Comparability of scalable, automated hMSC culture using manual and automated process steps

Automation will likely to play a key role in the development of scalable manufacturing processes for cell-based therapies. In this study, we have compared the effects of manual centrifugation and automated non-centrifugation cell culture process steps, performed using TAP biosystems’ CompacT SelecT automated cell culture platform, upon hMSC morphology, number, viability, surface marker expression, Short tandem repeat (STR) profile, and paracrine function. Furthermore, the comparability between flow cytometry analyses of hMSCs, performed at multiple sites, was investigated. No significant difference in hMSC growth and characteristics was observed between cells cultured using either the manual centrifugation process step or the automated non-centrifugation process step, in which residual dissociation agent is carried over. However, some variability in paracrine activity was observed between hMSCs cultured using alternative process steps. It is also apparent that differences in analytical methods can influence the inter-laboratory reproducibility of hMSC flow cytometry analysis, although differences in culture may also contribute to the variability observed in the expression of 2 of the 8 surface markers examined. This novel investigation into the effects of these two key process steps will help to improve the understanding of the influence of automated cell culture upon various cell culture parameters, as well as upon process comparability.


Introduction
Mesenchymal stromal cells (MSCs) were first described by Friedenstein et al. in the 1970s as an adherent, non-haematopoietic cell type, present in the bone marrow, with the capacity to form fibroblastic colonies in vitro, and were assigned the name 'colony-press immune reaction [4]. Although currently any definition of the various MSC therapeutic effects is lacking, recent years have seen the initial developments in the characterisation of MSC mode of action with the ISCT proposing a series of assays capable of determining human MSC immune regulatory properties [5]. However, substantial progress is required before in vivo MSC function is fully understood and comprehensive potency assays for each of the MSC putative mechanisms of action can be developed.
Considerable hurdles also exist in the manufacture of hMSCbased therapies, including the requirement for a large number of cells for cell therapy, and the significant workload, cost and variation that is associated with manual cell culture [6]. The development of automated adherent cell culture platforms, for example TAP Biosystems' CompacT SelecT (Royston, UK), has made automation a viable alternative to manual cell culture processes. However, the examination of the way in which the differences between manual and automated cell culture processes may affect relevant cell culture parameters is limited to a few studies and further investigation is required [7][8][9][10].
Many of the process steps within the CompacT SelecT automated hMSC passage protocols developed in the present study are very similar to that of manual passage protocols. However, these two culture methods differ in one key process step. During manual cell culture, once the dissociation agent (e.g. Trypsin EDTA) has been applied and the cells have been incubated, the enzyme is neutralised, the suspension is centrifuged and the supernatant is aspirated in order to isolate a cell pellet. However, during the automated cell culture process, once the dissociation agent has been applied, it is immediately poured off so that only a residual amount remains. The cells, and remaining enzyme, are then incubated and media is then applied in order to neutralise the dissociation agent.
The specific effect of this dissociation agent carryover and lack of centrifugation in the automated cell culture process has yet to be investigated. Therefore, in this study, we aim to compare the effects of centrifugation and non-centrifugation process steps upon hMSC number, viability, surface marker expression, Short tandem repeat (STR) profile, and paracrine function. In order to facilitate direct comparison and explore process comparability, only the centrifugation and dissociation process steps differed between the manual and automated culture methods in order to minimise any manual variation. This novel investigation into the effects of these two key process steps will help to improve the understanding of the influence of automated cell culture upon various cell culture parameters and may enable further optimisation of this process in the future.
Furthermore, the present study will explore the comparability between independent flow cytometry analyses of hMSC surface marker expression. It is apparent that significant inter-operator variation and differences in methods of analysis exist between users and laboratories, and that these may contribute to the variability in the results of flow cytometry analyses observed between laboratories [11]. Therefore, this study compared the results of hMSC flow cytometry analyses performed at two independent laboratories, in order to examine the effect of inter-laboratory variability.
Additionally, as part of the flow cytometry comparability investigation, a hMSC culture protocol transfer between Loughborough University (UK) and LGC (Teddington, UK) was performed. However, as no CompacT SelecT was available at LGC, the automated centrifugation and non-centrifugation protocols utilised at Loughborough University were replicated manually. This determined whether any observed differences in hMSC characteristics between automated centrifugation and non-centrifugation process steps were maintained in the manual versions of these process steps. This therefore allowed for the comparability between manual and automated versions of the centrifugation and non-centrifugation process steps to be investigated.
The present study will also examine the expression of various cytokines and growth factors which have been associated with a number of the more established MSC modes of action, including immune modulation, angiogenesis, and anti-apoptosis. In particular, the hMSC-mediated secretion of prostaglandin E-2 (PGE-2) and vascular endothelial growth factor (VEGF) into the cell culture medium, as well as the hMSC-mediated indoleamine 2,3 dioxygenase (IDO) activity, after pre-treatment, was measured prior to, and after, culture using each of the alternative process steps.
Culture expanded hMSC IDO secretion, stimulated by proinflammatory interferon gamma (IFN-␥) [12], has been strongly linked to the suppression of T, B and dendritic cells [13]. Natural killer (NK) cell function has also been found to be inhibited by both IDO and PGE-2 secretion by MSCs [14]. The suppression of T cell response, through the upregulation of IDO, has been associated with the depletion of tryptophan and the accumulation of toxic metabolites [15]. More recently, IDO has been found to be a central effector of MSC T cell suppressive function and a strong correlation between magnitude of IDO expression and suppression of T cell proliferation has been discovered [16].
PGE-2 has been highlighted as a central mediator of the inhibitory effects of MSCs upon immune cells [17]. It has been discovered [18,19] that, in response to pro-inflammatory mediators, MSCs secrete PGE-2 which requires intimate association or cell to cell contact to bind to the EP2 and EP4 receptors of resident macrophages. The binding of PGE-2 to host macrophages drives the transition of these cells from their classical M1, pro-inflammatory phenotype towards an M2, anti-inflammatory phenotype in which these macrophages secrete anti-inflammatory mediators, including IL-10 and IL-1 receptor agonist [19].
The hypoxic culture and preconditioning of hMSCs, in which ≤1% O 2 is commonly used, has been found to induce the overexpression of VEGF, an established pro-angiogenic factor [20][21][22][23][24][25]. A number of these studies have identified VEGF as the most critical factor responsible for the angiogenic properties of MSCs in vivo [21][22][23][24][25][26]. Furthermore, the expression of this protein by MSCs has also been associated with an anti-apoptotic effect upon host cells, improving cell survival in a number of studies [21,24]. Therefore, it is clear that the expression of PGE-2, IDO and VEGF are crucially important for the anti-inflammatory, immune modulatory, angiogenic and anti-apoptotic MSC modes of action. Thus, the measurement of these paracrine factors in the present study, after hMSC priming, may allow for the determination and comparison of the functional activity of hMSCs cultured using either manual or automated process steps.

CompacT SelecT preparation & calibration
Prior to the performance of any automated protocol on the TAP Biosystems' CompacT SelecT platform, the machine was prepared for use by ensuring a sufficient number of pipette tips were loaded, sufficient new T175 flasks were available, that adequate volume of reagents were loaded aseptically, and that the required sterile plastic tubing (Watson-Marlow Pumps, Falmouth, UK) was connected to allow for reagents to be pumped using the peristaltic pump system.
In order to ensure that the required volumes of reagent are dispensed throughout each protocol, a calibration step is performed prior to the performance of each CompacT SelecT protocol. Briefly, the plastic tubing was primed and a small volume of reagent was dispensed into a BD Falcon TM T175 tissue culture flask (BD Biosciences, San Jose, USA). The flask was then exported and the contents were weighed on digital scales to determine the volume of reagent dispensed, assuming that 1 mL of reagent weighs 1 g. This value can then be entered into the CompacT SelecT software in order to calibrate the peristaltic pump system, which adjusts the subsequent dispensing steps accordingly.

Creation of a hMSC working bank
The hMSCs utilised in the present study were previously derived from bone marrow mononuclear cells (BM-MNCs), purchased from Lonza (Basel, Switzerland), by colleagues in the Centre of Biological Engineering (Loughborough, UK). Briefly, the BM-MNCs were plated on tissue culture plastic and non-adherent cells were removed through culture over 2 passages. The isolated hMSCs were then cryopreserved (P0), and one vial of these cells was utilised as the source material for the generation of the hMSC working bank in the present study.
The initial hMSC vial (P0), containing 1 × 10 6 cells, was removed from cryostorage, thawed, centrifuged, and the supernatant aspirated in order to remove the cryopreservant. The cells were then re-suspended in pre-warmed DMEM high glucose (HG) Glu-taMAX (Life Technologies, Thermo Fisher Scientific, Waltham, USA) supplemented with 10% Qualified Fetal Bovine Serum (FBS) (Gibco ® , Life Technologies) and manually seeded into a BD Falcon TM Barcoded T175 flask (BD Biosciences). Subsequently, complete medium exchanges were performed every 3 days after seeding or passage. Once 80% confluency of the culture had been confirmed through visual examination, after approximately 6 days, the mother flask was manually passaged into four new daughter flasks. Once the 4 daughter flasks had reached 80% confluency, each of the flasks were imported into the CompacT SelecT automated cell culture platform and passaged into 16 granddaughter flasks using an automated passage protocol. Once these granddaughter flasks reached approximately 80% confluency, the flasks were exported from the CompacT SelecT and manually prepared for cryopreservation by washing and dissociating cells. The enzyme (Trypsin 0.05% EDTA, Life Technologies) was then neutralised with complete medium and the suspension was centrifuged (276 RCF × 5 min) in order to isolate a cell pellet. Finally, the cell pellet was re-suspended in 90% FBS and 10% dimethyl sulfoxide (DMSO) (Sigma Aldrich, St Louis, USA) and 2 × 10 6 cells were transferred into each cryovial. Each cryovial was stored at −80 • C for 24 h, in a Mr. Frosty freezing container (Thermo Fisher Scientific), before being transferred into cryostorage. The cells generated for this working cell bank represent 'Baseline' hMSCs (P2), to which the hMSCs cultured utilising centrifugation or non-centrifugation process steps were compared.

hMSC centrifugation culture method
For each of the four experimental runs, one cryopreserved hMSC vial (2 × 10 6 cells) was thawed in a 37 • C water bath, re-suspended in DMEM HG GlutaMAX TM supplemented with 10% FBS, centrifuged (400 RCF × 5 min), and the supernatant aspirated in order to remove the cryopreservant. The cells were then re-suspended in complete medium, transferred into a barcoded T175 flask and imported into the CompacT SelecT, at which point an automated seeding protocol was initiated, during which 7 × 10 5 cells were seeded into a new barcoded T175 flask (P3). Confluency was examined daily using microscopy and a complete medium exchange was performed 72 h after seeding and each passage. Once the culture reached 80% confluency, after approximately 6 days, a hMSC pre-centrifugation CompacT SelecT automated protocol was utilised to detach the cells from the mother flask and to obtain cell count, viability and aggregation data using the Cedex automated cell counter. The mother flask was then "borrowed" from the CompacT SelecT and the cell suspension was centrifuged (400 RCF × 5 min) in order to isolate a cell pellet. After the cell pellet was re-suspended in fresh culture medium, the cells were transferred back into the mother flask which was then imported back into the CompacT SelecT and a hMSC post-centrifugation automated protocol was performed in order to obtain cell count, viability and aggregation data and to passage the mother flask at a 1:4 split ratio, seeding four new daughter T175 flasks with 7 × 10 5 cells (P4). Once each of the 4 daughter flasks reached approximately 80% confluency, the preand post-centrifugation protocols were again utilised in order to passage the 4 daughter flasks into 16 new granddaughter flasks (P5) and to obtain cell count, viability and aggregation data. Once the granddaughter flasks reached approximately 80% confluency, all 16 flasks were pooled, using a hMSC pool automated protocol, and the cells were isolated using centrifugation (400 RCF × 5 min), re-suspended in cryopreservation medium (90% FBS & 10% DMSO) and cryopreserved in a sufficient number of cryovials.

hMSC non-centrifugation culture method
For each of the four experimental runs, similar thawing, centrifugation, and re-suspension processes as in the hMSC centrifugation culture method were utilised in order to remove the cryopreservant and suspend the cells in complete culture medium. Once again, the cells were transferred into a barcoded T175 flask and imported into the CompacT SelecT, at which point an automated seeding protocol was again initiated, and 7 × 10 5 cells were seeded into a new barcoded T175 flask (P3). Similarly to the centrifugation method, confluency was examined daily using microscopy and a complete medium exchange was performed 72 h after seeding and each passage. However, once the cells reached 80% confluency, after approximately 6 days, a hMSC noncentrifugation protocol was utilised, in which an enzyme pour-off step was utilised to dissociate the cells. During this dissociation process, Trypsin EDTA was applied, ensuring that the entire surface of the T175 flask is coated, and was immediately poured off, leaving only a residual layer of enzyme coating the flask which was then incubated in order to dissociate the cells. The hMSC non-centrifugation protocol also obtained cell count, viability and aggregation data and passaged the mother flask into four daughter flasks, seeding each with 7 × 10 5 cells (P4). Once each of the 4 daughter flasks reached approximately 80% confluency, the noncentrifugation protocol was again utilised to passage each of the 4 flask into 16 new granddaughter flasks (P5) and obtain cell count, viability and aggregation data. Finally, once each of the 16 granddaughter flasks reached approximately 80% confluency, all 16 flasks were pooled and the cells were isolated using centrifugation (400 RCF × 5 min), re-suspended in cryopreservation medium (90% FBS & 10% DMSO) and cryopreserved in a sufficient number of cryovials.

Surface marker expression
2.2.1. Loughborough University flow cytometry analysis of hMSCs cultured using CompacT SelecT automated centrifugation & non-centrifugation process steps Baseline (P2), centrifugation (P5), and non-centrifugation (P5) hMSCs were thawed, seeded in separate T175 flasks at a density of 4 × 10 3 cells/cm 2 (7 × 10 5 cells/T175 flask), and cultured, as described previously, until they reached approximately 80% confluency (∼6 days). Once confluent, the cells were dissociated, using trypsin EDTA, a cell count was performed, and 1 × 10 5 cells were plated per well in the relevant number of wells of a 96-well plate. The cells were prepared for flow cytometry using the relevant reagents, antibodies and isotype controls from the BD Stemflow TM kit (BD Biosciences) as indicated in the manufacturer's instructions. Each sample, plated with the relevant antibody cocktail, antibody drop-in, or isotype control, was repeated in triplicate. To determine the immunophenotype of baseline, centrifugation, and non-centrifugation hMSCs, multicolour flow cytometry was performed utilising positive and negative antibody cocktails for the primary CD markers that constitute the ISCT minimum criteria for hMSCs [3]. This included the positive expression of CD105, CD90, CD73 and CD44; and the negative expression of CD34, CD45, CD11b, CD19 and HLA-DR. The utilisation of pre-conjugated antibody cocktails allowed for multiparameter analysis of surface marker expression at the single cell level. This method has been previously validated by colleagues at Loughborough University [27].
The hMSC samples were analysed on the BD FACSCanto II (BD Biosciences) which was operated using the FACSDiva software version 6.1.3. For analysis, the flow cytometry data was exported in FCS 3 format and analysed using FlowJo software v10.

Comparability in surface marker expression
LGC hMSC culture using manual centrifugation & non-centrifugation process steps for flow cytometry analysis P2 hMSCs ('Baseline', Working Bank) were transferred from Loughborough University to LGC. From P3 to P5, hMSCs were cultured and expanded using manual protocols mimicking that of the manual (centrifugation) and automated (non-centrifugation) process steps performed at Loughborough University. In order to manually perform both the centrifugation and non-centrifugation culture processes, the automated protocols, as described previously, were replicated as closely as possible, with similar reagents and process steps, but without the use of the CompacT SelecT automated cell culture platform. In contrast to the automated processes, cell counting was performed using the Vi-Cell XR (Beckman Coulter, Brea, USA) during both the manual centrifugation and noncentrifugation processes, rather than using the Cedex automated cell counter. P5 cells were pooled at 1 × 10 6 cells per ml in DMEM with 10% FBS and 10% DMSO and cryopreserved in liquid nitrogen.

LGC flow cytometry analysis of hMSCs cultured using manual centrifugation & non-centrifugation process steps
Briefly, pre-banked hMSCs (P5) were analysed directly after cryopreservation in liquid nitrogen and prepared by thawing, removing the cryoprotectant, and fixing in 1× cell fixation buffer (BD Biosciences), incubating in Fc blocking reagent (BioLegend, San Diego, USA). Next, 2 × 10 5 cells were incubated with the selected group of antibody/isotype control cocktails in staining buffer (BioLegend) and washed before being transferred into 5 ml polystyrene round-bottom flow cytometry tubes (BD Biosciences).
To identify the correct phenotypic pattern of BM isolated hMSCs, immunophenotyping was performed by flow cytometry utilising the most commonly used monoclonal antibodies based upon the literature (see Fig. 3 and Appendix 1.1 of Supplementary material). The expression level of different surface markers was analysed by direct immunolabeling approach. Staining was performed on passage 5 cells from each condition (post-cent and non-cent) using the selected group of antibody cocktails with the corresponding isotype hMSC samples were analysed on the BD FACSCanto II (BD Biosciences), operated through the FACSDiva software version 6.1.3, and exported the data in FCS 3 format for analysis using FlowJo software v10.

Cell line authentication (CLA) & short tandem repeat (STR) profiling
STR profiling was performed by LGC Standards (Teddington, UK) through their commercially available cell line authentication (CLA) service. Samples were prepared for transport and CLA analysis by following the protocol provided by LGC Standards. One cryovial of hMSCs from the Working Cell Bank (Baseline, P2) (2 × 10 6 cells), 1 cryovial of post-centrifugation hMSCs (P5) (2 × 10 6 cells), and 1 cryovial of non-centrifugation hMSCs (P5) (2 × 10 6 cells) were removed from cryostorage, thawed and microfuged (73 RCF × 5 min). The supernatant was then aspirated and a PBS wash was applied, in order to remove any cryopreservant. The samples were again microfuged and the supernatant was again aspirated in order to isolate the cell pellets. Finally, the samples were re-suspended in 400 L of transport buffer, provided by LGC Standards, which lysed the cells and preserved the gDNA. The samples were then packaged and sent to LGC Standards, where the CLA analysis was undertaken and the STR profiles for each sample were compared. Fig. 2. Process diagram illustrating the differences between the CompacT SelecT manual (centrifugation) and automated (non-centrifugation) hMSC culture process steps.  Table of selected surface markers for LGC hMSC cell immunophenotyping. Antibodies were categorised in groups as "generic", "stress", "MSC positive (+)" and MSC negative (−) markers. Pre-conjugated antibodies were selected to generate four groups of antibody cocktails.
2.5. Paracrine functionality assays 2.5.1. hMSC pre-treatment Three cryovials of hMSCs from the Working Cell Bank (Baseline, P2) (2 × 10 6 cells), three cryovials of post-centrifugation hMSCs (P5) (2 × 10 6 cells), and three cryovials of non-centrifugation hMSCs (P5) (2 × 10 6 cells) were thawed and 1 × 10 6 cells from each cryovial were seeded into nine separate T175 flasks with DMEM HG GlutaMAX TM with 10% FBS. These nine T175 flasks were divided into three groups of three T175 flasks, with each group containing one flask from each of the baseline, post-centrifugation, and non-centrifugation conditions. The three groups consisted of 'inflammatory pre-treatment' (IN-PT), 'hypoxic pre-treatment' (HY-PT), and no treatment (NT) groups. After three days of culture, a media exchange was performed for each of the 9 flasks. The medium was aspirated from all hMSC flasks, and the flasks in each of the groups were specifically pre-treated: • The IN-PT hMSCs were treated with DMEM HG GlutaMAX TM with 10% FBS, as well as 10 ng/mL IFN-␥ (Sigma-Aldrich) and 15 ng/ml TNF-␣ (Sigma-Aldrich) for a further 72 h. These conditions were selected based upon the ISCT guidelines for immunological characterisation of MSCs [5]. • The HY-PT hMSCs were treated with DMEM HG GlutaMAX TM with 10% FBS, which had been pre-conditioned at 1% O 2 overnight, and were incubated at 37 • C at 1% O 2 for a further 48 h. These conditions were selected based upon established hypoxic hMSC preconditioning methods [28,22,29,25]. Chacko et al. [25] demonstrated, through hypoxic pre-treatment for 72 h, that VEGF expression was greatest after 48 h of pre-treatment. • The NT hMSCs were treated with DMEM HG GlutaMAX TM with 10% FBS and incubated at 37 • C at 5% CO 2 for a further 72 h.
After the hMSC populations were exposed to either the IN-PT, HY-PT or NT conditions, the conditioned medium from each flask was collected and stored at −80 • C.

Kynurenine quantification & IDO activity measurement
The enzyme indoleamine 2,3-dioxygenase (IDO) converts tryptophan to kynurenine, and the following assay was used to photometrically determine the concentration of kynurenine in inflammatory pre-treated (IN-PT) or no treatment (NT) conditioned medium samples from each of the baseline, post-centrifugation, and non-centrifugation hMSC groups. This assay has been successfully utilised in a number of previous studies in order to quantify IDO activity for a number of cell types, including hMSCs [30][31][32][33].
150 L of each of the conditioned medium samples, as well as a fresh DMEM with 10% FBS sample, were pipetted into separate 5 mL Eppendorf tubes (Eppendorf, Hamburg, Germany) and 50 L of 30% trichloroacetic acid (TCA) (Sigma-Aldrich) was added to each. Each of the samples was then incubated for 15 min at 50 • C and microfuged at 10,000 g for 5 min in order to hydrolyse N-formylkynurenine to kynurenine. 75 L of the supernatant from each of the samples was transferred to a 96-well plate. A serial dilution of 100 m kynurenine solution (Sigma-Aldrich) was used as a standard (100 m to 0 m) to allow for the generation of a stan-dard curve. 75 l of Ehrlich's reagent (1% p-dimethylbenzaldehyde in glacial acetic acid) (Sigma-Aldrich) was added to each well and the plate was incubated for 10 min at room temperature. Finally, the fluorescence of each well was determined by measuring the absorbance at 492 nm. Each sample was measured in duplicate and a repeat of the assay was performed in order to increase the reliability of the data. Sample data was normalised against the average absorbance of two culture medium (DMEM HG GlutaMAX TM with 10% FBS) blanks. Culture medium normalised values were then standardised utilising the kynurenine standard curve and the average values were then generated.

PGE-2 quantification
The prostaglandin E2 enzyme-linked immunosorbant assay (ELISA) was performed following the manufacturer's directions (Life Technologies TM , Invitrogen TM Novex ® Prostaglandin E2Human ELISA Kit). All buffers, tracers and antibodies were prepared as described in the manufacturer's directions. Prostaglandin E2 standard was prepared at the recommended concentrations, however, culture medium rather than Tris buffer was used for the dilution of the standard curve. This is recommended in the manufacturer's directions if performing an assay of culture medium samples. Two blank wells, two non-specific binding wells, three maximum binding wells, one total activity well, sixteen standard wells (8 standards in duplicate), and eighteen sample wells (6 wells in triplicate) were prepared as indicated in the manufacturer's guidelines. After incubation for 90 min in the dark, on an orbital shaker, the absorbance was measured using a plate reader at a wavelength of 410 nm.

VEGF quantification
The VEGF enzyme-linked immunosorbant assay (ELISA) was performed following the manufacturer's directions (Life Technologies TM , Invitrogen TM Novex ® Human VEGF ELISA Kit). All buffers, tracers and antibodies were prepared as described in the manufacturer's directions. VEGF standard was diluted in standard diluent buffer and the manufacturer's directions were followed in order to generate a standard curve. Two chromogen blank wells, two zero wells, sixteen standard wells (8 standards in duplicate), and eighteen sample wells (6 wells in triplicate) were prepared as indicated in the manufacturer's guidelines. After a final incubation of 30 min at room temperature, in the dark, the absorbance was measured using a plate reader at a wavelength of 450 nm. VEGF concentration in each sample was determined by plotting the absorbance of samples against the VEGF standard curve. In order to correct for the 1:2 dilution of samples performed during the assay, VEGF concentrations of each sample were multiplied by 2.

Statistical analyses
Experimental data regarding the viable cell density, viable cell yield, and viability of pre-centrifugation, post-centrifugation and non-centrifugation hMSCs within each passage was assessed using one-way analysis of variance (ANOVA) multiparameter analysis, utilising the IBM SPSS statistical software (Armonk, USA), to determine significant differences. One-way ANOVAs were also used to assess the significance of differences in the standard deviations (SD) of the viable cell densities, viable cell yields, and viabilities of hMSC pre-centrifugation, post-centrifugation, and non-centrifugation samples, in each of the four batches, from the second passage. Two-way ANOVAs were used to assess significant differences in the viable cell density, viable cell yield, and viability of pre-centrifugation, post-centrifugation and non-centrifugation hMSCs across all passages. Finally, One-way ANOVAs were used to determine the significance of any differences in the kynurenine, PGE-2, and VEGF concentrations of untreated and pre-treated baseline, post-centrifugation, and non-centrifugation hMSCs in their respective assays. The cut-off value for statistical significance (p) was set at 0.05. Tukey's honest significant difference (HSD) Post-Hoc tests were used to perform multiple comparisons of the hMSC growth and paracrine functionality data.

Morphology
One basic, preliminary method of characterising a cell population is through visual examination of a culture under an inverted light microscope with a digital camera attachment (Nikon Eclipse Ts100, Nikon, Tokyo, Japan). Although, this method does not give an accurate or quantifiable measure of the identity or characteristics of the cells, it can be used as a rudimentary method of verifying that the cells have not differentiated or undergone any substantial transformations. Fig. 4 demonstrates that no difference in morphology was observed between hMSCs cultured utilising manual or automated process steps, and that both hMSC populations exhibited a spindle shaped, fibroblast-like morphology. This  CD90 Expression level (%)
Non-Cent. would suggest that neither the centrifugation step in the manual process, nor the trypsin carryover in the automated process, had a substantial effect upon the morphology of the hMSCs.

Surface marker expression
The ISCT minimal criteria for the phenotypic expression of hMSCs indicates that ≥95% of the cell population should express CD105, CD73 and CD90, and that ≤2% of the cell population should express CD45, CD34, CD11b, CD19 or HLA-DR [3]. Differences in the expression of hMSC surface markers, between automated and manually-replicated passage protocols, were examined by comparing flow cytometry analyses performed at two laboratories using either BD Biosciences Stemflow TM kit (Loughborough University) or in-house prepared antibody cocktails (LGC).
As demonstrated in Fig. 5, the multicolour analysis performed at Loughborough University allowed for the demonstration of the coexpression of multiple hMSC positive markers, as well as the lack of co-expression of multiple negative markers and haematopoietic antigens, on single cells. This method conforms to that recommended in the ISCT minimal criteria [3]. Furthermore, no significant difference in positive or negative expression was observed between hMSCs cultured utilising alternative process steps, and between early (P2) and late (P5) passage hMSCs, with all populations expressing ≥95% positive marker expression and ≤2% negative marker expression (see Appendices 2.1 and 2.2 of Supplementary material).

Comparability in surface marker expression
The immunophenotype of hMSCs cultured and analysed by LGC met the majority of hMSC surface marker expression criteria, in concordance with Loughborough University's results (see Fig. 6 and Appendix 3.1 of Supplementary material). Significant expression of the primary hMSC positive markers (CD105, CD73 and CD90) was found. However, both CD105 and CD146 expression was lower than expected in the LGC analysis, with CD105 expression falling below the threshold set within the ISCT minimal criteria. The low expression of CD146 (MCAM) may reflect heterogeneity in the expression of this marker between individual donors. Alternatively, lower CD146 expression may reflect the later passage number of hMSCs used for flow cytometry analysis as this marker has been associated with cells that possess a shorter doubling time [34].
The typical hMSC negative markers CD45, CD11b and HLA-DR conformed to the ISCT minimal criteria, with expression observed at <2%. However, discordant CD34 expression was observed between the two laboratories, with higher percentage expression of CD34 in LGC's analysis compared to that of Loughborough University. There is significant uncertainty regarding the CD34 expression of hMSCs which has been identified as a hMSC marker in adult adipose tissue [35]. Although CD34 negativity is outlined in the ISCT minimal criteria, further research on these cells has indicated that the lack of CD34 expression may be an in vitro artefact and these cells may in fact be CD34+ in their in vivo niche [36].
In addition to the ISCT minimal phenotypic criteria for the identification of hMSCs, it has been identified that these cells can stain positive for a number of surface markers, including CD9, CD29, CD44, CD63, CD99, CD106, CD146 and negative for CD14, CD34, CD45 and CD133 [37].
LGC flow cytometry analysis also identified that a high proportion of hMSCs expressed generic and stress-related surface markers CD9, CD29, CD63, as well as partial expression of CD146 and CD98 (Fig. 6).
The LGC flow cytometric analysis of hMSCs, cultured using either the manually replicated centrifugation or non-centrifugation process steps over two passages, revealed a similar level of surface marker expression in both conditions. Finally, the inter-laboratory comparison of hMSC surface marker expression, suggests that the utilisation of alternative methods of analysis may be a source of variability in the results obtained from flow cytometry. This highlights the need for standardisation and calibration of flow cytometry methodologies and equipment. However, differences in culture protocols between sites, with LGC manually replicating the automated process steps, may have also influenced this variation in surface marker expression.

Cell line authentication (CLA) analysis & short tandem repeat (STR) profiling
Short tandem repeat (STR) profiling is a method used to amplify and compare specific loci on the DNA of multiple cell populations, and is often used in forensic analysis [38,39]. An STR is a unit of multiple nucleotides which are repeated many times in sequence along the length of a DNA strand, and by counting the number of these repeating units at specific loci within the DNA, an individual profile of the sample can be generated. The STR profile analysis used in this study was provided by LGC Standards as part of their cell line authentication (CLA) service, and this is used to measure the difference in STRs at 16 loci within the genome between samples. The output of each of these analyses is an electropherogram (EPG) illustrating the relative expression of STRs at each loci, as indicated by peaks in fluorescence on the EPG.
In the present study, STR profiles were generated for each of the baseline, post-centrifugation and non-centrifugation samples, and the post-and non-centrifugation EPGs were compared to that of the baseline in order to determine whether the manual or automated process steps had any significant effect upon the hMSC STR profile. Fig. 7 demonstrates that no significant difference was observed between the short tandem repeat (STR) profiles of the baseline, post-centrifugation and non-centrifugation hMSCs, and that sample each expressed peaks in STRs at the same loci.
However, comparison of the EPGs of the baseline, noncentrifugation, and post-centrifugation hMSC samples highlights possible differences in peak sizes and levels of STR expression, which may therefore indicate differences in the number of copies of the STR at each locus. However, it is also apparent that variations in fluorescence may occur between experimental runs, due to differential amplifications of the alleles or primer-binding site mutations [40], and therefore examining differences in peak sizes may be redundant. Therefore, variations in peak size between hMSC samples provided in this experiment were not considered to represent a significant change in the identity of a cell sample and each of the samples were considered to have identical STR profiles.

Viable cell yield
The chart above (Fig. 8), illustrates the average viable hMSC yield of centrifugation and non-centrifugation flasks from all four batches over two passages, as well as the viable cell number prior to seeding each of the mother flasks for each experimental run. These results demonstrate that no significant difference between viable hMSC yield obtained utilising pre-centrifugation, postcentrifugation, and non-centrifugation process steps was observed in each passage or across all passages. Therefore, it can also be concluded that the presence of residual trypsin carried over dur-ing the automated dissociation process step had no detrimental effect upon the number of obtainable cells compared to when this residual dissociation agent was removed using a manual centrifugation process step, commonly utilised in traditional cell culture protocols.
Furthermore, Fig. 8 may indicate a decrease viable hMSC yield after the process of centrifugation. This could suggest that that further optimisation of the centrifugation step is required, as excessively high relative centrifugal force (RCF) may induce cell death and insufficient RCF may not create an adequate cell pellet. However, statistical analysis does not support this trend and no significant difference between pre-and post-centrifugation viable cell yield was observed (p = 0.524). Furthermore, the RCF and duration of centrifugation utilised in the present study are often recommended in industrial cell culture protocols.
Additionally, the results of the present study indicate that the viable cell yields in P4 were significantly greater than those in P5 (p = 0.036), regardless of whether the manual or automated process steps were utilised. This implies that the growth rate of the hMSCs decreased over two passages, which supports the findings of previous investigations [9]. This may suggest that the cells had begun to enter senescence over the duration of their culture over two passages, and thus their growth rate began to decrease.
Finally, although no apparent difference in viable cell yield between hMSCs cultured utilising manual or automated process steps, in each batch over two passages, was determined, the standard deviations data (Fig. 8) indicates that greater variability in viable cell yield may have been observed in non-centrifugation hMSC samples compared to pre-or post-centrifugation hMSC samples. This may suggest that the utilisation of the non-centrifugation process step produces less consistent viable cell yields compared to when the centrifugation process step is employed and samples are measured either pre-or post-centrifugation. However, no significant difference between the standard deviations of the viable hMSC yield data generated using each of the process steps was found, and therefore, it cannot be concluded that any difference in variability between process steps was observed.
Although, the viable cell yield data illustrates a non-significant trend for greater variation when utilising the non-centrifugation process step compared to the centrifugation process step using the CompacT SelecT platform, it is likely that a fully automated cell culture process will be more consistent in comparison to a fully manual cell culture process. A reduction in variability is a typical outcome of the automation of manual cell culture processes and has been demonstrated in a number of previous studies [41,8,10].
It is likely that any increased variability within the automated process step can be attributed to differences between the mixing steps utilised in the manual and automated process steps. During the manual centrifugation process step, after the cells have been centrifuged, the cells are re-suspended in fresh culture medium and thoroughly mixed to ensure that a single cell suspension is generated. However, during the automated dissociation process step, the cell suspension is re-suspended using a defined number of automated mixing steps, which may not have been sufficient to generate an evenly mixed single cell suspension for each hMSC population. This failure to generate a homogenous single cell suspension in flasks undergoing a passage is likely to have created differences in the number of cells seeded into new daughter flasks. Also, the inadequate mixing of hMSC suspensions when utilising the non-centrifugation process step may have also led to sampling process errors. Without sufficient re-suspension of hMSC populations, it is likely that heterogeneity in cell counts would exist within a suspension, when multiple samples are taken from within a single suspension. Additionally, this variability may also be observed between suspensions, when a single sample is taken from multiple suspensions, and this inconsistency may be due to differences in cell aggregation, settling and adherence. Therefore, further optimisation studies may be required to develop a protocol with sufficient mixing steps to generate a single cell suspension and to allow for more consistent hMSC seeding.

Live cell aggregate size
In the chart above (Fig. 9), the average number and size of viable hMSC aggregates from the pre-, post-& non-centrifugation samples in batches 1-4, across two passages, are plotted. This data demonstrates that after centrifugation, an increase in the number of cells present in smaller aggregate sizes occurred. This would imply that the centrifugation process step may have reduced the larger aggregates of cells into smaller aggregates or into single cells, which is often reported with mammalian cells [42]. However, the lack of a greater number of cells in larger aggregates in the pre-centrifugation samples may not support this conclusion. Although, the centrifugation process step may reduce the obtainable viable cell yield, when the standard deviations of the pre-and post-centrifugation viable cell yields are compared, it appears that it may reduce the variability in the cell count. This may be attributed to the increased number of small aggregates observed after the centrifugation process step which may therefore lead to more reliable cell counts and more consistent seeding of daughter flasks.
It is also apparent that no difference in live cell aggregate size was observed when either the centrifugation or non-centrifugation process steps were utilised. However, the larger standard deviations in live cell aggregate size after the non-centrifugation process step may support the viable cell yield data and demonstrates the variability observed when the non-centrifugation process step was utilised.

Viability
The figure above (Fig. 10) summarises the mean viabilities for hMSC samples after seeding, pre-centrifugation, postcentrifugation, and non-centrifugation, across all four batches, and over two passages. These results indicate that no significant difference in the viability of hMSC cultures was observed between centrifugation and non-centrifugation cell culture process steps, which would therefore suggest that the enzyme carryover in the non-centrifugation process step had no detrimental effect upon cell viability. It is also apparent that no significant difference between pre-centrifugation and post-centrifugation hMSC viability was demonstrated, and therefore that the process of centrifugation had no significant effect upon hMSC viability. Furthermore, no significant difference in the variation of hMSC was observed between process steps or passages, and this variation was approximately 4% (±2%) viability in each group.

Kynurenine quantification & IDO activity measurement
The expression of IDO by hMSCs has been associated with the immunosuppressive properties of these cells, and this enzyme has been found to be one of the key bioactive factors responsible for the suppression of T, B, dendritic and NK cells [13,14,16]. Waterman et al. [43] discovered that, after toll-like receptor 3 (TLR3) priming, hMSCs exhibited an immunossupressive phenotype during which these cells demonstrated an increased IDO expression which acted to suppress T lymphocyte activation. Meisel et al. [30] previously described the mechanism through which the hMSC secreted IDO  inhibits T-cell activation, with their findings indicating that the expression of IDO, which converts tryptophan to kynurenine, by hMSCs significantly inhibited allogeneic T-cell responses, that the addition of tryptophan restored T-cell proliferation, and therefore that it is the IDO-mediated catabolism of tryptophan to kynurenine that is responsible for the inhibitory effects of hMSC secreted IDO upon T-cell activity. Also, more recently, Francois et al. [44] determined that tryptophan inhibition abolished hMSC suppression of T cell proliferation, that the level of IDO expression was correlated to hMSC immunosuppressive potential, and that hMSC secreted IDO caused monocytes to differentiate into immunosuppressive macrophages.
However, it is generally accepted that hMSCs must be stimulated by pro-inflammatory cytokines, particularly IFN-␥, in order to upregulate IDO expression [45,30]. The data from the present study, in which kynurenine concentration was measured photometrically, supports this hypothesis and it was determined that untreated hMSCs did not constitutively express IDO and that pre-treatment with inflammatory cytokines, specifically IFN-␥ and TNF-␣, was required to stimulate the expression of IDO by hMSCs. As untreated hMSCs did not express IDO, and therefore did not induce significant breakdown of tryptophan or the resultant increase in kynurenine concentration, no significant difference between untreated baseline, centrifugation or non-centrifugation hMSCs was observed. Furthermore, as inflammatory pre-treatment significantly increased IDO activity in all hMSC samples, the results of the present study indicate that hMSCs cultured utilising either manual or automated process steps expressed IDO in response to IFN-␥ stimulation, as is typical of hMSCs [12].
The results of this IDO assay may also indicate a nonsignificant trend for the kynurenine concentration in pre-treated Centrifugation (p = 0.763) and non-centrifugation (p = 0.273) hMSC conditioned medium to be reduced compared to that of pretreated Baseline hMSC conditioned medium, which would suggest that the IDO activity of the hMSCs decreased over two passages, regardless of the process step utilised. Additionally, Fig. 11 also illustrates a non-significant trend for a greater kynurenine concentration in inflammatory pre-treated centrifugation hMSC conditioned medium compared to that of inflammatory pre-treated non-centrifugation hMSC conditioned medium (p = 0.943), which would suggest that the IDO activity of pre-treated hMSCs cultured utilising the centrifugation process step was greater than that of pre-treated hMSCs cultured utilising the non-centrifugation process step. However, given the large standard deviation of the baseline hMSC conditioned medium samples and the lack of significance of these trends, further optimisation of the kynurenine assay is required in order to further understand IDO enzyme activity.
Therefore, from the IDO assay performed in the present study, it can be concluded that untreated hMSCs do not constitutively express IDO and that inflammatory pre-treatment is required in order to induce IDO secretion. It can also be proposed that the culture process, regardless of the process step utilised, may have a detrimental impact upon hMSC IDO expression after inflammatory pre-treatment, and that the automated non-centrifugation process step may have a slightly greater detrimental impact upon IDO activity compared to the manual centrifugation process step.
Given the correlation between IDO production and hMSC immunosuppressive potential [44], any effect of the culture process upon hMSC IDO expression may significantly impact their clinical efficacy, which would represent a major drawback for the manufacture of hMSC-based therapies. The IDO activity observed in all inflammatory pre-treated hMSC conditioned medium samples was sufficient to be of clinical relevance, with Tattevin et al. [46] previously reporting that a kynurenine concentration of ≤20 M was correlated with an increased anti-inflammatory response in sepsis patients. Therefore, as the results of the present study indicate that each of the inflammatory pre-treated hMSC conditioned medium samples exhibited a kynurenine concentration of >20 M, it is clear that each of the inflammatory pre-treated hMSC populations produced a clinically relevant immunosuppressive effect.
In addition to the inconsistency of the IDO assay, it must also be recognised that the innate variability between hMSC populations may have been likely to influence the results of this assay. For example, variability between donors, including age and gender, has been found to significantly influence hMSC gene expression and immune-suppressive function [47,48]. It has also been observed that the expansion of hMSCs in monolayer culture can lead to alterations in gene expression, therefore creating differences in gene expression profiles between populations [49,50]. Therefore, the heterogeneous nature of these cells may have contributed to the non-significant variability in IDO activity, between hMSCs cultured utilising each of the alternative process steps, observed in the present study.

PGE-2 Quantification
The PGE-2 mediated immunosuppression of hMSCs was first observed in a mouse sepsis model in which bone marrow derived stromal cells were administered [51]. It has since been determined that pro-inflammatory mediators activate sensors on the MSC membrane causing upregulation of the expression of COX 2 , and other arachidonic acid pathway components, by the MSCs, which in turn increases their PGE-2 secretion [19]. This PGE-2 negative feedback loop, for which intimate association or cell-cell contact may be required, stimulates the transition of resident macrophages from their classical pro-inflammatory phenotype towards an anti-inflammatory phenotype, in which the cells secrete anti-inflammatory mediators such as IL-10 and IL-1 receptor agonist [19]. PGE-2 has also been found to suppress dendritic cell maturation, T-lymphocyte activation, T-lymphocyte proliferation, and T-lymphocyte cytokine secretion [52,53]. Additionally, blocking the synthesis or activity of PGE-2, using either indomethacin or PGE-2 blocking antibody, has also been found to prevent the immunosuppressive effects of MSCs and allow lymphocyte proliferation to be restored [54,55].
The concentration at which PGE-2 is secreted by hMSCs may be critical in the magnitude of immunosuppressive response. Solchaga and Zale [55] discovered that PGE-2 accurately correlated with the immunosuppressive capacity of hMSCs, and thus that increased PGE-2 expression resulted in a greater immunosuppressive effect. Therefore, any differences in the concentration of PGE-2 secreted by baseline, centrifugation, or non-centrifugation hMSCs in response to pro-inflammatory cytokines in the present study may be significant and may represent a substantial change in the immunosuppressive function of these cells.
In previous experiments, Najar et al. [53] reported that, hMSCs derived from bone marrow, Wharton's jelly, and adipose tissue constitutively produced PGE-2. The results of the present study (Fig. 12) suggest that although baseline hMSCs may constitutively secrete PGE-2, the concentration at which the PGE-2 was produced was very low in comparison to inflammatory pre-treated cells. The data also indicates that the constitutive expression of PGE-2 in untreated cells may decrease with culture over two passages, using either manual or automated process steps over two passages, as the concentration of PGE-2 in the conditioned medium of untreated centrifugation (p = 0.625) and non-centrifugation (p = 0.643) hMSCs may have been lower than in that of the untreated baseline hMSCs, although this was not a significant trend. Previous studies have identified that the stimulation of hMSCs with certain factors increases their PGE-2 secretion, for example, the pro-inflammatory cytokine TNF-␣ has been found to be correlated with a significant increase in the expression of PGE-2 by hMSCs [55]. In the present study, the results of the PGE-2 ELISA illustrate that each of the pre-treated hMSC samples exhibited a significantly increased concentration of PGE-2 compared to the untreated hMSC samples, demonstrating that, similarly to IDO, an inflammatory microenvironment is required to stimulate significant PGE-2 expression. The increased secretion of PGE-2 by all pre-treated hMSC groups therefore indicates that, regardless of the cell culture process steps utilised, all of the cells demonstrated an immunosuppressive response to an inflammatory microenvironment. Previous experiments have reported a similar magnitude of PGE-2 secretion in mice and human MSCs after alternative pretreatment and co-culture [56,57].
Furthermore, it is apparent that the inflammatory pretreatment of the non-centrifugation hMSCs induced a significantly greater secretion of the immunomodulatory factor PGE-2 compared to that of baseline (p = 0.00) and centrifugation (p = 0.00) hMSCs. Also, no significant difference between the PGE-2 expression of pre-treated baseline and centrifugation hMSCs was observed (p = 0.885). These findings contradict the non-significant trends observed in the IDO assay, and suggest that the culture of hMSCs over two passages, utilising either centrifugation or non-centrifugation process steps, had no effect upon the PGE-2 secretion, and therefore the immunosuppressive potential, of hMSCs. This data may also indicate that the utilisation of an automated process step may promote an increased production of PGE-2 compared to baseline or centrifugation hMSCs.
Therefore, it can be concluded that pre-treatment with proinflammatory cytokines was required to induce significant PGE-2 secretion in all hMSC groups, although culture over two passages may decrease constitutive PGE-2 expression. It was also observed that, contrary to the non-significant trends observed in the IDO assay, culture over two passages using either the manual or the automated culture process steps did not decrease the hMSC immunosuppressive response to pro-inflammatory cytokines, and that the utilisation of the non-centrifugation process steps may in fact increase the secretion of PGE-2 in response to an inflammatory microenvironment. It could be hypothesised that this increased PGE-2 concentration in the conditioned medium of pre-treated non-centrifugation hMSCs was due to an increase in the immunosuppressive capacity of these cells, or perhaps due to a more pro-inflammatory environment in the culture when utilising the non-centrifugation process steps, resulting in a greater antiinflammatory response. However, further investigation would be required to investigate the effects of the non-centrifugation process steps upon the PGE-2 secretion of inflammatory pre-treated hMSCs. It must also be acknowledged that, as with the IDO assay, the heterogeneity of hMSCs may have influenced the results of this PGE-2 ELISA.

VEGF quantification
It has previously been demonstrated, both in vitro and in vivo, that hMSCs have the capacity to stimulate new blood vessel formation, and it has therefore become widely recognised that these cells have a pro-angiogenic function [58,59,4,17,60]. As discussed previously, vascular endothelial growth factor (VEGF) has been strongly linked to the angiogenic potential of hMSCs [61][62][63][64]. Additionally, it has been proposed that this growth factor plays a critical role in hMSC pro-angiogenic function [21][22][23][24][25][26].
A significant body of research regarding the preconditioning of hMSCs in low oxygen concentrations has identified that, although hMSCs constitutively express VEGF in normal oxygen concentrations, exposing these cells to a low partial pressure of oxygen, or hypoxia, stimulates the increased expression of VEGF [20][21][22][23][24][25]65]. Therefore, it has been concluded that by pre-treating hMSCs in hypoxic conditions the expression of VEGF by these cells will be increased, which will in turn improve their pro-angiogenic function. A number of investigators have demonstrated the increased angiogenic potential of hypoxia preconditioned hMSCs and have found that these pre-treated cells increase endothelial cell growth in vitro [21,24] and improve perfusion in rodent hind limb ischaemia models [21,22].
The results of the VEGF ELISA performed in the present study concur with the literature and demonstrated that all untreated hMSC populations constitutively expressed VEGF. However, it was also observed that both untreated centrifugation (p = 0.06) and non-centrifugation (p = 0.00) hMSCs secreted significantly lower concentrations of VEGF compared to untreated baseline hMSCs. This may suggest that, regardless of the process step utilised, the culture of these cells over two passages decreased their VEGF expression under normoxic conditions (20% O 2 ). However, it has previously been demonstrated that both rat and human MSCs maintain, or even increase, their VEGF secretion over multiple passages [26,66]. Additionally, this same pattern was not observed when the hMSCs were cultured under hypoxic conditions. Fig. 13 also illustrates that hypoxic pre-conditioning, at 1% O 2 , did not significantly increase hMSC VEGF production. No significant difference in VEGF concentration of hMSC conditioned medium samples was observed between untreated centrifugation and untreated non-centrifugation hMSCs compared to all hypoxia pre-treated hMSCs. Furthermore, it is apparent that untreated baseline hMSCs secreted significantly greater concentrations of VEGF compared to untreated centrifugation (p = 0.06), untreated noncentrifugation (p = 0.02), hypoxia pre-treated baseline (p = 0.00), hypoxia pre-treated centrifugation (p = 0.039), and hypoxia pretreated non-centrifugation (p = 0.00) hMSCs. Although substantial evidence supporting increased VEGF expression by hypoxic preconditioned hMSCs exists, a failure to demonstrate this pattern has been previously reported in which hMSCs were cultured at 1% O 2 as a monolayer [67].
Additionally, these results illustrate that hypoxia pre-treated centrifugation hMSCs may have produced a greater angiogenic response to hypoxic preconditioning compared to hypoxia pretreated baseline (p = 0.09) and non-centrifugation hMSCs (p = 0.05) by secreting a greater concentrations of VEGF into the culture medium. From this finding, it could be proposed that the manual centrifugation process step has the capacity to increase the expression of VEGF by hMSCs. However, although hypoxic preconditioning may facilitate this increase cooperatively with the centrifugation process, as this finding was not observed in the untreated centrifugation hMSC samples, this hypothesis cannot be supported and therefore further investigation is required.
As discussed previously, with regards the results of the IDO assay and PGE-2 ELISA in the present study, it must be acknowledged that, despite the observation of differences in VEGF expression between hMSCs cultured using each of the process steps, the heterogeneous nature of hMSC populations may significantly influence these results. Variability in hMSC characteristics can often be observed due to donor variability or inconsistency in the manufacturing process [68], leading to heterogeneity between populations.

Conclusions
From the present study, it can be concluded that no significant difference in hMSC morphology, surface marker expression, STR profile, viable cell yield, and viability was observed between hMSCs cultured using either the manual centrifugation step or the automated non-centrifugation process step, in which residual dissociation agent is carried over. Surface marker profile was concordant for the majority of hMSC markers between flow cytometry analyses performed at two independent laboratories; however, differences in culture and analysis methods between laboratories may have contributed to the variability observed in two of the surface markers. Additionally, all hMSC samples responded to proinflammatory pre-treatment (TNF-␣ & IFN-␥) by expressing PGE-2 and IDO, although none of the hMSC samples significantly upregulated VEGF expression in response to hypoxic pre-treatment (1% O 2 ). Although differences in PGE-2, VEGF and IDO expression, in response to hypoxic and pro-inflammatory pre-treatment, were observed between hMSCs cultured utilising manual and automated process steps, the innate heterogeneity of hMSCs may contribute to these findings. Finally, the results of the present study also suggest that the utilisation of the automated non-centrifugation process step may result in greater variability in viable hMSC yield.

Conflicts of interest & role of the funding sources
This research, in which TAP Biosystems' CompacT SelecT automated cell culture platform was utilised, was funded, in part, by TAP Biosystems as well as Loughborough University and the Engineering and Physical Sciences Research Council (EPSRC). Neither Loughborough University, as a funding body, nor the EPSRC were involved in the preparation of this publication. However, Dave Thomas, Product Manager at TAP Biosystems, advised on study design and data interpretation, in addition to providing research supervision for this work by acting as an industrial supervisor. The terms of this arrangement have been reviewed and approved by Loughborough University.
acknowledge the contribution of Thomas Heathman for providing assistance with the hMSC Flow Cytometry analysis performed at Loughborough University.