Evaluation of chimeric antigen receptor of humanized rabbit‐derived T cell receptor‐like antibody

Abstract

T‐cell receptor (TCR)‐like Abs that specifically recognize antigenic peptides presented on MHC molecules have been developed for next‐generation cancer immunotherapy. Recently, we reported a rapid and efficient method to generate TCR‐like Abs using a rabbit system. We humanized previously generated rabbit‐derived TCR‐like Abs reacting Epstein–Barr virus peptide (BRLF1p, TYPVLEEMF) in the context of HLA‐A24 molecules, produced chimeric antigen receptor (CAR)‐T cells, and evaluated their antitumor effects using in vitro and in vivo tumor models. Humanization of the rabbit‐derived TCR‐like Abs using the complementarity‐determining region grafting technology maintained their specificity and affinity. We prepared a second‐generation CAR using single‐chain variable fragment of the humanized TCR‐like Abs and then transduced them into human T cells. The CAR‐T cells specifically recognized BRLF1p/MHC molecules and lysed the target cells in an antigen‐specific manner in vitro. They also demonstrated antitumor activity in a mouse xenograft model. We report the generation of CAR‐T cells using humanized rabbit‐derived TCR‐like Abs. Together with our established and efficient generation procedure for TCR‐like Abs using rabbits, our platform for the clinical application of humanized rabbit‐derived TCR‐like Abs to CAR‐T cells will help improve next‐generation cancer immunotherapy.

The humanized TCR‐like antibody had a similar specificity and affinity for BRLF1/A24 as the original rabbit antibody. The TCR‐like CAR‐T cells generated from the humanized rabbit‐derived TCR‐like antibody demonstrated efficient cytotoxicity in vitro and in vivo.

Abbreviations

ADGRV1

adhesion G‐protein coupled receptor V1

AFP

alpha‐fetoprotein

BiTE

bispecific T‐cell engager

BLAST

Basic Local Alignment Search Tool

BRLF1p

BRLF1 peptide

CAR

chimeric antigen receptor

DPEP

dipeptidase

EBV

Epstein–Barr virus

FCM

flow cytometry

Fv

variable fragment

hr

humanized rabbit‐derived

IFN

interferon

ISAAC

immunospot array assay on a chip

SCT

single‐chain trimer

TCR

T‐cell receptor

1. INTRODUCTION

In recent years, cancer immunotherapy has developed rapidly and is recognized as the fourth treatment for cancer following surgery, chemotherapy, and radiation therapy. Among cancer immunotherapies, CAR‐T cells and BiTE have been applied clinically, and are markedly effective in the treatment of B cell hematologic malignancies. ¹ , ² Most CAR‐T cell and BiTE therapies target cell surface antigens, such as CD19, but not intracellular proteins that are expressed in cancer cells, even though approximately 95% of tumor antigens are intracellular proteins. ³ Thus, a strategy to target intracellular tumor antigens in order to adapt CAR‐T and BiTE therapies for a wide range of cancers ⁴ is necessary.

Cytoplasmic proteins are degraded by the proteasome and the produced peptides are complexed with MHC molecules, which are transported to the cell surface through the MHC class‐I pathway. The MHC/peptide complex is then recognized by the TCR on T cells. ⁵ Recently, a group of Abs called TCR‐like Abs, which recognize MHC/peptide complexes in the same manner as TCRs, was developed. ⁴ , ⁶ Thus, TCR‐like Abs that recognize intracellular antigens as p/MHC in the same manner as TCR have been investigated for targeting intracellular tumor antigens. ⁷ , ⁸ T‐cell receptor‐like Abs have been conventionally generated using hybridoma or phage display technology. ⁷ , ⁸ It has been difficult to efficiently produce TCR‐like Abs with high affinity and high specificity. In addition, the production of functional recombinant p/MHC for immunization and the screening of the obtained TCR‐like Ab clones are complicated. Therefore, the application of TCR‐like Abs in cancer immunotherapy is limited. ⁷ , ⁸ , ⁹ , ¹⁰

Recently, we established two techniques for efficiently generating TCR‐like Abs. The first is the generation of large amounts of functional soluble p/MHC in a SCT format. ¹¹ , ¹² The second is the efficient screening of TCR‐like Abs using an Ab‐secreting cell screening system named ISAAC. ¹³ , ¹⁴ , ¹⁵ By combining these two technologies, we generated recombinant TCR‐like mAbs from SCT‐immunized rabbit B cells, and confirmed their cytotoxicity in vitro in BiTE format. ¹⁴

The BRLF1 protein is expressed from a group of viral genes (early genes) necessary for viral genome replication, ¹⁶ and associated with nasopharyngeal carcinoma. ¹⁷ , ¹⁸ We previously analyzed EBV derived peptide‐specific TCRs in the context of HLA‐A24, in which BRLF1p/HLA‐A24‐specific TCRs were identified. ¹⁹ Thus, we selected BRLF1p/HLA‐A24 as a target for the generation of TCR‐like Abs and generated rabbit‐derived TCR‐like Abs against BRLF1p/HLA‐A24. ¹⁴ However, as these TCR‐like Abs originate from rabbits, they cannot be directly clinically applied due to their antigenicity for humans. In the present study, we humanized previously generated rabbit‐derived TCR‐like Abs against BRLF1p/HLA‐A24 as a model, applied them to produce CAR‐T cells, and evaluated their function in vitro and in vivo.

2. MATERIALS AND METHODS

T‐cell receptor‐like CAR‐T cells based on humanized rabbit‐derived TCR‐like Abs were generated using PBMCs from healthy donors and used for in vitro and in vivo assays. Informed consent was received from all volunteers in accordance with a protocol approved by the ethics review board of the University of Toyama (approval no. R2019177). In addition, we validated the cytotoxic activity of TCR‐like CAR‐T cells in an experimental tumor model using NSG mice. K562‐A24 cells transduced with luciferase gene and a mini gene encoding BRLF1 were implanted intravenously into NSG mice (n = 6/group). Two and 7 days later, CAR‐T cells were injected intravenously. The growth of tumor cells was measured using an in vivo imaging system (IVIS Lumina II; Perkin Elmer). The experiments were approved by the Committee on Animal Experiments at the University of Toyama (approval no. A2019MED‐38).

Further details on the materials and methods used for our study, including Abs, antigens, peptides, measurement of Ab affinity, cell lines, FCM analysis of humanized TCR‐like Abs, construction of TCR‐like CAR, retrovirus production, generation of TCR‐like CAR‐T cells, FCM analysis of TCR‐like CAR‐T cells, cytotoxicity assay in vitro, T cell activation and cytokine production assay, in vivo tumor model and bioluminescent imaging, and identification of candidate potential mimotopes are available in the Appendix S1.

3. RESULTS

3.1. Humanization of rabbit‐derived BRLF1p/HLA‐A24‐specific TCR‐like Abs

In the previous study, we generated four TCR‐like mAbs (#19‐03, #21‐28, #21‐34, and #23‐35) that recognized BRLF1p/HLA‐A24 specifically. ¹³ In this study, we humanized rabbit‐derived TCR‐like mAbs using complementarity‐determining region grafting technology. ²⁰ We then assessed the specificity of the hrTCR‐like Abs by analyzing their binding to a T2 cell line expressing HLA‐A*24:02 (T2‐A24) pulsed with different peptides using FCM. All hrTCR‐like Abs (#h21‐28, #h21‐34, #h19‐03, and #h23‐35) specifically bound to T2‐A24 cells in a BRLF1p‐dependent manner (Figure 1A). Among the four hrTCR‐like Abs, #h21‐34 showed weak binding to BRLF1p‐pulsed T2‐A24 cells.

FIGURE 1.

Functional characterization of humanized BRLF1/A24 T‐cell receptor (TCR)‐like Abs and generation of humanized rabbit‐derived (hr)TCR‐like chimeric antigen receptor (CAR)‐T cells. (A) Binding of hrTCR‐like Abs to T2‐A24 cells pulsed with indicated peptides. The binding was analyzed using flow cytometry (FCM). Representative data of three independent experiments are shown. (B) Affinity of BRLF1‐dtSCT‐specific hrTCR‐like Abs to BRLF1‐dtSCT (KD) was determined using Scatchard plots. The x‐axis indicates concentration (nM) of dtSCT and the y‐axis indicates the absorbance at 450 nm. The KD value is the average of three independent experiments. (C) Schematic diagram of the TCR‐like single‐chain variable fragment (scFv)‐CAR constructs, containing the CD8 hinge region (CD8 hinge), CD28 transmembrane and intracellular signaling domains (CD28cyto), and the CD3ζ signaling domain (CD3ζcyto). The internal ribosomal entry site (IRES) and EGFP (GFP) were also linked in the expression vectors. The scFv of CAR was constructed by linking the variable regions of humanized BRLF1/A24 TCR‐like Abs in order with V_H to V_L (HL) or V_L to V_H (LH) sequences. (D) Binding of indicated recombinant peptide/MHC (dtSCT) or peptide to cell surface TCR‐like CAR determined by FCM. As control CAR‐T cells, AGIA‐specific CAR was expressed on T cells. Each result is representative of at least two independent experiments. AFP, alpha‐fetoprotein; PE, phycoerythrin

Next, we examined the Ab binding affinity (KD) of TCR‐like Abs for BRLF1p/HLA‐A24 (an SCT with a disulfide trap format, BRLF1‐dtSCT) ¹¹ , ¹⁴ by ELISA and Scatchard plot analyses as previously described. ¹⁴ , ¹⁵ , ²¹ , ²² The affinity of the four hrTCR‐like Abs ranged from 5.1 × 10⁻⁷ to 6.3 × 10⁻⁹ M (Figure 1B). Compared with the affinity of the parental rabbit TCR‐like Abs (1.4 × 10⁻⁹ M for #21‐28, 4.4 × 10⁻⁸ M for #21‐34, 1.4 × 10⁻⁸ M for #19‐03, and 2.6 × 10⁻⁸ M for #23‐35), ¹⁴ #h21‐28, #h19‐03, and #h23‐35 showed corresponding affinity after humanization, whereas the affinity of #h21‐34 decreased. This suggests that, except for #h23‐34, three hrTCR‐like Abs (#h23‐35, #h21‐28, and #h19‐03) maintained their specificity and affinity after humanization. Thus, we selected these three hrTCR‐like Abs for further analysis.

3.2. Generation of CAR‐transduced T cells using humanized BRLF‐1/A24 TCR‐like Abs

In order to induce cell death of target cells using hrTCR‐like Abs, we produced hrTCR‐like CAR constructs from the hrTCR‐like Abs (Figure 1C). Several reports suggested that the order of heavy chain and light chain Fv in CAR affects the functional strength in CAR‐T cells. ²³ , ²⁴ Thus, we designed CARs of two different orders: V_H to V_L (HL) and V_L to V_H (LH) to compare their function (Figure 1C). We retrovirally transduced these CARs into human PBMCs. Flow cytometry revealed that the transduction efficiency of CAR to T cells was approximately 20%–60% (Figure S1A). Next, we confirmed the specificity of the CARs using FCM with dtSCT. All hrTCR‐like CAR‐T cells (#h21‐28 HL/LH CAR‐T, #h19‐03 HL/LH CAR‐T, and #h23‐35 HL/LH CAR‐T) specifically bound dtSCT in an antigen‐dependent manner (Figures 1D and S1B).

3.3. Cytotoxicity of BRLF1/A24 TCR‐like CAR‐T cells in vitro

Next, we analyzed the cytotoxicity of hrTCR‐like CAR‐T cells against target cells in vitro. We cocultured peptide pulsed T2‐A24 cells with hrTCR‐like CAR‐T cells. All hrTCR‐like CAR‐T cells showed cytotoxicity against T2‐A24 cells in a BRLF1p‐dependent manner (Figures 2A and S1C). We also cocultured peptide mini gene‐expressing K562‐A24 cells with CAR‐T cells. All hrTCR‐like CAR‐T cells were cytotoxic against BRLF1 mini gene‐expressing K562‐A24 cells (Figures 2A and S1C). However, #h19‐03 HL/LH CAR‐T cells and #h23‐35 HL/LH CAR‐T cells also showed cytotoxicity against AFP mini gene‐expressing K562‐A24 cells (Figure S1C). Thus, #h21‐28 HL/LH CAR‐T cells were cytotoxic in a peptide‐dependent manner. We selected #h21‐28 HL/LH CAR‐T cells for further analysis.

FIGURE 2.

Cytotoxic effects of humanized rabbit‐derived T‐cell receptor (hrTCR)‐like chimeric antigen receptor (CAR)‐T cells. (A) In vitro cytotoxic activity of the hrTCR‐like CAR‐T cells against indicated target cells at the indicated effector/target (E/T) ratio. AGIA‐specific CAR‐expressing T cells (Control CAR‐T) and nontransduced PBMCs were used as negative controls. Data are representative of three independent experiments and shown as mean ± SD. (B) Interferon‐γ (IFN‐γ) secretion by hrTCR‐like CAR‐T cells stimulated with indicated target cells for 24 h. IFN‐γ in the culture supernatants was measured by ELISA in triplicate. Data are shown as mean ± SD of triplicate experiments. (C) CD69 upregulation determined using flow cytometry (FCM). hrTCR‐like CAR‐T cells or control CAR‐T were cultured with indicated target cells for 24 h. CD69 expression was analyzed by FCM. Results are representative of at least two independent experiments

3.4. Activation of TCR‐like CAR‐T cells and cytokine release against BRLF1/A24 expressing cells

Second, we examined the antigen‐specific activation of hrTCR‐like CAR‐T cells. To this end, #h21‐28 HL or #h21‐28 LH CAR‐T cells were cocultured with peptide‐pulsed T2‐A24 cells and mini gene‐expressing K562‐A24 cells (K562‐A24/BRLF1 or K562‐A24/AFP), and then we evaluated CAR‐T cell activation based on secretion of IFN‐γ. ²⁵ Both #h21‐28 HL and #h21‐28 LH CAR‐T cells cocultured with target cells secreted IFN‐γ in an antigen‐dependent manner (Figure 2B).

We further analyzed T cell activation by the expression of CD69 ²⁶ on the cell surface of CAR‐T cells. Upregulation of CD69 on CAR‐T cells was observed in an antigen‐dependent manner when #h21‐28 LH CAR‐T cells were incubated with target cells (Figure 2C), whereas CD69 expression was upregulated on #h21‐28 HL CAR‐T cells by coculturing with K562‐A24/AFP cells. Therefore, #h21‐28 LH TCR‐like CAR‐T cells were activated by BRLF1/A24‐expressing target cells.

3.5. Cytotoxicity of #h21‐28 LH CAR‐T cells in vivo

Next, we evaluated whether #h21‐28 LH CAR‐T cells showed cytotoxicity toward K562‐A24/BRLF1 cells in vivo. NSG mice were intravenously injected with K562‐A24/BRLF1 cells (day 0). After 2 or 7 days, #h21‐28 LH TCR‐like CAR‐T cells or control CAR‐T cells were intravenously injected into mice through the tail vein (Figure 3A). On days 2, 7, and 10 we analyzed tumor growth in mice by monitoring with luciferase imaging. ²⁷ The #h21‐28 CAR‐T cells suppressed tumor growth compared with control CAR‐T cells and the tumor‐only group (Figures 3B,C and S2). Therefore, the hrBRLF1/A24 TCR‐like CAR‐T cells showed antitumor effects on BRLF1/A24‐expressing K562 cells in vivo.

FIGURE 3.

In vivo cytotoxic assay using humanized rabbit‐derived T‐cell receptor (hrTCR)‐like chimeric antigen receptor (CAR)‐T cells. (A) Schematic representation of hrTCR‐like CAR‐T cytotoxicity assay in vivo. Luciferase‐expressing K562‐A24/BRLF1 cells were injected intravenously into NSG mice on day 0. On days 2 and 7, #h21‐28 CAR‐T cells (36.8% to 77.4% CAR⁺) or control CAR‐T cells (20.7% to 69.4% CAR⁺) were injected intravenously into mice. (B) K562‐A24/BRLF1 tumor expansion in NSG mice was determined based on mouse whole‐body bioluminescence. Tumor growth was evaluated in three groups: #h21‐28 LH CAR‐T, control CAR‐T, and tumor only (n = 6 in each group). Kinetics of tumor growth are indicated. Data are presented as the average bioluminescence of six mice ± SE and p‐values were calculated using one‐way ANOVA test (*p < 0.05, ** p < 0.01). (C) Bioluminescent images of mice on days 2, 7, and 10 show that tumor expansion was suppressed in mice treated with #h21‐28 LH CAR‐T cells

3.6. Validation of cross‐reactivity of hrTCR‐like CAR‐T cells

We investigated the cause of the antigen‐independent cytotoxic activity observed in hrTCR‐like CAR‐T cells. First, we analyzed the binding of hrTCR‐like Abs to K562, K562‐A24, K562‐A24/BRLF1, or K562‐A24/AFP cells by FCM. Binding of #h21‐28, #h19‐03, and #h23‐35 hrTCR‐like Abs to K562‐A24/BRLF1 cells but not to K562, K562‐A24, or K562‐A24/AFP cells was observed (Figure S3A). CD69 upregulation on all CAR‐T cells was observed by coculture with K562‐A24/BRLF1 cells, but not observed by coculture with K562 cells (Figure S3B). CD69 upregulation on #h21‐28 LH CAR‐T cells was not induced by coculture with either K562‐A24 or K562‐A24/AFP cells. However, CD69 upregulation on #h21‐28 HL, #h19‐03 HL, #h19‐03LH, #h23‐35 HL, and #h23‐35 LH CAR‐T cells was induced when cocultured with either K562‐A24 or K562‐A24/AFP cells (Figure S3B). CD69 upregulation on CAR‐T cells was only observed when cocultured with BRLF1p‐pulsed T2‐A24 cells, but not with irrelevant peptide‐pulsed T2‐A24 cells (Figure S3C). These results show that TCR‐like CAR‐T cells did not respond to HLA‐A24 itself, but to the endogenous antigen of K562 cells.

3.7. Exploring potential cross‐reactivity of hrTCR‐like CAR‐T cells

We attempted to identify the potential cross‐reactive antigen for hrTCR‐like CAR‐T cells. To this end, we first determined the key amino acid residues in the epitope recognized by hrTCR‐like Abs using alanine‐substituted peptides. Amino acids at positions 2 and 9 were not altered because these positions are the canonical anchor residues. ²⁸ Flow cytometry analysis showed that L5 was indispensable for the binding of hrTCR‐like Abs. Remaining residues showed partial roles in the binding (Figure 4A).

FIGURE 4.

Epitope mapping of individual humanized rabbit‐derived T‐cell receptor (hrTCR)‐like Abs and discovery of cross‐reactive peptides. (A) BRLF1 peptides with alanine substitutions at positions 1, 3, 4, 5, 6, 7, and 8 were synthesized and pulsed onto T2‐A24 cells, and the binding of individual Ab was examined by flow cytometry (FCM). (B) Recognition motif of BRLF1p/A24 TCR‐like Abs. X indicates any amino acid. (C) Candidate cross‐reactive peptides of BRLF1p/A24 TCR‐like Abs. Mismatched amino acid residues are shown in red. (D) The indicated BRLF1‐like peptides were pulsed onto T2‐A24 cells and bindings of hrTCR‐like Abs were examined by FCM. (E) #h21‐28 LH CAR‐T or control CAR T cells were cocultured with T2‐A24 cells pulsed with BRLF1‐like peptides to analyze the cytotoxic activity of CAR‐T cells. (F) Interferon‐γ (IFN‐γ) secretion by #h21‐28 LH CAR‐T cells stimulated with BRLF1‐like peptide‐pulsed T2‐A24 cells. All results are representative of at least two independent experiments, and all data are shown as mean ± SD of triplicate experiments

We next undertook BLAST homology and NetMHC4.0 searches of human proteins based on the results of epitope recognized by #h21‐28 using alanine substituted peptides. We used the following criteria to determine the candidates: first, the Y2, L5, and L9 or F9 residues had to be 100% identical, and second, the remaining six amino acid residues had to show more than 50% identity because Y2 and L9 or F9 were canonical anchor amino acid residues, L5 was indispensable for binding to #h21‐28, and the remaining amino acids were dispensable but had partial roles in the binding (Figure 4A,B). As a result, DPEP2, ADGRV1, and DPEP3 derived peptides were identified (Figure 4C). The FCM analysis showed that #h21‐28 TCR‐like Ab did not bind to DPEP2, ADGRV1, or DPEP3 derived peptide‐pulsed T2‐A24 cells, whereas #h19‐03 Ab substantially bound to DPEP3 peptide‐pulsed T2‐A24 cells and #h23‐35 Abs weakly bound to DPEP2 peptide‐pulsed T2‐A24 cells (Figure 4D).

Finally, we tested the cytotoxicity of #h21‐28 LH CAR‐T cells against DPEP2, ADGRV1, or DPEP3 derived peptide‐pulsed T2‐A24 cells. The #h21‐28 LH CAR‐T cells showed no cytotoxicity against DPEP2, ADGRV1, or DPEP3 derived peptide‐pulsed T2‐A24 cells (Figure 4E). Additionally, #h21‐28 LH CAR‐T cells cocultured with DPEP2, ADGRV1, or DPEP3 derived peptide‐pulsed T2‐A24 cells did not secrete IFN‐γ (Figure 4F). These results suggest that the risk of potential cross‐reactivity of #h21‐28 LH CAR‐T cells is very low.

We also tested cross‐reactivity to other HLA types because virus‐specific T cells are cross‐reactive to many other HLA types. ²⁹ Because HLA‐A*24:02 is the most frequent, followed by HLA‐A*02:01 and A*11:01 in the Japanese population, ³⁰ we generated K562 cells that stably express HLA‐A*02:01 or HLA‐A*11:01 (Figure S4A, K562‐A02 or K562‐A11, respectively) and analyzed the reactivity of #h21‐28 TCR‐like Ab and #h21‐28 LH CAR‐T cells to them. The FCM analysis showed that #h21‐28 TCR‐like Ab did not bind to K562‐A02 or K562‐A11 cells (Figure S4B). Also, CD69 expression was not upregulated on #h21‐28 TCR‐like CAR‐T cells when cocultured with K562‐A02 or K562‐A11 cells (Figure S4C). These results suggest that the risk of cross‐reaction is low with #h21‐28 TCR‐like CAR‐T cells, at least for HLA‐A*02:01 and HLA‐A*11:01 of highly frequent HLA types in the Japanese population.

3.8. #h21‐28 CAR‐T cell response to EBV‐infected lymphoblastoid cells endogenously expressing EBV antigen and HLA‐A*24:02

Finally, we examined whether #h21‐28 CAR‐T cells could respond to EBV‐transformed lymphoblastoid cells endogenously expressing the EBV antigen and HLA‐A*24:02 (JTK‐LCL cells). We first analyzed the binding of #21‐28 TCR‐like Abs to JTK‐LCL cells by FCM. The rabbit #21‐28 TCR‐like Ab bound to JTK‐LCL cells, but humanized #21‐28 TCR‐like Ab did not (Figure S5A). We then analyzed CD69 expression and IFN‐γ production by coculture of #h21‐28 LH CAR‐T cells with JTK‐LCL cells. When #h21‐28 LH CAR‐T cells were cocultured with JTK‐LCL cells, CD69 expression was not induced (Figure S5B). Furthermore, #h21‐28 LH CAR‐T cells did not produce IFN‐γ when cocultured with JTK‐LCL cells (Figure S5C). Thus, we could not observe any response of #h21‐28 LH CAR‐T cells to EBV‐infected lymphoblastoid cell line cells that endogenously express EBV antigen and HLA‐A*24:02.

4. DISCUSSION

Recently we developed a system to efficiently generate rabbit‐derived TCR‐like Abs. ¹⁴ In this study, we humanized rabbit‐derived TCR‐like Abs that recognize BRLF1/A24 as a model for hrTCR like CAR‐T cells. The humanized TCR‐like Ab had a similar specificity and affinity for BRLF1/A24 as the original rabbit Ab. Moreover, the hrTCR‐like CAR‐T cells generated from the hrTCR‐like Ab showed efficient cytotoxicity in vitro and in vivo. In addition, there have been no reports of the generation and evaluation of CAR‐T cells from humanized rabbit mAbs. Our study provides new applications of rabbit Abs using Ab engineering technology.

One of the major generation methods for TCR‐like Abs is the selection of antigen‐specific Ab secreting B cells using the hybridoma method, ⁷ , ⁸ FACS‐based method, ³¹ or our ISAAC system. ¹⁴ We previously reported that our method is the most efficient among these. ¹⁴ Another major method is phage‐display libraries technology. ⁷ , ⁸ This method is highly selective toward the desired antigen and the objective Abs are achieved within a relatively short time. In the case of using human‐derived large naïve libraries, humanization of the Ab is not necessary. However, Abs of naïve libraries are not affinity‐matured in vivo because the antigen cannot be immunized to humans, and thus have moderate to low affinity (∼5 × 10⁻⁸ to 3 × 10⁻⁷ M). ⁷ , ⁸ In contrast, our TCR‐like Ab had high affinity ¹⁴ (Figure 1B) because of immunization to rabbit. In the case of using mouse‐derived immunized libraries, the Ab affinity is often higher than that of naïve libraries, ⁷ , ⁸ but humanization is necessary. In comparison with our method, we used rabbit Abs that generally show high affinity compared with conventional mouse Abs. ³² , ³³ Thus, it is expected that Abs with a high affinity can be obtained compared with immunized mouse‐derived libraries.

We evaluated the differences in functional consequences of HL and LH structures of the single‐chain variable fragment region of CARs, and found a difference in specificity. In both CAR‐T cells of #h21‐28 HL and #h21‐28 LH, binding to the antigen and production of IFN‐γ were observed in an antigen‐dependent manner (Figure 2B). However, #h21‐28 HL CAR‐T cells upregulated CD69 expression in an antigen‐independent manner when cocultured with K562‐A24/AFP cells (Figure 2C). This suggests that #h21‐28 HL CAR‐T cells cross‐react with peptides, causing off‐target effects, and indicates that the Fv order of CAR‐T cells affects the cross‐reactivity. Further analysis of the effects of the order of other Fv on the specificity of CAR‐T cells is required.

Monoclonal Abs with a high affinity are being investigated for diagnostic and therapeutic applications. ³² , ³⁴ However, in the case of CAR‐T, the antitumor effects are higher for CAR with a low affinity (1.4 × 10⁻⁸ M) than for CAR with a strong affinity (3.3 × 10⁻¹⁰ M). ³⁵ The affinity of humanized rabbit #h21‐28 was 7.0 × 10⁻⁹ M (Figure 1B). Thus, the affinity of humanized rabbit #h21‐28 is moderate.

T‐cell receptor‐like Abs have potential off‐target effects by cross‐reacting with sequence‐related peptides. ³⁶ The off‐target effects in normal cells are often a problem with TCR‐like Abs. ³⁷ , ³⁸ Recently, Kurosawa et al. reported that the #25‐8 TCR‐like Ab that recognized WT1p/HLA‐A02 was antigen‐specific at the protein level, but #25‐8 CAR‐T cells showed off‐target effects at the cellular level. This potentially dangerous cross‐reactivity was confirmed through analysis using #25‐8 CAR‐T cells targeting WT1‐negative HeLa/A02 cells. ³⁹ Similar to their report, there was no binding to K562 and K562‐A24 cells with the hrTCR‐like Abs (Figure S3A). However, #h21‐28 HL, #h19‐03 HL/LH, and #h23‐35 HL/LH CAR‐T cells responded to K562 cells in an HLA‐A24 dependent manner (Figure S3B). As K562 cells are free of the EBV‐associated nuclear antigen, ⁴⁰ the antigen‐independent CAR‐T cell response to K562‐A24 could be due to a cross‐reaction for endogenous antigen of K562 cells. The alanine substitution experiment and BLAST search indicated the potential cross‐reactive peptides derived from DPEP2 and DPEP3.

This study has the following limitation. First, although BRFL1p/A24 was chosen as the model in this study, we could not show any effect of #h21‐28 LH CAR‐T cells on EBV‐infected lymphoblastoid cell line cells that endogenously express EBV antigen and HLA‐A*24:02 (JTK‐LCL cells). The rabbit #21‐28 TCR‐like Ab bound to JTK‐LCL cells, suggesting that BRLF1p/HLA‐A24 is expressed on the cell surface in JTK‐LCL cells. However, the humanized #21‐28 TCR‐like Ab did not bind to JTK‐LCL cells, and there was no obvious response of #h21‐28 LH CAR‐T to JTK‐LCL cells. This could be due to the decrease in affinity of the TCR‐like Ab by humanization. To overcome this, we need to obtain and validate more TCR‐like Abs, or develop a humanization method that does not reduce the affinity of the Abs. Furthermore, we could not validate EBV‐infected tumor cell lines because we could not find a tumor cell line endogenously expressing BRLF1 protein. At this time, it is not possible to directly advance the hrTCR‐like CAR‐T cells generated in this study to clinical applications. Therefore, we consider that the generation of hrTCR‐like CAR‐T cells targeting antigens that can be used in actual clinical practice as our future work. Second, we only generated and validated TCR‐like Abs that recognized BRLF1p/HLA‐A*24:02. We have not investigated TCR‐like Abs that recognize other peptides presented on HLA‐A*24:02 or different HLA alleles. We next need to evaluate whether the same technique can be applied to generate TCR‐like Abs that recognize other peptides on different HLA alleles or tumor antigens and validate their functions in CAR‐T cells.

In conclusion, we generated rabbit‐derived TCR‐like Abs and humanized them. The hrTCR‐like CAR‐T cells showed effective cytotoxicity. Our previous ¹² , ¹⁴ and present studies determined a series of processes for the efficient generation of TCR‐like Abs using rabbits, their successful humanization, and their effects in CAR‐T cells in vitro and in vivo, which will be useful for generating a new cancer immunotherapy strategy targeting intracellular cancer antigens.

AUTHOR CONTRIBUTIONS

All authors designed the experiments. T.N., E.K., Y.H., and T.O. performed experiments and analysis. T.N., E.K., H.H., A.M., T.O., and H.K. interpretated the data. T.N.,T.O., and H.K. wrote the manuscript.

DISCLOSURE

The authors report no conflict of interest. Yoshihiro Hayakawa is an Editorial Board Member of Cancer Science.

APPROVAL OF THE RESEARCH PROTOCOL BY AN INSTITUTIONAL REVIEW BOARD AND INFORMED CONSENT

Informed consent was received from all volunteers in accordance with a protocol approved by the ethics review board of the University of Toyama (approval no. R2019177).

ANIMAL STUDIES

The in vivo experiments were approved by the Committee on Animal Experiments at the University of Toyama (approval no. A2019MED‐38).

Supporting information

Figure S1–S6

Appendix S1

Nakamura T, Kobayashi E, Hamana H, et al. Evaluation of chimeric antigen receptor of humanized rabbit‐derived T cell receptor‐like antibody. Cancer Sci. 2022;113:3321‐3329. doi: 10.1111/cas.15478