A green approach to obtain stable and hydrophilic cellulose-based electrospun nanofibrous substrates for sustained release of therapeutic molecules

Stable and (bio)-compatible nanofibrous matrices showing effective incorporation and release of nonsteroidal anti-inflammatory drugs (NSAIDs) hold a huge potential in tissue regeneration and wound healing. Herein, a two-step, water-based and needleless electrospinning method is used to fabricate thermally cross-linked multifunctional nanofibrous substrates from a hydrophilic cellulose derivative, i.e. carboxymethyl cellulose (CMC), and polyethylene glycol (PEG) with an in situ incorporated NSAID, diclofenac (DCF). Electrospun bi-component blend nanofibers, strongly linked together by ester bonds, with different degrees of cross-linking density are achieved by varying the concentrations of butanetetracarboxylic acid (BTCA, a green polycarboxylic cross-linker) and the sodium hypophosphite (SHP) catalyst, and the temperature. The results demonstrated that not only the dimensional stability and swelling properties could be better controlled but also the morphology, fiber diameter, surface area, pore volume, pore size, and functionality of the cross-linked nanofibers. Release kinetics of DCF from the nanofibrous substrates are controlled and prolonged up to 48 h, and the overall released mass of DCF decreased linearly with increasing cross-linking degree of BTCA and SHP. Fitting of release data using various kinetic models revealed that the release of DCF follows a non-Fickian (diffusion and erosion controlled) to Fickian mechanism (only diffusion-controlled process). Cell viability testing based on crystal violet dyeing showed that the DCF-incorporating nanofibers have excellent biocompatibility and no toxic effect on human skin fibroblast cells. Overall, the reported DCF-incorporating nanofibrous substrate demonstrates high potential to be used as a smart drug delivery system in wound healing, especially due to its noninvasive characteristics.


Introduction
Nanobers of polymeric materials have been in focus for several years because of their unique morphology and internal architecture. Due to their superior features such as high surface areato-volume ratio, nanoscale diameter, porous structure, and exibility over other forms of the same materials, polymer nanobers are used in several areas of applications, including packaging, electronics, membranes, air ltration, dye scavenging, semiconductors, etc. [1][2][3][4][5][6][7] In addition, they hold huge potential in wound healing, especially in cell growth, owing to their resemblance to extracellular matrices (ECM) and extracellular macromolecules such as collagen. 8 Apart from cell adhesion and proliferation, polymer nanobers have been used for the release of various therapeutic molecules, including proteins, peptides, DNA, and drugs. [9][10][11][12] Consequently, numerous materials, active ingredients, and polymers (synthetic and natural) were exploited as potential drug delivery systems for the treatment of several diseases including cancer, 13 chronic wounds, 10 periodontal disease, 14 etc. However, control over the rate and quantity of drug release is critical for successful treatment of specic wound type (e.g., chronic). In this context, a plethora of studies on fabrication of drug incorporated polymer nanobers were performed and proven to be a promising sustained and smart drug delivery system during the last years. [15][16][17] Needless electrospinning technique is commonly used to obtain the mentioned ultrane bers (micro to nano-meter diameters) in larger quantities (10 g h À1 ) from polymeric solution or polymer blends containing active drug molecules. 18 Even though stable nanobers blends can be easily made from hydrophobic (e.g., cellulose acetate, 19 polycaprolactone, 20 etc.) or mixture of hydrophobic and hydrophilic polymers (e.g., poly(lactic-co-glycolic acid) (PLGA)/polyethylene glycol (PEG), 21 etc.), they are not suitable for incorporation and release of water soluble drug such as diclofenac (DCF), the nonsteroidal anti-inammatory drug (NSAID), which is oen used for the treatments of pain and inammatory diseases. 22 As a result, the blended nanobers from several hydrophilic polymers such as chitosan, gelatin, carboxymethyl cellulose (CMC), polyethylene glycol (PEG), silk, cellulose nanocrystals, collagen, etc. have been studied and reported elsewhere. 15,23,24 Amongst several others, CMC, a derivative of cellulose (naturally occurring biopolymer), is oen used over synthetic polymers in wound healing applications for the controlled release of drugs, due to its hydrophilicity, biocompatibility, extensive uid uptake capacity, degradability, and inexpensiveness. However, CMC alone provide challenges in electrospinning, which can be overcome by incorporation of hydrophilic, non-immunogenic, and non-toxic spinning agent such as PEG. The latter provides adequate chain entanglement in the liquid state, and thus resulting in uniform ber formation. Even though the blends of CMC and PEG are electrospinnable, 25 the hydrophilic nanobers substrate still possess some disadvantages; as they can be rapidly dissolved/destroyed upon contact with wound liquid environment, preventing controlled and prolonged drug release. Thus, besides the uid uptake capacity, certain dimensional stability of the nano-brous materials in the wound-mimicking environment must be ensured. One way to resolve this issue is the usage of chemical cross-linkers such as aldehydes, carbodiimides, 26 epichlorohydrin, diglycidyl ether, glycidylmethacrylate, etc. 27 However, these cross-linkers have limited applications in wound healing, due their toxicity towards human skin cells. Therefore, multifunctional polycarboxylic acid (e.g. 1, 2, 3, 4 butanetetracarboxylic acid, BTCA), which is eco-friendly, water soluble, and non-toxic, is suitable candidate for cross-linking of CMC/PEG nanobers. While the CMC/PEG based hydrogels have been used for drug delivery and wound dressing/healing applications, 28,29 the systematic and detailed studies on electrospun nanobers from the same polymers for the release of DCF and growth of human skin cells such as broblast are very few. Especially, no studies can be found on the thermal crosslinking and stabilization of CMC/PEG nanobers with BTCA through a sustainable, greener and two-steps approach.
The aim of our study was therefore to prepare water insoluble CMC/PEG nanobrous substrates for the controlled and prolonged delivery of DCF. To achieve this, the DCF incorporated CMC/PEG blend aqueous solution was mixed with BTCA and SHP at different concentrations. The viscosity, conductivity and surface tension of these solutions consisting of DCF or no DCF were investigated in detail before electrospinning. The electrospun biocomponent CMC/PEG nanobers incorporated with and without DCF were characterized in terms of chemical composition, morphology, pore size, pore volume, ber diameter as a function of BTCA and SHP concentration and (cross-linking) temperatures by using several modern analytical tools. The crosslinking efficiency of BTCA and its impact on bre swelling capacity and stability was investigated by conventional swelling and quartz crystal microbalance with dissipation (QCM-D) measurements, respectively. In vitro release kinetics of DCF, at physiological pH, from nanobrous substrates cross-linked with different BTCA concentration were performed, and the release data was best tted to different kinetic models. Prepared nanobrous substrates were also tested in regard of their potential applicability in biomedical applications. This was done by determining the cytotoxicity of drug incorporated nanobrous substrates towards human skin derived broblasts. The obtained results from these studies should lay a strong basis; which could be potentially used to build multifunctional and stable nanober system for the targeted delivery of several therapeutic agents in the eld of wound healing, and tissue repair, in general.
2.2 Preparation of samples for electrospinning 5 g of CMC was added to 95 g of water (5 wt%) and stirred with a 3-bladed mechanical propeller stirrer (Ika, Germany) at 2000 rpm until a homogenous solution was obtained ($24 h). In the same way, 5 wt% PEG solution was prepared. The crosslinker BTCA and the catalyst SHP at four different concentrations (3, 5, 7 and 10 wt%based on the weight of CMC) were prepared by dissolving appropriate amount of BTCA or SHP in MilliQ water. The above prepared four types of solutions (CMC, PEG, BTCA and SHP) were then used for creating formulations for electrospinning. Briey, CMC (5 wt%) and PEG (5 wt%) solutions were mixed in a 1 : 1 ratio. To this, 1 ml of BTCA and SHP, both prepared at four concentrations (3, 5, 7 and 10 wt%), were added and stirred for 1 h. In total 5 samples were prepared, and they are designated as BTCA_0%, BTCA_3%, BTCA_5%, BTCA_7% and BTCA_10%, respectively. The control sample BTCA_0% i.e. CMC/PEG contains no BTCA and SHP. To these ve solutions, DCF (10 wt%, based on the weight of CMC, dissolved in water) was added and stirred with a 3-bladed mechanical propeller at 2000 rpm for 1 h. These new ve samples prepared with DCF were designated as DCF_BTCA_0%, DCF_BTCA_3%, DCF_BTCA_5%, DCF_BTCA_7%, and DCF_BTCA_10%, respectively.
Prior to electrospinning, all these ten prepared solutions were characterized in regard of their viscosity, surface tension and conductivity using viscometer (Fungi lab), photogoniometer (OCA35 Data Physics) and conductivity meter (Mettler Toledo), respectively.

Needless electrospinning
Electrospinning of all solutions was conducted using a pilotscale needless electrospinning apparatus ElMarco Nanospider NS LAB 500 (Czech Republic). In comparison to widespread single needle electrospinning setup, with low production rate 0.1-1 g h À1 , 30 the electrospinning equipment used in this study, has the production rate of up to several 10g h À1 . Nanospider is a needle-less electrospinning apparatus with a high voltage power supply (up to 80 kV), feeding unit (a bathtub with rotating electrodecylinder or wire) and a grounded collector (cylinder or wire electrode). In the electrospinning process, we used stainless steel cylinder electrode. All nanobrous samples were electrospun at constant conditions; distance between the electrode and the collector plate set at 17 cm, accelerating voltage at 60 kV, time of electrospinning was 40 min, while temperature and humidity of working environment were set at 20 C and 30%, respectively. Electrospun nanobers were collected on a 100% polypropylene (PP) nonwoven fabric (Pegatex® S), used as a support material, manufactured with spun bond technology and supplied by Pegas nonwovens s.r.o. (Znojmo, Czech Republic).
For electrospinning on QCM Au-crystals, a specially designed sample holder made from Teon was used and mounted together with the QCM crystals on the PP fabric. 31 The role of the sample holder is to avoid deposition of nanobers at the edges and the backside of the crystal. For the electrospinning experiments, the same parameters as mentioned above were used, but the electrospinning of solutions were performed for 10 s.
Crosslinking of electrospun nanobers. All the electrospun samples were peeled-off from the PP support and cut into 6 cm Â 6 cm. The QCM crystals with nanobers attached to the fabric were also removed. Subsequently, they were transferred to a glass Petri dish and dried in an oven at 80 C for 5 min. Then they were kept at three different temperatures (120 C, 140 C, 160 C) for 15 min for the cross-linking of nanobers.

Scanning electron microscopy (SEM)
The morphology of electrospun nanobers before and aer crosslinking was analysed by Field Emission Scanning Electron Microscopy (FESEM), by using a Carl Zeiss FE-SEM SUPRA 35 VP electron microscope. The images were recorded with an acceleration voltage of 1 keV at approximately 4.5 mm working distance. Electrospun samples were xed on a conductive carbon tape attached to a metal microscope holder and sputtered with a thin layer of palladium using a Benchtop Turbo sputtering device (Denton Vacuum, USA). The average diameter of cross-linked and non cross-linked nanobers was measured directly from selected SEM images using ImageJ soware and is given as an average value for each sample calculated from at least 20 measurements.

Attenuated total reection-Fourier transform infrared (ATR-FTIR) analysis
The ATR-IR spectra of nanobrous samples were measured using a PerkinElmer FTIR System Spectrum GX Series-73565 at a scan range of 4000-650 cm À1 . A total of 32 scans were performed at all measurements with a resolution of 4 cm À1 .

Swelling studies
A weighed amount of dry electrospun nanobrous substrates was put in sphere lattice (sieve pouch) and immersed in water at 37 C for 24 h. Aer swelling, nanobrous substrates were taken out, le to drain for 1 min and wiped with lter paper to remove surface bound water and weighed again. Obtained sample weights were used to determine the water absorption capacity (WAC) swelling degree by the following equation: where WAC is the equilibrium water absorbency (%), W e and W d are the weight of swollen and dry nanobrous substrates, respectively.

Porosity determination
Nitrogen adsorption/desorption measurements were performed on a Micromeritics TriStar II 3020 porosimeter (Micromeritics, USA) using a Brunauer-Emmett-Teller (BET) model for surface area evaluation. All samples (before and aer crosslinking) were cut into small pieces and degassed with nitrogen gas at 40 C prior to the measurements to remove volatile compounds. Average values of BET measurements were calculated from three repetitions of each sample measurement. Statistical analysis. All numerical values are given as mean AE SD. Statistical analysis was performed using SPPS Statistics 25 (IBM Corp. Armonk, NY, USA). A one-way ANOVA followed by a post-hoc Bonferroni test was carried out. P-values < 0.05 were considered statistically signicant.

Quartz crystal microbalance with dissipation (QCM D)
Adsorption as a function of the resonance frequency change of an oscillating quartz sensor was monitored using the QCM-D technique (model E4 from Q-Sense, Gothenburg, Sweden). The QCM-D instrument determines changes in frequency (f) and dissipation (D) of an oscillating quartz crystal. Deposition of mass or changes in the rigidity of material on the crystal surface can be detected. Negative frequency shis (f) indicate a deposition of mass whereas positive dissipation shis (D) are caused by a reduced rigidity of the coating. The QCM-D measurements were conducted at the fundamental frequency of 5 MHz and its overtones. A detailed description of the QCM-D technique can be found elsewhere. 32,33 QCM-D Au-crystals coated with electrospun nanobers (cross-linked and non cross-linked) were mounted in the QCM ow cell. Aer establishing a stable change in frequency and dissipation shi in air, MilliQ-water was introduced into the QCM chambers for 30-45 min. The ow rate of water and temperature were kept at 0.1 mL min À1 and 37 AE 0.1 C, respectively, throughout all experiments.

In vitro drug release
In vitro drug release studies of DCF loaded electrospun nano-brous substrates were performed using an Automated Transdermal Diffusion Cells Sampling System (Logan System 912-6, Somerset, USA). Drug loaded samples were cut into 1 cm Â 1 cm and placed on the top of a sterilised polyethylene terephthalate (PET) mesh. Sterilization was performed using UV light with 30 minute exposure. The function of PET mesh was solely to assure constant wetting of the samples, as well as the prevention of their sinking into the Franz diffusion cell. The receptor compartment of the cells was lled with PBS (purchased from Sigma-Aldrich, Germany in the form of tablets) and its temperature was maintained at 37 C. During the dissolution testing the medium was stirred continuously with a magnetic bar. Samples were collected over a period of 48 h at different time intervals, while the released/dissolved concentration of DCF in the receptor medium was determined by UV-Vis spectrophotometer (Cary 60 UV-Visible Spectrophotometer, Agilent, Germany) by quantication of the absorption bands at 276 nm (characteristic absorption maxima for DCF). The withdrawn sample volumes were replaced by fresh PBS. Due to sample withdrawal, followed by sample dilution through media replacement, sink conditions were assured. In calculation of concentrations using the Beer-Lambert law, this dilution was accounted for. All release studies were performed in three parallels.
The DCF loading content of the electrospun substrates were determined by the decrease of DCF concentration using UV-Vis spectroscopy against a standard curve at 276 nm.

DCF loading content ð%Þ ¼
weight of the DCF in substrate weight of the substrate Â 100 (2)

Biocompatibility studies
The electrospun samples were cut to dimensions of 1 cm Â 1 cm and sterilised under the same procedure as described in Section 2.9. Advanced Dulbecco's Modied Eagle's Medium (DMEM) cell culture medium with 5 wt% fetal bovine serum (FBS) (Life technologies, Thermo Fisher Scientic Inc., Germany) was added to each sample and incubated for 24 h at 37 C in 5 wt% CO 2 atm. The samples (supernatants of the starting samples) were then added to a P96 microlitre plate with skin broblast cell culture (ATCC-CCL-110, Detroit 551, LGC Standards, UK). The following dilutions were used: 1 : 2, 1 : 4, 1 : 8, 1 : 16 and 1 : 32, each in 4 replicates. For the control, Advanced DMEM with 5 wt% FBS was used. The samples with cell cultures were incubated for another 24 h at 37 C in 5 wt% CO 2 atm. Aer 24 h the cells were stained with 0.1% crystal violet and the absorbance was measured at 595 nm. The obtained results were subjected to statistical analysis as described in Section 2.7.

Conductivity, viscosity and surface tension
It is well-established that the morphology and diameter of electrospun nanobers can be substantially inuenced by the properties of the electrospinning solution such as conductivity, viscosity and surface tension, etc. 34 Therefore, prior to the formation of nanobers, the said properties of CMC/PEG solution admixed with the cross-linker and catalyst (BTCA/ SHP) at different concentrations (3, 5, 7 and 10 wt%) were investigated and the results are presented in Fig. 1. The addition of BTCA to CMC/PEG solution at all four concentrations does not change the surface tension and the observed values are close to ca. 60 mN m À1 (ESI, Table S1 †). However, it does inuence the conductivity and viscosity of solution. The measured conductivity of CMC/PEG solution (BTCA_0%) is due to the -COO À groups of CMC. Further increase of conductivity of the solution with increasing BTCA concentration is because of the -COO À groups of BTCA; as the latter possess four -COO À groups in its structure. This phenomenon can efficiently prevent the formation of beads and increase the continuous spinning of uniform and small bre diameter of CMC/PEG due to increased charge density on the surface of the ejected jet in electrospinning eld. 35 The inuence of added BTCA on the viscosity of the CMC/ PEG solution can be clearly seen in Fig. 1. It is expected that the addition of BTCA would lower the pH of CMC/PEG solution from neutral to acidic, leading to reduced electrostatic repulsion between the CMC chains, and thus increased viscosity. Although the measured pH of the solution is 2.5, the viscosity of the solution is slightly reduced aer the addition of BTCA from 3 to 5 wt%. This can be explained, rstly, not all the -COO À groups of CMC are protonated at lower concentration of BTCA and secondly, due to the weak interactions and entanglements between the chains of PEG, or CMC or CMC-PEG, which is caused by the added BTCA molecules. This might allow the polymer molecules to align much easier. However, at higher BTCA concentration (7-10 wt%), the viscosity of the solution is increased, which can be explained by reduced solubility, electrostatic repulsion between the CMC chains, and increased interactions between CMC and PEG chains.

Morphology, diameter and surface area of electrospun nanobers
It is known that CMC, due to its carboxylic functional groups, increases the charge density of the solution in ejected jet. This, usually, causes strong elongation forces because of selfrepulsion of excess charge density under electric eld, leading to the formation of straight and regular nanobers with thinner diameter. 36 On the other hand, the uncharged polymer, PEG alone leads to the formation of bers of thicker diameter, owing to its solution's unbalanced charging and jet stability under the electric eld. 37 Clearly, creating a balance between the surface tension of ejected jet and electrical force is crucial to create a regular and uniform electrospun nanobers; which can be overcome by using the combination of CMC and PEG, as proven in our previous studies. 28,29 Fig. 2 shows the morphology of non cross-linked and cross-linked electrospun nanobers with 0, 5 and 10 wt% BTCA at 160 C. Results from other BTCA concentrations and temperatures, and drug incorporated samples are given in the supporting information (Fig. S1 †). A beadless, long and continuous nanobers are observed for BTCA-free and nonheated CMC/PEG (0 wt%) sample. Aer cross-linking with BTCA and incorporation of DCF (ESI, Fig. S2 †), the electrospun nanobers still kept their uniformity, morphology and structure. In addition, the presence of drug (ESI, Fig. S2 †) and BTCA (at all concentrations) on the ber's surface could not be seen, suggesting that the BTCA as well as drug molecules are homogeneously embedded into the CMC/PEG nanobers. However, the BTCA-free CMC/PEG electrospun substrates are water soluble (see Fig. 6), which makes them unsuitable for long-term applications such as drug delivery and cell growth. To surmount this downside, BTCA was added to CMC/PEG solution and subjected to thermal cross-linking as stated above.
Since studies such as cell growth and drug release can be affected by ber size, it is important to control the average diameter of ber; the formation of it is a function of materials choice and electrospinning processing parameter, for example cross-linker concentration, viscosity, etc. The effect of BTCA concentrations and temperatures on the ber's average diameter (Fig. 3A) and ber distributions (Fig. 3B-D) can be easily noticed. Firstly, for non cross-linked samples (white bar, Fig. 3A), the average ber diameter at BTCA_3% is slightly reduced compared to BTCA-free CMC/PEG electrospun nano-bers (diameter: 188 AE 4 nm) and then it increased to 178 AE 5 nm for BTCA_10%. This data ts well with the surface area results where the reduction of average ber diameter resulted in increased specic surface area and vice versa (Table 1). It seems that electrospinning solution concentration, one of the critical parameters, plays a key role in controlling the ber diameter. Since the concentration is directly related to viscosity (besides other parameters) of the electrospinning solution, it is suggested that the decreased viscosity and increased conductivity of the solution resulted in reduced average ber diameter, improved ber size, and formation of a thin and interconnected ber network/structures. 38 Secondly, in general, with increasing temperature from 120 C to 160 C, the average ber diameter is increased for each BTCA concentration except for BTCA_10%. Maximum average ber diameter (334 AE 6 nm) is obtained for BTCA_3% at 160 C comparing to non-heated sample of the same concentration, and BTCA-free (0 wt%) sample. It is obvious that cross-linking with BTCA and increase of temperature lead to increased average ber diameter, and also affected the distribution of ber diameters (Fig. 3B-D). This can be attributed to the conformational changes of the polymer chains, which are then linked together by ester-bonds, leading to nanobers with larger diameters. Aer cross-linking with BTCA_3%, the average ber diameter of electrospun sample does not increase substantially regardless of the temperatures. They, for example, remained almost constant (ca. 195 nm) at 160 C for each BTCA concentration. The BET analysis performed for electrospun nanobers before and aer thermal cross-linking with BTCA_3-10% at 160 C is shown in Table 1. The statistical analysis showed that the surface area and pore volume of non-crosslinked samples is signicantly different at higher BTCA concentration (5-7 wt%) compared to BTCA_0% (control sample). Whereas in the case of pore size no signicant differences are found for non cross-linked samples. Even though the reduction of surface area, average pore volume and pore size are obvious aer cross-linking with BTCA, the observed values of BET parameters (surface area to pore size) are not statistically different when the BTCA concentration is increased from 3 to 10 wt% than that of control sample. The reduction of pore size and pore volume for cross-linked samples can be explained that the junctions between the nanobers brought closer by the ester-bond linkage and the ber mat is denser as the result of cross-linking. This behavior is consistent with the previously reported papers. 39,40    (-CH 2 stretching), respectively. The peaks that are centered between 1000 and 1200 cm À1 are attributed to -C-OH stretching vibrations and the C-O-C glycosidic bond of CMC. 26,41 All characteristic peaks of PEG are observed at 1148, 1101, 1062, and 958 cm À1 (C-O-C stretching vibration), 1456 cm À1 (-CH 2 scissoring), 1341 cm À1 (-CH 2 deformation mode), 1240 cm À1 (-C-O stretching), and at 2880 cm À1 (C-H symmetric stretching), 42 respectively. BTCA gives a sharp absorption peak at 1690 cm À1 , corresponding to -C]O group of the carboxylic acid groups. In addition, the broad peak and sharp peak centered at 3000 cm À1 and 1400 cm À1 are due to -OH and -CH 2 stretching mode. The other two peaks centered at 1295 cm À1 and 933 cm À1 are associated to -C-O stretching and -OH out-of-plan deformation of H-bonded carboxylic groups. 43 Although all characteristic peaks of neat polymer are observed for non-heated CMC/PEG nanobers added with BTCA_10%, the -C]O peak at 1690 cm À1 from the contribution of BTCA is absent. This suggest that the cross-linker BTCA is well-shielded by the polymeric networks of CMC and PEG.

Infrared analysis
ATR-FTIR spectra of electrospun nanobers cross-linked with BTCA_10% at different temperatures (120-160 C) is shown in Fig. 4B. Comparing the mentioned spectra before and aer cross-linking, the emergence of several new peaks is obvious. The most noticeable change occurred in the region between 1500 cm À1 and 1800 cm À1 for the cross-linked samples. Along with the carboxyl carbonyl peak at 1590 cm À1 , a new peak arises at 1720 cm À1 , which is attributed to ester carbonyl peak. The latter peak is absent for both non-heated BTCA_10% and BTCA_0% samples (data not shown). Similar phenomenon is also noticed for other BTCA_x (x ¼ 3, 5 and 7 wt%, ESI, Fig. S3 †). The observed new peak at 1720 cm À1 conrms that the carboxylic acid groups of BTCA are crosslinked to hydroxyl groups of CMC and PEG via ester bonds. Comparing other reported works that employed toxic chemical agents for cross-linking and stabilization of nanobers, 27 the reported approach in this work possess certain advantages in terms of green nature, simplicity, and non-toxicity, respectively.
Interestingly, the intensity of peak at 1720 cm À1 is increased with the raise in temperature from 120 C to 160 C, an indication that more number of ester-bonds or ester-linkages are created between BTCA and CMC/PEG. As a result, we evaluated the ester cross-linkages semi-quantitatively by measuring the Fig. 4 (A) ATR-FTIR absorption spectra of neat PEG, Na-CMC and BTCA, and non-heated CMC/PEG nanofibers with added BTCA_10%. (B) Spectrum of electrospun nanofibers cross-linked with BTCA_10% at different temperatures, (C) the absorption peak intensity ratio (1720 cm À1 / 1594 cm À1 ) of electrospun samples cross-linked with different BTCA concentrations and temperatures, (D) spectrum of pure DCF and DCF incorporated electrospun nanofibers cross-linked with different BTCA concentrations at 160 C. ester carbonyl and carboxyl carbonyl peak intensity ratio (1720 cm À1 /1590 cm À1 ). 44 As shown in Fig. 4C, the ratios of the mentioned peaks intensity increased not only with increasing concentrations of BTCA but also with increased temperature, demonstrating that increasing of both parameters (concentration of BTCA and temperature) favors the esterication process. The intensity ratio is two-fold higher at 160 C for each BTCA concentration when comparing to samples cross-linked at 120 C. Results from these semi-quantitative data conrms that both CMC and PEG can be esteried or cross-linked easily and efficiently either by increasing of BTCA concentration or temperatures.
Also, there are changes in the absorption pattern of glucopyranose ring of CMC i.e. in the region 800-1450 cm À1 as can be seen in Fig. 4B. The observed peak at 1413 cm À1 (-CH 2 bending vibration) and 842 cm À1 (C-O-C stretching at b-(1/4) glycosidic bonds) is associated with the crystalline and amorphous structure of cellulose, as reported by several others. 45,46 The absorption peak intensity at 1413 cm À1 is increased while it is decreased substantially for the peak at 842 cm À1 with increasing temperatures. This behavior is more pronounced, especially, at 160 C, where the amorphous peak (842 cm À1 ) is almost disappearing. A most likely scenario is that a rearrangement of the amorphous and crystalline regions in the CMC nanobers structure occurred, leading to increased crystalline domains with concomitant decrease of amorphous domains upon exposing the CMC to higher temperatures.
ATR-FTIR spectrum of DCF incorporated samples is shown in Fig. 4D. For pure DCF, all characteristic absorption peaks are observed at 3253 cm À1 (-NH stretching of the secondary amine, supporting information, Fig. S3 †), 1571 cm À1 (-C]O stretching of the carboxyl ion), 1500 cm À1 (-C]C aromatic ring stretching), 1450 cm À1 (-CH 2 bending), at 1280 and 945 cm À1 (-C-O-C stretching), and at 745 cm À1 (-C-Cl stretching). 47,48 All samples cross-linked with different concentrations of BTCA showed the -C-O-C and -C-Cl stretching peaks of DCF at 1280 cm À1 and 745 cm À1 , respectively, indicating that DCF is successfully incorporated into the nanobers. 49 Also, the characteristic peaks of CMC and PEG are not shied, implying the absence of any chemical interactions of DCF with polymer functional groups, meaning that the drug molecules are physically incorporated into the nanobrous matrices. The physical incorporation of DCF has certain advantageous that it can be diffused and released into the medium without any hinderance from the polymer matrix.

Wettability, swelling and stability of electrospun nanobrous substrates
The wettability, water absorption capacity (WAC) and dimensional stability of electrospun nanobrous substrates crosslinked with BTCA_x (x ¼ 0, 3, 5, 7, 10 wt%) at 160 C were investigated. The static water contact angle SCA(H 2 O) of CMC/ PEG nanobrous substrate cross-linked with 0 and 10 wt% BTCA at 160 C was not possible to measure as they are too hydrophilic; as conrmed by the rapid absorption (ca. 1 s) of water droplet when placed it on the substrate (see ESI, Fig. S4 †).
The WAC of BTCA-free (0 wt%) nanobrous substrate, heated at 160 C, is not given in Fig. 5 since they are dissolved rapidly and completely when immersed in aqueous swelling medium. Interestingly, the WAC of nanobrous substrates is increased with increasing BTCA concentration, but it is more pronounced for cross-linked (BTCA_7-10%) samples. This seems that the nanobers (substrates) become more and more hydrophilic aer BTCA addition and cross-linking. According to Basu et al. the addition of PEG to CMC does not increase WAC at neutral pH. 50 This leads to the conclusion that the deprotonated COO À groups of CMC and BTCA, at neutral pH, contribute mainly to the water uptake. Besides to their high WAC (BTCA_7%: 2500%, BTCA_10%: 5500%) and hydrophilicity, the nanobrous substrates are stable and are not dissolved in water even aer 48 h (see the photo images, Fig. 5 inset). At lower BTCA concentrations (3-5 wt%), no remainders of the nanobrous substrates or any visible residues could be seen already aer 28 h, an indication that the cross-linking with lower amounts of BTCA (3-5 wt%) are inefficient, leading to a complete dissolution of CMC/PEG polymer chains. This is also in good agreement with IR results where the maximum ester carbonyl bonds are formed at higher BTCA concentrations (7-10 wt%) at 160 C (Fig. 4C).
In addition to the above conventional approach, a highly sensitive and complementary technique i.e., QCM-D, which is oen employed to study water interaction capacity of supported nano-sized cellulose lms, 51,52 was used to analyse the swelling and stability of the electrospun nanobers. In Fig. 6, the QCM-D change in frequency (A) and dissipation (B) for the electrospun nanobers, cross-linked with different BTCA concentrations at 160 C, upon interaction with water is shown. For BTCA-free sample (0 wt%), as soon as the water interacts with the sample surface a steep increase of frequency shi for the rst few minutes followed by a steady state (no change in frequency shi) already before 5 min is observable. The shi in dissipation, used to measure the rigidity of the adsorbed layer, in particular, is increased initially and then decreased to negative value (DD 3 : À44 Â 10 À6 ) followed by the steady state. An increase of frequency shi and decrease of dissipation shi upon contact with water indicates the removal of loosely attached nanobers to the QCM crystals or the dissolution of polymer nanobers or the combination of both. This is also reected for samples cross-linked with lower BTCA concentrations (3-5 wt% BTCA), where increased frequency shi is observed, but lower than the value observed for BTCA-free (0 wt%) sample. The increased dissipation shi can be related to the swelling of undissolved polymers, which remained on the crystal surface. These ndings can be further conrmed from the SEM results. The latter shows the formation of a thin layer of nanobers from BTCA-free sample (C), which is due to a shorter time (10 s), used for electrospinning of the solutions. This was done because the coating of nanobers on the QCM Au-crystals should be thin enough to guarantee proper sensitivity of the QCM-D. Aer exposure to water, the cross-linked nanobers are not visible anymore on the crystal surface for 3 wt% (E) and 5 wt% (F) BTCA as that of BTCA-free (0 wt%) non cross-linked electrospun samples (D). This conrms that the applied 3 to 5 wt% BTCA is still too low to create sufficient ester-linkages and form water insoluble CMC/PEG nanobers, as also noticed in the case of swelling and stability analysis (Fig. 5). Interestingly, for the samples cross-linked with higher BTCA concentrations, the negative frequency shi (7 wt%: À9 AE 2 Hz, 10 wt%: À209 AE 6 Hz) is observed while the increase of dissipation shi is positive when they were brought into contact with water. The observed negative frequency shi indicating that the nanobers are not dissolved and rmly attached to the crystal surface. This is further proven by SEM data where the nano-bers are still visible on the crystal surface in the case of BTCA_7% (G), and even more are present for BTCA_10% (H). Not like in the case of lower BTCA concentrations, the nano-bers are covalently and tightly attached together by maximum number ester-linkages at the highest BTCA concentration (7-10 wt%), leading to water insoluble materials. As the BTCA concentration is increased from 7 to 10 wt%, the negative frequency shi and positive dissipation shi is also increased, indicating that the swelling capacity of the nanober layer is increased, and thus the layer becomes increasingly viscoelastic.

In vitro drug release studies
Ideally, drug loaded wound healing materials should facilitate sustained drug delivery (for the time until wound dressing exchange) and maintain minimal drug-skin contact to overcome local irritation, itching and allergies. Therefore, the use of ECM resembling nanobrous substrates, which are in threedimensional (3D) form and stable at physiological environments, are benecial because of its capacity to effectively control diffusion and sustain the release of NSAID therapeutic molecules. The in vitro release of DCF in percentage (%) and cumulative amount (mg cm À2 ) as a function of time in PBS at pH 7.4 from nanobrous samples cross-linked with different BTCA concentrations at 160 C are shown in Fig. 7A and B. To verify if there is any interference of CMC/PEO nanobers in the absorbance of DCF incorporated samples, the UV-Vis spectra of BTCA_10% (drug free) and DCF_BTCA_10% were recorded for comparison and the results are shown in ESI (Fig. S5 †). The presence of no peak at 276 nm for BTCA_10% sample suggest that their interference in the DCF incorporated sample can be neglected. The drug loading content for both non cross-linked and all cross-linked samples were determined to be in the range of ca. 30%. Even though the release pattern of DCF is similar for all cross-linked samples from varying concentrations of BTCA, the DCF is released in several steps for each BTCA concentration. Release of DCF is due to the diffusion of water into the nanobrous matrix, leading to swelling of the polymer, as demonstrated by swelling studies (Fig. 5 and 6). Swelling of the polymer enabled the encapsulated drug to diffuse out and be released into surrounding media. In this context, highly swollen matrices should release higher amounts of the drug. All samples burst in the initial stages (rst few minutes) of drug release, which can be caused by the adsorbed DCF on the surface and pores of nanobers during the electrospinning process. 53 Even though the differences in the overall released mass of DCF (Fig. 7B) is small during 5-30 min (burst period), they continued to increase for each BTCA concentration till the end of the measurements (48 h). For example, the release mass of DCF is approximately 6% (0.0015 mg cm À2 ) for the rst 5 min. Aer 2 h, the calculated released masses of DCF are in the following order: 0.01505 (BTCA_3%) > 0.01355 (BTCA_5%) > 0.01196 (BTCA_7%) > 0.00983 (BTCA_10%). The released mass at this point is in the range of ca. 50% for each BTCA concentration. All cross-linked systems continue to release DCF and no steady-state is reached within 48 h. But the percentage release prole (Fig. 7A) showed that all incorporated DCF is released from all samples within this time frame, which corresponds to the time until most of the wound dressings are exchanged in clinical wound care. The overall released mass of DCF from the non cross-linked and cross-linked samples with different concentrations of BTCA at 160 C is depicted in Fig. 7C. It is obvious that the overall release rate (Fig. 7B) and mass (Fig. 7C-D) is higher at 3% BTCA (mDCF: 0.029 AE 0.002 mg cm À2 ) and slow down with increasing concentration of BTCA. The latter, also, resulted in the linear reduction of release rate/mass with a correlation coefficient of 0.99 (Fig. 7D). It is expected that the strongly swollen and hydrated CMC/PEG nanobrous substates, which are cross-linked with higher concentration of BTCA (7-10 wt%, see Fig. 6), would lead to more released mass of DCF. However, an opposite behavior is noticed as shown in Fig. 7B. The reduction of overall released mass as a function of BTCA concentrations can be related to the porous structure of nano-brous substrates. It is suggested that increased specic surface area and smaller pore size (Table 1), in addition to the increased dense network structure, limited the dissolution and diffusion of DCF through cross-linked nanobrous network of CMC/PEG. In addition, the cross-linking density of BTCA can inuence the release rate of DCF. It is proposed that the low cross-linking density at lower BTCA concentrations, as conrmed by ATR-FTIR results (Fig. 4C), facilitated maximum diffusion, and thus leading to quick and higher released DCF mass. This can be further supported by swelling studies where the samples cross-linked with 3-5 wt% BTCA dissolved rapidly upon contact with water (Fig. 5). In contrast, the electrospun samples with high cross-linking density, achieved using 7-10 wt% BTCA, maintained the morphology and structural integrity/network of the nanobrous substrates during the entire drug release period, allowing limited diffusion and release of DCF. This also means that the swollen 3D network of nanobrous substrates is not collapsed or dissolved (like in the case of BTCA_3% and BTCA_5%) upon contact with release medium. QCM-D stability studies and SEM results also conrmed that the morphology and 3D network of electrospun samples cross-linked samples with 7-10 wt% BTCA are retained when they are brought into contact with water. Overall, these results lead to the conclusion that the release rate or kinetics can be well-controlled by tuning the cross-linker concentration. On the other hand, while no linear trend can be found for the non cross-linked samples, the overall released mass is lower for 3-5 wt% BTCA, higher at 7% BTCA, and become comparable with 10% BTCA cross-linked samples (Fig. 7C). In addition to stability issue, the handling of non cross-linked electrospun nanobrous substrates is difficult especially in release or cell culturing medium. Therefore, no further studies have been carried out for non crosslinked samples.
The release data was further evaluated and tted by linear regression analysis according to three different kinetic models to understand the release mechanism of DCF from BTCA crosslinked CMC/PEG nanobrous substrates. The kinetic models such as the rst-order (eqn (3)), Higuchi-model (eqn (4)) and Korsmeyer-Peppas model (eqn (5)) gave the best t and the highest square of correlation coefficient (r 2 ) value for the release data out of the ve kinetics models, which were considered for the data tting. The latter was done for the release data of DCF from 0-300 min. When compared with the other above mentioned two kinetic models, the rst order model resulted in the lowest r 2 value, and therefore not considered to explain the release kinetics of DCF. As can be seen in Fig. 8 B-C and Table 2, the tting results show that the release kinetics t is more consistent with Higuchi model following the Korsmeyer-Peppas model. While the Higuchi model gave the best t and the highest r 2 value (0.966-0.985) for all BTCA cross-linked samples, the Korsmeyer-Peppas model showed the best t and the highest r 2 value (0.973-0.976) only up to 5 wt% BTCA crosslinked samples. Moreover, the calculated r 2 value is slightly lower for the latter model. Overall, the obtained best ts and the highest correlation coefficient values for both models indicate that the tting results can be correlated well with the real relationship between the response data parameters.
In general, the First-order kinetics is employed to relate the drug release where the release kinetics are concentration depended.
where Q is the amount of released drug in time 't', Q 0 is the initial concentration of the drug and K is the rst-order rate constant.
The Higuchi model uses pseudo-steady-state assumptions to describe the release kinetics of drug from several types of pharmaceutical forms (e.g., porous matrix) based on Fick's law, showing a square root of time dependent process. 54 where Q is the amount of drug released at time 't' and K is the Higuchi constant. On the other hand, the Korsmeyer-Peppas model, derived from Fick's law, is mainly used to describe the release kinetics of drug from polymeric. This model can be dened as 54 Where Q t and Q f is the amount of drug released in time t and released aer innitive time, K is the kinetic constant and the exponent 'n' is the release characteristic index, which is used to characterize the release mechanism. If the exponent 'n' is less than 0.45, then the release mechanism is Fick diffusion and it is primarily based on drug diffusion. The drug release and the release mechanism are non-Fick diffusion control when 0.45 < n < 0.89. When n > 0.89, the drug release is dominated by the combination of both diffusion and erosion-controlled process. For cross-linked samples with 3-5 wt% BTCA, the highest r 2 value is obtained with Higuchi model, conrming the t 1/2 dependence of the drug release, which is purely the characteristic of Fickian diffusion mechanism. This means that the drug release is predominantly controlled by diffusion. On the other hand, the probability of drug release also via non-Fickian diffusion cannot be ruled out since the nano-brous substrates cross-linked with lower BTCA concentration (3-5 wt%) were degrading within the time period of drug release. This can lead to release of drug simultaneously and rapidly from both the surface and bulk part of matrix, meaning that as the nanober is degraded or eroded, the drug is released from the entire volume of the matrix. In such a case, the drug is free to be released through diffusion mechanisms as well, in addition to the erosion controlled, as it is supported by the diffusional exponent 'n' values derived from the Korsmeyer-Peppas model, which are above 0.45. Also, the several breaks observed in the rst derivatives proles (Fig. 7D) derived from the release data depicts that multiple mechanisms are involved in the course of DCF release. This means that release of DCF from the hydrophilic electrospun nanobers is caused not only by a "simple" diffusion-controlled mechanism but is oen additionally accompanied by a combination of swelling followed by erosion of polymer chains when come into contact with a buffer solution. For 7 and 10 wt% BTCA cross-linked samples, which are stable and highly swollen, the release of drug is clearly the Fickian i.e. diffusion-controlled type as it is revealed from the highest r 2 value according to the Higuchi model and from the exponent 'n < 0.45' value.

Biocompatibility studies
The nal set of experiments in this study was related to the testing of the prepared nanobrous substrates potential to be used in biomedical applications, especially in treatment of wounds or skin tissue repair. For this purpose, we tested the inuence of cross-linked nanobers with different BTCA concentrations on the human skin derived broblast viability (based on crystal violet dyeing). Of course, the same test was initially used to evaluate the biocompatibility of the materials, but we immediately saw that these pose no harm to the used cells, but even promoted cell growth. The initial set of experiments was performed using the standard procedure to dilute the initial samples using the binary system (1 : 2, 1 : 4, 1 : 8,.), but since the sample have proven biocompatible already or the lowest of the dilutions (1 : 2), only these are included here for discussion (Fig. 9). All other results are shown in the supporting information (see Fig. S6 †). From Fig. 9 it can be immediately observed that the viability of cells is signicantly improved when exposed to pure DCF compared to control sample (pure cell growth media -ADMEM + 5 wt% FBS). Similarly, the DCF incorporated BTCA_7% and BTCA_10% samples, which were stable water ( Fig. 5 and 6), showed signicant differences in the viability in comparison with control and other CMC/PEG samples. Since it is known that cells "prosper" in more stable environments (with the smallest possible material degradation during their growth/exposure), 55 the  observed higher viabilities in the more cross-linked (highly stable) samples, seems logical. It is proposed that the highly cross-linked nanobrous materials with BTCA_7% and BTCA_10% are less prone to degrade in the cell growth media than those of samples cross-linked with 0 to 5 wt% BTCA (that degraded or dissolved in water within 28 h). Obviously, the less degradation products, the higher the viability, which indirectly suggests that the degradation products of the prepared electrospun materials negatively affects the cell viability (although still no toxic effects were observed for neither of the samples). Based on the same amount (30%) of incorporated drugs in all samples, it is assumed that the drugs in the nanobers did not signicantly inuence the cell viability as that of BTCA at higher concentration. Finally, all the results prove the suitability of the prepared materials to be further tested in relation to their potential application in either dermal wound healing or in skin tissue engineering applications in general.

Conclusions
Needles electrospinning technique was used for the fabrication of diclofenac (DCF) incorporated and stable nanobrous substrates from the aqueous solution containing carboxymethyl cellulose (CMC), polyethylene glycol (PEG), and 1,2,3,4-butanetetracaboxylic acid (BTCA) respectively. Increasing, especially higher, concentrations of BTCA and temperatures favored the formation of ester-bonds and cross-linking of nanobers, since the BTCA is activated more in the presence of sodium hypophosphite monohydrate catalyst at higher temperature. The thermally cross-linked samples showed decreased average ber diameters, pore volumes and pore sizes, whereas the specic surface area of the nanobers are increased. The stability and swelling capacity of the cross-linked nanobrous substrates increased with increasing BTCA concentrations as proven by swelling and quartz crystal microbalance studies. Although a fast (burst) release followed by a diffusion controlled release of DCF are observed, the release kinetics are of the same type for all cross-linked nanobrous samples. A linear decrease in the overall released mass is observed during the period of release time with increasing BTCA concentration, although the opposite behavior is anticipated. Results from the release data tted using three kinetic models showed that the release of DCF follows the non-Fickian diffusion mechanism i.e. both diffusion and erosion controlled process at the lowest BTCA concentration based on the exponent 'n < 0.45' value derived from the Korsmeyer-Peppas model, although the Higuchi model gave the highest correlation coefficient r 2 value. At the highest BTCA concentration, a clear diffusion control (Fickian mechanism) is observed according to the calculated 'n' value which is lower than 0.45. Even though no proliferation of broblast cells is induced, neither the polymers, cross-linker and nor the high concentration of drug decreased the viability of cells, a positive sign that the applied components are biocompatible. Overall the obtained results demonstrated that the developed stable electrospun nanobrous system should have high advantages for the effective incorporation of all types of hydrophilic therapeutic molecules and give them a prolonged and controlled release for at least two days, so that the entire system can be even applied to treat chronic wounds, and in regenerative medicine in general.

Conflicts of interest
There are no conicts to declare. Fig. 9 Viability of human skin derived fibroblasts after exposure to the DCF incorporated electrospun nanofibrous substrates cross-linked with different concentrations of BTCA at 160 C. The shown results were calculated relative to control (pure cell growth media). The dashed red lines correspond to the calculated confidence intervals for the control sample. Values are expressed as percentage of the means AE SD. Statistical significance was defined as *P < 0.05 compared to control sample (ANOVA test).