C57BL/6 mice (purchased from the Australian Animal Resource Centre) or inbred C57BL/6 transgenic for green fluorescent protein (GFP) under the control of the ubiquitin promoter (C57BL/6-GFP) were used. All animal experiments were approved by the University of Queensland Animal Ethics Committee.

LSK and MSC populations were isolated from C57BL/6 or C57BL/6-GFP mice as previously described by our group [40]. All experiments involving MSCs were performed at passage 8–12. LSKs and MSCs were characterized by morphology, cell surface phenotype and functional capacity as previously published by our group [40].

MSCs were induced into the osteogenic lineage in vitro as follows: 2 × 10⁴ MSCs were seeded in 24-well plates, grown to confluence and cultured for 21 days in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with dexamethasone (0.1 μM), β-glycerol phosphate (100 mM), l-ascorbate-2-phosphate (10 mM), calcium chloride (4 mM), 10% fetal calf serum (FCS) and gentamycin (40 μg/mL, Pfizer, New York, NY, USA). These were fixed in 4% paraformaldehyde (PFA) and stained for the presence of calcified osteoid deposits with Alizarin red S solution [40,41]. Undifferentiated MSCs and MSCs induced into osteoblasts were further characterised according to their gene expression of HSC niche markers including angiopoietin 1 and 2, stem cell factor, jagged-1 and stromal cell-derived factor 1 (CXCL12).

Cell sorting and immunophenotype analysis was performed by flow cytometry using fluorochrome-labeled rat-anti mouse monoclonal antibodies (all at 1–2.5 μg/mL) as follows: c-kit allophycocyanin (APC; 2B8; BD, Franklin Lakes, NJ, USA), Sca-1 phycoerythrin cyanine-7 (PE Cy7; D7; BD), CD45 APC (30-F11; BD), CD31 PE (MEC13.3; BD), CD44 PE (IM7; BD), CD11b PE (M1/70; BD), F4/80 Pacific Blue (BM8; eBioscience, San Diego, CA, USA), Gr-1 APC Cy7 (RB6-8C5; BD), CD45R/B220 Pacific Blue (RA3-6B2; BD) and CD5 APC (53–7.3; BD). A biotinylated lineage cocktail (containing CD5 (53–7.3; BD), CD45R/B220 (RA3-6B2; BD), Gr-1 (RB6-8C5; BD) and F4/80 (BM8; eBioscience)) was also used for staining of haematopoietic cells with streptavidin Pacific Blue (Invitrogen, Carlsbad, CA, USA) secondary staining. Cell sorting was performed by fluorescence-activated cell sorting (FACS) on a FACS-ARIA (BD) and immunophenotyping was performed on an LSRII (BD) with results analysed in FlowJo Version 7.5 (Tree Star, Ashland, OR, USA).

RNA from was extracted with a Qiagen RNeasy Mini Kit (Qiagen, Venlo, Netherlands) using the manufacturer’s protocol. RNA was pre-incubated with DNase I (Invitrogen) and reverse transcription was performed with oligo-dT and Superscript III (Invitrogen) as described in the manufacturer’s instructions. Quantitative real-time polymerase chain reaction (qRT-PCR) was performed on cDNA using ABsolute™ QPCR SYBR^® Green (ABgene, Waltham, MA, USA). Product size and primer sequences used are shown in Table 1.

MSCs were seeded into a 24-well plate and allowed to reach confluence. Osteoblasts were induced from MSC as described above and used for experiments after 2–3 weeks of differentiation. 1 × 10³ purified LSKs (500 cells/cm²) were seeded on top of stromal cells in static cultures for 7 days in BioWhittaker X-vivo 10 medium (Lonza, Basel, Switzerland) supplemented with 20% FCS and gentamycin (40 μg/mL). After this period the resultant cell populations were removed from culture, counted and analysed by flow cytometry for lineage-specific haematopoietic markers and LSK markers.

Female C57BL/6 recipient mice were administered a split dose of total body irradiation (2 × 550 cGy) 3 h apart via a self-contained Gammacell 40 Exactor (GC-40E) irradiator using a ¹³⁷caesium radiation source. Twenty-four hours after irradiation, recipients were intravenously injected via the retro-orbital sinus with the appropriate cell populations resuspended in saline supplemented with 2% heat-inactivated FCS and DNase I (10 μg/mL, Roche, Basel, Switzerland). All mice were monitored by weight and clinical scoring over the first 28 days of the experiments and drinking water was supplemented with antibiotics and antifungal agents for the first 14 days. For analysis of functionality of the initial purified LSK population, BM cells and LSKs were isolated as above from C57BL/6-GFP mice. The c-kit depleted cells were sourced from the negative fraction of c-kit magnetic separation (cells devoid of HSCs) and the vehicle group received cell excipient only. For competitive repopulation transplants, 200 GFP⁺ LSKs or the equivalent number of cells produced in culture from 200 LSKs, were used as the donor population together with 4 × 10⁵ whole bone marrow (BM) cells from C57BL/6 as the competitor cells. Where performed, MSCs and osteoblasts were physically removed from harvested co-cultures prior to transplantation using flow sorting for GFP⁺ haematopoietic cells. For secondary transplants, BM was harvested from primary recipients, pooled and 2 × 10⁶ cells transplanted into lethally irradiated secondary recipients. All treatment groups were monitored for health and well-being by daily scoring. Tail bleeding was performed 4, 6, 8, 10 and 12 weeks post-transplant and flow cytometry analysis of GFP was used to detect the level of donor (GFP⁺) and competitive (GFP⁻) total cells, myeloid cells (CDllb⁺) and B cells (CD45R/B220⁺) in the recipient blood. T cell engraftment was not examined because of the known ability of recipient T cells to survive high-dose total body irradiation. Donor populations were deemed to be engrafted if greater than 1% GFP⁺ cells were observed in blood or BM. Blood, BM and spleen counts were performed on a Sysmex KX-21N Cell Counter (Sysmex Corporation, Kobe, Japan).

Snap frozen tissue was crushed into a powder using a pestle and mortar on dry ice and digested by overnight incubation with Proteinase K at 56 °C. The genomic DNA was then extracted using a QIAamp DNA Mini Kit (Qiagen) according to the manufacturer’s protocol (RNA was removed by RNase treatment). The genomic DNA was assessed for the presence of genomic GFP using the primers and fluorescent probe as follows: 5′end: CTGCTGCCCGACAACCA. 3′end: TGTGATCGCGCTTCTCGTT fluorescent probe 5′end: CCCAGTCCGCCCTGAGCAAAGAC. The number of GFP⁺ cells was then calculated using a standard curve of the amount of GFP genomic DNA from known cell numbers.

Either a Mann-Whitney test or Kruskal-Wallis test with Dunn’s Multiple Comparison post-test was used to test statistical significance between groups. This was performed using GraphPad Prism version 4.03 for Windows (GraphPad Software, La Jolla, CA, USA). Results were deemed statistically different if p < 0.01. All continuous data were expressed as mean ± 1 standard deviation (SD).

LSK and MSC populations were characterised as per Cook et al. (2012) [40]. Briefly, LSKs were shown to be enriched for a true repopulating population of HSC by transplantation of 200 LSKs into lethally myeloblated recipients and achieving 100% survival for >16 weeks (data not shown). MSCs had a classical fibroblast morphology and were Sca-1⁺, CD44⁺, CD45⁻, CD11b⁻ and CD31⁻. MSC were also shown to be able to differentiate into osteoblasts, adipocytes and chondrocytes [40,41] (data not shown).

The ability of MSCs to differentiate into osteoblasts was confirmed by qRT-PCR for lineage-specific markers. It was shown that MSC-derived osteoblasts expressed higher levels of markers for early osteogenic development, namely RUNX2 and osterix, when compared to undifferentiated MSCs (Figure 1A,B). Furthermore, MSC-derived osteoblasts expressed higher levels of osteocalcin (also known as bone gamma-carboxyglutamate protein), a marker of mature osteoblasts (Figure 1C) [12].

MSCs and osteoblasts were also assessed by qRT-PCR for their expression of HSC niche markers as an indication of their potential to support HSC growth and proliferation in vitro. A panel of niche-specific markers was selected which included angiopoietin-1, angiopoietin-2, stem cell factor, jagged-1 and SDF-1. All assayed genes were detected in both MSC and osteoblast populations; however, significantly higher expression of all genes was observed in osteoblasts (Figure 1D–H). This suggested that both populations had the propensity to support HSC growth in vitro but that osteoblasts might be a superior candidate for this.

For co-culture of LSKs on MSCs or osteoblasts monolayers, only serum was used: No haematopoietic growth factors or any other proteins were added. Co-cultures were optimised for haematopoietic cell expansion with X-vivo-10 medium supplemented with 20% FCS shown to be optimal for proliferation. 1000 LSKs were seeded on confluent MSC or osteoblasts monolayers. Haematopoietic cell expansion occurred on both undifferentiated MSCs and on osteoblasts (Figure 2A,B). After 5 days, cobblestone-like areas of adherent haematopoietic cells were observed (Figure 2A). After 7 days, over 95% of haematopoietic cells were found to be adherent to the stromal layer (Figure 2B). A mean 788-fold increase (±291-fold) in total haematopoietic cells on MSC monolayers was observed with an approximate two-fold higher total expansion (1513 ± 449-fold) when LSKs were co-cultured with osteoblasts compared to undifferentiated MSCs (p = 0.0013) (Figure 2C). This equated to a mean doubling time for haematopoietic cells on MSCs and osteoblasts of 17.5 h and 15.9 h respectively. The expansion appeared to be MSC- or osteoblast-specific, since LSKs co-cultured in the presence of adipocyte-differentiated MSCs showed no expansion of LSKs (data not shown), suggesting that marrow adipocytes are negative regulators of HSC expansion and/or survival, and confirming the findings of two previous studies [42,43]. Similar proportions of LSK cells were found to be present whether undifferentiated MSCs or osteoblasts were used (Figure 2D). However, as a result of the higher total cell expansion on osteoblasts, an increased yield of LSK was obtained in osteoblast/LSK co-cultures (p = 0.003) (Figure 2E). When characterising these expanded cells it was found that a majority of the expanded haematopoietic cells were lineage-committed (Figure 2F). The degree and type of haematopoietic lineage commitment was similar regardless of whether the supporting monolayer was of undifferentiated MSCs or osteoblasts, with a majority of the lineage-committed cells being of the myeloid lineage (Figure 2F).

CD marker profile/phenotype is not always indicative of cell function and thus flow cytometry data cannot be used to identify functional HSC numbers (i.e., those with the ability to repopulate the BM). As such, an in vivo competitive repopulation model was used to compare the functional capacity of 200 LSK expanded on either undifferentiated MSCs or osteoblast monolayers to that of 200 freshly isolated LSKs. At 12 weeks after transplantation we found that LSKs co-cultured for 7 days on MSCs or osteoblasts contained functional HSCs after primary transplantation into myeloablated recipient mice (Figure 3B–D). There was no significant difference between the co-cultured cells and the freshly isolated LSK cells in these experiments (Figure 3B–D), indicating that functional HSCs had been maintained in co-cultures with MSC and osteoblasts despite the absence of exogenous cytokines. As an additional test for the functional capacity of putative HSCs, BM cells were taken from animals that had received a primary transplant of co-cultured LSKs and were transplanted into myeloablated syngeneic secondary hosts. At 12 weeks after total body irradiation and cell infusion mean multi-lineage haematopoietic reconstitution was similar to that following primary transplantation (Figure 3F–H). Multi-lineage reconstitution was again evident for total haematopoietic cells, B cells and myeloid cells (Figure 3F–H) at a similar level to the fresh LSK recipient controls. This reinforced the results from the primary transplants indicating that HSC were maintained during culture with MSC and osteoblast monolayers. However, HSC were not expanded, as engraftment was not increased. The fact that reconstitution in myeloid and B lineages was equivalent proves that these cultured LSK cells were capable of long-term multi-lineage reconstitution. If a bias in either myeloid or B lineage had been observed, the conclusion would be that HSC clonal subtypes with different differentiation potentials (such as α-HSC, β-HSC, γ-HSC and δ-HSC as defined by Benz et al., 2012 [44]) were differentially selected on MSC or osteoblasts. This was not the case in the transplants shown. Therefore, co-cultures on MSC or osteoblasts are likely to maintain all HSC regardless of their clonal subtype.

Surprisingly, mice receiving LSKs co-cultured on osteoblasts, while showing haematopoietic engraftment at similar levels to that of freshly isolated LSKs, displayed significantly poorer survival than those receiving LSKs co-cultured on MSC (p = 0.0062) (Figure 3A) due to haematopoietic failure between days 6–10 following transplantation. This was despite all mice receiving 4 × 10⁵ healthy competing BM cells, which theoretically contained 4–6 long-term repopulating HSCs (enough to ensure haematopoietic reconstitution and survival). Thus, we concluded that transplantation of LSK-osteoblast co-cultures compromised survival of recipients of the primary transplants and that this in turn compromised survival of recipients of the secondary transplants (Figure 3A,E respectively). This suggested that the self-renewal capacity of normal (GFP⁻) HSCs in the survivors of the primary transplants was compromised.

We hypothesised that co-culture of LSKs on osteoblasts compromised HSC homing and engraftment. One possibility to explain this was that residual osteoblasts, harvested from the co-cultures, were trapped in the lungs and prevented HSCs accompanying them from reaching sites of haematopoiesis including the bone marrow and spleen. An alternative possibility was that the normal homing mechanisms of HSC was in some way impaired by their co-culture and administration with osteoblasts.

Using MSCs from mice transgenic for luciferase, we have previously shown that intravenously injected MSCs are detectable in the lungs within minutes of injection [45]. Thus, we hypothesised that large, matrix-secreting osteoblasts bind HSCs tightly prior to injection and that theses complexes becomes lodged in the small capillaries of the lungs, thus preventing optimal homing of both expanded and competitor-repopulating HSCs to sites of haematopoiesis. To explore this hypothesis, GFP⁺ osteoblasts or GFP⁺ MSCs were injected intravenously into lethally irradiated C57BL/6 recipients with unlabelled co-cultured haematopoietic cells and the lungs harvested at 1 day, 3 days or 12 days post-injection. GFP⁺ MSCs and GFP⁺ osteoblasts were detected by qRT-PCR for genomic GFP in harvested lung tissue at 1 and 3 days after injection, but were not detectable at 12 days after injection (Figure 4A). These findings are very similar to those using luciferase transgenic MSCs [45].

We next investigated whether MSC and osteoblast entrapment in the lungs was preventing HSC migration to the marrow and spleen. This was performed in a similar manner to the primary transplants (Section 3.3) by transplanting freshly isolated GFP⁺ LSKs or GFP⁺ LSKs co-cultured with unlabelled MSCs or osteoblasts together with whole BM competing cells. Assessment of leucocyte counts in the blood, BM and spleen was performed at the typical time of death in the osteoblast co-cultured transplant group (day 8–14 post-transplant). This revealed no observable differences between these groups in cellular reconstitution of blood or BM at day 8 post-transplant (Figure 4B,C). However, at day 8 post-transplant significantly fewer leucocytes were found in the spleens of recipients that had received HSCs co-cultured with osteoblasts compared to recipients of fresh LSKs (p < 0.001) and LSK/MSC co-culture recipients (p < 0.01) (Figure 4D). Since this was a competitive transplant experiment, it appeared that osteoblasts from the co-culture prevented both co-cultured HSCs and competitive unmanipulated HSCs from reaching the spleen and proliferating there. Thus, we concluded that the cause of death in the osteoblast/LSK co-cultured recipients was due to an impairment of short-term reconstitution, particularly in the spleen, rendering recipients susceptible to pancytopenia and subsequent infection. We next explored the question of homing to sites of haematopoiesis (marrow and spleen) of these different populations.

The above experiments did not show whether the impairment of short-term splenic reconstitution was due to compromised splenic proliferation of haematopoiesis or compromised homing to the spleen. However, since this was a competitive transplant model (i.e., recipients were co-transplanted with 4 × 10⁵ unmanipulated BM cells—A dose that normally allows short- and long-term repopulation), we hypothesised that this was due to compromised homing of both transplanted populations. To explore this, we assessed the homing of fresh GFP⁺ HSCs, LSKs co-cultured with MSCs and LSKs co-cultured with osteoblasts to the spleen at 16 h post-transplant, a time point at which detectable proliferation was unlikely. This revealed that the transplantation of LSK/osteoblasts co-cultured cells led to significantly reduced homing to the spleen (p = 0.0245) when compared to fresh LSK and LSK-MSC co-culture recipients (Figure 4E). Similar findings were made for BM homing (Figure 4F, p = 0.0154). In conjunction with the low splenic leucocyte counts, this implied that the transplanted competitive BM cells with repopulation ability cultured on osteoblasts did not reach the spleen early post-transplant. This led to complete failure of homing and engraftment of HSC in 50% of the animals and death within 10 days of transplant (Figure 3A). The other 50% of recipients survived for the duration of the study and showed a similar BM reconstitution to that of fresh LSK and MSC-cultured LSK (Figure 3B–D). Hence, we propose that the co-transplantation of osteoblasts significantly inhibits and/or prevents migration of cells with repopulation ability to both the spleen and BM. In order to attempt to confirm this, we then removed osteoblasts from the co-culture prior to injection of the co-cultured LSK.

Seven day co-cultures of GFP⁺ LSKs with unlabelled MSCs or with osteoblasts were initiated as before. The total resulting cells were harvested and the GFP⁺ haematopoietic cells were isolated by FACS sorting. These cells were then transplanted with whole BM competitors as before. Transplants depleted of osteoblasts now showed 100% survival at 30 days postransplant—The same result as with fresh LSKs and LSKs cultured with undifferentiated MSCs (Figure 5A). These findings showed that removal of osteoblasts from cell suspensions prior to infusion allowed short-term reconstitution. However, removal of MSCs or osteoblasts before transplantation paradoxically compromised the engraftment of the co-culture-expanded donor GFP⁺ haematopoietic cells (Figure 5B–D). Only a single mouse showed donor GFP⁺ myeloid engraftment in the MSC-cultured LSK transplanted group (Figure 5D). This suggested that HSC were lost during the sorting process. Since the cells were mono-dispersed prior to sorting, this loss was likely due to shear damage in the sorting process. Alternatively, the removal of osteoblasts may have allowed the “released” LSKs to differentiate more and thus lose their self-renewal potential.