Strategies to Address Chimeric Antigen Receptor Tonic Signaling

Abstract

Adoptive cell transfer using chimeric antigen receptors (CAR) has emerged as one of the most promising new therapeutic modalities for patients with relapsed or refractory B-cell malignancies. Thus far, results in patients with advanced solid tumors have proven disappointing. Constitutive tonic signaling in the absence of ligand is an increasingly recognized complication when deploying these synthetic fusion receptors and can be a cause of poor antitumor efficacy, impaired survival, and reduced persistence in vivo In parallel, ligand-dependent tonic signaling can mediate toxicity and promote T-cell anergy, exhaustion, and activation-induced cell death. Here, we review the mechanisms underpinning CAR tonic signaling and highlight the wide variety of effects that can emerge after making subtle structural changes or altering the methodology of CAR transduction. We highlight strategies to prevent unconstrained tonic signaling and address its deleterious consequences. We also frame this phenomenon in the context of endogenous TCR tonic signaling, which has been shown to regulate peripheral tolerance, facilitate the targeting of foreign antigens, and suggest opportunities to coopt ligand-dependent CAR tonic signaling to facilitate in vivo persistence and efficacy. Mol Cancer Ther; 17(9); 1795-815. ©2018 AACR.

Figure 1

Iterative design of first, second, third and fourth generation CARs. CARs are modular fusion receptor dimers that comprise (from N-terminus to C-terminus) an extracellular targeting moiety (typically an scFv) fused to a spacer (such as an IgG1 hinge & CH₂-CH₃ domains), a transmembrane domain (such as CD8α or CD28) and a signalling endodomain. First generation CARs fused the scFv to a CD3ζ, CD3ε or FcγR activation domain. Second generation CARs contain an additional intracellular costimulatory domain (such as CD28, 4-1BB, OX40 or ICOS) to recapitulate signal 2 for T-cell activation. Third generation CARs combine two or more costimulatory domains in cis. Fourth generation CARs are engineered with an activation inducible element such as an NFAT-responsive expression cassette to facilitate secretion of a transgenic cytokine such as IL-12. CSD, costimulatory domain; ICD, intracellular domain; NFAT, nuclear factor of the activated T-cell; scFV, single chain variable fragment; TMD, trans-membrane domain. Figure 1 and figures 3-5 are original and have been created specifically for this article.

Figure 2

Endogenous TCR tonic signalling facilitates T-cell differentiation & effector function. Circulating naive T-cells interact with steady state dendritic cells (DCs) in secondary lymphoid organs. High affinity interactions between the TCR and MHC presenting self-peptide mediate peripheral T-cell tolerance, clonal editing, anergy and Treg induction. Low to intermediate affinity interactions enhance basal TCR tonic signalling via CD3ζ and ZAP70 phosphorylation leading to a reduction in the T-cell activation threshold prior to encountering foreign antigen. Subsequent encounters with activated DCs result in enhanced clonal proliferation, cytokine release, cytotoxic granule formation (via hedgehog signalling and upregulation of RAC1) and differentiation to an effector phenotype. Non-MHC-mediated T-cell / DC interactions, such as the binding of adhesion molecules (not illustrated) further facilitates tonic signalling by inducing a transient increase in intracellular Ca²⁺, cAMP and ERK phosphorylation, strengthening T-cell responses to foreign antigen. Adapted from Garbi, N. et al. Tonic T-cell signalling and T-cell tolerance as opposite effects of self-recognition on dendritic cells, Current Opinion in Immunology 22, 601–608 (2010) [12], with permission from Elsevier.

Figure 3

(a): Tonic signalling correlates with CAR surface expression and can be addressed by optimal selection of the CAR promoter during lentiviral transduction. Frigault et al. found that c-Met or mesothelin-directed second generation CARs comprising an IgG4-derived hinge, CD28 CSD and CD3ζ underwent continuous proliferation during ex vivo expansion in the absence of ligand or exogenous growth factors [21]. Continuous proliferation correlated with CAR surface expression and required CD28 costimulation. A diverse array of cytokines and chemokines were significantly upregulated, including IL-2. Also upregulated were the transcription factors T-bet, GATA3 and EOMES (a hallmark of terminal effector differentiation) as well as the pro-survival protein Bcl-xL. CAR surface expression was reduced using a truncated PGK promoter during lentiviral transduction, reducing tonic signalling and improving anti-tumour efficacy and persistence in vivo. (b): CAR tonic signalling can induce T-cell exhaustion mediated by the upregulation of inhibitory molecules and can be reversed by substitution of the intracellular costimulatory domain. Utilising a GD2-directed second generation CAR comprising an IgG1-derived hinge and CH₂-CH₃ spacer, CD28 TMD/CSD fused to CD3ζ, Long et al. were able to demonstrate that ligand-independent tonic signalling during ex vivo expansion relied upon scFv interactions, causing CAR aggregation in cell surface punctae and the upregulation of cell surface inhibitory receptors including PD-1, LAG-3 and TIM-3 leading to an exhausted phenotype and increased apoptosis [18]. The deleterious impact of this tonic signalling could be reversed by substituting the CD28 CSD with 4-1BB. GD2.BBζ CAR T-cells exhibited reduced expression of exhaustion-associated molecules and an upregulation of pathways implicated in response to hypoxia, cellular metabolism and negative regulation of apoptosis. (c): 4-1BB costimulation can mediate tonic signalling and enhanced proliferation during ex vivo expansion. Milone et al. have demonstrated that during ex vivo expansion using anti-CD3/anti-CD28 coated magnetic beads, CD19.BBζ CAR T-cells exhibited a prolonged blast phase associated with higher rates of proliferation than corresponding 28ζ and 28BBζ CARs [62]. Enhanced proliferative capacity (but not persistence) was lost approximately 2 weeks following bead expansion. BBζ CARs produced both IL-2 and IFNγ (albeit at a lower level than 28ζ CARs) and significantly reduced levels of IL-4 and IL-10, consistent with skewing to a Th1-like phenotype. The picture is suggestive of an interaction between the 4-1BB costimulatory ICD and downstream mediators of TCR activation. The authors suggest that dysregulation of CD3ζ ITAM phosphatases (such as SHP1 or PTPH1) may be playing a role. The possibility of scFv domain swapping in this CD19 FMC63 model also remains uncertain. (d): 4-1BB costimulation can facilitate CAR tonic signalling via TRAF2 and NF-κB leading to Fas-related AICD, exacerbated by self-amplification at the level of the CAR promoter. Contrary to Long et al. [18], Gomes-Silva et al. have reported that a second generation CD19-directed CAR comprising a CD8α stalk and TMD, 4-1BB and CD3ζ ICDs expanded poorly ex vivo due to tonic signalling mediated by an interaction between the 4-1BB ICD and TRAF2 [73]. This led to activation of NF-κB, upregulation of Fas and Fas ligand and ICAM-1, ultimately causing caspase-8-mediated AICD. An additional effect on the γ-retroviral LTR promoter was also noted, causing a positive feedback loop via CAR self-amplification. This phenotype could be eliminated by mutating the TRAF2 binding site on 4-1BB at the expense of effective costimulation. Interestingly, the addition of a CD28 CSD was able to restore ex vivo expansion, overcoming the adverse effects of 4-1BB tonic signalling. Likewise, the insertion of an IRES element upstream of the LTR or transducing the CAR with a lentiviral vector and the EF-1α promoter reduced tonic signalling and restored function. (e): Alterations to the hinge and spacer domain can exacerbate tonic signalling, causing constitutive ligand-independent proliferation, terminal differentiation and poor migration in vivo. Watanabe et al. demonstrated that a second generation anti-PSCA CAR containing an IgG1 hinge and CH₂-CH₃ spacer linked to a CD28 CSD and CD3ζ was liable to bind to FcγRI and FcγRII expressed on monocytes and macrophages, resulting in pulmonary sequestration in vivo and poor trafficking into implanted tumours in NSG mice [20]. Substituting the spacer framework to IgG2 abrogated FcγR binding and improved CAR T-cell trafficking in vivo. However, the CH₂-CH₃ spacer was found to mediate CAR tonic signalling independent of ligand during ex vivo expansion, leading to constitutive proliferation, terminal differentiation to an effector memory phenotype and senescence. Utilisation of a shorter spacer could ameliorate tonic signalling without compromising cytotoxicity and improved in vivo efficacy. Figure 1 and figures 3-5 are original and have been created specifically for this article.

Figure 4

(a): Depiction of single chain variable fragment & oligomers. scFvs are inherently unstable structures due to non-covalent interactions between the heavy and light chains. They are liable to form oligomers, particularly at extremes of pH and temperature, due to domain swapping and framework interactions. Outside of their use in CARs a variety of conformations have been demonstrated, dependent upon the relative length of the peptide linker, with shorter linkers conducive to multimer formation. (b): Engineering the scFv to improve stability. The scFv lends itself to protein engineering to optimise stability and prevent oligomerisation. The primary objective is to strengthen the VH:VL interface. Options include (i) glycosylation to counter hydrophobic motifs and improve solubility; (ii) addressing the net charge of the antibody scaffold by substituting residues on either side of the CDRs; (iii) adding disulphide bridges; (iv) utilising computational modelling to improve the stability of the VH:VL interface (e.g. by substituting residues to add hydrogen bonds or to fill gaps); and (v) reverting hypermutations in framework regions to germline. VH, heavy chain; VL, light chain; FR, framework region; H1-3 & L1-3 represent complementary determining regions in the heavy & light chains respectively; Asp, aspartic acid; Trp, tryptophan; Tyr, tyrosine. Figure 1 and figures 3-5 are original and have been created specifically for this article.

Figure 5

Potential strategies to address the negative effects of CAR tonic signalling. (a) Optimal selection of the extracellular targeting moiety +/- engineering of the scFv or substitution with camelid-derived nanobodies or non-immunoglobulin based scaffolds; (b) optimisation of the hinge and spacer; (c) optimal selection of costimulatory endodomains; (d) utilising pharmacological agents to reverse or prevent negative consequences of tonic signalling (e.g. Akt inhibitors to prevent terminal effector differentiation +/- metabolic features of T-cell exhaustion); (e) engineering CAR T-cell metabolism (e.g. overexpressing PGC1α or impairing its degradation; (f) reconfiguring costimulation by overexpressing costimulatory molecules or ligands such as 4-1BBL or (g) CD28, potentially optimised by knocking down expression of Cbl-b, an E3 ubiquitin-protein ligase, that promotes anergy by regulating PI3K access to CD28; (h) optimising interactions with endogenous TCR components, which may contribute to CAR tonic signalling; (i) recapitulating or enhancing T-cell / DC interactions to lower the activation threshold for cytotoxicity; (j) preventing constitutive IL-2 production and Treg induction by mutating the CD28 binding site for Lck; (k) optimal selection of target ligand, autologous APCs, T-APCs or EATCs expressing target ligand may also facilitate CAR T-cell expansion and persistence in vivo; (l) utilising small molecule gated CARs e.g. by incorporating an FKBP/FRB* heterodimerizing module in the presence of a rapamycin analogue; (m) utilising blocking monoclonal antibodies to target inhibitory immune checkpoints; (n) utilising switch CARs (e.g. PD-1/CD28); (o) optimal selection of the expression vector and promoter, e.g. using non-LTR (SIN) lentiviruses, mRNA or transposon delivery; (p) co-expressing tethered cytokine fusion molecules (such as IL-15/IL-15Rα); (q) exploiting inside-out signalling to integrins to facilitate T-cell migration & bystander tumour cell targeting; (r) utilising Tet-off systems for temporal control of CAR expression; and (s) utilising CRISPR Cas9 to direct CAR expression specifically to the T-cell receptor α constant (TRAC) locus. Figure 1 and figures 3-5 are original and have been created specifically for this article.