A humanized TCR retaining authentic specificity and affinity conferred potent anti‐tumour cytotoxicity

Summary

The affinity of T‐cell receptor (TCR) determines the efficacy of TCR‐based immunotherapy. By using human leucocyte antigen (HLA)‐A*02 transgenic mice, a TCR was generated previously specific for human tumour testis antigen peptide MAGE‐A3_112–120 (KVAELVHFL) HLA‐A*02 complex. We developed an approach to humanize the murine TCR by replacing the mouse framework with sequences of folding optimized human TCR variable domains for retaining binding affinity. The resultant humanized TCR exhibited higher affinity and conferred better anti‐tumour activity than its parent murine MAGE‐A3 TCR (SRm1). In addition, the affinity of humanized TCR was enhanced further to achieve improved T‐cell activation. Our studies demonstrated that the human TCR variable domain frameworks could provide support for complementarity‐determining regions from a murine TCR, and retain the original binding activity. It could be used as a generic approach of TCR humanization.

Abbreviations

CAR

chimeric antigen receptor

CDRs

complementarity‐determining regions

ELISpot

enzyme‐linked immunosorbent spot

FBS

fetal bovine serum

HAT

high‐affinity TCR

HLA

human leucocyte antigen

IBs

inclusion bodies

IL‐2

interleukin‐2

k_a

association rate constant

k_d

dissociation rate constant

K_D

equilibrium dissociation constant

LDH

lactate dehydrogenase

PBMCs

peripheral blood mononuclear cells

pHLA

peptide HLA complex

SA

streptavidin

scFv

single‐chain antibody

SPR

surface plasmon resonance

sTv

single‐chain TCR variable‐fragment

TCR

T‐cell receptor

Introduction

T‐cell‐based immunotherapy has provided a great hope in combating cancer.1, 2 The anti‐tumour response is driven by the T‐cell receptor (TCR) that can specifically recognize antigenic peptide human leucocyte antigen (HLA) complex (pHLA) presented on tumour cells.3, 4 Antigen‐specific T‐cell therapy has demonstrated good efficacy on melanoma.5, 6 Recently, it has been shown that high‐affinity TCRs (HATs) can give rise to high‐avidity T‐cells, which demonstrate improved anti‐tumour responses.7, 8, 9, 10 T‐cells expressing NY‐ESO‐1‐specific HATs provided not only safe but also effective anti‐tumour response in clinical trials of melanoma, myeloma and synovial cell sarcoma.2, 11, 12, 13 These data suggest that the efficacy of T‐cell therapy depends on the TCR affinity in a defined range.

However, HATs specific for human antigen are not readily available directly from human immune repertoires, as T‐cells with HATs for self‐epitopes have been depleted during the thymic selection to remove auto‐reactivity.14 On the other hand, CD8⁺ T‐cells expressing HATs will be activated in a CD8‐independent manner. As mouse CD8 is unable to interact efficiently with the human HLA α3 domain, relative high‐affinity murine TCRs specific for human tumour antigen may be isolated from HLA transgenic mice, and developed for redirecting anti‐tumour T‐cell cytotoxicity.15, 16 Murine TCRs specific for cancer antigens should have good potential for application in the immunotherapy of cancer.

But the prospective immunogenic murine HAT may activate host immune response, and result in potency loss of the T‐cells transferred with the murine TCRs. Treatments with T‐cells transferred with murine TCRs specific for tumour‐associated antigen lead to patient‐generated antibodies against the TCRs,17, 18 and the antibodies inhibited the function of murine TCR‐transfected T‐cells. In addition, previous studies have demonstrated that immune responses against non‐human sequence can ultimately limit the effective function by eliminating adoptive chimeric antigen receptor (CAR) transferred T‐cell (CAR‐T), which display the murine original single‐chain antibody (scFv).19, 20 Therefore, immune‐mediated rejection is a major challenge for solid tumour immunotherapy, as it requires the persistence of transferred T‐cells. On the other hand, it is well known that the generation of humanized antibody has overcome the problem of murine antibody immunogenicity, made great achievement in clinic and, as a result, many humanized antibodies are now available on the market.21, 22 The humanized antibody was constructed by complementarity‐determining regions (CDRs) grafting, in which only the CDRs as the smallest antigen specificity units of the mouse antibody were grafted on the most similar human v‐gene frameworks.23 Although CDR grafting retained their capacity to bind the antigen, the affinity was usually diminished.24, 25 This may be due to the altered CDR conformation after being grafted onto the human framework.26 As the TCR also has the antigen pHLA to be contacted by six CDRs, in which α and β chains contribute three of them, respectively, we postulated that murine TCR can be humanized with a similar CDR grafting methodology, whilst the problem of affinity loss may be tackled during the process.

MAGE‐A3 has been expressed frequently in human tumours, and its expression is associated with poor prognosis.27, 28, 29 A murine TCR specific for the MAGE‐A3 antigen peptide has been generated from an HLA‐A*02 transgenic mouse.30 The murine TCR‐transfected human T‐cells have demonstrated potent anti‐tumour cytotoxicity and been used in clinical trials.1, 30 As the functional avidity between a T‐cell and its target cell is predominantly dependent on the TCR affinity, and cancer cells often present extremely low‐density epitopes for escaping immune surveillance, further affinity enhancement by several in vitro designs is required for optimal therapeutic applications.30, 31

In this study, based on the principle of CDR grafting,32 we constructed the humanized TCR in which the CDRs of the murine TCR SRm1 were grafted to human variable region fragments to reduce the immunogenicity. Considering the stability of the framework after CDR grafting, we also introduced point mutations to optimize key interaction by computer modelling.33 We demonstrated that the SRm1 humanized with stability‐optimized human TCR frameworks (g13t) showed almost 2·5‐fold higher affinity than that of the parent murine TCR. The humanized MAGE‐A3 TCR (SRm1g13t)‐transfected T‐cells showed enhanced cytotoxicity. The affinity of humanized TCR was optimized further by phage display after converting the TCR to a single‐chain TCR variable‐fragment (sTv).34, 35 Our study suggests that the CDR grafting strategy used for TCR humanization can enable the humanized TCR to retain the specificity and affinity of the parent murine TCR. Potent T‐cell activation could be generated with improved affinity of the TCR by directed molecular evolution.

Materials and methods

Construction and expression of TCR α and β chains

We selected templates including frameworks of previously optimized human αβ TCR sequences of g13t (derived from TRAV21*01 and TRBV6‐5*01) with good homologous scores for the murine counterparts, and used computer modelling to identify the key residues that supported the CDRs and could stabilize the TCR structure. The variable domains of humanized TCR SRm1g13t were constructed by mutating several amino acids in SRm1 variable region of α chain (SRm1a) (V11L, T12N, L13V, T14P, M20S, L21I, V43R, H45D, L46P, N47G, E48K, G75R, S84D, K91I, S92E, S93R, A94I, L96P, S97N, A100G, L101T, Y103F) and β chain (SRm1 b) (V4I, M7T, K13V,R14K, M15T, L20T, L48Q, G75R, I90R, L91I, A94V, N97S, Q98D, T99S, S100A, V101L, F103L) (Table 1). Variable regions (Fig. 1a) were fused with human constant region by overlapping polymerase chain reaction. The TCR fragment genes were synthesized by GenScript and cloned to pET‐28a vector (Novagen) after codon optimization for Escherichia coli expression system. The recombinant plasmids were transformed into BL21 (DE3) competent cells, after sequencing (Igebio), TCR α and β chains were overexpressed as inclusion bodies (IBs) by inducing with 1 mm IPTG at 37°, 250 rpm for 4 hr.

Table 1.

Design of SRm1g13t sequence

Figure 1.

Construction and production of humanized T‐cell receptors (TCRs) in vitro. (a) The design for different TCR expression cassettes. All these genes were cloned into pET‐28a vector. Vα, variable domain of α chain; Cα, constant domain of α chain; Vβ, variable domain of β chain; Cβ, constant domain of β chain. (b) The gel filtration chromatography of in vitro refolded SRm1 and SRm1g13t eluted with phosphate‐buffered saline. The desired fractions were collected and pooled. (c) The gel filtration chromatography of refolded peptide human leucocyte antigen (pHLA) (MAGE‐A3) and biotinylation efficiency analysis of purified pHLA (MAGE‐A3). The HLA‐A*0201 and β ₂ m inclusion bodies (IBs) were purified firstly, and then denatured protein products were refolded in vitro. The refolded products were purified by AKTA system. The desired fractions were collected and concentrated to do a biotinylation assay, and the reaction efficiency was checked on sodium dodecyl sulphate–polyacrylamide gel electrophoresis (SDS–PAGE). Lane M: protein ladder (NEB, P7703V); Lane 1: streptavidin (SA); Lane 2: pHLA (MAGE‐A3) monomer; Lanes 3–5: the mole ratio of SA to pHLA (MAGE‐A3) monomer is 1 : 1, 1 : 4, 1 : 8.

Protein refolding and purification

The soluble TCRs were produced as described previously.36 In brief, the purified TCR α and β IBs were denatured in 6 m guanidine‐HCl with 15 mm dithiothreitol at 37° for 30 min. The denatured IBs were diluted to a final concentration of 50 mg/l of α and 30 mg/l of β together in a refolding buffer containing 100 mm Tris‐HCl (pH 8·1), 0·4 m l‐arginine, 5 m urea, 2 mm EDTA, 6·5 mm β‐mercaptoethylamine and 1·87 mm cystamine. The refolding mixture was dialysed in deionized water at 4° overnight, then dialysed against 10 mm Tris‐HCl (pH 8·1) twice. The proteins were eluted with a gradient buffer (10 mm Tris‐HCl, pH 8·1/0–1 m NaCl) flowing through an anion‐exchange column QHP (GE Healthcare, Uppsala, Sweden) with AKTA system (GE Healthcare, Uppsala, Sweden). All the collected fractions were analysed on 12% sodium dodecyl sulphate–polyacrylamide gel electrophoresis (SDS–PAGE) under reducing and non‐reducing conditions, the desired fractions were pooled and concentrated with 10 kD Amicon Ultra‐15 Centrifugal Filter Unit (Merck Millipore, Cork, Ireland).

The biotinylated and non‐biotinylated pHLA complexes were also produced as described previously.37 The MAGE‐A3 peptide (KVAELVHFL) (ChinaPeptides) was dissolved in dimethylsulphoxide, and 6 mg HLA‐A*0201 and 4 mg β ₂ m IBs were diluted in 1 ml volume with two separate tubes in an injection buffer (pH 4·2) containing 3 m guanidine HCl, 10 mm sodium acetate and 10 mm EDTA. The prepared HLA‐A*0201 and β ₂ m IBs were injected into 200 ml folding buffer (pH 8·3) containing 0·1 m Tris‐HCl (pH 8·0), 0·4 m l‐arginine, 2 mm EDTA, 5 mm reduced glutathione, 0·5 mm oxidized glutathione and 0·2 mm PMSF for incubation overnight at 70 rpm and 4°. An aliquot of HLA‐A*0201 IBs was injected into the folding buffer every third day. The whole folding reaction was then dialysed in deionized water at 4° overnight, then dialysed against 20 mm Tris‐HCl (pH 8·1) twice. The proteins were eluted with a gradient buffer (20 mm Tris‐HCl, pH 8·1/0–1 m NaCl) flowing through an anion‐exchange column QHP (GE Healthcare) by AKTA system (GE Healthcare).

To obtain the biotinylated pHLA, the fractions containing purified products were mixed, concentrated and buffer‐exchanged with 20 mm Tris‐HCl (pH 8·1). The proteins were adjusted to a final concentration of 40 μm and biotinylated with BirA enzyme (Avidity, bulk BirA) at 60 rpm and 20° for 24 hr. The reaction was diluted with 20 mm Tris‐HCl (pH 8·1) and purified further by loading onto a QHP column and a Superdex 75 10/300 GL column (GE Healthcare).

Determination of TCR affinity with surface plasmon resonance (SPR) analysis

The affinity of soluble TCRs was determined by Biacore T200 (GE Healthcare) analysis as described previously.38 The CM5 chips were coated with streptavidin (SA) using amine coupling, and biotinylated pHLA was captured on the active channel. Soluble TCRs were injected sequentially through the reference and active channels at various concentrations or a single concentration. After subtracting the background signals, the dissociation rate constant (k _d), association rate constant (k _a) and equilibrium dissociation constant (K _D) were determined with the Biacore T200 evaluation software to fit the data using the 1 : 1 binding model. In order to high‐throughput estimate the affinities, the kinetic constants of the soluble affinity‐enhanced TCRs were obtained with single concentration. In this case, their R _max values were set with the formula of R _max = level of captured pHLA × MW_TCR/MW_pHLA × 0·7 to constant values prior to fitting.

Generation of HAT mutants

The selection of HAT by phage display was previously described.34 Briefly, we first designed the sTv (SRm1a g13t‐linker‐SRm1b g13t) of humanized TCR containing a flexible linker: GGGSEGGGSEGGGSEGGGSEGGTG (Table 2), and cloned it into a phage display vector. Then mutations were introduced into the CDR3 of α and β chains to construct libraries containing diversities of 1·32 × 10⁸ for both αCDR3 and βCDR3, and the actual diversity of each library could only be retained with larger individual colony‐counting than the theoretical one.

Table 2.

Design of sTv sequence

HATs were selected by panning the phage libraries with biotinylated pHLA, and were then analysed by the monoclonal phage ELISA and inhibition phage ELISA as described previously.34 Briefly, after several rounds of selection, the individual colonies were inoculated into a deep 96‐microwell plate containing 200 μl 2 × TY medium supplemented with 100 μg/ml ampicillin and 2% glucose. After incubation for 3–5 hr at 250 rpm and 37°, a total of 10⁹ helper phages were added and incubated for 30 min at 37° without shaking. The microwell plate was shaken at 250 rpm and 37° for another 30 min. The culture medium was replaced with 200 μl 2 × TY medium containing 100 μg/ml ampicillin and 50 μg/ml kanamycin, and incubated at 250 rpm, 30° overnight. The microwell plate was spun at 2000 g and 4° for 10 min to separate the phage and cells, the phage‐containing supernatant was used to screen high‐affinity binders. To screen binders, biotinylated pHLA was captured on two 96‐well ELISA plates coated with SA. The additional non‐specific protein‐binding sites on the ELISA plates were blocked with 300 μl 2% MPBS per well for 1 hr at room temperature. Phage mixtures were prepared with 100 μl supernatant and 100 μl 2% MPBS. Wells in one plate (for phage ELISA analysis) were added with 100 μl 1% MPBS, and the second plate (for inhibition phage ELISA) was added with 1% MPBS containing 400 nm pHLA, then 100 μl of the phage mixtures was added to the wells prepared as above. Then, 200 μl phage/MPBS solution was added to 96‐well ELISA plates to bind specific phage. Binding of sTv displaying phage to biotinylated pHLA was detected with 100 μl 1% MPBS containing 5000‐fold dilution of HRP‐anti‐M13 (GE Healthcare, 27‐9421‐01) and the TMB developing kit (Beyotime, P0209). The positive clone should have lower signal in the inhibition phage ELISA than the phage ELISA. The selected clones were analysed by sequencing, the sTv constructs were converted to the TCR heterodimer format by simply exchanging the CDR3 sequences of the wild‐type SRm1g13t TCR, and then the proteins were expressed as IBs, processed with refolding and purification for the SPR analysis.

Cell culture

NCI‐H1299 (MAGE‐A3⁺/HLA‐A*0201⁻) non‐small lung cancer cells and U266B1 (MAGE‐A3⁺/HLA‐A*0201⁺) haematological malignancy cells were purchased from ATCC, and NCI‐H1299‐A2 (MAGE‐A3⁺/HLA‐A*0201⁺) was constructed by transferring NCI‐H1299 with the HLA‐A*0201 gene in‐house. All these target cells were cultured with RPMI 1640 medium (Gibco) supplemented with 10% fetal bovine serum (FBS). T2 (HLA‐A*0201⁺) was cultured with IMDM medium (Gibco) supplemented with 10% FBS.

Peripheral blood mononuclear cells (PBMCs) were prepared with Ficoll‐Hypaque density gradient centrifugation using human lymphocyte isolation buffer. CD3⁺ T‐cells were isolated from PBMCs with CD3 microbeads (Cat: 130050101, Miltenyi Biotec) by positive selection. The purified CD3⁺ T‐cells were incubated in 10% FBS‐RPMI 1640 medium plus 50 U/ml interleukin (IL)‐2 (Cat:200‐02, PeproTech).

TCR transfection of CD3⁺ T‐cells

The full‐length TCR α and β chains were codon optimized for human cell expression system and cloned into the RNA expression vector pGEM‐4Z individually. The TCR α and β chains mRNA were generated by using mMESSAGE mMACHINE (Cat: AM134555, Ambion). The CD3⁺ T‐cells stimulated with human T‐activation anti‐CD3/28 beads (dynabeads, Cat:11131D, Gibco) at a cell to beads ratio of 1 : 1 were cultured in 12‐well plates with 10% FBS‐RPMI 1640 medium plus 50 U/ml IL‐2 for 48–72 hr. The stimulated CD3⁺ T‐cells were electroporated with mRNA using an ElectroAquarePorator (Lonza) and were tested for the expression of introduced TCRs overnight.

Cytotoxicity assays

The CytoTox96 Non‐Radioactive Cytotoxicity Assay kit was used to determine the cytotoxicity activity in triplicate. In brief, a total of 1 × 10⁵ cells/well TCR transfection positive CD3⁺ T‐cell at an effector to target ratio (E:T) (T2 cells loaded with MAGE‐A3 peptide, U266B1, NCI‐H1299‐A2 or NCI‐H1299) of 5 : 1 were cultured in a 96‐well plate for 24 hr at 37°, 5% CO₂. The detection of lactate dehydrogenase (LDH) release in supernatants was carried out following the manufacturer's instructions, and results were calculated as:

Cytotoxicity = (Experimental – Effector Spontaneous – Target Spontaneous)/(Target Maximum – Target Spontaneous).

Enzyme‐linked immunosorbent spot (ELISpot)

For the interferon (IFN)‐γ ELISpot assays, a total of 2 × 10³ CD3⁺ T‐cells cultured with 1 × 10⁴ target cells were incubated overnight at 37°. The plates were evaluated using an automated ELISpot reader, and developed according to manufacturer's instructions after culturing for 24 hr.

Staining transfected T‐cells with pHLA tetramers and antibodies

The pHLA (MAGE‐A3) complex was used to produce tetramer linking PE, the ratio of SA‐PE (Cat: 554061, BD) to pHLA (MAGE‐A3) was 1 : 4, SA‐PE was divided into 10 aliquots, each of them was added into the pHLA (MAGE‐A3) and incubated for 10 min at room temperature individually. The transfected CD3⁺ T‐cells were stained with an APC‐labelled anti‐CD8 antibody (Cat: 301104, Biolegend) and the PE‐labelled pHLA (MAGE‐A3) tetramer. FITC‐labelled anti‐CD3 antibody (Cat: 300406, Biolegend) and APC‐labelled antibody against the constant region of mouse TCR β chain (Cat: 109212, Biolegend) were also used to detect the TCR expression of transfected CD3⁺ T‐cells.

Intracellular cytokine staining

A total of 1 × 10⁵ cells/well CD3⁺ T‐cells were co‐cultured with 2 × 10⁵ cells/well target cells in 10% FBS‐RPMI 1640 in 48‐well plates. Golgi stop‐Brefeldin A (Cat: 420601, Biolegend) was added 6 hr before the detection. Cells were stained with an anti‐CD3 antibody, and the CD3⁺ cells were gated and stained, respectively, with antibodies of anti‐IFN‐γ (Cat: 502509, Biolegend), anti‐tumour necrosis factor (TNF) α (Cat: 557647, BD) or anti‐IL‐2 (Cat: 554567, BD) for intracellular detection of the cytokines according to the BD cytofix/cytoperm kit (Cat: 554714, BD).

Statistical analysis

All graphing and statistical analyses were performed using Graphpad Prism 6. Unpaired two‐tailed t‐test was performed using the Student's t‐test; significance is indicated as *P < 0·05, **P < 0·01, ***P < 0·001, NS means P > 0·05.

Results

Design and production of humanized TCR and pHLA

The murine TCR SRm1 specific for human tumour testis antigen peptide MAGE‐A3_112‐120 (KVAELVHFL) HLA‐A*02 complex [pHLA (MAGE‐A3)] was humanized on the basis of a folding optimized human TCR g13t. The resultant humanized TCR was named as SRm1g13t. Full‐length α and β chains of SRm1 and SRm1g13t TCRs were cloned into pET‐28a vector (Fig. 1a) for the expression of both TCRs, which were refolded in vitro and purified with the purity over 90% analysed by SDS–PAGE (Fig. 1b). The theoretical molecular weight of these TCRs was about 45 kDa, which was consistent with the bands in SDS–PAGE under the non‐reducing condition (data not shown). There were two bands in reducing SDS–PAGE, the molecular weights were about 22 kDa and 28 kDa that coincided with the theoretical molecular weight of the TCR α and β chains (data not shown). To evaluate the affinity of SRm1 and SRm1g13t, the target pHLA was also refolded and purified in vitro (Fig. 1c), and biotinylated to prepare pHLA tetramer for the cell assays. The biotinylation of pHLA (MAGE‐A3) was confirmed with the native gel analysis with an efficiency of about 95% (Fig. 1c).

Affinity determination of murine and humanized TCRs with SPR analysis

To evaluate the pHLA (MAGE‐A3) recognition and binding strength, the affinities of TCR SRm1 and SRm1g13t were determined by Biacore T200 with the global fitting method. Both TCRs showed a character of the single phase binding, and the pHLA (MAGE‐A3) binding affinity of SRm1 was indicated by a K _D of 5·9 × 10⁻⁴ m (Fig. 2). Although the humanized antibody usually showed decreased affinity in comparison with the parent murine molecule, the humanized TCR SRm1g13t, with a K _D of 2·4 × 10⁻⁴ m for pHLA (MAGE‐A3) binding, showed better affinity than that of the parent TCR SRm1 (Fig. 2). It was observed that two TCRs had similar retention times of 1·8 seconds and 1·9 seconds, as well as similar k _d figures of 3·8 × 10⁻¹second⁻¹ and 3·6 × 10⁻¹second⁻¹ for SRm1 and SRm1g13t, respectively. However, the humanized version SRm1g13t had a quicker on‐rate with a k _a of 1·5 × 10³ M⁻¹second⁻¹ in comparison to SRm1 with a k _a of 6·4 × 10² M⁻¹second⁻¹.

Figure 2.

Surface plasmon resonance (SPR) analysis of the binding affinity between T‐cell receptors (TCRs) and the peptide human leucocyte antigen (pHLA) (MAGE‐A3). The binding activity of these TCRs was analysed by Biacore T200. CM5 Biacore chips were coated with streptavidin (SA) using amine coupling, then the biotinylated pHLA (MAGE‐A3) was captured on the active channel. After blocking non‐specific binding sites of both the reference and active channels with 50 mm biotin, SRm1and SRm1g13t were injected sequentially through the reference and active channels at various concentrations, and buffer alone was used as a blank control. The affinities of TCRs were determined by multiple cycle kinetics. And the analytic concentrations were 5, 10, 20, 40, 80, 160 μm for SRm1, and 2·5, 5, 10, 20, 40, 80, 160 μm for SRm1g13t. After subtracting the blank control signals, the data were fitted with a 1 : 1 binding model using Biacore T200 evaluation software to obtain the kinetic constants (k _a, k _d and K _D).

Cytotoxicity analysis of T‐cells transfected with the murine and humanized TCRs

CD3⁺ T‐cells were transfected with full‐length TCR α and β chains mRNA for testing the antigen‐specific recognition. To reduce mispairing with endogenous TCR chains, we used murine constant regions, which were shown to demonstrate non‐immunogenicity,17 to construct both murine and humanized TCRs (Fig. 3a). The transfected CD3⁺ T‐cell population expressing expected murine TCR β chain was about 90%. The pHLA (MAGE‐A3) tetramer‐positive cells were taking just over 7% of the cell population for SRm1g13t, and over 4% for SRm1, which was consistent with the difference of TCR affinities for the antigen. The tetramer and CD8 double‐positive populations were 3·8% and 1·5% for SRm1g13t and SRm1, respectively (Fig. 3b), whilst the total CD8‐positive populations were similar between the two transfected cells.

Figure 3.

SRm1g13t‐transfected CD3⁺ T‐cells showed potent function. (a) Schematic illustration of RNA expression vectors encoding SRm1 and SRm1g13t expression cassette. (b) Flow cytometric analysis of T‐cell receptor (TCR)‐transfected CD3⁺ T‐cells. Dot plots showing the FACS profiles of CD3⁺ T‐cells stained with anti‐human‐CD8‐APC antibody and peptide human leucocyte antigen (pHLA) (MAGE‐A3) tetramer‐PE, or anti‐human CD3‐FITC and anti‐mouse‐TCR beta‐APC antibodies. (c) Recognition of tumour cells by the TCR‐transfected CD3⁺ T‐cells. The CD3⁺ T‐cells transfected with SRm1 or SRm1g13t were co‐cultured for 24 hr with different tumour cells. The release of interferon (IFN)‐γ was measured by IFN‐γ enzyme‐linked immunosorbent spot (ELISpot) assay, and values indicated the mean of duplicate samples. (d) Specific killing of tumour cell lines by TCR‐transfected CD3⁺ T‐cells. Cytotoxicity activity of TCR‐transfected CD3⁺ T‐cells was determined by lactate dehydrogenase (LDH) release after 24 hr. Polyclonal CD3⁺ T‐cells, obtained from a healthy donor, were transfected and used at 75 000 cells per well at an effector to target ratio of 5 : 1. (e) Intracellular cytokine staining for interleukin (IL)‐2, tumour necrosis factor (TNF) α and IFN‐γ of SRm1‐ or SRm1g13t‐transfected CD3⁺ T‐cells after antigen‐specific stimulation. SRm1‐ or SRm1g13t‐transfected CD3⁺ T‐cells were co‐cultured with NCI‐H1299‐A2 overnight, the cells were then collected and stained for the detection of cell surface CD3 molecule expression and intracellular IL‐2, TNF α or IFN‐γ expression. The plots were gated on CD3⁺ T‐cells, and the percentage of positive cells was indicated in the upper‐right quadrant. A representative experiment was shown.

In order to compare T‐cell function after transfection with SRm1g13t or SRm1 mRNA, the CD3⁺ T‐cells were cultured with the cells of T2 loaded with the MAGE‐A3 peptide, U266B1, NCI‐H1299‐A2 or NCI‐H1299, respectively. The SRm1g13t‐transfected CD3⁺ T‐cells produced higher levels of IFN‐γ than that of SRm1 cells for the first three target cells in the list, whilst both SRm1g13t‐ and SRm1‐transfected cells showed no recognition for the NCI‐H1299 tumour cells, which were HLA‐A*0201 negative, and indicated the responses were specific in a HLA‐restricted fusion (Fig. 3c). We also measured the specific lysis of tumour cells by using the transfected CD3⁺ T‐cells. SRm1g13t‐ and SRm1‐transfected CD3⁺ T‐cells exhibited similar lytic function against antigen‐positive U266B1 and NCI‐H1299‐A2 tumour cell lines (Fig. 3d).

To compare the cytokine production further, the percentage of SRm1g13t‐ and SRm1‐transfected CD3⁺ T‐cells that produce IL‐2, TNF α and IFN‐γ was determined by intracellular cytokine staining after overnight co‐culture with NCI‐H1299‐A2. The percentages of IL‐2‐ and IFN‐γ‐producing CD3⁺ T‐cells were almost doubled in number by SRm1g13t‐transfected cells in comparison to that of SRm1‐transfected cells (Fig. 3e). However, the TNF α‐producing population was stained with about an equal percentage in both SRm1g13t‐ and SRm1‐transfected CD3⁺ T‐cells (Fig. 3e). All these data demonstrated that the humanized TCR could confer CD3⁺ T‐cells with potent anti‐tumour cytotoxicity, even better than the parent murine TCR.

Affinity enhancement of humanized TCR with phage display

Previous studies have shown that TCR affinity optimization is needed for the T‐cells to achieve required activity. In order to optimize the SRm1g13t affinity by M13 phage display in vitro, we detected pHLA (MAGE‐A3) binding of the displayed SRm1g13t sTv to confirm the success of sTv display in a phage ELISA (Fig. 4). We generated the SRm1g13t phage libraries containing random mutations introduced into CDR3 of TCR α and β chains. The theoretical diversity of mutants in two phage libraries is equal to 1·32 × 10⁸, while the actual built TCR α and β libraries both contained nine times greater single colonies of 1·2 × 10⁹, demonstrating adequate diversity. After several rounds of selection, many clones were selected by monoclonal phage ELISA and inhibition phage ELISA. Based on the sequencing results of the selected clones, three unique clones in αCDR3 library and six unique clones in βCDR3 library were identified (Table 3). The mutations of αCDR3 F106L and βCDR3 E115G were observed among the selected αCDR3 or βCDR3 loops. Then the mutated clones were cloned into a soluble expression vector for expressing the sTv and analysing the k _d by Biacore T200 (Table 3). On the basis of these data, we found that the selected mutants always showed slower off rates compared with that of SRm1g13t sTv for improving the binding. We believe that further improved affinity should be generated from the combination of high‐affinity variants with these unique sequences.

Figure 4.

Antigen‐binding sensitivity and specificity of single‐chain T‐cell receptor (TCR) variable‐fragment (sTv) phage particles. Specific peptide human leucocyte antigen (pHLA) (MAGE‐A3) at different concentrations and non‐specific pHLA were coated in microtitres, and 10¹⁰ purified sTv and helper phage particles were applied to each well for phage ELISA. The same amount of helper phage was used as a negative control. The bound phages were detected with HRP‐conjugated anti‐M13 antibody. The binding signal was analysed at OD ₄₅₀. The phage ELISA was performed as described above, except serial dilutions of purified specific pHLA (MAGE‐A3). The non‐specific pHLA was used at the concentration of 200 nm.

Table 3.

Summary of output from sTv phage display selection and SPR determination of k _d (1/second)

Determination of the affinity for TCR variants

A total of 18 TCR combinational mutants was constructed with the CDR sequences of the sTvs and cloned into pET‐28a vector for IBs expression and refolding. The purity of each purified TCR was determined by SDS–PAGE, and showed over 90% pure for all TCR variants (Fig. 5a). Moreover, the molecular weights of the two bands in reducing SDS–PAGE were about 22 kDa and 28 kDa, which coincided with the theoretical molecular weight of TCR α and β chains (Fig. 5b). In order to select a mutant with the best affinity from these TCRs, we analysed the affinities of 18 HATs with a single concentration method (Fig. 5c). The method was verified by comparing the results of clone A31C2‐B31G1 affinity measurement, which were obtained with both the single concentration method and the standard single‐cycle kinetics. The results showed that both methods give practically the same affinities with K _D values of 1·4 × 10⁻⁶ m for the single‐cycle kinetics and 1·3 × 10⁻⁶ m for the single concentration method (data not shown). We demonstrated that the affinity of 18 combinational humanized TCRs was higher (K _D ~ 1 μm) than that of SRm1g13t (K _D > 200 μm; Fig. 5c; Table 4). A32E6‐B32G3 showed the highest affinity, with an increment of almost 500‐fold of that for SRm1g13t. In order to analyse the cellular function of the HAT, CD3⁺ T‐cells were transfected with the genes of combinational clones of A31D5‐B32G3 or A32E6‐B32G3 with K _D values of 1·1 × 10⁻⁶ m or 5·2 × 10⁻⁷ m, respectively.

Figure 5.

High‐affinity T‐cell receptors (HATs) refolded in vitro. (a and b) Coomassie‐stained sodium dodecyl sulphate–polyacrylamide gel electrophoresis (SDS–PAGE) of purified T‐cell receptors (TCRs). (a) After in vitro refolding, all TCRs were ultimately purified by gel filtration chromatography, and the desired fractions were collected and pooled to be analysed in non‐reducing SDS–PAGE. The non‐reduced TCRs run as a single band of 45 kDa. (b) The purified TCRs were analysed in reducing SDS–PAGE. The reduced TCRs run as two separate bands of 22 and 28 kDa, respectively. Lane M: protein marker; Lane 1: A31C2‐B31G1; Lane 2: A31C2‐B32B5; Lane 3: A31C2‐B32B6; Lane 4: A31C2‐B32C6; Lane 5: A31C2‐B32G3; Lane 6: A31C2‐B32G7; Lane 7: A31D5‐B31G1; Lane 8: A31D5‐B32B5; Lane 9: A31D5‐B32B6; Lane 10: A31D5‐B32C6; Lane 11: A31D5‐B32G3; Lane 12: A31D5‐B32G7; Lane 13: A32E6‐B31G1; Lane 14: A32E6‐B32B5; Lane 15: A32E6‐B32B6; Lane 16: A32E6‐B32C6; Lane 17: A32E6‐B32G3; Lane 18: A32E6‐B32G7. (c) Biacore T200 analysis of purified TCRs bound to the peptide human leucocyte antigen (pHLA) (MAGE‐A3) immobilized on CM5 sensor chip. The biotinylated pHLA (MAGE‐A3) was captured on the active channel, and all the TCRs were determined by single concentration (0·5 μm) kinetics. After subtracting the blank control signal, the data were fitted with a 1 : 1 binding model using Biacore T200 evaluation software to obtain the kinetic constants (k _a, k _d and K _D). The kinetic constants of the 18 TCRs were obtained by setting their R _max to constant values (using the formula: R _max = level of captured pHLA × MW_TCR/MW _{p HLA}× 0·7) before fitting. MW, molecular weight.

Table 4.

SPR analysis between different clones and pHLA (MAGE‐A3)

Enhanced function in HAT‐transfected CD3⁺ T‐cells

To test the specificity, we transfected T‐cells with combined HAT variants A31D5‐B32G3 or A32E6‐B32G3 using the RNA expression vector pGEM‐4Z. The positive percentages of designated TCR were about 90% (Fig. 6a) for SRm1g13t‐, A31D5‐B32G3‐ or A32E6‐B32G3‐transfected CD3⁺ T‐cells. The positive percentages of tetramer staining were over 30% in the two SRm1g13t mutant‐transfected CD8⁺ T‐cells, and much higher than that of SRm1g13t (Fig. 6b), which indicated the affinity‐enhanced tetramer binding activity. In addition, A31D5‐B32G3‐ and A32E6‐B32G3‐transfected CD3⁺ T‐cells produced IFN‐γ levels higher than SRm1g13t significantly on MAGE‐A3⁺/HLA‐A*0201⁺ tumour cell lines of U266B1 and NCI‐H1299‐A2, but marginally on the MAGE‐A3 peptide‐loaded T2 cells (Fig. 6c). All these TCR‐transfected CD3⁺ T‐cells did not recognize the NCI‐H1299, showing no binding to the HLA‐A*0201‐negative cells with the affinity‐enhanced TCRs. We also measured the percentage of high‐affinity variant‐transfected CD3⁺ T‐cells for production of IL‐2, TNF α and IFN‐γ by intracellular cytokine staining. The A31D5‐B32G3‐ and A32E6‐B32G3‐transfected CD3⁺ T‐cells produced IL‐2, TNF α and IFN‐γ levels much higher than that of SRm1g13t (Fig. 6d). These results indicated that HAT‐transfected CD3⁺ T‐cells had enhanced activities against the target cells.

Figure 6.

Enhanced function in high‐affinity T‐cell receptor (HAT)‐transfected CD3⁺ T‐cells. (a and b) Flow cytometric analysis of T‐cell receptor (TCR)‐transfected CD3⁺ T‐cells. (a) Dot plots showing the FACS profile of transfected CD3⁺ T‐cells stained with anti‐human CD3‐FITC and anti‐mouse‐TCR beta‐APC antibodies. (b) Dot plots showing the FACS profile of transfected CD3⁺ T‐cells stained with anti‐human‐CD8‐APC antibody and peptide human leucocyte antigen (pHLA) (MAGE‐A3) tetramer‐PE. (c) Recognition of tumour cells by the TCR‐transfected CD3⁺ T‐cells. CD3⁺ T‐cells transfected with SRm1g13t, A31D5‐B32G3 or A32E6‐B32G3 were co‐cultured for 24 hr with different tumour cells. The release of interferon (IFN)‐γ was measured by IFN‐γ enzyme‐linked immunosorbent spot (ELISpot) assay, and values indicated the mean of duplicate samples. (d) Intracellular cytokine staining for tumour necrosis factor (TNF) α, interleukin (IL)‐2 and IFN‐γ of different affinity TCRs‐transfected CD3⁺ T‐cells after antigen‐specific stimulation. SRm1g13t‐, A31D5‐B32G3‐ or A32E6‐B32G3‐transfected CD3⁺ T‐cells were co‐cultured with NCI‐H1299‐A2 cells overnight. Cells were then collected and stained for the detection of cell surface CD3 expression and intracellular IL‐2, TNF α or IFN‐γ expression. The plots are gated on CD3⁺ T‐cells, and the percentage of positive cells was indicated in the upper‐right quadrant. A representative experiment was shown.

Discussion

Using HLA‐A*02 : 01 transgenic mice, human cancer antigen‐specific TCRs can be often isolated from the mouse CD8⁺ T‐cells. Because the mouse CD8 interacts inefficiently with human HLA α3 domain, T‐cell‐possessing HAT will be stimulated and selected in a CD8‐independent manner during the T‐cell cloning processes.16 Obviously, such features should facilitate constant success of T‐cell cloning. Murine TCRs specific for human gp100, p53, MAGE‐A3 and CEA have been tested in clinical trials. We believe that the conceivable limitation of murine TCR clinical usage is the potential immunogenicity in subsequent TCR‐transferred T‐cell (TCR‐T) applications, especially when repeated doses are required for optimal therapeutic effects. Immunogenicity of murine TCRs has been verified, and it was found that some patients treated with mono‐doses could develop antibodies against the variable domains of the TCRs.17 Antibodies specific for murine TCR against cancer antigen CEA were observed at 3 and 4 months post‐treatment, and these antibodies also inhibited the secretion of IFN‐γ by PBL expressing the CEA TCR.18 Although current TCR‐T trial data with mono‐doses‐transfected T‐cells have not given a clear answer whether the presence of anti‐murine TCR antibody can have an impact on the anti‐tumour response, we believe that the prevention of immunogenicity may be needed for future trials. Moreover, studies have shown that the immunogenicity of murine scFv can generate anti‐mouse scFv antibodies for the elimination of CAR‐Ts and result in limited anti‐tumour effects,19, 20 and efforts have been made to humanize the murine‐derived scFv for reducing the immunogenicity in CAR‐Ts therapy.39 It is inevitable that the immunogenicity of murine TCRs may be a critical element to activate host immune response, and affect clinical outcomes of the therapeutic murine TCRs.

The immunogenicity of murine TCR could be diminished by humanization of the murine protein, on the basis of successful applications of humanized antibody in the past decades. We constructed the humanized TCR in which the CDRs of the murine TCR were grafted to human variable region fragments to reduce the immunogenicity. As the murine constant domain has demonstrated non‐immunogenicity, we may use it to improve the expression of introduced TCR and reduce mis‐pairing of introduced TCR with endogenous TCR chains.40, 41 Our studies have shown that the SRm1g13t could be expressed well on T‐cell surface, surprisingly, conferring better T‐cell activation than that of the parent murine TCR SRm1. This occurrence might be due to SRm1g13t conferring the TCR with higher affinity and quicker on rates but similar off rates when contacting the pHLA compared with SRm1, which led to more serial triggering to activate T‐cells.42

The templates used for the humanization were optimized with a new design by point mutations for improved stability and refolding (a pending patent by Yi Li). The stability‐optimized human TCR frameworks will maintain the stability of humanized TCR for further affinity improvement. Because the stability of TCR is related to the expression and refolding efficiency, the expressions of TCR SRm1g13t and SRm1 on T‐cell surface were measured to be 91·8% and 94·2%, respectively; however, refolding efficiency in vitro of SRm1g13t was better than SRm1 with about 29% and 13%, respectively (data not shown). Therefore, the folding optimized human v‐domains did play a role in improving the TCR stability and refolding efficiency.

The efficacy of TCR‐based immunotherapy is associated with the T‐cell avidity for the antigen on target cells, and animal models have also shown that only high‐avidity T‐cells can mediate tumour regression.43, 44 Recently, it has been shown that T‐cell avidity is primarily dependent on the TCR affinity,7, 8, 10, 45 and usually determines the levels of the cytokine release, cytotoxicity and anti‐tumour response. This might be explained that affinity‐enhanced TCR could enable the T‐cell to be activated with less copy pHLA complex presented by the cancer cells.13, 18, 46, 47 As one of the tumour's escape mechanisms is to reduce the presentation of pHLA on the tumour cells, further affinity enhancement may be a powerful and attractive approach for optimal therapeutic applications.48, 49 Reports have demonstrated that the affinity of TCRs can be optimized for better anti‐tumour activity by several in vitro designs.31, 34, 50, 51 In our study, we selected the high‐affinity humanized TCR mutants by phage display. All the combinational mutants showed higher affinity than that of SRm1g13t. Of note with the phage library, we also isolated the mutation of A111T for improved affinity of the TCR, which was reported previously by other groups.30 According to the sequence, there was an interesting pattern that the mutation of F106L of αCDR3 and E115G of βCDR3 was observed among the unique clones. Positions 106 of αCDR3 and 115 of βCDR3 were located at the edge of CDR3 loops, and the residues at these positions were more likely responsible for supporting the CDR to adopt a certain conformation, rather than directly interacting with the antigen peptide based on the existing structures of TCR and pHLA complexes studied previously.52, 53, 54 In addition, the combination of any βCDR3 clone with A32E6 clone has higher affinity in comparison to that with A31C2 or A31D5 clones (Table 4). This suggested that αCDR3 might play a major role during the interaction between the TCR and pHLA. However, further studies especially with detailed structural analysis should be the best way to divulge reliable mechanisms.

Previous studies demonstrated that continuous TCR signalling is required for synapse maintenance and full effector function.55 This is consistent with our data that HAT conferred T‐cells with better avidity and demonstrated better cytokine secretion of the TCR‐transfected CD3⁺ T‐cells. However, over a certain threshold, the affinity of TCR may not contribute to enhanced tetramer binding and T‐cell activation.56, 57, 58, 59 It is important for HAT to retain pHLA specificity, as off‐target toxicity should be avoided for the therapeutic applications.12, 60 The high‐affinity clones A31D5‐B32G3 and A32E6‐B32G3 were shown to retain their specificity for HLA‐A*0201 or the MAGE‐A3 antigen peptide (data not shown).

Although it is much easier to generate a murine TCR against human antigen, the immunogenicity may be a limitation for the usage. We used a pair of optimized human TCR g13t VαVβ domains to construct SRm1g13t for humanizing the murine TCR SRm1 specific for the antigen of HLA‐A*02/MAGE‐A3:112‐120 peptide (KVAELVHFL). The humanized TCR SRm1g13t exhibited higher affinity than that of murine parent TCR SRm1 and retained better anti‐tumour function. Further affinity enhancement was achieved by directed molecular evolution using phage display libraries, and the high‐affinity variants of A31D5‐B32G3 and A32E6‐B32G3 conferred the transfected CD3⁺ T‐cells with improved cellular function against the target cells. These results suggested that the technology of CDR grafting on folding optimized human TCR domains could be used to humanize murine TCRs generated from HLA transgenic mice.

Disclosures

The authors declare that they have no conflict of interest.