Machine learning reveals bilateral distribution of somatic L1 insertions in human neurons and glia

Abstract

Active retrotransposons in the human genome (L1, Alu and SVA elements) can create genomic mobile element insertions (MEIs) in both germline and somatic tissue 1 . Specific somatic MEIs have been detected at high levels in human cancers 2 , and at lower to medium levels in human brains 3 . Dysregulation of somatic retrotransposition in the human brain has been hypothesized to contribute to neuropsychiatric diseases 4,5 . However, individual somatic MEIs are present in small proportions of cells at a given anatomical location, and thus standard whole-genome sequencing (WGS) presents a difficult signal-tonoise problem, while single-cell approaches suffer from limited scalability and experimental artifacts introduced by enzymatic whole-genome amplification 6 . Previous studies produced widely differing estimates for the somatic retrotransposition rates in human brain 3,6-8 . Here, we present a highly precise machine learning method (RetroSom) to directly identify somatic L1 and Alu insertions in <1% cells from 200× deep WGS, which allows circumventing the restrictions of whole-genome amplification. Using RetroSom we confirmed a lower rate of retrotransposition for individual somatic L1 insertions in human neurons. We discovered that anatomical distribution of somatic L1 insertion is as widespread in glia as in neurons, and across both hemispheres of the brain, indicating retrotransposition occurs during early embryogenesis. We characterized two of the detected brain-specific L1 insertions in great detail in neurons and glia from a donor with schizophrenia. Both insertions are within introns of genes active in brain (CNNM2, FRMD4A) in regions with multiple genetic associations with neuropsychiatric disorders 9-11 . Gene expression was significantly reduced by both somatic insertions in a reporter assay. Our results provide novel insights into the potential for pathological effects of somatic retrotransposition in the human brain, now including the large glial fraction. RetroSom has broad applicability in all disease states where somatic retrotransposition is expected to play a role, such as autoimmune disorders and cancer. yielded ~59,900 (95% CI: 55,100-64,700) false positive MEI detections ( Fig. 1D and Extended Data Fig. 1A).

Increased burden of ultra-rare structural variants localizing to boundaries of topologically associated domains in schizophrenia

“…We chose not to analyze rare mobile element insertions because variant calling for these variants appear to be noisy from our 30× WGS and there was a lack of external dataset or analytic approach for the need of quality control. Increased somatic L1 insertions have been recently reported in neurons of schizophrenia patients using postmortem brain tissues 45 . The detection of somatic L1 insertions required very deep WGS (e.g.…”

“…200×) and tailored analytic methods (e.g. machine learning 45 ). For similar reasons, we also chose not to evaluate translocations and complex SVs in this study as we feel that these variants can be better detected from WGS using long-insert jumping libraries, deeper coverage, and targeted capture of breakpoints 46 .…”

Cas9 targeted enrichment of mobile elements using nanopore sequencing

“…Another study showed that a recurrent intronic deletion results in the exonization of an Alu element that is found in 6% of families with mild hemophilia A in France 13 . Somatic L1 retrotransposition can occur in neuronal progenitor cells [14][15][16][17][18] , indicating a possible role for L1s in the etiology of neuropsychiatric diseases 19 . In addition, a mutagenic L1 insertion that disrupted the 16th exon of the APC gene has been shown to instigate colorectal tumor development 9 .…”

Identification and Genotyping of Transposable Element Insertions From Genome Sequencing Data

“…As an alternative to single‐cell sequencing with its high experimental costs and amplification bias, some recent studies have used a bulk sequencing approach with high sequencing coverage to detect low‐level mosaic somatic insertions in the brain and other tissues. RetroSom is a machine learning−based tool that detects somatic L1 and Alu insertions from ultra‐high‐depth WGS, and its application to ∼200× bulk WGS of sorted neurons and glia revealed two brain‐specific somatic L1 insertions at ∼1% mosaicism (Zhu et al., 2019). Furthermore, HAT‐seq, a PCR‐based targeted bulk sequencing approach, identified somatic L1 insertions with low‑level mosaicism in neurons and non‐brain tissues (Zhao et al., 2019).…”