The effect of medium supplementation and serial passaging on the transcriptome of human adipose-derived stromal cells expanded in vitro

Background For adipose-derived stromal cells (ASCs) to be safe for use in the clinical setting, they need to be prepared using good manufacturing practices (GMPs). Fetal bovine serum (FBS), used to expand ASCs in vitro in some human clinical trials, runs the risk of xenoimmunization and zoonotic disease transmission. To ensure that GMP standards are maintained, pooled human platelet lysate (pHPL) has been used as an alternative to FBS. ASCs proliferate more rapidly in pHPL than in FBS, with no significant change in immunophenotype and differentiation capacity. However, not much is known about how pHPL affects the transcriptome of these cells. Methods This study investigated the effect of pHPL and FBS on the ASC transcriptome during in vitro serial expansion from passage 0 to passage 5 (P0 to P5). RNA was isolated from ASCs at each passage and hybridized to Affymetrix HuGene 2.0 ST arrays for gene expression analysis. Results We observed that the transcriptome of ASCs expanded in pHPL (pHPL-ASCs) and FBS (FBS-ASCs) had the greatest change in gene expression at P2. Gene ontology revealed that genes upregulated in pHPL-ASCs were enriched for cell cycle, migration, motility, and cell-cell interaction processes, while those in FBS-ASCs were enriched for immune response processes. ASC transcriptomes were most homogenous from P2 to P5 in FBS and from P3 to P5 in pHPL. FBS- and pHPL-gene-specific signatures were observed, which could be used as markers to identify cells previously grown in either FBS or pHPL for downstream clinical/research applications. The number of genes constituting the FBS-specific effect was 3 times greater than for pHPL, suggesting that pHPL may be a milder supplement for cell expansion. A set of genes were expressed in ASCs at all passages and in both media. This suggests that a unique ASC in vitro transcriptomic profile exists that is independent of the passage number or medium used. Conclusions GO classification revealed that pHPL-ASCs are more involved in cell cycle processes and cellular proliferation when compared to FBS-ASCs, which are involved in more specialized or differentiation processes like cardiovascular and vascular development. This makes pHPL a potential superior supplement for expanding ASCs as they retain their proliferative capacity, remain untransformed and pHPL does not affect the genes involved in differentiation in specific developmental processes. Electronic supplementary material The online version of this article (10.1186/s13287-019-1370-2) contains supplementary material, which is available to authorized users.


Background
Adipose-derived stromal cells (ASCs) could constitute a novel therapeutic option for the treatment of several diseases and are increasingly being assessed in clinical trials for this purpose [1][2][3]. Most clinical trials make use of ASCs that have been expanded ex vivo via several rounds of passaging in order to obtain adequate cell numbers [4,5]. In the laboratory, ASCs are traditionally expanded in medium supplemented with fetal bovine serum (FBS); however, it has been reported that ASCs expanded in FBS cause immune reactions when given to human patients [2,[6][7][8]. However, for these cells to be considered safe for patient use, they need to adhere to good manufacturing processes (GMPs), in which non-defined and animal-related products are eliminated [2,9]. As a result, several investigators have moved away from using FBS and have instead investigated the use of human alternatives such as pooled human platelet lysate (pHPL) [10][11][12]. Most studies compare the criteria as set out by the Mesenchymal and Tissue Stem Cell Committee of the International Society for Cellular Therapy (ISCT) and International Federation of Adipose Therapeutics and Sciences (IFATS) when comparing FBS to pHPL [6,10,[13][14][15]. These criteria include ASC adherence to plastic, immunophenotypic surface marker expression and the ability to differentiate into bone, fat, and cartilage [5,13]. The use of pHPL as a medium supplement has advantages over FBS. It has thus been reported that when the cells are expanded in pHPL, their innate characteristics are unaltered and proliferation is increased during expansion [10,12,16]. However, it is well known that experimental conditions, such as medium supplementation, can have an effect on gene expression [15,[17][18][19]. It is therefore important to demonstrate that the cells are safe for use in patients by measuring the effect of the medium supplementation at the level of gene expression. In this study, we assessed the changes in ASC gene expression that occur during serial passaging by comparing cells expanded in FBS versus pHPL.

ASC characterization
ASCs were characterized by surface marker expression (immunophenotype) and the ability to differentiate into adipocytes. Immunophenotype was assessed on SVF and at each passage (P0 to P5) using methods previously described [22]. ASCs were induced to differentiate into adipocytes at P5, and adipogenesis was measured using methods previously described [17,22]. Data and experimental design (Additional file 1: Figure S1) can be found in Additional file 1.

RNA isolation and quality
ASCs were expanded in FBS or pHPL and RNA was isolated at each passage. At confluence, the cells were dissociated using trypLE and counted. Thereafter, 1 × 10 6 cells were centrifuged (300g) and the resultant pellet was washed using phosphate buffered saline (PBS). RNA was isolated using the RNeasy Minikit (Qiagen, Hilden, Germany) according to the manufacturer's instructions, and quantified on a NanoDrop® ND 1000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). RNA purity was assessed at an absorbance optical density (OD) ratio of 260/280 and 260/230. RNA integrity and quality were assessed using a TapeStation® 2200 (Agilent Technologies; Santa Clara, CA, USA) together with RNA ScreenTape® and Sample Buffer kit (Agilent Technologies, Santa Clara, CA, USA) according to the manufacturer's instructions. Sample read-out was compared to a TapeStation® RNA ladder. RNA that had absorbance OD ratios greater than 2 and RIN values greater than 8 was used for downstream applications.

Microarray gene expression analysis
Total RNA (100 ng) isolated from ASCs expanded in FBS or pHPL from P0 to P5 was used for first-and second-strand cDNA syntheses, followed by the synthesis and amplification of complementary RNA (cRNA) by in vitro transcription using an Affymetrix GeneChip® WT PLUS Reagent Kit according to the manufacturer's protocol. Amplified cRNA was purified using magnetic purification beads. Thereafter, 15 μg of purified cRNA was used to synthesize second cycle single-stranded cDNA (ss-cDNA) and subsequently followed by another purification step. Purified ss-cDNA (5.5 μg) was fragmented, labeled, and used to prepare a hybridization cocktail. Hybridization was performed using the Affymetrix Gene-Chip® Hybridization Wash and Stain Kit according to the manufacturer's protocol. The hybridization cocktail was hybridized to Affymetrix GeneChip® Human Gene 2.0 ST arrays. Arrays were placed in an Affymetrix GeneChip® Hybridization Oven-645 rotating at 60 rpm at 45°C for 17 h, after which they were washed and stained in an Affymetrix GeneChip® Fluidics Station-450Dx before being scanned in an Affymetrix GeneChip® Scanner-7G. The output Affymetrix CEL files, which have intensity values for all probes present on the scanned arrays, were used for further analysis. The Robust Multiarray Analysis algorithm [23] in the Affymetrix Expression Console™ was used to perform background correction, summarization, normalization, and the calculation of probe set expression values. Finally, the Affymetrix Transcription Analysis Console™ was used to calculate the fold change of each probe set or transcript cluster identifier number and mapped to the corresponding gene. Only differentially expressed genes (DEGs) that had a fold-change ≥ 2 or ≤ − 2, a p value > 0.05, and an FDR > 0.5 were used for downstream analysis. The fold-change of each gene represents the change in gene expression seen between two samples or conditions being compared and is based on the signal measured.

Functional analysis
The DEGs for the different samples were used for functional analysis to determine significantly enriched pathways and processes using the g:GOSt functional enrichment analysis tool on the g:Profiler web server [24].
Functional analysis of the DEGs by gene ontology (GO) classification revealed that genes that were significantly upregulated at the different passages were enriched for certain biological processes (BP), cellular components (CC) and molecular functions (MF). Only the top 5 significant GO terms will be discussed here. From P0 to P5, pHPL-ASCs were enriched for GO terms such as developmental processes, cell cycle processes, cellular proliferation, and extracellular matrix and structure organization. FBS-ASCs were enriched for GO terms such as cell proliferation, adhesion, extracellular matrix and structure organization, cardiovascular and vascular development, structure morphogenesis, and other developmental processes (Table 1; Additional file 3).
GO classification of upregulated genes in FBS-ASCs revealed they were significantly enriched for cell migration and motility from P0 to P1, while those for P1 to P2 and P2 to P3 were mostly enriched for immunological responses and processes. Genes that were upregulated from P3 to P4 and P4 to P5 were not enriched for any GO terms (Table 2; Additional file 6). Genes that were downregulated from P0 to P1 and P1 to P2 were enriched for system and developmental processes, while those from P2 to P3 were enriched for immune subunit and protein assembly. In contrast, downregulated genes from P3 to P4 and P4 to P5 were not enriched for any GO terms.
For pHPL-ASCs, GO terms significantly enriched for in upregulated genes were immune responses from P0 to P1, regulation of developmental processes and stimulus responses from P1 to P2, RNA binding regulation and transcription factor activity from P2 to P3 and regulation of cardiovascular processes from P3 and P4. Genes that were upregulated from P4 to P5 were not enriched for any GO term (Table 3; Additional file 7). Genes that were downregulated from P1 to P2 were significantly enriched for cell cycle processes, from P2 to P3 for cardiovascular processes, while downregulated genes from P0 to P1, P3 to P4, and P4 to P5 were not enriched for any GO term.
We next undertook to evaluate the extent to which the ASC transcriptome at each passage (P1 through to P5) differs from its original state (SVF) at P0 when expanded in either FBS or pHPL, and to functionally characterize such changes using GO classification. This was done by comparing gene expression at each passage (P1 to P5) to that of the "original" seeded ASCs (SVF) at P0. For FBS-ASCs, 292, 514, 591, 685, and 737 genes were significantly upregulated while 273, 288, 350, 427, and 426 genes were significantly downregulated from P1 to P5 ( Fig. 3a and Additional file 8). For pHPL-ASCs, 297, 861, 848, 891, and 863 genes were significantly upregulated while 46, 700, 262, 427, and 523 genes were significantly downregulated from passage P1 to P5 ( Fig. 3b and Additional file 9).
GO terms significantly enriched for in upregulated genes at each passage (P1 to P5) when compared to P0 in FBS-ASCs (Table 4; Additional file 10) or pHPL-ASCs (Table 5; Additional file 11) were specific to immune responses and processes. GO terms specific to developmental processes were enriched for in the downregulated genes in FBS-ASCs at each passage (P1 to P5) when compared to P0 (Table 4; Additional file 10). For pHPL-ASCs, downregulated genes at P1 were not enriched for any GO term, while those of all the subsequent passages (P2 to P5) were enriched for cell cycle processes and developmental processes.
We observed during serial passaging that the ASC transcriptomic profile stabilizes (minimal change in DEGs between adjacent passage numbers) from P2 for FBS (Fig. 2a) and P3 for pHPL (Fig. 2b). This could mean that ASC cultures are more homogenous from P2 to P5 and from P3 to P5 when expanded in FBS and pHPL respectively.
From the list of DEGs obtained at each passage (P1 to P5) when compared to P0 for both the FBS-and pHPL-ASCs (Additional files 8 and 9), we observed that ASCs showed gene expression signatures that were unique at each passage (P1 to P5) which was independent of the medium supplementation (FBS or pHPL) used during in vitro expansion (Additional file 12). This unique passage-specific gene expression profile constitutes the DEGs that were common to both pHPL and FBS at each passage number. Equally, if the passage-specific gene expression profile (DEGs common to both FBS-and pHPL-ASCs at each passage) is excluded at each passage number, the remaining DEGs represent unique FBS-ASC and pHPL-ASC passage-specific gene expression profiles (Additional file 12).
In total therefore, there were 118 DEGs that constituted the FBS-ASC-specific gene expression profile, which is almost 3 times more than the 43 DEGs of the pHPL-ASCspecific gene expression profile (Additional file 14). Functional analysis of the pHPL-ASC-specific gene expression signature by GO classification showed that neither upnor downregulated genes were enriched for any biological process, while the FBS-ASC-specific gene expression signature showed upregulated genes that were significantly enriched for cell migration and cell movement processes, while the downregulated genes were significantly enriched for the regulation of cell communication, signal transduction and cell proliferation processes.
Since the passage-specific gene expression profile consists of the common genes expressed by both FBS-and pHPL-ASCs at each passage, the genes that are common to all these passage-specific profiles will then constitute an ASC gene expression profile that is not affected by medium supplementation or cell passage number. There   SLC16A10, SPARCL1, SPP1, TM4SF18, TMEM176B,  TNFRSF1B, TREM1, TREM2, TYROBP, and VSIG4) and 5 downregulated genes (F2RL2, FGF5, GALNT5, RAB3B, and SLC9A7) that constitute this subset of genes that were consistently differentially expressed from P1 through to P5. This set of genes therefore represents a unique in vitro ASC transcriptome profile that was neither affected by medium supplementation nor cell passage number (Additional file 14). GO classification of these genes revealed that they are significantly enriched for normal cellular processes like response to stimulus and stress, defense, and inflammatory responses and vesicle-mediated transport.

Discussion
Adipose-derived stromal cells (ASCs) are being assessed for their safety and efficacy in numerous clinical trials [6,14,25]. Traditionally, these cells are expanded in medium containing FBS, which is known to have several disadvantages such as the transmission of zoonotic diseases and the stimulation of immune reactions in patients [26,27]. This has been circumvented by changing from animal products to either clinical-grade, GMPcompliant, or human alternative products [28]. One such change has been to supplement culture medium with either serum-free media or human blood components. The use of different medium supplements has been well documented and all show comparable immunophenotypic profiles and differentiation capacities while having marked differences in proliferation capacity [6]. The advantage of pHPL over these alternatives lies largely in the ability to pool platelets from multiple donors.
Furthermore, it has been shown that ASCs expanded in pHPL retain their immunophenotypic characteristics and their ability to differentiate into bone, cartilage and fat [2,6,16]. One of the biggest advantages of using pHPL for ASC expansion is the marked increase in proliferation, which in turn makes the time required for expansion to therapeutic numbers considerably shorter [12,22]. However, not much is known about the effect of pHPL has on the transcriptome, proteome, and secretome of these cells, which may impact on the outcome of clinical trials. This study has made use of microarray technology to examine the effect of pHPL on the ASC transcriptome during serial expansion in vitro, by comparing gene expression patterns in cells serially expanded in FBS or pHPL from P0 to P5. Overall, the transcriptome of ASCs expanded in pHPL or FBS was most different at P2, the point at which the maximum number of genes were differentially expressed (811 and 707, respectively; Fig. 1). Most genes that were upregulated in pHPL-ASC were significantly enriched for biological process such as cell cycle, cell division, and proliferation. This supports a previous study by Glovinski et al., in which changes in the expression of genes involved in cell proliferation and development were observed for ASCs expanded in pHPL [12]. This likewise confirms findings from other studies which have shown an increase in ASC proliferation in pHPL [16,29]. For ASCs expanded in FBS, our findings are consistent with the observation that numerous genes involved in extracellular matrix formation are upregulated [30,31].
It is well documented that ASCs are a heterogeneous population as revealed by differences in transcriptome, proteome, and secretome between subpopulations within the ASC mixture [32][33][34]. The initial subset of adherent cells seeded in culture (P0) is a heterogeneous population; after passaging and prolonged expansion, the population becomes more homogenous [35]. Work performed by several groups has shown that the heterogeneity of ASCs during the expansion process remains between subpopulations and between individual cells within the same subpopulation [32,36,37]. Furthermore, it has been   established that serial passaging affects ASC gene expression profiles [29]. Global gene expression profiles could therefore be used as a tool to study ASC heterogeneity at different passages. The more homogenous the cultures are at different passages, the fewer the number of DEGs will be between them.
We have investigated the effect of serial passaging on the ASC transcriptome by comparing FBS-ASC and pHPL-ASC cultures at each passage to those of the previous passage. We observed that the transcriptome was relatively stable from P2 to P5 for cells expanded in FBS and from P3 to P5 for cells expanded in pHPL as is evident from the relatively low number of DEGs obtained between these passages. This suggests the ASC cultures become homogenous at the transcriptome level earlier in FBS (P2) than in pHPL (P3). Interestingly, the genes upregulated significantly in FBS-ASCs were enriched for biological processes involved in immune and inflammatory responses. These findings are similar to those reported by Kim et al., where genes involved in Table 3 Top 5 enriched GO terms for significantly up-and downregulated DEGs for pHPL-ASCs between subsequent passages. Related to Fig. 2b   inflammatory and immune responses, and cell migration and homing [19], were upregulated in ASCs expanded in FBS. They further postulated that the upregulation of these genes was due to the high cell density at the time of cell harvesting and could be the reason why FBS-expanded ASCs might be effective in treating graft-vs-host disease and damaged tissues. On the other hand, human clinical trials that have made use of ASCs expanded in FBS have reported adverse immune responses in patients after administration [2,[6][7][8]38]. This could be due to the upregulation of these inflammatory and immune response genes. Genes that were downregulated in ASCs expanded in FBS at early passages (P0 to P1) were enriched for biological processes involving tissue development. Other studies have reported similar findings [30] which may explain why differentiation into adipocytes is reduced at later passages [18,39]. Surprisingly, genes that were upregulated in pHPL-ASCs at earlier passages were also enriched for immune and inflammatory response processes. This could be due to the presence of immune cells in the early passages and may not be related to the serum used. To further explore the possible presence of immune cells in early passages (P0), we compared each passage (P1 to P5) to P0. It was observed that upregulated genes were significantly enriched for immune and inflammatory responses irrespective of the supplementation used, while the downregulated genes were enriched for tissue developmental and cell cycle and division processes. To assess serum-specific transcriptional changes (where genes were differentially expressed based on the serum supplementation used), we normalized gene expression at all other passages to the passage at which the transcriptome stabilizes (P2 for FBS-ASCs and P3 for pHPL-ASCs). For FBS-ASCs, the upregulated genes were enriched for immune and inflammatory responses; this supports the findings obtained when we compared each passage to the previous passage and each passage to P0. This may suggest that FBS-ASCs express genes that are involved in immune reactions; however, the functional implications of this in clinical or in vivo settings will need to be explored further. Genes that were downregulated in FBS-ASCs were enriched for structure, organ, and tissue developmental processes suggesting that ASCs have greater differentiation potential at earlier passages such as P2. For ASCs expanded in pHPL, upregulated genes were enriched for DNA and RNA regulation processes, BMP pathway signaling, and cell cycle and cell division processes. These findings suggest that proliferation may not decrease with increased passaging as indicated by Shahdadfar et al. [15] and could provide therapeutic numbers more readily than other human alternatives and FBS. ASCs showed passage and serum-specific gene expression profiles. The passage-specific gene expression profile which is comprised of the DEGs that are common to both pHPL and FBS at each passage might reflect the in vitro serial passaging effect on the ASC transcriptome. The serum-specific gene expression signature at each passage (P1 to P5) may be reflective of the FBS or pHPL effect on the ASC transcriptome at that time period in culture (passage number) during the serial expansion process.
There were 118 and 43 genes that were differentially expressed in ASCs throughout the serial expansion process in FBS and pHPL respectively. This might indicate an ASC transcriptome profile that is specific to the medium supplementation (FBS or pHPL) used during cell expansion, irrespective of passage number. Thus, a serum-specific signature could potentially be used to identify the medium supplement (FBS or pHPL) in which the cells were previously expanded. This in turn could inform decision making in terms of the downstream clinical/research applications of these cells. There were fewer DEGs obtained for the pHPL-ASC-specific gene expression signature (43 genes), which is 1/3 the number of DEGS observed in FBS-ASCs (118 genes). The pHPL-ASC-specific gene expression signature was not enriched for any biological processes unlike the FBS-ASC-specific expression signature. This could mean that pHPL has no significant effect on the ASC transcriptome during in vitro serial passaging and suggests that pHPL might be a better medium supplement than FBS for in vitro cell expansion. Furthermore, downregulated genes in the FBS-ASC-specific gene expression signature were enriched for cell proliferation processes. This supports the observation that ASCs grow slower in FBS when compared to cell-expanded pHPL.
Finally, we observed that ASCs have a unique in vitro transcriptome profile, which is independent of cell passage number and/or medium supplementation. This consists of a set of genes that are always expressed by ASCs in vitro at any given time in culture during the   [40]. PECAM-1 has been reported to be expressed by ASCs especially during early passages [41,42]. ITGAM is another gene shown to be expressed by ASCs at low levels up to P3, and exhibits greater than 70% isoform switching between experimental conditions [43]. CD53 and TREM1 have been reported recently as novel marker genes expressed by adipogenic progenitor preadipocyte cells and BCL6B by osteochondrogenic progenitor preadipocyte cells from mouse bone marrow [44]. Furthermore, a novel subpopulation of human adipose tissue-resident macrophages (ATMs) located in the interstitial spaces between adipocytes has been shown to express CD14, which upon culturing to P3 is lost, at which point the cells display an expression profile which is similar to ASCs [45]. Therefore, the expression of CD14 by ASCs in this study suggests the presence of a heterogeneous population of ASCs that contains this novel subpopulation of ATMs which persisted beyond P3 in culture. The entire process of obtaining a product for clinical purposes should adhere to the GMP guidelines. The use pHPL for the expansion of ASCs in vitro is one of many steps required. In this study, we made use of defined, clinical-grade reagents and the expansion of the ASCs was performed under sterile conditions. Isolation and expansion of ASCs in a closed system to further reduce the risk of contamination would provide a robust clinical GMP-complaint process.

Conclusion
This study highlights differences in the transcriptome of ASCs expanded in pHPL versus FBS, which could be used to guide their application in the clinical setting. ASCs expanded in FBS were enriched for immune and inflammatory responses, whereas ASCs expanded in pHPL were enriched for cell cycle, proliferation, and cell division. Our findings suggest that the differentiation capacity of ASCs is likely to be greater at earlier passages and that ASCs expanded in pHPL are likely to retain their proliferative capacity during prolonged expansion. These findings also suggest pHPL may be a superior supplement for expanding ASCs to therapeutic numbers without influencing the expression of genes involved in differentiation of specific developmental processes. Furthermore, we found that even though ASCs expanded in pHPL had a greater proliferation capacity, they were not enriched for genes specific to transformation. While these findings provide novel insights into potential markers for ASCs, some of the individual genes and groups of genes mentioned in this study need to be further investigated. Finally, to further compliment these findings, we believe that the proteome and the secretome of ASCs expanded in pHPL or FBS should also be studied.