1. Of Mice and Men: How Temporal and Other Biological Factors Affect Interpretability of Animal Models March 1, 2013 Gene G. Kinney, Ph.D. Prothena Biosciences American Society for Experimental NeuroTherapeutics | 15 th Annual Meeting

2. Disclosure Name of Commercial Interest Bristol-Myers Squibb; Merck & Co.; Elan Pharmaceuticals; Janssen Alzheimer Immunotherapy Prothena Biosciences, Inc. American Society for Experimental NeuroTherapeutics | 15th Annual Meeting Type of Financial Relationship Former employee and current shareholder (Elan and Bristol-Myers Squibb) Current employee and shareholder

3. Learning Objectives Using animal models of Alzheimer’s disease and anti-Aβ immunotherapy as a case study, illustrate: The strengths and limitations of animal model results for predicting clinical efficacy The importance of iterative translation during clinical development American Society for Experimental NeuroTherapeutics | 15th Annual Meeting ≠ http://3.bp.blogspot.com/- 8mjLTwgy_3w/UG2cW1Da9VI/AAAAAAAAEME/Mw3nTKKAWoA/s1600/dem entia_83248459.jpg

4. Exemplifying the obvious Most cases of Alzheimer’s disease are diagnosed § at age ≥ 65 yrs with a median survival of 4-8 years post-diagnosis It is generally believed that Alzheimer’s disease pathology can begin ≥ 20 yrs prior to diagnosis In a “modestly aggressive” mouse model of AD (PDAPP) the average age for onset of pathology is ~12 months of age* The average lifespan of a PDAPP mouse is ~18-32 mos * some deficits generally attributable to soluble Aβ aggregates are observed pre-pathology § Traditionally a differential diagnosis based on dementia. Recent move to more biologically-based definitions. 4 http://ars.els-cdn.com/content/image/1-s2.0-S0924977X02001037-gr1.jpg

5. Alzheimer ’ s disease biology First identified in 1907 by Alois Alzheimer based on post-mortem evaluation of demented patients – Extracellular plaques and intracellular tangles; but cause or consequence? Amyloid β Tau plaques tangles Kinase inhibitors (phosphorylation) Microtubule stabilization Production inhibitors Immunotherapeutics Clearance enhancers Anti-aggregate and “ plaque buster ” Brunden et al., Nature Reviews Drug Discovery 8, 783-793 (October 2009) Auguste D Alois Alzheimer 5

6. Development of the PDAPP mouse model enabled initial immunotherapy studies Focus on pathological contributions to disease Development of the first plaque forming mouse model was enabling for the testing of putative therapeutic approaches designed to be “disease modifying” Plaque formation Loss of synaptic integrity Non-TgPDAPP Loss of neuronal integrity Games et al., Nature, 373:523, 1995 6

7. Anti-Aβ immunotherapy: Background Anti-Aβ immunotherapy proposed as a therapeutic approach for the removal of Aβ from the CNS (Schenk et al, Nature, 400:173, 1999) UTC = untreated controls SAP = immunization with serum amyloid P PDAPP mice immunized between 6 wks and 13 mos ImmunizedVehicle Reductions in dystrophic neurites and astrogliosis also observed following immunization 7

8. Translational aspects of anti-Aβ immunotherapy 8

9. Anti-Aβ immunotherapy: Background Elan/Wyeth anti-Aβ vaccine (AN-1792) demonstrated activity on some endpoints in clinical trials – Reduction of senile plaque and cognitive benefit on NTB 9

10. Anti-Aβ immunotherapy: Background AN-1792 trials were discontinued following reports of meningoencephalitis in ~6% of the active treatment group – T-cell mediated response to the self-antigen is a likely cause 10

11. Translational aspects of anti-A β immunotherapy 11

12. Next generation approaches include passive and active immunization Schenk, Nature Reviews Neuroscience 3, 824-828 (October 2002) 12

13. TY11/15 3D6 Median Aβ burden values following passive immunization: 6 mo treatment 10.23% 0.69% Treatment% control TY11/15100 3D6 7 PDAPP mice (N>30/group) treated with 3 or 10mg/kg/week at 12-18 months of age Similar effects on: Neuritic dystrophy Brain Aβ levels (ELISA) Seubert et al., Neurodegenerative Dis, 5:65, 2008 13

14. The discovery of the Pittsburgh Compound-B (PiB) tracer allowed for PET imaging of amyloid deposition in AD patients 14

15. Bapineuzumab decreases amyloid burden in human subjects using [ 11 C] PiB PET imaging Rinne JO et al., Lancet Neurol., 9:363, 2010 N=28 patients with mild to moderate AD Randomized to intravenous bapineuzumab (N=20) or placebo (N=8) in three ascending dose cohorts (0.5, 1.0, or 2.0 mg/kg) 15

16. MRI observations consistent with transient edemic events were observed following Bapineuzumab treatment Salloway, S. et al. Neurology 2009;73:2061-2070 69-year-old APOE ε 4 homozygote Female Treated with bapineuzumab 1.0 mg/kg IV. Remained asymptomatic despite the appearance of multiple areas of ARIA evident on MRI. Patient was redosed at 0.5 mg/kg and followed for over 2 years without recurrence of ARIA Reported by Salloway et al., 2009 (Phase 2 data) Anti-A β edemic events appear sensitive to Bapineuzumab dose and APOE ε4 gene dose Events were transient in some cases and did not recur following redosing in a limited data set 16

17. Clinical evidence suggests that ARIA may be related to Aβ mobilization 17

18. Translational aspects of anti-A β immunotherapy 18

19. Preclinical studies designed to further understand the biology of ARIA The brain capillary network appears to be involved in antibody induced amyloid removal A β deposition on brain vasculature may impair vessel integrity, recovery following removal of vascular A β Key proteins involved in resolution of edemic events may be modulated during the process of immunotherapy induced clearance of A β Zago et al., Alz. & Dement., In Press Normal Vascular Aβ 19

20. More work to do – MMSE; ADAS-Cog; CDR-SB etc. How is the disease defined? – Dementia, Pathology, mix? Staging of disease -Spatial reference learning / memory; Episodic memory Alzheimer’s diseasehAPP Tg mouse models Stage at Diagnosis Generally believed to be advanced pathology at point of diagnosis Not applicable; diagnosis based on pathological involvement Optimal point of intervention Extensive efforts to identify earlier intervention times All currently completed disease modification trials have studied mild/moderate patients Data suggests that earlier intervention is more effective Late intervention may still be effective for some experimental approaches Alzheimer’s diseasehAPP Tg mouse models Dementia Assessed by broad based cognitive tools that may also incorporate function (e.g., MMSE, ADAS-Cog; CDR-SB); mixed dementia also possible Models are primarily driven by pathology endpoints; assessment of cognition tends to be very precise (e.g., spatial reference learning/memory; episodic memory, etc.) PathologyLikely mixed (e.g., tau, α-synuclein, etc) Homogenous population, pathology driven in specific tissues through use of promoters; often involves point mutations for more aggressive phenotype 20

21. Even more work to do – MMSE; ADAS-Cog; CDR-SB etc. How do key aspects of intervention therapy translate? – Dementia, Pathology, mix? -Spatial reference learning / memory; Episodic memory Alzheimer’s diseasehAPP Tg mouse models Magnitude of insol. Aβ reduction Modest reduction from baseline in treated patients. Continued accumulation in placebo patients. ~70 to ≥90% reductions depending on time of intervention relative to pathology onset Dose Level Dose of bapineuzumab (and other agents?) may be limited by ARIA Clear relationship between dose level, time of intervention and duration of treatment 21

22. Contextual Fear Conditioning has been developed as a sensitive functional cognitive measure Comery et al., J. Neurosci. 2005;25:8898-8902 1mAmp 120s 2s 120s 2s 30s Training (Day 1) Context Shock 22

23. Common Clinical Assessment Tools 23 http://www.mdanderson.org/publications/conquest/issues/2008-fall/old-drugs-new-possibilities-conquest-fall-2008-old-drugs-new-possibilities.html http://ginny-livingwithlyme.blogspot.com/2012/10/spinal-tap-3.html http://www.nimh.nih.gov/images/pubs/neuro-pet.jpg

24. Summary Although many (if not most) human diseases cannot be fully recapitulated by animal models, key aspects of the disease process are reproduced – Key breakthroughs in biological understanding of the disease processes – Useful models for testing interventional therapy Animal models are most effectively utilized when there is a full understanding of the limitations around the translation of the models to human disease – How is “efficacy” defined both pre-clinically and clinically – Clinical trial design limitations must be considered Heterogeneity of human subjects Limitations around invasive endpoint assessment Potential disconnect between “biological” and “pathological” disease definitions Clinical development will often result in unpredicted findings necessitating iterative translational approaches – AN-1792 immunization in AD patients resulted in meningoencephalitis in ~6% of the active treatment group New approaches to avoid Aβ directed T-cell responses are being tested clinically (passive and active immunization) – Reports of radiologically identified edemic events following passive immunization have led to additional preclinical studies using the PDAPP mouse model 24

25. Acknowledgements Elan Pharmaceuticals / Prothena Biosciences / Janssen AI Robin BarbourBob Brashear Manuel ButtiniMing Chen Dora GamesHenry Grajeda Michael GrundmanTerry Guido Stefan HeylenKaren Khan Mike LeeEnchi Liu Ruth MotterDale Schenk Sally SchroeterPeter Seubert Eric YuenWagner Zago Wyeth / Pfizer Davinder Gill Steven Jacobsen Tom Comery Menelas Pangalos 25