David L' Rimm M'D', Ph'D

1
Beyond IHC Accurate, reproducible and
quantitative measurement of protein analyte
concentrations in fixed tissue
David L. Rimm M.D., Ph.D Professor, Dept. of
Pathology Yale University School of Medicine
2
Disclosure

• I am a consultant, stockholder and scientific
  founder of HistoRx
• I am an author on the Yale held patent on the
  AQUA technology.

3
Use of analyte measurement to determine patient
management
Clinician suspects possible diabetes
Clinician suspects possible breast cancer
Obtain tissue sample (blood)
Obtain tissue sample (core biopsy)
Measure Blood glucose levels in mg/dl (objective
measurement)
Make Histologic Dx then measure estrogen receptor
levels (subjective judgment)
Treat with appropriate therapy
Treat with appropriate therapy
4
Fluorescence Microscopy for In situ Measurement
of Protein Expression

• Assessment of features on the basis of molecular
  interactions rather than morphology-based
  (contrast generating) edges (feature extraction).
• Not biased by post hoc human selection of the
  area selected for scoring.
• Broader dynamic range than chromagenic methods
• Emissive signal rather than absorptive to
  facilitate multiplexing

5
AQUA method of objective analyte measurement on
a tissue slide based on co-localization
Step 1 Mask (define region of interest, exclude
stroma, blank space, etc) colocalization with
Cytokeratin for carcinoma Step 2 Define the
numerator (target) and denominator (compartment)
S target intensity in compartment pixels
Numerator
Concentration
AQUA score
Denominator
S compartment pixel area
Step 3 Calculate the AQUA score
Step 4 Convert to absolute concentration or
normalize to set of uniform standards
6
Cytokeratin
Tumor Mask
Generating the AQUA score
Estrogen Receptor
TMA
WTS
TMA-Tissue Microarray WTS-Whole Tissue Section
S target intensity in compartment pixels
Combine DAPI image and cytokeratin image then
cluster to assign each pixel to a subcellular
compartment
AQUA score
S compartment pixel area
7
HT 29 SUM 159 BT 20 ZR 751 BT 474 T47D MCF7 Puro
0 Puro .01 Puro .1 Puro .5 Puro 1 Puro 5 MB 231 H
1666 MB 436 CHO SKBR3 MB 435
ERa levels by western
1 ng 2.5 3.5 5 7.5
10
rER
ERa
b-tubulin
AQUA score
ERa levels in cell lines by AQUA score
Alley Welsh
8
Limit of detection is 50 pg/mg
Use these cell lines (present on TMAs) to convert
patient ER AQUA scores from YTMA 49 to an ER
concentration (pg/ug)
Alley Welsh
9
HER-2 ELISA vs. AQUA Score
Anthony McCabe, PhD (JNCI, December 2005)

• With optimal antibody titer, AQUA scores are
  directly proportional to molecules of protein
  per cell by ELISA
• Amplification status is highly correlated to AQUA
  scores (plt0.0001)

10
AQUA analysis Normalization/Standardization
Example HER2
Operator 1
Operator 2
Mach. 1
Mach. 3
Mach. 2
Mach. 1
Stain 1
Stain 1
Stain 1
Stain 1
CV4.3 Cumulative
HER2 by AQUA analysis by 3 different operators
on 3 different machines with three different
batches.
PM2
PM1
PM1
PM3
PM1
PM1
PM1
10
11
The HistoRx AQUA platformHardware, Software,
Reagents for Quantitative Pathology

• Automated Fluorescence Microscopy
• Expanded dynamic range of measurement
• Multi-parametric
• Commercially available with about 18 current
  placements worldwide
• AQUA analysis software compatible with .tiff
  images
• In use by more than a dozen Pharma companies for
  drug development
• AQUA software now available on Aperio scanscope
  FL platform
• US Patent 7,219,016

12
Limit of detection is 50 pg/mg
Use these cell lines (present on TMAs) to convert
patient ER AQUA scores from YTMA 49 to an ER
concentration (pg/ug)
Alley Welsh
13
The Yale Breast Cancer Cohort YTMA 49 A Tissue
Micro Array with 640 (320 Node Negative and 320
Node Positive) Breast Cancer Cases, Controls and
Cell Lines
14
What is the ER false negative rate?
Yale Cohort (1962-1982)
Misclassification rate 16.7 False negative
rate 6.7 False positive rate 10
Confirmation of AQUA analysis ER- positive/IHC
ER-negative patients respond to tamoxifen in
progress in NSABP B14 and TEAM trial
Alley Welsh
15
Comparison of traditional IHC to AQUA analysis --
Reproducibility
Estrogen Receptor
Path 1 v. Path 2 Kappa 0.482 (plt0.001) Path 1
v. Path 3 Kappa 0.444 (plt0.001) Path 2 v. Path
3 Kappa 0.400 (plt0.001)
Positive/Negative concordance 92-95
Mark Gustavson
16
BCCA Cohort ER vs AQUA ER
Local ER status
ER -
ER
17
SWOG 9313 Local ER vs AQUA ER as a function of
HER2
18
Misclassification in YTMA 130
Linear conversion to pg/ug
19
Key Issues for Accurate Measurement on Protein on
Slides (limitations of the substrate and the
analytical reagents)

• Pre-analytic variables (cold ischemic time)
• Antibody validation

20
Whole section analysis of core biopsy vs.
resection to assess affect of ischemic time on
stability
Number of fields examined on each tumor
Yalai Bai and Eirini Pectacides
21
Lower Expression of pAKT in the resections than
in CNBs
Wilcoxon Signed Ranks test p0.004
Yalai Bai
22
Lower Expression of pERK in the resections than
in CNBs
Wilcoxon Signed Ranks test p0.38
Yalai Bai
23
Decreased Level of ER Expression In Surgical
Resections
Wilcoxon Signed Ranks test p0.014
Yalai Bai
24
Cytokeratin shows variability as a function of
tumor differentiation, but is generally lower in
resections than in CNBs
Wilcoxon Signed Ranks test p0.23
Yalai Bai
25
Steps for Normalization to Cytokeratin as a
method to adjust for global protein degradation
seen during cold ischemic time
1. Calculate a tissue specific Protein
degradation ratio
TR 70 CNB 100
0.7
For example
2. Then adjusted Target AQUA score by dividing
by the degradation ratio.
Target AQUA 250pg/ug CK normalization ratio
0.7

For example
Adjusted Target AQUA 250/0.7 357 pg/ug
Bx Biopsy TR Tumor Resection CK Cytokeratin
Yalai Bai
26
P0.0137
P0.0273
27
5, 26 7, 12
8, 22
11, 20
9,10
11, 11
25, 23
17, 23
10, 26
22, 19
P0.0039
P0.0059
AQUA Score in TRs is adjusted
28
P0.0645
P1.0000
AQUA Score in TRs is adjusted
29
P0.0049
P0.0020
AQUA Score in TRs is adjusted
30
P0.375
P0.0098
AQUA Score in TRs is adjusted
31
9, 31 15, 28 12, 11 9, 15 15, 26
5, 25 5, 29 18, 21
P0.1094
AQUA Score in TRs is adjusted
32
Toward Intrinsic Normalization

• Received NCI Biospecimen/SAIC contract to develop
  method for normalization of pre-analytic
  variables for protien assessment on FFPE
• Testing a series of housekeeping and hypoxia
  sensitive markers as normalization controls

33
Key Issues for Accurate Measurement on Protein on
Slides

• Pre-analytic variables (cold ischemic time)
• Antibody validation

34
Antibody Validation (reproducibility)
HER2 on serial sections
R0.966
c-Met Antibodies
35
EGFR antibodies tested
Elsa Anagnostou
36
R20.006
R20.05
R20.21
R20.03
R20.0007
R20.003
R20.0005
R20. 008
R20.17
R20. 61
Elsa Anagnostou
37

• Validate C17 polyclonal Ab
• Run WB on known positive controls (A431 and
  BaF3-HER3) and multiple NSCLC cell lines
• Test a different LOT on YTMA97 and correlate with
  previous LOT
• Run WB with new LOT and compare with previous

Elsa Anagnostou
38
Figure 1
R20.94
R20.96
R20.96
R20.95
R20.96
R20.95
Alley Welsh
39
QC Protocol for Antibody Validation

• Test on cell line series with Western blot
  correlation
• Test on non-expressing cell line with transfected
  with analyte
• Test on high expressing cell line with and
  without siRNA knockdown
• Test for lot to lot reproducibility

40
Example of validation of mTOR antibody
H1299
AS20
Elsa Anagnostou
41
Objective To develop a quantitative,
reproducible and easily applicable protein-based
test for prediction of NSCLC sub-classification
(Adenocarcinoma vs Squamous Cell Carcinoma)
42
M2. Antibody Validation and Assay Reproducibility
Assay Reproducibility for CK13
Cell Line Controls (AQUAWB)
Representative AQUA output in cell lines
H1666
H1355
A431
CK13
R0.91
y0.81x24.14
HT29
A431
Assay Reproducibility for EGFR
Cell Line Controls (AQUAWB)
Representative AQUA output in cell lines
H1355
H1666
A431
EGFR
R0.91
y14.66x94.14
A431
HT29
43
M2. Antibody Validation and Assay Reproducibility
Assay Reproducibility for CK5
Cell Line Controls (AQUAWB)
Representative AQUA output in cell lines
HCC2279
HCC193
H1666
H1299
A549
A431
R0.97
y56.74x-35.3
HT29
A431
CK
Assay Reproducibility for TTF1
Cell Line Controls (AQUAWB)
Representative AQUA output in cell lines
HCC2279
HCC193
H1299
A549
R0.95
H1299
HCC193
y6.73x-17.69
44
Training Set, Greek YTMA-140 Cohort (n280)
Table 2 Final reduced model
Model equation P(AC)1/(1e-(9.065-0.814CK130.67
1TTF1-0.685CK5-0.393EGFR))
1000 bootstrap samples
Model equation P(AC)1/(1e-(9.679-0.899CK130.71
1TTF1-0.687CK5-0.427EGFR))

• 4 key markers
• Cytokeratin 13
• TTF1
• Cytokeratin 5
• EGFR

45
Validation set 1, Yale YTMA-79 Cohort (n170)
Scores of specimens in the testing cohort
ROC curve for AC classification
SCC
AC
Sensitivity
AUC0.984
Specimen Score
1-Specificity

• Cut off point for classification between AC and
  SCC set at 0.4
• Sensitivity92.3, Specificity96.6, Area under
  the curve (AUC)0.984

46
Validation set 2, Moffitt/Bepler Cohort (n180)
ROC curve for AC prediction
Scores of specimens in the validation cohort
AC SCC
Sensitivity
AUC0.989
1-Specificity
Specimen Score

• Validation of the cut off point set at 0.4
  (testing cohort-YTMA79)
• Sensitivity 97.2, Specificity 94.8, Area
  under the curve (AUC)0.989

47

• PCA analysis was performed to identify the most
  prominent directions of the 28-dimensional data
• Interestingly AC and SCC clustered together
  however AS did not form a distinct group

Three dimensional plot of principal component
analysis
Three dimensional plot of rotated components
(rotation methodVarimax)
SCC
AC
AS
Rotations were used to better align the
directions of the factors with the original
variables so that the factors may be more
interpretable. The Varimax method tries to make
elements of this matrix go toward 1 or 0 to show
the clustering of variables
48
Thanks to
Rimm Group Robert Camp Elsa Anagnostou Bonnie
Gould Rothberg Veronique Neumeister Seema
Agarwal Huan Cheng Maria Baquero Alley
Welsh Jason Hanna Jennifer Bordeaux Bill
Bradley Tassos Dimou Summar Siddiqui Liz
Killiam Yalai Bai
Tissue Microarray Facility Lori Charette Joe
Salame Sudha Kumar Peter Gershkovich
Former Rimm Group Marisa Dolled-Filhart Maciej
Zerkowski Kyle DeVito Tony McCabe Greg
Tedeschi Mark Gustavson Carola Zalles Aaron
Berger Sharon Moulis Malini Harigopal Chris
Moeder Jena Giltnane Eirini Pectasides
Yale Collaborators Annette Molinaro Gina
Chung Harriet Kluger Ruth Halaban Lyndsay
Harris John McClaskey
Outside Yale Gerold Bepler (Moffitt) Elaine
Alarid (UW) David Huntsman (BCCA) Dan Hayes (U
Mich)
49
www.tissuearray.org
Rimm Lab Summer 08
50
Tissue Microarray (TMA) Construction
Dolled-Filhart and Rimm (2002) Principles and
Practice of Oncology Technology Update