528 volume 46 | number 6 | june 2014 | nature genetics
favors loss of PARK2 so commonly in so many types of tumors? As cyclin D1–CDK4 and cyclin D1–CDK6 kinase complexes phosphorylate RB and thereby regulate the G1-phase restriction
point2, PARK2-defective tumors might be less dependent on growth factors for cell cycle pro-gression. In addition, cyclin E1 promotes DNA replication and its deregulation causes DNA
damage5,12; hence, its overabundance upon PARK2 loss might trigger replication stress and fuel tumor progression through genomic instability12,13 (Fig. 1). Furthermore, is there a biological explanation for the cancer-associated defects that anticorrelated with PARK2 altera-tions, such as loss of BRCA1? As tumors harbor an enhanced load of DNA breaks12,13 and hence depend on the BRCA1-mediated DNA repair pathway(s) that also require CDK2 activity14, this anticorrelation might reflect a common pathway. However, as both cyclin E1 overexpression and lack of BRCA1 cause chromosomal instability, it is plausible that concomitant deregulation of these factors would result in a suprathreshold extent of genomic instability, thereby selecting against such dual defects in tumors. And, finally, as chemical inhibitors of the cyclin D–dependent (CDK4 and CDK6) and cyclin E–dependent (CDK2) kinases are emerging as potential anti-cancer drugs15, the status of PARK2 might pro-vide a biomarker for the selection of patients who may benefit from such treatment.
COMPETING FINANCIAL INTERESTSThe authors declare no competing financial interests.
1. Morgan, D.O. The Cell Cycle: Principles of Control (New Science Press, London, New York, 2007).
2. Bartek, J. et al. Curr. Opin. Cell Biol. 8, 805–814 (1996).
6. Gong, Y. et al. Nat. Genet. 46, 588–594 (2014).7. Beroukhim, R. et al. Nature 463, 899–905 (2010).8. Veeriah, S. et al. Nat. Genet. 42, 77–82 (2010).9. Ciriello, G. Genome Res. 22, 398–406 (2012).10. Lücking, C.B. et al. N. Engl. J. Med. 342,
1560–1567 (2000).11. Sarraf, S.A. et al. Nature 496, 372–376 (2013).12. Bartkova, J. et al. Nature 434, 864–870 (2005).13. Halazonetis, T.D. et al. Science 319, 1352–1355 (2008).14. Falck, J. et al. EMBO Rep. 13, 561–568 (2012).15. Bruyère, C. & Meijer, L. Curr. Opin. Cell Biol. 25,
772–779 (2013).
Figure 1 PARK2 in cyclin regulation and cancer. Functional PARK2 forms the CRL complexes PCF4 and PCF7 that target cyclin D1 and cyclin E1, respectively, for ubiquitination and proteasome-mediated turnover (left). Loss-of-function mutations or deletions of PARK2 result in aberrant accumulation of the proto-oncogenic cyclin D1 and cyclin E1 and contribute to tumorigenesis (right). a, a crystallin; D1, cyclin D1; E1, cyclin E1; Ub, ubiquitin.
D1D1D1
D1
E1E1E1
E1
Wild type
Proteasome
Cyclin D1 andcyclin E1 turnover
Mutated or deleted
PARK2
PARK2
UbUb
UbUb
UbUb
αβ
UBC
PCF4
CUL1FBX4
PARK2 UBC
PCF7
CUL1FBXW7
PARK2
AberrantCDK2, CDK4 andCDK6 activities
Deregulatedcell cycle
Replicationstress
Genomicinstability
Normalcell cycle Cancer
mental deficiency. Considerable phenotypic overlap of symptoms in imprinted diseases has for a long time suggested that there might be cross-talk between imprinted loci. On page 551 of this issue, Nissim Benvenisty and colleagues demonstrate that a paternally expressed long noncoding RNA (lncRNA) known as IPW has a role in modulating
in the next generation. Several congenital syndromes have been linked to mutations and epimutations in imprinted genes. PWS (MIM 176270) is such an imprinted dis-order and occurs owing to loss of genes on the paternal chromosome 15q11-q13. It is characterized by delayed neuropsychomo-tor development, overeating, obesity and
Genomic imprinting is an epigenetic process that marks a subset of genes in the germ line for monoallelic parent-of-origin expression
Cross-talk between imprinted loci in Prader-willi syndromeAdele Murrell
Prader-willi syndrome (Pws) is caused by loss of paternally expressed genes at an imprinted locus on chromosome 15, including the long noncoding Rna IPW. a new study identifies a critical role for IPW in modulating the expression of maternally expressed genes in trans, which has important implications for the understanding of imprinted gene networks.
Adele Murrell is in the Department of Biology and Biochemistry, Centre for Regenerative Medicine, University of Bath, Bath, UK. e-mail: [email protected]
nature genetics | volume 46 | number 6 | june 2014 529
Figure 1 IPW links the paternally expressed PWS domain on chromosome 15 to the MEGs in the DLK1-DIO3 domain on chromosome 14. IPW recruits the histone lysine methyltransferase G9A to the imprinting control region (IG-DMR) in the DLK1-DIO3 locus. G9A (green circles) converts monomethylation of H3K9 (H3K9me1) to H3K9me3 (purple star), thereby modulating (dampening) the expression of the already active MEGs.
MaternalIG-DMR
Paternal
Chr. 14
Chr. 15
PWS domain
DLK1-DIO3DLK1 DIO3
UBE3A ATP10A
MEG3
FRAT3 MKRN3
MAGEL2 NDN
IC transcript
U-exons SNURF-SNRPN
snoRD116-HG snoRD115-HGIPW
G9A
RTL1 RBII-36-HG (BSR) MicroRNA-HG (MIRG)
the expression of maternally expressed genes (MEGs)1.
What makes this finding remarkable is that IPW, a lncRNA with an unknown function in the etiology of PWS, is located within the crit-ical PWS-associated region on chromosome 15, whereas the MEGs (MEG3, MIR370 and MIR409) whose expression it suppresses are located in the imprinted DLK1-DIO3 locus on chromosome 14 (Fig. 1). The authors made this discovery when they compared gene expression in induced pluripotent stem cells (iPSCs) derived from the fibroblasts of individuals with PWS and iPSCs derived from normal fibroblasts. PWS-iPSCs were derived from individuals who had various deletions in the PWS region and therefore only had maternal copies of the genes in this region. As expected, PWS-iPSCs did not express normally active paternal genes. These cells did, however, show higher expression of vir-tually all known MEGs and microRNAs from the DLK1-DIO3 locus. Upregulation of MEG expression at the DLK1-DIO3 locus did not involve loss of imprinting, and a sequence known as the IG-DMR that controls imprint-ing in this region showed normal levels of methylation in PWS-derived cells. Indeed, higher expression of MEGs occurred from the already active maternal alleles.
IPW fine-tunes dosageLoss of imprinted genes in the PWS locus thus leads to an effect in trans of increased expres-sion of imprinted genes in the DLK1-DIO3 locus, suggesting that a gene within the PWS locus normally reduces the expression of the genes in the DLK1-DIO3 locus. Benvenisty and colleagues went on to show that IPW serves this function. They showed that IPW recruits the lysine methyltransferase G9A to the IG-DMR,
resulting in methylation of lysine 9 on histone H3 (H3K9), and this regulation seems to take place on both alleles, as trimethylation of H3K9 (H3K9me3) is not found to be allele spe-cific in normal embryonic stem cells (ESCs). PWS-derived cells had low levels of H3K9me3 at the IG-DMR. In rescue experiments, ecto-pic expression of IPW in iPSCs derived from PWS fibroblasts restored H3K9me3 levels and reduced MEG expression. It is interesting that the role of IPW is not to switch off MEG expres-sion or to maintain silencing; instead, it seems that, by facilitating the methylation of H3K9, IPW modulates and reduces expression of the MEGs in the DLK1-DIO3 locus. This find-ing suggests that, at specific windows during development, lower levels of MEG expression are required and imprinting alone is not enough to reduce dosage. It is further curious that IPW originates from a host lncRNA for small nucleolar RNA (snoRNA) and that the MEGs it modulates in the DLK1-DIO3 region are also precursors for snoRNAs.
The imprinted lncRNA clubIPW is one of several lncRNAs associated with an imprinted locus. Many lncRNAs interact with chromatin-modifying proteins owing to their secondary structure and can recruit chromatin-modifying complexes to specific genomic regions. In this sense, lncRNAs behave as adaptors or intermediar-ies between the DNA sequence and chroma-tin modifiers. In mice, the lncRNA Kcnq1ot1 associates with G9a and the polycomb repressive complex 2 (PRC2) to regulate the expression of other genes in the Dlk1-Dio3 locus in cis2,3. The function of IPW is remi-niscent of those of an increasing number of lncRNAs (not necessarily imprinted) that recruit chromatin-modifying complexes
for regulation in trans, such as HOTAIR4,5 and H19 (ref. 6). H19 was one of the first imprinted noncoding RNAs to be identi-fied7 but has only recently been shown in mice to regulate other imprinted genes within an imprinted gene network in trans. It does this by recruiting Mbd1 lysine meth-yltransferase complexes to their regulatory elements, which results in an increase in H3K9me3 levels8.
Imprinted gene networksIn mice, the imprinted gene Plagl1 (also known as Zac1), which encodes a transcrip-tion factor, regulates the expression of several other imprinted genes9. Evidence that an imprinted gene network also exists in humans comes from recent work in human prostate cancer cell lines10. In this work, it was reported that aberrant expres-sion for several imprinted genes (PLAGL1 (ZAC1), MEG3, NDN, CDKN1C, IGF2, H19 and KCNQ1OT1) could be rescued by the ectopic expression of PLAGL1 (ref. 10). Although the DLK1-DIO3 locus is represented in the gene network, it seems that the PWS locus is not conspicuously featured. The PWS locus has a prominent role in brain function, and it is perhaps part of an imprinted network other than the imprinted network described thus far that has a role in fetal growth. Future work will be needed to discover whether there are additional imprinted networks that regu-late brain development and behavior and to determine what function lncRNAs have in these networks.
What next?Overall, it seems that there is still plenty to learn about genomic imprinting. Imprinted genes continue to represent a useful system in which to study epigenetic gene regulation. Similarly, there is still a lot to learn about gene regulation by lncRNAs. IPW does not seem to discriminate between methylated and unmethylated DNA at the IG-DMR, which is to be expected if the interaction of lncRNAs with DNA is homology based. But we still lack direct evidence for lncRNA-DNA hybridization, and further research focusing on the mechanisms by which lncRNAs inter-act with DNA is needed.
COMPETING FINANCIAL INTERESTSThe author declares no competing financial interests.
1. Stelzer, Y., Sagi, I., Yanuka, O., Eiges, R. & Benvenisty, N. Nat. Genet. 46, 551–557 (2014).
2. Pandey, R.R. et al. Mol. Cell 32, 232–246 (2008).3. Wagschal, A. et al. Mol. Cell. Biol. 28, 1104–1113
530 volume 46 | number 6 | june 2014 | nature genetics
features of easiRNA-generating miRNAs, evi-dent in the authors’ data set, are likely involved. RNA quality control (RQC) mediated by 5′–3′ and 3′–5′ exonucleases normally presents a robust barrier to the pervasive access of RDR6 to miRNA-cleaved fragments7; however, RQC operates poorly on RNA substrates lacking both a 5′ cap and a 3′ polyA tail. Such substrates can be generated by multiple, discrete miRNA cuts known to stimulate RDR6-dependent tasiRNA production6,7. Single-cut miRNAs can also trig-ger RDR6 activity if they are 22 nt instead of 20 or 21 nt in length (the cognate size range of most miRNAs) or if their precursors display specific structural features8,9 (Supplementary Fig. 1). Simultaneous multiple cuts by miRNAs are infre-quent, and only a fraction of single-cut miRNAs would possess the necessary attributes to attract RDR6 on matching, reactivated TE transcripts. Additionally, given that plant miRNA action is poorly tolerant to imperfect base-pairing5, easi-RNA production triggered by miRNAs is prob-ably more an exception than the rule, and this feature likely contributes to shaping the easiRNA landscape observed in ddm1 mutant plants.
Serendipity versus adaptive coevolutionExperiments conducted in ddm1 mutants draw an ideal picture of easiRNA production that poorly reflects the spatiotemporal confine-ment of stresses or developmental programs that would normally reactivate TEs in wild-type plants (Fig. 1a). For instance, a transient defi-cit in DDM1 naturally promotes easiRNA production in pollen vegetative nuclei, where only a fraction of the ∼50 predicted easiRNA- generating miRNAs are expressed1,10. Moreover, many prevalent easiRNA-generating miRNAs are highly conserved across plant species, with their basic roles in the acquisition or mainte-nance of cell identity confining their accumula-tion within single tissues or even single cells in a manner inconsistent with their mobility11,12 (Fig. 1a). Thus, a model for easiRNA production
via mechanisms resembling those of easiRNA biogenesis6. Notably, tasiRNAs enable control of a much wider range of related transcripts than individual miRNAs alone, a feature well-tailored to the predicted role of easiRNAs.
Creasey et al.4 thus hypothesized that miRNAs could similarly initiate RDR6 action on epigenetically reactivated TE transcripts, a situation artificially created in ddm1 mutant Arabidopsis. The global easiRNA levels found in ddm1 mutants were indeed lower in ddm1-dcl1 double mutants and, as expected, were nearly abolished in ddm1-rdr6 double mutants. Differential small RNA profiling in ddm1 versus ddm1-rdr6 mutant inflorescences identified many potential easiRNA-generating miRNAs, for which 3,662 possible targets were bioinformatically predicted among the 3,903 annotated Arabidopsis TE genes. This high initial estimate likely reflects the relaxed miRNA-pairing parameters used for target prediction and was indeed halved following parallel analysis of RNA ends (PARE) set to retrieve discrete miRNA cleavage products on a genome-wide scale. Further mapping of potential 21- and 22-nt easiRNAs, validated with a low-read threshold, finally showed that approximately one-third of all TE gene transcripts possibly undergo miRNA-mediated cleavage and simultaneously spawn easiRNAs in ddm1 mutant inflorescences.
Although likely an overestimation, this figure concurs with previous findings that only 15 of ∼320 TE subfamilies contribute substantially to easiRNA production in ddm1 mutants yet occupy, alone, ∼25% of the total TE space2. However, it is clear from the authors’ data that only some sequences within this subset, or even within a given TE subfamily, are associated with easiRNAs. TEs themselves are unlikely to contribute substantially to this apparent specificity in targeting, as transposons with dissimilar evolutionary states or proliferation mechanisms spawn easiRNAs. Instead, atypical
Intertwined epigenetic processes protect plant genomes from the deleterious effects of trans-posable elements (TEs). In Arabidopsis thaliana, some TEs are silenced via cytosine methyla-tion and chromatin compaction mediated by DNA METHYLTRANSFERASE-1 (MET1) and DECREASED DNA METHYLATION-1 (DDM1), but silencing can be reset develop-mentally or inactivated by stress. Therefore, backup mechanisms exist to contain epigeneti-cally reactivated TEs, notably via conversion of their transcripts into double-stranded RNA (dsRNA) by RNA-dependent RNA polymerase 6 (RDR6). The Dicer-like RNase III enzymes, DCL4 and DCL2, process the dsRNA into populations of 21- and 22-nt easiRNAs that, upon loading into AGO1 and AGO2, are thought to guide post-transcriptional gene silencing of many reactivated TEs1–3 (Supplementary Fig. 1). How RDR6 initi-ates easiRNA biogenesis specifically on TE RNAs has remained mysterious, but new work reported by Robert Martienssen and col-leagues in Nature 4 could provide part of the answer by implicating endogenous microRNAs (miRNAs) in this process.
Exception to the ruleMost plant miRNAs are produced by DCL1 from primary transcripts containing imper-fect self-complementary fold-back regions. They regulate mRNAs usually bearing a single miRNA-complementary target site by promoting their endonucleolytic cleavage5. Cleaved RNA fragments are normally degraded, but, under some circumstances, RDR6 uses them as templates to initiate the production of trans-acting siRNAs (tasiRNAs)
exploring new models of easiRna biogenesisAlexis Sarazin & Olivier Voinnet
although silent transposons in plants can be reactivated by stress or during development, their potential deleterious effects are prevented by transposon-derived epigenetically activated small interfering Rnas (easiRnas). a new study shows how serendipitous interactions between reactivated transposons and endogenous microRnas might initiate easiRna biogenesis, establishing an unexpected link between these two classes of silencing small Rnas.
Alexis Sarazin and Olivier Voinnet are in the Department of Biology at the Swiss Federal Institute of Technology Zürich, Zürich, Switzerland. e-mail: [email protected]
4. Guttman, M. et al. Nature 458, 223–227 (2009).5. Tsai, M.C. et al. Science 329, 689–693 (2010).6. Gabory, A. et al. Development 136, 3413–3421
