Characterization of Nanofibers for Tissue Engineering: Chemical Mapping by Confocal Raman Microscopy

Abstract

Nanofiber scaffolds are used in bioengineering for functional support of growing tissues. To fine tune nanofiber properties for specific applications, it is often necessary to characterize the spatial distribution of their chemical content. Raman spectroscopy is a common tool used to characterize chemical composition of various materials, including nanofibers. In combination with a confocal microscope, it allows simultaneous mapping of both spectral and spatial features of inhomogeneous structures, also known as hyperspectral imaging. However, such mapping is usually performed on microscopic scale, due to the resolution of the scanning system being diffraction limited (about 0.2 – 0.5 micron, depending on the excitation wavelength). We present an application of confocal Raman microscopy to hyperspectral mapping of nanofibers, where nanoscale features are resolved by means of oversampling and extensive data processing, including Singular Value Decomposition and Classical Least Squares decomposition techniques. Oversampling and data processing facilitated evaluation of the spatial distribution of different chemical components within multi-component nanofibers.

Graphical Abstract

1. Introduction

Raman spectroscopy has been a useful tool for material characterization for decades. Raman scattering involves interaction of an incident photon with a molecule, and therefore it is sensitive to vibrational transitions present in a sample; it allows chemical analysis by nondestructive optical means, without reagents, staining, or sample preparation required. As a result, it has been used in a wide variety of applications, such as investigation of biocompatible polymers [1], textile fibers for forensic purposes [2], strain-induced conformation transitions in silkworm fibers [3], carbon-based materials, including carbon nanotubes [4–6], and bismuth black in pigments [7].

Scaffolds made from polymer nanofibers are used in tissue engineering for functional support of growing tissues [8]. Various nanofiber properties (polymer content and distribution, fiber thickness, mechanical stiffness, and so on) can have a significant effect on cell behavior and tissue development [9–10]. These parameters would, ideally, be fine-tuned for a specific application. However, nanofiber synthesis is a complex process affected by both production settings and environmental factors [11], and therefore it is necessary to characterize fibers generated under new conditions.

As a method of nanofiber characterization, Raman spectroscopy typically has been used in combination with imaging by scanning electron microscopy (SEM) [12–19], transmission electron microscopy (TEM) [13, 16, 20–23], x-ray diffraction (XRD) [13] or atomic force microscopy (AFM) [9]. Such studies either rely on sample homogeneity when the spectral signature is examined [12–17], or use the non-chemical properties of different fiber components, such as density [23] or surface roughness [9], to determine their spatial distribution.

Recent improvements in instrumentation lead to the development of confocal Raman microscopy (CRM), a technique of chemical imaging that combines confocal scanning with Raman spectroscopy. Spatial maps produced in this manner contain Raman spectra, rather than just light intensity, for each pixel of the image. Spatial resolution of CRM is diffraction limited, and ranges about 200 – 500 nm, depending on the wavelength of the excitation laser (typically 400 – 1000 nm). A number of studies that used CRM for chemical imaging of micro- and nanoscale structures, for example, cotton fibers [24], carbon nanotubes [25–27], and cells [28–30]. However, these experiments did not attempt to resolve any features smaller than diffraction limit.

For nanofibers with complex structure, confirming the spatial distribution of their chemical components is highly nontrivial. Usually, it is impossible to distinguish different components visually, and indirect methods, such as described in [9] and [23], may not be applicable. More specifically, the use of contact angle measurements [9] and Transmission Electron Microscopy (TEM) measurements [23] require that the polymers that are contained within the nanofibers have different hydrophobicity or density. CRM, on the other hand, only requires that their spectra are sufficiently different, which is almost always the case with nanofibers made out of different polymers.

Tissue engineering provides a renewable tissue source for transplantation [33]. It is a developing field combining interdisciplinary areas of engineering, biology, chemistry and the synthetic tissue growth. Scaffolds made from polymer nanofibers are used in tissue engineering for functional support of growing tissues [8]. These scaffolds are designed to promote the functional growth and development of complex tissues through physiological imitation [34]. Multiple studies have documented the use of nanofibers for bone regeneration [35], wound healing [36] and drug delivery [37].

However, there are challenges in fabricating a scaffold that promotes the cell attachment and polarization. Also, it has not been yet possible to produce a material which is durable enough to keep its structural integrity, and at the same time maintain the degradation rate corresponding to the time frame of the developing tissue. When scaffolds are designed, there are requirements that depend upon tissue-specific applications, such as mass transport properties, mechanical properties and biological interactions [38].

The common scaffolding techniques used in tissue engineering are micro-particles [39], electrospinning [8, 18], hydrogels [40] and polymer sponges [41]. The fabrication of nanofibers through electrospinning process creates a two-dimensional environment that yields high surface area, which mimics the basement membrane of in vivo tissue [42–43]. The electrospinning method has proven very effective for many difficult tissue engineering applications, such as heart [44]. Typical polymers used in electrospinning are polymethylmethacrylate (PMMA), polyvinyl alcohol (PVA), polylactic acid (PLA), and polylactic-co-glycolic acid (PLGA). The PLGA is a member of the polyurethane family, used both for comparatively soft and hard tissues (stiffness ~20MPa to 2.0GPa, varying with the fiber diameter). Various nanofiber properties, such as polymer content and distribution, fiber thickness, and mechanical stiffness, can seriously affect cell behavior and tissue development [9–10]. These parameters should be fine-tuned for a specific application. However, nanofiber synthesis is a complex process, affected by both production settings and environmental factors [11], and therefore it is necessary to characterize fibers generated under new conditions.

Here, we concentrate on applications of confocal Raman microscopy to analyze the structure of inhomogeneous nanofibers, consisting of different types of polymers, which is increasingly popular for tissue engineering applications. Confocal Raman microscopy allows direct chemical detection of various polymers, but the resolution of this technique tends to exceed the nanofiber thickness. As a result, just mapping the nanofibers using the CRM is insufficient, and further data processing (chemometrics) is required to extract the information about the component distribution within nanofibers [31–32]. In those cases, the nanoscale features can be resolved by means of oversampling and extensive data processing, including Singular Value Decomposition (SVD) and Classical Least Squares (CLS) techniques.

2. Materials and Methods

2.1. Raman micro-spectroscopy

Raman spectroscopy is a very effective technique used to identify various materials by their chemical composition [45]. It is based on inelastic scattering of light by molecules, yielding information about the vibrational modes, structure, and other properties of the sample. This non-destructive technique that can be employed in real time, with high selectivity and sensitivity [46]. Its utility and performance have been enhanced by integration of the modern optical and photonic hardware, such as fiber optics, miniaturized lasers and CCD cameras [47–49]. Raman spectroscopy allows chemical analysis by optical means, without reagents, staining, or sample preparation. As a result, it has been used in many diverse projects, such as investigation of biocompatible polymers [1] and carbon-based materials, including carbon nanotubes [4–6].

HORIBA’s LabRAM HR Evolution confocal scanning systems with 473 nm excitation laser was employed to collect the spectral data from pure PGS fibers and PLGA polymer films, and then the core/shell fiber samples. HORIBA’s XploRA Plus confocal scanning microscope (excitation wavelength 532 nm) was used to generate spectra of blend and emulsion nanofibers.

Linear scans were made either along (0.1 μm step) or across (0.05 μm step) individual fibers.

2.2. Chemical Mapping

Since the size of the nanofibers is approximately on the order of the size of the scanning beam, it is generally impossible to directly image the fiber structure. Moreover, the fiber will be contained within the entire confocal depth of the microscope. Thus, the detected nanofiber structure in this case will be essentially one dimensional (for a line scan), where the entire thickness of the fiber will generate Raman signal. Therefore, one needs to assume that the fiber cross-section is spherically symmetric (see Fig. 1).

Figure 1.

Laser beam scan across a nanofiber (left to right): (A) the beam begins to overlap with the shell, producing the shell Raman signature; (B) the beam begins to overlap with the core, producing the Raman signatures from both core and shell polymers; (C) the beam passes the core, so only shell Raman signature is produced; (D) the beam passes the fiber, so neither of the polymers are detected. The fiber and the laser beam are shown in a cross-section.

If a core/shell nanofiber [32] is scanned across with TEM00 Gaussian laser beam, as the beam mover from the position shown in Fig. 1A to the position shown in Fig. 1B, the fiber will first produce Raman signal from only the shell. In practice, the Raman signature of the shell polymer appears above the noise in the spectra, once there is enough interaction of material for the beam to produce the signatures. As the beam is scanned across the fiber, the shell signal gradually increases. Then, when the fiber core is also illuminated by the laser beam, the Raman spectra will contain both the spectrum of the shell and the spectrum of the core. Once the laser beam passes the core completely, it will only detect the signal from the shell again (Fig. 1C). Finally, when the laser beam no longer overlaps the shell (Fig. 1D), no spectra from either polymer are observed. From Fig. 1 one may deduce that if this detection modality is applied, the thickness of the fiber is the distance between the beam position in Figs. 1A and 1D, minus the laser beam waist diameter (spot size). Since the physical thickness of the fiber can be determined from SEM measurements, one can calculate the “actual” laser spot size from the beam positions which, when barely overlapped with the nanofiber, are capable of generating detectable Raman signal. This information can then be used to determine the thickness of the core within the nanofiber – a parameter of crucial importance for the fiber characterization that cannot be determined by any other approach.

2.3. Preparation of PGS/PLGA nanofibers

The main steps of this procedure were previously reported by us [31]. Briefly, Poly (glycerol-sebacate) (PGS) and Poly (lactic-co-glycolic) (PLGA) polymer solutions were independently drawn through a co-axial spinneret capillary, which are then spun to generate nanofibers with a core of PGS and shell of PLGA. The shell material essentially acts as a driving force for the core material to form a fiber structure, which allows incorporating both materials with different properties into nanofibers. The nanofibers were electrospun and deposited on a substrate, producing a fiber mat.

2.4. Preparation of elastin-PLGA nanofibers (Control)

The electrospinning solution was prepared as 4% elastin/4% PLGA (w/w) in hexafluoroisopropanol (HFIP) and sodium chloride (1% w/w) was added to avoid beading. Using a micro-syringe pump connected to a 3-mL syringe, the electrospinning solution was pumped through a (25 G) needle at a rate of 6 μL/min, and a distance between the needle tip and the collector ground of 15 cm. High voltage (14 kV) was applied to both the needle tip and aluminum foil-coated connector. The fibers were collected on 13 mm glass coverslips on foil.

2.5. Homogenous incorporation of growth factor by blend electrospinning (Blend)

One commonly used method to incorporate biologically active ingredients into nanofibers is by Blend electrospinning, which results in homogenous distribution of all ingredients within the fibers. Here, the epidermal growth factor (EGF) was incorporated into elastin-PLGA fibers as follows: 100 μg of EGF from murine submaxillary gland was reconstituted in 0.2 mL of 1% BSA in PBS to obtain a final concentration of 0.5 μg/ μL of EGF. A volume of 30 μL of EGF/BSA solution was added to a mixture solution of 0.75 g 12% elastin in HFIP and 0.75 g PLGA in HFIP to yield a final concentration of 4% elastin/4% PLGA and 10 ng/mg of EGF. Sodium chloride (1% w/w) was added to the solution before electrospinning to minimize beading. The solution was electrospun in a similar fashion to the control nanofibers.

2.6. Internal incorporation of growth factor by emulsion electrospinning (Emulsion)

Emulsion electrospinning usually yields internal incorporation of the biologically active ingredients. The EGF was incorporated into elastin-PLGA fibers by emulsion electrospinning as follows: 30 μL of (0.5 μg/ μL) solution of EGF in 1%BSA solution was added dropwise to 0.5 g of chloroform solution containing 2% (w/w) ethyl cellulose and mixed by vortexing for 5 minutes to yield a creamy emulsion. Ethyl cellulose (EC) was used as a surfactant. This emulsion was introduced dropwise to a mixture solution of 0.5 g of 12% elastin/HFIP solution and 0.5 g of 12% PLGA solution. The emulsion was further mixed by vortexing for 5 minutes. The final concentrations are 10 ng EGF/1 mg emulsion, 4% elastin (w/w) and 4% PLGA (w/w). The emulsion system was electrospun using the same parameters as blend electrospinning. This process was expected to produce fibers with embedded “islands” of EGF.

2.7. SEM Characterization of the scaffolds

Scanning electron microscopy (SEM) imaging of the scaffolds was carried out using a Zeiss 1550 field emission scanning electron microscope (Leo Electron Microscopy Ltd., Cambridge, UK; Carl Zeiss, Jena, Germany). The scaffolds were mounted on 1 cm² stubs and coated with approximately 5 nm of gold-palladium to minimize sample charging. Images were captured using an in-lens detector, 1–5 kV acceleration voltage and a working distance of 2–6 mm. The microscope annotation software was used to apply scale bars, and ImageJ software was used to measure the fiber diameters from calibrated images. At least 4 scaffolds of each type were imaged and 200 nanofibers analyzed for average fiber diameter measurements.

3. Results

3.1. Core/shell nanofiber characterization by confocal Raman microscopy with nanoscale resolution

PGS/PLGA (polyglycerol-sebacate / polylactic-co-glycolic acid) nanofibers represent an important subset of fibers that are used for tissue engineering applications. The PGS core enhanced biocompatibility, while the PLGA shell improved mechanical properties of the polymer scaffold. Thus, the availability of simple and non-destructive method of fiber mat characterization is one of the key requirements for a successful manufacturing of tissue engineered constructs [31, 50].

When the laser beam was focused on the fiber center, the intensity of Raman signal from both PLGA and PGS components was at a maximum, because the amount of each polymer was highest at this location. Also, the fiber was scattering the greatest amount of Raman signal back towards the detector. Away from the center, the intensity of the Raman signal decreased rapidly.

We have collected the spectra pure PLGA fibers and PGS polymer films, as shown in Fig. 2A. Since, in the core/shell PGS/PLGA fibers, PGS was embedded inside the fiber, detecting its presence there was much harder. As a proof of concept, we have purposely altered the fiber manufacturing process to increase the amount of PGS within a fiber. This resulted in the formation of PGS-rich “droplets” along the fibers (Fig. 2B). Upon measuring the Raman spectra from these PGS-rich areas, we indeed observed a clear signal from both polymers, as indicated in Fig. 2C. Specifically, the spectrum obtained from this area contained both 872 cm⁻¹ and 908 cm⁻¹ Raman peaks. While the former peak is associated with PLGA from the shell, the latter is due to PGS in the center of the fiber.

Figure 2.

(A) Raman spectra obtained from pure PGS fibers and PLGA films; (B) the image of the fiber with PGS-rich “droplet” in the center of the core/shell fiber (scale bar is 2 μm); (C) Raman spectrum obtained from the center of the core/shell fiber, demonstrating that both polymers are present in the “droplet” spectrum.

Figure 3 shows the results of Raman-based analysis applied to a core/shell nanofiber used for polymeric scaffolds. Figure 3A shows the optical image of the fiber. While the amount of PGS was much lower than what is seen in PGS-rich area in Fig. 2, it is still detectable in the center of the fiber (see Fig. 3B). In order to determine the distribution of each polymer within the nanofiber, we first located the PGS signal from the core, then moved the laser beam to one side of the fiber until no PGS (908 cm⁻¹ peak) was detected. The fiber was then scanned across, and the spectra at each step were collected and plotted as a two-dimensional hyperspectral image, as seen in Fig. 3D. Here, the spectrum intensity is plotted for the relevant part of the spectrum.

Figure 3.

Raman spectroscopic imaging of a nanofiber: (A) the optical image of the nanofiber, showing the scan area (scale bar is 2 μm); (B) spectrum from the center of the fiber, showing presence of both PGS and PLGA, and spectrum from the edge of the fiber, where the PGS peaks are absent; (C) Singular Value Decomposition (SVD) scatter plot, showing clear separation of core and shell spectra; (D) hyperspectral Raman image of the scan across the fiber (x-axis: relevant portion of the Raman spectrum, y-axis: laser beam position as it is scanned across the fiber) and the structure of the fiber based on the SVD analysis.

In order to process this hyperspectral data set, we applied the Singular Value Decomposition (SVD) analysis. SVD is a well-known method of partitioning the spectra from a hyperspectral image into distinct groups, based on the major contributing spectral line shapes (components) [51]. Once the spectra have been classified, the assigned groups are mapped back onto the one-dimensional scan line, to identify regions of different chemical composition within the sample [52]. Figure 3C shows each spectrum plotted as a single point on an SVD scatter plot (i.e. the magnitude of one SVD component versus another). For analysis, we limited the spectral region to the 800–1000 cm⁻¹ range, since this is where the greatest difference between the spectra of the two polymers was observed. SVD components 1 and 2 were found to closely mimic the sum of PGS and PLGA spectra, as they included both 872 cm⁻¹ and 908 cm⁻¹ peaks. SVD component 3, however, did not contain 908 cm⁻¹ peak, and thus was associated with PLGA shell only. Figure 3D shows the mapping of the spectra (hyperspectral image) on the fiber, and the components identified using SVD.

This method was used to calculate the thickness of the PLGA shell and the PGS core, even though the features of the fiber could not be directly resolved because of the large laser spot size. The width of the core and shell regions on both sides of the fiber appeared wider by the laser spot size (about 500 nm), since the shell region spectrum was still present, even if the laser spot touched the fiber only in part. Despite these difficulties, the core size was determined using the SVD-based map of the core and shell regions, together with the total fiber diameter measurements obtained by SEM. The core diameter in Fig. 3 was about 75% of the fiber thickness (400 nm), or approximately 300 nm.

3.2. Localization of epidermal growth factor in double emulsion electrospun nanofibers

A solution containing a blend of epidermal growth factor (EGF), ethyl cellulose (EC), elastin and PLGA was electrospun to produce nanofibers for tissue engineering applications. Another type of nanofiber was made by adding an emulsion of EGF and EC to a solution of elastin and PLGA, and emulsifying the resulting solution again before electrospinning. These nanofiber scaffolds were used to control the delivery of EGF to salivary gland cells. The average nanofiber diameter was about 0.37 μm. Controls - droplets of pure EGF, EC and PLGA on aluminum foil - were used to measure the Raman spectra of individual nanofiber components.

Pure PLGA, EC and EGF spectra (Fig. 4A) were used as basis components in the Classical Least Squares (CLS) linear regression analysis [53]. CLS fitting is based on the assumption that a spectrum of a mixture of ingredients is a combination of the spectra of pure ingredients. It is most suitable for the samples of known composition and known spectra of pure components. The method fits a complex spectrum with the linear sum of the spectra of individual components (basis) by finding the coefficients that minimize the difference between the target spectrum and the fit in the least squares sense. These coefficients represent contributions of each component. The spectrum from the center of the blend nanofiber, shown in Fig. 4B, contained all three materials, as can be seen from the CLS fit (the sum of PLGA, EC and EGF components weighed by their respective contributions) closely matching the actual measured spectrum (blend fiber). The CLS analysis was performed at each point of a linear scan across the blend fiber. Fig. 4C, representing the changing contribution of pure components to the blend fiber spectrum at different points of the scan, showed relatively smooth increase of all components towards the fiber center, corresponding to the amount of fiber material enclosed within the excitation volume, without drastic changes in the chemical content within the blend. These results confirmed that the blend electrospinning produced nanofibers with uniform distribution of EGF within the fiber.

Figure 4.

(A) Normalized Raman reference spectra of PLGA, EC and EGF films; (B) spectrum from the center of the blend fiber, containing all three materials, and the result of CLS fit based on PLGA, EC and EGF components, (C) CLS analysis of the scan across blend nanofiber, indicating a nearly uniform blend of PLGA, EC and EGF components. 3-point moving average was applied to reduce the random noise.

The emulsion fiber (Fig. 5A) was expected to contain “islands” of EGF embedded within. CLS analysis of a linear scan across the emulsion fiber (Fig. 5C) showed that one side of the scanned section did not contain EGF, while the other side of the fiber clearly included traces of EGF. Fig. 5B shows the scan taken along the emulsion fiber. The variation of the EGF component (Fig. 5D) is a sign of the presence of EGF islands along the fiber. These results indicated that emulsion electrospinning embedded EGF “islands” within the fibers.

Figure 5.

Raman mapping of emulsion nanofibers: (A) Optical image showing sample surface mapping recorded perpendicular to the emulsion fiber (scale bar is 5 μm); (B) Optical image showing sample surface mapping recorded along the emulsion fiber (C) distribution of the CLS component EGF, indicating the presence of an EGF “island” in the emulsion. The arrows indicate the location of the boundaries of the fiber (based on the EC signature, which is not shown); (D) distribution of the CLS component EGF, indicating the presence of multiple EGF “islands” in the emulsion. 3-point moving average was applied to CLS component to reduce the random noise.

4. Discussion

Confocal Raman microscopy (CRM) combined with data processing algorithms can be successfully applied to the investigation of sub-micron structure of polymer nanofibers used as synthetic scaffolds in tissue engineering applications. The presence of core and shell material in the PGS/PLGA nanofibers was confirmed by direct observation of Raman signatures, while the core/shell distribution within the fiber, as well as the spatial extent of the core, were determined with the help of the SVD algorithm, despite the fact that these features of the fiber could not be resolved directly by the confocal Raman microscope. Also, the spatial distribution of the EGF load within different types of electrospun nanofibers was determined using the CLS regression.

In some cases, nanofiber characterization produces unique challenges, such as 3D imaging of nanofiber composites or structures. Due to the scale difference between the structures and the nanofiber features that have to be resolved, and the limited ability of most techniques to image in depth, one has to sacrifice resolution, volume, or time it takes to obtain the images, to optimize 3D image quality and throughput [26]. The methods described here allow us to circumvent these limitations to image small-scale differences in nanofiber composition. A disadvantage of CRM is that even with the methods described here, the resolution is still limited to hundreds of nanometers, and nanofiber scaffolds can have even smaller diameters. To summarize the advantages of CRM are that it is effective in imaging thin core-shell nanofibers and structures when the core and shell polymers are of the of similar density or have very similar hydrophobicity. It also does not require a vacuum, so the nanofibers can be analyzed with non-destructive imaging with cells grown on them, which is essential for assessing biological responses to engineered scaffolds.

5. Conclusions

Nanofiber characterization is crucial to confirm and fine tune various nanofiber properties that optimize nanofiber scaffolds for tissue growth. It can be performed by different techniques, depending on the parameters that have to be determined. In most cases, a combination of techniques is used to find all required nanofiber properties.

CRM is of particular interest, since it can be used for 3D mapping of chemical composition of multi-component nanofibers, by simultaneously acquiring both spectral and spatial information about the sample. Such mapping is usually performed on microscopic scale; however, nanoscale resolution can be achieved by oversampling and data processing (for example, using SVD). In many cases, this is the preferred way to confirm the spatial distribution of different chemical components within multi-component nanofibers, which is particularly important for nanofiber scaffolds used in bioengineering for functional support of growing tissues.

Nanofiber scaffolds are used for functional support of growing tissues.

Raman spectroscopic mapping can be used to study the structure of nanofiber scaffolds.

Nanoscale features can be resolved by oversampling and extensive data processing.

Oversampling/data processing confirms the structure within multi-component nanofibers.