Proteomic dysregulation plays an important role in the spread of diffusely infiltrating gliomas, but several relevant proteins remain unidentified. Digital spatial processing (DSP) offers an efficient, high-throughput approach for characterizing the differential expression of candidate proteins that may contribute to the invasion and migration of infiltrative gliomas.
Diffusely infiltrating gliomas are associated with high morbidity and mortality due to the infiltrative nature of tumor spread. They are morphologically complex tumors, with a high degree of proteomic variability across both the tumor itself and its heterogenous microenvironment. The malignant potential of these tumors is enhanced by the dysregulation of proteins involved in several key pathways, including processes that maintain cellular stability and preserve the structural integrity of the microenvironment. Although there have been numerous bulk and single-cell glioma analyses, there is a relative paucity of spatial stratification of these proteomic data. Understanding differences in spatial distribution of tumorigenic factors and immune cell populations between the intrinsic tumor, invasive edge, and microenvironment offers valuable insight into the mechanisms underlying tumor proliferation and propagation. Digital spatial profiling (DSP) represents a powerful technology that can form the foundation for these important multilayer analyses.
DSP is a method that efficiently quantifies protein expression within user-specified spatial regions in a tissue specimen. DSP is ideal for studying the differential expression of multiple proteins within and across regions of distinction, enabling multiple levels of quantitative and qualitative analysis. The DSP protocol is systematic and user-friendly, allowing for customized spatial analysis of proteomic data. In this experiment, tissue microarrays are constructed from archived glioblastoma core biopsies. Next, a panel of antibodies is selected, targeting proteins of interest within the sample. The antibodies, which are preconjugated to UV-photocleavable DNA oligonucleotides, are then incubated with the tissue sample overnight. Under fluorescence microscopy visualization of the antibodies, regions of interest (ROIs) within which to quantify protein expression are defined with the samples. UV light is then directed at each ROI, cleaving the DNA oligonucleotides. The oligonucleotides are microaspirated and counted within each ROI, quantifying the corresponding protein on a spatial basis.
Diffusely infiltrating gliomas are the most common type of malignant brain tumor in adults and are invariably lethal. The propensity for glioma cells to migrate extensively in the brain is a major therapeutic challenge. The mechanism by which they spread involves directed migration and unchecked invasion. Invasive glioma cells have been shown to exhibit tropism and migration along white matter tracts1, with recent research implicating demyelination of these tracts as an active, protumorigenic feature2. Invasion is mediated by an epithelial-to-mesenchymal transition, in which glioma cells acquire mesenchymal properties by reducing the expression of genes encoding extracellular matrix proteins and cell adhesion molecules, amplifying migration and facilitating propagation through the tumor microenvironment3,4,5.
At the molecular level, disruption of several proteins that confer cellular stability and interface with immunogenic components has been demonstrated6. Infiltrative gliomas are known to undergo suppression of proteins with anti-apoptotic (e.g., PTEN) properties7. They also overexpress proteins that promote evasion of the host immune response (e.g., PD1/PDL1)8. The dysregulation of these complex pathways enhances tumorigenicity and increases malignant potential.
Within samples of invasive glioma, the aim was to evaluate the differential expression of proteins key to cell growth, survival, and proliferation, and to microenvironment structural integrity between invasive and non-invasive components. Additionally, we sought to study the differential regulation of proteins with an active immunogenic role, offering insight to the mechanism by which compromised host immune defenses may enhance the proliferative and invasive potential of gliomas. This is especially relevant given the recent breadth of research demonstrating how immune markers and drivers of dysregulation in malignancy can serve as targets of immunotherapy. Identifying viable therapeutic targets among the many proteins involved in immunosurveillance and reactivity requires a highly sensitive and comprehensive approach.
Given the wide array of candidate proteins that can be studied, we sought a method akin to immunohistochemistry but with enhanced data processing efficiency. Within the field of cancer biology, DSP has emerged as a powerful technology with important advantages over alternative tools for proteomic analysis and quantification. The hallmark of DSP is its high-throughput multiplexing capability, allowing for simultaneous study of several different proteins within a sample, marking an important distinction from standard but lower-plex technologies such as immunohistochemistry (IHC)9,10. The multiplex feature of DSP does not compromise its fidelity as a quantitative and analytical tool, as demonstrated by studies comparing DSP to IHC. When used for proteomic quantification of non-small cell lung cancer specimens, for example, DSP has been shown to have similar results to IHC11. Additionally, DSP offers customizable regional specification, in which users can manually define regions within which to perform proteomic analysis. This presents an advantage over whole-section multiplex methods10,12. In a single round of processing, DSP thus offers multiple layers of analysis by surveying several protein targets across multiple regions of interest.
DSP has applications in several different pathological settings. DSP is especially advantageous in oncologic analysis, as spatial variation can correlate with cellular transformation and differential protein expression. For example, DSP has been used to compare the proteomic profile of breast cancer to the adjacent tumor microenvironment. This carries important implications for understanding the natural history of this tumor and its progression, as well as potential response to treatment13. Additional contexts illustrating the versatility of DSP include spatial quantification of protein diversity in prostate cancer14, association of immune cell marker expression with disease progression in head and neck squamous cell carcinoma15, and demonstration of an epithelial-mesenchymal gradient of protein expression distinguishing metastatic from primary clear cell ovarian cancer16. By implementing DSP, we characterize the spatial topography of proteins that could impact tumorigenesis and invasion of gliomas.
The protocol outlined below follows the guidelines of the Dartmouth-Hitchcock Human Research Ethics Committee. Informed consent was obtained from the patients whose tissue samples were included in this study. See the Table of Materials section for details related to all materials, reagents, equipment, and software used in this protocol.
1. Slide preparation17
2. Semi-automated IHC system preparation and software configuration (for loading and running of slides)17
3. Antibody incubation and nuclei staining17
4. Fluorescence visualization, ROI identification, and UV photocleavage on the DSP instrument17
5. Protein readout17
6. Data analysis17
Figure 4 shows the representative results from a DSP experiment performed on samples of glioblastoma. A heat map is presented, illustrating one of the methods by which to capture data visually using the DSP software. Rows represent protein targets, and each column corresponds to a region of interest. A color range of blue to red denotes low to high expression, respectively. Variability of color within a row reflects regional protein heterogeneity and suggests a possible spatial association with differential protein expression. For example, in the present experiment, S100 and CD56 are universally high because they are neural markers.
Markers with the most variability included B7-H3, Ki-67, CD44, and fibronectin. These markers have important associations with tumor proliferation, migration, and metastasis, respectively. It is thus reasonable that they would exhibit regional variability between samples of malignant core, malignant invasion, and non-malignant tissue. Numerical data representation is also possible; a file containing the expression value of each protein marker within an ROI in table form can be exported for mathematical and statistical analysis. This expression value is produced by the digital counting feature available through DSP, which quantifies the microaspirated PC-oligos within each ROI (Figure 1 and Figure 2).
Figure 1: Defining regions of interest in tissue microarray from glioblastoma biopsies. Top, Hematoxylin and eosin-stained section of tissue microarray containing several 2 mm diameter core biopsies from a total of three glioblastoma patients. Bottom, regions of interest defined on fluorescence image. Please click here to view a larger version of this figure.
Figure 2: Photocleavable oligonucleotides. Antibody or antisense oligonucleotide probes for protein or RNA targets, respectively, are covalently linked to oligonucleotides with photocleavable linkers. Tissue sections are stained with the probes. UV light is then directed at ROIs to release the PC-oligos, which are digitally counted. This figure was modified from9, with permission of Springer Nature (copyright 2020). Abbreviations: PC-oligos = photocleavable oligonucleotides; ROIs = regions of interest. Please click here to view a larger version of this figure.
Figure 3: DSP procedure. 1) Tissue processing: A tissue slide is stained with oligo-conjugated antibody or RNA probes. 2) ROI selection: The tissue slide is imaged, and ROIs are delineated, either manually or automatically based on certain fluorescence patterns. 3) Cleavage of conjugated oligonucleotides: UV light is directed at the ROIs, resulting in cleavage of the photocleavable oligonucleotides. 4) PC-oligo collection: The PC-oligos are aspirated into a microcapillary tube. 5) Plating: Aspirated PC-oligos are deposited into a microtiter plate. 6) Repeat: Steps 3-5 are repeated for each ROI; between each cycle, meticulous washing is performed. 7) Quantification: Spatially resolved pools of PC-oligos can be hybridized to fluorescent barcodes; this allows for digital counting of up to approximately 1 million binding events per ROI. Alternatively, PC-oligos can be quantified via NGS, in which the entire microtiter plate is pooled into a single tube and sequenced. The reads are then translated into digital counts and mapped back to their original ROI, creating a visual map of protein or RNA expression within each tissue section. This figure was modified from9, with permission of Springer Nature (copyright 2020). Abbreviations: DSP = digital spatial processing; PC-oligos = photocleavable oligonucleotides; ROIs = regions of interest; NGS = next-generation sequencing. Please click here to view a larger version of this figure.
Figure 4: Representative heatmap from a DSP experiment. Columns represent individual regions of interest. Rows represent a protein target. Low expression is shown in blue, and high expression is shown in red. Abbreviation: DSP = digital spatial processing. Please click here to view a larger version of this figure.
GeoMx Immune Cell Profiling- human | Pan-Tumor Module |
PD-1 | MART1 |
CD68 | NY-ESO-1 |
HLA-DR | S100B |
Ki-67 | Bcl-2 |
Beta-2 Microglobulin | EpCAM |
CD11c | Her2 |
CD20 | PTEN |
CD3 | ER-alpha |
CD4 | PR |
CD45 | |
CD56 | |
CD8 | |
CTLA4 | |
GZMB | |
PD-L1 | |
PanCk | |
SMA | |
Fibronectin | |
Rb IgG | |
Ms IgG1 | |
Ms IgG2a | |
Histone H3 | |
S6 | |
GAPDH |
Table 1: A representative sample of proteins that can be assessed using predesigned antibody panels.
Probe U Working Pool | ||||||
Module 2 Probe R | Other modules | DEPC-treated water (microliters) | total volume (microliters) | Probe U Master Stock (microliter) | DEPC-treated water (microliter) | total volume |
2 | … | 16.5 | 2 | 14.5 | 16.5 | |
4 | … | 33 | 4 | 29 | 33 | |
6 | … | 49.5 | 6 | 43.5 | 49.5 | |
8 | … | 66 | 8 | 58 | 66 |
Table 2: Hybridization codes and Probe R and Probe U working pools. Abbreviation: hyb = hybridization; DEPC = diethyl pyrocarbonate.
Given the diversity of proteins that could potentially influence the aggressiveness of gliomas and the notion that several of these proteins remain undiscovered, a high-throughput protein quantification method is an ideal technologic approach. Additionally, given that spatial data in oncologic samples often correlates with differential expression18, incorporating spatial profiling into the protein quantification approach allows for more effective analysis.
The high-throughput approach of DSP also enables it to be used in a shotgun-like approach, which is ideal for discovering potential novel biomarkers of disease and response to therapy. In two studies of melanoma, DSP was used to evaluate the responses to combination therapy with ipilimumab and nivolumab versus nivolumab monotherapy19 and to neoadjuvant combination therapy with ipilimumab and nivolumab versus standard adjuvant treatment20. In both studies, DSP profiling of a variety of protein targets post treatment demonstrated differential relative expression of key immunologic markers, suggesting a role for potential biomarkers of therapeutic response.
DSP also facilitates the determination of the ultimate pathological significance of complex, molecular-level processes. For instance, a distinctive feature of invasive gliomas is their propensity to migrate directly through the extracellular matrix (ECM), and their migratory path is often guided by vascular and white matter tracts1,2. Variability in protein expression of the tissue microenvironment is a hallmark of the epithelial-to-mesenchymal transition that drives invasion. Several studies have examined how gliomas can upregulate matrix metalloproteinases (MMPs), resulting in the breakdown of the surrounding ECM and increasing invasiveness21,22,23. By visualizing and quantifying the cumulative, downstream effect of this differential regulation, DSP provides a widescale, holistic analysis.
A key determinant of the malignant potential of gliomas is their ability to interact with and manipulate host immune defenses. Similar to other solid tumors, gliomas employ a variety of mechanisms to evade host immunosurveillance and disrupt the activation of proteins critical to mounting an immune response. Recent advancements in immuno-oncology have focused on identifying immunogenic proteins that are expressed or exploited by tumors. At the forefront of these developments are the use of T cells to detect immunologically active tumor antigens against which oncolytic viruses can be directed24,25, and the development of checkpoint inhibitors against T cell inhibitory proteins (e.g., PD1/PDL1 and CTLA4) to potentiate host immunity against tumors26,27. Other stromal populations that figure prominently into the glioma microenvironment include macrophages, microvascular endothelial cells, immune cells, and a recently identified subpopulation termed 'glioma-associated stromal cells' (GASCs)28. The interactions of these cells with chemokines, cytokines, and components of the ECM have important implications for tumor proliferation, invasion, and response to treatment, and thus serve as important targets for regional quantification and comparison through technologies such as DSP.
The DSP protocol is user-friendly and allows for multiple layers of customization. As much of the quantification process is automated, the main user-dependent steps are geared toward attaining the highest possible sensitivity and specificity of target detection. Key steps within the protocol thus include determination of a stain for initial fluorescent visualization of the sample, delineation of ROIs, and selection of the antibody panels.
When designing a staining protocol, it is important to first identify a target that is likely to demonstrate quantifiable differences between various regions of the sample, corresponding to a critical characteristic or behavior of the tissue of origin. Next, an agent that will effectively bind the target must be selected. For example, if one hopes to visualize regions of hyperplasia or nuclear atypia, H&E or a nuclear stain may be chosen; alternatively, if aiming to visualize changes in the expression of a gene known to be associated with malignant or premalignant tissues, a fluorophore-conjugated antibody may be chosen. The tumor cells themselves can be specifically labeled in the case of IDH1-R132H mutant tumors. The present protocol could be modified by the addition of fluorescently labeled IDH1-R132H as a morphologic marker (protocol step 3.2). Up to four stains per sample may be used. Once visualization of the stained tissue has occurred, ROIs are designated. Selection of ROIs should reflect where significant differences in protein target expression are expected to occur. Key considerations when selecting ROIs include size (10-600 µm diameter), shape (circle, rectangle, or user-drawn), and segmentation (further demarcating sub-regions within a single ROI, offering an additional potential layer of analysis). These variables should be optimized to most effectively capture spatial variation in target expression that is anticipated based on the selected antibody panel.
Selection of the antibody panel is of critical importance, as differential antibody expression ultimately serves as the study end point. When performing this step, it is important to consider which markers may exhibit expression that correlates either positively or negatively with stages along the spectrum of tumor development, from benign, to preinvasive, to invasive or malignant. It is likewise important to carefully consider properties of both the tissue of origin and the tumor type, as both of these features may influence the expression of certain markers.
Because much of the process is automated, any flaws in the procedure are generally due to hardware or software malfunction, offering limited user troubleshooting capability. Issues that may arise include disruption of the built-in quality control (QC) measure (designed to flag and potentially omit data suggestive of low signal-to-noise ratio, low counts, and inadequate target detection), aberrant software updates, and procedural issues (e.g., premature termination of various steps). For these issues, it is recommended that the user contact the manufacturer for remote support. Alternatively, if the user encounters intrinsic data issues (e.g., low nuclei count overall, low variability in expression between ROIs), they may consider repeating earlier steps in the procedure (e.g., repooling data from the original TMA, redefining ROIs). It is also important to consider variations in tissue density as a potential confounder of antigenic production and thus antibody expression; hence, tissues of similar densities should aim to be selected for analysis.
Control measures are built into the protocol. The program provides both an internal positive and negative control that account for variations in hybridization of the PC-oligos to the fluorescent barcodes. Housekeeping proteins offer another positive control that can additionally serve as a normalization factor on the basis of cellularity.
The ability to define ROIs within which quantification occurs is the foundation of the spatial profiling capability of DSP. This customizable demarcation of ROIs marks a critical and distinctive step within the protocol. The option to create an ROI as small as a single cell allows for regional precision in protein or nucleic acid quantification, and the ability to define multiple ROIs and quantify their contents in tandem through multiplex analysis yields a high-throughput approach.
When ROIs are selected to represent pathologically distinct regions within a glioma sample (e.g., necrotic, perivascular), proteomic profiling and regional comparison may reveal mechanisms of tumor propagation and progression. For example, IHC has been used to regionally quantify hypoxia-inducible factor (HIF-1α) expression in glioblastoma samples and has revealed higher expression near necrotic areas compared to perivascular zones29. This is consistent with several studies indicating a role for hypoxia in the tumorigenesis of glioma.
Regional variation has also been widely studied among immune cell components of the tumor microenvironment. Macrophages/microglia constitute the predominant immune cell population in gliomas and have been comprehensively studied on a regional basis. Subpopulations of tumor-associated macrophages (TAMs) have been shown to play different roles in tumor progression depending on their location within the tumor and microenvironment. Those within the tumor invasive edge break down the basement membrane, promoting spread; those in hypoxic areas exert an angiogenic effect, facilitating growth; those in proximity to tumor vasculature secrete EGF and associated factors to direct stromal tumor cells toward blood vessels, driving metastasis30. These critical, spatially dependent properties demonstrate the value of regional proteomic data in cancer biology and represent an important additional application of DSP to the characterization of immune cell populations within a tumor and its microenvironment.
The technique has a few limitations, including cost and the relatively small number of targets that are available in the core antibody panels. In addition, although subcellular resolution is not possible with DSP, it will be possible with upcoming new technologies31. Despite these limitations, DSP is a powerful technique for certain types of targeted, previously unanswerable questions in glioma cell biology. This innovative technology presents an exceptional opportunity for uncovering new biological perspectives through precise assessment of various protein targets.
The authors have nothing to disclose.
The authors acknowledge the support of the Laboratory for Clinical Genomics and Advanced Technology in the Department of Pathology and Laboratory Medicine of the Dartmouth Hitchcock Health System. The authors also acknowledge the Pathology Shared Resource at the Dartmouth Cancer Center with NCI Cancer Center Support Grant 5P30 CA023108-37.
BOND Research Detection System | Leica Biosystems, Wetzlar, Germany | DS9455 | Open detection system containing open containers in a reagent tray |
BOND Wash | Leica Biosystems, Wetzlar, Germany | AR950 | 10X concentrated buffer solution for washing fixed tissue |
Buffer W | NanoString, Seattle, WA | contact company | Blocking reagent |
Cy3 conjugation kit | Abcam, Cambridge, UK | AB188287 | Cy3 fluorescent antibody conjugation kit |
GeoMx Digital Spatial Profiler (DSP) | NanoString, Seattle, WA | contact company | System for imaging and characterizing protein and RNA targets |
GeoMx DSP Instrument BufferKit | NanoString, Seattle, WA | 100471 | Buffer kit for GeoMX DSP (including buffers for sample processing and preparation) |
GeoMx Hyb Code Pack_Protein | NanoString, Seattle, WA | 121300401 | Controls for running GeoMX DSP experiemtns |
GeoMx Immune Cell Panel (Imm Cell Pro_Hs) | NanoString, Seattle, WA | 121300101 | Protein module with targets for human immune cells and immuno-oncologic targets |
GeoMx Pan-Tumor Panel (Pan-Tumor_Hs) | NanoString, Seattle, WA | 121300105 | Protein module with targets for multiple human tumor types and for markers of epithelial-mesenchymal transition |
GeoMx Protein Slide Prep FFPE | NanoString, Seattle, WA | 121300308 | Sample preparation reagents for GeoMX DSP protein analysis |
IDH1-R132H antibody | Dianova, Hamburg, Germany | DIA-H09 | Monoclonal antibody against human IDH1 R132H |
LEICA Bond RX | Leica Biosystems, Wetzlar, Germany | contact company | Fully automated IHC stainer |
Master Kit–12 reactions | NanoString, Seattle, WA | 100052 | Materials and reagents for use with the nCounter Analysis system |
nCounter Analysis System | NanoString, Seattle, WA | contact company | Automated system for multiplex target expression quantification (to be used with GeoMx DSP) |
TMA Master II | 3DHistech Ltd., Budapest, Hungary | To create the tissue microarray block |