Human PBMCs (Human Peripheral blood mononuclear cells) and HLA-A2+ T2 cells (can be efficiently loaded with exogenous peptides as they lack the transporter associated with antigen processing) were cultured in RPMI 1640 media (Lonza) supplemented with 10% FCS (Merck), 2mM L-Glutamine (Gibco) and 100U/ml Penicillin/Streptomycin (Invitrogen). HEK293T (Human Embryonic Kidney Epithelial) packaging cells were cultured in IMDM media (Lonza) supplemented with 10% FCS and 2mM L-Glutamine and 100U/ml Penicillin/Streptomycin.

Human PBMCs were obtained from healthy volunteers via National Health Blood Transfusion Service (Approved by UCL Research Ethics Committee, Project ID: 15887/001). 48 hours prior to retroviral transduction, bulk PBMCs were activated at 1x10⁶ cells/ml with 20ml/ml anti-CD3/CD28 dynabeads (Gibco) and 30U/ml Roche IL-2. In some experiments transduced human T cells were stimulated in vitro for 7 days with T2 cells presenting the TCR recognised peptides, followed by second round of peptide stimulation for 7 days.

For T cell functional assays, PBMCs were MACS (Magnetic-Activated Cell Sorting) sorted for CD8+ T cells following the manufacturer’s instructions (Miltenyi Biotech). Sorted CD8+ T cells were activated at 1x10⁶ cells/ml with 20ml/ml anti-CD3/CD28 dynabeads and 30U/ml Roche IL-2.

The following antibodies were used for the flow cytometric analysis of the cells: anti-human antibodies CD3-PE-Cy7 (SK7, BioLegend), TCRα/β-PE (IP26; Invitrogen), IgG1-PE (Invitrogen), IL-2 APC (MQ1–17H12, eBioscience), IFN-g-PE (B27; Invitrogen), VLHDDLLEA Tetramer-PE (BML), PD-1-BUV661 (BD), LAG3-BUV803 (Invitrogen), TIGIT-AlexaFlour700 (Invitrogen). Other antibodies used were anti-murine CD19-eFluor450 (1D3; Invitrogen), V5-APC (rabbit polyclonal; abcam) and purified myc (Bio-Rad). Live/Dead-eFluor780 (Invitrogen) was used to identify live cells. Peptides used for the T cell functional assays were: pCMVpp65 (NLVPMVATV) for the CMV TCR and pHA1 (VLHDDLLEA) for the HA-1.m2 and HA-1.m7 TCRs. The pHA2 (YIGEVLVSV) peptide was used as a control peptide.

DNA constructs were cloned into retroviral pMP71 vectors. For the CD3 constructs, IRES GFP was placed at the 3’ end and viral 2A sequences were used to separated constructs with polycistronic genes. The TCR constructs consisted of a TCRα chain, a viral P2A sequence, a TCRβ chain, a viral T2A sequence, and truncated murine CD19. A V5 tag was incorporated upstream of the TCR α variable domain and two myc tags were placed upstream of the TCR β variable domain.

Retrovirus production and the transduction of primary human T cells were performed as previously described (7). For the retroviral co-transduction of HEK293T cells, 2x10⁵ HEK293T cells were seeded into wells of a 24 well tissue culture treated plate in 500ul IMDM media and incubated for 30 mins at 37°C. After ensuring the cells were attached, media was discarded carefully and 250ml of each of the stated CD3 and TCR viral supernatants were added. Transduction was performed by centrifugation at 32°C, 2000rpm, 90 mins, no brake. The viral supernatant was discarded, and each well was supplied with 2ml fresh IMDM. 48–72h post transduction, cells were stained for Live/dead, and the antibodies stated in the text. Data was collected by LSRFortessa (BD Biosciences) and the analysis was done by FlowJo software.

3x10⁵ T2 cells were loaded with the stated concentrations of relevant or irrelevant peptide for 2 hours at 37°C, washed and then co-cultured with 7–10 days post-transduced 3x10⁵ CD8+ T cells for 18 hours. Assays were conducted in a 96 well plate, round bottom in a volume of 250 ul/well of RPMI media supplemented with Brefeldin A (Merck) at 1 ug/ul. Cells were stained for surface markers and washed. Fixation and permeabilization of the cells was performed by BD Cytofix/Cytoperm Kit as per manufactuer’s instructions. and then stained for IL-2 and IFN-γ for 1 hour at 4°C. Data was acquired on a LSRFortessa and analysed using FlowJo software.

7–10 days post-transduced CD8+ T cells were employed in killing assays. HLA-A2⁺ T2 cells were loaded with cognate peptide were labelled with 0.02uM CFSE whilst cells loaded with control peptide were labelled with 0.2uM CFSE. Following peptide loading for 2h, T2 cells were mixed at 1:1 ratio and 1x10⁵ transduced CD8+ T cells were co-cultured with 1x10⁵ mixed T2 cells for 18 hours. Cells were stained with anti-human CD3 Ab and Live/Dead antibodies, and data acquisition was done by LSRFortessa and analysed by FlowJo software. Antigen specific killing of T cells was calculated as % Specific Killing = 100- [(Relevant/Irrelevant T2 cells with T cells)/(Relevant/Irrelevant T2 cells with no T cells)]*100.

We designed a retroviral vector cassette encoding all four human CD3 chains separated by 2A sequences, and GFP separated by an IRES element ( Figure 1A ). We also produced vectors that contained only one CD3 chain (ζ or ε or δ or γ) and a control vector containing IRES GFP but no CD3 components. GFP expression was used to demonstrate successful transduction of primary human T cells, and TCR surface expression was determined in gated cells expressing similar levels of GFP ( Figures 1B, C ). Using antibodies specific for human α/β TCR and human CD3ε we found that the expression levels of TCR in T cells transduced with the vector encoding all CD3 chains was approximately 5-fold higher than in T cells transduced with the GFP control vector ( Figure 1C ). The transduction of individual CD3 chains revealed that ζ was able to increase TCR expression by nearly 3-fold, while transduction with ε resulted in only modest improvement of TCR expression. Surprisingly, provision of additional γ or δ significantly decreased TCRα/β surface expression in human T cells ( Figure 1D ). This is most likely due to a competition for CD3ε whereby high levels of γ leads to high levels of γ/ε dimers and a reduction in δ/ε dimers, while high levels of δ reduces γ/ε dimer formation. The reduction of either δ/ε dimers or γ/ε dimer impairs the assembly of and surface expression of intact TCR/CD3 complexes in T cells with excess γ or δ, respectively. The observed changes in TCRα/β surfaces expression were mirrored by similar changes in CD3ε, although the intensity of CD3ε staining was less than the TCR staining. CD4+ and CD8+ T cells showed similar responses to the provision of exogenous CD3 constructs. Increases in TCR and CD3 expression levels were seen in both T cell subsets following the provision of all four CD3 chains or CD3ζ and CD3ε alone ( Figure 1E ).

Next, we tested whether all CD3 chains were required for optimal expression increase of endogenous TCRs in primary human T cells. We first generated 4 vectors that contained three CD3 chains but lacked either γ,δ,ε or ζ. Transduction of T cells revealed that the lack of ζ significantly reduced the ability of the introduced CD3 chains to increase TCR surface expression ( Figures 2A, B ). In contrast, the lack of γ or δ did not significantly reduce the level of TCR upregulation that is seen when all four CD3 chains were transduced into T cells, whereas significantly less TCR expression was observed when ε was the missing in the transduction vector ( Figures 2A, B ). This suggested that the endogenous levels of CD3ζ are the most rate limiting CD3 chain in determining TCR surface expression, followed by CD3ε, while endogenous CD3γ and δ are relatively abundant and do not limit TCR/CD3 assembly and surface expression. This interpretation is compatible with the observation that the transduction of T cells with a vector encoding only ζ,ε and lacking γ,δ resulted in substantial upregulation of TCR surface expression ( Figures 2A, B ).

The surface expression profiles of CD3ε showed a similar pattern as the expression profiles of TCRα/β ( Figure 2C ). Interestingly, transduction of T cells with the vector containing only γ,δ,ζ resulted in a 2-fold increase in CD3ε surface expression compared to that of T cells transduced with control, indicating that surplus endogenous ε chains were available for the assembly of additional TCR-CD3 complexes.

The previous experiments have identified which CD3 chains are rate limiting and which are abundant in primary human T cells. In the next set of experiments, we wanted to determine which CD3 chains are essential to achieve TCR surface expression. For this purpose, we used the HEK293T (Human Embryonic Kidney) cells that lack endogenous CD3 and TCR. We used vectors expressing all four CD3 chains, or various chain combinations as illustrated and described in Figures 3A, C . We also used a vector encoding an HLA-A0201-restricted TCR specific for CMV, where the α and β chains were tagged with V5 and myc epitopes, respectively, to measure TCR expression with antibodies specific for these tags (8) ( Figure 3B ). Co-transduced HEK293T cells were analysed for GFP and CD19 expression, to identify GFP+CD19+ positive cells that were successfully co-transduced with both the CD3 and the TCR vectors ( Figure 3C ). Co-transduction of all four CD3 chains and the TCR resulted in efficient TCR α/β surface expression, whereas co-transduction of GFP control vector and TCR did not result in TCR surface expression ( Figure 3C ). We then tested vectors with three CD3 chains and found that the absence of the δ or ε chain nearly abolished TCRα/β surface expression ( Figures 3C–E ). In contrast, the effect of absence of the γ or ζ chain was less detrimental as TCR expression was retained, although much reduced compared to the expression achieved with all four CD3 chains ( Figures 3C–E ). We observed that the CD3 δ and ε chains were sufficient to achieve low levels of TCRα/β expression on the surface of transduced cells ( Figures 3C–E ). As before, the expression pattern of CD3ε in the co-transduced HEK293T cells mirrored the observed TCRα/β expression ( Figures 3D, E ). Finally, tetramer staining studies showed that the surface expression levels of TCRs ( Figure 3E ) correlated with the efficiency of tetramer binding ( Figures 3F, G ). The data also showed that tetramer staining required a certain threshold of TCR expression, which was only reached with all four CD3 chains and with CD3ζ,ε,δ. Together, these data suggest that CD3 δ/ε dimers are particularly important in initiating the assembly of a TCR-CD3 complex, which can then associate with γ/ε dimers and ζ/ζ homodimers for efficient migrating from the ER to the cell surface. In the absence of γ/ε and ζ/ζ some assembled TCRα/β-CD3δ/ε complexes seem to escape ER degradation and reach the cell surface, although at much lower levels than fully assembled TCR-CD3 complexes containing all CD3 chains.

As primary human T cells contained surplus CD3γ, δ, ε chains, we explored whether the provision of additional ζ chains in the context of TCR gene therapy would be sufficient to enhance the expression levels and the antigen-specific function of introduced TCRs. Hence, we placed the human CD3ζ gene downstream of three different TCRs specific for the minor histocompatibility antigens HA-1.m2 and HA-1.m7, and for the pp65 antigen of CMV ( Figure 4A ). The TCR genes were tagged with the V5 and myc epitopes to measure TCRα and β expression, and CD19 was used to identify transduced T cells. For each TCR we also tested 3 variants that contained defined amino acid changes in the framework of the TCR variable domains that we previously demonstrated to improve TCR expression levels (7, 8).

Figure 4B shows a representative experiment of human T cells transduced with the HA1.m2 wild type TCR or with the three variants V1–3 (top panels). The bottom panels show expression levels of each TCR in the presence of additional CD3ζ. The data show that for each TCR, the inclusion of CD3ζ increased the percentage of T cells that express both introduced TCR chains (V5+myc+) and reduced the percentage of cells that are single positive for V5 or myc. In these single positive T cells the introduced V5-tagged α chains have mispaired with endogenous β chains, and the myc-tagged β chains have mispaired with endogenous α chains. Figure 4C shows that additional ζ consistently increased the percentage of T cells expressing the introduced α and β chains for all 12 TCRs tested. In addition to the increase in T cell numbers expressing both TCR chains, exogenous CD3ζ also increased the level of surface expression in transduced cells ( Figure 4D ), which correlated with increased expression levels of CD3ε ( Figure 4E ). Figure 4F shows the summary of TCRα, β and CD3ε expression levels of all 12 TCRs in the absence and presence of CD3ζ.

As additional CD3ζ augmented the density of the introduced TCRs on the surface of human T cells, we tested whether this improved antigen specific T cell function. CD8+ T cells expressing either TCR-only, or the TCR+ζ combination of the HA-1.m2, HA-1.m7 or CMV-specific TCR were challenged with titrated concentrations of relevant peptide. In these functional experiments we compared the wild type TCR with the variant 3 modifications in the framework of the V-regions, because we previously showed that the V3 modifications were most effective in increasing TCR expression and function (7). This allowed us to test whether addition of CD3ζ can further improve the functionality of the V3 TCRs. Figure 5A shows that the addition of CD3ζ improved the antigen-specific IFN-γ and IL-2 production of the wild type HA1.m7 TCR and also of the V3 variant. The combination of V3+CD3ζ was most efficient in enhancing the poor antigen-specific response of the wild type HA1.m7 TCR. The analysis of the HA1.m2 and CMV TCRs confirmed that the V3 modification combined with CD3ζ was most effective in enhancing antigen-specific functionality of wild type TCRs, although the CMV-V3 TCR mounted a strong response in the absence of additional CD3ζ ( Figure 5B ). We also compared antigen-specific cytotoxicity of human T cells transduced with TCR only, or TCR+CD3ζ. Interestingly, the killing activity of the HA-1m7 TCR and its V3 version did not increase in the presence of CD3ζ ( Figure 5C ), although cytokine production of these TCRs was substantially improved by CD3ζ ( Figure 5B ). For the wild type HA-1m2 and CMV TCR the killing activity and cytokine production were both improved by CD3ζ to a similar extent, while the V3 versions of these TCRs showed enhanced cytokine production but little or no improvement in cytotoxicity ( Figures 5B, C ).

The pooled cytotoxicity data of the three wild type TCRs showed a significant increase of cytotoxicity at all peptide concentrations, particularly at low concentrations ( Figure 5D ). In contrast, the killing activity of human T cells transduced with the modified V3 TCRs was unchanged at high peptide concentrations, and increased killing was only detectable at low concentrators ( Figure 5E ).

Together, the data indicate that the effect of additional CD3ζ on cytotoxic activity varied substantially between the 3 different TCRs tested. The data also showed that the ability of CD3ζ to increase antigen-specific killing is particularly noticeable at low antigen concentration.

In a final set of experiments, we tested whether addition of CD3ζ in TCR transduced primary human T cells resulted in the upregulation of T cell exhaustion markers PD1, LAG3 and TIGIT. Human T cells were transduced with the CMV TCR without or with CD3ζ, and then stimulated at day 0 with the TCR-recognised CMV-peptide for 7 days, followed by another round of peptide stimulation for an additional 7 days. Flowcytomerty was used to determine PD1, LAG3 and TIGIT expression prior to peptide stiulation at day 0, and after each round of petide stimulation at day 7 and 14. The data in Figures 6A–C show the exhaustion marker analyses of gated CD8+ T cells, indicating that CD3ζ did not significantly alter the percentage of cells expressing PD1, LAG3 and TIGIT.