Structured Immuno-Oncology Combination Strategies To Maximize Efficacy

Jun Wang MD, PhD

Senior Medical Director

Immunotherapy Combinations

Roche Cancer Immunotherapy Franchise

1 Disclosures

• Employee of Roche • The information contained herein may refer to the use of products for indications other than those approved and/or listed in the Summary of Product Characteristics/USPI, or relating to molecules currently undergoing experimental trials • The issues addressed are not meant to suggest that the product be employed for indications other than those authorised/approved The Cancer-Immunity Cycle

3 Priming and activation 4 Trafficking of T‐cells to tumors

Active T cell Active T cell 5 Infiltration of Accessing the tumor T‐cells into Dendritic Initiating and propagating tumors cell anti‐cancer immunity TUMOR MICROENVIRONMENT 6 Recognition Cancer‐cell recognition and initiation of cytotoxicity of cancer Antigens cells by T‐ Apoptotic tumor cell 2 Cancer antigen cells presentation

Tumor cell 1 Release of cancer cell 7 Killing of cancer cells antigens The role of PD-L1 in the tumor microenvironment Modified from Chen and Mellman, Immunity 2013 Broad Pan-Tumor Potential with anti-PD-L1/PD-1 inhibitors: Approximate Monotherapy ORR in All-comers

So what is behind these differences acrossBroad tumor activity, types, and but between only a individualssubset of inpatients the same benefit indication?

Modified from D. Chen, BioScience Forum, 2015 PD-L1 is a Critical Source of Immune Suppression in Cancer: Differential PD-L1 Expression Patterns are Seen in Patients Both tumor & immune cells Immune cells only – (IC) Tumor cells only – (TC) (TC & IC) AdaptiveIntrinsic Adaptive & Intrinsic

Predictive of benefit in Predictive of benefit in lung bladder cancer cancer (ORR/PFS/OS)2 (ORR/OS)1

WCLC 2015 1IMvigor 210 ECC 2015, 2POPLAR ECC 2015 POPLAR (2L/3L NSCLC Post-Platinum Progression): mOS Efficacy increasing with higher PD-L1 expression

Updated analysis (Event / N=70%): Minimum follow-up 20 months

Updated median OS (95% CI), mo

Atezolizumab Docetaxel Subgroup (% of enrolled patients) n = 144 n = 143

0.45 TC3 or IC3 (16%) NE (9.8, NE) 11.1 (6.7, 14.4) 0.50 TC2/3 or IC2/3 (37%) 15.1 (8.4, NE) 7.4 (6.0, 12.5) 0.59 TC1/2/3 or IC1/2/3 (68%) 15.1 (11.0, NE) 9.2 (7.3, 12.8) 0.88 TC0 and IC0 (32%) 9.7 (6.7, 12.0) 9.7 (8.6, 12.0)

ITT (N = 287) 0.69 12.6 (9.7, 16.0) 9.7 (8.6, 12.0)

0,10.2 11 2 Hazard Ratioa

In favor of In favor of docetaxel Smith, et al. ASCO 2016

a Stratified HR for ITT and unstratified HRs for PD-L1 subgroups; NE, not estimable; Data cut-off: December 1, 2015 6 OAK (2L/3L NSCLC POST-PLATINUM PROGRESSION) OS BY PD-L1 EXPRESSION Median OS, mo On-study Prevalence Atezolizumab Docetaxel Subgroup 0.41 n = 425 n = 425 16% TC3 or IC3 20.5 8.9 0.67 31% TC2/3 or IC2/3 16.3 10.8 0.74 55% TC1/2/3 or IC1/2/3a 15.7 10.3 0.75 45% TC0 and IC0 12.6 8.9

0% 20% 40% 60% 80% 100% 0.73 100% ITTa 13.8 9.6

0.20,2 1 2 Hazard Ratioa

In favor of In favor of atezolizumab docetaxel

7 aStratified HR for ITT and TC1/2/3 or IC1/2/3. Unstratified HR for subgroups. TC, tumor cells; IC, tumor-infiltrating immune cells; OS, overall survival. Barlesi et al, Atezolizumab Phase III OAK Study. http://tago.ca/9Hh Yesterday---Oct 18, 2016 IMvigor210: Cohort 2 (Platinum Treated Bladder Cancer) Ongoing & durable responses across all subgroups

IC2/3 IC1/2/3 Alla IC1 IC0 (n = 100) (n = 207) (N = 310) (n = 107) (n = 103) ORR: confirmed IRF RECIST 28% 19% 16% 11% 9% v1.1 (95% CI) (19, 38) (14, 25) (12, 20) (6, 19) (4, 16) CR rate: confirmed IRF RECIST 15% 9% 7% 4% 2% v1.1 (95% CI) (9, 24) (6, 14) (4, 10) (1, 9) (0, 7)

Median follow-up: 17.5 months (range, 0.2+ to 21.1 mo)

• 71% of responses (35/49) were ongoing – 86% of CRs ongoing • mDOR was not yet reached in any PD-L1 IC subgroup (range, 2.1+ to 19.2+ mo)a

CR as best response PR as best response First CR/PR

as Best ResponseasBest Treatment discontinuationb c Patients With CR or PR PR or CR With Patients Ongoing response

0 2 4 6 8 10 12 14 16 18 20 Months a Per IRF RECIST v1.1 b Discontinuation symbol does not indicating timing. c No PD or death only. Data cutoff: Dreicer, et al 2016 ASCO Mar. 14, 2016. 9 May 18, 2016 PD-L1/PD-1 targeted therapies work best in inflamed tumors vs. non-inflamed tumors

“Inflamed” “Non-inflamed” • Tumor-infiltrating Lymphocytes • PD-L1 expression • CD8+ T cells • Genomic instability • Pre-existing immunity

Typically respond Typically DO NOT favorably to respond to checkpoint inhibition checkpoint inhibition

Converting from ‘non-inflamed’ to ‘inflamed’ is likely to be only 1 piece of the puzzle…. Using clinical patient data to understand why some patients respond better than others

Atezolizumab Phase I data: Urothelial Bladder Cancer Progressive Disease (PD) Why do many patients not respond? • No pre-existing immunity? • Inability to induce immunity? • Multiple negative regulators?

Stable disease (SD) • Insufficient T cell immunity? • Microenvironment barriers? • Can combination therapy improve the quality of response? • Should we be driving these patients to deeper responses?

Partial/Complete responders (PR/CR) What are the drivers of single-agent responses? • Pre-existing immunity? • Decreased negative regulators? • Is continuous treatment required? T. Powles et al (2014) Nature • How can PRs be enhanced to CRs Besides PD-L1 expression in TCs vs. ICs, there are also differences in the tumor microenvironment with respect to the “The T-Cell Army” vs. “The Tumor”

3 General Scenarios

The T-cell army by Jerome Groupman The New Yorker, April 23rd, 2012

D.S. Chen 2014 EORTC-NCI-AACR Molecular Targets Meeting The T-Cell Army vs. The Tumor: Scenario 1 – “Armed and Ready”

Cancer

T-cell Army Baseline On-Treatment

PD-L1 CD8 PD-L1 CD8 Herbst et al. Nature 2014 Herbst et al. Nature 2014

D.S. Chen 2014 EORTC-NCI-AACR Molecular Targets Meeting The T-Cell Army vs. The Tumor: Scenario 2 – The T-Cell Barrier (“Can’t Cross the River”)

Cancer

T-cell Army

CD8 PDL1

Herbst et al. Nature 2014

D.S. Chen 2014 EORTC-NCI-AACR Molecular Targets Meeting The T-Cell Army vs. The Tumor: Scenario 3 – “The Immune Desert”

Cancer

No T-cell Army

CD8 PDL1

Molinero, et al. unpublished data on file

D.S. Chen 2014 EORTC-NCI-AACR Molecular Targets Meeting Findings: The Tumor Immunity Continuum

Inflamed Non-Inflamed

Pre-existing Immunity Excluded Infiltrate Immunologically Ignorant

Mutational Load

TILs Angiogenesis Proliferating CD8 T cells/IFN Reactive stroma Tumors/ PD-L1 MDSCs Low Class I

Respond favorably to Convert to inflamed phenotype with combinations checkpoint inhibition

Hegde PS et al. (2016) Clin Canc Res The Cancer Immunity Cycle can be further grouped to help us to understand tumor mechanisms of resistance

RECRUIT / INFILTRATE (vasculature ) Non- ACTIVATE Inflamed (central) Non- Inflamed KILL CANCER CELLS (tumor) Inflamed

Source: Chen DS, Mellman I. Immunity. 2013. Overlaying the 3 immune blockade phenotypes on the Cancer Immunity Cycle

• How can we modulate tumor specific antigen presentation to promote T-cell IMMUNE EXCLUDED specific tumor responses CD8+ T cells • Can we generate new T cell responses accumulated therapeutically? but have not efficiently infiltrated RECRUIT / INFILTRATE (vasculature IMMUNE DESERT ) • What are the factors CD8+ T cells Non- are absent ACTIVATE Inflamed preventing T cell from tumor (central) infiltration? and its Non- periphery Inflamed KILL CANCER CELLS INFLAMED (tumor) Inflamed CD8+ T cells infiltrated, but are non- • How do we enhance functional checkpoint inhibitor effects?

Source: Chen DS, Mellman I. Immunity. 2013. Cancer Immunotherapy

The Future: Driving towards Personalized Cancer Immunotherapy An Example of Targeting a Mechanism of Immune Escape Tumor Cell Recognition and T cell Activation: T-cell bispecifics (TCBs)

PRIMING AND ACTIVATION 4 T‐CELL TRAFFICKING

• Anti-OX40 • Anti-41BB • Anti-CTLA4 • Anti-cytokine 3 T‐CELL INFILTRATION • Anti-CD27 5 • Anti-VEGF • Anti-Ang2/VEGF

ANTIGEN PRESENTATION

• Anti-CD40 • Neo- • IFN-α epitope 2 CANCER CELL RECOGNITION • Oncolytic virusesvaccine • Bi-specifics• CAR-T

6 • ImmTACs • BiTes

ANTIGEN RELEASE

• EGFRi • Chemotherapy CANCER CELL RECOGNITION

• ALKi • HDAC • Anti-PDL1 • IDOi • A2Ai

• BRAFi • Radiotherapy • Anti-PD-1 • Anti-TIGIT • IDO/TDOi 1 7 • MEKi • Anti-CSF- • Anti-TIM3 1R • Anti-LAG3

Chen & Mellman. Immunity 2013 Substantial investments in underlying technology Novel antibody platforms supporting research

Monoclonal Antibody drug Glyco-engineered Fab fragments Antibodies with antibodies (MAb) conjugates (ADC) antibodies modified Fc part

Antibody cytokine (1) Bispecific (2) Bispecific (1) T cell (2) T cell fusion proteins antibodies (biMAb) antibodies (biMAb) bispecifics bispecifics (FP) An Example of maximizing the value in CIT: Novel assets and combinations cergutuzumab amunaleukin Immunotherapy portfolio emactuzuma aFAP-IL2v FP chemo SERD b aCD40 taselisib aOX40

ipataserti IDOi b aCEA/CD3 TCB

aCD20/CD3 TCB

aTIGIT

Launched portfolio

Combination approved Chemo combination approved

Combination in development idasanutlin Chemo combination in development azacitidine Roche NME late stage polatuzumab vediotin Roche NME early stage

Non-Roche apporved drugs Status: June 2016 lenalidomide

emactuzumab (aCSF-1R); cergutuzumab amunaleukin (aCEA-IL2v FP); vanucizumab (aAng2/VEGF); polatuzumab vediotin (aCD79b ADC); taselisib (PI3Ki); ipatasertib (AKTi); SERD (selective estrogen receptor degrader); idasanutlin (MDM2 antagonist); Venclexta in collaboration with AbbVie; Gazyva in collaboration with Biogen; Alecensa in collaboration with Chugai; Cotellic in collaboration with Exelixis; Zelboraf in collaboration with Plexxikon; polatuzumab in collaboration with Seattle Genetics; ipatasertib in collaboration with Array Biopharma; IDOi in collaboration with NewLink; daratumumab in collaboration with Janssen (J&J) uueVision Future approaches immunotherapy through personalizedcancer all tumortypes across cure rates Increasing cancer Increasing Cure Rate tzlzmb(2016) Atezolizumab (2014) (2014) (2011) (2011-2016) Checkpoint Inhibitors Monotherapy cotellic) (chemo, bev,K, + Atezolizumab (chemo or…) Nivolumab + +(chemo or …) Pembrolizumab +Ipi (2015) Nivo (2015-2020) Combine withExisting Tx tzlzmb+(muo1 muo2) 1,Immuno +(Immuno Atezolizumab …) 2, Immuno 1, (Immuno Competition + (2020-2025) Expand BeyondCheckpoint Inhibitors xsub-groups Dx Multiple combostargetedat Personalized CIT (2025+) at today Where weare With progress comes increasing complexity Example: Implications on clinical trials

Targeted Targeted medicines + Chemotherapy medicines immunotherap y

+

Clinical Individual patients trial Unspecified / Patient sub-groups / populatio Large / Medium Medium - Small n/size Comprehensive genomic Need for Single disease No diagnostics sequencing & Dx marker response monitoring Phase-less Developme Phase I, II, III Phase I, II, III basket / umbrella nt process studies Vision: personalized cancer immunotherapy algorithms FMI 2nd Gen platform- FOUNDATIONCI

DNA repair/proliferation EMT Checkpoint mono Reactive Stroma TGF Algorithm 1 Disease subtypes Checkpoint chemo combo Algorithm 2 Tregs FOUNDATIONCI Viral load MHC

Chemokines Mutational load Myeloid cell Angiogenesis

Hypothetical

• Atezo trials, legacy trials in CRC, TNBC, Gastric, GBM, Ovarian Cancer, immune landscaping efforts across tumors Vision: personalized cancer immunotherapy algorithms

*Possible hypothetical algorithm:

Evaluate tumor: is the tumor inflamed?* Y N Inflamed Non-inflamed

1 2 3 4 1 2 3 4 Strong PD-L1 Weak PD-L1 No PD-L1 No identifiable Are T cells at MHC No T No identifiable immune tumor loss? cells? immune targets targets periphery?

Are suppressive IDO/kyneurinin Tumor antigen Antigen myeloid cells present? expressed? expression? experienced?

Anti-PDL1/PD1 Anti-PDL1/PD1 Anti-PDL1/PD1 Anti-PDL1/PD1 Anti-PDL1/PD1 Anti-PDL1/PD1 Anti-PDL1/PD1 Anti-PDL1/PD1 plus plus plus plus plus plus plus Anti-CSF1R IDO inhibitor Chemo Anti- T cell Anti-OX40 Chemo Radiotherapy angiogenics bispecifics Anti-CTLA4 Radiotherapy Targeted Anti-CD40 Targeted therapy Anti-CEA-IL2v therapy Vaccines

Chen CRI EATI CIMT AACR 2015 Summary

• Differential PD-L1 expression is observed between tumor and immune cells in patients, and is associated with distinct biologies and clinical outcomes

• There are several distinct PD-L1 expression patterns observed in individuals: 1. “Armed and Ready” 2. “The Immune Desert” 3. “The T-Cell Barrier”/”Can’t Cross the River”

• Emerging clinical data suggests agents targeting various aspects of the cancer immunity cycle may alter the tumor micro-environment, leading to increased anti-tumor responses and potentially enhanced clinical responses

• However, further understanding around the biology, individual immune status & tumor micro-environments is needed, which may help further optimize our future combination strategies Acknowledgements

Genentech/Roche thanks all the patients/families and physicians who have participated in our clinical trials to advance our scientific understanding of cancer immunotherapies

Partial listing of the /Roche “T Cell Army”: • Jeong Kim • Roel Funke Genentech/Roche • Ina Rhee • Daniel Waterkamp Cancer Immunotherapy Committee • Lisa Damico Beyer • Jeff Wallin • Cathi Ahearn •Ed Cha • Jose Saro • Dietmar Berger • Marcella Fasso • Daniel Brickman • Daniel Chen Atezolizumab Franchise •PritiHegde Team • Michele Livesey • Cathi Ahearn • Priti Hegde • Stuart Lutzker • Daniel Brickman • Lisa Liermann • Ira Mellman • Daniel Chen • Sushil Patel • William Pao • Mark Davis • Alan Sandler • Alessandra Polara • Gregg Fine • Paul Woodard • Kevin Sin • Bill Grossman • Eric Hoefer • Robin Taylor • Pablo Umana

Genentech and Roche Investigators, Cancer Immunotherapy Working Group and Biomarker Members