Highlights
The Keystone Non-Coding RNA Meeting drew some very big names in ncRNA research circles to Snowbird, Utah. Our guest reporter, Musa Mhlanga from the Institute for Molecular Medicine in Portugal, was on hand to make sure that we didn’t miss out on any of the action. Take a look at his coverage of this exciting event:
Keystone Symposium on Non-Coding RNA Overview
The joint Keystone meetings on Non-Coding RNA and Eukaryotic transcription, kicked off with a series of combined keynote lectures before undergoing a “meiotic separation” into the separate tracks. In many respects, this meeting had the feeling of what it must have been like to be a young scientist in the early 1980s. When one considers that there are at least as many mapped lncRNAs as there are coding RNAs, it definitely feels like a brave new world is emerging, where we actually only know about 2% of what the genome does and the rest can only be decrypted by interpreting phenotypes that are sometimes difficult to determine. This means lots of new exciting lncRNA functions are yet to be discovered by young scientists.
In this meeting the current mainstream view of the gene would come under serious challenge. With new terminology emerging (such as naming the atomic unit for genetics “the transcript”, as opposed to “the gene”) the community is beginning to re-craft our paradigms on how information is stored and transmitted in the genome. Many of the posters and talks at the non-coding RNA meeting attempted to recast their views on the field this way. Several succeeded greatly, and anyone who attended, from editors of major journals to post-docs, were treated to a number of interesting stories.
Importance of Non-coding RNA
John Mattick, University of Queensland
For John Mattick, conference co-organizer, and an early proponent of the importance of non-coding RNA, this was truly a “coming out” moment. He wove a compelling narrative of how the “one gene, one enzyme” paradigm came to obscure our thinking about the genome. This was best crystalized in a quote from a 1950 letter from Nobel laureate Barbara McClintock to Marcus Rhoades “Are we letting a philosophy of the protein-coding gene control (our) reasoning? What then is the philosophy of the gene?” As Mattick puts it, the misgivings McClintock highlighted half a century ago would then be propagated to the present day. Indeed the protein-centric view reflects a mechanical orientation that has led to a number of subsidiary assumptions. At its extreme, it has led to assigning the status of junk to 98% of the human genome. Today a more nuanced view is emerging where gene number does not necessarily scale with developmental complexity. Instead what does correlate is the increasing functions of non-coding RNA that are derived in small part from what we still call genes and in large part from the part we used to call junk.
The Epigenetics of lncRNA
Howard Chang, Stanford University
Perhaps the resurgence of interest in lncRNAs can be traced back to when Howard Chang and John Rinn published their seminal paper on how active and silent domains of Hox genes were regulated by lncRNAs. As Dr. Chang aptly puts it, genes need to remember their environment and what they have been exposed to. Epigenetics is a preservation of gene memory over time. In his talk at this meeting, Chang would enter into the mechanistic details of how lncRNA could assist in epigenetics with the example of how the adapter protein WDR5 is able to read multiple inputs.
The Chang lab has shown this protein to have an increased half-life as a result of HOTTIP RNA transcripts. Such expression by HOTTIP in turn increases WDR5 chromatin occupancy. Ubiquitination of WDR5 is stabilized by HOTTIP with the lncRNA appearing to play an unexpected role in physically blocking the proteasome from degrading ubiquinated proteins that are bound to certain lncRNAs. Indeed, specific short sequence motifs embedded within HOTTIP RNA are able to confer RNA-protein binding interactions that influence protein localization, however maintaining the length (and not the sequence) of the long RNA context is necessary to block ubiquitination. The most elegant example of this that Chang gave was by transferring a short sequence from HOTTIP to LacZ RNA that allowed the latter to bind WDR5 and stabilize its [WDR5] proteasomal degradation. Conversely if the sites where WDR5 binds this short sequence of RNA are mutated, the protein is no longer stabilized and its ubiquitination status targets it for degradation even in the presence of HOTTIP.
Finally Chang demonstrated how important the stabilization of WDR5 is to ES cells. If an inducible shRNA targeting WDR5 is “knocked-in” to ES cells or if WDR5 with the mutated RNA binding domain are introduced into ES cells, those cells lose their capacity for self-renewal and many transcriptional targets of WDR5 are not expressed. Since RNA binding is essential for its activity, chromatin occupancy of the protein is also lost as is H3K4me3 status at promoters. This represents one of the few examples of a protein other than LSD1, PRC2 and G9A through which lncRNAs are able to act. The relationship of WDR5 to HOTTIP seems to be highly intimate and it was unclear from Chang’s talk whether other short sequences within other lncRNAs could interact with the WDR5 domain similarly.
The Cautionary Tale of Malat1
David Spector, Cold Spring Harbor
Malat1/NEAT2 is a well known lncRNA having been discovered in 1994. Noted for expression levels that are greater than Beta-actin, it is in every cell and localized to large speckles in the nucleus. Malat1’s nuclear staining pattern co-localizes with the splicing factor SF2ASF and it has been found to bind to unmethylated polycomb2 protein.
These findings consolidated the evidence that it was performing some key function in the nucleus and perhaps in general homeostasis. As David Spector put it in his talk, “I would have bet a lot of money that Malat1 does something extremely important.” That comment portents what would happen next. After his group “floxed out” over half of the 6.7kb lncRNA and backcrossed the mice six generations, they would find that Malat1 was completely dispensable for growth & development. All major organs where it was previously abundantly expressed (liver, lung, testis, brain, prostate etc) would show no gross, or subtle defects or disorganization. In the nucleus the deletion of Malat1, did not result in any change to the level of phosphorylation of splicing factors when measured using various techniques including RNA-seq. Neither did any global gene expression patterns change in mice -/- for Malat1. One subtle difference could be identified was in the brain where 12 genes in the cortex were upregulated1.5-2.3 fold of which 5 were near the Malat1 locus. NEAT1, a lncRNA mapping near Malat1, was up-regulated in the liver.
Then the Spector lab took a different tack, focusing on what Malat1 was originally identified for: Metastasis associated in lung adenocarcinoma transcript 1. This is where the results that they found would correlate strongly with the lab of Sven Diedrichs who had also been searching for a function of Malat1. The Spector lab used specially designed antisense oligonucleotides (ASO) to knockdown Malat1 in wildtype animals. They found that it in these animals the growth of human non-small cell lung carcinoma was delayed in EBC1 primary tumors. Further the use of ASOs against Malat1 was able to inhibit lung metastasis of EBC1 tumors.
The Diedriechs lab would take a different route to finding a role for Malat1. They first asked whether they could predict which early stage tumors would metastasize based upon the expression of lncRNAs. Through this approach they identified LuCaIR1 which is overexpressed in non small cell lung carcinoma in 56/57 lung patient tumors. Malat1 was also recovered in this analysis and found to be essential for metastasis in lung cancer. Further they found that knockouts of Malat1 resulted in no motility in a cell culture scratch assay. It also seems to affect splicing in lung cancer cells. The Diedriechs lab made their own mouse model for Malat1, using a zinc finger nuclease approach to introduce an RNA destabilizing agent such as a poly A sequence into the Malat1 locus. They found these mice to be immune to cancer.
After Spector’s recounting of the entire Malat1 episode, one comes away with the feeling that a cruel joke was played on a gifted and dedicated scientist. The series of systematic failures despite rigorous experimentation was clearly hard for the audience to bear. Several postdocs and thousands of dollars later, it turns out that Malat1 is a housekeeping lncRNA that, when deleted, spares you lung cancer! With mouse behavioral studies outstanding, it is hoped that some more important aspect of its function will emerge then. If not, it will be the primary exemplar of what can go horribly wrong when you “bet the farm” on an abundantly expressed lncRNA.
Hearts & Wings in Mammals
Carla Klattenhoff, MIT
Perhaps the most intriguing aspect of lncRNAs is how just a single lncRNA expressed in the right cell type can radically alter entire tissues and body forms. HOTAIR and HOTTIP were previous examples of this and their effects were in affecting the Hox locus and complex patterning of digits and extremities. The two stars of this topic at the meeting were short talks selected from the abstracts on lncRNAs Braveheart and MIDAIR. Carla Klattenhoff from Laurie Boyer’s Lab at MIT presented Braveheart, a lncRNA that is required to generate beating embryoid bodies from mouse ES cells. Elegantly demonstrating that Braveheart most likely sits epistatic to most genes in cardiomyocyte formation such as MESP1 but downstream of Brachyury, she presented data on how the knockdown of Braveheart results in no defects in ES cell colonies or in the cell cycle of ES cell bodies.
Braveheart is antisense to a microRNA cluster which includes mir143/145. The expression appears to be linked to expression of mir143/145 since Braveheart’s knockdown results in no expression of those microRNAs or troponinT. Intriguingly inhibiting the expression of mir143/145 has no effect on beating embryoid bodies. Braveheart appears to function at the mesodermal precursor of the cardiac transition and localizes to both the cytoplasm and the nucleus. Its depletion in cardiomyocytes results in the loss of cardiac fate. Since near automatic cardiomyocyte differentiation is common to human and mouse ES cells there would be an expectation that this lncRNA is conserved in both species. Alas it is not, further obscuring what unique role it could be playing in mice that is absent in humans.
Nicola Illing, University of Cape Town
Enter MIDAIR, a lncRNA that may explain why bats are the only known winged mammals that can fly. This story begins in Africa at the Univ. of Cape Town where Nicola Illing and her lab would go to search caves at the southern tip of the continent for bats to understand how their wings evolved. Bats have hind legs with similar patterning to most other mammals. However their forelimbs or “hand” are webbed, with a unique arrangement of the digits and with skin attaching all of them. Upon isolating bat forelimbs in developing embryos the Illing lab were able to identify an EST (expressed sequence tag) that was only expressed in the webbing of the bat “hand” and not the hind limbs.
Analysis of this EST led to the discovery that it is derived from a transcriptional start site of a protein-coding gene, Meis2 conserved in humans, mice and bats. However the lncRNA is transcribed from a distinct non-coding region of Meis2 in bats and was christened MIDAIR. In humans the coding and noncoding transcripts overlap almost completely and thus are essentially the same with no functional lncRNA appearing to be formed. Both the coding and non-coding RNAs co-express in the bat forelimb. Much more characterization needs to be done, but the current indicators point to MIDAIR being the missing link between flight in bats and the absence of it in all other mammals. All this accomplished by a single non-coding RNA under 1000 bases long.
Convergent Evolution – Seeing is Believing
John Rinn, Harvard University
Dr. Rinn wove a tale of how convergent evolution can be in harmony with the independent arising of synonymous functions. The centerpiece of this talk was lncRNA-Rap1, a regulator of adipogenesis. The unique feature of Rap1 is that it occurs as 150bp short sequences in 12 instances in the mouse genome, and as 7 such instances in the human genome. There are 47 species where the lncRNA-Rap1 occurs. Where it gets more intriguing is that convergent evolution preserves its 5’ sequence in primates, but in mouse and other species such as cows, the lnc-Rap1 sequences have arisen independently.
This all suggests that lncRNA-Rap1 has an important function that is evolutionarily conserved probably through its secondary structure. When the Rinn lab did functional analysis of Rap1, they found it to be completely nuclear and single molecule FISH analysis by Arjun Raj’s lab at Penn and Biosearch Technologies revealed well-separated dots of introns and exons in the nucleus. This pattern was similar in mouse and human cells. Strikingly, two large clouds of lncRNA Rap1 stained in the nucleus with scattered dots visible elsewhere. Functional analysis of proteins that may be co-precipitating with lncRNA Rap1 found it to co-precipitate with hnRNP-U and SPT16, with the former being involved in chromosomal architecture. Studies performed using RIP-Seq and knockdowns of hnRNP-U and of lncRNA-Rap1 looked at which genes were regulated by both and revealed a very high concordance between the two phenotypes. Since each exon of lncRNA-Rap 1 has a promoter, is DNAse sensitive and has a CTCF site, it looks like this lncRNA is implicated in RNA and genomic architecture. There is certainly much more we will hear about lncRNA-Rap1.
Finally, Rinn announced the creation of a blog/website “noncodarnia”, to document noncoding RNA progress whose name is inspired by the Chronicles of Narnia.
Long Range RNA Interactions: Wormholes Through the Nucleus
Job Dekker, UMass Medical School
Through the efforts of several research groups but in particular those of Job Dekker and Nancy Kleckner, it is now commonly accepted that there is frequent transcriptionally dependent long-range contact and communication between genomic loci. This contact may involve regulatory elements located far from the target gene suggesting that the linear order of genes on a chromosome does not necessarily reflect functional relationships. Indeed looping between genes and regulatory elements seems to be a common occurrence. Just how frequently this is the case is what Dekker’s talk would be about.
Using a subset of 628 genes his lab examined what types of elements loop to a promoter, what their relationship to gene expression was, where they were located relative to each other, if the interaction could skip genes and finally if structural genomic features such as CTCF could block looping. Dekker used a small subset of genes because asking such questions at genomic scale is beyond the capability of the chromosome-capture and conformation technique, 5C that he previously developed and would use. In his small peek at such interactions in the human genome, the findings were of great interest. He found on average 1.4 million long-range interactions per cell. Of these (in the 4 ENCODE cell lines he examined) 50% of looping interactions were cell-type specific, with looping enriched in ES cells.
Dekker further grouped these interactions into four classes; promoter, enhancer, CTCF and unclassifiable. When these interactions were promoters-enhancer based they resulted in transcription and expression was greater if they were looped. Thousands of looping interactions occur in just 1% of the genome with interactions correlated to transcriptional status with the most frequent size being 120kb. The topology of this looping was categorized into three further classes with the majority of looping interactions occurring upstream of the transcription start site (TSS). 27% of the looping elements interacted with the nearest TSS. 48% of looping interactions were with the nearest active genes. 80% of looping interactions were able to skip over a CTCF site, with not all CTCF bound elements acting as insulators. That number fell to 60% when it they are bound by cohesin.
These data enabled the construction of a looping interaction network of all the “worm holing” in the nucleus. The next logical step is to assess the temporal nature of these looping interactions. That presents a steep challenge to current techniques and will require a significant technological leap.
**EpiGenie would to thank Musa Mhlanga, who is a Research Group Leader in the Gene Expression & Biophysics Unit at the Institute for Molecular Medicine in Lisbon, Portugal for providing this conference coverage.