Background Tumor-immune microenvironment (TiME) profiling methods, such as antibody-based labeling and RNA sequencing, have increased the complexity of analyses possible but are costly, time-intensive, and do not faithfully capture structural or post-translational modifications that change functional behavior of critical proteins. Here, we demonstrate rapid, single-cell differentiation and functional state characterization in the TiME using vibrational spectroscopy with metasurface optics (VISMO) and machine learning.
Methods Our metasurfaces are composed of nanoscale silicon resonators that localize electric fields into subcellular regions, enhancing Raman scattering interactions by ~104 across the cell surface without cell heating. We collect ~1000 single cell Raman spectra of various melanoma and immune cell types (YUMM1.7, YUMMER1.7, B16F10, LN6-Engleman, RAW macrophages) on our metasurfaces. Using machine learning, we ascertain vibrational peaks corresponding to relevant cell membrane molecular features (eg. AXL and GAS6).
Results We show a classification accuracy of >94% across cell types (YUMM1.7 and B16F10, RAW macrophages) and >80% across parent and derivative lines (YUMM1.7, YUMMER1.7). We also obtain an average of 92% sensitivity and 98% specificity amongst all cell lines. By developing feature analysis software, we extract pertinent Raman peaks for classification, including AXL expression, an important indicator of cancer progression and metastasis. Then, we extend our characterization to multicellular, co-cultured samples, showing how melanoma-immune cell interactions preserve certain spectral features while introducing new features via cell fusioning. By stimulating cells with growth and inhibiting factors (eg. LPS and IL-10), we show our VISMO platform can be used to monitor melanoma cell plasticity and macrophage polarization changes. Finally, we show cellular Raman spectra acquired from formalin-fixed, paraffin-embedded (FFPE) samples can enable local cell identification.
Conclusions Digitally dewaxing of VISMO from FFPE specimens demonstrates high similarity scores against unpreserved counterparts. By extending our findings to whole tumor FFPE and frozen clinical specimens, we are enabling a label-free tumor profiling assay that could greatly advance clinical biomarker efforts and impact precision treatment decisions of cancer patients.
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.