Type I atelocollagen (from bovine amnion, pepsin-digested, Memphis Veterans Affairs Medical Centre’s Research Service, US), HA (hyaluronic acid, sodium salt, Thermo Fisher Scientific, US), and elastin (from bovine neck ligament, Merck Life Science Ltd, UK), were selected as scaffold materials. EDC hydrochloride (Thermo Fisher Scientific Ltd, UK) and NHS (Sigma-Aldrich Ltd, UK) were chosen as crosslinking agents.

Pepsin-digested collagen was first made into an aqueous solution to undergo fibrillogenesis. Lyophilised collagen was dispersed into 0.05 M acetic acid solution (pH 3.2) and subsequently added into tris-buffered saline (TBS, pH 7.6) of equal volume with gentle stirring. The pH of the mixture was adjusted to a physiological level, and the mixture was transferred to incubate at 37°C until the formation of collagen fibres [25].

EDC and NHS (ratio 2:1) were dissolved in acidic deionised water (pH 4.8). The fibrillar collagen suspension was poured into the crosslinking solution and left to react for 4 hours with gentle stirring. The mixture was then washed first with 1.0 M NaCl and then with deionised water to remove unreacted chemicals [26].

The crosslinked collagen suspension was cast into polytetrafluoroethylene (PTFE) moulds of various dimensions and frozen at –20°C overnight, and the frozen samples were freeze-dried in a Christ I-5 freeze-dryer (Martin Christ Gefriertrocknungsanlagen GmbH, Germany) for 24 hours under 5 Pa vacuum to obtain porous foam-like single-layer scaffolds.

The collagen-HA and elastin (in the form of fibrillar elastin gel) layers were prepared according to the previous studies [26]. Freeze-dried crosslinked collagen scaffolds were immersed in 0.4% wt adipic acid dihydrazide (ADH)-modified HA solution and the resultant HA-infiltrated collagen structure was washed before moulding. The fibrillar elastin gel used is a novel elastin developed within the group, it is a gel-formed elastin in between insoluble elastin and soluble α-elastin. To assemble the tri-layer structure, a multi-step casting method was adopted: A layer of collagen was loaded into the mould to one-third of the height, then a freshly washed pre-fabricated collagen-HA layer of equal thickness was gently placed on top, lastly the remaining one-third volume was filled with elastin gel. The entire structure was frozen at –20°C overnight and then freeze-dried to obtain porous tri-layer scaffolds.

The occurrence of fibrillogenesis was visualised using SEM (Zeiss), and the accelerating voltage was set at 5 kV. The collagen suspensions before and after fibrillogenesis induction were deposited onto aluminium pin stubs and left air-dry in a near vacuum before coating with platinum (Leica EM ACE600). The characteristic D-spacing pattern was analysed with ImageJ (National Institutes of Health, US).

The microstructure of the scaffolds was characterised by SEM (Zeiss) with the accelerating voltage being 10 kV. Samples were carefully sectioned into 1–2 mm slices using a sharp surgical scalpel and coated with 5 mm platinum. For single-layer scaffolds, samples were cross-sectioned; for tri-layer scaffolds, samples were sectioned along the axial plane to demonstrate the layered structure. For the pore morphology study, the pore area was assumed to be circular and analysed using ImageJ. The circularised pore diameter (d) was obtained with Equation 1, where A is the measured pore area.

The chemical structure of the single-layer scaffolds was revealed by FTIR (Varian Excalibur FTS 3500 FT-IR Spectrometer) with the resolution and aperture setting being 4 cm^-1 and 0.5 cm^-1 at 2,000 cm^-1. The spectra obtained were processed and analysed using OriginLab.

Uniaxial compression tests were conducted on all scaffolds using a DMA Q800 (TA Instruments, US) at room temperature. The cylindrical samples (N = 5) were subjected to a 0.3 N/min force rate until 3 N was reached. Tensile and storage moduli mentioned supplementarily in the Discussion section were also obtained by DMA Q800. For viscoelasticity testing, the rectangular samples (N = 5) were subjected to 0.3 N preload and 0.1% constant strain under tension, with the tensile frequency oscillating at 0.1, 0.3, 1.0, 3.0, and 10.0 Hz. For tensile testing, the rectangular samples (N = 5) were subjected to a 0.3 N/min force rate similar to compressive testing until failure.

For tri-layer scaffolds, unidirectional three-point bend tests were performed using the DMA Q800 in both the WC (with curvature, sample bent towards the collagen layer) and the AC (against curvature, sample bent towards the elastin layer) directions. The rectangular samples (N = 5) were subjected to a 0.3 N/min force rate over a 20 mm span. Young’s modulus was calculated from the slope of the linear elastic region of the resultant stress-strain curves.

An easy and adequately accurate “self-deflection” test was developed within the group and was further improved by the author to test the bending properties of the hydrated samples [17]. It is the method that relies on linking the bending stiffness (EI, with E being the bending modulus and I being the second moment of inertia of a rectangular beam) to the maximum deflection (d_max) of the structure under uniformly distributed load (q) by the beam theory (Equation 2). In this method, the load is provided by the gravity of the phosphate-buffered saline (PBS) absorbed, which is expressed by the mass change/solution absorption before and after saturation. Figure 3 depicts the measurement of the maximum height (h) from the sample central plane to its lowest point and the sample thickness (t) to calculate d_max using Equation 3.

The rectangular samples (N = 5) were pre-saturated with 0.1 M PBS at room temperature for an hour and then placed on a U-shaped metal plate with a span of 20 mm (l) to allow free deflection. With the mass of the sample being recorded before and after saturation, the mass change (Δm) was calculated, and the bending modulus can therefore be calculated using Equation 4, where g is the acceleration of gravity (9.8 m/s²). The final equilibrium state of deflection was captured by a camera (Canon G7X Mark III, Japan) and the measurements were conducted using ImageJ. For tri-layer scaffolds, the tests were conducted in both the WC and AC directions.

Uniaxial compression tests in the hydrated state were carried out only on tri-layer scaffolds due to limited resources. The test conditions were kept the same as the non-hydrated tests for direct comparison, and the hydration was achieved by testing the pre-saturated cylindrical samples in PBS submersion.

The atelocollagen was demonstrated to achieve fibrillogenesis successfully via the distinct fibre structure formation captured in Figure 4. The characteristic D-spacing pattern of collagen fibres was prominently displayed, and the spacing was measured to be 69.3 nm (N = 43), coinciding with the range (64–71 nm) reported in various literature [27–29].

Non-crosslinked scaffolds were not sufficiently stiff to be sectioned, hence only crosslinked collagen-only scaffolds were characterised. Scaffolds with a porosity of over 97% were successfully produced, meeting the tissue engineering requirement of 90% [30]. Figure 5 presents representative SEM images at different magnifications of the collagen-only scaffolds, exhibiting desired interconnective and intact open pore morphology. Figure 6 showcases the seamless transitions between layers as well as the overview of three layers. Both mainly contain collagen, the layers collagen-only and collagen-HA show a subtle transition yet detectable by a change in the pore wall structure, i.e., the collagen-HA layer exhibits greater pore size (220 ± 142 μm) than the collagen-only layer (187 ± 83.0 μm) with less pore wall integrity and pore uniformity. The transition between the collagen-HA and elastin layer is more visibly marked by the flake-like feature of the elastin layer. The pore size of the collagen-only scaffolds was calculated to be 292 ± 41.4 μm with a wall thickness of 15.4 ± 4.40 μm. Interestingly, the average pore size of the collagen layer of the same making in the tri-layer structure, however, saw an approximate 40% decrease (187 ± 83.0 μm).

The FTIR spectra of the collagen scaffolds contained all five characteristic peaks of pure collagen, respectively, the amide A and B bands associated with N-H stretching, amide I signal resulted from C=O stretching and N-H bending, amide II peak containing N-H bending and C-N stretching signals, and the triplet peak amide III [31, 32]. It can be concluded that the fabrication method had no negative impact on the chemical backbone of collagen. Moreover, the amide III/1,450 cm^-1 intensity ratio remained at ~1, indicating the preservation of the collagen triple helical structure [33].

Figure 7 presents the FTIR spectra of both non-crosslinked and crosslinked collagen scaffolds with normalisation against 2,930 cm^-1, which is a result of C-H stretching and is unaffected by EDC/NHS chemistry [34]. Both the amide I band at 1,633 cm^-1 and the ester band at 1,161 cm^-1 confirmed that the crosslinking chemistry was effective and unharmful to the collagen backbone structure [35, 36]. The relative crosslinking extent was further studied to examine the contributions of the amide and ester bonds by comparing the relative intensity before and after crosslinking, and the result is visualised in Figure 8. After partial crosslinking, the amide bond intensity saw a 1.9 times improvement. In comparison, the ester bond intensity increased by 1.3 times, indicating that the amide bond played the leading role in the EDC/NHS crosslinking chemistry of collagen.

In the non-hydrated state, EDC/NHS crosslinking was demonstrated to have a positive impact on the compressive stiffness of collagen-only scaffolds, with a 1.4-fold improvement in the compressive modulus (from 29.1 ± 4.24 kPa to 39.7 ± 1.68 kPa). It can be observed from Figure 9 that both the collagen-only and tri-layer scaffolds shared the stress-strain curve features of lightly crosslinked polymers, with the stress plateau showing a noticeably positive slope. The compressive behaviour of the other two layers was characterised outside of the scope of this paper but included for comparison. It can be interpreted from the curves that the much stiffer collagen-HA layer and elastin layer mainly contribute to the stiffness of the tri-layer structure, in other words, the addition of the two layers increased compression resistance, resulting in a compressive modulus of 112 ± 32.9 kPa.

J-curve deformation is preferred because it allows the structure the flexibility to undergo normal physiological movement and the deformation resistance when excessive force is applied. After hydration, the tri-layer scaffolds exhibited a J-shaped stress-strain response under compression (Figure 10), confirming the biomedical potential of the design. Hydration effectively extended the linear region from 12% strain to 45% strain; however, it drastically reduced the compressive stiffness (expressed by Young’s modulus) from 112 ± 32.9 kPa to 0.907 ± 0.0434 kPa.

According to Equations 2 and 3, the hydrated bending modulus of the collagen-only scaffold was calculated to be 6.46 ± 1.15 kPa. The theoretical bending modulus in the non-hydrated state was calculated to be ~135 kPa using the preliminary tensile and compressive moduli obtained from experiments. It can be seen that hydration drastically reduced scaffold stiffness; however, it is still deemed ideal for that softer matrix (~6–14 kPa) is reported to better suit cell seeding [9].

For tri-layer scaffolds, the bending response was examined in both WC and AC directions due to the layered nature of the designs. The scaffolds are expected to be stiffer and more bending-resistant in the AC direction to effectively prevent blood backflow and the stronger compression resistance of elastin compared to collagen is predicted to help achieve that. The results of the bending moduli are organised in Table 1. It can be seen that in both non-hydrated and hydrated states, the scaffolds are stiffer in the AC direction, coinciding with the anisotropicity discovered in native heart valve leaflets [17]. The results again confirmed that hydration reduces scaffold stiffness by a great amount, in this case, 95 %.

This study has successfully fabricated and characterised the proposed tri-layer scaffold design with type I atelocollagen, HA, and fibrillar elastin gel. The design closely mimicked the tri-layer structure of the native aortic valves, and the bending anisotropy was successfully introduced. The proposed fabrication method, freeze-drying, and the crosslinking chemistry, EDC/NHS partial crosslink, were proven to be effective while able to preserve the collagen backbone structure. The design demonstrated a suitable microstructure for cell-related activities with varying pore sizes and interconnected pores. Under physiological conditions, the structure experienced the desired J-shaped curve under compression. Overall, the proposed tri-layer scaffold design showed potential for future developments as a tissue engineering heart valve.