Shooting for multiplexed pathology with Orion

Abstract

Pathological diagnosis relies on morphological assessment of tissue using histological staining and molecular phenotyping through immunostaining that must be performed on separate tissue sections. Orion is a newly reported methodology that facilitates multiplexed immunostaining with histological staining on the same slide.

Histology, the study of cellular morphology and organization in the cells’ native tissue environment dates back to the invention of the microscope, when the resolution limitations of the human eye were first overcome. Much of early histology was focused on understanding tissue structure, anatomy and roles in bodily functions, but it later evolved into pathology, the study of abnormal histology. In this context, hematoxylin and eosin (H&E) staining has been the gold-standard methodology, used for over 150 years to evaluate differences in tissues and diagnose disease. This approach is based on global pattern recognition across the tissue sample, rather than at the single-cell level, and relies on the pathologist’s expertise to recognize and understand patterns in the tissue for disease (including cancer) diagnosis. In contrast, immunostaining approaches enable molecular analysis at the single-cell level, with immunohistochemistry being a routine pathology tool. The promise of highly multiplexed immunostaining to complement genomic analyses for improved diagnosis and treatment of disease has resulted in a plethora of technological advances to enable the detection of 10–100 antigens on a single slide.

Global tissue analysis using H&E and single-cell immunohistochemistry approaches could provide improved understanding of tissue dysfunction and disease. However, combined staining and imaging of the same tissue slide is subject to substantial challenges because both techniques require the use of colorimetric dyes that cannot be readily visualized together. In this issue of Nature Cancer, Lin et al.¹ present a multiplexed immunostaining technology, termed Orion, that achieves one-shot multiplexed immunofluorescence followed by H&E staining on the same tissue section to generate matched molecularly specific immunostaining and pathologically meaningful H&E staining (Fig. 1).

Fig. 1 |. A comparison of existing highly multiplexed immunostaining strategies to the Orion platform.

a, Many multiplexed immunostaining strategies utilize conventional antibody-staining techniques (immunohistochemistry or immunofluorescence), including application of indirect (primary and secondary), directly conjugated and DNA-barcoded antibodies. Highly multiplexed image data are generated by repeating rounds of staining, imaging and signal removal using chemical bleaching, enzymatic or UV cleavage. Increasing the number of cyclic rounds leads to reduced tissue integrity and cell loss, rendering the resulting slides unusable for accurate pathological evaluation by H&E. b, Mass spectrometry imaging (MSI) uses metal-conjugated antibodies (each with its own unique isotope), applied to a tissue section in a single ‘master mix’. An ion beam is scanned across small regions of interest within the tissue section, which releases secondary metal isotope ions that are detected and quantified using a time-of-flight (ToF) mass spectrometer. Despite decreasing the staining time by eliminating multiple rounds of staining and imaging compared to conventional cyclic immunostaining strategies, this method is limited in resolution due to the ion beam spot size and cannot be used for whole-slide imaging due to the length of the scan time. c, The Orion platform developed by Lin et al.¹ provides a simplified immunostaining strategy in which 18-plex immunostaining and imaging are performed in a single step to preserve tissue integrity for subsequent H&E staining and pathological evaluation of the same tissue section. These data are combined, and image feature models are generated using machine learning algorithms to improve predicted patient outcomes as compared to existing pathological tools used in clinical practice.

Highly multiplexed immunostaining technologies rely on two main methodologies: (i) conventional antibody-staining techniques utilized in a cyclic fashion² and (ii) mass spectrometry imaging (MSI) using antibodies labeled with rare earth metals^3–5 (Fig. 1a,b). Cyclic immunostaining methods including cyclic immunohistochemistry and immunofluorescence have seen widespread adoption due to the stains’ detectability through conventional microscopy. To generate highly multiplexed immunofluorescence or immunohistochemistry images, cycles of staining, imaging and signal removal (for example, antibody stripping^6,7 or fluorophore bleaching^2,8) are required. Although effective, these protocols generally require weeks to complete and are destructive to the tissues. MSI methods, such as CyTOF⁴ and MIBI³, do not require cycling, resulting in improved tissue preservation. However, the resolution of MSI is restricted by laser-scanning spot size, creating a trade-off between single-cell resolution and whole-slide scanning due to the required scan times for high-resolution imaging of large tissue areas. To address these limitations, hybrid techniques have been developed that utilize unique antibody tags, such as DNA barcodes, analogous to the rare-earth-metal-tagged antibodies⁹. This permits the use of a single immunolabeling step with a ‘master mix’ of antibodies, decreasing overall cyclic immunostaining time. Antibody barcoding techniques (for example, Nanostring^10,11, CODEX¹², DNA exchange imaging¹³ and Immuno-SABER¹⁴) facilitate highly multiplexed immunostaining using non-destructive signal removal techniques. However, these methods still require cycles of staining and imaging, and most also rely on removal of the coverslip between cycles, which can result in substantial cell and tissue loss as cycle number increases. To date, all existing methods are difficult to integrate with gold standard H&E. Cyclic immunostaining methods (immunohistochemistry or immunofluorescence) cause substantial tissue changes, resulting in poor-quality H&E staining not suitable for pathological analysis after multiple rounds of staining. In addition, the trade-off of MSI between single-cell resolution and whole-tissue scanning has largely resulted in the sampling of small regions of interest within tissues, making this strategy difficult to integrate with the global tissue view of H&E.

In response to these challenges, Lin et al.¹ developed Orion (Fig. 1c), an imaging approach that required innovation in both the optical platform and the selection of fluorophores to permit the imaging of up to 18 channels simultaneously. The authors combined narrow-bandpass emission filters, for improved specificity, with high-powered excitation lasers and complementary metal-oxide semiconductor detectors for improved sensitivity. The Orion microscope was equipped with seven distinct laser lines that permitted the excitation of up to 18 fluorophores that were spectrally separated using a combination of the selected narrow-bandpass filters. Spectral extraction to isolate signal from individual fluorophores was subsequently completed during image processing. After testing more than 100 probes from different sources, the authors selected 18 fluorophores on the basis of emission wavelengths in the 500–875-nm range, high quantum efficiency and photostability, as well as compatibility with one another in high-plex imaging panels. These 18 fluorophores were directly conjugated to commercially available antibodies to characterize the immune infiltrate, proliferative cell state, as well as epithelial and endothelial cells, resulting in a panel of labeled antibodies suitable for studying the microenvironmental and architectural features of epithelial tumors and their adjacent normal tissues. Notably, because tissue processing was minimized, H&E images obtained using standard automated slide staining after first employing the Orion staining technique, were deemed visually equivalent by trained pathologists to serial sections that had not undergone highly multiplexed immunostaining.

The authors validated the Orion approach through the collection of three data types: (i) whole-slide imaging of human tonsil and human lung cancer, (ii) tissue microarray imaging of up to 30 different types of human tissues from normal, non-neoplastic tissue to common tumor types and (iii) whole-slide imaging of 74 archival resection specimens of colorectal cancer, for which they compared immunostaining data to those obtained through a well-established cyclic immunofluorescence (cyCIF) method. The image data generated using Orion were found to be similar both qualitatively and quantitatively in all three data types. Furthermore, the authors demonstrated the ability to perform two rounds of staining using the Orion platform, with H&E staining after immunostaining resulting in both high-quality immunofluorescence and H&E images. When these data were combined and analyzed using machine learning, the authors showed that multiplexed image-based models of CRC improved the prediction of progression-free survival compared to previously reported immunoscoring methods¹⁵. Thus, the authors’ approach combines high-plex immunofluorescence staining at single-cell to subcellular resolution with H&E imaging of the same cells to generate whole-slide format imaging of tissues that is suitable for clinical translation compared to other more labor-intensive and destructive cyCIF strategies.

The validated Orion platform provides the potential to combine >150 years of histopathological knowledge about normal and diseased tissues with molecular data derived from multiplexed immunofluorescence imaging that could be used to expand our knowledge of the origins and progression of disease and ultimately develop better-targeted, more precise treatment strategies for patients. This technology can be readily translated to the growing field of machine learning algorithms to further our comprehension of tissues in health and disease. Importantly, Orion is supported by open-source software and is compliant with Open Microscopy Environment and Minimum Information about Tissue Imaging data standards, making the platform readily available to other investigators for community use to improve disease diagnosis and prognosis, particularly for cancers.

Although the Orion approach simplifies the multiplexed imaging protocol and reduces the number of cycles to preserve tissue integrity for subsequent H&E, using the one-shot immunolabeling method diminishes the number of antibodies that can be used in a single panel (18–36 for one or two cycles) compared to other cyclic immunostaining technologies. For some applications, a reduced panel will be sufficient, but for a more in-depth view of cell state and signaling within the tumor microenvironment, a different multiplexed method will be required to facilitate higher-plex tissue interrogation. Furthermore, optimization of the antibody panel, including selection of fluorophore labels to enable accurate spectral unmixing, will be necessary if changes in biomarker targets are desired from the initially developed panel reported herein, which limits the flexibility of the Orion approach compared to existing multiplex methodologies. Lastly, whether using a serial H&E section, instead of using the same tissue section for multiplex immunofluorescence, would result in similar findings (that is, improved progression-free survival prediction) remains to be determined.

In summary, the work by Lin et al.¹ presents an innovative immunostaining technique that circumvents some of the known challenges of existing cyclic immunostaining techniques, which are limited by tissue destruction and the spectral resolution of conventional microscopes. Exploration of same-slide immunostaining and H&E-stained tissues using the Orion platform is poised to enable the quantification of whole-tissue patterns in the context of single-cell biomarker composition to improve understanding of disease.