Reframing cell therapy for cancer

Abstract

Technologies for engineering immune cells to recognize features on cancer cells are transforming oncology. Synthetic biologists now expand this approach by conferring therapeutic functions to nonimmune cells, and by programming cells to sense and respond to a new class of physiological cues.

Engineering human cells to serve as living therapeutics represents a promising frontier in medicine. To date, most effort and success in this field has focused on reprogramming immune cells to recognize specific features on the surface of cancer cells, such that when these modified immune cells are introduced into a cancer patient, the therapeutic cells kill the cancer cells while sparing healthy tissues. In a typical implementation of this strategy, a sample of the patient’s T cells–the cells that carry out targeted killing–is collected and then manipulated by introducing a gene encoding a chimeric antigen receptor (CAR). The CAR is an engineered protein that enables the immune cell to recognize a specific, pre-determined marker (antigen) known to be present on the surface of that patient’s cancer cells (Fig. 1a). These so-called CAR-T cell therapies are rapidly entering medical practice; hundreds of CAR-T cell clinical trials are currently underway, two such products received FDA approval in 2017, and tens of billions of dollars have been invested in this space.^1–3

Figure 1 |. Comparison across frameworks for engineering cells to respond to cancer.

(a) Traditional CARs are expressed in T cells and recognize antigens on the surface of cancer cells, triggering native T cell functions. (b) Kojima et al.⁶ take an alternative approach, using a relatively portable sensor-processor circuit to enable nonimmune cells to sense and kill cancer cells, coopting native JAK-STAT signaling. (c) Chang et al.⁷ extend the CAR approach to sense soluble ligands, including cytokines that accumulate at tumor sites. CAR, chimeric antigen receptor; JAK, Janus kinase; STAT, signal transducer and activator of transcription.

Figure 1 was created in part by artistic staff at Nature Chemical Biology to accompany this commentary. Therefore, the figure is not reproduced here. The original figure is available alongside the original article on the publisher’s website at DOI: https://doi.org/10.1038/nchembio.2573

Fully realizing the potential of engineered cell therapies, however, will require overcoming a number of challenges. First, manufacturing CAR-T cell therapies remains challenging and expensive, with the cost of treatment estimated at $475,000 per patient.⁴ While CAR-T cell therapies have been applied most successfully to treat liquid cancers such as leukemias, treatment of solid cancers has proven more challenging, in part due to the lack of unique surface antigens and the presence of tumor-associated immune suppression.⁵ Finally, effective immunotherapy requires striking a balance between safety and efficacy, and we are still learning how to best achieve these goals for each patient.

Two new synthetic biology technologies may help to address these needs. Recently, Kojima et al. reported that non-immune cells may be used in place of T cells in order to build engineered cells capable of targeted cancer cell killing via injection of an enzyme which converts a pro-drug into a toxic active product.⁶ This alternative to the CAR strategy could ultimately benefit the cost and challenge of therapeutic manufacturing, while providing a novel modality for tumor killing. Secondly, while all existing CARs recognize antigens displayed on the surface of target cells, in this issue, Chang et al. report that CARs may be engineered to confer recognition of soluble species.⁷ This finding is both surprising--providing new insights into how CARs function--and useful, since as the authors demonstrate, CAR-T cells may now be engineered to recognize and become activated by the soluble proteins that accumulate at immunosuppressive tumor sites.

To program non-immune cells to recognize and kill tumors, Kojima et al. developed a system for engineering “killer” cells to express a defined genetic payload upon binding to a target cell (Fig. 1b). Since CARs would not function in most cells, which lack signaling components found in T cells, the authors built both a synthetic receptor and compatible intracellular circuitry. The sensing system comprised: (1) signaling chains derived from a chimeric cytokine receptor, which signals through the native JAK-STAT pathway but includes external domains derived from an antibody that binds the tumor antigen HER2, and (2) an inhibitory chain comprising the CD45 intracellular domain, which blocks JAK-STAT signaling in a proximity-dependent fashion. In this setup, binding of the killer cell to the target cell displaces the inhibitory chain from the signaling chains, which then induce signaling via STAT6 (Fig. 1b). Expression of these components in nonimmune Human Embryonic Kidney (HEK) cells yielded target cell contact-induced expression of reporter genes from an engineered STAT6-responsive promoter. The implementation of this overall concept was refined over several rounds of tuning to improve performance (fold-induction of transgene expression upon target binding). To endow the engineered cells with an effector function, the reporter gene was replaced with a cargo gene comprising a cell penetrating peptide fused to the enzyme FCU1, which converts the prodrug 5-FC into the toxic 5-UMP. Engineered in this way, HEK “killer” cells recognized HER2-expressing HEK “target” cells to render 5-FC toxic. Importantly, this strategy was portable; when implemented in human mesenchymal stem cells, which are used extensively in clinical applications, this kill circuitry conferred killing of breast cancer cells via recognition of HER2. Ultimately, this sense-kill module could be implemented in any cell with compatible JAK-STAT signaling, and the cargo gene could be exchanged to evaluate various candidate therapeutic modalities (Fig. 1b).

Despite the popularity of CARs, how their signaling is initiated is not entirely clear; Chang et al. generated new insights by employing the “build to understand” principle to investigate whether CARs can be built to recognize soluble ligands. The authors first generated model CARs and found that various soluble ligands can trigger CARs, so long as the ligand induces receptor dimerization. Moving towards a translationally relevant target, this approach was extended to generate CARs that respond to the immunosuppressive, homodimeric cytokine TGF-ß. When these CARs were expressed in T cells, the effect of TGF-ß on the engineered cells was converted from inhibitory to stimulatory, effectively rewiring the response of the T cells to a cytokine found at many solid tumor sites (Fig. 1c). Notably, the authors observed that soluble ligands trigger CARs via a mechanism that involves mechanical coupling between extracellular and intracellular receptor domains, which may also be true of CARs that recognize surface antigens. These findings expand the repertoire of strategies for building cells that sense and respond to their environment in customizable and useful ways.

Each of these studies highlights the fact that CAR-T cells represent just the leading edge of a wave of engineered cell products that will confer new therapeutic modalities and address diverse medical needs. Achieving this vision will require new technologies, greater understanding of clinical realities such as patient-to-patient variation, and even new conceptual models for linking technology development to clinical performance.^8,9 The existence of multiple knobs for tuning the performance of a synthetic biology therapy is certainly a blessing, but it also poses challenges for achieving safety and efficacy, and for designing clinical trials and protocols that best benefit patients. As engineered cell products traffic from lab bench to bedside, addressing these challenges is a defining opportunity for this moment in medicine and bioengineering.