Abstract

   Human bone marrow-derived stromal (skeletal) stem cells (BM-hMSC) are
   being employed in an increasing number of clinical trials for tissue
   regeneration. A limiting factor for their clinical use is the inability
   to obtain sufficient cell numbers. Human embryonic stem cells (hESC)
   can provide an unlimited source of clinical grade cells for therapy. We
   have generated MSC-like cells from hESC (called here hESC-stromal) that
   exhibit surface markers and differentiate to osteoblasts and
   adipocytes, similar to BM-hMSC. In the present study, we used
   microarray analysis to compare the molecular phenotype of hESC-stromal
   and immortalised BM-hMSC cells (hMSC-TERT). Of the 7379 genes expressed
   above baseline, only 9.3% of genes were differentially expressed
   between undifferentiated hESC-stromal and BM-hMSC. Following ex vivo
   osteoblast induction, 665 and 695 genes exhibited ≥ 2-fold change (FC)
   in hESC-stromal and BM-hMSC, respectively with 172 genes common to both
   cell types. Functional annotation of significantly changing genes
   revealed similarities in gene ontology between the two cell types.
   Interestingly, genes in categories of cell adhesion/motility and
   epithelial–mesenchymal transition (EMT) were highly enriched in
   hESC-stromal whereas genes associated with cell cycle processes were
   enriched in hMSC-TERT. This data suggests that while hESC-stromal cells
   exhibit a similar molecular phenotype to hMSC-TERT, differences exist
   that can be explained by ontological differences between these two cell
   types. hESC-stromal cells can thus be considered as a possible
   alternative candidate cells for hMSC, to be employed in regenerative
   medicine protocols.

   Keywords: Pluripotent stem cells, Bone marrow stromal cells,
   Osteoblastic differentiation, Illumina bead microarray, Bioinformatic
   analysis

Highlights

     * •
       hESC-derived MSC-like cells were compared to immortalised BM-MSC.
     * •
       Comparison was performed using microarrays on non-induced and OB
       induced cells.
     * •
       Analysis demonstrated close hierarchical relationships and
       molecular phenotypes.
     * •
       90.7% of genes were similarly expressed in non-induced cells.
     * •
       73% of OB induced genes for both cell lines correlated with GO
       ontology analysis.

1. Introduction

   Human bone marrow stromal stem cells (BM-hMSC) are adult, multipotent
   stem cells present within the non-haematopoietic compartment of the
   bone marrow. BM-hMSC are capable of differentiating into mesoderm-type
   cells e.g. osteoblasts, adipocytes ([37]Abdallah and Kassem, 2012), and
   chondrocytes ([38]Puetzer et al., 2010, review) under appropriate ex
   vivo conditions, and can form heterotopic bone and bone marrow organ in
   vivo ([39]Abdallah et al., 2005). hMSC are being introduced for
   regenerative therapy in a number of clinical conditions including
   repair of bone ([40]Quarto et al., 2001) and cartilage defects
   ([41]Wakitani et al., 2002). In addition, the BM-hMSC secrete a large
   number of factors that exert anti-inflammatory and immune modulatory
   effects and thus BM-hMSC have been used in phase I/II clinical trials
   of tissue regeneration following cerebral haemorrhage ([42]Heile et
   al., 2009) and in graft-versus-host disease (GvHD) ([43]Le Blanc et
   al., 2010). One of the limiting factors for wider use of hMSC in
   therapy is the inability to obtain a sufficient number of cells needed
   for clinical applications. This is due to the small number of hMSC
   recovered from bone marrow aspirates and their limited proliferative
   potential during ex vivo expansion ([44]Stenderup et al., 2003).

   Human embryonic stem cells (hESC) are pluripotent stem cells (PSC) that
   can be maintained ex vivo in a self-renewing state and therefore
   represent an unlimited source for generating differentiated cells for
   regenerative therapies ([45]Volarevic et al., 2011). Significant
   progress has been made in designing and validating efficient
   differentiation protocols of hESC into a number of cell lineages
   including osteoblastic cells ([46]Abdallah Basem et al., 2011,
   [47]Irion et al., 2008). Additionally, new technologies allowing the
   expansion of pluripotent stem cells in bioreactors, without enzymatic
   interventions, will allow a large number of PSC to be generated prior
   to differentiation and transplantation ([48]Chen et al., 2014). While
   the most interesting application of hESC-derived differentiated cells
   is as a universal off-the-shelf product for allogenic use in cell
   therapy, further research about the in vivo functions of hESC-derived
   differentiated cells need to be conducted both in animal models and in
   humans.

   However, concerns have been raised regarding the ability of the ex vivo
   differentiation of hESC to induce a differentiated phenotype similar to
   that observed in tissue-specific cells ([49]Cahan and Daley, 2013). The
   phenotype of hESC-derived differentiated cells have traditionally been
   described based on a limited number of in vitro expressed lineage
   specific markers ([50]D'Amour et al., 2005, [51]Hay et al., 2008,
   [52]Kroon et al., 2008). Side-by-side comparisons of hESC-derived
   differentiated cells and bona fide tissue resident differentiated cells
   have rarely been conducted ([53]Irion et al., 2008). Four published
   reports have previously compared mesodermal lineage differentiation of
   hESC with tissue specific/resident cells. Raynaud et al. investigated
   cardiomyocytes derived from hESC in comparison with bone marrow
   mesenchymal cells ([54]Raynaud et al., 2013); Barbet et al., Bigdeli et
   al. and Barberi et al. all based their comparison of mesenchymal-like
   (MSC-L) cells, derived from hESC, and bone marrow derived cells at
   baseline only ([55]Barbet et al., 2011, [56]Bigdeli et al., 2010,
   [57]Barberi et al., 2005). Interestingly, Barbet et al. demonstrated
   that MSC-like cells derived from hESC clustered closer with BM-hMSC as
   compared to MSC-like derivatives with hESC ([58]Barbet et al., 2011).
   Thus, data examining differences between MSC-like cells and BM-hMSC
   during osteoblastic induction are absent.

   The availability of large-scale methods for gene profiling, such as DNA
   microarrays, offers an opportunity to define the “molecular phenotype”
   of ex vivo cultured cells. Additionally, data mining using
   bioinformatic tools can provide a functional understanding of pathways
   and processes enriched in differentiated cell lineages and allow
   comparisons between different cell populations ([59]Klimanskaya et al.,
   2004, [60]Larsen et al., 2010, [61]Lu et al., 2007, [62]Marei et al.,
   2011).

   We have developed an ex vivo culture method to isolate, propagate and
   differentiate stromal cells obtained from hESC (termed here
   hESC-stromal) into osteoblastic and adipocytic cells ([63]Harkness et
   al., 2011). At a cellular level, hESC-stromal exhibited comparable
   cluster of differentiation (CD) surface markers, and differentiated to
   the osteoblastic and adipocytic lineages using the same differentiation
   protocol employed in hMSC-TERT ([64]Harkness et al., 2011). In the
   present study, we employed DNA microarrays to compare the molecular
   phenotype of hESC-stromal to a bona fide bone marrow-derived,
   immortalised, hMSC. In addition, an in depth analysis of the
   differentiated phenotype of hESC derived MSC-like cells has been
   obtained. Our results demonstrate that these two cell populations are
   closely related at both the cellular and molecular levels.

2. Methods

2.1. Cell culture

   hESC-stromal cells were derived as previously described ([65]Harkness
   et al., 2011). Briefly, stromal fibroblast-like cells were isolated
   from an in house Odense3 embryonic stem cell line ([66]Harkness et al.,
   2011), grown under feeder free conditions in the presence of mouse
   embryonic fibroblast (MEF) conditioned media, and cultured on
   hyaluronic acid-coated culture plates (100 μg/ml; HA; Calbiochem-Merck,
   Darmstadt, Germany) ([67]Harkness et al., 2011). These cells were
   termed hESC-stromal cells and were routinely cultured in high glucose
   DMEM (Invitrogen, Taastrup, Denmark) with 10% FBS (PAA, Pasching,
   Austria). As a model for BM-hMSC, we employed hMSC-TERT ([68]Simonsen
   et al., 2002) that exhibit a stable cellular and molecular phenotype
   comparable to that of primary BM-hMSC ([69]Al-Nbaheen et al., 2013).
   hMSC-TERT were routinely cultured in MEM (Invitrogen) with 10% FBS
   (PAA, Pasching, Austria).

   Ex vivo osteoblast (OB) differentiation was performed using osteoblast
   induction medium containing β-Glycerophosphate (10 mM;
   Calbiochem-Merck), l-ascorbic acid-2-phosphate (50 μg/ml;
   Sigma-Aldrich, Brøndby, Denmark), dexamethasone (100 nM hESC-stromal,
   10 nM hMSC-TERT; Sigma-Aldrich) and calcitriol (1,25 hydroxy-Vitamin
   D[3;] 10nM). For samples collected for microarray analysis media were
   changed every 2 days until day 6 for hESC-stromal (n = 3 independent
   experiments, at passages 6 and 7) or day 7 for hMSC-TERT (n = 1
   experiment, passage 48). Data for validation of array gene expression
   was generated from 6 independent experiments and plotted using the SEM
   ([70]Fig. 4); data for initial comparison of hESC-stromal and hMSC-TERT
   undifferentiated and undergoing OB induction were calculated from 3
   independent experiments ([71]Fig. 1). The OB differentiated cells are
   termed: hESC-stromal-OB and hMSC-TERT-OB.

Fig. 4.

   [72]Fig. 4
   [73]Open in a new tab

   Validation of microarray data by RT-PCR (n = 6 independent
   experiments): (A) genes common to hESC-stromal-OB (array—grey bar;
   PCR—black bar) and hMSC-TERT-OB (array—grey bar; PCR—white bar) up or
   down regulated over 2 FC (p ≤ 0.01); (B) OB genes unique to
   hESC-stromal-OB; (C) OB genes unique to hMSC-TERT-OB. Data is presented
   as fold change.

Fig. 1.

   [74]Fig. 1
   [75]Open in a new tab

   Comparison of undifferentiated and in vitro osteoblast (OB)
   differentiated hESC-stromal and hMSC-TERT. (A) Gene expression of
   osteoblast lineage markers in hESC-stromal-OB ■ and hMSC-TERT-OB Inline
   graphic (n = 3 independent experiments) and ALP activity normalised for
   cell number in hESC-stromal-OB and hMSC-TERT-OB (n = 3 independent
   experiments; *p < 0.05); (B) Hierarchical clustering of cell samples
   following Illumina bead array.

2.2. Alkaline phosphatase (ALP) activity measurements

   ALP activity was quantified as previously described ([76]Qiu et al.,
   2010) using a 1 mg/ml solution of P-nitrophenylphosphate
   (Sigma-Aldrich, Brøndby, Denmark) in 50 mM NaHCO[3] with 1 mM MgCL[2],
   pH 9.6, at 37 °C for 20 min. Activity was stopped using 3 M NaOH,
   absorbance of each reaction (λ[max] = 405 nm) was measured using a
   FLUOstar Omega plate reader (BMG Laboratories, Ramcon A/S, Birkerod,
   Denmark) and ALP activity was normalised to cell number (n = 3
   independent experiments, 6 wells/experiment). Cell number was
   determined using a CellTiter-Blue Cell Viability assay, according to
   the manufacturer's instructions (Promega, Nacka, Sweden). A two-tailed,
   paired T-Test statistical analysis was used to establish if significant
   differences could be determined between the normalised ALP activity of
   hESC-stromal-OB and hMSC-TERT-OB at 6 days of induction.

2.3. RNA isolation and real-time quantitative RT-PCR

   Total RNA was isolated using TRIzol (Invitrogen) as previously reported
   ([77]Harkness et al., 2011) and cDNA was generated using a revertAid H
   minus first strand cDNA synthesis kit (Fermentas, St. Leon-Rot,
   Germany) according to the manufacturer's instructions. RT-PCR was
   performed on an ABI Step one PCR machine (Applied Biosystems-Life
   Technologies, Carlsbad, CA, USA). Data were normalised to the geometric
   mean of the reference genes (β-Actin, B2M, HPRT, UBC1) and analysed
   using a comparative Ct method where Δ-CT is the difference between the
   CT values of the target and the geometric mean of the reference genes.
   Primer sequences ([78]Supplementary Table 1) were designed using
   Primer-BLAST.

2.4. Illumina® bead chip microarray

   Following RNA extraction, samples were purified using a GenElute
   mammalian total RNA miniprep kit (Sigma-Aldrich; according to the
   manufacturer's instructions). Five hundred nanograms of total RNA per
   sample were used for biotin-labelled cRNA production using a linear
   amplification kit (Ambion). Hybridizations, washing, Cy3-streptavidin
   staining, and scanning were performed on the Illumina BeadStation 500
   platform (Illumina®) according to the manufacturer's instruction. cRNA
   samples were hybridised onto Illumina® human-8 BeadChips version 3. All
   basic gene expression data analyses were carried out using the
   BeadStudio® software 3.0. Raw data were background-subtracted,
   normalised using the “rank invariant” algorithm and then filtered for
   significant expression on the basis of negative control beads. To
   demonstrate the relationship between the different samples,
   hierarchical clustering was performed using the Illumina BeadStudio®
   software. Correlation was used as the distance measure, and clustering
   was performed with the average linkage method.

2.5. Functional gene enrichment analysis

   Data analysis for functional enrichment was performed using the
   Database for Annotation, Visualization and Integrated Discovery (DAVID)
   ([79]Huang et al., 2008, [80]Huang et al., 2009). Genes were selected
   for comparison if they had a detection threshold p value < 0.01. Under
   OB induction, genes that had a fold change (FC) of ≤− 2 or ≥ 2 compared
   to the corresponding control were employed in the analysis. Fold change
   from undifferentiated hESC-stromal over hMSC-TERT, hMSC-TERT over
   hESC-stromal and fold change of OB induced/non induced for each cell
   line were analysed using DAVID. Data from DAVID was plotted using the
   one tail Fisher exact probability value calculated by DAVID. MetaCore®
   v6.14 was used to perform pathway enrichment analysis on genes
   significantly expressed (p < 0.01). Gene Ontology (GO) ossification
   data were downloaded from the GO database
   ([81]http://www.geneontology.org/GO.downloads.ontology.shtml).
   Ossification genes, relative to undifferentiated cells, differentially
   regulated during OB differentiation were identified (> 2 FC or <− 2 FC)
   in either hMSC-TERT-OB (n = 17), hESC-stromal-OB (n = 15) or in both
   hESC-stromal-OB and hMSC-TERT-OB (n = 11).

   Studies published here did not require ethical approval further to that
   published in the original derivation articles ([82]Simonsen et al.,
   2002, [83]Harkness et al., 2010).

3. Results

3.1. Cellular phenotype

   Cultured hESC-stromal and hMSC-TERT exhibited a fibroblast-like
   morphology at baseline and cuboidal morphology following ex vivo OB
   differentiation, respectively (data not shown). Both hESC-stromal and
   hMSC-TERT responded to OB induction by upregulation of OB marker genes:
   RUNX2, ALPL, COL1A1, BGLAP and SPP1 as well as ALP activity
   ([84]Fig. 1A). Both cell types formed heterotopic bone and bone marrow
   organ when implanted subcutaneously in immune deficient mice as
   previously reported ([85]Harkness et al., 2011).

3.2. Comparison of molecular phenotype of undifferentiated hESC-stromal vs.
hMSC-TERT cells at baseline

   Microarray analysis identified 7379 expressed genes (a gene was
   considered to be expressed if the p-value of detection threshold is
   ≤ 0.01). Gene lists, used for GO BP and MetaCore® analyses as well as
   comparison with GO database, were established by the following
   criteria: undifferentiated genes regulated ≥ 2 FC of
   hESC-stromal/hMSC-TERT with a detection p-value of ≤ 0.01; OB induced
   gene lists were established for each cell line of OB
   induced/undifferentiated ≥ 2 FC with a detection p-value of ≤ 0.01.
   Hierarchical clustering demonstrated a close relationship between
   undifferentiated hESC-stromal and hMSC-TERT ([86]Fig. 1B). The majority
   of genes demonstrated similar expression levels in both cell types with
   9.3% of total expressed genes differentially regulated (353 genes
   differentially up-regulated (FC ≥ 2) and 334 down-regulated (FC ≤ − 2))
   between the two cell lines. Functional enrichment analysis for gene
   ontology (GO) biological processes (BP) revealed, in hESC-stromal the
   highest enrichment scores in categories of cell adhesion, mesodermal
   tissue developmental and cell motion ([87]Fig. 2A). In comparison, GO
   BP categories for cell division, response to steroid hormone stimulus
   and positive regulation of apoptosis were highly enriched in hMSC-TERT
   ([88]Fig. 2B). An overview demonstrating the distribution of genes
   (non-induced and OB induced) is shown in the Venn diagrams in
   [89]Supplementary Fig. 1A–D.

Fig. 2.

   [90]Fig. 2
   [91]Open in a new tab

   GO functional enrichment of hMSC-TERT and hESC-stromal cells over 2 FC
   (detection threshold p ≤ 0.01). (A) GO biological process categories of
   undifferentiated hESC-stromal cells/hMSC-TERT show an increased
   annotation to developmental genes suggesting an increased capacity for
   multi-lineage differentiation as compared to hMSC-TERT; (B) in
   comparison undifferentiated hMSC-TERT/hESC-stromal demonstrate an
   increased GO BP annotation to cell cycle/mitosis categories; (C) GO
   functional enrichment of genes up and down regulated during osteogenic
   differentiation unique to hESC-stromal-OB (n = 493); (D) GO functional
   enrichment of up and down regulated genes unique to hMSC-TERT-OB
   (n = 523).

Supplementary Fig. 1.

   [92]Supplementary Fig. 1
   [93]Open in a new tab

   (A–D) Venn diagrams demonstrating the distribution of genes from
   non-induced and OB induced samples. (A) In non-induced samples 353
   genes were demonstrated to be up-regulated in hMSC-TERT/hESC-stromal
   and 334 genes in hESC-stromal/hMSC-TERT out of a total of 7379 genes
   expressed. (B) Diagram demonstrating the categorization of genes (up
   and down regulated, p < 0.01, ≥ 2 FC) during OB induction. 523 genes
   were unique to hESC-stromal-OB, 493 genes were unique to hMSC-TERT-OB
   and 172 genes were common to both cell types. (C) Genes up-regulated
   and (D) genes down-regulated during OB induction: 231 up-regulated
   genes and 262 down-regulated genes were unique to hESC-stromal; 335
   up-regulated and 188 down-regulated genes were unique to hMSC-TERT;
   genes common to both cell types totaled 91 (up-regulated) and 81
   (down-regulated). (E) GO BP annotations of genes (over 2 FC; detection
   threshold p ≤ 0.01) common to hESC-stromal and hMSC-TERT during OB
   differentiation (n = 172).

3.3. Comparison of molecular phenotype of hESC-stromal-OB vs. hMSC-TERT-OB

   Prior to selecting a time point during OB induction for microarray
   analysis, hMSC-TERT and hESC-stromal, undergoing differentiation
   induction, were compared using ALP activity and ALP gene expression as
   a measure for osteoblast lineage differentiation. From these
   preliminary experiments d6 of hESC-stromal-OB and d7 of hMSC-TERT-OB
   were selected as being the most comparable time points (data not
   shown). In order to detect whether hESC-stromal and hMSC-TERT employ
   similar biological processes during ex vivo OB differentiation, we
   compared hESC-stromal-OB and hMSC-TERT-OB utilising the following four
   bio-informatic approaches.

   First, osteoblast differentiation regulated genes were compared between
   hESC-stromal and hMSC-TERT. Comparison of fold induction (OB
   induced/undifferentiated) identified a comparable number of genes both
   up and down regulated: 695 genes differentially regulated (FC ≤ − 2 or
   ≥ 2) in hMSC-TERT-OB and 665 genes in hESC-stromal-OB. Among these, 172
   genes (≈ 30%) were common to both cell types following differentiation
   suggesting a common OB differentiation program. Employing the DAVID
   tool for GO functional annotation of BP, the highest enriched GO
   categories of these 172 genes included mitosis, response to estradiol
   stimulus, insulin receptor signalling and regulation of apoptosis
   ([94]Supplementary Fig. 1E). In addition, the top 10 enriched GO
   categories for each cell type exhibited similarities e.g. cell
   adhesion, angiogenesis, cytoskeletal organisation, response to hormone
   stimulus and regulation of apoptosis ([95]Fig. 2C and D). Conversely,
   differences in GO categories were also observed. GO categories for
   epithelial-to-mesenchymal (EMT) transition and cell morphogenesis were
   unique for hESC-stromal-OB ([96]Fig. 2C) whereas hMSC-TERT-OB
   ([97]Fig. 2D) were enriched in GO BP categories for cell cycle
   processes, mitotic processes and response to oxygen levels. Data lists
   detailing genes annotated to the top 10 categories are presented in
   Supplementary Table 2.

   Second, we examined the separate annotations for genes that were up- or
   down-regulated during OB differentiation in the two cell types.
   hESC-stromal-OB demonstrated 231 genes and hMSC-TERT showed 335 genes
   that were up-regulated ≥ 2 FC during OB differentiation, and among
   these 91 genes were common between the two cell types
   ([98]Supplementary Fig. 2). Common GO categories were present e.g.
   proliferation, response to hormone stimulus, regulation of apoptosis
   and regulation of cell adhesion. For genes down-regulated during OB
   differentiation, 262 genes and 188 genes were found in hESC-stromal-OB
   ([99]Supplementary Fig. 3B) and hMSC-TERT-OB ([100]Supplementary
   Fig. 3C), respectively. Among these 81 genes were common between the
   two cell types ([101]Supplementary Fig. 3A) and were enriched for GO
   categories: mitosis and regulation of phosphorylation. The list of
   differentially up-regulated or down-regulated genes in hESC-stromal-OB
   and hMSC-TERT-OB are given in Supplementary Table 2.

Supplementary Fig. 2.

   Supplementary Fig. 2
   [102]Open in a new tab

   GO functional enrichment of biological processes up-regulated OB genes
   (p ≤ 0.01; ≥ 2 FC). (A) GO functional enrichment of genes commonly
   expressed in both hESC-stromal-OB and hMSC-TERT-OB (n = 91); (B) GO
   functional enrichment of genes unique to hESC-stromal-OB (n = 231); (C)
   GO functional enrichment of genes unique to hMSC-TERT-OB (n = 335).

Supplementary Fig. 3.

   Supplementary Fig. 3
   [103]Open in a new tab

   GO functional enrichment of down-regulated OB genes annotated to
   biological processes (p ≤ 0.01; ≤− 2 FC). (A) GO functional enrichment
   of genes commonly expressed in both hESC-stromal-OB and hMSC-TERT-OB
   (n = 81); (B) GO functional enrichment of genes unique to
   hESC-stromal-OB (n = 262); (C) GO functional enrichment of genes unique
   to hMSC-TERT-OB (n = 188).

   Third, to test for the presence of a common “osteoblastic
   differentiation signature” we curated an ‘ossification’ GO dataset that
   identified 293 genes. Of these 293 genes, we found that 135 did not
   pass the p-value detection threshold of ≤ 0.01 in either
   hESC-stromal-OB or hMSC-TERT-OB. Of those genes that were expressed
   (p ≤ 0.01), 115 (73%) were expressed equally in both cell types, 25
   (16%) genes were uniquely expressed in hESC-stromal-OB and 18 (11%)
   genes uniquely expressed in hMSC-TERT-OB (Supplementary Table 3).

   Finally, we compared pathways enriched in hESC-stromal-OB and
   hMSC-TERT-OB using MetaCore®. All expressed genes (detection p
   value ≤ 0.01) in hMSC-TERT-OB and hESC-stromal-OB were employed. The
   top 10 enriched pathways for hESC-stromal-OB and hMSC-TERT-OB are shown
   in [104]Fig. 3A and B, respectively. While common genetic pathways for
   both cell types included EMT and CSF signalling, their function may
   differ. For example, hESC-stromal-OB demonstrated enrichment for GM-CSF
   with a role in formation and release of inflammatory cytokines whereas
   hMSC-TERT-OB was enriched in G-CSF with a role in formation and release
   of granulocytes from bone marrow. In addition, both cell types
   demonstrated enrichment of the EMT pathway through induction of TGFB
   signalling. Interestingly, hMSC-TERT-OB demonstrated additional EMT
   pathway induced via MAPK and SMAD signalling.

Fig. 3.

   Fig. 3
   [105]Open in a new tab

   Enrichment analysis (MetaCore®) for unique and common pathways: (A)
   pathways unique to hESC-stromal-OB; (B) pathways unique to
   hMSC-TERT-OB. Both annotations demonstrate categories involved in EMT
   and TGFβ pathways suggesting these are important to both cell types
   during differentiation.

3.4. Validation of microarray data of hESC-stromal-OB and hMSC-TERT-OB

   We chose 24 genes (FC ≤ − 2 or ≥ 2) that were found among the core
   “ossification” gene category for validating the microarray data using
   real-time qRT-PCR ([106]Fig. 4A–C). We included genes present in both
   hESC-stromal-OB and hMSC-TERT-OB: CLEC3B, SRGN, IFITM1, PTN, MN1, EXT1,
   CBS and IGFBP5, genes expressed uniquely in hESC-stromal OB: DCN, COMP,
   MGP, P2RX7, SOX9, GLI2, PTGS2, IL6 or genes expressed uniquely in
   hMSC-TERT-OB: TGBF3, TGBF2, PRKD1, CEBPB, ID1, ECM1, GJA1, MMP2. As
   seen in [107]Fig. 4A–C, qRT-PCR revealed a close correlation of qRT-PCR
   measurements with those from DNA microarray data.

4. Discussion

   In the present manuscript, we have compared the molecular phenotype of
   hMSC either derived from hESC or from bone marrow. While there were
   some quantitative differences between the gene expression and cellular
   responses, the two cell populations exhibited a high degree of homology
   in relation to undifferentiated gene expression and qualitative
   responses to OB differentiation induction. In addition, we identified a
   common molecular signature and common core of genes and genetic
   pathways that are employed by both cell types during in vitro OB
   differentiation.

   We used an immortalised cell line (hMSC-TERT) to compare with
   hESC-stromal cells undifferentiated and under osteoblastic induction.
   The hMSC-TERT cell line has been extensively characterised by our group
   and has been demonstrated as a good predictive model for the biological
   behaviour of primary hMSC ([108]Abdallah et al., 2005, [109]Simonsen et
   al., 2002). In addition, publications from other groups comparing
   immortalised versus primary cells have also demonstrated retention of
   basic cell characteristics ([110]Boerma et al., 2006, [111]Ouellette et
   al., 2000). However, the expression of ectopic hTERT, and thus
   immortalisation, is indicative that genes associated with cell cycle
   will be up-regulated in comparison with primary cell lines. Thus,
   over-representation of cell cycle categories in undifferentiated
   hMSC-TERT biological processes GO is not unexpected. We expected that
   the most up-regulated genes observed in non-induced hMSC-TERT would be
   associated with mitosis, however, while cell cycle genes were
   overrepresented, the majority of the upregulated genes, annotated to
   mitosis, were between 2–3 FC with only 3 genes demonstrating > 3 FC
   upregulation when compared to hESC-stromal (INCEP, CCNF, NEK2). These 3
   genes were not found within the top 50 hMSC-TERT genes which were
   differentially annotated in hESC-stromal. Functional clustering of the
   top 50 genes demonstrated enrichment in categories: response to hormone
   stimulus and skeletal system development. In comparison functional
   clustering of the top 50 genes in hESC-stromal revealed enrichment for
   categories of: smooth muscle proliferation and muscle organ
   development. This suggests that these differences reflect differences
   of the ontology of the two cell types (adult versus embryonic) rather
   than immortalisation caused by overexpression of hTERT in hMSC-TERT.

   Data analysis determined that 90.3% of genes expressed were similar
   between undifferentiated hESC-stromal and hMSC-TERT. Our data, thus,
   corroborates previously published reports comparing the molecular
   phenotype of hESC-derived MSC-like cells and bone marrow-derived hMSC
   ([112]Raynaud et al., 2013, [113]Barbet et al., 2011, [114]Bigdeli et
   al., 2010, [115]Barberi et al., 2007). The authors demonstrated a close
   relationship between hESC-MSC-like cells and BM-hMSC employing
   principal component analysis (PCA) and hierarchical clustering
   ([116]Raynaud et al., 2013). Barbet et al. identified differentially
   expressed genes utilising a Taqman low-density array and reported that
   hESC-MSC clustered more closely with BM-hMSC than with undifferentiated
   hESC ([117]Barbet et al., 2011). Barberi et al. utilised affymetrix
   oligonucleotide arrays on a CD73^+ sorted population derived from hESC
   and compared these with BM-hMSC ([118]Barberi et al., 2005). Similar to
   our data the authors found enrichment for genes associated with
   vascular development and inflammatory response/wound healing. Finally,
   [119]Bigdeli et al. (2010) reported similarities in the enrichment of
   osteoblastic genes in undifferentiated hESC-MSC and BM-MSC.

   Other studies have examined the molecular phenotype of osteoblastic
   cells differentiated from hESC i.e. hESC-MSC-OB as compared with
   hiPSC-MSC-OB derived from induced pluripotent stem cells (iPSC)
   ([120]de Peppo et al., 2013) and demonstrated the ability of both cell
   types to undergo in vitro osteoblastic differentiation. Global DNA
   microarray analysis of hESC-MSC and hiPSC-MSC demonstrated similarities
   and differences in both undifferentiated cells and in cells undergoing
   OB differentiation. However, the limitation of these studies is the
   absence of comparison with bona fide osteoblastic cell populations e.g.
   osteoblastic cells derived from BM-hMSC as we did in our current study.
   Our comparison with a Gene Ontology (GO) derived osteogenic gene set
   demonstrated that 73% of osteogenic genes found in hESC-stromal-OB and
   hMSC-TERT-OB were similar. Within this common list, 11 genes were
   regulated over 2 FC (p ≤ 0.01) of which only 1 gene (STC1) was
   differentially regulated. In addition, of the 172 genes in the
   OB-induced cells (FC ≤ − 2 or ≥ 2; p ≤ 0.01) which were commonly
   expressed in both cell types, 86% of the genes were similarly
   expressed. Analysis of genes which were differentially regulated during
   OB induction (FC ≤ − 2 or ≥ 2; p ≤ 0.01) in hMSC-TERT-OB (n = 523) and
   in hESC-stromal-OB (n = 493), showed that the majority (96.7%
   hMSC-TERT-OB; 97% hESC-stromal-OB) did not demonstrate annotations for
   GO ossification. Data presented here therefore demonstrates a common
   set of genes for OB induction shared between hESC-stromal and
   hMSC-TERT. The large gene sets not exhibiting osteogenic annotation
   demonstrates that osteogenesis involves an activation of a complex
   network of genes related to basal biological processes.

   A large number of genes in both hESC-stromal-OB and hMSC-TERT-OB were
   annotated as related to epithelial-to-mesenchymal transition (EMT). EMT
   is a mechanism for generation of primary mesenchyme during
   embryogenesis ([121]Kalluri and Weinberg, 2009) and is orchestrated
   through Wnt signalling ([122]Kalluri and Weinberg, 2009, [123]Ullmann
   et al., 2008). Among the 130 genes regulated during EMT ([124]Groger et
   al., 2012), 68 genes were present in hESC-stromal and 65 genes in
   hMSC-TERT. Our study corroborates findings from previous studies
   reporting that development of MSC-like cells in hESC cultures was
   mediated through EMT as evidenced by differential regulation of
   E-cadherin and N-cadherin ([125]Ullmann et al., 2007), expression of
   EMT-associated genes ([126]Hwang et al., 2008) and flow cytometric
   analysis of E-Cadherin ([127]Boyd et al., 2009).

   We observed that the differentiated progeny of hESC, hESC-stromal
   cells, did not express pluripotent markers e.g. POU5F1, SOX2, NANOG,
   DNMT3B, GABRB3, and TDGF1 suggesting the loss of pluripotency and
   providing strong evidence that the hESC-derived differentiated cells
   will not form teratomas or tumours when implanted in vivo. In support
   of this, tumour formation was absent when hESC-stromal were
   subcutaneously implanted in immune compromised mice ([128]Harkness et
   al., 2011).

   Our data demonstrate similarities in the molecular phenotype of
   undifferentiated and OB differentiated hESC-stromal cells when compared
   with hMSC encouraging the use of hESC-stromal cells as an alternative
   source of hMSC for regenerative medicine protocols.

   The following are the supplementary data related to this article.
   Supplementary Table 1

   Primer sequences used for validation of microarray data.
   [129]mmc1.docx^ (17.8KB, docx)

   Supplementary Table 2 GO annotation from DAVID for genes regulated over
   2 FC (p ≤ 0.01).

   Supplementary Table 3 list of GO ossification genes present in
   hESC-stromal-OB and/or hMSC-TERT-OB.
   [130]mmc2.xlsx^ (91.3KB, xlsx)

Competing interests

   The authors have no competing interests.

Author contributions

   LH, MK, AA, JA were responsible for the experimental concept and
   design; LH and AJ carried out the laboratory experiments and data
   analysis; RAD and JA performed the microarray and data analysis; NAT
   and MRW carried out the bioinformatic analysis; LH and NAT prepared the
   manuscript; all authors read and approved the manuscript.

Acknowledgements