Advertisement

Imaging flow cytometry reveals that granulocyte colony-stimulating factor treatment causes loss of erythroblastic islands in the mouse bone marrow

  • Joshua Tay
    Affiliations
    Stem Cell and Cancer Group, Mater Research–University of Queensland Translational Research Institute, Woolloongabba, QLD, Australia
    Search for articles by this author
  • Kavita Bisht
    Affiliations
    Stem Cell Biology Group, Mater Research–University of Queensland Translational Research Institute, Woolloongabba, QLD, Australia
    Search for articles by this author
  • Crystal McGirr
    Affiliations
    Stem Cell Biology Group, Mater Research–University of Queensland Translational Research Institute, Woolloongabba, QLD, Australia
    Search for articles by this author
  • Susan M. Millard
    Affiliations
    Bones and Immunology Group, Mater Research–University of Queensland Translational Research Institute, Woolloongabba, QLD, Australia
    Search for articles by this author
  • Allison R. Pettit
    Affiliations
    Bones and Immunology Group, Mater Research–University of Queensland Translational Research Institute, Woolloongabba, QLD, Australia
    Search for articles by this author
  • Ingrid G. Winkler
    Correspondence
    Ingrid Winkler, Translational Research Institute, Mater Research–The University of Queensland, 37 Kent Street, Woolloongabba, QLD 4102, Australia
    Affiliations
    Stem Cell and Cancer Group, Mater Research–University of Queensland Translational Research Institute, Woolloongabba, QLD, Australia
    Search for articles by this author
  • Jean-Pierre Levesque
    Correspondence
    Offprint requests to: Jean-Pierre Levesque, Translational Research Institute, Mater Research–The University of Queensland, 37 Kent Street, Woolloongabba, QLD 4102, Australia
    Affiliations
    Stem Cell Biology Group, Mater Research–University of Queensland Translational Research Institute, Woolloongabba, QLD, Australia
    Search for articles by this author
Open AccessPublished:February 08, 2020DOI:https://doi.org/10.1016/j.exphem.2020.02.003

      Highlights

      • Nine-channel imaging flow cytometry can quantitate EBIs.
      • Most cell aggregates co-expressing macrophage and erythroblast markers are not EBIs.
      • EBI central macrophages are negative for CD11b and Ly6G, but CD11b+Ly6G+ granulocytes are frequently found at the periphery of EBIs.
      • G-CSF treatment in vivo leads to a profound reduction in EBI frequency in the bone marrow.
      The erythroblastic island (EBI) is a multicellular structure forming an erythropoietic niche consisting of a central macrophage surrounded by a rosette of maturing erythroblasts. Since their discovery more than 60 years ago, simultaneous quantification and visualization of EBIs remain difficult. Although flow cytometry enables high-throughput quantification of cell aggregates co-expressing macrophage and erythroblast markers, it cannot visually confirm whether the aggregates are genuine EBIs. While immunofluorescence microscopy allows visualization of EBIs, its low throughput limits its use for quantification. In the current study we employed nine-channel imaging flow cytometry (IFC) to develop a method to directly visualize and quantify EBIs in the mouse bone marrow. We found that EBI central macrophages do express F4/80, VCAM-1, and CD169, but not CD11b or Ly6G, and that CD11b+Ly6G+F4/80 granulocytes are found associated at the periphery of 40%–60% EBIs. Furthermore, we show for the first time using IFC that in vivo treatment with the hematopoietic stem cell-mobilizing cytokine granulocyte colony-stimulating factor (G-CSF) reduced EBI frequency in the bone marrow by more than 100-fold. These results indicate that mobilizing doses of G-CSF cause a collapse of EBIs in the bone marrow.
      The final stages of erythropoiesis are thought to take place in a multicellular functional unit called the erythroblastic island (EBI). EBIs were first identified in the human bone marrow (BM) in the late 1950s by Marcel Bessis [
      • Bessis M.
      L'îlot érythroblastique. Unité fonctionelle de la moelle osseuse.
      ], and later in the mouse BM [
      • Berman I.
      The ultrastructure of erythroblastic islands and reticular cells in mouse bone marrow.
      ] and spleen [
      • Sadahira Y
      • Mori M
      • Kimoto T
      Isolation and short-term culture of mouse splenic erythroblastic islands.
      ] and rat BM [
      • Mohandas N
      • Prenant M.
      Three-dimensional model of bone marrow.
      ]. EBIs were characterized as a cluster of maturing erythroblasts rosetting around a reticular cell rich in ferritin and ferrogenous micelles in their cytoplasm [
      • Bessis M.
      L'îlot érythroblastique. Unité fonctionelle de la moelle osseuse.
      ,
      • Berman I.
      The ultrastructure of erythroblastic islands and reticular cells in mouse bone marrow.
      ], later identified as a macrophage [
      • Mohandas N
      • Prenant M.
      Three-dimensional model of bone marrow.
      ,
      • Le Charpentier Y
      • Prenant M
      [Isolation of erythroblastic islands. Study by optical and scanning electron microscopy (author's transl)].
      ]. Various subsequent studies have found that maturing erythroblasts express cell adhesion molecules that interact with cell adhesion counter receptors expressed at the surface of the EBI central macrophage [
      • Jacobsen RN
      • Perkins AC
      • Levesque JP
      Macrophages and regulation of erythropoiesis.
      ] and that targeted disruption of these adhesive interactions leads to reduced frequency of EBIs and reduced erythropoiesis [
      • Lee G
      • Lo A
      • Short SA
      • et al.
      Targeted gene deletion demonstrates that the cell adhesion molecule ICAM-4 is critical for erythroblastic island formation.
      ,
      • Sadahira Y
      • Yoshino T
      • Monobe Y
      Very late activation antigen 4-vascular cell adhesion molecule 1 interaction is involved in the formation of erythroblastic islands.
      ,
      • Wang Z
      • Vogel O
      • Kuhn G
      • Gassmann M
      • Vogel J
      Decreased stability of erythroblastic islands in integrin beta3-deficient mice.
      ,
      • Wei Q
      • Boulais PE
      • Zhang D
      • Pinho S
      • Tanaka M
      • Frenette PS
      Maea expressed by macrophages, but not erythroblasts, maintains postnatal murine bone marrow erythroblastic islands.
      ]. EBI macrophages play roles in erythroblasts proliferation, maturation, and most importantly erythroblast enucleation during their terminal differentiation into reticulocytes [
      • Toda S
      • Segawa K
      • Nagata S
      MerTK-mediated engulfment of pyrenocytes by central macrophages in erythroblastic islands.
      ,
      • Klei TRL
      • Meinderts SM
      • van den Berg TK
      • van Bruggen R
      From the cradle to the grave: the role of macrophages in erythropoiesis and erythrophagocytosis.
      ]. Conditional gene deletion approaches have confirmed that EBI macrophages express VCAM-1 (which binds to integrin α4β1 expressed by erythroblasts) [
      • Ulyanova T
      • Jiang Y
      • Padilla S
      • Nakamoto B
      • Papayannopoulou T
      Combinatorial and distinct roles of α5 and α4 integrins in stress erythropoiesis in mice.
      ], Mer tyrosine kinase receptor [
      • Toda S
      • Segawa K
      • Nagata S
      MerTK-mediated engulfment of pyrenocytes by central macrophages in erythroblastic islands.
      ], and macrophage erythroblast attacher (Maea) [
      • Wei Q
      • Boulais PE
      • Zhang D
      • Pinho S
      • Tanaka M
      • Frenette PS
      Maea expressed by macrophages, but not erythroblasts, maintains postnatal murine bone marrow erythroblastic islands.
      ]. It has also emerged that EBI macrophages express the tissue-resident macrophage antigen CD169 [
      • Li W
      • Wang Y
      • Zhao H
      • et al.
      Identification and transcriptome analysis of erythroblastic island macrophages.
      ,
      • Seu KG
      • Papoin J
      • Fessler R
      • et al.
      Unraveling macrophage heterogeneity in erythroblastic islands.
      ], and specific in vivo depletion of CD169+ macrophages in mice leads to a profound reduction of medullary and splenic erythropoiesis [
      • Chow A
      • Huggins M
      • Ahmed J
      • et al.
      CD169+ macrophages provide a niche promoting erythropoiesis under homeostasis and stress.
      ,
      • Jacobsen RN
      • Forristal CE
      • Raggatt LJ
      • et al.
      Mobilization with granulocyte colony-stimulating factor blocks medullar erythropoiesis by depleting F4/80+VCAM1+CD169+ER-HR3+Ly6G+ erythroid island macrophages in the mouse.
      ]. Unfortunately none of these cell surface antigens is specific to EBI macrophages, and they cannot be used to prospectively isolate EBI macrophages to purity [
      • Li W
      • Wang Y
      • Zhao H
      • et al.
      Identification and transcriptome analysis of erythroblastic island macrophages.
      ].
      Various methods have been devised to isolate EBIs from the BM or spleen and involve suspension of cells from the tissue in buffers, sedimentation by gravity to enrich in multicellular aggregates, and plating into cultures dishes or glass slides. Although these methods enrich EBIs from cell suspensions, they do not enable quantification of the frequency of these EBIs in the BM or spleen, which precludes studies on the effect of drugs or stress on EBI numbers. Conventional flow cytometry methods have also been devised to quantify EBIs in the BM or spleen [
      • Chow A
      • Huggins M
      • Ahmed J
      • et al.
      CD169+ macrophages provide a niche promoting erythropoiesis under homeostasis and stress.
      ,
      • Jacobsen RN
      • Forristal CE
      • Raggatt LJ
      • et al.
      Mobilization with granulocyte colony-stimulating factor blocks medullar erythropoiesis by depleting F4/80+VCAM1+CD169+ER-HR3+Ly6G+ erythroid island macrophages in the mouse.
      ]. They are based on suspension of BM or spleen in buffers containing calcium and magnesium to maintain cell–cell adhesive interactions and flow cytometry analyses of cell clusters (high forward scatter width) rather than single cells. These methods are based on the assumption that cell aggregates containing both erythroid makers (such as Ter119 in the mouse) and macrophage markers (such as F4/80 in the mouse and CD169) contain EBIs that can be subsequently sorted for microscopy evaluation [
      • Chow A
      • Huggins M
      • Ahmed J
      • et al.
      CD169+ macrophages provide a niche promoting erythropoiesis under homeostasis and stress.
      ,
      • Jacobsen RN
      • Forristal CE
      • Raggatt LJ
      • et al.
      Mobilization with granulocyte colony-stimulating factor blocks medullar erythropoiesis by depleting F4/80+VCAM1+CD169+ER-HR3+Ly6G+ erythroid island macrophages in the mouse.
      ,
      • Fraser ST
      • Midwinter RG
      • Coupland LA
      • et al.
      Heme oxygenase-1 deficiency alters erythroblastic island formation, steady-state erythropoiesis and red blood cell lifespan in mice.
      ]. Although conventional flow cytometry shows that CD169+ macrophage depletion or administration of stressors such as granulocyte colony-stimulating factor (G-CSF) or FMS-like tyrosine kinase 3 ligand cause a dramatic reduction in medullary erythropoiesis and in the number of detected EBIs in the BM [
      • Chow A
      • Huggins M
      • Ahmed J
      • et al.
      CD169+ macrophages provide a niche promoting erythropoiesis under homeostasis and stress.
      ,
      • Jacobsen RN
      • Forristal CE
      • Raggatt LJ
      • et al.
      Mobilization with granulocyte colony-stimulating factor blocks medullar erythropoiesis by depleting F4/80+VCAM1+CD169+ER-HR3+Ly6G+ erythroid island macrophages in the mouse.
      ,
      • Jacobsen RN
      • Nowlan B
      • Brunck ME
      • Barbier V
      • Winkler IG
      • Levesque JP
      Fms-like tyrosine kinase 3 (Flt3) ligand depletes erythroid island macrophages and blocks medullar erythropoiesis in the mouse.
      ], this technology does not enable direct visual validation that these cell aggregates co-expressing erythroid and macrophage markers are actual EBIs.
      The relatively recent advent of imaging flow cytometry (IFC) technology, whereby an immunofluorescence micrograph is taken by a charge-coupled device (CCD) camera for each event as it passes through the flow cell, enables the direct validation of flow cytometry data with respect to cell morphology and distribution and antigen distribution within every single-cell aggregate detected by the machine. IFC has been recently adapted to detect EBIs in the mouse and rat BM in steady state [
      • Seu KG
      • Papoin J
      • Fessler R
      • et al.
      Unraveling macrophage heterogeneity in erythroblastic islands.
      ]. With a limited number of fluorescence channels, these authors confirmed that EBI macrophages express F4/80, VCAM-1, and CD169, but are devoid or have very low expression of CD11b in the mouse [
      • Li W
      • Wang Y
      • Zhao H
      • et al.
      Identification and transcriptome analysis of erythroblastic island macrophages.
      ,
      • Seu KG
      • Papoin J
      • Fessler R
      • et al.
      Unraveling macrophage heterogeneity in erythroblastic islands.
      ].
      In the current study we took advantage of more advanced 12-channel IFC machines to analyze simultaneously seven cell surface antigens in addition to the nucleus and morphological parameters, and developed a method to quantify the number of EBIs in the mouse BM before and after treatments that inhibit medullar erythropoiesis, such G-CSF administration [
      • Jacobsen RN
      • Forristal CE
      • Raggatt LJ
      • et al.
      Mobilization with granulocyte colony-stimulating factor blocks medullar erythropoiesis by depleting F4/80+VCAM1+CD169+ER-HR3+Ly6G+ erythroid island macrophages in the mouse.
      ]. With this new IFC quantification method, we were able to visualize, identify, and quantify the frequency of EBIs from the mouse BM and demonstrate that in vivo treatment with G-CSF leads to a >100-fold reduction in the frequency of EBIs in the BM and that EBI macrophages do express F4/80, VCAM-1, and CD169 but not CD11b or Ly6G. Furthermore, the presence of CD11b and Ly6G antigens in EBI-containing cell aggregates was not due to the EBI macrophages but to granulocytes found at the periphery of 40%–60% EBIs, further illustrating the difficulty of identifying EBIs and EBI macrophages by conventional flow cytometry on cell aggregates.

      Methods

      Mice and drug administration

      C57BL/6 mice were purchased from the Animal Resource Centre (Perth, Australia). All mice used were male and between 8 and 12 weeks of age. Mice were housed at The University of Queensland Biological Resources Facility at the Translational Research Institute. Procedures were approved by The University of Queensland Health Sciences Animal Ethics Committee (AEC No. 327/16). Granulocyte colony-stimulating factor (G-CSF, Filgrastim, Amgen, Thousand Oaks, CA) was administered at 125 µg/kg body weight subcutaneously twice daily for 4 days while control mice received an equivalent volume of saline.

      Preparation of EBI samples from BM

      Extraction and staining of BM aggregates were performed as described [
      • Seu KG
      • Papoin J
      • Fessler R
      • et al.
      Unraveling macrophage heterogeneity in erythroblastic islands.
      ], with minor modifications. Briefly, harvested femurs were flushed gently multiple times with a 25G needle and 1-mL syringe containing ice-cold Iscove's modified Dulbecco's medium (IMDM) supplemented with 20% fetal bovine serum (FBS), 100 U/mL penicillin, 100 μg/mL streptomycin, 2 mmol/L L-glutamine (all from Thermo Fisher Scientific, Waltham, MA) to dislodge most of the BM cells. Cell aggregates were gently pipetted through a 1-mL pipet until most BM macroscopic aggregates were no longer visible to the naked eye, and then filtered through a 70-µm cell strainer (Sigma Aldrich, St Louis, MO). Leukocytes in suspension were counted with an Ac•T Diff hematology analyzer (Beckman Coulter, Brea, CA), and a BM suspension containing 107 leukocytes was fixed with 4% paraformaldehyde in phosphate-buffered saline (PBS) for 10 min at room temperature. Cell suspensions were washed twice in PBS containing 2% newborn calf serum (NCS) and stained with 5 µmol/L Hoechst 33342 for nucleus visualization; anti-CD169-fluorescein isothiocyanate (FITC, clone 3D6.112, diluted 1:50); anti-CD71-phycoerythrin (PE, clone R17217, diluted 1:100); anti-CD11b-PE-CF594 (clone M1/70, diluted 1:100); anti-Ter119-peridinin-chlorophyll-protein (PerCP)-cyanine ((Cy)5.5, clone Ter119, diluted 1:40); anti-VCAM-1-PE-Cy7 (clone 429 [MVCAM.A], diluted 1:50); anti-F4/80-allophycocyanin (APC, clone BM8, diluted 1:150); and anti-Ly6G-APC-Cy7 (clone 1A8, diluted 1:100) antibodies (all from BioLegend, San Diego, CA) resuspended in Fc block (culture supernatant from 2.4G2 hybridoma specific for mouse CD16/CD32 produced in-house) for 1.5 hours at 37°C in the dark on a shaker. Stained cells were washed once in PBS containing 2% NCS and resuspended in 20–25 µL of the same buffer for imaging flow cytometry. Single color controls and fluorescence minus one (FMO) controls (Supplementary Figure E1, online only, available at www.exphem.org) were stained the same way with conjugated antibodies or dyes as described above.

      Analysis of EBI by IFC

      Samples were acquired within 6 hours of preparation, using INSPIRE software on an Amnis ImageStreamX Mk II equipped with 405-, 488- and 642-nm lasers (Luminex, Austin, TX, see Supplementary Table E1 [online only, available at www.exphem.org] for more details about instrument set up). Cell aggregates (Figure 1A ) have a more extended three-dimensional volume that protrudes below and above the focal plane of the two cameras of the IFC machine. Thus, the gradient root mean square (RMS) feature (which indicates the degree of the picture in focus) was set at 30–60 units for both brightfield channels 1 and 9 (Figure 1B) instead of >45 units to include events in moderate focus. With the classifier Area brightfield range set to 300–2500 µm [
      • Berman I.
      The ultrastructure of erythroblastic islands and reticular cells in mouse bone marrow.
      ], ∼10,000 aggregate events per mouse were collected on a low-flow-rate setting using the 60 × objective and analyzed using IDEAS 6.2 software (Luminex) (Figure 1C). Compensation of spectral spillover was performed manually using single color controls for each fluorophore in each experiment. Positive gates for each marker were set against the aggregate population in the respective FMO controls (Supplementary Figure E1). True EBIs were defined as possessing an F4/80+VCAM-1+CD169+ central macrophage surrounded by at least five Ter119+CD71+ or Ter119+CD71 erythroblasts (Figure 1C-G and Figure 2A-C) [
      • Lee SH
      • Crocker PR
      • Westaby S
      • et al.
      Isolation and immunocytochemical characterization of human bone marrow stromal macrophages in hemopoietic clusters.
      ]. True EBIs were identified in two successive steps as follows. Images of Hoechst 33342+ F4/80+Ter119+CD71+CD169+VCAM-1+ gated erythroblast-macrophage aggregates (GEMAs) (Figure 1C-G) were manually screened for potential EBIs defined as an F4/80+ cell clustering with three or more Ter119+ erythroblasts to reduce the number of events to analyze in detail (Figure 2A-B). These potential EBIs were then visually re-screened in detail for true EBIs with at least five Ter119+ erythroblasts and a central F4/80+ macrophage expressing CD169 and VCAM-1 antigens as defined above (Figure 2C).
      Figure 1
      Figure 1Imaging flow cytometry (IFC) gating strategy for enrichment of erythroblastic islands (EBIs). (A) Cell classifier gates were set on Aspect Ratio_Brightfield (BF) and Area_BF features on all events to identify singlet (aspect ratio: 0.8–1, area: 50–200 µm2) and aggregate (area: 300–2500 µm2) populations. (B) Cell aggregates in focus were gated on using gradient RMS of both BF channels (1 and 9). (C) Only cell aggregate events were acquired for further data analyses. (D–G) EBIs were enriched by first gating on Hoechst33342+ F4/80+ Ter119+ CD169+ CD71+ VCAM-1+ aggregates called gated erythroid–macrophage aggregates (GEMAs).
      Figure 2
      Figure 2Visualization of EBIs by IFC. (A–C) Representative example photomicrographs of GEMA (Hoechst33342+F4/80+Ter119+CD169+CD71+ VCAM-1+ aggregates), potential EBI, and true EBI events. Potential EBIs were selected for possessing an F4/80+ cell with three or more Ter119+ erythroblasts, while true EBIs were defined as possessing a central F4/80+CD169+VCAM-1+ macrophage surrounded by five or more CD71+ and/or Ter119+ erythroblasts. Markers superimposed in the merged photomicrographs are indicated in the respective color channels. (D) Representative scatterplot overlay of GEMA (from F), potential EBI, and true EBI populations. (E) GEMA events (from F) are mostly not true EBIs. Each dot represents a separate mouse, and columns represent means ± SD. n = 6 mice pooled from multiple experiments.
      The number of true EBIs per 10 million BM cells (N10mil) was calculated with the formula
      N10mil=Nraw×CfemurPfemur×Panalyzed×107cells


      where Nraw = number of true EBIs counted on IDEAS software (assuming analyses of all aggregates in sample), Cfemur = number of BM cells per femur, Pfemur = proportion of femur (by volume) stained with antibodies, and Panalyzed = volumetric proportion of BM sample run.

      Statistics

      Statistical differences were calculated with a two-tailed Mann–Whitney non-parametric test using Prism 7 (GraphPad, La Jolla, CA).

      Results

      Visualization of EBIs by IFC

      Intact cell aggregates containing EBIs were prepared by gentle flushing of BM, fixation, staining, and analysis by IFC, adapting the previously reported method [
      • Seu KG
      • Papoin J
      • Fessler R
      • et al.
      Unraveling macrophage heterogeneity in erythroblastic islands.
      ]. On the INSPIRE acquisition software, large cell aggregates were gated and acquired based on brightfield area feature (Area_BF), which reflects the overall size of the cell aggregates; the brightfield gradient RMS feature on both cameras, which reflects average focus of the image; and the brightfield aspect ratio (Aspect ratio_BF), which reflects the ratio between the long axis and short axis of the aggregates (Figure 1A–C). As reported by Seu et al. [
      • Seu KG
      • Papoin J
      • Fessler R
      • et al.
      Unraveling macrophage heterogeneity in erythroblastic islands.
      ], true EBIs were defined as possessing an F4/80+VCAM-1+CD169+ central macrophage surrounded by Ter119+CD71 and/or Ter119+CD71+ erythroblasts. On the IDEAS analysis software, EBIs were enriched by gating sequentially on Hoechst33342+ F4/80+ Ter119+ CD169+ CD71+ VCAM-1+ cell aggregates (Figure 1D–G) and named gated erythroid–macrophage aggregates (GEMAs) from here on.
      Visual inspection of events in the GEMA gate revealed that most events were not actual EBIs (Figure 2A), highlighting the problem with relying on conventional flow cytometry to quantify cell aggregates such as EBIs. Indeed, many GEMAs contained one to two Ter119+ erythroblasts, or F4/80+ monocytes or macrophages that were not in direct contact with erythroblasts, or punctate F4/80 staining that may represent fragments of monocyte or macrophage plasma membrane attached to another cell (Figure 2A). GEMA events were then manually screened for potential EBIs, defined as an F4/80+ cell with three or more Ter119+ erythroblasts (Figure 2B), and further screened a second time for true EBIs with at least five erythroblasts as defined above (Figure 2C). Following this double manual screening strategy, we found that EBI central macrophages always expressed F4/80, CD169, and VCAM-1 (Figure 2C) but did not express CD11b or Ly6G (Figure 3A), consistent with the report from Seu et al. [
      • Seu KG
      • Papoin J
      • Fessler R
      • et al.
      Unraveling macrophage heterogeneity in erythroblastic islands.
      ]. Visual inspection of aspect ratio and area of brightfield analyses of GEMA, potential EBI, and true EBI populations revealed significant variation in the size and shape of clusters within each of these categories, ruling out these parameters as a distinguishing EBI feature (Figure 2C,D).
      Figure 3
      Figure 3Granulocytes are frequently associated with EBIs. Representative photomicrographs of true EBI events containing adjacent (A), non-adjacent (B), or no CD11b+F4/80Ly6G+ granulocyte(s) (C). Granulocytes are marked with white triangles. Markers superimposed in the merged photomicrographs are indicated in the respective color channels. (D) A large proportion of true EBIs contain adjacent CD11b+F4/80Ly6G+ granulocytes. Data are means ± SD. n = 6 mice pooled from three independent experiments. (E) Representative scatterplot overlay of GEMAs, true EBIs with adjacent granulocytes, and true EBIs containing adjacent granulocytes.
      We found true EBIs are a relatively small subset of GEMAs (7.6 ± 2.8%) (Figure 2E). EBIs contained various proportions of Ter119+CD71+ immature erythroblasts and Ter119+CD71 mature erythroblasts. Interestingly, overlays of intensity plots of GEMA, potential EBI, and true EBI events plotted following the same gating strategy template used in Figure 1, revealed that true EBIs have higher expression of Ter119, F4/80, CD71, CD169, and VCAM-1 relative to GEMAs, while intensities of Ly6G, CD11b, and Hoechst33342 and magnitude of brightfield area were similar for true EBIs and GEMAs, indicating that these parameters cannot be used to enrich for true EBIs (Supplementary Figure E2, online only, available at www.exphem.org).

      Granulocytes are frequently found adjacent to EBIs

      Seu et al. [
      • Seu KG
      • Papoin J
      • Fessler R
      • et al.
      Unraveling macrophage heterogeneity in erythroblastic islands.
      ] previously reported that EBI macrophages are CD11b and estimated that 80% of EBIs contain at least one associated CD11b+ myeloid cell, although the type of myeloid cell was not further investigated. We confirmed that EBI macrophages are CD11b or CD11blow (Figure 3A), and we sought to further clarify the identity of these EBI-associated CD11b+ myeloid cells. Using IFC, we found that ∼60% of EBIs analyzed (2,650 EBIs in total from three independent experiments) had at least one adjacent CD11b+ cell at the periphery of the EBIs, and most of these were CD11b+F4/80Ly6G+ granulocytes (as illustrated in Figure 3A), with ∼45% of EBIs containing at least one adjacent granulocyte (Figure 3D). Importantly only EBIs with adjacent CD11b+ myeloid cells (within one cell diameter from the edge of the EBI) were manually counted as CD11b+ cell-containing EBIs (Figure 3A), and we excluded from this count photomicrographs (events) containing CD11b+ cells that were not immediately adjacent to the contained EBI as illustrated in Figure 3B. A flow plot overlay of EBIs with or without adjacent granulocytes confirmed that all EBIs with adjacent granulocytes were found in the CD11bhiLy6Ghi population (Figure 3D). However, some EBIs were detected as CD11b+Ly6G+ because the granulocyte was captured in the image but not directly connected or associated with the EBI as exemplified in Figure 3B, explaining why some true EBIs without adjacent granulocytes were still detected as CD11b+Ly6G+ in Figure 3E. This again cautions against conventional flow cytometry in analyses of cell aggregates, as without direct visualization of the detected events, this technique cannot distinguish whether granulocytes are adjacent or peripheral to the EBIs or integrated within the EBIs.

      G-CSF treatment leads to loss of EBIs in the BM

      We have previously reported that G-CSF treatment in vivo leads to suppression of erythropoiesis in the BM [
      • Jacobsen RN
      • Forristal CE
      • Raggatt LJ
      • et al.
      Mobilization with granulocyte colony-stimulating factor blocks medullar erythropoiesis by depleting F4/80+VCAM1+CD169+ER-HR3+Ly6G+ erythroid island macrophages in the mouse.
      ]. To examine the effect of G-CSF treatment on EBI numbers in the BM, we treated cohorts of mice with subcutaneous recombinant human G-CSF or an identical volume of saline for 4 days. We found that the fluorescence intensity of macrophage and erythroid markers (F4/80, VCAM1, CD169, CD71, and Ter119) in total aggregates was reduced following G-CSF treatment in vivo (Supplementary Figure E3, online only, available at www.exphem.org), consistent with our previous findings obtained with conventional flow cytometry [
      • Jacobsen RN
      • Forristal CE
      • Raggatt LJ
      • et al.
      Mobilization with granulocyte colony-stimulating factor blocks medullar erythropoiesis by depleting F4/80+VCAM1+CD169+ER-HR3+Ly6G+ erythroid island macrophages in the mouse.
      ]. Examination of the individual GEMAs from G-CSF-treated mice revealed that the macrophages within these GEMAs had more circular and less reticulated morphology (Figure 4A) compared with EBI macrophages from the BM of saline-treated mice (Figures 2C and 3A–C). Most GEMAs from G-CSF-treated mouse BM possessed fewer than two intact Ter119+ erythroblasts in direct contact with the macrophage or punctate Ter119+ spots that may represent fragments of erythroblast plasma membrane (Figure 4A). Therefore, G-CSF treatment alters macrophage morphology in the GEMAs isolated from the BM, possibly reflecting a change in macrophage function or a loosening of the adhesive interactions between EBI macrophages and erythroblasts. IFC analyses clearly showed that the frequencies of GEMAs and true EBIs were reduced more than 5-fold and 137-fold, respectively, in the BM of G-CSF-treated mice compared with saline-treated controls (Figure 4B,C).
      Figure 4
      Figure 4G-CSF treatment leads to loss of EBIs in the BM. (A) Representative example photomicrographs of macrophage aggregates in the GEMAs after 4 days of G-CSF treatment in vivo. Most macrophages are in contact with no to two erythroblasts or fragments of erythroblasts and multiple granulocytes. Markers superimposed in the merged photomicrographs are indicated in the respective colors. (B,C) Quantification of GEMA and true EBI frequencies in the BM of mice treated with saline or G-CSF for 4 days. n = 6 mice per group pooled from three independent experiments. **p < 0.01 (Mann–Whitney test).

      Discussion

      Visualization of EBIs by IFC has been reported in two studies [
      • Li W
      • Wang Y
      • Zhao H
      • et al.
      Identification and transcriptome analysis of erythroblastic island macrophages.
      ,
      • Seu KG
      • Papoin J
      • Fessler R
      • et al.
      Unraveling macrophage heterogeneity in erythroblastic islands.
      ]; however, these studies did not extend this method to achieve quantification of EBI frequency. Herein we achieved for the first time quantification of verified “true” EBIs by IFC and demonstrated that this can be applied to measure the effect of an in vivo treatment (herein G-CSF) on the frequency of EBIs in the BM. This novel quantitative approach can be applied to measure the effect of any in vivo treatment on EBI frequency in bone marrow, spleen, or any other hematopoietic tissue of mouse, rats, and humans, provided adequate panels of appropriate fluorescence-tagged antibodies specific to erythroblasts and macrophages are available in these species. The phenotypic identity of EBI central macrophages has been elusive since the discovery of EBIs more than six decades ago [
      • Bessis M.
      L'îlot érythroblastique. Unité fonctionelle de la moelle osseuse.
      ]. Until recently, identification of EBI central macrophages by flow cytometry relied on a combination of F4/80, CD169, and VCAM-1 markers, which are also expressed by other macrophages in the BM and other tissues [
      • Li W
      • Wang Y
      • Zhao H
      • et al.
      Identification and transcriptome analysis of erythroblastic island macrophages.
      ,
      • Seu KG
      • Papoin J
      • Fessler R
      • et al.
      Unraveling macrophage heterogeneity in erythroblastic islands.
      ,
      • Jacobsen RN
      • Forristal CE
      • Raggatt LJ
      • et al.
      Mobilization with granulocyte colony-stimulating factor blocks medullar erythropoiesis by depleting F4/80+VCAM1+CD169+ER-HR3+Ly6G+ erythroid island macrophages in the mouse.
      ]. Confocal imaging approaches for EBI are also hampered by the limited channels available and low throughput data. In this study, nine-channel IFC was used for the first time to evaluate quantity and cellular composition of EBIs isolated from mouse BM in homeostasis and after G-CSF treatment. IFC combines the best of both confocal microscopy, to actually visualize the EBIs, and flow cytometry, to generate high-throughput quantitative data visualized as both immunofluorescence photomicrographs and flow cytometry plots, allowing more accurate quantification of EBIs. We utilized the more stringent criteria of at least five erythroblasts for classification as true EBIs because the majority of EBIs reportedly possess at least five erythroblasts [
      • Li W
      • Wang Y
      • Zhao H
      • et al.
      Identification and transcriptome analysis of erythroblastic island macrophages.
      ], and doing so reduces false positives of untethered erythroblasts randomly imaged together with F4/80+ cells.
      It has recently been reported that the use of mice with a green fluorescent protein (GFP) reporter knocked in the erythropoietin receptor gene Epor enables the identification of EBIs by IFC and that EBI macrophages express Epor [
      • Li W
      • Wang Y
      • Zhao H
      • et al.
      Identification and transcriptome analysis of erythroblastic island macrophages.
      ]. The advantage of our method described here is that it can be performed in any mouse strain as it does not rely on the expression of the Epor-eGFP knocked-in allele used by Li et al. to identify EBIs [
      • Li W
      • Wang Y
      • Zhao H
      • et al.
      Identification and transcriptome analysis of erythroblastic island macrophages.
      ]. Our method is particularly advantageous to study EBIs in gene knock-out mice as the Li et al. method [
      • Li W
      • Wang Y
      • Zhao H
      • et al.
      Identification and transcriptome analysis of erythroblastic island macrophages.
      ] would require very tedious intercrossing and selection breeding to introduce the Epor-eGFP knocked-in allele into these knock-out strains. In the case of genes located in the same chromosome 9 as the Epor gene, the generation of mice carrying the Epor-eGFP allele and the two knock-out alleles by intercrossing would be very difficult. With our method, there is no need for additional selective breeding to count EBIs in the bone marrow of genetically modified strains.
      Our improved nine-channel IFC protocol revealed that almost half of all EBIs possess at least one adjacent granulocyte at the periphery of the EBIs, which further illustrates the difficulty of quantifying EBIs by conventional flow cytometry. Granulocytes have never been reported to be associated with EBIs and have no known roles in erythropoiesis. Our observation that most granulocytes associated with EBIs are peripheral to the EBIs, rather than within the core rosette of erythroblasts surrounding the central macrophages, suggests indeed that these granulocytes are bystanders rather than essential to the erythroblast maturation process. This finding serves as a warning in the interpretation of cell aggregate data by conventional flow cytometry as the latter method can lead to the misappropriation of antigens to actual EBI macrophages as exemplified here with Ly6G.
      We and others have previously reported that mobilizing doses of G-CSF reduce erythroblasts and EBI macrophage numbers in the BM but not in spleen [
      • Jacobsen RN
      • Forristal CE
      • Raggatt LJ
      • et al.
      Mobilization with granulocyte colony-stimulating factor blocks medullar erythropoiesis by depleting F4/80+VCAM1+CD169+ER-HR3+Ly6G+ erythroid island macrophages in the mouse.
      ,
      • Nijhof W
      • De Haan G
      • Dontje B
      • Loeffler M
      Effects of G-CSF on erythropoiesis.
      ], though all analyses were based on conventional flow cytometry, and the actual effects on EBIs and EBI macrophages were inferred based on the co-expression of Ter119, F4/80, VCAM-1, CD169, and ER-HR3 antigens. In this study, we confirm that true visually validated EBIs detected in the BM by IFC are more than 100-fold reduced after 4 days of G-CSF treatment. Among the relatively few GEMAs remaining in the BM of G-CSF-treated mice, macrophages displayed a swollen morphology with reduced reticulation. Thus, our data herein together with previous studies suggest that G-CSF treatment profoundly alters BM macrophage function in erythropoietic niches (EBIs), HSC niches [
      • Winkler IG
      • Sims NA
      • Pettit AR
      • et al.
      Bone marrow macrophages maintain hematopoietic stem cell (HSC) niches and their depletion mobilizes HSCs.
      ,
      • Christopher MJ
      • Rao M
      • Liu F
      • Woloszynek JR
      • Link DC
      Expression of the G-CSF receptor in monocytic cells is sufficient to mediate hematopoietic progenitor mobilization by G-CSF in mice.
      ], and at the endosteum [
      • Winkler IG
      • Sims NA
      • Pettit AR
      • et al.
      Bone marrow macrophages maintain hematopoietic stem cell (HSC) niches and their depletion mobilizes HSCs.
      ,
      • Christopher MJ
      • Rao M
      • Liu F
      • Woloszynek JR
      • Link DC
      Expression of the G-CSF receptor in monocytic cells is sufficient to mediate hematopoietic progenitor mobilization by G-CSF in mice.
      ], with profound repercussions on medullar erythropoiesis, hematopoiesis, and endosteal bone formation [
      • Jacobsen RN
      • Forristal CE
      • Raggatt LJ
      • et al.
      Mobilization with granulocyte colony-stimulating factor blocks medullar erythropoiesis by depleting F4/80+VCAM1+CD169+ER-HR3+Ly6G+ erythroid island macrophages in the mouse.
      ,
      • Winkler IG
      • Sims NA
      • Pettit AR
      • et al.
      Bone marrow macrophages maintain hematopoietic stem cell (HSC) niches and their depletion mobilizes HSCs.
      ,
      • Christopher MJ
      • Rao M
      • Liu F
      • Woloszynek JR
      • Link DC
      Expression of the G-CSF receptor in monocytic cells is sufficient to mediate hematopoietic progenitor mobilization by G-CSF in mice.
      ]. It remains to be determined whether G-CSF treatment causes reprogramming of the EBI central macrophages in the BM into macrophages incapable of maintaining EBIs. Such a mechanism could involve downregulated expression of macrophage cell adhesion molecules interacting with osteoblasts such as VCAM-1 [
      • Sadahira Y
      • Yoshino T
      • Monobe Y
      Very late activation antigen 4-vascular cell adhesion molecule 1 interaction is involved in the formation of erythroblastic islands.
      ,
      • Ulyanova T
      • Jiang Y
      • Padilla S
      • Nakamoto B
      • Papayannopoulou T
      Combinatorial and distinct roles of α5 and α4 integrins in stress erythropoiesis in mice.
      ], Maea [
      • Wei Q
      • Boulais PE
      • Zhang D
      • Pinho S
      • Tanaka M
      • Frenette PS
      Maea expressed by macrophages, but not erythroblasts, maintains postnatal murine bone marrow erythroblastic islands.
      ], αvβ3 integrin [
      • Wang Z
      • Vogel O
      • Kuhn G
      • Gassmann M
      • Vogel J
      Decreased stability of erythroblastic islands in integrin beta3-deficient mice.
      ], Mer tyrosine kinase receptor [
      • Toda S
      • Segawa K
      • Nagata S
      MerTK-mediated engulfment of pyrenocytes by central macrophages in erythroblastic islands.
      ], or Eph receptor B4 [
      • Hampton-O'Neil LA
      • Severn CE
      • Cross SJ
      • Gurung S
      • Nobes CD
      • Toye AM
      Ephrin/Eph receptor interaction facilitates macrophage recognition of differentiating human erythroblasts.
      ]. This question will be difficult to address because as shown herein, the G-CSF treatment causes the dissociation of EBIs, and as such the macrophages that were at the center of these EBIs become intractable once the EBI is dissociated. Future studies integrating time-lapse imaging of EBIs cultured with G-CSF and time-course IFC analyses of EBIs from BM and extramedullary organs after G-CSF treatment may shed more light on this question.

      Conclusions

      This study presents the first nine-channel IFC method to identify and quantify EBIs in both steady and stress states. We found that granulocytes are frequently present at the periphery of EBIs and that mobilizing doses of G-CSF lead to >100-fold reduction of EBI frequency in the BM. The same method can be applied to investigate changes in EBI number and cellular composition in different stress states such as inflammation, infection, and anemia to better understand the dynamics of EBIs and EBI macrophages in response to stress. Adaptation of the antibody cocktail in this method will also allow further phenotypic characterization of EBIs and their central macrophages in various hematopoietic tissues (BM, spleen, liver), in different species (mice, rats, humans), and under different physiological conditions.

      Acknowledgments

      We thank the Translational Research Institute, University of Queensland , and Mater Foundation for enabling this research by providing an excellent research environment and core facilities. This work was supported by National Health and Medical Research Council (Australia) Senior Research Fellowships APP1044091 and APP1136130 to JPL and APP1108352 to IGW and by additional funds from the Mater Foundation (JPL, IGW). The Translational Research Institute is supported by Therapeutic Innovation Australia (TIA) . TIA is supported by the Australian Government through the National Collaborative Research Infrastructure Strategy (NCRIS) program.

      Conflict of interest disclosure

      The authors declare no competing financial interests.

      Appendix. Supplementary materials

      References

        • Bessis M.
        L'îlot érythroblastique. Unité fonctionelle de la moelle osseuse.
        Rev Hematol. 1958; 13: 8-11
        • Berman I.
        The ultrastructure of erythroblastic islands and reticular cells in mouse bone marrow.
        J Ultrastruct Res. 1967; 17: 291-313
        • Sadahira Y
        • Mori M
        • Kimoto T
        Isolation and short-term culture of mouse splenic erythroblastic islands.
        Cell Struct Funct. 1990; 15: 59-65
        • Mohandas N
        • Prenant M.
        Three-dimensional model of bone marrow.
        Blood. 1978; 51: 633-643
        • Le Charpentier Y
        • Prenant M
        [Isolation of erythroblastic islands. Study by optical and scanning electron microscopy (author's transl)].
        Nouv Rev Fr Hematol. 1975; 15: 119-140
        • Jacobsen RN
        • Perkins AC
        • Levesque JP
        Macrophages and regulation of erythropoiesis.
        Curr Opin Hematol. 2015; 22: 212-219
        • Lee G
        • Lo A
        • Short SA
        • et al.
        Targeted gene deletion demonstrates that the cell adhesion molecule ICAM-4 is critical for erythroblastic island formation.
        Blood. 2006; 108: 2064-2071
        • Sadahira Y
        • Yoshino T
        • Monobe Y
        Very late activation antigen 4-vascular cell adhesion molecule 1 interaction is involved in the formation of erythroblastic islands.
        J Exp Med. 1995; 181: 411-415
        • Wang Z
        • Vogel O
        • Kuhn G
        • Gassmann M
        • Vogel J
        Decreased stability of erythroblastic islands in integrin beta3-deficient mice.
        Physiol Rep. 2013; 1: e00018
        • Wei Q
        • Boulais PE
        • Zhang D
        • Pinho S
        • Tanaka M
        • Frenette PS
        Maea expressed by macrophages, but not erythroblasts, maintains postnatal murine bone marrow erythroblastic islands.
        Blood. 2019; 133: 1222-1232
        • Toda S
        • Segawa K
        • Nagata S
        MerTK-mediated engulfment of pyrenocytes by central macrophages in erythroblastic islands.
        Blood. 2014; 123: 3963-3971
        • Klei TRL
        • Meinderts SM
        • van den Berg TK
        • van Bruggen R
        From the cradle to the grave: the role of macrophages in erythropoiesis and erythrophagocytosis.
        Front Immunol. 2017; 8: 73
        • Ulyanova T
        • Jiang Y
        • Padilla S
        • Nakamoto B
        • Papayannopoulou T
        Combinatorial and distinct roles of α5 and α4 integrins in stress erythropoiesis in mice.
        Blood. 2011; 117: 975-985
        • Li W
        • Wang Y
        • Zhao H
        • et al.
        Identification and transcriptome analysis of erythroblastic island macrophages.
        Blood. 2019; 134: 480-491
        • Seu KG
        • Papoin J
        • Fessler R
        • et al.
        Unraveling macrophage heterogeneity in erythroblastic islands.
        Front Immunol. 2017; 8: 1140
        • Chow A
        • Huggins M
        • Ahmed J
        • et al.
        CD169+ macrophages provide a niche promoting erythropoiesis under homeostasis and stress.
        Nat Med. 2013; 19: 429-436
        • Jacobsen RN
        • Forristal CE
        • Raggatt LJ
        • et al.
        Mobilization with granulocyte colony-stimulating factor blocks medullar erythropoiesis by depleting F4/80+VCAM1+CD169+ER-HR3+Ly6G+ erythroid island macrophages in the mouse.
        Exp Hematol. 2014; 42: 547-561.e544
        • Fraser ST
        • Midwinter RG
        • Coupland LA
        • et al.
        Heme oxygenase-1 deficiency alters erythroblastic island formation, steady-state erythropoiesis and red blood cell lifespan in mice.
        Haematologica. 2015; 100: 601-610
        • Jacobsen RN
        • Nowlan B
        • Brunck ME
        • Barbier V
        • Winkler IG
        • Levesque JP
        Fms-like tyrosine kinase 3 (Flt3) ligand depletes erythroid island macrophages and blocks medullar erythropoiesis in the mouse.
        Exp Hematol. 2016; 44: 207-212.e204
        • Lee SH
        • Crocker PR
        • Westaby S
        • et al.
        Isolation and immunocytochemical characterization of human bone marrow stromal macrophages in hemopoietic clusters.
        J Exp Med. 1988; 168: 1193-1198
        • Nijhof W
        • De Haan G
        • Dontje B
        • Loeffler M
        Effects of G-CSF on erythropoiesis.
        Ann NY Acad Sci. 1994; 718: 312-325
        • Winkler IG
        • Sims NA
        • Pettit AR
        • et al.
        Bone marrow macrophages maintain hematopoietic stem cell (HSC) niches and their depletion mobilizes HSCs.
        Blood. 2010; 116: 4815-4828
        • Christopher MJ
        • Rao M
        • Liu F
        • Woloszynek JR
        • Link DC
        Expression of the G-CSF receptor in monocytic cells is sufficient to mediate hematopoietic progenitor mobilization by G-CSF in mice.
        J Exp Med. 2011; 208: 251-260
        • Hampton-O'Neil LA
        • Severn CE
        • Cross SJ
        • Gurung S
        • Nobes CD
        • Toye AM
        Ephrin/Eph receptor interaction facilitates macrophage recognition of differentiating human erythroblasts.
        Haematologica. 2019 June 13; ([Epub ahead of print.])https://doi.org/10.3324/haematol.2018.215160