Read on Twitter
Excited to share our latest work, where we describe an approach to systematically address a pressing (in our opinion) need in epigenetics: How do we identify functional (ie. causal) disease-associated epigenetic variation? [1/18] https://doi.org/10.1101/2020.11.08.373456
This is led by Teresa Luperchio and @LeandrosBoukas in collaboration with @BjornssonL [2/18]
Our approach relies on leveraging the Mendelian Disorders of the Epigenetic Machinery: monogenic disorders caused by loss-of-function variants in epigenetic regulator genes. See for example [3/18]
https://doi.org/10.1146/annurev-genom-090613-094245
https://doi.org/10.1101/gr.239442.118
https://doi.org/10.1146/annurev-genom-090613-094245
https://doi.org/10.1101/gr.239442.118
... and a review by @ummiciptasari and Bokhoven [3.5/18] https://doi.org/10.1093/hmg/ddaa175
What's special about MDEMs? In our opinion, two things:(1) They are diseases where the disruption of normal epigenetic modification patterns is *causally* involved in pathogenesis (2) Even though they have different causative genes, they have many phenotypes in common [4/18]
On the sharing of phenotypes: some of these disorders are very hard to differentiate in the clinic. This figure (from XX) shows some known overlap [5]
![On the sharing of phenotypes: some of these disorders are very hard to differentiate in the clinic. This figure (from XX) shows some known overlap [5]](https://pbs.twimg.com/media/EmdzQbEWEAAXsTj.jpg "On the sharing of phenotypes: some of these disorders are very hard to differentiate in the clinic. This figure (from XX) shows some known overlap [5]")
We hypothesize that this phenotypic overlap is a consequence of overlap at the epigenomic level: the mutations disrupt distinct genes but lead to common epigenomic abnormalities which in turn give rise to the common phenotypes, by creating common gene expression abnormalities [6]
![We hypothesize that this phenotypic overlap is a consequence of overlap at the epigenomic level: the mutations disrupt distinct genes but lead to common epigenomic abnormalities which in turn give rise to the common phenotypes, by creating common gene expression abnormalities [6]](https://pbs.twimg.com/media/EmdzXkVWMAMG0H2.png "We hypothesize that this phenotypic overlap is a consequence of overlap at the epigenomic level: the mutations disrupt distinct genes but lead to common epigenomic abnormalities which in turn give rise to the common phenotypes, by creating common gene expression abnormalities [6]")
If that is true, then a joint analysis of multiple MDEMs should provide a catalog of such common abnormalities, which should be enriched for causal epigenetic hits [7/18]
![If that is true, then a joint analysis of multiple MDEMs should provide a catalog of such common abnormalities, which should be enriched for causal epigenetic hits [7/18]](https://pbs.twimg.com/media/EmdzbOCW4AEbjEl.jpg "If that is true, then a joint analysis of multiple MDEMs should provide a catalog of such common abnormalities, which should be enriched for causal epigenetic hits [7/18]")
As proof-of-principle, we study mouse models of Kabuki types 1/2 and Rubinstein-Taybi syndromes, and do ATAC- and RNA-seq on CD19+ B cells to gain insights into the immune dysfunction (IgA deficiency, abnormal B cell maturation); everything done in multiple replicates [8/18]
Inspired by recent work in covariate-powered multiple testing (IHW: https://dx.doi.org/10.1038/nmeth.3885, functional FDR), we develop a new approach for testing for overlap between differential features from different experiments [9/18]
And using this approach, we find widespread sharing of chromatin accessibility abnormalities across the three disorders [10/18]
![And using this approach, we find widespread sharing of chromatin accessibility abnormalities across the three disorders [10/18]](https://pbs.twimg.com/media/EmdzuebXMAAR_gT.jpg "And using this approach, we find widespread sharing of chromatin accessibility abnormalities across the three disorders [10/18]")
And often find gene expression changes downstream of promoter accessibility changes [11/18]
![And often find gene expression changes downstream of promoter accessibility changes [11/18]](https://pbs.twimg.com/media/EmdzyakXYAEEoYH.jpg "And often find gene expression changes downstream of promoter accessibility changes [11/18]")
Genes downstream of commonly disrupted promoters are more likely to have dysregulated expression than those with promoters disrupted only in one or two disorders [13/18]
![Genes downstream of commonly disrupted promoters are more likely to have dysregulated expression than those with promoters disrupted only in one or two disorders [13/18]](https://pbs.twimg.com/media/Emdz6U1XMAUIJAM.jpg "Genes downstream of commonly disrupted promoters are more likely to have dysregulated expression than those with promoters disrupted only in one or two disorders [13/18]")
And we provide two specific examples of how this widespread dysregulation likely leads to known immune phenotypes
1. For abnormal B cell maturation: Expression dysregulation of many transcription factor genes [14/18]
![And we provide two specific examples of how this widespread dysregulation likely leads to known immune phenotypes1. For abnormal B cell maturation: Expression dysregulation of many transcription factor genes [14/18]](https://pbs.twimg.com/media/Emd0Es1W8AInLKf.png "And we provide two specific examples of how this widespread dysregulation likely leads to known immune phenotypes1. For abnormal B cell maturation: Expression dysregulation of many transcription factor genes [14/18]")
1. For abnormal B cell maturation: Expression dysregulation of many transcription factor genes [14/18]
2. For IgA deficiency: Expression dysregulation of many genes known to be important for IgA production [15/18]
![2. For IgA deficiency: Expression dysregulation of many genes known to be important for IgA production [15/18]](https://pbs.twimg.com/media/Emd0IyfWMAA1_m7.png "2. For IgA deficiency: Expression dysregulation of many genes known to be important for IgA production [15/18]")
The last example is surprising: We know KS1 has IgA deficiency. In the clinic KS1 and KS2 are essentially indistinguishable (although KS2 is very rare, so we know less about it). We also know that KS2 has some immune problems. [16/18]
Together, this suggests that KS2 should have IgA deficiencey, just like KS1. But our analysis suggests that KS2 is different. So we assayed the mice and it turns out that KS2 mice do not have igA deficiencey. [17/18]
![Together, this suggests that KS2 should have IgA deficiencey, just like KS1. But our analysis suggests that KS2 is different. So we assayed the mice and it turns out that KS2 mice do not have igA deficiencey. [17/18]](https://pbs.twimg.com/media/Emd0WWUXcAQt-XN.png "Together, this suggests that KS2 should have IgA deficiencey, just like KS1. But our analysis suggests that KS2 is different. So we assayed the mice and it turns out that KS2 mice do not have igA deficiencey. [17/18]")
We think the study of the Mendelian Disorders of the Epigenetic Machinery is a principled approach for systematically mapping functional disease-associated epigenetic variation. And the breadth of MDEM manifestations suggests that many phenotypes can be studied this way [18/18]
If you're attending, @LeandrosBoukas is giving a talk on this study at Epigenomics of Common Disease 2020 #Epigenomics2020
