Three-dimensional (3D) culture systems are critical to investigate cell physiology and to engineer tissue grafts. In this study, we describe a simple yet innovative bioreactor-based approach to seed, expand, and differentiate bone marrow stromal cells (BMSCs) directly in a 3D environment, bypassing the conventional process of monolayer (two-dimensional [2D]) expansion. The system, based on the perfusion of bone marrow–nucleated cells through porous 3D scaffolds, supported the formation of stromal-like tissues, where BMSCs could be cocultured with hematopoietic progenitor cells in proportions dependent on the specific medium supplements. The resulting engineered constructs, when implanted ectopically in nude mice, generated bone tissue more reproducibly, uniformly, and extensively than scaffolds loaded with 2D-expanded BMSCs. The developed system may thus be used as a 3D in vitro model of bone marrow to study interactions between BMSCs and hematopoietic cells as well as to streamline manufacture of osteoinductive grafts in the context of regenerative medicine.
Bone marrow stromal cells (BMSCs) have received increasing experimental and clinical interest, owing to their surprising degree of plasticity [1–3] and their potential use for treatment of genetic  or immunologic  pathologies. In the field of regenerative medicine, BMSCs have been most extensively used for bone repair because their default pathway seems to be osteogenic . This has led to encouraging findings in heterotopic models [7, 8], in orthotopic implants [9, 10], and in a few clinical cases . Given their low frequency among bone marrow–nucleated cells (approximately 0.01%), BMSCs are typically selected and expanded by sequential passages in monolayer (two-dimensional [2D]) cultures. However, 2D-expanded BMSCs have a dramatically reduced differentiation capacity compared with those found in fresh bone marrow [12, 13], which limits their potential use for therapeutic purposes [6, 14].
Reasoning that a three-dimensional (3D) culture system may represent a more physiological environment than a Petri dish for a variety of cells [15, 16] and that fluid flow is an important component for seeding and culturing BMSCs in 3D environments [17, 18], we aimed in this work at developing an innovative procedure to seed and expand BMSCs directly into porous 3D scaffolds under perfusion. We demonstrated that perfusion of bone marrow–nucleated cells through the pores of 3D ceramic scaffolds resulted in the efficient expansion of clonogenic BMSCs and in the generation of highly osteoinductive grafts. Moreover, the developed system allowed us to coculture BMSCs with hematopoietic cells and to support hematopoiesis.
Materials and Methods
Bone Marrow Cell Culture
Bone Marrow Aspirates
Bone marrow aspirates (20- to 40-ml volumes) were obtained from eight healthy donors (36–54 years old) during routine orthopedic surgical procedures in accordance with the local ethical committee (University Hospital Basel) and after informed consent. Nucleated cells were isolated from aspirates by Ficoll density-gradient centrifugation. The initial number of BMSCs, defined as the number of fibroblast colony-forming units (CFU-F) in the fresh marrow aspirates, averaged 21 ± 7 per 105 nucleated cells.
Unless otherwise stated, medium (α-modified Eagle's medium) containing 10% fetal bovine serum was supplemented with 5 ng/ml fibroblast growth factor-2, 10 nM dexamethasone, and 0.1 mM L-ascorbic acid-2-phosphate to increase BMSC proliferation and osteogenic commitment [8, 19]. In some experiments, medium was alternatively supplemented with 2 ng/ml interleukin-3, 10 ng/ml stem cell factor, and 20 ng/ml platelet-derived growth factor-bb to support maintenance of hematopoietic cells in culture  (hematopoietic medium).
Using a perfusion bioreactor system we previously developed for cell seeding of 3D scaffolds , an average of 18.4 ± 6.6 million freshly isolated bone marrow–nucleated cells were perfused through 8-mm-diameter, 4-mm-thick disks of porous (total porosity, 83% ± 3%; poresize distribution: 22%, <100 μm; 32%,100–200 μm; 40%, 200–500 μm; 6%, >500 μm) hydroxyapatite ceramic (Engipore; Fin-Ceramica Faenza, Faenza, Italy, http://www.fin-ceramicafaenza.com) at a superficial velocity of 400 μm per second (previously determined to result in efficient and uniform cell seeding). Based on CFU-F assays of five marrow aspirates, an estimated average of 4.8 ± 2.6 × 103 BMSCs was perfused through each disk, corresponding to 4 BMSCs per cm2 of ceramic surface area. Such clonogenic BMSC seeding density was previously described to prolong BMSC lifespan and differentiation potential . After 5 days (cell seeding phase), harvested medium was plated in tissue culture dishes to quantify the fraction of CFU-F not seeded. Fresh medium was then added to the system, and the cell-ceramic constructs were perfused for an additional 14 days (cell expansion phase) at a velocity of 100 μm per second (previously determined to support cell viability throughout the scaffold thickness), with medium changes twice a week. As a control, bone marrow–nucleated cells from each donor were plated on tissue-culture dishes (2D expansion) using the same initial cell number/surface area as in the 3D ceramic disks and cultured for 19 days without passaging, with the same schedule of medium changes as for the 3D culture.
Bone Formation Assays
Constructs from four independent experiments, after the cell seeding or cell expansion phases of 3D culture, were implanted ectopically in recipient nude mice (CD-1 nu/nu, 1 month old; Charles River Laboratories, Sulzfeld, Germany, http://www.criver.com/index.html) in accordance with institutional guidelines. As a control, we implanted ceramics seeded with 2D-expanded BMSCs at the same density as measured in the corresponding 3D cultured constructs after the cell expansion phase. Seeding of 2D-expanded BMSCs was performed by static loading of a cell suspension. We previously reported that the fraction of cells retained in the scaffolds after seeding by static loading was similar to that obtained using the described perfusion device, although cells seeded statically were less uniformly distributed .
Quantitative Assessment of Bone Tissue Formation
Eight weeks after implantation, constructs were fixed in 4% formalin, decalcified (Osteodec; Bio-Optica, Milan, Italy, http://www.bio-optica.it), paraffin embedded, and sectioned at six different levels (5-μm-thick sections at 600-μm intervals). For each cross-section, stained by hematoxilin/eosin, six images (covering most of the total cross-sectional area) were used to quantify the amount of bone tissue normalized to the total available pore space, as previously described . The uniformity of bone tissue formation was quantitatively determined from the average (x) and standard deviation (s) of the bone amounts measured in each cross-section  as follows:
Scanning Electron Microscopy
Constructs cultured in the 3D system after the cell expansion phase were fixed in 4% formalin, dehydrated, critical point dried, and coated with 20 nm of Au. Scanning electron microscopy observation was performed with an ESEM XL 30 (Philips, Amsterdam, The Netherlands, http://www.philips.com) with 10-kV acceleration.
mRNA was extracted using TRIzol (Invitrogen, Carlsbad, CA, http://www.invitrogen.com), treated with DNAse, and retrotran-scribed into cDNA, as previously described . Polymerase chain reaction was performed and monitored with the ABI Prism 7700 Sequence Detection System (PerkinElmer/Applied Biosystems, Rotkreuz, Switzerland, http://www.perkinelmer.com), and expression levels of genes of interest (bone sialoprotein [BSP], collagen type I [CI], and osteopontin [OP]) were normalized to the 18S rRNA. Previously determined levels of expression of the genes of interest in human osteoblast cultures, also normalized to 18S rRNA , were used as reference.
After the cell expansion phase in the 3D culture system, cells were extracted from the ceramic pores by perfusing a solution of 0.3% collagenase and 0.05% trypsin/0.53 mM EDTA at 400 μm per second. Extracted cells were assessed for the ability to form fibroblastic and hematopoietic colonies and characterized by flow cytometry, as described below.
CFU-F assays of expanded cells were performed by plating four cells per cm2 in tissue culture dishes. After 10 days of culture, cells were fixed in 4% formalin and stained with 1% methylene blue, and the number of colonies was counted.
Hematopoietic Colony-Forming Unit Assay
Hematopoietic colony-forming unit assays were performed as previously described  to quantify the following types of hematopoietic clonogenic cells: neutrophils, macrophages, burst-forming-unit-erythroid, and granulocyte-erythroblast-macrophage-megakariocyte. Briefly, 2.5 × 105 cells per ml were cultured in medium containing 1.75 U/ml erythropoietin, 2.625 ng/ml granulocyte-colony stimulating factor, 40 U/ml granulocyte macrophage colony stimulating factor, 40 U/ml interleukin-3, and 62.5 ng/ml stem cell factor. After 14 days, the colonies were classified and counted.
Fluorescence-Activated Cell Sorting Analysis
Cell suspensions were incubated with antibodies against CD105 (Serotec), STRO-1, BSP, CI, OP (all from Developmental Studies Hybridoma Bank, Iowa City, IA, http://www.uiowa.edu/∼dshbwww), nerve growth factor receptor (NGFR), or CD45 (both from Becton, Dickinson and Company, Franklin Lakes, NJ, http://www.bd.com) and analyzed using a FACSCalibur flow cytometer (Becton, Dickinson and Company). Reactions with anti-BSP, -OP, or -CI were proceeded by membrane per-meabilization with BD Cytofix/Cytoperm Plus Kit (Becton, Dickinson and Company). Positive expression was defined as the level of fluorescence greater than 95% of corresponding isotype-matched control antibodies.
Results and Discussion
BMSC Expansion Under 3D Perfusion
Using a bioreactor system recently developed for efficient and uniform seeding of anchorage-dependent cells into 3D scaffolds , we perfused the nucleated cells of human bone marrow aspirates in alternate directions through the pores of disk-shaped ceramic scaffolds, and we hypothesized that BMSCs would attach to the ceramic substrate and proliferate. The number of BMSCs perfused through each scaffold, estimated by CFU-F assays, averaged 4.8 ± 2.6 × 103 cells. Medium was first changed after 5 days (cell seeding phase), which resulted in the elimination of the non-attached cell population, containing negligible numbers of CFU-F (<1% of those seeded in the scaffolds). Fresh medium was further perfused for an additional 14 days (cell expansion phase), during which time the total number of cells, monitored by Alamar blue, was found to increase at a nearly exponential rate (Fig. 1). At 19 days, the number of BMSCs found within the ceramic pores, calculated as the CD105+ fraction of the extracted cells, averaged 9 ± 3 × 105 cells for each scaffold. These data demonstrate that BMSCs can be seeded and extensively expanded (average of 8.2 ± 0.9 doublings in 19 days) by perfusion of bone marrow cell suspensions through 3D porous scaffolds, thereby avoiding typical 2D expansion.
Bone Formation by Expanded BMSCs
The osteoinductivity of the constructs resulting from BMSC seeding and expansion in the porous ceramic under perfusion (total of 19 days culture) was verified by ectopic implantation in nude mice. Reproducible, extensive, and markedly uniform bone formation was found in implanted constructs from four out of four independent experiments, performed using aspirates from different donors. Mature lamellar bone, organized in typical bone/marrow ossicles , filled an average of 52.1% ± 7.7% of the total available pore space and was distributed throughout the scaffold volume with high uniformity (Fig. 2). In contrast, when 2D-expanded BMSCs from the same donors were loaded into ceramic scaffolds at the same density as measured in the corresponding 3D cultured constructs, bone tissue was formed in only one of the four experiments. Moreover, in those constructs positive for bone formation, bone tissue filled only 9.6% ± 2.7% of the total available pore space and was localized to scattered peripheral regions (Fig. 2). The increased osteoinductivity of constructs generated using the developed system may have been supported by the ceramic substrate used for BMSC expansion , the 3D cell–cell interactions during culture , the regimen of fluid flow applied [17, 26], or combinations of these variables that remain to be further elucidated. Interestingly, constructs implanted immediately after the cell seeding phase, in which BMSCs were attached to the ceramic but had not significantly expanded, were never osteoinductive. This suggests that a critical density of osteoprogenitor cells is necessary to initiate bone formation and points out the limit of approaches based on direct implantation of scaffolds mixed with bone marrow aspirates, especially considering the known variability in the number of BMSCs per aspirate volume .
We then preliminarily characterized the morphology, phenotype, and clonogenicity of cells seeded and expanded within the developed 3D system. Scanning electron microscopy indicated the formation of a stromal-like tissue within the ceramic pores, consisting of a 3D network of spheroidal cells in contact with heterogeneously shaped fibroblastic cells (Fig. 3A). The mRNA expression levels of genes encoding for the osteoblast-related proteins BSP, CI, and OP averaged, respectively, 3.6%, 35.3%, and 48.0% of those previously quantified in human osteoblast cultures  (Fig. 3B). Levels were similar to those measured in 2D-expanded BMSCs and lower than those measured in BMSCs after osteogenic differentiation . Fluorescence-activated cell sorting analyses indicated that 68% ± 18% of the cells extracted from the ceramic scaffolds were positive for CD105, a surface marker typically expressed by cells of the mesenchymal lineage (Fig. 3C). These CD105+ cells expressed low levels of STRO-1 (proposed as a marker of early mesenchymal progenitors ) and NGFR (proposed as a marker of multipotent BMSCs [29, 30]) and high levels of BSP, OP, and CI (Figs. 3D–3H). The percentage of CD105+ cells capable of forming a fibroblastic colony (CFU-F) was markedly higher after expansion in the 3D than in typical 2D cultures (29.4% vs. 10.7%, respectively). Taken together, these data suggest that BMSCs generated in the developed 3D system were neither early undifferentiated mesenchymal precursors nor fully differentiated osteoblast-like cells but comprised a large population of clonogenic osteoprogenitor cells. Future studies should address whether changes in the substrate used (e.g., scaffold composition or architecture), flow rate, and culture medium composition will regulate the phenotype, proliferation, and multilineage differentiation capacity of the expanded BMSCs.
Hematopoietic Cell Characterization
The finding that a substantial fraction of the cells cultured in the developed 3D system was not of the mesenchymal lineage, as suggested by the rounded morphology and demonstrated by the lack of expression of CD105, induced us to investigate whether both hematopoietic and mesenchymal cells were cocultured within the ceramic pores. Indeed, in the engineered constructs we found cells positive for CD45, a surface marker of hematopoietic cells, at percentages (30% ± 15%) equivalent to those of cells negative for CD105 (Figs. 4A–4I). It is likely that cocultured hematopoietic cells, possibly including CD14-positive adherent macrophages, regulated the phenotype of BMSCs  and played a critical role in determining the osteoinductivity of the constructs, possibly by maintaining a higher fraction of clonogenic BMSCs. It has been described that upon transplantation into a host animal, BMSCs form an ectopic ossicle in which bone cells, myelo supportive stroma, and adipocytes are of donor origin where as hematopoiesis and the vasculature are of recipient origin . Considering that in our 3D system human hematopoietic cells were coimplanted with BMSCs, future studies should aim at determining whether human cells contributed to hematopoiesis in this model.
We next hypothesized that, through the addition of specific medium supplements, the developed 3D culture model allows the regulation of the relative proportions of hematopoietic and mesenchymal cells. Using supplements typically used for culture of hematopoietic cells (i.e., interleukin-3, stem cell factor, and platelet-derived growth factor-bb, hematopoietic medium) , the fraction of CD45+ cells found after 19 days of 3D culture was increased to more than 90% (Fig. 4I) whereas BMSC proliferation capacity was still sustained (average of 4.5 ± 0.7 doublings in 19 days). Interestingly, the use of this culture medium further increased the percentage of CFU-F within CD105+ cells from 29.4%–38.8% and generated relevant fractions of hematopoietic CFUs, including those with a mixed phenotype, indicative of early multilineage progenitor populations (Fig. 4J). Remarkably, the use of the same medium supplements in 2D cultures was not able to modulate the fractions of hematopoietic/mesenchymal cells nor their clonogenicity, possibly due to the fact that most of the non-adherent cells were not entrapped within the 3D niches of the ceramic or newly formed stromal-like tissue and were thus discarded during medium changes. This evidence further highlights the potential of the developed culture system, in which the 3D configuration under perfusion flow provides an extension of the concept of stromal feeder layer for the support and development of hematopoietic cells [23, 32] and thus modifies standard paradigms for culture of bone marrow cells.
Our study validates the simple but innovative concept that BMSCs can be seeded and expanded by perfusion culture through the pores of 3D scaffolds starting from minimally processed bone marrow aspirates and avoiding 2D culture expansion. The developed approach was used for the reproducible, spatially uniform, highly efficient, and simplified manufacture of osteoinductive grafts. Incorporating in the system features like automated medium change, monitoring and control of pH, gases, and metabolites are likely to lead to the development of a closed system for the automated and controlled production of autologous BMSC-based bone substitutes. Compared with previously proposed perfusion systems [17, 33], the elimination of the 2D culture would allow for a one-phase, streamlined procedure that could thus generate engineered bone grafts at reduced costs and make them commercially viable against alternative off-the-shelf osteoinductive materials (e.g., based on the delivery of growth factors). In this context, however, scaling up of the procedure to clinically relevant sizes will have to address the challenge of maintaining cell viability in larger constructs, both during in vitro culture and upon grafting.
Beyond the relevance in the field of bone tissue engineering, our results validate the developed process as a first step toward ex vivo tissue engineering of bone marrow as a model to investigate proliferation, differentiation, and interactions among different types of bone marrow cells in a more physiological environment than previously established systems (e.g., Petri dishes or spinner flasks ). The developed culture system may be further explored for the expansion under perfusion of CD34+ hematopoietic stem cells from bone marrow or cord blood within an engineered 3D stromal network. Finally, the same paradigm of bypassing 2D expansion by direct 3D perfusion culture may be used for the engineering of other 3D tissues and organs.
We would like to thank Raffaella Arbicò, Andrea Barbero, Marcel Dueggelin, Anna Marsano, Anca Reschner, and Silvia Scaglione for assistance and cooperation in conducting this research and Walter Dick, Oliver Frank, and Stefan Schären for providing human bone marrow aspirates. We are grateful to Roberta Martinetti (Fin-Ceramica Faenza) for the generous supply of Engipore ceramic scaffolds.