Assessing Graft Rejection by Automated C4d and CD34 Quantitation and Co-Localization

Assessing Graft Rejection by Automated C4d and CD34
Quantitation and Co-Localization
J.E. Tomaszewski1 T. Baradet1, C.H. Hoyt2, J.R. Mansfield2, and M. Feldman1
1) Pathology and Laboratory Medicine, University of Pennsylvania Health System, PA; 2) PerkinElmer, Hopkinton, MA
.
Abstract
Analysis of CD34 / C4d labeled cardiac FFPE biopsy sections
Multispectral Data Acquisition
Background. Rejection is the major cause of graft failure,
and if the injury to the organ is severe, it may not recover;
prompt diagnosis of acute rejection is therefore important,
with the monitoring of capillary C4d deposition being a
reliable early indicator of humeral rejection. Testing
requires taking two biopsies, one for frozen-section
analysis with immunofluorescence (IF) and the other for
formalin fixation and visual assessment with histochemical
stains. Currently, IF on frozen sections is the standard for
immunochemical evaluation of transplant biopsy
specimens. IF labeling is not often used for formalin-fixed,
paraffin-embedded (FFPE) specimens because of its
inherent autofluorescence, which makes IF imaging and
marker quantitation difficult. Spectral imaging can
overcome autofluorescence interference, enabling the use
of FFPE sections for assessment and eliminating the need
for dual biopsies. Additionally, vessels from biopsies from
transplants can be marked using one label (e.g. CD34) and
C4d using another then automated morphology-based
image analysis and quantitation can be used to obtain an
objective assessment of rejection.
Automated assessment of percent co-localization
of C4d within CD34 in vessels
Spectral Imaging and Unmixing
Conclusion. Automated quantitation of dual-labeled (CD34
and C4d) FFPE cardiac biopsy specimens can be achieved
using spectral imaging and morphologic image analysis
software, and gives a good correlation with results from
frozen sections and against visual assessment. This
methodology shows promise for becoming a routine
method for clinical assessment of organ transplant biopsies
and is amenable to studies of archival tissue.
High
High
High
High
High
High
High
High
High
Medium
Medium
Medium
Medium
Medium
Medium
Medium
Medium
Low
Low
Low
Low
Low
Low
Low
Low
Negative
Negative
Negative
Negative
Negative
Negative
Negative
Negative
90%
80%
80%
90%
80%
70%
80%
60%
70%
5%
10%
10%
15%
20%
25%
10%
40%
5%
2%
15%
10%
30%
10%
5%
5%
0%
0%
0%
0%
0%
0%
0%
0%
Vessel C4d Mean
Intensity (Counts)
81.9%
45.7%
60.2%
11.3%
10.0%
57.1%
2.3%
78.0%
2.6%
3.5%
0.4%
0.2%
0.2%
46.8%
31.2%
3.7%
84.0%
51.0%
0.0%
11.9%
2.7%
41.7%
0.0%
0.8%
4.9%
6.2%
3.6%
5.8%
1.0%
0.8%
3.2%
31.9%
2.0%
100.4
23.4
38.8
11.3
8.9
48.5
5.4
43.8
7.2
4.2
4.3
1.4
2.9
37.3
18.1
4.8
41.1
22.5
2.7
11.4
6.3
21.5
3.2
3.7
6.3
9.4
6.3
7.8
4.7
1.4
8.4
17.3
5.4
Sample 1480: 90%
Sample 23767: 0%
C4d image
C4d image
Multispectral imaging and automated analysis workflow
Comparison of Visual and Automated Assessments
100.00%
R = 0.510
90.00%
80.00%
70.00%
60.00%
50.00%
40.00%
30.00%
20.00%
10.00%
0.00%
0%
20%
40%
60%
80%
100%
Visual Assessment of Frozen IF
Mean Intensity per Group
60.000
50.000
40.000
30.000
20.000
10.000
0.000
Negative
Low
Med
High
3.412
9.706
14.264
35.034
Mean Intensity
p-values
Neg
Low
Med
Low
0.033
Med
0.051
0.246
High
0.012
0.029
0.062
Visual assessment of frozen samples
and automated assessment of FFPE
samples are shown in the table at
left. The images from two
representative FFPE samples are
shown above. A comparison of the
results of automated assessment vs
visual assessment is shown in the
graph. The overall correlation is
poor, with some clear outliers,
particularly in the High category.
80%
70%
60%
50%
40%
30%
20%
10%
0%
-10%
-20%
Percent Positivity
p-values
Neg
Low
Med
Spectral signal
isolation
Labels:
Green: C4d
Red: CD34
Pattern-recognition
image analysis
Color Code:
Color Code:
Green = C4d
Red = CD34
Blue = DAPI
Green = scar
Red = muscle
Blue = blank
Vessel
Segmentation
Autofluorescence
isolated and
removed
Percent Positivity per Group
70.000
-10.000
Nuance™, TRIO™ and
Vectra™ Multispectral
Imaging Systems
RGB Representation
of Spectral Cube
Automated Assessment of FFPE IF
1480
16004
16157
1665
3508
4855
7679
16785
6437
14777
8606
19729
4855
13204
15265
26633
17122
23492
15681
15943
16223
17395
18105
5844
9780
15265
18543
23767
23951
6946
11203
3434
7704
Visual
Automated
Assessment % Positivity
Percent Positivity of C4d
Results. High-quality IF images from FFPE specimens were
obtained using spectral imaging to remove
autofluorescence. Automated morphologic analysis of the
images identified vessels and quantified the C4d intensities
within those regions. Results from FFPE specimens were
comparable to those from frozen, and the image-based
objective measure of rejection status gave good correlation
to visual assessment.
Category
Mean Intensity of C4d
Design. A cohort of 33 matched formalin-fixed and frozen
cardiac biopsy specimens were sectioned and stained for
CD34 and C4d, and IF images of the FFPE samples were
acquired using a spectral imaging system. Multispectral
images were analyzed using an automated morphologic
image analysis software package to assess C4d staining
levels in capillaries. An image-based objective measure of
rejection status was then developed using a trainable
pattern-recognition-based image analysis tool. The
resulting algorithm automatically identifies capillaries and
measures C4D deposits in capillary walls and immediately
surrounding the capillaries. Automated measures were
compared to visual assessments.
Sample
Color Code:
Green = C4d +ve
Red = C4d -ve
Multispectral images of FFPE cardiac biopsy specimens were unmixed into their
constituent components, removing the autofluorescence and greatly increasing both
legibility and quantitative accuracy. The images were automatically segmented into
healthy muscle; scar and other tissue; and blank space. Within the healthy muscle
regions, each CD34 expression vessel was scored for C4d.
Negative
Low
Med
7%
14%
21%
Low
0.193
Med
0.121
0.297
High
39%
High
0.010
0.040
0.136
A comparison of the automated assessment of the FFPE section results differences between the four groups (Negative, Low, Medium and
High, as determined from visual assessment of the frozen section) shows that while there is a difference in the average value for each group,
the high standard deviation prevents many of the differences from being statistically significant. However, the differences between High and
Negative or Low are statistically significant for both the mean intensity of C4d staining and Percent Positivity. In addition, the difference
between all non-control groups (Low, Medium and High) and the Negative group are all statistically significant for Mean Intensity.
Conclusions
• Dual-color labeling of FFPE heart biopsies is easily achievable
• Multispectral data acquisition enables the elimination of interfering
autofluorescence for quantitative analyses
• Automated assessments of C4d/CD34 co-localization give a reasonable
correlation with results from visual assessment of frozen sections
• There are some differences between the visual assessment of frozen
sections and the image-based assessment of FFPE sections. This potentially
is due to differences between the two biopsies.
• This methodology shows promise for routine clinical transplant assessment
PerkinElmer, Inc., 68 Elm Street, Hopkinton, MA USA (800) 762-4000 or (+1) 203 925-4602 www.perkinelmer.com
Similar pages