IHC uses a primary antibody to target proteins of interest. Target proteins usually reflect the type and function of cell populations. A secondary antibody conjugated to an enzyme such as horseradish peroxidase is then applied before applying a chromogen substrate of the enzyme that will develop a colour upon enzymatic activity. This method enables amplification of the signal for the detection of the target protein with light microscopy (28). Additionally, IHC can be performed effectively on small tissue samples such as those obtained via EBUS (29).

Immunofluorescence (IF) uses a similar principle to chromagen-based IHC and can also be used to assess the TME. Here, instead of an antibody bound to an enzyme, antibodies are conjugated to fluorophores that can be detected by fluorescence microscopy (30). IF can either be performed directly where the primary antibody is attached to a fluorophore or indirectly where the fluorophore is attached to a secondary antibody which recognises the primary antibody of interest. Indirect IF allows for signal amplification (30).

While routinely used in clinical practice, IHC only allows for the assessment of a one or two proteins, and IF up to four, which limits comprehensive TME analysis. Multiplex immunohistochemistry (mIHC) has emerged as a tool for simultaneous detection of multiple biomarkers. However, there are several challenges to doing this accurately in a time-efficient manner. For example, performing co-staining with standard chromogen-based IHC is possible, however staining strategies are needed to overcome antibody cross-reactivity. Furthermore, due to overlap of chromogenic spectra, only up to 3 antibodies can be assessed this way. High-dimension multiplex IHC offers an alternative where antibodies and chromagen are stripped after each cycle of staining (31). This method enables staining of up to 15 proteins but is labour-intensive and cycles of staining/antibody stripping can cause tissue degradation (32). In addition, there is no robust analytical method to quantify cell populations with this approach (33). Similarly, while multiplex IF can assess up to 4 antibodies simultaneously, there are significant limitations. The most important considerations are spectral overlap of fluorophores and autofluorescence of the tissue (33).

Several methods such as tyramide signal amplification (OPAL™ mIHC), DNA barcode technology (CODEX ™, InSituPlex ™), image cyclic staining technology (MACSima ™), imaging mass cytometry (IMC) and Multiplex Ion Beam Imaging (MIBI) have been developed to provide more thorough assessment of the TME (33–38). DNA barcoded technologies and image cyclic staining technologies both have a paucity of published data (36). As a result, this paper will focus on OPAL™ mIHC, IMC, and MIBI. MIBI and IMC can be classified under mass cytometry-based imaging techniques as both use metal isotopes bound to antibodies. OPAL mIHC allows for simultaneous assessment of 4-9 antibodies (33, 39), while next generation IHC methods such as IMC and MIBI can assess up to 40 antibodies simultaneously. These techniques allow for assessment of both immune cell populations, their function and spatial arrangement in the TME (37, 38).

OPAL mIHC uses a tyramide signal amplification (TSA) system. This method uses a primary antibody to target the protein of interest and a secondary antibody that is conjugated to horse radish peroxidase. A tyramide-conjugated fluorophore is catalysed by horseradish peroxidase and will covalently bind to tyrosine residue in secondary antibody and tissue with the protein of interest. The primary and secondary antibodies can then be stripped, and the process repeated several times for different proteins of interest (33). OPAL mIHC allows for simultaneous assessment of up to 9 proteins in formalin-fixed and paraffin-embedded (FFPE) tissue samples (39).

Once stained slides are scanned, digital images are analysed using various software packages to obtain both spatial and quantitative information of the TME. Several software packages are available for analysis including commercial software such as HALO (Indica Labs, Albuquerque, New Mexico, USA) (40) and INFORM (Akoya Bioscience, Menlo Park, California, USA) (33), and open-source software such as QuPATH (41). Whole slide analysis is possible with mIHC. Analysis involves cell segmentation and annotation, cell phenotyping and spatial analysis. Development of machine learning models has allowed for the automation of some image analysis (42). Machine learning methods have also been developed to quantify immune cells in TME (43).

Like mIHC, mass cytometry-based imaging techniques have also been used to define the TME. These techniques utilise target antibodies attached to metal isotopes to assess multiple target proteins simultaneously. Two such techniques are MIBI and IMC.

MIBI uses secondary mass spectrometry to image primary antibodies that are coupled with rare metal isotopes. Firstly, FFPE sections are stained with rare earth metals conjugated primary antibodies. These are then loaded on to the MIBIscope and rasterized with a xenon duoplasmatron primary ion beam. This results in the release of secondary ions which are subsequently analysed with a magnetic sector mass spectrometer with time-of-flight (TOF) analysis. Multiple isotopes can be assessed simultaneously with most devices analysing up to 40 isotopes at a time. The resultant data provides a two-dimensional map of the distribution of each isotope and its corresponding antibody and its related epitope (37). Analysis of MIBI images can then be performed via computational pipelines that allow for automation of most steps (42, 50).

Like MIBI, IMC utilises rare earth metal labelled antibodies to stain FFPE specimens. Areas of interest within these specimens are then ablated by a laser, and the resulting tissue plumes are assessed by time-of-flight cytometry to determine isotope abundance (38, 51). IMC can analyse a similar number of antibodies as MIBI (52), however repeat analysis is not possible on the same section of tissue due to complete ablation (53). Mass cytometry-based imaging has considerable advantages over mIHC. Autofluorescence and spectral overlap are not a concern with mass-cytometry imaging and staining of the section in one single step ensures tissue integrity compared to mIHC where cyclic staining can lead to tissue degradation (37).

Flow cytometry uses fluorophore-conjugated antibodies to target molecules of interest. A suspension of cells bound to antibodies is exposed to a laser, one cell at a time. This causes fluorophore excitation which leads to emission of light at a particular wavelength that can be detected by a sensor. Multiple antibodies can be assessed simultaneously (up to 24 with some devices that offer spectral unmixing) (60). Flow cytometry allows for assessment of thousands of events per second (60). It can assess large numbers of cells quickly and provide extensive data with regards to cell populations and their function (60). The main limitations of flow cytometry are the limited number antibodies that can be assessed at one time and the overlap in emission spectra of some fluorophores meaning that antibody combinations need to be selected carefully (60).

Mass cytometry uses antibodies conjugated to heavy metal isotopes to target the molecules of interest. Once the cell suspension has been labelled with these antibodies, it is introduced into a mass spectrometer where an argon plasma creates an ionised cloud of atoms that is enriched for heavy metals. The heavy metal ions are then separated by time-of-flight mass spectrometer by their mass to charge ratio. This is then detected as an electrical signal at the terminal gate, the ion count reflecting the expression of the corresponding target molecule. Mass cytometry allows for the simultaneous assessment of up to 40 antibodies (59). Mass cytometry has been validated against flow cytometry for assessment of both peripheral blood mononuclear cells (PBMC) and tumour tissue in human cancer studies (69). Mass cytometry provided similar analysis to flow cytometry with respect to cell populations, function, activation, and exhaustion. Fewer cell numbers are also needed for analysis when compared to traditional flow cytometry (69).

RNA-seq techniques can vary depending on the platform that is used; however, most platforms involve RNA extracted from cells of interest, followed by RNA fragmentation. Improvements in library preparation techniques has allowed for improved assessment of degraded tissue, including FFPE blocks. Reverse transcription is then performed on RNA fragments to obtain complementary DNA (cDNA) which then undergoes PCR amplification to create sequencing libraries. A DNA polymerase is then used to sequence the libraries (82).

There are many pipelines available for computational analysis once sequencing is completed. While the specific type of analysis will depend on the purpose of the experiment, most analysis will follow the following broad steps. Quality control is first performed to determine the quality of sequencing and eliminates low quality reads. This is followed by read alignment where clean RNA reads are mapped to a human reference transcriptome. Once aligned, the short reads are assembled into transcripts. The expression of the transcript can then be quantified. This will routinely generate an expression matrix that can then undergo further statistical analysis (82).

Several bioinformatic analysis techniques have been developed to determine TME composition and include deconvolution techniques (83), gene set enrichment analysis (84) and other techniques that can quantify immune populations such as xCell (85). Deconvolution methods such as CIBERSORT (83) estimate cell populations based on overall signature (83) while gene set enrichment analysis determines differences in defined gene sets and can assess different gene signatures for immune cells (84). Whereas xCell uses a gene signature to map to 64 corresponding stromal and immune cells (85). These analysis techniques can be applied to large data sets, including publicly available data sets such as The Cancer Genome Atlas (TCGA) data base or datasets deposited on the Gene Expression Omnibus (GEO).

In addition to bulk RNA sequencing, single-cell RNA sequencing can profile the transcriptome at a single cell resolution (86). Analysis of different gene expression between individual cells can potentially identify rare populations that would not be detected with pooled analysis. This can be particularly useful for determining heterogeneity in immune cell populations in the TME (82).

Single-cell RNA sequencing first requires single cell isolation, which can be done by different methods including, limiting dilution (87, 88), micromanipulation (87, 88), flow activated cell sorting (89), laser capture microdissection (90), and microfluidic technology (91). Once single cell isolation has been performed, cells are lysed, and RNA then undergoes reverse transcription to cDNA. This is then followed by amplification and sequencing. While there are pros and cons of each technique, this will not be addressed here. Overall, when compared to bulk RNA sequencing more cells are required due to the loss of cells at time of isolation, lysis and mRNA capture (86, 92).

Analysis pipelines for single-cell RNA sequencing differs from bulk RNA sequencing as there is increased noise as well as technical and biological variability. Once quality control and normalisation have been implemented, analysis can be performed at a cellular level to cluster and classify cell types. Gene level analysis can then determine differences in differentially expressed genes between clusters to infer differences in function between cell clusters (86, 92). With both bulk RNA sequencing and single cell RNA sequencing, spatial information that would be available with IHC methods is absent. Spatial transcriptomics, which was first described in 2016 can be used to overcome this limitation (81).

Spatial transcriptomics uses positional barcodes to map cDNA to particular areas of tissue, which will then demonstrate variations in gene expression with respect to spatial orientation (81). To our knowledge spatial transcriptomic methods have not been performed on small biopsy samples.