1. 許駿 醫師 台大醫學院 腫瘤醫學研究所 台大醫院 腫瘤醫學部 台灣大學 癌症中心醫院
TJCC Mini-symposium for cancer immunotherapy
06 May 2017
Biomarker research: what we have learned from trials of immune checkpoint inhibitors
許駿 醫師
台大醫學院 腫瘤醫學研究所
台大醫院 腫瘤醫學部
台灣大學 癌症中心醫院

2. Conflict of interest disclosure
Research grant Ministry of Science and Technology (Taiwan), Celgene
Honorarium (speaker/advisor/travel) AstraZeneca, Bayer, Bristol-Myers Squibb, Eli Lilly, MSD, Novartis, Roche, TTY Biopharm

3. Landmark immunotherapy approval by FDA
Merkel cell
RCC
Melanoma
NSCLC
Bladder ca.
Hodgkin’s
HNSCC
Prostate
ALL
* Accelerated approval
Avelumab*
IL-2
Nivo.
Nivo.*
Ipi. + Nivo.*
IFN
Peg-IFN
IL-2
Ipi.
Ipi.
Pembro.*
Pembro.
T-Vec
Faster drug development/ approval
Extended indication
Nivo.
Atezo.
Pembro.
Pembro.
Nivo.*
Atezo.*
BCG
Atezo.*
Durva.*
Nivo.*
Pembro.*
Pembro.*
Nivo.
Sipuleucel-T
Blinatumomab*
1990
1992
1998
2010
2011
2014
2015
2016
2017
Chemokines/ cytokines
Cell therapy/ vaccines
Immune agonists
Non-specific
Checkpoint inhibitors
Ipilimumab (anti-CTLA4) nivolumab, pembrolizumab (anti-PD1)
Atezolizumab, avelumab, durvalumab (anti-PD-L1)
(accessed 02 May 2017; Modified from: Craft JE, ASCO 2016 and Hoos Nat Rev Drug Discov 2016)

4. Biomarkers research: what we have learned from trials of immune checkpoint inhibitors?
Types of biomarkers
(arbitrary)
Examples
Static
Mutation loads (carcinogen exposure, mismatch repair deficiency, etc.)
Specific genetic defects: β-catenin, EGFR, JAK1/2, PTEN loss, MDM2/4 , etc.
Dynamic
PD-L1 expression (tumor cells, immune cells)
Immune-response gene expression
Chaotic
Gut microbiome
Predictors for (1) efficacy; (2) resistance; (3) ‘hyper-progression’

5. What we have learned from trials of immune checkpoint inhibitors
What we have learned from trials of immune checkpoint inhibitors? Anti-PD-1/ anti-PD-L1 as examples
Objective response rate >50% (e.g., refractory Hodgkin’s disease)
Objective response ca. 20% (e.g., NSCLC, HNSCC, urothelial ca.)
Objective response rate < 10% (e.g., pancreatic adenocarcinoma, colon cancer)

6. New hope for refractory cancers (1) Hodgkin’s disease
23 patients
Nivolumab 3 mg/kg q 2 weeks
210 patients
Pembrolizumab 200 mg q 3 weeks
Cohort A
Cohort B
Cohort C
Ansell SM, et al. N Engl J Med 2015; 372: Chen R, et al. J Clin Oncol (E-pub Apr. 2017)

7. What we have learned from trials of immune checkpoint inhibitors
What we have learned from trials of immune checkpoint inhibitors? Anti-PD-1/ anti-PD-L1 as examples
Objective response rate >50% (e.g., refractory Hodgkin’s disease)
Objective response ca. 20% (e.g., NSCLC, HNSCC, urothelial ca.)
Objective response rate < 10% (e.g., pancreatic adenocarcinoma, colon cancer)

8. New hope for refractory cancers (2) Cancers with mismatch-repair (MMR) deficiency
Immune effector cells
Mutation loads
Le DT et al. N Engl J Med 2015; 372:

9. The mutation load theory:
High mutational loads → neoantigens → host antitumor immune response → biomarkers
Melanoma, lung cancers, MMR deficient cancer
0.01
0.1 1 10
100
1000
Frequently
Somatic mutation prevalence per megabase
Regularly
Occasionally
Schumacher TN et al. Science 2015; 348: 69-74

10. Neoantigens → host immune response in melanoma
由同一病患的T 細胞中辨識可誘發免疫反應的neoantigens
Carreno BM et al. Science 2015; 348: 803-9

11. Mutation/neoantigen loads and heterogeneity
‘All mutations are not created equal’
Phylogenetic analysis based on multi-regional exome sequencing
Hiley C, et al. Genome Biol 2014; 15: 453 McGranahan N, et al. Science 2016; 351:

12. Mutation/neoantigen loads, heterogeneity, and response to checkpoint inhibitor therapy
High clonal neoantigen loads and low heterogeneity (< 5% sub-clonal antigens) predicted durable response.
Sub-clonal neoantigens (spontaneous or induced) were not associated with treatment efficacy.
Multiple immune evasive mechanisms, e.g., MHC loss.
McGranahan N, et al. Science 2016; 351:

13. What we have learned from trials of immune checkpoint inhibitors
What we have learned from trials of immune checkpoint inhibitors? Anti-PD-1/ anti-PD-L1 as examples
Objective response rate >50% (e.g., refractory Hodgkin’s disease)
Objective response ca. 20% (e.g., NSCLC, HNSCC, urothelial ca., and many others)
Objective response rate < 10% (e.g., pancreatic adenocarcinoma, colon cancer)
Broad-spectrum anti-tumor activity
‘Universal’ mechanism of action?
‘Universal’ biomarker?

14. Kerr K, et al. J Thor Oncol 2015; 10: 985- 9
PD-L1 expression as biomarkers: Anti-PD-1/ anti-PD-L1 in non-small cell lung cancer
Kerr K, et al. J Thor Oncol 2015; 10:

15. 感謝 台大醫院腫瘤醫學部 高祥豐醫師 提供

16. Teng MW, et al. Cancer Res 2015; 75: 2139-45

17. PD-L1 expression: tumor cells vs. immune cells (1)
PD-L1 expression: tumor cells vs. immune cells (1)? Atezolizumab in NSCLC
Fehrenbacher L, et al. Lancet 2016; 387:

18. PD-L1 expression: tumor cells vs. immune cells (2)
PD-L1 expression: tumor cells vs. immune cells (2)? Atezolizumab in NSCLC
Rosenberg JE, et al. Lancet 2016; 387:

19. PD-L1 expression: tumor cells vs. immune cells (3)
PD-L1 expression: tumor cells vs. immune cells (3)? Pembrolizumab in HNSCC
IFN- 6-gene signature
CXCL9, CXCL10, IDO1, IFNG, HLA-DRA, STAT1
Composite score calculated by averaging normalized† values for each gene

20. Challenges of interpreting PD-L1 expression (1)
Constitutive
Genetic and epigenetic regulation
Induced
‘immunogenic’ tumors or therapy
Tumor-specific T cell response and interferon-γ production
Value of baseline (pre-treatment) samples questionable
Serial tumor sampling seldom feasible
Modified from Ribas A, et al. J Exp Med 2016; 213:

21. Challenges of interpreting PD-L1 expression (2)
Tissue markers: dynamic and heterogeneous
Functional imaging:
Non-invasive
PD-1/PD-L1 labeling
Immune cell labeling
Sensitivity/ specificity
McLaughlin J, et al. JAMA Oncol 2016; 2: 46-54

22. EGFR activating mutations associated with poor efficacy of anti-PD1/antiPD-L1 therapy
Treatment
Patient no.
Overall survival (mon)
EGFR mutation (%)
EGFR wild-type (%)
Nivolumab vs. docetaxel
582
12.2 vs. 9.4 14 58
Pembrolizumab vs. docetaxel
1034
10.4 vs vs. 8.5 8 85
Atezolizumab vs. docetaxel
287
12.6 vs. 9.7 6 51
Lee CK et al. J Thor Oncol 2017; 12: 403-7

23. Genetic aberrations of lung cancer and the immune microenvironment
EGFR-mutant
(N = 62)
KRAS-mutant
(N = 65)
P value
PD-L1 positivity
PD-L1+ ≥ 50%
7 (11%)
11 (17%)
0.449
PD-L1+ ≥ 5%
10 (16%)
20 (31%)
0.062
CD8+ TILs (image-based) per mm2
Median
185.1
330.1
0.011
(Range)
(6.1–1,161.9)
(8.5–2,567.3)
Concurrent PD-L1 expression & CD8+ TILs
PD-L1+ ≥ 50% & high CD8+ TILs
2/46 (4.3%)
10/56 (18%)
0.061
PD-L1+ ≥ 5% & high CD8+ TILs
15/56 (27%)
0.003
Gainor JF, et al. Clin Cancer Res 2016; 22:

24. -catenin pathway activation correlates with T-cell exclusion in melanoma
T-cell exclusion is caused by impaired priming of anti-tumor T cells and reduced numbers of CD103+ dermal dendritic cells.
Wnt/-catenin signaling induces expression of ATF3, which suppresses CCL4 and thereby interferes recruitment and activation of CD103+ dendritic cells.
Spranger S. et al Nature 2015;523:

26. Does ‘hyper-progression’ after immunotherapy really happen?
Review of tumor growth kinetic ratio (TGKR) in 34 HNSCC patients treated with anti-PD1/anti-PD-L1
TGKR > 2 more common in patients with cervical LN metastasis and regional recurrence
Saâda-Bouzid E. et al, 2017
NGS data from 155 stage IV cancer patients who received immunotherapies
TTF <2 months associated with MDM2/4, EGFR, or DNMT3A alterations
MDM2/MDM4 amplification (4 of 6) and EGFR alterations (3 of 10) associated with ‘hyper-progression’ (TTF <2 months, >50% increase in tumor burden,>2-fold increase in progression pace)
Kato S, et al. 2017
Saâda-Bouzid E. et al Ann Oncol 2017 (E pub) Kato S, et al. Clin Cancer Res 2017 (E pub)

27. Chen DS, Mellman I. Nature 2017; 541: 321-30

29. Anti-PD-1-based combination therapy for NSCLC
KEYNOTE 021G
CHECKMATE 012 Tx
Chemo*
Chemo* + Pembro
Chemo* +
Nivo n 63 60 15
ORR
29%
55%
47%
PFS
8.9 months
13.0 months
(HR: 0.53, p=0.10)
6.8 months OS
HR: 0.90, p=0.39
19.2 months
PD-L1 status and ORR
PD-L1>1%: ORR: 54%
PD-L1<1%: ORR: 57%
PD-L1>1%: ORR: 48%
PD-L1<1%: ORR: 43%
Langer C, Lancet Oncol 2016;17: 1497–508
Rizvi NA, J Clin Oncol 2016;34:
* Carboplatin + pemetrexed
感謝 台大醫院腫瘤醫學部 高祥豐醫師 提供

30. Atezolizumab (anti-PD-L1) + bevacizumab (antiangiogenic) vs
Atezolizumab (anti-PD-L1) + bevacizumab (antiangiogenic) vs. sunitinib for renal cell carcinoma (1)
McDermott D, et al. AACR 2017

31. Atezolizumab (anti-PD-L1) + bevacizumab (antiangiogenic) vs
Atezolizumab (anti-PD-L1) + bevacizumab (antiangiogenic) vs. sunitinib for renal cell carcinoma (2)
McDermott D, et al. AACR 2017

32. Atezolizumab (anti-PD-L1) + bevacizumab (antiangiogenic) vs
Atezolizumab (anti-PD-L1) + bevacizumab (antiangiogenic) vs. sunitinib for renal cell carcinoma (3)
McDermott D, et al. AACR 2017

33. T-cell complexities = more drug targets

34. Challenges of future immunotherapy trial design and biomarker development
Selection of the ‘optimal’ combination modalities
Identification of (immune) therapeutic targets
Comprehensive vs. specific biomarkers
Development of preclinical models
Collaboration beyond the biomedical community

35. ‘I do not fancy I know what I do not.’
You can't connect the dots looking forward; you can only connect them looking backward.
You have to trust that the dots will somehow connect in your future.
Steve Jobs, Stanford commencement address, 2005
‘I do not fancy I know what I do not.’
Socrates, in Plato’s Apology

36. 謝謝大家！
感謝 盧建宏醫師提供