Singh et al. reply

Abstract

Replying to: H. F. Jørgensen, Z.-F. Chen, M. Merkenschlager & A. G. Fisher Nature 457, 10.1038/nature07783; N. J. Buckley, R. Johnson, Y.-M. Sun & L. W. Stanton Nature 457, 10.1038/nature07784 (2009)

In contrast to the comments made by Jørgensen et al.¹ and Buckley et al.², our experiments showed that REST maintains the self-renewal and pluripotency of mouse embryonic stem cells (mESCs)³. Two recent papers support our work: ref. 4 indicated that REST is indeed in the network that regulates ESC self-renewal and pluripotency and ref. 5 showed that mESCs with lower REST levels derived from a mouse model of Down’s syndrome have decreased levels of self-renewal markers and a higher propensity towards differentiation, even when cultured in the presence of LIF. We note that Buckley and Stanton also recently concluded that REST is part of the Oct4–Sox2–Nanog regulatory network and has “a key role in the maintenance of the ESC phenotype”⁶. We proposed that REST represses a set of microRNAs that potentially target self-renewal genes. At least one of them, miR-21, represses self-renewal, probably by destabilizing the messenger RNAs of Sox2 and/or Nanog (not Tbx3 or c-Myc as suggested by Buckley et al.²). In this model, changes in the cellular environment that counter this function of REST or stimulate the mRNA levels of Sox2 or Nanog could minimize the effect of REST. For this reason, in our study we used mESCs with a low passage number, and we cultured them without feeder cells to avoid possible contributions of the feeder cells or an adaptive response to high passage.

Main

One possible reason for the discrepancies between our study and the studies by Jørgensen et al.¹ is the use of feeder cells in most of their experiments (see Fig. 1d of Jørgensen et al.). Indeed, our additional experiments showed that culturing mESCs with feeder cells severely dampens the requirement of REST-mediated maintenance of mESC self renewal and pluripotency (S.K.S., manuscript in preparation). However, to rule out adaptive responses, Jørgensen et al.¹ also performed REST knockdown experiments in feeder-independent “wild-type” mESCs (46C and OS25) using either siRest or shRest (Fig. 1e of Jørgensen et al.¹). However, 46C and OS25 are not wild-type cells and were generated from the E14TG2a wild-type cells by genetic manipulation (46C has an insertion in Sox1 gene and OS25 has an insertion in Sox2 and Oct4 genes)⁷. It is unclear how these manipulations affected REST function and the use of these cells and not the wild-type cells probably explains their contradictory results (we used wild-type E14TG2a mESCs). In some experiments, Jørgensen et al.¹ used retinoic acid (see Fig. 1e of Jørgensen et al.). Additionally, we did not use retinoic acid in our experiments. Buckley et al.² mentioned that their data (mostly relative RT–PCR) are derived using feeder-free conditions. One potential reason for the discrepancy between our data and theirs could be high cell passage number or different cell type, among other experimental variations. For instance, their earlier report⁶ mentioned above, in which they did find a critical role of REST in mESC self-renewal, used E14 ESCs instead of the HM1 cells used here. The absence of relative difference in Mash1 and Neurog1 levels in wild-type versus Rest^+/- mESCs is not surprising because neuronal differentiation will cause a notable reduction in REST protein levels^8,9 (and consequent high Mash1 and Neurog1 expression) in both wild-type and Rest^+/- cells. Our REST-miR-21 ChIP and functional data are specific, reproducible and significant. Buckley et al.² cited their high-throughput ChIP-PET and ChIP-qPCR analysis using our published primers³ to indicate that REST does not bind to the miR-21 gene chromatin in mESCs. We published multiple locus-specific primer sets only for conventional ChIP but not ChIP-qPCR analysis (although these were available on request). The primers used in our ChIP-qPCR studies would provide good controls for binding³, because binding of REST is weaker on miR-21 than on miR-124 chromatin and requires higher substrate. REST-miR-21 functional assays will also help resolve the discrepancies.

Buckley et al.² cited reports that used cell types other than mESCs. Moreover, their cited references 8 and 9 measure neither REST protein levels nor the direct effect of REST on miR-21. In contrast to Buckley et al.’s cited ref. 7, we note that our ref. 10 found that miR-21 expression was higher in differentiated than in undifferentiated mESCs, supporting our conclusions. Both communications cited Loh et al.¹¹ to suggest that knockdown of Rest using siRest did not reduce self-renewal and did not induce loss of self-renewal markers in mESCs. This paper neither measured percentage self-renewal nor the level of the self-renewal regulators after siRest treatment and cannot be used to counter our conclusions. Both communications also cited Chen et al.¹² to indicate that Rest^+/- or Rest^-/- mutant mice show germline transmission/gastrulation. This situation is similar to many other self-renewal regulators, such as LIF, LIF-receptor β, gp130 (also known as Il6st), Stat3 and c-Myc¹³ and could be relevant during diapause¹⁴ (see Supplementary Information of our paper for details).