Targets

One of the attractive features of CARs is that their antibody specificity allows them to recognize structures on intact proteins, carbohydrates, and lipids. As a result, the target antigens do not need to be processed and presented by MHC molecules, and any molecule on the cell surface can be easily targeted by CARs. The list of CAR targets evaluated includes but is not limited to CD19 , CD20 , and others for hematologic malignancies; Her-2 , MUC-1 , CEA , PSMA , and others for breast cancer, colon cancer, prostate cancer, and other solid tumors; and VEGF-R2 and KDR for tumor neovasculature. Essentially, any mAb can be engineered to make a CAR. For those targets where it is difficult to raise conventional antibodies in mice or other species, phage display and other technologies have been used to generate the antigen binding portion of the CAR . Therefore, CAR-based therapeutic options are limited only by the availability of mAbs.

Design

The general CAR design described previously consists of an scFv fragment fused to the hinge region of CD8, which is fused to the intracellular signaling regions from Fcγ or CD3ζ molecules ( Figure 18.1 ). Any CAR with this basic design is considered to be a “first-generation” CAR. Although first-generation CARs effectively transfer antitumor reactivity to mouse and human T cells, they have limited proliferative potential and persistence in vivo . A potential explanation for this is that CARs bypass all of the normal costimulatory signals that T cells use for survival and memory formation. One solution to these problems was the inclusion of a costimulatory cassettes (e.g., CD28, 4-1BB, OX40, or ICOS) in the CAR construct ( Figure 18.1 ). By having two signaling cassettes, one T cell activating and one costimulatory, T cells engineered with second-generation CAR constructs have enhanced proliferative capacity and persistence in vivo compared to T cells engineered with first-generation CA’s . Recently, a third-generation CAR has been developed that includes a third signaling cassette. In addition to the CD3ζ signaling cassette, two costimulation cassettes (CD28, 4-1BB, OX40, and/or ICOS) are included . T cells engineered with third-generation CARs have superior function, proliferative capacity, and survival in vivo compared to T cells engineered with second-generation CARs .

Another consideration in CAR design is the relative affinity of the scFV. In general, Fab fragments have higher affinity for antigen than TCRs have for peptide MHC complexes . However, the use of phase display libraries has led to high-affinity mAbs for CAR generation . These high-affinity CARs can bind to very low levels of antigen, making the CAR gene-modified T cells extremely potent antitumor effectors.

In Vivo Studies

T cells engineered with CARs have been extensively evaluated in vivo . In mice, it is clear that CAR gene-modified T cells can treat established tumors. Mouse T cells transduced to express first-generation CARs can reject established mouse tumors and human xenografts in vivo . In addition to using CAR-engineered T cells, several studies report using CARs to engineer hematopoietic stem cells . Mice that were engrafted with CAR-engineered hematopoietic stem cells were resistant to subsequent tumor challenge . One unexpected observation in these experiments was that a CAR can be expressed and function in non-T cell subsets such as cells of the myeloid lineage and NK cells . These studies indicate that CARs are effective therapeutic agents and have the added advantage of functioning in multiple effector lineages.

Although promising results were obtained with first-generation CAR gene-modified T cells, there were clear limitation in persistence noted in vivo . In cancer patients, CD3ζ chain downregulation and modulation of CD8 coreceptor chains have been implicated in T cell dysfunction. The advantage of using CARs to redirect the specificity of T cells is that the resulting CAR-expressing T cells are not dependent on many of the signaling molecules (CD3, CD4, or CD8) that are required for TCR signaling. However, the disadvantage is that CAR-expressing T cells may not get the necessary costimulatory signals needed for T cell survival in vivo . The second-generation CAR, which includes a CD28 signaling cassette, solves this problem ( Figure 18.1 ). TNP CAR transgenic mice expressing a CD28 cassette or a truncated CD28 cassette lacking the CD28 intracellular signaling domain demonstrated that costimulation was essential for full activation of CAR-expressing T and NK cells . Moreover, CD28 costimulation led to improved effector function and resistance to apoptosis of CAR-expressing T cells . Direct comparisons of first- and second-generation CAR-engineered T cells found that second-generation CAR-expressing T cells persisted longer in vivo and had better antitumor activity than first-generation CAR-expressing T cells . Recent studies with third-generation CARs indicate that multiple costimulation signaling domains are an improvement over one (second generation) or none (first generation) . The combination of CD137 (4-1BB) and CD28 in these third-generation CARs led to enhanced persistence and rejection of large tumors . Collectively, these studies indicate that the addition of costimulation leads to improved function and therapeutic efficacy of CAR-expressing T cells in animal models.

Several different CARs have been used to genetically modify T cells for use in clinical trials in patients with cancer and infectious diseases. The first clinical trials used a first-generation CAR targeting carbonic anhydrase IX (CAIX) for patients with renal cell carcinoma and folate binding protein (FBP) for patients with ovarian cancer . These studies demonstrated that it is feasible to prepare CAR-engineered autologous T cells for patient treatment. In both studies, no objective clinical responses were observed, and the gene-modified T cells failed to persist long term. Although no serious adverse events were observed using the FBP CAR gene-modified T cells, there was grade 2–4 liver toxicity in the patients treated with CAIX CAR gene-modified T cells. This toxicity was due to the expression of the CAIX antigen on the epithelial cells of the bile duct . In a later trial, a patient was treated with a third-generation CAR (ssFv/CD28/4-1BB/CD3ζ) targeting ERB B2 (Her-2) . The patient rapidly developed respiratory distress and died 5 days later. It was found that there was low-level ERB B2 expression on the epithelial cell of the lung, and the toxicity was presumably due to destruction of the normal lung epithelium . These results indicated that selection of the target and probably the affinity of the CAR are critical considerations for pursuing clinical trials using CAR gene-modified T cells.

Whereas there have been adverse events associated with CAR gene-modified T cells, there have been many encouraging clinical trials reported. Clinical trials have been reported using T cells expressing CARs targeting CD19, CD20, GD2 (Brenner), and others . The best characterized of all of these CARs targets CD19 on B cell lymphomas. Several groups have conducted clinical trials with gene-modified T cells and have obtained very promising results . In all cases, the only normal cells found to express CD19 are normal B cells, and the CD19 CAR can impact on number of circulating B cells in patients after treatment, leading to B cell aplasia . Other side effects reported in a few patients included tumor lysis syndrome and cytokine storm . The conclusions from these clinical trials indicate that T cells expressing anti-CD19 CAR were well tolerated and resulted in few adverse events.

Targets

Since we first reported engineering normal human PBL-derived T cells with TCR genes, the number of TCR genes evaluated has grown to include TCRs that target other melanoma differentiation antigens (gp100 and tyrosinase ), cancer testis antigens (NY-ESO-I and MAGE ), viral antigens (CMV , EBV , HIV , HCV , and others), and other tumor antigens (CEA , p53 , HA-1/HA-2 , and others). As with CARs, the potential targets of TCR gene-modified T cells are limited only by the availability of antigen-reactive T cell clones and TCR genes. Unlike isolating antibodies for generating CARs, raising T cell clones against many target antigens on tumors has been difficult. This problem can be circumvented by using T cell clones isolated in HLA-A2 transgenic mice that recognize human tumors or by using phage display to generate single-chain TCRs . Like CARs, the spectrum of TCR gene-modified T cells is limited only by the availability of TCR genes.

Factors Influencing TCR Assembly and Function

The proper function of a TCR gene-modified T cell requires adequate expression of the introduced TCR α and β chains, correct pairing of the TCR chains, and proper pairing with the elements of the CD3 complex. Initially, vectors containing multiple promoters ( Figure 18.2A ) , IRES elements ( Figure 18.2B ) , or multiple vectors ( Figure 18.2C ) were used to ensure adequate expression of both TCR chains. Recently, the use of 2A self-cleavage peptides has ensured that both chains are equally expressed from a single promoter ( Figure 18.2D ) . As shown in Figure 18.3 , 2A linked TCR chains generally give higher and more uniform TCR expression compared to having the chains expressed separately using individual promoters. It has also been shown that high TCR transgene expression can be achieved by codon optimizing the TCR α and β chain genes . The use of 2A sequences and codon optimization also has the advantage of increasing TCR expression without altering the TCR sequence.

Structure of viral vectors containing T cell receptor genes.

The most common delivery vehicles for transferring TCR genes are retroviral or lentiviral vectors. The basic vector configurations for expressing TCR genes shown depicted are based on our studies. The specifics for each group may be slightly different, but the general vector structure and elements such as the LTRs (black), additional promoters (white), IRES element (blue), 2A self-cleavage peptide (orange), TCR α chain (red), and TCR β chain (blue) are similar. Note that the arrows indicate promoter activity. Some vectors have additional elements for drug selection genes (neo ^r ) or cell tracking/marker genes (EGFP). (A) The viral vector depicted uses the promoter in the 5′ LTR to drive expression of the upstream chain (TCR α) and a second internal promoter to drive expression of the downstream chain (TCR β). (B) Expression of the downstream chain (TCR β) in this vector is driven by an IRES element. The promoter in the 5′ LTR drives the expression of the upstream chain (TCR α). (C) Some investigators create two vectors to deliver TCR genes. The structure of each vector is the same with promoter in the 5′ LTR driving the expression of each (TCR α or TCR β). (D) The fusion of the TCR α and TCR α into a single protein that is linked with a 2A self-cleavage peptide allows for equal expression of the two TCR chains. The promoter in the 5′ LTR drives the expression of this TCR α/TCR β protein, which is subsequently cleaved inside the cell into the two free chains.

Structure of the viral vector impacts on expression of the T cell receptor genes.

A comparison of T cells transduced with two different vectors configured for expressing TCR α and β chains is shown. T cells transduced with a retroviral vector with the TCR chains fused with a 2A self-cleavage peptide (as in Figure 18.2D ) have high and uniform TCR expression (blue line in top panel). T cells transduced with a retroviral vector with each TCR chain expressed from a different promoter (as in Figure 18.2A ) have lower and variable TCR expression (blue line in bottom panel). Isotype control staining in both panels is depicted (red line).

High and equal expression of the TCR α and β chains is not adequate to ensure a functional TCR is expressed on the surface of T cells. TCR expression is limited by the components of CD3 complex. Because each mature T cell has its own TCR, there is competition for the CD3 complex in the T cell surface, thus limiting TCR transgene expression. Also, pairing between the endogenous and exogenous TCR chains further limits the amount of introduced TCR expressed. The issue of TCR assembly has been addressed in several ways. We noticed that human TCR genes have better expression and function when their human constant regions are substituted with mouse constant regions . It was determined that inclusion of elements of the mouse TCR constant regions led to more efficiently pairing between the TCR chains and with the CD3 complex . Altering the glycosylation of the TCR , adding cysteine residues to the constant regions to promote interchain disulfide bonds , and inclusion of leucine zippers have been shown to facilitate proper pairing of the introduced TCR chains, leading to enhanced TCR expression and function.

Another critical factor that influences a TCR’s ability to redirect the specificity of T cells is affinity. During our initial studies, we found that our HLA-A2-restricted, MART-1-reactive TCR could engineer normal CD8 ⁺ T cells to recognize peptide-loaded targets and melanoma cells, whereas TCR-transduced CD4 ⁺ T cells could recognize peptide-loaded targets but not melanoma cells . This dependence on the CD8 coreceptor for tumor cell recognition led us to postulate that if a TCR could be identified with high affinity, we could generate “better” CD8 ⁺ T cells and MHC class I-restricted CD4 ⁺ T cells . We and subsequently others were successful in identifying CD8-independent TCRs that could transfer MHC class I reactivity to CD4 ⁺ T cells . These high-affinity TCRs have the added benefit of requiring a lower level of expression to activate T cells, circumventing some of the expression issues described previously.

Naturally occurring CD8-independent TCRs are relatively rare, so they are difficult to find. It has been shown that some of the previously described modifications designed to enhance expression and function can also convert a CD8-dependent TCR to a CD8-independent TCR . Also, tumor-reactive T cell clones in transgenic mice expressing human HLA-A2 molecules can express CD8-independent TCRs . However, the field has generally taken the approach of modifying the CDR 3 regions of existing TCRs to convert them to high affinity and CD8 independence . High-affinity TCRs continue to be identified and evaluated for their use in TCR gene transfer studies.

There are clear advantages to using CD8-independent/high-affinity TCRs. First, lower TCR transgene expression is needed to activate the TCR gene-modified T cells. Second, CD8-independent/high-affinity TCRs are able to generate a novel population of MHC class I-restricted CD4 ⁺ T cells . However, the question of need for CD8-independent/high-affinity TCRs for engineering T cells has not been resolved. We investigated the impact on TCR affinity and tumor cell recognition by T cells . We found that by increasing the stability of a low-affinity TCR/peptide/MHC trimolecular complex, which essentially mimics increasing the affinity of the TCR, we got better antitumor reactivity. However, increasing the stability of a high-affinity TCR/peptide/MHC trimolecular complex did not improve T cell reactivity, suggesting there is an upper limit to which TCR affinity impacts T cell function. We argued that improved signaling in the T cell led to better function in T cells expressing high-affinity TCRs . The importance of T cell biology rather than TCR affinity makes sense given the TIL studies and the impact of immune suppression on T cells. The affinity of the TCRs is no different on TIL cells in the tumor lesion versus TIL cells infused into patients following ex vivo expansion, but their impact on disease can be quite different. Similarly, T cells bearing high-affinity receptors can be rendered inactive in the presence of Treg cells or myeloid-derived suppressor cells. Despite these arguments, it still makes sense to use high-affinity TCRs for TCR gene transfer studies if they are available.

In Vivo Studies

Adoptive transfer of TCR gene-modified T cells has been evaluated in mice. Infusion of TCR-transduced splenocytes led to rejection of tumors and decreased viral load, indicating that the TCR gene transfer approach could be effective in vivo . In addition, there is a report of engineering Treg cells to suppress unwanted immune responses . Most reports found no adverse events in mice , but recently there have been indications that TCR gene-modified T cells could lead to adverse events, presumably due to TCR mispairing . Although there remains debate in the field regarding the impact of TCR mispairing on adverse events, there are solutions to the potential problem of off-target reactivity of TCR-transduced T cells. Anything that facilitates pairing of the introduced TCR chains, such as introducing disulfide bridges ( Figure 18.4 ) , leucine zippers , or mouse constant regions ( Figure 18.4 ) into human TCRs , would limit mixed pairing. Two other clever approaches have been described that limit mixed pairing. It has been shown that when TCR αβ chains are introduced into γδ T cells, there is no mixed chain pairing, leading to effector cells with less chance of promoting off-target toxicity . Another approach to limit mixed chain pairing involved altering the structure of the introduced TCR chains . As shown in Figure 18.4 , a normal TCR consists of an α and a β chain, with each chain containing one variable region and one constant region. By fusing the Vα region to the Vβ and Cβ regions, a three-domain chain is generated ( Figure 18.4 ). In order for this construct to express on the surface of T cells, a Cα single-domain chain must be coexpressed. Because neither of these chains can pair with the endogenous TCR and only with each other, this would be an effective approach for eliminating adverse events due to chain mispairing.

Strategies to prevent TCR mixed-chain pairing.

A normal TCR consists of two chains, each having a single variable region (Vα, dark blue; Vβ, red) and one constant region (Cα, light blue; Cβ, tan). One strategy to prevent mixed pairing is to exchange the human constant regions (solid) for the corresponding mouse constant regions (striped). Another strategy to prevent mixed pairing is to introduce cysteines into the constant regions to promote disulfide bonds (green line) between the correct TCR chains. Another strategy to prevent mixed pairing is to construct a three-domain chain consisting of Vα linked to Vβ linked to Cβ and a second chain consisting of only Cα. These chains can only pair with each other and not with the endogenous TCR chains.

TCR gene-modified T cells have also been evaluated in humans. The first two clinical trials reporting the use of TCR gene-modified T cells to treat cancer patients were published in 2006 . The first trial used T cells engineered with the MART-1-reactive TCRs we first described . Allogeneic anti-MART-1 TCR-transduced T cells were injected into melanoma tumor lesions . The goal of this study was to destroy the treated lesion and induce systemic anti-melanoma immunity through cross priming. The second trial used T cells engineered with a different MART-1 TCR for systemic infusion into lymphodepleted stage IV melanoma patients . Both of these trials had objective clinical responses (1 of 15 and 2 of 17 patients, respectively) and indicated that TCR gene-modified T cells were generally safe and well tolerated by the patients.

There have been several additional reports of clinical trials using TCR gene-modified T cells in cancer patients. One report targeted the melanoma differentiation antigens MART-1 and gp100 . The anti-MART-1 TCR used in this trial had higher affinity than the two used in the previous trials, and there was an increase in the frequency of clinical responses (6 of 20 patients treated) . The anti-gp100 TCR reported in this study was of mouse origin and was selected for CD8 coreceptor-independent tumor cell recognition . Of the 16 patients treated with gp100 TCR-transduced T cells, 3 had objective clinical responses. Although no off-target adverse events were reported, the use of T cells engineered with these high-affinity TCRs was implicated with the observed toxicity that was associated with the destruction of normal melanocytes . Many patients had melanocyte destruction in the skin, eye, and ear, resulting in vitiligo, uveitis, and hearing impairment, respectively. Vitiligo (skin depigmentation due to melanocyte destruction) is not reversible with any known treatment and is considered a common side effect in patients treated with interleukin-2 (IL-2) and other forms of immunotherapy . Vision and hearing changes were unexpected side effects of the therapy that were resolved in almost all patients naturally or with use of topical steroids . Although high-affinity TCRs did lead to improved clinical outcomes for melanoma patients, the adverse events need to be closely monitored in similar studies in the future.

A major advantage of the TCR gene-modified T cells for treating cancer patients is that it is possible to treat any patient with autologous tumor-reactive T cell regardless of their natural capacity to develop an antitumor immune response. Raising CEA-reactive T cells with antitumor reactivity for patient treatment has proven to be challenging. A clinical trial was reported using a mouse anti-CEA TCR to engineer human T cells to treat patients with advanced colon carcinoma . All three patients treated had a transient decrease in serum CEA levels, and one patient had a partial clinical response. However, all patients had severe inflammatory colitis with grade 2 (one patient) and grade 3 (two patients) diarrhea. Despite the interesting clinical results, due to toxicity associated with this trial, enrollment of additional patients was halted.

The adverse events associated with some TCR gene transfer trials led to testing targets considered to be “tumor specific.” One attractive group of targets is the cancer testis antigens . TCRs have been isolated that target NY-ESO-I, MAGE-A3, MAGE-A4, and other cancer testis antigens . T cells engineered to express an NY-ESO-I TCR have been used to treat patients with melanoma and synovial cell sarcoma . Objective clinical responses were reported in 5 of the 11 melanoma patients and 4 of the 6 synovial cell sarcoma patients treated with NY-ESO-I TCR-transduced autologous T cells . None of the patients exhibited any of the adverse events (skin, eye, and hearing) reported in the previous trials. Therefore, it is clear that depending on the antigen targeted by TCR gene-modified T cells, the TCR gene transfer approach can be a safe and potentially effective treatment for advanced malignancies.