Development of a bioreactor system for pre‐endothelialized cardiac patch generation with enhanced viscoelastic properties by combined collagen I compression and stromal cell culture

Treatment of terminal heart failure still poses a significant clinical problem. Cardiac tissue engineering could offer autologous solutions for the replacement of nonfunctional myocardial tissue. So far, soft matrix construction and missing large‐scale prevascularization prevented the application of sizeable cardiac repair patches. We developed a novel bioreactor system for semi‐automatic compression of a collagen I hydrogel applying 16 times higher pressure than in previous studies. Resistance towards compression stress was investigated for multiple cardiac‐related cell types. For scaffold prevascuarization, a tubular cavity was imprinted during the compaction process. Primary cardiac‐derived endothelial cells (ECs) were isolated from human left atrial appendages (HLAAs) and characterized by fluorescence‐activated cell sorting (FACS) and immunocytology. EC were then seeded into the preformed channel with dermal fibroblasts as interstitial cell component of the fully cellularized patch. After 8 days of constant perfusion culture within the same bioreactor, scaffold dynamic modulus and cell viability were analyzed. Endothelial proliferation and vessel maturation were examined by immunohistochemistry and transmission electron microscopy. Our design allowed for scaffold production and dynamic culture in a one‐stop‐shop model. Enhanced compression and cell‐mediated matrix remodeling induced a significant increase in scaffold stiffness while ensuring excellent cell survival. For the first time, we could isolate HLAA‐derived EC with proliferative potential. ECs within the central channel proliferated during flow culture, continuously expressing endothelial markers (CD31) and displaying basal membrane synthesis (collagen IV, ultrastructural analysis). After 7 days of culture, a complete endothelial monolayer could be observed. Covering cells aligned themselves in flow direction and developed mature cell–cell contacts.


Abstract
Treatment of terminal heart failure still poses a significant clinical problem. Cardiac tissue engineering could offer autologous solutions for the replacement of nonfunctional myocardial tissue. So far, soft matrix construction and missing large-scale prevascularization prevented the application of sizeable cardiac repair patches. We developed a novel bioreactor system for semi-automatic compression of a collagen I hydrogel applying 16 times higher pressure than in previous studies. Resistance towards compression stress was investigated for multiple cardiac-related cell types.
For scaffold prevascuarization, a tubular cavity was imprinted during the compaction process. Primary cardiac-derived endothelial cells (ECs) were isolated from human left atrial appendages (HLAAs) and characterized by fluorescence-activated cell sorting (FACS) and immunocytology. EC were then seeded into the preformed channel with dermal fibroblasts as interstitial cell component of the fully cellularized patch. After 8 days of constant perfusion culture within the same bioreactor, scaffold dynamic modulus and cell viability were analyzed. Endothelial proliferation and vessel maturation were examined by immunohistochemistry and transmission electron microscopy.
Our design allowed for scaffold production and dynamic culture in a one-stop-shop model. Enhanced compression and cell-mediated matrix remodeling induced a significant increase in scaffold stiffness while ensuring excellent cell survival. For the first time, we could isolate HLAA-derived EC with proliferative potential. ECs within the central channel proliferated during flow culture, continuously expressing endothelial markers (CD31) and displaying basal membrane synthesis (collagen IV, ultrastructural ABBREVIATIONS: 3D, three dimensional; AA, atrial appendage; B(P)EL, bovine serum albumin (BSA) polyvinyl alchohol essential lipids; CD, cluster of differentiation; CM, cardiomyocyte; CPC, cardiomyocyte progenitor cell; CSC, cardiac stem cells; DAB, 3,3 0 -diaminobenzidine; DAPI, 4 0 ,6-diamidino-2-phenylindole; DMEM, Dulbecco's modified eagle medium; EC, endothelial cell; ECM, extracellular matrix; EGM-2, endothelial growth medium 2; FACS, fluorescence-activated cell sorting; FCS, fetal calf serum; G 0 , storage modulus; G 00 , loss modulus; G*, complex shear modulus; Heart transplantation is still considered the gold standard treatment for patients suffering from terminal heart failure (Carrier & Perrault, 2014). For years, Eurotransplant has reported increasing numbers of patients on the waiting list facing a nearly constant pool of available donor hearts (Eurotransplant International Foundation: Annual Report, 2014). Cardiac tissue engineering could provide a solution to this problem of organ shortage. Different strategies have been pursued to create functional heart muscle tissue primarily implementing either cell-or scaffold-based approaches (Haraguchi, Shimizu, Yamato, & Okano, 2012).
When considering scaffold-based strategies, collagen type I has proven an ideal scaffold for cardiac tissue engineering on numerous accounts (Perea-Gil, Prat-Vidal, & Bayes-Genis, 2015). Moreover, mechanical dehydration of collagen hydrogel could create a stiffened network to match the alternating strains imposed on the intercellular matrix by cardiomyocyte (CM) contractility. Following a distinct physicochemical model described by Hadjipanayi et al. (2011), plastic compression (PC) of collagen gels not only greatly enhances mechanical properties like stiffness and tensile strength but also has been shown to positively influence cell migration (Serpooshan et al., 2013) and proliferation of epidermal, corneal, and urothelial cells in complex 3D models (Ajalloueian, Zeiai, Fossum, & Hilborn, 2014;Hu et al., 2010;Mi, Chen, Wright, & Connon, 2010). Most groups so far used fixed weight plates with a maximum force generation of approximately 1.4 kN/m 2 (Serpooshan et al., 2013) for compression. There is a proportional interrelation of collagen pore size and fibroblast-induced gel contraction (Serpooshan, Muja, Marelli, & Nazhat, 2011) favoring higher compression rates for stable shape retention. Furthermore, according to Ghezzi, Muja, Marelli, and Nazhat (2011) increasing collagen fiber density exerts a positive effect on fibroblast growth induction. Hence, an elaborate PC protocol with high compression rates is justified for ideal scaffold construction.
Another major issue to date in cardiac tissue replacement is the immediate and long-term sustenance of engineered tissue upon implantation. For example, Kawamura et al. (2012) reported limited cell survival in human-induced pluripotent stem cell (hiPSC)-derived CM sheets after transplantation in a pig model of chronic myocardial infarction. Therefore, instant nutrient and oxygen supply through an intrinsically preformed vascular network is mandatory for complex tissues exceeding 200 μm in thickness (Benavides et al., 2015). Two domains must be considered when dealing with in vitro prevascularization: (1) tube/network formation and (2) endothelial lining. Regarding (1), capillarization can be accomplished by selforganizing effects of admixed endothelial cells in coculture (Sekine et al., 2008;Shimizu, 2014) or micropatterning within suitable scaffolds (Choi et al., 2007;Zheng et al., 2012). However, direct surgical anastomosis with a human host vasculature requires the provision of a larger connecting vessel inside the transplant. Vollert et al. (2014) successfully generated endothelialized, perfusable channels of up to 500 μm in diameter within their engineered heart tissue using alginate spacer tubes. Other groups followed a top-down approach by reendothelializing decellularized porcine small bowel segments yielding full-size vascular networks which could be attached to host vessels via arterial and venous stubs (Andree et al., 2014;Mertsching et al., 2009;Schanz, Pusch, Hansmann, & Walles, 2010). In terms of (2), endothelial cell (EC) choice of origin is fundamental for graft-host and cell-cell communications. ECs are known to play a decisive role in allograft rejection, for example, via monocyte recruitment and T-cell stimulation (Al-Lamki, Bradley, & Pober, 2008), while at the same time exhibiting a certain tissue-specific heterogeneity concerning organ maintenance and cytokine responsiveness (Molema, 2010;Nolan et al., 2013). These data suggest an advantage using autologous, cardiac-derived ECs to optimize the cellular interplay for tissue engineered heart patch transplants.
We assumed that human left atrial appendages (HLAAs) could be harnessed for sufficient EC supply. These remnants of embryonic development can be accessed and excised easily during open or endoscopic heart surgery granting an additional benefit in reducing the incidence of postoperative cerebrovascular events, at least in patients with a low CHA2DS2-VASc score (Kato et al., 2015). In the past, mostly cardiomyocyte progenitor cells (CPCs) have been harvested from atrial appendages (AAs) fueling multiple animal and in-man studies. CPCs were tested for their potential of local cardiac repair after myocardial infarction in rats (Sakai et al., 1999), minipigs (Fanton et al., 2015), and humans (Chugh et al., 2012) yielding some promising results over the past two decades. In contrast, no efforts have been made so far to isolate stromal and EC populations of AAs, let alone introduce them to tissue engineering. Because coculturing of different cell types is known to exert positive effects on their self-organization potential (Czajka & Drake, 2015), it is conceivable that a proper mixture of endothelial and non-endothelial "support cells" could promote the growth and maturation of engineered cardiac tissues and a developing vascular network within.
In the present study, we aimed at building a scalable "cardiac- 2.2 | Cell isolation and differentiation 2.2.1 | Human dermal fibroblasts HDF were isolated as previously described by Moll et al. (2013). In brief, skin biopsies were washed with phosphate-buffered saline (PBS) (Sigma-Aldrich, St. Louis, MO, USA), connective tissue and fat were removed, and tissue was cut in strips of even size. Tissue was digested in dispase (2 U/ml; Thermo Fisher Scientific, Langenselbold, Germany) for 16 to 18 h at 4 C. Epidermal and dermal layers were separated with tweezers; dermal layers were cut into smaller pieces and digested for 45 min at 37 C in collagenase (500 U/ml; Serva, Heidelberg, Germany). Pieces were plated in Dulbecco's modified eagle medium (DMEM; high glucose; Thermo Fisher Scientific) with 10% FCS (Bio&SELL GmbH, Feucht, Germany). Outgrowing cells were cultured and split with Trypsin (Thermo Fisher Scientific) at some 80% confluency.

| Primary cardiac cells
HLAAs were excised during routine cardiac surgery with patients' consent and processed within 8 h. Cells were isolated as previously described (Messina et al., 2004;Smith et al., 2007)

| hiPSC-derived CMs
hiPSC-derived CMs were kindly provided by the Edenhofer group (Kadari et al., 2015). Cells had been differentiated from AR1034ZIMA hiPSC clone 1 and cultured in RPMI1640 (Thermo Fisher Scientific) with 2% B-27 supplement (Thermo Fisher Scientific), β-mercapoethanol F I G U R E 1 Fabrication of collagenous scaffold. (a) Compression of collagen I by a linear motor within a custom-made bioreactor. (b) Schematic drawing of fabrication process. Dissolved collagen I with interstitial cells was cast into the bioreactor core module with topped cylinder and a central needle placeholder. After solidification, plastic compression was performed. The patch was then transferred to dynamic culture. A peristaltic pump was used to induce a constant medium flow of 1 ml/min. After 24 h of primary culture, the needle was removed, and endothelial cells were seeded. (f.c. 100 μM; Thermo Fisher Scientific), L-ascorbic acid (f.c. 0.5 mg/ml; Sigma-Aldrich), and 1% P/S.

| Fabrication of cell-seeded PC collagen scaffolds
Collagen I was isolated following a protocol from Dieterich et al. (2002)  After 24 h, the medium reservoir was detached and stored at 37 C.
The lower compartment of the reactor was closed, the central placeholder removed, and noncytotoxic silicone tubing was attached. HAAEC suspension was injected in a "pump and suck" manner retaining some 400 μl of a 1 * 10 6 cells per milliliter suspension inside the channel. The tubing was closed and cells were allowed to settle for 1 h. The bioreactor was turned upside down for another hour to ensure ubiquitous EC attachment. The whole seeding procedure was performed twice to ensure adequate EC seeding density. After reconnection to the roller pump system patches were incubated under permanent perfusion conditions for a maximum of 7 days. Half of the reservoir medium was changed twice a week. At the end of the culture period, patches were gently removed from the bioreactor and further processed.

| Examination of PC effects on mechanical properties and cell survival
To comprehensively cover viscoelastic properties of our PC collagen scaffold, we chose to determine storage and loss moduli (G 0 and G 00 ) over Young's modulus. Rheological characterization was performed on freshly compressed patches (0D) and on patches cultured for 8 days (1 day of initial scaffold remodeling + 7 days of alleged channel endothelialization). Patches for both time points were manufactured with and without interstitial cells (HDF and control, respectively). For each setup (0D, 0DF, 8D, and 8DF), three patches were examined.
G* thereby closely resembles a Maxwell model of viscoelasticity with a serial setup of a pure viscous dampener and a pure elastic spring.
To determine the impact of compression on embedded cells, we digested freshly compressed patches (10 ml collagen, 5 ml GNL, and

| Qualitative MTT test of whole scaffolds
Viability testing of scaffold-incorporated cells was done via a qualitative MTT test. The cultured scaffold was taken out of the reactor and transferred into a 1 mg/ml MTT in VascuLife ® VEGF-Mv solution.
After 90 min of incubation at 37 C, the scaffold was washed thoroughly with PBS+ until no MTT residues were visible.

| Transmission electron microscopy
For transmission electron microscopy (TEM) samples were fixed as previously published by Helmprobst, Frank, and Stigloher (2015) for 1 h in 2.5% glutaraldehyde. Preparation of fixed patches followed the same protocol. Briefly, samples were washed, incubated with 2% OsO4, and contrasted with 0.5% uranyl acetate as described by Reynolds (1963).

| Human AAs are a reliable EC source for later autologous applications
To establish our prevascularized scaffold model, ECs were isolated from HLAAs. HLAA were excised during surgery, processed, and cells taken into culture (Figure 3a). Processed biopsies gave rise to WCM, a mixture of different cell types. We performed magnetic-activated cell sorting (MACS) for CD31 with WCM harvested from each specimen individually to obtain pure ECs. The resulting cell populations, CD31-negative cells and HAAEC as well as the unsorted WCM, were further investigated by FACS and immunofluorescence staining. Ki67 staining was applied to analyze the proliferative potential of all three cell populations from different donors (Table 2)

| DISCUSSION
In this study, we were able to devise a bioreactor system for combined PC of collagen I and subsequent 3D dynamic cell culture in a single integrated setup.
F I G U R E 2 Rheological examination of patches: (a) Vibratory rheological strain curves before and after 8-day incubation period, with and without fibroblasts. Two time points were examined: directly after compression (0D) and after 8 days of culture (8D). Shear modulus (G*; III) was calculated from elastic storage (G 0 ; I) and loss (G 00 ; II). (b) Bar chart of G 0 , G 00 , and G* at 0.2 % strain. 0D, 0DF, and 8D patches showed no significant difference regarding G 0 or G 00 . 8DF patches showed a significant increase in overall stiffness (*p %3C 0.05). (c) Collagen I has been broadly used for several tissue engineering applications. It is one of the most prominent proteins in various connective tissues throughout the human body and thus highly biocompatible with low immunogenicity. However, collagen I gels tend to be quite susceptible towards mechanical stress due to a high water content. Brown, Wiseman, Chuo, Cheema, and Nazhat (2005) (Brown et al., 2005;Cheema & Brown, 2013) or with a weighted plate (Drechsler et al., 2017;Hu et al., 2010;Witt et al., 2019). Our standardized, semi-automatic compression strategy yielded a considerably higher (about 16 times) compressive force than in previous reports by using a motor-driven approach instead of fixed weight plates (Serpooshan et al., 2013). While our method was able to limit further collagen shrink-  (Ghezzi, Marelli, Muja, & Nazhat, 2012). Consequently, we also tested for surgical handling in a hands-on test of "cut and sew" by an experienced cardiac surgeon ( Figure S3).
As stated earlier, the complex interplay of all cardiac cell types is required to create a unique environment of closely related mechanic and paracrine networks. Thus, balanced coculture models are the premise for developing a fully grown myocardial patch. We postulated a cell-specific resistance towards mechanical stress during our PC cycle.

Results of cell survival analyses indicated two distinguishable groups:
While interstitial cells such as HDF, CD31-negative cells, and CM displayed high compression resistance, vascular-associated cells, that is, HAAEC and pericytes, were more sensitive. This matches general cell culture findings: EC often exhibit a more sensitive behavior regarding the processes of medium exchange, cell passaging, and freezing/thawing than more resilient cell types. High survival numbers of CM could be explained by their unique physiological properties.
Due to the demands of permanent contraction and relaxation CM exhibit a suitable cytoskeleton and membrane configuration to withstand substantial mechanical strains (Sequeira, Nijenkamp, Regan, & van der Velden, 2014). However, they are sensitive towards collagenase A treatment, becoming obvious after correction of survival rates for digestion effects.

Isolation of cardiac-derived cells, namely, cardiac stem cells (CSC),
has gathered more and more attention as first clinical trials of stem cell therapy for myocardial infarction have been registered (Gyongyosi, Haller, Blake, & Martin Rendon, 2018). In our work, we focused on HLAA-derived cells. HLAAs are a remnant of embryonic development attached to the left atrium (Al-Saady, Obel, & Camm, 1999;Regazzoli et al., 2015). So far, murine LAA were used for isolation of cardiac progenitor cells (CPC) (Leinonen et al., 2013). Approaches in human, for example, isolation of a new subtype of CSC, utilized right AAs (Koninckx et al., 2013). Preparation of HLAA gave rise to a cellular mix termed WCM which was divided in endothelial CD31-positive  (Hahn & Schwartz, 2009). These findings suggest that proper medium flow and the resulting shear stress should positively influence growth and integrity of the endothelial layer. We chose a low flow velocity of 1 ml/min to avoid initial flush detachment of the ECs. This flow leads to a shear stress of some 0.77 dyne/cm 2 . Most methods published so far using microvascular networks with gravity-driven medium flow reported a shear stress range from 0.1 (Zheng et al., 2012) to 5 dyne/cm 2 (Chan et al., 2014;Chrobak et al., 2006). HAAEC in our model showed continuous proliferation over time and were metabolically active, suggesting appropriate dynamic culture conditions. Likewise, immunofluorescent 3D reconstruction of the channel highlighted adaptation of HAAEC to the applied perfusion conditions. Cells  (ECM) as indicated by basal lamina residues. It is known that EC can express most components of the vascular basement membrane themselves, but the final assembly requires heterotypic cell-cell contacts (Davis & Senger, 2005).

| CONCLUSIONS AND OUTLOOK
We present a new method of manufacturing a large-scale collagenbased scaffold. Motor-driven PC of collagen I hydrogel yields a viscoelastic scaffold, more likely to resemble in vivo tissue stiffness. A traversing endothelialized channel and successful seeding of various cell types under dynamic culture conditions make this patch a versatile tool for complex tissue engineering efforts. Even though our choice of collagen work-up and cells aimed at constructing a cardiac repair patch, our principal strategy is not limited to this specific tissue type.
The method could be adapted for a different organoid framework design. Possible modifications include but are not limited to changes in compression cycle length and force development, cell composition, and cytokine application due to the modular architecture of our bioreactor system.

ACKNOWLEDGMENTS
This work has been funded by the Federal Ministry of Education and Research (funding numbers 01EO1004 and 01EO1504). We acknowledge support by the K-Centre VascAge (COMET program-Competence Centers for Excellent Technologies) funded by the Austrian Ministry for Transport, Innovation and Technology (FE).
Open access funding enabled and organized by Projekt DEAL.

CONFLICT OF INTEREST
None declared.

ETHICS STATEMENT
Direct humanization for autologous transplantation without necessity of animal-based transplants. All tissues explanted under ethical vote of the university clinic Wuerzburg number 182/10.

SUPPORTING INFORMATION
Additional supporting information may be found online in the Supporting Information section at the end of this article.