Antigen sensitivity of CD22-specific chimeric T cell receptors is modulated by target epitope distance from the cell membrane

Abstract

We have targeted CD22 as a novel tumor-associated antigen for recognition by human CTL genetically modified to express chimeric T cell receptors (cTCR) recognizing this surface molecule. CD22-specifc cTCR targeting different epitopes of the CD22 molecule promoted efficient lysis of target cells expressing high levels of CD22 with a maximum lytic potential that appeared to decrease as the distance of the target epitope from the target cell membrane increased. Targeting membrane-distal CD22 epitopes with cTCR⁺ CTL revealed defects in both degranulation and lytic granule targeting. CD22-specific cTCR⁺ CTL exhibited lower levels of maximum lysis and lower antigen sensitivity than CTL targeting CD20, which has a shorter extracellular domain than CD22. This diminished sensitivity was not a result of reduced avidity of antigen engagement, but instead reflected weaker signaling per triggered cTCR molecule when targeting membrane-distal epitopes of CD22. Both of these parameters were restored by targeting a ligand expressing the same epitope but constructed as a truncated CD22 molecule to approximate the length of a TCR:pMHC complex. The reduced sensitivity of CD22-specific cTCR⁺ CTL for antigen-induced triggering of effector functions has potential therapeutic applications, as such cells selectively lysed B cell lymphoma lines expressing high levels of CD22 but demonstrated minimal activity against autologous normal B cells, which express lower levels of CD22. Thus, our results demonstrate that cTCR signal strength – and consequently antigen sensitivity – can be modulated by differential choice of target epitopes with respect to distance from the cell membrane, allowing discrimination between targets with disparate antigen density.

Introduction

We have previously demonstrated that human CD8⁺ T cells can be modified by single chain antibody (scFv)-based chimeric T cell receptors (cTCR) to promote redirected lysis of CD20⁺ targets and that such cells have potential efficacy for human tumor therapy. (1–3). CD20 is expressed on a large majority of B cell malignancies (4, 5) and is a proven target for antibody mediated immunotherapy (6–10), but these same features may also limit its utility as a tumor associated antigen for recognition by cTCR⁺ CTL. The anti-CD20 monoclonal antibody rituximab is universally administered to patients with relapsed or refractory non-Hodgkin’s lymphoma and persists in the serum at high concentrations for many months following cessation of treatment, which may interfere with antigen recognition by CD20-specific cTCR⁺ CTL (8). Indeed, we previously showed that pre-treatment of patient lymphoma samples with clinically significant concentrations of the anti-CD20 mAb 1F5 abrogates target lysis by CD20-specific cTCR⁺ CTL in vitro (1). Additionally, normal B cells express high surface densities of CD20 (11) and although this can be overcome with antibody mediated immunotherapy by administering saturating doses of the mAb, this large amount of self-antigen may lead to deletion or anergy of adoptively transferred antigen-specific CTL (12).

To attempt to circumvent these limitations of targeting CD20 as a tumor associated antigen, we have investigated CD22 as an alternative antigen for recognition by cTCR⁺ CTL. CD22 is expressed on 60 – 70% of neoplastic B cells (13) and is detected at a lower copy number than CD20 on normal B cells (~30,000 CD22 vs. 100,000 – 150,000 CD20 molecules/cell) (14, 15), which might limit the induction of anergy or deletion by normal B cells depending on the antigen sensitivity of the responding T cell. Additionally, as CD20 and CD22 are distinct antigens, the possibility exists for combined immunotherapies with anti-CD20 mAb and adoptive transfer of CD22-specific CTL, potentially yielding additive or synergistic activities against B cell malignancies.

When targeting CD22, however, the large extracellular domain of this molecule must be considered. Whereas CD20 is a tetraspanin-like protein with a small extracellular domain (16, 17), CD22 is comprised of seven immunoglobulin-like domains that provide a number of membrane-distal epitopes that can be recognized by distinct mAbs (18, 19). Recent work has revealed that activation of canonical TCR chains is critically dependent on the size of the MHC ligand being recognized, with signaling attenuating sharply when the TCR:peptide-MHC (pMHC) ligand pair size exceeds wild-type dimensions (20). The mechanism underlying this phenomenon as proposed by the “kinetic segregation model” involves the inability of extended length T cell:target interaction sites to exclude the large extracellular domain-containing phosphatases CD45 and CD148 from the synaptic contact point, as normally occurs in the regions of tight T cell:APC membrane apposition generated by standard TCR:pMHC interactions (21). This loss of phosphatase exclusion can then lead to inefficient phosphorylation of the TCR complex and result in inefficient signaling (21). It has also recently been shown that chimeric immunoreceptors exhibit diminished signaling efficiency as the distance of the epitope from the target cell membrane increases – albeit to a lesser extent than seen with canonical TCR (22). These results suggest that the choice of epitope targeted on the CD22 molecule might influence signaling efficiency and potential therapeutic activity.

To better understand the impact on cTCR signaling of large cTCR:ligand pair sizes and to investigate targeting CD22 as a tumor associated antigen for CTL mediated immunotherapy, we generated two CD22-specific cTCR. One cTCR contains an engineered single chain chimeric mAb (scFv) that binds to the first immunoglobulin-like domain of the CD22 molecule (far from the cell membrane) while the second cTCR was constructed with an scFv recognizing the third immunoglobulin-like domain, which is situated closer to the cell surface. Our findings demonstrate that the signal strength delivered by CD22-specific cTCR is modulated by the distance of the target epitope from the target cell membrane, with diminished antigen sensitivity observed when targeting membrane-distal epitopes. As a result of this phenomenon, CD22-specific cTCR⁺ T cells are able to discriminate between targets expressing varying densities of CD22 when targeting this molecule at epitopes far from the target cell membrane – killing B cell lymphoma cell lines while demonstrating minimal lytic activity against normal B cells.

Materials and Methods

Antibodies

cTCR expression was analyzed with a PE-conjugated polyclonal goat anti-human Fcγ specific IgG (Jackson Immunoresearch) or with a FITC conjugated polyclonal goat anti-mouse F(ab′)₂ specific IgG (Sigma). CD20 and CD22 expression were analyzed with monoclonal mouse anti-human mAbs (BD Pharmigen): PE-anti-CD20 Quantibrite^™ (clone L27), PE-anti-CD22 (clone S-HCL-1), and PE-Cy5-anti-CD22 (clone HIB22). The RFB4 binding epitope of CD22 was detected with the PE conjugated mouse anti-human CD22 mAb clone RFB4 (Chemicon). cTCR were stimulated with polyclonal purified goat anti-human Fcγ specific IgG (Jackson Immunoresearch).

Cell lines and primary cells

The Daudi, Raji, and Ramos B cell tumor cell lines and the Jurkat T cell line were obtained from ATCC. The Phoenix G cell line was obtained from Gary Nolan (Standford University, CA). Human PBMC were obtained from normal donors and positively selected for the CD8 or the CD19 antigens using immunomagnetic selection to obtain CD8⁺ T cells and CD19⁺ B cells, respectively (Miltenyi). The HD39 hybridoma line was a gift from Edward Clark (University of Washington, Seattle, WA).

Molecular constructs

The CD20 and CD22 antigens were cloned by RT-PCR (One-step RT-PCR kit, Invitrogen) from total RNA obtained from PBMC (RNeasy, Qiagen). The Ig1-2 and Ig3-7;1-2 CD22-based ligands were generated by overlap extension PCR (PfuUltra™, Stratagene). The cDNAs encoding the HD39 variable light and heavy immunoglobulin domains were obtained by RT-PCR amplification of total RNA obtained from the HD39 hybridoma cell line as previously described (23, 24). The variable light and heavy immunoglobulin cDNA sequences were linked by overlap extension PCR with a DNA sequence encoding the peptide linker: GSTSGGGSGGGSGGGGSS. The RFB4 scFv template DNA was a gift from David Fitzgerald (NIH, Bethesda, MD). The HD39 and RFB4 scFv cDNAs were exchanged with the Leu16 scFv region of the Leu16 cTCR construct obtained from Michael Jensen (City of Hope, Duarte, CA) using overlap extension PCR. The β-interferon scaffold attachment region was placed 3′ of the cTCR gene in reverse orientation as previously described (25). Each DNA construct was cloned into the LZRS pBMN-z MMLV vector obtained from Gary Nolan (Standford University, CA).

Retroviral transduction and cell culture

The Phoenix G packaging cell line was transfected with LZRS pBMN vectors containing genes of interest using Lipofectamine 2000^™ (Invitrogen). Stable transfectants were obtained following selection in 2 μg/mL puromycin (Invitrogen) and cultured in 175 cm² culture flasks (Corning). Retroviral supernatants were obtained over two days and precipitated with polyethylene gycol (PEG, MW 8,000, Sigma). A 5x PEG solution was added to retroviral supernatant for a final concentration 8 g/dL and allowed to precipitate for 24 hours at 4° C followed by centrifugation at 1,500 × g for 1 hour at 4° C. The pellets containing concentrated retrovirus were resuspended in 500 μL of RPMI each. 3 × 10⁶ Jurkat T cells were transduced by spin-fection with 1 mL of PEG-concentrated retroviral supernatant containing 4 μg/mL polybrene (Sigma) for 1 hour at 32° C, 1,500 RPM. The transduction cultures were then incubated at 37° C for 3 hours, washed, and resuspended in culture medium. Jurkat T cell clones expressing varying densities of CD20 and CD22 were obtained by limiting dilution cloning following transduction. CD8⁺ T cells were stimulated with 30 ng/mL OKT anti-CD3ε (Ortho Biotech) and 50 U/mL rhIL-2 (Chiron) and at 72 hours, 3 × 10⁶ CD8⁺ T cells were transduced by the same procedure as for Jurkat T cells, with the addition of 50 U/mL rhIL-2. Primary human T cells were maintained by stimulation with irradiated PBMC and LCL and anti-CD3ε every 14 days as previously described (26). CTL expressing high densities of the cTCR were obtained FACS^® sorting or by positive immunomagnetic selection using PE-conjugated polyclonal goat anti-human Fcγ specific IgG followed by anti-PE magnetic beads (Miltenyi). Miltenyi MS columns were used to enrich cells bound by anti-PE magnetic beads according to the manufacturer’s instructions.

CTL degranulation Assay

2 × 10⁵ T cells were stimulated by either plate-bound anti-Fcγ antibodies (200 μg/mL) or by 2 × 10⁵ target cells (1:1 E:T ratio) in a volume of 200 μL RPMI for 3 – 4 hours at 37° C. Supernatants were obtained and analyzed using the N-α-benzyloxycarbonyl-l-lysine thiobenzyl ester (BLT) esterase assay as previously described (27). Benzyloxycarbonyl-l-lysine thiobenzyl ester and 5,5′-dithiobis(2-nitrobenzoate) were purchased from Sigma. Briefly, 50 μL of culture supernatant were added to 100 μL of a solution of BLT and 5,5′-dithiobis(2-nitrobenzoate) at room temperature and absorbance at 405 nm was measured at 30 minutes to determine the relative amount of esterase activity in the samples. Maximum release was determined by lysing T cells with 5 μL of 1% IGEPAL CA-630 (Sigma). Percent degranulation was determined by the following formula:

$$\%\text{degranulation}=[({\text{A}}_{405,\text{sample}}-{\text{A}}_{405,\text{non}-\text{stimulated}\text{T}\text{cells}})/({\text{A}}_{405,\text{lysed}\text{T}\text{cells}}-{\text{A}}_{405,\text{non}-\text{stimulated}\text{T}\text{cells}})]\ast 100$$

Chromium Release Assay

Chromium release assays were performed as previously described (1). Briefly, 2 × 10^{3 51}Cr-labeled targets were incubated with CTL in 200 uL of RPMI + 10% FCS for five hours, after which 30 μL of supernatant was assayed for the presence of radioactivity. Maximum release was determined by lysing target cells with 100 μL of 5% IGEPAL CA-630 (Sigma). The percentage of target lysis was calculated according to the following formula:

$$\%\text{Lysis}=[(\text{Experimental}{}^{51}\text{C}\text{r release }-\text{ Spontaneous }{}^{51}\text{C}\text{r release})/(\text{Maximum }{}^{51}\text{C}\text{r release}-\text{Spontaneous }{}^{51}\text{C}\text{r release})]\ast 100$$

cTCR and antigen quantitation

The number of CD20 and CD22 molecules expressed by Jurkat clones, tumor cell lines, and autologous B cells was estimated as antibody binding sites (ABS) using Quantibrite^™ beads (BD Pharmingen) and the PE labeled anti-CD20 mAb L27 and the PE labeled anti-CD22 mAb H-SCL-1 (BD Pharmingen) with known fluorescence to protein ratios at saturating concentrations according to the manufacturer’s instructions. The number of cTCR expressed by CD8⁺ T cells was estimated using FITC-conjugated Quantum^™ microbeads (Bangs Laboratories, Inc.) and saturating concentrations of FITC-conjugated polyclonal goat anti-mouse F(ab′)₂ specific IgG (Sigma) with known fluorescence to protein ratio.

Flow cytometric target lysis assay (7-AAD assay)

CTL were labeled with 10 μM DDAO-SE (Molecular Probes) and target cells were labeled with 1 μM CFSE (Molecular Probes) and incubated at a 30:1 effector to target ratio (3 × 10⁵ T cells; 1 × 10⁴ target cells) in a 96 well plate for four hours at 37° C. The cells were then resuspended in 7-AAD (Molecular Probes) at 2 μg/mL and CSFE⁺ DDAO⁻ 7-AAD⁺ dead target cells were enumerated by flow cytometry. The proportion of each dead target cell in the absence of T cells was subtracted from each sample value to obtain specific lysis values. CTL were regularly used on days 12 – 14 of 14 day restimulation cycles.

cTCR down-modulation assay

CTL were labeled with 10 μM DDAO-SE (Molecular Probes) and target cells were labeled with 1 μM CFSE (Molecular Probes) and incubated at a 1:1 effector to target ratio (2 × 10⁵ T cells; 2 × 10⁵ target cells) in a 96 well plate for four hours at 37° C. The cells were then resuspended in a saturating concentration of PE-conjugated polyclonal goat anti-human Fcγ specific IgG (anti-cTCR) and the mean fluorescence intensity (MFI) of cTCR staining on DDAO⁺ CFSE⁻ T cells was determined by flow cytometry. Percent cTCR down-modulation was calculated by the following formula:

$$\%\text{cTCR down}-\text{modulation}=100\ast [1-\text{stimulated MFI}/\text{unstimulated MFI}).$$

Statistical Analysis

Chromium release curves of cTCR⁺ CTL in response to Daudi lymphoma line cells were compared using a two-tailed, paired T-test using GraphPad Prism software.

Results

The maximum lytic efficiency of cTCR⁺ CTL correlates with the distance of the target antigen epitope from target cell membrane

We generated two CD22-specific chimeric T cell receptors based on our previously described CD20-specific cTCR (denoted Leu16), consisting of a single chain antibody (scFv) antigen recognition module, an IgG1 Fc spacer promoting dimerization, the CD4 transmembrane domain, and the CD3ζ cytoplasmic domain as the signal transducer (Fig. 1A). The scFv sequences for these cTCR were derived from the CD22-specific mAbs RFB4 and HD39 which by epitope mapping bind to the third and first immunoglobulin-like domains of CD22 respectively ((18, 28) and data not shown; Fig. 1B). Thus, the HD39 cTCR:CD22 ligand pair is expected to result in the largest distance between the T cell and target, followed by the intermediate sized RFB4 cTCR:CD22 ligand pair, and finally the relatively small Leu16 cTCR:CD20 ligand pair recognizing the short tetraspanin-like CD20 molecule.

Figure 1. Maximum lytic efficiency of cTCR⁺ CTL correlates with the distance of the target antigen epitope from target cell membrane.

(A) Schematic of cTCR structure. scFv: single chain antibody, IgG1 Fc: CH1 hinge and constant heavy domains 2 and 3 of IgG1, CD4TM: CD4 transmembrane. (B) Epitope mapping of HD39 and RFB4 scFv binding sites of CD22. CD20 is shown to scale. (C) Flow cytometric measurement of cTCR expression. cTCR or vector transduced CTL were stained with PE-labeled goat anti-human Fcγ-specific polyclonal IgG. Black solid: empty vector, Black line: Leu16, Grey line: RFB4, and Dashed line: HD39. (D) Target lysis assay using ⁵¹Cr-labeled CD20⁺CD22⁺ Daudi cell line as targets. The data are representative of 4 experiments. Points represent individual samples. * p < 0.05, *** p < 0.0001: paired T-test, two tailed. Dashed line: Leu16 vs. RFB4 (*), Solid black line: RFB4 vs. HD39 (***), Grey solid line: Leu16 vs. HD39 (***).

The three cTCR constructs were transduced with a retroviral vector into human CD8⁺ T cells and the resulting cell lines were selected for similar cTCR expression by fluorescence activated cell sorting (Fig. 1C). These cell lines were then tested for lytic activity against the CD20⁺ CD22⁺ B cell lymphoma line Daudi in a standard chromium release assay. The lytic activity of transduced CTL lines correlated with the size of the cTCR:ligand pair, with the Leu16 CD20-specific cTCR⁺ CTL line producing the highest maximal level of target lysis (~ 75%), followed by the RBF4 (~ 60%) and HD39 (~ 40%) CD22-specific cTCR⁺ CTL lines (Fig. 1D). This descending hierarchy of lysis was reproducible in four independent experiments, suggesting that the diminished lytic efficiency of the CD22-specific lines might result from the relatively large distances of their target antigen epitopes from the target cell membranes.

The diminished cytotoxicity of CD22-specific cTCR⁺ CTL is not due to insufficient target antigen density, insufficient cTCR expression density, or low affinity of scFv

While the differential target lysis induced by the three cTCR correlated with cTCR:ligand pair size, these differences could reflect other factors such as target antigen density or cTCR scFv affinity. To assess these possibilities, we compared the antigen sensitivity of target lysis induced by the three cTCR expressed at similar copy number (~10⁵ cTCR/cell) in response to Jurkat clones expressing varying densities of CD20 and CD22 (Fig. 2A). The numbers of CD20 or CD22 antibody binding sites (ABS) per Jurkat target cell clone were estimated using saturating concentrations of PE-labeled mAbs with known fluorescence to protein ratios, with the observed median fluorescence intensities compared to a standard curve constructed using microbeads possessing known numbers of conjugated PE molecules. The Leu16 CD20-specific cTCR⁺ CTL demonstrated half maximal lysis (EC₅₀) at ~15,000 ABS/cell, whereas the CD22-specific cTCR+ CTL demonstrated a lytic EC₅₀ of ~ 50,000 ABS/cell (Fig. 2B.1 vs. 2B.2 and 2B.3). The CTL expressing the different cTCR exhibited differential maximum target lysis of the Jurkat clones with Leu16 ~75%, RFB4 ~35%, and HD39 ~20% (Fig. 2B.1, 2B.2, and 2B.3, respectively). Thus, although the CD20-specific cTCR demonstrated a three-fold higher antigen sensitivity than the CD22-specific cTCR based on requirement for ABS/cell, a descending hierarchy of lytic potential remained evident at saturating antigen densities. This suggested that differential antigen sensitivity of the cTCR or insufficient antigen expression by target cell lines did not account for the differences in the maximum lytic efficiencies observed among the three cTCR.

Figure 2. Lysis defect of CD22-specific cTCR is not due to insufficient target antigen density, insufficient cTCR expression density, or low affinity of scFv.

(A) Flow cytometry of transduced CD22⁺ Jurkat clones with varying surface density. Cells were stained with PE-H-SCL-1 anti-CD22 mAb. Six clones out of 12 to 15 total per antigen type are shown. Similar results were obtained for CD20⁺ Jurkat clones. (B and C) Antigen sensitivity of target lysis and cTCR down-modulation. 7-AAD lysis assay and cTCR down-modulation assay were performed as described in Materials and Methods. The expression density of the RFB4 cTCR is five fold greater on the CTL line used in figure 2C compared with the RFB4 cTCR⁺ CTL used in 2B.2. Points represent the average of duplicate samples in this and all following dose-response experiments. (D) Relationship between antigen density and cTCR down-modulation. Numbers of cTCR expressed per T cell were estimated as described in Materials and Methods. Data from cTCR down-modulation assay were converted from percentages to absolute numbers of down-modulated cTCR. Numbers of cTCR down-modulated were plotted vs. numbers of CD20 or CD22 ABS (estimated as described in Materials and Methods) for antigen densities below the cTCR down-modulation inflection point. Data are representative of at least 3 experiments.

One potential explanation for the reduced lytic activity of the cells targeting CD22 is that the CD22-specific cTCR were expressed at insufficient copy numbers to promote lysis as efficiently as the CD20-specific cTCR. This could occur if the avidity of target antigen recognition by the CD22-specific cTCR was lower than that of the CD20-specific cTCR, requiring a greater cTCR expression density to induce maximal cTCR:Ag interaction and signaling according to the law of mass action as previously noted for canonical TCR (29, 30). Previous work has demonstrated that down-modulation of canonical TCR in response to strong pMHC agonists can be correlated with the sensitivity of antigen recognition and effector functions (31, 32). Furthermore, it has been demonstrated that low TCR expression density results in triggering too few receptors to promote maximal effector function (32). CD3ζ-based chimeric receptors have been shown to down-modulate similarly when triggered, suggesting cTCR down-modulation could also be used as a measure of T cell activation and antigen recognition sensitivity (33, 34). We therefore analyzed antigen induced cTCR down-modulation to compare the relative avidity of antigen recognition among the three cTCR⁺ CTL lines, selected for expression of similar numbers of cTCR (~10⁵ cTCR/cell). Similar to target lysis, cTCR down-modulation followed a sigmoid relationship with respect to antigen density for all three lines, with half-maximal cTCR down-modulation occurring at very similar antigen densities (Fig. 2B). This suggested that the functional avidity of antigen recognition was similar among the three cTCR. At ~10⁵ cTCR/cell, the target cell antigen density required for half-maximal target lysis induced by the Leu16 cTCR was lower than that required for half-maximal cTCR down-modulation, suggesting that maximum lysis induction required only a portion of the cTCR population to be triggered and down-modulated, similar to findings with canonical TCR (35). In contrast, the CD22-specific cTCR demonstrated overlapping target lysis and cTCR down-modulation curves, suggesting that activation of the all of the cTCR expressed by these CTL was required to induce the observed level of target lysis. These results implied that the CD22-specific cTCR were promoting lysis less efficiently than the Leu16 cTCR despite recognizing antigen and down-modulating similarly at similar target antigen densities. Therefore, CD22-specific cTCR expression density might be limiting maximal lysis, possibly due to diminished lysis per triggered TCR rather than insufficient cTCR affinity.

To test the hypothesis that insufficient cTCR expression density was limiting the maximum lysis induced by CD22-specific cTCR at saturating antigen densities, an RFB4 cTCR⁺ CTL line was generated that expressed five fold greater numbers of the cTCR (~5 × 10⁵ cTCR/cell) than the RFB4⁺ CTL lines used in previous experiments by sorting the transduced CTL shortly after transduction using an immunomagnetic selection method. Furthermore, we performed the analysis earlier in the restimulation cycle (day 7 – 9), when the LTR driven cTCR expression was higher than at later time points in the restimulation cycle when the previous experiments were performed (day 12 – 14, data not shown). The RFB4 cTCR⁺ CTL line expressing greater numbers of cTCR required more antigen molecules expressed per target cell to induce half-maximal cTCR down-modulation, however neither the antigen sensitivity nor the maximum efficiency of target lysis increased (Fig. 2B.2 vs. 2C). This finding suggests that cTCR expression density was not limiting the diminished lytic antigen sensitivity or maximum efficiency of the RBF4 cTCR since maximal target lysis occurred at a lower antigen density than was necessary to down-modulate all the cTCR.

We noted that the sigmoid relationship between antigen density and cTCR down-modulation can be approximated by an exponential relationship at values below the inflection-point on a logarithmic axis (data not shown). When plotting the number of cTCR down-modulated vs. antigen density on a linear axis, the relationship is linear and close to 1:1 for all three cTCR below the inflection point, i.e., where availability of cTCR is not limiting (cTCR down-modulated:Ag ratios: Leu16 = 0.61 ± 0.20, RFB4 = 1.1 ± 0.33, HD39 = 0.45 ± 0.13, average of three experiments) (Fig. 2D). This linearity has been noted by Viola et al. when triggering canonical TCR with high affinity anti-CD3 mAb (32) but contrasts with the logarithmic relationship between pMHC ligand density and αβ TCR triggering which reflects the phenomenon of serial triggering in this setting (with a single pMHC triggering as many as 200 TCR) (36). This implies that the observed impaired maximum lytic potential of the CD22-specific cTCR does not derive from insufficient antigen binding efficiency as all three cTCR recognize antigen and down-modulate with similar sensitivity. These results also suggest that cTCR do not serially engage antigen molecules, potentially accounting for relatively low extent of cTCR down-modulation and target lysis observed in response to targets expressing fewer than 10⁴ antigen equivalents.

Targeting membrane-distal epitopes produces both a degranulation and granule targeting defect

The previous results suggested that a signaling defect might result from the relatively large cTCR:ligand pair sizes of the CD22-specific cTCR. It remained possible, however, that targeting membrane-distal epitopes could impair efficient targeting of lytic granules onto target cells, resulting in diminished lysis. We therefore analyzed the antigen sensitivity and efficiency of cTCR-induced degranulation in response to CD20 and CD22. The Leu16 CD20-specific cTCR induced maximal degranulation at relatively low antigen density (Fig. 3). In contrast, the HD39 CD22-specific cTCR induced very limited degranulation evident only at high antigen densities, suggesting a signaling defect from this cTCR. An intermediate phenotype was observed with the RFB4 CD22-specific cTCR: the maximum degranulation achieved was similar to that observed with the Leu16 CD20-specific cTCR (RFB4 = 54%, Leu16 = 60%, average of three experiments), but the sensitivity was much lower, requiring an antigen density for half-maximal degranulation similar to that observed for the HD39 cTCR (Fig. 3). Thus, despite promoting degranulation nearly equivalently to the Leu16 cTCR at saturating antigen densities, the RFB4 cTCR promoted lysis much less efficiently (Leu16 = 71% vs. RFB4 = 39%; averaged over three or more experiments, Fig. 2B.1 vs. 2B.2). These data suggest targeting membrane-distal epitopes results in both a degranulation and a granule targeting defect.

Figure 3. Targeting membrane distal epitopes produces both a degranulation and granule targeting defect.

Degranulation of cTCR⁺ CTL in response to Jurkat clones expressing varying densities of CD20 or CD22. Degranulation was assessed by BLT-esterase activity as described in Materials and Methods. Data from three experiments are overlaid.

Targeting membrane-distal epitopes results in diminished signaling per triggered cTCR

The preceding data suggested that while the three cTCR recognize antigen and down-modulate with similar avidity, the strength of the signal produced by the Leu16 CD20-specific cTCR is greater than that generated by the CD22-specific cTCR responding to wild-type (WT) CD22. Plotting the percent degranulation vs. the number of cTCR down-modulated revealed that both the Leu16 CD20-specific cTCR and RFB4 CD22-specific cTCR responded similarly after the threshold for activation was reached, but that the signaling efficiency of the Leu16 cTCR was ten-fold greater per cTCR than the RFB4 cTCR (Fig. 4A). These data suggest that ~10⁴ Leu16 cTCR must be triggered to achieve maximum degranulation, a value similar to that reported to be required for maximum production of IFN-γ by CTL responding via the endogenous TCR chains to pMHC (32). In contrast, triggering of far more RFB4 cTCR was required to produce the same level of degranulation (~10⁵ cTCR, average of four experiments). Similar values were observed for target lysis (Fig. 4B). As the HD39 cTCR was only partially down-modulated in response to WT CD22 (Fig. 2B.3 and data not shown) we could not directly compare the relationship between degranulation and the number of cTCR down-modulated with the values obtained for the Leu16 and RFB4 cTCR. Thus, while the RFB4 cTCR could be down-modulated similarly and at high antigen density promoted degranulation equivalently to the Leu16 cTCR, the strength of signal delivered per cTCR appeared to be far weaker.

Figure 4. Targeting membrane distal epitopes results in diminished signaling per triggered cTCR.

cTCR down-modulation was determined in simultaneous experiments in which target lysis and degranulation were also measured. The number of cTCR expressed per CTL was estimated as described in Materials and Methods and converted to the number of cTCR down-modulated. Percent degranulation (A) or percent target lysis (B) in response to a particular antigen density was plotted against the number of cTCR down-modulated in response to the same antigen density. One representative experiment of three is shown.

Targeting a truncated CD22-based ligand restores sensitivity and efficiency of degranulation and target lysis

We next investigated the role of ligand size on the antigen sensitivity and efficiency of lysis and degranulation promoted by CD22-specific cTCR. Activation of TCR by immobilized antibodies has been shown to exclude the large extracellular domain-containing phosphatase CD148 from the T cell contact site, which is essential for efficient signaling. In contrast, soluble anti-TCR antibodies failed to promote exclusion, contributing to inefficient signaling (37). We first tested degranulation in response to stimulation of the RFB4 and HD39 cTCR with plate-bound Fcγ-specific polyclonal goat anti-human antibodies. Unlike what was observed with recognition of targets expressing CD22, the maximum values of degranulation achieved with the RFB4 and HD39 cTCR were similar, demonstrating that the HD39 cTCR could promote efficient signaling if properly triggered (Fig. 5A). To determine if the signaling defect observed following antigen recognition by the HD39 cTCR could be corrected if it was recognizing a smaller ligand, we next generated a truncated CD22 molecule in which the five membrane-proximal immunoglobulin-like domains were deleted and the molecule contained only the first two immunoglobulin-like domains and the full transmembrane and intracellular sequences of CD22 (Fig. 5B). This molecule, denoted Ig1-2, contains the HD39 epitope in the first Ig-like domain but lacks the RFB4 epitope in the third Ig-like domain of CD22. Jurkat T cells were transduced with either WT CD22 or Ig1-2 and unsorted populations expressing similar intermediate levels of each ligand were used to test RFB4 and HD39 cTCR⁺ CTL lines for degranulation. As expected, the RFB4 cTCR was unable to respond to the Ig1-2 molecule but induced degranulation in response to WT CD22. In contrast, the HD39 cTCR induced degranulation in response to the Ig1-2 molecule to an extent similar to that observed with RFB4 cTCR responding to WT CD22 but again induced little degranulation in response to full-length CD22 (Fig. 5C). To determine if targeting the truncated ligand improved the antigen sensitivity of degranulation compared to that observed when targeting WT CD22, we generated a panel of Jurkat T cell clones with varying expression densities of Ig1-2 as previously performed with CD20 and CD22 WT molecules. The HD39 cTCR⁺ CTL line degranulated to a much greater extent in response to the Ig1-2⁺ Jurkat clones compared to its response to the WT CD22⁺ clones, consistent with the hypothesis that ligand size modulates the sensitivity of signaling leading to degranulation (Fig. 5D). Target lysis was also enhanced, suggesting that reducing the ligand size also rescued the granule targeting defect observed when targeting membrane-distal epitopes of CD22. We found that the lytic antigen sensitivity and maximum lysis observed for the HD39 cTCR⁺ CTL line targeting Ig1-2 and the Leu16 cTCR⁺ CTL line targeting CD20 were very similar, suggesting that cTCR:ligand pair size and not an HD39 cTCR-intrinsic defect or the nature of the epitope being targeted was responsible for diminished signaling seen in response to WT CD22 (Fig. 5E). These data suggest that targeting membrane-distal epitopes diminishes the signal strength produced by cTCR, thereby diminishing the sensitivity of antigen recognition.

Figure 5. Targeting a truncated CD22-based ligand restores sensitivity and efficiency of degranulation and target lysis.

(A) RFB4 or HD39 cTCR⁺ or empty vector transduced CTL were incubated on goat anti-human Fcγ-specific IgG-coated plates (200 μg/mL) for two hours in 100 μL RPMI. Supernatants were obtained and degranulation was determined in a BLT esterase assay. Average and standard deviation of duplicate samples are shown and are representative results from one of three experiments. (B) Schematic of truncated CD22-based ligand Ig1-2. (C) Degranulation in response to unsorted transduced Jurkat cells expressing similar, intermediate densities of WT CD22 or Ig1-2. Average and standard deviation of duplicate samples are shown and are representative of two experiments. (D) Degranulation (BLT assay) or target lysis (7-AAD assay) by HD39 cTCR⁺ CTL in response to Jurkat clones expressing varying densities of WT CD22 or Ig1-2. Representative results from one of three experiments are shown. (E) Overlay of data from three experiments in which target lysis by Leu16 or HD39 cTCR⁺ CTL was analyzed in response to Jurkat clones expressing varying antigen densities of CD20 or CD22 WT and Ig1-2, respectively.

Membrane-distal target epitopes are not inhibitory in the presence of membrane-proximal ones

Previous experiments employing artificially elongated mutants of the adhesion molecule CD48 expressed on target cells demonstrated that increasing the CD48:CD2 ligand pair size inhibited interactions of the much smaller TCR:pMHC ligand pair, as evidenced by diminished TCR down-modulation and IL-2 production by T cells (38). This finding suggested that the artificially enlarged mutant CD48:CD2 ligand pair might inhibit TCR signaling by generating an enlarged effector:target intermembrane space, interfering with the coaptation of the smaller cTCR:ligand pairs. A similar phenomenon of signal inhibition has been observed with T cells targeting artificially elongated pMHC molecules. This inhibition was found to correlate with an increased intercellular distance between target and effector cell membrane (20). The signal inhibition in this setting is proposed by the kinetic segregation model to arise from lack of exclusion of large extracellular domain-containing phosphatases from TCR microclusters as a result of this increased intermembrane distance, which could result in diminished TCR phosphorylation and signal strength (21). Thus, previous work has suggested that close apposition of target and effector cell membranes is necessary to ensure robust signaling during antigen recognition. Consequently, simultaneously targeting small ligands in the presence of large ones might inhibit efficient signaling by preventing this close interaction.

CD22, like many other proteins, is expressed as long and short isoforms, which for CD22 differ by two Ig-like domains. The smaller isoform, however, represents less than one percent of total cellular CD22, suggesting that responses to the larger isoform will predominate (39). As the HD39 cTCR responded weakly to the large WT CD22 molecule and strongly to the truncated Ig1-2 molecule, we wished to model how a CTL expressing a cTCR would function if it encountered a target cell concurrently expressing a large and small isoform of the same protein. We selected a Jurkat clone demonstrating very high expression of the Ig1-2 truncated CD22 molecule, transduced it with WT CD22, and sorted for expression of the RFB4 epitope at very high intensity to obtain a Jurkat line that was simultaneously WTCD22^Hi and Ig1-2^Hi. This cell line, denoted WT:Ig1-2, was lysed by the RFB4 cTCR⁺ CTL line as efficiently as in previous experiments using CD22WT^Hi Jurkat clones (Fig. 6A vs. Figs. 2B.2, 2C, 4B), affirming high functional expression of the full length WT CD22 molecule. The HD39 cTCR⁺ CTL line lysed both the WT:Ig1-2 and the Ig1-2 Jurkat clone with similar efficiency, suggesting that the presence of a membrane-distal epitope is not necessarily inhibitory if the same epitope is also located proximally (Fig. 6A). Next we tested if efficient signaling could be inhibited by the presence of a large extracellular domain above a membrane-proximal epitope. This might sterically limit antigen recognition and possibly impede close effector:target membrane apposition by preventing the clearance of a large extracellular domain protein from the target:effector contact site, normally comprised by small extracellular domain proteins (21). We therefore generated a rearranged ligand, denoted Ig3-7;1-2, that contained the third through seventh Ig-like domains of CD22 followed by the first two Ig-like domains now placed proximal to the membrane, and the full transmembrane and intracellular sequences of CD22 (Fig. 6B). Jurkat T cells were transduced with this ligand and sorted for high expression of the RFB4 epitope equivalent to cells selected for high expression of WT CD22. Lysis of this cell line by the CD22-specific CTL was then analyzed. The RFB4 cTCR⁺ CTL line lysed the Ig3-7;1-2⁺ cell line half as efficiently as the WT CD22⁺ Jurkat cell line, consistent with the increased distance of the third Ig domain from the membrane (Fig. 6C). In contrast, the HD39 cTCR⁺ CTL lysed the Ig3-7;1-2⁺ cell line as efficiently as the Ig1-2⁺ cell line, suggesting that large extracellular domains above a membrane-proximal epitope are not inhibitory to efficient signaling. Our data suggest that membrane-distal epitopes or large extracellular domains above target ligand epitopes are not intrinsically inhibitory to efficient cTCR signaling provided that a membrane-proximal ligand is present. Therefore, targeting a membrane-proximal epitope on a tumor cell would not necessarily be impeded by the presence of a larger isoform on the same cell containing the epitope in a membrane-distal position.

Figure 6. Membrane distal target epitopes are not inhibitory in the presence of membrane proximal ones.

(A) Lysis of CD22 WT^Hi Ig1-2^Hi or Ig1-2^Hi Jurkat cell lines by RFB4, HD39, or empty vector transduced CTL. Lysis was analyzed by the 7-AAD lysis assay. (B) Schematic of WT CD22, Ig1-2, and rearranged Ig3-7;1-2 CD22-based ligand. (C) Target lysis of Jurkat cell lines expressing equivalent densities of WT CD22, Ig1-2, and Ig3-7;1-2 ligands by RFB4, HD39, or Leu16 cTCR⁺ CTL. Lysis was analyzed by the 7-AAD lysis assay. Data are representative results from one of three experiments. Average and standard deviation of duplicate samples are shown for (A) and (C).

Diminished lytic sensitivity of CD22-specific cTCR targeting WT CD22 permits selective lysis of tumor cell lines while demonstrating minimal activity against normal, autologous B cells

Previous investigations have demonstrated that CTL expressing low affinity cTCR (40) or demonstrating low cTCR expression densities (41) can promote selective lysis, discriminating between targets with disparate antigen densities. We hypothesized that the decreased antigen sensitivity of the CD22-specific cTCR responding to the membrane-distal epitopes in WT CD22 might permit selective lysis of tumor cells and not normal B cells, which express fewer CD22 than CD20 molecules (14, 15). We therefore tested the three cTCR⁺ CTL lines for lytic activity against the CD20⁺ CD22⁺ tumor cell lines Raji and Ramos and against autologous B cells selected for CD19-expression by immunomagnetic sorting. While the CD20-specfic line lysed both tumor lines and autologous B cells efficiently, the CD22-specific lines demonstrated minimal cytotoxicity against autologous B cells (Fig. 7A). The RFB4 cTCR⁺ line demonstrated intermediate lytic activity and lysed the Raji, Ramos, and Daudi tumor cell lines (~30%, ~40%, and ~60%, respectively, Fig. 7A and 1D), but mediated nearly undetectable lysis of autologous B cells (Fig. 7A). In contrast, the HD39 cTCR⁺ CTL line demonstrated minimal cytotoxicity toward autologous B cells and relatively low cytotoxicity against the Raji, Ramos, and Daudi tumor cell lines (~10%, ~15%, and ~40%, respectively, Fig. 7A and 1D).

Figure 7. Diminished lytic sensitivity of CD22-specific cTCR targeting WT CD22 permits selective lysis of tumor cell lines and weak responses to normal, autologous B cells.

(A) Target lysis of the B cell lymphoma lines Raji or Ramos and of CD19 positively selected autologous B cells by Leu16, RFB4, and HD39 cTCR⁺ CTL or empty vector transduced CTL was assessed by the 7-AAD assay. Average and standard deviation of triplicate samples shown and are representative of four experiments. (B) cTCR down-modulation in response to tumor cell lines or autologous B cells was assessed by the cTCR down-modulation assay. For comparison, the cTCR down-modulation dose-response curve generated by cTCR⁺ CTL responding to Jurkat clones expressing varying CD20 or CD22 expression densities is superimposed. Data are representative of two experiments.

To further examine the basis for this differential lysis by the RFB4 cTCR, we investigated cTCR down-modulation in response to tumor cell lines and autologous B cells as a function of target antigen density – as a measure of activation – and compared those values to dose-response curves for cTCR down-modulation generated using CD20- and CD22-transduced Jurkat clones. Normal autologous B cells expressed amounts of CD20 and CD22 very similar to those reported by others: 200,000 CD20 and 25,000 CD22 molecules/cell, respectively (Fig. 7B) (14, 15). Both tumor cell lines and autologous B cells maximally induced down-modulation of the Leu16 cTCR, presumably by virtue of the high expression of CD20 (normal B cells: ~200,000, Daudi: ~400,000, Raji: ~380,000, and Ramos: ~ 300,000 CD20 molecules/cell; average of two experiments, Fig. 7B). In contrast, autologous B cells, which express CD22 at a lower density, only modestly induced down-modulation of the CD22-specific cTCR (~35% RFB4, ~25% HD39 cTCR down-modulation). Tumor cell lines induced down-modulation of the CD22-specific cTCR to intermediate values (~45 – 60% RFB4, ~35 – 45% HD39), consistent with the greater lytic activity exhibited by the CD22-specific cTCR in response to lymphoma cells compared to autologous B cells and with the relatively higher CD22 expression on tumor cell lines (Daudi: ~ 85,000, Raji: ~ 80,000, Ramos: ~ 40,000 molecules/cell; average of two experiments, Fig. 7B). These data suggest that the RFB4 cTCR may allow selective lysis of CD22^Hi tumor cells in lymphoma patients without lysing normal B cells to a significant extent and furthermore might improve T cell persistence and function by limiting undesired interactions between cTCR⁺ CTL and normal B cells, which might lead to anergy or deletion.

Discussion

We have investigated for the first time the potential utility of the CD22 surface molecule as a therapeutic target for the treatment of malignant B cell neoplasms and have identified insights into antigen recognition by cTCR⁺ CTL that will likely aid the design of future cTCR constructs. CD22 can be targeted as a tumor associated antigen expressed by malignant B cell neoplasms, allowing efficient lysis of CD22^Hi targets by CD22-specific cTCR⁺ CTL, but these T cells show diminished antigen sensitivity relative to CD20-specifc cTCR⁺ CTL as a result of targeting membrane-distal epitopes of CD22. However, due to this diminished antigen sensitivity CD22-specific cTCR⁺ CTL preferentially lyse tumor cell lines overexpressing CD22, while reacting weakly to normal B cells.

We noted that as the size of the cTCR:ligand pair increased, the efficiency of target lysis decreased (Figs. 1D, 2B, and 7A). This is consistent with studies examining primary T cells and T cell hybridomas responding to pMHC ligands of varying size (20), in which ligands exceeding the native TCR:pMHC dimensions severely diminished signaling in primary T cells. Recently, a similar correlation was shown between cTCR:ligand pair size and signaling efficiency with cTCR targeting membrane-proximal or distal epitopes of carcinoembryonic antigen (22). However, the diminution in lytic potential was less pronounced in that system than in ours, suggesting that differential flexibility of target molecule extracellular domains may modulate this phenomenon. We have extended those previous findings with mechanistic studies demonstrating that targeting membrane-distal epitopes diminishes the strength of the signal delivered per cTCR and produces a granule targeting defect without affecting the avidity of antigen recognition leading to cTCR engagement and down-modulation. Taken together, these data indicate that both canonical TCR and chimeric TCR are subject to ligand size-based signal attenuation, and suggest that a similar mechanism may be responsible for this phenomenon.

The kinetic segregation model suggests that increased TCR:ligand pair size allows the large extracellular domain-containing phosphatases CD45 and CD148 to gain access to TCR microclusters during antigen recognition, dephosphorylating the TCR and preventing efficient signal transduction to occur (21). An implicit consequence of this model is that a greater number of TCR must be triggered to generate a given quantity of phosphorylated TCR and signal strength when targeting membrane-distal epitopes as compared to membrane-proximal ones. Far fewer Leu16 CD20-specific cTCR needed to be triggered and down-modulated than RFB4 CD22-specific cTCR in order to promote maximum degranulation or target lysis (Fig. 4), which is consistent with the higher antigen sensitivity of the CD20-specific cTCR (Fig. 2B.1 vs. 2B.2). Since both these cTCR recognize antigen and are down-modulated with similar avidity following antigen engagement (Figs. 2B and 2D), the RFB4 cTCR must deliver a weaker signal per triggered cTCR. Targeting membrane-distal epitopes therefore might generate signals similar to partial agonists (which are characterized by partial CD3ζ phosphorylation (31, 36, 42)), consistent with evidence that recognition of artificially elongated pMHC molecules produces deficient CD3ζ phosphorylation (20) and that diminished signaling is observed with ligands that poorly phosphorylate CD3ζ (31, 42). The HD39 cTCR demonstrated EC₅₀ values for lysis and degranulation nearly identical to those observed for the RFB4 cTCR (Figs. 2B, 3), suggesting the HD39 cTCR also promotes a weaker signal than the Leu16 cTCR. As the HD39 cTCR can promote efficient lysis and degranulation in response to cells expressing the same epitope located more proximal to the membrane (Figs. 5C, D, and E), the observed signaling defects do not reflect insufficient cTCR expression density or intrinsic signaling defects of this particular cTCR. Instead, the data imply that cTCR signaling strength, and as a result antigen sensitivity, is decreased as the target epitope distance from the target cell membrane increases.

Based on the kinetic segregation model, we hypothesized that membrane-distal epitopes might inhibit efficient signaling in the presence of membrane-proximal ones by competing for cTCR binding and by generating an enlarged effector:target intermembrane distance, permitting phosphatases to enter the formed microcluster and dephosphorylate activated molecules of the signaling cascade – similar to mechanisms proposed for antagonistic pMHC molecules (43–47). Moreover, membrane-distal epitopes might actually be preferentially engaged compared with proximal ones, as these epitopes may be more accessible. In contrast, we found that the HD39 cTCR promoted equivalent lysis of target cells expressing a truncated CD22-based ligand in the presence or absence of WT CD22 containing the epitope distal to the membrane (Fig. 6A), suggesting efficient cTCR recognition of the truncated ligand. We also found that the HD39 cTCR could promote equivalent lysis of targets expressing either the truncated Ig1-2 CD22 molecule or a rearranged CD22 molecule containing a membrane-proximal epitope with a large extracellular domain above it (Fig. 6C). This suggests that lack of exclusion of a protein with a large extracellular domain from effector:target contact points does not necessarily inhibit efficient signaling. Our results thus demonstrate that cTCR can produce efficient signaling if a membrane-proximal ligand is present, regardless of the presence of membrane-distal ones.

Our results could be consistent with the kinetic segregation model if cTCR could preferentially engage membrane-proximal ligands and exclude ligands with distal epitopes from regions of tight membrane apposition generated by this preferential engagement, similar to exclusion of molecules such as CD45 and CD148 during αβ TCR-mediated antigen recognition (37, 48). This process might be thermodynamically favored by allowing coaptation of the smaller, topologically equivalent adhesion molecule CD2 with CD58 (in humans) or with CD48 (in mice). A similar phenomenon has been observed with enhanced CD28:B7 interactions occurring in the presence of interactions of the equivalently sized CD2:CD48 molecules but not the larger LFA-1:ICAM-1 molecules (49). This suggests that a similar topological sorting might be possible for cTCR targeting differentially sized ligands. Even in the absence of CD2:CD58 or CD2:CD48 interactions, the high affinity of scFv cTCR may be sufficient to rivet the target and effector cell membranes together such that large cTCR:ligand pairs and phosphatases with large extracellular domains are excluded, generating a microenvironment conducive to phosphorylation and efficient signal transduction.

Our investigation of the effect of cTCR:ligand pair size on the antigen sensitivity and efficiency of degranulation and target lysis revealed that the RFB4 and the HD39 CD22-specific cTCR demonstrate a granule targeting defect when engaged by WT CD22. While both the Leu16 and RFB4 cTCR mediate maximal degranulation at saturating antigen densities (Fig. 3), the maximum lysis induced by this degranulation by the Leu16 CD20-specific cTCR at saturating antigen density is almost twice that achieved by the RFB4 CD22-specific cTCR. A similar lytic granule targeting defect has recently been described from blockade with mAb of LFA-1:ICAM-1 interactions (50), which had no effect on degranulation but rather interfered with granule polarization, resulting in a roughly 50% reduction in lysis. More recently, confocal microscopy revealed that blockade of LFA-1:ICAM-1 interactions prevented exclusion of CD45 from regions of TCR signaling (51), as reported with canonical TCR chains targeting epitopes engineered to be more membrane-distal than a classical pMHC complex (20). These results suggest that targeting membrane-distal epitopes with cTCR may preclude proper LFA-1:ICAM-1 interactions, resulting in a defect in granule polarization and targeting. Such lack of adhesive interactions may also inhibit tight apposition of CTL and target membranes, which normally occurs during target lysis by CTL (52). These data suggest that strategies to modify cTCR to enhance signal strength may still fail to completely overcome the diminished target lysis observed when cTCR⁺ CTL target membrane-distal epitopes.

Analysis of cTCR antigen sensitivity suggests that even when targeting ligands of optimal size, cTCR require very large numbers of antigen molecules (~ 10⁴ ligands) to achieve half-maximal lysis values (Figs. 2B.1 and 5E) compared to αβ TCR which can require fewer than ten agonistic pMHC/cell (53, 54). Our data suggest that, unlike αβ TCR, cTCR do not serially engage antigen molecules, likely due to the high affinity and low dissociation rates of scFv compared with αβ TCR (Fig. 2D). Serial engagement and triggering of αβ TCR by pMHC has been hypothesized to amplify signal strength in the presence of few pMHC ligands (32, 36), and consequently loss of serial triggering should diminish the sensitivity of responses such as lysis to target antigen expression. As serial triggering may reflect activation of ~ 200 TCR per a single pMHC (36), loss of serial engagement could theoretically result in a decrease in antigen sensitivity of two orders of magnitude. Previous work has suggested that loss of serial engagement resulting from long TCR:pMHC association times can lower antigen sensitivity (55, 56). These results suggest that the high affinity of our cTCR may limit their antigen sensitivity by abrogating serial triggering.

Finally, we have defined conditions under which CTL can be genetically modified to express cTCR recognizing CD22 and selectively target malignant B cells for lysis, while reacting weakly to normal B cells. The RFB4 cTCR, which binds an epitope at an intermediate distance from the target cell membrane relative to the Leu16 and HD39 cTCR, promoted relatively efficient lysis of tumor cell lines while not lysing autologous B cells to a great extent (Figs. 1D, 7A). The failure of cells expressing the RFB4 cTCR to respond strongly to normal B cells may mitigate induction of anergy or deletion of CTL expressing this receptor, allowing enhanced anti-tumor activity in lymphoma patients with high numbers of circulating normal B cells. Our data suggest that differential choice of epitopes may allow modulation of antigen sensitivity, and that as shown previously, a greater degree of discrimination between targets of disparate antigen density can be achieved by cTCR with lower sensitivity (40). The adhesion molecules NCAM, L1-CAM, E-selectin, VCAM-1, and N-cadherin and some cell surface receptors such as PDGFRα contain large extracellular domains and are overexpressed in many cancers originating in the lung, CNS, prostate, breast, ovary, kidney, colon, and other tissues (57–63). Therefore targeting membrane-distal epitopes of overexpressed adhesion molecules or other large extracellular domain-containing molecules with cTCR might represent a novel strategy to modulate cTCR sensitivity and promote selective lysis of tumors, while sparing healthy tissue.

Although this approach takes advantage of antigens that are expressed at higher levels in tumors compared to normal cells, we recognize that by lowering the antigen sensitivity and maximum lytic potential of cTCR⁺ CTL, the outgrowth of tumors expressing lower target antigen densities (i.e., similar to normal cells) might be promoted. In vivo studies will need to be performed to determine if attenuating reactivity to self-antigens abundant on tumor cells will provide sufficient advantage to a cTCR⁺ CTL to permit tumor eradication while circumventing deletion or anergy induced by antigen-expressing normal tissues, and therefore justify the diminished antigen sensitivity resulting from targeting membrane-distal epitopes on tumors. In conclusion, our results suggest that targeting surface antigens with engineered chimeric T cell receptors is subject to basic constraints imposed by the underlying biology of T cells and that a greater understanding of these factors may allow design of receptors that exploit the properties of cTCR to achieve selective and efficient lysis of malignant cells with relative preservation of healthy tissue. We are optimistic that such reagents may eventually improve the management of patients with recurrent B cell neoplasms and other malignancies.