Additive manufacturing of scaffolds with sub-micron filaments via melt electrospinning writing

Poly--caprolactone (PCL) was purchased from Sigma-Aldrich Co. LLC and its molecular weight distribution analysed by SEC (detector: Jasco, RI 2031 plus/eluate: chloroform): M_n = 35 × 10³ gmol⁻¹/M_w = 83 × 10³ gmol⁻¹/M_z = 14 × 10⁴ gmol⁻¹/PDI = 2.4. The corresponding viscoelastic properties were measured with plate-to-plate (d_p = 20 mm/distance l_p = 1 mm) oscillating rheometry (MCR 301, Anton Paar, Ostfildern-Scharnhausen, Germany) with 20% deformation and 10 rad s⁻¹ with increasing temperature from 80 °C up to 130 °C with 0.025 °C s⁻¹.

Figure 1(B) shows a schematic of a custom-built machine used to perform MEW in this study. It contains a high voltage source (DX250R, EMCO, Hallein, Austria) controlled by voltage divided measurement (Digit Multimeter 2100, Keithley, Cleveland, USA), a pneumatically regulated melt feeding system (regulator, FESTO, Berkheim, Germany) and a planar movable aluminum collector plate (XSlide, Velmex, New York, USA) triggered by G-code (MACH 3 CNC software, ARTSOFT, Livermore Falls, USA). The electrical heating system was proportional-integral-derivative-regulated (TR400, Delta-t, Bielefeld, Germany) to assure a stable melt temperature profile. Two advantages of MEW are that ventilation for the device is not required, nor is the removal of toxic solvents after fabrication [21, 25].

The instrument parameters were systematically investigated to achieve sub-micron filaments and deposited as boxes with minimal geometric deviations. The following parameters were varied: accelerating voltage (2.0–10.0 kV), collector distance (1–10 mm), heating temperature (80–120 °C), feeding air pressure (0.5–4.0 bar), spinneret diameters (21 G; 23 G; 25 G; 27 G; 30 G; 33 G) and axis velocity (1000–9000 mm min⁻¹).

Box-structure scaffolds were melt electrospun by alternating layer deposition via 0° and 90° layers (15 × 15 mm²) with turning loops to optimize the deposition behaviour of the jet as shown in figure 1(C). The turning loops help with fibre placement and reduce structural defects near the edges. For SEM, DSC, XRD and in vitro research, 50 layers (100 fibres overlap at intersections) were deposited in each direction (0° and 90°) and the filament spacing was programmed to be 100 μm (figure 1(D)).

The MEW box scaffolds were characterized with SEM (Supra 25, Zeiss, Oberkochen, Germany). For each sample, the filament diameter and mesh-widths in 0° and 90° direction were measured with n = 50, n = 100 respectively, after gold-sputtering. To observe surface roughness, AFM (FlexAFM, Nanosurf, Langen, Germany) measurements were performed five times with 5 μm in length on uncoated single filaments.

The crystallization behaviour of these ultrafine filaments was also investigated with XRD (D5005, Siemens, Berlin, Germany). To exclude diffraction effects induced by the lattice structure of the boxes, uniaxial aligned filaments were prepared with 100 μm filament spacing and analysed in Bragg–Brentano geometry, using Cu–Kα radiation (λ = 0.154 18 nm) with a voltage of 40 kV and a tube current of 40 mA. Scans were performed in a 2Θ range from 10 to 60° with a step size of 0.02° and a measuring time of 8 s per step. In addition to this, the uniaxial samples were irradiated at planar tilted angles (0°, 30°, 60°) to investigate a possible anisotropic crystallinity. Afterwards the sample was molten at 85 °C for 30 min and analysed in the same manner to assure constant experimental conditions. To investigate the validity of the results, an unaligned sample was prepared with lowered axis velocities (800 mm min⁻¹) and measured three times.

Furthermore, the melting behaviour of the electrospun and raw material were characterized with DSC (204 F1 Phoenix, Netzsch, Selb, Germany). Therefore, similar masses (m_mesh = 6.26 mg and m_raw = 6.13 mg) were heated from −120 °C up to 90 °C twice and held for 5 min. Heating and cooling rates were performed with 10 K min⁻¹ respectively −10 K min⁻¹. As a control, DSC measurements with slow heating rates (2 K min⁻¹) were performed. To investigate changes in morphology and quenching effects, additional filaments and box scaffolds were heat-treated at 55 °C (below the crystallization onset temperature) for 18 h and analysed with AFM, XRD and DSC.

Cell adhesion behaviour on MEW scaffolds was investigated using primary human mesenchymal stromal cells (hMSCs) from trabecular bone that were isolated from the femoral head of patients undergoing total hip arthroplasty [26]. HMSCs were cultivated in DMEM/Ham's F-12 + GlutaMAX^TM-I medium supplemented with 10% fetal calf serum, 100 U ml⁻¹ penicillin, 100 mg ml⁻¹ streptomycin (all Gibco®, Life Technologies GmbH, Darmstadt, Germany) and 50 μg ml⁻¹ L-ascorbic acid 2-phosphate (Sigma-Aldrich Chemie GmbH, Schnelldorf, Germany). HMSCs were seeded on the three different scaffold types at passage two at a density of 3.1 × 10⁴ cells cm⁻² and incubated for 10 days in vitro (DIV) at 37 °C in a humidified atmosphere with 5% CO₂. For in vitro experiments, MEW scaffolds with 50 layers of filaments and mesh-widths of 90 and 150 μm were deposited onto microscope slide coated with NCO-sP(EO-stat-PO) as described previously [27]. This coating is used to adhere the scaffold to the glass substrate as well as reducing protein adsorption on the surface [28, 29].

The constructs were fixed with 4 wt% paraformaldehyde, permeabilized with 0.2 vol% triton-X 100 and stained with phalloidin, tetramethylrhodamine B isothiocyanate (Sigma-Aldrich Chemie GmbH, Schnelldorf, Germany) and Hoechst 33342, trihydrochloride trihydrate (Molecular Probes, Life Technologies GmbH, Darmstadt, Germany) according to instructions. Inverse fluorescence microscopy using ApoTome and Z-stack function (Axio Observer.Z1 with ApoTome.2, Zeiss, Göttingen, Germany) and Zeiss software ZEN pro 2012 were used to obtain fluorescent images. Single pictures of optical sections of the Z-stacks were merged using ImageJ version 1.49 k.

With its relatively broad molecular distribution (PDI = 2.4) and moderate molecular weight (M_w = 83 × 10³ g mol⁻¹) the PCL used in this study was readily melt electrospun. As would be expected of a molten linear polymer, the loss modulus G'' = 1.1 × 10⁴ Pa dominated the storage modulus G'' = 9.6 × 10² Pa at 84 °C. Numerous parameters investigated resulted in a regular frequency 'pulsing' that was a significant barrier to fabricate uniform structures (figure 2(A)). This filament pulsing varied in frequency of 0.1–0.01 Hz, often every minute. Care was needed to adjust the instrument parameter, to balance the forces (electrohydrodynamics; gravity, surface tension) and mass flow (air pressure, nozzle diameter) pulling and pushing the molten polymer through the spinneret [33]. Interestingly, SEM imaging of such pulsed samples revealed numerous nano-fibres with diameters below 100 nm (figure 2(B)). While these were a by-product of fibre pulsing, it is notable that such nano-fibres are possible using MEW. Hence, MEW does not only allow for precise morphology control, but may eventually be performed with much smaller polymer fibres.

Zoom In Zoom Out Reset image size

At the intersection of fibres, the height of the box scaffold is the greatest due to fibre overlaying. Structural breaks in the filaments can result and fibre sagging occurs increasingly with height (figure 2(C)). This could result in an incremental growing bending momentum and mainly tensile load acting on the fibres. Thus, breaks with increasing size were observed with a higher amount of deposited layers near fibre crossings, as is shown in figure 2(D). However such structural defects are influenced by the stored thermal energy of the solidifying fibres during MEW. When molten, these fibres can flow under a mechanical load and break—the ends are capped with 'droplets', circled blue in figure 2(D).

The smallest, highest quality diameter filaments were produced with the following parameters: accelerating voltage = 2.9 kV; collector distance = 1.5 mm; heating temperature = 109 °C (corresponding to a temperature of the spinneret = 84 °C); feeding air pressure = 2.8 bar; spinneret diameter = 33 G; velocity v = 5500 mm min⁻¹. As a result of these dynamic balanced and constant instrument parameters highly defined structures were consistently fabricated. Due to the high precision and almost defect free morphology of the filament structures the centres of the resulting scaffolds are optically transparent, even with the stacking of 50 filaments (figure 3). Surrounding the transparent region is an opaque one—here the acceleration of the stage/fibre is still in progress and fibres are close together. Instead of straight-formed filaments, sinusoidal-formed fibres or lateral shifted piles were deposited within the opaque region. At the edge of the clear region are poorly spaced, yet accurately stacked filaments (figure 3(D)). Figure 3(E) also presents quantitative measurements using SEM for the top diameter fibres. The average upper-side filament diameter was 817 ± 165 nm after 50 layers in each direction (0° and 90°) of stacking, into box scaffolds with distances of 100.6 ± 5.1 μm. Figure 3(F) shows an overview of the uniform structured mesh, while figure 3(B) documents the exact stacking of the layer by layer controlled deposition process and allows a view on the ultrafine filaments. SEM images also show that the filament surface is smooth and homogeneous. The relatively small deviations in filament diameter indicated only low oscillation of dynamic process balance dominated by the electrical field (E = 1.9 kV mm⁻¹). As shown by the SEM measurements, filament sagging occurred at the intersection of fibres since the height of the fibres here increased faster than the mid-point of each filament.

Zoom In Zoom Out Reset image size

A consequence of aligned fibres, light scattering effects dominate uniform interaction between light and material in the middle range of the sample. When looking at the tilted samples, the filaments of the box structures induced the refraction of different wavelengths of the incident light (figure 4(A)), less towards the edges where they are less oriented. However large regions of uniform structures can be made from such sub-micron filaments. The breakage of filaments at larger stacking heights can still be seen in figure 4(C) and is a processing challenge to overcome in future studies with sub-micron filaments.

Zoom In Zoom Out Reset image size

Assuming similar flow effects of molten polymers in melt spinning technologies, electrohydrodynamic jet stretching during solidification should orient polymer chains along the length of the filament [34, 35]. The resulting crystalline regions would then form perpendicular to the filament. This assumption could be confirmed through AFM analysis of single filaments (figures 5(A) and (B)). Due to the slight flow of molten polymer during and after deposition, an elliptic fit was set to analyse the shape. The integral of the filament indicates the cross section in order to compensate the inability of the AFM tip to measure spaces under overhangs. The cross section integral (2a = 1.3 μm/2b = 0.6 μm/A = 0.6 μm⁻²) of the deformed filament is equal to a circular filament of 850 nm in diameter.

Zoom In Zoom Out Reset image size

In agreement with SEM, AFM shows that the as-formed filaments showed a smooth surface with an average surface roughness of R_a = 23 ± 3 nm and a root mean square roughness of R_q = 28 ± 4 nm. In contrast to this, the surface roughness of single filaments changed significantly with R_a = 134 ± 36 nm/R_q = 153 ±31 nm due to annealing at 55 °C. At this temperature, the amorphous region flowed due to gravity, leaving the lamella clearly exposed, revealing information about the polymer solidification conditions and the macromolecular orientation. During the electrically induced acceleration of the jet, partially crystalline regions are forming via flow induced crystallization process [36, 37] and thereby increase the elastic modulus G'. A significant stretching of the electrified jet and anisotropic polymer chain orientation can be assumed during the process. This polymer alignment (figure 5(C)) and the inhibition of relaxation diffusion movement of macromolecules with further cooling and crystallization of the jet leads to an ordered structure of the lamellae.

For DSC measurements (figures 6(A) and (B)), the released specific thermal energy of the MEW structures ω_mesh = 125 J g⁻¹ was compared to the cut bulk PCL pellets. In contrast, the raw PCL material showed additional 30% higher specific melting energy ω_raw = 171 J g⁻¹. This result supports the assumption of quenched PCL filaments with incomplete crystallized and aligned regions combined with the influence of the high specific filament surface. Additionally, the melting onset T_mo and melting peak temperature T_mp of the electrospun material was 60.6, 70.2 °C respectively, and were shifted compared to the raw PCL with 55.5, 62.0 °C respectively. This could be the result of smaller sized crystalline areas enabled by high quenching rates, mainly at the high specific surface of the jet. All samples showed similar thermally properties during the 2nd heating cycle. Further, similarities between the 1st heating cycle and the 2nd heating cycles of the electrospun material were noticed. This shows an analogy of melting point and specific melting heat flow of the MEW processed material and PCL quenched with a rate of −10 K min⁻¹.

Zoom In Zoom Out Reset image size

To investigate an adverse effect of the fibrous morphology with difference in specific surface, additionally DSC measurements with slow heating rates (2 K min⁻¹) were performed. These investigation yielded similar results with T_mo of 56.1, 59.8 °C respectively, and ω_mesh = 148 J g⁻¹, ω_raw = 175 J g⁻¹ respectively, for filaments and raw material. Equally to the melting temperature, the glass transition temperatures T_g of the electrospun material were measured at lowered values: −60.7 °C compared to the raw and aged materials (1st cycle raw PCL: −64.1 °C/2nd cycle raw PCL: −64.6 °C/2nd cycle mesh: −64.6 °C). This could be the result of quenching induced and mechanically stored energy, enabling increased molecular movement at lower temperatures.

XRD revealed a peak shift of the (110)-plains from the ultrafine filaments with 21.43° to the sheet material with 21.37° and from 22.02° to 21.97° (figures 6(C) and (D)). Further, the integral peak area as degree of the crystallinity was about 60% of the fibrous structures compared to the thermally treated analogue. This was most probably due to high quenching of the crystalline structure enabled by the high surface area of the electrified jet. In contrast to the changes of the (110)-plains the (200)-peaks showed similar behaviour in position and integral which further indicates an anisotropic influence of the MEW fabrication method and plastic deformation of the semi-crystalline PCL structure. Since the background intensity profile of both measurements was equivalent, possible geometric reasons for the different peak profiles of filaments and sheet could be excluded. This conclusion was verified by the testing of unaligned samples and the angle depending XRD-measurements of uniaxial samples, which confirmed the anisotropic crystallinity. In contrast to the 0° sample measured along the filaments (I = 1900 cps), the 30° and 60° tilted samples showed lowered (110) peak heights (I = 1200 cps, I = 1500 cps respectively). Additionally, due the heat treatment every sample showed a slight increase in peak height and integral which confirmed residual amorphous phases.

With handling such fine filament scaffolds in vitro, practical considerations such as maintaining structural integrity during washing phases is a key property. While larger filament scaffolds can be immersed directly into a well plate, the ultrafine fibres here are affected by mechanical forces due to changing liquids during culture and staining. We therefore used NCO-sP(EO-stat-PO)-coated microscope slide as collectors for fibre deposition and scaffold production [27, 29, 38] (figures 7(A) and (B)). While detachment was a big problem from untreated surfaces, the box scaffolds predominantly remained mechanically attached to the NCO-sP(EO-stat-PO) surface even for up to 10 DIV including media changes. Cell adhesion was directed to the scaffold rather than attaching to the NCO-sP(EO-stat-PO) surface (figures 7(C)–(F), allowing for proper fluorescence microscopy of cells adherent to the scaffolds only. Filaments remain primarily straight and were also stained blue by Hoechst labelling due to adsorption. The cells were able to bridge across the box pores, and could fill them with time (figure 7(F); supplementary information). We can therefore use such layered ultrafine filament scaffolds for future in vitro experiments and more detailed biochemical analysis without the risk of mixed cellular populations adhering to the scaffold or to the substrate, respectively.

Zoom In Zoom Out Reset image size