Cotton-like micro- and nanoscale poly(lactic acid) nonwoven fibers fabricated by centrifugal melt-spinning for tissue engineering

Biodegradable materials in the form of nonwoven fibers have attracted increasing attention for tissue engineering applications because they offer large surface areas and interconnected networks. In this study, cotton-like nonwoven poly(lactic acid) (PLA) fibers were successfully fabricated by centrifugal melt-spinning. The effects of centrifugal speed and secondary melt-spinning processing on the morphology, mechanical properties, and cell compatibility of the fibers were investigated. Scanning electron microscopy, differential scanning calorimetry, and Fourier-transform infrared spectroscopy (FTIR), as well as cell culturing of MC3T3-E1 were used in this study. The results showed that centrifugal speeds from 350 to 1500 rpm satisfied the needs for fiber formation. The PLA fibers we prepared had three-dimensional structures with extensive diameter distribution from the nanoscale to several tens of micrometers, large pore sizes, and high porosities, significantly different from fibers produced by electrospinning. The fiber diameters and mechanical properties could be manipulated by controlling the centrifugal speed. The finest fibers were generated at 900 rpm with average diameters of 3.47 ± 3.48 μm. The fibers created by centrifugal melt-spinning exhibited lower cytotoxicity and higher cell proliferation than those obtained by electrospinning.


Introduction
Because they offer large surface-area-to-volume ratios and a wide range of morphologies and geometries in threedimensional polymeric scaffolds, ber-based biodegradable materials have attracted increasing attention for many applications in recent years. The past decade has seen increasingly rapid advances in the eld of biodegradable polymer bers employed as tissue engineering scaffolds. 1 Although most materials in tissue engineering are porous foam scaffolds produced by the particulate leaching method, gas foams, and freeze-drying, the large surface-area-to-volume ratio, exibility, and high permeability of bers grant them potential in tissue engineering applications. [2][3][4] Electrospinning has been extensively employed as a technique to generate scaffolds for tissue engineering. The nanoscale and nonwoven structures of electrospun bers, characterized by inherent porosities and random arrangements, can mimic that of extracellular matrices (ECMs), which is necessary for any in vivo tissue engineering application. However, the average pore size of such bers is less than 1 mm, much smaller than the actual cell size of 5-20 mm. Pore sizes below the average cellular diameter block cell migration within the structure. Thus, cells inevitably failed to penetrate electrospun nanobrous scaffolds. 5 Moreover, in electrospinning, the residual solvent residing in the bers can be detrimental to various cellular activities, compared to those produced by solvent-free processes, which restricts the applications of electrospun brous scaffolds or mats in tissue engineering applications. 6,7 Thus, in this study, we attempted to attain larger pore sizes to facilitate inltration and cellular in-growth, while avoiding solvent usage. Inspired by cotton candy machines, in this work, we manufactured an inexpensive centrifugal spinning system for the high-rate and low-cost synthesis of bers with diameters ranging from nanometers to micrometers in scale. Based on this system, cotton-like nonwoven bers of PLA were fabricated successfully. In this study, we focused on the effects of the centrifugal melt-spinning parameters on the diameter distribution of the bers. In addition, the physical and mechanical properties and cell compatibility of the bers were studied.

Materials and methods
Centrifugal melt-spinning device Scheme 1 shows a schematic of the centrifugal melt-spinning system developed in our present study. The apparatus consists of a rotary disk, heating unit, electromotor, and other electro-circuit-controlling devices retted from a cotton candy machine. The centrifugal speed is between 350 and 2000 rpm based on the electromotor. The temperature of the disk can be controlled from 20 to 300 C. The PLA powder lls the heated disk through the lling inlet, melts gradually, and fuses into bers under centrifugation at different speeds before collection by a rotary drum. PLA with the viscosity-average molecular weight (M v ¼ 90 595) was synthesized in our laboratory.

Preparation of nonwoven PLA bers
Before processing, the PLA powder was dried at 70 C for 4 h in a convection oven to avoid hydrolytic degradation during manufacturing. The dried PLA powder (denoted "RAW") was poured into the heated rotary disk through the lling inlet. According to our undocumented experimental data, a successional and higher yield of PLA bers could be obtained for the central and edge temperatures of the rotary disk of 220 C and 180 C, respectively. The melt temperatures differed, determined by the intrinsic thermodynamic properties of the raw materials. Three kinds of bers were achieved by different rotary speeds. The speeds were set as 350, 900, and 1500 rpm, denoted as LS, MS, and HS, respectively. The primary bers (denoted 1-LS, 1-MS, and 1-HS) were gathered by a rotating drum. Then these primary bers were used as the raw materials to produce secondary bers under the same speeds by repeating the procedure described above; the products are denoted 2-LS, 2-MS, and 2-HS, respectively. Electrospun PLA bers were used as a control group; these were produced under conventional conditions (concentration of polymer in chloroform: 8% (w/v); voltage: 20 kV; distance: 15 cm).

Morphological and diameter characterization
The morphologies of the PLA nonwoven bers were characterized by digital photography and eld-emission scanning electron microscope (FESEM, Philips XL30). The ber samples were plated with a thin layer of gold before FESEM observation. The FESEM images were analyzed with NIH Image J soware (provided by the National Institute of Health, USA) to determine the ber diameters. 1000 bers for each group were measured.

Tensile testing
Mechanical testing of the bers was performed on an Instron 1121 (US) universal testing machine at room temperature and the relative humidity of 47%. Tensile measurements were performed with the crosshead speed of 5 mm min À1 and initial grip separation of 20 mm. In order to ensure the comparability of different samples, the weight and length of the bers were set as 0.4 g and 3.0 cm uniformly and respectively to ensure that the cross-sectional areas would be 0.095 cm 2 , theoretically. An average of four individual tensile determinations was performed for each sample; the mean and standard deviation of these four determinations are Scheme 1 Schematic of the centrifugal melt spinning apparatus, consisting of (a) a rotary disk, (b) heating circuit, (c) electromotor, and other electro-circuit-controlling devices refitted from a cotton candy machine. The centrifugal speed is between 350 and 2000 rpm, according to the electromotor. The temperature of the disk can be controlled from 20 to 300 C. The polymer powder fills the rotary disk through the filling inlet (d), melts instantly, fuses into fibers under centrifugation at different speeds, and collects on a rotary drum (e).
presented. To characterize the ber structure, the tensile strength, elastic modulus, and stress-strain behaviors were tested on the tensile tester.

Intrinsic viscosity molecular measurement
The viscosity-average molecular weight (M v ) was measured using gel permeation chromatography (GPC) at 30 C. The extents of decrease in the M v were dened by the differences between the M v of the PLA before (M v1 ) and aer (M v2 ) centrifugal melt spinning, i.e., (1)

Fourier-transform infrared spectroscopy (FTIR)
In order to identify the chemical structural differences in the bers, the infrared (IR) spectra was obtained using a Bruker Vertex Â70 Fourier transform IR (FTIR) spectroscope by the viscosity method in a dilute polymer/chloroform solution of 3 mg mL À1 PLA.

Differential scanning calorimetry (DSC)
DSC scans were performed in the temperature range from 0 to 220 C at a heating rate of 20 C min À1 and the nitrogen ow was set to 50 mL min À1 using a DSC Q100 (TA Instrument, USA). From the thermograms, the crystallinities of the samples were recorded using the Universal Analysis 2000 program available from TA Instrument. All measurements were taken from the calorimetric date for the rst heating cycle, because the bers in the second heating scan show no differences. 2 The degree of crystallinity (X DSC,c ) was estimated considering the ideal melting enthalpy of 93.7 J g À6 according to the following equation:

Cytotoxicity test
The cytotoxicity of the bers was determined based on the viability of MC3T3-E1 cells in material extracts using a 3-(4,5dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide dye (MTT, Sigma-Aldrich) assay. The ber specimens from centrifugal melt-spinning and electrospinning were immersed into Dulbecco's Modied Eagle Medium (DMEM) containing 10% serum and incubated at 37 C for 72 h to obtain the material extracts. The ratio of the ber specimen to the medium was 0.5 g/25 mL. The material extracts were used as the cell culture media to perform the MTT assay. The MC3T3-E1 mouse pre-osteoblasts were purchased from the Shanghai Institute of Cell Biology, Chinese Academy of Sciences (CAS). The cells were seeded into a 96-well tissue culture plate (TCP, Gibco) at a density of 1 Â 10 4 cells per well and were cultured at 37 C and 5% CO 2 under static conditions for up to 24 h. The medium of each well was then replaced with 200 mL of the material extract and the cells were incubated for another 20 h. TCP was set as the control and 6 replicates were used for each group. 20 mL of MTT (5 mg mL À1 in phosphate-buffered saline (PBS)) was added to each well, aer which the cells were incubated for an additional 4 h. The medium was then removed and 150 mL of dimethyl sulfoxide (DMSO) was added to each well to solubilize the converted dye. The absorbance values at 450 nm were measured on a multifunctional microplate scanner (Tecan Innite M200).

Cell proliferation and cell morphology
The cellular attachment and cell viability of the bers were determined using the MTT assay. The bers and electrospun specimens were lled into a 48-well plate at 0.15 g per well.
Before seeding, the bers were pre-wetted by submersion in a 70% aqueous ethanol solution overnight and then immersion in media for 2 h in an incubator at 37 C. 7,8 Aer 1, 3, 7, and 14 days of incubation, 100 mL of MTT was added and incubation was continued for 6 h. 750 mL of acidied isopropanol (0.4 mol L À1 ) was added to each well and incubated at 37 C for 15 min to solubilize the converted dye. The solution in each well was mixed 200 mL samples were obtained and transferred to another 96-well plate. The optical densities were measured at 540 nm wavelength on a multifunctional microplate scanner (Tecan Innite M200). TCP was set as the control and three replicates used for each group. A morphological study of the MC3T3-E1 cells cultured on different samples was performed aer 14 days of cell culturing, using FESEM. The cell-ber constructs were rinsed three times with PBS and xed in 3% glutaraldehyde for 24 h. The samples were further rinsed in PBS and dehydrated with increasing concentrations of ethanol (50%, 60%, 70%, 80%, 90%, and 100%) for 30 min each. Finally, the cell-ber constructs were freeze-dried for 2 d before FESEM observation.

Statistical analysis
All experiments were performed in triplicate at the minimum. All quantitative data is expressed as the mean AE the standard deviation. Wherever appropriate, comparisons of the means were performed using Origin 8.0 soware (OriginLab Corporation, USA), with p < 0.05 considered statistically signicant.

Morphology and diameters
The diameter distribution, pore size, and orientation of bers are critical for tissue engineering applications. 8 PLA is one of the most promising biodegradable polymers in tissue engineering; PLA bers prepared by different methods, such as melt spinning, [9][10][11][12][13][14][15] solution spinning, 16,17 and electrospinning, [18][19][20][21] have been reported. In this work, we manufactured a centrifugal melt-spinning system for fabricating nonwoven bers of PLA. The morphologies of the bers, observed by optical microscopy and FESEM analysis, are shown in Fig. 1-3. Fig. 1 shows that the nonwoven PLA bers formed by the centrifugal melt spinning method are loose and cotton-like, signicantly different from the ake of electrospun PLA bers. For the primary melt-spinning process (Fig. 1B), the bers of 1-LS, fabricated at a low rotation speed, are rough and non-homogeneous. The neness and homogeneity of the primary bers is increased gradually as the centrifugal speed increases. However, no similar tendency appears in the ber morphology for the secondary melt-spinning process (Fig. 1C). The 2-MS bers appear ner and more homogeneous than the other two secondary melt-spinning samples.
Low-magnication FESEM observation shows that the 1-HS This journal is © The Royal Society of Chemistry 2018 and 2-MS bers exhibit the greatest neness and homogeneity (Fig. 2), which is consistent with the optical observations. Furthermore, the high-magnication FESEM observations indicate that the bers form a reticular and interconnected three-dimensional structure (Fig. 3). By centrifugal melt spinning, desirable three-dimensional structures can be formed, but similar structures are difficult to obtain by electrospinning.
The size distributions of the different bers are analyzed as shown in Fig. 4 and Table 1.
For the primary bers (Fig. 4), the diameters of 1-LS, 1-MS, and 1-HS have ranges of 0.10-101.61 mm, 0.12-72.22 mm, and  This journal is © The Royal Society of Chemistry 2018 the increased rotation speed of 1500 rpm, the diameters of the 2-HS bers are increased. The diameter distribution of the 2-HS bers does not follow the tendency of increasing diameter with increasing centrifugal speed. Because the viscosity of the spinning uid is positively associated with the crystallinity and the amount of chain scission, lower-viscosity precursor uids facilitate PLA ber formation, which supports the smaller diameters of 2-LS and 2-MS relative to the primary ones. However, in preparing the 2-HS bers, the balance between the centrifugal force and viscosity of the PLA is broken; thus, the less viscous PLA uid cannot resist the strong centrifugal force produced by the high centrifugal speed, and therefore, 2-HS contains coarser bers. In short, based on these, we can fabricate ideal nano-and microbers with various diameters. The physical and chemical properties could be adjusted by varying the centrifugal speed and raw material.
The centrifugal melt-spun bers have more extensive ber diameter distributions, ranging from single nanometers to dozens of micrometers, unlike the narrow nanoscale distribution of electrospun bers (the measurement of nanoscale bers was handicapped by technology restrictions). Two techniques of electrospinning and needle punching have been used to combine the benecial properties of nanobers and microbers in three-dimensional porous structures. 21 However, in this study, we fabricated nano-and microbers using a single technique. In other words, it was easier in preparation process to obtain nanoand microbers by this method. In addition, the pore sizes of the bers are clearly larger than the size of cells, and the ber orientation is three-dimensional and interconnected.
The diameter distribution of bers is an important surface morphology parameter of scaffolds for cell contact, affecting cell proliferation and differentiation. 22,23 Takahashi et al. found that the number of mesenchymal stem cells (MSCs) attached to non-woven fabrics was increased with increasing ber diameter and that both the alkaline phosphatase (ALP) activity and the osteocalcin content of MSCs, as bone differentiation markers, peaked for ber diameters of $9-12 mm. 8 Nanoscale bers have been widely used for tissue engineering because they can mimic the nanoscale geometry and topology of ECM structures. We  nd that bers with single diameter distributions are unable to meet many tissue-engineering needs. According to our calculations, the advantages of both nano-and microscale structures can be provided by a wide diameter distribution, with new advantages in addition. For example, scaffolds composed of bers with wide diameter distributions may show better permeabilities than those of nanoscale bers as well as higher surface areas than those of microscale bers, both of which facilitate gradual tissue ingrowth. Moreover, the loose compact nanoscale networked structure mimics ECMs, while microscale bers provide the mechanical support. Therefore, the PLA bers prepared in this study should have much potential in tissue engineering applications because they have broad diameter distributions ranging from nanometers to dozens of micrometers that can be adjusted by changing the centrifugal speed. Using such bers, we have fabricated three-dimensional scaffolds for bone tissue engineering in our laboratory. [24][25][26] Mechanical properties The mechanical properties of the bers fabricated at rotation speeds of 350, 900, and 1500 rpm are presented in Fig. 5.
The tensile strengths of the primary bers are increased signicantly with increasing centrifugal speed (1.51 AE 0.28, 2.34 AE 0.32, and 3.31 AE 0.12 MPa for 1-Ls, 1-MS, and 1-HS, respectively), and their elastic moduli follow the same trend (1-Ls < 1-MS < 1-HS). However, among the secondary bers, the tensile strength of 2-LS is 2.01 AE 0.15 MPa, but those of 2-MS and 2-HS are lower. Under the same centrifugal speeds, the tensile strengths of 2-MS and 2-HS are clearly smaller than those of 1-MS and 1-HS (Fig. 5 and Table 2).
The elastic moduli of the secondary bers are 10.23 AE 2.33 MPa, 17.65 AE 2.58 MPa, and 8.75 AE 0.76 MPa, signicantly lower than those of the primary bers.
The biomechanical properties of scaffolds are essential to support the attachment, proliferation, and differentiation of cells. In this study, the primary bers have higher tensile strengths and elastic moduli than the secondary bers; however, the secondary bers have greater strain capacities. We attribute this to the crystallinity reduction during the secondary melt-spinning process and the viscosity-average molecular weight reduction by random chain scission reactions, intermolecular and cyclic oligomerization, and/or Paper transesterication, as conrmed later by FTIR. The dramatic reduction in crystallinity during the melt-spinning process is a signicant but inevitable defect of this preparation method, because the process requires high temperatures.
Thermal degradation during melt spinning Table 3 shows that the viscosity-average molecular weights M v of the primary bers are decreased by approximately 12.69%, 8.59%, and 1.85% compared to the raw material, respectively. Simultaneously, the M v of the secondary bers are decreased by approximately 22.28%, 16.41%, and 5.88% compared to the raw material and by 10.98%, 8.55%, and 4.11% compared to the primary bers.
As shown in Fig. 6, FTIR analysis indicates no obvious differences between the raw material and 2-LS bers; no new hydroxyl groups, which would be the products of ester hydrolysis in all bers, are formed from carboxylic acid and alcohols.
The thermal decomposition and stability are important PLA ber properties determining the reprocessing performance and degradation of the nal materials. The decreases in viscosityaverage molecular weight show tendencies similar to that of the thermal stability. Higher centrifugal speeds correspond to faster PLA cooling and shorter durations in the heated disk.

DSC
The DSC thermal properties of the bers are shown in Fig. 7. The thermal transitions of the bers are important to determine the processing and end-use treatment of products containing them. In contrast to the RAW material, the meltspun PLA bers show greater variation in thermal transition. The DSC thermogram of the bers shows T g values of $60 C irrespective of ber formation phase ( Fig. 7A and B). An exothermic peak representing cold crystallization (T c ) is observed between 83 C and 90 C, and the melt temperature (T m ) is shied to lower temperatures between 159 C and 165 C. Unlike T g and T c , T m decreases slightly as the spinning speed is decreased. As the speed increases, the ber crystallinity increases, which elevates the melting point.

Cytotoxicity and cell viability
The cytotoxicity of the material extracts and the cell proliferation of MC3T3-E1 cells cultured on the bers are shown in Fig. 8. The melt-spun bers show no signicant differences in cytotoxicity, but they are less cytotoxic than the electrospun specimens (Fig. 8A). During the rst week of culturing, no signicant difference is observed in cell proliferation between the melt-spun bers and control (TCP), but proliferation is higher for the melt-spun than the electrospun bers (Fig. 8B).
The effect of ber diameter in the fabric constituting the scaffold on the cell behavior has been examined by multiple studies. [3][4][5] In this study, the biocompatibility of the melt-spun bers was assessed by examining their potential to support the growth and proliferation of MC3T3-E1 in the bers. Since the melt-spun bers are produced by a solvent-free process, the Paper route is more eco-friendly and less toxic than that using electrospun specimen, as identied by the results of the cytotoxicity analysis. Theoretically, smaller ber diameter distributions correspond to larger surface areas, which can provide greater cell permeation. Surprisingly, the cell proliferation rates among the bers show no signicant differences, possibly because the pores of the crude bers are more conducive to cell ingrowth. This assumption can also explain the high cell proliferation rate on the melt-spun bers aer 14 days of culturing, compared to TCP. The results here differ from those reported by Takahashi et al. 8 In their study, MSC was demonstrated to attach, proliferate, and differentiate on PET non-woven fabrics with various diameters and porosities. However, the species and diameter distributions of the bers differ, making the results incomparable.

Cell morphology
As shown in Fig. 9, the morphologies of the MC3T3-E1 cells grown on the 2-MS and 2-HS bers are analyzed using FESEM. 2-MS and 2-HS were selected for cell morphology analysis because the differences in diameter were the most obvious and the molecular measurements and thermodynamic properties were the most similar between 2-MS and 2-HS. The effect of cell morphology on the bers was thus entirely determined by the One inherent limitation of electrospinning is the relatively poor cellular inltration into the depth of the scaffold due to small pore size and high ber packing densities. 27 The novel method of centrifugal melt spinning was developed in this study to fabricate biodegradable polymer bers with a diameter distribution of 1-30 mm. Specically, the pore sizes and orientations of the centrifugal melt-spun bers are advantageous for cell penetration of the ber scaffolds.

Conclusions
Solvent-free PLA nonwoven bers with different diameter distributions and three-dimensional intersecting structures were prepared by a centrifugal melt-spinning method. The This journal is © The Royal Society of Chemistry 2018 centrifugal melt spinning parameters were investigated to optimize the physical and chemical properties of the bers. The melt-spun bers were cotton-like with broad diameter distributions and interconnected three-dimensional structures. The physical and chemical properties could be adjusted by varying the centrifugal speed and raw material. The variety of achievable physical and chemical properties, particularly in the diameter distributions, pore sizes, and intersecting structures, would be useful for tissue-engineering scaffold applications. In addition, the centrifugal melt device can be applied to both dry and wet spinning, indicating signicant potential. The results of cell experiments suggested that the bers produced by our method have lower cytotoxicity and greater proliferation than the electrospun specimens do, although no signicant differences appeared among the various centrifugal melt-spun bers.
The cell differentiation properties of the centrifugal melt-spun bers and the optimized production process remain for further investigation.

Conflicts of interest
There are no conicts to declare.