GMP compliant isolation of mucosal epithelial cells and fibroblasts from biopsy samples for clinical tissue engineering

Engineered epithelial cell sheets for clinical replacement of non-functional upper aerodigestive tract mucosa are regulated as medicinal products and should be manufactured to the standards of good manufacturing practice (GMP). The current gold standard for growth of epithelial cells for research utilises growth arrested murine 3T3 J2 feeder layers, which are not available for use as a GMP compliant raw material. Using porcine mucosal tissue, we demonstrate a new method for obtaining and growing non-keratinised squamous epithelial cells and fibroblast cells from a single biopsy, replacing the 3T3 J2 with a growth arrested primary fibroblast feeder layer and using pooled Human Platelet lysate (HPL) as the media serum supplement to replace foetal bovine serum (FBS). The initial isolation of the cells was semi-automated using an Octodissociator and the resultant cell suspension cryopreservation for future use. When compared to the gold standard of 3T3 J2 and FBS containing medium there was no reduction in growth, viability, stem cell population or ability to differentiate to mature epithelial cells. Furthermore, this method was replicated with Human buccal tissue, providing cells of sufficient quality and number to create a tissue engineered sheet.


Scientific Reports
| (2021) 11:12392 | https://doi.org/10.1038/s41598-021-91939-0 www.nature.com/scientificreports/ been shown to contain a non-human form of sialic acid (Neu5Gc) which elicits adverse immune responses in humans 24 . Thus the concerns arising from the use of xenogeneic material in patients has led to research into alternatives for the growth of squamous epithelial cells, such as growth arrested dermal fibroblasts 25 , MSC's 26 or omitting feeder layers all together 27 . Buccal mucosa has been used as a source of non-keratinised squamous epithelial cells, either as a graft 28,29 or as a tissue engineered product 30 . Various methods for procurement and culture of mucosal cells, such as outgrowth 31 or enzymatic digestion 18 , use of feeder layers 32 or tissue culture plastic coatings 19 and various media 33 have been reported. Green and Rheinwald initially cultured keratinocytes on growth-arrested 3T3 J2 17 , which have been shown to also promote the growth of oral epithelial cells 34 .
The objective of this study was to engineer a GMP-compliant process for the isolation and expansion of primary, non-keratinised squamous epithelial cells from buccal mucosa for potential clinical use. Here, we used porcine aerodigestive mucosa for proof-of-principle work. The pig is an ideal large animal to model upper aerodigestive research questions, their mucosa has comparable morphological characteristics to human aerodigestive mucosa 35 . Once an appropriate method had been achieved it was repeated with human buccal biopsies.

Results
Fibroblasts. Removal of the epithelial sheet from the lamina propria was investigated on both supraglottal and buccal porcine mucosa. After incubation in neutral protease for 16 ± 3 h at 4 °C followed by 10 min at room temperature the sheets were easily removed.
Digestion of the lamina propria to produce fibroblasts requires 24 h in collagenase at 37 °C or using the "h_skin_1" protocol for the Octodissociator a 3 h 37 °C enzymatic digestion before dissociation. The cells were then seeded into FBS-non-essential amino acid (NEAA) medium. The incubation time and methods were initially investigated using porcine supraglottal tissue, the viable cell number obtained was slightly greater for the automated method, providing significantly more cells by passage three (Fig. 1A), however there was no significant difference in cumulative population doublings between the three methods tested (Fig. 1B), suggesting that there were more cells obtained. The automated method was used on porcine buccal tissue and compared to the supraglottal, there was no significant difference between the two tissue types in the number of cells achieved up to passage 3 (Fig. 1C). All further experiments used the automated method and buccal tissue.
The composition of fibroblast media was investigated using buccal fibroblasts. Media containing FBS with NEAA, HPL with NEAA, FBS and HPL only were compared, by passage 2 there were significantly greater cells in the FBS medium compared to those containing NEAA. There was no significant difference between FBS and HPL. However there was no significant difference in population doublings achieved suggesting that a lower cell number was obtained at p0 due to varying adhesion (Fig. 1D,E).
When obtaining fibroblasts from human buccal pinch biopsies, there was no cell growth although cells were present in the digested material, therefore adhesion was investigated in porcine buccal fibroblasts.
FBS increased adhesion compared to DMEM HPL. Although not statistically significant, BI Media plus fibronectin coating provided the most adhesive surface (Fig. 1F). These cells grew up to p7 achieving 5.5 ± 2.2 population doublings and a cell number of 1.4 × 10 8 ± 1.43 × 10 8 , they had positive vimentin and negative α smooth muscle actin (αSMA) staining suggesting that they maintain their fibroblast phenotype (Fig. 1G,H).
The incubation time for 10 µM MMC growth arrest of the buccal fibroblasts was investigated using xCELLigence to measure cell growth rate by electrical impedance. Two hours exposure to 10 μm MMC induced fibroblast cell stasis in the absence of significant decline in cell number and was used for feeder layer creation (Fig. 1I).
The final protocol for the fibroblasts was automated digestion followed by 3 days culture in Bi media on Fibronectin coated plates, then transferring to DMEM plus HPL for further media changes and passages.

Epithelial cells. Buccal mucosal tissue was used for all epithelial cell experiments as a direct comparison to
human tissue. Digestion of the epithelial sheets in TrypLE was investigated. Buccal epithelial sheet was either, manually minced and incubated for 30 min at 37 °C, triturating every 10 min or placed into the enzyme cocktail and automatically dissociated using the Octodissociator program "37_TDK_1". There was no significant difference between the cell numbers obtained by the two methods ( Fig. 2A). For all further experiments the Octodissociator was used to standardise the protocol.
The composition of medium was investigated. We compared Epithelial medium containing 10% FBS, 5% HPL and 5% HPL plus 5 µM Y-27632. Cells grown in FBS-containing medium were flat, did not form the expected cobblestone pattern and did not survive to passage 1. HPL provided more viable cells, but after passage 2 these cells perished. The addition of Y-27632 improved the appearance of the colonies formed when viewed under a light microscope (Fig. 2E,F). There were significantly more cells grown in HPL + Y-27632 by passage 3, with a significantly higher viability of 96 ± 2.7%. These cells had undergone significantly more population doublings at 3.5 ± 1.7 by passage 3, with an average of 7.16 ± 4.3 days per population doubling (Fig. 2B-G). 5 µM Y-27632 was used in all further experiments.
We propose that autologous fibroblast feeders and epithelial cells can be obtained from the same biopsy. The removed epithelial cell sheet requires cryopreservation to allow development of an autologous feeder layer. The epithelial sheets were removed and cryopreserved directly in 4.5% HSA with 10%DMSO and 10% Y-27632. The epithelial cells were cryopreserved as a sheet to reduce operator error during digestion. Those frozen without Y-27632 did not survive. All cells from here have been cryopreserved prior to digestion.
Comparison of conditions. The epithelial cells created with this protocol were compared to those created by the industry standard protocol of 3T3 J2 feeder layers with FBS + Y-27632, as well as 3T3 J2 feeders with HPL + Y-27632 and primary feeder layers with FBS + Y-27632. There was no significant difference in cumulative   Primary FBS cells achieved by passage 3. However there was a significant decrease in viability of cells grown on 3T3 J2 feeders with FBS medium at passage 2 ( Fig. 3A-D). The comparable growth characteristics were confirmed when the epithelial cells were grown on the xCELLigence impedance plates, where we saw no significant difference in cell index (measure of cell density) at 160 h, although those grown in FBS-supplemented media achieved a higher cell index regardless of which feeder layer they were seeded on ( Fig. 3E-G).
As the cells showed comparable growth patterns in each method, we extended the study to include analysis of epithelial stem cells. Colony formation efficiency was measured and this was significantly reduced by passage 3 for those cells grown on Primary feeders in HPL-supplemented media (Fig. 4A). However, analysis of the frequency of holoclones there was no significant difference between the groups suggesting that there was a reduction in transiently amplifying cells in cultures maintained on primary feeder cells but no loss of epithelial stem cells (Fig. 4B). This is supported by the retention of the cell subset within the buccal epithelial cells which www.nature.com/scientificreports/ expressed the stem cell marker p75. There was no significant difference in the percentage of p75+ cells between the four culture conditions (Fig. 4C).
The cells were then grown in a non-contact co culture to investigate their ability to create a cohesive cell sheet. All conditions allowed the epithelial cells to form cell-cell junctions (Fig. 5a) and differentiate into multi-layered cell sheets (Fig. 5b), with HPL 3T3 4.3 ± 0.6, FBS 3T3 5 ± 1, Primary HPL 2.7 ± 0.6 and Primary FBS 3.3 ± 0.6 layers created, and were negative for the keratinised epithelial cytokeratin's 1/10 (Fig. 5c).  Holoclones colonies larger than 10 mm with round edges were counted, there was no significant difference between the conditions (b). P75 was measure using Flow cytometry at passage 3. There was no significant difference between the conditions (c). Data shows mean plus SD, n = 3. extracted, treated with MMC as above and, after seeding, progressed to form a confluent monolayer (Fig. 6A).

Scientific Reports
The epithelial cells grown on the primary fibroblast feeder layer were able to form the distinctive cobblestone pattern (Fig. 6B) and were able to form adherens junctions and differentiate into a multi-layered cell sheet at ALI in a non-contact co-culture (Fig. 6C-E) with 4 ± 2.8 layers. The epithelial cells continue expanding up to passage 3 where 2.6 ± 2.6 population doublings (Fig. 6F) had occurred with a population doubling time of 8.4 ± 6.5 days (Fig. 6G), maintaining a viability of 91.8 ± 7.2 (Fig. 6H) with a total cell number of 1.2 × 10 7 ± 1.34 × 10 7 (Fig. 6I) from an interpolated starting cell number of 4.6 × 10 5 ± 2.5 × 10 5 . The two samples both maintained p75 stem cells with 13.13 ± 1.7 and 6.4 ± 1.8 percent of basal cells. We estimate that this protocol applied to a human buccal tissue biopsy of no more than 1 cm 2 would provide sufficient cells to manufacture a tissue engineered cell sheet of 17 cm 230 .

Discussion
Specifications for a regulatory compliant epithelial cell tissue engineered product for clinical use include expansion of cells to required number, retention of high quality, such as intercellular junction formation, and maintenance of differentiation capacity. Results from this study suggest that both epithelial and fibroblast cells can be obtained from a buccal biopsy in a GMP-compliant manner, which are not inferior to the current gold standard method yet avoid the need for a master cell bank tested to ICH Q5 standards. All products used within this protocol are available in GMP compliant versions or are risk assessed and approved for use by the European Medicines agency (EMA).
Obtaining both fibroblast and epithelial cell types from a single biopsy limits the effect on the patient. Oral biopsies are easy to obtain, under local anaesthetic with little morbidity or scarring 4 . The use of autologous cells for transplantation removes the need for immunosuppression 36 .
Removal of an epithelial sheet from the lamina propria by Neutral protease II (also known as Dispase II) 37,38 , has been used at various incubation times, temperatures 36,[38][39][40] , and concentrations 36,37 then either removed with forceps 41 or scraped 42 . 4u of Neutral protease II for 16 h at 4 °C plus 10 min at room temperature allowed the whole sheet to be removed using forceps. Fibroblasts are obtained by collagenase digestion 30,43 or outgrowth 44 . Digestion was chosen as it is quicker. Digestion of the epithelial cells is carried out in trypsin-EDTA solution 37,38 , TrypLE is an equivalent animal free alternative. The GentleMACS was used to standardise the homogenisation of the tissue during digestion. The initial fibroblast digestion was carried out using supraglottal tissue to increase the tissue samples available from a limited number of animals. The two tissues were compared with no significant difference in viable cell number or population doublings achieved. All further investigation using Buccal tissue.
FBS is a problematic component for many reasons, the potential risk of transmission of xenogenic antigens and pathogens, animal welfare issues as well as climate change issues associated with cattle, batch variation and limited volume 16 . HPL is a suitable replacement, it is created from expired platelets from apheresis or platelet rich plasma. Large pooled batches reduce batch variation 16 . It has been shown to aid corneal surface healing and epithelialization 45 .
Fibroblast growth was improved by the addition of HPL, although adhesion was reduced. The mechanisms for these observations are not fully understood, although fibronectin and vitronectin in FBS may contribute 46 . This lack of adhesion may be overcome by the use of a human fibronectin as a coating and a richer medium for the 1st 3 days to allow for attachment 47 . Fibroblast in a wound healing environment may undergo differentiation into myofibroblasts 48 . The cells here did not contain αSMA a marker for myofibroblasts, but did show the fibroblast marker vimentin suggesting that fibroblasts were obtained and maintained from this method.
Growth arrest was achieved after two hours of MMC exposure for both porcine and human fibroblasts. This is similar to observations for human oral fibroblast feeder layers treated between 2.5 49 and 3 h 50 . After 5 days, the cells start to undergo apoptosis thereby reducing contamination risk to the final product.
The addition of the ROCK inhibitor Y-27632 provided cells with improved morphology and growth. Y-27632 has been used previously to aid the proliferation of epithelial cells 51 . Its suggested mode of action is reduction of differentiation, allowing the cells to retain stem cell-like characteristics both with 52 and without 3T3 J2 feeder cells 53 or to reduce anoikis 54,55 and epithelial mesenchymal transition 56 due to its action on the actin www.nature.com/scientificreports/ cytoskeleton 57 . Eye drops have been used to treat corneal damage 58 . It has also been shown to improve the cell viability of embryonic stem cells after cryopreservation. Buccal epithelial cells cryopreserved in HSA plus DMSO without Y-27632 were not viable. The addition of a cryopreservation step in a cell therapy is important, allowing the production to be timed in alignment with the patient treatment; reducing any additional waiting or further surgical procedures. Impedance readings show the growth of adherent cells by the increase in impedance as more cells cover the surface of the electrodes. However epithelial cells form electrically tight barriers once their tight junctions have formed. The results in Fig. 3 suggest that these junctions are created earlier in the cells grown in FBS medium. As these are both undefined the exact reason for this is no yet known. However it appears that the junction formation catches up as the junctions are apparent in Fig. 5.
Although 3T3 J2 have been utilised for many years, there is little understanding on the exact reason for their superiority as even other murine embryonic fibroblast do not support epithelial culture in the same manner 59 . The 3T3 FBS model appears to provide the best cell sheet with a greater number of cell layers and definite differentiation, as expected for the current gold standard protocol, there was no detrimental effect seen in retaining stem cells between the conditions and all cells were able to form differentiated multi-layered cell sheets at ALI. Porcine buccal epithelial cells have been shown to have lower CFE capacity compared to human buccal epithelial cells, however they maintain their ability to form cell sheets 60 . P75 is a buccal epithelial stem cell marker 61 . It is these stem cells/holoclones which have been shown to promote the survival of a cell sheet in vivo 23 .
As there was no significant difference between the conditions, human buccal biopsies were digested and the cells tested with primary feeder layers and HPL medium. The human cells have a slower growth rate than the porcine, seen by the lower population doublings achieved 2.95 ± 2.6 vs 4.3 ± 1 by passage 3. Increased population doubling time of 8.4 ± 6 days (Human) vs 5.4 ± 2 days (Porcine) lead to a lower cumulative cell number being achieved by passage 3, 1.2 × 10 7 ± 1.4 × 10 7 (Human) vs 3.4 × 10 7 ± 2.9 × 10 7 (Porcine). The porcine cells are taken from 1 week old pigs, whereas the human tissues where taken from adults and it is established that the growth rate of cells decreases with age. However, despite this, the number of epithelial cells cultured would be sufficient to create a tissue engineered construct of 17 cm 262 for the reconstruction of the upper aerodigestive mucosa. Which is larger than the spilt skin grafts of 2 × 6 cm used for paediatric laryngotracheal stenosis 63 . All epithelial cells created were able to differentiate at ALI suggesting that they were good quality cells.
Porcine cells made a good model here as they gave comparable results to the human tissue and for the testing of the cell sheet within a pig which is a suitable animal for upper aerodigestive disorders 30 .
In conclusion, we have developed a prototype process for the creation of mucosal epithelial cells with sufficient quality and quantity to form the basis of a clinically useful product, without recourse to cells or reagents that would present regulatory issues or clinical risk.

Conclusion
Both fibroblasts and squamous stratified epithelial cells can be obtained from a small, easily accessible, buccal biopsy and grown in entirely regulatory-compliant conditions. These cells are able form cell-cell junctions and maintain their stem cell niche and differentiation capacity. This prototype protocol forms a useful platform for the generation of clinically useful tissue-engineered mucosal replacements for unmet, and imperfectly met, surgical needs.   Colony formation efficiency (CFE). Triplicate feeder layers were set up in a 6 well plate and 500 epithelial cells added, these were incubated for 14 days at 37 °C, 5% CO 2 . All colonies or holoclones, colonies greater than 10 mm diameter with smooth edges, were counted and CFE calculated ((colony number/seeded cell number) × 100). Data represents 3 independent cell lines in triplicate. Creation of Non-contact co-cultures. 2 × 10 4 feeder cells were seeded into the base of the well. 1.2 × 10 5 epithelial cells were seeded onto the top of a Transwell cell culture inserts, 0.4 µm pore size. These were submerged in Eps-media for 3 days, then brought to ALI for a further 14 days, media changed three times a week. Data represents 3 independent cell lines in triplicate.

Methods
Immunofluorescent staining. The Transwell membranes and fibroblasts grown on sterile coverslips were fixed in 4% paraformaldyde, the Transwells were paraffin embedded, and sectioned to 4 µM. They were either, stained with H&E and examined under light microscope (Nikon eclipse Ti-E) or immunofluorescently (IF) stained. For IF staining the sections were deparaffinised using xylene and rehydrated. They underwent heat induced antigen retrieval in boiling Tris-EDTA-Tween buffer ( www.nature.com/scientificreports/ (sc-9090, Santa Cruz Biotechnology Inc, Dallas, USA), and Cytokeratin 1/10 (sc-53251, Santa Cruz), or Citrate buffer (1.92 g anhydrous citric acid, 0.5% tween 20, pH6) for E-Cadherin (ab15148, Abcam, Cambridge, UK) for 20 min. The fibroblasts were permeabilised for 15 min in 0.1% triton x-100 in PBS. Then both were blocked for 1 h in 1% BSA, 0.1% Tween 20 in PBS and incubated with primary antibodies (vimentin, ab8069, Abcam and αSMA 1A4/asm-1 Bio-Techne Ltd, Abingon, UK), overnight at 4 °C. Then incubated with Secondary antibodies anti-mouse Alexa 488 and anti-rabbit Alexa 633 (Life Technologies) for one hour in the dark at room temperature. Cell nuclei were visualised using DAPI (diamindino-2-pheylindole) and mounted using Prolong Gold antifade reagent. These were visualised using a fluorescent light microscope (Olympus Ix70) equipped with a red filter cube (excitation 600, emission 650 nm) a UV filter cube (excitation 350 nm, emission 450 nm) and a green filter cube (excitation 450 nm, emission 550 nm) or a laser scanning confocal Microscope system (Nikon C2 with LU-N4 laser unit) imaging of DAPI (405 nm) Alexa 488 (488 nm) and Alexa 633 (640 nm) and displayed with pseudo-colouring. The images were merged using Image J. Data for Fibroblast represents 3 independent cell lines in triplicate. Data for porcine samples represents 3 independent cell lines with a single Transwell. Data for Human samples represents 2 independent cell lines in triplicate.
Statistical analysis. Data were tested for normality using the Shapiro-Wilk test. Parametric data are shown as means ± standard deviation (SD). The difference between two groups was tested for significance using paired T test and the differences between > 2 groups was tested for significance using two way ANOVA with multiple comparisons. Non-parametric data are shown as median ± interquartile range, and the difference between groups was determined by Kruskal-Wallis with Dunn's correction for multiple comparisons. All data were analysed using Prism (GraphPad Software, San Diego, USA). P < 0.05 were considered statistically significant.