Introduction

Melanoma is an immunogenic malignancy with an elevated tumor mutational burden (TMB) and a highly active immune tumor microenvironment (TME) [1]. While historically associated with poor clinical outcomes, late-stage metastatic melanoma patients have seen a significant improvement in prognosis with the advent and clinical implementation of immune checkpoint inhibitors (ICIs) [2,3,4]. The favorable response to these therapies has been linked to a number of factors, such as abundance of neoantigens and the degree of infiltrating lymphocytes within the melanoma TME [5,6,7]. In recent years, spatial imaging technologies have been increasingly used to study the immune infiltration of melanomas, with the goal of understanding the complex interplay between stromal and tumor cells, as well as distinct immune and tumor-infiltrating lymphocyte subsets (TILs), and their effect on ICI response.

Improved approaches to characterize the melanoma TME carries a degree of urgency with clear clinical significance as more than 50% of late-stage metastatic melanoma patients treated with ICIs will either not respond or develop progressive disease [8]. With the recent approval of anti-LAG-3 ICIs for the treatment of metastatic melanoma, a better understanding of the role of the melanoma TME and mechanisms mediating patient response to anti-PD-1 monotherapy compared to combination therapy with anti-CTLA-4 or LAG-3, is clearly needed [9]. Furthermore, results from recent melanoma neoadjuvant clinical trials, SWOG S1801 (NCT03698019) and NADINA (NCT04949113), are poised to redefine management of high-risk resectable stage III melanomas [10]. Molecular profiling studies of pre- and post-treated melanomas in the neoadjuvant setting provides an efficient platform to understand mechanisms of therapy response with measured pathological response available in a relatively short period of time. Such melanoma study designs provide a critical opportunity to identify biomarkers to personalize adjuvant therapy and improve patient outcomes [11]. In this review, we will provide a brief survey of current multiplexed imaging technologies and highlight recent studies that have employed such approaches to characterize the spatial immune landscape of melanomas to better understand ICI responses.

Multiplexed imaging technologies

Limitations in conventional multiplexed tissue staining and antibody-based imaging, such as hematoxylin and eosin (H&E), immunohistochemistry (IHC) and immunofluorescence (IF) approaches has previously limited the scope and scale of melanoma TME research. However, studies have shown the improved predictive power of biomarkers of ICI response when incorporating spatial TME information, such as cell-type specific expression in specific regions of melanomas [5, 12]. Such studies have reinforced the need to employ more advanced highly multiplexed approaches for translational research.

To address these limitations, various multiplexed spatial imaging methods have been developed to study the melanoma TME in search of new clinically predictive biomarkers over the past decade (Table 1). Multiplexed IF and cyclical multiplexed immunofluorescence techniques such as OPAL-7 [13], tyramide signal amplification (TSA) [14], t-CyCIF (Tissue-based Cyclic Immunofluorescence) [15], or MILAN (Multiple Iterative Labeling by Antibody Neodeposition) [16] can image from 6 to 60 markers using repeated cycles of antibody deposition and imaging. Among these, the more highly-multiplexed technologies have enabled researchers to extract more cell-type specific spatial information, such as t-CyCIF, which makes use of repeated cycles of low-plex fluorescence staining and imaging for up to 60 markers [15]. T-CyCIF has recently been applied to the melanoma TME to evaluate sequential melanoma biopsies obtained from the same patient, revealing co-evolution of differences in immune composition among different clonal lineages, while another study on metastatic melanoma lymph nodes used a 45-plexed t-CyCIF antibody panel to classify prognostically relevant immune-cell neighborhoods [17, 18]. T-CyCIF has also been combined with other multiplexed technologies including PickSeq and NanoString to spatially characterize melanoma cellular neighborhoods, their transition from precursor states to melanoma in situ and invasive tumor, and the immune-suppressive environment along the tumor-stromal boundary [19].

Table 1 Imaging methods used to characterize the melanoma spatial microenvironment and example studies
Full size table

Recent advancements in lower-cost spatial immune profiling technology include bright-field multiplexed imaging, which enables imaging of triple-plexed bright-field colors for clinical and histopathological diagnostic assays using standard pathology-lab equipment [22]. Another advancement is nine-color multiplexed immunofluorescence imaging that has been applied to characterize cell phenotype diversity and immunosuppression patterns within the TME of malignant pleural mesothelioma [33].

In contrast, mass-spectroscopy-based imaging and analysis methods, such as MALDI (Matrix Assisted Laser Desorption/Ionization), DESI (Desorption Electrospray Ionization), SIMS (Secondary Ion Mass Spectrometry), and MIBI-TOF (Multiplexed Ion Beam Imaging by Time-of-Flight MS), can detect up to thousands of analytes including metabolites on a sample without using repetitive imaging cycles [29, 34, 35]. For example, MIBI-TOF, which can image up to 36-plex markers (or in theory, up to 100 markers) and uses a tunable ion-beam that can be adjusted to various sample depths and image resolutions, has been applied to image cellular phenotypes and tissue structure, revealing regional variability in tumor cell phenotypes in several different breast cancer subtypes including triple-negative breast cancer [35, 36]. CyTOF-IMC, another recent addition to the spatial imaging toolkit, can detect up to 40 proteins on a single sample of formalin-fixed paraffin-embedded (FFPE) tissue and has recently been applied to characterize the melanoma TME [23,24,25,26].

Exponential development in the spatial transcriptomics imaging field has enabled researchers to spatially image transcripts from up to 20,000 different protein-coding genes at mid to high-resolution, using technologies including NanoString [37], Visium [38], and MER-FISH [39]. Generally, spatial transcriptomics technologies fall into two categories. First, those that spatially segment tissue into a limited number of relevant cell types and interrogate large numbers of transcripts across those selected groups of cells, albeit without achieving single-cell resolution, such as NanoString [37, 40]. Second, those that provide single-cell, or close-to single-cell resolution for a more limited number of transcripts, such as MER-FISH [40]. In the intermediary zone are technologies that include Visium, which allow for lower-resolution whole transcriptome sequencing of ‘patches’ of 5–50 cells [38, 41]. These technologies have provided detailed insights into the TME in the context of ICI treatment for melanoma liver metastases. A comprehensive analysis of cutaneous and uveal melanoma liver metastases identified differences in the TME between these melanoma subtypes, such as PD-L1 expression and the ratio of exhausted CD8 T cells to several other T cell subsets [30]. High-plex spatial RNA profiling has been instrumental in revealing cell type-specific biomarker expression in the TME during early melanoma evolution, including prominent expression of S100A8 and S100A9 in melanoma-adjacent keratinocytes [42]. Furthermore, these technologies have been used to examine the immune landscape of tumor-associated lymph nodes in melanoma, leading to the identification of potential biomarkers of CD11c activation that could be clinically relevant for predicting survival outcomes [37].

Spatial transcriptomics platforms, including NanoString and Visium, have greatly advanced our understanding of gene transcription by facilitating the detailed analysis of mRNA levels at a high spatial resolution [37,38,39,40,41]. Of note, transcriptomic-based technologies do possess inherent limitations as some isoforms may not be accurately differentiated by these platforms and mRNA levels are not always congruent with protein expression. Thus, protein-based methodologies cannot be wholly substituted by spatial transcriptomics alone. Integration of spatial proteomic and transcriptomic profiling allowed by currently available platforms has already shown promise in offering a more holistic view of the cellular landscape [23, 24]. However, the field still faces significant challenges in merging these data streams, largely due to technological limitations in comprehensively measuring both the transcriptome and proteome simultaneously.

CyTOF studies in cutaneous melanoma

Flow cytometry has long been a cornerstone in the field of cellular analytics, providing a means to measure, quantify and analyze characteristics of single cells in suspension in a multiplexed workflow. The Helios CyTOF platform introduced the use of rare earth metal isotopes for antibody labeling and integrated mass spectroscopy into flow cytometry as an analysis technique, thus allowing for separation of cells based on their mass-to-charge ratio and eliminating issues with spectral overlap from traditional flow cytometry platforms [43, 44]. This significantly expanded the capabilities of flow cytometry technology and allowed for the simultaneous analysis of up to 60 parameters using metal-tagged antibodies, albeit without preserving spatial orientation of cells within tissues [44,45,46]. The technology used to develop the Helios CyTOF platform paved the way for the development of the Hyperion CyTOF-IMC platform that can multiplex image up to 40 markers using rare earth metal isotope-conjugated antibodies [47].

CyTOF-IMC can detect antibodies at sub-cellular resolution simultaneously on the same tissue slide using high-frequency laser ablation of antibody-tagged metal isotopes in a low-dispersion laser ablation chamber, for either FFPE or fresh tissue [47]. In brief, tissues are labeled with metal-tagged antibodies using standard immunohistochemistry methods, air dried, laser-ablated pixel by pixel, and transported by a mixed argon and helium stream to the CyTOF mass cytometer for measurement (Fig. 1) [47]. Cellular images generated by CyTOF-IMC undergo a two-step analytical process. Images are first segmented into individual cells using a variety of software. Next, cells are phenotyped using methods based either on unsupervised clustering (e.g. Phenograph) [48], FlowSOM [49], or probabilistic classification models (e.g. Astir) [50] (Table 2 provides a non-exhaustive overview of some of the more commonly-used tools available for these tasks). Advanced spatial analysis, including the calculation of median intercellular distances, and contact enrichment, can be performed using tools such as imcRtools [51] (see [52] for a more comprehensive review of analytical pipelines).

Table 2 Summary of different segmentation and classification tools for CyTOF-IMC data
Full size table
Fig. 1
figure 1

Illustration of the data acquisition workflow used for CyTOF-IMC: (1) One ~ 5 μm-thick section of a tissue microarray (TMA), usually composed of 1 mm2 melanoma cores (2) is stained with a cocktail of antibodies labeled with metal isotopes (colored asterisks). (3) Samples are ablated with a high energy laser in a rastered pattern, and the resulting ionized isotope plumes are analyzed by a mass cytometer, which returns the number and type of metal isotopes per pixel (one pixel = 1micron x 1micron). (4) Each antibody results in a single image per sample, and together are layered to construct a multi-image stack. Since 2014, CyTOF-IMC has been utilized to spatially analyze a wide array of tissues and tumor-types, including breast, lung, and melanoma, as well as to identify biomarkers of response to treatment in breast cancer (including HR and HER2) or immunotherapy (anti PD-1 or anti-CTLA4) in melanoma [24, 26, 47, 60, 61]

Full size image

Two of the earliest publications to use CyTOF to characterize melanoma patient samples utilized the Helios platform to characterize peripheral blood mononuclear cells (PBMCs) or immune profiles of tumor biopsies from patients treated with anti-PD-1 monotherapy, or combined anti-PD-1 and anti-CTLA-4 [62, 63]. One early study conducted on stage IV melanoma patients receiving anti-PD-1 immunotherapy found that the frequency of CD14 + CD16 − HLA-DRhi monocytes in PBMCs prior to therapy initiation was a strong indicator of progression-free and overall survival, suggesting their potential use in informing treatment decisions [62]. Another study identified activated T-cell signatures and T-cell populations in responders to both treatment modalities (anti-PD-1 and anti-CTLA-4) using a panel of multiplexed antibodies to characterize immune cell populations [63]. Further analysis revealed an EOMES + CD69 + CD45RO + effector memory T-cell phenotype that was significantly more abundant in responders to combined immunotherapy than non-responders. The gene expression profile of this population was associated with longer progression-free survival in patients treated with single-agent therapy and showed greater tumor shrinkage in both treatments, revealing insights into response and resistance mechanisms to ICIs [63]. These early CyTOF flow cytometry studies provided valuable information regarding abundance of immune cell populations present in the melanoma TME. However, the advent of the Hyperion IMC has facilitated the multiplexed characterization of spatial relationships of cells and biomarkers within the melanoma TME using the CyTOF platform.

Four recent studies have utilized CyTOF-IMC to directly characterize the melanoma TME (Table 3). Published in 2021, CyTOF-IMC was employed to analyze the TME of patients with metastatic melanoma who received ICI, and to identify indicative factors of treatment response [25]. Rather than segmenting cells, the authors used a newly designed version of the AQUA software to measure marker intensity in molecularly defined compartments. Multivariable analyses revealed significant associations of 12 markers with progression-free survival, and 7 markers with overall survival, which included b2-microglobulin [25].

Table 3 Studies using CyTOF-IMC to characterize the melanoma TME, including a brief summary and sample sizes
Full size table

In our study, published in 2022, we utilized CyTOF-IMC to profile more than 230,000 individual cells from 5 benign nevi and 67 melanomas and identified melanoma, lymphocyte, macrophage/monocyte, and stromal cell populations, allowing for in-depth spatial quantification of the melanoma microenvironment [26]. While prior studies have shown that the abundance of TILs in the TME is associated with better prognosis and response to ICIs [64,65,66], we demonstrated that within the pre-treatment melanoma TME, it was the abundance of proliferating antigen-experienced cytotoxic T-cells (CD8 + CD45RO + Ki67+) and their proximity to melanoma cells that best informed on ICI responses in our analyzed cohort [26]. This indicated that a shorter distance between melanoma and nearby antigen-experienced cytotoxic T-cells within the TME was linked to a favourable response, highlighting the potential of CyTOF-IMC to quantify spatial cell-cell interactions for ICI biomarker studies.

Two other studies were published in 2022 that utilized CyTOF-IMC but combined this approach with other modalities to investigate the immune environment of melanoma. Specifically, researchers investigated the role of regulatory and pro-inflammatory cytokine-expressing B cells in patients with melanoma, utilizing flow cytometry, CyTOF-IMC, single-cell RNAseq, immunofluorescence staining and transcriptomic analysis [23]. The group found that patients had enhanced circulating regulatory B cell populations and reduced pro-inflammatory B cell populations compared to healthy volunteers [23]. They also found that cytokine-expressing B cells in the melanoma TME assembled in clusters and interacted with T-cells and T-regulatory cells via various signaling pathways [23]. Patient-derived B cells were found to promote T-regulatory cell differentiation and T-helper cell proliferation in a TGF-β-dependent manner, an effect further enhanced with anti-PD-1 checkpoint blockade [23]. These findings highlight the bidirectional crosstalk between B and T cell subsets with immunosuppressive attributes in the context of melanoma. In another CyTOF IMC study, researchers reported the use of multiplexed mass cytometry–based imaging of protein markers and RNA transcripts to investigate the chemokine landscape and immune infiltration in metastatic melanoma samples [24]. The study showed that tumors lacking immune infiltration had low levels of antigen presentation and markers of inflammation, and were devoid of most of the profiled chemokines [24]. In contrast, infiltrated tumors expressed high cytokine levels, with CXCL9 and CXCL10 being localized in patches associated with dysfunctional T-cells expressing the B lymphocyte chemoattractant CXCL13 [24]. This study also found that T-cells play a role in B cell recruitment and potentially in B cell follicle formation, and that the formation of tertiary lymphoid structures may be accompanied by naïve and naïve-like T-cell recruitment, which can contribute to antitumor activity. These two CyTOF-IMC articles demonstrate the potential of combining multiple spatial ‘omic’ technologies to make important TME discoveries.

Future studies combining multiple spatial ‘omic’ technologies (e.g. CyTOF-IMC, NanoString, Visium, ATAC-Seq, microbiome analysis, and plasma metabolite analysis) will undoubtedly lead to a more deeper understanding of TME biology. One challenge with integrating multiple profiling approaches is that often these technologies require different biological material inputs (i.e. plasma, tissue sections, or tissue lysate). Already, studies have expanded IMC technologies to detect mRNA and proteins with single-cell and spatial resolution in melanoma and breast tumours, capturing three mRNAs along with a panel of proteins to reveal correlations between mRNA and proteins at both the single-cell and cell population levels [24, 67]. Furthermore, we recently interrogated microbiome composition, plasma metabolite makeup, and employed CyTOF-IMC to analyze the immune landscape of melanoma tumors prior to fecal microbiota transplantation (FMT) in order to more comprehensively characterize the immune infiltrates in responsive and non-responsive patients to combination FMT and ICI therapy [68].

Limitations

Spatial imaging proteomic and transcriptomics technologies, while powerful, are newer techniques that often grapple with limitations and compromises. These include balancing single-cell resolution with high multiplexing capabilities, limitations in availability of required reagents such as high-quality antibodies and probes, as well as challenges associated with high costs and slow data acquisition times coupled with the complexity of data analysis. For example, analyzing and interpreting the large and complex datasets produced by CyTOF-IMC requires sophisticated tools and computational expertise. Recent advances in machine learning have allowed for improved cell segmentation and classification in the analysis and interpretation of large and complex datasets produced by CyTOF-IMC. For example, a recent study utilized CyTOF-IMC and deep learning to analyze the TME of lung adenocarcinoma samples from 416 patients, profiling 1.6 million cells using a combination of classical and modern machine learning based computer vision algorithms in order to reveal distinct immune lineages and activation states with clinical correlates, and demonstrating the potential of artificial intelligence in predicting patient progression using a single 1-mm2 tumor core [57, 61]. Finally, as the technology for spatial imaging and transcriptomics continue to advance, analytical tools, imaging analysis methods and data processing pipelines will need to keep pace. These rapid advances provide both great opportunities and present unique challenges to researchers, as consolidation of best-practices for application and data analysis are continually evolving to meet the needs of the ever-advancing armamentarium of spatial imaging technologies.

Conclusion

The future of single-cell spatial profiling will be driven by the development of novel technologies that can overcome current limitations and provide more nuanced insights into single-cell dynamics. These advancements will enable multiplexed, cell-type specific, and tumor location-specific analysis of biomarkers, which could significantly enhance their predictive power. Furthermore, the integration of 3D spatial imaging using technologies such as CyTOF-IMC may provide a more comprehensive view of the cellular landscape. Ultimately, the integration of 3D biomarker imaging across multi-omic platforms would significantly improve our understanding of the TME [69].

Within the melanoma TME research field, these single cell imaging advancements are poised to have an immediate and profound impact. With the recent approval of anti-LAG-3 for the treatment of late-stage metastatic melanoma, identification of biomarkers that can inform on prescription of monotherapy anti-PD-1, versus combinations that include anti-CTLA-4 and LAG-3, are urgently needed. Moreover, single cell technologies amenable to predicting response in the neoadjuvant setting could significantly advance personalized medicine by tailoring adjuvant treatments to individual patient profiles [11].

As our review has underscored, recent studies utilizing single cell imaging approaches are just beginning to illuminate the cellular composition and functional dynamics of the melanoma TME. This area of investigation has the potential to uncover novel biomarkers for diagnosis and prognosis, particularly in the context of ICI response, and to identify new targets for therapeutic intervention.

Search methodology

In this review, a search of peer-reviewed scientific articles and literature reviews was completed using PubMed with the following search terms: (“Melanoma”) AND (“Tumor micro-environment” OR “Tumour micro-environment” OR “Tumor microenvironment” OR “Tumour microenvironment” OR “Cellular Microenvironment” OR “Cellular Micro-environment” OR “Tumour immune microenvironment” OR “Tumor immune microenvironment” OR “Tumour immune micro-environment” OR “Tumor immune micro-environment”) AND (“spatial landscape” OR “spatial”), within the date range of 2014–2023 (the earliest CyTOF-IMC paper being released in 2014), retrieving 82 results. After full-text review, we included 3 papers on melanoma spatial imaging using CyTOF-IMC, and captured an additional 6 papers on other melanoma spatial imaging technologies. We also applied a snowballing search methodology using the references cited in the articles identified in the literature search, locating an additional 1 pertinent paper (see Table 3). For the snowballing search, articles were limited to those including CyTOF-IMC or other relevant spatial imaging technologies, melanoma or another relevant cancer, or other pertinent articles. Each identified item was assessed for relevance by a member of the study team and final calls were made by the senior author. This review is intended to summarize high-impact, relevant and recent literature on spatial imaging technologies, specifically CyTOF-IMC, for the tumor microenvironment and immune landscape in melanoma, and is not meant to be exhaustive.