iRNA Webinar 11.2022 with Christina Akirtava, (McManus’s lab, CMU): False-positive IRESes, mRNA annotation errors, and a paradigm “unshift” in mammalian development

On Nov. 22, 2022, Christina Akirtava from Prof. Joel McManus’s lab at Carnegie Mellon University presented her and Gemma May’s latest work where they investigated how errors in genome annotations can lead to false positives. They’re intriguing story was also recently published in a Proceedings of the National Academy of Sciences article (Akirtava, May, & McManus; PNAS 2022 ).

Christina started by explaining the key findings of Xue et al. 2015 that declared a paradigm shift in mammalian development. She explained how bicistronic assays, which are prone to false-positives, were used to measure translational effects of putative IRESs found in the transcript leaders of Hox genes. This work was followed Leppek et al. 2020 who used cryo-EM to observe which region of the transcript leader was interacting with the small ribosomal subunit. The cryo-EM results lead to more bicistronic assays, the results of which were interpreted as putative IRES activity. The Barna group combined these results with other published findings of hyperconserved transcript leaders that showed similar conservation patterns to the Hoxa9 transcript leader. They concluded that the putative IRES region they found was conserved among mouse, human, and zebrafish.

Next, Christina presented her computational modeling of the predicted structures among the 3 organisms and found that the predicted structures are different among them and the putative IRES structure is not conserved or formed in human or zebrafish. Looking at publicly available datasets and annotations, specifically ENCODE RNA-seq data, showed much shorter transcript leaders for Hoxa9 than used by the Barna group, alongside Hoxa10/Hoxa9 fusion reads, in line with the findings of Ivanov et al. 2022 that show Hoxa9 5’ transcript leader is misannotated. Additional evidence for the short transcript leader was provided by the Barna lab in 2017 where they performed Ribo-seq and RNA-seq on mouse embryos showing translation initiation at the annotated protein start codon, and not before.

Using bicistronic assays with controls for monocistronic transcription and alternative splicing, Gemma May and Christina showed that the putative IRES region was demonstrating independent promoter activity, in line with earlier published results by Ivanov et al. 2022. Gemma and Christina mapped the transcription start site of Hoxa9 transcript leader to the annotated promoter sequence, once again supporting a shorter version of the Hoxa9 transcript leader. They also tested additional Hox putative IRESs and found they exhibit independent promoter activity as well. These putative IRESs appeared to have been misannotated as well, given data that indicated short transcript leaders, similarly to the Hoxa9 transcript leader. Further, Christina and her colleagues show that the sequence that was required for the observed bicistronic reporter activity (observed by Xhu et al. 2015) is a conserved transcription factor binding site that can be bound by multiple transcription factors. Finally, using monocistronic reporters with mutated and wild type promoters, Christina and her colleagues show that the originally reported bicistronic results were a false positive.

After concluding the above IRESs are false positives associated with misannotation, Christina expanded her search to other reported hyper conserved transcript leaders with putative IRES activity. She found that many of them are misannotated similarly to the Hox genes’ transcript leaders she described. Most of these regions had overlaps with ENCODE promoters or enhancers, a smaller percentage overlaps with 3’ splice sites.

Overall, Christina’s work highlighted the prevalence of misannotations and possible consequences of those as well as the importance of careful experimental controls to avoid misinterpretation. More generally, the webinar emphasized our need as Computational Biologists for careful work on both the computational and experimental ends.

It was a pleasure to host Christina and get to know more about her work. You can find the video recording of the seminar here: