High-Content Analysis

Advances in High Content Analysis:
Phenotypic Drug Discovery
Bev Isherwood
High Content Biology, Discovery Sciences, AstraZeneca
ELRIG Drug Discovery 2012, Manchester
Outline
•
High Content Analysis
•
Evolution of HCA
•
Technology and Platforms
•
Applications
• Target identification and validation
• Phenotypic Screening & Toxicity Prediction
• Physiologically relevant models for MoA studies
• Tissue-based & automated immunohistopathology
•
Observations & Trends
2 | B Isherwood | 06 September 2012
Innovative Medicines | Discovery Sciences
High-Content Analysis
“An automated cell biology method drawing on optics, chemistry, biology
and image analysis to permit rapid, highly parallel biological research”
Access to complex phenotypes
intractable by non-HCS methods
Represents
targets in
their native
environment
Cell differentiation
3 | B Isherwood | 06 September 2012
Cellular movement
e.g. speed of migration
Tissue organisation
Subpopulations
Heterogeneity
Spatial events
e.g. translocation
Morphological alterations
e.g. cytoskeletal
Innovative Medicines | Discovery Sciences
Evolution of HCA
High-content approaches underpin significant assay development and technology
delivered to projects at AstraZeneca. Can be used to embrace complexity of biology
Cytotoxicity,
Protein
Expression
1M compound
Cell-based HTS
2002
2004
4 | B Isherwood | 06 September 2012
2006
Primary Cells,
Timecourses, 3D
2008
Supercellular Structures,
Phenotypic Profiling, Stem Cells,
Drug Combinations, Ex Vivo
Tissue Imaging
2010
2012
Innovative Medicines | Discovery Sciences
High Content Biology Facility
• Latest HCA platform technology is enabling the application of assays with
increased physiological relevance earlier in drug discovery
Functional ex vivo &
multidimensional imaging
Plate-based Imaging - wide field & confocal
live cell chambers & liquid handlling
Long-term Culturing
& Plate Processing
Imaging Cytometers
Kinetic Imaging & Primary
Tissue Culture
Image Analytics, Data
Processing, Data Management
Molecular Profiling
5 | B Isherwood | 06 September 2012
Innovative Medicines | Discovery Sciences
Application of HCA to Drug Discovery
Frontloading
Toxicity Prediction
MoA & prioritization & reprofiling
of compounds for clinical testing
in Physiological Models of
Disease
Discovery
Target
selection
Lead
Generation
Lead
Optimisation
Validation of Targets &
Pathways
in Physiological
Models of Disease
6 | B Isherwood | 06 September 2012
Pre clinical
POM
Development
GMD
POP
Phenotypic Screening
Paradigms
POC
Phase III /
Launch
Product
maint.
Mechanistic understanding of
compound action using
Automated Histopathology
Innovative Medicines | Discovery Sciences
Physiologically Relevant Models
• Human primary tissue key to implementation of HCA in drug discovery
Primary Human Myoblasts/ Satellite Cells
Quadriceps muscle (vastus lateralis) dissected post mortem
differentiation
protocol
Yield > 5 x 105 cells/g
Myf5+ Satellite cells/ myoblasts
Muscle myotubes stained for
myosin-heavy chain
Human Adipose Derived Stem Cells (ADSCs)
Subcutaneous adipose tissue
Isolate
subcutaneous
adipose tissue
Aspirate and
wash fat
Digest with
collagenase
Isolate
stromal-vascular
fraction
ADSCs
Adipogenic differentiation
7 | B Isherwood | 06 September 2012
Innovative Medicines | Discovery Sciences
Example: Organotypic model of Angiogenesis
•
•
Formation of new vessels from pre-existing vasculature
Multistage process involves endothelial cell proliferation,
migration & differentiation; challenging to model in vitro
•
Role in many pathological processes including tumour
formation, macular degeneration, psoriasis
• Assessment of vessel density & morphology in a human primary
• Mean Tube Area
HDF:HUVEC co-culture organotypic model
Hoechst
CD31
Image Analysis
Tubule
Node
Secreted Growth
Factor Analysis
Isherwood B. (2011) International Drug Discovery
8 | B Isherwood | 06 September 2012
• Tube Area Covered
• Average Tube Thickness
• Nodes (& Area Covered)
• Tube Length Per Set
• Mean Tube Length
• Branchpoints,
Segments
• Mean & Total Node Area
• Connected Sets
• Total Tube Length &
Area
• Nuclei Count Sum
• bFGF
• PlGF
• sFlt-1
• VEGF
Innovative Medicines | Discovery Sciences
Higher throughput screening
• Laser Scanning Cytometer: Acumen Explorer (TTP LabTech Ltd)
• Identify modulators without detailed MoA characterization
- low resolution
- whole-well scanning
- fast (~9 min/plate)
- analysis on-the-fly
Dose-response:
Mycophenolic
Acid
% Inhibition Total Tube Area
Scatter Plot
100
80
60
40
20
0
5E-6
1E-5
5E-5
0.0001
0.0005
0.001
Dose
IC50= 396 nM
9 | B Isherwood | 06 September 2012
Innovative Medicines | Discovery Sciences
0.005
0.01
0.02
Multiparametric Profiling
• High-content imaging instruments
• Detailed profiling of MoA of vascular modulating compounds upon
tubule structure & other cellular processes (i.e. cell cycle):
10 | B Isherwood | 06 September 2012
Innovative Medicines | Discovery Sciences
Automation of co-culture models
• Hepa-filtered enclosed Agilent Biocel system for automated screening:
Cell plating
Assay ready plate transfer
Reagent addition
Analyte capture
Manual
pIC50 6.23
TGFBR Inhibitor (A83-01)
3 nM
Automated
pIC50 6.35
3 nM
Univariate analysis of Z’ & pIC50
11 | B Isherwood | 06 September 2012
Innovative Medicines | Discovery Sciences
Custom developed automated Image Analysis
• Enables contextual & relational analysis of image objects
• Objects can be custom defined to include biological knowledge
12 | B Isherwood | 06 September 2012
Innovative Medicines | Discovery Sciences
Custom developed automated Image Analysis
• Enables contextual & relational analysis of image objects
• Objects can be custom defined to include biological knowledge
1 – Pre-processing
13 | B Isherwood | 06 September 2012
Innovative Medicines | Discovery Sciences
Custom developed automated Image Analysis
• Enables contextual & relational analysis of image objects
• Objects can be custom defined to include biological knowledge
2 – Finding Vessels
Tubule Area
Parameters
Quantified
from Vessel
Mask
14 | B Isherwood | 06 September 2012
Innovative Medicines | Discovery Sciences
Custom developed automated Image Analysis
• Enables contextual & relational analysis of image objects
• Objects can be custom defined to include biological knowledge
3 – Skeletonization
Tubule Length &
Branchpoint
Parameters
quantified from
Skeletonization
15 | B Isherwood | 06 September 2012
Innovative Medicines | Discovery Sciences
Custom developed automated Image Analysis
• Enables contextual & relational analysis of image objects
• Objects can be custom defined to include biological knowledge
4 – Nuclei Segmentation
Nuclei Count &
Morphology
Parameters
quantified from
segmentation
16 | B Isherwood | 06 September 2012
Innovative Medicines | Discovery Sciences
Custom developed automated Image Analysis
• Enables contextual & relational analysis of image objects
• Objects can be custom defined to include biological knowledge
4 – Classification
Nuclei
classified
by Cell
type
17 | B Isherwood | 06 September 2012
Innovative Medicines | Discovery Sciences
Highly flexible workflow- automated data pipeline
• 90 compound screen monitoring 2 λ & 16 sites/well = 86,400 images
• Multiple parameters (~15) = 1,296,000 data points…
• Combines & manipulates data from different sources using pipeline tools:
Datastore
Input
Compound Information
Cell Plate Information
Analysis Results
Dataset
18 | B Isherwood | 06 September 2012
Tools
- Normalization
- QC
- Stats; PCA
Output
- Spotfire
- AZ Screening Dbase
- Pipeline Pilot Viewer
Innovative Medicines | Discovery Sciences
Highly flexible workflow- interactive QC
Interactive image QC
S c a tte r P lo t
10-DEBC... 1400W di... 2-Methox... A 205804
% Inhibition of total tube
area (normalized to controls
)
100
50
0
-50
-100
-150
-200
-250
-300
Anti--a2b... CL 82198... Caveloin -
A 83-01
AG 213
AG 825
ALLN
API-2
Angiostatin
DAPT
DMPQ di...
HA 1100 ...
IFN
Caveolin + Colchicine Combreta... Concana... Cytochala...
100
50
0
-50
-100
-150
-200
-250
-300
100
50
0
-50
-100
-150
-200
-250
-300
Dose-responses for Total Tube
area of 90 molecule VM screen:
DR 2313
Doxorubi...
DuP 697
Etoposide FGF-basi... Fumagillin
GW 5074
Genistein
JTE 013
KF 38789
Ki 8751
L-NIL hydr... L-NIO hyd... LY 364947 Lactacystin Leupeptin... Levamisol
Linomide
100
50
0
-50
-100
-150
-200
-250
-300
100
50
0
-50
-100
-150
-200
-250
-300
MG 132
PD 98059
Marimastat Methyl 2,5... Monastrol
PDGF
PI 103 hy...
PI 828
Mycophe... NSC 237...
PP 1
PP 2
NU 1025
PP 3
Nimesulide
ODQ
PD 1667...
Peptide F9 Perhexile... Rapamycin
100
50
0
-50
-100
-150
-200
-250
-300
S1P
SB 2021... SB 2035... SB 2167... SB 2390... SB 4152... SB 4315... SCH 280... SEW 2871 SP 6001...
100
50
0
-50
-100
-150
-200
-250
-300
SU 4312
SU 5416
Suramin h...
Taxol
Thalidomi...
Thiolutin
Tranilast
U0124
U0126
VEGF
100
50
0
-50
-100
-150
-200
-250
-300
Vinblastin... Vincristin... Wortman... Y-27632 ... ZK 164015 a-hVEGF... anti-avb3 ... cyclo (43... cyclo (PL... normal go...
100
50
0
-50
-100
-150
-200
-250
-300
1... 1...
1...
0...
0...
0...1... 1...
1...
0...
0...
0...1... 1...
1...
0...
0...
0...1... 1...
1...
0...
0...
0...1... 1...
1...
0...
0...
0...1... 1...
1...
0...
0...
0...1... 1...
1...
0...
0...
0...1... 1...
1...
0...
0...
0...1... 1...
1...
0...
0...
0...1... 1...
1...
0...
0...
0...
Dose
Log [Concentration]
(μM)
(Dose-responses for 13 other
parameters available)
19 | B Isherwood | 06 September 2011
0
100 %
Colour by % Inhibition
of Nuclei Count
Innovative Medicines | Discovery Sciences
Highly flexible worfklow- integrated statistics
Data Normalization
3. Activation/
Differentiation
2. Migration
p38MAPK
Outlier Removal (Controls)
Raf/ERK Signalling
Dimension Reduction
Active Compounds
Identified
Clustering
Src
TGFBR Signalling
Rho Kinase
VEGF Signalling
1. Survival
TKI
mTOR/PI3K/Akt
Microtubule Effects
•
•
20 | B Isherwood | 06 September 2012
85% data explained PC1-3 (96% PC1-6)
68 compounds identified as active at, at least a
single dose (by Mahalanobis Distances,
thresholded at 0.1% significance level)
Innovative Medicines | Discovery Sciences
Phenotypic Assays for Drug Discovery
• HC phenotypic analysis of cell-based assays to complement traditional
target-centric Lead Identification programs
Cell Cycle
Morphology
Apoptosis
Phenotypic fingerprint (morphology assay)
MCF7 cells & Aphidicolin
dose
Caie P. et. al. (2010) Mol. Cancer.Ther. 9(6) p1913-26
21 | B Isherwood | 06 September 2012
Innovative Medicines | Discovery Sciences
Temporal Effects in Cell Based Assays
• Increased dimensionality of kinetic profiling enables analysis of temporal effects and
optimization of endpoint assays
HUVECs fluorescently labelled with lentiviral GFP vector
Isherwood B. et. al. (2011) Pharmaceutics. 3 p141-170
22 | B Isherwood | 06 September 2012
Innovative Medicines | Discovery Sciences
Combinations Profiling
• Increased dimensionality of kinetic profiling enables analysis of temporal
effects. In addition to measurement of phenotypic responses, additional
data can derive further understanding of compound mode of action:
Kinetic Imaging
Phase Contrast
• DEVD-NucViewTM 488 caspase substrate to monitor onset of apoptosis
in live cell assay (no wash steps)
Caspase Positive [Feature_1]
Staurosporine (1 μM) & DMSO Control (0.1%)
23 | B Isherwood | 06 September 2012
Fluorescence
Mask
Innovative Medicines | Discovery Sciences
Kinetic Profiling: Apoptosis
Compound A
Compound B
Caspase Positive [Feature_1]
Example Combination Output:
T54
Combination
Compound 1
Compound 2
Summary statistics used to obtain measure of synergistic effects- allows
comparison across cells from different patient backgrounds
24 | B Isherwood | 06 September 2012
CI=13.5
Innovative Medicines | Discovery Sciences
Imaging cells in 3D
• 3D Tumour Cell Transwell Invasion Assay (2-photon confocal)
Top
20μm
HT1080 cells
Media (0.2% FCS)
40μm
60μm
80μm
Fibrillar collagen/
Matrigel
100μm
% No. cells
invading
beyond 60m
70
60
120μm
50
40
Media (10% FCS)
Hoechst:
Cell number
Calcein/Sytox:
Cell Viability
Tagged Proteins: Functional Biosensor
140μm
30
20
160μm
10
0
180μm
Control
200μm
25 | B Isherwood | 06 September 2012
Innovative Medicines | Discovery Sciences
• Automated Z-stack capture & Image Analysis
HT1080 cell invasion through collagen
depth
Hoechst nuclear stain
• Invasion
quantified using
DefiniensXD
platform
Number of kits
available
14
Combinations:
12
% No. cells 10
invading
8
beyond 60m
6
4
2
0
Carragher NO. (2009) Clin. Exp. Metastasis 26(4):381-97
26 | B Isherwood | 06 September 2012
control
AZX
(0.01M)
AZY
(1M)
AZX (0.01M)
+
AZY (1M)
Innovative Medicines | Discovery Sciences
Imaging cells in 3D
• Air-Liquid Interface culture model of airway epithelium
• Monitor effects of smoke on epithelial breakdown and progenitor division
post fixation & ICC:
control
+ smoke
Cross-section of culture labelled for proliferation and
differentiation
• Classify cell phenotypes & spatial
relationships using definiens
• Smoke drives epithelial cell proliferation,
causes mixing of immature and mature
cells, decreases tight junction proteins
27 | B Isherwood | 06 September 2012
Innovative Medicines | Discovery Sciences
Tissue HCA
• Beta-cell proliferation a key endpoint in diabetes research
• Gold-standard process utilses isolated rodent islets. Requires process of
isolation and dispersion. Numerous washes and poor adhesive properties
of dispersed islets reduces cell number
Adhesive slide-based assay
Isolated pancreatic islets
Dedicated Slide Scanner
Reduced timelines
(months > weeks)
and islet (hence
animal) number
required
28 | B Isherwood | 06 September 2012
DAPI & EdU
Innovative Medicines | Discovery Sciences
Automated Immunohistopathology
• Pancreatic beta-cell mass in type 2 diabetes
Low resolution scanning can report beta cell mass
Traditional methods
involve cell counting
by hand on DAB
stained sections.
29 | B Isherwood | 06 September 2012
Innovative Medicines | Discovery Sciences
Automated Immunohistopathology
• Dedicated slide imaging coupled with advances in automated image
analysis. Example of pancreatic and beta-cell analysis:
10x image of
nuclei, insulin &
glucagon
Classification
view of
exocrine, alpha
& beta cells
• Time saving: From 1.5 h to 25 min per animal
• Improved Data Quality: From 10 images per animal to whole sample (>100)
• Detailed cellular resolution (e.g. stain intensity, morphology-islet area and
shape) as well as tissue area & cell count
Gorman T (2012) ELRIG Drug Discovery, Poster
30 | B Isherwood | 06 September 2012
Innovative Medicines | Discovery Sciences
Key Observations
• Range of complementary imaging
systems, fast and multimodal enables
platform selection to meet a breadth of
applications
In Vitro
Phenotypic
HCA
Clinical
systems
pathology
Preclinical
Imaging
• Application is dependent upon integration with
other specialties- advanced cell biology, liquid
handling & automation, informatics and data processing,
image analysis techniques
• Prosecution of Phenotypic functional endpoints to complement
traditional target-centric approaches. Powerful technique to allow
quantification from physiologically relevant models of disease.
31 | B Isherwood | 06 September 2012
Innovative Medicines | Discovery Sciences
Future Challenges & Prospects
• Increasing use of organotypic systems:
primary tissue & stem cell biology- robust
analysis & calibration at scale
• Development and adoption of novel
modalities & biosensors to monitor
e.g. small molecule location,
additional functional endpoints and
multiplexed options
Intensity Image
Lifetime Image
Yoyo-1
Talbot C et.al. (2008) J. Biophotonics. 1 p514-521
32 | B Isherwood | 06 September 2012
Innovative Medicines | Discovery Sciences
Future Challenges & Prospects
Colchicine (0.1 μM)
• Full exploitation of
information rich data using
machine vision technology or
single cell analysis developed
using bioimage benchmark
collections
Isherwood B. et. al. (2011) Pharmaceutics. 3 p141-170
• Linkage of phenotype with pathway & disease
through integration of data from diverse sources
e.g. genomics, proteomics & systems pathology
33 | B Isherwood | 06 September 2012
Innovative Medicines | Discovery Sciences
Acknowledgements
High Content Biology
James Pilling
Sam Peel
Sandeep Daya
Anesh Sitaram
Lauren Drowley
Mark Roberts
Wendy Vernon
Tracy Gorman
Claire Priest
Elizabeth Mouchet
Anne Camm
Linda MacCallum
Gillian Barrett
Dave Nicholls
Shaun Hawley
R&DInformation
Jeremy Campbell
34 | B Isherwood | 06 September 2012
Kirk Schroeder
Vince Groppi
Dyke McEwen
Beth Leslie
Brad Neagle
Neil Carragher
Ed Ainscow
Peter Caie
Rob Eagle
Jo Francis
Rebecca Walls
Tom Houslay
Michael Sullivan
Patrick O’Shea
Paul French
Clifford Talbot
Sunil Kumar
Discovery Statistics
Mike Dymond
Susan Lovick
Grant Zimmermann
Jeb Ledell
Oncology iMED
Sandra Brave
Hannah Greenwood
Claire Barnes
Paul Wylie
Sarah Payne
KingsMill Hospital
David Walsh
Innovative Medicines | Discovery Sciences
Advances in High Content Analysis:
Phenotypic Drug Discovery
Bev Isherwood
High Content Biology, Discovery Sciences, AstraZeneca
ELRIG Drug Discovery 2012, Manchester

Download Report