• IMC allows for successful multiplex imaging of the TME in FFPE tissues.
• Images can be resolved at the single cell level with simultaneous
measurement of multiple membrane/cytoplasmic and nuclear markers
(including cell signaling and functional states), which can then be clustered
into relevant phenotypic subsets and re-analyzed in their spatial context on
the original tissue section images.
• By identifying distinct immune subsets and mapping their interactions, this
technology may potentially allow TME-based tumor subclassification and
identification of novel biomarkers.
• Additional studies on larger sample sizes of CHL and NLPHL are ongoing,
including “neighborhood” analyses to comprehensively define cell-cell
interactions in the TME.
• In classic Hodgkin lymphoma (CHL) and nodular lymphocyte predominant Hodgkin
lymphoma (NLPHL), composition of the tumor microenvironment (TME) varies
significantly across and within histologic subtypes and affects disease prognosis.
• Characterization of the TME is currently hindered by the phenotypic complexity of T cell
subsets, macrophage and other myeloid subtypes, and stromal/ vascular components.
Assessment of the spatial relationships of tumor cells and immune cell subsets is
difficult using conventional techniques that either require tissue disruption (flow
cytometry, gene expression profiling) or limit analysis to one to few markers per tissue
section (immunohistochemistry, immunofluorescence).
• Using metal-tagged antibodies on FFPE tissue sections, the Fluidigm Hyperion imaging
mass cytometry (IMC) combines a mass cytometer with a laser ablation system,
enabling simultaneous immunophenotyping by >40 markers on a single slide with
subcellular resolution (1 um).
• The complex multidimensional datasets generated by IMC can be analyzed by a
variety of data analysis tools to segment and classify individual cells, cluster similar
phenotypic subsets and determine their spatial relationships.
• In this study, we demonstrate the feasibility of this approach to analyzing the TME of
Hodgkin lymphomas.
Highly Multiplexed Analysis of the Tumor Microenvironment in
Hodgkin Lymphomas by Imaging Mass CytometryImran N. Siddiqi, Erik Gerdtsson, Monirath Hav, Mohan Singh, Parvesh Chaudhry, Wendy Cozen, James Hicks, Peter Kuhn, Akil A Merchant
University of Southern California, Los Angeles, CA-90033, USA
CD15
This work was supported by Ming Hsieh Institute, STOP Cancer, and National Institutes of
Health/ National Cancer Institute.
CD68
CD15
Antibody Clone TagBCL2 EPR17509 146NdBCL6 K112-91 147Sm
CD134/OX40 Polyclonal 151Eu
CD16 EPR16784 146Nd
CD183 (CXCR3) G025H7 142Nd
CD185/CXCR5 RF8B2 153Eu
CD194 (CCR4) 205410 149Sm
CD20 H1 161GdCD206 5C11 169Tm
CD279 (PD-1) NAT105 175Lu
CD3 Polyclonal, C-Terminal 170Er
CD30 JCM182 165HoCD31 C31.3 144NdCD34 QBEND/10 152SmCD4 EPR6855 156GdCD45RA HI100 155GdCD45RO UCHL1 173YbCD68 KP1 159TbCD8a C8/144B 162DyEphrinB2 EPR10072(B) 166ErFoxP3 236A/E7 163Dy
Granzyme B EPR20129-217 167Er
HLA-DR YE2/36 HLK 174YbICOS D1K2T 148NdKi-67 B56 168ErLAG-3 D2G40 153EupERK1/2 D13.14.4E 171Yb
pStat3 [Y705] 4PSTAT3 158Gd
T-bet D6N8B 145NdTim3 D505R 154SmVimentin RV202 143Nd
Figure 1: Malignant Reed Sternberg cell co-excessing CD30 (blue) and CD15 (cyan) surrounded
by CD4+ helper T cells (magenta), CD8+ cytotoxic T cells (green) and CD68+ macrophages
(orange) in classical Hodgkin lymphoma (A). Malignant LP cells (blue) surrounded by PD1+
(magenta) CD4+ helper T cells (orange), CD8+ Cytotoxic T cells (green) and CD68+ macrophages
(cyan) in nodular lymphocyte predominant Hodgkin lymphoma (B).
Figure 2. After segmentation
of the images single cell
feature data was extracted in
HistoCAT[1]. The
multidimensional data was
visualized using the tSNE
algorithm[2] to generate two
dimensional plots with the
default t-SNE parameters
(initial dimensions, 110;
perplexity, 30; theta, 0.5). We
used PhenoGraph version 0.2
for unsupervised clustering to
identify the multiple cell
populations.
1. Amir, A.D. et al. Nat. Biotechnol. 31: 545–552 (2013).
2. Levine J.H. et al. Cell. 162(1):184-97 (2015).
3. Schapiro D. et al. Nat. Methods. 14(9):873-6 (2017).
4. Vitozanotelli & Bernd Bodenmiller (2017). A flexible image segmentation pipeline for
heterogenous multiplexed tissue images based on pixel classification.
https://github.com/BodenmillerGroup/ImcSegmentationPipeline/blob/20170915_imc_semin
ar/documentation/201709_imctools_guide.pdf
IMC Workflow
1) Design
panels using
Antibodies conjugated
to metal tags
2) Stain
FFPE tissue
sections or Tissue
micro arrays(TMA)
3) Laser Ablate/Image
1mm2 tissue sections using the
Hyperion Imaging System
(Fluidigm). The data generated
by the instrument is converted
into 1 megapixels images.
4) Image Segmention for Single Cell data
CellProfiler is utilized to preprocess the images for iterative Ilastik
pixel classification. Using the Ilastik framework, a supervised random
forest based learning algorithm allows for dividing the pixels into
different classes e.g. nuclear, cytoplasmic/membranous and
background.
5) Data Analysis
The images along with the mask are analyzed with HistoCAT,
developed for IMC data analysis. In addition, the features of each
cell can be imported into the R environment for data visualization,
statistical analysis, and clustering of cell phenotypes.
Introduction Antibody Panel Classic Hodgkin Lymphoma
Figure 1: (A) Merged image showing manual identification of malignant Reed Sternberg (RS) cell
co-expressing CD30 (blue) and CD15 (cyan) with associated CD4+ helper T cells (magenta),
CD8+ cytotoxic T cells (green) and CD68+ macrophages (orange). (B) Close-up image showing
TME interactions with malignant RS cells.
PC 17: RS cells
PC11: Th1/ TIM3
PC14: exhausted T
PC10:?Macrophage
CD206+ Ki67+
PC1: Th cells
PC5: B cells
PC13: Macrophages
PC3: cytotox T
PC16: T regs
PC2: B cells (naïve/ memory)
PC4: B cells
Cluster
1
2
3
4
5
6
7
8
9
10 11 12 13 14
Cluster Presumptive Cell Types (?)
Th [CD3 dim, CD45 RO dim, CD4 dim]
Naïve, mem B [CD20, CD45RA, CXCR5+]
Cytotox T [CD3, CD8, CD45RO]
B cells [CD20 dim, CD45RA dim, CXCR5-]
B cells [CD20 dim, CD45RA dim]
? [CD8, CD45RO dim, CD45RA dim]
2 populations? [ICOS, CD3 dim, CD45RA dim]
Myeloid [CD16, CD206, CCR4]
B cells [CXCR5, Vim, CD20 low]
M2? [CD206, Ki67]
Th1 [Tim3, CD4 dim, CD3 dim]
TFH [CD3, CD4, CD45RO, CXCR5]
Macrophage [CCR4, CD68, CXCR3, CXCR5]
Exhausted T [CD3, CD4, Lag3, CD45RO]
? [vim low]
T regs [FOXP3, CD4]
RS cells [CD15, CD30, Ki67, Tbet, Tim3, Vim]
Cytotox T [Granzyme B]
? [CD4, Tbet, CD3]
CC
R4
CD
15
CD
16
CD
20
6C
D2
0C
D3
0C
D3
CD
45
RA
CD
45
RO
CD
4C
D6
8C
D8
CX
CR
3C
XC
R5
Ep
hrin
B2
FO
XP
3G
rz B
ICO
SK
i67
La
g3
Tb
et
Tim
3V
ime
ntin
A BC
Nodular Lymphocyte Predominant Hodgkin Lymphoma
Figure 3. (A) CHL PhenoGraph. Unsupervised algorithm identifies 19 phenotypic clusters (represented by different colors) in the assayed regions. Putative cell types are identified by their
overall marker expression patterns (arrowed). (B) Heatmap representing ranges of expression of each marker on Phenograph clusters and corresponding presumptive cell types. (C)
Comparison of manually gated RS cells compared to backgating of Phenograph cluster 17 (PC17) on the source image. PC17 gating identifies higher numbers of RS cells in the image.
1
2
3
4
5
6
78
9
10
11
13
14
15
16
12
17
1819
BCL2 BCL6 CCR4 CD20 CD31 CD3
CD45RA CD45RO
CD4 CD68
CD8 CXCR3 FOXP3 GRZ B
ICOS Ki67 Lag3 PD1
Vimentin
Tbet Tim3
1234567891011121314151617 Cluster
30 µ m
30 µ m
Figure 4. NLPHL was analyzed with a
similar panel of markers. Cell segmentation,
tSNE analysis and Phenograph clustering
(not shown) identified 17 phenotypic
clusters. (A) Heatmap showing patterns of
marker expression in the various
Phenograph clusters (PC). PC #7 (red
square) shows a pattern of reactivity most
consistent T follicular helper cells (TFH;
BCL6+, CD3+, PD1+, CD4). (B) PC7
backgated and projected back on the source
image clearly identifies TFH cell rosettes
around LP cells. (C) Merged image
identifies LP cell (CD20, red) and associated
CD4 (green) and CD8 (blue) T-cells).
LP
CD4
CD4
CD8
A
B C
Summary
Acknowledgments
References
DisclosuresThe authors do not report relevant disclosures.
