Stable colony-stimulating factor 1 fusion protein treatment increases hematopoietic stem cell pool and enhances their mobilisation in mice

All procedures complied with the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes and were approved by The University of Queensland (UQ) Health Sciences Ethics Committee. C57BL/6 (CD45.2⁺) mice were sourced from Australian Resources Centre. Congenic B6.SJL-Ptprc^a Pep^c/BoyJ mice (B6.SJL CD45.1⁺), competitor transgenic red fluorescent protein (RFP) mice (generated by Professor Patrick Tam, Children’s Medical Research Institute, Sydney, Australia, and derived from TgN (ACTB-DSRed.T3) Nagy ES cells [21]), MacGreen mice containing Csf1r-enhanced green fluorescent protein (GFP) transgene [22] and C57BL/6-Tg(UBC-GFP)30^Scha/J GFP reporter mice were supplied from in-house breeding colonies. All mice used were 10-week-old females housed under specific pathogen-free conditions.

C57BL/6 mice were treated daily with either CSF1-Fc (1 mg/kg) or saline subcutaneously for 4 consecutive days as previously described [15]. Mice were killed at 7 and 14 days post-first saline or CSF1-Fc injection and peripheral blood [23], spleen [24], liver [25] and femoral bone [26] harvested for flow cytometry assessment, immunohistochemistry and colony-forming unit (CFU) assays. HSC mobilisation was performed 14 days post-first CSF1-Fc treatment by 3 consecutive bi-daily intraperitoneal injections of saline or G-CSF (125 µg/kg; Filgrastim, Amgen, Thousand Oaks, CA) with the same tissues collected the day following the last G-CSF injection.

The CFU assays were performed as per Forristal et al. [27]. Briefly, the cytokine mix was prepared using optimal concentrations of recombinant mouse IL-3 (1000 U/ml), IL-6 (1000 U/ml), stem cell factor (SCF) (100 ng/ml) and granulocyte macrophage colony-stimulating factor (GM-CSF) in IMDM-supplemented media, and 60.9 µl of cytokine mix was added into 35 mm Petri dishes (Nullcon). In total, 2 µl of leukocyte suspension from 1 ml of sterile BM suspension from one femur, 10 µl of sterile 10 ml leukocyte suspension generated from half spleen and 10 µl of neat sterile blood were seeded on cytokine mix and covered with 1 ml of methyl cellulose. Colonies were counted after 7 days of culture incubated at 37ºC in 5% CO₂.

Competitive grafts were generated using 2 × 10⁵ BM from C57BL/6 donor mice treated with saline or CSF1-Fc (as per above) and equal numbers of BM cells from naïve UBC-GFP ‘competitor’ mice as described [28]. The BM cells were injected intravenously into lethally irradiated 10-week-old C57BL/6 recipient female mice. In another competitive transplant assay, C57BL/6 donor mice were treated with combination of CSF1-Fc or saline (as per above) plus G-CSF and whole heparinised blood was collected via cardiac puncture. In total, 20 µl of whole blood was then mixed with 2 × 10⁵ BM competitor cells from naïve RFP mice to generate the competitive graft [28]. The competitive grafts were intravenously injected into lethally irradiated 10-week-old B6.SJL CD45.1⁺ recipient mice. Blood was collected from recipients at 8-, 12- and 16-week post-transplant to quantify chimerism. Content in repopulation units (RU) was calculated for each individual recipient mouse according to CD45.2⁺ donor blood chimerism at 16-week post-transplantation using the following formula: \(RU = \frac{{\left[ {D*C} \right]}}{{\left[ {100 - D} \right]}}\) where D is the percentage of donor CD45.2⁺ B and myeloid cells and C is the number of competing CD45.1⁺ BM RUs co-transplanted with the donor cells (C = 2 for 200,000 competing BM cells). RUs were then converted to per ml of blood [29]. Multilineage reconstitution was confirmed using a standard blood leucocyte lineage panel as previously described [30].

Standard hematology analyser blood differential was used to enumerate blood cell counts. Myeloid lineage [7, 31], HSPC [7, 32], B cell subsets [33, 34] and red blood cell maturation [32, 35, 36] phenotyping were performed as detailed in Additional file 1: Table S1. Cell acquisition was performed on Beckman Coulter’s CyAn™ ADP Analyser (Beckman Coulter, USA), Cytoflex Analyser (Beckman Coulter, USA) or BD LSRFortessa™ X-20 (BD Biosciences, USA) for panels specified in Additional file 1: Table S1. Data analysis was performed using the FlowJo software (Tree Star Data Analysis Software, Ashland, OR).

For imaging flow cytometry BM was harvested and enriched for c-KIT⁺ (CD117) cells using magnetic activated cell sorting (MACS, Miltenyi Biotec) as previously described [37]. Enriched cells were stained using the antibody panel detailed in Additional file 1: Table S2 prior to image acquisition on a 3 laser AMNIS ImageStreamX MkII (Luminex Corporation) using the following settings: low flow rate, 40 × objective, 60 mW 405 nm laser power, 30 mW 488 nm laser power, 50 mW 642 nm laser power and the side scatter laser turned off. Data were compensated and analysed using IDEAS software (Life Science Research).

BM was harvested and enriched for c-KIT + (CD117) cells using MACS as above. All cell sorting was performed using a MoFlo Astrios cell sorter (Beckman Coulter, USA), and HSPC subpopulations were gated as shown in Additional file 1: Fig. S5. Monocytes were sorted as F4/80⁺CD115⁺ cells as shown in Additional file 1: Fig. S1. Bone marrow-derived macrophages (BMM) were generated as previously described [38] and collected at day 7 of culture. Total RNA was extracted from sorted cells, BMM and total BM using the RNeasy Kit (Qiagen, UK) according to the manufacturer’s protocol. First-strand cDNA was synthesised using the iScript cDNA synthesis kit (Bio-Rad, USA). Quantitative PCR (qPCR) was performed using Maxima SYBR Green/ROX qPCR Master Mix (Thermo Scientific, USA) on a sequence detection system 7500 (Applied Biosystems, USA) and analysed by 7500 System SDS Software (version 1.3.1) using the following primers: mCsf1r-forward: 5′ CTGGGAGATCTTCTCGCTTGGT 3′; mCsf1r-reverse: 5′ CTCCAGGTCCCAGCAGGACT 3′. Relative expression to the housekeeping gene Hprt (mHprt-forward: 5′ GCCCCAAAATGGTTAAGGTTGC 3′; mHprt -reverse: 5′ AACAAAGTCTGGCCTGTATCCAAC 3′) was calculated using the power delta Ct method.

For immunohistochemistry 5-µm sections of left hind limb and spleen, sampled from each block at 3 sectional depths 100 µm apart, were deparaffinized and rehydrated prior to staining with rat anti-F4/80 monoclonal antibody (Abcam, UK) or isotype control as previously described [39]. Sections were scanned at 40X magnification using Olympus VS120 slide scanner (Olympus, Japan) and analysed for percent area of chromogen staining in the entire section using Visiopharm VIS 2017.2 Image analysis software (Hørsholm, Denmark). Representative images were collected on an Olympus Bx50 microscope and cellSens standard software 7.1 (Olympus, Tokyo, Japan). For immunofluorescence, spleens were snap frozen in liquid nitrogen and embedded in Optimal Cutting Temperature (OCT; VWR, UK). Frozen sections (5 μm) were sequentially stained with anti-mouse F4/80-Alexa Fluor (AF) 647 (Abcam, UK), anti-mouse CD45R (B220)-AF488 (Biolegend, USA), anti-mouse CD169-AF594 (Biolegend, USA) or anti-mouse CD3ε-AF647 (Biolegend, USA). For Ki67 staining, paraffin tissues were dewaxed and retrieved using Diva Decloaker (Biocare Medical, USA) and sequentially stained with rabbit-anti mouse Ki67 (Abcam, UK) followed by species specific antibody coupled to AF594, followed with anti-mouse F4/80-AF647 (BD Bioscience, USA). All immunofluorescent images were counterstained with 4′,6-diamidino-2-phenylindole (DAPI; Thermo Fisher Scientific, USA). Immunofluorescence images were acquired on the Upright Motorized Olympus BX63 Upright Epifluorescence Microscope (Olympus Life Science, Australia). For morphometric analysis, digital microscopy images were analysed using ImageJ software (http://​rsb.​info.​nih.​gov/​ij/​) as previously described [40]. Briefly, all images were coded and assessed blindly at least 3 sectional depths. Background intensity thresholds were applied using an ImageJ macro which measures pixel intensity across all immunostained and non-stained areas of the images, and these were converted to percent staining area per mm² of tissue [40].

CSF1-Fc treatment was previously demonstrated to increase F4/80⁺ and Gr-1⁺ BM cells at day 5, after 4 consecutive daily treatments [15]. However, sustained impacts and resolution kinetics of CSF1-Fc treatment on BM myeloid cells and broader impacts on other hematopoietic lineages were not mapped. Given CSF1-Fc remains substantially elevated in circulation for 72 h post-delivery, peak treatment effects could be delayed by at least 3 days post-treatment cessation. Consequently, CSF1-Fc was administered to female C57BL/6 mice daily for 4 days as described [15] and BM, spleen and blood were examined at 7 and 14 days post-initiation of treatment (Fig. 1a). F4/80⁺ resident BM macrophages and monocytes were present throughout the BM in saline-treated mice (Fig. 1b) including perivascular (closed arrows) and endosteal (open arrows) regions known to be enriched for HSC niches [41]. Seven days after initial CSF1-Fc treatment the frequency of F4/80⁺ cells and their ramification in BM was appreciably increased (Fig. 1b) and could be quantified as a 1.8-fold expansion in F4/80 staining area (Fig. 1c). This BM F4/80⁺ macrophage/monocyte expansion was transient and returned to baseline levels by day 14 (Fig. 1b, c). Flow cytometry confirmed an absolute increase in BM monocytes at day 7 which returned to baseline by day 14 (Fig. 1d; Additional file 1: Fig. S1a). There was no change in BM granulocyte number compared with saline at either time point (Fig. 1e; Additional file 1: Fig. S1b). BM B cell number was decreased at day 7 and rebounded to supraphysiologic levels at day 14 (Fig. 1f; Additional file 1: Fig. S2a). BM T cells were not immediately affected by treatment but were modestly increased at day 14 (Fig. 1g; Additional file 1: Fig. S2a). Blood changes mirrored BM observations as there was a transient 41% and 28% reduction of total white and red blood cell counts/ml at day 7 post-CSF1-Fc. Monocytosis was evident at day 7 but had resolved by day 14 (frequency relative to live CD45⁺ cells: saline 3.83 ± 0.98%; day 7 post-CSF1-Fc 10.3 ± 3.15%, p = 0.002; day 14 post-CSF1-Fc 4.5 ± 0.73%). Peripheral B cells were also transiently and significantly reduced at day 7 post-CSF1-Fc (frequency in live CD45⁺ cells: saline 42.81 ± 5.86%; day 7 post-CSF1-Fc 18.26 ± 2.78%; p < 0.000001) but rebounded to supraphysiologic frequency by day 14 compared to saline (52.4 ± 3.81%; p = 0.01). Peripheral T cells and granulocyte frequency were unchanged by CSF1-Fc treatment at the examined time points (not shown). Overall, CSF1-Fc treatment stimulated a robust but transient expansion of BM monocytes and macrophages, and while it disrupted B lymphopoiesis, this was transient and rapidly corrected by a rebound expansion of BM lymphocytes.

Gow et al. previously reported an increase in spleen weight at day 5 after the same CSF1-Fc treatment regimen in transgenic MacGreen mice. This was associated with increased number of splenic GFP⁺ myeloid cells as well as increased area of the red pulp and marginal zone [15]. In situ assessment of the prolonged effects of CSF1-Fc treatment on splenic myeloid populations revealed overt disruption of gross splenic morphology at day 7 after CSF1-Fc treatment in non-transgenic mice (Fig. 2). There was dramatic decrease in both B220⁺ B cells (Fig. 2a) and CD3⁺ T cells (Additional file 1: Fig. S3a and c) with minimal identifiable white pulp on standard histology (not shown). There was also a complete loss of CD169⁺F4/80^low/− [42] marginal zone metallophilic macrophages (Fig. 2a) and a distinguishable marginal zone (Fig. 2a and Additional file 1: Fig. S3a). Disrupted splenic architecture was accompanied by sustained splenomegaly with spleen weight increased threefold 7 days after CSF1-Fc injection and remained 1.4-fold larger at 14 days post-initial treatment (Fig. 2b). Despite splenomegaly the number of Ki67⁺ proliferating cells present in spleen was substantially reduced at day 7 post-CSF1-Fc treatment (Additional file 1: Fig S3b and e). This contrasted to control and day 14 post-CSF1-Fc-treated spleens in which Ki67⁺ cells were common in both the red and white pulp (Additional file 1: Fig S3b and e). There was a substantial increase in F4/80⁺ macrophage-stained area within red pulp at day 7 post-treatment (Fig. 2a, c). In normal spleen, red pulp macrophages do not express CD169 [42, 43] and this was confirmed in saline-treated mice (Fig. 2a, e). Notably, at day 7 in CSF1-Fc-treated mice, the majority of red pulp macrophages in the expanded spleens of CSF1-Fc-treated mice acquired CD169 expression (Fig. 2a, d, e).

While the gross disorganisation of the spleen resolved by day 14, the altered F4/80⁺CD169⁺ red pulp macrophage phenotype persisted (Fig. 2a, d, e). Additionally, CD169⁺ macrophages were also detected within the CD3⁺ T cell zones of the white pulp in CSF1-Fc-treated mice in contrast to saline-treated controls (Additional file 1: Fig. S3a and d). Flow cytometry showed a significant decrease in absolute abundance of splenic monocytes (2.8-fold), granulocytes (3.3-fold), B cells (4.2-fold) and T cells (3.2-fold) at day 7 in CSF1-Fc-treated mice which resolved by day 14 (Fig. 2f–i). Overall, CSF1-Fc treatment temporarily disrupted splenic architecture, characterised by expansion of red pulp macrophages at the expense of other mature leukocytes, but induced a prolonged increase in spleen size, without overt in situ proliferation, accompanied by a persistent alteration in red pulp macrophage phenotype.

CD169 is expressed by HSC [5, 7] and erythroid [44, 45] niche macrophages in the BM. Given both the sustained increase in spleen size and shift in splenic red pulp macrophage phenotype to express CD169 [5, 7], we assessed whether CSF1-Fc treatment induced extramedullary hematopoiesis and/or impacted HSPC populations in BM, spleen and liver at 7 and 14 days post-treatment. HSPC functional subsets were subdivided based upon signalling lymphocyte activation molecule (SLAM, CD150) marker [46] expression within lineage negative, c-KIT⁺ and Sca1⁺ cells (LSK): CD48^negCD150⁺ self-renewing HSC, CD48^negCD150^neg non-self-renewing multipotent progenitors (MPP) and CD48⁺ hematopoietic progenitor cells (HPC). Committed progenitor cells were gated as lineage negative, c-KIT⁺ and Sca1^neg cells (Additional file 1: Fig. S5). This strategy provides improved precision of long-term repopulating HSC segregation from other HSPC populations compared to a previously reported study of CSF1 actions on HSPC [47].

To reassess the potential for direct action of CSF1-Fc on HSC proposed by others [47], we analysed Csf1r mRNA and protein expression in SLAM marker gated HSPC populations (Fig. 3). Consistent with single cell transcriptional profiling indicating Csf1r mRNA was expressed predominantly in committed progenitors [48], Csf1r mRNA was undetectable in sorted HSC or MPP, but expression was low in HPC and committed progenitors (Fig. 3a). The relative amount of RNA was considerably greater in sorted monocytes, total BM or in vitro generated BM-derived macrophages (Fig. 3a) [49‐51]. HSC and MPP did not express detectable CSF1R by standard flow cytometry, whereas it was detected on a small subpopulation of HPC and committed progenitors (Fig. 3b). Using imaging flow cytometry assessment of BM HSPC gated using the same LSK SLAM marker strategy isolated from MacGreen reporter mice [22], cell surface CSF1R protein (CD115) was undetectable in HSC, MPP and HPC (Fig. 3c, d), with the latter reflecting lower sensitivity of this technique versus standard flow cytometry. CD115 surface expression was detected on 6% of committed progenitors (Fig. 3c, d). Notably, 35% and 24% of HPC and committed progenitors, respectively, expressed the cytoplasmic MacGreen GFP reporter driven by the Csf1r-promoter (Fig. 3c, d), consistent with activation of Csf1r transcriptional machinery in these cells. Less than 1% of gated HSC events had detectable GFP expression (Fig. 3c), and when present, it was associated with debris artefact on the cell surface, not homogenous expression of GFP throughout the cytoplasm (Fig. 3d). In total, 7% of MPP expressed endogenous cytoplasmic GFP (Fig. 3c, d). Considering this with the lack of Csf1r mRNA expression (Fig. 3a), it suggests that a small number of MPP are transcriptionally primed to initiate Csf1r expression but that this has not yet progressed to native gene expression, or that it is below detectable levels of quantitative RT-PCR. These data confirm that HSC and MPP do not express CSF1R protein and therefore administered CSF1-Fc will not act directly on these cells.

Nevertheless, CSF1-Fc treatment induced dynamic changes in organ distribution and number of phenotypic HSC and MPP. BM HSC and MPP were significantly decreased at 7 days post-treatment (Fig. 4a). MPP returned to normal (Fig. 4a), but BM HSC numbers rebounded to 1.7-fold above baseline at day 14 (Fig. 4a) post-CSF1-Fc. By contrast, BM HPC was unchanged by CSF1-Fc treatment (Fig. 4a). The CSF1-Fc-induced reduction of detectable BM phenotypic HSC and MPP at day 7 could be a consequence of downregulation of the key marker c-KIT in response to local environment changes triggered by CSF1-Fc [23, 52]. To confirm that number of HSC residing in BM was genuinely reduced by CSF1-Fc treatment, colony-forming unit (CFU; Fig. 4a) and competitive transplantation assays were performed (Additional file 1: Fig. S6a). There was no significant difference in CFU potential in BM from saline or day 7 post-CSF1-Fc-treated animals indicating HPC number and function was not impacted (Fig. 4a). However, competitive transplant confirmed loss of hematopoietic repopulation potential compared to saline-treated donor BM (Additional file 1: Fig. S6b) supporting actual decrease in long-term repopulating HSC resident in BM.

Unexpectedly, CSF1-Fc treatment induced a sustained increase of both phenotypic HSC and MPP in the spleen with these populations remaining elevated at day 14 (Fig. 4a). Splenic HPC was substantially increased at day 7 but returned to normal by day 14 (Fig. 4a). The treatment also led to a transient but striking increase in phenotypic HSC and MPP in the liver at 7 days post-treatment which resolved by day 14 (Fig. 4a), indicating liver had adapted to supported temporary lodgement of HSC, while BM and spleen were diverted to supporting elevated myelopoiesis (Figs. 1, 2). No change was noted in liver HPC number (Fig. 4a). Elevated HSPC mobilisation into blood was not evident at either time point examined after CSF1-Fc treatment (Additional file 1: Fig. S4). In the spleen, CFU activity aligned with the large and sustained increase in phenotypic HSPC driven by CSF1-Fc treatment (Fig. 4a). Overall, CSF1-Fc treatment produced a transient decrease in HSPC residence in BM at day 7 followed by a compensatory overshoot in the HSPC pool in BM and spleen at 14 days after initial treatment.

To determine whether the increased splenic HSC pool observed at 14 days post-CSF1-Fc treatment was an indirect consequence of compensatory extramedullary hematopoiesis, myeloid progenitors [7, 31] (Additional file 1: Fig. S7), lymphoid progenitors [32, 33, 53] (Additional file 1: Fig. S7), erythroblasts and reticulocytes [35, 36] were assessed in BM, spleen and liver at day 14 post-initial CSF1-Fc or saline treatment. No effect on common myeloid progenitor (CMP) number was observed at this time point in either BM or spleen, whereas they were increased in the liver (Fig. 4b). Granulocyte macrophage progenitor (GMP) number was not impacted in BM but was elevated in spleen and liver (Fig. 4b), and this is likely attributable to sustained CSF1-Fc treatment impacts on myeloid lineage dynamics. Megakaryocyte erythroblast progenitors (MEP, gating see Additional file 1: Fig. S7) and common lymphoid progenitors (CLP) were unchanged at day 14 in the BM and spleen but increased in the liver (Fig. 4b). These observations indicated extramedullary hematopoiesis is supported within the liver but not spleen at day 14 post-CSF1-Fc treatment.

The rebound in B cell populations noted in Figs. 1 and 2 in BM and spleen is likely attributable to significantly increased Pre-Pro B cells and Pre-B cell progenitor subsets in the BM (Additional file 1: Figs. S2 and S8a) at day 14 with no change in the spleen (not shown), where progenitor B cells do not reside under homeostatic conditions [34, 54]. All 3 transitional B cell populations that are normally located within spleen [33, 34] were increased at day 14 after CSF1-Fc treatment (Additional file 1: Figs. S2 and S8b) indicative of systemic elevation of B lymphopoiesis. Erythropoiesis, as indicated by number of proerythroblasts (Additional file 1: Fig. S8c) and erythroblasts (Additional file 1: Fig. S8d), was elevated in BM but not in the spleen when compared to saline control at day 14 post-CSF1-Fc. Reticulocytes number was equivalent to baseline in both BM and spleen (Additional file 1: Fig. S8e). Overall the data indicate that the increased splenic HSC pool was not an indirect consequence of extramedullary hematopoiesis during recovery from CSF1-Fc treatment. A more likely explanation consistent with induced expression of CD169 (Fig. 2a) is that the treatment has changed splenic capacity to support HSPC homing and/or temporary lodgement. On this basis, with appropriate timing, CSF1-Fc treatment has the potential to expand available phenotypic long-term repopulating HSC for mobilisation.

To test CSF1-Fc as a potential donor conditioning regimen, mice were treated with CSF1-Fc for 4 days followed by a HSPC mobilising regimen of G-CSF initiated at day 14 (CSF1-Fc + G-CSF, Fig. 5a). Consistent with a previous study [4], G-CSF treatment reduced F4/80⁺ macrophage frequency in BM (Fig. 5b, c) but had no significant impact on F4/80⁺ macrophage frequency in spleen (Fig. 5d, e). These responses were not altered by CSF1-Fc conditioning (Fig. 5b–e). Spleen weights of CSF1-Fc + G-CSF-treated mice were significantly higher when compared to the other 3 treatment groups (Fig. 5f). For simplicity, only G-CSF-treated groups as the key comparator groups are shown in subsequent figures. Saline only or CSF1-Fc only controls harvested at day 17 confirmed minimal change in data trends compared to day 14 CSF1-Fc treatment outcomes (Figs. 1, 2, 3, 4) with representative key examples provided in Additional file 1: Fig. S9. Pre-CSF1-Fc treatment did not change monocyte frequency in BM (Fig. 5g) or spleen (Fig. 5h) of G-CSF-treated mice. BM granulocyte frequency was significantly reduced (Fig. 5i), while splenic granulocyte frequency was significantly increased (Fig. 5j) by the combination therapy compared to G-CSF treatment alone.

No significant change in HSC, MPP or HPC number in BM of G-CSF mobilised mice was induced by pre-treatment with CSF1-Fc (Fig. 6a–c). However, the number of phenotypic HSC (2.5-fold), MPP (2.5-fold) and HPC (3.8-fold) mobilised by G-CSF was each increased in the blood of CSF1-Fc pre-treated mice compared to saline + G-CSF mice (Fig. 6d–f). Similar relative increases in HSC, MPP and HPC were detected in spleen in mice pre-treated with CSF1-Fc (Fig. 6g–i). The effect of the CSF1-Fc pre-treatment on progenitor cell mobilisation was confirmed by CFU assays using BM, blood or spleen (Fig. 6j–l).

The function of mobilised HSC from saline and CSF1-Fc pre-treated mice was assessed using a competitive transplantation assay (Fig. 7a). Recipients that received a mobilised blood graft from saline + G-CSF control donors had detectable but low frequency of donor-derived blood leukocytes over 16 weeks post-transplant (Fig. 7b), whereas the frequency of donor-derived blood leukocytes (Fig. 7b) was significantly increased when it was sourced from donors that had CSF1-Fc + G-CSF treatment. Calculation of repopulating units within the grafts showed a significant increase in the number of repopulating units/ml in the blood of mice mobilised with G-CSF +CSF1-Fc compared to saline + G-CSF (Fig. 7c). Pre-treatment with CSF1-Fc did not bias the multilineage reconstitution potential of G-CSF-mobilised HSC (Fig. 7d). These results confirm that prior treatment with CSF1-Fc improved G-CSF-induced stem cell mobilisation efficiency and reconstitution potential.