Spatially-offset Raman spectroscopy for monitoring mineralization of bone tissue engineering scaffolds: feasibility study based on phantom samples

Using phantom samples, we investigated the feasibility of spatially-offset Raman spectroscopy (SORS) as a tool for monitoring non-invasively the mineralization of bone tissue engineering scaffold in-vivo. The phantom samples consisted of 3D-printed scaffolds of poly-caprolactone (PCL) and hydroxyapatite (HA) blends, with varying concentrations of HA, to mimic the mineralisation process. The scaffolds were covered by a 4 mm layer of skin to simulate the real in-vivo measurement conditions. At a concentration of HA approximately 1/3 that of bone (~0.6 g/cm3), the characteristic Raman band of HA (960 cm−1) was detectable when the PCL:HA layer was located at 4 mm depth within the scaffold (i.e. 8 mm below the skin surface). For the layers of the PCL:HA immediately under the skin (i.e. top of the scaffold), the detection limit of HA was 0.18 g/cm3, which is approximately one order of magnitude lower than that of bone. Similar results were also found for the phantoms simulating uniform and inward gradual mineralisation of the scaffold, indicating the suitability of SORS to detect early stages of mineralisation. Nevertheless, the results also show that the contribution of the materials surrounding the scaffold can be significant and methods for subtraction need to be investigated in the future. In conclusion, these results indicate that spatially-offset Raman spectroscopy is a promising technique for in-vivo longitudinal monitoring scaffold mineralization and bone re-growth.


Introduction
A common treatment for major bone damage is autologous bone grafting [1]. Autologous grafting is not the optimal solution because the quality and quantity of bone grafts that can be harvested is sometimes not sufficient to meet demand, it increases the risk of infection, and can lead to haemorrhaging, cosmetic disability, nerve damage, and a loss of function. An alternative to autologous grafts is the use of tissue engineered scaffolds that can be implanted in the defect to offer a 3D structure to support and stimulate the regeneration and repair of the bone [2]. In vivo models where scaffolds are implanted in critical bone defects in large animals (e.g. sheep) have been commonly used to model the healing process in humans [3]. These studies provide a better understanding of the bone repair process in order to optimise the physical and chemical properties of the material to reduce healing time and improve the quality of the newly formed bone. For a critical bone defect, the damage to the bone is so great that the body fails to repair the bone correctly [4]. What will constitute a critical defect depends on the size of the bone damaged [5,6], and the age and health of the patient [7,8].
When a defect reaches this critical size, the body fails to fill the defect with extracellular matrix (e.g. collagen) and to mineralise; instead, the exposed damaged bone is repaired leaving an indent or hole in the tissue with repaired sides, leading to bone tissue that is weaker than before the damage [9]. When using a scaffold, the quality of the repaired bone depends on the ability of cells to populate and remodel the scaffold. Thus, longitudinal data regarding mineral deposition within the scaffold is desirable and important for optimizing the physiochemical properties of the scaffold.
Typically, the quality of the repaired bone is evaluated by end-point histological tests. However, histology is destructive and therefore can be used only at the end time-point (typically 4-6 weeks after implantation). 3D micro-computed tomography (μCT) is commonly used to analyze the morphology and mineral density of newly formed bone in animal models [10], including for in-vivo longitudinal studies [11,12]. μCT has also been combined with single photon emission computed tomography (SPECT) in order to obtain more detailed molecular information during the course of bone formation and remodeling [13]. Although this technique requires radioactive SPECT probes, a study using synthetic hydrogel scaffolds implanted in critical size calvarial defects generated in mice, showed that in-vivo longitudinal data regarding morphology and bone density agreed with end-point histological and μCT evaluations [13].
Raman spectroscopy (RS) is a non-destructive spectroscopic technique that has high chemical specificity and does not require exogenous labels or probes [14]. RS has been widely used for the analysis of bone tissue [15][16][17] and bone tissue engineering scaffolds [18][19][20][21][22]. Spatially-offset Raman Spectroscopy (SORS) is a variant RS technique that is able to recover molecular information of bone in-vivo transcutaneously [23,24]. Recently, we demonstrated the feasibility of using SORS to measure Raman spectra of hydroxyapatite (HA) powder buried in layers of polymer and ceramic tissue engineering scaffolds as thick as few millimetres, covered by 1 mm thick skin layer [25]. While these feasibility studies indicated the potential of SORS for measuring in-vivo longitudinal data from small animal model studies, a better understanding of the SORS signals is required in order to optimise the instrumentation for in-vivo measurements (i.e. a hand-held probe) and support the data analysis. By increasing the maximum offset of the device and having control over the size of the collection points the sensitivity of the device was optimised.
Here we have developed a series of phantom samples to mimic the mineralisation of scaffolds implanted in a large animal critical bone defect and the measurement conditions for in-vivo longitudinal study. The samples consisted of 3D printed composite scaffolds with polycaprolactone (PCL) and hydroxyapatite (HA) microparticles, for which the concentration of HA varied to simulate different degrees of mineralisation [26]. The experiments were carried out using a table-top SORS instrument based on a digital micro-mirror device (DMD) [27], that allowed flexible adjustments of the spatial offsets in order to optimise the measurement conditions and develop the design of a future fibre-optics SORS probe that could be used in real animal studies.

Spatially-offset Raman spectroscopy (SORS) instrument
The DMD-based SORS instrument was equipped with a 785 nm wavelength laser (Xtra II, Toptica). A 100 mm focal length 2-inch diameter lens was used to focus the laser beam on the sample (120 mW power, spot size ~0.5 mm) and to collect the backscattered Raman photons. After passing through a dichroic filter that blocked the elastically scattered photons, the Raman photons were focused with a lens on a software-controlled DMD (size 14.4 mm x 8.8 mm, resolution 1920 x 1080 pixels, DLP6500 Texas Instrument). As the DMD was located in a plane conjugate to the sample, it allowed the selection of multiple spatially offset collection points (0. 22 Fig. 2(a)-Multiple h a square hole (10 mm x 10 mm x 10 mm) which mimicked the critical defect. Behind the Teflon slab, a 5 mm thick layer of polystyrene (PS) was placed in order to give an indication of whether the femur bone at the back of the defect would contribute to the measured SORS signal. After the insertion of the scaffolds (sizes 10 mm x 10 mm x 2mm) in the Teflon hole, the phantom sample was covered by a layer of pig skin (sourced from a local retail outlet) cut to 4 mm thickness, with a large enough surface area to cover the face of the phantom. While this phantom closely resembled the in-vivo measurement conditions, the materials used were selected to have similar light scattering properties to bone [2] but distinctive Raman bands to allow an understanding of the contributions of various regions of the phantom to the measured SORS spectra. This is particularly important for SORS measurements where different parts of the sample can be probed depending on the value of the spatial offset. Figure 2(e) presents the Raman spectra of each material used for the phantom. The spectrum of HA has a strong band at 960 cm −1 (phosphate band) [15][16][17]. The PCL has Raman bands assigned to the C-O-C vibrations as a triplet at 1067 cm −1 , 1098 cm −1 , and 1110 cm −1 , and CH 2 bands at 1300 cm −1 and 1445 cm −1 [4]. Teflon and PS have strong Raman bands at 734 cm −1 and 1004 cm −1 respectively. Therefore, the use of these two materials for the bone phantom effectively allows us to identify the contributions of the sides (Teflon) and back (PS) of the critical defect to the measured SORS spectrum without interfering with our measurement of the 960 cm −1 band from HA. Because the molecular composition of pig skin changes with depth, Raman spectra were measured from the top epidermis/dermis part (bands at 970 cm −1 and 1300 cm −1 assigned to collagen) and from the adipose tissue at the bottom part (strong 1368 cm −1 band) [30,31].

Data analysis
First, all spectra were normalised to minimum 0 and maximum 1. The difference between the SORS spectra of the PCL:HA scaffold and PCL-only scaffold (control sample) at each offset value was then calculated using an in-house iterative algorithm. To eliminate errors caused by small baseline or intensity variations (likely due to small differences in optical scattering properties between samples), a correction factor in the form of a 2nd order polynomial was included in the subtraction algorithm. This polynomial was determined by minimising the square difference between the two SORS spectra in an iterative method. The spectral region 930-980 cm −1 containing the main HA band at 960 cm −1 was excluded from the minimisation. A number of 30 iterations was used, as this was observed to lead to stable solutions in all cases.  These resu be used to qua a given spatial h the increased PCL:HA ratio he intensity of scaffold cont he intensity of ever, for a give ignal) was con g. 3, for a given m PCL:HA sca cm −1 assigned   Fig. 7 presents the calculated ratio between the intensity of the 960 cm −1 band in the difference spectra (PCL:HA minus PCL-only scaffolds) and the 1445cm −1 band in the spectra of the sample containing HA (I 960 /I 1445 ). The results show that the relative intensity of the 960 cm −1 band increased when the concentration of HA and the spatial offset increased. The highest sensitivity and detection resolution for HA was observed when using the 4 mm spatial offset. However, once the outer PCA:HA layer of the scaffold reached the 1:4 ratio, the I 960 /I 1445 ratio seemed to plateau, regardless of the value of the spatial offset (higher than 0), indicating a decrease in detection resolution for HA.

Conclusions
The use of tissue engineering scaffolds for stimulating bone re-growth in critical bone defects is a promising way to improve bone healing; however, monitoring bone growth in situ remains a challenge. Using phantom samples based on 3D-printed PCL:HA scaffolds we investigated the feasibility of SORS to monitor mineralisation of bone tissue engineering scaffolds in large animal models. The tests investigating the detection limits for mineralisation showed that SORS is able to detect HA concentrations at an order of magnitude lower than that found in living bone, even through a 4 mm thick layer of skin (mimicking in-vivo transcutaneous measurements). These low concentrations are only detected when the HA was located immediately under the skin surface. As the HA concentration increased so did the depth at which the HA was detectable. For the highest concentration of HA, the detection depth increased to 4 mm. Bone has a higher concentration of HA than any of the scaffolds investigated here and a 2 mm thick layer produced a stronger signal across all depths as seen in Fig. 4. The I 960 /I 1445 band ratio can be a useful parameter to attempt longitudinal quantification of HA concentration, but would require a means to measure the thickness of the skin layer, which can vary during the 4-6 week duration of the real in-vivo experiments. Nevertheless, such changes in skin thickness may be measured using complementary techniques, such as ultrasound imaging.
For the experiments mimicking the uniform scaffold mineralisation, setting a spatial offset larger than 2 mm allowed sensitive detection of HA. Similar results were observed for the inward gradual mineralisation model. Nevertheless, the highest sensitivity and detection resolution for HA was observed when using the 4 mm spatial offset (largest offset possible with current instrument), which indicates that even the earliest mineralisation stage included in this study was detectable (i.e. only the outer 2 mm layer of the scaffold contained HA at a concentration ~10x lower than bone). The results also show that the surrounding walls of the bone defect also contributed to the measured SORS spectra (Teflon signal), that may overlap the Raman bands from the HA in the scaffold. For this reason, it would be advantageous to start the acquisition of SORS spectra as soon as the scaffold was implanted in order to establish a baseline SORS spectrum and quantify the changes in the HA signals during the 4-6 week period of bone re-growth. The ability to follow the bone healing process on the same animal will provide higher quality data with ethical and economic benefits from reducing the number of animals used during the research.

Funding
The Engineering and Physical Sciences Research Council [grant number EP/L025620/1]).