Abstract

   The human bone marrow (BM) gives rise to all distinct blood cell
   lineages, including CD1c+ (cDC2) and CD141+ (cDC1) myeloid dendritic
   cells (DC) and monocytes. These cell subsets are also present in
   peripheral blood (PB) and lymphoid tissues. However, the difference
   between the BM and PB compartment in terms of differentiation state and
   immunological role of DC is not yet known. The BM may represent both a
   site for development as well as a possible effector site and so far,
   little is known in this light with respect to different DC subsets.
   Using genome-wide transcriptional profiling we found clear differences
   between the BM and PB compartment and a location-dependent clustering
   for cDC2 and cDC1 was demonstrated. DC subsets from BM clustered
   together and separate from the corresponding subsets from PB, which
   similarly formed a cluster. In BM, a common proliferating and immature
   differentiating state was observed for the two DC subsets, whereas DC
   from the PB showed a more immune-activated mature profile. In contrast,
   BM-derived slan+ non-classical monocytes were closely related to their
   PB counterparts and not to DC subsets, implying a homogenous prolife
   irrespective of anatomical localization. Additional functional tests
   confirmed these transcriptional findings. DC-like functions were
   prominently exhibited by PB DC. They surpassed BM DC in maturation
   capacity, cytokine production, and induction of CD4+ and CD8+ T cell
   proliferation. This first study on myeloid DC in healthy human BM
   offers new information on steady state DC biology and could potentially
   serve as a starting point for further research on these immune cells in
   healthy conditions as well as in diseases.

   Keywords: dendritic cells, non-classical monocytes, bone marrow,
   peripheral blood, microarray analysis, cytokines

Introduction

   The human bone marrow (BM) compartment is the main site of
   hematopoiesis. It harbors early progenitor cells as well as more
   differentiated cells that will enter the circulation in a later stage
   for homing to secondary lymphoid organs or effector sites to mediate
   their immune functions. The developmental pathway of the dendritic cell
   (DC) lineage in the BM has been extensively studied in mice
   ([37]1–[38]12). Later, human homologs of DC were identified and
   recently the differentiation from human BM precursors into DC was
   described ([39]13–[40]23). Increasing knowledge on the sequential
   development of common and early progenitors to committed and more
   differentiated precursors may support the understanding of normal and
   aberrant hematopoiesis. This is for instance exemplified by selective
   cell lineage depletion in primary immunodeficiencies such as seen in
   patients with GATA2 mutations ([41]24–[42]26). These patients clearly
   show reduced numbers of natural killer cells, B cells, and monocytes
   and in addition a total absence of all DC subsets. In contrast, the
   level of granulocytes is normal, indicating a block in differentiation
   from progenitor to the more developed stages. Only recently, the
   immediate precursor of human DC has been identified which may now
   explain these DC-restricted depletions. Lee and colleagues developed a
   culture system in which they were able to define the human DC
   developmental pathway ([43]18). From a common human
   granulocyte-monocyte-DC progenitor to a more committed human
   monocyte-DC progenitor, they finally could identify the direct human
   common DC progenitor (hCDP). The presence of hCDP is restricted to the
   BM compartment where it gives rise to a migratory conventional DC
   precursor (pre-CDC) that then produces the two major types of human
   myeloid DC, CD1c+, and CD141+ (now called cDC2 and cDC1, respectively)
   ([44]19, [45]27). Transcriptional heterogeneity within the pre-CDC pool
   further revealed the existence of pre-commitment for cDC2 or cDC1 DC
   ([46]20). Examination of different human tissues showed that these
   pre-CDC were not only present in the BM but could also be found in
   peripheral blood (PB) and tonsils. In contrast, the immediate monocyte
   precursor was only detected in BM. In addition to the information on
   sequential precursor production, data on activation state and
   functional capacities in different stages of cell development may give
   further insights into healthy and altered immune responses. It is
   assumed that immature/precursor forms of DC, including cDC2 and cDC1,
   leave the BM and circulate through the PB to the lymphoid tissues.
   Previously published data indeed support the fact that these
   circulating precursors are less mature and differentiated than their
   counterparts in peripheral tissues ([47]15, [48]24, [49]28).

   In human, the effect of environmental signals on DC function in
   different tissues, such as spleen, thymus, PB, lung, and skin was
   studied before by Heidkamp and colleagues ([50]29).

   However, information on developmental state and functional differences
   of cDC2 and cDC1 DC between the human BM and PB compartment is not
   available.

   In this study, we focused on myeloid DC in the healthy human BM
   compartment in steady state conditions and compared them with their PB
   equivalents. Slan+ non-classical monocytes, previously recognized as a
   DC-like subset and, therefore, probably more closely linked to
   conventional DC than classical monocytes, were also included in this
   analysis. Detailed genome-wide transcriptional analysis as well as
   comparative functional validation showed clear compartment-related
   differences. Hierarchical clustering implied a close relationship of
   BM-derived cDC2 and cDC1 DC. Both subsets clustered away from their PB
   equivalents. Differentiation pathways, cell proliferation, and cell
   cycle control genes were most prominent in BM DC, whereas immune
   effector-related transcripts were over-expressed in PB subsets. In
   contrast, transcriptional profile and immunological function of slan+
   non-classical monocytes were less dependent on the environment as a
   homogenous profile in both compartments was demonstrated. BM- and
   PB-derived slan+ non-classical monocytes clustered closer together than
   with the DC subsets from their respective compartments. Thus, under
   steady state circumstances both myeloid DC subsets potentially form a
   residential BM population that are in an earlier developmental phase
   than their PB equivalents, whereas slan+ non-classical monocytes are
   possibly PB-derived and patrol the BM in an effector role.

Materials and Methods

Sample Processing and Cell Isolation

   Human BM was obtained after written informed consent from patients
   undergoing cardiac surgery in the VU University Medical Center
   Amsterdam, The Netherlands. Peripheral blood (buffy coat) was purchased
   from the Sanquin blood supply service (Amsterdam, The Netherlands). The
   study was approved by the local ethical committee and in accordance
   with the declaration of Helsinki. Mononuclear cells (MNCs) from both PB
   as well as from BM were isolated by density centrifugation using
   Ficoll-Paque medium (GE Healthcare, Uppsala, Sweden). For functional
   tests, conventional DC subsets (cDC2 and cDC1) were sorted from fresh
   samples with a BD FACSAria™ flow cytometer (BD Biosciences, San Jose,
   CA, USA), using a monoclonal antibody (mAb) cocktail consisting of
   CD1c-PE-Cy7 (eBioscience, San Diego, CA, USA), CD11c-PerCP-Cy5.5,
   CD14/CD19-APC-H7 (all, BD Biosciences), CD141-APC (Miltenyi Biotec,
   Utrecht, The Netherlands), and HLA-DR-V450 (BD Biosciences). Both cell
   subsets were identified in the CD11c+ and HLA-DR+ cell fraction. cDC2
   cells were also gated as CD14−/CD19− (Figure S1A in Supplementary
   Material). For better recovery of viable cells, slan+ non-classical
   monocytes were isolated based on M-DC8 expression using a magnetic
   isolation kit (Miltenyi Biotec). For application in microarray
   experiments, all cell subsets were flowcytometrically sorted from four
   individual PB and four BM donors (non-paired samples). After an
   enrichment step, comprising magnetic depletion of CD3+ and CD19+ cells
   using microbeads (Miltenyi Biotec), cells were incubated with a mAb
   cocktail consisting of M-DC8-FITC (Miltenyi Biotec), CD16-PE (Beckman
   Coulter, Brea, USA), CD11c, CD1c, CD141, CD14, CD19, and CD45-KO (BD
   Biosciences, Figure S1B in Supplementary Material). Sorted cells
   routinely showed >92% purity and were stored in Trizol (Life
   Technologies, Carlsbad, CA, USA) until further use for RNA extraction.

Cell Frequencies

   The enumeration of different cell subtypes in BM and PB was performed
   after lysis of erythrocytes (BD Pharm Lyse, BD Biosciences). Samples
   were incubated with a mix of mAb: M-DC8-FITC, CD303-FITC (Miltenyi
   Biotec), CD16-PE, CD11c-PerCP-Cy5.5, CD1c-Pe-Cy7, CD141-APC,
   CD14-APC-H7, CD19-APC-H7, HLA-DR-V450, and CD45-KO and analyzed on a
   flow cytometer (FACSCanto-II, BD Biosciences). Data analysis was done
   using FlowJo software (Tree star, Ashland, OR, USA). After debris and
   doublet exclusion, myeloid cell subsets were identified in the CD45+
   compartment and further gated on CD1c, CD141, M-DC8, CD303, or CD14
   expression. HLA-DR and CD11c were used for further selection of cell
   populations (Figure [51]1A). To exclude contamination of B cells in the
   CD14+ monocyte gate, back gating in CD45 vs side scatter plots was used
   (data not shown in figure). Frequencies were calculated as percentages
   of CD45+ MNCs.

Figure 1.

   [52]Figure 1

   [53]Figure 1
   [54]Open in a new tab

   Dendritic cell (DC) and monocyte enumeration and baseline expression
   levels of (co-) stimulatory molecules in BM and PB. (A) Gating strategy
   for DC and monocyte subsets. CD45+ MNCs were gated after debris and
   doublet exclusion. Thereafter, cell subsets were identified based on
   the expression of CD1c (cDC2), CD141 (cDC1), or M-DC8 (slan+
   non-classical monocytes). Classical monocytes were selected based on
   their high CD14 expression. (B) Frequencies of slan+ cells, monocytes,
   and the myeloid DC subsets cDC2 and cDC1 in human BM as compared to
   human PB. Percentages were calculated from the total CD45+ mononuclear
   cell fraction. Individual results and median percentages are shown for
   17–21 (PB) and 13–20 (BM) healthy donors. Monocytes formed the main
   population in both compartments (PB, 16.0% and BM, 12.5%), followed by
   cDC2 (PB, 0.53% and BM, 0.60%) and slan+ non-classical monocytes (PB,
   0.92% and BM, 0.39%). In BM as well as in PB, cDC1s were only present
   in low percentages (PB, 0.043% and BM, 0.081%). (C) Paired frequencies
   of PB and BM subsets, derived from one donor. A Wilcoxon matched-pairs
   signed rank test was performed to compare the PB and BM compartment
   (n = 7 donors, cDC2, and monocytes: no significant differences, cDC1
   p = 0.03 and slan+ p = 0.02). (D) Baseline expression levels of HLA-DR
   (n = 7) and the co-stimulatory molecules CD80 (n = 6) and CD86 (n = 7)
   on BM-derived cell subsets. Individual experiments and median values
   are shown. (E) A comparison of baseline expression levels of three
   different activation markers between cell subsets from the PB
   (n = 14–23) and BM (n = 6–7) compartment. Histograms of the different
   markers of a representative donor and bar graphs of repeated
   experiments are shown. Median MFI values of individual stainings of the
   different cell subsets are compared between the two compartments. For
   CD80 and CD86, FMO controls are shown as a dotted line. Abbreviations:
   BM, bone marrow; PB, peripheral blood; WBC, white blood cells; MNC,
   mononuclear cells; MFI, median fluorescence intensity; FMO,
   fluorescence minus one. *p < 0.05, **p < 0.01, ***p < 0.001, and
   ****p < 0.0001.

Preparation of cRNA and Gene Chip Hybridization

   RNA was isolated from cDC2 (18,000–265,000 cells), cDC1 (1,000–17,300)
   and slan+ non-classical monocytes (9,000–150,000) and amplified using
   the Ovation Pico WTA System V2 (NuGen, San Carlos, CA, USA).
   Subsequently, RNA was labeled with the Encore Biotin Module Kit (NuGEN)
   according to manufacturer’s protocols. Five microgram of the resulting
   labeled cDNA from each sample was hybridized to Human Transcriptome
   Arrays 2.0 microarrays (Affymetrix) and signals were scanned by
   Affymetrix GeneChip Scanner 3000 7G.

Microarray Data Analysis

   Data analysis was performed as previously described ([55]30). Briefly,
   microarray signals were normalized using the RMA algorithm in the
   Affymetrix^® Expression Console software (Affymetrix^®) and quality
   checks and background noise correction were performed. Qlucore Omics
   Explorer 3.2 (Qlucore AB, Lund, Sweden) and Perseus were used for
   transcript filtering, identification of differentially expressed genes
   (DEGs; by performing an ANOVA analysis and using q-values of <0.05),
   heat map visualization, and Gene Set Enrichment Analysis (GSEA).
   Pathway analyses and interaction networks were created within the
   STRING v10.0 database ([56]http://string-db.org/). The microarray data
   have been deposited in the GEO public database under Accession Number
   [57]GSE112770.

Cell Staining, Maturation, and Cytokine Secretion

   The intracellular expression of the proliferation marker Ki-67 was
   assessed in different cell subsets of the myeloid lineage by using a
   mAb panel consisting of the surface markers CD1c, CD11c, CD14, CD19,
   CD45, CD141, M-DC8, and HLA-DR, and, a PE-conjugated mAb against the
   intracellular Ki-67 marker (BD Pharmingen). Unstimulated cell subsets
   of eight paired BM and PB samples were analyzed. After erythrocyte
   lysis, cell surface staining was performed. Thereafter, cells were
   fixed with phosphate-buffered saline containing 1% paraformaldehyde
   (PFA) and permeabilized by means of BD FACS Lysing solution (BD FACS
   Lysing Solution, BD Biosciences). Subsequently, the Ki-67 mAb was added
   and its intracellular expression in the different cell subsets was
   evaluated by flow cytometry (FACSCanto-II, BD Biosciences).

   Baseline expression levels of co-stimulatory (CD80 and CD86) and minor
   histocompatibility (MHC)-class II (HLA-DR) molecules were evaluated
   before isolation of BM and PB cell subsets. Assessment of maturation
   capacity upon TLR-ligation by DC subsets and slan+ non-classical
   monocytes from both compartments was performed as described previously
   ([58]30). Briefly, cells were isolated and pre-incubated for 4 h in
   Iscove’s Modified Dulbecco’s Medium supplemented with 10% fetal calf
   serum, 100 IU/ml penicillin, and 100 µg/ml streptomycin (all Gibco,
   Paisley, United Kingdom) at 37°C in a 5% CO[2] incubator. Thereafter,
   they were left unstimulated or were stimulated overnight with a
   combination of TLR ligands, i.e., LPS (100 ng/ml, Sigma-Aldrich, St.
   Louis, MO, USA) and R848 (3 µg/ml, Enzo Life Sciences, Farmingdale,
   USA) for cDC2 and slan+ non-classical monocytes, or Poly I:C (25 µg/ml,
   Sigma-Aldrich) and R848 for cDC1. The expression of respective markers
   was again measured by flow cytometry. Supernatants of these cultures
   were used for the detection of different cytokines by human enhanced
   sensitivity cytometric bead array (CBA) flex sets (BD Biosciences).

Allogeneic Mixed Leukocyte Reaction (MLR)

   The capacity of DC subsets and slan+ non-classical monocytes to induce
   T cell proliferation was tested in an allogeneic MLR as described
   previously ([59]30, [60]31). In brief, carboxyfluorescein succinimidyl
   ester (CFSE, Life Technologies, 1 μM)-labeled PB lymphocytes (PBL,
   i.e., CD14 depleted cell fraction) were co-cultured for 5 days with
   isolated and stimulated BM- or PB-derived slan+ monocytes or cDC2 in a
   ratio of 1:10. CFSE-dilution in CD4+ and CD8+ T cells was measured by
   flow cytometry.

Statistical Analysis

   Graphpad Prism 6 software (San Diego, CA, USA) was used for
   non-microarray statistical data analysis. For two-group comparisons a
   non-parametric Mann–Whitney U test was performed, whereas for matched
   sample analyses a Wilcoxon signed rank test was used. For statistical
   differences a p-value of <0.05 was considered significant.

Results

Quantitative Differences in Subsets Between the BM and PB

   Dendritic cell and monocyte subsets were enumerated in the mononuclear
   cell fraction of fresh PB (n = 17–21) and BM (n = 13–20) samples from
   non-paired healthy donors (HD). Cell populations were characterized by
   the expression of the cell-identifying markers CD1c and CD141 for the
   myeloid DC subsets. cDC2 were further defined as CD19−/CD14− and
   HLA-DR+/CD11c+. Slan+ cells, recently identified as the non-classical
   monocytes ([61]30, [62]32), were characterized based on their CD11c and
   M-DC8 expression. Backgating in CD16 confirmed the expression of this
   marker on their cell surface (not shown). In contrast, classical
   monocytes did not express CD16, and were identified based on the high
   expression of CD14 (Figure [63]1A). B cell contamination was ruled out
   by creating back gating plots using discriminative CD45 expression and
   side scatter position of monocytes (not shown). In both compartments,
   all cell subsets could be found. Among these cells, CD14+classical
   monocytes accounted for the main population in PB as well as in BM
   (Figure [64]1B; PB, 16.0%; BM, 12.5%). Slan+ non-classical monocytes
   showed elevated numbers in PB compared to BM (PB, 0.92%. BM, 0.39%,
   p = 0.0095). In contrast, for cDC1, the frequency was significantly
   increased in the BM compared to PB (PB, 0.043%. BM, 0.081%,
   p < 0.0001). Comparable frequencies were detected for BM- and
   PB-derived cDC2 (PB, 0.53%; BM, 0.60%). In a set of 7 paired frequency
   analyses (PB and BM derived from the same donors), similar results were
   found for the different cell populations as compared to the larger
   unpaired cohort. For cDC1, increased percentages were seen in BM,
   whereas for slan+ non-classical monocytes all one-to-one comparisons
   showed elevated numbers in the PB compared to the BM (Figure [65]1C).

   Extensive immunophenotypic information for DC and monocyte subsets in
   PB is available from previously published research ([66]27, [67]30,
   [68]33–[69]38). However, for human BM DC and monocytes these data are
   very limited ([70]29). Therefore, the expression levels of important
   co-stimulatory and antigen-presenting molecules were assessed in
   BM-derived cells and compared to the expression levels on their PB
   counterparts. Similar to our observations for PB cell subsets ([71]30),
   CD86 and HLA-DR were clearly expressed (at baseline, i.e., unstimulated
   expression) on DC and monocyte subsets. The expression of CD86 was
   higher on BM-derived monocyte subsets compared to BM-derived DC,
   whereas HLA-DR showed elevated levels on DC subsets (Figure [72]1D).
   Baseline expression levels of these markers were minimally different
   between BM and PB (Figure [73]1E). Only the HLA-DR expression on BM
   cDC2 was significantly higher compared to PB cDC2. For monocytes, CD86
   levels were significantly higher in PB.

Transcriptomic-Based Clustering of DC Subsets but Not of Slan+ Non-Classical
Monocytes Largely Depends on the Compartment They Reside in

   For a better insight into the function of myeloid DC subsets (cDC2 and
   cDC1) in the BM compartment and to potentially identify
   compartment-related transcriptomic differences, an extensive
   genome-wide profiling study was performed. Both DC subsets as well as
   slan+ non-classical monocytes were sorted from fresh PB and BM of four
   individual HD. A total of 58,036 transcripts were found to be
   expressed. Principal component analysis (PCA), using all expressed
   transcripts, showed a strong similarity of cell subsets between donors
   (Figure [74]2A). After applying an ANOVA test, 24,191 genes were found
   to be differentially expressed between all subsets in both
   compartments. Minimal spanning tree analysis revealed a
   compartment-based separation of cDC2 and cDC1. In contrast, slan+
   non-classical monocytes from both compartments clustered closer
   together and separated from the DC subsets, irrespective of compartment
   (Figure [75]2B). Hierarchical clustering confirmed the separation for
   the conventional cDC2 and cDC1 subsets. Slan+ non-classical monocytes
   from PB and from BM grouped together (Figure [76]2C). Furthermore,
   slan+ non-classical monocytes were more closely related to PB DC than
   to BM DC subsets. The same analysis was repeated for DC only, without
   using slan+ monocytes. PCA and hierarchical clustering of all genes and
   10114 DEGs showed similar results (Figure S2 in Supplementary
   Material). These findings suggest that transcriptional expression in
   myeloid DC subsets is more influenced by their environment (BM or PB)
   than by their intrinsic subset characteristics. In contrast, gene
   expression in slan+ non-classical monocytes appeared to be less
   dependent on environmental factors.

Figure 2.

   Figure 2
   [77]Open in a new tab

   Principal component analysis (PCA) and hierarchical clustering of cell
   subsets in bone marrow (BM) and peripheral blood (PB). Relationship
   between slan+ monocytes and cDC2 and cDC1 (myeloid DC) in the two
   different compartments: PB vs BM. For each subset, isolated cells of
   four PB and four BM donors were used. (A) PCA for all expressed
   transcripts and (B) PCA for genes that were differentially expressed
   between the cell subsets. The length of the lines between the dots
   corresponds to the degree of similarity (minimal spanning tree
   analysis). Percentages shown indicate the degree of total variation
   within each component. (C) Hierarchical clustering, based on
   differentially expressed transcripts, of replicate samples.

DC Subsets Display a Less Immune-Active Transcriptional Signature Than Slan+
Non-Classical Monocytes in the BM

   Recent gene expression profile studies mainly focused on PB-derived
   cell types. For human BM-derived monocytes and DC, limited, if any,
   transcriptional information is available. Therefore, detailed gene
   expression profiles were analyzed in BM-derived slan+ non-classical
   monocytes, cDC2, and cDC1. From all expressed BM genes (58,459),
   differentially expressed transcripts between the three subsets were
   listed. Only if a gene was differentially over- or under-expressed in
   both comparative analyses (e.g., slan+ vs cDC2 and slan+ vs cDC1), it
   was considered to be subset-specific. Similar to our previous findings
   for PB subsets ([78]30), BM-derived slan+ non-classical monocytes
   showed the highest number of unique transcripts. In total, 2,142 genes
   were differentially expressed in slan+ non-classical monocytes as
   compared to both DC subsets. From these DEGs, 965 were over-expressed
   and 1,177 were under-expressed. cDC2 showed 31 over-expressed and 14
   under-expressed transcripts compared to the other two subsets. For cDC1
   69 genes were found to be over-expressed and 54 to be under-expressed.
   Clustering of these DEGs in a heat map identified four gene clusters
   (Figure [79]3A, lower panel). Absolute expression levels of the
   identified DEGs offered further details of the possible biologically
   relevant differences between the subsets (Figure [80]3B). Consistent
   with our PB data ([81]30), the top 100 over-expressed genes for slan+
   non-classical monocytes contained recognized monocytic and
   immune-related genes, such as FCGR3A/B, TLR4, C5AR1/2, C3AR1, CR1, and
   CX3CR1. For cDC2, 25 genes with highest FC values are shown and for
   cDC1 60 genes. Also for the two BM-derived DC subsets well-known
   transcripts were found (CD1c, CD163, MRC1, and CSF3R for CD1c DC and
   XCR1, CLEC9A, and IRF8 for CD141 DC), indicating a consistent
   expression of subset-defining genes in BM and PB. Due to the limited
   number of over-expressed transcripts in cDC2 and cDC1, further pathway
   and interaction analyses were only performed for slan+ non-classical
   monocytes in order to find relevant biological context for these DEGs.
   Again, the top 100 highest FC values were used in this approach. They
   were imported in the STRING v10.0 database and a gene ontology (GO)
   term enrichment analysis was executed. The most enriched pathways
   mainly covered immune-related biological processes. As found for
   PB-derived slan+ non-classical monocytes ([82]30), their BM
   counterparts also appeared to be highly involved in complement-mediated
   immune responses (Figures [83]3C,D). Enrichment analysis for
   under-expressed DEGs pointed to a lower involvement in adhesion and
   antigen processing and presentation (Figure [84]3D). Yet again, this
   shows great overlap with previous PB findings ([85]30). Network
   interaction analysis of over-expressed DEGs showed a strong interaction
   of complement-related proteins as well as TLR-mediated signaling
   pathways (Figure [86]3E). For under-expressed DEGs a strong network of
   MHC-class II molecules was found (Figure [87]3F).

Figure 3.

   [88]Figure 3

   [89]Figure 3
   [90]Open in a new tab

   Differentially expressed genes (DEGs) between slan+ monocytes and cDC2
   and cDC1 in the bone marrow (BM) compartment. After identification,
   DEGs were further analyzed for the three cell subsets in the BM
   compartment. (A) Cluster analysis of transcripts that are
   differentially expressed between the subsets are presented in a heat
   map. Each row represents a single transcript. Relative expression
   levels for four donors per subset are shown. z-scores were calculated
   in one row to show the relative expression values per donor compared to
   the mean of the entire row (z-score −2: low expression, z-score +2:
   high expression). Four different clusters were generated. In green,
   DEGs that were over-expressed in cDC2 are shown, whereas DEGs
   over-expressed in cDC1 are within the blue cluster. Under- and
   over-expressed transcripts for slan+ monocytes are shown in yellow and
   gray, respectively. (B) Coding and over-expressed transcripts, sorted
   on mean FC values, are shown for all cell subsets. For slan+ monocytes
   the top 100 is used, whereas for the other two cell subsets 25 and 60
   transcripts are shown (cDC2 and cDC1, respectively). For each
   transcript, the mean absolute expression level of four donors (1: low
   expression, 10: high expression) is color coded in the given cell
   subset. Q-values for both comparisons (i.e., slan+ vs cDC2 and slan+ vs
   cDC1) are shown. (C) Pathway analysis of over-expressed transcripts in
   slan+ monocytes compared to both DC subsets. The top 100 highest FC
   values from over-expressed coding transcripts were imported into the
   STRING v10.0 bioinformatics tool and significant enriched pathways were
   identified. From this list, the top 10 with lowest FDR (out of 62) of
   enriched biological processes are shown. (D) Enriched KEGG pathways for
   over- and under-expressed transcripts in slan+ monocytes compared to DC
   subsets. Top 100 highest FC values of coding transcripts were used from
   the over- and under-expressed gene lists. An interaction network
   analysis of involved transcripts is displayed for over-expressed genes
   (E) as well as for under-expressed genes (F) in slan+ monocytes.
   Disconnected nodes are hidden and the line thickness indicates the
   strength of data support. Abbreviations: FC, fold change; FDR, false
   discovery rate; GO, gene ontology; ID, identification; KEGG, Kyoto
   encyclopedia of genes and genomes.

Gene Expression Profiles Are Dependent on Their Environment

   To further delineate the distinct gene expression profiles for cDC2 and
   cDC1 in relation to their compartment (and in a lesser extent for slan+
   monocytes), DEGs between PB and BM were identified for all three
   individual subsets. The highest number of differentially expressed
   transcripts between BM and PB was found for cDC2: in total 4,019 DEGs,
   of which 2,439 DEGs were over-expressed in BM and 1,580 over-expressed
   in PB. For cDC1 743 genes were over-expressed in BM and 436 in PB.
   Slan+ non-classical monocytes showed 998 genes to be over-expressed in
   BM and 847 in PB. Clustering of relative expression levels in each
   subset is depicted in Figure [91]4A. In order to translate these
   identified DEGs into a relevant biological meaning, lists of over- and
   under-expressed genes were loaded onto the STRING v10.0 pathway tool.
   For all three subsets, but most distinctly for cDC2 and cDC1, cell
   cycle and division processes were highly enriched in BM-derived
   subsets. Also metabolic cellular responses were significantly
   overrepresented in BM-derived cells (Figure [92]4B). Genes that were
   over-expressed in PB subsets were mainly involved in immune-related
   pathways, indicating a more differentiated and immunologically
   functional profile for cells circulating in this compartment (Figure
   [93]4C). To further confirm these findings a GSEA was performed for BM
   as well as for PB subsets. Using Kyoto Encyclopedia of Genes and
   Genomes pathways from the Broad Institute database an enrichment of
   pathways involved in cell cycle and replication was found for BM
   subsets. In contrast, immune-related pathways were enriched in
   PB-derived subsets (Figure [94]4D). Again, confirming a distinct
   transcriptional profile of cell subsets in BM and PB.

Figure 4.

   [95]Figure 4

   [96]Figure 4
   [97]Open in a new tab

   Transcriptomic comparison between cell subsets of the bone marrow (BM)
   vs the peripheral blood (PB) compartment. Gene expression profiles were
   compared between the BM and PB for each individual cell subset. (A)
   Number of differentially expressed transcripts found between the two
   different compartments (e.g., slan+ in BM vs slan+ in PB).
   Differentially expressed genes (DEGs) were clustered per row by using
   z-scores for each transcript, thereby reflecting relative expression
   levels. Heat maps for each individual subset are shown. For each cell
   subset, a pathway enrichment analysis was performed in which all DEGs
   were used that were over-expressed in BM (B) or over-expressed in PB
   (C). Top 10 enriched biological processes for both compartments and
   each individual subset are shown. (D) Gene set enrichment analysis for
   BM- and PB-derived cell subsets. The top 10 of enriched KEGG pathways
   are shown together with four selected enrichment plots. Abbreviations:
   FDR, false discovery rate; GO, gene ontology; ID, identification; NES,
   normalized enrichment score.

Generation of a Gene Signature for DC Subsets That Is Environmentally
Dependent

   To establish a DC signature for BM as well as for PB, over-expressed
   gene lists (from the comparison BM vs PB) from both myeloid DC subsets
   were used. Since slan+ non-classical monocytes clearly form a
   non-DC-like population, they were excluded from this analysis. Using
   gene descriptions and GO biological process terms (e.g., cell cycle and
   immune response), transcripts involved in cellular and immunological
   processes were first selected for cDC2 as well as for cDC1. Relative
   expression levels of these transcripts are shown for BM and PB for both
   DC subsets (Figure [98]5A). From this selection, it is clear that
   transcripts that are involved in DNA replication, differentiation
   (i.e., FLT3), and cell cycle processes were evidently over-expressed in
   BM DC compared to PB DC. In contrast, cytokine signaling, adhesion, and
   endocytosis are terms that were overrepresented in PB DC. For PB cDC2,
   also co-stimulatory molecules such as CD40, CD80, and CD83 were
   expressed in higher levels compared to their BM equivalents. To further
   search for biologically relevant, and compartment-specific, DC gene
   signatures, Venn diagrams were created using gene lists of BM DC and PB
   DC. This revealed an overlap in over-expressed genes of 581 (22.3%)
   between BM cDC2 and BM cDC1 and 131 (6.9%) between PB cDC2 and PB cDC1
   (Figures [99]5B,C), indicating a greater similarity between BM DC than
   between PB DC. Next, both overlapping genes lists were used for further
   pathway and network analysis in STRING v10.0. Pathways involved in cell
   cycle processes and cell division represented the overlapping
   transcripts in BM DC and formed the main biological function of these
   DC in this compartment (Figure [100]5D). In contrast, the overlapping
   transcripts in PB-derived DC could be found in pathways involving cell
   activation, differentiation, and antigen-receptor mediated signaling
   (Figure [101]5E). A network of these common PB-derived DC transcripts
   was generated (Figure [102]5F). Since the number of overlapping
   transcripts in BM DC was too high, a network could not be created.

Figure 5.

   [103]Figure 5

   [104]Figure 5
   [105]Open in a new tab

   Compartment-specific DC gene signatures for cDC2 and cDC1. Both myeloid
   DC subsets were further analyzed to find compartment-specific DC gene
   signatures. (A) Transcripts were filtered based on gene descriptions
   and GO biological process terms by using cell cycle- and immune
   response-related terms. Relative expression for cDC2 and cDC1 in both
   bone marrow (BM) as well as peripheral blood (PB) are shown for
   selected genes. Venn diagrams were used to obtain overlapping genes
   between cDC2 and cDC1 in the BM (B) and PB (C) compartment. (D) The 581
   commonly over-expressed transcripts in BM-derived DCs were used in a
   pathway enrichment analysis. (E) Similarly, overlapping transcripts
   (131 in total) of PB cDC2 and cDC21 were used in the STRING pathway
   analysis. (F) An interaction network of the PB transcripts was
   generated. In total, 85 genes could be annotated. The minimum required
   interaction score was set to medium confidence (0.400), disconnected
   nodes were hidden and the thickness of the lines indicates the strength
   of data support. Abbreviations: FDR, false discovery rate; GO, gene
   ontology; ID, identification.

Functional Analyses Confirm Transcriptional Findings

   From the transcriptional profiling as described above it became clear
   that gene expression in myeloid DC subsets, and to a lesser extent in
   slan+ non-classical monocytes, was influenced by the compartment in
   which they reside. For functional validation of these findings, the
   expression of the Ki-67 protein, an intracellular marker for cell
   proliferation, was evaluated in a paired analysis of eight matched PB
   and BM samples. Consistent with our transcriptional analyses, for all
   subsets the percentage of Ki-67-positive cells was higher in BM than in
   PB and overall much higher in the DC subsets than in the slan+
   non-classical monocytes. This was consistent for all donors (Figure
   [106]6A). Furthermore, DC-like functions, such as maturation capacity
   upon TLR-ligation, were tested for PB- and BM-derived cell subsets.
   Median fluorescence intensity (MFI) of the co-stimulatory molecules
   CD80 and CD86, and, of HLA-DR at baseline and after overnight culture
   in the presence or absence of LPS and R848 (a combination of TLR
   ligands that has synergistic effects on cytokine production) ([107]30,
   [108]39), were compared and fold change values were calculated (Figure
   [109]6B). Due to low cell yield, cDC1 could not be tested. The
   up-regulation of CD80 and CD86 was lower in BM-derived cDC2 as compared
   to their PB equivalents. In contrast, slan+ non-classical monocytes
   from BM and PB displayed a similar ability to enhance the expression
   levels of these molecules.

Figure 6.

   [110]Figure 6

   [111]Figure 6
   [112]Open in a new tab

   Functional validation of transcriptional findings in myeloid DC and
   slan+ non-classical monocytes. (A) Ki-67 expression in peripheral blood
   (PB) and bone marrow (BM) subsets. A representative plot of one donor
   and percentages of positive cells of eight matched samples are shown. A
   Wilcoxon matched-pairs signed rank test was used to compare the
   compartments. (B) Maturation capacity upon TLR-stimulation. Isolated PB
   or BM cDC2 and slan+ monocytes were either left unstimulated or were
   stimulated overnight with LPS + R848. Expression levels of CD80, CD86,
   and HLA-DR were assessed by flow cytometry. The fold change (FC) in
   median fluorescence intensity (MFI) values at baseline and after
   overnight culture (unstimulated or stimulated condition) was
   calculated. Median FC values for 6–10 PB and 3–4 BM (cDC2) or 8–12 PB
   and 3–4 BM (slan+) experiments are shown. (C) Cytokine secretion assay.
   Culture supernatants of PB-derived cDC1 (n = 3), cDC2, and slan+
   (n = 6) cells and BM-derived cDC1 (n = 2), cDC2, and slan+ (n = 4)
   cells, unstimulated and stimulated (cDC2 and slan+: LPS + R848, cDC1:
   Poly I:C + R848), were analyzed for the presence of different cytokines
   by cytometric bead array (CBA, pg, picogram). (D) Allogeneic mixed
   leukocyte reaction (MLR). CFSE-labeled peripheral blood lymphocytes
   (PBL) were co-cultured with PB or BM-derived cDC2 or slan+ monocytes.
   The percentage of CFSE-diluted T cells was determined by using flow
   cytometry. Median values of four different experiments are shown.
   *p < 0.05, **p < 0.01, and ***p < 0.001.

   Next, the cytokine profile of slan+ non-classical monocytes and myeloid
   DC from the PB and BM compartment was assessed. Cells were stimulated
   overnight and by using CBA kits different cytokines were analyzed in
   culture supernatants. For stimulated cDC2 a remarkable difference was
   found between the PB and BM compartment (Figure [113]6C).
   Pro-inflammatory cytokines, in particular the Th-1-polarizing cytokine
   IL-12p70, were secreted in very low amounts by BM DC as compared to the
   corresponding subset in PB. A same trend was seen for cDC1 in
   stimulated conditions. Slan+ non-classical monocytes secreted
   negligible levels of IL-12p70 in PB and BM. In both compartments, they
   mainly produced IL-1β, IL-6, IL-8, and TNF, without significant
   differences between these two sites. Cell survival percentages in these
   cultures could not be an underlying explanation for differences in
   cytokine secretion, since cultures showed equal live cell percentages
   for PB and BM-derived subsets (Figure S3 in Supplementary Material). To
   test the T cell priming and proliferation induction capacity,
   stimulated cDC2 and slan+ non-classical monocytes from both
   compartments were used in an allogeneic MLR. Again, cDC1 could not be
   used because of low cell numbers. In agreement with the transcriptional
   results, PB-derived cDC2 showed a superior induction capacity of T cell
   proliferation in CD4+ T cells as well as in CD8+ T cells compared to
   the BM-derived cDC2 (Figure [114]6D; Figure S4 in Supplementary
   Material). In contrast, but also confirming transcriptional data, slan+
   non-classical monocytes did not show a significant difference between
   the compartments.

Discussion

   Dendritic cell subsets arise in the BM and leave this compartment in a
   more differentiated, but immature form, in order to supply the
   peripheral DC pool. The entire human DC development pathway has
   recently been elucidated and DC progenitors can now be
   immunophenotypically identified ([115]40). The direct progeny of the
   circulating DC progenitors (pre-CDC) consists of the well described
   CD1c+ (cDC2) and CD141+ (cDC1) myeloid DC subsets. These two subsets
   have been widely studied in a healthy and diseased context, in PB and
   secondary lymphoid organs such as tonsils, lymph nodes, and spleen
   ([116]35, [117]41–[118]44). Their exact developmental state and
   immunological role in the human BM compartment, however, have not been
   explored before. Therefore, we carried out a genomic as well as a
   functional study in which we compared both DC subsets and the slan+
   non-classical monocytes from BM to their PB counterparts.

   The frequency of different subsets in the BM and PB already showed
   compartment-related differences. cDC1 were more frequent in the BM
   compartment than in the PB, whereas slan+ non-classical monocyte levels
   were clearly higher in PB compared to BM. Initially, slan+
   non-classical monocytes were considered to belong to the DC lineage
   ([119]38, [120]45), but recent insights firmly place them in the
   monocyte population (also confirmed in our previous study) ([121]30,
   [122]32). The immunophenotypic analysis performed in our current study
   also supports this reallocation for BM-derived slan+ non-classical
   monocytes. They showed significantly lower expression levels of HLA-DR
   as compared to the myeloid DC subsets and similar levels as compared to
   their PB counterparts and CD14+ classical monocytes. Additional
   evidence is provided by our transcriptional findings. BM-derived slan+
   non-classical monocytes clearly clustered away from the cDC2 and cDC1
   and showed the most distinct gene signature with the highest number of
   DEGs. These differentially expressed transcripts could be translated
   into pathways that were related to involvement in innate immune
   responses and a low capacity for DC-like functionalities such as
   antigen processing and presentation, comparable to PB findings
   ([123]30). For BM DC subsets, over-expressed genes confirmed their
   lineage commitment and a consistent subset-specific gene expression
   profile between PB and BM. However, beside these conserved transcripts,
   large differences in gene expression were found between the two
   compartments. In a hierarchical clustering, DC from one compartment
   clustered closer together than that they did with their equivalent in
   the other compartment, indicating a location-dependent transcriptional
   profile. Both cDC2 and cDC1 showed similar enrichment patterns in BM
   and PB. In BM, a network of cell cycle and differentiation transcripts
   was found to be over-expressed, whereas in PB, immune-related effector
   genes were overrepresented. The development of DC is highly dependent
   on Fms-related tyrosine kinase 3 ligand ([124]46–[125]48). Its
   receptor, Flt3 (CD135), is expressed in all stages, but highest levels
   can be found in the most immature forms. From our transcriptional data,
   both subsets showed over-expression of FLT3 in the BM compartment,
   thereby supporting the hypothesis that BM-derived cDC2 and cDC1 are in
   a more immature and differentiating state than their PB equivalents.
   Validation of the transcriptional findings by evaluating the expression
   levels of the proliferation marker Ki-67, further confirmed the
   FLT3-responsive and proliferative state of BM DC as compared to PB DC.
   From a functional perspective, a similar conclusion can be drawn. DC
   from the BM compartment showed a lower activity in DC-like
   functionalities as compared to matched DC from PB. Upon TLR-triggering,
   using a known synergistic combination of LPS and R848, BM-derived cDC2
   were less able to up-regulate co-stimulatory molecules (in particular
   CD80), which was reflected in their low capacity for T cell priming.
   The induction of CD4+ and CD8+ T cell proliferation was substantially
   lower for BM cDC2 than for PB cDC2. Additionally, the ability of BM DC
   to secrete pro-inflammatory cytokines was inferior to PB DC. In
   particular, the production of IL-12p70, a cytokine specifically
   secreted by highly specialized DC subsets, was nearly absent in BM
   subsets. For slan+ non-classical monocytes the differences between the
   two compartments was less evident. Their non-DC-like character was
   confirmed in both compartments, suggesting minimal effects of the
   location and a more stable and differentiated state of this cell type.

   In conclusion, this study provides the first transcriptional and
   functional findings for myeloid DC subsets in the human BM. In this
   compartment, DC harbor more immature and less differentiated features
   and are not equipped yet to fully exert DC-like functions. These data
   reveal the BM primarily as a DC developmental location rather than an
   effector site. Slan+ non-classical monocytes, on the other hand, seem
   to enter the BM as effectors from the PB in accordance to their more
   differentiated state and patrolling immune function. In several
   BM-derived diseases, such as certain hematological malignancies, DC
   function, and frequency is disturbed potentially causing altered immune
   responses that may be involved in the pathogenesis of the disease.
   However, most of these studies focused on PB-derived DC, and only some
   studied the equivalent forms of DC in the BM compartment
   ([126]49–[127]57). Since disease development in these disorders
   probably arises in the BM compartment itself, it is of high interest to
   study altered immune processes in this compartment rather than only in
   the PB. Results from this study on healthy BM DC may thus serve as a
   baseline for further immunological research on aberrant DC function in
   such disorders.

Ethics Statement

   The study was approved by the local ethical committee. All subjects
   gave written informed consent in accordance with the Declaration of
   Helsinki.

Author Contributions

   Contribution: NL-K and KL performed experiments and analyzed data.
   NL-K, KL, TW, HB, SK, TG, ML, and AL interpreted data. NL-K, KL, TG,
   ML, and AL designed research. NL-K wrote the paper.

Conflict of Interest Statement

   The authors declare that the research was conducted in the absence of
   any commercial or financial relationships that could be construed as a
   potential conflict of interest.

Supplementary Material

   The Supplementary Material for this article can be found online at
   [128]https://www.frontiersin.org/articles/10.3389/fimmu.2018.01619/full
   #supplementary-material.
   [129]Click here for additional data file.^ (2.2MB, PDF)
   [130]Click here for additional data file.^ (1,015.3KB, PDF)
   [131]Click here for additional data file.^ (1.8MB, PDF)
   [132]Click here for additional data file.^ (1.1MB, PDF)

References