The Bumpy Road to CAR Activation

Main Text

A rapidly emerging immunotherapy approach is the use of chimeric antigen receptor (CAR) T cells. While adoptive transfer of CAR T cells has seen striking results in some malignancies, its efficiency has been tempered in others due to limited expansion, persistence, and tumor homing of the transferred cells.¹ Another clinical limitation is the serious toxicities that may arise following CAR T cell tumor recognition (reviewed in Sadelain et al.²). This raises a very important issue that must be resolved for proper cell engineering: are we fully harnessing the biology of artificial T cell signaling for effective immunotherapy? A recent paper in Science Signaling by Salter et al.³ demonstrates that the signaling cascades initiated by synthetic CARs cannot be predicted entirely by their design and sheds light on the differences between the phosphorylation of the proteomes of CD28/CD3ζ compared to 4-1BB/CD3ζ CAR T cells.

The last decade has seen the emergence of antitumor therapeutic development. The Nobel Prize in Physiology or Medicine 2018, awarded to James P. Allison and Tasuku Honjo, has added to the enthusiasm toward immunotherapy. Their work has created a major shift in our understanding of the immune system recognition of malignancies and how we can manipulate this new weapon to our advantage. We are indeed entering a new era where immunotherapy will soon complement standard radiotherapy or chemotherapy regiments for cancer treatment.

CAR T cells are engineered T cells expressing fusion proteins, mostly combining an antigen-specific single-chain fragment (scFv) coming from a monoclonal antibody with T cell receptor (TCR) intracellular signaling domains. In vitro studies have demonstrated that first-generation CAR T cells, containing only a CD3ζ moiety, support T cell activation and target cytotoxicity, but with very limited persistence and antitumor efficacy following adoptive transfer.⁴ Second-generation CARs, therefore, incorporated a two-signal model of T cell activation by modifying the CARs to include a CD28 or 4-1BB (CD137) costimulatory domain that provides signals for T cell effector function, proliferation, and, more importantly, persistence.5, 6 Nevertheless, in recent years, these CAR constructs have shown variable effects in vivo. Therefore, the authors of the new study aimed to assess whether differences between CD28/CD3ζ and 4-1BB/CD3ζ CAR T cells were attributable to divergent T cell activation pathways. Human primary T cells were transduced with modified lentiviral vectors encoding CD19- or ROR1-specific CD28/CD3ζ CARs and identical CD19- or ROR1-specific 4-1BB/CD3ζ CARs fused to a nine-amino acid Strep-tag II (STII) sequence in the extracellular CAR hinge. Canonical T cell signaling events were then evaluated following STII microbead stimulation. Using an elegant phosphoproteomic approach with liquid chromatography-tandem mass spectrometry (LC-MS/MS), they identifed many novel phosphoprotein signaling events in stimulated CARs that were not identified using previous methods. Of note, both CD28/CD3ζ and 4-1BB/CD3ζ CARs could promote phosphorylation of endogenous CD28 following stimulation. Surprisingly, they concluded that patterns of protein phosphorylation were very similar in cells expressing either of these CAR constructs. What stood out was rather a more intense ZAP-70 and CAR CD3ζ phosphorylation in stimulated CD28/CD3ζ cells, which was associated with a general increased phosphorylation at a greater number of sites than in 4-1BB/CD3ζ CAR T cells. Differences in CAR signal intensity were partially attributed to greater Lck association with CD28/CD3ζ, which could be altered by mutating tyrosine and proline residues in the CD28 signaling domain (Figure 1).

Figure 1.

Differences in CAR Signaling Intensity and Kinetics Lead to Divergences in Tumor Control

Stimulation of CD28/CD3ζ and 4-1BB/CD3ζ leads to similar phosphorylation events that comprise both canonical CD28 and 4-1BB signaling elements. However, what strikingly distinguishes both CARs is the more expedited and robust activation of CD28/CD3ζ compared to 4-1BB/CD3ζ CAR T cells. This will induce a more differentiated state in these cells, promoting exhaustion and potentially leading to inadequate tumor control. On the other hand, 4-1BB CAR T cells will exhibit a memory phenotype and show efficient tumor eradication, even at low dose.

Signaling strength during physiological T cell activation relies on three signals: TCR engagement, costimulation, and cytokines.⁷ This will then initiate transcriptional programs that dictate effector cell differentiation, memory formation, and proliferation (reviewed in Kaech and Cui⁸). However, the data by Salter et al.³ indicate that CAR signaling only partially mimics endogenous TCR activation since phosphorylation of CD3δ, ε, or γ ITAMs were not detected in the CAR T cells. This should be taken into account when trying to generate artificial receptors. Combining various stimulatory domains may not properly recapitulate physiological signaling and thus may be unpredictable in their behavior.

There have been reports of constitutive phosphorylation of the CAR CD3ζ domain leading to tonic signaling in first-generation CAR constructs.⁹ This was linked to clustering of CAR molecules at the cell surface and thought to promote T cell exhaustion.¹⁰ Thus, the functional state of a given CAR following ex vivo expansion could differ according to its design and should be assessed accordingly. Salter et al.³ further showed that Raji tumor control in NOD-scid IL2rγ^null (NSG) mice treated with a low dose of CD28/CD3ζ CAR T cells was severely impaired despite the accumulation of higher levels of these cells in murine bone marrow compared to animals treated with a low dose of 4-1BB/CD3ζ CAR T cells.

The impaired tumor control was associated with an induction of stronger and more immediate effector functions, but also greater expression of the inhibitory receptor PD-1. It has been shown that in vitro T cell priming can drive cells toward an exhausted profile, resulting in limited T cell expansion and concomitant apoptosis.¹¹ Proliferation and tumor cytotoxicity of infused CAR T cells could, in part, depend on their exhaustion status prior to patient transfer, as has been seen in other T cell based therapies.¹² In addition, phenotypic and functional T cell exhaustion can be differentially regulated depending on culture conditions and/or duration, thus proposing that a more comprehensive monitoring of in vitro expanded CAR T cells could enhance in vivo proficiency of an immunotherapeutic product.¹³ Hence, one can hypothesize that encoding a CD28 signaling domain on a CAR may lead to excessive stimulation and thus promote exhaustion, thereby compromising in vivo efficacy.

In the light of this, the authors suggest that creating safer and more potent CAR T cells might be carried out by modifying CAR construct design, but real success is impeded by our lack of control of synthetic CAR signaling. Future research should focus on determining whether these phosphorylation differences persist in second-generation CAR constructs after multiple rounds of stimulation and whether this could be overcome by adding exogenous cytokines during CAR T cell activation and proliferation. To summarize, through this quite elegant observation of CAR T cell activation strength and kinetic variations, the findings by Salter et al.³ may have a high impact on future CAR design and how we have to evaluate in greater depth their intrinsic mode of activation in order to develop more precisely controlled therapeutics.

Conflicts of Interest

The author declares no competing interests.