Dear Editor

The contribution of cellular senescence to the aging of animals is still a controversial issue. Human diploid fibroblasts can be cultured in vitro for a finite number of divisions, after which they undergo a metabolic condition called replicative senescence.1,2 This condition is characterized by the absence of response to mitogenic stimuli and elevated levels of cyclin-dependent kinase inhibitors,3 which in turn could be responsible for the presence in these cells of dephosphorylated retinoblastoma protein and low E2F transcription factor activity.4,5 Furthermore, there is convincing experimental evidence that telomere shortening plays a key role in the establishment of the senescent phenotype.6,7 Cells derived from old individuals share with in vitro senescent cells some cellular and molecular phenotypes, but it is not clear whether these phenotypes are completely overlapping. The data currently available show that normal cells from aged donors have a proliferative potential lower than those taken from young individuals8,9,10,11,12 and, consistently, cells from patients suffering from diseases characterized by a precocious senescence, such as Werner's syndrome, Down's syndrome and Hutchinson–Gilford progeria, show a significant impairment of proliferative potential compared to cells from healthy donors of comparable ages.13,14,15,16,17 Other results, however, showed that, in a large series of normal subjects of various ages, the maximal population doubling of skin fibroblasts is completely independent from the age of the donor.18 Furthermore, in vivo aged cells, like in vitro senescent fibroblasts, show the appearance of the senescence-associated β-galactosidase activity,19 and the accumulation of cyclin-dependent kinase inhibitors, p21waf1 and p16.3,20,21

A direct approach to address the possible relationship between cells undergoing in vitro replicative senescence and cells taken from old subjects is to compare their gene expression profiles.

In a recent paper, the expression profile of human fibroblasts taken from old subjects has been compared to that of cells taken from young individuals,22 showing that there are numerous genes expressed at different levels in these two types of cells. We selected some of these genes to analyze whether the differences of their expression observed in fibroblasts taken from old vs young individuals could be similarly observed with the appearance of in vitro acquired replicative senescence.

The cells used for this study were: (i) IMR-90 human embryo fibroblasts (EF); (ii) two populations of skin fibroblasts taken from young donors of 7 and 9 years (CRL-1474 and CRL-7469), indicated as YF1 and YF2, respectively; (iii) two populations of skin fibroblasts taken antemortem from old individuals of 92 and 96 years of age (AG04064A and AG04059B) indicated as OF1 and OF2, respectively. Proliferation capacity of these cells was examined at different Population Doubling Level (PDL) by measuring BrdU incorporation, during 48 hour incubation in the presence of the nucleoside. EF cell BrdU incorporation is >90% until 50 PDL, and thereafter decreases, reaching a mean value of 5% at 60 PDL (EF60). YF1 and YF2 cells have a percentage of BrdU incorporation very similar to that of EF cells at 25 PDL (EF25), which remains elevated until 35 PDL. OF1 and OF2 cells maintained in culture up to 20 PDL showed more than 85% of BrdU positive cells, but this percentage decreases after a few more PDL (only 6% of BrdU positive cells after 25 PDL).

We also measured the levels of p21waf1 mRNA and the number of SAβ-gal positive cells; p21waf1 mRNA is low in EF28, YF1 and YF2 cells, in which SAβ-gal positive cells are quite absent, while EF60 and OF cells contain several fold more p21 mRNA than EF28 and YF cells and a high percentage of SAβ-gal positive cells (⩾50%). In conclusion these results are in agreement with the observations reported by others, indicating that cells taken from old individuals are similar to the cells that acquired the senescent phenotype in vitro: in fact, they have a limited proliferation potential, show increased levels of p21waf1 mRNA and most of them are SAβ-gal positive.

To further examine this similarity we analyzed by quantitative real-time PCR the levels of expression of some genes in the various cell populations under examination. Table 1 shows that the two mRNAs encoding prostaglandin endoperoxide synthase (PTGS1) and fibromodulin (Fmod) are significantly increased in OF1 and OF2 cells compared to YF1, in agreement to that reported in the paper of Ly et al.22 The mRNA levels of these genes in YF2 cells are very similar to those present in YF1 cells (data not shown). In particular, we found that OF120 cells compared to YF1 cells, on three different RNA preparations each measured in triplicate, have 16-fold more PTGS1 mRNA and 20-fold more Fmod mRNA. These differences were confirmed when we analyzed the RNA from OF125 cells (slowly- or not-dividing cells, see above), although the differences were less strong. On the contrary, the concentration of PTGS1 mRNA in senescent EF60 cells is similar to that observed in EF28 cells, while Fmod mRNA shows an opposite change, being decreased by more than sevenfold in EF60 compared to EF28 cells.

Table 1 Gene expression levels in in vitro senescent cells and in cells from old individuals
Full size table

Another experimental system that has been used to induce replicative senescence in vitro is based on the exposure of fibroblasts to low doses of oxidants.23,24,25 EF28 cells, grown in the continuous presence of 100 μM diethylmaleate (DEM), a glutathione depleting agent, stop growing after about 6 days, as demonstrated by their BrdU incorporation (about 5%). Accordingly, p21 mRNA levels significantly increased and many cells became SAβ-gal positive (>45%) (data not shown). In EF28 cells, after 6 days of mild oxidative stress, PTGS1 and Fmod mRNA levels are both decreased about six- and four-fold, respectively, compared to untreated cells and this modification is again opposite to that observed in OF cells compared to both YF populations. Therefore, the behaviour of PTGS1 and Fmod mRNAs demonstrate that the molecular phenotypes of in vitro induced senescence and in vivo acquired aging are distinguishable.

Two other genes analyzed are cathepsin C (CatC) and metalloproteinase (HME). As shown in Table 1, the first is expressed at similar levels in both EF28 and EF60 cells and in EF28 cells exposed to low concentrations of DEM, and no significant difference is seen comparing both OF1 and OF2 cells with the counterparts taken from young donors or with the no longer dividing OF125 cells. HME mRNA was always increased, both in OF cells compared to YF, in EF28 vs EF60 and in EF28 cells following mild oxidative stress. Therefore, HME mRNA modifications appear to be similar to those observed for p21waf1: they cannot distinguish between in vitro acquired senescence and in vivo cellular aging.

Five other genes showed a behaviour completely different from those described above. Table 1 shows that the mRNAs encoding cyclin A (CycA), cyclin F (CycF), thymidylate synthase (TS), hepatocyte nuclear factor-3/fork head homolog 11A (HFH-11A) and cyclin-sensitive ubiquitin carrier protein (Ucar) are many fold decreased in EF60 cells compared to EF28. Similarly, EF28 cells exposed to low doses of DEM displayed a significant decrease of these mRNAs after 2 days of treatment, except for TS mRNA whose level decreases starting from day 4. It is worth noting that the extent of the decreases observed in the cells treated for 6 days with DEM are, in many cases, several fold higher that those observed in EF60 cells.

Surprisingly, OF1 and OF2 cells compared to YF1 and YF2 cells did not show any significant change in the expression of these mRNAs (Table 1, columns 1,2). On the contrary, OF1 cells, grown until they have reached 25 PDL (OF125) and no longer incorporating BrdU, show a significant decrease of the mRNA levels of all the five genes (Table 1, column 3). Therefore, the levels of these five genes appear to be modified only in cells which have ceased to grow, regardless their origin.

In summary, these experiments demonstrate that it is possible to distinguish the molecular phenotype of human fibroblasts which have acquired the senescent phenotype in vitro, either by repeated in vitro passages or by mild oxidant treatment, with that of fibroblasts taken from old individuals (i.e. in vivo acquired senescence). Two examples of the differences between in vitro and in vivo acquired senescence are the genes encoding prostaglandin endoperoxidase synthase and fibromodulin. Their opposite behaviours in the two senescent phenotypes support the hypothesis that other genes could also be regulated in a similar fashion.

The existence of a gene expression profile characteristic of in vitro senescent cells, regardless of their origin, is suggested by the changes observed in the expression of some other genes, including cyclin A, cyclin F, thymidylate synthase, hepatocyte nuclear factor-3/fork head homolog 11A and cyclin-sensitive ubiquitin carrier protein. All these genes encode proteins that play a role in cell cycle progression, and it is reasonable that in a cell, no longer dividing but still surviving, many of the genes involved in S-, G2- and M-phases of the cell cycle are downregulated.