A current focus of cancer research is the development of immunostimulatory therapeutic antibodies, based on the notion that increasing the number or activity of cytotoxic T lymphocytes, and in particular those directed against tumour-associated antigens, may achieve effective immune mediated tumour rejection. Unlike other techniques, such as the use of antibodies specific to tumour cells themselves, this approach does not rely on antigen-specificity, and accordingly is subject to risks of autoimmunity as a consequence of non-specific activation of T cells [1].

The risks associated with such an approach have been placed firmly in the spotlight following the recent highly publicised suspension of Phase-I clinical testing of TGN1412 (see [2] for TGN1412 Investigator’s Brochure). This monoclonal antibody, one of a class of ‘superagonists’ for the immune co-stimulatory receptor CD28, was developed by TeGenero Immune Therapeutics (Würzburg, Germany) for the treatment of both autoimmune diseases [3] and leukaemia [4, 5]. Use of such antibodies for treatment of both autoimmunity and malignancy seeks to exploit the opposing functions of the antibody; in the former, the aim is the expansion of regulatory T cells for the suppression of self-directed immune responses, whereas in the latter, the aim is expansion and activation of cytotoxic T cells directed against tumour antigens. Shortly after being injected with the TGN1412, all six healthy volunteers were admitted into intensive care with rapid onset grade-IV toxicity, suffering from multi-organ failure, caused by cytokine release syndrome [6].

One man’s agonist...

The natural ligands for CD28 are the B7 family of antigens (B7.1/CD80 and B7.2/CD86), found on antigen presenting cells (APCs) [7]. Under natural conditions, T cell activation requires two separate stimulatory signals: (1) T cell receptor (TCR) engagement of antigen presented by MHC, and (2) binding of co-stimulatory molecules such as B7 to its cognate receptor CD28 [8, 9] (Fig. 1a). Binding of B7 to CD28 causes expansion, proliferation and cytokine release by T cells [10, 11]. However, if either signal is provided with sufficient intensity, as for instance by superantigen stimulation of TCR (Fig. 1b), or antibody mediated super activation of CD28 (Fig. 1c), T cell activation can be achieved without the need for the second signal [12–16]. Therefore, TGN1412 mediated activation of CD28 should be able to induce antigen independent, polyclonal, T cell activation. Indeed, in pre-clinical studies for the treatment of B-cell chronic lymphocytic leukaemia, TGN1412 has been reported to stimulate polyclonal T cell activation [4, 5].

Fig. 1
figure 1

APC-mediated activation of CD4 and CD8 T cells involves MHC/antigen-mediated stimulation of TCR, as well as the activation of co-stimulatory signalling pathways (a). However, potent stimulation of TCR (e.g. by superantigens) (b) or CD28 (e.g. by TGN1412) (c) can bypasses the need for the second signal

Full size image

However, during the development of TGN1412, a number of studies were published reporting that superagonistic anti-CD28 antibodies could induce the preferential activation of CD4+CD25+ T cells (Tregs), rather than polyclonal activation of conventional T cells [3, 17–19]. Tregs suppress the activation of T cells, including autoreactive T cells that have remained in the mature T cell population by evading thymic elimination and/or interaction with tolerising APCs [1, 20–22]. Therefore, preferential expansion of Tregs would result in active suppression of autoreactive T cells. In contrast, the polyclonal expansion of conventional T cells would cause the activation of autoreactive T cells and induction of autoimmune responses [23, 24].

Other studies have independently established that expansion of Tregs is highly dependent on the activation of CD28 [25, 26]. Nonetheless, the preferential expansion of Tregs by antibody mediated activation of CD28 is extremely dose-dependent; in one study employing an experimental model of autoimmune encephalomyelitis, a dose of 0.5 or 1.5 mg/kg elicited preferential expansion of Tregs [27]. However, this preferential expansion of Tregs was no longer detectable at the higher dose of 5 mg/kg [27].

A similar example of dose sensitivity has been illustrated in clinical trials of recombinant IL-2/diphtheria toxin conjugate DAB389IL-2 (also known as denileukin diftitox or ONTAK) for the vaccination mediated immune therapy of metastatic renal cell carcinoma patients. Whilst this IL-2 conjugate preferentially eliminated Tregs, the effects were highly dose dependent and transient [28]. Thus, common molecular targets that can be used both for preferential expansion of Tregs and for polyclonal expansion of conventional T cells can be expected to be highly dose sensitive, resulting in dramatically different outcomes at different doses.

Finding the right dose

The recent anti-CD28 trial has also served to highlight the problems that can arise when designing animal safety studies to test a humanised antibody. Herein lies a conundrum; if a non-humanised antibody is used in animal safety studies, it might not betray potential toxic effects that may arise as a consequence of humanisation of the antibody. Therefore, such a safety study would be incomplete and uninformative of the corresponding human dose. By contrast, if a humanised antibody is used in animal studies, it may be neutralised swiftly, thus masking the harmful side effects of the drug, making it a challenging task to properly estimate the appropriate dose range for human use. This is an important concern in the case of monoclonal antibodies, as they are likely to be highly dose dependent and species specific [29]. This conflict may be partially resolved by the development of transgenic animal models, able to express human orthologues of the relevant genes, providing a more realistic model for testing of humanised antibodies [30].

Moreover, given the documented possible autoimmune side effects of this class of therapies, and the fact that they are designed for administration to patients with altered immune status (cancer or autoimmunity), the wisdom of using healthy volunteers in the TGN1412 trial is questionable. However, whilst patients may have proved to be a more suitable choice, the limits of feasibility and difficulties in recruitment of suitable patients may restrict this choice in practice.

Lessons from CTLA-4

Similarly to CD28, the cell surface protein CTLA-4 also binds B7. In contrast to the action of CD28, CTLA-4 engagement of B7 actively inhibits T cell stimulation, inducing apoptosis or anergy in the activated T cells (Fig. 2a) [31–34]. Although it is presented on the T cell surface at a lower density than CD28, CTLA-4 has a ∼ 20-fold higher affinity for the B7 ligand [35]. Activation of naive T cells results in increased cell surface expression of CTLA-4 [36], thus creating a negative feedback mechanism to moderate their activation. CTLA-4 is constitutively expressed on the surface of CD4+CD25+ Tregs [37, 38]. Therefore, CD28 and CTLA-4 play antagonistic roles in the regulation of T cells [39, 40], maintaining a fine balance between CD28-mediated activation and CTLA-4-mediated inhibition of T cell activation.

Fig. 2
figure 2

B7 ligation of CTLA-4 (a) triggers apoptosis or energy in T cell populations. Inhibition of CTLA-4 signalling (b) prevents apoptosis/anergy in the activated T cell population

Full size image

Although it is difficult to detect CTLA-4 expression on resting T cells, in stimulated cells it is rapidly trafficked to the cell surface, whereupon CTLA-4 ligation effectively blocks even the earliest events in T-cell activation, including cytokine production and cell proliferation [32, 41]. The key regulatory role of CTLA-4 in ‘dampening’ of an immune response is demonstrated clearly in CTLA-4 knockout mice, which suffer from fatal organ destruction as a result of lymphoproliferation [42, 43]. Moreover, mutations in CTLA-4 in humans are implicated in several autoimmune diseases such as type-I diabetes [44]. Inhibitory antibody mediated blockade of CTLA-4 (Fig. 2b) in murine models is effective in reducing tumour size, but these models also show autoimmune side effects ranging from loss of skin pigmentation (vitiligo) to allergic encephalomyelitis [45, 46]. Autoimmune side effects have also been documented in human clinical trials of an inhibitory anti-CTLA-4 antibody in a number of different human cancers [47–50].

Anti-CTLA-4 trials thus provide an important point of comparison with the anti-CD28 therapeutics. Despite the fact that CTLA-4 is expressed constitutively on CD4+ CD25+ Tregs, there is experimental data to suggest that CTLA-4 blockade also promotes an increase in the levels of conventional T cells [51]. Hence, both immunosuppressive and enhanced autoimmune effects should be considered as possible sequelae of CTLA-4 blockade therapy. Arguably, there is no currently available technology for the reliable preferential expansion of Tregs over conventional T cells.

It should be noted that anti-CTLA-4 therapies only target activated T cells, due to the absence of CTLA-4 on the surface of naive T cells. By contrast, superagonistic anti-CD28 antibodies can act upon both naive and antigen-experienced T cells [52, 53], and therefore have the potential for a much more powerful and indiscriminate immune response.

Tethering the beast

The magnitude and rapidity of the response observed in the TGN1412 trial does not appear to be compatible with the activation of antigen-specific cytotoxic T lymphocytes, as this would not produce observable side effects until much later. The swift and severe side effects exhibited in the TGN1412 trial are much more consistent with the rapid release of a variety of pro-inflammatory cytokines – known as a “cytokine storm” [12], possibly due to CD28 mediated polyclonal stimulation of T cells. Interestingly, a recent study examining superantigen-mediated toxic shock, in which similar rapid responses are observed, has reported a biphasic pattern of cytokine response, with an early rise in TNF-α suggested to be responsible for swift manifestation of symptoms, rather than the later cytokine surge [16]. Certainly, the clinical situation is likely to be much more complex, and a previous study of cytokine storm-related sequelae following systemic IL-2 administration for renal cell carcinoma has indeed reported both quantitative and qualitative variations in the pattern of cytokine expression among individual patients [54]. Nonetheless, the potent responses manifested in all six previously healthy volunteers receiving TGN1412 suggest the promiscuous activation of an overriding mechanism.

Both anti-CTLA-4 and anti-CD28 therapies rely on releasing free antibody into the body, exacerbating the risk of a runaway immune response, since the kinetics of the interaction are pushed in favour of widespread expansion. By contrast, strategies which aim to modulate antigen specific T cells, for instance tumour cell vaccines generated by gene transfer mediated expression of B7.1 (CD80) on the surface of tumour cells, are distinct in that they have a tethered co-stimulator, as opposed to soluble, free, antibody [55–57].

While polyclonal expansion of T cells is observed in B7.1 over-expressing transgenic mice [58], B7.1 on the surface of tumour cells can activate only the T cell subsets that are able to engage the MHC-antigen on these cells. This allows a more selective stimulation of T cells that are able to access the B7 expressing tumour cell vaccine, whilst retaining the CTLA-4 mediated negative feedback loop for prevention of run-away T cell activation.

Concluding remarks

Despite the tremendous therapeutic potential of antibody-mediated T cell regulation, the recent TGN1412 trial has highlighted a number of key issues surrounding design of clinical trials and potential risks associated with the manipulation of broad ranging immunological responses. There are of course a host of other potential factors that may have caused the severe side effects in the TGN1412 trial, including manufacturing and/or dosing errors. However, it appears that there is an all too fine balance between boosting an immune response, and triggering autoimmunity. Given the inherent sensitivity of immune therapy based strategies, every care must be taken to increase the safety of such clinical trials. On a practical level, this entails using recruitment and dosing schedules to allow the assessment of at least the most severe/acute side effects in individual patients before proceeding with larger patient cohorts. In addition, careful risk/benefit analyses should feature centrally in the choice between the use of healthy versus patient volunteers. The TGN1412 trial has also highlighted the need for detailed attention to the relevance of pre-clinical animal studies, particularly with reference to both the dose and specificity of humanised antibodies. Development of safe and effective means of manipulation of regulatory T cells for therapeutic applications remains an important target in drug development. Only once the selective manipulation of regulatory T cells can be consistently achieved in vivo, can the full potential of this technology be finally realised.