Reducing Late-Stage Biologic Failure with Immune Assays

Harald Kropshofer
Contributor: Harald Kropshofer
Posted: 11/28/2013

Dr. H. Kropshofer is Senior Personalized Healthcare Leader in the Pharmaceutical Division of F. Hoffmann-La Roche. He is located in the head quarter in Basel, Switzerland. He is globally accountable for clinical immunology and clinical immunosafety in late stage development – with a particular focus on risk assessment and mitigation related to immunogenicity, hypersensitivity and first infusion related side-effects of therapeutic proteins in the context of cardiovascular, neurological diseases and in oncology.

Dr. Kropshofer is biochemist and immunologist by training. Prior to joining Roche in 2001, he was Research Leader at the German Cancer Research Center in Heidelberg, Germany and at the Basel Institute for Immunology, Switzerland. He is Associate Professor in cell biology at the University of Heidelberg since 2000.

Dr. H. Kropshofer has more than sixty publications to his credit in peer-reviewed journals, and is a regularly featured speaker at international conferences in Europe and the USA

 Gerald Clarke:           So, Harald, could you give us a little bit of background as to your history in the field?

Harald Kropshofer:   Right, my pleasure to do so. So, I’m working in this field quite some time. I spent almost ten years in academia at different locations and I was leading research labs, dealing with immunological questions in terms of interference of drugs with the immune system and vice versa. Then, when I left the German Cancer Research Centre in 2000, I made it into the field of drug development by joining Hoffmann-La Roche and this was first in a research centre: the Basel Centre for Immunology and then after two years, I moved directly into non-clinical drug development and drug safety. And after building up an immunosafety section there I joined the late stage development and have been doing personalised healthcare since three and a half years now, here at Roche having a global responsibility, overseeing many drugs, mainly biotherapeutics because they most often involve questions around immunotoxicology. It’s less the small molecules. It’s more all the other drugs.

Clarke:           So, there have been some high profile cases of antibodies causing severe immune reactions. What are the risks and how can they be minimised?

 Kropshofer:    Oh, yes. So, the original thought was that, in particular, monoclonal antibodies, which are used for therapeutic intervention would be less of a concern, because there are immunoglobulins in our body; there is a huge amount in circulation and the thought was, we are just mimicking immunoglobulins, which are endogenously synthesised and have their functions there. So, there shouldn’t be a lot of toxicology, with regards to immunotoxicology. But this view changed gradually in the past decade.

 The first strong indication that also monoclonal antibodies can do quite harm to the human body was the Northwick Park event in 2006, when six healthy volunteers were exposed to a monoclonal antibody the first time in the Phase 1 trial and when those young volunteers experienced acute serious type hypersensitivity reactions, it turned out to be a cytokine storm and this made our field strongly aware that even if you do everything right, in terms of you obey all the rules, you do all the pre-clinical work, you do all testing of different species, so in this context, with the TeGenero case, all species you can envisage were really tested quite rigorously on any immunotoxicological side effects. Nevertheless, none were seen in animals. It turned out that the humans are different.

 And this is a key point: humans are different, so to what extent can we really mitigate those risks, which are somewhere hidden in the structures, in the formulation or in both when we administer patients to this type of drug? To be honest and clear, monoclonal antibodies are, of course, not at the forefront in causing immunotoxicological issues. But it is far more than we originally anticipated when the first monoclonal therapeutic antibodies were designed and nowadays we are trying to exploit those learnings from early failure, but also late stage failure and most times it is late stage failure, which prevents monoclonal antibodies from coming effectively into patients, because in Phase 3, it turns out to be an issue with the immune system.

So, most failures turn out late, unfortunately, which is, of course, a big, big issue for the Pharma industry and this is because you need, sometimes, a high number of patients to be exposed to such a drug to really see those patients who react because of the genotype, because of the individual setup in a way, which is unexpected. Most of the time it’s immune system, which causes the problem and the medication strategies we have in place and we need to have in place are, of course, manifold. We try as best as we can to do preclinical studies and to see which organs may do harm, but in particular, the immune system is an organ, which is apparently behaving very differently from the immune system of, for example, monkeys, rats or mice and this is a key learning from our past, so that we cannot rely only on preclinical studies, which are of a regulatory mast.

 But we have to use our patience and we have to use the data we are collecting from patients, which are showing those adverse events in order to mitigate the risk for the next generation of patients and this is what we are doing. So, in other words, it’s not so much that we can exploit only the route of a bench-bedside, but it is more now bedside-bench because we learn and by learning from failure, from data we collect on patients afflicted with adverse events, that we set up new strategies and new avenues to go to minimise the risk and to make our drugs safer.

 Clarke:           So, you’ve done work on whole blood assays to assess cytokine response. How can this work help in predictive toxicology?

 Kropshofer:    Right. I spent quite some time in the non-clinical arena to make a new type of essay available to predict the risk for first infusion or first injection related adverse events, such as cytokine release syndrome, but this is exactly the adverse event, which was a root cause for those TeGenero related symptoms, which was hypertension, bronchospasm, oedema, and whole skin events after those six healthy volunteers were first exposed to a monoclonal antibody targeting the CD28 molecule and we were thinking hard: what can we do to avoid this from happening again if we would have, at Roche, a monoclonal antibody targeting a similar receptor?

 Fortunately, we had not a single event so far going this direction. Nevertheless, we wanted to be proactive in setting up a test now and this test uses, fresh human whole blood from healthy volunteers or from patients, depending on the indication, depending on the setup and we are trying to do this now since quite a while, since 2008. We set up such a assay in our preclinical assay platforms, to run every biologic, every monoclonal through this type of test in order to see whether there are blood samples, which give rise to cytokine release. So, the test concept is quite easy. You can do the same test, by the way, with animal blood and most of the time, you will see no response. Response means you will not see any cytokines to be released in the presence of the drug.

 So, the concept is that you just use fresh human whole blood, add the drug of question, add a concentration, which is reflecting the peak exposure in man and we do this before we go into Phase 1 clinical trials and we just see whether there is, out of 40/50 different samples, any individual who is reacting in terms of release of typical proinflammatory cytokines, which is IL-6, TNF?, IL-8 or IL-10. So we have a whole battery of end points included in this type of assay and as long as we see a no-cytokine release, we are happy and this is a go signal, this is green light for moving to the next step. As soon as we would see a cytokine release, meaning an elevation of the typical proinflammatory cytokines, we would have to go back and see what quality of the drug or the drug formulation gives rise to such a response and we are quite happy, so far, that we did that because, by using this test, we had the good news that not a single monoclonal antibody or biotherapeutic drug we brought into Phase 1 so far gave rise to any of those acute cytokine release related adverse events.

 Clarke:           How can knowledge of T-cell-dependent responses to protein therapeutics aid in predicting immune responses?

 Kropshofer:    Right, this is now the other branch of the immune system; before with the cytokines, they are often just released by the innate immune cells, which is monocyte, neutrophils and NK cells, while the other half of the adverse events we do see in clinical development comes from an unwanted response of the adaptive immune system, which is B lymphocytes and/or T lymphocytes. So, since T lymphocytes, which is CD4 plus T helper cells, essentially, are activated first and give help to B cell activation, finally, leading, for example, to the formation of antibodies directed against a drug protein, we focus on the T cell activation testing because it’s the first step in this cascade of events leading, at the end, to formation of anti-drug antibodies, complement activation and/or other adverse events like hyposensitivity reactions.

 So, we want to see whether we can really prevent CD4 helper T cells to get activated. So, again, we use this test system pre-clinically before we move into Phase 1 and expose patients to our monoclonals or other therapeutic proteins including peptides. And the test principle here is the same as what we see normally in lymph nodes, happening before we get an anti-drug antibody response, which is a clonal expansion of a certain specific CD4 positive helper T lymphocytes. They are activated by interaction with specific, professional antigen-presenting cells, which are the sentinels of the immune system. They are called dendritic cells because of the dendrites they are exposing to the solvent. And those dentritic cells, they are located at all interfaces of our body with the external milieu.

 So, for example, in mucosa, of course, in the blood and, of course, particularly enriched in the upper layers of the skin. So, those dendritic cells, for example, dermal dentritic cells, they are the ones who get the chance to move to local lymph nodes and to activate the T lymphocytes there and so we have just tried to mimic this scenario by isolating dendritic cells from individuals because those dendritic cells, they differ from individual to individual, with regards to the expression of an immune receptor, which is key, which is the MHC Class 2 molecules or HLA human leukocyte antigens. And the human leukocyte antigens expressed on dendritic cells in very high abundance, they are really different from individual to individual, so we are very polymorphic in this respect and this also differentiates us from other species; this HLA molecule, it has a specificity, each HLA molecule has a certain specificity in binding peptides derived from a drug protein, for example. And so it is those HLA molecules that really make one of the key differences between a protein becoming immunogenic in men, but not becoming immunogenic in species because of the species difference across HLA molecules.

 So, these HLA molecules, they are supposed to bind peptides and present them to T lymphocytes, so the T lymphocytes get activated by the interaction of a T cell receptor with those HLA peptide complexes. Now, the question is whether a protein gives rise to the appropriate peptide. Dendritic cells test that by, as soon as they get exposed to a protein, they take up the protein, chop it into peptides and load it into HLA molecules and we mimic exactly this by isolating those dendritic cells from individual to individual and try to bring them together in a test tube with autologous T lymphocytes, so peripheral T lymphocytes from the blood of the same donor where we get the dendritic cells from.

 And we do this from healthy volunteer to healthy volunteer or from patient to patient and can see whether there is peptide being really presented by the dendritic cells and whether those peptide HLA complexes give rise to activation of T lymphocytes. So, what we measure at the end is, in absence or presence of a drug, whether T lymphocytes proliferate and/or secrete T lymphocyte-specific cytokines and this can be done in conventional assays, which just count the number of T lymphocytes before and after exposure to the drug. This is the T cell activation assay and this is a way to see whether a protein has the capacity to stimulate individual-wise, the T cell population of patients or healthy volunteers or not.

 Clarke:           So, how can we successfully integrate these predictive methodologies into drug development?

 Kropshofer:    Right. This is a more tricky question because we realised, in the past, doing these type of approaches now for almost ten years, that it is always a case by case decision, depending on the indication, depending on the drug molecule, be it a monoclonal antibody, be it a protein, be it a peptide or whether we are in an indication which is really at higher risk as an autoimmunity indication or an oncology indication where most of the time, patients are immunosuppressed so that the risk is reduced. Other parameters, which shape the battery of tests we need to do preclinically and clinically is, for example, the mode of administration. So, we should be pretty clear: as long as we go, for example, with a monoclonal antibody into an intravenous administration mode in an oncology indication, of course we have a low risk scenario. Why? Because monoclonal antibodies per se, are at the lower scale of the overall risk. For example, peptides are at the higher scale in the risk assessment because they are really less folded so they are much more eager to get in contact with those HLA molecules of dendritic cells and are much more likely to activate the immune system in an unwanted way. So, going back to monoclonals in an intravenous setting in the context of an oncology patient population, is really a kind of lower risk scenario because the monoclonal is quite robust nowadays, being human or humanised, so that there are very few regions, which are really giving rise to T cell activation, for example.

 And the next step is, of course, that an IV, intravenous administration mode, as compared to a subcutaneous administration mode exposes the drug less to the immune system because the critical first step, which is dependent on dendritic cells is less pronounced because in the blood stream there is less of those dendritic cells, while dendritic cells are highly enriched in the skin, as mentioned before. So, an oncology indication on top, which is the patients who are immunosuppressed most of the time, again reduces the risk. This is, of course, totally different if we would move into a subcutaneous mode of administration with a protein or a peptide drug in an asthmatic patient population. Here the risk is higher because allergic patients, asthmatic patients, they have lower thresholds for activation of the immune system. The subcutaneous mode exposes the drug more to the immune system and, yes, a protein often is more a difficulty for the immune system to generate epitopes, which are key for T and B lymphocyte activation, for example.

 So, we tailor our approaches, which I have outlined before T cell activation testing prediticion tools for T cell epitopes and B cell epitopes. We tailor this, depending on the risk. So, it’s a risk-based approach, of course, as also outlined by the regulators in US and in Europe at the moment. So, when we have, for example, such a high-risk scenario, we are really, very busy in not only doing our conventional battery of preclinical tox studies in monkeys and in rats or in other animal species in parallel, we really try to get a good oversight on human cell-based tests and on the tests I outlined before, in order to make sure that first infusion or first injection cytokine release is not an issue, that we keep the formation of anti-drug antibodies low, which, at the end, not only may impose safety issues, but, which may also give rise to neutralising antibodies, for example, which then make a drug inefficacious because of the neutralisation effect.

 So, in other words, it is always a case-by-case decision, which type of tests we run and we are not only running tests, as I outlined at the beginning of this conversation, at the preclinical stage. Of course, we do as much as we can preclinically, but often, there is absolutely no issue, not in terms of safety or in terms of efficacy in a Phase 1 or in a Phase 2 trial. This is often the case because the patient numbers are not sufficient to see issues or/and that time of observation is not long enough. This all is in place in Phase 3. So, telling you, frankly, that it is often in Phase 3 that we have very few patients who show a strange behaviour in terms of immunological reactions and we try, again, to collect and store as much as we can, samples for making our analysis retrospectively; so that we have the samples proactively and retrospectively so that we are really in good shape if there is an issue, that we do have the samples in the right quality at the right time point prior and after an event, so that we can look into new parameters. For example, IGE. So, the presence or absence of IGE or, for example, in terms of hypersensitivity reactions, we then can look into whole blood again and look into basophils, for example, which are in the blood and which are activated, maybe, giving rise to hypersensitivity reactions, so that we can do in vitro or ex vivo testing also during our Phase 3 clinical trial, in order to see what is root cause. Is it the drug or is it the formulation or is it a combination of all that?

Clarke:           So, briefly, what changes do you think we will see in this field in the coming 18 months?

Kropshofer:    Right. I think, the changes will go in the other direction of exploiting more the clinical samples and to explore new avenues of what end points are telling us are predictive for all sorts of adverse events, safety related immunotoxicological events. I think, there is a huge new field waiting for us. It’s also the genomic profiling, the sequence efforts we are doing, so we will, within the next two years, we will have the capacity to look more often into the whole genome sequence of patients showing such strange behaviour in terms of individual responses of the immune system and we try to better understand, is it a polymorphism, for example, in certain cytokine genes? Or a new polymorphism in HLA, which predisposes such patients to such events?

 Sometimes in oncology indications, it may just be ten or 20 cases of hypersensitivity reactions, which make a drug programme stop at the end because this is really unacceptable to, for example, a diabetes population or another non-oncology indication. So, we need to find out much more individualised, so patient-by-patient, what makes those patients different from the other patients showing absolutely no adverse events? And here, genotyping is one thing, but we also have to look more into the cells, so we have to take out primary cells from those patients, blood-derived cells and looking into expression profiles and into cytokine profiles and into their interaction with the innate and adaptive immune system in order to understand the difference.

 So, I believe we will not expand into more sophisticated preclinical animal studies. I do see a way forward if we are successful in making mice more human, but it can only be an additional help; it will not be the only help. The real steps bringing us forward is the patient, is the healthy volunteer showing the adverse event at the level of the human immune system and the reason for this is the human immune system is too different from any other species and so, on top of that, we, as humans, are among us, too different, so we have our different genomic equipment and our receptors are different. We learn about snips and we learn about individual difference at that molecular level and we need to do that in the next future so that maybe in a decade, we are in better shape in the diagnosis and prognosis of immunotoxicological side effects of drugs and I’m sure we will make it, but it’s still quite a challenge.

 If you would like to find out more about this topic and the predictive toxicology summit next February 24th–26thin London, England, visit us at


 Please note that we do all we can to ensure accuracy within the translation to word of audio interviews but that errors may still understandably occur in some cases. If you believe that a serious inaccuracy has been made within the text, please contact +44 (0) 207 368 9482 or email

Harald Kropshofer
Contributor: Harald Kropshofer
Posted: 11/28/2013


Crowne Plaza Zurich, Zurich, Switzerland
June 17 - 19, 2018