To investigate the effect of senescence on differentiation ability of ASCs, we derived clonal cultures of ASCs from single freshly isolated cells [3]. The adipogenic differentiation capacity of these clones in early passage cultures (passage 4; P4) has been described previously [10]. Three clones, designated B1, B2 and B3, were cultured to senescence. Senescence was reached at passage P18 (i.e., after 48 population doublings) for clone B1, P16 (44 population doublings) for clone B2 and P17 (46 population doublings) for clone B3. Senescence was manifested by cell cycle arrest, as judged by the absence of cell division on the basis of cell counts upon at least two rounds of subculture after these passage numbers were reached (data not shown), and by typical enlarged cell morphology (Fig. 1A). Passage number achieved at senescence is referred to as Ps hereafter.

To evaluate the differentiation capacity of early passage vs. senescent cells, all clones were induced to differentiate toward the adipogenic pathway for three weeks, starting either at P7 or at Ps. All P7 clones differentiated efficiently, on the basis of Oil Red-O staining (Fig. 1B). Differentiation, however, was impaired in senescent cells (Fig. 1C, Ps). Note that differentiation was elicited for each clone with 20,000 cells/cm² both at P7 and at Ps, so we could not attribute differences in the extent of adipogenesis to differences in cell density.

Quantitative analysis of expression of adipogenic genes in undifferentiated P7 and Ps cells by real time reverse transcription (RT) polymerase chain reaction (PCR) corroborated the reduced ability of senescent ASCs to commit to adipogenesis. In undifferentiated P7 cells, expression of peroxisome proliferator-activated receptor gamma 2 (PPARG2) occurred at low levels and was not significantly altered in senescent cells in any of the clones (P > 0.05; Fig. 2A). Expression of fatty acid-binding protein 4 (FABP4) [28] mRNA, also detected at low levels at P7, was unaltered at Ps in clone B2 (P > 0.1), but was upregulated 3-fold in clones B1 and B3 (P < 0.01 for each clone, relative to undifferentiated cells; Fig. 2A). Transcripts for lipoprotein lipase (LPL) were undetectable at P7 or Ps in clones B2 and B3, consistent with previous observations [10], but were expressed at a low level in both P7 and senescent cells in clone B1 (Fig. 2A). Lastly, leptin (LEP) mRNA was undetected in clone B1 but was upregulated 5- to 6-fold at senescence in clones B2 and B3 (P < 0.01 relative to undifferentiated cells; Fig. 2A). It is noteworthy however, that all transcripts detected in undifferentiated P7 ASCs were expressed at very low levels, as illustrated by elevated Ct values (> 35) despite high real time PCR efficiency (data not shown).

Three weeks of adipogenic stimulation initiated at Ps promoted upregulation of all genes relative to undifferentiated cells in all clones (P < 0.01 for any gene examined), except for LEP which remained unchanged in clone B3 (P > 0.1; Fig. 2B). Extent of gene upregulation varied between clones, reflecting differences in the ability of each clone to commit to adipogenesis even in late passage culture (P < 0.01 between clones, for each gene). This is similar to what we previously observed with the same clones at passage 4 [10]. Note, however, that a significant part of this variation was due to the low levels of transcripts detected in undifferentiated cells, small variations in which can result in large differences in relative upregulation upon differentiation (small variations in PCR efficiency also affect quantification of mRNA levels [29]). Although transcriptional activation took place, LPL mRNA levels in clones B2 and B3, and LEP mRNA level in clone B3 could not be quantified as these genes were not expressed in undifferentiated cells (see Fig. 2A). Furthermore, FABP4 and LPL mRNA levels in ASCs after three weeks of adipogenic stimulation at P7 were higher than after induction of differentiation at Ps (P < 0.05 for FABP4 and P < 0.005 for LPL; one-sample t-test), while those of PPARG2 and LEP were not different (P > 0.05, one-sample t-test; Fig. 2C). Impaired lipid accumulation and transcriptional activation of adipogenic genes during culture of ASCs to senescence illustrate a reduced commitment to adipogenesis, despite stable or moderately upregulated transcript levels in undifferentiated cells.

To determine whether reduced transcriptional level of adipogenic genes in senescent ASCs had an underlying epigenetic basis, we assessed the DNA methylation status in the LEP, PPARG2, FABP4 and LPL promoters by bisulfite sequencing. Bisulfite conversion of DNA relies on the deamination of unmethylated cytosines into uracils, while methylated cytosines are protected from deamination. PCR converts uracils to thymidines and sequencing of several PCR products cloned in bacteria enables a quantitative assessment of the position and number of methylated cytosines in the region examined [30]. Regions analyzed by bisulfite sequencing in the LEP, PPARG2, FABP4 and LPL promoters are shown in Fig. 3A. We first validated the significance of the regions examined by showing that the extent of methylation was consistently lower in all promoters in differentiated SGBS (Simpson-Golabi-Behmel Syndrome) human adipocytes (0–45% methylated CpGs) than in peripheral blood T cells (24–72% methylation; Fig. 3B; P < 10^-4 for each promoter; t-test).

DNA methylation profiles of undifferentiated senescent cells are shown in Fig. 4A and illustrated for all individual clones in Additional file 1, Supplementary Fig. 1. The LEP, PPARG2 and LPL promoters were largely hypomethylated in all clones examined, with percentages of methylation ranging from 7 to 29% (Table 1, Undifferentiated). We found that the preferred areas of methylation in each locus we previously identified in early passage (P4) cells [10] were maintained in Ps cells (Fig. 4A). Furthermore, as in P4 cells, methylation profiles were mosaic in senescent cells, both between ASC clones and within a clonal population (Fig. 4A). The FABP4 promoter displayed greater variation in global methylation with clone B2 being 25% methylated, while B3 was 75% methylated (Fig. 4A; Table 1, Undifferentiated; Additional file 1, Supplementary Fig. 1C). Note that although adipogenic differentiation potential was evaluated in ASCs at P7 in this study, early passage methylation status was previously determined at P4 [10]; however no or minimal methylation differences were anticipated between P4 and P7, relative to senescence (P16-18).

Methylation profiles of each promoter in senescent cells (Fig. 4B, red lines) were compared to those obtained earlier [10] in the same ASC clones cultured to P4 (Fig. 4B, green lines). Methylation profiles remained overall similar during culture to senescence (Fig. 4A, B). However, two CpGs were found to be hypermethylated in FAPB4 (CpGs 1 and 2; P < 0.01 and P < 0.001, respectively; t-tests), owing to hypermethylation in clones B1 and B3 (Additional file 1, Supplementary Fig. 1C). Moreover, we found one CpG to be hypomethylated in LPL (CpG 6; P < 0.001; t-test; Fig. 4B) also owing to demethylation of this cytosine in clones B1 and B3 (Additional file 1, Supplementary Fig. 1D). When global percentages of methylation were compared between undifferentiated individual ASC clones at P4 and Ps, however, LEP, PPARG2 and FABP4 were found to be hypermethylated upon senescence in clone B1, while clones B2 and B3 only showed LEP and FAPB4 hypermethylation, respectively (Table 1, Undifferentiated). LPL promoter methylation was reduced between P4 and Ps in clone B1 and remained stable in the other clones (Table 1, Undifferentiated).

These results indicate that overall, adipogenic promoters retain the largely hypomethylated profile that characterizes ASCs. Nevertheless, a few CpGs become hyper- or hypomethylated upon senescence in the FABP4 and LPL promoters, respectively, albeit not in all clones.

We next examined the methylation profile of adipogenic promoters after adipogenic stimulation in senescent cells and compared these profiles to those of cells differentiated at P4 [10]. Methylation patterns were similar in the LEP promoter in differentiated P4 and Ps cells (Fig. 4C, D; Table 1, Differentiated), with only two CpGs, No. 5 and 21, being hypermethylated at Ps (P < 0.01; t-tests; Fig. 4D). Upon examination of individual clones, however, clones B2 and B3 displayed enhanced methylation at senescence (P < 0.001; t-tests; Table 1, Differentiated) due to hypermethylation of several CpGs (Additional file 1, Supplementary Fig. 2A). In PPARG2, enhanced methylation was seen at Ps compared to P4 (Fig. 4C, D; Table 1, Differentiated; P = 0.011; t-test) due to hypermethylation of CpGs No. 4, 5 and 6 (Fig. 4D; P < 0.001 for each cytosine; t-tests; Table 1; Additional file 1, Supplementary Fig. 2B). The FABP4 and LPL promoters displayed remarkably similar methylation profiles in Ps and P4 differentiated cells (Fig. 4C, D), with only clone B3 showing FABP4 hypermethylation in differentiated Ps cells (Table 1) due to enhanced methylation of CpG 2 (Additional file 1, Supplementary Fig. 2C). Finally, we found that methylation profiles of all promoters were overall similar in undifferentiated and in adipogenic-stimulated Ps cells (Fig. 5, showing a direct comparison of data of Figs. 4B and 4D, red curves), except for hypermethylation of CpG 5 in the LEP promoter (Fig. 5; P < 0.01; t-test). We concluded that overall, methylation patterns in adipogenic promoters are similar between adipogenic-stimulated cells in early passage and at senescence. However, as with undifferentiated cells, specific CpGs appear to be hyper- or hypomethylated in a promoter-specific and cell clone-specific manner.

We occasionally noted the substitution of a cytosine for a thymidine (C:T substitutions) in some LEP bisulfite sequences, consistently at CpG No. 4 (Fig. 4A, C, blue marks). Sequences were unambiguous and displayed either a methylated C or an unmethylated C (the latter read as A, as shown in Additional file 1, Supplementary Fig. 3A) or, unexpectedly a T (read as a T; Additional file 1, Supplementary Fig. 3B). To account for this substitution, it has been proposed that CpG methylation increases the mutation rate from CpG to TpG or CpA, as suggested by the lower number of CpGs observed over that expected in mammalian genomes [31,32].

In summary, methylation profiles at adipogenic promoters are generally maintained upon clonal culture of ASCs to senescence. Nevertheless, at the level of individual clones, methylation may be differentially affected in a gene-independent manner, but this does not necessarily correlate with transcriptional changes. Adipogenic stimulation of early passage or senescent cells does not markedly affect adipogenic promoter DNA methylation.

Little is known on alterations in DNA methylation in MSCs during in vitro aging and how these may relate to changes in differentiation potential. We examined three clonal lines of ASCs with an ability to differentiate toward adipogenic [10], osteogenic and endothelial pathways [11,13]. Similarly to cultured bone marrow MSCs [20,33], all ASC clones cultured to senescence displayed reduced adipogenic differentiation potential in vitro, on the basis of lipid synthesis and transcriptional upregulation of, primarily, FABP4 and LPL relative to the same clones in early passage cultures. Interestingly however, basal transcription of these genes was unaltered (PPARG2, LPL) or moderately enhanced (FABP4, LEP) in undifferentiated cells at Ps. Thus, reduced transcription efficiency in differentiated senescent cells cannot be attributed to lower transcript levels in the undifferentiated state.

What may be the basis for reduced adipogenic potential in senescent ASCs? First, PPARG2, a master regulator of adipogenesis [34], is upregulated to a level similar in early passage and senescent cells, thus we cannot explain reduced adipogenesis by PPARγ insufficiency. Second, adipogenic stimulation of senescent cells upregulates LPL expression, albeit to a lower level than in early passage cells. A suboptimal transcription level is likely to affect translation, preventing further phenotypic changes. Of note, transcription without detectable protein expression seems to be a feature of undifferentiated ASCs [3]. Third, transcription or synthesis of transcriptional regulators of adipogenic genes, such as C/EBPα [34], BMP4 [35] or ADD1/SREBP1c [36] may be compromised in senescent cells and may account for adipogenesis deficiency despite the putative availability of binding sites (see below). Indeed, failure to express C/EPBα results in inability to develop white adipose tissue in vivo [37,38]. Fourth, adipogenic failure may also be due to impaired downregulation of GATA2/3, pre-adipocyte factor 1 or Wnt10b, whose expression blocks cells in a pre-adipocyte stage [39].

To determine whether there was an epigenetic basis for reduced differentiation potential in senescent cells, we examined the DNA methylation profiles in the LEP, PPARG2, FABP4 and LPL promoters. Senescent cells maintained the overall mosaic hypomethylation of LEP, PPARG2 and LPL previously reported in P4 cells [10]. FABP4 exhibited hypermethylation in two undifferentiated ASC clones (B1, B3) at senescence. However, this did not correlate with reduced transcription potential upon adipogenic stimulation because clone B2, which is not hypermethylated at senescence, showed similar a reduction in transcription level to clones B1 and B3. Mosaic methylation has been reported in other primary cell cultures [26,27,40] and is likely the result of stochastic errors in maintenance methylation which occur over many rounds of replication. On the basis of our observations, therefore, it is difficult to attribute any change in promoter methylation to an effect of senescence per se.

Identification of relationships between the most hypomethylated areas in adipogenic promoters and binding sites for transcription factors argues that most, but not all, these sites remain available during culture. In the LEP promoter, methylation of CpGs No. 4–6 in clones B2 and B3 covers a recognition site for the methyl-CpG binding protein Kaiso [41,42]. Moreover, two Sp1binding sites span CpGs No. 15 and 16 while CpG No. 17 lies in a site occupied by an unidentified factor [43]. CpGs No. 15–17 are unmethylated at P4 and at senescence, presumably leaving these binding sites available. Additionally, CpG No. 21 lies within a C/EBPα recognition site flanked by E boxes for putative ADD1/SREBP1 binding, all of which are important for LEP expression [43,44]. In contrast to CpGs No.15–17, CpG No. 21 is hypermethylated in senescent ASCs, but only in the context of adipogenic stimulation, in clones B1 and B3. CpG No. 21 hypermethylation may account for limited or undetectable LEP upregulation in adipogenic stimulated senescent cells. Moreover, we found that LEP transcripts increased 5–6-fold in undifferentiated senescent clones B2 and B3, but not in clone B1, which displayed heaviest methylation on CpG No.21. Therefore, methylation of CpG No.21 correlates with impaired LEP transcriptional upregulation in senescent cells in the context of adipogenic stimulation.

Methylation profiles of the LPL and FABP4 promoters further illustrates a functional relationship between DNA methylation and transcription factor binding. The LPL promoter contains a PPAR responsive element between CpGs No. 7 and 8 [45,46], a sterol-responsive element acting as binding site for SREBP1c [36], and between CpGs No. 9 and 10, a CAAT box and a binding sequence for Oct1 and TFIIB. Notably, CpGs No. 7–11 are consistently unmethylated in both early and late passage ASCs, thereby enabling transcriptional activation of LPL upon differentiation even in senescent cells. Availability of transcription factor binding sites in senescent ASCs also yields for the FABP4 promoter, which is strongly activated upon adipogenic stimulation. Notably, this promoter harbors a C/EBPα binding site close to CpG No. 4 [47], the only cytosine unmethylated regardless of ASC donor [10], cell clone and differentiation state in early passage and in senescent cells.

Stochastic methylation in senescing ASCs is reminiscent of random methylation patterns established in human intestinal crypt stem cells in aging individuals [48,49]. Aging-related random methylation has been proposed to occur as a result of failure of maintenance methyltransferase activity, and/or exposure to environmental factors and health of the individual [24,50,51]. Our results provide evidence that methylation changes detected in senescent ASCs do not correlate with transcriptional defects and reduced differentiation potential. Likewise, there is to date little evidence linking DNA methylation of lineage-specific genes and aging-related impairment of stem cell function (tissue homeostasis) [20], stochastic changes in gene expression [52] and genome integrity [53]. Nevertheless, comparing aging to in vitro replicative senescence should be made with caution, as age of MSCs does not always affect differentiation potential in vitro [54].

In summary, our results argue that reduction of adipogenic potential in senescent ASCs may relate to transcriptional defects upon adipogenic stimulation. However, it cannot in general be attributed to specific methylation changes in adipogenic promoters, and thereby to availability of transcription factor binding sites regulated by CpG methylation. One exception to this contention is one CpG in the LEP promoter (CpG No. 21 in the present study), whose senescence-related methylation may impair upregulation of the gene upon adipogenic stimulation. There also seems to be a correlation between regions of CpGs hypomethylation and important transcription factor binding sites.

Cells with a CD45^-CD31^-CD34⁺CD105⁺ phenotype (ASCs) were isolated from the stromal vascular fraction of human adipose tissue as described [3]. In short, lipoaspirates obtained from the hip or thigh regions of female donors below 40 years of age were digested with collagenase and DNase I, and adipocytes were separated from stromal vascular cells by centrifugation. Stromal cells were strained through sieves and CD45^- cells were retrieved by negative selection. CD31^- cells were in turn negatively selected using FITC-conjugated anti-human CD31 antibodies and anti-FITC microbeads. Isolated ASCs are CD34⁺CD105⁺so no negative selection was performed for these markers [3]. For clonal expansion, single freshly isolated ASCs were cultured in 48-well plates in DMEM/F12 medium containing 50% fetal bovine serum. After ~16 h, the medium was replaced with DMEM/F12/10% fetal bovine serum and after 3 weeks, colonies were passaged by trypsinization. The three ASC clones used in this study (clones B1, B2, B3) have been previously characterized for their ability to differentiate toward adipogenic, osteogenic and endothelial cell lineages [10,11,13].

Cells were cultured to confluency in DMEM/F12/10% serum and stimulated for 3 weeks with 0.5 mM 1-methyl-3 isobutylxanthine, 1 μM dexamethasone, 10 μg/ml insulin (Novo Nordisk) and 200 μM indomethacin (Dumex-Alpharma). To visualize lipid droplets fixed cells were stained with Oil Red-O as described earlier [3].

Total RNA was purified using the DNAse-containing RNeasy Micro Kit (Qiagen). RT-PCR was carried out from 500 ng total RNA using the Iscript cDNA synthesis kit (BioRad). Quantitative RT-PCR reactions were performed in triplicates from two separate cDNA preparations on a MyiQ Real-time PCR Detection System using IQ SYBR^® Green (BioRad). RT-PCR primers used have been published earlier [10]. SYBR^® Green PCR conditions were 95°C for 4.5 min and 40 cycles of 95°C 30 sec, 60°C 30 sec and 72°C 30 sec, and transcript levels calculated [29] using GAPDH as a normalization control [10]. RT-blank control PCRs showed no product amplification for all genes examined in this study (data not shown). Relative transcripts levels were compared using paired t-tests and statistical significance expressed as a two-tailed P value with P < 0.05 considered as significant. When relevant, transcript levels were compared to a "reference" level (level 1) using one-sample t-tests.

Methylation of four adipogenic gene promoters was analyzed by bisulfite sequencing. The LEP promoter region examined, and known to be regulated by methylation [44], covered nucleotides 2719–2937 (GenBank U43589) and spanned 27 potentially methylated cytosines within nucleotides -203 to +16 with respect to the transcription start site (TSS). The PPARG2 promoter region [55] spanned nucleotides 108–587 (GenBank AB005520) and included 6 CpGs within nucleotides -620 to – 140 with respect to the TSS. The FABP4 (GenBank NM_001442) promoter region encompassed 4 CpGs within nucleotides -478 to -65 with respect to the TSS. The LPL promoter region (GenBank X68111) spanned bases 1321–1777 and included 11 CpGs within nucleotides -398 to +59 with respect to the TSS. Sequence positioning was identified using Ensembl [56].

For bisulfite conversion, DNA was purified by phenol-chloroform-isoamylalcohol extraction. Cells were lysed for 10 min in lysis buffer (10 mM Tris-HCl, pH 8, 100 mM EDTA, 0.5% SDS) and digested with 0.1 mg/ml Proteinase K overnight. Bisulfite conversion was performed using the MethylEasy DNA Bisulfite Modification kit as detailed by the manufacturer (Human Genetic Signatures) [10]. Conversion rates, assessed by conversion of all cytosines in non-CpG dinucleotides to uracils (reading as thymidines after PCR), was 100% (data not shown). Converted DNA was used fresh or stored at -20°C. Converted DNA was amplified by PCR using primers commercially available (Human Genetic Signatures) for LEP. Primers for PPARG2, FABP4 and LPL were designed using Methprimer [57] and previously reported [10]. For PPARG2, FABP4 and LPL, PCR conditions were 95°C for 7 min and 40 cycles of 95°C 1 min, 54°C 2 min and 72°C 2 min, followed by 10 min at 72°C. For LEP, nested PCRs were performed, each with 95°C for 3 min and 30 cycles of 95°C 1 min, 50°C 2 min and 72°C 2 min, followed by 10 min at 72°C. PCR products were cloned into E. coli by TOPO TA cloning (Invitrogen) and reverse-sequenced (MWG Biotech).

Methylation data are represented as filled circles (methylated CpG) and empty circles (unmethylated CpG) for each bacterial clone obtained. Each row of circles represents methylation pattern based on the sequence of one cloned PCR product (i.e., one genomic allele). Global methylation percentages for a given promoter were compared using paired t-tests. Numbers of methylated cytosines for a given CpG were compared between cell populations using paired t-tests and significance was expressed using two-tailed P values.