**3. Results and Discussion**

Inspired by techniques such as freeze-drying [12] and salt-leaching [13] where ice crystals or salt crystals are used as porogens, it was hypothesized that it should be possible to use a similar approach to modify the macroscale form of fibers collected using conventional electrospinning. This setup, termed "sacrificial core electrospinning", could be used to produce devices for potential cell and tissue encapsulation applications. As presented in Figure 1, the setup was simple and extends conventional systems though inclusion of a sacrificial core (a sodium chloride pellet) rotated directly in front

of a grounded plate. Fibers deposited onto the pellet conform to the shape and size of the pellet. After electrospinning, the sacrificial core pellet can be dissolved by suspending in ultrapure water. Whilst this results in the creation of a hollow capsule with a port for cell or tissue injection, the electrospun fibers do not provide sufficient internal support and the device is susceptible to collapse (Figure S1). To overcome this issue, a custom-made 3D printed supporting frame was developed (Figure S2) and used both as a former for the sacrificial core and to provide internal support once the core has been dissolved (it is worthwhile to mention here that use of 3D printed frame without salt embedding was not efficient - check Figure S3). Using this approach, it was possible to successfully fabricate devices with varying length, width (Figure 2a), shape (Figure 2b) and thickness (Figure 2c).

**Figure 2.** Sacrificial core electrospun products: With sacrificial core-based electrospinning, it was possible to prepare PA66 macro-capsules with variable size (**a**), shape (**b**), and thickness (**c**).

Whilst conventional electrospinning techniques enable control to be gained over the microscale properties of fiber diameter and density, the use of a printed framework and salt core facilities the ability to control the macro-scale features of the fabricated construct. Of the 21 parameters commonly observed to influence fiber formation and the properties of the resultant product, we studied the effects of solvent type, polymer concentration, flow rate, applied voltage, tip-to-collector distance and voltage–to–distance ratio using polyamide 66 (PA66 or nylon 66) (Table 1 and Figure 3) [14]. In accordance with the literature, electrospunPA66 dissolved in formic acid (FA) resulted in thinner fibers (121 ± 12 nm), in comparison to PA66 electrospun in 2,2,2-Trifluoroethanol (TFE, 388 ± 127 nm) or 1,1,1,3,3,3-Hexafluoro-2-propanol (HFIP, 967 ± 127 nm) (Figure 3a). Differences in the evaporation rate of the solvents, as well as the conductivity and viscosity of the resultant solutions may have also contributed to the differences observed in fiber diameter [14]. Similarly, an increase in the PA66 concentration from 5 to 7.5 to 10% (*w*/*v*) resulted in an increase in fiber diameter from 486 ± 159 nm to 967 ± 127 nm to 1908 ± 258 nm respectively (Figure 3b). The viscosity and elastic modulus of the solution increases with polymer concentration and this may be attributable for the changes in the fiber diameter observed [14]. The change in fiber diameter in relation to voltage was found to follow a slight inverse relationship, with fiber diameters of 1113 ± 258 nm, 967 ± 127 nm and 904 ± 247 nm recorded at 10, 15 and 20 kV (Figure 3c). These changes may be attributed to the differences in volumetric charge density associated with the applied voltage, which would significantly affect the initial jet diameter at the tip of the Taylor cone [14]. Whilst the distance (Figure 3d) and the voltage–to–distance ratio (Figure 3e) do not seem to have any significant influence on fiber diameter, we noticed a change in fiber morphology in relation to the voltage–to–distance ratio. Finally, the fiber diameters were found to be 937 ± 189 nm, 967 ± 127 nm, 1165 ± 412 nm and 1292 ± 194 nm, when the flow rates were 0.025, 0.050, 0.075 and 0.100 mL/min respectively (Figure 3f). It has previously been noted that, with other parameters constant, the lowest flow rate typically results in a small jet radius that yields the thinnest

fibers, whereas higher flow rates typically result in a larger jet radius and yield thicker fibers [14]. Our observations were in agreement with earlier reports [14–16].

**Figure 3.** Effect of electrospinning parameters on fiber morphology: scanning electron microscopy (SEM) images of electrospun PA66 samples showing the effects of solvent type (**a**), polymer concentration (**b**), voltage (**c**), distance (**d**), voltage–to–distance ratio (**e**), and flow rate (**f**). Corresponding plots of fiber diameter are presented in (**a'**) to (**f'**) respectively. Fiber width was analyzed by manually measuring the width of individual fibers against a given scale using Image J software (n = 25). # and ‡ denotes statistical differences at *p* < 0.01 and *p* < 0.001 respectively. Scale bar = 2 μm.

The bulk properties of the polymer were compared before and after the electrospinning process. Electricity, a physical property, does not typically affect the material properties of the polymer unless reactive additives are present in the polymer solution. In the current approach, the sodium chloride core was in direct contact with the deposited fibers within the electric field. Attenuated

total reflectance—Fourier transform infrared spectroscopy (ATR-FTIR) revealed similar spectral patterns with characteristic peaks related to N–H stretching at ~3330 cm−1, CH2 stretching at ~2840 cm<sup>−</sup>1, amide I at ~1650 cm−1, amide II at ~1545 cm−1, and amide III at ~1370 cm−<sup>1</sup> [15,17] (Figure 4a). A simultaneous TG/DTA investigation also suggested comparable heat flow patterns with a characteristic peak melting temperature (*T*m) at ~264.5 ◦C for pristine and ~262.5 ◦C for electrospun PA66, and decomposition patterns with the onset decomposition temperature of ~380 ◦C for pristine and ~373 ◦C for electrospun PA66 samples (Figure 4b,c). The spectral peak values were in good agreement with previously reported values for PA66 [16,18]. Collectively, ATR-FTIR and TG/DTA studies confirmed that the sacrificial core electrospinning was similar to the typical electrospinning process and did not influence the bulk properties of the polymer used in this study.

**Figure 4.** Material characteristics before and after sacrificial core electrospinning: Attenuated Total Reflection-Fourier-transform infrared spectroscopy (ATR-FTIR) (**a**), and thermogravimetry /differential thermal analysis (TG/DTA) (**b**—DSC, **c**—TGA) confirmed that the use of sodium chloride within the collector has no significant influence on the bulk properties of the polymer (PA66).

Since the devices fabricated using the sacrificial core electrospinning technique are intended to be used in cell and tissue encapsulation applications, it is of prime importance to verify the free movement of biomolecules, such as cytokines and growth factors, across the capsule walls. To investigate this, fabricated capsules were independently loaded with three model proteins: insulin, albumin and immunoglobulin G (Ig G) (1 mg/mL in PBS), with molecular weights of 5.8, 66 and 150 kDa, respectively, and incubated in protein-free PBS. Following overnight incubation at 37 ◦C, the capsule was removed, and the solution taken from around the capsule was assayed using the Bradford method. The presence of protein was detected in the solutions from all tested capsules (see Figure 5). However, minor differences were observed between the solutions from the 5, 7.5 and 10% (*w*/*v*) PA66 capsules. This may be due to the differences in the total surface area of the fibrous constructs. However, these differences were negligible. The results suggested that, in contrast to conventional hydrogel-based encapsulation systems, where the rate of biomolecule diffusion is restricted, the electrospun membranes produced in this study allow free diffusion of biomolecules. This may be beneficial for cell and tissues encapsulation applications that require rapid biomolecule transport [19].

**Figure 5.** Protein diffusion study: Biomolecules such as insulin (**a**), albumin (**b**), and Ig G (**c**) loaded (1 mg/mL) into the sacrificial core electrospun capsules, were found to diffuse through the membrane into the incubation buffer. \* denotes statistical differences at *p* < 0.05.

Finally, the ability of the electrospun capsule to support cells was assessed *in vitro* using two different model cell types: adherent fibroblasts (from human dermis) and non-adherent lymphocytes (Jurkat cells from human peripheral blood). In each case, 5.0 × <sup>10</sup><sup>5</sup> cells were loaded into the capsule and the cell-laden capsule was incubated in a CO2 incubator. After 3 d of incubation, cell viability was indirectly measured using the metabolic Alamar blue assay. Cells cultured in a well plate under similar conditions were treated as the positive control. The purpose of the study was to investigate whether the porous structure of the capsule posed a barrier to diffusion that would affect the cell viability. Cell viability remained unchanged across all capsules tested (Figure 6). These results suggest that the electrospun membranes posed no limitation to the exchange of gases, nutrients and metabolic wastes essential for cell survival. We also collected the spent medium and performed further Alamar blue assays to determine if the encapsulated cells were released through the porous wall of the capsule. Both the fibroblasts and lymphocytes were not detected in the spent media, indicating that the porosity of the device (Figure S4) was efficient enough to restrict access to cells (Figure 6). The results indicate that the resultant electrospun device could be potentially used in various cell and tissue encapsulation applications. Considering the lymphocyte-based results, the capsule could also be used as an immuno-isolating system; however, detailed experiments are required to prove the immuno-protective properties of the device.

**Figure 6.** Cell encapsulation study: (**a**) Adherent cells (human dermis fibroblasts, HDF) and (**b**) non-adherent cells (lymphocytes from human peripheral blood, Jurkat cells) were encapsulated in the electrospun capsules. The number of cells present at day 3 remained unaffected in comparison to cultures on standard tissue culture-treated polystyrene dishes. Spent media showed no detectable cell activity, suggesting that the capsules effectively compartmentalized the cells. No statistical differences were found within the capsule and spent medium groups.

Previously, Lathuilièreet al., generated flat sheet devices consisting of an outer polymeric frame, porous membranes and a reinforcement mesh assembled using ultrasonic welding [20]. These devices were used for the implantation of genetically engineered allogeneic cells and passive immunization against amyloid-β. Similarly, Nyitrayet al., reported a two-step heat-sealing process to fabricate a polycaprolactone-based thin-film device and used this device to transplant pancreatic islets [21], whilst Park et al., described the assembly of a bacterial cellulose, collagen and alginate-based composite device for encapsulation of neuronal cells [22]. Recently, David et al., showed how a Polytetrafluoroethylene-based bilaminar device (Theracyte) can be used in the encapsulation of ovarian allograft to restore ovarian endocrine function [23]. While these reports suggest that there are a wide range of potential applications for bio-encapsulation technology, they also reveal the unmet need for an efficient macro-encapsulation device and illustrate the complexity of fabrication processes currently being investigated. In contrast, the sacrificial core electrospinning approach presented in this report is relatively simple as it follows well-known electrospinning principles and there is no complicated assembly or other processing steps involved. It is also versatile as it allows the fabrication of a device with a wide range of macro- and micro-scale features. The resultant device can be used as a cell-based drug delivery device or as a cell reservoir to treat conditions such as diabetes, neurological disorders and sensory diseases; however, the true efficiency of the device can only be determined by further *in vivo* studies.
