The genetic engineering of T cells through the introduction of a chimeric antigen receptor (CAR) allows for generation of tumor-targeted T cells. Once expressed by T cells, CARs combine antigen-specificity with T cell activation in a single fusion molecule. Most CARs are comprised of an antigen-binding domain, an extracellular spacer/hinge region, a trans-membrane domain and an intracellular signaling domain resulting in T cell activation after antigen binding.
We performed a search of the literature regarding tumor immunotherapy using CAR-modified T cells to provide a concise review of this topic.
This review aims to focus on the elements of CAR design required for successful application of this technology in cancer immunotherapy. Most notably, proper target antigen selection, co-stimulatory signaling, and the ability of CAR-modified T cells to traffic, persist and retain function after adoptive transfer are required for optimal tumor eradication. Furthermore, recent clinical trials have demonstrated tumor burden and chemotherapy conditioning before adoptive transfer as being critically important for this therapy. Future research into counteracting the suppressive tumor microenvironment and the ability to activate an endogenous anti-tumor response by CAR-modified T cells may enhance the therapeutic potential of this treatment.
The ability of the immune system to recognize and eradicate cancer is well established. This principle is most clearly demonstrated in the context of allogeneic hematopoietic stem cell transplantation for hematologic malignancies [1, 2]. However, in the allogeneic setting, the benefit of ‘graft versus leukemia’ is offset by the complications of graft versus host disease (GVHD). In the context of autologous ‘adoptive cellular therapy’ (ACT), the risk of GVHD is minimal and the eradication of cancer using autologous tumor-infiltrating lymphocytes (TILs) has been demonstrated in patients with melanoma . However, the process by which tumor reactive TILs are isolated and expanded is technically difficult, labour intensive and time consuming. To overcome these obstacles and broaden the application of ACT, many studies have reported the development of methods for genetically engineering T lymphocytes to target tumor cells. These strategies include engineering T cells with a chimeric antigen receptor (CAR) that redirects T cell specificity and function . Pioneering studies by Gross et al.  demonstrated the proof principle that expression of an antibody derived single chain variable fragment (scFv) coupled to a T cell signaling domain will redirect T cell specificity and function. With the advent of efficient methods of human T cell modification and the ability to rapidly expand these tumor-targeted T cells, this strategy has become a feasible treatment option [6-9]. In addition to CAR modification, an alternative method of generating tumor-targeted T cells is by the introduction of a T cell receptor (TCR) α and β chains with known antitumor specificity . However, this method of T cell modification is beyond the scope of the present review. The present review aims to focus on CAR design and the elements required for the successful eradication of malignancies by CAR-modified T cells. In addition, we discuss the findings obtained in early clinical trials testing CAR-modified T cells, as well as the future challenges to this evolving field.
Advantages of CAR-modified T cells
There are several advantages to utilizing CAR-modified T cells for cancer immunotherapy. CARs recognize tumor antigens in a human leokocyte antigen (HLA)-independent manner . This allows CAR-modified T cells to overcome the tumor's ability to escape immunodetection by down-regulation of HLA molecules on the cell surface [10, 11]. Furthermore, because tumor targeting is HLA-independent, the use of CARs is applicable to a broad range of patients irrespective of HLA-type. Targeting of tumor antigens by CAR-modified T cells is applicable to any cell surface antigen, including proteins, carbohydrates and glycolipids for which a monoclonal antibody can be generated. This enables CAR-modified T cells to respond to a broader range of targets compared to the native TCR. CAR modification can redirect the specificity of most T cell subsets, including CD4, CD8, naïve, memory or effector T cells. This is critically important because naïve and antigen experienced T cells have different functional capacities that may make them more or less favorable for use in adoptive cell therapy . The introduction of one or more T cell activating signals is possible with CAR modification, which may enhance the ability of T cells to proliferate, persist and lyse targeted cells. Furthermore, the genetic modification of T cells is not limited to activating signals. T cells can be engineered to deliver potent anti-tumor immunomodulators (e.g. cytokines) to the hostile tumor microenvironment, which may enhance their anti-tumor effect . Finally, and most significantly, the ability to generate a large quantity of tumor-specific T cells in a relatively short period of time makes this strategy feasible for use in the clinical setting [9, 14].
CAR-modified T cells: targeting
The basic design of a CAR includes a tumor-associated antigen (TAA)-binding domain (most commonly a scFv derived from the antigen-binding region of a monoclonal antibody), an extracellular spacer/hinge region, a trans-membrane domain and an intracellular signaling domain [4, 15]. The target antigen chosen for CAR specificity is a critical determinant for the effectiveness and safety of the genetically-engineered T cell. The ideal target antigen is expressed only on tumor cells and will produce the greatest effect if it is required for tumor cell survival. This type of TAA will have little chance for immune editing and tumor escape because it is critical for survival of the tumor . Unfortunately, the expression of most TAAs is not restricted to tumor cells, leading to undesirable effects of ‘on target/off tumor’ toxicity. Historically, CAR-modified T cells have only targeted antigens expressed on the target cell surface. However, CARs that target intracellular antigens (via HLA-restricted presentation) have also been developed [17, 18].
One widely utilized target of CAR-modified T cells is the CD19 antigen expressed not only on almost all normal B cells, but also most B cell malignancies . Because CD19 is not expressed on hematopoietic stem cells, the toxicity of targeting this antigen is limited to B cell aplasia after treatment with anti-CD19 CAR-modified T cells [20-22]. This ‘on target/off tumor’ toxicity, in the form of B cell aplasia, is considered to be a tolerable side-effect of this therapy. Some other examples of targets for CAR-modified T cells include ERRB2 (HER-2/neu) for the treatment of breast and prostate cancer [23, 24], prostate-specific membrane antigen (PSMA) for the treatment of prostate cancer [25, 26], carboxy anhydrase-IX (CAIX) for the treatment of renal cell carcinoma [27, 28], Lewis Y for the treatment of lung and ovarian tumors [29, 30], carcinoembryonic antigen for the treatment of colon cancer [31, 32], folate-binding protein, folate receptor or MUC-CD for the treatment of ovarian cancer [33, 34], and the diasialoganglioside GD2 for the treatment of neuroblastoma . Although this is not an exhaustive list, it clearly illustrates the wide range of tumors that can be targeted by CAR-modified T cells.
Although most CARs utilize a scFv for tumor targeting, some studies report the development of CARs consisting of receptor ligands fused to intracellular signaling domains. Examples include a polypeptide against vascular endothelial growth factor receptor (anti-VEGFR2), an integrin binding peptide (anti-αvβ6), heregulin (anti-Her3/4 receptor) or interleukin (IL)-13 mutein (anti-IL13 receptor-α) [36-39]. In yet another approach Zhang et al.  have demonstrated the anti-tumor efficacy of a CAR containing the NKG2D receptor coupled to the CD3 zeta (CD3ζ) chain. In this manner, genetically-modified T cells will target the ligands of NKG2D (e.g. MIC A/B) that are overexpressed on a number of tumors. In an attempt to develop a ‘universal’ CAR, Ang et al.  have developed an anti-fluorescein isothiocyanate (anti-FITC) CAR. In this manner, FITC-conjugated therapeutic molecules (e.g. monoclonal antibodies, ligands, or nucleic acid based aptamers) that are specific to one or more TAAs can be employed to target the tumor of interest. T cells genetically-modified to express the anti-FITC CAR will then target the FITC-labeled tumors. This method allows the targeting of multiple TAAs that will limit the ability of tumors to evade targeting by down-regulating a single target antigen . Furthermore, this approach allows the treatment of several tumor types with one CAR.
Attempts at improving CAR design and T cell activation have involved increasing the affinity of the antibody-recognition domain. It has been postulated that increased CAR antigen affinity will improve the targeting of tumors with low antigen expression and amplify T cell activation. However, Chmielewski et al.  illustrated that the maximum activation of T cells via CAR antigen-binding was independent of the antigen-binding affinity . It was also noted that low affinity receptors could discriminate between tumors with low and high expression of the target antigen . This characteristic could be used to reduce the severity of ‘on target/off tumor’ toxicity, in that CAR-modified T cell activation would be limited to cells with overexpression of the target antigen. In addition, careful consideration must be used in selecting CARs with high antigen affinity. When antigen binding is too avid, the effector T cell might be unable to engage multiple targets thereby limiting its effectiveness .
Because most currently designed CARs are based on scFv derived from murine antibodies, these foreign molecules may elicit an undesired anti-CAR response by the host. This anti-CAR effect was demonstrated by Lamers et al.  who found both antibody and cell mediated responses to cells expressing CARs. A strategy to prevent this outcome includes humanization of scFv fractions or utilizing human antibodies to generate CARs . The optimal format for developing the antigen-binding domain of a CAR remains to be established and warrants further preclinical and clinical investigation.
CAR-modified T cells: signaling
The optimal T cell activation signaling domains incorporated into a CAR remain a topic of debate. First-generation CARs mediate T cell activation through immunoreceptor tyrosine-based activating motif of the CD3ζ chain or the FcεRIγ . The CD3ζ signal was found to provide the requisite ‘signal 1’ resulting in T cell activation, target cell lysis, modest IL-2 secretion and in vivo anti-tumor function [46-48]. Subsequent to these initial reports, the anti-tumor potential of T cells modified with first-generation CARs has been demonstrated in preclinical studies [47-51]. Despite these in vivo results, the anti-tumor effect of first-generation CARs is limited. Suboptimal stimulation of the T cell with only ‘signal 1’ results in T cell anergy, leading to poor cytokine secretion, poor proliferation and eventual apoptosis of the genetically-modified T cell [52, 53]. Furthermore, tumor eradication was predicated on the expression of co-stimulatory molecules (e.g. B7.1/CD80) on the tumor cell surface .
To enhance CAR activation signals and overcome the limitation of first-generation CARs, second-generation CARs were developed that incorporated co-stimulatory domains. The most well studied T cell costimulatory receptor is CD28, which interacts with the B7 family molecules, B7.1 and B7.2, located on the surface of target cells [54, 55]. According to the current model of T cell activation, CD28 provides a second activation signal (co-stimulation; ‘signal 2’) that augments ‘signal 1’ from the TCR/CD3 complex [56, 57]. Costimulation by CD28 enhances T cell proliferation, IL-2 synthesis and expression of the anti-apoptotic protein Bcl-xL . To replicate this endogenous T cell activation, second-generation CARs were designed to deliver both ‘signal 1’ from the CD3ζ domain and ‘signal 2’ from a CD28 signaling domain. Maher et al.  tested a CD28 containing second-generation CAR targeted to PSMA. When compared with a first-generation anti-PSMA CAR (signal 1 only), the second-generation CAR led to enhanced IL-2 production, increased proliferation in response to antigen and sustained lysis of PSMA + targets by genetically-modified T cells. In a preclinical model of B cell malignancy, Brentjens et al.  demonstrated enhanced eradication of established tumors using a CD28 containing second-generation CAR compared to a first-generation CAR specific for the same antigen (CD19). Several studies have similarly demonstrated increased proliferation, increased cytokine production, up-regulation of anti-apoptotic proteins (e.g. Bcl-xL) and delayed activation-induced cell death with CD28 containing second-generation CARs [32, 59-63]. In addition, CD28 containing second-generation CARs may enable genetically-modified T cells to counteract the inhibitory effects of the ‘hostile tumor microenvironment.’ Studies have demonstrated that T cells genetically-modified with a CD28 containing CAR proliferate and express nuclear factor kappa-B despite the repressive effects of both transforming growth factor (TGF)-β and T regulatory cells (Tregs) [64, 65]. The persistence of genetically-modified T cells is also enhanced when utilizing CD28 containing second-generation CARs .
Second-generation CARs have also been developed using alternative co-stimulatory molecules. Coupling CD3ζ signaling with other B7 family members [e.g. inducible costimulation (ICOS)] or tumor necrosis factor receptor (TNFR) family (e.g. CD137/4–1BB or CD134/OX-40) has been described . Finney et al.  demonstrated enhanced antigen-induced cytokine [e.g. IL-2, IFN-γ, TNF-α, granulocyte-macrophage colony stimulating factor (GM-CSF)] production by T cells modified to express B7 family and TNFR family containing second-generation CARs . Co-stimulation with CD28 produced the highest level of IL-2, and B7 family co-stimulation resulted in higher levels of IFN-γ, TNF-α and GM-CSF compared to TNFR containing CARs. All second-generation CARs enabled resting T cells to express the anti-apoptotic protein Bcl-2 and enhanced proliferation in response to antigen compared to first-generation CARs. Target cell lysis was enhanced by the inclusion of CD28, ICOS and CD134, with ICOS co-stimulation exhibiting the highest in vitro target cell lysis . However, the incorporation of ICOS signaling may result in suboptimal IL-2 production, which may detract from using this co-stimulatory domain in CAR design . Our group also compared a similar panel of second-generation CARs that included CD28, DAP10, OX-40 and 4-1BB co-stimulatory domains . In these studies, CD28 containing CARs demonstrated superior in vitro proliferation and cytokine secretion compared to the alternative constructs . In addition, CD28 containing second-generation CARs had enhanced in vivo antitumor activity compared to first-generation CARs in the setting of tumors that fail to express costimulatory ligands .
By contrast, Milone et al.  compared the in vivo anti-tumor efficacy of CD28 versus CD137 (4–1BB) second-generation CARs in an acute lymphoblastic leukemia (ALL) xenograft model of disease . In their study, a CD19 targeted CD137 containing second-generation CAR had an enhanced anti-leukemic effect, improved persistence and an antigen-independent activation of T cells resulting in improved efficacy after adoptive transfer . However, in a mesothelioma tumor model, Carpenito et al.  also demonstrated equivalent anti-tumor efficacy for both CD28 and 4-1BB containing second-generation CARs. Other studies have also reported similar results with CD137 containing second-generation CARs, noting that CD137 co-stimulation results in improved T cell survival, activation-induced cell death resistance, increased expression of anti-apoptotic proteins (e.g. Bcl-xL), sustained proliferation and persistence, enhanced cytokine production, and increased antigen-specific tumor cell lysis [71, 72].
In addition to the work on CD28 and 4-1BB, alternative methods to provide co-stimulation have been developed. The incorporation of CD244, a NK cell receptor, into the CAR design resulted in the acquisition of a cytolytic effector memory phenotype and augmented CAR-mediated responses compared to first-generation CAR (CD3ζ only) modified T cells [73, 74]. However, comparisons between CD28, CD137 and CD244 containing second-generation CARs have yet to be performed. In an alternative method Stephan et al.  demonstrated potent auto- and trans-costimulation of T cells modified to express a first-generation CAR and co-transduced to express co-stimulatory ligands (e.g. CD80 and 4-1BBL). T cells modified in this manner have increased proliferation, cytokine secretion, in vitro survival, in vivo expansion, in vivo persistence and eradication of systemic malignancy in a mouse model of disease .
In an attempt to further optimize CAR design several groups have developed ‘third-generation’ CARs that provide signal 1, signal 2 and an additional costimulatory signal (e.g. CD28/4-1BB/CD3ζ signaling). Comparisons between second-generation CARs and third-generation CARs have demonstrated conflicting results. Several studies have reported enhanced cytokine production, T cell survival and anti-tumor efficacy for T cells that express a third-generation CAR [70-72, 76, 77]. Zhong et al.  have demonstrated enhanced cytokine production, improved in vivo T-cell survival, enhanced tumor elimination, improved PI3K/AKT pathway activation, enhanced Bcl-xL expression and reduced T cell apoptosis for a third-generation CAR (CD28/4–1BB/CD3ζ signaling) against PMAS (anti-PSMA) . Pule et al.  demonstrated sustained in vitro proliferation, enhanced IL-2 production and the ability to maintain cytolytic function after repeated antigenic stimulation for a CD28-OX40 containing third-generation CAR. Wilkie et al.  compared third-generation CARs (CD28/4-1BB/CD3ζ versus CD28/OX-40/CD3ζ) targeting MUC1 (expressed on breast and ovarian tumors). In their study, T cells modified with either third-generation CAR had in vitro cytotoxicity equivalent to T cells modified with a CD28 containing second-generation CAR directed against the same target. However, CD28/OX-40 containing third-generation CAR-modified T cells had improved in vitro IFN-γ secretion . Unfortunately, a direct comparison between the in vivo anti-tumor efficacies of these third-generation CARs was not performed.
It must be noted that differences between second- and/or third-generation CARs may not be solely attributable to the signaling domains incorporated in their design. Rather, the difference in antigen-binding domains (i.e. scFv), method of transduction (lentivirus versus retrovirus), tumor model (systemic versus subcutaneous), route of T cell administration (intravenous versus intraperitoneal versus intra-tumoral) and culture conditions, as well as a number of other variables, could account for the differences found in these studies. Accordingly, the optimal combination of T cell activation signaling is still debated. However, several studies are currently testing second- and third-generation CARs in clinical trials and a more thorough understanding should be forthcoming.
CAR-modified T cells: trafficking
Genetically-modified tumor-targeted T cells must traffic to the site(s) of disease to effectively eradicate disease. The ability of CAR-modified T cells to localize to the site(s) of tumor has been demonstrated using a dual bioluminescent imaging of genetically-modified T cells and tumor cells in a mouse model of cancer . CAR-modified T cells accumulated at most site(s) of the systemic tumor and persisted over time . More significant are the findings of recent clinical trials showing that T cells genetically-modified to express a second-generation CAR localize to sites of disease [20, 22, 66]. Our group has demonstrated CD19 targeted T cells expressing a CD28 second-generation CAR trafficked to several sites of disease (e.g. bone marrow, lymph nodes, liver) after adoptive transfer in patients with chronic lymphocytic leukemia (CLL) . Savoldo et al.  have also demonstrated the trafficking of CAR-modified T cells to a skin lesion 2 weeks after ACT in a patient with non-Hodgkin lymphoma (NHL) . T cells targeted to the same antigen (CD19) but genetically-modified with a first-generation CAR (CD3ζ only) or a CD28 containing second-generation CAR were simultaneously infused into patients. Strikingly, only T cells modified with the CD28 containing second-generation CAR were found to traffic to the tumor site . In another study using a 4-1BB containing second-generation CAR, genetically-modified T cells were found in the bone marrow of CLL patients after adoptive transfer . Taken together, these results are consistent with the ability of CAR-modified T cells to localize to site(s) of disease.
Although studies reported above are promising, it remains possible that genetic engineering and ex vivo expansion of T cells may alter the expression of one or more chemokine/cytokine receptors necessary for trafficking. One strategy for enhancing the trafficking of T cells to site(s) of disease is through the genetic expression of chemokine receptors. Several studies have demonstrated this principle through the expression of CXCR2 (CXCL1 receptor) and CCR4 (CCL17 receptor) in CAR-modified T cells [81, 82]. As this therapeutic strategy evolves, finding the optimal methods that enable T cells to traffic to site(s) of tumor will be critical for successful tumor eradication.
CAR-modified T cells: persistence
Tumor-targeted T cells must persist for a sufficient period of time to result in successful tumor elimination. In ACT utilizing TILs, the persistence of adoptively transferred T cells was correlated with an improved clinical response . Conditioning chemotherapy can enhance the persistence of adoptively transferred T cells through a number of mechanisms, as demonstrated in preclinical models and in the setting of ACT using TILs in patients with melanoma [84, 85]. Our group recently verified the principle of CAR-modified T cell persistence after conditioning chemotherapy . An initial cohort of CLL patients treated with CD19 targeted T cells and without conditioning chemotherapy had limited to undetectable persistence. By contrast, a subsequent cohort of patients who received previous conditioning chemotherapy (cyclophosphamide) demonstrated evidence of genetically-modified T cells in the bone marrow for up to 6 weeks after adoptive transfer . Conditioning chemotherapy may also reduce the patient's disease burden, which will enhance the persistence of the adoptively transferred T cells. In one study, a lower disease burden was correlated with T cell persistence in patients with metastatic neuroblastoma treated with CD171-targeted T cells . This finding was also demonstrated in our study, with an inverse relationship between the persistence of genetically-modified cells and the peripheral blood tumor burden .
The link between T cell phenotype and persistence has also been investigated in the context of ACT. After antigen exposure, naïve T cells can develop into one of two memory subsets termed effector memory (TEM) or central memory (TCM) . Naïve and antigen experienced T cells have different functional capacities that may make them more or less favorable for use in adoptive cell therapy . In nonhuman primates, adoptive transfer of a CMV specific CD8+ T cell clones derived from a TCM (CD62L+) but not from a TEM (CD62L–) can establish persistent T cell memory . In a recent study of CAR-modified viral specific and nonviral specific T cells, the persistence of adoptively transferred T cells was predicated on an increased frequency of helper (CD4+) and central memory (CD45RO + CD62L+) cells within the infused product .
The persistence of modified T cells may also be contingent on the signaling domain incorporated into the CAR. Savoldo et al.  demonstrated enhanced persistence (up to 6 months) of T cells modified with a CD28 containing second-generation CAR compared to first-generation modified T cells targeting the same antigen (CD19). Other studies support the ability of T cells modified with a CD28 containing second-generation CAR to persist in the blood and bone marrow after adoptive transfer into patients [20, 21]. In comparison Kalos et al.  demonstrated the persistence of CD19 targeted T cells containing a 4-1BB second-generation CAR for > 9 months after adoptive transfer. Patients exhibited a significant benefit from CAR-modified T cells and this may be the clearest indication that the persistence of CAR + T cells may be required for optimal clinical responses.
The administration of IL-2 may also enhance the persistence of genetically-modified T cells. Till et al.  treated seven patients with indolent NHL with T cells modified to express a first-generation CAR targeted to CD20. The persistence after infusion of modified T cells was enhanced in patients who received low-dose subcutaneous IL-2 . However, although patients in the trial had minimal toxicity after the infusion of low-dose IL-2, the use of exogenous IL-2 is tempered by the toxicity profile seen in other studies .
One strategy to enhance the persistence of CAR + T cells is through the use of virus-specific T cells. In this manner, co-stimulation of CAR-modified T cells occurs through the engagement of the native TCR against viral antigens. Pule et al.  reported on the safety of this approach and demonstrated enhanced survival of CAR-modified autologous Epstein–Barr virus specific cytotoxic T lymphocytes (CAR-CTLs) compared to activated T cells (CAR-ATC) when modified with a distinguishable first-generation CAR against the same tumor antigen (GD2) . Although, initially, CAR-CTLs survived in the circulation at higher level than CAR-ATCs, this difference was lost by 6 weeks after infusion . In a recent update of this trial, the persistence of either CAR-CTLs or CAR-ACTs beyond the 6-week timepoint was not reported as being different . The persistence of either modified T cell population was contingent on the frequency of helper (CD4+) and central memory (CD45RO + CD62L+) cells within the infused product . However, despite these findings, it should be noted that the use of genetically-modified viral specific T cells is still an attractive method for broadening ACT using allogeneic donor derived virus-specific T cells that reduced risk of GVHD after adoptive transfer .
Factors affecting the persistence of adoptively transferred T cells include the use of conditioning chemotherapy, patient tumor burden, cytokine supplementation, co-stimulatory signaling, T cell phenotype, and possibly the use of viral specific T cells. The optimal method for the persistence of CAR-modified T cells is yet to be fully defined; however, it is clear that efficacy of ACT is contingent on the persistence of the T cells after adoptive transfer.
CAR-modified T cells: function
Optimal tumor eradication by CAR-modified T cells requires the ability to maintain cytolytic function after adoptive transfer. The hostile tumor microenvironment includes Tregs, myeloid derived suppressor cells and several immunosuppressive molecules/cytokines (e.g. TGF-β) that inhibit the anti-tumor function of adoptively transferred or endogenous T cells [10, 94]. Early clinical trials testing CAR + T cell therapy had disappointing results, which could have been a result of the suppressive effects of this hostile tumor microenvironment [86, 90, 91]. It has been shown that T cells modified to express a CD28 containing second-generation CAR were resistant to in vitro inhibition by Tregs . However, that study used inducible Tregs (iTregs), which have unstable Foxp3 expression . Natural Tregs have been demonstrated to inhibit the in vitro and in vivo anti-tumor function of T cells modified to express a second-generation CAR . Conditioning with irradiation or chemotherapy can reduce the number of Tregs allowing gene modified T cells to eradicate tumor cells. This principal was demonstrated by our group using cyclophosphamide to restore the antitumor activity of CAR-modified T cells despite targeted inhibition of natural Tregs in a mouse model of cancer .
Tumor eradication by CAR-modified T cells remains the best measure of retained anti-tumor function (clinical responses reported in clinical trials are discussed below). A surrogate indication of retained function after adoptive transfer is the development of ‘on target/off tumor’ toxicity (e.g. B cell aplasia and/or hypogammaglobinemia when targeting the CD19 antigen). Recent trials utilizing anti-CD19 CARs have demonstrated B cell aplasia in the peripheral blood after adoptive transfer [20-22]. Elevation of pro-inflammatory cytokine profiles is another surrogate marker for T cell function after adoptive transfer. This was demonstrated by Kalos et al.  in CLL patients by an increase in several pro-inflammatory cytokines in the peripheral blood and bone marrow that was correlated with peak expansion of CAR-modified T cells. Adoptively transferred T cells also maintained their ex vivo capacity to degranulate in response to target antigen as assessed by CD107a surface expression .
CAR-modified T cells: toxicity
The implications of ‘on target/off tumor’ toxicity can be far more serious than B cell aplasia. One patient with metastatic colon cancer died 5 days after lymphocyte depleting chemotherapy (cyclophosphamide and fludarabine) followed by infusion of modified T cells targeted to ERBB2 . It was speculated that a large number of infused modified T cells localized to the lung, resulting in the release of pro-inflammatory cytokines after the recognition of low levels of ERBB2 on lung epithelia . The resulting pro-inflammatory cytokine release triggered pulmonary toxicity, multi-organ failure and eventual death of the patient . In another trial, liver toxicity was reported in five out of 11 patients with renal cell carcinoma treated with T cells modified to express a first-generation CAIX specific CAR [28, 98]. It was concluded the toxicity was the result of modified T cells targeting bile duct epithelial cells that had low levels of CAIX expressions . These studies highlight the need for target antigens that are uniquely expressed on cancer cells at the same time as sparing normal tissue.
At our center, we safely infused CD19 targeted T cells into three CLL patients without conditioning chemotherapy, although our first patient who received conditioning chemotherapy (cyclophosphamide) followed by CAR-modified T cells died 2 days after T cell infusion . After a thorough review of all clinical data, including an autopsy, the likely cause of death in this patient was infection and not the infused modified T cells . Despite these findings, our clinical trial was modified to divide the infusion of T cells over 2 days to enhance safety. Subsequent to this modification, we have infused an additional five patients with CAR-modified T cells after chemotherapy without any further adverse events .
First-generation CAR-modified T cell trials
Several centers have conducted clinical trials testing the anti-tumor efficacy of first-generation CAR-modified T cells. Overall, early clinical trials using first-generation CAR-modified T cells have failed to demonstrate significant clinical benefit [28, 86, 89-92, 98, 100]. Furthermore, significant variation exists between these trials including the targeted antigen, the gene transfer method, the method of ex vivo expansion, cell dose, the use of exogenous IL-2, and the use of conditioning therapy. Nevertheless, we have gained invaluable insight into the factors that constitute effective ACT using CAR-modified T cells.
Lamers et al. [28, 98] reported their experience regarding patients with metastatic clear cell renal cell carcinoma treated with T cells expressing an anti-AIX CAR with and without subcutaneous IL-2. Infusions of the CAR-modified T cells resulted in liver toxicity in five out of the eight patients infused. It was concluded that the hepatic toxicity seen in this trial was most likely related to targeting of CAIX on bile duct epithelial cells . Three additional patients were treated without liver toxicity on a modified protocol that included the infusion of an anti-CAIX monoclonal antibody (to block CAIX on normal liver tissue) before T cell infusion . The persistence of CAR-modified T cells was limited for all patients on this trial and no objective clinic responses were observed. Significantly, patients developed both a humoral and cellular anti-CAR response, which could help explain the limited persistence of modified T cells .
Kershaw et al.  described the treatment of 14 ovarian cancer patients with CAR-modified T cells directed against the α-folate receptor. Subcutaneous IL-2 was given to a majority of patients (n = 8) and adverse events were consistent with the toxicity profile of IL-2. Again, modified T cells had limited persistence and limited tumor localization as determined using 111Indium-labeled T cells. No objective responses were observed and an inhibitory factor developed in three patients (out of six tested) which, when tested, reduced the in vitro anti-folate receptor tumor response of modified T cells .
Park et al.  reported the adoptive transfer of CAR-modified CD8+ clones directed against the L1-cell adhesion molecule (L1-CAM; CD171) in six patients with neuroblastoma. No clinical responses were observed; however, increased T cell persistence (as measured by quantitative polymerase chain reaction (Q-PCR) of peripheral blood) was noted in a patient with limited disease burden compared to patients with larger tumor burdens .
Till et al.  reported on the adoptive transfer of CAR-modified T cells targeting CD20 in seven patients with refractory or relapsed indolent NHL. In the first cohort, no clinical responses were seen and the persistence of T cells was limited. In the second cohort, who received subcutaneous IL-2, the persistence of modified T cells was enhanced. One patient in the second cohort exhibited a partial remission and one patient had decreased metabolic activity of their tumor as measured by positron emission tomography .
Louis et al.  and Pule et al.  treated 19 patients with neuroblastoma using two populations (viral and nonviral specific) of T cells modified with a CAR directed against GD2. Three patients with active disease (bone or bone marrow disease) achieved complete remission after adoptive transfer, of which two remain in complete remission (1 and 4 years after T cell infusion). The persistence of modified T cell was contingent on the increased frequency of helper (CD4+) and central memory (CD45RO + CD62L+) cells within the infused product. However, over time, viral specificity had no impact on the persistence of genetically-modified T cells . Furthermore, the persistence of modified T cells correlated with a superior clinical outcome . Modified T cells persisted at low levels for up to 192 and 96 weeks for nonviral and viral specific CAR-modified T cells, respectively, as determined by Q-PCR of peripheral blood .
Jensen et al.  treated four patients with recurrent NHL with CAR-modified T cells. Two patients with diffuse large B cell lymphoma were treated with cloned CD8+ CTLs expressing a CD20-specific CAR after autologous hematopoietic stem cell transplant. Two patients with follicular lymphoma were treated with polyclonal CAR-modified T cells expressing a CD19-specific CAR and low dose subcutaneous IL-2. Modified T cell persistence was again limited; cellular anti-transgene immune responses was noted for two patients, and no objective clinical responses were found .
Second-generation CAR-modified T cell trials
Based on promising preclinical data, several centers initiated clinical trials using T cells modified with second-generation CARs. Kochenderfer et al.  described the treatment of one NHL patient with a CD28 containing second-generation CAR directed against the CD19 antigen. This patient received conditioning therapy that included cyclophosphamide and fludarabine followed by modified T cells and IL-2. This therapy resulted in partial remission of the patient's lymphoma (for up to 32 weeks), B-cell aplasia (for up to 39 weeks) and the persistence of CAR-modified T cells shown by Q-PCR of the blood (for up to 27 weeks) .
Savoldo et al.  treated six patients with relapsed or refractory NHL with CAR-modified T cells. In this trial, nonpreconditioned patients received simultaneous infusion of two autologous T cell products (first-generation and CD28 containing second-generation CARs) both specific for CD19. Superior persistence, expansion and trafficking to a site of disease (cutaneous skin lesion) for CD28 containing second-generation CAR-modified T cells were demonstrated . However, patients did not show evidence of sustained tumor regression in this cohort.
Our group described the treatment of nine patients (eight with CLL and one with ALL) with CD28 containing second-generation CAR-modified T cells directed against the CD19 antigen . T cell infusion was well tolerated in all but one patient who died 2 days after modified T cell infusion . Six patients received conditioning chemotherapy with cyclophosphamide before T cell infusion. The persistence of modified T cells was enhanced with previous chemotherapy and inversely proportional to the peripheral blood tumor burden . Three out of four evaluable patients with bulky CLL exhibited a response to treatment with conditioning chemotherapy followed by modified T cell infusion, including one patient who had a marked clinical and radiographic reduction in lymphadenopathy. The one patient treated with ALL exhibited a B cell aplasia before allogeneic hematopoietic stem cell transplantation. Modified T cells were also found to traffic to sites of disease (bone marrow, lymph nodes, liver) and retained ex vivo cytotoxicity after adoptive transfer as previously discussed .
Kalos et al.  and Porter et al.  June and colleagues described the treatment of three CLL patients with a 4-1BB containing second-generation CAR targeting CD19. In these studies, patients received conditioning chemotherapy before T cell infusion. Strikingly, modified T cells were demonstrated to expand (by up to 10 000-fold) after adoptive transfer, traffic to site of disease (bone marrow) and were detected at high levels (by flow cytometry and Q-PCR of peripheral blood and bone marrow) for up to 6 months post-infusion . All three patients exhibited a response after treatment (two complete remission and one partial remission), and the development of a tumor lysis syndrome correlated with elevation of peripheral blood CAR-modified T cell number and pro-inflammatory cytokine levels . In addition, B-cell aplasia, decreased plasma cell number and hypogammaglobulinemia were evident in these patients .
Third-generation CAR-modified T cell trial
As noted above, one patient with colon cancer was treated with a third-generation (CD28/4-1BBCD3ζ) targeting ERBB2 . The patient died 5 days after lymphocyte-depleting chemotherapy (cyclophosphamide and fludarabine) followed by infusion of modified T cells. The ‘on target/off tumor’ toxicity from the modified T cells lead to a clinically significant release of pro-inflammatory cytokines, resulting in pulmonary toxicity, multi-organ failure and the eventual death of the patient .
The ultimate promise of tumor immunotherapy is the cure of cancer without the toxicity of conventional treatments. The preclinical and early clinical results of CAR-modified T cells have given hope that this aim is within reach. The treatment of cancer with CAR-modified T cells has several advantages: HLA independent recognition of target antigens, broad applicability to most patients, ability to circumvent ‘tumor escape’ and the ability to rapidly deliver a population of tumor-specific T cells. The successful application of this technology will require the identification of target antigens that are uniquely expressed on tumor cells, thereby minimizing the risk of toxicity. In addition, as demonstrated in recent clinical reports, a prerequisite for the success of this therapy is the in vivo persistence of CAR-modified T cells after adoptive transfer [20-22]. We have shown that multiple injections of modified T cells can artificially increase T cell persistence and enhance anti-tumor efficacy in a mouse model of cancer . However, this technology should allow for a single injection of T cells which engraft, proliferate, persist and retain targeted cytotoxic function for a lifetime. Otherwise, this technology is ultimately an expensive and ineffective intervention. Continued investigation into the elements that govern the persistence of tumor-targeted T cells is essential. Thus far, signaling, tumor burden, conditioning chemotherapy, T cell phenotype and the use of supplementary cytokines have all been implicated. Furthermore, as noted in the present review, several of the early trials have demonstrated the development of an immune response against the adoptively transferred cells [91, 98, 100]. Understandably, this phenomenon has limited the persistence and efficacy of adoptively transferred cells. Moving forward, the development of less immunogenic CARs (e.g. humanized scFvs), promotion of tolerance and/or optimizing immunosuppression will need to be investigated for the successful application of this therapy.
Although recent studies have demonstrated success for this technology in hematologic malignancies, the application of this therapy to solid malignancies may need further development. Although early research has focused on the ability to reliably generate tumor-targeted T cells, it is possible that CAR-modified T cells will be rendered ineffective upon entering the suppressive tumor microenvironment. Thus, the future of this therapy is in the generation of CAR-modified T cells that resist the anergy and apoptosis that occurs for all immune effectors within the tumor microenvironment. Furthermore, more than overcoming the hostile tumor microenvironment, this therapy should have the capacity to recruit an endogenous anti-tumor response. Targeting a single antigen on a tumor cell may not only initially result in a decrease in tumor burden, but may also select for the expansion of tumor cells lacking this target antigen. Therefore, although genetic modifications of T cells may allow us to specifically target the tumor and potentially overcome the hostile tumor microenvironment without recruitment and activation of the endogenous immune system, it is likely that this therapy will be insufficient to cure the majority of patients. What is further required is the activation of an endogenous anti-tumor response (e.g. TILs, NK cells, innate immune system) by the CAR-modified T cells. If effective, it is plausible that this activation will induce epitope spreading against several antigens expressed by the tumor, thereby reducing the ability of tumors to escape eradication.
Although, scientifically, this field has significantly progressed in a relatively short period of time, the success of all cancer therapies is only measured by the impact that they have on patients in the clinic. To fully define the efficacy of this therapy, multi-institutional clinical trials will need to be conducted, which require a significant expenditure of labour and funding. Unfortunately, these expenditures can be prohibitive and the currently available funding mechanisms inadequately cover their costs. However, with the recent landmark responses seen in smaller single institutional studies, there is a new hope that resources will be made available for the development of this promising therapy.
With this review, we have attempted to outline the qualities that translate into successful tumor immunotherapy using CAR-modified T cells (Figure 1). Thus far, the target antigen, signaling domain, ability to traffic, persistence, retained function, patient disease status and conditioning regime have all been critically important. We conclude that the future of this field lies in the development of genetically-engineered tumor target T cells that can overcome the hostile tumor microenvironment and recruit an endogenous anti-tumor response. The final hurdle for researchers in this field is the implementation of clinical trials and securing of the funding needed to complete them. As this technology progresses and more reports of successful eradication of malignancy occur, these hurdles should be easily overcome.
Supported in part by grants from the National Institutes of Health (CA-138738, CA-59350), Alliance for Cancer Gene Therapy, Damon Runyon Clinical Investigator Award (R.J.B.), The Annual Terry Fox Run for Cancer Research (New York, NY) organized by the Canada Club of New York, Kate's Team, Mr William H. Goodwin and Mrs Alice Goodwin and the Commonwealth Cancer Foundation for Research and the Experimental Therapeutics Center of MSKCC, Geoffrey Beene Cancer Foundation, William Lawrence and Blanche Hughes Foundation, and the St Baldrick's Foundation Post Doctoral Fellowship in Childhood Cancer (K.J.C.). The authors would like to thank Joe Olechnowicz for his skilled work in reviewing this manuscript. The authors declare that there are no conflicts of interest.