Original Article

Gene Therapy (2011) 18, 62–72; doi:10.1038/gt.2010.127; published online 14 October 2010

CD28 cosignalling does not affect the activation threshold in a chimeric antigen receptor-redirected T-cell attack

M Chmielewski1, A A Hombach1 and H Abken1

1Zentrum für Molekulare Medizin Köln, and Tumorgenetik, Klinik I für Innere Medizin, Uniklinik Köln, Köln, Germany

Correspondence: Dr H Abken, Zentrum für Molekulare Medizin Köln, Uniklinik Köln, Robert-Koch-Str. 21, D-50927 Köln, Germany. E-mail: hinrich.abken@uk-koeln.de

Received 24 March 2010; Revised 5 August 2010; Accepted 6 August 2010; Published online 14 October 2010.



Adoptive immunotherapy of cancer using chimeric antigen receptor (CAR)-engineered T cells with redirected specificity showed efficacy in recent trials. In preclinical models, ‘second-generation’ CARs with CD28 costimulatory domain in addition to CD3ζ performed superior in redirecting T-cell effector functions and survival. Whereas CD28 costimulation sustains physiological T-cell receptor (TCR)–CD3 activation of naïve T cells, the impact of CD28 cosignalling on the threshold of CAR-mediated activation of pre-stimulated T cells without B7–CD28 recruitment remained unclear. Using CARs of different binding affinities, but same epitope specificity, we demonstrate that CD28 cosignalling neither lowered the antigen threshold nor the binding affinity for redirected T-cell activation. ‘Affinity ceiling’ above which increase in affinity does not increase T-cell activation was not altered. Accordingly, redirected tumor cell killing depended on the binding affinity but was likewise effective for CD3ζ and CD28–CD3ζ CARs. In contrast to CD3ζ, CD28–CD3ζ CAR-driven activation was not increased further by CD28–B7 engagement. However, CD28 cosignalling, which is required for interleukin-2 induction could not be replaced by high-affinity CD3ζ CAR binding or high-density antigen engagement. We conclude that CD28 CAR cosignalling does not alter the activation threshold but redirects T-cell effector functions.


immunotherapy; T cell; chimeric antigen receptor; CD28


APC, antigen-presenting cell; CAR, chimeric antigen receptor; IFN-γ, interferon-γ; IL-2, interleukin-2; MHC, major histocompatibility complex; PE, phycoerythrin; scFv, single-chain fragment of variable region; TCR, T-cell receptor; XTT, 2,3-bis(2-methoxy-4-nitro-5-sulphonyl)-5((phenyl-amino)carbonyl)-2H-tetrazolium hydroxide.



For use in adoptive immunotherapy, T cells can be redirected with predefined specificity by engineering with chimeric antigen receptors (CARs) that consist of an antibody-derived binding domain in the extracellular and a T-cell activation domain in the intracellular moiety.1 The binding domain consists of a single-chain fragment of variable region (scFv) antibody linked either directly or through an extracellular spacer to the intracellular CD3ζ chain. Although CD3ζ signalling alone is sufficient to initiate redirected activation of pre-stimulated T cells,2, 3 an additional costimulatory signal is required for full T-cell activation, prevention from activation-induced cell death and anergy. CD28 is the prototype of a family of costimulatory molecules that is physiologically engaged by binding to ligands on antigen-presenting cells (APCs). However, most therapeutically targeted cells, including tumor cells, do not express the CD28-activating ligands B7.1 (CD80) or B7.2 (CD86) that are physiologically expressed on APCs. To provide appropriate costimulation, the intracellular CD3ζ chain was covalently linked to the costimulatory domain of CD28 or to other members of the CD28 family including 4-1BB (CD137) and OX40 (CD134) in the same CAR molecule.4, 5, 6, 7, 8 This ‘second-generation’ CAR has the advantage that redirected T-cell activation is accompanied by suppressing inhibitory and/or strengthening stimulatory signals. Accordingly, Beecham et al.9 demonstrated that CD28 costimulation sustains survival and prolonged polyclonal expansion of CAR-engineered T cells. CAR-mediated CD28 cosignalling induces interleukin-2 (IL-2) that is used in an autocrine fashion by redirected T cells to increase their amplification without the need of B7–CD28 engagement.7 CD28–CD3ζ CAR signalling moreover counteracts transforming growth factor-β1-mediated repression in T-cell proliferation.10 Both sustained survival and increased amplification result in a prolonged redirected T-cell response and in an improved antitumor attack, making second-generation CARs favorable for clinical use.11, 12, 13

In physiological activation of naïve T cells, CD28 recruitment by B7 engagement sustains formation of the immunological synapse, thereby lowering the antigen threshold and increasing T-cell activation. In physiological T-cell activation, the activation efficiency correlates with the T-cell receptor (TCR) avidity to the cognate peptide–major histocompatibility complex (MHC) complex.14, 15, 16, 17, 18 Optimal CD28 costimulation is provided by high-avidity engagement by dimeric B7.1 expressed on APCs, followed by dimer dissociation that facilitates downregulaton of CD28 and B7.1 internalization.19 CD28 costimulation induces at least two independent activation pathways: one is integrated with TCR signalling in the context of synapse formation and is mediated through transcriptional enhancement and the second is mediated through increasing mRNA stability.20 CD28–CD3ζ CAR-redirected T-cell activation, however, differs from physiological TCR–CD28 activation in so far that the CAR delivers the costimulatory signal independently of B7 recruitment to the triggering synapse. Although CD3ζ CAR-mediated T-cell activation strongly depends on the affinity of the antibody-derived binding domain,21 it remains unclear whether CD28 cosignalling in second-generation CARs without B7 recruitment alters the affinity-dependent threshold in redirected T-cell activation. The issue is of major relevance with respect to the fact that most target antigens for redirected adoptive immunotherapy are not exclusively expressed on tumor cells but broadly on a variety of healthy tissues, although at lower levels. A lower activation threshold because of CAR CD28 cosignalling would negatively impact the selectivity of a redirected T-cell attack.

To address the impact of CD28 cosignalling, we redirected human peripheral blood T cells toward ErbB2 (Her2/neu) by CD28–CD3ζ and CD3ζ signalling CARs, respectively, both sets with different binding affinities but the same epitope specificity. Data revealed that CD28 cosignalling does not impact the activation threshold or ‘affinity ceiling’, above which an increase in affinity does not increase T-cell activation but T-cell-mediated effector functions. CD28 costimulatory CARs can increase cytokine secretion, but cannot increase sensitivity toward target cells expressing intermediate or low ErbB2 densities. Moreover, CD28 costimulation, which is required for IL-2 secretion, cannot be replaced by high-affinity binding or high antigen densities in the presence of a CD3ζ CAR. This is of major relevance for the adoptive immunotherapy of cancer, as lowering the affinity ceiling and antigen-dependent threshold of CAR-redirected T cells would have made healthy cells expressing low levels of antigen potential targets for T-cell attack, resulting in autoimmunity.



The second-generation CARs with intracellular CD3ζ fused to the CD28 costimulatory domain simultaneously deliver CD3ζ and CD28 signalling on antigen engagement. We addressed whether CD28 cosignalling impacts the affinity threshold of CAR-driven T-cell activation. We made use of a panel of scFvs targeting the same ErbB2 epitope with different affinities, that is, Kd=3.2 × 10−7 to Kd=1.5 × 10−11M (Table 1). Two sets of CARs with either CD3ζ or combined CD28–CD3ζ signalling domains were generated (Figure 1a) and retrovirally expressed human peripheral blood T cells (Figure 1b). Cells were pre-activated through CD3–IL-2 in order to allow retroviral transduction.

Figure 1.
Figure 1 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

Expression of recombinant anti-ErbB2 CARs on the surface of T cells. (a) Schematic diagrams depicting the modular composition of the anti-ErbB2 CARs with intracellular CD3ζ and combined CD28–CD3ζ signalling domains, respectively. For comparison, CD8–CD3ζ CARs were generated by replacing the CD28 signalling domain by the intracellular CD8 chain of same length. (b) Human peripheral blood T cells were engineered by retroviral gene transfer with the respective CARs. To record CAR expression, cells were stained with a PE-conjugated anti-CD3 antibody and a fluorescein isothiocyanate (FITC)-conjugated anti-IgG1 antibody that binds to the CAR extracellular IgG Fc domain and analyzed by flow cytometry. The CAR expression levels on transduced T cells, as indicated by mean fluorescence intensities, were nearly identical. The number of CAR-expressing cells was adjusted by titration with non-transduced T cells in each of the following experiments.

Full figure and legend (112K)

To confirm the redirected specificity of CAR-mediated activation, T cells genetically engineered with CARs of different specificities, that is, for ErbB2 and CD30, respectively, were coincubated with the respective target cells, that is, SK-OV-3 cells of ErbB2+ CD30 and MyLa cells of ErbB2 CD30+ phenotype (Figure 2a). Anti-ErbB2 CAR-engineered T cells were activated by SK-OV-3, but not by MyLa cells, indicated by dose-dependent increase in interferon-γ (IFN-γ) secretion and lysis of SK-OV-3 cells, whereas MyLa cells were not lysed (Figure 2b). Vice versa, T cells with the anti-CD30 CAR were activated by MyLa, but not by SK-OV-3 cells.

Figure 2.
Figure 2 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

CAR-redirected T-cell activation is antigen specific. (a) Expression of ErbB2 and CD30 on the surface of target cells. SK-OV-3 and MyLa tumor cells were analyzed for ErbB2 and CD30 expression using the anti-ErbB2 and the anti-CD30 antibody HRS3 (bold lines), respectively. An isotype-matched antibody served as control (thin lines). Bound antibodies were detected by fluorescein isothiocyanate (FITC)-conjugated anti-mouse IgG antibody and recorded by flow cytometry. (b) CD3+ T cells were engineered with the anti-ErbB2 CAR C6-B1D2-scFv-Fc-CD3ζ (no. 714) and the anti-CD30 CAR HRS3-scFv-Fc-CD28–CD3ζ (no. 926), respectively, adjusted to the same number of CAR-expressing cells and cocultivated (0.078–1 × 104 receptor-grafted T cells per well) with ErbB2+ CD30 SK-OV-3 and ErbB2 CD30+ MyLa cells (5 × 104 cells per well), respectively. IFN-γ in the culture supernatants was determined by enzyme-linked immunosorbent assay (ELISA), and viability of tumor cells was determined by the XTT-based viability assay.

Full figure and legend (96K)

To address the affinity-dependent T-cell activation in the presence of CD28 cosignalling, T cells were engineered with the panel of anti-ErbB2 CARs of different affinities, each triggering through the CD3ζ and the combined CD28-CD3ζ domains, respectively. On coincubation with ErbB2high SK-OV-3 cells, IFN-γ secretion was substantially enhanced in T cells with CD3ζ–CD28 cosignalling CAR compared with CD3ζ-only CAR (Figure 3a). Increase in binding affinity above Kd=1.6 × 10−8M, however, did not further increase IFN-γ secretion. ‘Affinity ceiling’ was found for both CD3ζ and CD28–CD3ζ CAR-redirected T-cell activation; noteworthy, the ‘ceiling affinity’ was the same for CD3ζ and CD28–CD3ζ signalling CARs, as summarized in Figure 3b.

Figure 3.
Figure 3 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

CD28 costimulation does not impact the affinity-dependent threshold in T-cell activation. (a) T cells were engineered with anti-ErbB2 CARs of different affinities and with CD3ζ (upper panel) and CD28–CD3ζ signalling domains (lower panel), respectively. Engineered T cells (0.6 × 103–10 × 103 CAR-expressing T cells per well) were coincubated with ErbB2+ SK-OV-3 cells (5 × 104 cells per well) for 48h. IFN-γ and IL-2 in the culture supernatants were determined by enzyme-linked immunosorbent assay (ELISA), and viability of SK-OV-3 cells was recorded by XTT-based viability assay. (b) IFN-γ and IL-2 concentrations in the culture supernatants of cocultures with 5 × 103 CAR T cells and the viability of SK-OV-3 cells on coincubation with 0.5 × 103 CAR T cells were plotted against Kd of the respective binding domain. Data were transformed from (a).

Full figure and legend (129K)

To address the impact of CD28 cosignalling on T-cell-mediated cytolysis, the viability of SK-OV-3 target cells was monitored in the presence of engineered cytolytic T cells. T cells triggered by CARs with higher binding affinities, that is, Kd <10−8M, lysed SK-OV-3 cells in a similar fashion, whereas T cells triggered by the low-affinity CAR executed cytolysis with lower efficiency (Figure 3a). However, the cytolytic efficacy was not substantially affected by CD28 costimulation.

CD28–CD3ζ CAR-triggered T-cell activation was accompanied by IL-2 secretion that was not recorded on CD3ζ CAR signalling, even not on high-affinity binding (Figure 3a). We conclude that high-affinity binding of CD3ζ CARs does not substitute for CD28 costimulation with respect to the induction of IL-2 secretion. High-affinity binding in the presence of CD28 cosignalling, however, did not further increase the amount of secreted IL-2. ‘Affinity ceiling’ occurred at the same threshold as observed for the induction of IFN-γ secretion, that is, Kd=10−8M. Taken together (Figure 3b), the data demonstrate that inclusion of the CD28 cosignalling domain, although necessary for induction of IL-2, enhanced IFN-γ release but did not substantially affect cytotoxicity. Noteworthy, CD28 did not alter the affinity threshold or ‘affinity ceiling’ in redirected T-cell activation compared with CD3ζ signalling only.

To rule out whether the altered CD3ζ position in the CD28–CD3ζ CAR compared with the CD3ζ CAR is responsible for IL-2 induction, we replaced the CD28 by the intracellular CD8 domain of same length. In contrast to CD28–CD3ζ CAR, CD8–CD3ζ CAR-redirected T cells did not produce IL-2 on engagement of ErbB2+ SK-OV-3 cells (Figure 4). The data indicate that CD28 cosignalling, and not the more membrane distal position of CD3ζ, is responsible for the induction of IL-2 secretion. Our conclusion is confirmed by the observation that T cells triggered by the mutated CD28Δ–CD3ζ CAR, in which the Lck-binding motif is deleted by two-point mutations, showed no IL-2 secretion, whereas IFN-γ secretion was not substantially altered (data not shown).

Figure 4.
Figure 4 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

CD28 cosignalling and not the altered position of CD3ζ in a CD28–CD3ζ CAR induces IL-2 secretion. T cells were engineered to express the anti-ErbB2 CAR with the CD3ζ, CD28–CD3ζ and CD8–CD3ζ signalling domains, respectively. Mock engineered T cells served as controls (w/o). CAR T cells (1 × 104) were coincubated for 48h with SK-OV-3 cells (2.5 × 104 cells), and IFN-γ and IL-2 were monitored by enzyme-linked immunosorbent assay (ELISA) in the culture supernatants.

Full figure and legend (38K)

In addition to the affinity, the amount of antigen has substantial impact on redirected T-cell activation, which may be affected by CD28 CAR cosignalling. On engineering with anti-ErbB2 CARs with or without CD28 and with high- or low-binding affinity, respectively, T cells were incubated on plates coated with increasing amounts of ErbB2. T cells triggered by high-affinity CARs required lower antigen amounts to induce IFN-γ compared with T cells with low-affinity CARs (Figure 5a). T cells with CD28–CD3ζ CARs, however, required the same antigen amount for activation as T cells with CD3ζ CARs, indicating that CD28 costimulation has no substantial impact on the antigen threshold for redirected T-cell activation.

Figure 5.
Figure 5 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

The antigen threshold for CAR-triggered T-cell activation is not altered by CD28 costimulation. (a) T cells with anti-ErbB2 CARs of different affinities (Kd=1.2 × 10−10 and 1.6 × 10−8M, respectively), and with CD3ζ or CD28–CD3ζ signalling domains, respectively, were incubated for 48h on plates coated with increasing amounts of ErbB2-Fc protein. IFN-γ in the supernatants was determined by enzyme-linked immunosorbent assay (ELISA). The amount of ErbB2 (concentration in the coating solution) was plotted against the amount of IFN-γ (percentage of maximum concentration) in the culture supernatant. T-cell activation by immobilized antigen was dependent on the binding affinity and the amount of antigen, both of which not altered by CD28 costimulation. (b) SK-OV-3, LS174T and Colo320 cells express different amounts of ErbB2; MyLa cells lack ErbB2 as recorded by flow cytometry. T cells were engineered with ErbB2-specific CARs of different affinities and with CD3ζ (lower panel) and CD28–CD3ζ signalling domains (upper panel), respectively. T cells (0.078–1 × 104 CAR T cells per well) were coincubated with the respective target cells (5 × 104 cells per well) for 48h. T cells without CAR (w/o) served as additional control. IFN-γ and IL-2 in the culture supernatants were determined by ELISA, and the viability of target cells was determined by XTT-based viability assay. (c) Data from (b) were plotted against the amount of ErbB2 (mean fluorescence intensity) expressed on target cells.

Full figure and legend (101K)

To address the situation when T cells encounter antigen on the surface of target cells, engineered T cells were coincubated with ErbB2high SK-OV-3 cells, ErbB2medium LS174T cells and ErbB2low Colo320 cells. Coincubations with MyLa cells lacking ErbB2 served as controls. On coincubation with ErbB2medium target cells, more T cells were required to induce same amounts of IFN-γ compared with incubation with ErbB2high cells (Figure 5b), indicating less efficient T-cell activation by lower amounts of antigen. Again, T-cell activation by a CD28–CD3ζ CAR required the same amount of ErbB2 as CD3ζ CAR-redirected T cells. On binding to low antigen cells like Colo320, CD28–CD3ζ costimulation did not induce IFN-γ secretion, even not on high-affinity binding. Coincubation of engineered T cells with ErbB2 MyLa cells did not induce a T-cell response, once again indicating that the specificity in CAR-mediated T-cell activation is not altered by high-affinity binding or CD28 cosignalling. Taken together (Figure 5c), the data indicate that there is a clear correlation between binding affinity and the amount of antigen required to induce T-cell activation. CD28 cosignalling, however, does not impact the antigen threshold required to initiate T-cell activation, independently of whether T cells engage antigen through high- or low-affinity binding. IL-2 induction required CD28 cosignalling that cannot be substituted by high-affinity binding or high amounts of antigen.

As CD28–CD3ζ signalling CARs bypass the requirement to recruit CD28–B7 complexes into the triggering synapse, we asked whether physical cell–cell interactions through B7–CD28 have additional impact on CAR-redirected T-cell activation. ErbB2+ LS174T tumor cells were engineered with B7-1 and B7-2 (Figure 6a). T cells were engineered with the CD3ζ or CD28–CD3ζ CAR, respectively (Figure 6b), and coincubated either with ErbB2+ B7 LS174T or with engineered ErbB2+ B7-1+ B7-2+ LS174T cells. CD3ζ CAR-driven T-cell activation was increased on binding to B7+ LS174T cells compared with LS174T lacking B7; however, CD28–CD3ζ CAR-driven T-cell activation, was not increased further (Figure 6c). The data indicate that additional B7–CD28 engagement has impact on CD3ζ CAR, but not on CD28–CD3ζ CAR-triggered T-cell activation.

Figure 6.
Figure 6 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

Recruitment of CD28–B7 increases CD3ζ but not CD28–CD3ζ CAR-triggered cytokine secretion. (a) LS174T and LS174T-B7 express similar high levels of ErbB2 and (b) T cells express ErbB2-specific CARs with CD28–CD3ζ and CD3ζ domains, respectively, in similar densities, as recorded by flow cytometry. (c) T cells engineered with the respective ErbB2-specific CAR (2 × 104cells) were coincubated for 48h with ErbB2+ LS174T cells without B7 (open symbols) and with B7-1 and B7-2 (closed symbols) (each 2 × 104cells), respectively. IFN-γ in the culture supernatants was recorded by enzyme-linked immunosorbent assay (ELISA). Engineered T cells without LS147T cell coincubation (w/o) served as controls.

Full figure and legend (128K)

To explore the cytolytic activity of CD3ζ and CD28–CD3ζ CAR-redirected T cells, we engineered T cells with CARs of high and low affinity (Kd=1.5 × 10−11 and 3.2 × 10−7M, respectively), both with CD3ζ and CD28–CD3ζ signalling domains, respectively. ErbB2+ tumors were induced in Rag-2-/-cγ--/- mice by subcutaneous transplantation of LS174T tumor cells. When tumors were established to a size of approximately 40–60mm3, engineered T cells were peritumorally applied in order to make sure that the same number of engineered T cells are in or near the tumor lesion. As shown in Figure 7, tumor growth was more efficiently repressed by high-affinity than by low-affinity CAR T cells. The same performance, however, was found for CD3ζ and CD28–CD3ζ CAR-redirected T cells; in particular, the redirected tumor cytolytic response by the low-affinity CAR was not increased by CD28 costimulation. This is in accordance to our observation that the cytolytic capacity of redirected T cells in vitro was not altered by CD28 costimulation (cf Figure 3a). Although this experimental outline excludes secondary factors like T-cell persistence, migration, proliferation and prevention of activation-induced cell death in recording T-cell-mediated tumor cell lysis, the long-term efficacy of an antitumor response is dependent on these factors and substantially modulated by CD28 costimulation.23, 24

Figure 7.
Figure 7 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

Tumor repression by engineered T cells depends on the binding affinity and is not altered by CD28 costimulation. ErbB2-transfected MC38 tumor cells were subcutaneously injected (0.8 × 106 tumor cells per mouse) into Rag-2/γc/ mice (six mice per group). Tumors were grown to a size of 40–60mm3. T cells (2 × 106) with ErbB2-specific CARs of different affinities (1.5 × 10−11 and 3.2 × 10−7M, respectively) and with CD28–CD3ζ and CD3ζ domains, respectively, were peritumorally injected at day 0. Tumor progression was daily monitored for up to 40 days. Tumor volumes were compared using one-tailed t-test.

Full figure and legend (57K)



In the adoptive immunotherapy of cancer, second-generation costimulatory CARs are currently entering clinical trials; however, the impact of CD28 cosignalling without B7 engagement on the threshold in CAR-redirected T-cell activation was so far little explored. Although B7–CD28 engagement lowers the threshold in physiological TCR–CD3 activation of naive T cells, we here revealed that CD28 cosignalling by a chimeric antigen receptor does not affect the antigen and affinity-dependent threshold in T-cell activation. This is particularly obvious in the situation of low-affinity antigen binding and of binding to low amounts of antigen. Our data are based on a panel of CARs targeting the same epitope of ErbB2 with different affinities, that is, from 3.2 × 10−7 to 1.5 × 10−11M. Redirected T-cell activation is more efficient by higher than by low-affinity binding; however, there is an ‘affinity ceiling’ of approximately 10−8M, at least in our example of anti-ErbB2 CARs, above which increase in affinity does not increase redirected T-cell activation. We previously reported ‘affinity ceiling’ for CD3ζ CAR-redirected T cells at the same affinity;21 therefore, it was unexpected that CD28 does not substantially impact the binding affinity-dependent T-cell activation. The observation has substantial consequences for the redirected immunotherapy of cancer, as a number of tumor-associated antigens used for redirected targeting are also expressed in healthy tissues, although at lower levels. A lower activation threshold due to CD28 cosignalling would have negatively impacted the selectivity of a CAR-redirected T-cell attack and increased the risk of autoimmunity. In vitro data were confirmed by CAR-redirected tumor cytolysis in vivo that was dependent on the binding affinity and not altered by CD28 costimulation (cf Figure 7). Although we here recorded short-term tumor cytolysis after administration of engineered T cells in the vicinity of the tumor, secondary parameters additionally impact the overall outcome in a long-term anti-tumor attack compared with CD3ζ CAR-driven T-cell activation. For instance, CD28 costimulation increases the expression of B-cell lymphoma 2 protein, preventing activation-induced cell death and anergy and thereby improving the long-term antitumor attack.25

CD28 cosignalling impacts the cytokine response of redirected T cells. IFN-γ is increased and IL-2 secretion is induced by CD28 cosignalling.6, 26 The altered pattern in cytokine secretion in the presence of CD28 cosignalling is not due to an altered CD3ζ activity as a consequence of the more distal membrane position in the CD28–CD3ζ CAR, as the CD8-CD3ζ CAR, which harbors the CD8 intracellular domain of same length in place of CD28, did not induce IL-2 or increase IFN-γ secretion. High-affinity CD3ζ CAR binding or high amounts of antigen cannot replace CD28 cosignalling for IL-2 induction.

In accordance with our data, Willemsen et al.27 previously reported that engagement of a MHC class I-restricted CAR with γ signalling domain increased cytokine secretion when combined with CD28 costimulation. A CAR targeting the same antigen with higher affinity showed increased cytokine secretion. Cytolysis by γ-CD28 CAR-modified T cells was improved compared with γ CAR T cells when target cells were loaded with small amounts of peptide antigen. Increasing antigen amounts, however, decreased the difference in cytolytic activities of γ versus γ-CD28 CAR T cells, in particular of the high-affinity CAR. Compared with MHC-loaded peptide antigen, the MHC-independently targeted antigen in our study is present in substantially higher amounts on the cell surface, which may explain the difference in the observations.

In addition to the binding affinity, the position of targeted epitope within the cognate antigen has major impact on the efficacy of CAR-driven T-cell activation.28 We previously explored the situation by redirecting T cells against the membrane distal and the proximal domain of carcinoembryonic antigen. The membrane-proximal carcinoembryonic antigen epitope more efficiently mediates CAR-redirected T-cell activation than the distal epitope. Targeting the distal epitope in a more membrane-proximal position in a modified carcinoembryonic antigen molecule, however, increased activation of engineered T cells. We conclude that the epitope position within the targeted molecule, in addition to the binding affinity, has an impact on the efficacy of T-cell activation.

Although a CD28–CD3ζ CAR provides CD28 cosignalling in the absence of B7–CD28 recruitment, additional CD28–B7 engagement did not further alter T-cell activation. In the presence of a CD3ζ CAR, however, B7 engagement increased IFN-γ secretion of redirected T cells, indicating that the physiological CD28 costimulation cooperates with CAR-driven T-cell activation. There are, however, several differences in the TCR–CD28 and CAR-driven T-cell activation. The avidity of physiological TCR–MHC interactions is much lower than those of the antibody domains used in CAR binding. Another caveat is the activation of naïve T cells, whereas CAR-modified T cells are pre-stimulated to allow retroviral engineering. On physiological T-cell–APC interactions, CD28–B7 binding potentates synapse formation by increasing TCR–CD3 density through approximation of the interacting membranes.29, 30, 31, 32 Signal amplification through receptor clustering results in the integration of the TCR with costimulatory signalling that becomes obvious when compensating for a weak TCR signal.33 Early events in T-cell activation are likely to require B7–CD28 engagement that helps to form the synapse, with the result that low-affinity ligands can also successfully initiate T-cell signalling.32 This implies that in low-avidity, T cell induction of T-cell activation is dependent on the extent of supramolecular clustering. The TCR binding threshold exhibits a relatively sharp cutoff between full activity and no activity that seems to be marginally modulated by CD8. Moreover, the peptide–MHC has to be presented in a form that allows the appropriate synapse formation on the T cell in order to initiate downstream signalling including IL-2 release.15 Costimulatory molecules are thought to increase stability during early stages in this process. Optimal CD28 costimulation requires high-avidity engagement by dimeric B7.1 on APCs, followed by dimer dissociation that facilitates CD28 downregulation and B7.1 internalization.19 TCR and CD28 costimulation downmodulates the TCR-induced cyclic AMP-mediated inhibitory signals through recruitment of a β-arrestin–phosphodiesterase 4 complex, leading to cyclic AMP degradation and thus allowing a full T-cell response to occur.34 In contrast to physiological TCR–CD28, the CD28–CD3ζ CAR drives T-cell activation independently of CD28 recruitment into the triggering synapse and provides CD28 cosignalling without the need of B7.1 engagement. The artificial fusion of the CD28 and CD3ζ signalling domains in a CAR facilitates Lck-mediated CD28 phosphorylation that binds and activates phosphatidylinositol 3-kinase for downstream signalling, resulting in IL-2 release.


Materials and methods

Cell lines and reagents

293T cells are human embryonic kidney cells that express the SV40 large T antigen.35 SK-OV-3 (ATCC (American Type Culture Collection) HTB77, Manassas, VA, USA), LS174T (ATCC CL 188)36 and Colo320 (ATCC CCL 220.1) are derived from various adenocarcinomas expressing different amounts of ErbB2. LS174T-B7 cells were obtained by transfection of LS174T cells with an expression vector for both B7-1 and B7-2.6 MyLa are CD30+ ErbB2 cutaneous lymphoma cells kindly provided by Dr K Kaltoft (University of Aarhus, Denmark). The MC38 murine tumor cell line was kindly provided by Dr M Neumaier (University Hospital Eppendorf, Hamburg, Germany) and transfected to express human ErbB2. OKT3 (ATCC CRL 8001) is a hybridoma cell line that produces the anti-CD3 monoclonal antibody OKT3. 293T cells were cultured in Dulbecco's modied Eagle's medium supplemented with 10% (v/v) fetal calf serum; all other cell lines were cultured in RPMI-1640 medium, 10% (v/v) fetal calf serum (all from Life Technologies, Paisly, UK). OKT3 antibody was affinity purified from hybridoma supernatants utilizing goat anti-mouse IgG2a antibodies (Southern Biotech, Birmingham, AL, USA) that were immobilized on N-hydroxy-succinimid-ester-activated sepharose as recommended by the manufacturer (Amersham Biosciences, Freiburg, Germany). C6-B1D2scfv-hIgG fusion protein was affinity purified from supernatants of transfected 293T cells utilizing the human IgG1-specific monoclonal antibody 4E3 (Southern Biotech) that has been immobilized on N-hydroxy-succinimid-activated sepharose (Amersham Biosciences). The phycoerythrin (PE)-conjugated anti-human CD3 antibody UCHT1 was purchased from Dako (Hamburg, Germany). The PE-conjugated goat anti-human IgG antibody F(antibody’)2 and the PE-conjugated goat anti-mouse IgG polyclonal sera were purchased from Southern Biotechnology. The mouse anti-human ErbB2 antibody Antibody-5 was purchased from Oncogene (Cambridge, MA, USA), the recombinant ErbB2-IgG Fc fusion protein from R&D Systems (Minneapolis, MN, USA) and the PE-conjugated mouse anti-human CD30 antibody Ber-H2 from Dako. The anti-human IFN-γ antibody NIB42, the biotinylated anti-human IFN-γ antibody 4S.B3, the anti-human IL-2 antibody 5344-111 and the biotinylated anti-human IL-2 antibody B33-2 were purchased from BD Bioscience (San Jose, CA, USA). The generation of the anti-ErbB2 scFv antibodies has been described in detail.37 The binding parameters of the anti-ErbB2 scFvs are summarized in Table 1.

Generation of CD8 cDNA

mRNA was prepared from T cells using ‘RNeasy Mini Kit’ following the manufacturer's instructions, and first-strand complementary DNA (cDNA) was synthesized using ‘Oligo-dT and Omniscript RT Kit’ (all from Qiagen, GmbH, Hilden, Germany). The CD8 cDNA was synthesized by PCR amplification using the primer oligonucleotides, 5′-ANTIGENCGATGGATCCATCTACATCTGGGCGCCCTTGGCC-3′ (sense) and 5′-GTCTGCGCTCCTGCTGAACTTCACTCTGACGTATCTCGCCGAAANTIGENGCTGGGCTT-3′ (antisense), thereby incorporating the BamHI restriction site.

Generation of recombinant CARs

The generation of the ErbB2-specific anti-ErbB2-scFv-Fc-CD3ζ CARs with the CD3ζ signalling domain has been described in detail.21 To generate the retroviral expression cassettes for the anti-ErbB2-scFv-Fc-CD28–CD3ζ CARs with combined CD28–CD3ζ signalling domain, the C6.5 scFv DNA and its derivatives (kindly provided by Dr J Marks, UCSF) (Table 1) were amplified by PCR using the following set of primer oligonucleotides (NcoI and BamHI restriction sites are underlined): 5′-CGTACCATGGATTTTGANTIGENGTGCANTIGENATTTTCANTIGENCTTCCTGCTAAT
CANTIGENTGCCTCAGTCATAATGTCTANTIGENACCGGCGATGGCCCANTIGENGTG-3′ (sense) and 5′-TTCTGGATCCGCACCTANTIGENGACGGTGACCTT-3′ (antisense). The amplified anti-ErbB2 scFv DNA replaced the scFv DNA of the BW431/26scFv-Fc-CD28–CD3ζ CAR, which was previously described.6 To generate CD8-containing CARs, the CD8 cDNA was assembled together with the intracellular CD3ζ cDNA and amplified by PCR using the primer oligonucleotides, 5′-ANTIGENCGATGGATCCATCTACATCTGGGCGCCCTTGGCC-3′ (sense) and 5′-TATCTTCTCGANTIGENTTAGCGANTIGENGGGGCANTIGENGGCCTGC-3′ (antisense), thereby incorporating the BamHI and XhoI restriction site, respectively (underlined). The amplified PCR product CD8–CD3ζ was excised with BamHI and XhoI and cloned between the BamHI and XhoI sites of the anti-ErbB2 CAR encoding vectors.21 T cells were transduced with recombinant retroviruses as described in detail elsewhere.35 Cells were analyzed by flow cytometry.

Immunofluorescence analysis

ErbB2 expression was determined by flow cytometry utilizing the anti-ErbB2 antibody, clone Ab-5 (Merck KGaA, Darmstadt, Germany), and an isotype-matched control antibody (BD Bioscience) (each 10μgml–1). Bound antibodies were detected by a fluorescein isothiocyanate-conjugated F(antibody)2-anti-mouse IgG antibody (5μgml–1; Southern Biotech). Expression of CD30 was monitored by the PE-conjugated anti-CD30 antibody Ber-H2 (Dako) and an isotype-matched control PE-conjugated antibody (BD Bioscience). T cells engineered with CAR were identified by simultaneous staining with a PE-conjugated anti-human IgG1 antibody (1μgml–1) and the fluorescein isothiocyanate-conjugated anti-CD3 antibody (UCHT-1, 1:20). Immunofluorescence was analyzed using FACSCalibur flow cytometer equipped with the CellQuest research software (Becton Dickinson, Mountain View, CA, USA).

CAR-mediated activation of engineered T cells

T cells were engineered by retroviral gene transfer with ErbB2-specific CARs containing the CD28–CD3ζ and the CD3ζ signalling domains, respectively, and 0.078 × 104–1 × 104 CAR-expressing T cells per well were cocultivated with target cells (5 × 104 cells per well). The culture supernatants were analyzed for IFN-γ by enzyme-linked immunosorbent assay. Briefly, IFN-γ was bound to the solid-phase anti-human IFN-γ antibody NIB42 (1μgml–1) and detected by the biotinylated anti-human IFN-γ antibody 4S.B3 (0.5μgml–1) (BD Bioscience). The reaction product was visualized by a peroxidase–streptavidin conjugate (1:10000) and ABTS as substrate. IL-2 in the culture supernatant was measured using enzyme-linked immunosorbent assay. Briefly, IL-2 was bound to the solid-phase coated anti-human IL-2 antibody 5344-111 and detected by the biotin-conjugated anti-human IL-2 antibody B33-2 (0.5μgml–1) (BD Bioscience). Specific cytotoxicity was monitored by an XTT (2,3-bis(2-methoxy-4-nitro-5sulphonyl)-5((phenyl-amino)carbonyl)-2H-tetrazolium hydroxide) reagent-based colorimetric assay, according to Jost.38 Briefly, engineered and mock-transduced T cells were cocultivated with ErbB2+ or ErbB2 tumor cells. After 48h, XTT (1mgml–1) (Cell Proliferation Kit II; Roche Diagnostics, GmbH, Mannheim, Germany) was added and incubated for 30 to 90min at 37°C. Reduction of XTT to formazan by viable tumor cells was monitored colorimetrically at an adsorbance wavelength of 450nm and a reference wavelength of 650nm. Maximal reduction of XTT was determined as the mean of six wells containing target cells only, and the background as the mean of six wells containing RPMI-1640 medium, 10% (v/v) fetal calf serum. The nonspecific formation of formazan due to the presence of T cells was determined from triplicate wells containing T cells in the same number as in the corresponding experimental wells. The number of viable tumor cells was calculated as follows:

Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author


The Rag-2/γc/ mice were obtained from Dr D von Laer (Applied Virology and Gene Therapy, Georg-Speyer-Haus, Frankfurt/Main, Germany). To test for tumor growth, cells of the ErbB2+ subclone MC38/ErbB2 were subcutaneously injected into Rag-2/γc-/ mice (0.8 × 106 tumor cells/mouse; six mice per group) and grown to volumes of approximately 40–60mm3. Engineered T cells with ErbB2-specific CAR were once intratumorally injected (2 × 106 T cells per mouse). Tumor progression was monitored for 40 days after treatment.

We confirm that all experiments were performed in accordance with relevant guidelines and regulations.


Conflict of interest

The authors declare no conflict of interest.



  1. Eshhar Z. The T-body approach: redirecting T cells with antibody specificity. Handb Exp Pharmacol 2008; 181: 329–342. | PubMed | ChemPort |
  2. Brocker T, Karjalainen K. Signals through T cell receptor-zeta chain alone are insufficient to prime resting T lymphocytes. J Exp Med 1995; 181: 1653–1659. | Article | PubMed | ISI | ChemPort |
  3. Brocker T. Chimeric Fv-zeta or Fv-epsilon receptors are not sufficient to induce activation or cytokine production in peripheral T cells. Blood 2000; 96: 1999–2001. | PubMed | ISI | ChemPort |
  4. Thislethwaite F, Mansoor W, Gilham DE, Hawkins RE. Engineering T-cells with antibody-based chimeric receptors for effective cancer therapy. Curr Opin Mol Ther 2005; 7: 48–55. | PubMed | ISI | ChemPort |
  5. Hombach A, Abken H. Costimulation tunes tumor-specific activation of redirected T cells in adoptive immunotherapy. Cancer Immunol Immunther 2007; 56: 731–737. | Article
  6. Hombach A, Wieczarkowiecz A, Marquardt T, Heuser C, Usai L, Pohl C et al. Tumor specific T cell activation by recombinant immunoreceptors: CD3zeta signaling and CD28 costimulation are simultaneously required for efficient IL-2 secretion and can be integrated into one combined CD28/CD3zeta signaling receptor molecule. J Immunol 2001; 167: 6123–6131. | PubMed | ISI | ChemPort |
  7. Hombach A, Sent D, Schneider C, Heuser C, Koch D, Pohl C et al. T cell activation by recombinant receptors: CD28 costimulation is required for IL-2 secretion and receptor mediated T cell proliferation but does not affect receptor mediated target cell lysis. Cancer Res 2001; 61: 1976–1982. | PubMed | ISI | ChemPort |
  8. Finney HM, Lawson AD, Bebbington CR, Weir AN. Chimeric receptors providing both primary and costimulatory signaling in T cells from a single gene product. J Immunol 1998; 161: 2791–2797. | PubMed | ISI | ChemPort |
  9. Beecham EJ, Ma Q, Ripley R, Junghans RP. Coupling CD28 co-stimulation to immunoglobulin T-cell receptor molecules: the dynamics of T-cell proliferation and death. J Immunother 2000; 23: 631–642. | Article | PubMed | ISI | ChemPort |
  10. Koehler H, Kofler D, Hombach A, Abken H. CD28 costimulation overcomes TGF-β mediated repression of proliferation of redirected human CD4+ and CD8+ T-cells in an anti-tumor cell attack. Cancer Res 2007; 67: 2265–2273. | Article | PubMed
  11. Pule MA, Savoldo B, Myers GD, Rossig C, Russell HV, Dotti G et al. Virus-specific T cells engineered to coexpress tumor-specific receptors: persistence and antitumor activity in individuals with neuroblastoma. Nat Med 2008; 14: 1264–1270. | Article | PubMed | ChemPort |
  12. Kershaw MH, Westwood JA, Parker LL, Wang G, Eshhar Z, Mavroukakis SA et al. A phase I study on adoptive immunotherapy using gene-modified T cells for ovarian cancer. Clin Cancer Res 2006; 12: 6106–6115. | Article | PubMed | ISI | ChemPort |
  13. Morgan RA, Dudley ME, Wunderlich JR, Hughes MS, Yang JC, Sherry RM et al. Cancer regression in patients after transfer of genetically engineered lymphocytes. Science 2006; 314: 126–129. | Article | PubMed | ISI | ChemPort |
  14. Schodin BA, Tsomides TJ, Kranz DM. Correlation between the number of T cell receptors required for T cell activation and TCR-ligand affinity. Immunity 1996; 5: 137–146. | Article | PubMed | ISI | ChemPort |
  15. Chervin AS, Stone JD, Holler PD, Bai A, Chen J, Eisen HN et al. The impact of TCR-binding properties and antigen presentation format on T cell responsiveness. J Immunol 2009; 183: 1166–1178. | Article | PubMed
  16. Holler PD, Kranz DM. Quantitative analysis of the contribution of TCR/pepMHC affinity and CD8 to T cell activation. Immunity 2003; 18: 255–264. | Article | PubMed | ISI | ChemPort |
  17. San José E, Borroto A, Niedergang F, Alcover A, Alarcón B. Triggering the TCR complex causes the downregulation of nonengaged receptors by a signal transduction-dependent mechanism. Immunity 2000; 12: 161–170. | Article | PubMed | ISI | ChemPort |
  18. Liu H, Rhodes M, Wiest DL, Vignali DA. the dynamics of TCR:CD3 complex cell surface expression and downmodulation. Immunity 2000; 13: 665–675. | Article | PubMed | ISI | ChemPort |
  19. Bhatia S, Sun K, Almo SC, Nathenson SG, Hodes RJ. Dynamic equilibrium of B7-1 dimers and monomers differentially affects immunological synapse formation and T cell activation in response to TCR/CD28 stimulation. J Immunol 2010; 184: 1821–1828. | Article | PubMed
  20. Sanchez-Lockhart M, Graf B, Miller J. Signals and sequences that control CD28 localization to the central region of the immunological synapse. J Immunol 2008; 181: 7639–7648. | PubMed |
  21. Chmielewski M, Hombach A, Heuser C, Adams GP, Abken H. T cell activation by antibody-like immunoreceptors: increase in affinity of the single-chain fragment domain above threshold does not increase T cell activation against antigen-positive target cells but decreases selectivity. J Immunol 2004; 173: 7647–7653. | PubMed | ChemPort |
  22. Adams GP. Improving the tumor specificity and retention of antibody-based molecules. In Vivo 1998; 12: 11–21. | PubMed |
  23. Holdorf AD, Kanantigenawa O, Shaw AS. CD28 and T cell co-stimulation. Rev Immunogenet 2000; 2: 175–184. | PubMed | ChemPort |
  24. Rudd CE, Taylor A, Schneider H. CD28 and CTLA-4 coreceptor expression and signal transduction. Immunol Rev 2009; 229: 12–26. | Article | PubMed | ChemPort |
  25. Kerstan A, Hünig T. Cutting edge: distinct TCR- and CD28-derived signals regulate CD95L, Bcl-xL, and the survival of primary T cells. J Immunol 2004; 172: 1341–1345. | PubMed | ISI | ChemPort |
  26. Alvarez-Vallina L, Hawkins RE. Antigen-specific targeting of CD28-mediated T cell co-stimulation using chimeric single-chain antibody variable fragment-CD28 receptors. Eur J Immunol 1996; 26: 2304–2309. | Article | PubMed | ChemPort |
  27. Willemsen RA, Ronteltap C, Chames P, Debets R, Bolhuis RLH. T cell retargeting with MHC class I-restricted antibodies: the CD28 costimulatory domain enhances antigen-specific cytotoxicity and cytokine production. J Immunol 2005; 174: 7853–7858. | PubMed | ISI | ChemPort |
  28. Hombach AA, Schildgen V, Heuser C, Finnern R, Gilham DE, Abken H. T cell activation by antibody-like immunoreceptors: the position of the binding epitope within the target molecule determines the efficiency of activation of redirected T cells. J Immunol 2007; 178: 4650–4657. | PubMed |
  29. Wülfing C, Davis MM. A receptor/cytoskeletal movement triggered by costimulation during T cell activation. Science 1998; 282: 2266–2269. | Article | PubMed | ISI | ChemPort |
  30. Viola A, Schroeder S, Sakakibara Y, Lanzavecchia A. T lymphocyte costimulation mediated by reorganization of membrane microdomains. Science 1999; 283: 680–682. | Article | PubMed | ISI | ChemPort |
  31. Bromley SK, Iaboni A, Davis SJ, Whitty A, Green JM, Shaw AS et al. The immunological synapse and CD28-CD80 interactions. Nat Immunol 2001; 2: 1159–1166. | Article | PubMed | ISI | ChemPort |
  32. Wülfing C, Sumen C, Sjaastad MD, Wu LC, Dustin ML, Davis MM. Costimulation and endogenous MHC ligands contribute to T cell recognition. Nat Immunol 2002; 3: 42–47. | Article | PubMed | ISI | ChemPort |
  33. Purtic B, Pitcher LA, van Oers NSC, Wülfing C. T cell receptor (TCR) clustering in the immunological synapse integrates TCR and costimulatory signalling in selected T cells. Proc Natl Acad Sci USA 2005; 102: 2904–2909. | Article | PubMed | ChemPort |
  34. Bjørgo E, Taskén K. Novel mechanism of signaling by CD28. Immunol Lett 2010; 129: 1–6. | Article | PubMed
  35. Weijtens ME, Willemsen RA, Hart EH, Bolhuis RL. A retroviral vector system ‘STITCH’ in combination with an optimized single chain antibody chimeric receptor gene structure allows efficient gene transduction and expression in human T lymphocytes. Gene Therapy 1998; 5: 1195–1203. | Article | PubMed | ISI | ChemPort |
  36. Boukamp P, Petrussevska RT, Breitkreutz D, Hornung J, Markham A, Fusenig NE. Normal keratinization in a spontaneously immortalized aneuploid human keratinocyte cell line. J Cell Biol 1988; 106: 761–771. | Article | PubMed | ISI | ChemPort |
  37. Schier R, McCall A, Adams GP, Marshall KW, Merritt H, Yim M et al. Isolation of picomolar affinity anti-c-erbB-2 single-chain Fv by molecular evolution of the complementarity determining regions in the center of the antibody binding site. J Mol Biol 1996; 263: 551–567. | Article | PubMed | ISI | ChemPort |
  38. Jost LM, Kirkwood JM, Whiteside TL. Improved short- and long-term XTT-based colorimetric cellular cytotoxicity assay for melanoma and other tumor cells. J Immunol Methods 1992; 147: 153–165. | PubMed | ISI | ChemPort |


We thank Drs GP Adams and JD Marks, UCSF, for providing us with the anti-ErbB2 scFv C6.5 and derivatives thereof, and Dr S Davis, University of Oxford, UK, for helpful suggestions to the manuscript.

This study was supported by grants from the Deutsche Krebshilfe, Bonn, the ATTACK Consortium of the European Union and the Köln Fortune Program of the Medical Faculty of the University of Cologne.