All materials used (unless specified) were purchased from Sigma-Aldrich (Italy). A 5% w/v gelatin solution was prepared by dissolving gelatin (Type A, G1890, 300 bloom strength) in deionized water. Porous gelatin scaffolds were prepared with a multi-step procedure. The solution was stirred for 1 h at 50°C, allowing complete dissolution, then casted in cylindrical shaped mold and physically gelled at room temperature. Samples were kept at 4°C for 1 h, and then at −20°C overnight. Gelatin samples were then freeze-dried (−50°C, 150 mBar) to obtain a porous structure as described elsewhere (Lien et al., 2009). Samples were swollen in deionized water and then cross-linked by immersion in a 10 mM glutaraldehyde (GTA) solution in 40% v/v ethanol/deionized water. The crosslinking reaction was controlled by keeping the glutaraldehyde/gelatin ratio (defined as molar concentration of GTA versus gelatin weight) constant. The scaffolds were immersed in GTA solution at 4°C for 48 h, until the cross-linking reaction occurred. Therefore, the samples were immersed in 0.1 M glycine solution in deionized water for 2 h at room temperature in order to stop any further cross-linking reaction (and remove any excess of GTA). Phosphate buffered solution (PBS) was sequentially used to rinse samples. Samples were kept at −20°C overnight, and finally freeze-dried (−50°C, 150 mBar) until all water content was removed. Samples were then stored at room temperature and sterilized with gas plasma before using.

Water absorption capability, porosity and stiffness of the porous gelatin scaffolds were evaluated using the procedure described elsewhere (Spinelli et al., 2012). Swelling ratio (Q) was calculated from the ratio of the weight of a dry (W₀) and a completely swollen (W_eq) sample (Brannon-Peppas and Peppas, 1990) returning the amount of adsorbed water. For the measurements, cryogel were swollen in deionized water at room temperature and weighted (W_i) at different time points until a swelling equilibrium was reached. A precision microbalance (AE240, Mettler, Italy) was used: in case of wet samples, blotting paper was used to remove the water in excess. Porosity was indirectly evaluated by the imbibition method (Mwangi and Ofner, 2004; Martucci et al., 2006), while pore size was directly measured by processing both Scanning Electron Microscopy (SEM) and optical microscopy acquisition of sample sections with ImageJ (Abramoff et al., 2004). Sample stiffness was measured by compressive mechanical tests. Prior to the tests, samples were completely swollen in deionized water, compressive tests were then performed using a Zwick-Roell Z005 Instron twin column-testing machine (Zwick Testing Machines Ltd., UK). Samples were compressed up to 5% of their initial length using a 0.01 mm· s⁻¹ strain rate; tests were performed with the samples partially immersed in water to preserve their hydration. Data were then post-processed and stress-strain curves were obtained. Samples stiffness was evaluated within 1% strain (first linear zone) of the stress-strain curve.

A computational fluid-dynamic (CFD) analysis of the modular chamber bioreactor with a porous scaffold was performed, assessing the perfusion and oxygenation of the millimeter sized scaffold. The analysis was performed using Brinkman and Incompressible Navier-Stokes equations were combined with reaction and diffusion equations using Comsol Multiphysics (COMSOL AB, Stockholm, Sweden). The system was then modeled with a porous domain representing the scaffold (porous section of 2 × 12 mm in size, with permeability of 1.68 × 10⁻¹⁰ m² and 90% porosity) placed on the bottom of the bioreactor perfusion chamber, which is represented by a fluid domain. A preliminary analysis was performed in order to verify scaffold oxygenation as function of bioreactor flow rate (Supplementary Figure 1A). Chosen an inner flow rate to a value of 200 μ L/min, fluid flow inside the perfusion chamber and scaffold's perfusion was analyzed (Supplementary Figure 1B).

hMSCs were purchased from Lonza and cultured in MSCGM™ BulletKit™ (hMSC basal medium, Lonza Japan Ltd, Tokyo, Japan). The cells were expanded to the desired number and used between passages 3 and 7. For the endothelial differentiation in 2D conditions, hMSCs were seeded onto tissue culture polystyrene plates (TCPS) and cultured for 7 days in endothelial differentiation medium [(EDM: hMSC basal medium supplemented with 50 ng/ml VEGF165 (R&D systems)]. hCMPCs were isolated from healthy donor atrial biopsies, cultured in basal medium (M199/EGM-2 (Lonza) (3:1) 10% FBS (Equitech-Bio Inc., Kerrville, Texas, USA), 1% MEM nonessential amino acids (from 50× stock; Life Technologies Gaithersburg, Maryland, USA) and 2% penicillin/streptomycin (from 50× stock; Invitrogen, Carlsbad, California, USA) and differentiated in cardiac differentiation medium (CDM: IMDM/Hams F12 (1:1, Life Technologies) with L-glutamine (Life Technologies), 2% horse serum (Life Tchnologies), nonessential amino acids, insulin–transferrin–selenium (Invitrogen), penicillin/streptomycin) as previously described (Smits et al., 2009a,b). Briefly, minced auricle were digested in collagenase A (1 mg/ml) for 2 h at 37°C while stirring. Afterwards, the solution was filtered through 40-μm cell strainer and centrifuged at 300 g for 5 min at RT. Then, the pellet was resuspended in cold buffer containing EDTA (2 mM) and 2% FBS and filtered through 40-μm cell strainer. The filtered cell population was subjected to magnetic cell sorting (Miltenyi Biotec, Sunnyvale, CA, USA) using Sca-1-coupled beads, following the manufacturer's protocol, in order to isolate Sca-1-like + progenitor cells. The cells were cultured on 0.1% gelatin- (Sigma-Aldrich, St Louis, Missouri, USA) coated plates in basal medium. For cardiac differentiation, cells were grown in differentiation medium supplemented with 5 μM 5-azacytidine (Wako Pure Chemical Industries, Ltd, Osaka, Japan) for 72 h. After induction, the medium was replaced with differentiation medium supplemented with 10⁻⁴ M ascorbic acid (Wako) and TGF-β 1 (1 ng/ml, PeproTech, Rocky Hill, NJ, USA) and changed every 3 days for up to 3 weeks. Human umbilical vein endothelial cells (HUVEC) were purchased from Lonza, grown in EGM-2 bullet kit media and used between passage 2 and 4. The media were replenished every other day.

The pGreenZeo lentiviral expression vector carrying the full-length copGFP gene under the control of cardiac Troponin T type 2 (TNNT2) promoter was purchased from SBI (System Biosciences Inc. Mountain View, CA, USA) and delivered into mammalian cells according to the manufacturer's instructions. Briefly, hCMPCs were transduced at an approximate MOI of 20 and incubated at 37°C overnight. The following day, half of the culture medium was replaced with fresh medium and 48 h after the transduction, the medium was discarded and fresh medium added. The cells (hereafter referred as cTNT-GFP hCMPCs) were cultured on TCPS or gelatin scaffolds in the presence of basal medium or CDM.

Prior to cell seeding, gelatin scaffolds were swollen in cell culture medium in Ultra-low Attachment 24-well plates (Corning® Incorporated) for 24 h. The cells (2.0 × 10⁵) were directly seeded on top of each scaffold in a small volume (100 μl) of appropriate medium and allowed to adhere for 60 min before adding the rest of medium. Then, the scaffolds were placed in the incubator for 12 h and transferred in another low-adhesion well. Cardiac TNT-GFP hCMPCs were cultured into scaffolds for 7 days in complete hCMPCs basal medium or CDM, while hMSCs-seeded scaffolds were cultured for 4 days in a mixture of Matrigel™/hMSCs basal medium (1:10) with or without VEGF (50 ng/mL) before being processed or co-cultured with pre-conditioned cardiac progenitors. The formation of vascular-like structures on gelatin scaffolds was followed by labeling hMSCs with the membrane-permeable live-cell labeling dye Calcein AM (Invitrogen) after 4 days of culture in EDM. Briefly, Calcein AM was added to hMSC-seeded scaffolds at a final concentration of 2 μ M for 30 min at 37°C and 5% CO₂ in the dark. Five scaffolds have been used for any single experiment. The experiments have been performed independently three times.

A commercial modular chamber bioreactor (Quasi-Vivo®, QV500, Kirkstall, United Kingdom) was used to perfuse the scaffolds. The main feature of this system is the ability to apply high flow rates and provide high nutrient turnover to cells without imposing high shear stress or turbulent flow (Mazzei et al., 2010). Pre-seeded scaffolds were gently transferred into the bioreactor chamber, and then the chamber was filled with cell culture media prior to perfusion. Two Quasi-Vivo® chambers and one mixing chamber (growth medium reservoir) were then connected in series to a peristaltic pump (High Performance Perista BIO-MINI PUMP, ATTO) giving a final volume of cell culture media of about 21 mL. The use of the mixing chamber guarantees medium oxygenation (Figure 1). To establish the optimal flow rate, the gelatin porous scaffold was included in the fluid-dynamic and oxygen transport model previously described (Mazzei et al., 2010), giving consistent results (Supplementary Figure 1). Cardiac TNT-GFP hCMPCs grown for 7 days onto gelatin scaffolds in the bioreactor were used to study the effect of dynamic culture conditions on cell growth, differentiation potential and gene expression profile. The medium was replenished every third day. For the generation of vascularized cardiac grafts, scaffolds were soaked for 12 h in diluted Matrigel™ (1:10 in hMSC basal medium) and switched to a new low-adhesion plate in EDM. The pre-conditioned scaffold was loaded with 2.0 × 10⁵ hMSCs and kept for 4 days in the incubator at 37°C, 5% CO₂. After a 2 week pre-conditioning passage in CDM on TCPS, 2.0 × 10⁵ cTNT-GFP hCMPCs were seeded onto the vascularized scaffold and allowed to attach for 24 h. Thence, the cell-loaded scaffolds were carefully moved to Quasi-vivo bioreactor chambers and perfused with CDM at 200 μl/min (the rate suggested by the computational model) for 7 days. Cells grown on TCPS and 3D scaffolds under static conditions were used as controls. Five scaffolds have been used for any single experiment. The experiments have been performed independently three times.

Total RNA was extracted by TRIZOL® Reagent according to the manufacturer's instructions (Invitrogen). One microgram of RNA for each sample, measured by NanoDrop 2000 (Thermo Scientific) and assessed on ReadyAgarose Precast Gel (Bio-Rad), was retro-transcribed using RT² First Strand Kit including DNAse treatment to remove genomic DNA (SA Biosciences Corp., USA). The resulting cDNA was diluted 1:10 in DNase/RNase-free water and the expression profile of genes involved in different pathways analyzed by the following RT² Profiler™ PCR Arrays (SA Biosciences Corp.): (i) RT² Profiler™ PCR Array Human Angiogenesis (PAHS-024Z); (ii) RT² Profiler™ PCR Array Human Extracellular Matrix & Adhesion Molecules (PAHS-013); (iii) qBiomarker iPSC PCR Array Cardiomyocytes differentiation (iPHS102); (for a comprehensive list of genes included in these arrays please refer to Supplementary Table 1). Real-time PCR was performed on ABI 7500 Real-Time PCR System (AB Applied Biosystems, Foster City, CA) using RT² SYBR Green/ROX qPCR Master Mix (SA Biosciences Corp.) and the following cycling parameters: 1 cycle at 95°C for 10 min; 40 cycles at 95°C for 15 s, and 60°C for 1 min. Data was analyzed by the ΔΔCt method with the PCR Array Data Analysis Web Portal (http://www.SABiosciences.com/pcrarraydataanalysis.php). The graphs show the mean values of obtained fold changes by analyzing independently two samples per each experimental condition.

Data are represented as mean ± SD. Student's T-test was used to analyze the data. Comparisons between different conditions were considered statistically significant for P < 0.05.

For a more extensive description of the Materials and Methods used, please refer to Supplementary Information.

Three dimensional gelatin scaffolds were characterized for their structure, water affinity, and stiffness. The analysis of the sections showed a distinctive bimodal pore size distribution. Although the average pore diameter ranged between 50–200 μm (with a few having a maximum size of 500 μm), the smaller pores were approximately double the size of the bigger pores (diameters of 61.62 ± 24.01 and 140.38 ± 19.51 μm, respectively; Figure 2A). Swelling tests were performed to assess the scaffold capacity to retain water. The scaffolds were monitored over 24 h of immersion in cell culture media, showing a complete rehydration after 8 h with a swelling ratio of about 10.92 ± 0.32 (Figure 2B). These data were also used to measure scaffold porosity, which was found to be approximately 90% (also confirmed in Spinelli et al., 2012). Finally, compressive mechanical tests were also performed on swollen samples (after 24 h in cell culture media), and on samples kept immersed in cell culture media and in the incubator for 1 week. Scaffold stiffness (evaluated within the first linear region of stress-strain plot, Figure 2C) was 10.5 ± 1.4 kPa, which was stable in time. After scaffold characterization and assessment of mechano-architectural properties, attention was dedicated to define the most suitable working conditions of the bioreactors to guarantee a homogeneous scaffold oxygenation in the bioreactor. Specifically, literature values of hypoxic limits (1%) (Cipolleschi et al., 1993) and oxygen consumption of stem cells (K_m = 0.001 mM, V_max = 1 pm/min/10⁴ cells) (Varum et al., 2011) were used to estimate the minimum oxygen concentration at the base of the scaffold as a function of flow rate and cell density, while the Wang and Tarbell approximation was used to calculate the shear stress at the pore walls. Results obtained from the computational models (reported in Supplementary Information) evidenced that—by using a flow rate of 200 μL/min—an oxygen concentration of 0.15 mM at the base of the scaffold (14%) and a shear stress of 1.10⁻⁶ Pa in the pores are obtained (Wang and Tarbell, 1995; Boschetti et al., 2006). The use of lower flow rates in the bioreactor may compromise oxygen supply rates to cells, especially as they proliferate, while higher flow rates may cause shear-related damage (Mazzei et al., 2010). Bioreactor parameters were set according to these results for all the following experiments.

Establishing a physiologically relevant system requires the development of 3D culture conditions resembling the in vivo microenvironment (Pampaloni et al., 2007). As reported above, the scaffolds here used were designed to have mechanical properties similar to those of cardiac tissue (Engler et al., 2007). Preliminary experiments demonstrated that culturing human hCMPCs on 3D gelatin scaffolds for 7 days induced a clear upregulation in the expression of genes (fold change ≥2, as compared to cells grown on TCPS) coding for proteins involved in cell-matrix remodeling (collagens (COL), laminins (LAMA), metalloproteinases (MMP) and their inhibitors (TIMP) and cell-cell adhesion (integrins) (Figure 3, for a complete list of the genes analyzed see Supplementary Table 1) and testified the actual activity of the cells in remodeling the scaffold.

The hCMPC reporter cell line (cTNT-GFP hCMPCs) was driven toward a cardiac phenotype in both 2D (Supplementary Figure 2) and 3D conditions, when stimulated with CDM (Figure 4A). Cell commitment was confirmed by the up-regulation of a number of cardiac-specific genes like actinin alpha 2 (ACTN2), adrenoceptor beta 1 (ADRB1), desmin (DES), cardiac troponin T type 2 (TNNT2), cardiac troponin I type 1 (TNNI1), myosin light chain (MYL) 2, 3, 7, myosin heavy chain b (MYH7) in cardiac progenitors stimulated on the scaffold as compared to their respective control (3D culture in basal medium). Moreover, when compared to cells cultured in CDM on TCPS, hCMPCs grown onto gelatin scaffolds encountered the up-regulation of few cardiac genes, confirming a superior performance of 3D vs. 2D culture conditions (Figure 4B, for a complete list of the genes analyzed see Supplementary Table 1). On the other hand, the ability of hMSCs to acquire an endothelial phenotype (Supplementary Figure 3) was exploited to obtain endothelial cells from hMSCs under 3D culture conditions. The scaffolds were treated with a mixture of EDM and Matrigel™ for 24 h. SEM analysis showed that the interconnected porosity of the scaffold was not modified by Matrigel™ treatment (highlighted by yellow dotted line in Figure 4C) and that the successful adhesion of human cells was achieved (Figure 4C, right). Consequently, hMSCs grown on 3D gelatin scaffolds in EDM displayed the up-regulation of a wide range of angiogenic markers as compared to cells cultured on scaffolds in basal medium (Figure 4D, for a complete list of the genes analyzed see Supplementary Table 1). Confocal analysis of 3D scaffolds after 4 day hMSC culture demonstrated the establishment that cells loaded with Calcein AM or expressing endothelial marker VCAM-1 colonized the whole scaffold, while aligning around the pores (Figure 4E).

The use of dynamic culture conditions using a bioreactor is thought to increase cell survival within a 3D structure by favoring inner core perfusion and catabolite removal (Cimetta et al., 2007; Burdick and Vunjak-Novakovic, 2009; Vozzi et al., 2011). Indeed, a slight although not significant increase in DNA content, indicative of cell proliferation, could be noticed 7 days after 3D cell culture in the bioreactor, as compared to static conditions (Figure 5A). Given the ability of human cardiac progenitor cells to proceed to cardiac maturation on 3D gelatin scaffolds, their capacity to activate the cardiac program when exposed to dynamic conditions in a modular bioreactor was assessed. For this experiment, cTNT-GFP hCMPCs were pre-committed in CDM for 2 weeks on TCPS. Afterwards, an aliquot of this cell population was seeded onto scaffolds and further stimulated for 1 week under static or dynamic conditions. The data confirmed that the commitment of cTNT-GFP hCMPCs could be achieved by CDM in dynamic conditions (Figure 5B) and that a minor although significant spontaneous commitment of hCMPCs was induced by 3D culture conditions per se, as also described in static culture (white columns). Although the dynamic culture conditions did not significantly modify the percentage of GFP-positive cells with respect to the static protocol, real time PCR array analysis demonstrated that dynamic stimulation—when combined with differentiation medium—could lead to the up-regulation of few cardiac-specific genes (Figure 5C, for a complete list of the genes analyzed see Supplementary Table 1).

Given the positive effect of dynamic culture conditions on cardiac gene activation, we sought to obtain vascularized cardiac 3D tissues by co-culturing hMSCs-derived vascular cells and pre-differentiated hCMPCs in the bioreactor system for 7 days, as described in Figure 6A. Gelatin sponges were first seeded with hMSCs under endothelial differentiation conditions. In the meantime, cTNT-GFP hCMPCs seeded on TCPS were challenged with CDM for 2 weeks, seeded onto hMSCs-loaded scaffolds and allowed to adhere for 24 h before being exposed to continuous medium flow. The analysis of the co-cultured constructs after 1 week testified massive scaffold colonization by the cells (Figure 6B). The scaffold thickness was not significantly modified by medium perfusion (data not shown) that affected, instead, cell migration inside the scaffold. In fact, the side of the colonized scaffolds directly exposed to flow showed a more uniform and extensive distribution of vascular-like cells (in red) and hCMPC-derived cardiomyocytes (green) inside the scaffold, while the cells were mostly distributed on the scaffold surface in the static control (Figure 6C). The expression of VCAM-1 and CD144 endothelial markers confirmed the presence of endothelial-like cells, while GFP expression testified the persistence of a number of cardiac-like cells within the bio-construct (Figure 6D). Sections from the perfusion group showed VCAM-1-positive cells aligned forming tube-like structures around the pores and contacting GFP-positive cells (Figure 6E). The extensive cell distribution within the scaffold accounted for the formation of a densely packed multicellular tissue derived from the two different stem cell types used (Figure 6F).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.