Multiplexed Ion Beam Imaging (MIBI)

Abstract

Background: An ability to characterize the cellular composition and

spatial organization of the TME using multiplexed IHC has been limited

by the techniques available. Here we show the applicability of MIBI™

technology for cell phenotype identification and spatial relationships

analysis across multiple tumor types.

Methods: FFPE samples from tumor TMAs were stained with a panel of

15 antibodies, each labeled with a specific metal isotope. Brightfield IHC

served as a comparator to establish that the labeling procedure did not

affect antibody affinity. In MIBI the stained section was scanned via

secondary ion mass spectrometry to image the tissue for target

expression. Leave-one-out (LOO) analyses (a single antibody left out of

the panel, compared to the full MIBI panel) showed negligible loss in

signal intensity of targets indicating an absence of interference between

antibodies. Multi-step processing and segmentation produced images of

the TME and the frequency of different cell subsets. Distances between

cell subsets were calculated using this data.

Results: Panel validation by LOO showed that all markers were

repeatable across 9 panels (R2 0.94 - 1.00). Control samples imaged at

study start and end showed consistent marker quantification (inter-run

R2 > 0.99), indicating MIBI staining and acquisition is robust and

reproducible. A total of 50 tumor specimens from 15 tumor types were

characterized for their immune profile and spatial organization (nearest-

neighbor). Most samples showed infiltrating cytotoxic T cells and

macrophages present amongst tumor cells. Spatial analysis of the TME

in two ovarian serous carcinoma samples showed differences in the

distance between tumor and immune cells, highlighting the degree of

mixing between tumor and immune cells. Identification of admixed

PD-L1+ macrophages and PD-1+ T cells in a urothelial carcinoma

sample allowed for calculation of distances to each other.

Conclusions: MIBI offers high-parameter multiplexing capability, at

sensitivity and resolution uniquely suited to understanding the complex

tumor immune landscape including the spatial relationship of immune

and tumor cells and expression of immunoregulatory proteins.

Multiplexed Ion Beam Imaging (MIBI) For Characterization of the Tumor

Microenvironment Across Tumor TypesJason Ptacek1, Darren Locke2, Rachel Finck1, Mary-Ellen Cvijic2, Zhuyin Li2, Murat Aksoy1, Jay Tarolli1, Yi Zhang1, Jessica Finn1

1Ionpath Inc. Menlo Park, CA 2Bristol-Myers Squibb Princeton, NJ

MIBI™ is a trademark of IONpath INC. For Research Use Only. Not for diagnostic use.

© Copyright 2019, IONpath.

MIBI Data Recapitulates IHC Staining Patterns Across

A Variety of Tissue Types

Single Cell Profiling of Immune Subsets of the TME

Spatial Distribution of Immune Cells and Tumor Cells

Differs Between Ovarian Cancer Samples

CD3

CD8

CD68

Keratin

CD8+PD-1+ T cells (red)

CD4+PD-1+ T cells (blue)CD4 and CD3 colocalized (green)

CD8 and CD3 colocalized (red)

CD45

Keratin

Tumor vs Immune

CD3

CD4

CD8

Immune Cell Subsets

CD4

CD8

PD-1

Immune Checkpoint Proteins

PD-L1 present on:

• CD68+ macrophages (blue)

Urothelial cancer sample shows

presence of TILs (cyan)

CD3

CD-68

PD-L1

Immune Checkpoint Proteins

• Patient 34 shows a compartmentalized organization of the TME versus

Patient 35 that shows mixing of immune cells and tumor cells.

• Calculation of nearest-neighbor distances between tumor cells and the

nearest immune cells further characterizes differences in TMEs.

Patient 34

Conclusions

• MIBI offers many features advantageous for the analysis of

complex tissues including the tumor microenvironment.

• MIBI control data indicates that staining and sample

acquisition is robust and repeatable across tissue types.

• The tissue architecture is defined at sub-cellular resolution

enabling segmentation that further enables single-cell

analysis and phenotypic characterization.

• Sensitivity afforded from the low background of MIBI allows

for markers with low expression such as PD-1 and FOXP3

to be readily detected.

• This study evaluating multiple tumor types shows the

possibilities of MIBI for patient stratification and immune profiling

of tumor samples under therapeutic forces.

Acknowledgements

We thank all patients who have donated samples for this investigation.

Patient 34

Patient 35

Patient 34Nearest-Neighbor Distances

Histogram of Nearest-Neighbor Tumor-Immune Distances

MIBI Data Reproducible For High, Medium and Low

Abundance Targets

Patient 35

A

B

C

D

E

F

Breast Carcinoma,

Triple Negative

B

Colonic

Adenocarcinoma

C D E

Prostatic

Adenocarcinoma

F

Urothelial Carcinoma

H

Cutaneous Melanoma

I

Endometrial

Adenocarcinoma

J

H

I

J

B C D E

F H I JG

G

Lung Adenocarcinoma

Lung Squamous Cell

Carcinoma

G

Non-Hodgkin

Lymphoma, Low Grade

A 100 𝜇m

A

Breast Carcinoma,

ER Positive

250 𝜇m

1 mm

• Example images of 10 tumor types comparing H&E images to MIBI.

• MIBI images showing a subset of markers analyzed by MIBI illustrate the

spatial organization of tumor cells and immune cells.

PD-L1+ cells

PD1+CD8+ T cells

PD1+CD4+ T cells

PD1+FoxP3+ T cells

Nearest-Neighbor Distances

NuclearMarker(s)

MembraneMarker(s)

Hybrid Multiplexed Image Watershed Segmentation

Target Clone Label Cell Expression

dsDNA 35I9 DNA 89Y All cells

beta-tubulin D3U1W 166Er All cells

HLA class 1 A, B, and C EMR8-5 160Gd All cells

Na-K ATPase alpha1 D4Y7E 176Yb All cells

Keratin AE1/AE3 165Ho Tumor cells, epithelial cells

CD3 D7A6E 159Tb T cells

CD4 EPR6855 143Nd T helper cells, some macrophages

CD8 C8/144B 158Gd Cytotoxic T cells

CD45 2B11 & PD7/26 175Lu Immune cells

CD56 MRQ-42 150Nd NK cells, neurons

CD68 D4B9C 156Gd Macrophages

FOXP3 236A/E7 146Nd Regulatory T cells

PD-1 D4W2J 148Nd T cells

PD-L1 E1L3N 149Sm Tumor cells, macrophages, APCs

SOX10 BC34 170Er Melanocytes, nerve cells

CD3

CD8

CD68

PD-1

PD-L1

Acq. Date: 5/10/18

Sample: Tonsil 1706

5/27/18

Tonsil 1707100 𝜇m

Phenotypic Classification

100 𝜇m

Titrate and test

antibody by IHC Conjugate

AntibodyCompare by IHC

Conjugated Ab v.

Unconjugated AbTitrate conjugated

antibody by MIBI

MIBI

Staining Panel For Phenotypic Analysis of the TME

• MIBI preserves spatial information allowing tumor-immune cell boundary and

distance determination.

• Segmentation uses multiplexed data and prior training data as ground truth.

• An ion beam rasters across the tissue liberating ions including the

isotopes bound to the tissue via the antibodies.

• Time-of-flight mass spectrometry separates the labels based on mass for

detection of markers present across the tissue.

• Antibody performance is assessed by IHC before conjugation to an isotope.

• Conjugated antibody performance is measured by MIBI and compared to IHC.

dsDNA

CD45

Keratin or SOX10

LOO Panel R2

Full - FOXP3 1.00

Full - PD-1 1.00

Full - PD-L1 1.00

Full - CD56 0.99

Full - CD68 1.00

Full - CD8 1.00

Full - Keratin 0.99

Full - Sox10 0.99

Maker Intensity Across Tonsil Replicates

Run 1 Ion Counts

Run 2

Ion C

ounts

Slope=0.99

R2=1.00

• Serial sections of lung adenocarcinoma were stained

with either the full panel or panels that lacked one

antibody (LOO panel).

• Regression analysis of signal intensity between Full

and LOO panels are summarized in the table.

• Images of keratin expression are similar between the

full panel and panels lacking FOXP3 or PD-1.

• Keratin signal is absent from the TMA stained with the

panel without keratin demonstrating the lack of

significant signal overlap across channels.

• Replicate serial sections of tonsil (shown), placenta, and melanoma acquired

by MIBI before and after acquisition of sample cohort show equivalent results.

• Segmented cells are assigned

a phenotype based on

combination of markers

• Distances between PD-L1+

cells and PD-1+ cells are

calculated

Single Cell Marker Expression

Cells (Cell Score)

Targ

et

Target expression across cells is

quantified by Cell Score, a metric

determined by the target’s signal

intensity and its distribution across a cell

No FOXP3

No Keratin No PD-1

Full Panel 100𝜇m

Keratin

Full Panel Ion Counts

Full

Pane

l –

FO

XP

3 I

on C

ounts

Slope=0.99

R2=1.00

Maker Intensity Between Panels