Practical Approaches to Tumor Xenograft Analysis
Post on 30-Dec-2015
39 Views
Preview:
DESCRIPTION
Transcript
Practical Approaches to Tumor Xenograft Analysis
Frank Voelker, Flagship Biosciences LLC
Trevor Johnson, Flagship Biosciences LLC
Veronica Traviglone, Infinity Pharmaceuticals
Igor Deyneko, Infinity Pharmaceuticals
www.flagshipbio.com
Outline
Introduction
Anatomy of a Xenograft
Defining the Approach to the Analysis
Analytical Strategies
Examples of Cases
Presenting FACTS for Target Tissue Analysis
www.flagshipbio.com
General Anatomy of a Xenograft
CD-31 Stain for Microvessels
www.flagshipbio.com
Chronic Inflammation and Peripheral Adipose Tissue
Adjacent tissue frequently contains host responses such as inflammation, adipose tissue or adnexal structures that need to be excluded from the analysis
www.flagshipbio.com
Extensive Necrosis and Degeneration
Necrosis and neoplastic cells in various stages of degeneration are common features of a xenograft.
CD31 Immunostain
www.flagshipbio.com
Host Response of Vascular Ingrowth Demonstrated with CD31 Immunostain
Even the smallest microvessels may be enclosed in adventitial tissue as an extension of stromal ingrowths
www.flagshipbio.com
Compare H&E and IHC Stains
H&E
CD31
www.flagshipbio.com
Compare Corresponding Areas of H&E and CD31 Stains
Compare microscopic characteristics of an H&E- stained section with comparable regions of an IHC CD31-stained section
This will provide valuable perspective regarding IHC target tissue staining and will allow more accurate identification of tissue classes.
H&E CD31
www.flagshipbio.com
Significant Xenograft Tissue Components
Neoplastic CellsTissue BiomarkerConnective Tissue StromaNecrosis and DegenerationCystic or Secretory Vacuoles Artifactual Space
Which of these do you want to include in your analysis?
www.flagshipbio.com
Some Examples of Analytical Endpoints
Neoplastic Cell AreaTotal Xenograft Area
Marker Area or Score Neoplastic Cell Area
A function of neoplastic cell abundance versus necrosis and/or stromal prominence
Immunohistochemistry expression of marker metabolism within neoplastic cell population
Number of Neoplastic NucleiNeoplastic Cell Area
An indication of mean neoplastic cell size
Targeted Cell NumberTotal Xenograft Area
Cell frequency within measured area
Number of MicrovesselsNeoplastic Cell + Stromal Area
Microvessel analysis
www.flagshipbio.com
What aspects of a marker determine treatment effect?
(Nwp/Ntotal)x(100) + Np/Ntotal)x(200) + Nsp/Ntotal)x(300) = “H Score” (For a maximum of 300)
Staining variation in tumors or other tissues frequently raises the question of whether to measure either percent area or average intensity of an immunostain.
In some cases, a solution to this problem is to measure score as an output convention encompassing both percent area and stain intensity.
www.flagshipbio.com12
Manual Use of Positive and Negative Pen Tools
Similar IHC staining of fibronectin and secretion droplets in this xenograft tumor with subsequent poor differentiation by the Genie™ classifier required the use of the negative pen tool to assist in quantitating fibronectin using the IHC deconvolution algorithm.
www.flagshipbio.com
Manual Use of Positive and Negative Pen Tools
Use of a 21UX Cintiq Wacom drawing board facilitates the use of positive and negative pen tools in manually delineating critical features of the xenograft.
www.flagshipbio.com
Using Genie™ to Segregate Neoplasm from Nontarget Tissue in a Xenograft Stained for Phospho-Histone 3
Central regions of the xenograft contain an interdigitating pattern of necrosis and connective tissue trabeculae too complex for manual exclusion .
Accurate segregation of neoplasm (green) from necrosis and connective tissue (red) was accomplished using Genie™
www.flagshipbio.com
The Rationale Behind Histology Feature Recognition
Using software to preprocess an image with the purpose of segregating target tissue components from nontarget tissue.
Aperio Genie™
Visiopharm
Subsequent analysis yields accurate data only from the target tissue component, and omits erroneous nonspecific results from nontarget tissue.
Feature recognition is valuable in xenograft analysis when target and nontarget regions are too intricately interwoven for manual exclusion.
www.flagshipbio.com
Strategy for Analyzing Successive Samples
Progression of analysis through sample series
Red = 1st ClassifierBlue = 2nd ClassifierGreen = 3rd Classifier* = Points when a new classifier was developed during the analysis
***
Xenograft neoplasms within a study are surprisingly heterogeneous even though derived from the same source, and it is difficult and time consuming to derive a common pattern recognition classifier appropriate for all neoplasms within a study. It is more expeditious and also yields more accuracy to develop new classifiers as the analysis progresses.
www.flagshipbio.com
Consistent Analysis of Study Samples Using Genie™
Subsequent Uniform Analysis of Segregated Target Tissue for area/intensity
Even though multiple classifiers are constructed for successive samples, the final analyses of target tissue components will be accurate and comparable if
the same algorithm threshold values are constantly maintained.
www.flagshipbio.com
Using Genie™ to Segregate Neoplasm from Nontarget Tissue in a Xenograft Stained for Phospho-Histone 3
Higher magnification reveals accurate separation of target from nontarget tissue.
The grayed-out nontarget tissue class consists of a mixture of connective tissue trabeculae and necrotic tumor cells.
www.flagshipbio.com
Assessing Fibronectin in a Xenograft Using Several Software Programs
Assessing the intensity and quantity of fibronectin as a marker in xenograft stroma
Aperio Deconvolution Mark-up
Visiopharm Mark-up
Original IHC Image
www.flagshipbio.com20
Assessing Angiogenesis in a Xenograft Using Several Software Programs
Use of the microvessel analysis algorithm to assess angiogenesis in a xenograft neoplasm in a mouse
Microvessel analysis provides important information regarding potential antineoplastic effects of pharmaceutical compounds
Original CD31-Stained Image
Aperio Microvessel Algorithm Mark-up
Visiopharm Mark-up
www.flagshipbio.com
Using the Microvessel Analysis Algorithm to Count Cyclin B1-Positive Target Cells in a Xenograft
Use of the microvessel analysis algorithm to assess macrophage populations in mouse xenograft neoplasms
Threshold algorithm parameters are modified to accommodate the smaller size and shape characteristics of the cells
www.flagshipbio.com
Use of Visiopharm™ Software to Analyze a Marker Separately in Cytoplasm and Nuclei of a Neoplasm
Discriminating between nuclear and cytoplasmic regions of a neoplasm allows separate biomarker intensity measurement for both nuclear and cytoplasmic markers (analysis with Visiopharm software).
www.flagshipbio.com
Distinctive or Special Staining Facilitates Target Tissue Pattern Recognition
H&E H&E Classifier
CD31 CD31 Classifier
First pass accuracy is much greater for the H&E-stained section than for the hematoxylin- counterstained IHC section
Current histology pattern recognition programs often have difficulty distinguishing target tissue profiles because of low contrast /nonspecific counterstaining
www.flagshipbio.com
Feature Analysis on Consecutive Tissue Sections (FACTS)
4. QC and pathologist
review
3. Image and ROI
registration
2. Automatedfeature
recognition
1. Consecutive tissue
sectioning
www.flagshipbio.com
Image and ROI Registration
Register images with <3% Feature Error Across multiple consecutive slides– Not a trivial problem given the potential histological and
biological differences between tissue sections and stains
Several registration techniques currently used throughout CT/MRI, CT/PET, CT radiology, ultrasound, and multispectral imaging– Typically elastic, morphing, because of image warping– Heavy use of image features or fiducials
Type of registration for best performance in FFPE tissue is heavily dependent on the application
3. Image and ROI
registration
www.flagshipbio.com
Typical Image Registration Process
Feature detection
– Distinctive objects to be aligned
Feature matching
– Finding the association between the distinct objects
Transform model estimation
– The mapping functions used to align the subsequent image to the reference image
– May be non-elastic (much simpler) or elastic
Image re-sampling and transformation
– Applying the mapping functions where non-integer coordinates can be correctly interpolated
www.flagshipbio.com
Non-Elastic Image to Image Registration
Perform color deconvolution to focus on a common Stain, typically Hematoxylin
– Helps reduce potential errors from strongly stained “positive” features
Use slide features and image intensities to adjust rotation and translation of selected images or regions
u(x,y) -> v(x,y)
Rotation: x’ = x•cos(θ) - y•sin(θ)
y’ = x•sin(θ) + y•cos(θ)
Translation: x -> x’
y -> y’
www.flagshipbio.com
Example of Slide Registration
www.flagshipbio.com
Annotated Feature to Feature Registration
More complex due to biological and histological differences between serial sections
Usually Elastic to count for these differences
– Important to note that the actual images are never morphed, just the annotations
Let the application determine the method
Phase Registration:
℮xp(2 ∏ i (u x’ + v y’)) = ƒ(f) ƒ(g) / | ƒ(f) ƒ(g) |
Mutual Information:
MI(X,Y) = H(Y) – H(Y | X) = H(X) – H(Y) – H(X,Y)
where H(X) = Ex(log(P(X))
www.flagshipbio.com
Wavelet Transform:
Decompose image into collection of appropriate wavelet (Spline, Haar, Etc.)
Filter image along rows and columns (high and low pass)
Find frequency coefficients and apply differential measurements
Annotation Feature to Feature Registration (cont.)
www.flagshipbio.com
Example of Feature Registration
www.flagshipbio.com
Acknowledgments
Dr. Steven Potts, Flagship Biosciences LLC
Dr. David Young, Flagship Biosciences LLC
Ms. Charlotte Aagaard Johnson, Visiopharm Inc.
Mr. Rob Diller, Flagship Biosciences LLC
Mr Erik Hagendorn, Flagship Biosciences LLC
Questions? Email contact: frank@flagshipbio.com
trevor@flagshipbio.com
top related