Tomorrow I have to give evidence at the enquiry about what went wrong at the Northwick Park Phase 1 study of the TeGenero superagonistic anti-CD28 antibody in March, when 6 normal volunteers suffered a severe systemic reaction that put them into the intensive care unit and almost killed them.
For those who haven’t been following the story, anti-CD28 is a costimulatory molecule on the surface of T cells. It is normally involved in the interaction between T cells and dendritic cells in an immune response. A protein antigen taken up by the dendritic cell is broken down into peptides which are then presented in a groove in the class II major histocompatibility (MHC) molecules on the surface of the dendritic cell. In performing the two functions, antigen capture and peptide presentation, the dendritic cell has to mature, losing its phagocytic function and developing the costimulatory molecules, CD80 and CD86, on its surface.
CD4 positive T cells are committed to react with specific antigens. The commitment lies within the T cell receptor (TCR) on the surface (this is the counterpart of the immunoglobulin molecule on B cells, and like immunoglobulin it is a different shape for every antigen). A CD4 positive T cell recognises the peptide in the MHC groove and moves in to dock with it. As the TCR engages with the MHC, various other surface molecules also link up, chief of which is CD80 with CD28. Stimulation and costimulation of the CD4 positive T cell cause it to proliferate and secrete cytokines like interleukin-2. The CD4 positive T cells is known as a helper cell and it aids other cells like B cells and cytotoxic (CD8+) T cells to attack the invader. The CD4+ reaction is controlled by a similar mechanism. When activated the CD4+ T cells produces a molecule CTLA-4 on its surface. This also reacts with CD80 on the dendritic cell, but this reaction has a suppressive effect. CTLA-4 is also expressed by CD4+, CD25+ T cells known as regulatory T cells, which also suppress an immune response by the same interaction.
Stimulating CD28 by itself does not stimulate T cells, the interaction of the TCR and MHC is also required. Instead of CD80, CD28 can be bound by antibodies to CD28, but most antibodies bind to a single CD28 molecule at a time and this is insufficient to build up a lattice of molecules on the surface of the T cells that causes stimulation. Antibodies against the TCR complex given together with an anti-CD28 can help to build the lattice. What TeGenero were able to do was produce an antibody that crosslinked CD28 and thus stimulate T cells without the need to ligate the TCR. In the event the antibody stimulated T cells too well and the volunteers almost died of an overactive immune system.
My contribution to this inquiry stems from experiments that we did in the 1990s. For many years my colleagues in Southampton, George Stevenson and Martin Glennie, have been working with antibody molecules, using chemical reactions to engineer the antibodies into different shapes (George worked in Oxford with Rodney Porter who won the Nobel Prize for discovering the structure of antibodies.) Antibodies have three business ends, 2 for attacking the target (known as Fabs) and one to bind to the various effector mechanisms that antibodies use to kill invaders with (such as killer cells, macrophages or complement). This end is known as Fc. You can think of three double pointed probes sticking out from a central hinge. A fairly simple chemical reaction will make them come apart at the hinge, and then they can be stitched together again, but not necessarily in the combinations that they were in originally. This we have made bispecific antibodies by taking one Fab from an anti-CD20 and one from an anti-CD16, and these can be made with or without an Fc.
The particular molecule that we tested was a trispecific with Fabs from anti-CD3, anti-CD2 and anti-CD28 stitched together. CD3 and CD2 are part of the TCR complex, and the purpose was to activate T cells by forming the lattice that I mentioned earlier. We examined this antibody in the test tube, and sure enough it was excellent at stimulating T cells. Both at making T cells divide and in making them secrete interleukin-2, it was 5 times as good as PHA, which is the standard reagent for stimulating T cells in the test tube.
We intended to use it in patients with melanoma, since we know that this is one of the few cancers in man that are definitely under immune control, and one of the few that responds to large doses of interleukin-2. We had experience of using interleukin-2 and we knew it produces very severe side effects. So we started our first-in-man tests at a very low dose. We decided to treat a patient with advanced melanoma who had no other available treatment and who had an obvious tumor on his skin that we could measure.
You all know the dose of rituximab – around 500 mg. Our starting dose was less than one two thousandth of that – 200 micrograms. Like rituximab, we gave it by slow intravenous infusion. It produced no effect. We gradually raised the dose to 1.2 mg. At this dose he had interleukin-2 type side effects – rash, fever, rigors, hypotension, muscle pain and rising serum creatinine. We stopped the trial. Although there was a marginal improvement in his tumor, it was clear that this treatment was no better than interleukin-2.
In our studies using various engineered antibodies, we have occasionally found other severe side effects. With an anti-CD37 x anti-CD64 bispecific used in advanced CLL we saw a patient with a similar cytokine storm, and with an anti-CD38-saporin immunotoxins used in intractable myeloma we saw transient blindness lasting 36 hours.
Of course none of this explains why the TeGenero antibody was so toxic in the volunteers when it was so harmless in mice and monkeys (which had a 500 times greater dose). I think that the answer to this lies in a paper from Nguyen et al in PNAS (May16 2006, 103:7765-70) They describe a molecule, Siglec-5, which is present on all animal T cells including chimpanzees, gorillas and orang-utans. Siglec-5 is an inhibitor of T-cell activation via the TCR.
From this I have drawn up some principles for the testing of biological agents in man.
1. Good as animal experiments are, there is no animal perfect model.
2. Biological therapy is increasingly targeted. There is no point in applying it in people who lack the target. It follows that trials in normal volunteers are inappropriate.
3. Agents are used first in incurable patients suffering from the target illness
4. Since the trials take place in individuals that have the target, they are designated phase I/II trials and look both at safety and efficacy.
5. One patient at a time is treated with time taken to evaluate the response
6. We give antibodies by slow intravenous infusion
7. We start at very low doses