We’ve rebranded some of our products, learn more ›

CODEX is now PhenoCycler
Phenoptics is now Phenolmager

From Stars to Cells: Johns Hopkins Researchers Discover Predictive Spatial Phenotypic Signatures with AstroPath

One of the biggest questions in immuno-oncology is figuring out how patients will respond to immunotherapy. Approaches like single-plex immunohistochemical stains don’t capture enough information to accurately predict response. To help solve this challenge, researchers at Johns Hopkins University (JHU) developed AstroPath, a novel platform which combines the Phenoptics multiplex immunofluorescence (mIF) workflow with sky-mapping algorithms derived from astronomy to perform deep spatial profiling of microscopic tumor sections.

In a groundbreaking study recently published in Science, the JHU team used AstroPath to build two-dimensional maps of the tumor microenvironment in metastatic melanoma, identifying a composite spatial phenotypic signature that is highly predictive of response to anti-PD-1 immune checkpoint inhibitor therapy and patient outcome.

The paper describes a strategy for multispectral multiplex immunofluorescence (mIF) assay development and image analysis for the generation of large, standardized datasets to facilitate immuno-oncology biomarker discovery.

Lessons from astronomy

Whole-slide mIF imaging produces massive datasets, presenting a challenge for data analysis. Astronomers struggled with the same problems a few decades ago, when they began building comprehensive, three-dimensional maps of the sky. The Sloan Digital Sky Survey, one such map, was built by Alexander Szalay, PhD, astrophysicist and Bloomberg Distinguished Professor in the Department of Computer Science at JHU.

Janis Taube, MD, MSc, a pathologist and Director of the Division of Dermopathology at JHU, worked with Dr. Szalay to apply lessons from astronomy to mIF image acquisition and analysis. They identified the need for a well-designed relational database that enables data consistency and efficient data querying. Human intervention also needed to be minimized, so data flows from instrument to database. And large datasets tend to be dominated by systematic, rather than statistical errors, requiring a conscious effort to identify and mitigate errors. These lessons guided the development of AstroPath.

From astronomy to pathology with AstroPath
AstroPath identifies a predictive spatial phenotypic signature for advanced melanoma

Once AstroPath was optimized for the generation of high-quality, whole-slide mIF datasets, the team got to work identifying features in the melanoma tumor immune microenvironment predictive of immunotherapy response.

They developed an Opal mIF panel for six markers:  PD-1, PD-L1, CD8, CD163, FOXP3, and SOX10 and performed imaging on the Vectra quantitative pathology system. The mIF panel was validated against the current clinical gold standard, chromogenic IHC, and was also verified in an independent cohort from a different academic institution.

“Akoya were the ones [with the staining reagents] that were closest to the clinic in terms of true clinical utility, and that’s really what I was looking for…Also in terms of the imaging microscopy equipment, they’re closest to clinical implementation,” Dr. Taube told GenomeWeb.

Using their validated panel with AstroPath, the authors were able to spatially map complex cell phenotypes, characterize PD-1/PD-L1 expression in situ, and determine the predictive value of these features for immunotherapy in advanced melanoma.

Assessing PD-1 expression intensity, the team showed that PD-1 expression increases as lymphocytes enter the tumor and that, among CD8+PD-1+ cells, those with PD-1low/mid are most likely to respond to PD-1/PD-L1 blockade compared with PD-1high cells. These results support findings from previous studies performed using in vivo animal models.

The team identified two key specific cell phenotypes predictive of response. The density of CD8+FoxP3+PD-1low/mid cells was closely associated with a positive response to anti-PD-1 immunotherapy, while the CD163+PD-L1negative myeloid phenotype was associated with a lack of response.

Area under the curve (AUC) of 6-plex mIF assay using a ranked, “hot spot” approach to HPF selection. The CD8+FoxP3+PD-1low/mid spatial phenotypic signature had the greatest predictive value with a sampling strategy of 20-25 HPFs.

A framework for biomarker discovery and development

In this study, the authors present a detailed, multi-step approach for mIF assay development and validation. Of note is the unbiased acquisition of whole-slide images, followed by standardized selection of high-powered fields (HPFs).

As the authors point out in the paper, previous mIF studies characterizing the tumor microenvironment in large cohorts have used a manual, operator-dependent approach to HPF selection. The operator-independent approach put forth here enables more robust biomarker identification and development.

Other important features of the multi-step approach include the benchmarking of mIF staining sensitivity against chromogenic IHC and the “single-marker” strategy for cell phenotyping (individually assigning cells positive or negative values for each marker) which improved cell segmentation accuracy.

Together, these approaches provide a framework for mIF assay development and data analysis that can be used to standardize results across research institutions – an important step towards clinical implementation.

Subscribe to Our Blog for Updates

By signing up you are agreeing to our Privacy Policy.