Multiplexed Imaging and Analysis

Cancer researchers are zeroing in on the cellular combatants that strategically assemble in the tumor microenvironment (TME), where cancer and immune cells try to kill, evade, and outwit each other. The TME is a prime example of the need for multiplexed imaging technology to detect multiple proteins within cells or tissues simultaneously, for revealing locations of different cell types and their expression levels of key genes.

Two common multiplexed imaging methods—immunohistochemistry using immunofluorescence, and mass cytometry—are achieving this. Immunofluorescence-based methods use specific antibodies to identify proteins expressed in cultured cells, or sections of frozen or fixed tissues. Innovation in molecular tools has expanded multiplexing capabilities, including fluorescent reporters with various emission spectra. However, the number of potential markers can be limited by spectral overlap of fluorophores, and autofluorescence interference. Mass cytometry, which identifies proteins using metal-tagged antibodies, can multiplex more markers but is less common. This article provides examples of both methods, and discusses challenges of multiplexed imaging analysis.

Immunofluorescence

Spatial biology using fluorophore-tagged antibodies is a tried-and-true method that relies on iterative rounds of staining for many targets. For example, the Cell DIVE system from Leica Microsystems can image over 60 markers at the single-cell level. “Cell DIVE is a precise, open multiplexing solution that lets your research dictate the level of automation required, which antibodies to use, and how to build your antibody panel to uncover unique cell phenotypes and their distribution within the tissue microenvironment,” says Katie White, Cell DIVE Product Manager at Leica Microsystems. Although Cell DIVE was optimized for FFPE tissues, users have also developed protocols for studying frozen tissues.

Cell DIVE assays are fully customizable, with flexible antibody sourcing—choose your own, or use Cell DIVE validated antibodies from their collaborators Cell Signaling Technology. Leica also worked with Advanced Solutions Life Sciences to develop an optional robotic platform, which automates the pipetting and slide movements during staining and imaging. “Cell DIVE is a purpose-built solution for tissue multiplexing—so any application that can benefit from a comprehensive analysis of many protein biomarkers in a single tissue sample could benefit from Cell DIVE,” says White. “Typical applications are in cancer research, drug development, and tissue mapping projects.”

Long a provider of imaging instrumentation and confocal microscopy platforms, Zeiss also focuses on image analysis software using AI. The ZEN software runs Zeiss imaging systems, includes image analysis tools with machine learning, and can incorporate customized deep learning models trained on the Zeiss arivis Cloud platform. “The arivis Cloud trained models are used to automate analysis and/or detection of structures that are difficult to segment with traditional methods,” says Samantha Fore, Product Marketing Manager for Life Sciences Laser Scanning Microscopy at ZEISS Research Microscopy Solutions.

Fore emphasizes the flexibility inherent in Zeiss’s wide range of offerings in imaging instrumentation and analysis software, which can benefit almost any imaging application. As an example, she notes a recent study from Birgitt Schuele’s lab at Stanford University. “Multiple rounds of fluorescently tagged neuro-related markers were applied using the Akoya CODEX^® technology and imaged on the ZEISS LSM 980 [confocal microscope],” says Fore. “This demonstrates the flexibility to perform multiplex applications at high resolution and in 3D, and perform the necessary downstream processing and analysis of these complex multiplex data sets.”

For analyzing whole histopathology slides, CellCarta offers a digital pathology and image analysis service that stains samples for biomarkers of interest, in addition to markers enabling image co-registration over multiple slides. “Our algorithm for Ultivue's IO8 panel combines seven immune cell markers with a CK/SOX10 mask to characterize up to 13 phenotypes in the tumor and stromal compartments,” says Sofie Daelemans, Scientific Team Lead for the Pathology Histology Imaging Quantification Unit at CellCarta. “In addition, we can use image co-registration to align the images of serial slides to PanCk or PanCK-CD8 to characterize any other biomarker in the tumor and stromal compartments.”

CellCarta’s algorithm offers different hands-on levels, for example, from full manual annotation to fully automated. “It is possible for sponsors or clients to have a different workflow for the specific technical procedure with other acceptance or exclusion criteria than the default,” says Daelemans. “There are possibilities with our current image analysis platforms to share images and annotations so further analysis can be executed by researchers.”

Mass cytometry

Unlike immunofluorescence-based imaging, mass cytometry can image tens of markers in a single run. For example, the Imaging Mass Cytometry™ (IMC™) with the Hyperion XTi™ Imaging System from Standard BioTools uses metal-tagged antibody labeling of tissue sections. A UV laser ionizes the metal tags, creating signals that are detected according to mass using cytometry by time-of-flight or CyTOF^®. “Discrete signals from each ionized metal tag are detected based on differences in mass instead of wavelength, resulting in precise data without interference from autofluorescence,” says Clinton Hupple, Director of Proteomics Product Management at Standard BioTools. IMC can simultaneously image over 40 markers in a single scan with subcellular resolution, which would take longer if done with multiple cycles of immunofluorescence.

Increasingly, researchers are employing spatial techniques to identify cellular populations involved in complex diseases. “Many institutions are using IMC to help understand response to immunotherapy in cancer, the pathogenesis of infectious disease, and the rejection of transplanted organs,” says Hupple. For example, researchers at Children’s Hospital Los Angeles recently used IMC to study individual immune cells in biopsies from 27 children with acute liver failure. They found that children who recovered from acute liver failure without needing a liver transplant had T cells expressing the protein PD-1. In contrast, this T cell population was absent in biopsies from children who needed a liver transplant. “After identifying more than 83,000 immune cells, investigators discovered this rare population of T cells expressing PD-1, which may be useful as a biomarker for predicting organ recovery,” says Hupple.

Analysis challenges

Analyzing multiplex imaging data can be challenging, with large data files, multiple rounds of imaging to capture different fluorophores, and 3D datasets. “Registering these rounds of imaging and the ability to segment and perform image analysis in dense 3D tissues often presents challenges to the end user,” says Fore.

The ZEISS arivis Pro (formerly arivis Vision4D) software is designed for complex image analysis using data from Zeiss or other microscopes. “Arivis Pro allows the user to be limitless in the number of channels, tiles, z-stacks, and more for datasets generated in their multiplex imaging workflows,” says Fore. “Customized deep learning AI models trained on arivis Cloud or publicly trained models such as Cellpose can be/are integrated with the software and can be used for segmentation of overlapping and intermixed cells in dense 3D tissues.” To address the challenges of using large datasets, Zeiss offers the ZEISS arivis Hub (formerly arivis VisionHub), a scalable image processing and data storage option that can relieve bottlenecks encountered during data processing.

CellCarta is increasingly using AI to reduce the user input required for analysis, such as automatic exclusion of artifacts. “We are seeing much higher accuracy in nuclear segmentation leading to better and more consistent results, especially in more challenging assays,” says Daelemans. “Thanks to AI we may no longer need masks like PanCK or SOX10 to perform regional segmentation as we can train the algorithm to recognize these ourselves.”

The analysis of mass cytometry data is also benefiting from AI. “Because IMC provides high-dimensional, multiplexed data, the image analysis approach can diverge from classical pipelines,” says Hupple. Standard BioTools offers a recommended pipeline for data analysis, which now includes integrated AI-driven analysis from Visiopharm^®. “We also recognize and support cytometry experts who are contributing to and developing tailored data analysis approaches, and provide a platform for users to see what’s being used in high-impact publications and learn from other Imaging Mass Cytometry experts,” says Hupple. Whether you collect multiplex imaging data using mass cytometry or immunofluorescence-based methods, new algorithms and the application of AI will likely boost your analytical power.