BALB/cAJcl mice were purchased from Clea Japan, Inc. (Tokyo, Japan). Enhanced green fluorescent protein (EGFP)-transgenic C57/BL6-Tg (CAG-EGFP) mice and wild-type C57BL/6NCrSlc mice were purchased from Japan SLC, Inc. (Hamamatsu, Japan). The procedures used in this study were approved by the Animal Care and Use Committee of Osaka City University Graduate School of Medicine, Osaka, Japan (approval number, 09023).

The PLLA microspheres (PLA-Particles), PLLA microspheres containing rhodamine B (PLA-Particles-redF) and PLLA microspheres containing magnetite (PLA-Particles-M) were purchased from Micromod Partikeltechnologie (Rostock, Germany).

Rat anti-CD31 antibody (clone: MEC13.3) was obtained from BD Pharmingen (San Diego, CA). The avidin-biotin complex (ABC) kit and the 3,3′-diaminobenzidine (DAB) substrate kit were purchased from Vector Laboratories (Burlingame, CA). Rat monoclonal anti-vascular endothelial growth factor (VEGF) antibody (clone: RM0009-2G02) was purchased from Angio-Proteomie (Boston, MA). Rabbit polyclonal anti-fibroblast growth factor-2 (FGF-2) antibody was obtained from Abcam (Cambridge, UK). Alexa Fluor 594-conjugated goat anti-rat IgG antibody and Alexa Fluor 594-conjugated chicken anti-rabbit IgG antibody were purchased from Invitrogen Life Technologies (Carlsbad, CA). 4′,6-Diamino-2-phenylindole dihydrochloride (DAPI) was obtained from Pierce Biotechnology (Rockford, IL). Enzyme-linked immunosorbent assay (ELISA) kits for VEGF, FGF-2, interleukin 1-β (IL-1β), monocyte chemotactic protein-1 (MCP-1) and stromal cell-derived factor-1 (SDF-1) were purchased from R&D Systems (Minneapolis, Minn). The GFP ELISA kit was obtained from Cell Biolabs (San Diego, CA). Barium sulphate was purchased from Sakai Chemical Industry (Osaka, Japan) and used as contrast material for three-dimensional computed tomography (3D-CT) angiography.

The dispersed HAp nanocrystals with an average diameter of 50 nm and rod-like shape were prepared with a wet chemical process and used after calcination with an antisintering agent—poly(acrylic acid, calcium salt)—at 800°C for 1 hour, as described in our previous reports [23]. HAp nanocrystal-coated PLLA microspheres were fabricated as described [24]. Briefly, the PLLA microspheres (0.1 g) were treated with alkali (pH 11.0; adjusted with 25% ammonia solution) for 1 h at room temperature to introduce carboxyl groups onto the surfaces, washed with water, and then dried under reduced pressure. The alkali-treated and dried microspheres were washed with ethanol and immersed in a 1.0% HAp ethanol dispersion for 1 h at room temperature with stirring. The HAp-coated PLLA was washed five times with ethanol by sonication for 3 min, dried under reduced pressure, and used as NS. Bare PLLA microspheres (LA) were used as controls. The NS and LA were characterised in terms of size, particle size distribution and morphology using scanning electron microscopy.

Isolation of BMNCs was performed as described [25]. Briefly, BMNCs were harvested from 8-week-old male mice [BALB/cAJcl or C57/BL6-Tg (CAG-EGFP)] by washing of the tibiae and femora with serum-free Dulbecco's Modified Eagle's Medium (DMEM) and separated using Ficoll-Paque Plus (GE Healthcare AB, Sweden).

Murine BMNCs (5×10⁶ cells) were incubated with each microsphere preparation (3,000 particles) at 37°C for 8 h. After incubation, microspheres were harvested using a cell strainer (35-µm nylon mesh; BD Falcon, Japan). The samples were fixed with 2.5% glutaraldehyde for 1 h and dehydrated with aqueous ethanol (30%, 50%, 70%, 90%, 99%, 100%) media and 100% n-butanol for 15 min in each step. They were subsequently lyophilized and observed under scanning electron microscopy (SEM) (JSM-6301F; JEOL Ltd., Tokyo, Japan) operated at 5 kV.

Eight-week-old male mice (BALB/cAJcl or C57BL/6NCrSlc) were used for the hind limb ischemia model. Unilateral hind limb ischemia was induced by resection of the left femoral arteries, veins and side branches under anaesthesia with sodium pentobarbital (50 µg/g) injected intraperitoneally (i.p.) as described [26]. BMNCs (5×10⁶ cells) were suspended in 100 µL of serum-free DMEM with or without 1.5 mg (3,000 particles) of NS or LA, and incubated at 37°C for 3 h. Immediately after operative resection of the artery and vein, 100 µL of the suspension including 5×10⁶ cells and/or 3,000 particles of each microsphere preparation was injected intramuscularly into the ischemic thigh muscle at 4 sites (25 µL/site) using a 22-gauge needle as described elsewhere [27]. Ischemic operation, cell transplantation and evaluation of limb necrosis were performed by separate operators in a blinded manner.

We collected whole thigh muscle tissues 3, 7 and 14 days after induction of limb ischemia. The tissue lysate was prepared using a modification of the technique described by Abbott et al. [28]. Briefly, tissue extracts from thigh muscle were prepared by homogenization and lysis with 50 mM Tris-HCl, 5 mM EDTA, 250 mM NaCl, 50 mM NaF, 0.5 mM Na₃VO₄, 10 mM β-GP, 0.5% NP-40 and protease inhibitor cocktail (Nacalai Tesque, Inc. Japan). The homogenate was centrifuged for 5 min at 10,000×g. The clear supernatant was collected and used for quantification of GFP, proangiogenic factors (VEGF, FGF-2, MCP-1, SDF-1 and IL-1β) and total protein content. Total protein content of each sample was determined by bicinchoninic protein assay. Proangiogenic factors and GFP were quantified using the respective ELISA kits in accordance with the manufacturer's instructions.

Thigh tissue sections were prepared using a modification of the technique described by Clausen et al. [29]. Briefly, mice were anaesthetized with pentobarbital (i.p.) and perfused through the left ventricle using 20 ml of saline followed by 100 ml of cold 4% paraformaldehyde (PFA) in phosphate buffer (PB). The thigh tissue was post-fixed in 4% PFA for 2 h. The tissue was then embedded in paraffin or immersed in 20% sucrose in PB overnight followed by freezing in OCT compound (Sakura Finetek, Co. Japan). Paraffin-embedded sections were used for immunohistochemical staining of CD31 as described [30]. Following deparaffinization and rehydration of the sections, endogenous peroxidase activity was blocked using 0.3% hydrogen peroxide before application of blocking serum. Sections were then incubated with anti-CD31 antibody. Immunohistochemical staining was performed by the avidin-biotin complex method (ABC Kit). Colour was developed with DAB. Sections were counterstained with hematoxylin (Sigma Diagnostics, MO). CD31-positive capillary numbers were counted in 4 fields of the injection site. The injection site landmarks were the microspheres (NS or LA). The frozen sections were incubated with anti-mouse VEGF antibody or anti-mouse FGF-2 antibody, followed by incubation with fluorescent secondary antibodies and counterstaining with DAPI. Immunofluorescence was observed using a fluorescence microscope (BZ-8000; Keyence, Osaka, Japan).

We used PLLA microspheres containing magnetite (PLA-Particles-M) as control LA and the core of NS to be detected by X-ray CT scan. Angiography was performed using the technique described by Zhuang et al. [31] with slight modifications. Briefly, BALB/cAJcl mice were anaesthetised with sodium pentobarbital (50 µg/g, i.p.) and injected with 5 µl/g barium sulphate suspension (90 w/v%) directly into the left ventricle. CT images were obtained using a micro-CT scanner (La Theta LCT200; Aloka, Tokyo, Japan) according to the manufacturer's protocol. The 3D data were constructed from sliced CT images by summing those images along the Z-axis, and vascular volume analysis was performed with an image analyzer (VGStudio MAX software; Volume Graphics, Heidelberg, Germany) according to the manufacturer's protocol.

Apoptotic transplanted BMNCs and host cells were evaluated 10 days after limb ischemia by TUNEL assay in frozen sections with an In Situ Cell Death Detection Kit, TMR Red (Roche Laboratories) in accordance with the instructions provided by the manufacturer.

Data are presented as means (SD). Statistical significance was evaluated by ANOVA and Scheffé's test for comparison and contrast between multiple groups. Plots of the estimated limb survival ratio after the operation were constructed by the Kaplan-Meier method and were compared using the log-rank test. In all analyses, P<0.05 was taken to indicate significance.

Additional descriptions of experimental procedures can be found in Methods S1.

NS are microspheres approximately 100 µm in diameter (Fig. 1A-a), the surfaces of which are coated with a monolayer of HAp nanoparticles 50 nm in diameter (Fig. 1A-b, -c, -d). To assess the cell adhesiveness of NS, SEM was performed after incubation of NS and bare PLLA microspheres (LA) as controls with murine BMNCs at 37°C for 8 h in vitro. The number of cells adhering to NS was much greater than that to LA (Fig. 1B-a and b). High-magnification SEM images showed active cell adhesion to NS (Fig. 1B-c).

To determine the colocalization of implanted cells with injected microspheres, BMNCs from EGFP-transgenic mice and rhodamine B-containing PLLA microspheres (orange) as a scaffold core or control microspheres were implanted into the ischemic hind limbs of C57BL/6NCrSlc mice (Fig. 2A). Few implanted BMNCs were observed around LA (Fig. 2A-a), while markedly larger numbers of cells were seen with NS (Fig. 2A-b) in ischemic thigh tissue 7 days after transplantation. Intramuscular levels of GFP derived from transplanted BMNCs were consistently and significantly higher in the group injected with NS than that injected with LA or BMNCs alone at 3, 7 and 14 days after implantation, while GFP levels were not significantly different between BMNCs alone and LA+BMNCs groups (Fig. 2B).

Hind limb ischemia in BALB/c mice was used as an intractable ischemia model as these mice show little spontaneous collateral vessel formation in response to ischemia with ischemic hind limb necrosis [1] (Fig. 3A). The limb survival ratios after the operation in each group were compared using Kaplan-Meier analysis and log-rank statistics (Fig. 3B). In this limb ischemic model, approximately 90% of mice treated with vehicle alone developed hind limb necrosis within 5 days after the operation. Injection of NS alone did not improve limb survival, while BMNC implantation slightly but significantly improved the limb survival ratio to about 20%. Implantation of BMNCs with NS (NS+BMNCs group) markedly improved limb survival compared with implantation of BMNCs alone, while co-injection of LA did not.

To determine whether NS enhances cell implantation-induced angiogenesis, capillary formation associated with implanted BMNCs derived from EGFP-transgenic mice was determined by immunostaining for the endothelial cell marker CD31 (Fig. 4A). The density of CD31-positive capillaries in mice implanted with NS alone (Fig. 4A) was the same as that in the group injected with vehicle alone (data not shown). In the LA+BMNCs group, most of the GFP-positive BMNCs disappeared from around the LA and were no longer colocalized with LA 7 days after transplantation (Fig. 4A), and capillary density around migrating BMNCs was somewhat increased. In contrast, capillary density was markedly increased when BMNCs were co-implanted with NS. Similar enhanced CD31-positive capillary formation in the NS+BMNCs group compared with the LA+BMNCs group was observed in another severe ischemia model in BALB/cAJcl mice (Fig. 4B-a, b, c). If BMNCs alone are implanted, there will be hardly any BMNCs in the implanted area after 7 days (Fig. 2B), thereby complicating the evaluation of capillary formation in the implanted area. Because BMNCs do not adhere to LA, the BMNCs implanted in the LA+BMNCs group will also migrate from the implanted area, and there will be hardly any BMNCs remaining in the implanted area after 7 days, similar to the BMNCs alone group (Fig. 2, 4A). For this reason, LA, which is evaluated as a target for comparison with an NS particle in this study, additionally serves as a marker of the BMNCs-implanted area.

Next, we performed X-ray 3D-CT angiographic analysis to determine collateral vessel formation and blood flow. Blood flow of the left hind limb was confirmed to be completely disrupted just after femoral artery resection (Fig. 5A-a). As our micro-CT equipment has a spatial resolution of 48-µm, NS or LA with a diameter of 100 µm could be readily detected, and were visualised as light green particles on 3D-CT images. No collateral arteries had developed 7 days after the operation in the vehicle (Fig. 5A-b) or NS alone groups (Fig. 5A-c). Collateral arteries were somewhat enhanced in the BMNCs alone (Fig. 5A-d) and LA+BMNCs groups (Fig. 5A-e). Interestingly, collateral arteries were well developed around NS in the NS+BMNCs group (Fig. 5A-f). The vascular volume of the ischemic area was measured with 3D-CT angiogram data (Fig. 5B). For objective evaluation of the development of collateral arteries in the ischemic areas, we measured the vascular volume within the lower part of the thigh, ranging from the center of femur's major axis to the end of femur, based on the arterial phase angiographic data. Resembling the development of microvessels not more than 10 µm in diameter (Fig. 4), the collateral arteries that were detectable with 3D-CT also showed more marked increase in the NS+BMNCs group as compared to that in the BMNCs alone group and the LA+BMNCs group (Fig. 5B).

To investigate which proangiogenic factors are involved in enhanced therapeutic angiogenesis in mice treated with NS and BMNCs, candidate cytokines, including VEGF, FGF-2, IL-1β, MCP-1 and SDF-1 were measured by ELISA in ischemic hind limb muscles of the BALB/c mouse. Intramuscular levels of these proangiogenic factors were unaffected by vehicle injection alone (data not shown). Intramuscular levels of VEGF and FGF-2 in the NS+BMNCs group were significantly higher than those in the other groups 3 and 7 days after the operation (Fig. 6-a, -b). Levels of IL-1β and MCP-1 in the NS+BMNCs group were slightly but significantly higher than those in the BMNCs group but not those in the LA+BMNCs group 3 days after the operation (Fig. 6-c, -d). However, the levels of IL-1β and MCP-1 were not significantly different among BMNCs alone, LA+BMNCs, and NS+BMNCs groups 7 days after the operation (Fig. 6-c, -d). The level of SDF-1 in the NS+BMNC group was comparable to those in other groups with ischemia at 3 days after the operation. Although SDF-1 levels in the BMNCs alone, LA+BMNCs, and NS+BMNCs groups were higher 7 days after the operation than those in the other groups treated without BMNCs, there were no differences among the 3 groups treated with BMNCs (Fig. 6-e). Importantly, while VEGF levels were normalized even in the BMNCs and LA+BMNCs groups 7 days after ischemic operation, the NS+BMNCs group showed prolonged elevation of VEGF level (Fig. 6-a). This effect may explain enhanced recovery of blood flow in NS+BMNCs group, since long-term exposure to VEGF was reported to be necessary to produce stable microvasculature that is not resorbed after withdrawal of VEGF stimulation [3].

To further clarify the roles of proangiogenic factors, we investigated the localization of VEGF and FGF-2 expression in the ischemic muscle. VEGF expression was mostly co-localized with implanted GFP-positive BMNCs around NS, with some diffusing out of the clusters of BMNCs and NS (Fig. 7A). In contrast, although FGF-2-expressing cells were detected around and within the clusters, many did not express GFP (Fig. 7B), which could be explained by the previous finding that transplanted progenitor cells stimulate resident cells to produce additional secreted factors [8].

We investigated the possibility that co-implantation of NS prevents apoptotic cell death of BMNCs, and thus contributes to prolonged cell localization. Large numbers of transplanted BMNCs were positive for TUNEL 10 days after the operation when injected alone, suggesting apoptotic cell death (Fig. 8A-upper half, B). Some TUNEL-positive BMNCs did not necessarily express abundant GFP possibly due to the process of cell death. In marked contrast, TUNEL-positive BMNCs were hardly detected when cells were transplanted with NS (Fig. 8A-lower half, B).

Finally, we evaluated biodegradability and non-biotoxic property of NS. The course of degradation of NS over 1 year period was followed by 3D-CT. NS was implanted in the thigh muscle of normal mice. The main degradation mechanisms of PLLA in vivo are random hydrolytic chain scission, enzymatic degradation and acceleration of the breakdown by free radicals released from activated phagocytic cells such as macrophages [32]–[34]. Reduction in its molecular weight reduces the strength of NS, resulting in cracking and decomposition [32]. NS volume detected by 3D-CT decreased gradually over 1 year period, although 35.1±8.0% of NS volume was still detected 12 months after implantation (Fig. S1). Next, histological examination was performed to confirm the biodegradability and non-biotoxic property of NS. At 7 days, 3 months, and 12 months after implantation, a small number of CD45 positive leukocytes were recruited around NS, and most of these cells were Mac3 positive macrophages (Fig. S2). The macrophage count around NS was lesser at 7 days than at 3 and 12 months after implantation. At 12 months after implantation, invasion by macrophages inside the decomposed particles was noted (Fig. S2). These data suggested that the recruitment of macrophages around NS was due to the mild foreign-body reaction to eliminate foreign matter [35]. In addition, although slight increase in the number of fibroblast and the amount of type I and type III collagens around the particles were noted 12 months after implantation, no intense chronic inflammation, which could promote the strongly enhanced fibrogenesis by fibroblast [35], [36], was observed (Fig. S3).