1 INTRODUCTION

Recent progress in spectral flow cytometry and single-cell mRNA sequencing revolutionized the way immune cells are characterized and allowed the discovery of novel (sub)populations, as well as the decoding of leukocyte differentiation pathways. Moreover, molecular biology techniques such as CRISPR-Cas9-mediated gene editing provided elegant ways to analyze the function of individual genes in defined cell types. However, as powerful as these techniques are, comprehensive insight into the function of any cell type is impossible without analyzing their three-dimensional environment by bioimaging. Only by bioimaging the position of leukocytes within organs or tissues can be seen, their migration patterns within or between organs can be tracked, encounters with other cells can be monitored, and even subcellular and molecular mechanisms of these interactions can be observed. In the last 20 years, we witnessed enormous advances in sample preparation protocols, the development of novel fluorescent probes, improvements in illumination devices, microscope optics, and detectors. These improvements were paralleled by an enormous progress in electronics, affordable computing power, and data analysis software. Today, scientists can choose among many different methodologies, each providing a tool to observe different aspects of leukocyte functions. While non-invasive imaging techniques enable in vivo tracking of leukocytes between different organs at the level of the whole organism,1 intravital microscopy (IVM) provides the resolution required to tracking individual cells in small regions of intact organs.2 On the other hand, entire organs from mice can be visualized 3-dimensionally (3D) using light-sheet microscopy, albeit at the cost of tissue fixation required to make the tissues transparent and thus permeable to light.3 Due to its simplicity, confocal microscopy remains the standard approach for qualitative and quantitative assessment of cell positioning and migration within tissue slices or position of organelles within the cell.4 The vast majority of confocal systems allow simultaneous detection of approximately 5 parameters, which does not allow distinction of specialized leukocyte subpopulations or simultaneous detection of several different cell types. This hurdle has recently been overcome by mass cytometry imaging. While its resolution levels are only about one third to one fourth as good as for confocal imaging, mass cytometry imaging can resolve up to 40 different parameters simultaneously on tissue sections stained by isotope-marked antibodies applying mass spectrometry and computer-assisted picture-assembly.5 Finally, several different super-resolution microscopy techniques surpassed the diffraction limit of light microscopes—a minimal distance beyond which two objects cannot be separated—and enabled imaging even of individual proteins and their interactions.6, 7

These diverse bioimaging methods have enormously contributed to our understanding of mechanisms by which dendritic cells (DCs) populate and survey peripheral tissues, recognize pathogens, mature and migrate into secondary lymphoid organs, where they activate T cells inducing protective or tolerogenic immune responses. Understanding DC biology is indispensable for the development of vaccines as well as for understanding why immune tolerance sometimes gets broken, leading to allergies, autoimmunity, or dampened immune responses against tumors.

Hence, this review aims to provide an overview of divergent bioimaging techniques that allow insight into different aspects of DC functions. At the same time, we hope to use this methodological overview as framework that will provide the current status of DC immunology and highlight the open questions that bioimaging can answer.

2 DC SUBSETS

In 1973, Ralph Steinman and Zanvil Cohn used then state-of-the-art phase-contrast microscopy to describe a novel cell type distinct from monocytes and macrophages and named it DC due to its long dendritic-shaped processes.8 However, the years following their discovery DC research was plagued with several issues. Some of those, such as the paucity of DC numbers and difficulty to dislodge them from other cells, required careful standardization of cell isolation protocols and the development of appropriate staining protocols for their identification by flow cytometry and histology. These approaches, however, revealed that DCs are not a uniform cell population but rather a group of cells with different origins and functions,9-11 further complicating their distinction from more abundant macrophages and monocytes. At the beginning of the millennium, DCs were separated from monocytes and macrophages based on their morphology, function, and/or a limited set of surface markers. Overall, cells with dendrites that expressed major histocompatibility complex (MHC) class II (MHC-II) and cluster of differentiation (CD)11c on the cell surface and/or migrated to the lymph nodes to present antigens to T lymphocytes were classified as DCs, while large phagocytic cells filled with vacuoles that were mainly expressing F4/80 were considered to be macrophages.12 However, none of these classification systems proved to be optimal and universal. While in the spleen CD11c appears to be exclusively expressed on DCs, it is not suited for DC separation from CD11c-expressing macrophages in the other organs, like the lungs,13 kidneys,14 or intestine.15 On the other hand, CD64, a selective macrophage marker in most of the tissues is also expressed on DCs in certain organs such as kidneys,16 or in inflamed lungs.17 Further, it was soon discovered that certain DC populations do not migrate,18, 19 as well as that macrophages have, beside phagocytic, also immunomodulatory and metabolic functions. The sole distinction on the basis of morphology is also not always suited, as intestinal macrophages possess transepithelial dendrites.20

At the same time, the development and standardization of transcriptomic techniques led to the formation of ImmGen, a consortium that summarized gene expression of mouse hematopoietic cells and their developmental precursors in a comprehensive compendium (www.immgen.org).21 This opened a possibility to classify leukocytes independent of other parameters and clearly indicated molecular signatures that distinguished DCs from monocytes and macrophages.22, 23 These studies were later validated by findings that steady-state macrophages are of yolk sac origin and are maintained by self-renewal,24-26 prompting re-classification of epidermal Langerhans cells, originally considered being DCs, to macrophages.27, 28 In contrast, monocytes and the majority of DCs originate in bone marrow from the monocyte-committed, common monocyte progenitors (cMop) and common DC progenitors (CDPs), respectively.16, 29-31 In 2014, Martin Guilliams and colleagues proposed to classify DCs, monocytes, and macrophages based on their cellular origin at the first level, which is then refined by their location, morphology, and/or function as a second level.32 Further transcriptome alignments, refined with single-cell RNA sequencing and combined with knock-out experiments revealed that CDPs differentiate into two main DC subpopulations: conventional (or classical) DCs (cDCs) and plasmacytoid DCs (pDCs) (reviewed in Refs. [33-37]). Monocyte-derived DCs (moDCs) represent the third group of DC, which share many functions with DC but arise from monocytes under certain (inflammatory) conditions.38-40 For the purpose of this review, we will briefly characterize those three DC subtypes (Figure 1 and Table 1), while for more detailed insight we point the reader to other recent reviews.34-37

Major mouse DC subsets (modified from Ref. [151]). Transcriptome analysis revealed that common DC progenitors (CDPs) differentiate into conventional (or classical) DCs (cDCs) and plasmacytoid DCs (pDCs).33-37 In the bone marrow, growth factor FMS-like tyrosine kinase 3 ligand (Flt3l) induces CDP differentiation into pre-cDCs that are released into the blood to seed peripheral and lymphoid organs to differentiate into cDCs.44-47 Pre-DCs stage appears to be already pre-committed for development into one of the two cDC subtypes: type 1 (cDC1) or type 2 (cDC2) cells.45, 48-50 Differentiation of cDC1s depends on transcription factors Irf8, Batf3, and Id2,12, 37 while activation of Relb, Rbpj, and Irf4 lead to cDC2 differentiation.63-66 Recent single-cell RNA sequencing experiments have indicated that cDC2 can be further split into at least 2 subgroups, cDC2A and cDC2B.71-73 pDCs are, on the other hand, continuously produced from CDPs bone marrow under control of transcription factors Irf8 and TCF4 (E2-2),82, 83 although some can be derived from FLT3+IL-7R(CD127)+CD117lo/int lymphoid progenitor cells (not depicted).87 Recently, “transitional” DCs (tDCs) that share phenotype and characteristics of pDCs and cDC2s have been described, albeit their ontogeny is still unclear.80, 97 Finally, under inflammatory conditions certain monocyte subpopulations can also differentiate into monocyte-derived DCs (moDCs).38-40 Indicated markers mark the expression of proteins that can be used to phenotype those DC subtypes in various mouse tissues.17, 74-81 Of note, the inflammation affects the expression of many depicted markers, making them reliable exclusively during steady state.17, 74 TABLE 1. Phenotype maker of human dendritic cell (DC) subpopulations cDC1 cDC2 pDC moDC

HLA-DR+

CD11clow

BTLAint/low

CD1c-/low

SIRPα/CD72a-

CD141++

CADM1+

Clec9a+

XCR1+

CD26+

CXCR3+

CD103+

HLA-DR+

CD11c+

BTLAlow

CD1c+

SIRPα/CD72a+

CD11b+

CD141-

CD1a+

Clec4a/DCIR+

CD103+ (intestine)

HLA-DR+

CD14+

CD303/Clec4c+

CD304/NRP1+

HLA-DR+

CD14+

CD11c+

CD206+

CD64+

SIRPα/CD72a+

CD1a+

DC-SIGN/CD209+

Abbreviations: cDC1 and cDC2, conventional DC type 1 and 2, respectively; moDC, monocyte-derived DCs; pDC, plasmacytoid DCs. 2.1 Conventional DCs

In the bone marrow, growth factor FMS-like tyrosine kinase 3 ligand (Flt3l) mediates CDP differentiation into pre-cDCs that also express, but do not depend on, the zinc-finger transcription factor Zbtb46.41-43 Pre-DCs are released from bone marrow, migrate through blood and seed peripheral and lymphoid organs, where they differentiate into cDCs.44-47 While some plasticity might still exist at the pre-DCs stage, it seems that they are already pre-committed to develop into one of the two cDCs: type 1 (cDC1) or type 2 (cDC2) dendritic cells.45, 48-50 These two types of cDCs, generated by activation of different transcription factors, have complementary functions. The cDC1s, which depend on the transcription factors Irf8, Batf3, and Id2,12, 37 are specialized for capturing antigens from dying cells and their cross-presentation in an MHC-I context to naive CD8+ T cells,51-57 production of interleukin (IL-)12,58 and presentation of cell-associated antigens to naïve CD4+ T cells.59-62 On the other hand, cDC2s, which require the transcription factors Relb, Rbpj, and Irf4 for differentiation,63-66 present antigens in an MHC-II context to naive CD4+ T cells inducing helper T cell-mediated immune responses.67-70 However, recent single-cell sequencing data indicate that cDC2 cells could comprise several separate lineages.71-73 In the spleen, cDC2s were split into two subpopulations, cDC2A and cDC2B, delineated by the presence or absence of transcription factor T-bet, respectively.71 Likewise, cDC2s from the lungs were also clustered into several subpopulations that seem to represent different stages of their maturation.72, 73 Whether these cDC2 subpopulations represent novel DC subtypes having different functions or whether they simply represent different activation or differentiation stages of one common cDC2 subtype will have to be established in future studies.

Heterogeneity of cDCs at the transcriptional level is even more obvious when they are phenotypically classified. Recently, careful characterization of mouse cDCs across different organs indicated that they can be identified as CD64−F4/80−CD3−CD19−B220−NK1.1−MHCII+CD26+CD11c−/+.74 Of note, CD26 was added as a marker to account for downregulation or low CD11c expression on some cDCs.75, 76 Further, cDCs can be split into cDC1s, which express XC chemokine receptor 1 (XCR1), the C-type lectin receptor DNGR-1/CLEC9A, and the cell adhesion molecule CADM1, and cDC2s that express high levels of CD11b and CD172a.74, 77, 78 Importantly, the same expression pattern is found on human and macaque cDCs.74

The analysis of cDCs is further complicated by the fact that the tissue to which they locate also affects their phenotype. Mouse spleen- or lymph node-resident cDC1s, for example, are CD8α+, while peripheral tissue cDC1s express CD103 (integrin αE).79 However, novel methods that enable simultaneous analysis of a large (>40) number of markers, such as mass cytometry and spectral flow cytometry, and novel bioinformatic tools for data analysis allowed the insight into the complexity of DC phenotypes across and within the tissues.74, 80, 81 The complexity is further increased during inflammation, reflecting different cDC activation stages.17, 74 It is likely that the diversity of cDCs is a consequence of evolutionary adaptation for induction of appropriate immune response to divergent pathogenic insults.

2.2 Plasmacytoid DCs

As cDCs, pDCs also originate from bone marrow, where their development from CDPs requires the transcription factors Irf8 and TCF4 (E2-2).82, 83 They are continuously produced and released into circulation to seed different organs, where they survive only for several days.84 In the periphery, pDCs are characterized as CD11c+MHCII+PDCA1+B220+ (mice) or CD11c+MHCII+CD123+CD303(BDCA2)+CD304(BDCA4)+ (humans) DCs.12, 35 They are small, round cells whose primary function is rapid local production of type I interferons and other chemokines and cytokines as a response to viral infection.35, 37, 85 However, accumulating data indicate that pDCs do not appear to be a homogenous population and that there is a pDC subpopulation capable of antigen cross-presentation to CD8 T cells (reviewed in Refs. [35, 86]). This heterogeneity might be attributed to the fact that some pDCs appear to be differentiating from FLT3+IL-7R(CD127)+CD117lo/int lymphoid progenitor cells that might have divergent functional properties than pDCs generated from CDPs.87 For the moment, it remains unclear whether the different activation status of pDCs or their functional specialization causes reported functional differences.

2.3 Monocyte-derived DCs

Inflammation or infection induces the differentiation of monocytes into a specific DC subset, monocyte-derived DCs (moDCs), that synergize with cDCs in the induction of T cell–mediated immune responses (reviewed in Refs. [36, 88]). Already in the last millennium, it has been discovered that this process can be mimicked ex vivo by differentiating mouse monocytes into DCs upon treatment with granulocyte-macrophage colony-stimulating factor (GM-CSF).89, 90 Soon after it was discovered that for human monocytes GM-CSF has to be supplemented with IL-4.91, 92 The possibility to in vitro differentiate DCs from bone marrow progenitors or monocytes boosted the DC field dramatically since it overcame the up to then notorious scarcity of these cells limiting many experimental setups. Apart from expediting basic research, ex vivo generated DCs also raised hopes for the generation of cell-based vaccines for cancer immunotherapy (reviewed in Ref. [93]). However, the results from clinical studies did not meet the expectations, and although immune responses were induced, the clinical benefit of the therapy was limited.94 A partial explanation for these disappointing results provided the findings that only certain monocyte subpopulations are capable of differentiating into moDCs.95, 96 Recently, a group around Bart Lambrecht carefully dissected the role of moDCs in several mouse models of respiratory viral infection and allergy.17 They found that moDCs represent a mixture of monocyte-derived cells and cDC2s, which could be separated using CD26 as a marker. Importantly, only the inflammatory cDC2s, but not monocyte-derived cells, were migratory and had antigen presentation capacity similar to cDC1 cells.17 These intriguing results require validation in other inflammation models involving other organs.

2.4 Are there more bona fide DC subsets?

In addition to the described DC subpopulations, a potential forth DC subset has been suggested by several groups based on detailed single-cell RNA sequencing analysis. Distinguished from pDCs according to high expression of the chemokine receptor CX3CR1, they were termed “transitional” DCs (tDCs) as they share phenotype and characteristics of both pDCs and cDC2s.80, 97 Their role in immunity is at the moment unclear, although there are indications that tDCs induce regulatory T cell responses.98

The following chapters will provide an overview of the bioimaging toolbox used to visualize different DC subpopulations and their functions. We will mainly focus on the results obtained with mouse cDCs. Data obtained with other DC subtypes (moDCs and pDCs) will be mentioned at places where cDC data are not available. Finally, we will also summarize data on human DCs, where possible.

3 A NEEDLE IN A HAYSTACK: OR HOW TO FIND DCS IN NON-INFLAMED TISSUE?

Reliable markers are the prerequisite for successful DC visualization, ideally distinguishing divergent DC subtypes between themselves and from other cell types. Thus, one of the initial aims in DC research was to search for the antibodies that specifically bind to the DC surface markers99, 100 and their number steadily increased during the last decade of the 20th century. At that time, the main tool for visualization of DC positioning either at homeostatic or pathological conditions was immunohistochemistry. For example, it helped to establish that, during homeostatic conditions, cDC1s (then known as lymphoid DCs) localize in the T cell-rich zones of the spleen and lymph nodes, whereas cDC2 (then known as myeloid DCs) are in the marginal zone bridging channels of the spleen.101, 102 Further, it helped to confirm that human skin DCs also constantly recirculate through the lymph and accumulate in T cell-rich regions of lymph nodes, while lymph node-resident DCs predominantly locate around the lymph node cortex.18

At the beginning of the millennium, however, it became clear that the use of individual markers does not allow unambiguous visualization of DCs. During that time, the development of the fluorescent probes bloomed and progress in molecular biology enabled novel techniques to tag proteins and introduce fluorescent genetic markers.101 Moreover, methods for the detection of fluorescent probes, including flow cytometry and confocal microscopy, were optimized and empowered with increasing computing powers. These new tools enabled the discovery of specific DCs markers, allowing their distinction from other cells as well as characterization of different DC subtypes.13, 74, 103, 104 The visualization of DCs also became more complicated, requiring simultaneous detection of at least 3-5 different antibodies and or genetic markers. Thus, confocal microscopy replaced immunohistochemistry as a method of choice for DC visualization. Combinations of fluorophore-labeled antibodies against carefully selected markers enable reliable imaging of different DC subtypes. For example, a combination of antibodies against MHC-II, CD11c, and CD103 allows localization of cDC1s, while antibodies against MHC-II, CD11c, and CD11b allow detection of cDC2s in the lung (Figure 2). In mice, confocal microscopy of tissues stained with different antibody combinations revealed localization of divergent DC populations in various organs, including spleen,105-107 lymph nodes,108-110 and lungs.111-113 More recently, confocal microscopy visualized interactions of interfollicular DCs with fibroblastic reticular cells important for their homeostatic proliferation and survival.114 This method also allowed insight into the DC positioning within human tissue and their interactions with T and B cells.81, 115, 116

Differentiation of dendritic cells (DC) from macrophages within the mouse lungs by fluorescent confocal microscopy. Frozen lung slices (8 µm) from naïve C57Bl/6 mice were fixed 10 minutes with ice-cold acetone, rehydrated for 5 minutes in TBS (PBS, 0.05% Tween), blocked (5% rat serum, 5% anti-CD16/CD32 antibody (clone 2.4G2, produced in-house) in TBS for 15 minutes, and incubated for 45 minutes at room temperature with primary antibodies: (A) anti-CD11c (PE-Cy7), anti-MHCII (APC), and anti-CD103 (PE); (b) anti-CD11c (PE-Cy7), anti-MHCII (APC), and anti-CD11b (PE). After washing with TBS, the sections were stained with DAPI (1 µg/mL) for 3 minutes, washed again with TBS, embedded in Mowiol and imaged using an Axioscan Z1 (Zeiss). (a) cDC1s are MHCII+CD11c+CD103+, while alveolar macrophages (AM) are MHCII−/lowCD11c+CD103−. (B) cDC2 are MHCII+CD11c+CD11b+, AMs MHCII−/lowCD11c+CD11b−, while interstitial macrophages (IM) MHCII−/lowCD11c−CD11b+. Please note autofluorescence of airway epithelium in PE channel

The need for staining with several antibodies to label a particular DC subpopulation was partially alleviated by the generation of transgenic mice that express fluorescent markers specifically in DCs. Several such mouse strains are currently available, each with its strengths and weaknesses (Table 2). Of note, transgenic mice strains have to be rigorously tested for the specificity of reporter expression to prevent misinterpretation of the data. For example, initial reports using CD11c-EYFP mice indicated an extensive DC network involved in the brain important for protection against viral infections.117, 118 However, detailed immunophenotyping of those YFP-expressing brain cells revealed that they are microglia rather than DCs.119 Similarly, while reporter gene expression in CD11c- and CX3CR1-transgenic mice suggested a dense DC network in the renal interstitium,120, 121 a majority of those cells were recently reported to be macrophages.14 Nevertheless, transgenic reporter mice represent valuable tools for research of DC maturation, migration, and interaction with other cells, as described in the following sections. Genetic labeling, however, only partially solved the need for simultaneous staining of several markers to define one DC subtype.

In the last decade, several methods for visualizing of up to 60 different markers emerged, circumventing the limited number of channels available in most conventional confocal microscopes. Serial staining immunofluorescence increases the number of samples that can be simultaneously imaged by iterative staining, imaging, and removing of fluorescence signal.122-124 However, serial staining immunofluorescence usually requires even week-long stain/wash/image/strip cycles that are prone to sample degradation over time.5 In this regard, an advanced iterative staining and chemical bleaching protocol were recently developed which prevents tissue destruction and reduces the fluorophore inactivation and antibody labeling steps to less than 1 hour.125 Another approach to expand the number of fluorophores that can be detected simultaneously on a confocal microscope was by combining imaging with spectroscopy. This led to the development of spectral imaging, a method that in addition to intensity also collects spectrum data for every pixel of the image. Resolution of spectrum data enables simultaneous detection of numerous fluorophores in large tissue areas, while the intensity of signal provides information of the abundance of maker expression.126, 127 This method is especially effective when used on antibody-stained tissues from transgenic reporter mice. The group of Ronald Germain demonstrated its full power and firmly established that migratory and resident cDC1 cells can predominantly be found within the T cell zone of the lymph node, migratory cDC2 cells in interfollicular regions, whereas resident cDC2 cells reside in the outer cortex and within the lymphatic sinus endothelium.128, 129 All those DC subpopulations are predominantly associated with lymph node blood vessels but locate to different parts of vascular trees so that there is little local mixing of cells.130 A similar approach was recently applied to describe the distribution of cDCs (and pDCs) in human tonsils.116

Another visualization method that can match flow cytometry in the number of simultaneously detected markers is mass cytometry imaging (MCI), a method that uses either imaging mass cytometry or multiplexed ion beam imaging.5 The basic principle of MCI is that antibodies are conjugated to stable isotopes instead of fluorophores. After staining, isotope abundance is detected in small (1 µm) parts of the tissue using a mass spectrometer. Subsequent computer-assisted reconstitution leads to the assembly of a high-dimensional image indicating positions of different cells. Due to the specificity of isotype detection, MCI allows simultaneous detection of up to 40 different parameters.131 Although the full power of this method is still being tested, imaging mass cytometry localized DCs around blood vessels in kidneys,14 and established their contribution to skin symptoms in dermatomyositis, a rare, systemic autoimmune disease.132 In a current pre-print, multiplexed ion beam imaging was used to monitor cell interactions within the healthy and virus-infected lymph nodes.133 Unfortunately, a broad use of MCI is hindered by long acquisition times due to evaporation of small tissue points (2-5 hours for 1 mm2),5 in addition to high costs for equipment and reagents.

An alternative could be oligonucleotide hybridization, which increases the number of markers that can be imaged with most standard 3-color fluorescence microscopes to up to 28.134, 135 In an approach known as Immuno-SABER, tissue is stained with a mixture of antibodies labeled with DNA barcodes that are then extended via primer exchange reactions into orthogonal single-stranded DNA concatemers generated in a preprogrammed manner.135 The signals are generated by binding of multiple fluorophore-labeled detection oligos in a sequential manner. In the in situ polymerization of fluorescent dNTP analogs (CODEX) method, tissues are stained with a mixture of antibodies labeled with uniquely designed oligonucleotide duplexes with 5′ overhangs.134 These antibodies are then detected by fluorescently labeled index nucleotides that are incorporated by in situ oligonucleotide extension. Polymerization proceeds sequentially so that in each step only two oligonucleotides (ie two antibodies) are labeled, detected, and bleached before the next set of nucleotides is polymerized. It remains to be evaluated how well these methods can be used for imaging of DC subtypes.

All so far mentioned visualization methods rely on the analysis of individual tissue sections. While it is possible to reconstruct a virtually three-dimensional (3D) image of an organ (or part of it) from thin tissue slices,136 these reconstructions are imperfect due to tissue loss and structural disruption during fixation, slicing, and staining. To circumvent this issue, a number of tissue clearing procedures have been developed that enable light penetration through the whole organ have been developed during the last decade.137, 138 By selecting an appropriate clearing protocol, it is possible to make tissue completely transparent and stain even large tissue blocks, such as whole lung lobes139 or parts of the spleen.140 Imaging of such samples is usually done on a light-sheet fluorescent microscope, in which a transparent sample is illuminated with the thin plane of light to image a series of optical sections that can be then 3D reconstructed.