Expansion of Human Mesenchymal Stromal Cells from Fresh Bone Marrow in a 3D Scaffold-Based System under Direct Perfusion

Mesenchymal stromal/stem cell (MSC) expansion in conventional monolayer culture on plastic dishes (2D) leads to progressive loss of functionality and thus challenges fundamental studies on the physiology of skeletal progenitors, as well as translational applications for cellular therapy and molecular medicine. Here we demonstrate that 2D MSC expansion can be entirely bypassed by culturing freshly isolated bone marrow nucleated cells within 3D porous scaffolds in a perfusion-based bioreactor system. The 3D-perfusion system generated a stromal tissue that could be enzymatically treated to yield CD45- MSC. As compared to 2D-expanded MSC (control), those derived from 3D-perfusion culture after the same time (3 weeks) or a similar extent of proliferation (7–8 doublings) better maintained their progenitor properties, as assessed by a 4.3-fold higher clonogenicity and the superior differentiation capacity towards all typical mesenchymal lineages. Transcriptomic analysis of MSC from 5 donors validated the robustness of the process and indicated a reduced inter-donor variability and a significant upregulation of multipotency-related gene clusters following 3D-perfusion- as compared to 2D-expansion. Interestingly, the differences in functionality and transcriptomics between MSC expanded in 2D or under 3D-perfusion were only partially captured by cytofluorimetric analysis using conventional surface markers. The described system offers a multidisciplinary approach to study how factors of a 3D engineered niche regulate MSC function and, by streamlining conventional labor-intensive processes, is prone to automation and scalability within closed bioreactor systems.


Introduction
MSC are receiving an increasing experimental and clinical interest, owing to the large degree of plasticity and the capacity to modulate the immune system or the phenotype of cancer cells [1]. Their use is thus advocated for treatment of various genetic, haematologic or immunologic pathologies and in the emerging field of regenerative medicine [2][3][4]. For most of these potential applications, given the low frequency among bone marrow nucleated cells (around 0.01%), MSC are typically expanded by sequential passages in monolayer (2D) cultures. However, this process is associated with a progressive reduction of their clonogenicity and multilineage differentiation capacity, and is often accompanied by cellular senescence [5,6].
Studies on different cellular systems have led to the concept that maintenance of 'early progenitor' properties generally requires a tissue-specific microenvironment or niche [7][8][9][10][11], which can hardly be resembled by the plastic substrate and 2D configuration of tissue culture flasks [12]. Various attempts have thus been reported to expand MSC in three-dimensional (3D) environments, based on suspension culture in the presence of dynamic flow [13,14], on microcarrier beads [15][16][17] or on a rotating bed bioreactor system [18,19]. Despite the promising results obtained, however, these approaches required an initial phase of MSC growth on plastic, which is intrinsically associated with the selection of the adherent cellular fractions, possibly already depleted of the less adherent earlier progenitors [20], and the loss of most hematopoietic lineage cells. Indeed, non-mesenchymal bone marrow cells were proposed to be involved in regulating MSC function [21] and have been demonstrated to enhance growth of MSC with clonogenic properties [22,23].
We previously reported that the continuous perfusion of freshly isolated human bone marrow cells directly through the pores of 3D ceramic-based scaffolds resulted in the reproducible generation of tissue constructs, which were highly osteogenic upon ectopic implantation in nude mice [24]. By eliminating the 2D culture step, the system not only streamlined the MSC culture process, but also supported the maintenance of hematopoietic lineage cells, including some of the early progenitors (i.e., CFU-GEMM), thereby establishing some features of the bone marrow niche [25].
In this study, we aimed at investigating the use of the above described 3D scaffold-based perfusion system for human MSC expansion. For this purpose, the generated constructs were enzymatically processed and the retrieved cells were phenotypically and functionally compared to those generated following conventional expansion protocols. Furthermore, a microarray analysis was introduced to identify potential new molecular markers and pathways differentially regulated as well as to validate the robustness of the process across different donor preparations.

MSC Culture
Using a perfusion bioreactor system described in [26] and now commercially available by Cellec Biotek AG (http://www. cellecbiotek.com), an average of 66610 6 freshly isolated bone marrow-nucleated cells were perfused for 5 days through 8-mmdiameter, 4-mm-thick disks of porous (total porosity, 83%63%; accessible surface area 3200 cm 2 ) hydroxyapatite ceramic (Engipore; Fin-Ceramica Faenza, Faenza, Italy, http://www. finceramicafaenza.com) at a superficial velocity of 400 mm per second. After 5 days, culture medium was replaced and perfusion culture was performed at a velocity of 100 mm per second for additional 14 days and changing the medium twice per week. In order to establish a comparison with the standard culture process, MSC expansion in 2D (in 56 cm 2 Petri dishes; BD Biosciences) was performed for up to 19 days without passaging using similar initial cell numbers/surface area and schedule of medium changes, as in the 3D expansion condition.

Cell Extraction
At the end of the expansion phase in the 3D culture system, cells were extracted by substituting the CM with a solution of 0.3% collagenase (collagenase) and perfusing the ceramic constructs for 40 min followed by 0.05% trypsin/0.53 mM EDTA solution (trypsin) for additional 15 min both at 400 mm per second. Extracted cells were subsequently sorted using anti-CD45-coated magnetic beads (Miltenyi Biotec, Auburn, CA), according to the manufacturer's instructions. 2D-expanded cells were retrieved by using the same enzymatic solutions, i.e. collagenase for 40 min and trypsin for 5 min. The fraction of dead cells, preliminarily assessed by assessed by Trypan blue exclusion (Sigma, Switzerland), was negligible (less than 3%), with no obvious differences between the experimental groups. Both CD45 + and CD45 2 viable cell populations were assessed for the ability to form fibroblastic colonies. The CD45 2 populations were further characterized by flow cytometry, gene expression by means of microarray analysis and quantitative real-time (QRT) PCR or tested for the multilineage differentiation capacity, as described below.
Clonogenicity (CFU-f) and flowcytometry assays CFU-f assays (n = 5) of bone marrow or expanded cells were performed by plating 4400 freshly isolated mononucleated cells or 4 expanded cells per cm 2 in tissue culture dishes, respectively. The procedure was optimized following preliminary experiments with serial dilutions of plated cells. After 14 days of culture, cells were fixed in 4% formalin, stained with 1% methylene blue and the number of colonies was counted.

RNA Extraction and Microarray analysis
Total cellular RNA (40 ug) was extracted from 2D or 3Dperfusion expanded CD45-cells, obtained from 5 independent experiments/donors, using RNeasy Micro kit (Qiagen, Valencia, CA) following the protocol supplied by the manufacturer. RNA were hybridized to Affymetrix Human HG-U133plus2 GeneChip arrays according to the manufacturer recommendations. All the data have been deposited in the Gene Expression Omnibus database with experiment series number GSE52896 available at http://www.ncbi.nlm.nih.gov/geo/query/acc. cgi?acc = GSE52896.
Arrays pre-processing and analysis were performed using R and the Bioconductor package (http://www.bioconductor.org/) and passed through array quality control using the AffyQCreport tool. Raw intensities were normalized using RMA and scaled to a 2% trimmed mean of 150. Probes with normalized expression values below 50 in both groups were filtered out. Differential gene expression was performed using Limma. Probes were annotated using the platform annotation file version 31 from NetAffx. Genes with a fold change higher than 2 and an adjusted P-value below 0.05 (Benjamini and Hochberg multiple testing correction) were considered regulated. Data from microarrays were analysed by Principal Component Analysis (PCA) using TM4 Multi Experiment Viewer (MeV), available at http://www.tigr.org/software/ tm4/mev.html in order to ascribe the overall variability of the sample to a limited number of variables.
To validate microarray data, the expression of a set of genes was evaluated by quantitative real-time (QRT) PCR ( Figure S2). Total RNA extraction, cDNA synthesis and real-time reverse transcriptase-polymerase chain reaction (RT-PCR; 7300 AB Applied

Database for Annotation, Visualization and Integrated Discovery (DAVID) and Cytoscape
Functional enrichment analysis for up-and down-regulated genes (2 fold with an adjusted pvalue below 10 22 ) was performed using the open-source web-based DAVID platform (http://david. abcc.ncifcrf.gov/) including Gene Ontology (GO) and Pathways categories. Enriched functional categories and pathways were clustered by gene overlap using Enrichment Map in Cytoscape [27,28] and labelled for recurrent keywords using the WordCloud plugin (http://baderlab.org/Software/WordCloudPlugin). In the generated Cytoscape diagram, the node size is proportional to the number of genes defining the node. Edges connect nodes that share common genes and edge thickness is proportional to the number of shared genes between nodes.

Multilineage differentiation assays
The osteogenic differentiation capacity was tested by culturing cells, obtained from 3 independent experiments/donors, for 2 weeks in CM further supplemented with 100 nM dexamethasone, 10 mM b-glycerophosphate, and 0.05 mM ascorbic acid-2-phosphate. After 2 weeks, cell layers were either stained with alizarin red solution to evidence mineral deposition or assessed for ALP activity normalized to cell numbers, as previously described [29]. Shorter culture time with respect to the commonly used in literature 3 weeks protocol was chosen in order to maximize the differences regarding the in vitro osteogenic differentiation capacity of MSC between the two experimental conditions.
The adipogenic differentiation capacity was tested by alternating cycles of cell culture with different media, including 10 mg/ml insulin, 10 mM dexamethasone, 100 mM indomethacin, and 500 mM 3-isobutyl-1-methyl xanthine (adipogenic induction medium) or 10 mg/ml insulin (adipogenic maintenance medium) as previously described [30]. After a total of 14 days, the presence of adipocytes was microscopically documented and quantified following Oil red-O staining.

Immunosuppression assay
The proliferation of CD4+ T cells, sorted from PBMCs of a healthy donor, in the presence of MSC was performed in 96-well plates following a method described [31]. Briefly, 2D-and 3Dperfusion expanded MSC were seeded at densities of 1250, 5000 and 20000 cells per well and allowed to attach at least 4 h at 37oC with RPMI1640 medium supplemented with 10% FBS, 10 mM HEPES buffer, 1 mM sodium pyruvate, 10000 U/ml penicillin and 10000 mg/ml streptomycin (all from GIBCO, Switzerland), before adding 100000 CD4+ cells in the presence of 1 ug/ml of the mitogen phytohemagglutinin (PHA; Remel Europe Ltd. Clipper Boulevard West, Crossways Dartford,Kent, DA26PT UK). After 56 h of co-culture, 1 mCi/well 3 H-thymidine (GE Healthcare, Little Chalfont, United Kingdom) was added to each well and incubated for additional 16 h. Cells were then harvested and the 3 H cpm counted by a scintillation beta-counter to measure the radioactivity in DNA recovered from the cells in order to determine the extent of cell division. Each condition was tested in triplicate.

Statistical analysis
For the microarray analysis, genes with a fold change higher than 2 and an adjusted P-value below 0.05 (Benjamini and Hochberg multiple testing correction) were considered regulated. Results are reported as mean 6 SD. Statistical analysis was performed with GraphPad Prism 4.0 (Graph Pad software, La Jolla, CA, USA). Differences were assessed using Mann-Whitney U-tests and considered statistically significant with P,0.05.

3D-perfusion expansion of freshly isolated MSC
Using a bioreactor system as described in [26] and graphically illustrated in Figure 1, total BM cells were perfused through the scaffold pores for 5 days (cell seeding phase), followed by perfusion of culture medium for further 14 days (cell culture phase). Based on the retrospectively calculated density of CFU-f from the five donors (0.08%60.02%) and assuming that all CFU-f attached to the ceramic scaffolds, an estimated average of 5.4610 3 MSC were perfused through each scaffold, corresponding to 1.6 MSC per cm 2 of ceramic surface area. This process resulted in the formation of stromal-like tissue structures, including cells of heterogeneous morphologies in physical contact with each other (Figure 2a). Instead, conventional cell culture in Petri dishes using similar cell density per surface area led to the generation of adherent cells, typical of the fibroblastic phenotype (Figure 2b). Enzymatic retrieval of the cells from both conditions and labelling for CD45 indicated the presence of a significantly higher percentage of cells of the hematopoietic lineage after expansion in 3Dperfusion as compared to 2D (19.365.7% vs 6.064.5% CD45+) (Figure 2c). The extent of MSC proliferation in the 3D perfusion system, assuming that all harvested CD45-cells (total of 1.36+/2 0.34610 6 cells/scaffold) were of the mesenchymal lineage and derived from the initial relative number of seeded CFU-f, retrospectively estimated by clonogenicity assays, was of 7.661.7 doublings, corresponding to about 0.4 doublings/day. MSC growth in plastic dishes within the same time frame of 19 days (total of 1.76+/20.46610 6 cells/dish) was significantly higher, corresponding to 0.74 doublings/day (Figure 2d). Based on the measured cell yields, the same numbers of cells occupying 56 cm 2 of a 10 cm diameter Petri dish could be expanded in ,0.2 cm 3 of scaffold volume.

Phenotypic characterisation of 3D-perfusion expanded MSC
In order to investigate the phenotype of the mesenchymal cells, retrieved cells were negatively sorted for the expression of CD45. Results displayed in Figure 3 indicate a large overlap in the cytofluorimetric profile of the two cell-expanded groups, without clear-cut differences in the presence or absence of specific cell populations. However, as compared to 2D-expansion culture, a lower percentage of MSC expanded by 3D-perfusion expressed CD90 78.2% vs 99.8%), CD105 (61.2% vs 98.9%), CD166 (87.1 vs 99%) and ALP (5.8% vs 18.5%), a marker associated with the osteoblastic differentiation of MSC. Moreover, slightly higher percentage of 3D-perfusion expanded MSC were positive for HLA-DR (22.8% vs 10.8%) or for CD146 (25.2% vs 11.6%) and SSEA-1 (11.4% vs 7.6%), which were proposed to be associated with progenitor cell properties [32][33][34][35][36][37].

Microarray analysis of 2D-and 3D-perfusion expanded MSC
In order to broaden the search of potential differentially expressed markers and to validate the robustness of the process across different donor preparations, the CD45-fractions of bone marrow cells expanded by 3D-perfusion or in 2D from 5 independent donors were profiled using expression microarrays. Exploratory analysis using PCA (data dimensionality reduction) was performed in order to reveal correlations between the samples [38]. By dot-plotting the data derived by the two experimental groups, it is possible to estimate the similarity between each sample as a function of the distance of each pair of dots. This analysis shows a striking separation of the samples from 2D and 3Dperfusion on Principal Component 1 (PC1), confirming that culture conditions represented the most influential factor in discerning among cell preparations (Figure 4a). Interestingly, samples derived from cells cultured in 2D were more spread along the PC2 axis as compared to 3D-perfusion ones, suggesting a higher inter-donor variability induced by 2D-expansion. After preprocessing, we identified 702 genes (343 up-regulated and 359 down regulated) with a fold change of 2 and an adjusted p-value of 10 22 . A list of the 10 more up-and down-regulated genes is reported in Table 1.
In order to investigate the pathways associated with the 2D versus the 3D cultures, we performed a Gene Ontology (GO) enrichment and cluster analysis using the online web-platform DAVID on the derived list of regulated genes. The statistically enriched pathways can be visualized as an enrichment map with nodes being pathways and edges representing the overlap in genes in these pathways (Figure 4c). The main GO categories increased in 3D-perfusion vs 2D-expanded MSC were the ''Monosaccha- Figure 1. Schematic overview of the experimental setup. Bone marrow aspirates were seeded into the 3D perfusion system and in conventional Petri dishes. After culture, cells from both systems were enzymatically retrieved and CD45-sorted cells using magnetic beads were analyzed as described. doi:10.1371/journal.pone.0102359.g001 rides metabolic processes (fructose and glucose)'', ''Chemokine activity'', ''Inflammatory response'', ''Response to hypoxia'' and ''Negative regulation of apoptosis'' (Table S1). Consistent with the multicellular tissue-like morphology observed during 3D-perfusion expansion, the GO functional categories ''Positive regulation of multicellular organismal process'' and ''Extracellular space'' were also significantly over-represented in the list of up-regulated genes in 3D culture. Conversely, GO categories related to ''Fat-related (phospholipid and sphingolipid) and organophosphate metabolic processes'' as well as to cytoskeleton, contraction and adhesion were found to be decreased in the 3D-perfusion vs 2D-expanded MSC (Table S2). Figure 4b displays a representation of the resulting GO categories and the identified pathways from Tables S1 and S2, linked to the underlying biological processes. Interestingly, ''Bone development'' was found to be up-regulated in the 2D-expanded cells and appeared to be consistent with the increased protein expression of ALP. Furthermore, one of the most significant 3D up-regulated geneset uncovered using GSEA analysis is the PluriNet [39], a matrix of global gene expression profiles of various types of stem cells, supporting the more ''stem cell'' like transcriptional footprint of our 3D-perfusion model. Also the geneset describing osteoblastic differentiation was downregulated in the 3D-perfusion vs the 2D-expanded MSC (Figure 4c).

Validation of the in vitro functionality of 3D-perfusion expanded MSC
We next investigated whether the differential gene expression accounting for multi-potency maintenance and differentiation was mirrored in the functionality of CD45-cells expanded by 3Dperfusion or 2D. The CD45-fraction of the 3D perfusion-cultured cells included a 4.3-fold higher percentage of clonogenic cells (Figure 5a) than that of cells expanded on plastic for the same time (respectively 17% vs 4%), suggesting a better preservation of progenitor cell features. This hypothesis was further confirmed by a more efficient multi-lineage differentiation capacity upon exposure to typical chondrogenic, osteogenic and adipogenic conditions, as determined by histochemical and quantitative biochemical assays (Figure 5b). Since the number of doublings by MSC expanded for 19 days under 3D-perfusion or 2D was different, cell populations were compared also using a shorter culture time in 2D (i.e., 14 days), leading to 9.1 total doublings and thus more similar to the 3D-perfusion group. Both the clonogenic cell fraction and the multilineage differentiation profile of the shorter expansion time in 2D (data not shown) were comparable to those determined for the longer expansion time. Notably, both 2D-and 3D-perfusion expanded MSC cells shared similar antiproliferating effects on activated CD4+ cells when co-cultured in vitro ( Figure S1). Lastly and as expected, CD45+ cells from both experimental groups did not contain adherent fibroblastic   clonogenic cells when re-plated in Petri dishes, confirming efficient magnetic depletion (data not shown).

Discussion
We have developed a system for the expansion of MSC which entirely bypasses the use of 2D surfaces by seeding and expanding fresh bone marrow preparations directly within the pores of 3D scaffolds under perfusion flow. As compared to the conventional 2D culture system, MSC expanded under 3D-perfusion (i) preserved better their early progenitor properties, as they maintained a higher clonogenicity and a superior multilineage differentiation capacity, (ii) did not lose their anti-proliferative function, based on a standard in vitro assay typically used to claim 'immunomodulation' properties, and (iii) displayed reduced interdonor variability and consistent upregulation of multipotencyrelated pathways, as assessed by transcriptomic analysis.
Identifying a strategy for efficient expansion of MSC preserving their functionality is a critical target towards fundamental mechanistic studies on their biological properties, as well as for their prospective clinical use in the field of tissue engineering and regenerative medicine [12]. Among several hurdles, the absence of phenotypic markers that uniquely identify populations of MSC with specific functions challenges the definition of a quality control during MSC culture [40]. Indeed, surface proteins typically used to characterize MSC [41] were not differentially expressed in cells expanded in 2D or by 3D-perfusion, indicating that they are not suitable to capture functional features related to superior clonogenicity and multilineage differentiation. Only a limited set of markers, including CD146 and SSEA-1, were expressed by a larger percentage of 3D-perfusion expanded MSC, consistent with the proposed association of those markers to earlier progenitor/ stem populations of MSC [34][35][36][37]. HLA-DR was found to be expressed in both conditions, likely due to the presence of FGF-2 in the culture medium [42,43]. The relatively higher HLA-DR expression observed in 3D-perfusion condition is consistent with previous observations on the effect of hematopoietic cells on MSC [43] and did not alter the anti-proliferative effects of MSC on Tcells. The broader impact of HLA-DR expression on the immunomodulatory properties of MSC is still subject of debate.
A genome wide comparison demonstrated a clear separation between the transcriptomes of MSC expanded in 2D or 3Dperfusion as evidenced by PCA; moreover, a reduced dispersion of 3D-perfused samples indicates that culture conditions can diminish the inter-donor variability that typically affects 2D cultures. Gene set enrichment analysis further demonstrated that following expansion under 3D-perfusion, MSC up-regulated or maintained a transcriptome profile similar to that of other stem cells, supporting the superior maintenance of the experimentally verified MSC multipotency. In this context, after the expansion phase bone related pathways were found down-regulated in the 3D-perfusion group, further indicating better preservation of an undifferentiated MSC phenotype, against the default progression towards the osteoblastic lineage [44]. Consistently, epidermal growth factor like ligands, which were highly upregulated in the 3D-perfusion dataset, were previously shown to be important for maintenance of osteoprogenitor cells at an undifferentiated stage [45]. Following the differentiation induction phase, 2D-expanded cells, displayed a limited osteogenic profile, despite their apparently more ''osteoblastic'' phenotype after expansion. Although an in vivo test of osteogenicity was not performed, the in vitro data seem to indicate that the spontaneous tendency to express osteoblastic genes does not necessarily reflect a superior efficiency of functional differentiation.
Amongst the highest differentially upregulated genes found in the 3D-perfusion dataset, several ones coded for toll-like receptors (TLR), interleukins (IL) and other chemokines which are known to be involved in processes of cell migration, tissue homeostasis and repair, as well as in the regulation of immunologic responses. In particular, the higher expression of TLR-2 together with IL-6 and IL-8 may indicate the activation of the receptor by its associated ligands, which has been previously proposed to regulate MSC multipotency [46]. The established 3D transcriptional profiles described here highlighted differential expression of several transmembrane related genes, which may represent a starting point for future studies to define novel markers for the prospective isolation of earlier MSC progenitors.
Some of the categories identified from GO enrichment analysis during 3D-perfusion expansion were related to hypoxia, negative regulation of apoptosis and cell metabolism. Previous studies reported the positive role of hypoxia, a physiological feature of the niche of MSC [47], on the cell maintenance in an undifferentiated state, with metabolic features associated with an extended and more genetically stable lifespan [48]. In the described 3Dperfusion culture system, oxygen gradients and thus hypoxic regions may have occurred as a result of the relatively low rate of fluid flow passing through compact areas of cell-laid ECM. Future studies will have to further explore the role of hypoxia by either using smaller sized scaffolds, thereby enhancing oxygen transport, or by performing 2D cultures in hypoxic conditions. For 2D expanded cells, cytoskeletal binding, contractile fiber and adherence junction pathways were up-regulated. These biomechanical ECM-induced processes were previously reported to influence cell fate [49,50] and induce osteogenesis of MSC, independently from the culture conditions [51]. Indeed, it has been shown that MSC sense the stiffness of their environment through physical contact and contraction of ECM proteins, which are deposited according to the rigidity of the underlying material surface [50]. Here, the up-regulation of these processes in 2D expanded MSC may be possibly explained by their continuous exposure to the rigid surface of plastic, in contrast to the 3Dperfusion system, where cells were progressively embedded within ECM (Figure 2), of most likely lower stiffness.
The two experimental conditions for MSC expansion differed in multiple parameters of various nature (e.g., 3D vs 2D configuration, ceramic vs plastic substrate, flow-induced shear vs static environment, maintenance vs loss of hematopoietic cells), which can hardly be de-coupled to establish appropriate controls. Thus, while confirming the influence of a dynamic 3D environment on MSC properties [52], the identification of the mechanisms leading to a more functional population of MSC when expanded under 3D-perfusion is beyond the scope of the present work. It is likely, however, that the 3D structure of the scaffold is instrumental to entrap various cells types, including hematopoietic cells [53], and supports the deposition and presentation of extracellular matrix signals known to positively regulate MSC expansion [54,55]. Based on the recent finding that typical stromal populations can form the niche to earlier, less adherent MSC progenitors which are removed with medium changes [20], it would be tempting to speculate that the stromal cell network generated within the scaffold offers the environmental cues required to support maintenance in culture of the earlier progenitors.
In the present study, a ceramic-based material has been used as a surface for initial adhesion and growth of MSC, in order to mimic some features of the mineralized trabeculae surrounding a marrow stromal tissue. It is likely that the use of materials of different composition, architecture and surface properties would provide different priming signals to marrow cells. Thus, the choice of the scaffold included in the perfusion chamber could represent a critical parameter of the system and at the same time an additional tool to dissect the role of specific factors in maintenance of MSC features. The 3D culture process critically requires the use of direct perfusion, initially in order to uniformly distribute cells throughout the scaffold pores [26] and later to efficiently nourish the cells down to the scaffold core. Moreover, the induced perfusion would also mimic the physiological role of interstitial fluid flow and associated mechanical shear in the bone environment [56][57][58]. Our previous study, though with animal derived BMSC and slightly different medium composition, has indicated the effect of continuous flow during culture in maintaining the presence of hematopoietic lineage cells [59]. Therefore, an experimental setup involving 3D cell cultures under static conditions or by perfusing pre-sorted CD45-cells from bone marrow preparations could identify the role of hematopoietic lineage cells in the maintenance of MSC functionality.

Conclusions
In this work we have proposed an unprecedented paradigm for human MSC expansion, which -unlike most so far reported methods -does not rely on plastic adherence to initiate the culture. The described system relies on the in vitro establishment of a 3D stromal environment as a biomimetic niche supporting MSC growth while better preserving their functional properties. The complete elimination of the labour-extensive serial passaging in monolayer and the use of a perfusion-based bioreactor open the perspective of a streamlined, automated and controlled MSC expansion within closed systems, possibly addressing not only cell quality issues but also cost effectiveness and standardization of the manufacturing process for clinical and industrial implementation [60]. From a scientific perspective, the culture method offers the possibility to systematically investigate how different parameters (e.g., scaffold composition, architecture and functionalization, flow rate) regulate the phenotype, growth and function of the generated cell populations, and could be used as an engineered 3D model of the bone marrow stromal environment to study physiological interactions among multiple cell types. Finally, the approach may be extended to other stem cell systems, of interest in fundamental research, molecular medicine and cellular therapy. Figure S1 Anti-proliferative effect of MSC, expanded either by 3D-perfusion or 2D, on CD4+ activated cells. (TIF) Figure S2 QRT-PCR evaluation of gene expression for selected genes for the preparations (prep) derived from five different donors to validate the microarray data. Legends: 3D represents the 3D-perfusion condition. (TIF) Table S1 Biological processes correlated with MSC genes that were up-regulated at least two-folds in 3Dperfusion as compared to 2D cultures. Terms are ordered according to their p-values. (DOCX)