On 2016 Apr 18, Duke RNA Biology Journal Club commented:

This comment is the summary of a discussion from our journal club meeting.

General Impressions: An impressively thorough paper which uses a combination of cell biology, biochemistry and high-throughput sequencing approaches to first identify lncRNAs associated with repressed chromatin, determine what genes a model lncRNA regulates and, finally, identify a specific structural mechanism by which this occurs. This paper lays the groundwork for future identification and characterization of chromatin associated lncRNAs. One overarching criticism with this work is it reads as two separate papers - one identifying the biological role that MEG3 plays in transcription regulation and a second developing the triple helix method of regulation.

Specific Points:<br> A technique, ChRIP-seq, was used to determine the global lncRNA environment of repressive chromatin. The design of the RIP protocol using two different proteins associated with repressed chromatin greatly simplified the final analysis by narrowing the pool of lncRNA targets to 70 unique lncRNAs. However, it was interesting that, even though both RIPs used 4SU crosslinking, the enrichment seen for T-to-C (or A-to-G) conversions was within the EZH2 pulldown and not the H3K27me3 pulldown. What are the 440 RNAs that specifically interacted with H3K27me3 but not EZH2 and why don’t they show crosslinking to the protein over input levels? While this was an interesting puzzle, the comparison that was done between the EZH2 crosslinked enriched RNAs and the chromatin enriched RNAs to provide a list of only 70 lncRNAs, was a clever way of finding chromatin associated RNAs that specifically bound to the PRC2 complex.

From the list of 70 candidate lncRNAs to study, MEG3 was selected. Even though the focus is on one lncRNA, a similar characterization pipeline could be used for the other RNAs identified through this technique. The cross-links identified through the T-to-C transitions were used as a starting point to identify a clear binding site to the protein. Luckily for them, the 4SU labeling, which usually over-crosslinks to its targets, shows only two identifiable clusters of crosslinks. Starting with crosslinks in the more conserved exon 3, they identified 9 bases that abolish ~50% of binding to EZH2 in vitro and in vivo, however, this would also imply there are separate sites on MEG3 that facilitate the other 50% of the binding. Additionally, the non-conserved sequences in MEG3 could fine-tune the binding to various proteins, including EZH2, in different tissues or organisms. Neither of these points is discussed further in the article but would be interesting to delve into if the ChRIP-seq analysis could be modified and applied to tissue samples.

After identifying MEG3’s putative site of binding to EZH2, the group did the obvious experiment and made knockdowns of both EZH2 and MEG3 to determine which genes were affected by RNA-seq experiments. An interesting addition might have been to use complete knockdown cells and a “rescue” with the mutant MEG3 from Fig 2 to provide insight into what genes that binding site specifically affects. They determine that both knockdowns show overlap for the TGF-beta pathway. This is further validated by using an orthogonal assay called ChOP-seq.<br> To explore the intricacies of the MEG3:TGF-beta gene interaction, Mondal and co-workers looked for sequence motifs within the MEG3-binding regions of the genome and discovered a GA-rich sequence motif. Interestingly, they also found GA-rich sequence motifs in the genomic binding locations of rox2 and HOTAIR, two well studied lncRNA known to bind the genome. This information suggests that GA-rich motifs are important for lncRNA localization across the genome.

Several previous studies explored the possibility that lncRNA form triple-helix structures with their target genome binding site using a combination of computational and experimental techniques; Mondal and co-workers used the Triplexator software, which is based on the binding rules needed for triple helix formation, as well as RNAse digestion assays. The Triplexator software identified several regions in MEG3 with high probability of forming triple helices, and those with the highest probability were GA-rich sequences. The assays combined GA-rich dsDNA probes and a target GA-rich segment of MEG3. After incubation, these solutions were separately treated with RNase A and RNase H. RNase A does not degrade ssDNA or dsDNA while RNase H degrades only ssRNA. The solution was sensitive to RNase A digestion, but resisted RNase H digestion. This indicates that the RNA present is not single stranded; however, the authors assume that a triple-helix structure is forming. Because they are not directly observing the structure, it is potentially possible that the RNA is displacing one of the DNA strands and forming an RNA:DNA double-helix. Mondal and co-workers use an anti- triplex dA.2rU antibody for their in vivo assays to confirm the presence of triple-helices, but there is no indication that this antibody was purchased from a commercial source, nor do they explain the method of raising the antibody in-house.

While this information indicates a RNA:DNA triple-helix structure, direct observation is necessary to confirm. Numerous methods could be used to assess the triple-helix structure: X-ray crystallography, nuclear magnetic resonance (NMR), small-angle x-ray or neutron scattering (SAXS/SANS), cryoelectron-microscopy (cryo-EM). As initially stated, we believe this work would have benefitted from splitting into two separate papers: one exploring the genomic binding of MEG3 to the TGF-beta pathway genes, and the other exploring the potential role of triple-helix formation in lncRNA binding.

On 2016 Apr 18, Duke RNA Biology Journal Club commented:

This comment is the summary of a discussion from our journal club meeting.

General Impressions: An impressively thorough paper which uses a combination of cell biology, biochemistry and high-throughput sequencing approaches to first identify lncRNAs associated with repressed chromatin, determine what genes a model lncRNA regulates and, finally, identify a specific structural mechanism by which this occurs. This paper lays the groundwork for future identification and characterization of chromatin associated lncRNAs. One overarching criticism with this work is it reads as two separate papers - one identifying the biological role that MEG3 plays in transcription regulation and a second developing the triple helix method of regulation.

Specific Points:<br> A technique, ChRIP-seq, was used to determine the global lncRNA environment of repressive chromatin. The design of the RIP protocol using two different proteins associated with repressed chromatin greatly simplified the final analysis by narrowing the pool of lncRNA targets to 70 unique lncRNAs. However, it was interesting that, even though both RIPs used 4SU crosslinking, the enrichment seen for T-to-C (or A-to-G) conversions was within the EZH2 pulldown and not the H3K27me3 pulldown. What are the 440 RNAs that specifically interacted with H3K27me3 but not EZH2 and why don’t they show crosslinking to the protein over input levels? While this was an interesting puzzle, the comparison that was done between the EZH2 crosslinked enriched RNAs and the chromatin enriched RNAs to provide a list of only 70 lncRNAs, was a clever way of finding chromatin associated RNAs that specifically bound to the PRC2 complex.

From the list of 70 candidate lncRNAs to study, MEG3 was selected. Even though the focus is on one lncRNA, a similar characterization pipeline could be used for the other RNAs identified through this technique. The cross-links identified through the T-to-C transitions were used as a starting point to identify a clear binding site to the protein. Luckily for them, the 4SU labeling, which usually over-crosslinks to its targets, shows only two identifiable clusters of crosslinks. Starting with crosslinks in the more conserved exon 3, they identified 9 bases that abolish ~50% of binding to EZH2 in vitro and in vivo, however, this would also imply there are separate sites on MEG3 that facilitate the other 50% of the binding. Additionally, the non-conserved sequences in MEG3 could fine-tune the binding to various proteins, including EZH2, in different tissues or organisms. Neither of these points is discussed further in the article but would be interesting to delve into if the ChRIP-seq analysis could be modified and applied to tissue samples.

After identifying MEG3’s putative site of binding to EZH2, the group did the obvious experiment and made knockdowns of both EZH2 and MEG3 to determine which genes were affected by RNA-seq experiments. An interesting addition might have been to use complete knockdown cells and a “rescue” with the mutant MEG3 from Fig 2 to provide insight into what genes that binding site specifically affects. They determine that both knockdowns show overlap for the TGF-beta pathway. This is further validated by using an orthogonal assay called ChOP-seq.<br> To explore the intricacies of the MEG3:TGF-beta gene interaction, Mondal and co-workers looked for sequence motifs within the MEG3-binding regions of the genome and discovered a GA-rich sequence motif. Interestingly, they also found GA-rich sequence motifs in the genomic binding locations of rox2 and HOTAIR, two well studied lncRNA known to bind the genome. This information suggests that GA-rich motifs are important for lncRNA localization across the genome.

Several previous studies explored the possibility that lncRNA form triple-helix structures with their target genome binding site using a combination of computational and experimental techniques; Mondal and co-workers used the Triplexator software, which is based on the binding rules needed for triple helix formation, as well as RNAse digestion assays. The Triplexator software identified several regions in MEG3 with high probability of forming triple helices, and those with the highest probability were GA-rich sequences. The assays combined GA-rich dsDNA probes and a target GA-rich segment of MEG3. After incubation, these solutions were separately treated with RNase A and RNase H. RNase A does not degrade ssDNA or dsDNA while RNase H degrades only ssRNA. The solution was sensitive to RNase A digestion, but resisted RNase H digestion. This indicates that the RNA present is not single stranded; however, the authors assume that a triple-helix structure is forming. Because they are not directly observing the structure, it is potentially possible that the RNA is displacing one of the DNA strands and forming an RNA:DNA double-helix. Mondal and co-workers use an anti- triplex dA.2rU antibody for their in vivo assays to confirm the presence of triple-helices, but there is no indication that this antibody was purchased from a commercial source, nor do they explain the method of raising the antibody in-house.

While this information indicates a RNA:DNA triple-helix structure, direct observation is necessary to confirm. Numerous methods could be used to assess the triple-helix structure: X-ray crystallography, nuclear magnetic resonance (NMR), small-angle x-ray or neutron scattering (SAXS/SANS), cryoelectron-microscopy (cryo-EM). As initially stated, we believe this work would have benefitted from splitting into two separate papers: one exploring the genomic binding of MEG3 to the TGF-beta pathway genes, and the other exploring the potential role of triple-helix formation in lncRNA binding.