Additional Lessons from the CD28 Superagonist TGN1412 Trial

CORRESPONDENCE LINK TO ORIGINAL ARTICLE LINK TO AUTHOR’S REPLY

any adverse effects”. Indeed, the design of the Phase I study, in which all the volunteers Storm forecasting: additional were dosed at 10‑minute intervals4, suggests that no practical consideration was given to lessons from the CD28 superagonist the possibility of CRS occurring. TGN1412 trial Failure to reconcile discordant preclinical data. The early decision by the Medicines and Health Regulatory Agency5 to publicly Christopher Horvath, Laura Andrews, Andreas Baumann, Lauren Black, release the IMPD, investigator’s brochure, Diann Blanset, Joy Cavagnaro, Kenneth L. Hastings, David L. Hutto, clinical trial protocol, assessment report Timothy K. MacLachlan, Mark Milton, Theresa Reynolds, Stan Roberts, and informed consent for TGN1412 (avail- Mark Rogge, Jennifer Sims, George Treacy, Garvin Warner and James D. Green able on the Citizens for Responsible Care and Research website) should be applauded This letter is in response to the Comment a failure to reconcile discordant preclini- because it facilitated independent review. article published in Nature Reviews cal data sets. Three, there was a failure to Such a review shows that disparate results Immunology by Thomas Hünig (The identify a pharmacologically relevant species were obtained from in vitro and in vivo storm has cleared: lessons from the CD28 and/or data set for the determination of the assessments of the superagonist activity of superagonist TGN1412 trial. Nature Rev. first‑in‑human dose. And, four, there was CD28‑specific antibodies in humans, rats Immunol. 12, 317–318 (2012))1. In his com- a failure to translate the preclinical data to and monkeys. mentary, Thomas Hünig attempts to explain estimate the anticipated pharmacological In vitro, T cell activation and cytokine “why the three sets of preclinical data … activity of the first‑in‑human dose. We shall release have been demonstrated for the used to support this Phase I clinical trial discuss these topics in order. monoclonal antibody 5.11A1 (a precur- had failed to predict the cytokine storm” in sor to TGN1412) in human systems and healthy volunteers. He describes some of the Failure to identify risk based on the intended for JJ316 (a rat CD28‑specific monoclonal findings in these test systems and concludes target and mechanism of action. With antibody homologous to TGN1412) in rat that they “all failed to provide evidence TGN1412, there were strong reasons to systems6,7. Based on the publicly released for the toxic potential of the antibody for proceed with caution based on the nature of information, it appears that no such demon- distinct and unrelated reasons”. He further the target and the intended and/or desired stration of in vitro pharmacological activity claims that “the disastrous outcome … may pharmacological activity: superactivation of TGN1412 was presented for human or have made a positive contribution to the of T cells. Comparable superactivation of monkey T cells in the IMPD or investigator’s future development of immunomodulatory T cells occurs naturally when bacterial brochure. In these documents, it appears drugs”, primarily by “teaching us the superi- superantigens bypass normal antigen pres- that the only data describing the activity of ority of the MABEL [minimum anticipated entation processes and directly co‑activate TGN1412 in human systems were the results biological effect level]-based approach MHC class II molecules and T cell recep- for in vitro T cell mitogenesis assays that over the NOAEL [no observed adverse tors (TCRs), resulting in fever, extensive were conducted with the precursor mono- effect level]‑based approach”. cytokine release, toxic shock and death clonal antibody 5.11A1 or with TGN1412 In sharp contrast to Thomas Hünig’s at plasma concentrations as low as 0.1 pg in the presence or absence of methotrexate. opinions, we believe that the serious adverse per ml (REF. 2). Historically, iatrogenic T cell The concentration–response relationships events were predictable based on the known mitogenesis, T cell superactivation and sud- for TGN1412 in human cell systems and CD28 receptor biology, the intended phar- den, but prolonged, depletion of circulating their relationship (or lack thereof) to those macological activity of TGN1412 and the T cells have been the hallmarks of immuno for TGN1412 in monkey systems or JJ316 in preclinical safety assessment programme. modulatory CD3‑specific monoclonal rat systems do not appear to have been dis- The purpose of a preclinical safety assess- antibodies, including OKT3 (Janssen-Cilag), cussed, limiting the ability to model poten- ment programme is to identify potential HuM291 (visilizumab; PDL BioPharma) tial effects of the administration of TGN1412 safety risks and to qualify and quantify these and others3. These reactions depend on to humans from the in vitro data. risks by using in vitro human data, in vitro the extent of receptor occupancy and can In rats, a single dose of 5 mg per kg of and in vivo animal data, and translational occur at remarkably low doses. Although JJ316 resulted in a 3- to 6‑fold increase in and mechanism-based knowledge of the TGN1412 was a first‑in‑class CD28‑specific spleen size and a 20‑fold increase in splenic intended pathway modulation to ‘predict’ superagonist, it was intended to induce regulatory T (TReg) cell numbers by day 3; in vivo human responses. We feel that the superactivation of T cells, much like circulating T cells were not evaluated8. these criteria were not met for TGN1412. the CD3‑specific antibodies that induce This was the expected pharmacological Specifically, the publicly available informa- cytokine-release syndrome (CRS). While outcome for a CD28‑specific superagonist tion in the investigational medicinal product cursory mention of the possibility of CRS antibody in a rat system, and was consist- dossier (IMPD) and/or the investigator’s was made in the IMPD and investigator’s ent with the in vitro activity. By contrast, brochure demonstrates deficiencies in four brochure, insufficient consideration was ascending single-dose pharmacokinetic and major areas of the preclinical safety assess- given to the lessons already learnt from pharmacodynamic studies with 5 to 50 mg ment. One, there was a failure to identify natural superagonists and comparator prod- per kg of TGN1412 in monkeys resulted risk based on the intended target, mecha- ucts. In fact, it was stated in the investigator’s in the saturation of CD28 receptors on nism of action and previous experiences brochure that “TGN1412 is expected to be T cells, but caused only minimal (<3‑fold) with comparator products. Two, there was well tolerated in humans and not to elicit and delayed (by >2 weeks) increases in

NATURE REVIEWS | IMMUNOLOGY www.nature.com/reviews/immunol

circulating TReg cell numbers and no dose- periphery of HIS mice could be particu- However, a discussion of the potential phar- related cytokine response. In the pivotal larly sensitive to stimuli”. We consider this macological effect of the first‑in‑human toxicology study in monkeys, repeated depletion of circulating human T cells to be dose was not apparent in the IMPD. In weekly doses of 5 mg per kg and 50 mg per analogous to that observed with numerous fact, there appears to be no discussion of kg of TGN1412 resulted in saturation of the CD3‑specific monoclonal antibodies and the pharmacokinetic or pharmacodynamic receptor at both doses. However, TGN1412 to be strongly suggestive of overwhelming effects of any of the proposed Phase I doses was reported to have induced the expan- CD28‑mediated T cell superactivation. As (0.1, 0.5, 2 and 5 mg per kg) and no projec- sion of T cell populations only at the 5 mg the authors considered this to be a new tion of the anticipated human therapeutic per kg dose, and then only in males (n = 3). finding, a thorough consideration of the dose or concentrations. The first‑in‑human There was no splenomegaly and splenic potential implications for humans and the dose was simply proposed as 0.1 mg per kg,

TReg cells were not evaluated. Thus, despite submission of a report to the regulatory or 1/500 of the NOAEL in monkeys. complete receptor occupancy, there was authorities was warranted before the Obviously, for successful implementa- no dose-responsive, biologically relevant initiation of the clinical trial. tion, a NOAEL approach requires that one pharmacological activity and certainly no As the examples from the IMPD and is able to ‘observe’ an adverse effect, which is demonstration of the desired superagonist other TeGenero publications illustrate, the not possible when pharmacological activity activity. This was consistent with the lack different in vitro and in vivo preclinical test is not present in the chosen preclinical spe- of in vitro activity against monkey cells systems gave conflicting results that were not cies. Hünig further implies that reliance on that was later demonstrated in 2007 (REF. 9). reconciled. Superagonism was demonstrated a NOAEL approach alone was the accepted When considered as a whole, the four data in vivo with JJ316 in rats and with 5.11A1 practice at the time, and that the subsequent sets in the regulatory filings (in vitro studies in HIS mice and in vitro with JJ316 on cells recommendation for and adoption of the in rats and humans and in vivo studies in expressing rat CD28 and with 5.11A1 on MABEL approach was a positive contribu- rats and monkeys) indicate that there was a cells expressing human CD28. However, tion of the TeGenero incident. This per- significant issue with the pharmacological pharmacological activity was not demon- spective is not consistent with the fact that relevance of monkeys. strated in monkeys, warranting further applicable regulatory guidance was in place In addition, there was one other data set investigation to reconcile this discrepancy. at the time for biotherapeutics products, for that was available to TeGenero before the which toxicity typically arises from exagger- initiation of the Phase I trial but that was Failure to identify a pharmacologically ated pharmacological effects. This guidance probably not available to regulators, as it relevant test system (or systems) for stressed that pharmacologically relevant was not apparent in the IMPD. A paper that first‑in‑human dose selection. As indicated test systems must be identified and used for included an author from TeGenero10 was above, expected pharmacological effects safety assessment11 and that the NOAEL- submitted for publication on 17 January were seen in rodent systems using precur- derived human equivalent (first) dose should 2006, just 1 month after the 19 December sor or homologous antibodies. However, be revised if a human equivalent dose based 2005 signature dates on the IMPD and within the regulatory filings for TGN1412, on a pharmacologically active dose (PAD; investigator’s brochure documents, and there were no in vitro pharmacology data the precursor to MABEL) would be a more was published online on 2 March 2006, just comparing the results for human cells sensitive indicator of potential toxicity12. 11 days before volunteers received doses of with those from monkey cells. Monkeys Reasons for additional conservatism 0.1 mg per kg of TGN1412. In this work, appeared to have been selected as the when selecting the first‑in‑human dose CD28 superagonism was evaluated in mice relevant species solely on the basis of the are specifically listed in the US Food and lacking a functional immune system that demonstration that TGN1412 binds to Drug Administration (FDA) guidance. had been irradiated and reconstituted with monkey CD28 (from receptor occupancy These reasons include: a new therapeu- human CD34+ fetal liver stem cells. These and tissue cross-reactivity studies) and the tic target, a long half-life of the product, ‘humanized’ mice (termed HIS mice) conserved sequence of the CD28 6‑amino- inadequate dose–response data and an develop immune systems containing all acid C’’D loop, without any data to indicate animal model (species) with limited utility the major human myeloid and lymphoid that TGN1412 exhibits relevant pharma- (pharmacological relevance)12. Additional cellular compartments, including human cological activity in monkeys. As relevant guidelines, in place since 1977, stipulate that CD28+ T cells and antigen-presenting cells pharmacological effects were not seen, an the first‑in‑human dose should not exceed expressing human Fc receptors10. In these alternative approach to first‑in‑human dose the projected human therapeutic dose13. mice, administration of as little as ~0.3 mg selection focusing on the data from other Subsequent analyses of the TGN1412 dose, per kg of 5.11A1 resulted, by day 6, in a assay systems would have been appropriate. using data available before the trial, showed 7‑fold increase in the overall number of that the 0.1 mg per kg first‑in‑human dose human thymocytes and a 20‑fold increase Failure to estimate the anticipated activity was estimated to result in a maximum in the number of mature thymocytes, which of the first‑in‑human dose. The nature of the plasma concentration of ~1 μg per ml and to was said to be consistent with the intended target, the pertinent historical experience be associated with near-complete receptor superagonist effect. However, the mice also with relevant comparator products and the saturation14–16. This would be predictably showed a “drastic” depletion of circulating conflicting preclinical data with TGN1412 expected to result in overwhelming T cell human CD28+ T cells from between day 3 would indicate a distinct likelihood of CRS. superactivation and risk of CRS, as is and day 60 after antibody administration10. This understanding would dictate that a common to T cell-activating biotherapeutics. The authors of this study pointed out that conservative approach should be taken and In conclusion, we believe that Hünig’s they had not observed such an effect of that the first‑in‑human dose for TGN1412 commentary perpetuates the mistaken belief 5.11A1 previously and they “hypothesized should be soundly justified on the basis that the TeGenero event was “completely that human [CD28+] T cells found in the of anticipated pharmacological activity. unpredicted by the preclinical studies”. We

NATURE REVIEWS | IMMUNOLOGY www.nature.com/reviews/immunol

believe that serious adverse events, including 1. Hünig, T. The storm has cleared: lessons from the CD28 superagonist TGN1412 trial. Nature Rev. CRS, were indeed predictable, or should have Immunol. 12, 317–318 (2012). been anticipated, based on the known CD28 2. Proft, T. & Fraser, J. D. Bacterial superantigens. Clin. Exp. Immunol. 133, 299–306 (2003). receptor biology and the intended pharma 3. Horvath, C. J. & Milton, M. N. The TeGenero incident cological activity of TGN1412. Failure to and the Duff Report conclusions: a series of unfortunate events or an avoidable event? Toxicol. reconcile inconsistencies in the available Pathol. 37, 372–383 (2009). data and to correctly apply a translational 4. Suntharalingam, G. et al. Cytokine storm in a phase 1 trial of the anti‑CD28 monoclonal antibody TGN1412. approach using the most relevant data for N. Engl. J. Med. 355, 1018–1028 (2006). selection of the first‑in‑human dose com- 5. Medicines and Healthcare Products Regulatory Agency. Clinical trial suspension: latest findings. pounded this problem. Our perspective is MHRA [online], http://www.mhra.gov.uk/NewsCentre/ that the evolution of regulatory guidance Pressreleases/CON2023515 (2006). 6. Lühder, F. et al. Topological requirements and and industry practice is best facilitated by signaling properties of T cell-activating, anti‑CD28 scientific discussions that include parties antibody superagonists. J. Exp. Med. 197, 955–966 (2003). experienced in pharmaceutical develop- 7. Tacke, M., Hanke, G., Hanke, T. & Hünig, T. ment and regulation, not by the response to CD28‑mediated induction of proliferation in resting T cells in vitro and in vivo without engagement of the disasters. Therefore, the true lesson of the T cell receptor: evidence for functionally distinct forms TeGenero incident is that it is essential that of CD28. Eur. J. Immunol. 27, 239–247 (1997). 8. Lin, C. H. & Hünig, T. Efficient expansion of regulatory the justification of the first‑in‑human dose T cells in vitro and in vivo with a CD28 superagonist. for biotherapeutics (and for small molecules) J. Immunol. 33, 626–638 (2003). 9. Stebbings, R. et al. “Cytokine storm” in the phase I involves the perspective of preclinical safety trial of monoclonal antibody TGN1412: better assessment experts experienced in the design, understanding the causes to improve preclinical testing of immunotherapeutics. J. Immunol. 179, implementation and interpretation of the 3325–3331 (2007). preclinical programmes for such therapies. 10. Legrand, N. et al. Transient accumulation of human mature thymocytes and regulatory T cells with CD28 Christopher Horvath is at INTICA Biomedical, USA. −/− −/− superagonist in ‘‘human immune system’’ Rag2 γc mice. Blood 108, 238–245 (2006). Laura Andrews is at Genzyme, a company of Sanofi, 11. International Conference on Harmonisation of USA. Technical Requirements for Registration of Pharmaceuticals for Human Use. ICH harmonized Andreas Baumann is at Bayer Healthcare, Germany. tripartite guideline S6: preclinical safety evaluation Lauren Black is at Charles River Laboratories, USA. of biotechnology-derived pharmaceuticals (ICH, 1997). Diann Blanset and James D. Green are at Boehringer- 12. US Food and Drug Administration Center for Drug Ingelheim Pharmaceuticals, Inc., USA. Evaluation and Research. Guidance for industry. Estimating the maximum safe starting dose in initial Joy Cavagnaro is at Access BIO, USA. clinical trials for therapeutics in adult healthy volunteers (FDA, 2005). Kenneth L. Hastings is at Sanofi SA, USA. 13. Association of the British Pharmaceutical Industry/ BioIndustry Association. Guideline for preclinical and David L. Hutto is at Eisai, Inc., USA. clinical testing of new medicinal products. Parts 1 & 2 (ABPI, 1977). Timothy K. MacLachlan and Mark Milton are at Novartis, 14. Sims, J. & ABPI/BIA Early Stage Clinical Trials Task USA. Force. Calculation of the minimum anticipated biological effect level (MABEL) and 1st dose in Theresa Reynolds is at Genentech, USA. humans. EMA [online], http://www.ema.europa.eu/ docs/en_GB/document_library/Presentation/2009/11/ Stan Roberts is at SAR Safety Assessment, USA. WC500010862.pdf (2007). 15. Lowe, P. J. et al. On the anticipation of the human Mark Rogge is at Biogen IDEC, USA. dose in first‑in‑man trials from preclinical and prior Jennifer Sims is at Integrated Biologix GmbH, clinical information in early drug development. Xenobiotica 37, 1331–1354 (2007). Switzerland. 16. Lowe, P. J., Tannenbaum, S., Wu, K., Lloyd, P. & George Treacy is at Janssen R&D, USA. Sims, J. On setting the first dose in man: quantitating biotherapeutic drug-target binding through Garvin Warner is at Alnylam Pharmaceuticals, USA. pharmacokinetic and pharmacodynamic models. Basic Clin. Pharmacol. Toxicol. 106, 195–209 Correspondence to C.H. and J.G. (2010). e-mails: [email protected]; Competing interests statement [email protected] The authors declare no competing financial interests. Disclaimer: The views expressed in this article are those of the authors and do not necessarily reflect FURTHER INFORMATION those of their employers. Citizens for Responsible Care and Research: http://www.circare.org/foia5/tgn1412.htm doi:10.1038/nri3192-c1 ALL LINKS ARE ACTIVE IN THE ONLINE PDF Published online 31 August 2012

NATURE REVIEWS | IMMUNOLOGY www.nature.com/reviews/immunol