Background Head and neck cancer (HNC) is a heterogeneous group of malignancies that arise from the mucosal surfaces of the upper aerodigestive tract. The tumor microenvironment (TME) of HNC is characterized by the presence of immune cells, stromal cells, and extracellular matrix components. A key feature of the TME is hypoxia, which promotes tumor growth, invasion, and metastasis by altering the expression of genes involved in angiogenesis, cell survival, and metabolism. Understanding the complex interplay between hypoxia and immune infiltrates in the TME of HNC is crucial for the development of novel therapeutic strategies for the treatment of this disease. by digital spatial profiling is an excellent method of probing the TME, but assessing large cohorts can be time consuming. Automating a profiling workflow to reduce hands-on time and region of interest (ROI) selection bias will enable exploration of large cohorts to identify mechanisms of action, potential drug targets, and biomarkers.
Methods We developed an optimized spatial multi-omic workflow to enable high-throughput spatial analysis on GeoMx® Digital Spatial Profiler (DSP) using the Whole Transcriptome Atlas (WTA) and immuno-fluorescent morphology markers: SYTO82 (nuclei), CAIX (hypoxia), pan-cytokeratin (epithelium), CD3 (T-cells). A.I.-based analysis (Oncotopix® Discovery) of serial section H&E images and GeoMx IF images was developed to identify ROIs for GeoMx collection. Immune hot and cold selection used leukocyte density; tumor/stromal interface selection used epithelial areas. Areas of illumination (AOI) were chosen using concentric CAIX expression gradients. Integrated analysis of digital images using Oncotopix Discovery and the whole transcriptome was done to assess the above TME compartments.
Results AI/Deep Learning based AOI segmentation reduced AOI selection time and improved accuracy of tissue compartment enrichment, especially between samples and tissue types. Automated development of hypoxia gradient-based AOI enabled a selection strategy not possible in the standard DSP software. Cell phenotyping using IF morphology scan was used to supervise cell deconvolution. DSP results correlated well with patient outcomes.
Conclusions This work shows that ROI-based spatial analyses can be used to explore the effects of hypoxia levels on immune infiltration in HNC. Automated AI-based ROI selection provides a means of sampling relevant tumor subtypes based on hypoxia and immune infiltrate criteria in an unbiased, reproducible manner, and can provide a standardized, automated method for selecting ROIs and segmenting AOIs across a cohort of mixed tissue types and pathological subtype, improving throughput.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.