Head-to-head comparison of two engineered cardiac grafts for myocardial repair: From scaffold characterization to pre-clinical testing

Cardiac tissue engineering, which combines cells and supportive scaffolds, is an emerging treatment for restoring cardiac function after myocardial infarction (MI), although, the optimal construct remains a challenge. We developed two engineered cardiac grafts, based on decellularized scaffolds from myocardial and pericardial tissues and repopulated them with adipose tissue mesenchymal stem cells (ATMSCs). The structure, macromechanical and micromechanical scaffold properties were preserved upon the decellularization and recellularization processes, except for recellularized myocardium micromechanics that was ∼2-fold stiffer than native tissue and decellularized scaffolds. Proteome characterization of the two acellular matrices showed enrichment of matrisome proteins and major cardiac extracellular matrix components, considerably higher for the recellularized pericardium. Moreover, the pericardial scaffold demonstrated better cell penetrance and retention, as well as a bigger pore size. Both engineered cardiac grafts were further evaluated in pre-clinical MI swine models. Forty days after graft implantation, swine treated with the engineered cardiac grafts showed significant ventricular function recovery. Irrespective of the scaffold origin or cell recolonization, all scaffolds integrated with the underlying myocardium and showed signs of neovascularization and nerve sprouting. Collectively, engineered cardiac grafts -with pericardial or myocardial scaffolds- were effective in restoring cardiac function post-MI, and pericardial scaffolds showed better structural integrity and recolonization capability.

should closely resemble the physiological myocardial extracellular matrix (ECM) properties as internal scaffold conformation, mechanics, and composition will modulate cellular differentiation, migration, and adhesion 4-8 . In this context, decellularized cardiac tissues provide a close match to the native, physiological microenvironment, as they preserve the inherent stiffness, composition, vasculature network, and three-dimensional framework 9 , and enable electromechanical coupling with the host myocardium upon implantation 10,11 . Herein, we tested two decellularized scaffolds generated from the myocardium and pericardium, as previously described 12,13 . The scaffolds were either repopulated with porcine adipose tissue mesenchymal stem cells (pATMSCs) or tested as cell-free scaffolds. Macro and micromechanics, structural characteristics and protein content of the decellularized scaffolds were evaluated prior to recellularization. The functional benefits associated with both the acellular and recellularized scaffolds were assessed in a preclinical MI swine model. Cardiac function was analyzed with magnetic resonance imaging (MRI), and accurate infarct size was obtained with late gadolinium enhancement.

Methods
Generation of decellularized myocardial and pericardial scaffolds. Myocardial tissue samples were isolated from porcine hearts (n = 22), and surgical samples of pericardial tissue were acquired from 90 patients (63 males, 27 females; mean age 69 ± 10 years; range 42 to 85 years) undergoing cardiac interventions at our institution, with apparently healthy pericardia and after signed consent. This study was revised and approved by the Germans Trias i Pujol University Hospital ethics committee, and all the protocols conformed to the principles outlined in the Declaration of Helsinki. Both cardiac tissues were decellularized, lyophilized, and sterilized as described elsewhere 12,13 . Recellularization of sterilized cardiac scaffolds. The sterilized acellular cardiac scaffolds were repopulated with pATMSCs obtained from porcine pericardial adipose tissue, as previously reported 13 , to generate both the engineered myocardial and pericardial grafts. Initially, the decellularized scaffolds were rehydrated with a mixture of 175 μL of peptide hydrogel RAD16-I (Corning, Corning, NY) and 175 μL of GFP + -pATMSCs (1.75 × 10 6 ) in 10% sucrose (Sigma-Aldrich, St Louis, MO), added dropwise on top of the scaffold. To promote hydrogel jellification, α-minimum essential medium eagle (α-MEM) (Sigma-Aldrich) was added over the scaffold. The produced cell-enriched grafts were maintained over one week under standard culture conditions (37 °C, 95% air, 5% CO 2 ), changing medium every 2 days.

Flow cytometry. Changes in cell viability were assessed by flow cytometry using the Annexin V Apoptosis
Detection Kit and the 7-amino-actinomycin D (7-AAD) viability staining solution (eBiosciences). Cells (2-5 × 10 5 ) were labelled according to the supplier's instructions. The percentage of viable cells was measured by using a Coulter EPICS XL flow cytometer with the Expo32 software (Beckman Coulter).

Measurement of micromechanical properties with atomic force microscopy (AFM). Micromechanical
properties of native myocardial tissue and decellularized and recellularized scaffolds were studied in 5 porcine hearts. 50 µm thick slices of fresh myocardium were obtained using a vibratome (0.01 mm/s blade velocity; 2 mm amplitude; VT 1200 S, LEICA, Germany) and attached to glutaraldehyde pre-treated glass slides and directly measured with atomic force microscopy (AFM). The decellularized and recellularized myocardial scaffolds (15 × 5 × 5 mm) were embedded in 3:1 (v:v) solution of optimal cutting temperature medium (OCT, TissueTek, USA) in phosphate buffered saline (PBS) 1x and frozen at −80 °C. For AFM measurements, 50 µm thick scaffold slices were cut using a cryostat (HM 560 CryoStar, Thermo Fisher, Madrid, Spain), placed on top of positively charged glass slides and stored at −20 °C until AFM measurements. To assess the effect of the different steps of the recellularization process in tissue stiffness, AFM measurements were performed in myocardial samples after lyophilization, sterilization and peptide hydrogel embedding (5 additional porcine hearts). A similar AFM measurement protocol was performed to measure the stiffness of human pericardial samples but in this case native pericardial stiffness was measured in 5 × 5 × 2 mm samples adhered with cyanoacrylate to a petri dish.
Before AFM measurements, the slices with OCT were rinsed several times with PBS until OCT was completely removed. Then, the slides were placed on the sample holder of a custom-built AFM system 12 attached to an inverted optical microscope (TE 2000, Nikon, Japan), and force-indentation (F − δ) curves were measured at RT using cantilevers with a polystyrene microsphere 4.5 μm in diameter attached at its end (nominal spring constant of 0.03 N/m, Novascan, Ames, IA). The actual spring constant of the cantilevers was calibrated with the thermal noise method. The micromechanical Young's modulus (E m ) was computed by fitting F − δ data with the Hertz contact model 12 where R is the radius of the microsphere and ν the Poisson's ratio (assumed to be 0.5).
For each slice, measurements were performed in two locations separated by ~500 µm. At each location, five measurements were taken separated by ~10 µm following a linear pattern. At each measurement point, five F − δ force curves with a maximum indentation of 1 µm were recorded. Myocardium and pericardium micromechanical stiffness was characterized as the average E m computed from the different force curves recorded in each sample.
Measurement of macromechanical properties with tensile testing. For macromechanical measurements, strips (~8 × 1 × 1 mm) of native, decellularized, recellularized, lyophilized, sterilized and hydrogel embedded myocardium and pericardium cut along the long-axis with a scalpel were used (n = 5 each). The strips were maintained at RT in Krebs-Henseleit solution (Sigma-Aldrich). For tensile testing, each strip was gently dried with tissue paper and its mass (M) measured. The unstretched length (L 0 ) of the strip was measured and its cross-sectional area (A) computed as SCIEnTIFIC REPORtS | (2018) 8:6708 | DOI:10.1038/s41598-018-25115-2 where ρ is the tissue density (assumed to be 1 g/cm 3 ). One end of the strip was glued with cyanoacrylate to a hook attached to the lever of a servocontrolled displacement actuator, which simultaneously stretched the strip and measured the stretched length (L) and the applied force (F) (300C-LR, Aurora Scientific, Ontario, Canada). The other end of the strip was glued to a fixed hook. Measurements were performed inside a bath with Krebs-Henseleit solution at 37 °C. The stress (σ) applied to the strip was defined as: Tissue stretch (λ) was defined as: Strips were initially preconditioned by applying 10 cycles of cyclic stretching at a frequency of 0.2 Hz and maximum stretch of ∼50%. L 0 was measured again and F − L data from 10 more cycles were recorded. The stiffness of the tissue was characterized by the macromechanical Young's modulus (E M ) defined as Stress-stretch (σ − λ) curves were analyzed with Fung's model, which assumes that E M increases linearly with stress as: where α·β is the value of E M at L 0 and α characterizes the strain-hardening behavior of the tissue. Eq. 6 leads to a stress-stretch exponential dependence where σ r and λ r define an arbitrary point of the σ − λ curve. The parameters of Eq. 7 were computed for each curve by non-linear least-squares fitting using custom built code (MATLAB, The MathWorks, Natick, MA). Using these parameters, E M was computed at unstretched length (λ = 1) and at 20% stretch (λ = 1.2). Macromechanical stiffness of each heart strip was characterized as the average E M computed from the last σ − λ curves recorded.
Proteomic analysis. The proteins from decellularized tissues were initially extracted and digested as detailed in the Supplementary methods. Peptides were analyzed in an Impact II Q-TOF mass spectrometer (Bruker, Billerica, MA) using the nano-LC dedicated CaptiveSpray source, coupled to a nanoRSLC ultimate 3000 system (Thermo Fisher). The peptide separation details are also found in the Supplementary methods. The mass spectrometer was operated in the Instant Expertise mode, consisting of a cycle of: (i) One MS spectrum in 250 msec, mass range 150-2200 Da; (ii) as many as possible MS/MS spectra in 1.75 sec, at 32-250 msec, depending on precursor intensity, mass range 150-2200 Da; and (iii) precursor exclusion after one selection. Searches against the Swissprot (2015-08) library of mammal protein sequences were performed using the Mascot 2.5 search engine (MatrixScience, Boston, MA), with a 7 ppm tolerance at the MS level and 0.01 Da tolerance at MS/MS level, and a false discovery rate fixed to a maximum value of 1%. The variable modifications were carbamidomethyl, Gln->pyro-Glu, acetyl, oxidation, and deamidation.   conditions. Data are presented as the mean ± SEM. **P < 0.001 between native pericardium and native myocardium. # P = 0.044 between recellularized pericardium and recellularized myocardium, using Student's t test. ## P = 0.013 and $ P = 0.005 between recellularized myocardium vs. native and decellularized myocardium, respectively, with ANOVA test with Tukey's post-hoc correction for multiple comparisons analysis. Morphometric analysis. Sacrifices were performed at 39.3 ± 7.5 days after MI with an intravenous injection of potassium chloride solution. Following a lateral thoracotomy, porcine hearts were excised. The left ventricular (LV) infarct area was measured in photographed heart sections obtained 1.5 cm distal to the artery ligation, as previously described 14 . Double-blind quantification was performed by two independent investigators using Image J software (version 1.48, National Institutes of Health, Bethesda, MD).
Histopathological and immunohistological analysis. Paraffin slices (4-μm) were stained with hematoxylin/eosin (H/E), Movat's pentachrome, Movat's modified (for simultaneous collagen and mucopolysaccharide acid staining 15 , and Gallego's modified trichrome to analyze histological changes and scaffold recellularization using a computer-associated Olympus CKX41 microscope (Olympus, Tokyo, Japan) with a ProgRes ® CF Cool camera (Jenoptik, Jena, Germany).  Pore size measurements. The pore size of both decellularized myocardial and pericardial scaffolds (n = 3 of each) was blindly evaluated using 10 randomly selected SEM images with a magnification range from 500 to 1500, and at least 10 different pores from each image were randomly chosen and measured to generate an average value. For each pore, the long and short axis lengths were determined using ImageJ software, and pore roundness was calculated as the ratio between the short and long axis.
Statistical analysis. Data are represented as the mean ± SEM. Statistical analyses were performed through the independent Student's t test for cell number and density, pore analysis, and myocardium vs. pericardium mechanical analysis; paired sample t-test was used to analyze cell viability experiments and one-way analysis of variance (ANOVA) with the Tukey's post-hoc correction for multiple comparisons was conducted for mechanical, MRI and infarct size data. SPSS 21.0.0.0 (SPSS, Inc, Chicago, IL) and SigmaPlot 13 (Systat Software Inc, San Jose, CA) were used for data analysis. Values were considered significant when P < 0.05 and tendency when P < 0.10.
The datasets generated and/or analysed during the current study are available from the corresponding author on reasonable request.

Accession Protein
The mechanical properties of cardiac tissue were evaluated at two different scales: macromechanics by tensile testing and micromechanics using AFM. Macroscopic stiffness of the native myocardium was 2.5 ± 0.5 kPa at unstretched length and increased to 14.0 ± 2.3 kPa at 20% stretch (Fig. 3C,D). No significant differences were found after decellularization and recellularization. The non-linear parameter α (Equation 6) was 13.1 ± 0.7 in the native myocardial tissue with no significant changes in the decellularized and recellularized scaffolds. Tensile tests did not reveal significant differences between native, decellularized, and recellularized samples neither at unstretched length (Fig. 3C) nor at 20% stretch (Fig. 3D). Interestingly, macroscopic stiffness did not change significantly when the acellular myocardial scaffold was lyophilized, sterilized, or embedded in RAD16-I hydrogel (Supplementary Figure S1A,B). Microscopic stiffness measured in the native myocardium with AFM (13.4 ± 0.9 kPa) was ∼5-fold higher than macroscopic stiffness found in unstretched strips. Similar to the macroscopic mechanical behavior, no significant changes were observed between decellularized, lyophilized, sterilized, or rehydrated myocardial scaffolds (Supplementary Figure S1C). In contrast to macroscopic measurements, microscopic stiffness of recellularized scaffolds (22.0 ± 2.2 kPa) was ∼2-fold stiffer than native tissue (13.4 ± 0.9 kPa) and decellularized scaffolds (11.6 ± 2.4 kPa) (respectively-test P < 0.05) (Fig. 3E).
There were no significant changes in macroscopic and microscopic stiffness between native, decellularized and recellularized pericardium (Fig. 3C-E). No significant differences were found in the unstretched length between pericardium and myocardium. However, native pericardial tissue and its recellularized scaffolds were ∼3-fold stiffer than the myocardium (Student t test P < 0.001 and P < 0.05, respectively) (Fig. 3D). Consistently, the non-linear parameter α (Equation 6) of the native (36.9 ± 4.5), decellularized (31.2 ± 3.6) and recellularized (23 ± 1.8) pericardium was ∼2 fold higher (P < 0.05) than the myocardium, reflecting a stronger strain-hardening behavior. In agreement with myocardium mechanical behavior, no significant changes were found when comparing data along the different recellularization steps (Supplementary Figure S1A-C).
Proteomic characterization. Decellularization of both cardiac tissues was associated with enrichment of matrisome/ECM-related proteins, with 7.7% and 9.8% of proteins in native myocardium and pericardium, respectively, and 25.2% and 41.1% in decellularized myocardial and pericardial scaffolds, respectively (Fig. 4A-D). For decellularized myocardium matrisome proteins, 24 of the 40 proteins were also in the native tissue, and 16 proteins were detected exclusively within the decellularized myocardium. In the pericardium, matrisome enrichment was more pronounced: only three proteins were common in the decellularized and native pericardia; while forty-eight proteins were only detected in the decellularized pericardium and there was just one unique protein in the native pericardium (Fig. 4E,F).
Comparative analysis of the generated acellular myocardial and pericardial tissues revealed they shared 40% of the matrisome proteins. Among them, the most remarkable proteins were the different collagen subtypes, responsible for maintaining the ECM structure and modulating cell differentiation; the laminin family and heparan  sulfate, mainly involved in cell adhesion; and fibronectin, which is involved in different cellular processes such as adhesion, survival, and differentiation. In addition, the number of unique proteins identified was notably higher for the pericardium, with up to 25 distinctive matrisome proteins vs. just 14 in the myocardium (Fig. 5A and Tables 1-3). Finally, subdivision of the ECM-proteins showed more proteins within each ECM protein subtype for the decellularized pericardium, except for the collagen subclass (Fig. 5B).  Supplementary Table S1. The LVEF changes between sacrifice and baseline indicated improved function for Per-MI and Per-ATMSCs (P < 0.001). Changes in the LVESV showed a trend towards an improvement in the studied groups (Fig. 6A). Similarly, LVEF changes between sacrifice and post-MI were also significant: (P = 0.004). In addition, LVESV changes between sacrifice and post-MI were significant (P = 0.031) (Fig. 6B). Changes in the absolute values at 40 days for LVEF (P = 0.017) and LVESV (P = 0.020) were also significant.
Immunohistological analysis. Correct adhesion of the implanted graft with subjacent myocardium was observed in all animals after sacrifice. After histological analysis, Movat's modified staining confirmed the collagen composition of the scaffolds (red), the presence of sulfated mucopolysaccharides (violet), red blood cells (yellow) as well as lymphocytes and plasma cells (green). Gallego's modified staining also confirmed the presence of collagen in scaffolds (brilliant green), elastic fibers (fuchsine violet) and squamous epithelium (yellow) (Fig. 7A). Moreover, scaffolds of all experimental groups showed functional blood vessels connected to host vascularization, which was confirmed by the presence of erythrocytes in the lumen, some even with SMA in the medial layer (Fig. 7A,B,D), and newly formed nerves within the scaffolds (Fig. 7B).

Discussion
Cardiac tissue engineering is mainly based on the combination of a biomimetic scaffold with a cell lineage of interest. Here, we characterized the structure as well as the macro and micromechanical scaffold properties after decellularization and recellularization of two biological scaffolds from myocardial and pericardial tissue. Importantly, no adverse events were associated with pericardium isolation from patients, following a safe cardiac surgery procedure previously reported by our group 13,18,19 . We also evaluated their proteome composition and the ability of the two engineered grafts (once recellularized with ATMSCs) to restore cardiac function. When using an acellular framework for the scaffold design, the ECM inherent properties, such as structure, mechanics, and protein composition, must be ideally maintained from the beginning. Matrix disruption due to decellularization agents must be minimized to recreate the native physiological milieu, as structure preservation has been related to cell differentiation, migration, and alignment 7,8 . Mechanical stiffness modulates cell adhesion, differentiation, proliferation, and migration 4,6,20 , as well as contractility 21 . We have performed a multiscale study of myocardial and pericardial mechanics. Stiffness measured at the macroscale is important to assess the mechanical matching between the ventricular wall and the attached patch. The similar macroscopic stiffness of the native tissue and its recellularized scaffold indicates homogeneous deformation of the patch and healthy ventricular wall during heart beating. On the other hand, stiffness measured with AFM characterizes the mechanical properties at the microscale, which is the scale at which cells sense their mechanical microenvironment. AFM probes micromechanics of the sample under unstretched conditions. Therefore, the ~5-fold higher stiffness we found with AFM (Fig. 3E) as compared to that measured at 0% stretch with tensile assays (Fig. 3C) suggests that cells sense a local niche substantially stiffer than the stiffness of the whole 3D matrix scaffold. This differential stiffness indicates that scaffold macromechanics is determined by the intrinsic micromechanical properties of the ECM as well as by the 3D topology of the matrix. Although no significant differences were found in macroscale stiffness at 0% stretch between myocardial and pericardial scaffolds (Fig. 3C), the more marked strain-hardening behavior exhibited by the pericardium (~2-fold higher α) resulted in a ~2-fold stiffer pericardial scaffold at 20% stretch (Fig. 3D). It should be noted that ventricular tissues are subjected to ~20% stretch during heart beating indicating that the decellularized pericardial tissue provides a stiffer scaffold under physiological conditions. According to our results, macro and micromechanics were well-maintained in both cardiac scaffolds following decellularization. No significant mechanical changes were observed after subjecting the decellularized scaffolds to lyophilization, sterilization and hydrogel embedment. This allows the conservation and storage of the decellularized scaffolds until later use. Of note, the matrix pore size was significantly larger in pericardial scaffolds, which correlated with the findings from a cardiomyocyte transverse section reported in a previous study 22 . Despite the addition of ATMSCs significantly increased myocardial micromechanics, the resulting stiffness for both recellularized myocardial and pericardial scaffolds was within the optimal range (10-20 kPa) to drive physiological processes such as cardiomyocyte maturation, contraction, and cardiac lineage commitment, as reported in the literature 23,24 . Therefore, seeding ATMSCs on top of the decellularized myocardium did not have a major impact on the matrix micromechanics; rather, it resulted in an overall preservation of the scaffold structure and mechanics following the recellularization process.
Regarding protein content, the native matrisome proteins were preserved in both cardiac tissues upon decellularization, which is consistent with prior studies 16 , and some of the main cardiac ECM components were identified, including: collagens (type-I, -III, -IV, -V, -VI), laminin, emilin-1, fibronectin, lumican, heparan sulfate proteoglycan, nidogen, and periostin. These key proteins act as mediators or effectors of many cellular processes, resulting in favored ATMSC adhesion and survival. Indeed, laminins, fibronectin, and type-I and IV collagens enhance cellular survival 25,26 ; whereas, fibronectin 27 , lumican 28 , and type-IV and -VI collagens 29,30 are associated with cell attachment through specific interactions with integrin cell receptors. Galectin-1, vitronectin, decorin, and tenascin-X are involved in cell adhesion 27,31,32 and were also identified in our decellularized pericardial scaffolds. This finding, along with the optimal pore diameter suitable for cell infiltration 22,33 , may explain the increased cell retention and penetrance observed in these scaffolds.
According to the 3 R principle (Replacement, Reduction and Refinement), we included previous animal experimentation models 13,19,34 in this work, performed by the same surgical team under identical conditions for comparison. We demonstrated that the implantation of the engineered cardiac grafts led to improvements in LVEF and/or LVESV, limited infarct size expansion, and coupling with the underlying myocardium, as indicated by the vascularization and innervation of the grafts. The benefits exhibited here by both recellularized scaffolds are greater than those reported for direct intracoronary or intramyocardial injection of ATMSCs 35 , or administration of an acellular myocardial 36 or pericardial scaffold 37 . The combination of the cardiac scaffolds with ATMSCs showed more improvement than the non-recellularized scaffolds, indicating a synergistic effect, as reported previously 38 . Scaffolds may be beneficial for triggering MI salvage, by providing a favorable microenvironment for the seeded ATMSCs, and for recruiting endogenous stem cells towards the infarct bed 39 , boosting both vascularization and cardiomyogenesis. The contribution of ATMSCs to MI recovery has also been previously described, and the most recent evidence points towards paracrine signaling, rather than direct effects of ATMSCs 40 . Paracrine action would be consistent with a low number of retained or seeded cells, as in our study, being able to promote effects 41 , such as vessel formation, protecting resident cardiomyocytes from apoptosis, and mobilizing resident stem cells to potentiate vascularization and cardiomyogenesis 42 . Remarkably, in the present work, both cell-free and ATMSC-recellularized cardiac scaffolds, regardless of their origin, were successfully integrated in host cells, undergoing neovascularization and neoinnervation. However, further studies on ECM signaling and/or host cell recruitment pathways are required to discern which mechanisms are responsible for these processes.

Conclusions
Two engineered cardiac grafts were successfully generated by combining ATMSCs with either myocardial or pericardial acellular scaffolds, which preserved structural framework in both cases. Nevertheless, superior preservation of mechanical intrinsic properties and higher enrichment of major cardiac ECM proteins was observed for pericardial scaffolds. Furthermore, in vitro, the recellularized pericardial scaffold showed higher cell retention and penetrance, and a bigger pore size compared to recellularized myocardial scaffolds. In the context of a preclinical MI model, the engineered grafts integrated with the underlying myocardium, with signs of neoinnervation and neovascularization, and improved cardiac function post-MI.