**Impact of Dual Cell Co-culture and Cell-conditioned Media on Yield and Function of a Human Olfactory Cell Line for Regenerative Medicine**

**Rachael Wood 1,2, Pelin Durali <sup>1</sup> and Ivan Wall 1,2,3,\***


Received: 4 March 2020; Accepted: 10 April 2020; Published: 12 April 2020

**Abstract:** Olfactory ensheathing cells (OECs) are a promising candidate therapy for neuronal tissue repair. However, appropriate priming conditions to drive a regenerative phenotype are yet to be determined. We first assessed the effect of using a human fibroblast feeder layer and fibroblast conditioned media on primary rat olfactory mucosal cells (OMCs). We found that OMCs cultured on fibroblast feeders had greater expression of the key OEC marker p75NTR (25.1 <sup>±</sup> 10.7 cells/mm2) compared with OMCs cultured on laminin (4.0 <sup>±</sup> 0.8 cells/mm2, *<sup>p</sup>* <sup>=</sup> 0.001). However, the addition of fibroblast-conditioned media (CM) resulted in a significant increase in Thy1.1 (45.9 <sup>±</sup> 9.0 cells/mm2 versus 12.5 <sup>±</sup> 2.5 cells/mm2 on laminin, *<sup>p</sup>* <sup>=</sup> 0.006), an undesirable cell marker as it is regarded to be a marker of contaminating fibroblasts. A direct comparison between human feeders and GMP cell line Ms3T3 was then undertaken. Ms3T3 cells supported similar p75NTR levels (10.7 <sup>±</sup> 5.3 cells/mm2) with significantly reduced Thy1.1 expression (4.8 <sup>±</sup> 2.1 cells/mm2). Ms3T3 cells were used as feeder layers for human OECs to determine whether observations made in the rat model were conserved. Examination of the OEC phenotype (S100β expression and neurite outgrowth from NG108-15 cells) revealed that co-culture with fibroblast feeders had a negative effect on human OECs, contrary to observations of rat OECs. CM negatively affected rat and human OECs equally. When the best and worst conditions in terms of supporting S100β expression were used in NG108-15 neuron co-cultures, those with the highest S100β expression resulted in longer and more numerous neurites (22.8 ± 2.4 μm neurite length/neuron for laminin) compared with the lowest S100β expression (17.9 ± 1.1 μm for Ms3T3 feeders with CM). In conclusion, this work revealed that neither dual co-culture nor fibroblast-conditioned media support the regenerative OEC phenotype. In our case, a preliminary rat model was not predictive of human cell responses.

**Keywords:** olfactory ensheathing cells; spinal cord injury; neural regeneration; cell therapies

#### **1. Introduction**

Spinal cord injury (SCI) is a devastating injury to the central nervous system (CNS) that affects 250,000–500,000 new people worldwide every year. People with an SCI are 2–5 times more likely to die prematurely [1], and lifetime costs are estimated between US\$1 and US\$5 million, excluding indirect costs such as a loss in wages [2]. The majority of patients will suffer partial or complete paralysis [3] and can also suffer from chronic neuropathic pain syndromes, which have a serious effect on the quality of life [4]. Following SCI, inflammation and cell death ensue, and eventually, scar tissue forms, which is the main impediment to spontaneous regeneration, as the axons that do sprout and re-grow are unable to path-find through the scar to reach their target [5–9].

Unlike the rest of the body, the CNS does not regenerate. An exception is the olfactory system, which retains its ability to regenerate throughout adult life, due to the presence of a special type of glial cell, the olfactory ensheathing cell (OEC) [10]. These cells have been studied for potential use in spinal cord repair [11] due to their natural role in regenerating and guiding olfactory receptor neurons from the peripheral nervous system (PNS) into the CNS. Mucosal OECs can be biopsied via a minimally invasive intranasal approach, and although mucosa OECs tend to have lower yields and purity, they are more clinically attractive than their counterparts, bulb OECs, which are found in the lining of the brain [12].

Characterisation of OECs is commonly based on the expression of the neurotrophic receptor, p75NTR, in cell populations isolated from the olfactory mucosa or bulb [11,13–15]. p75NTR is a receptor that induces neurite outgrowth and cellular survival [16]; however, it is not a definitive marker for OECs [17]. In addition to p75NTR, glial cell marker S100β is commonly used as a positive marker for OECs [18], and Thy1.1 and fibronectin (Fn) are used as markers for contaminating fibroblasts [11,19–21].

For an OEC therapy to be developed, more understanding is needed of how to control the growth of functional subsets that express glial markers p75NTR and S100β and also how accurately these putative identity markers are predictive of function. If we better understand the markers that predict regenerative potential, we can then make strides toward enriching cells expressing those markers.

Due to the rapid onset of damage that occurs during an SCI, surgical options have to be enacted within 24 h of the injury [22], and by three weeks, the majority of the damage has occurred, so it is important to administer any cell therapy before this point [23]. A key issue with OECs is that their yield and purity variance cannot be predicted. Due to this, allogeneic cell therapy is seen as the most promising option as an off-the-shelf therapy that can be administered in the vital period before severe secondary damage sets in [24,25]. With the variability between patients, there would be no guarantee that enough functional cells could be produced in the required time frame, let alone all the release testing carried out to ensure the patient's safety. Previously, our group generated conditionally immortalised human mucosa-derived OEC cell lines [26] toward the goal of producing an off-the-shelf therapy for SCI. In this study, we aim to compare our human cell line against primary rat cells to determine the best conditions for OEC culture and identify the relevance of animal studies in this area.

#### **2. Materials and Methods**

#### *2.1. Primary Cell Isolation and Culture*

Mucosae were isolated following the optimum protocol previously identified [27]. The mucosae were dissected from three adult female Sprague-Dawley 200–250g rats, which were euthanized by carbon dioxide asphyxiation (Schedule 1 method [28]) according to the U.K. Animals (Scientific Procedures) Act 1986. Each rat provided two mucosae. Each mucosa was placed in DMEM/F12 media to be transported to the laboratory. The mucosae were placed in Hanks' Balanced Salt Solution (HBSS, Gibco Life Technologies, Gaithersburg, MD, USA) with 1% P/S (10,000 units penicillin, 10 mg streptomycin/mL) in a petri dish and washed by gently wiping each mucosa across a spatula. After washing, the mucosae were placed on a petri dish and cut up into small pieces, then placed in a 5 mL dispase II (2.4 units/mL, Sigma-Aldrich, UK) solution for 45 min at 37 ◦C in order to digest the tissue enzymatically.

The mucosa and dispase solution was placed in the centrifuge for five minutes at 400× *g*. The supernatant was discarded, and the cells and tissue were re-suspended in 5 mL collagenase (Type I solution, 0.05%, Sigma-Aldrich, UK) for 15 min at 37 ◦C. Every five minutes, the collagenase was taken out of the incubator and mechanically triturated briefly before being placed back in the incubator. At the 15 minute point, the collagenase suspension was placed in the centrifuge for five minutes at 400× *g*. The cells were re-suspended in 7 mL DMEM/F12 media (2% FBS, 1% P/S) in a T25 flask at 37 ◦C in 5% CO2 for 24 h. All cell cultures were carried out at 37 ◦C and 5% CO2.

The cells were placed in a tissue culture flask for 24 h as a differential adhesion step. The purpose of this step was to decrease the amount of contaminating fibroblasts in the culture. Fibroblasts have a faster adhesion time than OECs. After 24 h, most fibroblasts will adhere to the tissue culture plastic, whereas OECs will still be in suspension [27]. After 24 h, the suspension was replated onto laminin-coated (20 μg/mL, Sigma-Aldrich, UK) wells. The wells were coated for four hours at 37 ◦C. The laminin was then removed, and the cells were plated while the matrix was still wet.

#### *2.2. Cell Line Generation and Culture*

The feeder layer (HuG418) was prepared from cells isolated from the human olfactory mucosa. The cells used were thawed from a pre-existing cell line prepared by Dr. Melanie Georgiou (UCL). The protocol followed by Dr. Melanie Georgiou was summarised in the paper by Pollock et al. [29]. The human mucosa OEC line was generated using the same technology; however, the purification steps were carried out before immortalisation to ensure a purer OEC population. The generation and culture conditions of this line (PA5) were covered by Santiago-Toledo et al. [26], and the culture conditions remained the same for HuG418.

#### *2.3. Feeder Layer Generation*

Feeder layers were prepared by removing the media from the T75 flask and replacing with 4 mL mitomycin C (MMC) (Sigma-Aldrich, UK) for 2 h at 37 ◦C to inactivate the cells. After two hours, the MMC was removed from the cells, and the cells were washed with PBS (Lonza). The cells were removed from the flask using Trypsin/EDTA (Sigma) and placed in the centrifuge for five minutes at 400× *g*. After discarding the supernatant, cells were re-suspended in DMEM/F12 and plated up at <sup>12</sup> <sup>×</sup> <sup>10</sup><sup>3</sup> cells/cm2. This density was used as it gave a good coverage of the well whilst allowing space for the OECs on the well surface. Feeders remained in the incubator for two days before any cells were plated on them.

#### *2.4. Collection of Conditioned Media*

Conditioned media (CM) was collected from HuG418 cells after two days of culture. The media was centrifuged for five minutes at 400× *g*. CM was mixed at 1:1 with fresh media and added to the required conditions.

#### *2.5. Culture of PA5 Cells*

For the experiments involving PA5 cells, wells were coated either with Ms3T3 feeders, poly-l-lysine (PLL), or laminin. PA5 cells were plated at 6000 cells/cm2 and cultured for five days. After five days, cells were fixed and stained for S100β and fibronectin. Neurotrophic factor 3 (NT-3) was added to some media conditions at 50 ng/mL as the literature indicated that NT-3 encourages the neurological repair phenotype in OECs [27].

#### *2.6. In Vitro Co-culture with NG108-15 Neurons*

NG108-15 neurons were either grown alone on Ms3T3 feeders (as a control) or on a layer of PA5 cells that were seeded onto either laminin or Ms3T3 feeders. NG108-15 neurons were seeded at 500 cells/cm<sup>2</sup> onto the laminin-coated plates or feeder layers. Co-culture was carried out for five days with a media change occurring on Day 3 following the protocol laid out by Santiago-Toledo et al. [26]. Cells were fixed with 4% PFA at room temperature (RT), and cells were stained to detect β-III tubulin.

#### *2.7. Immunocytochemistry (ICC)*

Cells were fixed with 4% paraformaldehyde (PFA, Sigma-Aldrich, UK) for 20 min at RT. The PFA was removed, and the cells were washed three times with PBS. Three washes with PBS were carried out in between every step of the immuno-staining in order to remove and dilute any traces of the previous reagent. One wash was defined as filling the well with PBS and leaving for five minutes. Zero-point-two-five percent Triton-X (Sigma-Aldrich, UK) was added to each well and left for 20 min at RT. All solutions were made up in PBS. Five percent goat serum solution (Dako) was used as the blocking solution for 30 min at RT. Primary antibodies were prepared at a dilution of 1:200 and added for 90 min at RT. The primary antibodies used were as follows: rabbit anti-p75NTR (Millipore), mouse anti-Thy1.1 (Millipore), rabbit anti-S100β (Dako), mouse anti-fibronectin (Sigma-Aldrich, UK), and mouse anti-βIII tubulin (Sigma-Aldrich, UK). The negative control was subjected to the same treatment as other wells, the exception being the exclusion of the primary antibody. The secondary antibodies (DyLight® 488 or 594, goat anti-rabbit IgG (H + L) and goat anti-mouse IgG (H + L), Vector Laboratories, Burlingame, CA, USA) were prepared at 1:200 and Hoechst (Sigma-Aldrich, UK) at 1:1000 and incubated for 45 min at RT. Fluorescent imaging was carried out on an EVOS FL microscope (Life Technologies).

Five images were taken per well and were taken in a cross formation around the centre of the well. Two to three wells were stained per condition. After imaging, each channel was examined to assess whether the images taken were representative of the well. Brightness and shutter speed were set by identifying a cell that was determined to be positive and ensuring the background was dark to prevent overexposure of the cells. These settings were held constant between images for an experiment to ensure the images could be directly compared.

Images were analysed using ImageJ. A colour threshold was set to identify a positive cell, and only cells that were identified to be above this threshold were counted as positive. Positive cells were counted, and the results were calculated as the proportion of cells positive for the marker and the yield of positive cells over the imaged area.

#### *2.8. Circularity Analysis*

A macro was written in ImageJ in order to analyse the morphology of the cells. Circularity was used as the defining value. The macro was written in JavaScript to automate a calculation method that already exists in ImageJ. The "adjust threshold" window opens automatically when the macro is run and can be adjusted to select the area of interest. Any cell that was larger than a set threshold (this value can be adjusted depending on whether the nucleus or the whole cell is being analysed) was discarded to prevent cell clusters being included in the calculation. In addition to this, any cells on the boundary of the image were discarded so they did not contribute to false readings. The circularity was calculated as in Equation (1), where *A* is the area (pixels2) and P is the perimeter (pixels).

$$\text{Circularity} = \frac{4A\pi}{P^2} \tag{1}$$

A circularity of 1 indicates a perfect circle, and a circularity of 0 indicates a straight line. When the macro was run, a printout of the circularities was displayed. These data were exported to GraphPad® and plotted as a histogram.

#### *2.9. Neuronal Growth Analysis*

According to the literature, five days of co-culture are optimal for observing enhanced neurite outgrowth compared with three days of co-culture [27]. After five days of co-culture, cells were labelled with antibodies against βIII-tubulin. Neuron number, neurite number, and neurite length were quantified using the NeuronJ plugin in ImageJ [30]. Neurites were traced and measured within the software and exported to GraphPad® for analysis.

#### *2.10. Statistical Analysis and Data Accessibility*

Data are presented as the mean ± standard error of the mean (SEM). For the majority of the data, one-way ANOVA was used to determine statistically significant differences, and the Bonferroni correction was carried out using GraphPad® software to calculate *p*-values. When there were only two sets of data to compare, the Shapiro–Wilk test was used to test for normality. When data were normally distributed, the Tukey test was used to test for significance. When data were not normally distributed, the Kruskal–Wallis test was used. The Bonferroni correction was not used for sets of data that only compared two conditions as there was no need to correct for multiple comparisons. For the comparison of histograms, the Kolmogorov–Smirnov test was carried out.

In this work, one experimental repeat (n) was defined as cells taken from different flasks/groups of animals.

Statistics are only reported in the text when there are significant differences at α = 0.05. In all figures, a single asterisk (\*) indicates *p* < 0.05, two asterisks (\*\*) *p* < 0.01, and three asterisks (\*\*\*) *p* < 0.001.

The datasets generated and analysed during this study are available from the corresponding author upon reasonable request.

#### **3. Results and Discussion**

#### *3.1. Human Feeders Encourage an Increase in p75NTR and Spindle-Shaped Cells in Rat OECs*

The ICC was quantified using yield (positive cells per mm2), but not purity, as the presence of the feeders in only some conditions would make any assessment of purity misleading. It can be observed from Figure 1B,E,H,K,M that the addition of CM significantly increased the expression of Thy1.1. When CM was added to OECs cultured on feeders, an increase in Thy1.1 over and above the higher expression induced by feeders with standard media was observed (25.7 <sup>±</sup> 12.4 cells/mm<sup>2</sup> on feeders with CM from 14.5 <sup>±</sup> 4.8 cells/mm<sup>2</sup> on feeders with standard media). The increase in Thy1.1 expression observed on laminin with CM (45.9 <sup>±</sup> 9.0 cells/mm2) indicated that HuG418-derived CM affected Thy1.1 expression more when the feeders themselves were not present. This could be due to the feeders supporting themselves. Where feeders were present, those cells could uptake some of the soluble factors present in the CM, leaving a lower concentration of soluble factors for the OECs. This would mean that there were more soluble factors in the media when the feeders were not present. Fibroblasts participate in paracrine signalling [31,32], and therefore, it follows that if they are not present to receive the factors, the factors present can significantly assist other cells present.

p75NTR was present in all conditions at a similar level, with the exception of laminin with standard media, which was significantly lower at 4.0 <sup>±</sup> 0.8 cells/mm2. The presence of CM increased the expression of p75NTR on laminin (19.7 <sup>±</sup> 7.6 cells/mm2). An increase in p75NTR expression without the presence of the feeders would imply that there was some form of paracrine signalling occurring. Whether this was due to FGF2 expression or a combination of other factors would require a deeper examination of the components in the CM. Without the higher expression of Thy1.1, CM with laminin would be a promising condition to take forward. Thy1.1, historically an undesirable marker of fibroblast phenotype in the OEC field, negated the positive effect of the increase in putative OEC marker p75NTR as it implied a higher level of fibroblast impurities that would need to be removed from the culture.

The morphology of S100β-positive cells was analysed using an automated circularity macro in ImageJ (Figure 1N–Q). Numerical analysis revealed that there was no significant difference between the distributions for feeders versus feeders with CM, and both conditions had a left skew (skew = 0.4 and 0.5, respectively). The laminin condition had a skew value of 0.3 and a kurtosis value of −0.8. When CM was added to the laminin condition, the distribution was significantly different from the standard laminin condition (Kolmogorov–Smirnov, *p* = 0.01). The distribution with CM had a less flattened distribution (kurt = −0.3 compared to −0.8) and a more pronounced left shift (skew = 0.7 compared to 0.3). This was a stronger shift than the feeder conditions. This was significant as this was the condition with the most Thy1.1-positive cells. The most likely option was that spindle-shaped OECs were also expressing Thy1.1, as well as S100β. Co-labelling of these antibodies was not carried out in this study due to the lack of consistency experienced with the separate S100β antibody, which meant the Thy1.1 and S100β antibodies used were raised in the same species. The literature indicated that it is possible that OECs do co-express p75NTR and Thy1.1 [33–36] and Thy1.1 and GFAP [34], so the assumption that Thy1.1 is a contaminating cell marker may need to be

re-examined. Due to the lack of putative OEC markers [37,38], these cells are poorly characterised, and this presents one of the major challenges of OEC research.

**Figure 1.** Fluorescent micrographs of primary rat olfactory mucosal cells (OMCs) cultured on laminin (**A–F**) and HuG418 feeders (**G–L**) in the presence (**D–F**, **J–L**) and absence (**A–C**, **G–I**) of HuG418 conditioned media and stained for olfactory ensheathing cell (OEC) biomarkers p75NTR and S100β and fibroblast marker Thy1.1. Positive cells were counted in ImageJ and calculated as the number of positive cells over the image area (**M**). OMCs cultured on laminin with standard media had the lowest yield for p75NTR, and the addition of conditioned media caused an upregulation of undesirable marker Thy1.1. Circularity was used as a measurement of morphology, and positive S100β cells were analysed for their circularity (**N–Q**). Cells on laminin with standard media and feeders with conditioned media gave more enlarged cell morphologies. The scale bars represent 400 μm. Data are the means ± SEM., n = 3. CM, conditioned media.

Although a differential adhesion step was used to remove rapidly adherent cells, leaving the slow attaching OECs in culture, the primary culture could still harbour contaminating cell types. This is demonstrated in Figure 2, where an entire well of OECs on feeders was imaged at 10× objective magnification (EVOS Life Technologies AMF4300) and the resulting images merged together using PanoramaPro2 in order to gain an understanding of the culture as a whole. It was observed that there were distinct areas of Thy1.1-positive colonies. When images were taken for counting, these images could yield data from 100% p75NTR-positive to 0% p75NTR-positive. This wide range of values resulted in a large standard deviation and therefore a large standard error of the mean across independent experiments (seen Figure 1M). These large errors were due to stochastic variability in the cultures due to the different cell subsets present, and these different cells could predominate in different regions of a single dish. From a cell manufacturing perspective, this created a significant challenge for creating a robust and well-characterised product.

**Figure 2.** Assembled fluorescent micrograph of a cell population derived from rat olfactory mucosa cultured on a human feeder layer of olfactory fibroblasts. Images were taken at 100× total magnification in a 24 well plate (diameter = 15.6 mm) and stitched together using PanoramaPro2. It can be seen from this stitched image that there are definite populations present in the well plate. There are distinct areas of Thy1.1 (red) that do not appear to be dispersed with the areas of p75NTR (green). This variety in present populations explains the large error bars found in the cell counts. When multiple views were taken for counting, the purity could range from 0–100%, which relates to a large standard deviation.

After staining, there was also a red/yellow ring around the periphery of the plate that was not associated with any positive cells. On close inspection of the well plates used, it was found that the bottom of the wells was rarely smooth, and often, there were circular scratches, presumably as a result of the manufacturing process. It was postulated that the antibody interacted with this non-smooth surface or was trapped in ECM deposited as a consequence of cell responses to the surface scratches and gave the circular staining [39]. This pattern was observed in several images; however, it was not until the full well was imaged that it was noticed that it made a complete circle.

#### *3.2. Comparison of Di*ff*erent Feeder Layers*

One striking observation in the S100β (detecting OECs) and Fn (fibroblasts) staining condition comparing Ms3T3 and HuG418 feeders (Figure 3) was that the OECs did not appear to grow directly over the feeder layer. Instead, they grew close to the fibroblasts (Figure 3C,D). This placement made sense if the cells were gaining benefits from cell-to-cell contact. If paracrine factors were involved, the OECs would not necessarily border the fibroblasts so closely [40].

No differences in p75NTR expression were observed on the different feeder layers (Figure 3A,B,G) with yields of 15.9 <sup>±</sup> 6.8 cells/mm2 and 10.7 <sup>±</sup> 5.3 cells/mm2 on Ms3T3 and HuG418, respectively. This may indicate that the OECs were benefitting from general cell-to-cell contact, not specifically anatomically-matched HuG418 cells. However, as both lines were comprised of fibroblasts, there could also be a common paracrine factor involved. OECs have been well documented to grow poorly in isolation, and when a low yield is obtained, they do not proliferate [41–43]. Additionally, they have been observed to survive better in a culture with a mixed population (normally OECs and olfactory

fibroblasts) [10,44–47]. This would also explain why the HuG418 CM did not have the anticipated beneficial impact on OECs' p75TNR expression as hoped as the cells themselves may be required. In the literature, it has been observed that there is a close relationship between OECs and olfactory fibroblasts. However, it does appear that this relationship is due to the physical contact as opposed to any paracrine factors present [48].

**Figure 3.** Cells were fixed after 14 days of culture and stained for OEC biomarkers p75NTR and S100β and fibroblast biomarkers Thy1.1 and Fn (**A–D**). Positive cells were quantified in ImageJ and graphed as the yield of positive cells in the image area (**E–G**). Circularity was used as a measurement of morphology, and positive S100β cells were analysed for their circularity (**H**,**I**). The scale bars represent 400 μm. Data are the means ± SEM, n = 3.

Thy1.1 expression was found to be significantly different between the two feeders (Kruskal–Wallis, *p* = 0.01). As observed in the ICC images and quantification, the presence of Ms3T3 downregulated the expression of Thy1.1 (yield of 14.6 <sup>±</sup> 4.1 cells/mm<sup>2</sup> on HuG418 and 4.8 <sup>±</sup> 2.1 cells/mm<sup>2</sup> on Ms3T3). This was important as it indicated that the Ms3T3 feeders were superior to the HuG418 when it came to supporting OEC growth and expression of p75NTR. OECs appeared to be supported by cell-to-cell contact and did not gain any benefit from paracrine factors. These results suggested that Ms3T3 feeders were superior to HuG418 cells as they resulted in fewer Thy1.1-positive cells whilst maintaining similar p75NTR expression levels.

These results, together with the results obtained with CM, indicated that OECs benefitted from cell-to-cell contact rather than paracrine signalling by HuG418 cells. Ms3T3 cells are commercially available as a GMP cell line (Sigma-Aldrich, UK). Therefore, their application as a feeder layer to enhance OEC phenotype during scalable manufacture is attractive to advance OECs towards the clinic. Potential safety concerns around the purity of the final clinical preparation could be addressed by purifying OECs with affinity-based removal of the mouse feeders. From this initial work with primary rat OECs, it was found that rat OECs have a more promising regenerative function when cultured

with feeders. However, there was no difference between human olfactory mucosa-derived feeders and mouse 3T3 cells in their ability to support OEC marker expression.

#### *3.3. Human OECs Behave Di*ff*erently to Rat OECs in Culture*

Studies in small animals are important to generate pre-clinical data and to gain an understanding of responses in physiologic systems before moving into humans. However, there are concerns that OECs in different animals behave differently [49–51], and animal cells and tissue do not accurately reflect the function or behaviour of human material. This is especially important when the length of the injury is considered. The gap to be bridged by the neurons in an SCI is much smaller in a rat compared to a human [50], and the ability of OECs to support neurons over this small distance may not be relevant on the human scale. Therefore, we generated conditionally immortalised cell lines from the human olfactory mucosa [26] and tested these under the same conditions in order to determine how translatable animal studies are.

ICC micrographs in Figure 4 revealed that Fn staining on laminin occurred in a completely different pattern from the other matrix conditions. In Ms3T3 and PLL conditions, the Fn deposition could be more easily traced back to the individual cells producing it. In contrast, on laminin, the staining pattern was more widely distributed. These distinct differences in the pattern were unexpected. Due to this staining pattern, cell counts and yield were not carried out for Fn due to the inaccuracy of determining the original cell responsible for the Fn. The reason for the different staining pattern on laminin is not known, and no references to similar occurrences could be found in the literature.

**Figure 4.** PA5 cells cultured on laminin or Ms3T3 feeders, in standard media or HuG418 conditioned media. Cells were cultured for five days and stained for S100β and Fn (**A–L**). S100β-positive cells were quantified in ImageJ and graphed as yield of positive cells in the image area (**M**). Conditioned media constantly gave lower yields of S100β, and laminin with standard media gave the highest yield. The scale bars represent 400μm. Data are the means ± SEM, n = 3.

S100β expression appeared to be more prominent in conditions with NT-3 compared with those conditions cultured in its absence and CM did not have any positive impact on S100β expression (Figure 4). From the three matrix conditions, it was found that the presence of Ms3T3 feeders was less beneficial for glial marker expression compared with PLL and laminin, which contrasted what was observed for rat OECs (yield of 133.7 <sup>±</sup> 24.0 cells/mm<sup>2</sup> on Ms3T3 with standard media compared to 184.8 <sup>±</sup> 27.7 cells/mm2 and 235.9 <sup>±</sup> 31.0 cells/mm2 for PLL and laminin with standard media, respectively). This was not necessarily surprising, as it has been found that there are several differences between the rat and human olfactory system [52] including in vitro growth, spontaneous immortalisation, and morphology [53,54]. Additionally, in terms of implantation and isolation, the olfactory mucosa in rats is yellow, whereas in humans, it cannot be discerned from the respiratory tissue [55]. This leads to a higher level of non-OECs being present in the implant, which may or may not assist in regeneration. It does cause concern that any previous work that has been carried out in rat is not translatable to human scale up [24,49].

Another difference between the rat and human OECs is that the human OECs do not appear to benefit from the presence of NT-3. Although the study was not carried out as part of this work, previous studies have shown that NT-3 significantly increased the expression of glial markers in rat cells [27]. For our human cell line, the presence of NT-3 seemed to increase the level of expression slightly, but not significantly.

The differences that occurred between rat and human OECs in regard to neurotrophic factors and matrix preference showed that more in depth study and understanding of human OECs are necessary to be able to predict cell behaviour. It also indicated that care needs to be taken when translating results from rat to human OECs. Two human cell lines were investigated (data not shown from the second cell line), and they both followed the same pattern with laminin being the highest performing condition in terms of yield and significantly higher than most of the conditions with CM. Due to this, human OECs were carried forward into the future work to ensure the conditions investigated were relevant to the clinic.

#### *3.4. Human OEC Co-culture with Neurons*

Neurite outgrowth in cultured neurons is considered an indication of neuro-regenerative potential [56–60], and therefore, neurite outgrowth was quantified to compare the cultures. Neurite number and length were normalised to the number of neurites and neurons. Neurons were cultured in isolation to ensure any neurite extension we observed could be related to the presence of the OECs.

Cells co-cultured on laminin with standard media had a higher number of neurites with longer extensions (Figure 5). Cells co-cultured on laminin and feeders with CM did not perform as well as cells cultured with standard media. No difference was observed between the neuron-only conditions on the different matrices for all measurements. This indicated that the neurons were responding to the PA5s as opposed to the culture conditions (matrix and media).

The conditions cultured with OECs performed best in regard to average neurite length (Figure 5S, one-way ANOVA, Bonferroni post-hoc, *p* < 0.001). Although co-culture on laminin with standard media performed better than those co-cultured with CM (49.7 ± 3.7μm versus 40.4 ± 1.6 μm), this did not result in any significant difference.

When the neurite length per neuron was examined (Figure 5T), the co-culture on laminin with standard media showed significantly longer neurites per neuron (76.0 ± 11.2 μm) compared with every other condition (43.6 ± 5.5 μm on laminin with CM and 41.0 ± 10.6 μm on Ms3T3 with CM, one-way ANOVA, Bonferroni post-hoc, *p* < 0.01). All conditions without OECs performed similarly to each other (22.8 ± 2.4 μm, 17.8 ± 2.3 μm and 17.9 ± 1.1 μm on laminin with standard media, laminin with CM and Ms3T3 with CM, respectively).

The number of neurites extended by each neuron revealed a similar pattern (Figure 5U) where the co-culture on laminin with standard media produced more neurites per neuron (1.54 ± 0.20) than the majority of the other conditions (one-way ANOVA, Bonferroni post-hoc, *p* < 0.001) with the exception of co-culture on laminin with CM (1.08 ± 0.14). Collectively, these data showed that the matrix and media condition that gave the highest S100β expression in OECs resulted in the best neuronal growth support.

**Figure 5.** PA5 cells were co-cultured with NG108-15 neurons for five days under different matrix and media conditions. NG108-15 cells were also cultured in isolation to ensure that any improvement observed could be reliably attributed to the PA5 cells as opposed to the matrix and media conditions. After five days of NG108-15 culture, cells were fixed and stained for βIII-tubulin and p75NTR (**A–R**). The NeuronJ plugin in for Image J was used to quantify the neurite outgrowth, and measurements were made for neurite length/neurite (**S**), neurite length/neuron (**T**), and neurites/neurons (**U**). The scale bars represent 200 μm. Data are the means ± SEM, n = 3.

The formation of neurites from neurons is vital to the functionality and development of the nervous system [61]. From these results, it can be seen that co-culture on laminin with standard media was the condition that provided the most support to the development of neurons in regard to the longest average neurite. In the absence of co-culture, the development of neurons was not well supported, and as a result, shorter neurites were observed. Previously, it was found that laminin with standard media was a promising condition for OEC proliferation and S100β expression. This suggested that when OECs were expressing higher levels of S100β and that they were able to provide better support to neurons. When S100β expression was lower, the presence of the OECs was still able to benefit the neurons.

When Schwann cells support peripheral nerve repair, it has been observed that an upregulation of S100β leads to identification of so-called "reactive" Schwann cells, which are responsible for axonal sprouting [62]. If OECs followed a similar behavioural pattern, it could be that the conditions with higher S100β were more capable of allowing axonal sprouting and therefore resulted in increased neurite numbers.

It has been observed in several studies that laminin is an extracellular matrix that is able to stimulate rapid neurite growth and has been directly linked to neurite outgrowth in vitro [63–65]. Studies have also shown that enhanced neurite outgrowth and preferred attachment was observed when neurons from the CNS were plated on laminin [66], and it is thought that the presence of laminin starts to create a more permissible environment for axonal extension [67]. The hostile environment present after SCI is a key part of why nerve regeneration does not happen spontaneously [24,68,69]. There was no significant difference observed between neurons cultured in isolation on laminin and on Ms3T3 feeders. This suggests there is more to the interaction than the preference for laminin. The combined effect of the favourable matrix and support cells could explain why laminin with standard media was the highest performing condition.

CM collected from HuG418 had a lesser effect on the average neurite length per neuron compared to standard media. This may be initially related to the lower S100β expression in the OECs in this condition. It would indicate that the CM from HuG418 does not have any beneficial soluble paracrine factors for the neurons. This lack of factors and therefore interaction between these two populations is not necessarily unexpected as although studies have shown a benefit in transplantation with a mixed population of OECs and fibroblasts (Keyvan-Fouladi et al., 2003 [44], Ramón-Cueto et al., 2000 [10], Raisman and Li, 2007 [47], Teng et al., 2008 [46]), it has been under the understanding that the fibroblasts support the OECs, not the neurons. Fibroblasts have not been pursued as a cell therapy option for nerve regeneration, and they are not believed to have specific properties that enhance the function of neurons [70,71]. Therefore, standard media resulting in longer neurites than conditioned media can be seen as an expected output.

#### **4. Conclusions**

We established that the behaviour of rat and human OECs did not follow the same patterns, and therefore, caution needs to be engaged when using pre-clinical data that have been carried out using rat models. When focusing on human OECs, we found that out of the conditions investigated, laminin with NT-3 was the best condition for protein expression and neurite extension. It would be valuable to investigate the components of the fibroblast conditioned media to further understand what aspects resulted in Thy1 upregulation and how the population reacts when Thy1 positive cells are removed.

**Author Contributions:** Conceptualization, R.W. and I.W.; data curation, R.W. and P.D.; formal analysis, R.W. and P.D.; funding acquisition, I.W.; investigation, R.W. and P.D.; methodology, R.W. and P.D.; project administration, R.W. and I.W.; supervision, I.W.; visualization, R.W.; writing, original draft, R.W. and I.W.; writing, review and editing, R.W., P.D. and I.W. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Biotechnology and Biological Sciences Research Council (BBSRC; Grant Number BB/K011154/1), the Bioprocessing Research Industry Club (BRIC), the UCL Overseas Research Scholarship, and the New Zealand Federation for Graduate Women (NZFGW).

**Acknowledgments:** We would like to thank the Biological Services Unit (BSU) at UCL for their help with culling the animals.

**Conflicts of Interest:** The authors declare no conflicting interest in this work.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Communication* **Challenges and Solutions for Commercial Scale Manufacturing of Allogeneic Pluripotent Stem Cell Products**

#### **Brian Lee 1,\*, Breanna S. Borys 2, Michael S. Kallos 2, Carlos A. V. Rodrigues 3, Teresa P. Silva <sup>3</sup> and Joaquim M. S. Cabral <sup>3</sup>**


Received: 29 February 2020; Accepted: 26 March 2020; Published: 28 March 2020

**Abstract:** Allogeneic cell therapy products, such as therapeutic cells derived from pluripotent stem cells (PSCs), have amazing potential to treat a wide variety of diseases and vast numbers of patients globally. However, there are various challenges related to the manufacturing of PSCs in large enough quantities to meet commercial needs. This manuscript addresses the challenges for the process development of PSCs production in a bioreactor, and also presents a scalable bioreactor technology that can be a possible solution to remove the bottleneck for the large-scale manufacturing of high-quality therapeutic cells derived from PSCs.

**Keywords:** allogeneic cell therapy; induced pluripotent stem cell; human embryonic stem cell; cell aggregate; expansion; differentiation; scalable manufacturing; scale up; single-use bioreactor; Vertical-Wheel; U-shaped vessel; computational fluid dynamics; shear stress; turbulent energy dissipation rates; homogeneous hydrodynamic environment

#### **1. Introduction**

With their potential to cure a wide variety of disease indications and address vast patient populations, allogeneic cell therapies derived from pluripotent stem cells (PSCs) are poised to revolutionize therapeutic medicines [1,2]. However, 2D planar manufacturing technologies that have been commonly used for small scale R&D and early stage clinical trials are inadequate and cost-prohibitive for production at the larger scales required for late-stage clinical trials and commercial manufacturing [3]. Single-use bioreactors are widely recognized as a feasible manufacturing solution but need to be optimized in order to meet the unique process requirements of PSCs [4]. One of the critical challenges for future success in commercializing allogeneic cell therapy products is establishing a scalable manufacturing technology that can reliably reproduce the yield and quality of PSC-derived products generated from small-scale R&D at larger scales sufficient for commercial manufacturing [5].

PSCs are mortal human cells that include specific cell types such as human embryonic stem cells (hESCs) and induced pluripotent stem cells (iPSCs). When cultivated in a 2D planar vessel, PSCs attach to a surface substrate and grow as a monolayer. In contrast, starting from single cells or small clumps in suspension bioreactors, PSCs that come into contact will naturally clump together to form spherical cell aggregates [6]. The formation of cell aggregates is required not only for the cell expansion phase but also for subsequent differentiation, which can be a multi-step process that directs the pluripotent cells to turn into a final target cell type for treating a particular disease.

Depending on a particular cell expansion and differentiation process, there is an optimal range of spherical cell aggregate sizes that can maximize the efficiency and production yield of expansion and differentiation processes in a bioreactor [7–9]. If a PSC aggregate becomes too large, nutrients and the growth or differentiation factors may be unable to evenly diffuse from the aggregate surface into its center, leading to unwanted cell death or heterogeneous cell populations during expansion or differentiation [10,11]. PSC aggregates that are too small may result in less efficient cell expansion and differentiation, thus lowering the yields of final target cells [12].

There are varying published examples of optimal cell aggregate sizes for different cell types and cell culture process steps. One experiment using hESCs showed that maximum viability and minimal cell apoptosis during expansion was achieved when average aggregate diameter was 300 μm after 7 days of expansion, while apoptosis (with the majority of necrotic cells in the centers of aggregates) peaked by day 14 if the average diameter reached 500 μm [13]. In another experiment, hESC aggregates with diameters of 400 μm detrimentally had half the concentration of oxygen in their centers compared to aggregates of 200 μm [10]. In a different experiment focusing on hESC differentiation, an average diameter of 450 μm was ideal for cardiogenesis, whereas diameters between 150–300 μm were ideal for endothelial cell differentiation [14]. For two different experiments involving hPSCs, an average aggregate diameter of 139 ± 26 μm was optimal for initiating differentiation into neural cells in one experiment [15], while an average diameter of 130 ± 40 μm was optimal for endoderm induction in the other experiment [16]. As these various experiments indicate, optimal cell aggregate size is highly dependent on factors such as cell type and expansion or differentiation process steps.

While PSC aggregate formation can be influenced by variables such as cell proliferation rate, cell–cell adhesion strength, and cell packing density, the hydrodynamic environment inside a bioreactor, which is created by the impeller used to continually mix the liquid media, also has a significant impact on determining aggregate size and, ultimately, cell viability [17]. The hydrodynamic environment is characterized by the two parameters of fluid shear stress and turbulent energy dissipation rate. Both of these parameters are inversely correlated with the average size of PSC aggregates: higher levels of shear stress and turbulent energy dissipation rate result in smaller sized aggregates, and vice versa. After the initial seeding of single cells or small preformed aggregates into a bioreactor, collisions due to the hydrodynamic environment will facilitate aggregate growth, either through the addition of single cells onto existing aggregates or the fusion of smaller aggregates into larger ones. At sufficiently higher agitation rates, the increased levels of shear stress will promote the breakage of loosely attached or temporarily agglomerated larger aggregates and thus limit their maximum possible size [18].

Fluid mixing in a bioreactor using a traditional horizontal-blade impeller creates a significantly uneven hydrodynamic environment. The highest levels of shear stress and turbulent energy dissipation rates will be near the tips of the rapidly spinning impeller, with decreasing gradients of these hydrodynamic parameters as the distance from the impeller increases [19,20]. Such a wide range of shear stress levels and energy dissipation rates in a horizontal-blade impeller bioreactor results in a non-uniform hydrodynamic environment, which in turn results in a wide variation in PSC aggregate sizes. In this scenario the size of cell aggregates will vary and the range of aggregate diameters in suspension becomes broad, which ultimately results in an inconsistent yield and quality of PSCs during expansion and differentiation steps. This variation in PSC aggregate size becomes more pronounced at larger sizes of horizontal-blade impeller bioreactors.

#### **2. Scalable Bioreactor Technology as Manufacturing Solution**

In contrast, the innovative Vertical-Wheel™ impeller, in conjunction with a distinct U-shaped vessel, provides a significantly more homogeneous hydrodynamic environment in Vertical-Wheel bioreactors. Computational fluid dynamics (CFD) analysis shows consistently low shear stress levels on all surfaces of the Vertical-Wheel impeller, as well as a narrow range distribution of turbulent energy dissipation rates throughout the U-shaped vessel (Figure 1) [21].

**Figure 1.** Computational fluid dynamics (CFD) analyses of shear stress on the surface of vertical-wheel impeller and range of turbulent energy dissipation rates in U-shaped vessel.

Fluid mixing using the Vertical-Wheel results in a very low variation in these two hydrodynamic conditions while still achieving the complete and continual suspension of PSC aggregates. The homogeneous hydrodynamic environment will result in a much tighter distribution of PSC aggregate size (Figure 2A). As a result, PSC aggregates in a Vertical-Wheel bioreactor will consistently have spherical shapes with similar diameters.

**Figure 2.** *Cont.*

**Figure 2.** (**A**) Projected distribution of cell aggregate sizes in Vertical-Wheel bioreactor (homogeneous) vs. horizontally stirred bioreactor (non-homogeneous) hydrodynamic environments. (**B**) inverse correlation between average pluripotent stem cells (PSC) aggregate diameter and agitation rate in Vertical-Wheel bioreactors.

In addition, the average diameter of PSC aggregates can be controlled by simply adjusting the agitation rate of the Vertical-Wheel impeller (Figure 2B). As agitation rate increases, turbulent energy dissipation rates and shear stress levels also increase while maintaining a homogeneous hydrodynamic environment, resulting in smaller average aggregate diameters and still narrow size distributions. The inverse is also true, with average aggregate sizes becoming larger as agitation rate is lowered. Starting with a relatively low agitation rate after seeding can promote initial aggregate spheroid formation. The agitation rate could later be increased to a speed that completely suspends larger particles and forms optimally sized cell aggregates, while also preventing unwanted fusion or agglomeration. Furthermore, the homogeneous hydrodynamic environment is maintained as bioreactor volume increases. Being able to predict the mixing properties and hydrodynamic environment at larger scales based on process development done at a small scale will be enormously beneficial for establishing a scalable manufacturing process for PSCs.

#### **3. Independent Biological Performance Data from Collaborators**

Biological studies in Vertical-Wheel bioreactors have confirmed the inverse correlation between agitation rate and average size for iPSC aggregates (Figure 3A). The uniformity of cell aggregate sizes at different agitation rates in small-scale Vertical-Wheel bioreactors, compared to those produced in horizontal spinner flasks, was also confirmed (Figure 3B) [22]. The observed narrow size distributions and uniform morphology also indicate that unwanted agglomeration into overly large or irregular spheroids is minimized across various agitation rates.

(**B**)

**Figure 3.** (**A**) Inverse correlation between average iPSC aggregate diameter and agitation rates in Vertical-Wheel bioreactor (0.1 L Scale). (**B**) Comparison of iPSC aggregate diameters and morphology with different agitation rates in Vertical-Wheel bioreactor vs. horizontal-blade spinner (0.1 L scale). Scale bar (200 μm).

As previously mentioned, the uniformity of PSC aggregates is essential not only for the efficiency of cell expansion, but also for directed differentiation, by minimizing the chance that growth and differentiation factors diffuse unevenly through aggregates, and therefore avoiding the undesired heterogeneity and lower quality of target cells. In another set of experiments, iPSC aggregates were successfully differentiated into neural cells, which then formed cerebellar organoids in suspension within small scale Vertical-Wheel bioreactors (Figure 4). After 35 days of a cell expansion and differentiation process, iPSC-derived organoids were efficiently maturated to GABAergic and Glutamatergic neurons in 0.1 L scale Vertical-Wheel bioreactors [23]. Other iPSC aggregates were successfully differentiated into beta cells, cardiomyocytes, or mammary organoids in suspension using different differentiation processes in various scales of Vertical-Wheel bioreactors (data not shown). In addition, studies to differentiate iPSCs into liver organoids in suspension are currently ongoing.

**Figure 4.** Directed differentiation of human iPSCs into cerebellar organoids in 0.1 L scale Vertical-Wheel bioreactor. Scale bar (100 μm).

While these various experiments were performed at relatively small scales (0.1 to 15 L) in Vertical-Wheel bioreactors, the homogeneous hydrodynamic environment created by the Vertical-Wheel impeller and U-shaped vessel has been modelled by CFD analysis up to 80 L scale and beyond. Therefore, it can be predicted that the formation of uniformly sized PSC aggregates will also be scalable in Vertical-Wheel bioreactors, resulting in efficient cell expansion and differentiation processes even at larger scales.

A PSC expansion process can start with a small cryopreserved vial of PSCs from a working cell bank, inoculating thawed cells into a small-scale bioreactor, and then expanding the cells by serially passaging them into progressively larger bioreactors. Using the scalable bioreactor system with a consistent hydrodynamic environment will improve the efficiency of serial passaging and total cell yield. Studies were performed in order to determine the consistency of PSC expansion in Vertical-Wheel bioreactors, from small-seed culture scale (0.1 L) to potential production scale (80 L). Freshly thawed iPSCs were inoculated into a 0.1 L Vertical-Wheel bioreactor and then four consecutive passages of iPSCs were performed using scale-down model bioreactors of identical size. These studies repeatedly demonstrated that the use of Vertical-Wheel bioreactors led to an average 32-fold expansion of iPSCs during 6–8 days of culture period per passage, achieving a cumulative cell expansion of greater than one million-folds in 28 days (Figure 5). The cells harvested at the end of the serial passage were of high quality, maintaining a normal karyotype, pluripotency, and the ability to form teratomas in vivo. This scale-down model process suggests that commercial scale production of over one trillion high quality iPSCs could feasibly be generated in a Vertical-Wheel bioreactor with 50 L working volume [22].

Various scaling factors can be used to the determine potential scalability of expansion and differentiation processes developed at a small scale in Vertical-Wheel bioreactors. Common parameters that would need to be kept constant during scale-up include Reynolds number, impeller tip speed, power input, maximum shear stress, velocity, and energy dissipation rate. A recent study has suggested that maintaining the volume average energy dissipation rate, determined through computational fluid dynamics simulations, is the best scale-up method for predicting the agitation rates that will sustain a desired average aggregate diameter with narrow size distribution [24]. Therefore, the agitation rates for each progressively larger Vertical-Wheel bioreactor can be calculated that will maintain the same homogenous hydrodynamic environment that was determined during process development in small-scale Vertical-Wheel bioreactors.

**Figure 5.** Successful Serial Passaging of iPSCs in 0.1 L Scale Vertical-Wheel Bioreactors.

#### **4. Conclusions**

The ultimate goal of providing allogeneic PSC-derived products reliably to vast numbers of patients requires a series of optimized unit operations at various scales to meet target manufacturing lot sizes. Numerous manufacturing processes need to be considered, including seed train, expansion, differentiation, cell harvest, wash, concentration, fill-finish, and cryopreservation. In particular, upstream cell culture processes represent the most challenging bottleneck to achieving robust and economical manufacturing at larger scales. With homogeneous hydrodynamic environments and unparalleled scalability, Vertical-Wheel bioreactors can serve as the innovative manufacturing platforms to enable commercial-scale production of PSC-derived therapeutic cells and to secure the supply of these novel therapeutic cell products to patients.

**Author Contributions:** B.S.B. and M.S.K. performed the experiments of PSC aggregate expansion, including serial passaging. C.A.V.R., T.P.S., and J.M.S.C. performed the experiments of differentiating PSC aggregates into various organoids. B.L. wrote the manuscript. All authors have given approval of the final version of the manuscript.

**Funding:** The research in this manuscript received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Electrospinning Live Cells Using Gelatin and Pullulan**

#### **Nasim Nosoudi 1,2,\*, Anson Jacob Oommen 2, Savannah Stultz 2, Micah Jordan 2, Seba Aldabel 2, Chandra Hohne 2, James Mosser 2, Bailey Archacki 2, Alliah Turner <sup>2</sup> and Paul Turner <sup>2</sup>**


Received: 3 February 2020; Accepted: 20 February 2020; Published: 22 February 2020

**Abstract:** Electrospinning is a scaffold production method that utilizes electric force to draw a polymer solution into nanometer-sized fibers. By optimizing the polymer and electrospinning parameters, a scaffold is created with the desired thickness, alignment, and pore size. Traditionally, cells and biological constitutes are implanted into the matrix of the three-dimensional scaffold following electrospinning. Our design simultaneously introduces cells into the scaffold during the electrospinning process at 8 kV. In this study, we achieved 90% viability of adipose tissue-derived stem cells through electrospinning.

**Keywords:** electrospinning; live-cell electrospinning; tissue engineering; cell seeding; high voltage; viability

#### **1. Introduction**

Tissue engineering aims to produce synthetic tissues that maintain, restore, or improve native tissue functions [1]. Engineers utilize the formation of both acellular scaffolds and scaffolds that are seeded with cells to accomplish these objectives. Acellular scaffolds are typically used to define a space for new tissues to develop [2]. These scaffolds serve as an extracellular matrix to promote cell adhesion and growth in vivo. Scaffolds with seeded cells have a greater significance because they closely mimic human tissues. It is essential for cell adhesion and migration to occur within these scaffolds. Overall, scaffolds are usually porous and created by various methods, such as electrospinning, phase-separation, freeze-drying, and self-assembly [3]. They enhance the body's ability to heal itself by providing a biodegradable matrix that can enable cells to grow [1].

Electrospinning is a quick and efficient way to produce scaffolds because it allows the scientist to control parameters such as the sizes of nanofibers and nanopores [4]. Other parameters can be constructed as well based on a careful selection of the polymer and an appropriate solvent [5]. During electrospinning, the polymer will dissolve in a volatile solvent and be loaded into a syringe. This liquid is extruded from the needle tip at a constant rate by a syringe pump. In addition, a positive or negative lead will connect to the needle-tip of the syringe while a ground lead is placed on a collector plate [6]. The distance between the syringe-tip and the collector plate will vary depending on the properties of the polymer solution. When the electrostatic force on the polymer solution overcomes the surface tension, a jet of the polymer solution will form and eventually travel towards

the collector plate [7]. As the jet flows towards the collector plate, the liquid will accelerate and deposit micro/nanofibers of the polymer on the collector plate [8].

Utilizing current methods, cells are seeded onto the scaffold after it has been formed. Cell seeding can be time-consuming because it requires three steps: creation of the scaffold, differentiation of the cells, and incorporation of the cells into the scaffold. Cell differentiation is already time-consuming and requires additional components, such as growth factors [9]. Another problem arises once these cells are differentiated and seeded: limited ability of cell diffusion into the scaffold [10]. Limited diffusion can produce a nonuniform distribution of cells that causes varied properties and cell densities within different areas of the scaffold. This is potentially detrimental to the longevity of the scaffold both in vitro and in vivo [11]. A possible method to increase cell dispersion in the scaffold is to directly incorporate the cells into the electrospinning process.

There is evidence suggesting that externally applied magnetic fields can affect cell differentiation. It is likely that the generated electric field affects the cell membrane [12]. When the membrane is forced to change shape, it will distort the cytoskeleton of the cell, which attaches the cell membrane to the nucleus. This change in the cytoskeleton will affect the expressed genes and cause the creation of different cell signals, which could induce differentiation [13]. Incorporating stem cells into the electrospinning process will expose them to an electric field that likely induces unique behaviors (i.e., cell differentiation) as previously reported while using lower voltages [14,15]. While this has the potential to be successful, there are many potential hazards to consider.

A concern in this process is that cells could not survive the voltage used in electrospinning. While an electric field could cause unique behaviors, an excessively large electric field could be detrimental to the viability of the cells. Moreover, the field could denature specific protein channels in the membrane, which causes irreparable cell damage [16]. Voltages will be kept as low as possible to prevent this from occurring. Typical electrospinning voltages range from 1 kV to 30 kV [17]. The applied voltage required to create a scaffold will vary depending on the polymer used.

Jayasinghe et al. made use of the coaxial electrospinning from immortalized human brain astrocytoma [18]; a year later, the author described its use from primary porcine vascular smooth muscle cells or rabbit aorta smooth muscle cells, while the protectant polymer was polydimethylsiloxane (PDMS) [19]. Yunmin et al. showed simultaneous bio-electro spraying of human adipose stem cells (ASCs) while electrospinning polyvinyl alcohol (PVA), but the study used two separate needles [20]. Recently, Hoare et al. used hydrazide-functionalized POEGMA (POH) and aldehyde-functionalized POEGMA (POA) along with poly(ethylene oxide) (PEO) to successfully encapsulate NIH 3T3 fibroblasts and electrospin them [21].

This research will test two different polymer combinations: collagen/poly(ethylene oxide) (PEO) and gelatin/pullulan. These polymer combinations will electrospin at around 8 kV of applied voltage. One other constraint for this experiment is the solvent used for dissolving the polymer. In current electrospinning methods, common solvents for collagen include 1,1,1,3,3,3-hexafluoro-2-propanol (HFP), 2,2,2-trifluoroethanol (TFE), or acids (tri-fluoro acetic acid (TFA), acetic acid, hydrochloric acid) [22,23]. These solvents could be toxic if cells are directly incorporated into the polymer–solvent solution. To overcome this restriction, the group will use cell media as the solvent.

Collagen was chosen as the initial polymer because it is the primary constituent of the body's natural extracellular matrix [24]. However, collagen is typically electrospun with an acetic acid solvent, which would likely cause cell death. No studies have attempted to show the success of electrospinning collagen with cell media as the solvent. Therefore, other polymers will also be utilized to determine which one creates the best scaffold. Moreover, collagen may degrade while electrospinning, but successful trials have been reported using different methods, including modifying the collagen surface with methyl methacrylate-*co*-ethyl acrylate [25]. Gelatin is simply denatured collagen; therefore, it can create scaffolds with the same success as collagen [26]. Because previous studies have determined that PEO increases the yield of uniform fibers when electrospun with other polymers, we decided to use it as well [27]. Pullulan and gelatin are commonly used together in hydrogels, and pullulan has shown antioxidant potential [28]. Therefore, the group will electrospin with pullulan, gelatin, or a combination of both. Adipose-derived stem cells (ADSCs) will be used due to their accessibility and potential for creating various terminally differentiated cells, such as osteoblasts, chondrocytes, adipocytes, and neurons [29]. The cells will be directly incorporated into the five polymer solutions prior to electrospinning. To group's knowledge, this is the first time that electrospinning live cells using these natural biocompatible polymers has been reported.

#### **2. Materials and Methods**

#### *2.1. Electrospinning Device*

Due to the nature of working with living stem cells, it was imperative to maintain sterile conditions throughout the entire spinning process. To maintain a sterile environment, the spinning took place under a sterile biological safety cabinet. The electrospinning device needed to withstand exposure to ultraviolet (UV) light, so it could be sterilized for at least 24 hours under the culture hood as the UV light was usually turned on the night before the experiment. Acrylic, which can handle UV exposure, was determined to be the material that the electrospinning device would be constructed with. Acrylic sheets and cement were used to construct the framework of the electrospinning device along with the necessary spinning supplies, such as a plate, voltage supply, electrical leads, and syringe pump.

#### *2.2. Cell Culturing and Electrospinning*

P2-P4 of adipose tissue-derived stem cells (hASCs) from Lonza (Walkersville, MD, USA) were used for cell cultures. Cells were plated in T75 culture-treated flasks with approximately 1 million cells per flask, and culture media was changed every 3-4 days for the duration of the culture. Three components make up the cell electrospinning solution: protectant, solvent, and cell pellet. Collagen, Poly(ethylene oxide 10,0000), pullulan, and gelatin powders were used as protectants. Poly(ethylene oxide) (Sigma), pullulan (Hayashibara Laboratories, Okayama, Japan), Type A gelatin from porcine skin (Electron Microscopy Sciences, Hatfield, PA, USA), and extracted collagen from rat tail were dissolved in solvent at concentrations of 2.5 mg/mL, 5 mg/mL, 10 mg/mL, 20 mg/mL, and 30 mg/mL. For PEO, gelatin, and pullulan, the powder was dissolved in serum-free culture media in the previously mentioned concentrations, while collagen acetic acid was used instead of serum-free culture media. The protectants and mesenchymal stem cell medium (solvent) were mixed again at the ratio of 1:1 by volume and placed on a stirring hot plate for 20–30 minutes to warm and mix. The solution was warmed to 40 degrees Celsius in the case of gelatin. The tube temperature reduced to 37 degrees Celsius. Cell pellets (1 <sup>×</sup> <sup>10</sup>6) were then added to the protectant solution. Cell electrospinning content was aseptically transferred to a sterile 10 ml syringe, and a sterile 18-gauge syringe needle tip was secured. The collector plate, which is a petri dish, was positioned 7.5 cm from the end of the needle tip. The syringe pump settings were adjusted to produce readings for a plastic 10 ml syringe pump. The pump rate was set to 30 μL/min and reduced at increments of 5 μL/min to determine the optimized pump rate for each cell electrospinning solution. Control cells using the same combination of gelatin, pullulan, gelatin/pullulan, collagen, or PEO were sprayed at the same rate on an empty petri dish without any voltage application. This procedure was done at room temperature, and electrospinning was never performed for more than 15 minutes.

#### *2.3. Viability Test*

The viability was investigated by a live/dead assay kit and fluorescence microscopy. Approximately 6 hours after electrospinning, the culture media was aspirated from each well. After incubation with calcein and ethidium (2 μM calcein and 4 μM ethidium in PBS) for 10 minutes at 37 ◦C, samples were washed with PBS and cells were imaged.

#### Cytotoxicity Test (Lactate Dehydrogenase (LDH) Activity)

The media was aspirated two days after spinning, and cells were washed with PBS. Lactate dehydrogenase or LDH (Cytotox96 kit, Promega, Madison) was performed on the attached cells according to the manufacture's protocol to look at the cell viability using cell lysate [30].

$$\text{Viability} \%= \frac{\text{Average OD of sample} \ast 100}{\text{Average OD of control}}$$

#### *2.4. Gene Expression by Reverse Transcription-Polymerase Chain Reaction (RT-PCR)*

Seven days after spinning, RNA was isolated according to the manufacturer's instructions for the RNeasy plus mini kit (Qiagen, Germantown, MD, USA), and RT-PCR was performed according to the instruction manual of the One-Step RT-PCR kit (Qiagen, Germantown, MD, USA). The selected pluripotential genes were SOX2 and OCT4.

#### *2.5. Immunocytochemistry*

Cellular morphology was visualized on Day 2 using fluorescence microscopy. Briefly, samples were fixed with 4% paraformaldehyde (PFA) in PBS (pH 7.4) for 15 min at room temperature (RT). After rinsing with PBS three times, the samples were placed in a permeabilization solution with 0.1% (v/v) Triton X-100 for 10 min and rinsed again with fresh PBS three times. The cells were incubated with Phalloidin 488 and DAPI (Life Technologies, Carlsbad, CA, USA) to visualize the f-actin and nuclei, respectively.

#### *2.6. Microscopy*

To observe the structure of the scaffold, Fluorescein Isothiocyanate (FITC)-conjugated gelatin was used. Electrospun cells/scaffolds deposited on microscope glass slides were imaged using an Olympus BX51 microscope equipped with an Olympus DP73 camera and CellSens software.

#### *2.7. CytoViva Microscopy*

To confirm that the cells were embedded within the scaffold, the cells were labeled using a green CMFDA cell tracker dye (Invitrogen, Oregon, Germantown, MD, USA) before electrospinning; they were labeled with DAPI afterward. Samples were imaged using CytoViva's patented enhanced darkfield transmitted light condenser (NA 1.2–1.4) coupled with CytoViva's proprietary Dual Mode Fluorescence (DMF) module. These components were configured on an Olympus BX51 upright microscope using an Olympus100X oil UPL Fluorite objective (NA 0.60–1.30) with adjustable iris objective optimized for darkfield imaging. The light source used was Prior Lumen 200 with a metal halide lamp and variable light attenuation. Optical images were captured using a DAGE-MTI XLMCT cooled CCD camera with a 7.4 μm pixel size.

#### *2.8. Fourier-Transform Infrared Spectroscopy (FTIR)*

Scaffold compositions were determined by loading the samples onto an attenuated total reflectance (ATR) attachment and using a Thermo Scientific Nicolet iS 50 FTIR (Thermo Fisher, Waltham, MA, USA). Data were plotted in MS Excel (Microsoft, Redmond, WA, USA).

#### **3. Results**

A scheme of the electrospinning process is shown in Figure 1A. Electrospinning was only observed at a concentration of 5 mg/mL at 8 kV (Table 1). Cells were detected 6 hours after electrospinning to observe attachment as a sign of viability. Most cells in the collagen scaffold were dead (i.e., stained red). The PEO scaffold had a lot of red cells floating in the petri dish. Gelatin, pullulan, and pullulan/gelatin

had good cell viability (i.e., stained fluorescent green), while the number of dead cells (stained red) was minor.


**Table 1.** Different concentrations that have been tried to electrospin cells.

**Figure 1.** (**A**) Schematic diagram of electrospinning setup for live cells; (**B**) lactate dehydrogenase (LDH) from cell lysate in electrospun and control groups; (**C**) LDH from cell lysate in electrospun (cell+ media) compared to Sprayed cell+ media. Results are normalized to control. \* shows *p*-value < 0.05.

The viability of cells in collagen was very low in both control and electrospun groups. Control cells were sprayed at the same rate on the petri dish without any voltage application. The 0.01% acetic acid that was used to dissolve collagen is probably the reason for low cell viability. PEO was dissolved in cell media and is biocompatible. However, very low cell attachment was observed in both control and electrospun groups, but the group realized this did not result from just the PEO. The PEO viability was 50% in the control group and less than 20% in the electrospun group. There is a significant difference between the control and electrospun group which is caused by the additional effect of electrospinning. In the electrospinning process, the polymer solution is exposed to shear stress, and dead cells in PEO can be the result of non-Newtonian fluids' behaviors and shear stress. When PEO was removed from the formulation, gelatin/cells' viability and attachment were good, and LDH on cell lysate showed 88% cell viability from the electrospun group compared to the control group (Figure 1B,C). However, when the sample was switched to pullulan/gelatin/cells, the group achieved

99% viability compared to the control, but the pullulan/cell scaffold had 91% viability. The *p*-value was not statistically significant among the three groups. To prove the role of the protectants (gelatin and pullulan), cells were electrospun with only culture media, and the cell viability reduced to 40% compared to cells and media that were sprayed with the same rate on the petri dish.

Seven days after electrospinning, Oil Red O, toluidine blue, and Alizarin Red S staining were used to study adipogenic, chondrogenic, and osteogenic differentiations. All cells were negative for Oil Red O, toluidine blue, and Alizarin Red S. Moreover, PCR data showed no significant change in SOX2 and OCT4 after electrospinning (Figure 2), which confirms stemness after and before electrospinning.

**Figure 2.** Gene expression in gelatin and gelatin/pullulan electrospun groups normalized to control. Control groups are cells cultured with 5 mg/mL gelatin or 5 mg/mL pullulan/gelatin.

To look at the cell alignment, the group used actin staining 2 days after electrospinning. Cell alignment was random as expected (Figure 3A,B). Images of the scaffold with FITC gelatin showed a porous structure which was later confirmed by CytoViva imaging as well. (Figure 3C).

A highly porous structure was observed after CytoViva imaging. It appears that the cells become embedded in these pores as confirmed by another CytoViva imaging, where cells were pre-stained with a cell tracker and DAPI. Those that house the cells were approximately 10 μm in diameter.

The band observed at 996 cm−<sup>1</sup> in the pullulan and electrospun samples, which is associated with C-OH bending vibrations at the C-6-position in the case of polysaccharide, indicates the strength of the interchain interactions via hydrogen bonding [31].

The primary hydroxyl groups at the C-6-position were available in the pullulan macromolecule (Figure 4C). However, there were no hydroxyl groups at the C-6-position in gelatin. This band can show the glycosylation between the gelatin and pullulan molecules or the formation of the interchain hydrogen bond in the composite fiber. The amide I (AmI) band at 1630 cm−<sup>1</sup> in pullulan/gelatin was the strongest among the three and shifted slightly to a higher wavelength, which can be associated with AmI sensitivity to hydrogen bonding at the C=O group [32,33]. Hydrogen bonding plays a significant role in the stabilization of protein secondary structure which can result from the presence of pullulan here [34].

This experiment was run at 8 kV, and the best concentration for Pullulan/gelatin was 5 mg/ml at the ratio of 1:1. Further studies are needed to look at the effect of higher voltage on cell viability and differentiation.

**Figure 3.** (**A**) Actin staining of adipose-derived stem cells in control and in (**B**) pullulan/gelatin/cells at 10×. Phalloidin 488 (green) labels actin, while DAPI (blue) labels the nucleus. (**C**) Cells were surrounded by FITC (green) conjugated gelatin.

**Figure 4.** *Cont*.

**Figure 4.** CytoViva and FTIR of the three scaffolds. (**A**) The cells were stained with CellTracker Green CMFDA (Invitrogen) at 2.5 μM for 1 h before electrospinning and were stained for DAPI after electrospinning. (**B**) CytoViva image of the cells and scaffold with no pre-staining. (**C**) FTIR of the pullulan, gelatin and gelatin/pullulan electrospun scaffolds.

#### **4. Discussion**

Uniaxial electrospinning with a single needle is a common technology for the fabrication of scaffolds that can provide the initial scaffold for tissue engineering applications. On the other hand, coaxial electrospinning facilitates the incorporation and preservation of bioactive substances, where the shell is often used to protect sensitive substances encapsulated in the core. In this new method, uniaxial electrospinning is incorporated with live cells. Polymers are used to protect the cells, and the cells are encapsulated in the polymers during the electrospinning process. In this study, the group used pullulan, gelatin, collagen, and PEO. Trials with collagen and PEO were unsuccessful as cell viability was not acceptable. Highly hydrated polymers, such as PEO, suppress cellular and molecular adhesions by providing a physical steric barrier [35,36].

This study proved that pullulan and gelatin could protect cells from high voltage damages. Pullulan and gelatin are biocompatible, water-soluble polymers that have been shown to be ineffective at changing phenotype, viability, and cell differentiation. Pullulan can quench reactive oxygen species [37] and be a great scaffold in combination with gelatin. Moreover, pullulan can increase the tensile strength of gelatin, which is very important in tissue engineering [38]. Its structural features, such as the presence of large amounts of hydroxyl groups in the main chain, make it an optimal polymer for creating scaffolds. Studies show that the increase in the pullulan content of a scaffold leads to an increase in viscosity and eventually a decrease in electrical conductivity [39]. The group believes that the composition used in this experiment for making scaffolds acted as a shield for live cells against electrical conductivity, which was shown by the viability studies.

CytoViva images showed a porous structure with cells embedded in it. Moreover, we observed single cells covered by pullulan and gelatin, which was confirmed by fluorescence microscopy. Studies have shown that integrin is an electric field-sensing protein on the cell surface [40]. In addition, gelatin attaches to cells via integrin. Blocking the sensing proteins may be the reason for protecting the cells from high-voltage damage. In general, gelatin can protect the cells by covering the essential structures required for cell function and viability [41]. Gelatin and pullulan are both water-soluble, and this scaffold will dissolve after incubation at 37 ◦C. These polymers protect the cells during electrospinning but eventually dissolve in the media which can be altered by biocompatible crosslinkers [42].

In this study, no significant difference was found in Sox2 and Oct4 gene expression before and after electrospinning, but it has been shown by numerous studies that low voltage electrical stimulation can affect gene expression of transforming growth factor-β (TGF-β), collagen type-I, alkaline phosphatase (ALP), bone morphogenetic proteins (BMPs), and chondrocyte matrix [43]. More experiments are needed to better understand the effect of 8 kV on stem cells, and the broader gene expression still needs to be studied.

#### **5. Conclusions**

The success of this project opens a new field of study within tissue engineering. The discovery that cells can be directly incorporated into the electrospinning process has many potential benefits within the tissue engineering realm. In this paper, electrospinning with an applied voltage of 8 kV was observed at a concentration of 5 mg/mL gelatin/pullulan, but there are many combinations of polymers and cell-types that can still be tested. The application of this design is endless, and many properties of cells may change with variations of voltages and materials. Although this method is limited to water-soluble polymers, using the core–shell technology allows the use of other polymers. By using the core–shell technology, the outer polymer can be loaded with cells and the inner core can be loaded with a polymer that is not water-soluble.

**Author Contributions:** S.A., J.M., and C.H. performed most of the gelatin studies. P.T., M.J., and A.T. performed most of the gelatin/pullulan studies. S.S., A.J.O., and B.A. helped in microscopy, CytoViva, and preparing the manuscript. N.N. initiated the research and participated in study design and analyses. N.N. edited the final version of this manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** Marshall University start-up fund.

**Acknowledgments:** I would like to thank Byron Cheatham from Cytoviva and David Ladle from Wright state university for Microscopy. Philippe Georgel, Charles C. Somerville and David Mallory who provided me lab space to finish this work.

**Conflicts of Interest:** We declare that the authors have no competing interests that might be perceived to influence the results and/or discussion reported in this paper.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Review*
