Recent advances in CAR-T cell engineering

After the successful binding target of CD19 was obtained in B cell ALL and DLBCL, studies were initiated to identify more effective binding targets. Successful targets include a B cell mature antigen (BCMA) for multiple myeloma (MM). The objective response rate of patients with MM to this target was 85% in an early trial, in which 15 of 33 patients (45%) achieved a complete response [12]. CAR-T cells engineered in this manner show exciting response rates for specific hematological malignancies, and a few phenomenal targets have been discovered, including CD20 and CD22 [13, 14] (mainly in B cell lineages). However, for many tumors, no suitable known target is currently available for CAR recognition and binding, resulting in modest outcomes in clinical trials. An ideal target should be highly specific for the tumor and have wide tumor coverage to ensure both safety and effective tumor clearance, and stability also plays a key role in the duration of the response [15]. However, few antigens meet all three requirements, and methods have been proposed to cover the shortage of targets.

One of the major challenges in identifying targets to ensure safety is avoiding “on-target off-tumor” effects: the nonspecific expression of a target antigen on healthy cells stimulates CAR-T cells and causes damage to healthy tissues, which may pose a threat to the patient’s life. For example, in malignancies of the myeloid lineage, antigens such as CD123 or CD33 are challenging targets to use because they are also expressed in vital bone marrow stem cells; therefore, although treatments targeting these antigens eliminate the tumor cells, the stem cells in the bone marrow are also killed, leading to myelosuppression. This intolerable side effect makes strategies targeting CD33 or CD123 a double-edged sword; therefore, two promising approaches have been proposed to take full advantage of this type of tumor-associated antigen (TAA). The first approach is based on an RNA carrier that expresses a CAR that avoids long-term myelosuppression. Preclinical experiments showed promising results, but a clinical trial showed no observable myelosuppression or antitumor effects on a cohort of 7 patients [16‐18]. The other approach is based on protecting the bone marrow stem cells using hematopoietic stem cell transplantation (HSCT) to eliminate the anti-CD123/CD33 CAR-T cells. Using this approach, leukemia cells are eliminated, but the CD33-silent hematopoietic stem cells are protected from apoptosis induced by the CAR-T cells. The feasibility of this approach has also been reported in preclinical studies [19], but its effectiveness remains to be verified in clinical trials. We undoubtedly hope to identify a more specific TAA to serve as the binding domain for a CAR that is able to kill tumor cells while minimizing damage to normal cells, but in practice, the discovery of this TAA remains the main challenge to the extensive use of CAR-T cells as a treatment for the majority of tumors and must be resolved in the coming years.

The interaction of a CAR and its binding target determines the tumor-killing effect and the proliferation of the constructed CAR-T cells, which depend on the density of the antigen, the binding affinity, and the binding strength of the antigen. In other words, the stability of the target antigen determines the persistence of the response to some extent. Malignant cells may downregulate the target antigen to avoid CAR-T cell-induced death, which is one of the major drivers of antigen-positive relapse [20]. One promising strategy to solve this complex problem involves lengthening CAR-T cell persistence by re-engineering the intracellular region, which is discussed later in the section describing the endodomain. Another mechanism by which tumors evade treatment is through their heterogeneity, which is particularly effective when the coverage of the target antigen is limited and is the main reason for antigen-negative relapse after CAR-T cell therapy because unpredictable mutations can cause a loss of the target antigen. Therefore, as explained above, the chosen antigen must be universally expressed in the tumor. A small number of residual tumor cells can cause the disease to relapse with full resistance to the CAR-T cells targeting the specific epitope. Fortunately, this type of relapse is overcome using dual-target or combined-target CAR-T cells [20]. The primary outcome of dual-target CAR-T cell therapy in B cell malignancies is summarized in Table 1, which indicates the responses to suitable dose levels compared to the responses observed after treatment with single-target CAR-T cells. Notably, in the clinical trials cited, antigen-negative relapse was rare, suggesting that dual CAR-T cell therapy is a promising method to prevent the relapse caused by antigen escape. As shown in the table, several methods are currently available to produce a dual-target CAR-T cell involving a bispecific CAR and two different CARs carried by one or two carriers, but more clinical data are needed to select the best method to achieve dual-target CAR-T cell therapy. In addition to B cell malignancies, trials of dual anti-CD19 plus BCMA CAR-T cells for MM have also delivered promising results, with 20 (95%) of 21 patients achieving an overall response, including nine (43%) with a stringent complete remission (CR), three (14%) with a CR, five (24%) with very good partial remission (VGPR), and three (14%) with PR [31]. Another clinical trial of CD38 plus BCMA CAR-T cells for patients with R/R MM resulted in 14 (87.5%) patients achieving an overall response, 8 (50%) achieving sCR, 2 (12.5%) achieving VGPR, 4 (25.00%) achieving PR, and 14 (87.5%) achieving a bone marrow MRD-negative status. The longest duration of a sCR exceeded 51 weeks, and 5 of 8 patients (62.5%) maintained sCR, 2 progressed to VGPR, and 1 to a PR. The PFS rate at 9 months was 75% [32]. Furthermore, the combined-target CD22 CAR-T cells have been proven capable of treating patients who are resistant to CD19 CAR-T cells [14]. Nine of ten patients who had previously received CD19-directed immunotherapy achieved CR, including all five patients who had enrolled with CD19dim- or CD19-negative B-ALL and one patient who was refractory to both CD19 CAR T cell and blinatumomab therapies. Although no side-by-side comparison of single-target CAR-T cells and dual-target CAR-T cells are available to compare their levels of efficacy, multitarget CAR-T cell therapy might improve clinical outcomes by decreasing the relapse rate. However, for dual CAR-T cell therapy, managing the target epitopes to include molecules other than those in B cell lineage diseases is the next challenge to minimize the possible damage caused by the “on-target off-tumor” effect and prolong persistence to avoid antigen-positive relapse. In contrast to patients carrying only the universal target antigen, a fraction of patients carrying a mutation that causes the expression of a specific antigen that is not normally expressed in the tumor-origin tissue, such as CD7 or Lewis Y on AML cells, is able to receive CAR-T cell therapy that is relatively highly specific and that does not harm the normal myeloid cells; for example, Lewis Y is an antigen expressed on T cells but not on normal myeloid cells. The first clinical trial of anti-Lewis Y CAR-T cells for AML, in which 5 patients were treated, revealed that only one patient achieved a transient CR, while two achieved PR and the other two maintained a stable disease throughout the trial. However, all the subjects died within 1 year because of disease progression [33]. Based on these results, a satisfactory therapeutic effect may difficult to achieve when treatments rely solely on these rare target epitopes, and combining these epitopes with other widely covered targets may be a more promising strategy.

To date, the identification of effective targets remains a slow procedure, but with the development of sequencing techniques, we remain optimistic that an effective target for each type of tumor will be discovered in the future.

The development of single-chain variable fragment (scFv) CARs was also revolutionary. Currently, most available CAR-T cell therapies have adapted murine CARs, which are recognized by the immune system as foreign antigens; this source of antigen is also the main reason for the short persistence of CAR-T cells. Humanized CAR-T therapy has been proposed to minimize the influence of murine CARs and prolong the persistence of CAR-T cells. In a clinical trial of patients with B-ALL, after 14 days of CAR-T cell infusion, 19/23 (82.6%) patients achieved CR/CRi, 12/23 (52.2%) patients achieved CR, and 18 (78.3%) patients were MRD-negative. The other 4 patients were evaluated as NR. One of the NR patients achieved CR after 2 months of infusion with anti-CD22 CAR-T cells [34]. Another similar clinical trial was able to evaluate the response of all 10 patients with R/R ALL recruited for the study, and all achieved CR; six remained in a CR state for more than 18 months without further treatment. Long-term persistence of humanized CAR-T cells was observed in most of the patients [35]. Although the primary outcome of the humanized CAR-T cells showed a similar response rate to murine CAR-T cells, the humanized CAR-T cells appeared to provide greater long-term benefits than the murine CAR-T cells by persisting longer. Furthermore, in another clinical trial, 4 of 5 patients with B-ALL who relapsed after treatment with murine CAR-T cells or who had no initial response to murine CAR-T cells successfully achieved CR after being infused with humanized CAR-T cells [36]. The human-derived CAR structure may be an efficient and promising method to prolong the persistence of CAR-T cells, and the longer persistence indeed improves the duration of the response.

As explained above, multiple-target CAR-T cell therapy may improve the antitumor efficiency and prevent the relapse caused by antigen loss. The CAR structure is separated and the sequence that encodes only the inner and connecting parts is implanted into T cells to increase the coverage provided by the manufactured T cells. Then, the binding sequence is injected into the patients, and a specific CAR target that depends on the molecule injected into the patients and achieves universal coverage is produced. The potential for universal CARs, which can shift the binding epitope through biotin/avidin and leucine zippers, has also become a topic of significant interest. Lohmueller et al. created AT-CARs using an affinity-enhanced monomeric streptavidin 2 (mSA2) biotin-binding domain that, when expressed on T cells, targets cancer cells coated with biotinylated antibodies, while Cho et al. presented a split, universal, and programmable (SUPRA) CAR system that simultaneously encompassed multiple critical “upgrades,” such as the ability to switch targets without re-engineering the T cells, fine-tune the T cell activation strength, and sense and logically respond to multiple antigens [37, 38]. However, antigen shifting requires the persistent activation of CAR-T cells. In addition, a bispecific T cell engager with activated T cells has proven effective in some patients with advanced tumors, and these treatments are “off the shelf” and inexpensive. Further improvements are needed in the inner part of the CAR to make in vitro manufacturing worthwhile.

The ectodomain of a CAR normally has a similar structure to a monoclonal antibody (mAb), namely, the Fc region in Ig. The Fc domain mediates the antigen-antibody reaction, leading to the elimination of antibodies in a normal immune reaction. However, the Fc region exerts a negative effect on CAR-T cell persistence and function. This outcome may result from Fc-FcγR interactions between CAR-T cells and other immune cells that lead to tonic signals that accelerate T cell aging. This interaction is prevented by blocking the aging caused by the hinge and increasing the flexibility of the CAR. For example, an initially designed IgG1 Fc region has a strong affinity for FcγR, which induces tonic apoptosis-promoting signals in the CAR-T cells, for which the CH2 region is required [39]. Furthermore, knock out of the CH2 region decreased the tonic signal and prevented activation-induced T cell death. This outcome is achieved by including only the CH3 segment of an Fc region or by replacing the whole Fc region with IgG2, which has a low affinity for FcγR. This strategy avoids Fc-FcγR interaction-induced exhaustion, which prolongs the persistence of the CAR-T cells. A CD19 CAR-T cell therapy in which the Fc region was deleted also showed a high binding affinity for CD19+ malignant cells in vivo [40]. Recently launched clinical trials will clarify whether this strategy enhances the persistence and antitumor activity.

The hinge domain can also be based on linkers of membrane receptors, providing a flexible and long connection between binding sites and contributing to the binding affinity. Ectodomain linkers of CD8 alpha and CD28 were compared in CD19 anti-CAR-T cells [41], and the CD8 alpha linker generated better results, as it led to lower levels of cytokine release and less activation-induced cell death. Moreover, in a recent study, the use of NGFR as a hinge resulted in a very low affinity for FcγR and suggested that NGFR potentially represents a great traffic marker for CAR-T cell detection [42]. The flexibility of a CAR is related to its length, which contributes to the affinity of a CAR and a target antigen. In a preclinical experiment, an IgG4 CH3 hinge domain was inserted to connect the CD28 linker, and the CAR-T cells showed increased growth, migration, and CD4 subtype expansion. Researchers have not clearly determined whether the antitumor efficacy of anti-CD19 CAR-T cells is increased after the implementation of this approach. However, the addition of linkers may result in a significant increase in the activity of anti-mesothelin CAR-T cells since some target epitopes, such as CD22, require additional flexibility to achieve optimal affinity. Thus, this approach represents a promising method to improve the effects of CAR-T cells. The hinge domain was previously neglected, but the aforementioned preclinical experiment has proven that a comparable hinge domain might play a vital role in modulating the binding affinity and signal transduction, particularly for targeting dim antigens or low-affinity malignant cells.

Innovative approaches to modifying the intracellular domain have been the primary driver of revolutionary changes in CAR-T cells across generations to date. First-generation CAR-T cells with only the ITAM segment of CD3 became obsolete because they lacked persistence and a proliferation capacity. Second-generation anti-CD19 CAR-T cells showed strong antitumor efficacy in patients with B cell ALL and NHL and were formulated to contain one endodomain of a stimulation epitope, such as CD28 or 4-1BB [4‐6]. Studies exploring the potential of costimulation domains are ongoing, and new costimulation domains, such as ICOS and CD27, have shown efficacy in eliminating tumor cells in preclinical experiments [48, 49]. However, these domains have consistently been shown to decrease inner tonic signaling, which causes T cell exhaustion, with 4-1BB exerting a greater effect than CD28 in preclinical and clinical trials. The CD28 costimulation domain responds quicker than the 4-1BB costimulation domain, but the population of CAR-T cells with the 4-1BB costimulation domain peaks at a higher level after a relatively slow expansion [50]. Additionally, a preclinical trial revealed that the 4-1BB costimulation domain promotes the formation of the memory Tcm subtype, while the CD28 costimulation domain promotes the formation of the Tem subtype [51], and thus, CAR-T cells with the 4-1BB costimulation domain persist longer than the CAR-T cells with the CD28 costimulation domain. In addition, compared with the CD28 CAR-T costimulaion domain, the 4-1BB CAR-T costimulation domain induces weaker tonic signaling, which means that exhaustion transpires more slowly for the CAR-T cells with the 4-IBB costimulation domain than for CAR-T cells with the CD28 costimulation domain [52‐54]. However, the persistence and relapse problems associated with CAR-T therapies, in general, are unable to be solved by a single costimulation domain. Therefore, the third generation of CAR-T cells was designed with two costimulation domains. Preclinical testing showed increased persistence, proliferation, and antitumor activities. In clinical trials of the third-generation anti-CD19 CAR-T cells with 4-1BB and CD28 endodomains, the primary response rate did not show enhanced efficacy compared to the second-generation CAR-T cells in treating subjects with R/R B cell malignancies. Furthermore, for patients with MRD+ or a small burden of B cell malignancies at the time of the infusion, third-generation CAR-T cells exhibited better proliferation [55‐58]. Additional preclinical experiments discovered more combinations, such as ICOS plus 4-1BB, TLR2 (Toll/interleukin-1 receptor domain of Toll-like receptor 2) plus CD28, and 4-1BB plus OX40, all of which exhibited better performance than the combination of CD28 and 4-1BB costimulation domains in vivo and in mathematical models [43, 59]. Despite these promising results, the rate of cytokine release syndrome (CRS) increased, which may be attributed to the repeated signal delivered by both costimulation domains. In summary, third-generation CAR-T cells may benefit from combining complementary costimulation domains. On the other hand, the simple addition of a costimulation domain may produce severe side effects and accelerate the aging of the CAR-T cells. Additional clinical trials are needed to assess and monitor the risk of side effects due to overactivation to realize the optimal potential of third-generation CAR-T cells.

The costimulation domains contribute substantially to the expansion and persistence of CAR-T cells for therapy. Currently, CAR-T cells with the CD28 and 4-1BB costimulation domains are the most commonly used CAR-T cell products. However, 4-1BB has several advantages compared with CD28. New costimulation domains have been discovered and tested, but the tonic signaling problem remains unresolved. The use of more than one costimulation domain may increase the tumor-killing activity and persistence of CAR-T cells, but a poor match might lead to increased tonic signaling, which would result in more severe side effects and faster T cell exhaustion.

With the expansion of clinical trials, checkpoint inhibitor therapy has shown the ability to reverse the exhaustion of CAR-T cells in the short-term in relapsed patients and to shape the tumor environment through the inducible release of transgenic immune modifiers. CAR-T cell therapies based on these principles are known as fourth-generation CAR-T cell therapies [60]. In addition, with the exception of immune modifiers, T cell engagers and some membrane receptors are also transduced with second-generation CAR-T cells, which are generally called next-generation CAR-T cells. These CAR-T cells are not limited by T cell cytotoxicity and represent a powerful tool to re-establish the immune system after infusion. As the most highly anticipated advancement to date, fourth-generation and next-generation CAR-T cells CAR-T cells are divided into two categories, cells that utilize secreted molecules and cells that utilize membrane receptors (Fig. 2). Both types are described below.