lead to off-target need for careful safety testing of any new CAR construct, also toxicity, as observed in our mouse model. Moreover, FcγR- when using proven scFv-binding domains and targets considered mediated activation combined with the potent activation signals to be safe. Four out of five mice treated with CAR-mRNA T cells in provided by the other CAR elements may provoke activation- the MRD model completely recovered after a few days, consistent induced cell death. Our observation that no full-spacer CAR T cells with a transient expression of the receptor. By contrast, animals could be identified in peripheral blood at three days after transfer receiving stably transduced T cells were killed within 1 week is consistent with this notion. Further, we observed in vitro that owing to persistent toxicity. The different outcomes of the T cells with a high CAR expression displayed characteristics of treatments exemplify that transient redirection may represent a stronger activation combined with a higher apoptotic fraction. In safer alternative to viral transduction. line with our data, a recent study from Hudecek et al.42 indicated Serious adverse events, not predicted through preclinical that an IgG4-based spacer caused activation-induced cell death testing programs, have been observed in trials with several CARs and impaired the expansion of CAR T cells in vivo. On the other and TCRs. These unpredicted events and the unexpected toxicity hand, a clinical pilot study with a CD20-CAR bearing an IgG1 observed in our animal experiment add weight to the argument hinge-CH2CH3 spacer in lymphoma patients showed in vivo for using mRNA transfection in ‘first-in-man’ trials. Importantly, the persistence of modified T cells at the tumor site up to 1 year, mRNA approach allows for a classical trial design based on dose albeit at low levels.43 Taken together, IgG1/IgG4-based spacers escalation of a non-replicating drug as opposed to virally appear to work well in some systems, but not in others. This may transduced cells where the ‘drug’ dose is increased in an reflect that a number of elements influence in vivo efficacy, uncontrolled manner in the patients. Further, the CAR expression including preconditioning regimes, target molecule density, level per cell can be fine-tuned by adjusting the mRNA choice of scFv, CAR signalling elements and T-cell phenotype. concentration at electroporation. The mRNA strategy may also Regarding CD19-specific CARs, constructs with IgG1-CH2CH3 be applicable beyond early clinical testing, in settings where a spacers have been used in a few clinical trials,30,31 but to our temporary treatment period is sufficient. Regarding CARs target- knowledge not in those that have achieved clinical ing CD19, it is desirable to avoid the long-term B-cell aplasia responses.1–3,5,29,32,33 It is not possible to know whether the lack associated with integrating vectors, but it is presently not known of clinical efficacy may be related to the CH2CH3-spacer as multiple whether a temporary treatment period could be curative. variables differ between the trials. Transient expression of CD19 CARs may be suitable up front of We observed no difference in activity between a series of allogeneic transplantation, for bringing patients with MRD after second or third generation fmc63 CARs containing different co- conventional therapy into complete remission. The potential of stimulatory domains (CD28, OX40 or CD28OX40), but the short- transient redirection of T cells in MRD has been demonstrated in 12aa spacer CAR appeared to yield a better anti-leukemia effect clinical studies with the CD19/CD3-bispecific antibody than the half-spacer variants. The reason for this moderate blinatumomab.45 In pediatric patients, the concerns over inser- difference is not clear, but points to the need to tailor the spacers tional mutagenesis and long-term B-cell aplasia are of particular for each CAR/target. Of note, only studies in humans can relevance. Although present data suggest that integrating vectors

Figure 5. Mouse xenograft study comparing CAR constructs with half or short IgG1-spacers combined with alternative signalling domains. (a) Twenty-six NSG mice were injected with 1 ×106 ffLuc-NALM-6 cells i.v on day 0. Four study groups (ﬁve mice in each group) received treatment with T cells expressing four different CAR constructs (short-28, half-28-OX, half-OX, half-28). The T cells were administered as 5 × 106 CAR T cells on days 1, 3, 8 and 11. Control mice received non-transfected T cells (n = 3) or phosphate-buffered saline (n = 3; 1 died under anaesthesia). (b) The animals were imaged at the indicated time points pre- and post injection. The right column (‘controls’) shows the three mice receiving mock-T cells (top) and the two mice receiving phosphate-buffered saline (bottom). (c) The graph shows the average ( ± s.d.) bioluminescence signal (counts per second per cm2) from mice in each study group. (d) Kaplan-Meier analysis on all CAR T-cell groups pooled together, compared with pooled control groups. CAR T-cell-treated mice had a signiﬁcantly prolonged survival compared with control mice (Po0.001; log-rank test).

Figure 6. CAR binding to FcγR and off-target activation mediated by macrophages. T cells were expanded for 10 days and transfected with mRNA encoding two alternative CD19 CAR constructs differing only in the spacer region: (i) CAR 19-short-28 (short-28 T) (ii) CAR 19-full-28 (full-28 T). Control T cells were electroporated without mRNA (mock T). Protein L/Streptavidin-APC staining indicated CAR expression in 490% of T cells, at similar levels between the two CAR constructs (not shown). (a) CAR binding to FcγR. Flow cytometry analysis of T cells stained 20 h after electroporation with GST-tagged soluble mouse FcγRI (CD64). Anti-GST FITC was used as secondary Ab. Overlay histogram (a) shows staining of the three T-cell populations (short-28, full-28, mock) and corresponding controls stained with only the secondary anti- GST FITC Ab (Sec. Ab only). The full-spacer CAR (full-28 T) was the only one to stain above the sec. Ab controls. (b,c) T-cell activation mediated by macrophages. Mouse macrophages (Macr) were obtained from NSG mice. T cells (short-28, full-28, mock) were harvested 20 h after electroporation and cultured for 16 h with/without macrophages. Three parallel cell cultures were performed for each condition. The T cells were stained for ﬂow cytometry with activation markers CD25 and CD69. Figure (b) shows representative zebra plots for CD4+ mock T cells (left), short-28 T cells (middle) and full-28 T cells (right) after 16 h culture with macrophages. Figure (c) shows the mean percentage of CD25high/CD69+ cells for each condition. CD4+ and CD8+ T cells were analyzed separately. Error bars indicate s.e.m. from triplicate cell cultures.

may be safe,18–20 the risk of insertional mutagenesis is of indicating a vaccination effect, in their ongoing mesothelin mRNA increasing relevance as the trend is to use T cells with stronger CAR trial, which represents the first clinical study with mRNA CARs. regenerative capacity 4,46 In the context of immunogenic cell death, a Th1-like cytokine The anti-leukemia efficacy observed in this mouse model profile, like the one observed in vitro and in mouse serum in our suggests that transient CAR expression after mRNA transfection study, may be of importance for directing the immune response. was sufficient for the T cells to home to leukemic sites and kill In conclusion, our study suggests that the CH2 domain in malignant cells. In patients, CAR-mediated killing of tumor cells commonly used IgG1-Fc spacers may have a detrimental effect on may also lead to immunogenic cell death.47 Interestingly, Carl CAR efficacy and cause toxicity, owing to FcR engagement. The June and co-workers27 have observed epitope spreading, findings point to a need for tailoring the spacer for each CAR

Gene Therapy (2015) 391 – 403 © 2015 Macmillan Publishers Limited IgG1 spacer abrogates CAR efficacy in vivo H Almåsbak et al 401 construct and highlight the importance of testing new CARs were used: CD8-FITC, CD4-PerCP and Annexin V-APC (eBioscience, San in vivo. Moreover, the results show that transiently modified T cells Diego, CA, USA); IFNγ FITC, CD107a PE-Cy5, CD4 BV510, CD8 BV711, CD25 are capable of controlling leukemia in vivo and support the BV786, CD69 PE-Cy5 and AnnexinV-FITC (BD Biosciences). Expression of the rationale for developing an mRNA-based platform for clinical use. CD19 CAR was determined by goat anti-human IgGFc-Cy5 antibody On the basis of the data reported here, we have initiated a first-in- (Jackson Immuno Research Europe Ltd, Newmarket, UK) or protein L-based 50 man trial evaluating repeated injections of mRNA CAR T cells in staining. FcR binding to CAR T cells was evaluated using soluble GST- acute lymphocytic leukemia patients. tagged mouse FcγRI (CD64), human FcγRI (CD64) or human FcγRIIA-H131 (CD32A),51 combined with anti-GST AF488 (Merck Millipore, Oslo, Norway).

MATERIALS AND METHODS T-cell functionality assays Ethical approval Flow cytometry assays. T cells were stimulated with target cells at effector: The study was approved by the Regional Committee for Medical Research target ratio of 1:3, or as stated. Measurement of intracellular IFNγ and Ethics (Oslo, Norway). All mouse experiments were conducted under an CD107a were performed as previously described.21 Annexin-V staining was Institutional Animal Care and Use Committee-approved protocol. performed according to the manufacturer’s protocol. Briefly, the cells were stained with surface monoclonal antibody, washed with Annexin binding Cell lines, primary T cells and electroporation buffer and stained with conjugated Annexin V for 15 min at room NALM-6 cells (DSMZ-no ACC 128; CD19+ human acute lymphocytic temperature. Finally, the cells were washed with Annexin binding buffer leukemia cell line) were cultured in RPMI-1640 with L-glutamine (PAA and analyzed by flow cytometry within 60 min. Laboratories, Pasching, Austria), Gentamicin (50 μgml− 1; Sanofi Aventis, Paris, France) and 10% fetal calf serum (PAA). Phoenix HEK 293 cells were Cytokine measurement by ELISA. T cells and target cells were resuspended − grown in DMEM (PAA) with 10% fetal calf serum and 100 U ml 1 penicillin/ in culture medium without IL-2, combined at the ratio of 1:1 and incubated streptomycin. Peripheral blood mononuclear cells were isolated from at 37 °C for 48 h. Supernatants were harvested and frozen at − 80 °C for healthy donor buffy coats using Lymphoprep (Axis-Shield, Oslo, Norway) subsequent ELISA following standard protocol (Mabtech AB, Nacka Strand, and cultured in CellGro DC medium (CellGenix, Freiburg, Germany) with 5% Sweden). Mouse serum samples were diluted in ELISA buffer. Data analyses human serum (PAA), 0.8% Mucomyst (AstraZeneca, London, UK), 50 μgml− 1 − 1 were performed with ReaderFit software (Hitachi Software Engineering Gentamicin and 100 U ml IL-2 (Novartis, Emeryville, CA, USA). T cells America Ltd, South San Francisco, CA, USA). were expanded from peripheral blood mononuclear cells as previously described21 using Dynabeads ClinExVivo CD3/CD28 (Life Technologies, Chromium release assay. Cytotoxic activity was measured in standard Oslo, Norway). Expanded T cells (day 10 post activation) were transfected 51 21 with CAR mRNA by use of electroporation, as previously described.21 Cr release assays, as previously described.

Vector construction Mouse macrophage stimulation of T cells The CAR constructs 19-IgGFc-CD28OX40ζ (kindly provided by Dr Martin Peritoneal macrophages were collected from NSG mice by lavage with cold Pule, University College London) and 19-CD28ζ (kindly provided by Dr phosphate-buffered saline 24 h after intraperitoneal injection of 2 ml 3% Claudia Rössig, University Hospital Muenster) were re-cloned from the Brewer's thioglycollate (Sigma-Aldrich, Oslo, Norway). Cells were washed, retrovirus vector SFG into the mRNA expression vector pCIpA102 as resuspended in RPMI1640+1+% fetal calf serum and plated in a 48-well previously described.21 Modified CAR constructs were made using the 19- plate (5x105 cells per well). Non-adherent cells were removed after 1 h IgGFc-CD28OX40ζ as template, via overlapping PCR using primers from incubation at 37 °C. The macrophages were incubated over night, before Eurofins, Ebersberg, Germany and Pfu Ultra II Polymerase (Agilent addition of T cells at the ratio of 1:1. T cells and macrophages were co- Technologies, Waldbronn, Germany). The primer sequences can be cultured for 20 h and stained with monoclonal antibodies against CD4, obtained from the corresponding author. The PCR products were cloned CD8, CD25 and CD69. into pENTR-D using the gateway system (Life Technologies) and moved into a pCIpA102-G gateway-compatible RNA vector following standard protocol. Mouse xenograft studies NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice were bred in-house. Six to In vitro synthesis of mRNA 8-week-old mice were injected in the tail vein with NALM-6 cells, as stated. The in vitro mRNA synthesis was performed essentially as previously The NALM-6 cells were engineered with a retroviral vector (kindly provided described.21 Anti-Reverse Cap Analog (Trilink Biotechnologies Inc., San by Dr Rainer Löw, EUFETS AG, Germany) to express firefly luciferase 52 Diego, CA, USA) or β-S-ARCA48 (generous gift from Dr A Kuhn, BioNTech and EGFP. Tumor growth was monitored by bioluminescent imaging AG, Mainz, Germany and Dr E Darzynkiewicz, Warsaw University, Poland) (Xenogen Spectrum system; Living Image v3.2 software; Advanced were used to cap the RNA. The mRNA was assessed by agarose gel Molecular Vision, Grantham, UK). Anesthetized mice were injected intra- electrophoresis and Nanodrop (Thermo Fisher Scientific, Waltham, peritoneally with 150 mg kg − 1 body weight of D-luciferin (Caliper Life MA, USA). Sciences, Hopkinton, MA, USA). Animals were imaged 10 min after luciferin injection. Retroviral transduction 6 Phoenix HEK 293 cells (1 × 10 per 6 cm dish) were plated before co- Statistical analysis transfection of packaging plasmids with the SFG-CD19 CAR vectors using Statistical analyses were performed using SPSS for Windows, version 18 FUGENE-6 reagent (Roche, Oslo, Norway). The supernatant was harvested 2 (IBM, Armonk, NY, USA). and 3 days after transfection, snap-frozen, and stored at − 80 °C. T cells Continuous data were described with median, mean and range. were expanded with Dynabeads as described previously21 and transduced at day 3 with the retroviral supernatant in retronectin-coated plates or in One-way ANOVA followed by SNK test was applied for comparing the ’ tubes with retronectin-coated M-450 Epoxy Dynabeads (kind gift from Axl distribution of data between three or more groups. Student s t-test Neuratuer, Life Technologies) essentially as published.49 After transduction, was used for statistical comparisons between two groups. Survival was the cultures were expanded for 7 days with IL-2 (100 U ml − 1) and 5% calculated using the Kaplan-Meier method. The log rank analysis was used human serum before adoptive transfer. for comparing survival between groups. All P-values given are two-tailed values. A P-value below 0.05 was considered significant. Flow cytometry Flow cytometry analysis was performed on a LSR II flow cytometer (BD Biosciences, Franklin Lakes, NJ, USA) and data were analyzed with the CONFLICT OF INTEREST Flow Jo Software (Tree Star Inc, Ashland, OR, USA). The following reagents The authors declare no conflict of interest.

© 2015 Macmillan Publishers Limited Gene Therapy (2015) 391 – 403 IgG1 spacer abrogates CAR efficacy in vivo H Almåsbak et al 402 ACKNOWLEDGEMENTS 23 Birkholz K, Hombach A, Krug C, Reuter S, Kershaw M, Kampgen E et al. Transfer of This work was supported by The Norwegian Cancer Society, Radiumhospitalets mRNA encoding recombinant immunoreceptors reprograms CD4+ and CD8+ legater, The Norwegian Health Region South East and the Research Council of T cells for use in the adoptive immunotherapy of cancer. Gene Ther 2009; 16: – Norway. We would like to thank Professor I Sandlie (University of Oslo and OUH) for 596 604. valuable advice and for providing soluble FcγRs. We also thank Dr AA Tveita (OUH) for 24 Rabinovich PM, Komarovskaya ME, Wrzesinski SH, Alderman JL, Budak-Alpdogan providing mouse macrophages and Ms. HJ Hoel, Ms. A Faane and Dr N Westerdaal T, Karpikov A et al. Chimeric receptor mRNA transfection as a tool to generate – (OUH) for laboratory assistance. We further thank Dr R Löw (EUFETS AG) for providing antineoplastic lymphocytes. Hum Gene Ther 2009; 20:51 61. 25 Zhao Y, Zheng Z, Cohen CJ, Gattinoni L, Palmer DC, Restifo NP et al. High- the reporter vector and LifeTechnologies for supplying M-450 Epoxy and CD3CD28 efficiency transfection of primary human and mouse T lymphocytes using RNA Dynabeads. Finally, we would like to thank Dr A Kuhn (BioNTech AG) for valuable electroporation. Mol Ther 2006; 13: 151–159. advice and Dr C Rössig (University Children's Hospital Münster) and Dr M Pule 26 Barrett DM, Liu X, Jiang S, June CH, Grupp SA, Zhao Y. Regimen-Specific Effects of (University College London) for providing CAR plasmids. RNA-Modified Chimeric Antigen Receptor T Cells in Mice with Advanced Leuke- mia. Hum Gene Ther 2013; 24:717–727. 27 Beatty GL, Haas AR, Maus MV, Torigian DA, Soulen MC, Plesa G et al. Mesothelin- REFERENCES specific Chimeric Antigen Receptor mRNA-Engineered T cells Induce Anti-Tumor 1 Kochenderfer JN, Rosenberg SA. Treating B-cell cancer with T cells expressing Activity in Solid Malignancies. Cancer Immunol Res 2014; 2:112–120. anti-CD19 chimeric antigen receptors. Nat Rev Clin Oncol 2013; 10:267–276. 28 Brentjens RJ, Riviere I, Park JH, Davila ML, Wang X, Stefanski J et al. Safety and 2 Kalos M, Levine BL, Porter DL, Katz S, Grupp SA, Bagg A et al. T cells with chimeric persistence of adoptively transferred autologous CD19-targeted T cells in patients antigen receptors have potent antitumor effects and can establish memory in with relapsed or chemotherapy refractory B-cell leukemias. Blood 2011; 118: patients with advanced leukemia. Sci Transl Med 2011; 3: 95ra73. 4817–4828. 3 Brentjens RJ, Davila ML, Riviere I, Park J, Wang X, Cowell LG et al. CD19-targeted 29 Brentjens RJ, Curran KJ. Novel cellular therapies for leukemia: CAR-modified T cells T cells rapidly induce molecular remissions in adults with chemotherapy- targeted to the CD19 antigen. Hematology Am Soc Hematol Educ Program 2012; refractory acute lymphoblastic leukemia. Sci Transl Med 2013; 5: 177ra138. 2012: 143–151. 4 Jensen MC, Riddell SR. Design and implementation of adoptive therapy with 30 Savoldo B, Ramos CA, Liu E, Mims MP, Keating MJ, Carrum G et al. CD28 costi- chimeric antigen receptor-modified T cells. Immunol Rev 2014; 257: 127–144. mulation improves expansion and persistence of chimeric antigen receptor- 5 Grupp SA, Kalos M, Barrett D, Aplenc R, Porter DL, Rheingold SR et al. Chimeric modified T cells in lymphoma patients. J Clin Invest 2011; 121: 1822–1826. antigen receptor-modified T cells for acute lymphoid leukemia. N Engl J Med 2013; 31 Jensen MC, Popplewell L, Cooper LJ, DiGiusto D, Kalos M, Ostberg JR et al. 368: 1509–1518. Antitransgene rejection responses contribute to attenuated persistence of 6 Kochenderfer JN, Dudley ME, Feldman SA, Wilson WH, Spaner DE, Maric I et al. adoptively transferred CD20/CD19-specific chimeric antigen receptor redirected B-cell depletion and remissions of malignancy along with cytokine-associated T cells in humans. Biol Blood Marrow Transplant 2010; 16:1245–1256. toxicity in a clinical trial of anti-CD19 chimeric-antigen-receptor-transduced 32 Kochenderfer JN, Wilson WH, Janik JE, Dudley ME, Stetler-Stevenson M, Feldman T cells. Blood 2012; 119: 2709–2720. SA et al. Eradication of B-lineage cells and regression of lymphoma in a patient 7PuleM,FinneyH,LawsonA.Artificial T-cell receptors. Cytotherapy 2003; 5:211–226. treated with autologous T cells genetically engineered to recognize CD19. Blood 8 Chmielewski M, Hombach AA, Abken H. CD28 cosignalling does not affect the 2010; 116:4099–4102. activation threshold in a chimeric antigen receptor-redirected T-cell attack. Gene 33 Cooper LJ, Jena B, Bollard C.M. Good T cells for bad B cells. Blood 2012; 119: Ther 2011; 18:62–72. 2700–2702. 9 Loskog A, Giandomenico V, Rossig C, Pule M, Dotti G, Brenner MK. Addition of the 34 Hombach A, Abken H. Costimulation tunes tumor-specific activation of redirected CD28 signaling domain to chimeric T-cell receptors enhances chimeric T-cell T cells in adoptive immunotherapy. Cancer Immunol Immunother 2007; 56: resistance to T regulatory cells. Leukemia 2006; 20: 1819–1828. 731–737. 10 Pule MA, Straathof KC, Dotti G, Heslop HE, Rooney CM, Brenner MK. A chimeric T 35 Hudecek M, Lupo-Stanghellini MT, Kosasih PL, Sommermeyer D, Jensen MC, Rader cell antigen receptor that augments cytokine release and supports clonal C et al. Receptor affinity and extracellular domain modifications affect tumor expansion of primary human T cells. Mol Ther 2005; 12:933–941. recognition by ROR1-specific chimeric antigen receptor T cells. Clin Cancer Res 11 Maher J, Brentjens RJ, Gunset G, Riviere I, Sadelain M. Human T-lymphocyte 2013; 19: 3153–3164. cytotoxicity and proliferation directed by a single chimeric TCRzeta/CD28 36 Wilkie S, Picco G, Foster J, Davies DM, Julien S, Cooper L et al. Retargeting of receptor. Nat Biotechnol 2002; 20:70–75. human T cells to tumor-associated MUC1: the evolution of a chimeric antigen 12 Hinrichs CS, Restifo NP. Reassessing target antigens for adoptive T-cell therapy. receptor. J Immunol 2008; 180:4901–4909. Nat Biotechnol 2013; 31: 999–1008. 37 Guest RD, Hawkins RE, Kirillova N, Cheadle EJ, Arnold J, O'Neill A et al. 13 Parkhurst MR, Yang JC, Langan RC, Dudley ME, Nathan DA, Feldman SA et al. The role of extracellular spacer regions in the optimal design of chimeric immune T cells targeting carcinoembryonic antigen can mediate regression of metastatic receptors: evaluation of four different scFvs and antigens. J Immunother 2005; 28: colorectal cancer but induce severe transient colitis. Mol Ther 2011; 19:620–626. 203–211. 14 Lamers CH, Sleijfer S, van Steenbergen S, van Elzakker P, van Krimpen B, Groot C 38 Bridgeman JS, Hawkins RE, Hombach AA, Abken H, Gilham DE. Building better et al. Treatment of metastatic renal cell carcinoma with CAIX CAR-engineered chimeric antigen receptors for adoptive T cell therapy. Curr Gene Ther 2010; 10: T cells: clinical evaluation and management of on-target toxicity. Mol Ther 2013; 77–90. 21: 904–912. 39 Hogarth PM, Pietersz GA. Fc receptor-targeted therapies for the treatment of 15 Morgan RA, Chinnasamy N, Abate-Daga D, Gros A, Robbins PF, Zheng Z et al. inflammation, cancer and beyond. Nat Rev Drug Dis 2012; 11: 311–331. Cancer regression and neurological toxicity following anti-MAGE-A3 TCR gene 40 Overdijk MB, Verploegen S, Ortiz Buijsse A, Vink T, Leusen JH, Bleeker WK et al. therapy. J Immunother 2013; 36:133–151. Crosstalk between human IgG isotypes and murine effector cells. J Immunol 2012; 16 Linette GP, Stadtmauer EA, Maus MV, Rapoport AP, Levine BL, Emery L et al. 189:3430–3438. Cardiovascular toxicity and titin cross-reactivity of affinity-enhanced T cells in 41 Hombach A, Hombach AA, Abken H. Adoptive immunotherapy with genetically myeloma and melanoma. Blood 2013; 122:863–871. engineered T cells: modification of the IgG1 Fc 'spacer' domain in the extracellular 17 Stauss HJ, Morris EC. Immunotherapy with gene-modified T cells: limiting side moiety of chimeric antigen receptors avoids 'off-target' activation and effects provides new challenges. Gene Ther 2013; 20: 1029–1032. unintended initiation of an innate immune response. Gene Ther 2010; 17: 18 Westwood JA, Murray WK, Trivett M, Shin A, Neeson P, MacGregor DP et al. 1206–1213. Absence of retroviral vector-mediated transformation of gene-modified T cells 42 Hudecek M, Sommermeyer D, Kosasih PL, Silva-Benedict A, Liu L, Rader C et al. after long-term engraftment in mice. Gene Ther 2008; 15: 1056–1066. The non-signaling extracellular spacer domain of chimeric antigen receptors is 19 Newrzela S, Cornils K, Li Z, Baum C, Brugman MH, Hartmann M et al. Resistance of decisive for in vivo antitumor activity. Cancer Immunol Res e-pub ahead of print 11 mature T cells to oncogene transformation. Blood 2008; 112: 2278–2286. September 2014. 20 Scholler J, Brady TL, Binder-Scholl G, Hwang WT, Plesa G, Hege KM et al. Decade- 43 Till BG, Jensen MC, Wang J, Qian X, Gopal AK, Maloney DG et al. CD20-specific long safety and function of retroviral-modified chimeric antigen receptor T cells. adoptive immunotherapy for lymphoma using a chimeric antigen receptor with Sci Transl Med 2012; 4: 132ra153. both CD28 and 4-1BB domains: pilot clinical trial results. Blood 2012; 119: 21 Almasbak H, Rian E, Hoel HJ, Pule M, Walchli S, Kvalheim G et al. Transiently 3940–3950. redirected T cells for adoptive transfer. Cytotherapy 2011; 13: 629–640. 44 Barrett DM, Seif AE, Carpenito C, Strong EP, June CH, Grupp S.A et al. Biolumi- 22 Barrett DM, Zhao Y, Liu X, Jiang S, Carpenito C, Kalos M et al. Treatment of nescent tracking of human and mouse acute lymphoblastic leukemia reveals advanced leukemia in mice with mRNA engineered T cells. Hum Gene Ther 2011; potent immunogenicity of luciferase in some preclinical models of leukemia. 22: 1575–1586. 52nd ASH Annual Meeting;4–7 December 2010; Orlando, FL, USA, 2010.

Gene Therapy (2015) 391 – 403 © 2015 Macmillan Publishers Limited IgG1 spacer abrogates CAR efficacy in vivo H Almåsbak et al 403 45 Topp MS, Gokbuget N, Zugmaier G, Degenhard E, Goebeler ME, Klinger M et al. vaccines in immature dendritic cells and induce superior immune responses Long-term follow-up of hematologic relapse-free survival in a phase 2 study of in vivo. Gene Ther 2010; 17: 961–971. blinatumomab in patients with MRD in B-lineage ALL. Blood 2012; 120: 49 Heemskerk B, Jorritsma A, Gomez-Eerland R, Toebes M, Haanen JB, Schumacher 5185–5187. TN. Microbead-assisted retroviral transduction for clinical application. Hum Gene 46 Lugli E, Dominguez MH, Gattinoni L, Chattopadhyay PK, Bolton DL, Song K et al. Ther 2010; 21: 1335–1342. Superior T memory stem cell persistence supports long-lived T cell memory. J Clin 50 Zheng Z, Chinnasamy N, Morgan RA. Protein L: a novel reagent for the detection Invest 2013; 123: 594–599. of chimeric antigen receptor (CAR) expression by flow cytometry. J Transl Med 47 Chmielewski M, Kopecky C, Hombach AA, Abken H. IL-12 release by engineered 2012; 10:29. T cells expressing chimeric antigen receptors can effectively Muster an antigen- 51 Berntzen G, Lunde E, Flobakk M, Andersen JT, Lauvrak V, Sandlie I. Prolonged and independent macrophage response on tumor cells that have shut down tumor increased expression of soluble Fc receptors, IgG and a TCR-Ig fusion protein by antigen expression. Cancer Res 2011; 71: 5697–5706. transiently transfected adherent 293E cells. J Immunol Methods 2005; 298:93–104. 48 Kuhn AN, Diken M, Kreiter S, Selmi A, Kowalska J, Jemielity J et al. Phosphor- 52 Loew R, Heinz N, Hampf M, Bujard H, Gossen M. Improved Tet-responsive pro- othioate cap analogs increase stability and translational efficiency of RNA moters with minimized background expression. BMC Biotechnol 2010; 10:81.

Supplementary Information accompanies this paper on Gene Therapy website (http://www.nature.com/gt)