Extracellular NGFR Spacers Allow Efficient Tracking and Enrichment of Fully Functional CAR-T Cells Co-Expressing a Suicide Gene

Chimeric antigen receptor (CAR)-T cell immunotherapy is at the forefront of innovative cancer therapeutics. However, lack of standardization of cellular products within the same clinical trial and lack of harmonization between different trials have hindered the clear identification of efficacy and safety determinants that should be unveiled in order to advance the field. With the aim of facilitating the isolation and in vivo tracking of CAR-T cells, we here propose the inclusion within the CAR molecule of a novel extracellular spacer based on the low-affinity nerve-growth-factor receptor (NGFR). We screened four different spacer designs using as target antigen the CD44 isoform variant 6 (CD44v6). We successfully generated NGFR-spaced CD44v6 CAR-T cells that could be efficiently enriched with clinical-grade immuno-magnetic beads without negative consequences on subsequent expansion, immuno-phenotype, in vitro antitumor reactivity, and conditional ablation when co-expressing a suicide gene. Most importantly, these cells could be tracked with anti-NGFR monoclonal antibodies in NSG mice, where they expanded, persisted, and exerted potent antitumor effects against both high leukemia and myeloma burdens. Similar results were obtained with NGFR-enriched CAR-T cells specific for CD19 or CEA, suggesting the universality of this strategy. In conclusion, we have demonstrated that the incorporation of the NGFR marker gene within the CAR sequence allows for a single molecule to simultaneously work as a therapeutic and selection/tracking gene. Looking ahead, NGFR spacer enrichment might allow good manufacturing procedures-manufacturing of standardized CAR-T cell products with high therapeutic potential, which could be harmonized in different clinical trials and used in combination with a suicide gene for future application in the allogeneic setting.

Chimeric antigen receptor (CAR)-T cell immunotherapy is at the forefront of innovative cancer therapeutics. However, lack of standardization of cellular products within the same clinical trial and lack of harmonization between different trials have hindered the clear identification of efficacy and safety determinants that should be unveiled in order to advance the field. With the aim of facilitating the isolation and in vivo tracking of CAR-T cells, we here propose the inclusion within the CAR molecule of a novel extracellular spacer based on the low-affinity nerve-growth-factor receptor (NGFR). We screened four different spacer designs using as target antigen the CD44 isoform variant 6 (CD44v6). We successfully generated NGFR-spaced CD44v6 CAR-T cells that could be efficiently enriched with clinical-grade immuno-magnetic beads without negative consequences on subsequent expansion, immuno-phenotype, in vitro antitumor reactivity, and conditional ablation when co-expressing a suicide gene. Most importantly, these cells could be tracked with anti-NGFR monoclonal antibodies in NSG mice, where they expanded, persisted, and exerted potent antitumor effects against both high leukemia and myeloma burdens. Similar results were obtained with NGFR-enriched CAR-T cells specific for CD19 or CEA, suggesting the universality of this strategy. In conclusion, we have demonstrated that the incorporation of the NGFR marker gene within the CAR sequence allows for a single molecule to simultaneously work as a therapeutic and selection/tracking gene. Looking ahead, NGFR spacer enrichment might allow good manufacturing procedures-manufacturing of standardized CAR-T cell products with high therapeutic potential, which could be harmonized in different clinical trials and used in combination with a suicide gene for future application in the allogeneic setting.
Keywords: car-T cells, car spacer, cell sorting, good manufacturing procedures-manufacturing, antitumor efficacy, suicide gene inTrODUcTiOn Chimeric antigen receptors (CARs) are synthetic biology mole cules constructed by fusing an extracellular antigenbinding moiety, commonly the singlechain variable fragment (scFv) of a tumorreactive monoclonal antibody (mAb), with intracel lular activatory units, usually the CD3 zeta chain, coupled with costimulatory endodomains from either CD28 or 41BB (1). Over the last 5 years, different US Institutions have shown impressive antitumor effects after infusing CD19 CART cells in patients with refractory Bcell malignancies, including chronic lympho cytic leukemia (2,3), Bcell acute lymphoblastic leukemia (4)(5)(6)(7) and nonHodgkin lymphomas (2,8). More recently, the approval of the first CART cell therapy to treat pediatric and young adult relapsed/refractory Bcell ALL has landmarked the beginnings of what will likely be a new era in cancer therapeutics.
Accumulating clinical data suggest that primary expansion and longterm in vivo persistence of CART cells are main determinants of the final therapeutic outcome. These properties are seemingly influenced by both CART cell and hostspecific factors. For instance, CAR designs including CD28 (9) and 41BB (10) costimulatory endodomains, as well as the frequen cies of stem (TSCM) and central memory (TCM) T cells in the final product (11), have both been shown to substantially contribute to a longlived phenotype. On the other hand, patient pre conditioning is recognized to promote CART cell engraftment (7,12), while contrariwise residual host immunity may cause their humoral and/or Tcell mediated rejection, especially if murine scFv sequences are used (7,13,14). Related to this, while using human scFv may drastically reduce the immunogenicity of synthetic CARs, prediction algorithms may be exploited to evaluate the potential of fusion sites between human components to provide immunogenic epitopes for Tcell immune responses, allowing their preventive modification (15). As CART cells are entering the commercial phase, investigators, regulators, and industrial stakeholders are dedicating increasing attention to the pharmaceutical aspects of this revolutionary type of treatment, including rationalization of good manufacturing procedures and indepth analysis of toxicology, pharmacokinetics, and pharma codynamics (16). These continuing efforts clearly require new, easy and informative methods for tracking and characterizing transgeneexpressing and, therefore, pharmacologically active T cells, both in the final CART cell product before infusion and, later, in treated patients.
Currently available tracking methods rely on qPCR (4, 5, 17) or on antibodies specific for either the CAR molecule itself (11,18) or a separate marker gene (7,8,19). Compared with PCR, antibodybased methods have the advantage of enabling not only the in vivo tracking of CART cells, but also the characteriza tion, at a singlecell level, of their differentiation, activation, and exhaustion statuses. In addition, they offer the unique possibility to enrich CART cells before infusion, allowing the design of more standardized CART cell therapies. In foresight, this pos sibility might crucially facilitate the translation of CART cells to the allogeneic setting, where coexpressing a suicide gene would necessarily require an enrichment step to remove unmodified alloreactive cells (20). Unfortunately, the antibodybased methods for CART cell marking developed so far have some limitations, especially in light of their potential use as universal enrichment tools. For instance, antiidiotypic mAbs already used for CD19 CARs (18) would need to be developed for each single specific ity and, if used for enrichment, are expected to unduly activate CART cells during ex vivo manipulation. On the other hand, separate immunomarker genes (7,8,19) reflect CAR expression only indirectly and may saturate the cargo capacity of currently available viral vectors, abating transduction efficiency, especially in the case of multicistronic cassettes (CAR, immunemarker and suicide gene).
A promising alternative to these approaches is the inclusion of an immunomarker sequence within the extracellular portion of the CAR molecule itself. In this study, we designed an innovative CAR spacer based on extracellular domains from the lowaffinity nervegrowthfactor receptor (NGFR), a marker gene already used in the clinic for the selection/tracking of transduced T cells. We then validated the antitumor efficacy of NGFRenriched CART cells specific for the CD44 isoform variant 6 (CD44v6), CD19, and CEA in clinically relevant in vivo xenograft mouse models. Additionally, we engineered T cells with a clinicalgrade bicistronic retroviral vector encoding for the NGFRspaced CD44v6 CAR and the thymidine kinase (TK) suicide gene and proved efficient sorting with clinicalgrade reagents, potent antitumor efficacy and optimal suicidability upon exposure to Ganciclovir. This NGFRspaced CD44v6 CAR Tcell product is currently at late stage of process development and these efforts have recently gained momentum by the EC through dedicated H2020 funding to support phase I/IIa clinical trial in patients with relapsed/refractory acute myeloid leukemia (AML) and multiple myeloma (MM).

MaTerials anD MeThODs construct generation
We used the lowaffinity NGFR gene as reference (P08138, TNR16_HUMAN). The NGFR wildtype long (NWL) construct contains the four TNFR cysteinerich domains and the serine/ threoninerich stalk. The NGFR wildtype short (NWS) con struct contains only the four TNFR cysteinerich domains. The NGFR mutated long (NML) construct contains the four TNFR cysteinerich domains and the stalk, but the fourth domain was largely deleted to avoid NGF signaling (21). The NGFR mutated short (NMS) contains only the four TNFR cysteinerich domains, with the mutated version of the fourth domain. NGFR spacers were synthetized by GeneArt (ThermoFisher) and cloned into an original CAR incorporating an IgG1derived CH2CH3 spacer, the CD28 transmembrane and costimulatory domains and the CD3 zeta chain (9). The same procedure was applied to generate NGFR isoformspaced CARs specific for CD19 and the carci noembryonic antigen. All constructs were expressed into SFG monocistronic retroviral vectors (22) under the direct control of viral LTR. In a specific set of experiments, NWL and NMS isoformspaced CD44v6 CARs were expressed into a clinical grade bicistronic retroviral vector in combination with the TK suicide gene (SFCMM3) (20). In these constructs, CAR genes were placed under control of the internal SV40 promoter in place of the ΔNGFR marker gene while TK was maintained under the control of viral LTR (see Figure 7B).

Transduction and culture conditions
Buffy coats from healthy blood donors were obtained after written informed consent and IRB approval. T cells were stimulated with CD3/CD28beads (CTS Dynabeads CD3/CD28, ThermoFisher), RVtransduced with two rounds of spinoculation, and cultured in RPMI 1640 (GibcoBrl), 10% FBS (EuroClone) with IL7/ IL15 (5 ng/ml, PeproTech). After 6 days, beads were removed and, after additional 3-7 days, CART cells were stained with the PEconjugated antiNGFR mAb C401457 and enriched with antiPE immunemagnetic beads (Miltenyi Biotech). In a specific set of experiments (Figure 7), T cells were stimulated with CD3/CD28beads and RVtransduced into a RetroNectin coated bag in XVivo 15, 3% plasma with IL7/IL15 (100 and 200 U/ml, respectively). The day after, beads were removed and, after additional 3 days, NWLisoformspaced CART cells were enriched with clinicalgrade antiNGFR immunemagnetic beads (CD271 Microbeads, Miltenyi Biotec), while NMSisoformspaced CART cells were enriched with the twostep procedure described above. CART cell expansion is expressed as fold increase over numbers before enrichment.

In Vitro Functional assays
Immunoenriched CART cells were cocultured at different E:T ratios with the following tumorcell lines: MM.1S myeloma cells, THP1 and HL60 myeloid leukemic cells, BV173 lymphoid leu kemic cells, BxPC3 pancreatic tumor cells, ALLCM leukemic cells [kindly provided by Fred Falkenburg, Leiden University Medical Center (23)], primary AML blasts and MM plasma cells (from our Institutional Biobank after IRB approval). T cells transduced with an irrelevant CAR were always used as control (CTRL, GD2, or CD19specific as indicated). Twentyfourhour supernatants were collected and subsequently analyzed by FACS using the LEGENDplex cytokine immunoassay (BioLegend). Fourday cocultures were analyzed by FACS using FlowCount Fluorospheres (BeckmanCoulter). The elimination index was cal culated as follows: 1 − (number of residual target cells in presence of target antigenspecific CART cells/number of residual target cells in presence of CTRL CART cells). In cell proliferation assays, immuneenriched CART cells were labeled with 0.2 µM carboxy fluorescein succinimidyl ester (CFSE, Invitrogen), washed, and stimulated with irradiated (10,000 rad) tumor cells at 1:5 E:S ratio. After 6 days, the cells were analyzed by FACS and proliferation expressed as the percentage of CFSEdiluting cells. In a specific set of experiments (Figure 7), the efficacy of the suicide gene machinery was evaluated by exposing PHA activated (2 µg/ml, SigmaAldrich) CAR T cells to increasing concentrations of GCV. Cell viability was analyzed 5 days after by trypan blue exclusion and calculated as follows: number of living cells in GCVtreated samples/number of living cells in untreated samples × 100. T cells transduced with the original SCFMM3 construct carrying TK and ΔNGFR as a separate selection marker were used as comparison.

studies in Xeno-engrafted nsg Mice
All mouse experiments were approved by the Institutional Animal Care and Use Committee (IACUC #468) of San Raffaele University Hospital and Scientific Institute and by the Italian Governmental Health Institute (Rome, IT). 6 to 8weekold NSG mice were obtained from the Jackson Laboratories. In the hightumor burden settings, mice were injected intra venously (i.v.) with tumor cells transduced with a lentiviral vector encoding for the secreted luciferase Lucia (24) (THP 1luc cells, 1.5 × 10 6 ; MM.1Sluc cells, 2 × 10 6 ) and, after 14 (THP1) or 28 (MM.1S) days, treated with immunoenriched CART cells (5 × 10 6 ) by i.v. infusion. Tumor progression was monitored weekly by bioluminescence using the QUANTILuc detection reagent (InvivoGen) and expressed as relative light units (RLUs), according to the manufacturer instructions. Circulating human Tcell counts were measured weekly by FACS using FlowCount Fluorospheres (BeckmanCoulter). Mice were sacrificed when RLUs were >4 × 10 7 for THP1luc cells or >4 × 10 5 for MM.1Sluc cells, or when manifesting clinical signs of suffering. In vivo cytokine concentrations were measured by FACS using the LEGENDplex immunoassay (BioLegend). In a specific set of experiments (Figure 7), mice were injected with THP1 cells and, after 14 days, treated with immunoenriched CART cells. At day 40, mice were sacrificed and weights of THP1infiltrated livers were analyzed. In the minimal residualdisease settings, mice were injected i.v. with either THP1 or ALLCM cells (5 × 10 6 ) and, after 3 days, treated with immuneenriched CART cells (5 × 10 6 ) by i.v. infusion. Circulating human T and leukemiccell counts were measured weekly by FACS using FlowCount Fluorospheres (BeckmanCoulter). At sacrifice (THP1, day 35; ALLCM, day 45), liver, bone marrow, and spleen were analyzed for the pres ence of residual leukemic cells.

statistical analysis
Statistical analyzes were conducted using Prism Software 5.0 (GraphPad). We analyzed the datasets with paired or unpaired Student's ttest, oneway ANOVA or the Logrank Mantel-Cox tests. Differences with a P value < 0.05 were considered statisti cally significant.

ngFr-spaced car-T cells can Be Tracked and enriched With anti-ngFr reagents
We recently developed a CAR specific for CD44v6 for targeting AML and MM, as well as multiple epithelial tumors (25). However, our original CAR construct was equipped with an IgG1CH2CH3 spacer (CH2CH3) that could reduce its antitumor activity in vivo due to undesired interaction with Fc receptors (FcR)bearing myeloid cells, as previously reported (26)(27)(28)(29)(30). A truncated ver sion of the lowaffinity NGFR lacking intracellular signaling components is currently used in the clinic for gene marking and enrichment of engineered T cells (20,31). Aiming at developing a construct that could simultaneously enable the enrichment and tracking of fully functional CD44v6 CART cells, we replaced the CH2CH3 spacer with the extracellular domain from the NGFR. In particular, we generated four isoforms of the NGFR spacer, namely the wildtype long (NWL), the wildtype short (NWS), the mutated long (NML), and the mutated short (NMS), differing in length and inclusion or not of a specific deletion abrogating NGF signaling (21) (Figure 1A; see Materials and Methods).
We cloned the different NGFR isoformspaced CD44v6 CAR constructs into retroviral vectors (22,25) under the control of viral LTR driving strong transgene expression and transduced primary T cells after activation with CD3/CD28beads and IL7/ IL15, according to a protocol that enriches for stem (TSCM) and central memory (TCM) T cells (32)(33)(34). After transduction, all iso forms could be identified on the cell surface using the antiNGFR mAb C401457 and could be enriched with immunomagnetic beads ( Figure 1B). Interestingly, different NGFR staining inten sities between isoforms (NWL > NMS ≫ NWS/NML) were paralleled by staining with Protein L, a reagent that specifically binds to scFv (35) (Figure 1C), suggesting that different surface CAR stability, rather than epitope accessibility, was responsible for these differences.
After enrichment, all NGFR isoformspaced CART cells expanded similarly to those enriched through CH2CH3 ( Figure 1D). Interestingly, however, NWLenriched CART cells displayed a lower proportion of TSCM cells compared to NWS, NML, or NMSenriched CART cells ( Figure 1E) and a higher proportion of T cells expressing HLADR and PD1, alone or in combina tion (Figure 1F), suggesting that the NWLisoform might have induced tonic signaling, as already reported for CH2CH3 spacers (29). As a confirmation, even in the absence of cognate antigen, NWLenriched CART cells displayed a basal level of prolif eration (Figure 1G), and a higher apoptotic rate ( Figure 1H) compared to CART cells enriched through other NGFR isoform spacers.

ngFr-enriched cD44v6 car-T cells are Fully Functional In Vitro
To verify the functionality of NGFRenriched CD44v6 CART cells, we tested them in coculture experiments with tumor cells expressing or not CD44v6. Similarly to cells enriched through the CH2CH3 spacer, NWL, NWS, and NMS, but not NMLenriched CART cells killed CD44v6+ THP1 myeloid leukemia cells and MM.1S myeloma cells, while spared CD44v6− BV173 lymphoid leukemia cells (Figure 2A). On the contrary, they expectedly failed to kill CD44v6−/FcR+ HL60 myeloid leukemia cells, indicating efficient abrogation of FcR recognition. Accordingly, NWL, NWS, and NMS, but not NMLenriched effectors secondarily proliferated upon coculture with CD44v6+, but not with CD44v6− tumor cells (Figure 2B).
To exclude that replacing the CH2CH3 spacer with NGFR spacers might have reduced antitumor reactivity in vitro, we chal lenged NGFRenriched CD44v6 CART cells against CD44v6+/ FcR− MM.1S cells at limiting E:T ratios. Importantly, NWL, NWS, and NMS, but not NMLenriched CART cells, were as cytotoxic as those enriched through CH2CH3 ( Figure 2C). The maintenance of potent antitumor reactivity was confirmed by cytokine production in response to MM.1S cells ( Figure 2D) and, most importantly, by efficient killing of both primary AML blasts and malignant plasma cells ( Figure 2E).
Importantly, while exposing NGFRenriched CD44v6 CART cells to CD44v6+ MM.1S cells induced secondary proliferation, add ing soluble NGF at concentrations capable of forcing neuronal tumorcell differentiation had no effect (Figure 3), ruling out a potential proliferative advantage of NGFRenriched CD44v6 CART cells upon encounter with its natural ligand NGF.

ngFr-enriched cD44v6 car-T cells Mediate Potent Therapeutic effects against high-leukemia and Myeloma Burdens in nsg Mice
We next sought to validate the antitumor efficacy of NGFR enriched CD44v6 CART cells by stresstesting them in vivo against hightumor burdens. Prior to that, in order to identify which NGFR isoform(s) to choose as candidate(s) for further vali dation, we screened them in a minimal residualdisease setting, i.e., in NSG mice challenged with THP1 cells 3 days earlier. After infusion, all NGFR isoformenriched CART cell products could be identified in the peripheral circulation by FACS ( Figure 4A). Notably, differences in CAR expression levels observed in vitro were confirmed in vivo. As expected, only mice treated with CH2CH3, NWL, NWS, or NMSenriched CD44v6 CART cells, but not those treated with NMLenriched CD44v6 CART cells, benefited from substantial antitumor effects, as indicated by normalization of THP1 cellinfiltrated liver weight ( Figure 4B).
We, therefore, proceeded to test NGFRenriched CD44v6 CART cells in vivo against highleukemia burdens by exclud ing those enriched through the NML isoform. To this aim, NSG mice were challenged with THP1 cells carrying a secreted luciferase that allows the monitoring of tumor growth by simply analyzing blood samples (24). After waiting 14 days for high leukemia burdens to develop, mice were randomized to receive NWL, NWS, or NMSenriched CART cells. All three NGFR isoformenriched CD44v6 CART cell products were capable of delaying leukemic progression ( Figure 4C) and significantly prolong animal survival (Figure 4D). According with the differ ences described in vitro, NMSenriched T cells proved the most efficacious, followed by NWLenriched T cells, which possibly   Figure 4E). In keeping with higher antitumor efficacy, only mice treated with NMS or NWLenriched CD44v6 CART cells experienced a transient increase in IL2 and TNFα the day after CART cells infusion.
We finally challenged NMSenriched CD44v6 CART cells against highmyeloma burdens and compared them with CH2CH3enriched CD44v6 CART cells. In NSG mice rand omized for hightumor burdens by means of the MM.1Sluc tech nology (see Materials and Methods), NMSenriched CART cells better engrafted and persisted longer than those enriched through CH2CH3 (Figure 5A), resulting in a superior ability to control myeloma progression ( Figure 5B) and to prolong diseasefree survival ( Figure 5C).

ngFr-spaced car-T cells are effective against Multiple Target antigens
Desirably, a novel molecular tool aimed at expanding CART cell applications should be functional across multiple target antigens.
To address this issue, we generated NWL and NMSspaced CAR constructs specific for CD19 and CEA. Importantly, after retro viral transduction, both constructs could be visualized by FACS and enabled CART cell enrichment with immunomagnetic beads (Figure 6A). Moreover, NWL and NMSenriched CD19 and CEA CART cells killed CD19+ and CEA+ tumor cells, respectively, while failed to kill cells that did not express the respective target antigen (Figures 6B,C). Similarly to CD44v6, CH2CH3enriched CD19 and CEA CART cells killed CD19 and CEA tumor cells expressing FcRs, confirming compromised specificity when using this specific CAR design. As a proof of concept of the antitumor efficacy of NGFRspaced CARs against a target antigen different from CD44v6, we xeno engrafted NSG mice with the CD19+ ALLCM lymphoid leukemia cells and later infused them with NWL or NMSenriched CD19 CART cells. In this setting, both NGFR isoformenriched CD19 CART cell products induced leukemia remissions, as ascertained in peripheral blood and lymphoid organs ( Figure 6D).

ngFr-spaced cD44v6 car T-cells can Be enriched With clinical-grade reagents
A NGFRenriched CD44v6 CART cell product coexpressing the TK suicide gene for switchingoff potential toxicities (20, 31) will soon be investigated in a phase I/IIa clinical trial in relapsed/ refractory AML and MM (ECfunded H2020 EURECART Consortium). During validation, we observed that, while all NGFR isoformspaced CD44v6 CARs bound the antiNGFR mAb C401457, only the NWLisoform was able to bind the antiNGFR mAb ME20.4 (Figure 7A), possibly as a consequence of conformational changes altering epitope accessibility. Since directly conjugated clinicalgrade immunomagnetic beads are based on ME20.4, in addition to NMS we decided to include into further process development also the NWLisoform. The two NGFR isoformspaced CD44v6 CARs were cloned along with the TK suicide gene into a clinicalgrade bicistronic retroviral vector (20), under the control of an internal SV40 promoter, which is far less potent than viral LTRs in driving transgene expression ( Figure 7B). After transduction, NWLspaced suicidal CD44v6 CART cells were enriched with directly conjugated immuno magnetic beads, while enriching NMSisoformspaced suicidal CD44v6 CART cells required a cumbersome twostep procedure, i.e., staining with the PEconjugated antiNGFR mAb C401457, followed by the use of antiPE immunomagnetic beads. In either case, suicidal CD44v6 CART cells were enriched to >90% purity, with yields of 40-60% ( Figure 7C). Quite surprisingly, however, at the end of the entire process, the disparities between the two CART cell products were completely abated. Indeed, NWL and NMSspaced suicidal CD44v6 CAR T cells displayed a similar dif ferentiation phenotype (Figure 7D), and most importantly, medi ated superimposable antitumor effects against highleukemia burden in vivo (Figure 7E), suggesting the attenuation of NWLinduced tonic signaling under the control of the internal promoter SV40. Last but not least, NWLspaced CD44v6 CAR T cells maintained efficient suicidability upon exposure to GCV, with an IC50 of 0.032 µM compared to 0.059 µM of the original clinicalgrade construct (Figure 7F).

DiscUssiOn
Despite widespread and justified excitement, the most recent results of cellular immunotherapy with CAR T cells specific for CD19 (2)(3)(4)(5)(6)(7)(8) raise important questions as to longterm antitumor efficacy and safety (36). Moreover, the successful application of CART cells to other hematological malignancies and, possibly, solid tumors remain to be demonstrated, as the choice of addi tional viable target antigens is so far limited.

FigUre 2 | Continued
If one takes a closer look at the CD19 CART cell products investigated so far in the clinic, beside differences in costimula tory endodomains, stimulation protocols and viral vectors, it is worth noting the significant intertrial and interpatient vari ability in terms of proportion of transduced cells, with average values ranging from 20 (3)(4)(5) to 60% (2,6) in different trials and  individual values ranging from 5.5 to 45.3% (5), from 5 to 60% (4), or from 55.1 to 76.8% (6) in different patients. Ideally, to better study the crucial factors influencing the efficacy and safety profiles, CARs would be designed to enable the in vivo tracking of receptorbearing T cells, their phenotypic characterization and possibly their reisolation for functional ex vivo analysis. In addition, the possibility to purify CART cells before infusion into patients would promote the design of standardized proto cols for CART cell therapies and facilitate future application in the allogeneic setting. Recently, it has been demonstrated that, though CARexpressing T cells are expected to exert antitumor effects in the absence of GVHD, the transfer of large numbers of nontransduced T cells is sufficient to increase its incidence and severity (37). A critical role of the costimulatory endodomain has been also highlighted, with 41BB proved more prone to induce lethal GVHD than CD28 (37). Therefore, even if recent clinical studies have reported minimal GVHD incidence in recipients of allogeneic CD19 CAR.28z T cells for the treatment of Bcell malignancies (38), it cannot be excluded a different outcome when changing the treatment schedule, tumor type and CAR construct. In this scenario, having the possibility to purify T cells expressing the CAR and possibly a suicide gene before infusion into patients may drastically increase the safety of the strategy.
Recently, introduction of Streptag II sequences into specific sites of synthetic CARs and natural TCRs of diverse specificities have been proposed for the easy identification and rapid purification  of fully functional engineered T cells (39). In this work, we proposed, as a valuable alternative, the inclusion within the CAR backbone of a novel extracellular spacer based on the lowaffinity NGFR. Moreover, we demonstrated the possibility to exploit the NGFR spacer to sort functional CART cells coexpressing the TK suicide gene using clinicalgrade vector and reagents, paving the way for a successful exploitation of this strategy in the allogeneic setting.
A truncated version of NGFR, lacking its intracellular signaling domain, has been extensively tested in clinical trials in combination   . Forty days later, mice were sacrificed and their liver weighted. THP-1 infiltrated liver weights are shown for the different cohorts (each symbol represents a single mouse and the median value for each group is reported). The dashed area depicts the range of normal liver weight from age/sex-matched normal NSG mice. A representative experiment (of n = 3) is shown. Results from a one-way ANOVA test are indicated when statistically significant (*P ≤ 0.05; **P ≤ 0.01). (F) Growth suppression of CD44v6 suicidal CAR T cells analyzed after activation with PHA and treatment with increasing concentrations of GCV (means ± SEM from a representative donor). T cells transduced with the original vector expressing the tk suicide gene and the selection marker ΔNGFR were used as comparison.

FigUre 7 | Continued
with other therapeutic genes (like the TK suicide gene), proving its suitability as a safe and effective marker for Tcell selection (20,31). Importantly, we here show that, when inserted within the CAR molecule, NGFR is incapable of signaling in response to soluble NGF, even at manifold supraphysiological concentra tions, eliminating concerns about uncontrolled outgrowth of NGFRenriched CART cells upon potential in vivo growthfactor encounter. In addition, when inserted into the CAR molecule, it does not interfere with the suicidal machinery upon coexpression with the TK suicide gene. CARs are modular structures, where all domains may affect CART cell performances (40). Accordingly, the different NGFR isoformspaced CARs investigated in our study showed functional disparities. For instance, the NWS and the NML design were less expressed than the others, possibly as a consequence of defective protein stability. Accordingly, they showed suboptimal antitumor activity. The NML design in particular failed to show antitumor activity in vitro, suggesting defective interaction with the target antigen. On the other hand, the NWLisoform CAR, which is the one expressed at higher levels and the unique that could be reproducibly enriched with clinicalgrade immunomagnetic beads directly conjugated with an antiNGFR mAb (clone ME20.4), suffers from tonic signaling, which led to premature Tcell aging and reduced antitumor acti vity in vivo. It was relevant, however, that changing the expression platform and the manufacturing procedure restored the in vivo antitumor activity of NWLisoform CART cells, underlining that multiple variables have to be taken into account when designing a new CAR molecule.
While spacers are required for CART cell targeting of tumor cell membraneproximal epitopes (41,42), the same spacers have been shown to have neutral (27,40) or even unfavorable (43) effects on the recognition of membranedistal epitopes. Interestingly, the NGFR extracellular portion has a modular structure, characterized by repeated TNFR cysteinerich domains, allowing the future evaluation of the possibility to adapt spacer length to different specificities by acting on this modular structure. In this work, for proofofconcept of the functionality of NGFRbased spacers, we chose as target antigen the CD44 isoform variant 6 (CD44v6), which is knowingly overexpressed in AML (44) and MM (45), as well as on a variety of epithelial tumors (46), and has been implicated in tumor progression and resistance to radiochemotherapy (47). This choice provides the unique opportunity of tackling multiple tumor indications with a single CART cell product, while simultaneously lower ing the probability of immune evasion, as observed with CD19 CART cells (48,49). We recently developed a CD44v6 CAR and proved its efficacy against AML and MM (25). Regretfully, however, our original CD44v6 CAR included an IgG1CH2CH3 spacer, which may unduly interact with FcRexpressing myeloid cells, driving offtarget activation of CART cells (26). As a consequence, it has been demonstrated that, once infused into NSG mice, CH2CH3spaced CART cells fail to exert significant antitumor activity because of their premature clearance from the blood stream upon interaction with soluble FcRs (30) or/and trapping in the lung upon interaction with resident phagocytes (27,29). Interestingly, while we did not observe this detrimental effect of the IgG1CH2CH3 spacer against lowtumor burdens (25), we clearly observe it against hightumor burdens, where possibly the number of CART cells available for in vivo tumor cell killing becomes limiting. Several approaches have been proposed to date for abating the offtarget effects of CH2CH3 spacers, including their partial or complete removal, or their mutagenization (27,29,30). Compared to them, however, while maintaining potent therapeutic effects against hightumor bur dens, the NGFR spacer technology has the advantage of being fully compatible with clinicalgrade immunomagnetic beads. Last but not least, similar results to that obtained with CD44v6 were achieved with two additional target antigens, CD19 and CEA, validating the potential universality of NGFR spacers as safe and effective tools for CART cell enrichment before infusion.
In summary, we have developed and validated a technology enabling the enrichment of CART cells through an NGFR spacer by means of clinicalgrade immunomagnetic beads.
A CD44v6 CART cell product implementing this technology and coexpressing the TK suicide gene for switchingoff potential toxicities (20,31) is currently at late stage of process develop ment and will soon be investigated in a phase I/IIa clinical trial in relapsed/refractory AML and MM (ECfunded H2020 EURE CART Consortium).

eThics sTaTeMenT
This study was carried out in accordance with the recom mendations of the "IACUC: Institutional Animal Care and Use Committee (IACUC #468). " The protocol was approved by the "Italian Ministry of Health. "