TGx-DDI Qualification of a Preclinical Biomarker C-Path - PSTC - PowerPoint PPT Presentation

TGx-DDI – Qualification of a Preclinical Biomarker C-Path - PSTC – RIKEN Meeting – Yokohama 2019 The HESI TGx-DDI Biomarker Qualification Working Group

HESI: International non-profit building science for a safer, more sustainable world. 150 Universities, Research Institutes, and Scientific Foundations 18 Countries 75 Government Agencies & Institutes 18 Months 70 Corporate Sponsors >1000 Scientists >77 Distinct Projects at HESI events in 2018 15 Scientific Committees 2

CREATE and TEST CONVENE PROVIDE decision- technology and scientific collaborations across makers with sound frameworks that can be academic, government, science for better, more used to protect humans NGO, clinical, and informed decisions. & the environment. industry scientists What HESI does... 3

Jiri Aubrecht, PhD Scientific Director, Takeda Heidrun Ellinger-Ziegelbauer, PhD HESI Senior Scientist, Bayer Pharma Biomarker Carole Yauk, PhD Research Scientist, Health Canada Andrew Williams, MSc Qualification Biostatistician, Health Canada Al Fornace, MD Consortium Professor, Georgetown University Julie Buick, Biostatistician, Health Canada Leadership Henghong Li, MD, PhD Assistant Professor, Georgetown University Syril D Pettit, DrPH ( representing HESI e- Executive Director, HESI STAR committee) Roland Froetschl, PhD Scientific Research Group Leader, BfArM Lauren Peel, BS Germany Scientific Program Manager, HESI

Positive findings of in vitro chromosome damage assays Assessment of relevance to human provides a challenge to sponsors and regulatory agencies Chromosome damage Cancer risk in humans Cancer in animals Point mutations DNA damage Genotoxicity testing Carcinogenicity testing • Evaluating genetox and carci data • Required for NDA • Required for IND • High sensitivity, but low • Mechanistic studies • 2-year bioassay • Genetox battery specificity • Epidemilogical studies • Cost:$3M/cmpd • Cost: $60K/cmpd • ~30% lead chemicals • IARC process • Time: 3 years • Time: 1-3 month positive for in vitro • Cost: ??? chromosome damage • Time: decades assays Genetics and health status  Genetic Susceptibility Non-genotoxic mechanisms  Disease state, stress  Proliferation Environment  Nuclear hormone receptor activation  Food  Epigenetics  Pollutants

TGx-DDI Context of use ICH S2(R1) Option 1 In vitro chromosome damage AMES In vivo micronucleus (positive) Drug Candidate • Chromosomal aberration (negative) (negative) • Micronucleus (CREST negative) TGx-DDI Results (Negative/Positive) WoE Assessment Consider TGx-DDI results and other data/assays relevant for assessment of genotoxic potential. Irrelevant Relevant

Dec 2009 - Letter of intent May 2011 – HESI notifies March 2004 - HESI TGx- to FDA to submit a FDA of plan to submit a biomarker development biomarker qualification genomic biomarker for project initiated plan qualification TGx-DDI July 2011 - Qualification Dec 2016 - B iomarker June 2017 - FDA responds plan briefing meeting at qualification data package with questions on FDA submitted to FDA submission Qualification: August 2017 – FDA moves August 2017 - HESI from prior qualification October 2017 Letter of A Long Path... responsed to FDA program to new program Support from FDA to HESI questions under 21st Century Cures legislation. Today – HESI team June 2018 – HESI Submits Oct 2018 - TGx-DDI developing final reports new Biomarker Qualification meeting at and completing additional Qualification Status Report FDA cross-lab technical to FDA validation.

Study Design Concentration and time point optimization – cytotoxicity (MTT) 4 and 24 h, 6 -10 concentrations Phase 2 – Expression of three stress response genes - ATF , Gadd45a, p21 (qRT-PCR 6 concentrations)  Main study validation set – Established statistical analysis pipeline Phase 1 – 44 compounds, 5 distinct mechanistic classes  Test system validation – Expression profile of all test substances, 4h – Comparability with previous studies (testside- treatment validation) Phase 3 – Cell culture (TK6) and microarray (human whole genome array, Agilent)  Validation studies – Cisplatin, 4 experiments, 4h treatment – Cross-laboratory/ cross-platform  Main study training set - 28 compounds DDI/ non-DDI – Case studies – calculation of biomarker (classifier panel) with  Prediction of substance class training set – optimization with external test set (caffeine, 3-NP and iPMS ) using different – Use of the biomarker (Classifier) on expression bioinformatic tools LoO, NS C, S VM profiles and prediction of DNA-damaging potential

Stres ess s response ge gene expres essi sion used for do dose se finding A B C Fold change Stress gene expression measured by qPCR D E Fold change

Classifier tra raining set et The biomarker was developed using a training set of DNA-damaging and non- DNA-damaging model compounds. Li et al. Env.Mol.Mut. 2015

Nearest shrunken centroids probability analysis Phase 1. Principle Component Statistical Analysis methods Two dimensional hierarchical clustering

Identifying the biomarker: Nearest Shrunken Centroids Probability Analysis  The class centroids are shrunk toward the overall centroids after standardizing by the within-class standard deviation for each gene.  This gives higher weight to genes whose expression is stable within the same class.  In the test cases, the standardized distance to the shrunken centroid is calculated and the class probability is Robert Tibshirani et al. PNAS 2002;99:10:6567-6572 determined.

Biomark rker r panel Entrez ID Entrez ID Gene Response p53 Gene Response p53 Symbol regulated Symbol regulated ❖ ❖ 59 ACTA2  yes 139285 FAM123B V - 64782 AEN  yes 283464 GXYLT1 V - 7832 BTG2  yes - 3008 HIST1H1E  57103 C12orf5  yes - HIST1H2B 3018 B  1026 CDKN1A  yes - HIST1H2B 8347 C  1643 DDB2  yes - HIST1H2B 8339 G  11072 DUSP14  yes - HIST1H2B 8346 I   144455 E2F7 yes - HIST1H2B 8342 M  9538 EI24 V yes - HIST1H2B 8341 N  Transcripts comprising 26263 FBXO22  yes - 8351 HIST1H3D  1647  yes - GADD45 A 3398 ID2 V the TGx-DDI biomarker 121457 IKBIP  yes - 80271 ITPKC  4193 MDM2  yes - 3708 ITPR1 V  23612 PHLDA3 yes - 353135 LCE1E  8493 PPM1D  yes - 9209 LRRFIP2 V  51065 RPS27L yes - 84206 MEX3B   50484 RRM2B yes - 79671 NLRX1 V  9540 TP53I3 yes - 5100 PCDH8   51499 TRIAP1 yes - 1263 PLK3   10346 TRIM22 yes - 5564 PRKAB1   91947 ARRDC4 - - 5565 PRKAB2   10678 B3GNT2 - - 5734 PTGER4 V 282991 BLOC1S2  - - 9693 RAPGEF2  84312 BRMS1L  - - 389677 RBM12B V 868 CBLB V - - 6400 SEL1L V 9738 CCP110  - - 6407 SEMG2  1052 CEBPD V - - 29950 SERTAD1  1062 CENPE  - - 4090 SMAD5  8161 COIL V - - 51768 TM7SF3  23002 DAAM1 V - - 608 TNFRSF17  196513 DCP1B  - - 10210 TOPORS V 79733 E2F8  - - 373856 USP41 

App pplyi ying the biomark rker Principle Component Two- Dimensional Probability Analysis Hierarchical Clustering Analysis

Cas Case e study: Accu ccurate predict ction of DDI DDI capacity of 3-Np Np, caffe feine and IPMS MS using TG TGx-DDI From Li et al. Env.Mol.Mut. 2015 15

Application in the pr pres esen ence of S9 9 metab abolic act ctivation sys yste tem: Accu ccurate predict ction of B(a (a)P )P, A AFB1 1 and Dexam amethas asone - 16 chemicals tested in presence of S9 metabolic activation systems and confirmed to yield accurate predictions From Buick et al. Env Mol Mut 2015 Yauk et al. Env Mol Mut 2016 16

Summary Ph Phase 1 1 TG TGx-DDI Biomarker to Predict ct D DNA D Damage-Inducing ( (DDI) C Chemical als The in vitro transcriptomic biomarker predicts the probability that an agent is DDI or non-DDI. DDI Non-DDI  Developed using human cells in culture (TK6 cells)  From exposure to 28 prototype DNA damage-inducing (DDI) and non-DDI chemicals  64 genes identified as being predictive of DDI potential TGx-DDI Publications for Methods Development, Validation, Application: Genes TGx-28.65 biomarker development and validation • Li, HH et al. Environ Mol Mutagen (2015) • Li, HH et al. PNAS (2017) Development of method for use of biomarker with metabolic activation system • Buick, JK et al. Environ Mol Mutagen (2015) • Yauk CL et al. Environ Mol Mutagen (2016) TGx-DDI Software development Agents • Jackson, MA et al. Environ Mol Mutagen (2017) Case study • Buick, JK et al. Mutat Res (2017) 17

44 Compounds Phase 2. Main 5 Mechanistic Classes study validation DNA microarrays

Summa mmary TG TGx-DDI b biomarker a accu ccurately identifies D DDI and n non-DDI a agents Class 1 – Direct DDI agents Class 2 – Indirect DDI agents Class 4 – Non-DDI agents Class 5 – IRRELEVANT in vitro positives + Metabolic activation TGx-DDI effectively identifies Class 5 agents Li et al., Development and validation of a high-throughput transcriptomic biomarker to address 21st century genetic toxicology needs. PNAS, 2017

Cross-laboratory Phase 3. Cross-platform Validation Studies Case studies

Cross-platform comparison of performance of TGx-DDI Li H-H et al. PNAS 2017; 114(51):E10881-E10889