Emerging biomarkers for immunotherapy in lung cancer Prof Keith M Kerr Department of Pathology Aberdeen University Medical School, Aberdeen Royal Infirmary Aberdeen, UK Why Do We Need Biomarkers for Immunotherapy? Precision medicine is a reality for many tumour types Avoidance of harm? There are toxicities from these drugs Is there a subgroup of patients who fair worse on I-O treatment? Alternative treatment would be better Not all patients respond to and benefit from these treatments Enrich the treatment population for benefit How many to treat, to get one response? Benefit is relative to standard of care Financial burden of expensive therapy I-O, immuno-oncology. 3 Disclosures Consultancy AbbVie, AstraZeneca, Boehringer Ingelheim, Bristol-Myers Squibb, Eli Lilly, Merck Serono, Merck Sharp & Dohme, Novartis, Pfizer, Roche, Ventana Honoraria (speaker) AstraZeneca, Boehringer Ingelheim, Bristol-Myers Squibb, Eli Lilly, Merck Serono, Merck Sharp & Dohme, Novartis, Pfizer, Roche, Ventana Therapeutic Aims & Assumptions Inhibit the interaction of PD-1 and PD-L1 In order to take the brakes off an existing tumourspecific immune response We hope this tumour-specific immune response actually exists This requires immunogenicity Which requires antigenicity (neoantigens) to be visible to the immune system A surrogate for probable neoantigen load is tumour mutational burden (TMB) Potential Biomarkers THIS is the drug target! PD-1~PD-L1 Interaction is present Specific immunity is available the tumour is inflamed Whole tumour mutation burden: Whole Exome Sequencing 2 Numbers of mutations in diagnostic panels 4 1
Other possible factors impacting therapy response Host factors General immune status Gut microbiome The Lottery Model The Cancer Immunogram Blank CU et al. Science 2016 PD1~PD-L1 inhibition is active There is an anti-tumour immune response to re-activate The tumour is immunogenic Tumour Micro-environment Other checkpoints Soluble immune inhibitors Inhibitory cell populations Inhibitory tumour metabolism Tumour invisible or impervious to immune killing For NSCLC, PD-L1 IHC Has the Ability to Enrich for Response and Treatment Benefit, however. PD-L1 IHC is not a perfect biomarker Not all positive patients respond Occasional responders in biomarker negative populations But no biomarker is perfect Expectations in the real, clinical world 7 In Advanced NSCLC, PD-L1 Expression by Immunohistochemistry Can Enrich for Response Response rates in the ITT population are 14% to 20% 1-6 Response rates in selected populations are 30% to 45% 2,4,5 Possible Confounders of Tumour Cell PD-L1 IHC Prediction Heterogeneity of expression potential for sampling error Cutoffs in a biological continuum Cohort in the PD-L1-negative group do less well on second-line immune checkpoint inhibitors 3,6 The required power of the biomarker to improve outcome depends upon the activity of the standard of care comparator. This differs between 1 st and 2 nd line therapy. ITT, intent-to-treat. 1. Brahmer J et al. N Engl J Med. 2015;373:123-135. 2. Spira AI et al. ASCO 2015; Abstract 8010.3. Borghaei H et al. N Engl J Med. 2015;373:1627-1639. 4. Herbst RS et al. Lancet.2015;387:1540-1550. 5. Reck M et al. N Engl J Med. 2016;375:1823-1833.6. Rittmeyer A et al. Lancet 2017;389:255-265 6 8 2
Binary Output Versus Biological Continuum Biomarker is ABSENT or at a LOW LEVEL 1% Biomarker is ABSENT You are unlikely to You are unlikely to Biological continuum of biomarker expression Biomarker is PRESENT You are likely to Same principle may apply to Tumour Mutation Burden, THRESHOLD CUTOFF Immune Gene Biomarker expression is PRESENT at an signatures INTERMEDIATE LEVEL and other You IO may biomarkers with a quantitative determination 50% 80% Biomarker is PRESENT at a HIGH level You are likely to 9 11 Binary Output Versus Biological Continuum Biomarker is ABSENT You are unlikely to Biomarker is PRESENT You are likely to Possible Confounders of Tumour Cell PD-L1 IHC Prediction Heterogeneity of expression potential for sampling error Cutoffs in a biological continuum Biomarker is ABSENT or at a LOW LEVEL 1% You are unlikely to THRESHOLD CUTOFF Biomarker is PRESENT at an INTERMEDIATE LEVEL You may 50% 80% Biological continuum of biomarker expression Biomarker is PRESENT at a HIGH level You are likely to Expression on tumour cells (TC) and/or immune cells (IC) Intrinsic induction of PD-L1: oncogenic pathway driven Other immune regulatory mechanisms are also active Technical issues Does the assay correctly identify patients? Did the assay work this time? 10 12 3
Intrinsic induction of PD-L1 expression Upregulation related to oncogenic pathway activation in the tumour JAK-STAT PI3K, MET Probably others due to crosstalk Implies PD-L1 expression that may not be immunologically active Five drug- PD-L1 IHC assay combinations? Trial validated assay for each drug? It is unlikely that labs will provide multiple tests Staining platform availability? One Trail validated assay for all drugs? Lab Developed Test? Laboratory developed tests (LDTs) Modification of the companion/complementary assay LDT built around a trial clone LDT built around another clone Drug Nivolumab 28-8 Pembrolizumab Atezolizumab Durvalumab Assay based on clone. 22C3 SP142 SP263 Avelumab 73-10 We have little idea, so far, how to identify these cases Teng MWL et al. Cancer Res 2013 Outcome of an IHC test is a function of the primary antibody AND the detection system used But life in complex.five most advanced PD1 axis inhibitors and their biomarker assays Drug Company PD-L1 Diagnostic Ab clone Nivolumab Bristol-Myers Squibb Staining Platform Clinically relevant cut offs * Biomarker status 28-8 (Dako) Dako Link 48 TC 1%, 5%, 10% Complementary Pembrolizumab Merck/MSD 22C3 (Dako) Dako Link 48 TC > 1, 50% Companion Atezolizumab Genentech/Roche SP142 (Ventana) Ventana BenchMark ULTRA Durvalumab Astra-Zeneca SP263 (Ventana) Ventana Avelumab Pfizer/ Merck Serono Benchmark TC > 1, 5, 50% IC > 1, 5, 10% TC 25% Complementary Not sure 73-10 (Dako) Dako Link 48 TC 1, 50, 80% Unknown Treatment populations defined by different assays and related cut-points are different 30/38 cases (78.9%) 26/38 cases (60.5%) 26/38 cases (60.5%) 20/38 cases (52.6%) Number of cases (%) with PD-L1 expression above the assay specific selected cut-off Stating the obvious but.. The cut-point stays with the DRUG, and not the assay Hirsch FR et al. JTO 2016 4
Several technical comparability studies have shown broadly similar results.. Blueprint 2: Cases above cut offs by different assays TC staining Across the spectrum of PD-L1 tumour cell expression, three assays show similar performance Dako 28-8, Dako 22C3, Ventana SP263 The Ventana SP142 assay stains fewer tumour cells Scheel A et al Mod Pathol 2016; Hirsch FR et al. JTO 2016; Rimm DL et al. JAMA Oncol 2017; Ratcliffe M et al Clin Can Res 2017; Adam J et al. Ann Oncol 2018 (Feb), Hendry S et al JTO 2018; Scheel AH et al, Histopathol 2018. Small number so only broad conclusions 22C3, 28-8 and SP263 more or less the same SP263 a little more sensitive? SP142 less sensitive 73-10 more sensitive Actual clinical data? TsaoMS et al. WCLC 2017 PD-L1 assays: same, different? SP263 may be a little more sensitive Hendry S et al. JTO March 2018 Tsao MS, Kerr KM, Hirsch FR et al, Blueprint 2 5
Are their clinical data to support the technical comparability of some assays? Alternative PD-L1 IHC biomarkers adequately predict outcome for 2 nd line Nivolumab in NSCLC Fujimoto D et al. JTO March 2018 Different scoring algorithms same class effect Same patients? Test Reliability? 6
Can 25 pathologists agree with each other about PD-L1 scoring? Tsao MS, Kerr KM, Hirsch FR et al, Blueprint 2 WCLC 2017 Performance of LDTs YES! For TC score (TPS%) NO! For IC score Fleiss Kappa: >0.90: excellent 0.75-0.9: good 0.40-0.59: weak 0.20-0.39: minimal <0.01-0.20: slight/none Frequently sub-standard 13/27 (48%) of LDTs failed to make the grade (k > 0.75) 1 Nordic EQA 22C3, 28-8, SP263-based Assay (77-92% pass rate) 22C3 or E1L3N clone LDTs (20% pass rate) UKNEQAS Trial validated assays - 100% pass LDTs - 35% Borderline; 65% Failed Although a good LDT can be developed An E1L3N-based LDT matched Dako assays after intensive validation Rimm DL et al. Lancet Oncol 2017 German QUIP programme Adequate performance of LDT Scheel A et al. Histopathol 2018 Koo TK & Li MY, J Chiropr Med 2016;15:155-63 1. Adam J et al. Ann Oncol 2017 Trial validated assay or go off piste? Therapeutic Aims & Assumptions Inhibit the interaction of PD-1 and PD-L1 In order to take the brakes off an existing tumourspecific immune response We hope this tumour-specific immune response actually exists This requires immunogenicity Which requires antigenicity (neoantigens) to be visible to the immune system A surrogate for probable neoantigen load is tumour mutational burden (TMB) Potential Biomarkers THIS is the drug target! PD-1~PD-L1 Interaction is present Specific immunity is available the tumour is inflamed Whole tumour mutation burden: Whole Exome Sequencing Numbers of mutations in diagnostic panels 28 7
Cancer immune system interaction Immune Gene Signatures and Atezolizumab Tumour inflammation and Immune inhibitory mechanisms Poplar Trial: 2L NSCLC IMPower 150 Trial: 1L NSCLC Arm B: Arm C: Landmark PFS, % atezo + bev + CP bev + CP 6-month 72% 57% 12-month 46% 18% HR, 0.505 (95% CI: 0.377, 0.675) P < 0.0001 Minimum follow-up: 9.5 mo 6.8 mo 11.3 mo (95% CI: 5.9, 7.4) (95% CI: 9.1, 13.0) 8-gene mrna signature: CD8A, GZMA, GZMB, IFNγ, EOMES, CXCL9, CXCL10 and TBX21 Teng MWL et al. Cancer Res 2013 CI, confidence interval; HR, hazard ratio; T eff, T-effector; IFN-, interferon-gamma. Adapted from Fehrenbacher L et al. Lancet. 2016;387:1837-1846. 31 Is the Tumour Inflamed? No actual evidence that the inflammation is tumour specific, but this is inferred and assumed Approaches in NSCLC have, so far, been rather complicated A molecular rather than a morphological approach to inflammation All of these factors are potentially confounded if inflammation is prognostic Morphological or Molecular Inflammation? Immune infiltrate or Immune desert? Where are the immune cells? 30 8
Immune desert Inflammed stroma Inflammed infiltrated Therapeutic Aims & Assumptions Inhibit the interaction of PD-1 and PD-L1 In order to take the brakes off an existing tumourspecific immune response We hope this tumour-specific immune response actually exists This requires immunogenicity Which requires antigenicity (neoantigens) to be visible to the immune system A surrogate for probable neoantigen load is tumour mutational burden (TMB) Potential Biomarkers THIS is the drug target! PD-1~PD-L1 Interaction is present Specific immunity is available the tumour is inflamed Whole tumour mutation burden: Whole Exome Sequencing Immune excluded a probabilistic neo-antigen lottery model Numbers of mutations in diagnostic panels 35 Morphology or Molecular Inflammation? Analysis of the Association Between TMB and PD-L1 Expression a CheckMate 026 TMB Analysis: Nivolumab in First-line NSCLC Immune infiltrate or Immune desert? Where are the immune cells? Which immune cells are present? Which immune cells matter? CD8 CD4 CD1a CD68 CD163 FoxP3 etc CD8 IHC Conde E et al, Histopathol 2018 TMB (no. of missense mutations) 1000 316 243 100 32 10 High TMB Low/medium TMB Immune cell presence or proximity? Multiplex IHC? 50 75 0 25 100 PD-L1 (% tumor expression) a Slide courtesy of John Haanen, NKI Image courtesy of Dr L Sholl, Boston, USA There was no association between TMB and PD-L1 expression in patients with 1% PD-L1 tumor expression a All patients had 1% PD-L1 tumor expression 9
Whole Exome Sequencing Versus Surrogate Panels for TMB Total exome mutations (mutations/mb) 100 50 10 1 Checkmate 026: Total Exome Mutations vs Genes in FoundationOne Panel 1,2 1 10 50 100 FoundationOne panel* (mutations/mb) Carbone DP et al. N Engl J Med. 2017;376(25):2415-2426. FM1 (mutations/mb) 40 30 20 10 0 0 POPLAR: Genes in FoundationOne Panel vs Whole Exome 3 25 50 75 WES (mutations/mb) Kowanetz et al., 2016, WCLC. 3 Surrogates of Surrogates of Surrogates of.of tumour mutational burden In several tumour types Polymerase E (POLE) mutations 1,2 Mismatch repair genes (MMR) 1 Microsatellite instability (MSI) 1,2 KRAS, STK11, TP53 and EGFR mutation in NSCLC 2,3...and by smoking history 1 All of these have been associated with better (enriched) responses to IO *Based on in silico analysis filtering on 315 genes in FoundationOne comprehensive genomic profile (Foundation Medicine, Inc, Cambridge, MA, USA) 1 FM1=FoundationOne; MB=megabase; TMB=tumor mutation burden; WES=whole-exome sequencing. 1. Frampton GM, et al. Nat Biotechnol. 2013;31:10231031. 2. Peters S et al. Oral presentation at AACR 2017. CT082. 3. Kowanetz M et al. Oral presentation at WCLC 2016. OA20.01. 37 1. Chabanon RM et al. Clin Cancer Res. 2016;22:4309-4321. 2. Tanvetyanon T. Transl Can Res. 2017;6:S424-S426. 3. Ross JS et al. Annal Oncol. 2017;28: https://doi.org/10.1093/annonc/mdx376.004. 39 Tumour Mutation Burden (TMB) Combinations of Biomarkers for Immunotherapy? Metabolism Other stuff How to measure it? Whole exome sequencing, Large multigene diagnostic panels What is a HIGH TMB? Numerous definitions, often above median or upper tertile >10 mutations per megabase is becoming established PD-L1 IHC Microbiome Surrogates of immune response Immune cells Gene signature T-effector and IFN- gene signature subgroups 2 There is no consensus on how this should be done All tumours 1 Tumour Mutational burden Neonatal clonal architecture and clinical benefit of immune checkpoint blockade 3 Surrogates of TMB 1. Adapted from Rizvi NA et al. Science. 2015;348:124-128. 2. Adapted from Fehrenbacher L et al. Lancet. 2016;387:1837-1846. 3. Adapted from McGranahan N et al. Science. 2016;351:1463-1469. 10
So far we have very little data on biomarker combinations or direct comparisons of different biomarkers in trials Conclusions TMB plus High PD-L1 IHC better than IHC alone? Checkmate 026 trial Immune gene expression signature (Teff) no better than PD-L1 IHC IMPower 150 trial PD-L1 assessment will remain Inflammation and the Tumour Microenvironment will be important but assessment needs to be clarified Generic or specific inflammation Cell types Cytokines Predictions of foreignness Actual TMB Surrogates of TMB Neoantigens What is coming? Combination therapies IO combinations IO plus chemotherapy IO in stage 3 disease IO as adjuvant or neo-adjuvant therapy Relative benefit versus toxicity (cost) for combination therapy >50% vs 1-49% vs <1% PD-L1 expressors 11