3D Electrospun scaffolds promote a cytotrophic phenotype of cultured primary astrocytes

Authors

  • Chew L. Lau,

    1. Florey Institute of Neuroscience and Mental Health, The University of Melbourne, Parkville, Australia
    Search for more papers by this author
    • These authors contributed equally to this work.
  • Michelle Kovacevic,

    1. Florey Institute of Neuroscience and Mental Health, The University of Melbourne, Parkville, Australia
    Search for more papers by this author
    • These authors contributed equally to this work.
  • Tine S. Tingleff,

    1. Florey Institute of Neuroscience and Mental Health, The University of Melbourne, Parkville, Australia
    2. Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark
    Search for more papers by this author
  • John S. Forsythe,

    1. Department of Materials Engineering, Monash University, Clayton, Australia
    Search for more papers by this author
  • Holly S. Cate,

    1. Centre for Neuroscience Research, Department of Anatomy and Neuroscience, The University of Melbourne, Parkville, Australia
    Search for more papers by this author
  • Daniel Merlo,

    1. Florey Institute of Neuroscience and Mental Health, The University of Melbourne, Parkville, Australia
    Search for more papers by this author
  • Cecilia Cederfur,

    1. Florey Institute of Neuroscience and Mental Health, The University of Melbourne, Parkville, Australia
    Search for more papers by this author
  • Francesca L. Maclean,

    1. Research School of Engineering, The Australian National University, Canberra, Australia
    Search for more papers by this author
  • Clare L. Parish,

    1. Florey Institute of Neuroscience and Mental Health, The University of Melbourne, Parkville, Australia
    Search for more papers by this author
  • Malcolm K. Horne,

    1. Florey Institute of Neuroscience and Mental Health, The University of Melbourne, Parkville, Australia
    Search for more papers by this author
  • David R. Nisbet,

    1. Department of Materials Engineering, Monash University, Clayton, Australia
    2. Research School of Engineering, The Australian National University, Canberra, Australia
    Search for more papers by this author
    • These authors contributed equally to this work.
  • Philip M. Beart

    Corresponding author
    1. Florey Institute of Neuroscience and Mental Health, The University of Melbourne, Parkville, Australia
    • Address correspondence and reprint requests to Philip M. Beart, Florey Institute of Neuroscience and Mental Health, The University of Melbourne, Parkville, Victoria 3010, Australia. E-mail: philip.beart@florey.edu.au

    Search for more papers by this author
    • These authors contributed equally to this work.

Abstract

Astrocytes are a target for regenerative neurobiology because in brain injury their phenotype arbitrates brain integrity, neuronal death and subsequent repair and reconstruction. We explored the ability of 3D scaffolds to direct astrocytes into phenotypes with the potential to support neuronal survival. Poly-ε-caprolactone scaffolds were electrospun with random and aligned fibre orientations on which murine astrocytes were sub-cultured and analysed at 4 and 12 DIV. Astrocytes survived, proliferated and migrated into scaffolds adopting 3D morphologies, mimicking in vivo stellated phenotypes. Cells on random poly-ε-caprolactone scaffolds grew as circular colonies extending processes deep within sub-micron fibres, whereas astrocytes on aligned scaffolds exhibited rectangular colonies with processes following not only the direction of fibre alignment but also penetrating the scaffold. Cell viability was maintained over 12 DIV, and cytochemistry for F-/G-actin showed fewer stress fibres on bioscaffolds relative to 2D astrocytes. Reduced cytoskeletal stress was confirmed by the decreased expression of glial fibrillary acidic protein. PCR demonstrated up-regulation of genes (excitatory amino acid transporter 2, brain-derived neurotrophic factor and anti-oxidant) reflecting healthy biologies of mature astrocytes in our extended culture protocol. This study illustrates the therapeutic potential of bioengineering strategies using 3D electrospun scaffolds which direct astrocytes into phenotypes supporting brain repair.

image

Astrocytes exist in phenotypes with pro-survival and destructive components, and their biology can be modulated by changing phenotype. Our findings demonstrate murine astrocytes adopt a healthy phenotype when cultured in 3D. Astrocytes proliferate and extend into poly-ε-caprolactone scaffolds displaying 3D stellated morphologies with reduced GFAP expression and actin stress fibres, plus a cytotrophic gene profile. Bioengineered 3D scaffolds have potential to direct inflammation to aid regenerative neurobiology.

Abbreviations used
ACTA2

skeletal muscle actin α2

BDNF

brain-derived neurotrophic factor

CXCL12

chemokine (C-X-C motif) ligand 12

EAATs

excitatory amino acid transporters

ECM

extracellular matrix

GFAP

glial fibrillary acidic protein

GSTA1

glutathione S-transferase α1

HMOX1

heme oxygenase (decycling) 1

PCL

poly-ε-caprolactone

TAGLN

transgelin

Repair of the injured CNS requires reconstitution of neural networks through axon extension and synapse formation. Synapses in the CNS are in close apposition with astrocytes, which are recognised as plastic cells playing key roles in maintaining brain function via energetics, anti-oxidant activity, trophic factor synthesis, neurovascular coupling and l-glutamate (Glu) homeostasis (Ridet et al. 1997; Maragakis and Rothstein 2006; Parpura et al. 2012). Astrocytic morphological change with plasticity would not be unexpected in a cell with a large cell body, an extensive and diverse arbour, tight junctions with other astrocytes and end-feet on the blood–brain barrier and vasculature (McMillian et al. 1994; Ridet et al. 1997; Panickar and Norenberg 2005). Not surprisingly astrocytes exhibit various morphological and biochemical changes in response to changes in their milieu induced by physiological and pathological events. Such changes occur across a continuum of events, being influenced by the prevailing extracellular milieu in disease and trauma and by the extent of multiple factors, which appear able to cause both short- and long-term responses (Ridet et al. 1997; Maragakis and Rothstein 2006). Whilst the terms astrogliosis and reactive gliosis can be found in the literature, astrocytes are now considered to exist in a variety of phenotypes that have pro-survival (‘cytotrophic’) and destructive (‘cytotoxic’) components (McMillian et al. 1994; Panickar and Norenberg 2005; Sofroniew and Vinters 2010). These phenotypes have been found in both the normal brain and brains affected by neurological diseases and there is recent evidence that astrocytes can dynamically change phenotypic components, an attractive target to promote endogenous repair if they can be directed into a cytotrophic phenotype supportive of neuronal survival and axon regrowth. Thus changing the physical and molecular phenotype of astrocytes may facilitate neural repair mechanisms (Maragakis and Rothstein 2006; Sofroniew and Vinters 2010).

Work from our laboratory, initially focusing on excitatory amino acid transporters (EAATs), documented in conventional cultured astrocytes that morphological changes in vitro induce alterations in biology and gene expression (Zagami et al. 2009; Lau et al. 2010, 2011, 2012; Sheean et al. 2013). The actin cytoskeleton and Rho GTPases (Rho, Rac, Cdc42) are fundamental determinants of cellular motility and migration (Le Clainche and Carlier 2008; Mattila and Lappalainen 2008). We found treatment of astrocytes with inhibitors of Rho kinase produced stellated morphology, less actin stress fibres and a shift in the F-/G-actin ratio to a predominance of G-actin (Lau et al. 2011). Here analyses of the astrocytic transcriptome confirmed major alterations to genes of the extracellular matrix (ECM), with elevated expression of EAAT2, brain-derived neurotrophic factor (BDNF) and key anti-oxidant genes (Lau et al. 2012) – findings suggestive of shift to a cytotrophic phenotype. Tissue engineering is another strategy where the use of scaffolds provides cues for cellular organisation, survival and function (Stevens et al. 2005; Teo et al. 2006) and in concert with materials science to manipulate surface chemistry, fibre alignment, diameter and inter-fibre spacing can morphologically replicate components of the ECM (Stevens et al. 2005; Teo et al. 2006). We found that the implantation of 3D electrospun poly-ε-caprolactone (PCL) scaffolds produced a delayed astrocytic response in the striatum in a rat model of traumatic brain injury (Nisbet et al. 2009). After 60 days, astrocyte numbers had returned to normal levels, but more importantly the PCL scaffold failed to elicit a prolonged foreign body reaction. Interestingly, neurites infiltrated into the randomly aligned scaffold during the peak in astrocytes activation, suggesting the infiltration was being promoted by astrocytes. This result led to our hypothesis that electrospun scaffolds may promote cytotrophic astrocytic phenotype. Here we have taken this postulate to a comparison in vitro of astrocytes cultured on electrospun 3D scaffolds and in conventional 2D mode. Investigations of astrocytic morphology and distribution on 3D electrospun scaffolds, as well as their ability to support cell survival, proliferation and functional outcomes of growth, demonstrated cellular penetration of astrocytic processes into the scaffolds and promotion of a cytotrophic phenotype; two key outcomes likely supportive of brain repair. Our evidence demonstrates that bioengineering astrocytes has great potential for regenerative neurobiology. A preliminary account of these findings was presented at the 43rd Meeting of the American Society for Neurochemistry (Beart et al. 2012).

Materials and methods

All experimentation was approved by the Ethics Committee of the Florey Institute for Neuroscience and Mental Health and was undertaken according to the guidelines of the National Health and Medical Research Council (NHMRC, Australia).

Preparation of poly-ε-caprolactone scaffolds

Poly-ε-caprolactone (PCL) was obtained from Sigma-Aldrich (molecular weight 70 000–90 000; St Louis, MO, USA). 2D PCL films were fabricated using a compression moulder heated to 80°C. The PCL was held between two polished stainless steel plates covered with Teflon release film and allowed to reach temperature for 2 min. A 20 kPa force was then applied to the plates for 5 min to fabricate the films before they were removed and quenched in an ice-water slurry. 3D electrospun scaffolds were fabricated as described previously (Nisbet et al. 2008) with minor modifications. Polymer solutions of 10% (w/v) were prepared for electrospinning by dissolving PCL in 2 mL of chloroform (Merck Pty Ltd., Kilsyth, Australia) and methanol (Merck Pty Ltd.) at a ratio of 3 : 1 (v/v). The solutions were placed into a glass syringe (10 mL) with a 21-gauge needle for electrospinning at a flow rate of 1.5 mL/h. A 15 kV accelerating voltage was used at the positive electrode, with a ground rotating mandrel (diameter = 5 cm) employed as the collector. The working distance was 18 cm and the collector was coated with aluminium foil for easy removal. Speeds of 200 rpm (orbit diameter 14 cm) were used to fabricate randomly oriented fibres and 4000 rpm (orbit diameter 14 cm) to fabricate aligned fibres. Once removed from the collector, PCL scaffolds where then cut into circles using a punch with diameters matching the dimensions of the wells (96- and 24-well plates). The scaffolds were sterilised in 70% ethanol for 15 min and washed with sterilised phosphate buffered saline pH 7.4 prior to use.

Scaffold characterisation

The morphology of the scaffolds was characterised using scanning electron microscopy (SEM). The samples were sputter coated with platinum at 20 mA for 1 min. All SEM images were taken under 3 kV with a working distance of 3.5 mm on a Zeiss UltraPlus FESEM (Oberkochen, Germany). The average diameters of the fibres were determined using ImageJ software (http://rsb.info.nih.gov.ij/, National Institutes of Health, version 1.46r) to measure a total of 20 fibres across four different samples and were 400 ± 110 and 450 ± 150 nm, for random and aligned scaffolds, respectively (Fig. 1).

Figure 1.

SEM images of a (a) random and (b) aligned electrospun poly-ε-caprolactone (PCL) scaffolds, with a mean diameter of 400 ± 110 nm and 450 ± 150 nm respectively. Scale bar = 1 μm.

Cell culture

All media used for primary cell culture and maintenance were from Gibco® Life Technologies (Mulgrave, Australia) unless otherwise stated. Primary cultures of astrocytes were established from the forebrain of post-natal day 1.5 C57 Black 6 mice (animal facilities of Florey Institute for Neuroscience and Mental Health) as previously described (O'Shea et al. 2006). After astrocytes had formed a confluent layer (10 DIV), flasks were shaken overnight (to remove cells other than astrocytes) in a Ratek Orbital Mixer Incubator (180 rpm, orbit diameter 25 mm, 37°C). Sub-culturing generated secondary cultures where astrocytes were seeded on to 24-well plates (polystyrene plate, glass coverslips or 2D PCL at 2 × 104 cells/well) or 96-well plates (polystyrene plate, random or aligned 3D PCL at 8 × 103 cells/well) depending on the experiment, and incubated in a humidified incubator at 37°C with 5% CO2. Scaffolds were held at the bottom of the well by sterile glass inserts. Preliminary experiments explored the optimal cell density, examining seeding densities of 4, 8 and 16 × 103 cells/well (Figure S1). A full medium change was performed to remove non-adherent cells and medium was subsequently changed every 3–4 days until cells were ready for use (4 DIV or 12 DIV following sub-culturing). Immunocytochemistry with glial fibrillary acidic protein (GFAP) revealed the presence of a monolayer of astrocytes (Apricò et al. 2004; O'Shea et al. 2006). Note: 4 DIV and 12 DIV following sub-culturing is equivalent to 14 DIV and 22 DIV but for simplicity will be referred to in this paper as 4 DIV and 12 DIV.

Immunocytochemistry

The immunocytochemical procedures have been described previously (Lau et al. 2011). Astrocytes grown on glass coverslips, 2D PCL films, aligned or random scaffolds were fixed in 4% paraformaldehyde in phosphate buffered saline for 10 min. Cells were incubated with primary antibodies GFAP (1 : 1000; Chemicon International Inc., Melbourne, Australia) at 4°C overnight, followed by secondary antibody (Alexa Fluor® 488; Molecular Probes® Life Technologies, Mulgrave, Australia) incubation for 3 h at ~23°C. Coverslips and scaffolds were mounted on glass microscope slides with DAKO fluorescence mounting medium (DAKO, Victoria, Australia), left to dry in the dark overnight and stored at 4°C. Fluorescence was visualised under an Olympus IX71 inverted microscope (Olympus Australia Pty Ltd, Melbourne, Japan). Digital images were acquired using an Olympus Camedia C-5050 Zoom digital camera attached to the Olympus IX71 inverted microscope.

Cytoskeletal staining

As described previously (Lau et al. 2011), concurrent labelling of F- and G-actin was obtained by staining with TRITC-conjugated phalloidin (1 : 1000; Sigma-Aldrich, Melbourne, Australia) and Alexa Fluor® 488-conjugated deoxyribonuclease I (DNaseI, 1 : 250; Molecular Probes® Life Technologies), respectively.

Confocal imaging

Astrocytes were prepared on scaffolds and immunoreacted with GFAP (see above). Nuclei were stained using Hoechst 33342 (5 μg/mL; Molecular Probes® Life Technologies) and the fluorescence was visualised under Leica SP8 Confocal Microscope (Leica Microsystems, Wetzlar, Germany). The scaffolds were imaged using reflection of a HeNe 633 nm laser. Z-stacks were obtained with optical sectioning at 0.2 μm.

Cell viability and function assays

Cellular viability was assessed using a MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) reduction assay according to the manufacturer's instructions (Lau et al. 2011). In addition a lactate dehydrogenase (LDH) assay was carried out using a Cytotoxicity Detection Kit (Roche, Sydney, Australia) according to the manufacturer's instructions (Lau et al. 2011).

l-Glutamate uptake: Activity of astrocytic EAATs was determined using [3H]-d-aspartate ([3H]-d-Asp) uptake (Apricò et al. 2004). In brief, cells were pre-incubated at 37°C for 5 min and then incubated with [3H]-d-Asp (50 nM, 5 min), with or without the unlabelled d-Asp (1 mM) in uptake buffer (135 mM NaCl, 5 mM KCl, 0.6 mM MgSO4, 1 mM CaCl2, 6 mM d-glucose, 10 mM HEPES, pH 7.5). Uptake was terminated by washing at 4°C, and accumulated radioactivity determined by scintillation spectrometry (O'Shea et al. 2006).

Western blot analysis

As described previously (Lau et al. 2010), cells were grown in polystyrene micro-well plates, with or without aligned and random PCL scaffolds for 12 DIV. Samples were pooled from eight wells (= 8 replicates), and total cell protein concentration was determined with the Bio-Rad Dc Assay Kit (Sydney, Australia) according to the manufacturer's instructions. Standard western blot protocols were carried out with 10 μg protein per lane (three lanes per condition, where each lane represents an independent experiment) and membranes were incubated with primary GFAP antibodies at 1 : 1000 (Promega, Melbourne, Australia) overnight at 4°C. Following washing, membranes were incubated with horseradish-peroxidase conjugated secondary antibodies (goat anti-rabbit IgG, 1 : 1000) for 3 h at ~23°C. Proteins were then visualised using enhanced chemiluminescence. As a control for protein loading, blots were subsequently probed for β-actin (primary antibody 1 : 10 000) using the same procedures. Densitometric analysis of western blots was performed using ImageJ software (http://rsb.info.nih.gov.ij/, National Institutes of Health, version 1.46r) to measure the area and density of proteins bands after subtracting the background of the autoradiographic film (Cimarosti et al. 2005).

Quantitative real-time PCR

RNA was extracted from astrocytes grown on polystyrene micro-well plates, 2D PCL, aligned and random electrospun scaffolds. Total RNA was extracted from four independent cultures with eight replicates per culture using the RNeasy Mini Kit (Qiagen, Alameda, CA, USA) according to the manufacturer's instructions. In order to quantify the RNA extract, 1.8 μL of RNA sample was used for spectrophotometric analysis using NanoDrop ND-1000 (NanaDrop Technologies, Inc., Wilmington, DE, USA). qRT-PCR was performed on an ViiA 7 sequence detection system (Applied Biosystems, Foster City, CA, USA) as previously described (Binder et al. 2008). Primers were designed using Primer Express 3.0 (Applied Biosystems, Melbourne, Australia). Sequences of primers and details of all procedures were published previously (Lau et al. 2012). To determine statistical significance for qRT-PCR data, 95% confidence intervals were constructed for single-variable analysis. All data are presented as mean ± SEM, and are expressed relative to the control condition (polystyrene or 2D PCL).

Data analysis

Specific details are given above. All values are mean ± SEM from replicate determinations from multiple independent experiments. Data were subjected to one-way anova followed by Dunnett's Multiple Comparison post-hoc Test using GraphPad Prism v.4.0 (GraphPad, San Diego, CA, USA).

Results

Dramatic change in astrocytic morphology on 3D PCL

Initial work explored optimal culture conditions by investigating GFAP immunolabelling and observations based upon length of GFAP-positive astrocytic processes and indicated a density of 8 × 103 cells/well yielded viable cultures (Figure S1). Under these conditions, astrocytes grown under control conditions on coverslips were evenly distributed at 4 DIV and appeared as flattened cells (Fig. 2a), with increased cell number at 12 DIV (Fig. 2b). When cultured on random scaffolds, astrocytes appeared as tight colonies (ratio of larger to smaller axis < 2) at 4 DIV (Fig. 2), and these colonies expanded to occupy the scaffold with condensed processes (typically 50–75 μm) by 12 DIV. By comparison, astrocytes grown on aligned scaffolds grew in colonies that were more rectangular (ratio of larger to smaller axis ≥ 2) (Fig. 2e at 4 DIV) than those growing on random scaffolds (Fig. 2c), with processes aligning with the fibres and by 12 DIV extending for lengths sometimes exceeding 300 μm (Fig. 2f). Morphological estimates of the length of GFAP-positive processes at 12 DIV were 65 ± 3 μm and 90 ± 3 μm (≥ 15 images, three independent experiments) for random and aligned scaffolds, respectively.

Figure 2.

Temporal changes in astrocytic morphology on different matrices. Primary astrocytes were sub-cultured on coverslips (a and b), random poly-ε-caprolactone (PCL) (c and d) and aligned PCL scaffolds (e and f) for 4 DIV (a, c and e) or 12 DIV (b, d and f) and stained for glial fibrillary acidic protein (GFAP) expression. Scale bar = 50 μm.

As actin is central to cellular morphology and plays a key role in regulating cellular responses to the extracellular matrix (Provenzano and Keely 2011), its arrangement on the 3D scaffolds was studied by staining astrocytes for its two forms, filamentous, F-actin, and globular, G-actin (Fig. 3). The intensity of G- and F- actin staining on 3D scaffolds was higher than on coverslips (2D). Furthermore, by 12 DIV F-actin staining in 2D was uniformly cobblestoned reflecting astrocyte morphology (Fig. 3a and c), whereas G-actin staining was barely detectable at both 4 DIV and 12 DIV (Fig. 3b and d), but was consistently expressed on both 3D scaffolds at both time points (Fig. 3f, h, j and l). The organisation of F- and G-actin stained colonies on random and aligned scaffolds reflected the pattern discerned by GFAP immunohistochemistry (Fig. 2).

Figure 3.

Effects of days in culture and poly-ε-caprolactone (PCL) scaffolds on F-/G-actin expression. Primary astrocytes were sub-cultured on coverslips (a–d), random PCL (e–h) and aligned PCL scaffolds (i–l) for 4 DIV (a, b, e, f, i and j) and 12 DIV (c, d, g, h, l and m) and stained for F-actin (TRITC-phalloidin; red) or G-actin (Alexa Fluor® 488-conjugated DNaseI; green). Scale bar = 100 μm.

Astrocytes on 3D scaffolds maintain viability but show decreased total protein and GFAP levels at 12 DIV

Prolonged culturing of astrocytes under control conditions and on aligned PCL scaffolds resulted in a significant reduction in mitochondrial function (Fig. 4a). Reduction in mitochondrial function, as shown by MTT activities, was similar at 12 DIV regardless of the culture conditions. In addition, nuclear condensation and fragmentation were not observed at 12 DIV, consistent with the absence of pyknotic nuclei (Figure S2). LDH, an extracellular measure of compromised cell membrane, was significantly higher in astrocytes grown on polystyrene (control) and aligned scaffolds at 4 DIV than in those cultured until 12 DIV (p < 0.05). Surprisingly, the LDH of cells grown on aligned scaffolds was higher than those grown on polystyrene (control, p < 0.05) (Fig. 4b). Total cellular protein increased between 4 DIV and 12 DIV, regardless of the surface on which cells were cultured and was taken as an indication of astrocytic proliferation within the 8 days (Fig. 4c). However, cellular protein concentration of astrocytes grown on the 3D scaffolds was significantly lower (~ 30–40% decreases at 12 DIV) than those grown on polystyrene (random p < 0.001 and aligned p < 0.05) at 12 DIV. [3H]-d-Asp uptake activity, which is an index of Glu transport and thus reflects astrocytic function (Beart and O'Shea 2007), was similar in all culture conditions. On the basis of the studies described above, 12 DIV was chosen as the timepoint for further investigation.

Figure 4.

Comparison of temporal changes in cell viability (a and b), total protein (c) and excitatory amino acid transporter (EAAT) activity (d) of astrocytes cultured on different matrices for 4 DIV and 12 DIV. *p < 0.05, **p < 0.01, ***p < 0.001. Comparisons were made using two-way anova with Bonferroni's post-hoc test. Data represent mean ± SEM. For 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT), lactate dehydrogenase (LDH) and [3H]-d-Asp uptake, n = 3 (4 DIV), n = 6 (12 DIV) from independent experiments. For total protein, = 5 (4 DIV) and = 9 (12 DIV) from independent experiments.

Western immunoblotting studies were undertaken for GFAP (Fig. 5), a main constituent of intermediate filaments in astrocytes, and an index classically increased in reactive astrocytes (Middeldorp and Hol 2011). The expression of GFAP in astrocytes grown on both random and aligned scaffolds decreased appreciably (~ 80% reductions) relative to those maintained in 2D culture at 12 DIV (polystyrene, p < 0.01).

Figure 5.

Effects of aligned and random poly-ε-caprolactone (PCL) scaffolds on glial fibrillary acidic protein (GFAP) protein expression. (a) Western blot demonstrates a significant decrease in GFAP (51 kDa) protein expression after 12 DIV on both aligned and random PCL scaffolds. Samples from three independent experiments. β-actin (41 kDa) was used as loading control. Quantification was based on optical density, measured using ImageJ. Data represent mean ± SEM. **Significantly different from control, p < 0.01, one-way anova with Dunnett's Multiple Comparison post-hoc test.

Astrocytic interaction with the 3D scaffolds

Confocal imaging indicated that GFAP-positive fibres penetrated into the fibre network of aligned and random PCL scaffolds (Fig. 6a and b respectively). Image analysis allowed the concurrent visualisation of GFAP-positive fibres and digital sectioning demonstrated that processes extended up to 10 μm into the scaffolds, including projecting to the base of the scaffolds (Fig. 6). Hoechst-labelled nuclei were also visualised within the 3D PCL, indicating that astrocytes were capable of migrating deep into the scaffolds. Astrocytes appeared not to show any obvious preference with respect to their capacities to penetrate into random or aligned scaffolds.

Figure 6.

Three-dimensional rendering of confocal images (Z-stacks). Astrocytes were grown for 12 DIV on (a) aligned poly-ε-caprolactone (PCL) scaffold and (d) random PCL scaffold. (b, c, e and f) showed digital sectioning, revealing the cross-section of the scaffold. Cells on the scaffold were labelled with glial fibrillary acidic protein (GFAP) (red) and Hoechst 33342 (blue). Divisions on x-/y-axes = 10 μm and on z-axis = 2 μm.

Gene expression

Given the extensive changes in astrocytic morphology and GFAP expression, we sought molecular insights into genes that might contribute to the transformation of phenotype. Our previous study (Lau et al. 2012) focused attention on select genes involved in (i) cell motility and pathfinding [skeletal muscle actin α2 (ACTA2), vinculin (VCL), chemokine (C-X-C motif) ligand 12 (CXCL12)], (ii) TGFβ-related migration [transgelin (TAGLN)] and (iii) astrocytic function, including Glu transport (SLC1A2 = EAAT2), BDNF and anti-oxidant [glutathione S-transferase α1 (GSTA1) and heme oxygenase (decycling) 1 (HMOX1)] activities. The mRNA expression of all of these genes was significantly greater in astrocytes cultured on the 3D scaffolds, regardless of their fibre alignment (Fig. 7), relative to conventional 2D cultures. PCL itself possessed properties that induced increased mRNA expression, since expression of ACTA2, VCL, CXCL12 and BDNF were increased in astrocytes grown on either 2D or 3D PCL (both random and aligned) scaffolds (Fig. 7). mRNA expression of TAGLN, a gene linked to transforming growth factor β-mediated cellular migration, in astrocytes grown on 3D-aligned scaffolds was significantly greater than astrocytes grown on control scaffolds and 2D PCL film, suggesting this increase is because of the 3D morphology and topography of the PCL scaffold (Fig. 7). Similar increases in mRNA expression of the major Glu transporter SLC1A2 and GSTA1, a key enzyme for glutathione synthesis, in conjunction with the BDNF data, suggest that the electrospun fibres had promoted a salutary astrocytic phenotype.

Figure 7.

Expression profiles of selected genes when astrocytes were grown on different matrices for 12 DIV (C, control; P, 2D poly-ε-caprolactone (PCL); A, aligned scaffold; R, random scaffold). Data are presented as mean ± SEM from four independent experiments, and are expressed relative to the control. 95% confidence intervals were constructed for single-variate analysis where *indicates significant changes compared to control, #indicates significant changes compared to 2D PCL.

Discussion

Astrocytes offer great potential for regenerative neurobiology because of the fundamental roles they play in the function of the CNS, participating in many physiological events including synaptic transmission, maintenance of neurotransmitter homeostasis, cerebral blood flow and as a site of anti-oxidant defence during stress. Astrocytic responses occur as a continuum that is graded according to the extent of the trauma or disease – when minor there is resolution and even the extreme scenario of glial scar formation is considered manageable by pharmacological intervention (Ridet et al. 1997; Mueller et al. 2009; Sofroniew 2009). Our interest in taking advantage of the biology of astrocytes to promote repair and regeneration was driven by our documentation of phenotypes in vitro (Lau et al. 2012) and in vivo (Nisbet et al. 2009) consistent with a pro-survival state. In these studies different interventions resulted in cytotrophic properties that potentially could result in beneficial outcomes. Here we have used engineered scaffolds to influence astrocytic phenotype with the ultimate goal of understanding how the inflammatory cascade might be tuned to generate a favourable outcome. Key findings emergent from our study were that astrocytes survived and migrated on 3D PCL scaffolds with a stellated morphology, adopting a phenotype where the reduced pattern of GFAP expression and overall gene profile were likely to be cytotrophic in a reactive milieu.

Cell morphology and behaviour

When astrocytes were cultured on the 3D scaffolds, they initially formed colonies with condensed processes and then adopted elongated and ramified cell morphologies. Given the relative increase in G-actin staining (i.e. fewer stress fibres) and an ~ 80% decrease in GFAP expression, our data indicate these cells are less stressed compared to cells grown in the conventional 2D culture system. The morphology of the cells changed dramatically as they extended processes over the substratum to occupy vacant space. This cytoskeletal event is considered an important determinant of cell growth and survival (O'Neill et al. 1986; Schnell et al. 2007; Horne et al. 2010). At 12 DIV we demonstrated that the survival and viability of astrocytes grown on the 3D PCL scaffolds was maintained with findings suggestive of adoption of a quiescent state after the initial proliferation and expansion (c.f. Fig. 4). While cellular proliferation was supported by 3D scaffolds, the large reduction in GFAP levels relative to only a ~ 30–40% decrease in total protein is indicative of a genuine shift in astrocytic phenotype on PCL scaffolds.

Given the tendency for astrocytes to form colonies as early as 4 DIV on the scaffolds, contact inhibition (or density-dependent inhibition of cell division) may be occurring in these conditions (Martz and Steinberg 1972). However, nutrient supply is thought to be more important than the effects of cell–cell contact, so the pattern of organisation of high-density cultures may be a contributing factor to their slower rate of cell proliferation. Thus cellular density and the reduced nutrient supply, as well as the toxic build up of waste products, likely influenced viability at the density of 16 × 103 cells/well (Owen and Shoichet 2010), where there were extensive overlapping layers of astrocytes. Moreover, for both types of 3D scaffolds GFAP-positive processes were significantly longer at the 8 × 103 relative to 4 × 103 cells/well at 12 DIV (Figure S2), suggesting this density was optimal in terms of cell–cell contact required for astrocytic growth. Density-dependent inhibition of cell division may help explain the reduction in glial scarring observed in rat striatum when treated with PCL scaffolds in vivo (Nisbet et al. 2009), considering astrocyte proliferation soon after injury has been associated with a destructive phenotype (Parish et al. 2002).

Cell–scaffold interactions: cytoskeletal evidence for increased cellular migration

Electrospun scaffolds are engineered to mimic the native ECM, and both the fibre orientation and topography have profound effects astrocytic shape, growth and function (Singhvi et al. 1994; Gerardo-Nava et al. 2009). The present results demonstrate that aligned scaffolds provide contact guidance cues that allow for longer and more directed process outgrowth (Z-stack, Fig. 6) even though migration and penetration of cell bodies into both random and aligned scaffolds was similar. This ability of electrospun scaffolds to influence the behaviour and directionality of astrocytic growth is thus of particular interest to CNS repair strategies attempting to target specific areas of injury and direct the outgrowth of regenerating axons (East et al. 2010). Astrocytes have been grown successfully on micro-grooved and -patterned substrates (Sorensen et al. 2007; Meng et al. 2012) but these topographies do not offer the 3D advantages of our scaffolds. Attempts to culture astrocytes on electrospun fibres have generally been hampered by issues related to poor proliferation and cytoskeletal stress (Cao et al. 2012; Kim et al. 2012), and our success shows the advantages of employing secondary cultures of astrocytes in combination with electrospun scaffolds. When this study was essentially complete the successful maintenance of astrocytes in 3D was reported on electrospun scaffolds (Puschmann et al. 2013; Zuidema et al. 2014). Our procedure has the important advantage over both these studies that it allows 3D culture over an extended time interval of more mature astrocytes under conditions where there is preservation of the cytotrophic phenotype with an extensively arbourised cellular architecture featuring numerous major and minor processes.

Normally astrocytes are grown in vitro as a 2D monolayer, which is very different to their 3D environment in vivo. Unlike the recent work by Zuidema et al. (2014), who undertook extensive characterisation of astrocytic architecture in 2D, we used confocal Z-stacks in conjunction with Hoechst staining, and cytochemistry for GFAP to demonstrate that cells migrate and penetrate through the scaffold from its top through its middle to the base. The mean fibre diameter of our electrospun fibres was appreciably less (0.4 vs. 2 μm) than that of Zuidema et al. (2014), who also employed fibronectin, conditions that are likely to encourage adhesion and extensive growth and migration (Wang et al. 2009) along their scaffolds. Importantly, the appreciable penetration of astrocytes processes into our biomatrices achieved one of the key aim of our study to produce 3D astrocytes. Interestingly, both scaffolds lead to promotion of Glu transport, but guided by our earlier evidence for cytoskeletal changes (Lau et al. 2012) we also demonstrated reduced expression of GFAP and actin stress fibres in conjunction with up-regulation of BDNF and anti-oxidant genes. These data suggest that the current difficulty in correlating healthy, stellated in vivo astrocytic phenotypes with flat, cobblestone-like morphologies displayed in vitro may be overcome by using electrospun 3D scaffolds as a substrate to induce more physiologically representative astrocytes for in vitro studies of cellular behaviour and morphology.

Cellular interaction with different substrates can activate different cellular and ECM mechanisms linked to the control of astrocytic cytoskeletal arrangement (Provenzano and Keely 2011). The actin cytoskeleton has been well established to have important roles in cell morphology and migration, with astrocytic process formation occurring because of depolymerisation of actin filaments via a shift from F-actin to G-actin (Kimelberg 2004; Lau et al. 2011). This study found that control cultures of primary astrocytes have a high F- to G-actin ratio and numerous focal adhesions. Increases in both F- and G-actin observed in astrocytes cultured on random and aligned fibres are most likely a reflection of process elongation on these scaffolds. These data suggest that control cells are at higher levels of cytoskeletal stress, possibly because of the non-compliant 2D substrate (glass/polystyrene) on which they were seeded, whereas PCL scaffolds more clearly reflect the elastic modulus encountered in vivo. This interpretation is also supported by our immunoblotting data of GFAP expression, which was decreased in astrocytes grown on both random and aligned scaffolds rather than up-regulated as would occur in inflammatory response occurring in CNS insults (Middeldorp and Hol 2011).

The pathways that regulate actin polymerisation are not fully characterised, however the GTPases (Rho, Rac and Cdc42) are known be involved (Goldman and Abramson 1990; Hall 1998; Etienne-Manneville and Hall 2002). Extensive work in our laboratory has shown that following treatment with the Rho kinase inhibitors, cultures of murine astrocytes shift from a flat, cobblestone phenotype to a stellate shape, with a concomitant decrease in F-actin and a proportional increase in G-actin (Lau et al. 2011) and adoption of a cytotrophic phenotype (Lau et al. 2012). Astrocytes grown on the PCL scaffolds showed similar increases in ‘healthy’ genes, especially the SLC1A2 (i.e. EAAT2) and anti-oxidant-linked GSTA1, indicating that 3D scaffolds promote a transition to a cytotrophic phenotype. Our observation of increased SLC1A2 is consistent with Zuidema et al. (2014), who also found EAAT1 expression was elevated on their bioscaffold. Expression of the neurotrophin BDNF was always elevated on PCL, including on both random and aligned scaffolds. Indeed, there is evidence that SLC1A2 and BDNF may be co-regulated in health and pathology (Xu et al. 2008; Gourley et al. 2012; Lau et al. 2012). The up-regulation of SLC1A2 expression found on the scaffolds was also found with ROCK inhibitors (Lau et al. 2012) and likely represents the homeostatic behaviour displayed by astrocytes to maintain this key transporter (Abe and Misawa 2003; Zagami et al. 2009; Lau et al. 2010) integral to the CNS defence against excitotoxicity. Key indices of the healthy phenotype would appear to be not only elevated expression of SLC1A2, but also of the major anti-oxidant gene, GSTA1, together with decreased GFAP and an accompanying increase in G-actin.

Conclusion

These results show that astrocytes cultured on electrospun 3D PCL scaffolds have extended processes that penetrate the scaffold. Importantly, decreased expression of the astrocytic intermediate filament marker GFAP and increased G-actin suggested astrocytes adopted biologies that were less stressed, whilst displaying elevated expression of genes indicative of a cytotrophic phenotype. This study points to the therapeutic potential of bioengineering strategies using 3D electrospun scaffolds of differing fibre orientations that direct astrocytes into phenotypes supporting brain repair.

Acknowledgments and conflict of interest disclosure

The authors would like to thank Ms Alexandra Rodriguez and Kiara Bruggeman for critical reading of the manuscript. This research was supported by funding from an NHMRC project grant (APP1020332). DRN was supported by an Australian Research Council Australian Postdoctoral Fellowship and subsequently by an NHMRC Career Development Fellowship (APP1050684). MKH and PMB were supported NHMRC Research Fellowships, APP1020401 and APP1019833, respectively. CLP is a Senior Medical Research Fellow supported by the Viertel Charitable Foundation, Australia. Access to the facilities of the Centre for Advanced Microscopy (CAM) with funding through the Australian Microscopy and Microanalysis Research Facility (AMMRF) is gratefully acknowledged.

All experiments were conducted in compliance with the ARRIVE guidelines. The authors have no conflict of interest to declare.

Ancillary