You'd Prefer An Argonaute

RNA-Guided RNA Cleavage by a CRISPR RNA-Cas Protein Complex

Caryn R. Hale, Peng Zhao, Sara Olson, Michael O. Duff, Brenton R. Graveley, Lance Wells, Rebecca M. Terns and Michael P. Terns

Cell 139 (5): 945-56, 25 November 2009.
doi:10.1016/j.cell.2009.07.040

This week’s ace summary and analysis by Robin Friedman:

The CRISPR (clustered regularly interspaced short palindromic repeats) system is a set of DNA sequences and associated genes involved in prokaryotic immune defense. Since it was discovered that CRISPR loci generate small (usually 25-60 nt) RNAs that often match phage or plasmid sequence, it has been tempting to make an obvious analogy with eukaryotic RNA interference, which can also be used to protect against viruses. However, what little was known about CRISPR mechanism pointed to a DNA-dependent mechanism of invader recognition and defense. One complicating factor is that there are at least nine distinct subtypes of the CRISPR system based on different sets of Cas (CRISPR-associated) genes. In this paper Hale et al. examine a subtype, Cmr, that had not previously been studied.

The authors first showed that in P. furiosis, there are two main species of CRISPR RNA product (termed psiRNAs), 39 and 45 nucleotides long. They purified protein complexes containing these psiRNAs and subjected them to mass spectrometry, finding seven members of the CRISPR-associated Cmr family. Sequencing of the psiRNAs revealed that each species had an 8-nucleotide “psi-tag”, consisting of the 3’ end of the constant repeat sequence, followed by unique guide sequence.

To test the mechanism of the psiRNA-Cmr complex, the authors used several synthetic constructs with sequence similarity to P. furiosis psiRNAs. The complementary RNA sequence was specifically cleaved at two spots, but the sense RNA sequence, unrelated RNA sequences, and a complementary DNA sequence were not cleaved. Truncations of these synthetic complementary RNAs showed that cleavage occurs in the same location, suggesting that the Cmr complex cleaves 14 nucleotides from the 3’ end of the psiRNA. Finally, the authors reconstituted the Cmr complex in vitro with recombinant proteins and synthetic psiRNA and recapitulated the cleavage behavior of the native complex. Only one of the six included Cmr proteins were dispensable for cleavage.

This is one of the simplest Cell papers I have seen. However, its few results are thoroughly proven. There are many questions left to answer about CRISPR function both in this model system and in others, but the scope of this paper is merely to show that the CRISPR system can function through RNA cleavage. This paper finally provides evidence strengthening the appealing analogy between CRISPRs and eukaryotic RNAi, which is sure to stimulate more interest in the system.

Distinct Argonaute-Mediated 22G-RNA Pathways Direct Genome Surveillance in the C. elegans Germline

Weifeng Gu, Masaki Shirayama, Darryl Conte, Jessica Vasale, Pedro J. Batista, Julie M. Claycomb, James J. Moresco, Elaine M. Youngman, Jennifer Keys, Matthew J. Stoltz, Chun-Chieh G. Chen, Daniel A. Chaves, Shenghua Duan, Kristin D. Kasschau, Noah Fahlgren, John R. Yates, Shohei Mitani, James C. Carrington and Craig C. Mello

Molecular Cell 36: 231-244, 1 October 2009.
doi:10.1016/j.molcel.2009.09.020

This week’s comprehensive summary/analysis by Michael Nodine:

Upon screening for mutants defective in RNAi, Craig Mello’s group found that the DICER-RELATED HELICASE-3 (DRH-3) gene was required for germline and soma RNAi. Gu et al. also found that DRH-3 was required for endogenous siRNA (esiRNA) production and/or stability.  When examining the requirement of DRH-3 for esiRNA production more closely, they found that DRH-3 was involved in the production of a specific class of esiRNAs, which they termed the 22G-RNAs due to their length and preference for a 5’ guanosine. 22G-RNAs were found to not have 5’-monophosphates, which are typically found in DICER products, and this observation led the authors to hypothesize that 22G-RNAs are RNA-dependent RNA polymerase (RdRP) products rather than DICER products. They then cloned small RNAs from wild-type and drh-3 samples using a 5’-independent ligation method, and found that 22G-RNAs mapped to ~50% of the protein coding genes annotated in the C. elegans genome.

Interestingly, 22G-RNAs tended to map to the 3’-ends of genes and there was less of a requirement for DRH-3 for 22G-RNAs derived from gene 3’-ends. Since the drh-3 mutants contained point mutations in the conserved helicase domain, this hinted at the possibility that DRH-3 may be part of an RdRP complex and may facilitate its movement along the RNA template by removing inhibitory secondary structures. Consistent with this idea, they found that two RdRPs were redundantly required for 22G-RNA production, and that these two RdRPs along with the tudor domain-containing protein EKL-1 interacted with DRH-3.

They then went on to find that worm-specific Argonautes (WAGOs) were redundantly required for 22G-RNA production. Genes, transposons, pseudogenes and cryptic loci were all found to be targets of 22G-RNAs, and components of the non-mediated decay (NMD) pathway were demonstrated to play a role in the biogenesis of at least a subset of 22G-RNAs. Gu et al. also demonstrated when and where 22G-RNAs function during worm development. WAGO-1 was localized to P-granules, which are localized just outside nuclear pores in the female germline and are thought to play a role in maternal RNA repression and storage. In addition, high-throughput sequencing and developmental northerns suggested that 22G-RNAs are enriched in the female germline and maternally inherited.

Thus, 22G-RNAs are key components of a surveillance pathway, which operates in the female germline and represses protein coding genes, pseudogenes and transposons. Presumably, incorrectly processed protein coding transcripts are targets for 22G-RNA biogenesis/action. However, it remains unknown how aberrant transcripts are recognized. Transcripts lacking poly(A) tails were previously demonstrated to be better substrates for C. elegans RdRPs in vitro, and incorrectly processed transcripts are better substrates for RdRP-dependent RNAi in plants. Fission yeast nucleotidyl transferases have been implicated in the recognition of aberrant transcripts by RdRPs and exosomes, and a homologous nucleotidly transferase, as well as a 3’-5’ exonuclease were found to be required for 22G-RNA production.  Based on these observations, the authors suggest that a nucleotidyl transferase and a 3’-5’ exonuclease, both of which were shown to be required for 22G-RNA production, may function in an exosome-like complex to recognize aberrant transcripts and/or recruit the 22G-RNA RdRP complex. Finally, 22G-RNA pathway components are subcellularly positioned just outside female germline nuclei. Based on their observations, the authors hypothesize that 22G-RNA components may ‘monitor’ the female germline transcriptome and thus function in the surveillance of maternally-inherited RNAs.

qiRNA is a new type of small interfering RNA induced by DNA damage

Heng-Chi Lee, Shwu-Shin Chang, Swati Choudhary, Antti P. Aalto, Mekhala Maiti, Dennis H. Bamford & Yi Liu

Nature 459: 274-277, 14 May 2009.
doi:10.1038/nature08041

This week’s aufschlussreiche summary and analysis by Anna Drinnenberg:

The term “quelling” refers to a posttranscriptional gene silencing phenomenon observed in Neurospora crassa, and was one of the first RNAi pathways to be described (Romano and Macino, 1992). Quelling is triggered by the expression of transgenes, also called “aberrant RNAs,” and results in silencing of both transgenes and cognate endogenous transcripts. It involves the production of double-stranded RNA (dsRNA) by an RNA-dependent RNA polymerase (QDE-1) using the transgenic mRNA as a template. Subsequently, one of two Dicer proteins (DCL-1 and DCL-2) cleaves the dsRNA substrate into small RNA duplexes that get loaded into an Argonaute effector complex containing QDE-2. After cleavage and exonucleolytic digestion of the passenger strand, the other siRNA strand functions as a guide strand in QDE-2 to degrade homologous endogenous transcripts. The physiological role of quelling is thought to be control of transposon expansion in order to preserve genomic integrity.

The authors of this study suggest a new physiological role for components of the quelling pathway in the response to DNA damage. During the process of studying the regulation of QDE-2 they noticed that the expression of QDE-2, at both the mRNA and protein level, is upregulated upon DNA damage caused by adding Histidine, EMS, or Hydroxyurea to the media. Immunoprecipitating QDE-2, they identified a new class of small RNAs ~21nt in length whose abundance is increased in the IP following DNA damage. Interestingly, these small RNAs appear to be shorter than the previously identified siRNAs (~25nt) of the quelling pathway that are produced by the same Dicer proteins (Catalanotto et al., 2004). It will be interesting to determine if the Dicer proteins, that are thought to act redundantly (Catalanotto et al., 2004), can produce small RNAs of different lengths or if the interaction with a cofactor could determine the cleavage interval on the dsRNA substrate.

Most of the small RNAs, which they referred to as “qiRNAs,” are derived from the sense and antisense strands of an rDNA array exceeding the regions that are transcribed into rRNA by Pol1, suggesting that a distinct transcript gives rise to the precursor (aberrant RNA) for the qiRNAs. They noticed that the production of aberrant RNA was not inhibited by thioelutin, a known inhibitor of RNA polymerases. In an attempt to identify the protein that produces the initial qiRNA precursor transcript from the rDNA array, they observed that QDE-1, already known to have RdRP catalytic activity, can also synthesize RNA transcripts using a DNA template. This is a very interesting observation and raises the question if RdRPs in other organisms also have DNA-dependent RNA polymerase activity. Such an activity would make the production of precursor RNAs for small RNAs independent from the canonical transcription pathway for the majority of other cellular RNAs.

In trying to assign a role to the qiRNAs, the authors noticed that the decrease in protein production upon DNA damage is partially blocked in QDE-1 and QDE-3 mutant strains. Moreover, QDE-1 and DCL-1/DCL-2 mutants show increased sensitivity to DNA damage reagents. These observations certainly provide a first hint of function of this pathway. A more detailed follow-up experiment could be a more precise demonstration that the qiRNAs in complex with QDE-2 directly downregulate rRNA transcripts (but this experiment was beyond the scope of this study).

Another recent publication from Cerere and Cogoni (Cecere and Cogoni, 2009) suggests that the small RNAs are involved in copy number control of the rDNA locus, possibly preventing recombination within the array. Changes in the heterochromatic state of the rDNA array could unify both observations: the block in downregulation of the transcripts and the increased recombination rate. Moreover, high-throughput sequencing of the small RNAs as well as RNAseq analysis in Neurospora crassa upon DNA damage might also identify other qiRNA sources and their potential targets.

References:
Catalanotto, C. et al. (2004). Redundancy of the two dicer genes in transgene-induced posttranscriptional gene silencing in Neurospora crassa. Mol Cell Biol 24, 2536-2545.

Cecere, G., and Cogoni, C. (2009). Quelling targets the rDNA locus and functions in rDNA copy number control. BMC Microbiol 9, 44.

Romano, N., and Macino, G. (1992). Quelling: transient inactivation of gene expression in Neurospora crassa by transformation with homologous sequences. Mol Microbiol 6, 3343-3353.

An RNA-dependent RNA polymerase formed by TERT and the RMRP RNA

Yoshiko Maida, Mami Yasukawa, Miho Furuuchi, Timo Lassmann, Richard Possemato, Naoko Okamoto, Vivi Kasim, Yoshihide Hayashizaki, William C. Hahn & Kenkichi Masutomi

Nature AOP, 23 August 2009.
doi:10.1038/nature08283

This week’s summary and sound analysis by Jenny Rood:

Summary:  The Masutomi lab, in collaboration with others, present evidence in this article for a novel functionality of the human telomerase reverse transcriptase catalytic subunit, hTERT, when it is bound to the RNA RMRP.  TERT has previously been shown to form a complex with the TERC RNA, providing the template for telomere elongation. This paper demonstrates the interaction of TERT and RMRP in vitro and in vivo and then argues that this complex serves as a RNA-dependent RNA polymerase whose eventual function is to synthesize double-stranded RMRP RNA for the siRNA pathway.

The authors first identify the TERT-RMRP interaction through immunoprecipitation of overexpressed tagged hTERT and then isolation and sequencing of the associated RNAs.  RMRP sequences appeared roughly as frequently as TERC sequences, together making up 5% of hits; a total of 38 RNA sequences were found to be associated with TERT in this assay.  This interaction is then confirmed by RT-PCR and northern blot.  Purifications with a variety of overexpressed TERT fragments show that the RNA interaction occurs in the N-terminal half of the protein, where TERC is also known to be bound, yet the TERT-RMRP complex fails to elongate telomeres in the PCR-based TRAP assay.

Under high salt conditions (approximately double physiological salt) in vitro, the authors are able to isolate an RNA species that is twice the length of the RMRP RNA that reacts with both sense and antisense probes to RMRP.  Moreover, truncation of the 3’ end of RMRP eliminates this product, supporting a 3’ back-priming mechanism and implicating TERT-RMRP as a RNA-dependent RNA polymerase (RdRP).  Cell lines lacking TERT (VA-13) also fail to form the double-stranded product, but overexpression of TERT in these lines rescues this phenotype.

Overexpression of RMRP, on the other hand, in cell lines containing TERT leads to a reduction in RMRP signal.  Short RNAs complementary to RMRP can also be isolated from these cell lines, implicating an siRNA mechanism.  Further evidence for the generation of RMRP siRNA through the RdRP activity of TERT-RMRP is provided by association of the short RMRP species with Argonaute 2 and rescue of the RMRP overexpression phenotype in Dicer knockdowns.

Comments and future directions:  The paper clearly demonstrates an interaction between TERT and RMRP RNA, and progresses towards a possible model for the function of this RdRP, but many questions remain.  The experiments showing the generation of siRNAs and their effects in vivo seem preliminary and rushed.  During discussion of this paper in journal club, it was mentioned that figure 4e might have been improved by including a negative control: does the VA-13 cell line, which lacks TERT, also lack RMRP small RNAs?

The paper mentions that both TERC and RMRP bind in the same region, but this region is defined very broadly as the N-terminal half of TERT.  It would be very interesting to have more structural evidence about the interaction of TERT with these two RNAs.  Are the same residues responsible for the binding of both RNAs?  Are these two RNA binding events mutually exclusive?  If so, what implications does this have for the function of TERT?

A known set of mutations in RMRP is responsible for the human disorder of cartilage-hair hypoplasia.  It would be interesting to determine if these mutations inactivate the TERT-RMRP RdRP activity, and how in turn this causes a pleiotropic disease.

Finally, on a broader note, the authors discuss that a back-priming mechanism inherently limits the number of possible products (verified by incubating TERT-RMRP with total RNA).  It would be useful to determine which products are made, and consequently, if these are also processed into siRNA, or what function they serve.  Similarly, it would be interesting to see if any other RNAs from the initial TERT screen for bound RNAs had similar functions to the TERT-RMRP complex.

In summary, this paper suggests a tantalizing new function for TERT in complex with a different RNA besides the canonical TERC that will likely provide many insights in the future.

Co-translational mRNA decay in Saccharomyces cerevisiae

Wenqian Hu, Thomas J. Sweet, Sangpen Chamnongpol, Kristian E. Baker & Jeff Coller

Nature 461 (7261): 225-229, 10 September 2009.
Nature AOP, 23 August 2009.
doi:10.1038/nature08265

This week’s exacting summary and analysis by David Weinberg:

In their recent Nature article, Jeff Coller and colleagues demonstrate that mRNA in decay in the budding yeast Saccharomyces cerevisiae can occur while the mRNA is still engaged with actively translating ribosomes.  Prior to this paper, the dogma in the field had been that ribosome dissociation was a necessary step before decapping.  The evidence for this exact model was a bit lacking: it was clear that translation initiation and decapping are competing processes since both require access to the cap structure, but there didn’t seem to be any indication that decapping would have to interfere with elongating ribosomes.  Indeed, co-translational mRNA decay had been previously hypothesized but no lab had demonstrated it.  Using the tools of budding yeast and some clever molecular biology, Jeff Coller’s lab is able to do just that.

The basic outline of the article is a series of very similar experiments showing that decaying mRNAs (i.e., deadenylated, decapped, and/or partially degraded) are associated with translating ribosomes using different combinations of knock-out and wild-type strains, and artificial and endogenous mRNAs.  The initial indication that the existing model for mRNA decay might be incorrect is that deadenylated mRNAs that accumulate in a decapping-defective strain remain on polyribosomes.  The same is true of decapped mRNAs that accumulate in an XRN1 knock-out strain.  The authors are careful to include enough control experiments to show that the polyribosome association suggested by sucrose gradients is actually due to bound ribosomes and not another macromolecular complex that might similarly alter the sedimentation properties of an mRNA.

Next, using both transcriptional turn-on and shut-off experiments the authors claim to show that decapping can occur when mRNAs are associated with translating ribosomes.  This point – that the associated ribosomes are actively translating when decapping occurs – is important to distinguish from ribosome-reloading following mRNA decapping.  The turn-on experiment is a bit bogus as their interpretation is based on the observation that at the first timepoint when a decapped mRNA accumulates, that decapped mRNA is on polyribosomes.  Given that they had 2 timepoints (20 and 60min), this conclusion seems like a stretch and could have been left out of the paper without any detriment.  However, the transcriptional shut-off experiment is convincing: adding cycloheximide at the same time that transcription is shut-off prevents the ribosome run-off on decapped mRNAs that is seen in the absence of cycloheximide.  Thus, the ribosomes observed to be associated with decapped mRNAs were in the act of translation.

Up to this point, all experiments had been performed in mRNA decay mutants in order to observe relatively rare decay intermediates.  Using an artificial mRNA in which rare codons are used to stall translating ribosomes, the authors demonstrate that their conclusions also hold in wild-type cells.  Of course, such an artificial mRNA can potentially give results that aren’t true of endogenous mRNAs that are efficiently translated.  The authors therefore conclude their paper with the ultimate demonstration of co-translational mRNA decay for two different (highly expressed) endogenous mRNAs in wild-type cells.

From their work, the authors conclude that a ribosome-free state is not required for mRNA decay and, in fact, mRNA decay can occur co-translationally.  This ability to initiate decay of mRNAs that are still being translated seems to provide a more rapid means of mRNA decay: rather than waiting for translation to finish or ribosomes to be actively removed, Xrn1 can begin to degrade the mRNA from the 5′ end immediately after decapping.  The paper also makes an evolutionary argument that such a decay mechanism would have the benefit that it would not interfere with residual translating ribosomes and, therefore, would prevent the production of truncated polypeptides.

While the science in this paper is (for the most part) convincing, its presentation is a bit frustrating.  At the end of the paper, the experiments that the reader should really care about – those performed with endogenous mRNAs in wild-type cells – are FINALLY shown.  But to get there the reader has to get through the artifact-prone experiments done in other settings.  Of course, these artifact-prone experiments lend further support to the final model.  However, I would have rather seen some space in the paper devoted to further characterization of the decay mechanism.  For example:
– Do the same principles hold for all mRNAs?  In particular, is this true of histone mRNAs that lack poly-A tails? Lower expressed mRNAs that are more difficult to detect?
–  Is Xrn1 decay of ribosome-associated mRNAs distributive or processive?  Are there cycles of degradation/dissociation as ribosomes finish elongating, or does Xrn1 remained engaged throughout?
– How does mRNA half-life relate to ribosome occupancy?  Is slow translation of an mRNA associated with slow decay?
I hope that we can find some answers to these questions in future work from the Coller lab.

Architecture and secondary structure of an entire HIV-1 RNA genome

Joseph M. Watts, Kristen K. Dang, Robert J. Gorelick, Christopher W. Leonard, Julian W. Bess Jr, Ronald Swanstrom, Christina L. Burch & Kevin M. Weeks

Nature 460 (7256): 711-716, August 2009.
doi:10.1038/nature08237

This week’s snappy summary and analysis by Vikram Agarwal:

In this article, Watts and colleagues demonstrate a method to systematically predict the RNA secondary structure of an HIV-1 viral genome.  Their technique, called SHAPE (2′-hydroxyl acylation analysed by primer extension), exploits the principle that nucleotides in a flexible conformation can react with an electrophile that subsequently blocks a primer extension reaction.  In this way, one can visualize and quantify the tendency of regions of RNA to participate in base pairing or remain unstructured, all at single-nucleotide resolution.  The technique was recently shown to accurately predict nearly 95-100% of bases in rRNA, tRNA, and several coding RNAs (Deigan et al., PNAS 2009).  The SHAPE reactivity of each base position is converted linearly into a pseudo-free energy term, which is incorporated into RNAstructure to fold the entire RNA.

Here the foldings reconstruct known motifs in the HIV genome, such as the gag-pol frameshift element, which are shown here to actually be constituents of more extensive motifs.  Most notably, the technique allows the group to predict an accurate structure for all coding regions, which have been formidable to characterize using traditional approaches.  The paper finds significant correlations between RNA secondary structure and protein secondary structure, with inter-protein linkers and protein-domain junctions often corresponding to highly structured RNA regions.  This result implicates such highly structured regions in modulating and offering time for protein folding during translation.  To investigate this possibility, the authors perform a ribosomal toeprinting experiment on two HIV-1 open-reading frames to test the hypothesis that local RNA flexibility influences the pausing of a ribosome as it scans the RNA message (Supplementary Figure 5).

Overall, this paper contributes a notable advance in the accurate characterization of RNA structures on a large scale.  It opens the door for parallelization of SHAPE analysis to characterize even larger RNA genomes at high-resolution, which would open a wealth of knowledge about how RNA motifs and other signals are interpreted by the cell.  Moreover, it suggests a link between RNA structure and protein folding.  However, this link requires a more thorough and direct investigation in the future.  The authors establish only a correlative relationship between the RNA and protein structure, and have yet to dissect the underlying causality of the process.  This may ultimately merit mutational experiments that modify RNA secondary structure to examine if regions within a protein are differentially folded via the fine-tuning of ribosomal processivity by the base pairing interactions of RNA.

Ars2 Links the Nuclear Cap-Binding Complex to RNA Interference and Cell Proliferation

Joshua J. Gruber, D. Steven Zatechka, Leah R. Sabin, Jeongsik Yong, Julian J. Lum, Mei Kong, Wei-Xing Zong, Zhenxi Zhang, Chi-Kong Lau, Jason Rawlings, Sara Cherry, James N. Ihle, Gideon Dreyfuss and Craig B. Thompson

Cell 138 (2): 328-339, July 24, 2009.
doi:10.1016/j.cell.2009.04.046

This week’s long summary and analysis by David Garcia:

Synopsis:
The authors demonstrate a clear role for the mammalian Ars2 protein in cell proliferation, as well as an interaction with the Cap-Binding Complex. Depletion of Ars2 reduces the miRNA or siRNA directed repression of two reporter constructs, and levels of mature miRNAs for two out of four miRNAs investigated. Ars2 also co-precipitates with Drosha, but not Dicer, and pre- or mature let-7 can rescue some or all of the loss of repression of the reporters associated with Ars2 depletion. The authors are less successful at directly connecting Ars2’s effect on RNAi that they observed to a potential role in primary-miRNA to pre-miRNA processing by Drosha. The processing assays should have tested more substrates, and their variability is not explained clearly enough.

Detail:
This paper aims to demonstrate a critical role for the mammalian protein Ars2 in cell proliferation, and associate this role with other evidence that it also affects the stability or processing efficiency of a couple of primary miRNA transcripts. While initially setting out to study how mammalian Ars2 imparts a resistance to arsenic oxide treatment, the authors discovered previous studies had focused on a partial clone of the protein. In this study, the full-length protein affects arsenic treatment in a manner opposite to what had been demonstrated before; its reduction has a profound influence on cell proliferation; and it contains a number of domains common to RNA binding proteins, and homology to the Arabidopsis protein SERRATE known to affect the processing of primary miRNA transcripts by Dicer-Like 1 (the plant Drosha).

The paper starts by showing that when Ars2 is knocked down using shRNAs, 3T3 cells die more slowly compared to controls when treated with arsenic oxide (Fig1B). Also observed when depleting Ars2 is a cell proliferation defect, using colony formation assays and counting population doublings (Fig1D,F). To investigate further, the authors generate a floxed allele of Ars2 in mESCs, and then generate mice from which they derive immortalized MEFs that they can infect with Cre expressing retrovirus. These Ars2 depleted cells exhibit the same proliferation defects as had been seen with the cell lines (Fig2B). Examination of various tissues from the Ars2 depleted adult mice show that in tissues that have relatively high cell proliferation and normally high expression levels of Ars2, like hematopoietic tissues, there is decreased cellularity or increased apoptosis (Fig2D). In contrast, there are no such differences observed between the transgenic and WT mice in other lower proliferating tissues.

To find out what other proteins Ars2 interacts with, a flagged-Ars2 is expressed, and co-precipitates with several components of the cap-binding complex (CBC), including CBP80, CBP20, and importin-alpha (Fig3A). Ars2 and the CBC interact with 7-Methyl-Guanosine capped RNAs (Fig4A), as would be expected for a component of the CBC. In addition, Ars2 shuttles between the nucleus and cytoplasm (Fig4B), like other CBC components.

Given SERRATE’s known role in miRNA biogenesis in Arabidopsis, the authors address whether the mammalian Ars2 in involved in RNAi. Two types of luciferase reporters are created: one with 3X let-7 (7mer) miRNA binding sites derived from the C. elegans lin-28 3 prime UTR; and another with a perfectly matched site for the let-7 miRNA (~22mer match). In HeLa cells, they first transfect siRNAs to knock down Ars2 or CBP80 or Ago2 (positive control), and then transfect the reporters to see how repression compares to cells treated with a control siRNA. The siRNAs against Ars2 and CBP80 inhibit let-7 directed repression at levels comparable to knockdown of Ago2 for both reporters, with a more pronounced effect on the miRNA sites reporter (Fig5A).

I am surprised knocking down Ago2 only affected repression ~2.5 fold for the perfect site reporter, and not much greater. Let-7 levels are very high in HeLa cells. However, there could be some complications of knocking down Ago2 with an siRNA. Unfortunately, while the authors state a reporter with a mutated let-7 seed site led to a loss of repression under all conditions, they do not show how this loss of repression compares to the loss they observed from Ars2, CBP80, and Ago2 siRNAs. This would have been a useful control to show.

Next the authors show that addition of excess let-7 duplex can rescue loss of Ars2, but not Ago2, for both reporters (Fig5B), suggesting Ars2 may affect some step in miRNA biogenesis before Dicer. Depletion of Ars2 or DGCR8 with two different siRNAs each led to a decrease in mature let-7 levels (~50% reduction)(Fig5C,D), and the same story for mature miR-21 (expressed very highly in HeLa). While they couldn’t detect pre-let-7 by Northern (I’m surprised by this), they did see a reduction in pre-miR-21 in Ars2 depleted HeLa cells. The reduction in pre or mature was not observed for all miRNAs they probed: there was no change for miR-30a or miR-16 (FigS3).

To investigate which step of miRNA processing is affected by Ars2 depletion, they immunoprecipitate a tagged Drosha and Dicer from 293T cells. Ars2 and CBP80 both co-precipitate with Drosha but not Dicer, and this interaction is not RNA dependent (Fig6A). Addition of excess pre-let-7 rescued some of the loss of repression from Ars2 or CBP80 depleted cells (but not Ago2 depleted)(Fig6C). Depletion of Ars2 also led to a decrease in pri-miR-21 levels by qRT-PCR, indicating that Ars2 may also influence the stability of Drosha substrates (Fig6D).

A pri-miRNA processing assay is employed to address the role of Ars2 in affecting the fidelity of Drosha mediated processing. In these assays, they compare the processing of an in vitro transcribed pri-miR-155 between different cell extracts. The authors state that there is an observed reduction in pri to pre processing in the MEF Ars2 KO cells compared to WT (Fig6F). However, it may be problematic to compare two different cell extracts that are almost guaranteed to differ in more than just Ars2 expression. In the Ars2 KO, the “correct” pre band mostly disappears, but a slightly smaller band also present in the WT intensifies, without explanation. They only show data for a single pri-miRNA, 155 (but say comparable results were obtained for miR-21). They quantitate the reduction in “correct” processing as 3-fold for miR-155 (Fig6G).

The paper ends by weakly connecting the observed effect of Ars2 on cell proliferation with its potential effect on miRNA processing. The authors first re-visit the proliferation point by showing that serum starved MEFs that exit the cell cycle lose Ars2 expression (Fig7A). They confirm Ars2’s proliferation dynamic expression in the hematopoietic cell line Bax-/- Bax-/- (Fig7C). A primary miRNA processing assay, comparing processing of pri-miR-155 (in vitro transcribed as before) between 10% serum grown or 0.1% serum starved MEF extracts, is claimed to show a decrease in pri to pre in the Ars2 depleted serum starved extracts (Fig 7D). While there seems to be an increase in heterogeneity in pre, the total reduction in pre in the serum starved cells is not robust. As before, the implication that we are to assume that variability between different cell extracts, from different cell lines, or grown under vastly different conditions, cannot contribute to variability in pri to pre processing is, I think, not well founded. Moreover, the authors only show results from the testing of one miRNA, miR-155. There is insufficient explanation for bands nearby those they deem as “correctly processed.” While the authors have shown clearly how Ars2 affects cell proliferation, interacts with the CBC, and interacts with Drosha, I don’t believe they have clearly explained how Ars2 affects RNAi with their proposed mechanism of affecting miRNA biogenesis. Hopefully there will be stronger data to support this hypothesis in future papers.

Cellular MicroRNA and P Bodies Modulate Host-HIV-1 Interactions

Robin Nathans, Chia-ying Chu, Anna Kristina Serquina, Chih-Chung Lu, Hong Cao, and Tariq M. Rana

Molecular Cell 34 (6): 696–709, June 2009.
doi:10.1016/j.molcel.2009.06.003

This week’s summary and scrupulous analysis by Anonymous:

The authors showed that a host microRNA (miRNA), mir-29a, targets the 3’ untranslated region (3’ UTR) of HIV-1 directly and negatively regulates viral expression. mir-29a has two other family members (with the same seed sequence) that are also able to repress viral expression. In addition, the paper suggested that this negative regulation contributes to the latency of HIV-1 via the following model: Upon integration into the host genome and transcription, viral mRNAs are transported out into the cytoplasm where translation of viral proteins and virus assembly usually takes place. If however, the HIV-1 3’ UTR is targeted by mir-29a, the viral mRNA is sequestered into P bodies, translation is suppressed and virus assembly is prevented. Activation could occur when viral mRNAs are released from P bodies after certain stimuli. If this model is true, then mir-29a-mediated regulation would be acting as a checkpoint from viral latency to activation.

While the authors convincingly demonstrated that mir-29a negatively regulates viral expression via a direct interaction with the HIV-1 3’ UTR, the evidence presented for P body sequestration is not as strong. The viral RNA is certainly sequestered into cellular foci but three of the four markers used to identify P bodies in the paper can also be found in stress granules. The authors made a point that the negative regulation is due to translational suppression and not mRNA destabilization but not much was done to probe that the latter was not taking place.

Still, the authors did a nice job tying the relevance of this miRNA-mediated regulation to viral latency. Aligning the 3’ UTRs of various HIV-1 subtypes, they found that the mir-29a target site in the 3’ UTR is highly conserved. This is true for most subtypes except the O group HIV-1 RNAs, in which the non-conserved nucleotides in the seed match region would abolish interaction with mir-29a. The authors highlight the fact that O group HIV-1 is endemic to certain parts of Africa and is typically 100-fold less infectious than the widely-circulating M group HIV-1, whose viral subtypes maintain the conserved nucleotides in the mir-29a seed match site. Hence it is plausible that mir-29a interaction and HIV-1 infection capability could be linked. If this is true, it is remarkable that the relatively recently-evolved HIV-1 is able to co-opt the more ancient mir-29a (conserved across humans, mice, rats, dogs and chickens) to modulate its own life cycle.

It is worth mentioning that while HIV-1 produces more than 30 mRNAs, almost all of them share the same 3’ UTR. Hence even a single miRNA-mediated downregulation would prevent other viral proteins, such as the viral transcription activator Tat, from being made. This, in turn, would reinforce latency. Having said that, mir-29a is not the first miRNA identified to target the 3’ UTR of HIV-1. A previous paper (ref.) found that several miRNAs (though mir-29a was not one of them) target the HIV-1 3’ UTR and contributes to latency in resting primary CD4+ T lymphocytes, which are the cells usually infected by HIV-1. In the subtype alignment described above, the 3’ UTR is well-conserved even beyond the mir-29a seed match site. It would be interesting to see if the seed matches of these miRNAs also coincide with these other highly-conserved segments.

Reference:
Cellular microRNAs contribute to HIV-1 latency in resting primary CD4+ T lymphocytes
Huang et al., (2007) Nature Medicine 13: 1241-1247.

Therapeutic microRNA Delivery Suppresses Tumorigenesis in a Murine Liver Cancer Model

Janaiah Kota, Raghu R. Chivukula, Kathryn A. O’Donnell, Erik A. Wentzel, Chrystal L. Montgomery, Hun-Way Hwang, Tsung-Cheng Chang, Perumal Vivekanandan, Michael Torbenson, K. Reed Clark, Jerry R. Mendell and Joshua T. Mendell

Cell 137 (6): 1005-1017, June 2009.
doi:10.1016/j.cell.2009.04.021

This week’s exemplary summary and analysis by Anonymous:

A recent report by Kota and colleagues describes the systemic delivery of a tumor suppressive miRNA, miR-26a, to a murine model of hepatocellular carcinoma (HCC). In this study, it was shown that miR-26a, a miRNA previously shown to be down-regulated by the c-Myc oncogene, functions as a tumor suppressor via induction of a G1 arrest in a human HCC cell line. This G1 arrest was ascribed to repression of miR-26 targets Cyclin D2 and Cyclin E2. Furthermore, they described an adeno-associated viral vector that would express miR-26a in combination with eGFP. This virus was then delivered to mice over-expressing c-Myc in the liver, which has previously been shown to induce liver tumors. Compared to tumor-bearing mice infected with a virus expressing only eGFP, tumor-bearing mice infected with the virus expressing both eGFP and miR-26a had a substantial reduction in tumor burden without gross toxicity to the liver and other organs. This provides the first example of systemic miRNA delivery suppressing an established tumor in vivo.

While this observation has important therapeutic implications, there are several limitations to the study. First, while the in vitro data suggests that miR-26a acts as a tumor suppressor via cell cycle arrest, the in vivo data for miR-26a causing cell cycle exit is unconvincing. In particular, the use of Ki67 intensity as a measure of cell cycle status is not legitimate, as Ki67 is either present in cycling cells or absent in non-cycling cells. Thus, it appears that the more likely mechanism of tumor suppression in vivo is apoptosis, due to the robust increase in TUNEL staining. However, it is not clear which targets of miR-26a contribute to increased apoptosis. Moreover, while this study proves that miR-26a can suppress early-stage c-Myc liver tumors, it is not clear whether miR-26a can suppress late-stage, invasive HCC. Since most patients with HCC show up with significantly advanced disease, examination of miR-26a delivery in more aggressive lesions would be an important next step.

Transfection of small RNAs globally perturbs gene regulation by endogenous microRNAs

Aly A Khan, Doron Betel, Martin L Miller, Chris Sander, Christina S Leslie & Debora S Marks

Nature Biotechnology 27 (6): 549-55, June 2009.
doi:10.1038/nbt.1543

This week’s cerebral analysis by Graeme Doran:

Investigating the application of small RNAs to destabilize specific mRNA targets, researchers have observed a variety of non-specific ‘off-target’ effects – alterations in the expression level of mRNAs that do not contain a perfectly complementary target site for the small interfering RNA transfected into cells. In a minority of cases, the target mRNA is actually stabilized!

Briefly, ‘off-target’ effects have been broadly understood to derive from 4 main sources:

a) Partial complementarity between the small interfering RNA and mRNAs in the cell.

b) Stress responses due to transfection of foreign RNA.

c) Downstream secondary gene expression changes due to silencing of a specific target.

d) Disruption of the endogenous miRNA regulatory network due to competition between the cellular miRNA machinery and exogenously supplied small interfering or short hairpin RNAs.

To date, conflicting experiments have suggested that some si/shRNAs may inhibit endogenous miRNA activity in some scenarios, particularly when the silencing RNAs are present at high levels. This may occur through saturation of either or both the RISC and DICER activities in the cell, depending on the type of small RNA used.

This study from Khan et al. attempts a broad survey of previously published small RNA transfection experiments.

The key evidence presented are:

1) Transfection of small RNAs into immortalised cell lines produces a consistent de-repression of genes that contain an endogenous miRNA target site within their 3’UTR.

2) The extent of de-repression is quantitative – that is it depends both upon the extent of endogenous miRNA repression on a specific UTR, and the concentration of siRNA used in the transfection.

3) Cellular ARGONAUTE/RISC activity is likely subject to loading competition between small RNAs within cells.

Compiling an extensive set of data leads to some brevity of description in the methods. The authors use 4-species conservation as a criterion for miRNA site prediction. One would expect that the observed de-repression effect would be independent of the seed site conservation, as it is commonly possible to see miRNA repression on non-conserved seed sites, and the authors do not make clear whether this signal was present for the non-conserved site data or not.  Further, the normalization of predicted target site sets is not well described. One might envisage that the ‘baseline’ gene sets with no endogenous or exogenous seed sites would be shorter on average than gene sets with 1 or more conserved sites. Shorter UTRs have less scope for regulation by miRNAs or other
RNA binding factors, and so maybe would have less regulation to perturb.

The core data (Figure 2) indicates a consistent and significant de-repression of UTRs that contain conserved (higher confidence) endogenous miRNA target sites, but the data is correlative rather than conclusive – and would have been greatly enhanced by a simple luciferase assay to precisely determine the contribution of individual sequences to repression/de-repression in the context described. As is, the model and the data fit each other, but other modes of de-repression are not discounted experimentally. Data suggesting that the de-repression is dose-dependent seem to rely on relatively small sets of genes selected specifically for responsiveness to siRNA treatment and thus this aspect of the story is less convincing. Furthermore, there is little proof that concentrations required by ‘good’ siRNAs to silence genes (1-10nM) have a significant de-repressive effect on miRNA targets, and so the importance of the observed effect in therapeutic situations where delivery concentrations are low is still an open question.

The core findings of this paper, and the methodology employed, open up some interesting questions about the basic biology of RISC competition in cells. How, for instance, do rapidly induced miRNAs (such as miR-21 upon SMAD pathway activation) interact with the endogenous pool of miRISC? Is there present a surplus of free ARGONAUTE for such instances, or is ARGONAUTE protein concentration limiting in the cell? Do endogenous siRNAs – generated on cell stress or viral infection – displace miRNAs from their silencing role, and what is the half-life of endo-siRNA loaded RISC in vivo? Tumors commonly downregulate miRNA activity, and DICER depletion enhances tumorigenesis.  Can this also be achieved by limiting the available miRNA RISC in tumor cells?

Broadly speaking this was a stimulating paper, and whilst I don’t think that it breaks new ground in considering ‘off-target’ effects of small RNAs within cells, it provides some of the most convincing evidence of miRNA target de-repression to date. And beyond this, it provides the thought provoking concept of RISC competition, new questions upon which to focus, and a validated method for analyzing miRNA target de-repression.

Argonaute HITS-CLIP decodes microRNA–mRNA interaction maps

Sung Wook Chi, Julie B. Zang, Aldo Mele & Robert B. Darnell

Nature, Advance Online Publication, June 17 2009.
Nature 460 (7254): 479-486, July 23 2009.
doi:10.1038/nature08170

This week’s deep summary and analysis by Noah Spies:

Despite the best efforts of numerous labs over the last decade, studying microRNA—messenger RNA interactions is still a slow and error-prone process of computational predictions based around sequence conservation (and a host of other sequence elements), supported by luciferase reporter assays, microRNA-transfection followed by micro-array or mass spec analysis, knockouts of microRNAs and components of their pathway, and other methods. So, it is with great excitement and some frustration that we receive this HITS-CLIP paper from the Darnell lab.

In a late 2008 paper, Darnell and colleagues developed the “HITS-CLIP” method, short for high-throughput sequencing of RNA isolated by cross-linking immunoprecipitation. This mouthful-of-an-acronym method involves using ultraviolet light to cross-link proteins to nucleic acids, allowing stringent immunoprecipitation of direct protein—RNA complexes. In Chi et al (2009), the Darnell lab has focused this method on identifying interactions between the workhorse of the microRNA-induced silencing complex, Argonaute (Ago), the microRNAs bound in Ago, and the mRNAs targeted by those microRNAs. The authors found that immunoprecipitation of Ago from cross-linked cells produced two populations of Ago-RNA complexes: (1) Ago—microRNA complexes, which run at about 110 kDa after partial RNase digestion, and (2) Ago-mRNA complexes, which run closer to 130 kDa. By isolating these RNA populations separately and sequencing using Illumina, the authors were able to globally identify both Ago—microRNA and Ago-mRNA interactions.

The difficulty in analyzing these data comes from the heterogeneous population of microRNAs: which Ago-mRNA sequence tag corresponds to an Ago with which microRNA loaded? The authors first use a clever approach they dub “in silico CLIP” to simulate distributing sequence tags across messenger RNAs based on mRNA expression. This simulation provides a background level for the number of tags that would be expected by chance to simultaneously overlap one another, forming clusters. The authors then identify significantly enriched mRNA-sequencing tag clusters, and show that tags in most of these clusters are tightly distributed around the center, giving a sharp peak. For each cluster, then, the authors can search for 6—8mer microRNA seed matches within the cluster, and suggest which microRNA bound which mRNA clusters.

There was a significant enrichment of clusters at both ends of 3′ untranslated regions (UTRs), as was expected given prior research that most functional microRNA targets are in these regions. This study also identified many Ago—mRNA clusters in coding sequence, although not above background, and in introns and intergenic sequence, though the authors did not explore the explanation that these may simply be the result of unannotated transcripts and retained introns.

This method has the advantage over computational predictions of identifying true Ago—mRNA interactions, but these interactions do not necessarily result in noticeable down-regulation of the messenger RNA. To begin to assess how often these Ago—mRNA interactions are productive, the authors transfected a brain-specific microRNA, miR-124, into the cervical cancer cell line HeLa, and then used HITS-CLIP to identify Ago—mRNA clusters. The authors found that those mRNAs apparently bound by miR-124, according to HITS-CLIP, were significantly downregulated following transfection when compared to those with miR-124 sites computationally predicted by TargetScan. This was true at both the protein and the mRNA level across all transcripts, as well as when only looking at the brain-expressed messenger RNAs at the mRNA level, although brain-expressed genes did not show a convincing down-regulation at the protein level.

The authors end with a faulty Gene-Ontology—based analysis, comparing HITS-CLIP to previously published microRNA-target predictions. For the most highly expressed microRNAs in their study, the authors analyzed enrichment of various GO categories in HITS-CLIP mRNA clusters with the associated seed site. They found significant enrichment for several of these microRNAs for several of these neuronal GO categories. In comparing these results to microRNA-target predictions, the authors compared mRNAs with and without predicted microRNA target sites. However, this ignores the fact that many genes have very little conservation in their 3′ UTRs, and hence could not be predicted as targets of any microRNA. A better comparison might take transcripts targeted by non-expressed microRNAs as the background set, and compare these to those predicted to be targeted by the highly expressed microRNAs.

In summary, this is an exciting and powerful new technique, which will quickly broaden our understanding of microRNA regulation. A few issues marred what could have been an exceptionally interesting paper. First, the authors seemingly randomly cherry-pick their data for each figure panel, sometimes choosing conserved microRNAs, sometimes non-conserved; sometimes those clusters present in all their replicates, sometimes only those in two or more replicates; sometimes the top 30 most-expressed microRNAs, sometimes only the top 20. These decisions may have been well founded, or the results were similar regardless of which data they chose, but without clear explanations of why they conducted their analyses the way they did, it is difficult to express confidence in the robustness of their results. Secondly, the HITS-CLIP method has a huge advantage over target prediction methods in being able to identify non-conserved target sites, and yet the authors restricted most of their analyses to only those conserved microRNA targets. Finally, the authors chose not to make the raw HITS-CLIP sequencing data readily available online (submission to the NCBI Short Read Archive is the standard for sequencing data, as GEO is the standard for micro-array data), although one can hope that this will be rectified in the near future.

Update 7/24/09:

As Dr. Darnell calls attention to in his comment below, all of the raw data and UCSC links are now available. For them, visit the Darnell Lab Ago HITS-CLIP website here.

Upstream open reading frames cause widespread reduction of protein expression and are polymorphic among humans

Sarah E. Calvoa, David J. Pagliarinia and Vamsi K. Mootha

PNAS 106 (18): 7507-7512, May 2009.
doi: 10.1073/pnas.0810916106

This week’s incisive summary and analysis by Robin Friedman:

Upstream ORFs (uORFs) generally consist of an AUG codon with an in-frame stop codon preceding the end of the canonical coding sequence (CDS).  The uORFs therefore can either be entirely upstream of the CDS or overlapping the start of the CDS.  uORFs have been shown to decrease CDS expression in many anecdotal cases, although translation of the CDS can still occur by leaky scanning or re-initiation.  Early analysis suggested that <10% of vertebrate mRNAs had upstream AUGs, but more recent computational predictions suggested that >40% of vertebrate genes have uORFs.  This study is the first to experimentally address the extent of uORF impact on a genome-wide scale.

The authors constructed a 5′ UTR dataset from refgene annotations, finding that 49% of human and 44% of mouse transcripts have at least one uORF.  They next examined high-throughput MS/MS datasets for steady-state protein quantification at a genome-wide level.  In each of four datasets, genes that have uORFs have lower protein expression than genes with no uORFs, even after normalizing to mRNA expression.  uAUG context, the distance from cap to uORF, uORF conservation, and the number of uORFs all affected this difference in protein expression, whereas uORF length and distance from uORF to CDS did not.

While the previous experiments show that uORF-containing genes have lower steady-state protein levels, they do not show a direct effect of uORFs on translation.  To test directly whether the uORFs affect translation, the authors created reporters with the 5’UTRs from randomly selected genes containing uORFs fused to luciferase.  Compared to a single-nucleotide-mutant that removes the uAUG, the luciferase activity was reduced ~50% in five randomly selected mouse genes, while the mRNA level, assayed by qPCR, was mostly unchanged.  For 10 mouse genes with MS/MS and conservation support for functional uORFs, the luciferase reporters showed 50-80% repression at the protein level.

Asking whether the uORFs could be involved in human polymorphism and disease, the authors queried dbSNP and the human gene mutation database for mutations that create or destroy uORFs.  There are 509 genes with polymorphic uORFs, and 14 with recorded mutations linked with disease.  Five of the polymorphisms were tested by qPCR, and the uORF was found to repress protein levels by 30-60%, while five of the disease-causing uORF mutations were found to repress by 70-100%.

This paper convincingly argues that uORFs are widespread in humans and have a widespread impact on protein expression.  Much of this impact is likely conserved and functional.  In addition, they provide interesting experimental support for the fact that uORFs typically repress at the translational level as opposed to through NMD and that CDS translation downstream of uORFs likely proceeds from leaky scanning rather than from re-initiation.  While the mechanism has not been elucidated on a genome-wide scale, this paper provides an refreshing look at an often-ignored but important contribution to translational control.

A Role for RNAi in the Selective Correction of DNA Methylation Defects

Felipe Karam Teixeira, Fabiana Heredia, Alexis Sarazin, François Roudier, Martine Boccara, Constance Ciaudo, Corinne Cruaud, Julie Poulain, Maria Berdasco, Mario F. Fraga, Olivier Voinnet, Patrick Wincker, Manel Esteller, Vincent Colot

Science 323 (5921): 1600-1604, March 2009.
doi: 10.1126/science.1165313

This week’s summary and expert analysis by Michael Nodine:

DNA methylation of transposable elements occurs through both RNAi-dependent and RNAi-independent mechanisms in plants. Methylation of transposable elements leads to their silencing and maintains genomic stability. Mutations in methylation components, such as the maintenance methyltransferase MET1 and the chromatin remodeler DDM1, lead to a loss of >70% of genomic methylation. The progeny from met1 x wild-type (WT) and ddm1 x WT crosses have reduced methylation despite these mutations being recessive. Furthermore, when these heterozygous plants are selfed and the MET1 and DDM1 loci are restored to the homozygous WT condition, several loci remain hypomethylated. Based on these findings, it has been proposed that once methylation is severely compromised it cannot be restored and thus is permanently lost. However, comparisons between different Arabidopsis accessions revealed that the methylation patterns of repetitive elements were similar across generations. This suggests that a mechanism exists to prevent permanent loss of DNA methylation. That is, there must be a way to specifically and robustly reestablish methylation.

In this study, Teixeira et al. set out to identify the mechanism that underlies this methylation reestablishment. First, they crossed the methylation defective mutant ddm1 with WT, recovered DDM1 F2 plants and selfed these plants for several generations. They then examined the methylation levels of several loci in the heterochromatic knob region of chromosome 4, and found that methylation was restored for ~50% of the repetitive loci examined (remethylatable sequences (R)), but not for the other ~50% (non-remethylatable sequences (NR)). The patterns of NR and R sequences were consistent between different independent lines. Remethylation did not occur in the F1 generation, but was progressive from the F2 generation onwards and led to silencing of transposable elements. In contrasts to previous models, these findings indicate that a robust and targeted remethylation process takes place.

The authors went on to demonstrate that cytosine remethylation occurred in all three sequence contexts (CG, CHG and CHH where H it A, T or C). They also found that NR sequences had stronger dependence on DDM1 for CHH methylation than did R sequences. Since RNAi components have important roles in CHH methylation, this observation led to the hypothesis that RNAi may be involved in remethylation of R sequences. To test this, they examined small RNA datasets and found that R sequences had a strong association with small RNAs especially 24-nt heterochromatic siRNAs (hc-siRNAs), which are involved in RNA-directed DNA methylation. Moreover, when they combined mutations in the RNAi machinery with ddm1 they observed an enhanced loss of methylation at both R and NR sequences suggesting that RNAi plays a role in the methylation at both types of loci. However, mutations in RNAi components (when not combined with ddm1) resulted in decreased methylation at R, but not NR, sequences. To demonstrate that RNAi plays a direct role in remethylation, they examined whether R sequences were remethylated when RNAi was compromised in the initial generation and found that only sporadic and inconsistent remethylation occurred in several independent progeny lines. Together, these results indicate that RNAi is involved in robust remethylation at specific loci.

Based on these findings, the authors propose that there are three types of methylated loci: those that 1) depend solely on maintenance methylation machinery (NR loci), 2) depend on both maintenance methylation and RNAi components (R loci), and 3) depend solely on RNAi components (unaffected in ddm1 mutants). Furthermore, they speculate that this mechanism may allow for the generation of epialleles with differences in transgenerational stability.

Although, the authors performed a thorough analysis of the 500 kb heterchomatic knob region of chromosome 4, their conclusions could have strengthened if they would have performed a more genome-wide bisulfite sequencing approach to test whether remethylation occurs on a large-scale in both euchromatic and heterochromatic regions. Furthermore, it would have been informative if they would have reported whether small RNA levels increase with each generation. This may have yielded insight into the mechanism behind the progressive nature of remethylation. Several outstanding questions remain. What features distinguish remethylatable vs. non-remethylatable sequences? Why does it take so many generations for remethylation to be re-established? Is there a benefit for the observed slow re-establishment vs. a more rapid one?