Mimicking nature: fabrication of 3D anisotropic electrospun polycaprolactone scaffolds for cartilage tissue engineering applications

Abstract There is a growing need to develop strategies capable of engineering the anisotropic cartilaginous fibrous network in vitro and consequently overcome the anatomical and functional restrictions of the standard medical procedures used for cartilage regeneration. In this work, we suggest a fabrication procedure to build 3D anisotropic multi-layered fibrous scaffolds. Polycaprolactone (PCL) was used as bulk material for the different electrospun layers (horizontally, randomly and vertically aligned) that were assembled and then structurally maintained by a biocompatible graphene-oxide-collagen (GO-collagen) microporous network. To validate the resourcefulness of the technique, four PCL-GO-collagen scaffolds with different anisotropic properties were produced and characterized by analysing their depth dependent morphological and mechanical properties.


INTRODUCTION
In the human body, each tissue and organ presents a specific composition and architecture responsible for providing biochemical and biophysical cues that ultimately modulate the cell response towards a precise biological function 1 . The partial or total discontinuity of this synergetic pathway due to factors such as aging, disease and/or injury can lead to severe health problems that may not be solved by neither natural regenerative processes nor via standard clinical procedures.
Thus, from this perspective, the successfully implementation of TE strategies deeply depends on the ability of the engineered scaffolds to mimic the chemistry and geometry of the injured extracellular matrix (ECM) and consequently trigger the cellular mechanisms that collective ly promote regeneration 2 . Biomaterials, especially polymers, are essential components in the biomimetic design of scaffolds since they not only work as bioactive interfaces capable of influencing cell activities like adhesion, proliferation and differentiation, but also offer the possibility to easily engineer composites with advanced 3D architectures 3 . Furthermore, polymers have recently started to be combined with carbon related nanomaterials like graphene and graphene oxide (GO), opening the doors to a new set of composites with enhanced and tunable chemical, electrical, thermal and mechanical properties capable of simulating specific cellular microenvironments 4-5 . Together with the presence of these biocompatib le and biologically active composites, the recent successful inclusion of additive manufacturing technologies like electrospinning and 3D printing in biofabrication strategies 6-7 has allowed to narrow the interconnection between function and structure in the next generation of TE scaffolds 8 . In fact, similarly to tissues in vivo, a proper geometry is crucial to provide a spatially organized 3D fibrous/porous network capable of promoting in vitro cell-scaffold interactions that stimulate the production of new ECM at a ratio compatible with the degradation of the engineered structure 1,8 .
Actually, several studies have already pointed out that the combination of suitable biophysical cues (e.g. chemical, mechanical properties) with 3D geometric features (e.g. porous network and architecture) can lead to a successfully implementation of complementary chemical and morphological gradients 9 , to a regulation of cell migration 10 and to a modulation of cell differentiation 11 . For example, in a recent study, Li et al. 12 have reported some advantages of 3D relatively to 2D silk fibroin/chitosan scaffolds, including the ability of the cells to show differe nt morphologies and enhanced migration and differentiation processes.
One of the major challenges of biomimetic designing TE scaffolds is to recreate in vitro the hierarchical organization of the fibrous collagen network present in the native cartilage, which is an anisotropic tissue responsible for guaranteeing not only an enhanced load-bearing and loadtransferring interface between bones, but also low friction during joint movement 13 . The depthdependent biomechanical properties of articular cartilage are related with the density and organizational arrangement of its ECM components, particularly the collagen fibrils, which are divided into three functionally complementary nanofibrous zones 14-15 . Indeed, the orientation of the collagen fibres goes from perpendicular to the subchondral bone surface in the deepest zone, to random in the middle zone and finally to parallel in the superficial region. These topographic cues are essential to modulate a suitable cell response because they deeply influence both morphology and function of the chondrocytes and consequently the production and maintena nce of the ECM elements. Therefore, the damage of this anisotropic fibrous structure often results in substantial musculoskeletal morbidity since the low proliferative rate and density of chondrocytes together with the avascularity of the cartilaginous tissue do not ensure an appropriate regeneratio n process 14, 16-17 . Alternatively, an enhanced nutrient diffusion in large biocompatible scaffolds can be easily accomplished by 3D printing/bioprinting fabrication 23 due to the possibility to combine the chemical characteristics of bioactive bioinks with accurate mechanical and spatial features [24][25][26] .
Other scaffolding approach involves the development of porous networks that are able to modulate the cell response by adjusting depth dependent architecture gradients related with chemical composition 27 or with pore size 28 and/or orientation 29 . In fact, sequential layers of horizonta lly, randomly and vertically aligned pores can successfully mimic some structural and biomechanica l properties of the superficial, middle and deep zones of the cartilaginous tissue, respectively, leading to an anisotropic scaffold capable of enhancing cellular organization 30

Materials
The PCL with a molecular weight of 80 000 Da, the collagen solution from bovine skin (6 mg mL - The GO aqueous dispersion (4.0 mg mL -1 , water dispersion) was purchased from Graphenea.

PCL-GO-collagen scaffolds fabrication
The and then rolled into spiralled cylinders within a GO-collagen hydrogel, which was fabricated accordingly to our previous work 41 . Briefly, the GO aqueous dispersion was directly mixed with the collagen solution in acidic medium (pH =2) and then rapidly shaken for 10 s in order to form a hydrogel. The weight % of collagen relatively to GO used was 24. The GO-collagen was then centrifuged (Thermo Scientific™ Heraeus™ Multifuge™ X1 Centrifuge) for 10 minutes at 10 000 RPM with the purpose of removing the excess of water. For the middle layer, an ethanol:wa ter (9:1, v:v) bath was used as a collector with a working distance of 10 cm. The electrospun fibres were collected from the bath after a period of 1.5 minutes of electrospinning.
Previously to the assembly, the three electrospun layers were immersed in water for 5 hours at room temperature. After this period, a cylindrical mould (high = 5 mm; diameter = 5 mm) was used for the assembly: the deep layer was placed into the bottom, followed by the middle layer and then by the superficial layer, which was engineered by pilling up cylinders (high = 100 μm; diameter = 5 mm) cut from the original PCL electrospun mesh. The mould was covered to guarantee the contact between the three fibrous layers and placed at -70°C during a period of 5 hours. The frozen 3D electrospun structure was then removed from the mould and rapidly immersed into the GO-collagen hydrogel placed in a second cylindrical Teflon mould (high = 5 mm; diameter = 10 mm) in order to originate a heterogeneous microporous network after a freezedrying process at -70°C (Telstar lyoQuest HT-40, Beijer Electronics Products AB) capable of surrounding the fibres and consequently ensuring the correct placement of the assembled layers.
The final PCL-GO-collagen scaffold presented a tri-layered structure (geometry 1) and its fabrication is summarized in Figure 1a. Additionally, this methodology was used to fabricate PCL-GO-collagen scaffolds with alternative fibrous geometric architectures by varying the sequence of the different layers during the assembly process ( Figure 1b): geometry 2 (superficial + middle layers), geometry 3 (superficial + deep layers) and geometry 4 (middle + deep layers). For these bi-layered PCL-GO-collagen scaffolds, each layer was 2.5 mm high.

Physicochemical characterization
Chemical analysis of the materials used were performed via attenuated total reflectance Fourier transform infrared (ATR-FTIR) in a Bruker Tensor 27 FT-IR spectrometer (Bruker Corporation).
This technique was also used to confirm the success of the self-assembly process of the GOcollagen hydrogel. The spectra were recorded between 4000 and 400 cm -1 , with a resolution of 4 cm -1 and 256 scans.
The swelling properties of the PCL-GO-collagen scaffolds were determined by immers ing the samples into distilled water for 24h at room temperature. The swelling ratio was calculated using 5 specimens of each architecture with the following equation: R = (Ws-Wd)/Wd, where R is the swelling ratio (mg/mg), Ws is the weight of the swollen scaffold and Wd is the weight of the dried scaffold.

Morphological characterization
The topography features of the three electrospun layers were evaluated individually using a scanning electron microscope (SEM) Hitachi SU 70 (Hitachi High-Technologies) after transversal cuts were made in the PCL-GO-collagen scaffold with the geometry 1. The dimensions of both fibres and pores were evaluated by direct analysis of ten SEM pictures for each sample. These analysis were only made in the PCL-GO-collagen scaffold with the geometry 1 since similar layers (only with different heights) were used for the fabrication of the other geometries.

Mechanical characterization
The compression properties of the PCL-GO-collagen scaffolds (all geometries) and the three electrospun layers (superficial, middle and deep) were tested by compressing the samples using a Shimadzu MMT-101 N (Shimadzu Scientific Instruments) with a load cell of 100 N. The samples were compressed at a rate of 1 mm min -1 after a pre-charge of 0.07 N and their compressive moduli were calculated by analysing the slope of the stress-strain curves at different strain values (5%, 10%, 15% and 20%). The ratio between compressive moduli at consecutive strains were used as a measure of densification 30 . For each scaffold and layer, 5 specimens were analysed.
The structural integrity of the PCL-GO-collagen scaffolds were studied by submitting one sample of each geometry to a fatigue test inside a bioreactor during 360 hours at 37 °C and 5% CO2. The samples were immersed in a DME/F-12 + 1% Pen-Strep + 10% FBS medium and submitted to a sinusoidal compression of 10% strain at 0.5 Hz during intervals of 3 h followed by 9 h of rest. The medium was exchanged tree times per week. The ∆stress was calculated with the equation: ∆stress (Pa) = (σmax-σmin), where σmax and σmin are correspondingly the maximum and minimum stress levels measured during each hour.

3.1.Chemical analysis
PCL, GO and collagen were the biomaterials selected for the fabrication of the anisotropic scaffolds since they are very popular in tissue engineering applications due mostly to the manipulation/stability properties 42 , functionalization potential 5 and biocompatibility 43 , respectively. The ATR-FTIR spectra of these materials are showed in Figure 2, where, as expected, the PCL spectrum (blue) presents its standard absorbance peaks 44  Indeed, it is possible to observe the O-H related peak centred at 3425 cm -1 in both GO and GOcollagen spectra. The other GO oxygen moieties can be identified in the peaks located at 1724 cm -1 (C=O), 1220 cm -1 (C-O epoxy) and 1047 cm -1 (C-O alkoxy). In the GO spectrum it is also possible to identify the C=C stretching vibrations at 1623 cm -1 . Relatively to collagen, the amide I and III displayed bands can be observed at 1627 cm -1 and 1232 cm -1 . Additionally, it is also noticeable the characteristic peak of the amide II in both collagen (1540 cm -1 ) and GO-collagen (at 1535 cm -1 ) spectra.

Morphological analysis
As it is shown in Figure 1, it was possible to assemble different fibrous 3D geometries by just varying the combination of the electrospun layers during the assembly process when a mould was were rolled up and implanted either with or without cells with the purpose of correcting bone defects in vivo [53][54] . For the fabrication of this layer, it was exploited the same electrospun mesh used for the superficial layer since its pores can work as supplementary channels for nutrient diffusion across the spiral branches 55 , which would be otherwise only ensured by the longitud ina l spaces between them. Actually, in this work, as the rolling process was carried out within a GOcollagen hydrogel, it was possible to guarantee not only the verticality of the electrospun fibres, but also that the same microporous network that surrounds the 3D electrospun system intercalates each fibrous spiral branches after lyophilisation ( Figure 3d). Thus, instead of empty spaces, the biocompatible GO-collagen porous arrangement may supply functional bridges between the fibrous walls capable of supporting cells and the newly synthetized ECM. Additionally, the GOcollagen network avoids the compact accumulation of fibres, offering an alternative strategy to other methodologies that involve, for example, the inclusion of microparticles into the electrospun mesh to safeguard an appropriate distance between the fibrous spiral branches 56 .

Mechanical analysis
Taking into account the average deformational behaviour of articular cartilage in vivo 14, 57 , the mechanical responses under compressive loading of the three electrospun layers used for the fabrication of the different PCL-GO-collagen geometries were studied at low strains (ε ≤ 20%).
As it is possible to see in Figure 4a, the three stress-strain curves of the electrospun layers showed linear regimes, indicating that, in all cases, most of the fibres are straight during compression and therefore the following typical events of both fibril 58 and porous 30 systems (buckling, collapsing and densification of the network) were not substantial for the compression regime applied. As expected, the deep layer proved to be significantly more resistant to compression relatively to the superficial and middle layers due to its vertical fibres (Figure 4b). Indeed, the compressive moduli changed with fibre alignment, decreasing from vertical to random to horizontal. This behaviour matches other reported fibrous and porous anisotropic scaffolds, which resistance to deformatio n is smaller when the direction of the force is perpendicular to the orientation of the fibres/pores.
Relatively to densification of the fibrous networks, it is possible to confirm the interpretation of the stress-strain curves since the values of the compressive moduli ratios are close to 1 (Figure 4c), indicating no considerable changes in the layers architecture. However, it is worthy mention that the ratio of compressive moduli is constantly superior in the superficial layer because of the higher probability of fibre-fibre contact due to its small pores. Thus, it is possible that for higher strains (ε > 20%) this layer will present more noticeable increasing of resistance during compression relatively to the middle and deep layers, where the larger pore size and the verticality of the fibers could better prevent the contact between fibers. The compressive properties of the electrospun layers are summarized in Table 1.
The combination of these electrospun layers towards the design and fabrication of 3D anisotropic fibrous architectures gave rise to multi-zonal PCL-GO-collagen scaffolds with compressive properties extremely dependent on the geometry used. Predictably, likewise the individual electrospun layers, all the PCL-GO-collagen scaffolds presented linear regimes for the strain interval applied (Figure 5a), indicating the maintenance of their structural and mechanica l integrity. The effect of anisotropy is evident by analyzing the compressive moduli in Figure 5b since the values for each scaffold lay between those calculated from the electrospun layers that shape each characteristic geometry. This is particularly noticeable regarding the bi-layered scaffolds because the lowest compressive moduli belongs to the geometry 2, which includes the superficial and middle layers into its fibrous architecture and, on the other hand, the mechanica l response of the PCL-GO-collagen scaffold with geometry 4 is the most prominent due to the combination of the two electrospun layers that better resist compression (middle and deep).
Although the influence of the deep layer is also notorious in the compressive response of the scaffold with geometry 3, the horizontal alignment of the fibers in its superficial layer does not provide an efficient resistance to compression. Thus, the PCL-GO-collagen scaffold with this geometry presents an intermediate behavior relatively to geometries 2 and 4. In its turn, the compressive moduli of the tri-layered PCL-GO-collagen scaffold (geometry 1) showed similar values comparatively to the bi-layered PCL-GO-collagen scaffold with geometry 2. This analogous response occurs because the superficial and middle layers represent 60% of the total high of the geometry 1, leading to a more noticeable resistance to compression of these two electrospun layers for low strains. Indeed, the PCL-GO-collagen scaffold with geometry 1 presented ratios of compressive moduli (Figure 5c) that indicate no considerable densification of the 3D structure and therefore no substantial elastic buckling of the vertically aligned fibers of its deep layer. In the same way, the PCL-GO-collagen scaffolds with the other geometries showed densification levels that coincide with their stress-strain curves, meaning no changes in the scaffold architecture induced by compression. All the mechanical testing was performed with swollen samples since water uptake capacity is not only an indispensable requirement to boost cartilage regeneration 39 , but also a crucial feature to mimic the physiology of the cartilagino us microenvironment during compression 14 . In fact, besides the physical supporting of the 3D anisotropic structures, an additional advantage of the inclusion of the GO-collagen porous network is to considerably improve the absorbing and retention of water of the scaffolds. In this way, from a functional point a view, it will be possible to both compensate the hydrophobicity and the nonswelling features of PCL, moreover, the swelling ability of the microporous structure guarantees that the PCL-GO-collagen scaffolds return to their original shape after compression. As it is possible to see from Figure 5d, the swelling ratios of the anisotropic PCL-GO-collagen scaffolds showed comparable values independently of the characteristic geometry used. This is probably due to the PCL nuclei of each scaffold present the same dimensions (5mm of diameter and 5 mm high) and, therefore, a similar proportion of GO-collagen porous network surround them.
However, it is possible to observe that the PCL-GO-collagen scaffold with a slightly higher swelling ratio is the one that presents the geometry 4, which it is in accordance with the combination of the larger porous network of the middle layer with a PCL fibrous spiral with branches intercalated with the GO-collagen porous network (deep layer). The compressive and swelling properties of the PCL-GO-collagen scaffold are summarized in Table 2.
There is a wide range of reported Young modulus for the natural cartilage (from 0.1 to 5 MPa) due to not only the complex anatomy and physiology of this anisotropic tissue, but also to its origin and experiment conditions 14, 45,59 . Taking this into account and considering the value 89.5 ± 48.6 kPa 34 to be the compressive modulus of natural cartilage, the here presented PCL-GO-collagen scaffolds showed a huge potential for cartilage TE applications since their compressive moduli matched this interval, independently of the geometry used. Indeed, these scaffolds offer the opportunity to study how complementary anisotropic 3D fibrous features and, consequently, alternative mechanical properties can affect the cell response towards cartilage regeneration. Specifically, in theory, the different geometries will be able to induce singular depth dependent cellular densities, orientations and morphologies that could lead to dissimilar localized ECM production elements 31,[34][35][60][61] and subsequently to alternative cellular platforms that might be selected depending on the specific requirements of the followed TE strategy. Additionally, the inclusion of in vitro mechanical stimulation provided by bioreactors in such TE approaches is a growing modality since cells can sense and transform external mechanical stimuli into biochemica l signalling pathways that can decisively impact cellular events such as ECM production and cell differentiation [62][63][64] . In this regard, the viability of PCL-GO-collagen scaffolds to integrate dynamic mechanical stimulation protocols was evaluated within a patented bioreactor apparatus 65 ( Figure   6a) capable of providing a controlled uniaxial mechanical loading strain and frequency. As it is possible to see from Figure 6b, the scaffolds were able to maintain their mechanical integrity after 360 hours of testing, as it is revealed by the constant value of ∆stress measured for the intercalated stimulation periods (8.21 ± 0.50 kPa). This result, although preliminary, is an excellent indicator for the ability of these scaffolds to positively respond to induced mechanical loading, leading to a wider range of promising cell culture conditions that can be combined towards an optimized cartilage regeneration.

CONCLUSION
In this study, we developed a viable methodology to fabricate functional 3D anisotropic electrospun scaffolds with customized fibrous geometries for cartilage TE applications. Indeed, by using two complementary electrospinning set ups, it was possible to engineer layers with fiber orientation and mechanical properties that were similar to their cartilaginous counterparts. Such layers were then assembled into distinct 3D scaffolds with alternative depth dependent topographic and mechanical properties, depending on the geometry applied. Therefore, the versatility of the     layer; Vassembling and freezing the layers in order to fabricate a three-layered electrospun scaffold; VIincorporating the anisotropic 3D electrospun structure within the GO-collagen hydrogel with the purpose of creating a microporous network able to support and involve the electrospun structure after freeze-drying; b) alternative geometric architectures used during the assembly process: from the right to leftgeometry 2 (superficial + middle layers); geometry 3 (superficial + deep layers) and geometry 4 (middle + deep layers).