Neonatal osteomacs and bone marrow macrophages differ in phenotypic marker expression and function

Abstract

Osteomacs (OM) are specialized bone-resident macrophages that are a component of the hematopoietic niche and support bone formation. Also located in the niche, are a second subset of macrophages, namely bone marrow-derived macrophages (BM Mφ). We previously reported that a subpopulation of OM co-express both CD166 and CSF1R (the receptor for MCSF), and that OM form more bone-resorbing osteoclasts than BM Mφ. Reported here are single cell qRT-PCR, CyTOF and marker-specific functional studies that further identify differences between OM and BM Mφ from neonatal C57Bl/6 mice. Although OM express higher levels of CSF1R and MCSF, they do not respond to MCSF-induced proliferation, in contrast to BM Mφ. Moreover, RANKL, without the addition of MCSF, was sufficient to induce osteoclast formation in OM, but not BM Mφ cultures. OM express higher levels of CD166 than BM Mφ, and we found that osteoclast formation by CD166^−/− OM was reduced compared to wild-type (WT) OM, whereas CD166^−/− BM Mφ showed enhanced osteoclast formation. CD110/c-Mpl, the receptor for thrombopoietin (TPO) was also higher in OMs, but TPO did not alter OM-derived osteoclast formation, whereas TPO stimulated BM Mφ osteoclast formation. CyTOF analyses demonstrated OM uniquely co-express CD86 and CD206, markers of M1 and M2 polarized macrophages, respectively. OM performed equivalent phagocytosis in response to LPS or IL-4/Il-10 which induce polarization to M1 and M2 subtypes, respectively, whereas BM Mφ were less competent at phagocytosis when polarized to the M2 subtype. Moreover, in contrast to BM Mφ, LPS treatment of OM led to the upregulation of CD80, an M1 marker, as well as IL-10 and IL-6, known anti-inflammatory cytokines. Overall, these data reveal that OM and BM Mφ are distinct sub-groups of macrophages, whose phenotypic and functional differences in proliferation, phagocytosis and osteoclast formation may contribute physiological specificity during health and disease.

INTRODUCTION

Macrophages are phagocytic immune cells known for their heterogeneity amongst different tissues. Apart from the cartilage, resident macrophages are phenotypically identified in almost every tissue of the body. Amongst these tissue resident macrophages in the bone marrow microenvironment, are two subsets of macrophages that normally reside in close proximity, known as osteomacs (OM) and bone marrow-derived macrophages (BM Mφ). OM are bone resident macrophages found lining the endosteal bone surface, and are currently characterized as CD45⁺F4/80⁺ cells⁽¹⁾. Functionally, OM are critical for promoting bone anabolism, remodeling, repair, and regeneration^(1-9), and sustaining hematopoietic stem cells (HSC)^(10,11). BM Mφ are found in the bone marrow and are similarly characterized as CD45⁺F4/80⁺ cells, but are functionally distinct. We recently reported that OM, in conjunction with osteoblasts and megakaryocytes support hematopoietic enhancing activity⁽¹⁰⁾. Another independent group has shown that lack of OM can lead to mobilization of HSC to the peripheral blood and can affect maintenance of the niche⁽¹¹⁾. We also demonstrated that OM act as osteoclast precursors and can form bone-resorbing osteoclasts in the presence of RANKL and MCSF⁽¹⁰⁾. Moreover, Wu et. al demonstrated that TRAP⁺ osteoclasts are colocalized near F4/80⁺ OM in mouse endosteal bone⁽⁷⁾.

Since both OM and BM Mφ are considered residents of the hematopoietic niche, it is essential to phenotypically distinguish OM from BM Mφ; however, to date, clear phenotypic markers are lacking. Previously, we identified in mice a small subgroup of cells, characterized as CSF1R⁺CD166⁺, that are present in OM but not BM Mφ populations⁽¹⁰⁾. In addition, we identified that CD169, a known macrophage marker, was present on both OM and BM Mφ. Several reports identify CD169⁺ macrophages as important for promoting erythropoiesis and long-term HSC engraftment^(12-15). However, depletion of these CD169⁺ cells was shown to affect both OM and BM Mφ numbers⁽¹⁶⁾, thus casting some doubt on the extent to which BM Mφ contribute to the competence of the niche.

In the current study, we performed extensive single cell genomic, proteomic, and advanced high dimensional mass cytometry (CyTOF) analyses to further identify distinguishing criteria between murine neonatal OM and BM Mφ. We also focused our efforts on identifying functional similarities and differences between OM and BM Mφ pertaining to their proliferation, polarization, phagocytosis, and their role as osteoclast precursors.

METHODS

Preparation of OM and neonatal calvarial cells (NCC) and neonatal bone marrow (NBM)

Fresh neonatal calvarial cells (NCC) and neonatal bone marrow (NBM) from 2-3 day old neonatal pups were prepared as previously described^(17,18). Briefly, calvariae were pretreated with 4mM EDTA in PBS for 30 minutes followed by sequential collagenase digestions (200U/mL). Fractions 3 to 5 (15 minutes each digestion) were collected as fresh NCC. As previously demonstrated^(10,19), these cells are more than 95% osteoblasts or osteoblast precursors and ~5% OM. NCC were used as the source of neonatal osteomacs, which were stained with CD45 and F4/80 and then flow sorted as described below. We used NCC, rather than long bones for OM preparation because NCC provided a greater number of OM for our studies. In addition, NCC preparations were free from contaminating bone marrow derived macrophages (BM Mφ); it was not possible to fully flush out the bone marrow from the long bones of neonatal pups. We used the same 2-3 day old decapitated pups to make NBM preparations, for subsequent isolation of BM Mφ. Briefly, hind limbs were dissected, the muscle separated from the bone, and the bone was cut into small pieces to gently crush and release the marrow from within. Bone marrow was lysed of red blood cells (RBC) using RBC lysis buffer (Invitrogen), stained with CD45 and F4/80 and flow sorted to isolate BM Mφ. All mice were on a C57BL/6J background and free of pathogens. Husbandry conditions included 12 hour light/dark cycles at controlled temperatures. Dams were fed standard chow. Neonatal pups 3-5 days of age were used for experiments, and males and females were combined (not sexed). Mice were maintained, bred, and handled following protocols approved by the Indiana University Institutional Animal Care and Use Committee (IACUC), in accordance with the NIH Guidelines for the care and use of laboratory animals.

Preparation of OMs and BM Mφ from adult mice

Bone marrow was flushed from the femur and tibia of 8-12 week-old adult mice with IMDM media using 18-G needles. Flushed bone marrow was RBC lysed and used as the source of adult BM Mφ. To prepare adult OM, after removing bone marrow, the bone tubes were digested with collagenase, as previously described for NCC isolation. Bone-resident Mφ (OM) and BM Mφ were then sorted from bone-resident cells and bone marrow cells by flow cytometry.

Cell staining, flow cytometry, and cell sorting

Freshly-isolated NCC and NBM were washed with stain wash (1% bovine serum in PBS) and stained for 15 minutes at 4°C. NCC and NBM were stained with CD45 and F4/80 (Biolegend) and sorted into CD45⁺F4/80⁺ OM or BM Mφ, respectively (SORP, FACSAria or FacsFusion, BD Biosciences). For CD110 staining, a biotinylated monoclonal antibody was purchased from Takara Bio, Japan. Streptavidin conjugated to Pacific Blue (Biolegend) was used as the secondary antibody. Cells were acquired on a FACS X20 (BD Biosciences) cell analyzer, and data were analyzed with FlowJo (BD Biosciences).

Single cell qRT-PCR

The procedure for data collection for single cell qRT-PCR has been previously described⁽²⁰⁾. Briefly, the C1 Single-Cell Auto prep System (Fluidigm) was used to capture single cells and prepare cDNA followed by fast gene expression analysis using EvaGreen on the BioMark HD System (Fluidigm). Instructions for the same are provided online (PN 100-4904 I1). Based on cell sizing, we primed a C1 IFC for PreAmp 10-17μm chip for OM and a 5-10μm chip for BM Mφ. Cells were resuspended at a concentration of 200,000 cells/ml and loaded onto the primed IFC which was run using the STA: Cell Load script to capture single cells. Captured single cells were reverse transcribed into cDNA and pre-amplified and the product was subjected to gene expression analysis. The BioMark HD system (protocol 68000088 K1) was used to perform 90 individual qRT-PCR reactions on the harvested product from every single cell captured on the IFC. The harvested product was loaded onto a primed 96.96 Dynamic array IFC which was prepared according to manufacturer’s instructions and then transferred to the BioMark HD where individual gene expression was quantified.

Singular Analysis Toolset on software R (version 3.0.2) available on the Fluidigm website was used to analyze the data files generated. Principal component analysis plots were used to determine outliers. Differentially expressed genes were identified between OM and BM Mφ using one-way ANOVA. Violin plots were created to further compare the two cell types.

Mass cytometry (CyTOF)

Antibodies used in this study are listed in Supplemental Table 1. All pre-conjugated and custom conjugated antibodies were purchased from DVS sciences, Fluidigm. Each antibody was titrated to determine the optimal concentration to label NCC and NBM. Briefly, the two cell sources were treated with 10ug/ml Brefeldin A (Sigma-Aldrich, St. Louis, MO) for 3 hours to inhibit protein transport. Following stimulation, 3.5 million cells were resuspended in PBS and stained with Cell-ID Cisplatin viability stain (DVS Sciences, Fluidigm). Cells were washed using stain wash (0.1% BSA, 0.1% Na-Azide, 10nM EDTA in PBS) and blocked using Fc-Receptor block (Biolegend). Cells were stained with extracellular metal-labeled antibodies at 4°C for 30 minutes, followed by a wash, fixation using 1.5% formaldehyde for 30 minutes and 2 washes with Maxpar Perm-S Buffer (DVS Sciences, Fluidigm). Next, cells were stained with intracellular antibodies for 30 minutes. Cells were washed and incubated overnight in 1:1000 Cell-ID Intercalator-Ir diluted in Maxpar Fix and Perm buffer (DVS, Sciences, Fluidigm). Samples were resuspended in 1X EQ Calibration beads (DVS, Sciences, Fluidigm) and acquired on a CyTOF 2 mass cytometer (DVS, Sciences, Fluidigm). The bead signature was used to normalize raw CyTOF data before analysis on Cytobank software. Raw data were gated on singlet viable cells based on DNA labeling with iridium (Ir191/193), event length and cisplatin (Pt195). NCC and NBM were then gated on CD45⁺F4/80⁺ cells to identify OM and BM Mφ, respectively. These gated cells were used to make viSNE plots and heatmaps.

Cell proliferation assays

CellTiter 96^® AQueous Non-Radioactive Cell Proliferation Assay (#G5421) from Promega was used, following instructions provided with the kit. Briefly, 25,000 OM or BM Mφ were cultured in 96-well plates for 1-2 days. 10μl of the phenazine methosulfate (PMS) was added to 2 mL of the 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium) solution (MTS). 20μl of this final solution was added to each well. Cultures were then incubated at 37°C for 3 hours. The amount of MTS reduced to a soluble formazan product was measured at 490nm absorbance using a BioTek Cytation 5 imaging reader. The quantity of formazan product is equivalent to the number of living cells in culture. In parallel studies, live cells were trypsinized, stained for trypan blue and manually counted using a hemocytometer.

Osteoclast differentiation

Osteoclasts were prepared as previously defined⁽¹⁰⁾. Flow sorted cells were plated at a density of 7,500-15,000 cells in 96-well plates. Neonatal OM and BM Mφ were differentiated in vitro using macrophage colony-stimulating factor (MCSF, 20 ng/ml) and receptor activator of NFκB ligand (RANKL, 80 ng/ml) (Peprotech) in αMEM plus 10% FBS for 5-7 days. Cells were fixed in 4% paraformaldehyde in PBS and stained with tartrate-resistant acid phosphatase (TRAP) according to manufacturer’s protocol (Sigma). Multinucleated, TRAP-stained cells with three or more nuclei were counted as osteoclasts. Cells were imaged with a Leica DMI4000B inverted microscope with Retiga EXi digital camera, and ImagePro software.

Phagocytosis assay

Phagocytosis assay was performed using pHrodo^™ Red BioParticles^® (Life Technologies). Briefly, 75,000-100,000 sorted neonatal OM and BM Mφ were plated in a 96-well plate and cultured overnight in complete αMEM medium (plus 10ng/ml of MCSF). Alternatively, phagocytosis was performed after treatment with lipopolysaccharide (LPS) endotoxin (for M1 polarization) or with a combination of interleukin-4 (IL-4) and IL-10 (for M2 polarization). After treatment for 24 or 48 hours, the supernatant from the wells was removed, and replaced with pHrodo^™ Red BioParticles^® resuspended in live imaging solution (Thermo Scientific). Culture plates were incubated at 37°C for 1–2 hours. Cells were collected and analyzed using ImageStream (Amnis) to collect images of phagocytosed particles.

RNA extraction and QPCR analysis

Total RNA was extracted from neonatal OM and BM Mφ using the RNeasy kit (Qiagen, Hilden, Germany) and used as a template for the synthesis of cDNA utilizing the Transcriptor First Strand cDNA Synthesis Kit (Roche, Basel, Switzerland). Based on previously publications, primer pairs (Supplemental Table 2) specific for Cd80⁽²¹⁾ Cd86⁽²¹⁾, Arg1⁽²²⁾ and Gapdh⁽²³⁾ were custom-ordered from Integrated DNA Technologies (Coralville, IA). The Fast SYBR^™ Green Master Mix (Thermo Fisher Scientific), together with primers, was added to each sample for the PCR reaction. Gene-specific primers and Taqman probes for IL-6, IL-10 and GAPDH were purchased from Thermo Fisher Scientific. These primer and probe sets, together with samples, were added to Taqman Fast Advance master mix for PCR reaction. All samples were run on the QuantStudio 6 Flex Real-Time PCR System (ThermoFisher Scientific). The values were calculated as 2^ΔCt (GAPDH). All samples were analyzed in duplicate.

Statistical analysis

Data are presented as the mean ± standard deviation unless otherwise indicated. All experiments, except the single cell qPCR, were performed at least 3 independent times. Statistical differences were determined using One-way Anova with Tukey Kramer or student’s t-test post hoc analysis. P-value was set as 0.05.

RESULTS

Multidimensional single cell analyses identify several distinguishing markers between OM and BM Mφ

Global differences between neonatal OM and BM Mφ were examined using the BioMark HD microfluidic platform. We analyzed 90 total genes (Fig 1), 48 of which were significantly different between OM versus BM Mφ. Of these genes, 40 were expressed at higher levels, while 8 genes were expressed at lower levels in OM compared to BM Mφ. For example, violin plots (Fig 1) demonstrate minimal adiponectin (ADIPOQ) mRNA expression in OM, whereas ADIPOQ expression was significantly higher in BM Mφ. On the other hand, expression of the thrombopoietin receptor, CD110 (also known as c-Mpl) was significantly higher in OM with minimal to no mRNA expression in BM Mφ. Other genes which were significantly elevated in OM include colony stimulating factor-1 receptor (CSF1R/MCSF1R) and colony stimulating factor 1 (CSF1/MCSF) amongst others. Raw data for the different genes examined, including genes that were upregulated, down-regulated or unchanged in OM versus BM Mφ are shown in Supplemental Table 3.

Figure 1. Single cell genomics identifies distinguishing genes between neonatal OM and BM Mφ.

70 freshly isolated single OM and 96 single BM Mφ were captured using the C1 integrated fluidic circuit. cDNA from these cells was analyzed by qPCR with the primers of 90 genes using the BioMArk HD platform. Data is represented using violin plots. Red asterisk (*) indicates significantly higher expression whereas blue asterisk (*) indicates significantly lower expression of that gene in OM versus BM Mφ (p<0.05). Only genes that were significantly upregulated or downregulated in OM versus BM Mφ are shown. The average mRNA expression of all 90 genes as well as their p-value is provided in Supplemental Table 3.

In order to determine proteomic differences between neonatal OM and BM Mφ, we performed single cell mass cytometry (CyTOF), which detects both surface and intracellular proteins. OM and BM Mφ were analyzed using a panel of 17 surface plus 13 intracellular antibodies (Supplemental Table 1). The panel of 30 antibodies was custom designed based on our single cell genomics data (Fig 1) and previous literature pertaining to macrophage expression of surface and intracellular markers associated with hematopoiesis regulation. In comparison to BM Mφ, OM showed significantly elevated expression of several proteins, including CD86, CD206, platelet factor-4 (PF-4), stromal derived factor-1 (SDF-1), and platelet derived growth factor-β (PDGF-β), among others, as shown in Supplemental Fig 1.

To identify unbiased subpopulations/clusters in OM versus BM Mφ, we next performed FlowSOM, which analyzes Flow cytometry data using self-organizing maps (SOM) to identify and visualize marker clustering in cells. In our hands, the staining for some intracellular markers reduced the expression of other key surface markers, possibly due to the permeabilization step involved. As a result, FlowSOM analysis of OM and BM Mφ was performed using a total of 17 surface markers (Fig 2A-B). Initial analysis of these data revealed the heterogeneity of OM and BM Mφ populations (Fig 2A). FlowSOM identified 7 unbiased subpopulations/clusters in OM and 6 in BM Mφ, based on the 17 surface antibodies (Fig 2A). Each cluster was then phenotypically characterized based on the same antibody panel. Only 1 small cluster overlapped between OM and BM Mφ (teal-colored cell cluster, blue arrow, Fig 2A), indicating that the two cell populations are phenotypically different.

Figure 2. Identification of proteomic differences between neonatal OM and BM Mφ.

Single cell suspensions of NCC and NBM were analyzed using a subpanel of 17 surface antibodies. Data were gated on CD45⁺F4/80⁺ cells to identify OM in NCC (left panel) and BM Mφ in NBM (right panel). A) Representation of the heterogeneity observed within subpopulations of OM and BM Mφ. Each subpopulation was characterized depending on their expression of the surface markers. The teal-colored cell cluster (blue arrow) indicates that the two cell populations are phenotypically different. B) Heat map demonstrating the expression of 17 antibodies on OM and BM Mφ, N=3-6. Blue arrows indicate genes that were subsequently analyzed for functional change. C) ViSNE plots indicate surface marker differences between neonatal OM and BM Mφ. D) Flow cytometric analysis (representative data) of freshly isolated neonatal OM and BM Mφ using CD45, F4/80, and CD110 antibodies (N=3).

We generated a heat map of all our surface proteins, which further confirmed phenotypic differences between OM and BM Mφ (Fig 2B). Consistent with previous characterization⁽²⁾, OM were Mac-2 low and expressed macrophage-associated markers such as CSF1R/MCSF1R, CD11b, CD14, CD166 and CD169 (Fig 2B). In fact, expression of CSF1R, CD166, CD86 and CD206 were significantly higher in OM compared to BM Mφ (Fig 2C). CD166 (ALCAM) is a marker important for the hematopoietic niche^(24-29), and our previous studies using both neonatal and adult derived OM and BM Mφ identified a small subgroup of OM that co-express both CSF1R and CD166, whereas this subgroup is not present in BM Mφ⁽¹⁰⁾. The unique CSF1R⁺CD166⁺ population was detected in both neonatal and adult OM and is at best 5% (if not less) of total OM (CD45⁺F4/80⁺ cells)⁽¹⁰⁾. Consistent with our earlier work using adult cells⁽¹⁰⁾, CSF1R mRNA and MCSF mRNA expression were significantly elevated in neonatal OM, compared to neonatal BM Mφ (Fig 1, 2B & Supplemental Table 3). Although mRNA expression of CD166 was lower in neonatal OM compared to neonatal BM Mφ (Fig 1, Supplementary Table 3), the protein level of CD166 (ALCAM) was higher in OM than BM Mφ, (Fig 2B and 2C)⁽¹⁰⁾, indicating differences in the regulation of CD166 at the translational level. As discussed below, CD166 differentially affects the ability of OM versus BM Mφ to form osteoclasts. In a separate publication, we also describe the importance of CD166 on OM function in the HSC niche^(25,29,30).

We analyzed neonatal OM and BM Mφ for the expression of CD86 (M1 macrophage marker) and CD206 (M2 macrophage marker), and both were found to be significantly elevated in OM, compared with BM Mφ (Fig 2C). Of interest, neonatal OM uniquely co-express both CD86 and CD206 (Fig 2B, 2C), whereas neonatal BM Mφ express some CD86 with low or undetectable CD206. While the focus of the current studies is on neonatal macrophages, we also examined the expression of CD86 and CD206 in 8-10 week-old adult mice (Supplemental Fig 2). Compared to neonatal OM, CD206 expression was significantly reduced in adult OM, although it was expressed on 95% of the sorted adult OM population. These findings indicate that that CD86⁺ CD206⁺ co-expression in neonatal OM and not neonatal BM Mφ, suggest these markers are unique distinguishing features of OM in the early growth and development of mice.

Previously, we reported CD110 (c-Mpl) is expressed in osteoclast progenitors, and is important in regulating skeletal homeostasis⁽³¹⁾. Given our single cell mRNA data that CD110 was more highly expressed in OM than BM Mφ (Fig 1), we sought to perform CyTOF analysis of CD110 in OM versus BM Mφ. However, the CD110 antibody failed to conjugate with the heavy metal, precluding CyTOF analyses. As an alternative, we performed flow cytometry to determine differences in CD110 expression in neonatal OM compared to neonatal BM Mφ. CD45⁺F4/80⁺ OM and BM Mφ were gated and analyzed for the expression of CD110 (Fig 2D). CD110 was found to be highly expressed in neonatal OM but not neonatal BM Mφ, revealing it to be a distinct surface marker on OM. Altogether, our single cell data identified the unique expression of CD166, CD110, and CD206 on neonatal OM. However, none of the identified markers were expressed on 100% of OM (or not expressed by some other cell type), precluding their use as exclusive selection markers. Nevertheless, as described below, OM and BM Mφ exhibit several different functional attributes.

OM proliferation is not enhanced by MCSF, in contrast to BM Mφ

The cytokine MCSF binds to its receptor CSF1R (MCSFR) to promote monocyte/macrophage proliferation⁽³²⁾. Since both OM and BM Mφ are subsets of macrophages that express CSF1R (Fig 1, 2A), we examined cell proliferation in freshly isolated cells and following MCSF-stimulation for 24 and 48 hours (Fig 3) using the MTS assay and cell counting. We demonstrated that freshly isolated (unstimulated) OM show a higher basal proliferation than BM Mφ after 24 hours (Fig 3A). However, OM proliferation was not further augmented by extrinsic MCSF stimulation for 24 or 48 hours (Fig 3A, 3B). On the other hand, in the presence of MCSF, an increase in BM Mφ proliferation was observed after 24 hours, and to a greater extent after 48 hours (Fig 3B). When compared to OM, BM Mφ showed higher overall proliferation after 24 and 48 hours in culture with MCSF. These findings suggest that the higher endogenous expression of MCSF found in OM versus BM Mφ (Fig 1, Supplemental Table 3), potentially dampens the MCSF-induced proliferative response of OMs.

Figure 3. OM exhibit increased basal but not MCSF-stimulated proliferation compared with BM Mφ.

Neonatal OM and BM Mφ were isolated using flow cytometry. 25,000 cells were plated in 96-well plates in the presence or absence of 20ng/ml MCSF. Cells growth was measured using an MTS proliferation assay or by direct counting of live cells using trypan blue in cultures. A) 24h, B) 48h cultured cells. Absorbance was measured at 490nm. Statistical analysis was performed using one-way ANOVA, N=3.

Functional importance of CD166 and CD110 in osteoclast formation by OM verus BM Mφ.

It is well established that monocyte/macrophage proliferation by MCSF preceeds osteoclast differentiation and maturation, and in culture both MCSF and RANKL are required to induce osteoclast formation⁽³³⁾. As previously reported⁽¹⁰⁾, and demonstrated in Fig 4B, neonatal OM show enhanced ability to differentiate into multinucleated bone resorbing osteoclasts, compared to BM Mφ⁽¹⁰⁾. Given neonatal OM endogenously express both MCSF and CSF1R, we examined if OMs could form osteoclasts when supplemented with only RANKL, without the addition of MCSF. As shown in Supplemental Fig 3, OM were able to form osteoclasts when cultured with RANKL only (no added MCSF), albeit fewer osteoclasts were formed (compare with Fig 4B). In contrast, no osteoclasts were formed when BM mφ were cultured with RANKL only (no added MCSF). These findings support a role for MCSF/CSF1R autocrine (and/or paracrine) signaling, leading to osteoclast formation in RANKL-stimulated OM cultures.

Figure 4. CD166^−/− OM and BM Mφ have different capacities to form osteoclasts.

Neonatal OM and BM Mφ were isolated by flow cytometry and cultured in the presence of MCSF and RANKL to induce osteoclast formation. After 5-7 days, mature multinucleated TRAP-stained osteoclasts were enumerated. WT and CD166^−/− neonatal mice were used as a source of OM and BM Mφ. A) Representative micrographs of a single experiment, with scale bar indicating 100μm. B) Representative data showing differences in osteoclast (OC) number between OM and BM Mφ from WT and CD166^−/− neonates. C) Effects of OM and BM Mφ plating density on osteoclast formation. 7x10³ (7K) and 15x10³ (15K) cells/well were plated in 96-well plates. Data are expressed as a percentage relative to the respective WT controls for OM and BM Mφ, and represent the average of 5-7 replicate experiments. Statistical analysis was performed using students t-Test or one-way ANOVA.

Based on our findings that OM express higher levels of CD166 protein (Fig 2B, 2C)⁽¹⁰⁾, we examined the role of CD166 in regulating osteclast formation in OM verus BM Mφ (Fig 4). Interestingly, OM derived from CD166^−/− neonates showed signficantly reduced osteoclast formation compared to OM from WT mice at two different cell plating densities (Fig 4). In contrast, BM Mφ from CD166^−/− neonates showed enhanced ability to form osteoclasts, which was evident at the higher plating density (with the lower plating density showing a similar trend). Together, these data suggest differences in CD166 signaling in OM verus BM Mφ populations, which affects the extent to which these cells can differentiate into osteoclasts in vitro.

We previously showed that thrombopoietin (TPO), which is the ligand for CD110 (c-Mpl), stimulates osteoclast formation of BM Mφ^(34,35). However, the effect of TPO on OM function has not been explored. Given CD110 expression was signficantly higher in neonatal OM versus neonatal BM Mφ (Fig 1 and 2), we used a similar osteoclastogenesis assay to examine if TPO differentially affects the ability of OM or BM Mφ to form osteoclasts. We cultured neonatal OM and neonatal BM Mφ in osteoclast media (containing MCSF and RANKL) supplemented with 100 ng/ml TPO until mature osteoclasts were formed (Fig 5). The TPO concentration was based on our previous studies demonstrating increased osteoclast formation in BM Mφ^(34,35). In agreement with our published report⁽³⁴⁾, TPO led to an increase in osteoclast formation by neonatal BM Mφ at two different plating densities (Fig 5B, 5C). However, no significant change in osteoclast formation was observed in OM cultured with TPO at either plating density (Fig 5B, 5C). These data indicate CD110 does not not play a major role in regulating osteoclast formation by OM. However, given that CD110 is more highly expressed in OM than BM Mφ (Fig 1, 2D and Supplemental Table 3), it is possible CD166 exerts other, yet-to-be identified, functions in OM.

Figure 5. OM express higher levels of CD110 but are not responsive to TPO-induced osteoclast formation.

Neonatal OM and BM Mφ were cultured in the presence of MCSF and RANKL, in the presence or absence of 100 ng/ml TPO. After 5-7 days, mature multinucleated TRAP-stained osteoclasts were enumerated. A) Representative micrographs of a single experiment. Scale bar indicates 100μm. B) Representative data of a single experiment showing differences in osteoclast (OC) number in OM versus BM Mφ treated without TPO (control, Ctrl) or with TPO. C) Effects of cell plating density (7x10³ and 15x10³ cells/well in 96-well plate) on osteoclast formation by OM and BM Mφ treated with TPO or with vehicle (Ctrl). Data are expressed as a percentage of the respective controls for OM and BM Mφ, and represent the average of 3-5 replicate experiments. Statistical analysis was performed using students t-Test or one-way ANOVA.

OM and BM Mφ differ in M1 versus M2 polarized phagocytosis

Phagocytosis by macrophage subtypes is required for immune function⁽³⁶⁾. To determine whether OM can phagocytose similar to BM Mφ, phagocytosis assays were performed uisng PE-conjugated E.coli BioParticles. Flow cytometry results indicate that both OM and BM Mφ fluoresce in the PE emission channel (Fig 6A-B). Images of these cells demonstrated that the BioParticles were intracellular in both OM and BM Mφ, indicating their engulfment (Fig 6A). Quantitation showed that 75% of OM and 78% of BM Mφ phagocytosed the PE-BioParticles (Fig 6B). These data suggest OM can functionally phagocytose foreign particles to a similar extent as BM Mφ.

Figure 6. Phagocytosis by OM is unchanged by M1 or M2 polarization.

OM and BM Mφ were cultured at 75x10³–100x10³ cells/well in αMEM plus 10% FBS with 20ng/ml MCSF. After 24h, the cells were incubated with 100μL of PE-conjugated E. coli BioParticles for 1h, and then analyzed for phagocytosis. A) Representative images (N=2 independent experiments) indicating both OM and BM Mφ phagocytose the BioParticles. Bright field images (top row) show that cells have engulfed the BioParticles (bottom row). B). Percentage phagocytosis of OM and BM Mφ. C-D) OM and BM Mφ were treated with either 100ng/ml of LPS or 10ng/ml IL-4 plus 10ng/ml IL-10 to induce polarization into M1 and M2 macrophages, respectively. Vehicle-treated cells were used as controls. After stimulation for 24h, the cell phagocytosis assay was performed as described above. C) Representative flow cytometry data shown in the form of dot plots for OM and BM Mφ that underwent phagocytosis. D) Percentage phagocytosis determined from flow cytometry data, normalized to untreated samples. Statistical analysis was performed using one-way ANOVA, N=3.

Next, we performed experiments to determine the importance of macrophage polarization in OM and BM Mφ phagocytosis. Conflicting data exists regarding the competence of M1 versus M2 macrophages in phagocytosis. Certain groups suggest that phagocytosis via M1 macrophages is more prominent⁽³⁷⁾, whereas others suggest vice versa^(38,39). Our CyTOF data (Fig 2) showed that neonatal OM uniquely co-express both CD86 and CD206, which are well-established markers of M1 and M2 macrophages, respectively. In contrast, BM Mφ primarily express CD86. We speculated that because OM co-express both M1 (CD86) and M2 (CD206) markers, they may possess functional attributes consistent with both M1 and M2 macrophages. To test this, we stimulated these cells with LPS for M1 polarization or IL-4 plus IL-10 for M2 polarization, and then determined changes in phagocytosis. As predicted, OM maintained their ability to phagocytose when stimulated with LPS (71 %) or IL-4/IL-10 (74%), which was similar to unstimulated (vehicle control) OM (75%) (Fig 6C-D). Although phagocytosis was not statistically different between untreated BM Mφ and LPS-polarized BM Mφ, we observed a significant reduction in BM Mφ phagocytosis after stimulation with IL-4/IL-10, which induces M2 polarization. The data indicate that OM perform equivalent phagocytosis irrespective of their M1/M2 polarity; however, BM Mφ are less competent at phagocytosis when polarized into the M2 subtype with IL-4/IL-10.

We also compared neonatal OM and BM Mφ for mRNA expression of additional M1 markers (CD86, CD80) and M2 markers (Arg1) after treatment with LPS for M1 polarization or IL-4/IL-10 for M2 polarization, respectively. Vehicle-treated control (naïve) OM expressed higher baseline levels of CD86 compared to BM Mφ, consistent with our CyTOF data (Fig 2). In contrast to CD86 (M1 marker) which was unchanged in OM treated with LPS (Fig 7A), LPS stimulation led to a significant increase in CD86 in BM Mφ, compared to naïve BM Mφ. Conversely, CD80 (M1 marker) was lower in naïve OM, compared to naïve BM Mφ (Fig 7B), and was significantly increased in LPS-stimulated OM but not BM Mφ. When polarized to the M2 subtype with IL-4/IL-10, we found that Arg1 (M2 marker) was upregulated in both OM and BM Mφ, although its expression was highest in IL-4/IL-10 stimulated BM Mφ (Fig 7C). Next, we examined if LPS altered the expression of interleukin-6 (IL-6) and IL-10, cytokines known to be involved in immune and anti-inflammatory responses^(40-43). Interestingly, OM showed a robust, LPS-induced increase in the expression of IL-6 (Fig 7D) and IL-10 (Fig 7E), while their expression in BM Mφ remained barely detectable with or without LPS. Taken together, these findings further highlight phenotypic differences between OM versus BM Mφ in the naïve state as well as after polarization into M1 and M2 macrophages, which may affect their functional activity in immune responses.

Figure 7. M1 and M2 marker expression in naïve and polarized OM and BM Mφ.

OM and BM Mφ were cultured as described in Figure 6. Naïve cells were treated with vehicle, 100ng/ml of LPS or 10ng/ml IL-4 plus 10ng/ml IL-10 to induce polarization into M1 and M2 macrophages, respectively. QPCR analysis was performed to examine the expression of different M1 markers (CD86, CD80) as well as a M2 marker (Arg1). A) CD86, B) CD80 and, C) Arg 1. D-E) LPS-stimulated OM and BM Mφ cultures were compared for the expression of IL-6 mRNA and IL-10 mRNA. Absolute mRNA expression (2^ΔCT) normalized to GAPDH are shown. Statistical analysis were performed using ANOVA, N=3 replicate experiments.

DISCUSSION

OM are identified as F4/80⁺ macrophages that line the endosteum, whereas BM Mφ are present in the bone marrow and are also defined as CD45⁺F4/80⁺ cells. Through our previous work and curent data, we demonstrate that in addition to F4/80, OM express other classical macrophage markers such as CD11b, CD14, CD68, and Mac-2 (Fig 2). All of these markers are common between OM and BM Mφ, making it difficult to determine which subset of macrophages is essential for hematopoietic function^(10-15), bone anabolism^(1-9), or osteoclastogenesis⁽¹⁰⁾. In this manuscript, calvariae-resident neonatal OM were further characterized and shown to be phenotypically and functionally different from neonatal BM Mφ.

Previous studies have identified CD166 as a distinguishing marker between neonatal OM and BM Mφ; however, CD166 is not expressed by all OM⁽¹⁰⁾. Through our multidimensional single cell studies, we identified several phenotypic differences between OM and BM Mφ, including the expression of CD110 and CD206 (Fig 1 & 2). Interestingly, a classification of macrophages exists which bifurcates CD86⁺ macrophages as M1 subtype and CD206⁺ macrophages as M2 subtype^(44,45). Although neonatal OM co-express both CD86 and CD206 (Fig 2C), this classification cannot be used to distinguish adult OM, which show lower levels of CD206. This finding may indicate that neonatal OM are at an earlier phase of differentiation than BM Mφ, and may reflect increased selectivity/plasticity of the early hematopoietic niche. In published studies⁽¹⁰⁾ using adult OM and adult BM Mφ, we identified a subgroup of OMs that are CSF1R⁺CD166⁺, thus providing a profile unique to adult OM, compared to adult BM Mφ.

Due to the identification of CSF1R, the MCSF receptor, on both OM and BM Mφ (Fig 1&2), we examined the effect of recombinant MCSF on OM and BM Mφ proliferation. Our data demonstrated higher baseline proliferation of neonatal OM compared to unstimulated BM Mφ at 24 hours (Fig 3), and that OM proliferation was not increased by recombinant MCSF, in contrast to BM Mφ which showed an MCSF-stimulated increase in proliferation at both time points examined (Fig 3). Based on the finding that OM express both MCSF (CSF1) and its receptor, CSF1R, we speculated that increased autocrine/paracrine signaling may potentially explain why basal OM proliferation exceeded that BM Mφ, and also why OM proliferation was not further stimulated by MCSF. To test this, we examined whether the addition of recombinant MCSF was obligatory for osteoclast formation by OM versus BM Mφ. While it has long been known that BM Mφ act as precursors for osteoclast formation, we recently reported that OM can be induced to form osteoclasts in the presence of recombinant MCSF and RANKL⁽¹⁰⁾. Consistent with MCSF/CSF1R autocrine/paracrine signaling, we found that OM were able to form osteoclasts even in the absence of recombinant MCSF (RANKL only added), whereas both MCSF and RANKL were required for BM Mφ to form osteoclasts (Supplemental Fig 3).

CD110 (c-Mpl), the receptor for TPO, is known to have a role in osteoclastogenesis⁽³¹⁾. CD110 (c-Mpl) is expressed on the osteoclast lineage and CD110^−/− (knockout) mice demonstrate an overall increase in the number of osteoclasts (as well as osteoblasts)^(31,35). We also reported that unsorted macrophages from bone marrow stimulated with TPO form more osteoclasts than controls^(34,35). In the current study, we demonstrated that although neonatal OM express higher levels of CD110 than neonatal BM Mφ (Fig 1, 2D and Supplemental Table 3), OM did not undergo TPO-stimulated increase in osteoclast formation, which was observed in the BM Mφ cultures (Fig 5). However, it is possible that TPO binding to CD110 may cause other functional changes in OM that are yet-to-be determined. We also examined the role of CD166, which we found was more highly expressed in neonatal OM than neonatal BM Mφ. Interestingly, OM isolated from CD166^−/− neonates show reduced ability to form osteoclasts, compared to WT OM (Fig 4). Although the in vivo role of OM in osteoclast formation remains to be established, it is interesting to speculate that CD166^−/− OM, due to their reduced ability to form osteoclasts, may contribute to the elevated bone mineral density observed in CD166^−/−mice ⁽³⁰⁾.

M1 and M2 is a sub-classification of macrophages wherein M1 macrophages are classically activated and pro-inflammatory; whereas, M2 macrophages are alternatively activated and anti-inflammatory⁽⁴⁶⁾. Recent evidence demonstrates that M2 macrophages are more efficient at osteoclastogenesis compared to M1 macrophages⁽⁴⁷⁾, which is consistent with our previous report⁽¹⁰⁾, and validated here (Fig 4, 5), that OM form more osteoclasts than BM Mφ. We also found that OM have phagocytotic activity, similar to BM Mφ, marking them as part of the immune system (Fig 6). Although no previous data has demonstrated OM phagocytosis, there have been instances of correlation between OM and efferocytosis of apoptotic cells within bone^(8,48). Data in Fig 6 further strengthen this analogy and identify OM as possessing a classical function of macrophages, namely phagocytosis. Given OM show phenotypic attributes of both M1 and M2 macrophages, while a sub-population of BM Mφ was M1 polarized, we stimulated both macrophage subsets with LPS (for M1 polarization) and IL-4/IL-10 (for M2 polarization) and determined changes in phagocytosis. Whereas no changes in phagocytosis were observed in OM after M1 or M2 polarization (Fig 6C-D), stimulation of BM Mφ with IL-4/IL-10 caused a reduction in phagocytosis. Previous research regarding the efficiency of M1 versus M2 macrophages in phagocytosis is divided^(37-39). This descrepency may be because of an overlap in genes expressed by M1 and M2 macrophages⁽⁴⁶⁾, or due to different laboratories using different markers to define their M1 and M2 populations. Our studies demonstrate that LPS-stimulated OM show increased expression of IL-6 and IL-10, cytokines known to have peitrophic actions in inflammatory responses, hematopoiesis and metabolism^(40-43). Our curent findings are consistent with a report that calvaria-derived OM may respond to LPS and release TNF-α⁽¹⁾.

As summarized in Table 1, we identified several functional similarities and differences between OM and BM Mφ. Our previous work demonstrated that OM interact with osteoblasts to enhance hematopoietic activity, and that the hematopoietic activity of OM could not be substituted by BM Mφ⁽¹⁰⁾. Further, other laboratories have shown that OM support osteoblast mineralization, bone formation and bone healing^(1-9). Through our current single cell studies, we identified several targets implicated in regulating hematopoiesis^(49-53), such as PF-4, SDF-1, and PDGF-β, that are more highly expressed in OM than BM Mφ. Interestingly, these findings support the possibility of different functional roles between OM and BM Mφ in the hematopoietic niche, and may also explain our previous data which demonstrated that OM, but not BM Mφ, support the hematopoietic enhancing activity of HSC⁽¹⁰⁾.

Table 1.

Summary of distinguishing markers and functional differences between neonatal OM and BM Mφ.

Characteristic	OM	BM Mφ
Expression of surface markers
• CD45	++	++
• F4/80	+++	++
• CD11b	+++	+++
• CD68	+	+
• Ly6G	++	+++
• Mac-2	+	+
• CSF-1R	+++	++
• CD14	+++	++
• CD169	+++	++
• CD166	++	−
• CD110	+++	−
• CD206	+++	−
• CD86	+++	+
• CD80	+++	++
Expression of intracellular markers
• SDF-1	+++	+
• PF-4	+++	+
• PDGF-β	+++	+
Megakaryocyte induced proliferation(10)	Yes	No
Extrinsic MCSF stimulated proliferation	No	Yes
Phagocytic Capability	Yes	Yes
Precursor for osteoclast differentiation(10)	Yes	Yes
	++	+
Size (μm)	~10-15	~8-10
Interact with osteoblasts and megakaryocytes to enhance hematopoietic activity(10)	Yes	No

Our current studies demonstrate OM are functionally capable of phagocytosis (Fig 6) and also act as osteoclast progenitors (Fig 4, 5)⁽¹⁰⁾, both of which are broadly similar to BM Mφ. However, OM are phenotypically and functionally very distinct from their BM Mφ counterparts (Fig 1, 2, Table 1). In addition, an interesting trend emerged related to our OM studies in that this macrophage subtype showed elevated M1 and M2 marker expression, and exhibited increased functional activity in the naïve state but showed little or no change in marker-selective cellular functions (proliferation, osteoclastogenesis, polarization) in response to exogenous stimuli. Taken together, these data suggest OM may perform more of a homeostatic function in physiology rather than a role in the acute response to disease.

Our investigations extend the phenotype and function of neonatal OM. However, additional studies are needed to identify distinguishing marker(s) that reliably identify these macrophage subtypes throughout development. The identification of clear distinguishing markers between OM and BM Mφ is necessary to design future mouse models which would be useful for in vivo functional studies to better determine the role of OM versus BM Mφ in immunity, inflammation, hematopoiesis, and bone homeostasis.

Supplementary Material

Supplementary Materials

Supplemental Figure 1: Multidimensional CyTOF analysis of neonatal OM and BM Mφ. Single cell suspensions of NCC and NBM were analyzed using a panel of 30 surface and intracellular antibodies. Data were gated on CD45⁺F4/80⁺ cells to indicate OM in NCC (upper panel) and BM Mφ in NBM (lower panel). (A) Heat maps indicating surface and intracellular differences between the two macrophage subtypes. (B) Representative ViSNE plots of PF-4, SDF-1 and PDGF-β. N=3-7.

Supplemental Figure 2: Flow cytometric analysis of neonatal versus adult OM and BM Mφ. OM were isolated by flow cytometry from neonatal or adult calvaria, while BM Mφ were isolated from neonatal or adult bone marrow. Cells were gated on CD45⁺F4/80⁺ and analyzed for CD86 and CD206 expression (N=3). Adult OM were isolated after collagenase digestion of marrow-flushed long bones from 8 week-old mice as previously reported.⁽¹⁰⁾ Representative data is shown. FMO: fluorescence minus one.

Supplemental Figure 3: Osteoclast formation by OM is independent of recombinant MCSF. Neonatal OM and BM Mφ were isolated by flow cytometry and cultured at 26x10³ cells/well in the presence of 80ng/ml RANKL (without added MCSF). After 5-7 days, mature multinucleated TRAP-stained osteoclasts were enumerated. Representative data showing differences in osteoclast (OC) number in OM versus BM Mφ cultures. No osteoclasts were formed in BM Mφ cultures in N=2-3 replicate experiments. Statistical analysis was performed using students t-Test.

Supplemental Table 1: List of metal-conjugated CyTOF antibodies.

Supplemental Table 2: QPCR primers.

Supplemental Table 3: Average mRNA expression in neonatal OM and BM Mφ based on single cell genomics.