Promising Chimeric Antigen Receptors for Non-B-Cell Hematological Malignancies, Pediatric Solid Tumors, and Carcinomas

CD33

The stem cell surface antigen CD33 was identified as the target of the antibody L4F3, which showed binding to and elimination of leukemic blasts (but preservation of stem/progenitor cells) in vitro upon the addition of complement. High impact studies using genetic lineage tracing and serial transplantation identified CD33 as a marker of stem cells capable of transferring leukemia to immunodeficient mice (so-called leukemia-initiating cells [LICs]) in some, but not all, patients. Higher-affinity anti-CD33 clones were used to engineer the approved antibody-drug conjugate (ADC) gemtuzumab ozogamacin. Initially in relapsed patients, then in combination with induction regimens, gemtuzumab showed efficacy in some trials, but not in others. Retrospective cross-trial comparisons showed that trial design and drug dosing varied along with toxicity; retrospective evaluation identified a subset of patients with low levels of CD33 expression that did not benefit from addition of gemtuzumab. Thus, CD33 was among the first recognized targets of antileukemic antibodies in AML.

Early studies in adoptive T-cell therapy applied a single-chain variable fragment (scFv) derived from a rat-anti-human CD33 antibody to a first-generation CAR design (i.e., CD33 scFv-CD28 transmembrane domain- $CD3\zeta$ signaling domain) to generate retroviral vector (RV)–transduced cytokine-induced killer cells. These cells showed modest in vitro efficacy but were not tested in patients. Subsequent studies using second-generation CAR designs (i.e., including a CD28 or 4-1BB costimulatory domain in addition to CD3 $\zeta$ ) showed improved cytokine release, higher rates of in vitro tumor killing, and, when injected into immunodeficient mice, elimination of tumor xenografts. These studies formed the basis of preclinical testing and FDA approval for clinical trials. Subsequent mouse studies have developed combination CRISPR/CAR strategies wherein hematopoietic cells are rendered resistant to CD33 CAR T-cells through inactivation of CD33 by CRISPR. This strategy would reduce or eliminate the concern for myelosuppression by CD33 CAR T-cells, but gene-edited hematopoietic stem cell transplants remain a technology still in development.

Despite the preclinical efficacy of CD33 CAR T-cells, patient experience has been limited. A single report of one 41-year-old male patient with AML describes the clinical experience following escalated doses of a cellular product manufactured from 90 mL of peripheral blood that was comprised nearly entirely of CD8 T-cells. Following treatment, the patient developed elevated levels of IL-6, TNF $\alpha$ , IL-8, IL-10, and IFN γ and worsened pancytopenia that fluctuated inversely with the patient's fever. Two weeks after infusion, the patients' marrow blasts had decreased from 50% to <6%, after which the blast fraction increased, resulting in relapse. Transient hyperbilirubinemia was observed but resolved without specific intervention. Levels of CAR T DNA measured in both peripheral blood and marrow remained high, implying CAR T-cell persistence. To gain insight to the mechanism of relapse, the authors found CAR T-cells recovered from the patient retained the ability to kill CD33 ⁺ target cell lines ex vivo . Nevertheless, the immunophenotype of the AML blasts at the time of relapse remained CD33 ⁺ , suggesting a mechanism of immune escape that was either specific to the blasts or the in vivo environment.

An alternative approach using an NK cell line transduced with a CD33 CAR construct showed in vitro efficacy and tolerability in three patients, but none achieved a complete remission without measurable residual disease. Despite transplantation in the two responding patients, both relapsed: one after 15 months and the second after 4 months. Other phase I trials are in progress, including a trial at the University of Texas MD Anderson Cancer Center, but have yet to report outcomes.

CD123

The IL-3 receptor (CD123) is an early stem cell marker that efficiently defines LICs. In humans, the fraction of AML cases expressing CD123 in early progenitor populations ranges from 70% to >95%. In vitro treatment of leukemic marrow samples with an antibody to CD123 eliminated the LIC population through targeted complement activation. This led to suggestions that, in most cases, all or nearly all of the LIC population was CD123 ⁺ .

Attempts to target CD123 in myeloid neoplasms include an engineered dual-affinity molecule CD123 × CD3 (flotetuzumab) that retargets T-cells as well as a ligand-toxin conjugate linking IL3 to diphtheria toxin (SL-401, tagraxofusp). Flotetuzumab showed modest efficacy in initial phase 1 studies, with 5 of 27 (18.5%) of patients treated at the recommended phase 2 dose showing complete response or complete response with incomplete count recovery. In a related neoplasm, blastic plasmacytoid dendritic cell neoplasm (BPDCN), high CD123 expression is a diagnostic hallmark and outcomes are very poor, with median overall survival ranging from 2 to 10 months depending on immunophenotype. A planned interim analysis of a single-arm phase 2 trial evaluated 47 patients with BPDCN treated with tagraxofusp found complete remission rates above 80%, high rates of subsequent stem cell transplantation rates, and 18 and 24 month survival rates of 59% and 52%, respectively. These data were sufficient for the FDA to approve the drug for BPDCN, and trials in other myeloid leukemias are ongoing.

The clinical experience with CD123 CAR T-cells is limited, but promising. Initial studies conducted at City of Hope Cancer Center (Duarte, CA), identified a second-generation CAR utilizing a CD28 costimulatory domain that, when transduced into T-cells via lentiviral vector (LV), showed activity against CD123-expressing targets, including primary patient AML samples in vitro. Mouse xenograft models showed transient reductions in tumor bulk following CAR T-cell administration and delayed mortality. In addition, treatment of cord blood samples with CD123 CAR T-cells did not reduce the number of colony-forming units, suggesting the safety of this approach.

The first reported trial of CD123 CAR T-cells treated seven patients, six with refractory AML following allogeneic stem cell transplant and one with BPDCN. Escalating doses (50 × 10 ⁶ cells, 2 pts; or 200 × 10 ⁶ cells, 4 pts) of CAR T-cells were administered following fludarabine plus cyclophosphamide lymphodepletion. Importantly the authors reported that there were no treatment-related cytopenias. One of the two patients treated at the lower dose achieved a CR (blast percentage 0.9%) after two infusions of CAR T-cells; two patients treated at the higher dose achieved a CR without MRD and were subsequently treated with a second allogeneic stem cell transplant. The remaining four patients did not respond to treatment.

Safety concerns, based on preclinical data that more potent CD123-targeted CARs prevented or impaired human hematopoiesis in mouse xenograft models, led researchers at the University of Pennsylvania to develop a system in which T -cells were electroporated with RNA encoding the CAR, resulting in transient activity. In a phase 1 clinical trial, involving seven patients with AML (four with prior allo-HSCT), only two of six patients received all planned doses due to manufacturing challenges, and all five patients who were treated with CAR T-cells had disease progression before day 28, with no observed expansion of CAR T-cells within the blood or reduction in the number of circulating CD123-positive cells. These results led the authors to indicate their intention to abandon the strategy of RNA-based CAR T-cell manufacture in favor of elimination of lentivirus-transduced CAR T-cells using alemtuzumab followed by rescue allogeneic transplant.

Others have focused on further preclinical design strategies that tune levels of expression and CAR affinity to distinguish tumor from normal cells or to incorporate inducible suicide switches. A single report of a safety switch containing CD123 CAR T enhanced with CD27 costimulation (CD123 scFv—CD28 transmembrane—4-1BB—CD27— $CD3\zeta$ —iCasp9) was tested in a single 47-year-old male patient. The patient had FLT3 ⁺ AML that had relapsed following allogeneic transplant. The patient was conditioned with cyclophosphamide prior to administration of 1.8 × 10 ⁶ CAR T-cells/kg. On day 8 postinfusion, the patient developed severe cytokine release syndrome (CRS), accompanied by high levels of IL-6 and TNF $\alpha$ that was controlled by tocilizumab. Bone marrow examination 20 days following infusion showed a decrease in blast fraction from 59% to 45%. The authors report that this result warrants further investigation.

Clinical trials investigating CD123 CAR T-cells are currently registered at MD Anderson Cancer Center, University of Pennsylvania, Cornell Weill Cancer Center and City of Hope in addition to 12 institutions in China. The use of dual-targeted CD123 × CD19 CAR T-cells has been reported in three patients with refractory acute lymphocytic leukemia and appears tolerable and efficacious (all three patients achieved a complete remission, two of which were MRD negative); however the relative contribution of CD19 and CD123 CAR targeting is unclear.

CLL1/CLEC12A

C-type lectin-like molecule 1 (CLL1/CLEC12A) is a surface glycoprotein that is expressed on approximately 90% of leukemic cells with early/progenitor phenotype. Both an unbiased phage-display/cloning approach and a comprehensive multiomics approach identified this protein as highly expressed on both leukemic blasts and LICs and absent or minimally expressed on hematopoietic stem cells. Immunophenotyping showed expression of CLL1 on committed CD33 ⁺ CD34 ⁺ myeloid cells in bone marrow and peripheral blood, specifically granulocytes and monocytes as well as mature and precursor dendritic cells, but not lymphocytes or NK cells. No other tissues analyzed expressed CLL1 mRNA. In addition, reengrafting normal HSC failed to express CLL1, whereas engrafting AML blasts showed expression of CLL1 with the ratio of CLL1 ⁺ to CLL1 ⁻ cells among CD34 ⁺ CD38 ⁻ cells predictive of the time to relapse.

Three groups have provided preclinical data describing CAR constructs targeting CLL1, which have activity against in vitro targets and prolonged survival of xenografted NOD/SCID/IL2-gamma knockout (NSG) mice. In one case, the authors used a standard second-generation CAR construct (scFv-CD8 transmembrane domain-41BB- $CD3\zeta$ ) ; in another, the authors used a murine costimulatory domains and a backbone that incorporated both CD28 and 41BB costimulatory domains ; and in the third case, the authors used a standard second-generation CAR construct with an added iCasp9 (inducible caspase-9) safety switch. All showed effective killing and tumor elimination. Using NSG mice that additionally express GM-CSF, IL3, and stem cell factor (NSG-S), Kenderian et al. showed CLL1-targeted CAR T-cells can eliminate a leukemic stem cell population that is resistant to chemotherapy.

In vivo studies with CLL1-targeted CAR T-cells are limited to a single case series of compound CD33-CLL1 CAR T-cells in which an LV encoding two complete CAR constructs joined by a cleavable linker (P2A) was used. A single pediatric patient with Fanconi anemia that had progressed to AML and was refractory to five cycles of chemotherapy was treated with lymphodepletion followed by two doses of compound CD33-CLL1 CAR T-cells (1 × 10 ⁶ per kg, days 1 and 2). On day 12, the patient's blast count was 98% and developed signs of mild CRS, and then 7 days later, the repeat marrow showed a complete remission without measurable residual disease and the patient was subsequently treated with a nonmyeloablative allogeneic stem cell transplant. Outcomes for additional patients reported to have been treated are not yet publicly available.

Lewis-Y/CD44

Early studies designed second-generation CAR constructs to the Lewis-Y antigen, a carbohydrate with increased expression on leukemic blasts. The numerous challenges of targeting carbohydrate antigens include the following: glycosylated structures are present on nearly all mammalian cells, detection of the target difucosylated oligosaccharide depends on the specificity of the antibody, and RNA expression and routine mass spectrometry cannot detect the specific stereochemistry recognized by anti-Lewis-Y antibodies. Nevertheless, differential glycosylation is common on tumors, and generation of positive cell lines (and a negative control) can be accomplished by introduction of the FUT2 gene responsible for fucosyltransferase activity. However, this enzyme may mediate glycosylation of multiple glycolipids, carbohydrates, and proteins. CD44 is one of several proteins modified by fucosylation, copurifies with Lewis-Y, and has been associated with stemness and invasive properties of various tumors, including ovarian and breast carcinoma, in part mediated by increased expression of adhesion molecules and increased secretion of metalloproteinases. CD44 binds to various glycans (including hyaluronic acid) and specific spliced isoforms, i.e., those that include exon 6, are further enriched on AML and myeloma and maybe required for bone marrow homing and in vivo tumorigenesis.

A second-generation CAR composed of an anti-Lewis-Y scFv fused to a CD28 transmembrane and costimulatory domain followed by $CD3\zeta$ showed activity against both ovarian cancer and AML cell lines in vitro and in mouse xenografts . Importantly, these studies demonstrated minimal off-target effects, as measured by lack of reactivity against autologous neutrophils that express approximately 20% the amount of Lewis-Y antigen that breast cancer cell expresses. A phase 1 trial enrolled five patients with relapsed AML. One patient died during an attempt at reinduction and did not receive CAR T-cells. Patient conditioning was with fludarabine (30 mg/m ² days 1–5) and high-dose cytarabine (2 g/m ² days 1–5), and the cell product was infused following marrow recovery in patients with residual disease. The four infused patients received between 1.5 and 9.2 × 10 ⁶ transduced CAR T-cells per kg. Two of the four patients had postinfusion reaction that included neutropenia in one patient and fevers and chills in a second. Two patients had stable disease (one with stable MRD that lasted 23 months, the other with relapse at day 49), and one patient has a transient reduction in blasts followed by relapse after 28 days. A fourth patient had a cytogenetic complete remission but relapsed 5 months after infusion. The fourth infused patient had no benefit. A fraction of the CAR T-cells labeled with indium showed tracking to sites of disease in the three patients with discernible benefit; furthermore, the two patients with the greatest benefit also had the most robust CAR T-cell expansion, though interestingly the patient with the complete response did not show high levels of serum cytokines. Samples obtained at relapse showed expression of Lewis-Y antigen at levels comparable with levels observed at diagnosis. None of the patients reported in the manuscript were described to have completed an allogeneic stem cell transplant following therapy, which may have provided some benefit.

Current trials of Lewis-Y-targeted CAR T-cells are limited to a single site in China and a trial by the original authors treating solid tumors. Improvements in CAR T-cell manufacture may lead to improved responses and outcomes. Severe CRS was not seen, and this may be interpreted either as evidence of safety, or of a somewhat inert product. Preclinical studies investigating CAR constructs targeting CD44v6 were reported, but these have not been explored in patients. The field seems to have paused, but interest in this CAR construct may be resurrected if improved designs could be identified using more sophisticated preclinical models.

NKG2D

In response to genotoxic or metabolic stress including malignant transformation, cells may increase expression of one or several of eight ligands bound by NKG2D, an activating protein receptor whose expression is mostly limited to cytotoxic lymphocytes, including NK cells, NKT-cells, $\gamma \delta$ T-cells, and CD8 ⁺ T-cells. NKG2D ligands all share structural homology to the $\alpha 1$ and $\alpha 2$ domains of MHC class 1 but exhibit extensive allelic diversity (i.e., one ligand, MICA, has 109 alleles).

While activation of endogenous NKG2D proceeds with induction of a switch in RNA isoform, which results in a protein capable of binding adaptor molecules that in turn activate PI3 kinase, the CAR design tested by the Nikifarow and colleagues is distinct: to maintain sensitivity to the wide variety of NKG2D ligands, the entire NKG2D protein (rather than an scFv) was used as the targeting moiety and the $CD3\zeta$ domain without a costimulatory domain was fused to the cytoplasmic tail. The recombinant NKG2D/ $CD3\zeta$ fusion protein bound the endogenous adaptor protein Dap10, which provides an additional costimulatory signal to RV-transduced T-cells. Preclinical studies showed robust cytokine production from transduced T-cells, lysis of NKG2D ligand-expressing tumor cells lines, and limited inhibition by soluble NKG2D ligands. Perhaps most interestingly, immunodeficient mice that had eliminated xenografted tumors following adoptive transfer of NKG2D-CAR-T-cells were resistant to tumor rechallenge. A recent phase 1 trial featured 14 patients, 12 of whom provided T-cells for CAR manufacture, of which 7 had AML. Patients with prior allogeneic transplant were excluded. Because inflammation may increase NKG2D ligand expression on normal tissues, no lymphodepleting chemotherapy was used prior to cell infusion as an additional safety parameter. No infusion reactions, CRS, neurotoxicity, colitis, or pneumonitis was observed, and all adverse events of grade ≥3 were attributed to underlying disease. No objective clinical response was seen in any patient, and all went on to receive additional therapy. Three patients experienced short-lived (3–6 months) stable disease under conditions in which progression was expected.

Because of the wide variety of tumors found to express NKG2D ligands, the safety of the initial clinical studies, and the efficacy seen in preclinical mouse studies, further investigations are ongoing.

CD30/CD37

Some TAAs are shared between T-cell and B-cell neoplasms, including CD30 and CD37. The CD30-targeted immunoconjugate brentuximab vedotin has shown efficacy in CD30-expressing T-cell lymphomas including angioimmunoblastic T-cell lymphoma and anaplastic large-cell lymphoma ; however, subtypes of peripheral T-cell lymphoma show variable expression of CD30 and decreased responses to brentuximab. These data suggest that CAR targeting of this antigen may be more potent. Constructs targeting the CD37 antigen have been shown to be effective in vitro but have not been tested in patients with T-cell malignancies.

CD30-directed CARs has been tested in two phase 1 trials and shown tolerability and success. The first trial, conducted in China, used an LV to deliver a CAR construct containing a 4-1BB costimulatory domain. Eighteen patients with non-Hodgkin lymphoma (one of which was anaplastic large-cell lymphoma) were infused following one of four chemotherapy conditioning regimens, the most frequent of which was fludarabine/cyclophosphamide; however, the patient with ALCL did not receive pretherapy conditioning. A 57% reduction in skin lesions was noted, and the patient was judged to have a partial response that lasted 3 months. Two patients had adverse events of grade ≥3, both of which were judged to be due to disease and/or prior chemotherapy. Thus, this treatment showed safety and promise.

The second trial included nine patients, two with anaplastic large-cell lymphoma. This CD30 construct used a CD28 costimulatory domain, CAR T-cells were manufactured via RV and were infused without chemotherapy conditioning. No adverse events related to CAR T-cell infusion were documented, and no cases of CRS were observed. Of nine patients, three achieved a complete remission following CAR T-cell therapy, two of which extended beyond 2 years. Three patients had transient stable disease or partial responses but progressed within 6 months, and three did not benefit from treatment. There was an inverse correlation between soluble CD30 at the time of infusion and peak CAR T-cell expansion, suggesting that soluble antigen may dampen CAR T-cell reactivity in vivo, even though soluble antigen did not block CAR T-cell reactivity in preclinical testing. The authors also examined PD1 expression on CAR T-cells but found no correlation between expression levels and response. In sum, expanding the indication of the CAR product already tested in NHL may be one strategy to deliver the promise of this technology to patients with T-cell leukemia.

CD7

While coexpression of a CAR protein and the cognate tumor antigen has been shown to lead to retention of the tumor antigen within the endoplasmic reticulum, an alternate approach is to intentionally delete the tumor antigen as well as the T-cell receptor from allogeneic T-cells constructing a “universal” product that would not result in graft-versus-host disease. While CD7 is expressed on ∼30% of AML cells and nearly all T-cell leukemias, there is controversy as to whether the population of LICs is CD7 positive or CD7 negative ; its coexpression on normal T-cells prevents production of CD7-directed CAR T-cells via the above mechanism. A recent study described preclinical data showing the feasibility of combining CRISPR-mediated gene editing and viral transduction to produce CD7 ^KO CD7 CAR T-cells. These cells were efficiently transduced (>70%), successfully gene edited, and reactive against tumor cell lines in vitro and in vivo . In murine models of T-cell and myeloid leukemias, the CAR T-cells showed protection. This approach may be more widely adopted, and a clinical trial is planned.

CD5

Early studies using CAR T-cells directed at CD5 and carrying a CD28 costimulatory domain showed resistance to fratricide through increased BCL-2 expression and downregulation of surface CD5 expression. These cells showed reactivity against CD5-expressing tumor cell lines and primary tumors; xenografted mice treated with CD5 CAR T-cells showed initial tumor clearance followed by recurrence. These studies are being followed up with further attempts at engineering CD5 ^KO T-cells similar to the description of CD7 above and with a clinical trial, which has not yet reported outcomes.

CD4

Pinz et al. generated CAR T-cells directed at the CD4 antigen by selecting CD8 ⁺ T-cells for transduction. Cells isolated from normal peripheral blood or from umbilical cord blood showed reactivity against CD4 ⁺ tumor cell lines and delayed tumor progression in mouse xenograft models. One challenge to this approach is that patients would be expected to develop immune suppression similar to that seen in the CD4-depleting viral infection HIV. Thus, the authors developed a transduced NK cell approach using the same CAR backbone but using a GMP-qualified NK cell line (NK-92) as the effector cell. These cells react similarly to CAR-transduced T-cells but persist in vivo for approximately 2 weeks. The authors report that transduction of NK-92 cells resulted in in vitro killing of CD4 ⁺ tumor cell lines and delayed tumor progression in mouse xenograft models but showed no toxicity to hematopoietic stem cells when assessed by methylcellulose colony formation. These results are being followed up in a clinical trial.

GD2

One of the most exciting recent advances in high-risk neuroblastoma came with the incorporation of immunotherapy targeting the neuroblastoma antigen disialoganglioside (GD2). This target is uniformly expressed on tumor cells, and a randomized phase 3 study incorporating the chimeric monoclonal antibody dinutuximab after consolidative myeloablative chemotherapy demonstrated a 20% improvement in event-free survival compared with patients receiving standard-of-care therapy. Toxicities associated with this therapy were significant, however, and up to 40% of patients experience grade 3 or greater neuropathic pain, likely from antibody binding and complement activation on peripheral nerves, which also express surface GD2.

Based on both the promise and limitations of antibody-based immunotherapies, CAR T-cells targeting GD2 have been a long-standing interest in neuroblastoma. Investigators at Baylor College of Medicine were the first to present preclinical models of a first-generation CAR (containing only the CD3ξ endodomain without any other costimulatory domains) with in vitro antineuroblastoma activity. The first clinical trial incorporating this GD2 CAR attempted to enhance costimulatory activity via engagement of the native TCR by transducing the CAR into Epstein-Barr virus (EBV)–specific CD8 ⁺ cytotoxic T lymphocytes (CTLs). Persistence of the adoptively transferred product was enhanced in EBV-specific CTLs compared with CAR-transduced polyclonal T-cells, and persistence was enhanced by the presence of CD4 ⁺ or CD45RO ⁺ CD62L ⁺ central memory T-cells in the adoptively transferred product. Of 19 patients treated, 2 patients experienced long-term remissions, but all patients eventually developed progressive disease, and time to progression correlated with CAR T-cell persistence.

The use of viral-specific T-cells is appealing because it leverages native TCR functionality and theoretically provides physiologic stimulation. However, preclinical data suggest that target engagement simultaneously through a CAR and a TCR can lead to decreased persistence and T-cell exhaustion. Most CAR constructs now use embedded costimulation. In preclinical studies, a third-generation CAR containing both the CD28 and OX40 costimulatory domains demonstrated that incorporation of tandem costimulation domains increased T-cell expansion and improved cytokine release. This third-generation CAR construct was evaluated in a single-institution phase 1 study at Baylor where patients with relapsed or refractory disease were treated in one of three cohorts: GD2 CAR T-cells alone, GD2 CAR T-cells following lymphodepleting chemotherapy with fludarabine and cyclophosphamide (Flu/Cy), and GD2 CAR T-cells following Flu/Cy and the anti-PD-1 checkpoint inhibitor pembrolizumab. While subjects who received lymphodepleting chemotherapy and checkpoint blockade had increased persistence when compared with those who did not receive these therapies, antineuroblastoma efficacy was limited in all cohorts with no objective responses seen. While persistence appeared to improve with the addition of costimulatory domains, clinical activity did not appear to improve compared with first-generation CARs or viral-specific CARs. Others have postulated that the scFv used for this CAR construct, 14g2a, can aggregate on the surface of CAR T-cells and lead to tonic signaling and T-cell exhaustion and that specific costimulatory molecules may accelerate this process.

In an effort to overcome obstacles encountered with the 14G2a scFv, including tonic signaling and a risk for antibody-mediated CAR rejection from patients with previous exposure to the 14g2a-based chimeric antibody dinutuximab, investigators have looked at alternate scFvs to target GD2. Based on promising preclinical activity of an alternative CAR construct incorporating the humanized murine scFv KM8138 fused to the CD28 costimulatory domain and the CD3ξ endodomain, investigators in the United Kingdom are enrolling patients with relapsed and refractory neuroblastoma to receive this second-generation CAR. Preliminary results of this trial presented in abstract format demonstrated minor clinical responses, limited CAR persistence, and CRS in one patient.

Despite the challenges faced by GD2 CAR, it is an active area of continued focus. Several approaches to improve persistence and antineuroblastoma activity are ongoing. Investigators in Shenzen, China, are evaluating a fourth-generation CAR (including CD27, CD28, and 4-1BB costimulatory domains fused to the CD3ξ endodomain) in patients with relapsed/refractory neuroblastoma and have presented early findings, which include two near-complete regressions of bulky disease and a 15% partial response rate in the first 34 patients treated. In addition, optimization of the GD2 scFv to improve binding affinity and decrease aggregation to improve antitumor activity has been evaluated. Investigators at the Children's Hospital of Philadelphia created two novel GD2 binders based off of the 14G2a construct. Changes to both the scFv linker and affinity structures were evaluated, and the high-affinity scFv GD2-E101K led to increased cytokine production and antineuroblastoma activity in vitro compared with 14G2a. However, in vivo xenograft experiments of GD2-E101K induced a fatal encephalopathy in all CAR-treated mice. This study highlights one potential risk of creating highly specific CAR binders in solid tumors, where TAAs usually have some normal tissue expression and off-tumor on-target toxicities remain a challenge.

L1-CAM (CD171)

Based on the known toxicity profile of GD2-targeted therapies and the concern that a CAR that is effectively able to effectively target GD2 may have significant neurotoxic effects, our group has evaluated an alternative CAR target in neuroblastoma, L1-CAM or CD171. L1-CAM expression has been associated with recurrent NB and found to correlate with tumor progression and metastasis in several types of cancer, including colon carcinoma, malignant glioma, cutaneous malignant melanoma, ovarian carcinoma, prostate cancer, renal cell carcinoma, and uterine carcinoma. Furthermore, L1-CAM has been found to participate in the regulation of tumor cell differentiation, proliferation, migration, and invasion.

City of Hope in collaboration with Fred Hutchinson Cancer Research Center (FHCRC) and Seattle Children's Hospital (SCH) performed the first-in-human pilot study and assessed the feasibility of isolating and the safety of infusing, autologous CD8 ⁺ cytolytic T lymphocytes coexpressing a first-generation CD171-specific CE7 CAR and the selection-suicide expression enzyme HyTK in children with recurrent/refractory NB. The CAR used in that trial was developed by using a single-chain antibody extracellular domain (scFv) derived from the L1-CAM-specific murine CE7 hybridoma. This chimeric immunoreceptor, CE7, is specific for an epitope in the extracellular domain of L1-CAM present on NB cells and, to a limited extent, on adrenal medulla and sympathetic ganglia. Meli et al. found that inhibition of glycosylation in NB cells leads to a loss of cell surface binding sites for mAb chCE7, suggesting that CE7mAb binds to an epitope on L1-CAM that is selective for tumor L1-CAM and dependent on glycosylation. Although this first clinical study was designed to maximize patient safety, tumor response was observed. No patient experienced grade 4 or 5 adverse events associated with the administration of the genetically altered T-cells, nor were overt toxicities to tissues known to express L1-CAM, such as the central nervous system, adrenal medulla, and sympathetic ganglia, observed. Four of the six patients treated in that study had detectable levels of transferred T-cells 1 week after the first infusion. Interestingly, the patient who experienced a partial tumor response following adoptive transfer showed persistence of the transferred T-cells up to 6 weeks after the second infusion. The patient was also the only patient to have minimal residual/stable disease upon administration of the first infusion. Among the six patients who underwent adoptive T-cell therapy, one experienced prolonged survival, succumbing to disease four-and-a-half years after the first infusion.

Based on these encouraging preliminary results, patients are being actively enrolled on a follow-up study targeting CE7 at SCH. The study is evaluating three different CAR constructs in three different therapeutic arms: a second-generation 4-1BB CE7 CAR (arm A), a third-generation 4-1BB-CD28tm CE7 CAR (arm B), and a second-generation 4-1BB CE7 CAR incorporating the full-length IgG4 spacer domain (arm C). The goal of this study is to evaluate the safety and efficacy of each CAR construct and, within the confines of a phase 1 trial, understand differences in persistence and efficacy of each of the CAR constructs. At the time of publication, enrollment is ongoing ( NCT02311621 ).

Other targets