In vitro elastogenesis: instructing human vascular smooth muscle cells to generate an elastic fiber-containing extracellular matrix scaffold

Elastic fibers are essential for the proper function of organs including cardiovascular tissues such as heart valves and blood vessels. Although (tropo)elastin production in a tissue-engineered construct has previously been described, the assembly to functional elastic fibers in vitro using human cells has been highly challenging. In the present study, we seeded primary isolated human vascular smooth muscle cells (VSMCs) onto 3D electrospun scaffolds and exposed them to defined laminar shear stress using a customized bioreactor system. Increased elastin expression followed by elastin deposition onto the electrospun scaffolds, as well as on newly formed fibers, was observed after six days. Most interestingly, we identified the successful deposition of elastogenesis-associated proteins, including fibrillin-1 and -2, fibulin-4 and -5, fibronectin, elastin microfibril interface located protein 1 (EMILIN-1) and lysyl oxidase (LOX) within our engineered constructs. Ultrastructural analyses revealed a developing extracellular matrix (ECM) similar to native human fetal tissue, which is composed of collagens, microfibrils and elastin. To conclude, the combination of a novel dynamic flow bioreactor and an electrospun hybrid polymer scaffold allowed the production and assembly of an elastic fiber-containing ECM.


Introduction
The extracellular matrix (ECM) is a complex assembly of structural and functional proteins that are maintained by the resident cells. The cells not only secrete and assemble the ECM, but also respond to cues from the matrix that can alter cell behavior and maintain homeostasis. Elastic fibers within the ECM are crucial for tissue resilience and elasticity [1][2][3]. They play a major role during normal organ development [4] and are important structural components ensuring the proper function of adult tissues such as heart valves and blood vessels [5,6]. During development, the soluble precur-sor named tropoelastin is aggregated on the cell surface and finally forms the mature elastic fiber in the extracellular space, assisted by proteins such as fribrillins, fibulins, fibronectin, lysyl oxidase and many more (figure 1) [3,7]; however, the exact process of elastic fiber assembly remains poorly understood.
Inducing elastogenesis in a tissue-engineered construct would not only enable the production of functional implants but also the generation of advanced human-based in vitro test systems. These elastic fibercontaining systems could either be used to further understand the process of elastic fiber assembly or to study processes that are attributed to elastic fiber In vitro elastogenesis: instructing human vascular smooth muscle cells to generate an elastic fiber-containing extracellular matrix scaffold degeneration, including ageing. However, the in vitro generation of elastic fibers represents a major bottleneck in tissue engineering. Neonatal or adult rat vascular smooth muscle cells (VSMCs) and neonatal human dermal fibroblasts are routinely used to synthesize tropoelastin in vitro, and although elastin deposition can be shown in two-dimensional (2D) cell cultures [1,8,9], there have been no reports of the generation of functional human-based elastic fibers in three-dimensional (3D) tissue-engineered constructs [10][11][12]. Although it has been previously described that biochemical and biophysical factors impact elastin gene and protein expression in 3D tissue-engineered constructs, elastic fiber assembly in the extracellular space has not been shown [13].
Bioreactor systems have been designed to mimic the physiological and tissue-specific in vivo environment [14][15][16][17]. Such systems can potentially assist in the induction of elastogenesis in controlled in vitro experimental settings. To mimic the cellular environment and thus guide cell proliferation and migration, 3D substrates have also been used [5,18,19]. Electrospinning is one method used to generate 3D substrates and has been studied intensively for tissue engineering applications. With electrospinning, it is possible to generate scaffolds that provide not only three-dimensionality, but also mimic the natural ECM architecture [20][21][22][23].
We hypothesize that an optimal engineered microenvironment providing three-dimensionality and defined biochemical features, combined with biophysical signals can instruct cells to produce their own functional elastic fiber-containing ECM. Therefore, we designed a bioreactor and 3D electrospun hybrid scaffolds to expose primary-isolated human VSMCs to defined shear stress and topographies. Comprehensive ultrastructural and immunohistological studies were performed to verify the presence of an elastic fibercontaining network, including elastogenesis-associated proteins.
2D experiments were performed using six-well culture inserts for dynamic culture and twelve-well culture inserts for the static control (Greiner Bio-One, Frickenhausen, Germany). Electrospun PEGdma-PLA scaffolds were used for 3D experiments. 2D membranes and 3D scaffolds were used either uncoated, or they were coated with a 1:5 mixture of 1% hyaluronic acid (HA; Advanced BioMatrix Inc., San Diego, USA) in PBS prior cell seeding, since HA had been reported to assists in tropoelastin crosslinking. For six-and twelve-well membranes, as well as electrospun scaffolds, we used 2 × 10 4 , 9 × 10 4 or 3 × 10 5 cells per substrate. Twenty four hours after cell seeding, the substrates were transferred to the bioreactor. Depending on the experiment, 3.2 ng ml −1 transforming growth factor beta 1 (TGFβ1; Sigma-Aldrich) was added to the VSMC medium (C-22062, Promocell, Heidelberg, Germany).

Bioreactor system
The bioreactor was designed using SolidWorks for computer-aided design (Solidworks2010, Dassault Systèmes SolidWorks Corporation, Ludwigsburg, Germany). Furthermore, we developed a computational model to characterize the hydrodynamics of the designed bioreactor. Both, shear stress and laminar flow physics (Reynolds number) were calculated by a numerical approximated solution of the Navier-Stokes-Equation. Implementation, meshing and calculation were performed in Comsol 4.0 (COMSOL Multiphysics GmbH, Berlin, Germany). The simulations were carried out using bioreactor geometry and dimension. Moreover, a generalized minimal residual method (GMRES) was applied as the solver. The initial value for temperature was set to 37 °C. Based on the simulations, pump settings were defined in order to generate specific shear stress conditions.
For the in vitro experiments, the one-chamber perfusion bioreactor was connected to a closed tubing system (SC0746; Ismatec, Wertheim-Mondfeld, Germany) containing a 30 ml medium reservoir. The medium was pumped from the reservoir through the system using a peristaltic pump with a CA8 pump head and cassette (Ismatec, Glattbrugg, Switzerland). Briefly, the medium entered the bioreactor system through the inlet port, transited the flow chamber (1.5 × 1 × 0.7 cm) where the cell-seeded substrate was clamped in, and left the reactor through the outlet port. Furthermore, a filter (6901-2502; Whatman GmbH, Dassel, Germany) allowed permanent gas exchange, while the medium was pumped through the system. In order to carefully adapt the cells to the flow without shearing them off, the flow rate was initially set at 0.74 ml min −1 and then adjusted to 1.48 ml min −1 on the second day. After applying constant shear stress, the substrates were harvested and analyzed.

Specific elastin autofluorescence detection
3D scaffolds and human elastin (SH476, EPC, Owensville, USA) were rinsed with DPBS and transferred to glass-bottom dishes. The Mai Tai laser of a LSM710 (Carl Zeiss GmbH) was used to excite the samples with a wavelength of 760 nm [28,29]. Emissions between 440 and 500 nm were detected and analyzed utilizing Zeiss software (Carl Zeiss GmbH).

Scanning and transmission electron microscopy
A scanning electron microscope (1530 VP Zeiss, Carl Zeiss GmbH) was used to visualize the fibers of the electrospun scaffolds, which were mounted onto stubs and sputtered with platinum. Cell-seeded scaffolds were fixed with 2% glutaraldehyde, dehydrated with ethanol and dried at room temperature before mounting and sputtering. Transmission electron microscopy (TEM) was performed as described before [30] on electrospun scaffolds, 6 d dynamically cultured electrospun scaffolds and 18 week human valves (UCLA IRB #05-10-093). Briefly, all samples were fixed with 3% glutaraldehyde in 0.1 M sodium cacodylate buffer (pH 7.4). The samples were treated sequentially with 1% osmium tetroxide and 2% tannic acid, both in buffer and 2% uranyl acetate in water. After dehydration with a graded series of methanol to propylene oxide, the scaffolds and valves were infiltrated and embedded in Epon and polymerized for 3 d at 60 °C. The blocks were then sectioned (60 nm) and counterstained with methanolic uranyl acetate and lead citrate. For imaging, a FEI Tecnai 12 transmission electron microscope was used at 120 kV.

Live-dead staining
Primary human dermal fibroblasts (hFs) from adult skin were used in order to determine a potential cytotoxic effect of the electrospun scaffolds. This system is certified and accredited according to DIN EN ISO 10993-5. hFs were isolated using collagen digestion and seeded into chamber slides (Sigma-Aldrich). In parallel, extracts of soluble polymer compounds were prepared by incubating the electrospun scaffolds in Dulbecco's Modified Eagle Medium (DMEM; 11965-092, Life Technologies GmbH). After 48 h, the scaffolds were removed and the seeded hFs were incubated for another 48 h with the generated extracts. Fluorescein diacetate (FDA; F7378, Sigma Aldrich) and propidium iodide (PI; P4170, Sigma Aldrich) enabled the discrimination between living (green) and dead (red) cells in order to determine cytotoxic material effects. Stock solutions of 5 µg ml −1 FDA in acetone and 0.5 µg ml −1 PI in DPBS were prepared and stored at -20 °C. For live-dead staining, extracts were removed and a solution of 105 µL FDA, 95 µL PI and 3150 µL DMEM (serum free) was added. After 15 min incubation at 37 °C, the staining solution was replaced by DPBS for washing. Subsequently, images were taken with an Axio Observer fluorescence microscope (Carl Zeiss GmbH).

Contact angle measurement
Wettability was determined by deposing a 2 µl drop of dH 2 O onto the scaffolds as previously described [22]. Measurements were recorded using a DSA25S (Krüss, Hamburg, Germany).

Statistical analyses
One-way analysis of variance (ANOVA) was carried out to compare data groups. A probability value of 95% (p < 0.05) was used to determine significance. All data are presented as mean ± standard deviation.

Scaffold and bioreactor design and evaluation
To create the 3D fibrous hybrid scaffolds, we electrospun PEGdma and PLA (figure 2(a) [22],). After UV-crosslinking of the dimethacrylate groups, a hydrophilic and non-cytotoxic scaffold with an average fiber diameter of 0.37 ± 0.08 µm was obtained (figures 2(a)-(d), [22]). A fluid-flow bioreactor was designed and fabricated to allow cultures of cell-seeded 2D membranes and 3D scaffolds to be exposed to defined fluid shear stress. A computer-aided design of the flow bioreactor showing the flow direction and the scaffold position is displayed in figures 3(a) and (b). To determine optimal fluid mechanics, simulations were performed (figures 3(c) and (d)) and a Reynolds number of approximately 1 was identified. The red stream lines in figure 3(c) indicate that no turbulences occurred, and a Reynolds number <2300 confirmed laminar flow conditions in the bioreactor. With a flow rate of 1.48 ml min −1 , a constant shear stress between 6 × 10 −4 and 11 × 10 −4 dyne cm −2 (equal to 6-11 × 10 −5 Pa) was applied over the entire scaffold.

Influence of shear stress, three-dimensionality and biochemical aspects
For cell culture experiments, 2D and 3D, uncoated and HA-coated substrates were seeded with primary isolated, CNN1-, αSMAand SM-myosin-expressing VSMCs and cultured for 3 or 6 d ( figure 4). We observed cell alignment on the 2D membranes, perpendicular to the direction of flow, starting at 6 d of dynamic culture (figures S1(a)-(h) (stacks.iop.org/ BMM/10/034102)). In general, under all culture conditions, the 6 d cultures revealed significantly higher ELN expression and protein production when compared to the 3 d cultures. Moreover, after 6 d, a significant shear stress-induced up-regulation of ELN and elastin protein was observed when comparing the dynamic bioreactor cultures to the static controls (elastin protein: 3.11 ± 0.34 (dynamic) versus 2.4 ± 0.21 (static), p = 0.026; ELN: 8.49 ± 2.68 (dynamic) versus 4.43 ± 0.11 (static), p = 0.0065; figures S1(i) and (j) (stacks.iop.org/BMM/10/034102)). No significant differences were detected between uncoated controls and HA-coated samples in either static or dynamic cultures (figure S1). Since TGFβ1 participates in matrix regulation [8], we added TGFβ1 [31] to the dynamic 2D and 3D cultures. After 6 d, the substrates were stained for elastin and quantified (figure 5 and figure S2 (stacks.iop.org/BMM/10/034102)). Significantly higher elastin production was detected in TGFβ1-treated 2D cultures compared to cultures without TGFβ1 ( figure S2(a)); however, no significant difference was observed for ELN expression in these samples ( figure S2(b)). Semi-quantification of the immunofluorescence images revealed a significant increase in elastin production on the 3D hybrid scaffolds compared to the 2D membranes (figure S2(c)); 7.33 ± 0.58 3D versus 4.15 ± 0.67 2D, p = 0.003). This increase was not seen at the gene level ( figure S2(d)).

Induction of elastogenesis and ECM development on 3D scaffolds
Interestingly, we observed deposition of elastin on the hydrophilic electrospun fibers, and elastin-positive fibers were detected in between cells (figures 5(a) and (b)). TEM further revealed the presence of long bundles of 'coated' microfibrils in between the cells (figures 5(c) and (d)). Indeed, electron-dense material could often be identified within these bundles ( figure 5(d)). Moreover, analyzing the cell-seeded fibrous electrospun matrix using multiphoton laserinduced autofluorescence imaging as previously described [29], emission patterns comparable to those of pure human elastin were detected (figures 5(e)-(j)). In contrast, this specific elastin autofluorescence and the corresponding emission patterns did not occur on the scaffold cultured under the same conditions without cells (figures 5(g) and (j)).
Elastic fiber formation depends not only on sufficient tropoelastin synthesis, but also on proper matrix assembly and crosslinking. We therefore performed immunofluorescence staining to determine the presence of important elastogenesis-associated proteins in our cultures and identified that after a 6 d dynamic culture of VSMCs on the 3D HA-coated scaffold, fibrillin-1 and -2, fibronectin, fibulin-4 and -5 and EMILIN-1 (figures 6(a)-(f)), as well as the cross-linking enzyme LOX (figure S3 (stacks.iop. org/BMM/10/034102)) were all produced. Although TGFβ1 and three-dimensionality had no significant impact on ELN expression, a significant increase of FBLN5 and FBN1 expression was observed by adding TGFβ1 (figures S2(e) and (f)). FBN1 and FN1

Discussion
It was previously demonstrated that mimicking physiological conditions such as shear stress [14] or three-dimensionality [32] supports tissue maturation. Therefore, we designed a fluid flow bioreactor system, which is suitable for the culture of 3D cellseeded tissue-engineered scaffolds. The defined and applied shear stress in this study differs from physiological blood luminal flow, but instead corresponds to the in vivo situation, where VSMCs are exposed to a very low transmural interstitial flow rather than direct shear stress [33]. VSMC alignment was observed after 6 d of culture, a phenomenon that has been previously described when VSMCs are exposed to laminar shear stress [33]. Furthermore, a shear stress-induced up-regulation of tropoelastin/elastin synthesis was observed comparing the dynamic cultures to the static controls. There was no significant impact on elastin expression due to HA coating of the substrates. The HA used in this study has a molecular weight of ~750 000 Da. Although high molecular weight HA (>10 000 Da) does not induce elastin precursor synthesis, this type of HA supports elastic matrix deposition and more efficient crosslinking [34]. Smaller HA molecules do enhance elastin precursor synthesis [35]; however, due to its proposed capacity to support tropoelastin crosslinking, the larger HA was used in our experiments. The addition of TGFβ1 increased elastin protein synthesis as well as FBLN5 and FBN1 expression, but did not enhance ELN expression. The obtained results indicate that TGFβ1 has no direct impact on ELN expression, but may be crucial for protein binding and elastic fiber assembly since TGFβ1 stabilizes elastin mRNA and facilitates tropoelastin recruitment for elastic fiber assembly [8,36,37].
In addition, we observed a positive effect on elastin protein expression as well as elastin deposition and new fiber formation on the 3D scaffolds. Lin et al described in their 4 d-study a direct impact of threedimensionality on ELN gene expression [31], which is in contrast to our experimental data based on which we did not identify such an effect of three-dimensionality on the ELN gene expression. We assume that these discrepancies are either due to the different culture periods or caused by the application of another type of material [31]. An explanation for the increased elastin protein level on 3D scaffolds is that the fibrous electrospun matrix has a higher surface area and therefore more proteins, growth factors and other signaling molecules can attach [5,20]. Based on these results we conclude that the HA-coated electrospun PEGdma-PLA scaffold allows bio-functionalization with elastin in vitro. More interestingly, immunohistological analyses revealed newly formed elastin-containing fibers in between the layers of VSMCs. Ultrastructural assessment of the newly assembled ECM confirmed the presence of long bundles of coated microfibrils. It has previously been described that both smooth and 'naked' microfibrils as well as microfibrils coated with accessory proteins such as tropoelastin and fibronectin exist in the ECM [38,39]. In our study, the microfibrils coated with protein could represent developing/ maturing elastic fibers.
Tropoelastin is the soluble ~70 kDa elastin precursor. Tropoelastin monomers coacervate under physio-logical conditions and are cross-linked in the presence of fibrillin-containing microfibrils to form insoluble elastic fibers [3,4,7,40]. The initial self-assembly of elastin and the significance of fibrillins and fibulins has been extensively studied [9,41,42], and the complex interactions between elastogenesis-associated proteins have been comprehensively reviewed [7,43]. Accordingly, the successful assembly of elastic fibers is highly dependent on the elastogenesis-associated proteins, which were all present in our study after 6 d of in vitro culture. In addition, transmission electron micrographs revealed a newly formed ECM comparable to the one seen in 18 week fetal heart tissues. Although the elastin detected in the tissue-engineered constructs was not as dense as seen in the native 18 week matrix, it must be considered that the observation was made after only 6 d in vitro. Our results display therefore a developing but not yet completely matured elastic fiber-containing ECM.

Conclusion
We conclude that the customized bioreactor described in this study, combined with a hybrid fibrous 3D polymer scaffold, is a promising tool to induce ECM formation, including the production of elastin and elastic fiber-related proteins, which are necessary for the assembly of mature elastic fibers. To our knowledge, this is the first report of the 3D in vitro generation of elastic fibers, which addresses a major bottleneck in the fields of tissue engineering and regenerative medicine. This technology can be utilized not only for in vitro elastogenesis studies, but moreover, our system can potentially serve as a platform technology to engineer elastic fiber-rich tissues and organs such as heart valves and blood vessels. In our study, we employed a biocompatible poly (ethylene glycol)-based scaffold that can be potentially bio-functionalized with human elastin and instruct cells to develop their own ECM. This is highly advantageous when aiming for applications in regenerative medicine and medical device development.

Limitations and future work
In this study, the described bioreactor system enabled the induction of elastogenesis on an electrospun PEGdma-PLA scaffold. The size of the flow chamber is sufficient when aiming for an introduction of elastic fibers within a 2D or 3D cell culture test system; however, the systematic up-scaling of this system will be required in order to generate clinically relevant tissues. Furthermore, in this study we used primary isolated human VSMCs. A limiting factor with these cells is that they are isolated from veins and arteries, which is unfavorable when producing patient-tailored implants due to the limited number and variation in quality. Therefore, the identification of alternative cell sources, for example by utilizing autologous adult stem cell-derived or induced-pluripotent stem cell-derived cardiovascular progenitor cells or VSMC, will be necessary.