The low expression levels of transgenes adjacent to telomeres compared to the levels of the same transgenes integrated at interstitial sites in our study demonstrates TPE in mouse ES cells, the first such report for normal mammalian cells. Following initial selection, expression levels of telomeric transgenes were, on average, 4.5-fold lower than their levels of expression at interstitial sites. Furthermore, expression levels of the neo and HSV tk genes at telomeric sites were 7- and 13-fold lower, respectively, than their expression levels from interstitial sites that did not contain telomeric repeat sequences. Further evidence for TPE in mouse ES cells was provided by the increased expression of the neo gene in the FS-1, FS-2, and FS-3 subclones, in which the transient expression of the I-SceI endonuclease had resulted in the loss of adjacent telomeres (Fig. ). This rules out a role for proximal mouse sequences in repression of the telomeric transgenes, since the mouse DNA sequences originally proximal to the telomeric plasmid now flank the integration site on both sides in these subclones.
Our results also demonstrate that complete silencing of the telomeric transgenes occurs in the absence of selection and that this silencing involves extensive methylation of DNA (Fig. , , and ). However, DNA methylation appears to be a secondary mechanism for suppression of the telomeric transgenes which is distinct from the mechanism responsible for their low levels of expression in cells kept under selection. This conclusion is supported by several lines of evidence. First, the very low expression levels of the telomeric transgenes in clones A211 and A405 occur even though there is no detectable or minimal methylation of the plasmid sequences. This also holds true for ES cell lines isolated from 3-day-old preimplantation embryos. Second, subclones of A211 and A405 that do not perform de novo methylation of the plasmid DNA and fail to silence the transgenes after prolonged growth without selection still express the transgenes at very low levels. Third, while treatment of cells with 5-AzaC reverses the DNA methylation and complete silencing of the telomeric transgenes, it does not alleviate their initial high level of repression. Finally, the increased expression of the neo
gene in the FS-1, FS-2, and FS-3 subclones following the loss of the adjacent telomere (Fig. ) is not accompanied by a decrease in DNA methylation (data not shown). Based on these observations, we believe that the complete silencing of genes located near telomeres in mouse ES cells occurs through a two-step mechanism, with an initial repression step mediated through a telomeric-specific chromatin conformation, which is then followed by complete silencing mediated by DNA methylation, similar to silencing in genomic imprinting and X chromosome inactivation (8
). Like these other forms of epigenetic silencing, the complete silencing of telomeric transgenes is also reversible, as demonstrated by the fact that demethylation of DNA following treatment with 5-AzaC results in both an increase in survival in G418 and an increase in expression levels.
We also demonstrated extensive silencing and DNA methylation of the telomeric transgenes in tissues and MEFs isolated from A405 transgenic mice. The telomeric transgenes are silenced in essentially all of the cells in these tissues since the DNA containing the transgenes is completely resistant to digestion with methylation-sensitive restriction enzymes, and MEFs were unable to form any colonies in G418 but were able to form large numbers of colonies in ganciclovir. As in the case with ES cell clones A211 and A405, this silencing was reversible, as MEFs were capable of forming colonies in G418 after treatment with 5-AzaC. In contrast, ES cell lines established from early embryos of the A405 transgenic mice grew well in G418 and were sensitive to ganciclovir. This indicates that the DNA in these ES cells was not extensively methylated in vivo and therefore appears to have been demethylated, similar to the rest of the genomic DNA in the preimplantation embryo (8
). However, the expression levels of the telomeric transgenes in these ES cell lines were still very low, similar to the levels expressed in clones A211 and A405. Combined, these results demonstrate the presence of a strong TPE in the mouse, which in combination with DNA methylation can lead to complete silencing of genes located near telomeres in both ES cells in culture and a variety of other cell types in vivo.
In some respects, our findings are consistent with studies of human tumor cell lines, which also utilized telomeric transgenes to demonstrate the presence of a substantial TPE (4
). However, we also found several significant differences between TPE in mice and that reported for human tumor cells. The TPE in human tumor cells was reversible by treatment with TSA, indicating that deacetylation of histones by type I and/or II histone deacetylases was involved. In contrast, we found that TSA did not produce a significant reversal of TPE in mouse ES cells. TSA did produce a slight increase in expression of the telomeric transgenes in clone A405. However, it is important that the transgenes in clone A405 are integrated into the promoter of the hnRNP A1 gene (unpublished observation), which is actively expressed in proliferating cells (7
) and therefore might influence the effect of TSA on TPE. Regardless, the increase in expression level of the transgenes after TSA treatment (2.2- and 1.5-fold for the neo
and HSV tk
genes, respectively) was still much less than the 10- to 50-fold increase in expression reported for the telomeric transgenes in a human tumor cell line (4
). This difference in response between mouse ES cells and human tumor cells to TSA may be an indication of different mechanisms for generating TPE in these cells, although it may also reflect differences in metabolism or uptake of TSA. One possible explanation is that the mouse homolog of yeast Sir2, Sir2α, is involved in TPE in mouse ES cells but not in human tumor cells. Like its counterpart in yeast, Sir2α is a class III histone deacetylase and is therefore not sensitive to TSA, and it yields the same array of products in vitro (1
). An inhibitor of Sir2, sirtinol, was found to have no effect on TPE in a human tumor cell line (31
). However, because the ability of sirtinol to inhibit mammalian Sir2 is not well characterized, the potential role of histone deacetylation by Sir2 in mammalian TPE remains to be investigated. Also, a likely scenario is the involvement of histone methyltransferases in mouse TPE. A recent report from María Blasco's laboratory demonstrates an epigenetic regulation of telomere length in the mouse through changes in the methylation status of histone H3 Lys9 and concomitant changes in the binding of certain chromobox proteins (19
Another striking difference between our results with mouse ES cells and studies of human tumor cell lines is in the role of DNA methylation in silencing of the telomeric transgenes. Although the expression levels of the telomeric transgenes in the mouse ES cells were very low even without DNA methylation, the complete silencing that occurred after growth without selection was accompanied by extensive DNA methylation. Similarly, complete silencing and extensive DNA methylation of the transgenes were also observed in vivo in various mouse tissues. The relationship between this complete silencing and DNA methylation was further demonstrated by the reversal of silencing upon treatment with 5-AzaC. In contrast, DNA methylation did not appear to play a role in TPE in the human tumor cell line studied by Koering and colleagues (31
). This lack of response may have been due to the relatively low level of 5-AzaC that was used in that study (1 μM versus the 3 μM concentration used in our study). Moreover, this difference in response to 5-AzaC might also reflect basic differences in the ability of mouse ES cells and human tumor cell lines to perform de novo methylation of DNA.
There are several possible explanations for the observed differences in TPE/silencing between mouse ES cells and human tumor cells. First, it is possible that there is a fundamental difference in the mechanisms of TPE/silencing in humans and mice. For example, the difference in the lengths of telomeres in inbred laboratory mice and humans (47
) could influence the mechanism of TPE/silencing. A more likely explanation, however, would be that the variations in the mechanism of TPE/silencing reflect fundamental differences between ES cells and somatic cells or between normal and tumor cells. It is clear from our data on MEFs and tissues obtained from A405 transgenic mice that mouse somatic cells are capable of maintaining the methylation patterns of the telomeric transgenes set at the embryonic stage. However, this is not an indication of de novo methylation capability in these cells. De novo methylation activity is present at high levels in mouse ES cells, as shown by the extensive DNA methylation of transgenes (32
). In contrast, somatic cells show limited de novo DNA methylation capabilities, which corresponds to the reduced expression of methyltransferase enzymes involved in de novo DNA methylation (12
). Moreover, a decrease in global DNA methylation is also observed in cancer cells (27
), which also have reduced levels of DNA methyltransferases involved in de novo methylation (13
). Thus, in view of the reduced de novo DNA methylation capability in somatic and cancer cells, it is not surprising that repression of the telomeric transgenes in human tumor cells was not followed by extensive DNA methylation (31
). Similarly, the difference in response between mouse ES cells and human tumor cells to TSA might reflect the fact that a homolog of yeast Sir2, Sir2α, is involved in TPE in ES cells but not in somatic and/or tumor cells.
Finally, an intriguing observation of our study is that telomeric transgenes are completely silenced and extensively methylated in vivo in somatic mouse tissue. Although DNA methylation and silencing of transgenes are common in transgenic mice, DNA methylation-associated inactivation of interstitial transgenes generally exhibits mosaicism, the frequency of which depends on the integration site and the nature of the integrated sequences (56
). The absence of mosaicism in the methylation and expression of the telomeric transgenes therefore shows that silencing at telomeres is exceptionally efficient. Consistent with this conclusion, the isolation of methylated DNA from human cells showed enrichment for subtelomeric DNA, demonstrating that the regions around telomeres are heavily methylated (11
). In contrast to the tissues in adult mice or MEFs derived from 13-day-old embryos, ES cells isolated from preimplantation embryos initially showed little or no silencing of the telomeric transgenes. In fact, prior to implantation, DNA in the early embryo is known to undergo extensive demethylation. Then, between implantation and gastrulation, the embryo undergoes global de novo methylation to reestablish the DNA methylation pattern, which will be maintained throughout the life of the somatic cells in the animal (25
). This expression of genes in the early embryo as a result of demethylation and their progressive silencing afterwards have been proposed to play an important role in development (25
). The observed expression of telomeric transgenes in ES cells isolated from preimplantation embryos is therefore consistent with the hypothesis that the regulation of expression of genes through TPE has a role in early development. The lack of methylation and complete silencing of the telomeric transgenes in human tumor cell lines may also mean that telomeric genes normally expressed early in development are also expressed in cancer cells, although, as mentioned above, this may also just reflect the lack of de novo methylation in human tumor cells. Understanding the molecular mechanisms responsible for TPE and silencing of telomeric genes may therefore provide new insights into both human development and cancer.