T-cell expressing chimeric antigen receptor (CAR T-cell) has emerged as promising therapy for hematological malignancies. Since 2017, four CAR T-cell products - axicabtagene ciloleucel (axi-cel), tisagenlecleucel (tisa-cel), lisocabtagene maraleucel (liso-cel) and brexucabtagene autoleucel - have received the approvement by US Food and Drug Administration (FDA) for treatment of relapsed/refractory (R/R) Non-Hodgkin B-cell lymphoma (B-cell NHL). At the same time, tisa-cel has been approved by the FDA for R/R acute lymphoblastic leukemia (ALL) treatment.1

One crucial aspect of CAR T-cell therapy is to establish and validate a reproducible manufacturing process of high-quality able to deliver clinical-grade CAR-T cell products. It is a prerequisite for the wide application of this technology. Product quality needs to be built-in within every step of the manufacturing process (Figure 1).

Figure 1.

Production of CAR T cells in a GMP facility.

(0.32MB).

However, before the clinical-grade platform validation, many preclinical assays have to be performed. There are several imperative experiments requested by different regulatory international agencies - for instance, FDA and European Medicines Agency (EMA) - as essential pre-step data to establish a robustic investigational new drug (IND) dossier. In Brazil, the Agência de Vigilância Sanitária (ANVISA) incorporated on its regulatory framework the same principles of FDA and EMA.2

There are great expectations for the imminent introduction of CAR-T therapy in Brazil. Its significant complexity, however, represents an important challenge for proper implementation and achievement of the same goals observed in other countries. This consensus is based on the expert opinion that appointed the minimum preclinical assays required to support the development of new CAR-T cell therapies.

With the increasing progress in genetic modification technologies, several approaches have already been tested to generate CAR-T cells. The two main methods are viral transduction (retroviral or lentiviral vector), or transfection with naked plasmid DNA, transposase DNA-mediated integration or mRNA by electroporation or by lipid-based transfection system.3 Commonly used vectors with their advantages and disadvantages are listed in Table 1. Most current studies use retroviral vectors (such as lenti- and retroviral vectors). Lentiviral vectors have become particularly attractive for clinical applications due to their ability to efficiently transduce most cell types, including non-proliferating cells, such as Naive T cells.4 Both γ-retroviruses and lentiviruses integrate semi-randomly into the host cell genome. Thus, presenting a risk of insertional mutagenesis and dysregulation of genes adjacent to the integration site5 and local hotspots.6

The development and use of non-integrative vectors, or ones that integrate at specific locations, is an important goal for CAR therapy. The development of new vectors has been driven mainly by the need to address the safety issue, in particular the problem of insertional mutagenesis.7

Non-viral vectors such as Sleeping Beauty (SB), piggyBac (PB) or Tol2 derived transposons typically require co-transfection of the transposon DNA with a transposase such as an expression plasmid or mRNA. Consequently, this results in genomic integration generally into AT-rich genomic regions.8 Although the integration profile of transposon can be considered biologically safe, recently two patients developed malignant lymphoma after a treatment with CAR-T cells anti CD19 using a PB vector.9 In addition, they are easier to produce on a large scale. In addition, they harbor greater transgenic capacity and generate fewer concerns about biosafety. However, an important difference between the various technologies is the duration of CAR expression in the modified cell. For longer expression (multiple weeks), viral transduction is usually employed, while mRNA electroporation results in transient expression lasting for about one-two weeks.

Most of the molecular targets for CAR-T cells are not exclusively expressed in tumor cells. Some of these molecules can be found expressed at different levels in tumor and healthy cells. Exploring scFvs recognizing the target with different affinities in CAR molecules can potentially allow the discrimination between healthy and tumor cells. Some groups have explored such approach, especially in the preclinical setting10 and reviewed in.11 Recent results showed potential toxicity due to on-target off-tumor recognition when using high affinity CARs,12 and this also advocates in favor of reduced CAR affinities in specific contexts. Reducing the affinity of anti CD19 scFvs has also shown potential benefits, such as delayed kinetics of T cell activation (and potentially less induction of CRS) without loss of tumor elimination capacity.13 One critical aspect of this approach is that some target antigens can be modulated in different cell types according to tissue inflammation status of pharmacological exposure. As an example, CD22, an ALL-target antigen, can be modulated with drugs both in vitro and in vivo, augmenting the recognition and killing by CD22-specific CARs.14 Such contexts make it challenging to fine tune target recognition based on the balance of scFv affinity and target molecule expression density.

Mouse models have been of pivotal importance in determining the efficacy and safety of CAR-T cells therapy. According to the nature of the tumor graft and of its recipient, mouse models can be categorized into four groups: a) syngeneic; b) human xenografts; c) immunocompetent transgenic, and d) humanized transgenic mouse models.56

Syngeneic or immunocompetent allograft mouse models use CAR‐T cells, tumors, and target antigens that are all murine‐derived. The high ground of this model is that the recipient has a fully physiologic immune system, thus representing an excellent tool to observe CAR‐T cells' interaction with other cells and elements of the immune system. Additionally, syngeneic models can reveal toxicities caused by the presence of tumor-associated antigens (TAAs) in healthy murine tissues, the so-called on-target off‐tumor toxicities. Nevertheless, as mouse biology does not always accurately recapitulate human biology and many important adverse events associated with CAR-T cells therapy were not observed in the syngeneic models, namely the cytokine release syndrome (CRS). In addition, the persistence in the circulation of murine CAR‐T cells is shorter compared to human CAR‐T cells.57 An important example of the contribution of the syngeneic models is provided by the studies of David Gilham's group showing that second-generation CARs with CD28 costimulatory domains induced B cell aplasia and chronic toxicity accompanied by an increase in CD11b+Gr‐1+ myeloid‐derived suppressor cells (MDSCs).58,59 The unexpected effect on MDSCs could be detected only in a syngeneic model.

The xenograft models are based on the injection of human tumors and CAR‐T cells into an immunocompromised mouse. Currently, most xenograft mouse models of CAR‐T therapy use the NOD‐SCID‐IL2rγnull (NSG) mouse strain.60 The parental NOD-SCID strain was obtained by crossing non‐obese diabetic (NOD) mice, which have impaired innate immunity, with severe combined immunodeficiency (SCID) mice, which are deficient in the adaptive immune system. NSG mice were developed by introducing a mutation in the IL‐2 receptor γ chain gene, which resulted in the loss of signal transduction from multiple cytokines and increased immunodeficiency compared to parental NOD-SCID mice. The xenograft models were frequently used to validate proof-of-concept studies. As in this model, human CAR-T cells interact with human tumoral tissue in the absence of an intact immune system, and xenograft models are tools to test the selectivity of more complex or novel CAR constructs. One area of interest is in developing CARs that are dependent on multiple TAAs for full activation to decrease the chance of off‐tumor effects or antigen escape. Examples of the new strategies of CAR-T cells evaluated in xenograft models are a) bispecific CARs targeting both CD19 and CD20 to treat lymphomas and acute lymphoblastic leukemias,61 b) masked anti‐EGFR CARs in which the scFVs were masked by a peptide with a linker that is cleavable by proteases expressed by tumor cells but not healthy tissues62 c) Switch-mediated activation and retargeting of CAR-T cells that targeted a specific peptide neo‐epitope introduced on a TAA‐binding antibody.63 The xenograft models were also used to understand the mechanisms of CAR-T cells exhaustion. In two second-generation human CAR‐T cells, exhaustion was prevented with 4‐1BB costimulation and exacerbated by CD28 costimulation.64

Patient‐derived xenograft (PDX) models are a specific subtype of xenograft model generated by implanting a primary tumor biopsy instead of injecting tumor cell lines in an immunocompromised host. The advantages of PDX models are to better capture tumor heterogeneity and to avoid exposure to artificial environments used in the in vitro experiments. In a hepatocarcinoma PDX model, Jiang et al. showed that it accurately predicted patient response and could be useful in determining the course of treatment, such as a combination of checkpoint blockade and CAR‐T therapies.65

Immunocompetent transgenic mice have been the less frequently used model employed in CAR‐T studies. Their utility relies on their potential for the evaluation of on‐target off‐tumor effects. These transgenic mice are engineered to lack the expression of a murine TAA (knockout) and to express a human TAA (knockin) in specific tissues and at a certain level. Pegram et al. reported that treatment with CD19-specific, CAR-T cells that were further modified to constitutively secrete IL-12 were able to eradicate established disease in the absence of prior conditioning.38 Moreover, the tumor elimination required both CD4(+) and CD8(+) T-cell subsets, autocrine IL-12 stimulation, and subsequent IFNγ secretion by the CAR(+) T cells. The authors proposed that adoptive therapy using CAR-targeted T cells modified to secrete IL-12 would obviate or reduce the need for conditioning regimens.38

Humanized transgenic mice are immunocompromised mice implanted with human immune cells in addition to human tumor and CAR‐T cells. The main advantage of this model is the fact that mice are tolerant to human cells yet have aspects of a human immune system. Several groups have used NSG mice transplanted with human CD34+ cells to look for CAR‐T toxicity against hematopoietic stem cells (HSPCs). In the humanized immune system (HIS) model, sublethally irradiated newborn BALB/c Rag2‐/‐γc ‐/‐ mice are injected with human CD34+ cells. The use of newborn immunodeficient mice was important because they support better T cell development, as they do not display the thymic involution and phagocytic activity against human immune cell engraftment seen in adult mice. Hombach et al. used a HIS model to test if anti‐CD30 CAR‐T cells could eradicate lymphoma without lasting B cell aplasia.66 CD30 is also expressed by HSPCs during activation but in lower levels than in lymphoma cells. This difference in expression level, as well as the expression of granzyme B inhibitor by HSPCs, may have contributed to the protection of the engrafted HSPCs from CAR‐T‐mediated toxicity. A major concern in humanized models is the fact that the rates of lymphoid and myeloid cells produced by human CD34+ cells, as well as the profile of the circulating cytokines, are distinct from those found in humans, and therefore do not reflect the human immune cell development.

From the above, it is clear that there is no single mouse model capable of encompassing all aspects involved in the efficacy and safety for CAR‐T therapy, but this is a moving field and there has been significant improvements in the humanization of murine immune systems. At present, we still need the association of multiple animal models that will provide complementary information of different aspects of CAR‐T treatment. Nevertheless, it is evident that animal models are essential tools to predict CAR‐T safety and efficacy in the clinic.

The science of CAR-T cell development and production has evolved at a fast pace in recent years. Although many biological aspects of CAR expressing cells are still being uncovered, minimal parameters for CAR-T cell function characterization are now widely accepted by the scientific community. We approached herein most of the important aspects for characterizing a functional and high-quality CAR-based cell product. The biological assays for evaluating such parameters are also impacted by new technological developments, and regular updates for the validation tests are expected.

It is important to note that new technologies are being developed and disruptive science is supporting CAR-T cell developments, so new aspects of CAR-T cell biology, such as surrogate markers for in vivo antitumor effectivity and exhaustion/dysfunction are likely to be incorporated as evaluation criteria in the near future.