Effect of donor variation on osteogenesis and vasculogenesis in hydrogel cocultures

Abstract To introduce a functional vascular network into tissue‐engineered bone equivalents, human endothelial colony forming cells (ECFCs) and multipotent mesenchymal stromal cells (MSCs) can be cocultured. Here, we studied the impact of donor variation of human bone marrow‐derived MSCs and cord blood‐derived ECFCs on vasculogenesis and osteogenesis using a 3D in vitro coculture model. Further, to make the step towards cocultures consisting of cells derived from a single donor, we tested how induced pluripotent stem cell (iPSC)‐derived human endothelial cells (iECs) performed in coculture models. Cocultures with varying combinations of human donors of MSCs, ECFCs, or iECs were prepared in Matrigel. The constructs were cultured in an osteogenic differentiation medium. Following a 10‐day culture period, the length of the prevascular structures and osteogenic differentiation were evaluated for up to 21 days of culture. The particular combination of MSC and ECFC donors influenced the vasculogenic properties significantly and induced variation in osteogenic potential. In addition, the use of iECs in the cocultures resulted in prevascular structure formation in osteogenically differentiated constructs. Together, these results showed that close attention to the source of primary cells, such as ECFCs and MSCs, is critical to address variability in vasculogenic and osteogenic potential. The 3D coculture model appeared to successfully generate prevascularized constructs and were sufficient in exceeding the ~200 μm diffusion limit. In addition, iPSC‐derived cell lineages may decrease variability by providing a larger and potentially more uniform source of cells for future preclinical and clinical applications.

an adverse effect of autologous bone grafting is the introduction of new skeletal defects, accompanied by donor site morbidity and chronic donor site pain in 28% of cases (Ross, Tacconi, & Miles, 2000). Future therapies for critical size bone defects would be a great improvement if they could eliminate the need for donor tissue while ideally maintaining the efficacy of autologous bone grafting. Regenerative medicine holds the promise to develop constructs outside of the body that can be created using the desired cell types, implanted at the defect site, and designed to meet the biomechanical demands.
The maximum size of vital engineered bone constructs has traditionally been limited by the poor diffusion of oxygen and nutrients to the core regions (reviewed by Carmeliet & Jain, 2000 and Rouwkema, Rivron, & van Blitterswijk, 2008). One potential strategy to solve this problem is to establish an integrated functional vascular network together with stimulation of osteogenesis and bone matrix formation, even without perfusion (reviewed by Teoh, 2015 andUnger et al., 2007).
One approach to create a vascular network in engineered bone tissue is to coimplant cell populations that can establish vasculature and can differentiate into osteogenic cells. Prior studies have emphasized the mutual importance of multipotent mesenchymal stromal cells (MSCs) in supporting vasculogenesis and the significance of endothelial progenitor cells (EPCs) in the stimulation of the osteogenic potential of MSCs, both in vitro and in vivo (reviewed by Frohlich et al., 2008, Guerrero et al., 2013, Li & Wang, 2013, Liu et al., 2012, Unger et al., 2007. A specific type of EPC named "endothelial colony forming cells" (ECFCs), especially those derived from cord blood, showed robust proliferative potential and inherent vasculogenic and angiogenic capacity and can contribute to de novo blood vessel formation in vitro and in vivo, in contrast to mature ECs (reviewed by Banno & Yoder, 2018). However, only few studies have addressed the simultaneous formation of in vitro prevascular networks and osteogenesis with these type of cells (Gawlitta et al., 2012, reviewed by Liu et al., 2015, Liu et al., 2012.
Several key steps towards clinical translation remain to be taken.
For example, the reproducibility and standardization of coculture protocols and outcomes of selected release criteria are essential in enabling quality control of the resulting prevascularized tissue constructs for clinical application. At present, it is not known how the choice and combination of cells from one or different donors affects coculture outcomes. Data regarding ECFC heterogeneity have only been rarely reported, whereas MSCs have been widely described to have heterogeneous characteristics, potentially influenced by isolation and culture methods, donor age, donor gender, and medical history (Bertram, Mayer, & Schliephake, 2005;Katsara et al., 2011;Phinney et al., 1999;Portalska et al., 2013;Schellenberg et al., 2011;Siddappa, Licht, van Blitterswijk, & de Boer, 2007;Stenderup, Justesen, Clausen, & Kassem, 2003). Nevertheless, the influence of heterogeneity of donor MSCs on their specific osteogenic potential has only been examined in monolayer culture models. Therefore, the first aim of the present study is to examine the effects of donor variation on the vasculogenesis and osteogenesis of a 3D coculture model of subcultured primary MSCs and ECFCs.
Translating 3D cocultures to the clinic by incorporation of (preferably) autologous cells raises another critical challenge: limited cell sources. Ideally, both MSCs and ECFCs would be derived from an autologous cell source to avoid an immunogenic response upon implantation. However, the isolation of ECFCs from the adult peripheral blood is inefficient compared with isolation from cord blood, which is usually an allogeneic cell source. To date, the use of (autologous) adult peripheral blood-derived ECFCs is considered unfavourable in the clinical setting because the prevalence of ECFC is extremely low (20 times lower than in cord blood), resulting in very low isolation yields that hamper the reproducibility and viability of possible therapies (reviewed by Banno & Yoder, 2018, Mund, Estes, Yoder, Ingram, & Case, 2012. Moreover, the angiogenic potential of peripheral blood-derived ECFCs appears substantially lower than that of their cord blood-derived counterparts (Ingram, Mead, Tanaka, et al., 2004, reviewed by Richardson & Yoder, 2011. Alternatively, induced pluripotent stem cells (iPSCs) could provide an unlimited source of clinically relevant, autologous endothelial cells with vasculogenic capacity. Consequently, our second goal was to evaluate the use of iPSC-derived endothelial cell precursors (iECs), differentiated from several independent iPSC lines in the Matrigel coculture model with MSCs, and assess their vasculogenic capacity and reproducibility in an osteogenic coculture model.
In the present study, a Matrigel coculture system in an osteogenic environment was developed in which donor dependency of vasculogenic and osteogenic cells and their behaviour could be assessed.

| Isolation of MSCs
MSCs were isolated from human bone marrow aspirates that were obtained from consenting patients (n = 7; aspiration procedure was approved by the local medical research ethics committee, University Medical Center Utrecht) that underwent different surgical procedures (indicated in Table 1). These isolates are referred to as MSC1 to MSC7.
Researchers were blinded to the medical history of the donors.

| Isolation of ECFCs
Cord blood of seven different donors was used (procedure was approved by the medical research ethics committee, University Medical Center Utrecht, informed consent was obtained from the mothers) as a source for the ECFC isolation (hereafter referred to as ECFC1 to ECFC7; researchers were blinded to the condition of the mother and donor child). Cord blood was diluted with PBS (1:3), and the mononuclear cell layer was retrieved after centrifugation (400 × g) on 1.077 g/ml Ficoll-Paque gradient (GE Healthcare). 20 × 10 6 MNCs were plated in a 50-μg/ml collagen type I-coated (BD Biosciences, rat tail) well of a six-well plate with 1 ml of complete endothelial growth medium-2 (EGM-2) containing Endothelial Basal Medium-2 + SingleQuots (Lonza), 100 U/ml-100 μg/ml PenStrep, and 10% heat-inactivated FBS. The medium was changed daily until day 7 and then three times per week. Between weeks 2 and 4, ECFC colony outgrowth was observed. When individual colonies expanded, but did not touch each other yet, the cells were trypsinized and replated into collagen type I-coated culture flasks at a density of 7,000 cells/cm 2 . Complete EGM-2 medium was used for subsequent cell expansion. After isolation, ECFCs were either expanded or frozen and used between passages 7 and 12.  Figure S1).

| In vitro MSC-ECFC cocultures in Matrigel
Cocultures were performed in growth factor-reduced Matrigel (354230, BD Bioscience). The samples were prepared by mixing  50 μl ODM, containing both cell types, with 50 μl Matrigel. Each sample of 100 μl gel/ODM contained a total cell volume of 625,000 cells (ratio of 4:1 MSCs to ECFCs) and was pipetted into a 12-well plate.
The mixture was allowed to form a hydrogel at 37°C for at least 1 hr, resulting in a hemispherically shaped construct, after which 1 ml ODM was added on top. The medium was changed every 3-4 days. On day 10 or day 21, hydrogels were fixed with formalin (10%) and stored in PBS (4°C) until further processing.
To assess prevascular structure formation in large constructs,  Table 2 (three donors for each cell type, n = 3 gels per combination). Table 3 indicates the combinations used to study the influence of MSC and ECFC variation on the osteogenic properties of the constructs (i.e. osteonectin expression and mineralization) (N = 3 independent repetitions, n = 2 gels per combination).

| Osteocalcin/osteonectin staining on paraffin sections
Upscaled cylindrical constructs and iEC-containing cultures were

| Von Kossa staining on paraffin sections
Von Kossa staining was used to visualize mineral deposition in the cocultures (

| Alkaline phosphatase activity in hydrogels
ALP was used to indicate early osteogenesis, as the activity is associated with committed osteoprogenitor cells. For detection of ALP activity (in the samples shown in Table 2) on day 10, the fluorescent Vector SK5100 kit was used. In order to link the ALP activity with the vasculogenesis in one half of a construct, the remaining half of the construct was used for the CD31 staining (Table 2). Constructs were permeabilized and incubated with the kit's Red Substrate in the dark followed by washing in demi water.

| Quantification of the prevascular structures
The influence of donor variation on angiogenic properties was further assessed by quantifying the total length of all CD31-positive structures in the images (length in pixels). Images of the CD31-stained cocultures (n = 3 per group) of Table 2 were captured (Olympus BX60, Cell-F software) and processed in Adobe Photoshop CS6. The levels were individually adapted by eliminating over-exposed and under-exposed pixels following conversion of the images to black and white. Subsequently, the images were inverted (brightness −150 and contrast 100). The resulting files were batch processed in the freeware programme "Angioquant" (Niemisto, Dunmire, Yli-Harja, Wei Zhang, & Shmulevich, 2005). All images had individually optimized processing settings for smoothening, segmentation with automatic thresholding, and pruning of structures below 25 pixels. Then, the total length of the prevascular structures (total length in one image of a construct at day 10) was assessed.

| Statistical analyses
Statistical analyses were performed with GraphPad Prism 6.01.
Angioquant outcomes were tested for the significant influence of donor variation on the total structure length with a two-way ANOVA with multiple comparisons of the mean, when varying the donor of the

| Identification and characterization of MSCs and ECFCs
As expected, the bone marrow-derived MSCs showed a fibroblast-like morphology, their multi-lineage potential was confirmed, and their CD-marker expression profile was consistent with accepted criteria for MSCs (Dominici et al., 2006;Figure S1a-d). ECFCs organized into their characteristic colonies with rounded cell morphology and exhibited a cobblestone morphology upon passaging. The ECFCs showed high expression of endothelial/haematopoietic stem cells markers ( Figure S2).

| Formation of 3D prevascular structures in MSC-ECFC osteogenic cocultures
Hemispherical 100 μl-sized cocultures of MSCs with ECFCs in Matrigel, cultured in osteogenic differentiation medium, produced CD31-positive 3D pre-vascular networks by day 10 (Figure 1a). At day 21, an even more extensive network with an increased number of junctions was observed (Figure 1b). Pseudopodial processes, on sprouting tip cells, were observed in the constructs at both day 10 and 21, indicating ongoing angiogenesis (Figure 1c). In addition, pericytic mural cells adjoining the CD31-positive prevascular structures were visualized by NG2 ( Figure 1d) and α-SMA detection ( Figure 1e).
The nuclear staining of the cocultures demonstrated that not all cells contributed to vasculogenesis or differentiated towards mural cells (Figure 1f). Most likely, these cell nuclei belong to cells which are committed to the osteogenic lineage (as presented in Figure S4).

Monocultures of either MSCs or ECFCs did not show prevascular net-
work formation or α-SMA-positive network formation ( Figure S3).

| Cocultures in larger 3D constructs
To confirm 3D vascular network formation in constructs of a consider-

| The effects of donor variation on prevascular network formation
All evaluated donor combinations (Table 2)  Similarly, by varying MSC donor while the ECFC donor was unchanged, a significant difference in the prevascular network length was observed (e.g., M1E1 vs M2E1 and M2E1 vs M3E1). From this, it can be concluded that both the MSCs' and ECFCs' donor individually influence the length of the prevascular structures, thus the angiogenic properties of the coculture. Overall, the specific donor combination of MSC-ECFC rather than one cell type has the overhand in determining the length of the pre-vascular network; that is, the cell types react differently when combined with cells from different donors.

| Osteogenic differentiation in MSC-ECFC cocultures
To confirm simultaneous osteogenic differentiation and vasculogenesis in the cocultures, the constructs analyzed in Figure 3 were also evaluated for ALP activity. Qualitative analysis confirmed that the nine coculture combinations displayed ALP activity throughout the construct ( Figure S4). Among these combinations, minimal variation in the intensity of the ALP staining was visible. An increase of ALP activity was found in MSC-ECFC cocultures compared with the corresponding MSC monocultures ( Figure S4, insets).
To further confirm osteogenic differentiation in the cocultures, ON expression was evaluated in the MSC-ECFC combinations    was also observed when larger constructs were generated, in which the size reached the often described diffusion limit of 200 μm (reviewed by Carmeliet &Jain, 2000, andRouwkema et al., 2008).
The addition of ECFCs to the MSCs enhanced the osteogenic differentiation and mineralization. In contrast to most coculture models (reviewed by Liu et al., 2015), processes (Bidarra et al., 2011, reviewed by Hirschi, Ingram, & Yoder, 2008, Loibl et al., 2014, Nassiri & Rahbarghazi, 2014 Thus, by deriving iECs from patient-derived iPSCs, the drawbacks of using cord blood-derived ECFCs can potentially be overcome. Moreover, iPSCs exhibit an unlimited proliferation potential and therefore present an attractive (autologous) cell source for future regenerative treatments (Samuel et al., 2013). Furthermore, iECs might be able to reduce variation as iPSC-derived endothelial cells were shown to be produced with high batch uniformity (Belair et al., 2015) and less variance than primary ECs (White et al., 2013), which has also been demonstrated in our iEC-MSC cocultures ( Figure 6).
So far, several research groups have reported successful creation of prevascular structures derived from iPSCs. iECs have been shown to repopulate decellularized tissue-engineered vascular structures (Margariti et al., 2012), (self)assemble into perfusable tubular structures (Belair et al., 2015;Zanotelli et al., 2016), and form 3D networks in vitro Orlova et al., 2014) and in vivo (Samuel et al., 2013). Orlova et al., (2014) succeeded in coculturing iPSC-derived ECs and iPSC-derived pericytes in a 2D environment, and demonstrated that iPSC-derived ECs were able to functionally integrate into embryonic zebrafish vasculature. Nevertheless, before iPSCs can be integrated in therapies in clinical settings, safety concerns must be addressed.
Overall, our study demonstrates that care should be taken when varying donors of cells in coculture models as donor variation can affect cell differentiation and thus the reproducibility of results. It is imperative that several important hurdles towards clinical translation of functional prevascularized bone constructs are taken. This does not only include the choice of (autologous/allogeneic) cell sources but also upscaling of size, reproducibility, and standardization of (co) culture protocols and release criteria. Our in vitro 3D coculture model is an accessible method to explore new regenerative strategies to overcome these hurdles and move towards in vivo applications.

SUPPORTING INFORMATION
Additional supporting information may be found online in the Supporting Information section at the end of the article.     (Table ) after 10 days of culture. All donor combinations showed ALP activity (red) to a similar extent. Images are representative of the triplicates and ALP activity of corresponding MSC monoculture controls can be found in the insets.