Engineering Heart Valve Interfaces Using Melt Electrowriting: Biomimetic Design Strategies from Multi‐Modal Imaging

Abstract

Interfaces within biological tissues not only connect different regions but also contribute to the overall functionality of the tissue. This is especially true in the case of the aortic heart valve. Here, melt electrowriting (MEW) is used to engineer complex, user‐defined, interfaces for heart valve scaffolds. First, a multi‐modal imaging investigation into the interfacial regions of the valve reveals differences in collagen orientation, density, and recruitment in previously unexplored regions including the commissure and inter‐leaflet triangle. Overlapping, suturing, and continuous printing methods for interfacing MEW scaffolds are then investigated for their morphological, tensile, and flexural properties, demonstrating the superior performance of continuous interfaces. G‐codes for MEW scaffolds with complex interfaces are designed and generated using a novel software and graphical user interface. Finally, a singular MEW scaffold for the interfacial region of the aortic heart valve is presented incorporating continuous interfaces, gradient porosities, variable layer numbers across regions, and tailored fiber orientations inspired by the collagen distribution and orientation from the multi‐modal imaging study. The scaffold exhibits similar yield strain, hysteresis, and relaxation behavior to porcine heart valves. This work demonstrates the ability of a bioinspired approach for MEW scaffold design to address the functional complexity of biological tissues.

Multi‐modal imaging techniques unravel the orientation of collagen fibers connecting different regions of the aortic heart valve, which is used to inspire novel design methods for melt electrowritten scaffolds. Biomimetic scaffolds exhibit continuous interfaces, gradient porosities, region‐specific layer numbers, and tailored fiber orientations in a singular design. Mechanical testing shows native tissue‐like behavior, including yield strain, hysteresis, and stress relaxation.

1. Introduction

Biological tissues are complex, multi‐phasic, heterogeneous, and hierarchical structures, that present exquisite properties ideally suited for their functionality.^{[ 1} , ² , ^{3 ]} Moreover, their interfaces play the critical role of not only physically connecting heterogeneous regions but also exhibiting graded structural, cellular, and mechanical features, and thus critically contributing to the overall functionality of the tissue or organ.^{[ 4 ]} Accordingly, when attempting to biofabricate scaffolds for these tissues, the interfaces represent a pivotal, yet challenging piece of the puzzle.^{[ 5 ]} The majority of research surrounding engineering of tissue interfaces relates to soft–hard tissue interfaces in areas such as the ligaments and tendons,^{[ 6 ]} cartilages,^{[ 7} , ^{8 ]} as well as dental,^{[ 9 ]} and craniomaxillofacial implants.^{[ 10} , ^{11 ]} In these applications, heterogeneous scaffolds have been achieved through advanced manufacturing techniques, such as multi‐axial extrusion,^{[ 12} , ¹³ , ^{14 ]} varying degrees of cross‐linking,^{[ 15 ]} two step phase separation,^{[ 16 ]} multi‐material bioinks,^{[ 17} , ^{18 ]} and through controlled spatial deposition of biomaterials with advanced 3D printing technologies.^{[ 19} , ^{20 ]}

Fibrous scaffolds and in particular electrospun meshes have been extensively investigated in the field of soft tissue engineering for applications such as skin,^{[ 21 ]} neural,^{[ 22 ]} vascular,^{[ 23 ]} or cardiac tissue,^{[ 24 ]} however there is a lacking body of comparable research toward interfacial design. Interfacing between regions has been achieved mostly on a layer‐by‐layer basis, by either altering print parameters^{[ 25} , ^{26 ]} or solution concentration^{[ 27} , ^{28 ]} during the fabrication process, cross‐linking afterward,^{[ 29 ]} or individual layer‐by‐layer assembly.^{[ 30} , ³¹ , ³² , ^{33 ]} The absence of interface complexity for electrospun scaffolds could be due to the lack of precise control over fiber orientation when manufacturing with this technology. New approaches to manufacturing micro‐fibrous scaffolds, such as melt electrowriting (MEW), may enable more sophisticated interface designs.

MEW is a highly precise additive manufacturing technique capable of printing complex fibrous scaffold constructs at sub‐micron resolutions.^{[ 34} , ³⁵ , ^{36 ]} The complexity and resolution characteristics of MEW have been beneficial for producing functional, biomimetic, soft tissue scaffolds for skin,^{[ 37} , ³⁸ , ^{39 ]} nerves,^{[ 40 ]} myocardium,^{[ 41 ]} cartilage,^{[ 42} , ⁴³ , ⁴⁴ , ^{45 ]} and aortic heart valves.^{[ 20} , ⁴⁶ , ^{47 ]} Similar to electrospinning, interfacing on a layer‐by‐layer basis has been shown with MEW.^{[ 44} , ^{47 ]} More recently, the ability to fabricate heterogeneous fibrous structures within a single layer using MEW has been demonstrated, marking a notable step forward in the field of fibrous scaffold fabrication.^{[ 20} , ⁴⁸ , ⁴⁹ , ⁵⁰ , ^{51 ]} Despite this, the concept of heterogeneous MEW printing is still in its infancy, and as a result, the design of the interface between heterogeneous regions has not been a primary focus of previous work.^{[ 48} , ⁴⁹ , ⁵⁰ , ^{51 ]}

Here we explore the potential for MEW to engineer complex, user‐defined, interfaces between heterogeneous regions of heart valve scaffolds. The complexity and heterogeneity of the aortic valve make it a good showcase to examine the technical capabilities of MEW in complex scaffold fabrication for soft tissue interfaces. The average human aortic valve undergoes over 30 million cycles per year, amounting to over 2 billion cycles during a 70 year lifespan.^{[ 52 ]} Such remarkable haemodynamic properties are enabled by the combined biomechanical behavior resulting from each of the structurally distinguishable regions of the valve as well as the interfaces between them.^{[ 53} , ^{54 ]} We have previously demonstrated MEW scaffolds with mechanical properties similar to those of physiological heart valve leaflets using a bioinspired design approach.^{[ 46 ]} This was enabled by the data available surrounding the relationship between the mechanical properties of the leaflets and collagen recruitment mechanisms during loading.^{[ 55} , ⁵⁶ , ^{57 ]} We now consider regions of the heart valve beyond the leaflets, potentially unlocking further applications of MEW to enhance valve functionality. Regions such as the commissures, inter‐leaflet triangles, and the interface between them, all play crucial roles in the biomechanical behavior of heart valves.^{[ 58} , ⁵⁹ , ^{60 ]} While there is a substantial body of literature surrounding the morphology and mechanical properties of these regions of the aortic valve, the specific nature of how they interface with one another, with respect to fibrous extracellular matrix (ECM) orientation, remains unknown.

Here, we applied multi‐modal imaging to examine the morphology of the interfacial regions of the aortic heart valve as the basis to establish a functional bioinspired design for MEW scaffold interfaces. Subsequently, three methodologies for interfacing heterogeneous scaffolds were investigated to understand their effect on tensile and flexural properties. Combining the native tissue investigation and study on interfacing methods, an integrated biomimetic MEW scaffold of the heart valve interfacial regio was designed, fabricated, and tested.

2. Results and Discussion

2.1. Morphological Analysis of the Aortic Heart Valve Interfaces

The aortic valve leaflets are responsible for the majority of the valve's overall functionality and have been the focus of previous studies.^{[ 58} , ⁶¹ , ^{62 ]} Moreover, other regions of the aortic valve are also known to significantly contribute to the valve's functionality.^{[ 58} , ⁵⁹ , ^{60 ]} These regions have been classified in differing ways;^{[ 63 ]} here we divide them into the leaflets, sinuses of Valsalva, annulus, commissures, and inter‐leaflet triangles (Figure  1A). Each region has different concentrations and orientations of ECM constituents, including collagen and elastin fibers which are of particular interest when considering the mechanical functionality of the valve.^{[ 54 ]} Collagen fibers are the main load bearing component of the ECM and are responsible for the anisotropic mechanical characteristics and J‐shaped stress–strain response of the tissue.^{[ 55} , ⁵⁶ , ^{57 ]} Elastin fibers play a role in the low‐strain performance of the valve but more importantly regulate the collagen fiber orientation, ensuring they return to their pre‐loaded state between cycles.^{[ 64 ]} While all components of the ECM play an important role in the valve functionality, our investigation focuses on the orientation of collagen fibers to inspire MEW scaffold design. It is thus important to understand what is already known regarding collagen fiber orientation in the heterogeneous regions of the aortic valve, which is outlined briefly below.

Figure 1.

Microstructure of the commissure and inter‐leaflet triangle of the aortic valve. A) Overview of the regions of the aortic valve. Scale bar = 10 mm. B) Reconstructed micro‐computed tomography of the aortic valve in near‐physiological systolic conditions, with blue sections representing where slices were taken for further microscopy. Viewed from the i) coronal (external), ii) sagittal, iii) axial, and iv) coronal (internal) directions. C) Images of vibratomed sections prior to second harmonic generation (SHG) imaging, with slices going deeper into the tissue from top to bottom. Specific regions correlating to SHG images are shown with dotted boxes. Scale bars = 2 mm. SHG images of D) the commissure and interface with the beginning of adjacent leaflets, E) commissure at a deeper slice, and F) inter‐leaflet triangle. Scale bars (D–F) = 100 µm.

The aortic valve leaflets contain three distinct layers: the fibrosa, spongiosa, and ventricularis (from aorta to ventricle). The tensile load bearing layers are the fibrosa and ventricularis, which consist mostly of circumferentially and radially aligned collagen fibers, respectively. The sinuses of Valsalva are similar in composition to the aortic wall, consisting of three distinctive layers where the inner (intima) and outer (adventitia) comprise of primarily longitudinally arranged collagen fibers, whereas the middle (media) layer is predominantly circumferentially arranged fibers.^{[ 62 ]} The annulus is a fibrous ring structure connecting the leaflets and sinus wall into the left ventricle, and consists mainly of highly dense, circumferentially aligned collagen bundles, resulting in relatively rigid mechanical properties.^{[ 62} , ^{65 ]} The commissure exists at the point where the free edge of the two leaflets meet. By helping transmit forces between the leaflets and the surrounding aortic root, the commissure plays a crucial function in supporting both the flexural properties required to open the valve in diastole as well as withstanding high tensile loads as the valve closes in systole.^{[ 66 ]} With respect to the microstructure, collagen fibers primarily extend radially from the leaflets, intertwining through the commissural region and anchoring into the aortic wall, enabling transmission of forces between the leaflet and root.^{[ 62 ]} The inter‐leaflet triangles are governed by the region below the commissure, between the leaflets, and above the annulus.^{[ 67 ]} The orientation of the fibrous microstructure is largely unexplored in the inter‐leaflet triangle, however, it is thought to primarily follow that of the annulus, that is, circumferentially. Notably, while the collagenous microstructure of these regions is known, the specific nature of how the collagen fibers orientate in the interfaces between these regions remains to be investigated. This could provide valuable insight to inform the design of scaffolds striving to mimic native fibrous morphology. Accordingly, we began by asking whether high resolution multi‐modal imaging of porcine tissue could provide insight into some of these unknowns.

Experimentation and analysis using porcine tissue has been adopted widely in cardiovascular research owing to the greater availability of tissues with anatomical and haemodynamic similarities to humans.^{[ 68} , ⁶⁹ , ^{70 ]} While Martin et al. showed different mechanical properties in porcine and human aortic tissue, it is unclear if these were due to age‐related effects.^{[ 71} , ^{72 ]} Importantly, when analyzing the collagenous microstructure in the aforementioned study, the regional fibrous orientation was consistent between human and porcine species, as well as in a study by Filova et al.^{[ 73 ]} Accordingly, porcine tissue has often been used for biomechanical studies,^{[ 74} , ^{75 ]} and to elucidate fiber alignment using microscopy.^{[ 76} , ^{77 ]}

Here, porcine aortic valve tissue was fixed using a hybrid immersion fixation and hydrostatic pressure distension method to preserve near‐physiological diastolic conditions (Figure S1, Supporting Information). First, we used micro‐computed tomography (micro‐CT) for the 3D reconstruction of the valve architecture to facilitate macroscale identification of regions of interest (Figure 1B). Next, tissue was sectioned longitudinally through the wall of the aortic root (Figure 1B), parallel to the direction of blood flow, at a thickness of 250 µm (Figure 1C) for further analysis via second harmonic generation (SHG) imaging. This orientation allowed for imaging of the commissure and inter‐leaflet triangle, the interface between them, as well as the interfaces with the leaflets, sinuses, and annulus. Collection of three slices into the aortic wall enabled for confirmation of orientation information in the third dimension. SHG is an optical microscopy technique ideally suited for imaging collagen from the tissue scale to molecular scale due to the second order non‐linear characteristic of collagen fibers.^{[ 78} , ^{79 ]}

The microstructure of specific heart valve regions was identified in SHG images (Figure 1D), including the commissure, and the fibrosa and ventricularis layers of two adjacent leaflets. In the commissure, large amounts of collagen fibers were densely interlaced and intertwined between the vertical and horizontal directions. Three distinct groups of collagen fibers travelled down from the top of the commissure into the leaflets. The left and right of these groups merged into the fibrosa layers of the two adjacent leaflets, which can be identified by their distinctly rounded structures with collagen fibers travelling into and out of the page, showed by a diminishing SHG signal. Interestingly, we observed that the center group of fibers travelled down from the top of the commissure for approximately 500 µm before splitting into the ventricularis layer of each leaflet, identifiable by the vertically aligned collagen. To our knowledge, the orientation and extent to which the commissure interfaces with the fibrosa and ventricularis layers of two adjacent leaflets was unknown until now. In a deeper image of the commissure, we observed highly aligned collagen fibers travelling in the horizontal direction from the sinus before joining into larger bundles as they turned to travel vertically downward, by which point the fibers were strongly aligned and closely bundled again (Figure 1E).

SHG imaging of the inter‐leaflet triangle revealed horizontally orientated bundles of collagen fibers that emerged from the leaflet and sinuses on either side (Figure 1F). Fibers regularly intertwined as they transitioned from horizontal, to diagonal and then primarily vertical orientation. The vertically aligned bundles exiting the top of the image continue into the commissural region. Changes in fiber direction were gradual and accompanied by fiber recruitment into discretised bundles as they pivoted before reaching vertical or diagonal alignment.

In order to validate SHG collagen fiber orientation data, correlative focused ion beam scanning electron microscopy (FIBSEM) imaging was conducted on the same samples. A specific region from the SHG images was selected and co‐registered to the FIBSEM slice (Figure 2A). At low magnification, the FIBSEM slice shows collagen fiber bundles filling the space between the valve fibrocytes (Figure 2B). At higher magnification, individual collagen fiber crosssections can be seen grouped together in bundles, travelling primarily circumferentially (C), as well as one bundle changing orientation from radial (R) to longitudinal (L) (Figure 2C). This agrees with observations from SHG imaging in Figure 2A, where collagen fibers primarily travelled circumferentially, perpendicular to the plane of the slice, whilst gradually turning to run longitudinally, parallel to the same plane as the slice.

Figure 2.

Correlative SHG and FIBSEM imaging of aortic valve commissure. A) Overview SHG image displaying location of FIBSEM imaging plane. Scale bar = 1 mm. B) Corresponding low magnification FIBSEM image taken in the circumferential direction displaying the aortic valve fibro‐cellular microstructure in the upper regions of the commissure. Scale bar = 10 µm. C) Corresponding high magnification FIBSEM image showing individual collagen fiber cross sections travelling in the circumferential (C) direction, and longitudinal fiber sections running in the radial (R) and longitudinal (L) direction. Scale bar = 1 µm.

From the multi‐modal imaging analysis of the porcine aortic valve, we can summarize the following: 1) The commissure consists of a complex interwoven network of collagen fibers from multiple directions, which then coalesce into a primarily vertical alignment at the core of the commissure; 2) in the inter‐leaflet triangle region, collagen fiber bundles exhibit a regular diagonal weave before coalescing toward primarily vertical alignment, bundling more as they transition into the commissure; and 3) fiber orientation changes across interfaces were gradual, and often characterized by a transition from uniformly aligned large fibrous sheets into more discrete bundling as fibers turn at more acute angles. Due to the complexities associated with developing a sample preparation method compatible with precise correlative imaging across modes, and the large time taken to execute this, a limitation of this study was n = 1 samples. The observations from this high‐resolution multi‐modal imaging analysis performed on the aortic heart valve interfaces were then used as inspiration for interfacing fibrous heart valve scaffold designs.

2.2. Strategies to Interface Multi‐Phasic Melt Electrowritten Scaffolds

MEW is a powerful additive manufacturing technology with the potential to leverage specific aspects of the microstructural features observed in native tissues, such as collagen orientation, into the design of complex biomimetic scaffolds.^{[ 41} , ^{46 ]} The printing of MEW scaffolds relies on the precise control over mainly five process parameters: temperature, pressure, voltage, working distance, and collector speed (Figure 3A), resulting in a highly controllable molten polymer jet of micrometric resolution that is deposited onto a collector in a direct‐writing mode. However, unlike in other 3D printers, MEW does not allow stopping and starting of extrusion during a print, as this would disrupt the Taylor cone and cause printing defects. Thus, the design strategy for interfacing the different architectures of multi‐phasic MEW scaffolds requires careful consideration, as this could impact their ultimate mechanical and biological functionality.^{[ 5 ]} Here, we systematically investigated the effect on scaffold behavior of three different interfacing methods compatible with MEW printing: overlapping, suturing, and continuous (Figure 3B). Previously, we have successfully used the former two methods to interface spatially heterogeneous regions with MEW,^{[ 20 ]} but the effect of the interface on the resulting scaffold was not examined.

Figure 3.

Types of MEW interfacing methods investigated. A) Schematic of MEW system and its process parameters. B) Schematic of overlap, suture, and continuous MEW interfacing methods. C) Digital microscope (top) and SEM (bottom) images of MEW PCL scaffolds with 1 mm pore square patterns and 0.5 mm pore diamond patterns interfaced with the overlap, suture, and continuous methods. Scale bars = 1 mm.

Three different patterns were chosen (squares, diamonds, and serpentines) to be combined using the three interfacing printing strategies to produce bi‐phasic MEW scaffolds (Figure 3C and Figure S2, Supporting Information). Scaffolds were fabricated from medical grade poly(ɛ‐caprolactone) (PCL) to make five‐layered, 10 mm by 40 mm scaffolds, comprising two 20 mm patterns and an interfacial region. The size of the interfacial region depended on the interfacing method used: overlaps were 1 mm, sutures were 2 mm, and the continuous method resulted in a gradual, undefinable transition region. To demonstrate the capability of the three methods to interface spatially heterogeneous scaffolds, the pore size for each pattern was varied between 1 and 0.5 mm. The square pattern was chosen to alternate pore size, while the diamond and serpentine pore size were kept constant at 0.5 and 1 mm, respectively. Pore sizes were chosen to facilitate ease of printing and delineation of the effect of interfacing method independent of the scaffold's applicability to tissue engineering. Control scaffolds (40 mm by 10 mm single pattern scaffolds with no interface) were also printed to test each pattern and pore size independently of any interface. Average fiber diameter was 26.09 ± 1.90 µm. Figure 3C shows exemplary images of each of the interface types for the square‐diamond scaffolds. Exemplary images of the serpentine‐diamond and serpentine‐square scaffolds are shown in Figure S2, Supporting Information.

We next assessed how well each interfacing method resembled the programmed print path by comparing their morphologies using light microscopy and scanning electron microscopy (SEM). Bridging of fibers, a known imperfection where fibers will deviate from their intended path onto an adjacent fiber due to electrostatic repulsion caused by residual charge,^{[ 80 ]} was identified in varying quantities across all scaffolds. Overlapping scaffolds exhibited the densest interfacial region, with some cases of bridging. The suturing method was more customizable, but for the chosen design presented slightly less fiber density and comparable amounts of bridging to the overlapping scaffolds. The continuous printing technique best matched the planned print path with fewer fiber bridging defects and the lowest density. These results were consistent with literature, as the amount of printing defects in MEW scaffolds are expected to increase in areas of higher fiber density as well as in areas where the jet changes direction abruptly.^{[ 81 ]} Thus, continuous interfaces produced more accurate prints and exhibited a gradual transition of densities between spatially heterogeneous regions, as opposed to a dense band containing more fibers.

2.3. Effect of Interfacial Printing Methods on the Mechanical Properties of Bi‐Phasic MEW Scaffolds

In order to assess the effect of the interfacial printing method on the mechanical properties, we next performed uni‐axial tensile testing on the bi‐phasic MEW scaffolds with loading perpendicular to the interface (Figure 4 ). The effect of clamping distance from the interface on mechanical results was first investigated and no significant difference was observed in the calculated stresses (Figure S3, Supporting Information). However, strain values were observed to vary significantly and thus related metrics such as yield strain cannot be compared. Subsequently, due to the limited strain testing range of the equipment, scaffolds were clamped closer to the interface on the side of the weaker scaffold. Figure 4B showed a representative stress–strain plot of the tensile test on the continuously interfaced serpentine‐square scaffold. The initial elastic deformation (i) followed by plastic deformation (ii) of the scaffolds, was accounted for almost entirely by the weaker pattern, in this case the serpentine. The fibers of the weaker scaffold then began to neck resulting in strain hardening, shown by a gradual increase in stress. When the scaffold reached its ultimate tensile strength (UTS), the scaffold ruptured (iii) indicated by the sharp decrease in stress corresponding to the strain at which the two patterns separated from each other. A similar stress–strain curve was common among the rest of the interfaced scaffolds (Figure 4C). Stress–strain plots and scaffold testing videos for the serpentine‐square (Video S1, Supporting Information), square‐diamond (Video S2, Supporting Information), and serpentine‐diamond scaffolds (Video S3, Supporting Information) all showed similar behavior.

Figure 4.

Uni‐axial tensile testing of bi‐phasic MEW scaffolds interfaced using either the overlap, suture, or continuous technique. A) Time‐lapse images through tensile testing of a continuously interfaced 1 mm pore serpentine and 0.5 mm pore square scaffold. Scale bars = 5 mm. B) Representative stress–strain plot of continuously interfaced serpentine‐square scaffold with labels corresponding to regions of i) elastic deformation, ii) plastic deformation, and iii) scaffold rupture. C) Combined representative stress–strain plots for the serpentine‐square scaffolds with each type of interface and control scaffolds. D) Young's modulus, yield strength, and ultimate tensile strength for each type of bi‐phasic scaffold with interface method and control scaffolds (average over n = 3, error bars represent one standard deviation, significant differences were assessed using one‐way ANOVA with Tukey's multiple comparisons test. Statistical significance is shown at *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, ****p ≤ 0.0001.

The Young's modulus of the interfaced scaffolds was dominated by the more elastic, meaning lower modulus, of the two interfaced patterns (Figure 4D). In all except one case (continuous serpentine‐square) the elasticity of the interfaced scaffold was either equal to or greater than that of the more elastic pattern. For both the square‐diamond and the serpentine‐diamond scaffolds, the Young's modulus was not affected by the interfacing method. Interestingly, interfacing the square‐diamond scaffolds resulted in a significantly more elastic scaffold than either of its constituents, irrespective of interfacing method. In the serpentine‐square scaffolds however, the suturing method was the only technique resulting in a more elastic scaffold. These results suggest that despite the majority of the elastic behavior coming from the more elastic pattern, both patterns contribute in some manner to the final scaffolds elasticity. The yield strength indicates the level of stress at which the scaffolds will start to deform permanently. Results showed that yielding was dictated entirely by the weaker pattern, meaning the pattern with lower yield strength. Similarly, the UTS of the scaffolds was primarily determined by the weaker of the two patterns bar one case (sutured square‐diamond). Overall, the interfacing method did not have a significant effect on uni‐axial mechanical properties due to the dominance of the weaker scaffold. However, other mechanical properties relevant to tissue‐engineered scaffolds may be impacted and thus should be investigated.

We next investigated how the printing method used to interface two different patterns would alter the flexural properties of the resultant bi‐phasic scaffolds. To do so, we built a flexural testing apparatus in accordance with ASTM D1388 (Figure 5A). Each of the scaffolds was pivoted immediately on one side of the interface, leaving one pattern fixed, and one flexing, from which the bending angle was measured, and subsequent flexural stiffness (G, measured in Nm) calculated. G values for the control scaffolds were also calculated. Lower G values indicate a more flexible scaffold. Flexural properties depended largely on the pattern of the flexing scaffold (Figure 5B). Serpentine scaffolds were the most flexible, followed by diamonds, 1 mm squares, and 0.5 mm squares. Note that as the bending angle, θ, trends to 0°, G trends to infinity, meaning very stiff scaffolds that barely flex had exponentially higher values for G. This was observed for both the square patterns, which have distinguishably higher values for G. Interestingly, the type of interface significantly influenced flexural stiffness. In particular, for stiffer patterns, both suture and continuous interfaces enabled greater flexure for the stiffer square patterns, with continuous interfaces enabling the greatest increase in flexure. Contrastingly, overlapping interfaces often had a detrimental effect on flexure. For already flexible patterns, such as the serpentine and diamond, the interface had a negligible effect. Comparing interfacing techniques overall, a correlation can be observed whereby continuous interfaces offer the greatest degree of flexure, followed by suture and then overlap. Interestingly, a similar trend was observed when visually analyzing fiber density in the interfacial region between methods. Hence, this may be a possible mechanism behind the observed flexural properties, whereby an interface with higher density and more sites of fiber fusion will result in greater flexural stiffness.

Figure 5.

Flexural testing of bi‐phasic MEW PCL scaffolds with different interfacing designs. A) i) Schematic of flexural testing apparatus with θ representing the bending angle measured. ii) Close‐up view of a bi‐phasic scaffold displaying the location of the pivot line. Scale bar = 1 mm. iii) Image of testing of multi‐phasic scaffold with serpentine pattern. Scale bar = 10 mm. B) Flexural stiffness data (average over n = 3, error bars represent standard deviation, two‐way ANOVA with Tukey's multiple comparisons test resulted in p < 0.0001 for effect of pattern, p = 0.0002 for effect of interface and non‐significant differences between individual means).

In summary, interfacing MEW scaffolds will, in most cases, not weaken the scaffold's mechanical performance and often result in equal or greater elasticity than the individual constituents. Furthermore, using a continuous interfacing approach is recommended when trying to achieve greater flexural properties. This is valuable within the context of interfacial tissue engineering, as specific flexural properties may be desired across the interface between two regions. In the context of a tissue‐engineered heart valve scaffold, a high degree of flexure is required for the interface between the leaflet and inter‐leaflet triangle, as large amounts of cyclic bending will occur. This strategy could therefore continue to be leveraged to further unlock the capabilities of MEW scaffolds.

2.4. User‐Generated MEW Scaffolds with Complex Continuous Interfaces

Writing G‐code for continuous printing of MEW scaffolds is not trivial, especially in the case of complex scaffold architectures with multiple regions. Thus, we created a novel software with a graphical user interface (GUI) to enable the rapid generation of spatially heterogeneous MEW scaffolds which includes tailorable interface functions while using continuous printing (Figure 6A). Two sections, labeled A and B, can be defined, with the ability to select different lengths, pore sizes, and patterns (including rectangles, serpentines, auxetic cubes, and auxetic stars), in each. A complex interface function can be defined, allowing the border between sections A and B to be dictated by a mathematical function while still being continuous, including sinusoidal, parabolic, arrowhead, and angled shapes. Furthermore, three techniques of varying the Y‐pore size between sections were available: joining, halving, and fanning. The first method, joining, causes each horizontal pair of fibers from section A to join together as they transition to section B, resulting in double the pore size, as well as doubling the number of fiber layers in section B (Figure 6B). This technique was inspired by the observations made from native tissue (Figure 1F), where small collagen bundles gradually combine to form thicker bundles, as observed within interfaces elsewhere in the body.^{[ 3 ]} Alternatively, the halving method deposits an additional layer of the section A scaffold at an offset equal to half the pore size, resulting in halving of the overall pore size of section A, while maintaining equal fiber layers between section A and B. The fanning method causes a gradient change in porosity throughout section B via a user defined “fanning height” (Figure 6B,C). Videos demonstrating the live print path of each of the exemplary scaffolds shown in Figures 6B and 6C are shown in Videos S1 and S2, Supporting Information, respectively. Here we have shown only some of the capabilities of the GUI to generate continuous, complexly interfaced, spatially heterogeneous MEW scaffolds. This novel software and GUI unlocked the ability for rapid, iterative design of a wide range of continuously interfaced scaffolds, offering a wide range of tunability of mechanical and morphological properties that could be highly beneficial to interfacial tissue engineering applications using MEW.

Figure 6.

GUI of a novel software for generating complex, continuous, spatially heterogeneous G‐codes. A) Screenshot of the GUI, displaying user input options available, drop‐down menu options, and corresponding schematics indicating the effect on G‐code design. B) i) Inputs used for exemplary MEW scaffold design employing the joining and fanning methods, ii) overview digital microscope image of MEW PCL scaffold, iii) high resolution image of 1 mm pore auxetic star pattern, and iv) high resolution of the fanning interface showing continuous transition from 1 mm pore to 2 mm. C) i) Inputs used for exemplary MEW scaffold design employing the diagonal and fanning methods, ii) overview digital microscope image of MEW PCL scaffold, iii) high resolution image of 1 mm pore diagonal serpentine pattern, and iv) high resolution of the fanning interface showing continuous transition from 1 mm pore to 2 mm. Scale bars = 2 mm.

2.5. Design, Fabrication, and Testing of a Bioinspired Aortic Heart Valve Interfacial Scaffold

Leveraging the novel microstructural information of the aortic valve interfaces gained from the multi‐modal imaging investigation, we next aimed to design a bioinspired MEW scaffold of the aortic heart valve interfacial region using complex continuous interfaces. First, orientation analysis of the collagen structures from Figure 1F enabled a quantitative understanding of the micro fibrous collagen orientation (Figure 7A[i,iv]). Gradual changes in fiber distributions were observed from 0° (horizontal), to peaks at ±50° (diagonal), and then gradual changes again to ±90° (vertical) fiber alignment. Interestingly, few fibers were entirely vertically orientated, with the majority being woven in a regular diagonal pattern with a slight vertical inclination. The intricate and complex woven fibrillar structure of the inter‐leaflet triangle region was then simplified into schematic form to facilitate scaffold design (Figure 7A[ii]). The pattern in the leaflet region was based on our previously published heart valve leaflet design.^{[ 46 ]} The region between the leaflets, as opposed to being discretised into commissure, inter‐leaflet triangle, and annulus, utilized a gradient diamond pattern (Figure 7B). The gradient transitions from primarily horizontal alignment at the bottom, which aimed to emulate the circumferential fibers present in the annulus, to a primarily vertical alignment at the top, mimicking the longitudinal fibers present at the commissure (Figure 1D,E). In fiber design we avoided entirely horizontal or vertical orientation, as straight MEW fiber patterns are more rigid and weaker than diamond patterns of the same porosity (Figure 4D). Similar distribution of fiber orientations was achieved when comparing the fibers in the MEW scaffold design with those previously analyzed in native tissue (Figure 7A[iv]). Hence, the resulting scaffold design combined the native tissue observations with continuous interfaces to create a biomimetic scaffold design with gradient porosities, region‐specific layer numbers, tailored fiber orientations, and spatially heterogeneous regions.

Figure 7.

Biomimetic MEW scaffold for the interfacial region of the aortic heart valve, including continuous interfaces, gradient porosities, region‐specific layer numbers, tailored fiber orientations, and spatially heterogeneous regions. A) Biomimetic design process; i) color‐mapped orientation analysis of second harmonic generation images of the inter‐leaflet triangle region, scale bars = 0.5 mm, ii) schematic representing collagen fiber orientation in the inter‐leaflet triangle, iii) resulting MEW scaffold and color‐mapped orientation analysis of interfacial region, scale bars = 2 mm, and iv) overlaid orientation distribution from inter‐leaflet triangle tissue and MEW scaffold. B) Biomimetic G‐code design with dimensions. C) G‐code path with corresponding time stamps displaying continuous print path. D) SEM image showing close‐up view of continuous fiber interface employing joining technique where fibers have fused upon joining paths. Scale bar = 100 µm.

Scaffolds were successfully fabricated from PCL with an average fiber diameter of 27.69 ± 4.66 µm. The continuous printing path (Figure 7C and Video S3, Supporting Information) shows the joining technique employed to transition continuously from five layers in the leaflet region to ten layers in the middle of the scaffold. In some cases, where a double print layer was deposited due to the joining method, fiber fusion occurred simultaneously, resulting in thicker fibers instead of two layers of fibers (Figure 7D). Average fiber diameter for non‐fused and fused fibers was 24.34 ± 0.75 and 33.27 ± 2.30 µm, respectively. Comparison of the cross‐sectional areas of fused to non‐fused fibers confirmed the fiber fusion phenomena with fused fibers having approximately double (187 ± 14%) the area. This was likely due to the high temperature of the deposited fiber being maintained caused by the very short time between fiber depositions on the same path. This was reminiscent of the morphological arrangement of collagen fibrils found in native tissue, where smaller collagen bundles join together to form larger, tightly packed bundles as they change orientation (Figure 1F).

The scaffold was then mechanically characterized under bi‐axial, physiologically relevant strains (Figure 8A). Strain‐related metrics were used to understand tissue‐like characteristics such as anisotropy and viscoelasticity. It is important to note that this scaffold contained the leaflets, inter‐leaflet triangle, annulus, and commissural regions, meaning that comparisons to independent region‐specific data would not be strictly physiologically accurate. Attempts to perform a similar loading regime on porcine tissues were made, however, due to difficulties with clamping the tissue to the mechanical tester and variable cross‐sectional areas, reliable quantifiable data was not available to use as a comparison (Video S7, Supporting Information). Interestingly, we noted that across all samples of porcine tissue (n = 12) failure never occurred in the interfacial region. Thus, the scaffold was compared against reported values for native tissue in regions as similar to the scaffold as possible (Figure S4, Supporting Information). In future studies, haemodynamic testing in a flow loop setting will ensure proper physiological loading of a complete heart valve scaffold.

Figure 8.

Bi‐axial mechanical characterization of MEW interfacial heart valve scaffolds. A) Bi‐axial tensile testing set up displaying circumferential and longitudinal testing directions. Scale bar = 5 mm. B) Young's modulus, yield strength, and yield strain. C) Hysteresis behavior from cyclical testing at constant strain showing representative stress–strain plots in both directions and quantification of hysteresis shown by the area under the unloading curve divided by the area under the loading curve for the same cycle. D) Representative relaxation behavior after four incremental increases in strain (2.5% and 5% in the longitudinal and circumferential direction, respectively), held for 1000 s, followed by quantification of relaxation percentage for each step (all data shows average over n = 3, error bars represent one standard deviation, significant differences were assessed using parametric ratio paired t‐tests, *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001), ns = not significant.

The ratio of strain rates in the circumferential to longitudinal direction was chosen as 2:1 (Figure S4, Supporting Information).^{[ 82 ]} When tested to failure, the scaffold exhibited similar Young's modulus in both directions, whereas the yield strength in the circumferential direction was approximately double that of the longitudinal direction (Figure 8B). Scaffold strength could be improved by adjusting layer number or fiber diameter. Strain–displacement vector mapping showed complex regional anisotropic deformation behavior (Video S8, Supporting Information), where the circumferential and longitudinal strains were accounted for primarily by the leaflet and inter‐leaflet triangle regions, respectively. However, strain–displacement vectors showed a gradual change in direction and magnitude between these regions, indicating a smooth transition of load across the interface. Notably, the yield strain in the circumferential and longitudinal directions was 27.7 ± 6.9% and 16.6 ± 5.4%, respectively, both of which were higher than the average maximum strains reported in literature across the different regions of the aortic valve (Figure S4, Supporting Information).^{[ 82} , ⁸³ , ^{84 ]} This implies the scaffold will remain within the elastic region and not plastically deform under physiological strains.

Hysteresis of the scaffold was then determined by cyclically testing the samples to a constant strain of 20% and 10% in the circumferential and radial directions, respectively, and calculating the ratio of the area under the unloading curve to the loading curve (Figure 8C). A constant low level of hysteresis (≈20% energy loss per cycle) was observed in both directions. Our data concurs with reported 17% hysteresis for porcine aortic heart valves.^{[ 85 ]} The relaxation behavior of the scaffold was characterized using 5% and 2.5% strain increments in the circumferential and radial directions, respectively (Figure 8D). The scaffold was then allowed to relax for 1000 s, an adequate time frame to exhibit the majority of relaxation behavior.^{[ 86 ]} At all strain levels, the scaffold exhibited rapid initial relaxation, before stabilizing to an asymptote, which was then reported as relaxation percentages. A consistent degree of relaxation was maintained in the circumferential direction of 27.6 ± 1.5%, whereas the radial direction exhibits initial large relaxation before stabilizing to 30.6 ± 1.7%. These values were remarkably similar to values reported for porcine aortic heart valve relaxation.^{[ 85 ]} In summary, the bioinspired aortic heart valve interface scaffold demonstrated characteristics reminiscent of native tissue including anisotropy, hysteresis, as well as yield strains within the range of native tissues.

3. Conclusion

In this work, for the first time, we unraveled the microstructural organization of collagen fibers in the interface regions of the aortic heart valve using high resolution correlative multi‐modal imaging. This was leveraged to inspire functional design of MEW scaffolds including biomimetic interfacing regions for heart valve tissue engineering applications. A systematic morphological and mechanical study into three different methods of interfacing bi‐phasic MEW scaffolds revealed that weaker regions dominate the tensile mechanical response. With regard to flexural stiffness, a property often overlooked in scaffold design, continuously interfacing bi‐phasic scaffolds showed enhanced flexibility of the resulting constructs. Crucially, a novel software and accompanying GUI enabled the rapid design and G‐code generation of not only an array of MEW complex scaffold designs, but also the continuous interfacial boundary across which these regions were connected. This software unlocks new capabilities of MEW for the field of interfacial tissue engineering. Subsequently, we designed, fabricated, and tested a bioinspired MEW scaffold for the aortic heart valve interfacial region, showing for the first time a scaffold containing continuous interfaces, gradient porosities, region‐specific layer numbers, and tailored fiber orientations in a singular biomimetic design. When tested under physiologically relevant bi‐axial conditions, the scaffold exhibited promising tissue‐like characteristics, including strain yield, hysteresis, and relaxation behaviour, all similar to native tissue. Future studies will focus on achieving a complete heart valve scaffold to evaluate its functionality in physiological, haemodynamic settings. Additional studies to confirm human tissue structures and analysis of tissue variability will add strength to the outcomes reported here. Design and printing methodologies developed in this study were inspired by the detailed investigation of native tissue, and thus we encourage similar investigations to advance and transform the field of biomimetic tissue engineering.

4. Experimental Section

Tissue Sourcing and Preparation

Collection and use of porcine animal tissue for this work was approved by the University of Western Australia Institutional Biosafety Committee (F 69199). Porcine hearts were sourced from the University of Western Australia Large Animal Facility, where tissue was excised within 2 h of euthanasia. Immediately after removal, hearts were dissected to leave the aortic roots with their valves, sinuses, the first few millimeters of the coronary arteries, and a portion of the ascending aorta.

Primary fixation of (n = 1) tissue was in fresh 4% paraformaldehyde solution (Cat#C007, ProSciTech, Australia) made in 0.1 molar phosphate buffer (pH 7.4). Fixation was conducted for 1 h at room temperature and equivalent to diastolic pressure (80 mmHg) with the hybrid immersion fixation and hydrostatic pressure distension apparatus shown in Figure S1, Supporting Information. Secondary fixation was via immersion overnight in 2.5% glutaraldehyde (Cat#EMS16400; ProSciTech, Australia) in 0.1 molar phosphate buffer (pH 7.4). Excess tissue was removed by dissection before storage in glutaraldehyde fixative solution at 4 °C.

Micro‐Computed Tomography

During imaging, tissue samples were kept moist under paper towel soaked with glutaraldehyde solution and placed inside a sealed zip lock bag. Micro‐CT imaging (Skyscan 1176, Bruker‐micro‐CT, Kontich, Belgium) was performed at a source voltage of 45 kV, source current of 556 µA, 86 ms exposure time, 34.81 µm pixel⁻¹ resolution, with a 0.2 mm aluminum filter, and 0.7° over 360° rotation with 2/image frame averaging enabled. The micro‐CT data was segmented and reconstructed using in‐house software before being brought into STAR‐CCM+ (v16.04.012‐R8, Siemens) for further smoothing and processing.

Second Harmonic Generation Imaging

A 5% agarose (Agarose LE, analytical grade, Promega, Australia) embedding solution was used at 65 °C. Tissue samples were removed from the fixative solution, and blotted dry, before being placed into a custom sized (≈30 mm cubed) Lego (The Lego Group, Denmark) mold that contained the entire volume of the tissue and minimized the amount of agarose solution used. The agarose solution was decanted around the tissue, ensuring minimal bubbles remained. Embedded tissue was cooled to room temperature, then stored at 4 °C overnight. For sectioning, the solid gel/tissue block was removed from the mold and superglued to the vibratome stage (Vibratome 3000, The Vibratome Company, St Louis, MO, USA). After sectioning the block face to the region of interest, three consecutive sections were taken at 250 µm thickness. Optical imaging was then conducted on these slices using an inverted A1RMP multi‐photon microscope (Nikon) equipped with 10×/0.45NA objective lens (Nikon) and tunable laser (10 mW output; 900 nm emission wavelength). All images were captured using NIS‐Elements AR software (v5.30.02; Nikon).

Electron Microscopy of Tissue

Electron microscopy samples were prepared from sections after SHG imaging by microwave assisted processing as outlined by Nguyen et al.,^{[ 87 ]} using a BioWave Pro microwave system (Pelco). Briefly, the samples were osmicated using the R‐OTO method,^{[ 88 ]} en bloc stained with aqueous uranyl acetate and lead aspartate, then dehydrated through graduated ethanol series (80%, 90%, 95%, 100%, and 100% v/v) and propylene oxide (100% and 100% v/v). Sample infiltration with Araldite 502/Embed 812 was performed via graduated concentration series in propylene oxide under vacuum (25%, 50% 75% 100%, and 100% v/v). The samples were polymerized at 60 °C for 48 h and trimmed on an ultramicrotome (Leica UC6, Leica Biosystems) in preparation for imaging on a FEI Helios Nanolab G3 CX DualBeam FIBSEM (Thermo Fischer Scientific). The target region was milled using a Gallium FIB current of 65nA at 30 kV, before backscatter electron imaging of the block‐face at an accelerating voltage of 2 kV in magnetic immersion mode.

MEW Scaffold Fabrication

Scaffolds were fabricated using an in‐house‐built MEW device as reported previously,^{[ 20} , ³⁸ , ⁴⁶ , ^{47 ]} from medical‐grade PCL (Purasorb PC 12, Corbion, The Netherlands). The specific processing parameters used were: air pressure‐driven extrusion at 100 kPa; through a 23G metallic needle; at a working distance of 3 mm; a translation speed of 400 mm min⁻¹; with voltage of 4.4 ± 0.1 kV applied to the spinneret and grounded collector plate; with 86 °C ring heater and 31.5 °C bed heater. Exact processing parameters used were varied slightly depending on the environmental conditions on the day. All bi‐phasic scaffolds used for tensile and flexural testing had dimensions of 10 mm height × 40 mm width (20 mm width per pattern).

MEW Scaffold Imaging and Measurements

MEW scaffolds were first imaged using a Hirox RH‐2000 digital microscope (Hirox, Europe) at a resolution of 4.51 µm pixel⁻¹. The accompanying software was used to measure fiber diameter at three locations across n = 3 scaffolds of each type. Scaffolds were then cut using scissors and fixed to SEM stubs using double sided carbon tape. Stubs were then sputter coated with gold, two at a time, for 60 s, using a JEOL Smart Coater (JEOL, USA). SEM images were then acquired using a JCM‐6000 Benchtop SEM (JEOL, USA) at acceleration voltage of 10 kV, 30× magnification, with high probe current and high vacuum mode.

Uni‐Axial Tensile Testing

Scaffolds were mechanically tensile tested uni‐axially across the interface using the CellScale Biotester (CellScale, Waterloo, Canada) equipped with a 1.5N load cell. Samples (n = 3 for each scaffold type) were secured with custom 3D printed miniature clamps and suspended in air at room temperature. Uni‐axial tests used a displacement rate of 1% length/min. Stress–strain curves were plotted using the generated force‐displacement data, where strain was defined as engineering strain (change in length/initial length). Initial length was kept constant at 15 mm for all calculations. Cross‐sectional area was quantified as the scaffold wall height, as opposed to the peak height. Six measurements were taken across n = 3 samples of each bi‐phasic scaffold. Cross‐sectional area was observed to vary within individual scaffolds between the two patterns and the interface. However, as the majority of deformation occurred in the weaker pattern, confirmed by visual analysis, the cross‐sectional area of the weaker pattern was used to calculate associated stresses. Averages for each pattern: serpentines = 1.06 ± 0.03 mm², diamonds = 1.25 ± 0.20 mm², 1 mm squares = 1.40 ± 0.21 mm², 0.5 mm squares = 1.16 ± 0.04 mm². Young's modulus was calculated from the slope of the stress–strain curve at the steepest linear region, the strain range for which depended on the sample/pattern. Yield strength was calculated as the point where the stress–strain curve plateaued. UTS was determined as the maximum stress reached for each sample.

Bi‐Axial Tensile Testing

Bi‐axial tensile tests were carried out using the same equipment as the uni‐axial tests, with a 5N load cell. A pre‐load of 50 mN was applied before every test to ensure a consistent initial state of the scaffolds. Young's modulus and yield strength were calculated using the same protocol as for uni‐axial tests. Cyclical testing was conducted for one pre‐load cycle (to establish the in‐vivo state of the material^{[ 89 ]}) and nine subsequent loading cycles, which were then plotted as stress–strain. Hysteresis was calculated as the ratio of the area under the unloading curve to the loading curve for each cycle. The relaxation behavior of the scaffold was characterized using 5% and 2.5% strain increments in the circumferential and radial directions, respectively (Figure 8D). The scaffolds were then allowed to relax for 1000 s, an adequate time frame to exhibit the majority of relaxation behavior.^{[ 86 ]} Relaxation percentages were reported as the difference between maximum and minimum stress for each step.

Flexural Testing

Flexural testing was conducted using the Peirce cantilever test method (ASTM D1388). A custom testing apparatus was designed in Fusion 360 (Autodesk, USA) and printed using Prusament PLA (Prusa Research, Czech Republic) with a Prusa MK3s+ (Prusa Research, Czech Republic). Angles were measured six times independently per flexing pattern, thrice in one direction and then thrice more with the scaffolds flipped over to account for the natural warping of the scaffold after printing.

Complex G‐Code Generator GUI and Path Visualization

An in‐house program was developed using Python (Python Software Foundation, USA) to incorporate continuous interface design into spatially heterogeneous G‐codes for MEW printing. A script generating a GUI was made, for which the desired parameters of the scaffold could be entered. Once entered the interface would execute other scripts that perform the necessary calculations to plot data for the desired printing path. This data was then converted by another script into a G‐Code Notepad file for input into MEW printer software. To permit visualization of G‐code, another custom Python‐based software was built that imported raw G‐code and produced both stationary and dynamic renders of the print path, with customizable scale, colors, line widths, and path speeds.

Orientation Mapping

Orientation analysis of SHG images and MEW scaffolds was done using the plugin OrientationJ,^{[ 90 ]} for the software Fiji (National Institute of Health, USA).^{[ 91 ]} Using the OrientationJ Analysis module, a hue‐saturation‐brightness color survey was calculated using a 2‐pixel cubic spine. Then the OrientationJ Distribution module was used to plot the distribution of orientations.

Statistical Analysis

All data was presented as mean ± standard deviation for n = 3 samples unless stated otherwise. For uni‐axial tensile testing data, significant differences were assessed using one‐way ANOVA with Tukey's multiple comparisons test. For flexural stiffness data, significant differences were assessed using two‐way ANOVA with Tukey's multiple comparisons test. For bi‐axial tensile testing data, significant differences were assessed using parametric, ratio paired t‐tests. Alpha was 0.05 for all tests. All statistical analysis was conducted using GraphPad Prism software (GraphPad, San Diego, USA).

Conflict of Interest

M.J.V., C.L. and E.D.J.P. have submitted a provisional patent related to this work. The remaining authors declare no conflict of interest.

Supporting information

Supporting Information

Supplemental Video 1

Supplemental Video 2

Supplemental Video 3

Supplemental Video 4

Supplemental Video 5

Supplemental Video 6

Supplemental Video 7

Supplemental Video 8

Vernon M. J., Lu J., Padman B., Lamb C., Kent R., Mela P., Doyle B., Ihdayhid A. R., Jansen S., Dilley R. J., De‐Juan‐Pardo E. M., Engineering Heart Valve Interfaces Using Melt Electrowriting: Biomimetic Design Strategies from Multi‐Modal Imaging. Adv. Healthcare Mater. 2022, 11, 2201028. 10.1002/adhm.202201028

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.