25 March 2017 New Targets Committee Quantitation with Whole Section Analysis – Xenograft Models in Oncology Drug Develpoment JSTP Meeting February 2010 David Young DVM DACVP DABT Flagship Biosciences LLC New Targets Committee

25 March 2017 New Targets Committee Presentation Outline Introduction to digital pathology and quantitative image analysis Biomarker development Basics of IHC analysis Image analysis – Concepts and tools Target tissue identification Case study – Use of image analysis in Oncology drug development IHC biomarker analysis – from xenograft to tumors Lessons from quantitative analysis of tumors New Targets Committee

Quantitative Analysis - The Big Advantage 25 March 2017 Quantitative Analysis - The Big Advantage Image analysis of digitized images provides practical, accurate and reproducible quantifiable measurements of cellular change, replacing subjective with objective evaluation New Targets Committee

Why Quantitative Image Analysis? 25 March 2017 Why Quantitative Image Analysis? Generally toxpath evaluations are sufficiently accurate and efficient that they need not be replaced by image analysis Minimal Mild Moderate Severe In some special cases, observed changes may be of such importance that objective image analysis with statistical significance is needed to demonstrate their validity New Targets Committee

Biomakers in Discovery Pathology 25 March 2017 Biomakers in Discovery Pathology Applications of Biomarker Assays Development work and pre-clinical models Use in clinical trials (patient selection, stratification) Retrospective analysis of clinical samples New Targets Committee

25 March 2017 New Targets Committee Biomarker Basics Tumor Based Proteins Immunohistochemistry (IHC) fluorescent in situ hybridization (FISH) Phospho- proteins Mutations Variants Blood/Serum Based DNA Germline Tumor shed (CTCs) Proteomics Single or multiple proteins New Targets Committee

25 March 2017 New Targets Committee IHC Scoring Basics +3 +2 +1 +3 +2 IHC scoring is based on a subjective interpretation of stain intensity New Targets Committee

IHC Staining Intensity Criteria 25 March 2017 IHC Staining Intensity Criteria +1 +3 +2 New Targets Committee

IHC Intensity Staining Criteria Shift 25 March 2017 IHC Intensity Staining Criteria Shift +1 +2 +3 +1 +2 +3 New Targets Committee

25 March 2017 25 March 2017 New Targets Committee IHC Scoring (H-Score) Proportion Score (PS) 1% 10% 30% 75% 100% Intensity Score (IS) 0 = negative 1 = weak 2 = intermed 3 = strong The pathologist scores staining features of cells (eg. cytoplasmic, nuclear, or membranous staining) by intensity of stain and percentage of stained cells Immunostaining was scored using a previously established scoring system developed by Dr. Craig Allred in our group. The proportion of stained tumor cells was estimated on a scale of 0 to 5: 0 represents no staining; 1 is less than 1/100 cells; 2 is 1/100 to 1/10 cells; 3 is 1/10 to 1/3; 4 is 1/3 to 2/3; and 5 indicates staining in more than 2/3 of the tumor cells. Staining intensity was rated as negative, weak, intermediate, or strong. A total IHC score was calculated by adding the proportion score and the intensity score. The range of values is 0, and 2 to 8. New Targets Committee New Targets Committee Page 10

25 March 2017 New Targets Committee Example of H-scoring H score = (1)x(PS1) + (2)x(PS2) + (3)x(PS3) Example: (1)x(20%) + (2)x(30%) + (3)x(50%) = 230 New Targets Committee

Subjective IHC Scoring – The ‘H Score’ 25 March 2017 Subjective IHC Scoring – The ‘H Score’ The H score puts a quantitative number on a subjective evaluation (semi-quantitative scoring) Does not distinguish between a high percentage of low to medium stained cells and a small percentage of strongly stained cells. Requires that the pathologist define low medium and high intensity levels. Is very dependent on the pathologist experience and subjectivity. New Targets Committee

Scoring by Quantitative Analysis 25 March 2017 Scoring by Quantitative Analysis Using quantitative image analysis - “H” Score evaluation is automatically calculated Aperio’s IHC Deconvolution Algorithm provides attribute outputs in the following similar formula: (Nwp/Ntotal)x(100) + (Np/Ntotal)x(200) + (Nsp/Ntotal)x(300) = “H” Score   Where: Nwp = Number of weakly positive pixels Np = Number of moderately positive pixels Nsp = Number of strongly positive pixels Ntotal = Total number negative + positive pixels 13 New Targets Committee

The importance of Object Recognition in the Future of Image Analysis 25 March 2017 The importance of Object Recognition in the Future of Image Analysis Use the lowest magnification necessary to visualize object New Targets Committee

Object Recognition Defines Analysis 25 March 2017 Object Recognition Defines Analysis New Targets Committee

Target Tissue Analysis 25 March 2017 In it’s Simplest Terms….. Count and measure simple structures/objects. Measure area of defined regions/stain. Measure intensities of stain as a percentage of defined regions. Combinations of 1, 2 and 3 above. New Targets Committee

Methods for Defining the Target Tissue for Analysis 25 March 2017 Methods for Defining the Target Tissue for Analysis Define the target tissues for analysis using common (eg H&E) or special (eg IHC) staining procedures and manual differentiation. Define the target tissues for analysis using histology pattern recognition tools Assist in defining target tissues in 1 and 2 above by using the positive and negative pen tools. A high degree of accuracy in target tissue definition will assure a high degree of accuracy in the final analysis. New Targets Committee

Some Guidelines for Analysis of Slides from Experimental Studies 3/25/2017 25 March 2017 Some Guidelines for Analysis of Slides from Experimental Studies Assure immediate optimal fixation for all tissue samples. Uniformity of handling as well as fixation time is important. Staining procedures for all slides in a study need to be performed simultaneously in a single batch to assure uniformity of stain. Sampling must be strictly representational as well as consistent. Care must be taken to assure exact uniformity of analysis with respect to anatomical location (eg. Tissue trimming, sectioning) Use a ‘practice’ subset of slides - A preliminary evaluation of image analysis tools between some slides of varying stain intensities will help assure that analysis values are established optimally for all slides in the study 18 New Targets Committee 18

Digital Pathologist’s Toolbox 25 March 2017 Digital Pathologist’s Toolbox Analysis Tools Positive Pixel Count Color Deconvolution IHC Nuclear IHC Membrane Co-localization Microvessel Analysis Preprocessing Utility Genie™: Histology Pattern Recognition New Targets Committee

25 March 2017 New Targets Committee Analytical Tools Area Based Analysis Pixel Count IHC Deconvolution Co-localization Cell Based Analysis IHC Nuclear IHC Membrane Angiogenesis Rare Event Analysis Rare Event Detection New Targets Committee

Genie™ - Histology Pattern Recognition 25 March 2017 Genie™ - Histology Pattern Recognition Histology pattern recognition software as a preprocessing machine - segregates target from nontarget tissue during analysis Los Alamos National Laboratory’s Genetic Imagery Exploration Analytical Result Analytical Result Analysis Tool Analysis Tool GENIE Preprocessing Primary Image Primary Image 21 New Targets Committee

Example of Preprocessing with Genie™ and Image Analysis 25 March 2017 Primary IHC image Genie™markup with selection of neoplasm 1 2 Final Aperio ImageScope deconvolution markup New Targets Committee

Example of Oncology Development and Use of Image Analysis 25 March 2017 Example of Oncology Development and Use of Image Analysis New Targets Committee

From primary tumor to distant metastasis 25 March 2017 Cancer Progression Hypothesis From primary tumor to distant metastasis New Targets Committee

Epithelial-Mesenchymal Transition (EMT) 25 March 2017 Epithelial-Mesenchymal Transition (EMT) Most solid tumors start with an epithelial phenotype External and internal signaling events trigger transition to mesenchymal phenotype Mesenchymal tumor cells invade neighboring tissue and into the vasculature to metastasize A B C A B epithelial EMT Invasion and metastasis of epithelial cancers utilize transition to a mesenchymal state (EMT) C mesenchymal Blood Vessel Adapted from Brabletz et al. (2005),Christofori (2006), Lee et al. (2006, Thiery & Sleeman (2006) Endothelial Cells Human cancers demonstrate a multiplicity of signaling repertoires by which cell survival and migration programs are achieved. Recent data highlight the conversion of epithelial cancer cells to a more mesenchymal-like state, a process termed epithelial-mesenchymal transition (EMT), to facilitate cell invasion and metastasis. Mesenchymal-like tumor cells gain migratory capacity at the expense of proliferative potential; the reverse conversion, a mesenchymal-epithelial transition (MET), is thought to be required to regenerate a proliferative state and form macrometastases resembling the primary tumor at distant sites. Brabletz T et al. Migrating cancer stem cells – an integrated concept of malignant tumor progression. Nature Rev 2005;5:744-749 Christfori C. New signals from an invasive front. Nature 2006; 41:444-450 Lee JM et al. The epithelial-mesenchymal transition: new insights in signaling, development, and disease. J Cell Biology 2006;7:973-981 Thiery JP, Sleeman JP. Complex networks orchestrate epithelial-mesenchymal transitions. Nat Rev/Mol Cell Biol 2006; 7:131-142 New Targets Committee

EMT - Potential Biomarkers and Targets 25 March 2017 EMT - Potential Biomarkers and Targets External Signals Transcriptional Reprogramming Slug Zeb Molecular Response Biological Consequence Kang, 2004Cell v118 p277-279 New Targets Committee

Cell Line Sensitivity to TKIs 25 March 2017 25 March 2017 Cell Line Sensitivity to TKIs Epithelial markers are maintained in Sensitive tumors Mesenchymal markers are maintained in Refractory tumors EMT markers appear to be a good predictor of erlotinib sensitivity in vivo Refractory Sensitive H460 Calu6 A549 H441 H292 E-cadherin Epithelial g-catenin Fibronectin Mesenchymal Vimentin GAPDH Adapted from Thomson et al., Cancer Res., 2005 New Targets Committee New Targets Committee Page 27

Clinical Correlation of TKIs 25 March 2017 25 March 2017 Clinical Correlation of TKIs In Advanced NSCLC in Patients with E-cadherin Positive Tumors 0.0 0.2 0.4 0.6 0.8 1.0 | | | | | Chemo Alone, E-cadherin pos (N=37) Erlotinib + Chemo, E-cadherin pos (N=28) | | | | | | Chemo Alone, All Patients (N=540) | | | Erlotinib + Chemo, All Patients (N=539) Progression-Free Rate | | | HR=0.37 p=0.0028 | | | | | | | | | | | | | Adapted from Yauch, Clin Cancer Res (2005) | | | | | | Weeks 20 40 60 80 E-cadherin Positive Patients had a Longer Time to Progression Comparing Combined EGFR-TKI (Erlotinib) with Chemotherapy to Chemotherapy Alone A retrospective analysis of a Phase III trial in advanced NSCLC for erlotinib plus chemotherapy compared with chemotherapy alone, showed that E-cadherin expression was a significant predictive marker for efficacy of EGFR inhibition by erlotinib, as measured by progression-free survival (Yauch et al. 2005). In this study (TRIBUTE), erlotinib did not add to the efficacy of chemotherapy. The loss of E-cadherin during tumor progression to a more mesenchymal state has both treatment and prognostic significance. Yauch RL et al. Epithelial versus mesenchymal phenotype determines in vitro sensitivity and predicts clinical activity of erlotinib in lung cancer patients. Clin Cancer Res 2005; 11(24):8686-8698 New Targets Committee New Targets Committee Page 28

IHC Assessment of EMT Biomarker E-cadherin 25 March 2017 25 March 2017 IHC Assessment of EMT Biomarker E-cadherin New Targets Committee New Targets Committee Page 29

Heterogeneity in Tumor Tissue – E-cad 25 March 2017 Heterogeneity in Tumor Tissue – E-cad New Targets Committee

Cell Culture - E-cadherin 25 March 2017 Cell Culture - E-cadherin New Targets Committee

Aperio Membrane Algorithm Changes 25 March 2017 Aperio Membrane Algorithm Changes Aperio Membrane v9  Modified membrane algorithm Threshold Type 0 - Edge Threshold Method Lower Blue Thresholding Upper Blue Thresholding 220 Min Nuclear Size (um^2) 10. 30. Min Nuclear Size (Pixels) 40 119 Max Nuclear Size (um^2) 2000 Max Nuclear Size (Pixels) 7914 Min Nuclear Roundness 0.1 0.7 Min Nuclear Compactness 0. Min Nuclear Elongation 0.5 Cytoplasmic Correction Yes Cell/Nucleus Requirement 0 - All Cells Min Cell Radius (um^2) 5. Min Cell Size (um^2) Max Cell Size (um^2) Min Cell Roundness Min Cell Compactness Min Cell Elongation Background Intensity Threshold 250 Weak(1+) Intensity Threshold 210 225 Moderate(2+) Intensity Threshold 140 170 Strong(3+) Intensity Threshold 85 95 Completeness Threshold 50 New Targets Committee

25 March 2017 NSCLC Criteria setup New Targets Committee

EMT Xenograft - E-cadherin 25 March 2017 Entire Specimen IHC Test box (3+) Percent Cells 71.83 50 65.67 (2+) Percent Cells 9.61 40 8.17 (1+) Percent Cells 18.53 10 26.16 (0+) Percent Cells 0.03 0.00 SCORE 253.24 240 239.51 New Targets Committee

25 March 2017 New Targets Committee NSCLC (E-cadherin) E-Cad Aperio IHC (3+) Percent Cells 68.60 50 (2+) Percent Cells 6.25 25 (1+) Percent Cells 24.54 20 (0+) Percent Cells 0.60 5 SCORE 242.84 220 New Targets Committee

Xenograft Model – Skin Tumors With GENIE Preprocessing 25 March 2017 Xenograft Model – Skin Tumors With GENIE Preprocessing New Targets Committee

Xenograft model – Selection of Genie Classifiers 25 March 2017 Xenograft model – Selection of Genie Classifiers New Targets Committee

Xenograft Model - Montage 1 25 March 2017 Xenograft Model - Montage 1 New Targets Committee

Xenograft Model – Genie Selection and Membrane Analysis 25 March 2017 Xenograft Model – Genie Selection and Membrane Analysis New Targets Committee

Xenograft Model – Analysis 25 March 2017 Xenograft Model – Analysis New Targets Committee

Can We Use the Whole Section? 25 March 2017 Can We Use the Whole Section? New Targets Committee

Montage 2 – Using Skin Classifier 25 March 2017 Montage 2 – Using Skin Classifier New Targets Committee

Xenograft Model – Whole Image Analysis 25 March 2017 Xenograft Model – Whole Image Analysis New Targets Committee

Xenograft E-cad Selections 25 March 2017 Xenograft E-cad Selections New Targets Committee

Results of Xenograft IHC Analysis 25 March 2017 Results of Xenograft IHC Analysis Manual subjective analysis vs GENIE assisted image analysis New Targets Committee

Tumor Specimens – Validation Set 25 March 2017 Tumor Specimens – Validation Set New Targets Committee

NSCLC - GENIE Classifiers 25 March 2017 NSCLC - GENIE Classifiers Tumor epithelium - Green Tumor stroma - Yellow Normal lung - Red New Targets Committee

25 March 2017 NSCLC - 37279 New Targets Committee

25 March 2017 New Targets Committee NSCLC - 37279 37279 Manual GENIE (3+) Percent Cells 55 50 (2+) Percent Cells 33 29 (1+) Percent Cells 12 21 (0+) Percent Cells 1 H-score 243 229 %+2 and +3 88 79 New Targets Committee

25 March 2017 NSCLC - 37409 New Targets Committee

25 March 2017 New Targets Committee NSCLC - 37409 37409 Manual GENIE (3+) Percent Cells 60 77 (2+) Percent Cells 20 5 (1+) Percent Cells 10 18 (0+) Percent Cells H-score 230 260 %+2 and +3 80 82 New Targets Committee

25 March 2017 NSCLC - 37321 New Targets Committee

25 March 2017 New Targets Committee NSCLC - 37321 37321 Manual GENIE (3+) Percent Cells 76 (2+) Percent Cells (1+) Percent Cells 24 (0+) Percent Cells 100 H-score 253 %+2 and +3 Cells (Total)   17 Complete Cells 13 New Targets Committee

Lessons Learned - Image Analysis – From Discovery to Clinical Trials 25 March 2017 Lessons Learned - Image Analysis – From Discovery to Clinical Trials Pre-analytical handling remains an unknown factor Pathologist must designate areas of interest GENIE needs to be best ‘refined’ to properly ID tissue Standarized IHC staining protocol CRITICAL Locking of algorithm for same staining protocol Consistent ‘scoring’ by image analysis Pathology review of slides is still required New Targets Committee

25 March 2017 www.flagshipbio.com Thank you New Targets Committee