Quantitative and functional interrogation of parent-of-origin allelic expression biases in the brain

J. Pérez, Nimrod D. Rubinstein, Daniel E. Fernandez, Stephen W. Santoro, Leigh A. Needleman, Olivia Ho-Shing, John J Choi, Mariela Zirlinger, Shau‐Kwaun Chen, Jun S. Liu, Catherine DulaceLife 2015 · DOI

TL;DR

This paper introduces a novel Bayesian statistical model (BRAIM) to precisely quantify parent-of-origin allelic expression biases in the brain, revealing a more complex landscape of genomic imprinting than previously understood. The study identifies 160 imprinted transcripts, including 41 new genes, and demonstrates that these biases are dynamic, tissue-specific, and functionally significant, particularly in brain development and neuronal cell-type regulation.

What Problem Does This Paper Address?

Genomic imprinting, where only one parental copy of a gene is expressed, is a key epigenetic mechanism influencing gene expression. While traditionally thought of as a binary on/off switch (monoallelic expression), emerging evidence suggests a spectrum of "parentally biased" expression. However, accurately quantifying these subtle biases, especially in the complex and heterogeneous environment of the brain, has been challenging. Previous RNA-seq studies often suffered from low resolution, inconsistent results, and a lack of rigorous validation. This paper aims to address these limitations by developing a robust computational framework to comprehensively map and functionally interrogate parent-of-origin allelic expression biases across the mammalian brain.

Methods

The authors developed a novel Bayesian Regression Allelic Imbalance Model (BRAIM). This model is designed to:

  • Account for transcript-level resolution: Unlike gene-level analysis, BRAIM quantifies expression for individual transcripts, capturing isoform-specific imprinting.
  • Handle biological and technical variability: It explicitly models factors like mouse strain (cross), sex, and age, as well as measurement error in RNA-seq data.
  • Provide robust statistical inference: It uses Bayesian variable selection to estimate the probability of an effect (parental bias) being significant.

To validate their findings, they employed pyrosequencing, a highly accurate method for quantifying allele-specific expression, on a subset of candidate genes. They also performed functional studies involving targeted genetic deletion of specific alleles of the Bcl-x gene in mice to assess the phenotypic consequences.

Datasets

The primary dataset used in this study was generated through RNA sequencing (RNA-seq) of 48 individual mouse cerebella. These samples were derived from F1 hybrids of reciprocal crosses between C57BL/6J and Cast/EiJ mouse strains, covering two developmental stages (postnatal day 8 and adult day 60) and both sexes.

For validation, pyrosequencing was performed on 12 individual cerebella for each tested gene, using a separate batch of F1 hybrids.

Functional validation involved genetically modified mice with brain-specific deletions of the maternal or paternal Bcl-x allele. Phenotypic analyses included histochemical staining (NISSL, immunofluorescence) and image analysis of brain sections.

Key Insights and Ideas

  • Beyond Monoallelic: Genomic imprinting is not just an on/off switch; a significant proportion of imprinted genes exhibit biased expression, with varying degrees of preference for one parental allele.
  • Dynamic Regulation: Parental bias is not static. It changes significantly with age, brain region, and even between brain and peripheral tissues, suggesting a highly dynamic regulatory process.
  • Apoptosis Pathway Enrichment: The programmed cell death (apoptosis) pathway is a prominent target of genomic imprinting, with imprinted genes playing roles in both promoting and inhibiting apoptosis.
  • Functional Significance of Bias: Even subtle parental biases can have profound functional consequences. The differential contribution of paternal and maternal Bcl-x alleles to brain development highlights this.
  • Clustering and Conservation: Imprinted genes often cluster together on chromosomes, and this organization, particularly for genes with weaker biases, appears to be evolutionarily conserved, suggesting functional importance.

Key Results

  • 160 Imprinted Transcripts Identified: The study identified 160 imprinted transcripts in the cerebellum, including 41 novel imprinted genes that were independently validated by pyrosequencing. This represents a ~30% increase in the number of known mouse imprinted genes.
  • High Precision: The BRAIM approach achieved a precision of approximately 93%.
  • Age-Related Changes: 59% of imprinted genes showed age-regulated imprinting (either in parental bias or overall expression), with many exhibiting stronger biases and higher expression at the developing P8 stage.
  • Spatial Regulation: Many imprinted genes showed distinct parental biases across 16 brain regions and 7 non-neural tissues, with the brain exhibiting more consistent and robust imprinting patterns.
  • Bcl-x Deletion Phenotype: Brain-specific deletion of the paternal Bcl-x allele (which is paternally biased and more highly expressed) resulted in a significant ~15% reduction in brain weight and loss of specific neuronal types (excitatory neurons in the cortex, Purkinje cells in the cerebellum). Maternal deletion had no significant effect.

Limitations and Future Work

  • Cell-Type Specificity: The RNA-seq data, derived from bulk tissue, cannot definitively distinguish whether parental biases are present in all cell types or result from averaging monoallelic expression in some cell types and biallelic expression in others. Future work could involve single-cell RNA-seq to achieve allelic resolution at the cell type level.
  • Mechanism of Bias: While the study demonstrates the functional impact of parental bias, the precise molecular mechanisms underlying these dynamic spatial and temporal regulations (e.g., DNA methylation, chromatin regulators) require further investigation.
  • Evolutionary Conservation of Bias: While micro-synteny conservation suggests functional importance, direct experimental validation of imprinting in other mammalian species for the identified genes would strengthen these conclusions.
  • Broader Functional Implications: The study focused on the apoptosis pathway and Bcl-x. Further exploration of other biological pathways enriched for imprinted genes and the functional consequences of their dysregulation is warranted.

Summary updated

Abstract

The maternal and paternal genomes play different roles in mammalian brains as a result of genomic imprinting, an epigenetic regulation leading to differential expression of the parental alleles of some genes. Here we investigate genomic imprinting in the cerebellum using a newly developed Bayesian statistical model that provides unprecedented transcript-level resolution. We uncover 160 imprinted transcripts, including 41 novel and independently validated imprinted genes. Strikingly, many genes exhibit parentally biased--rather than monoallelic--expression, with different magnitudes according to age, organ, and brain region. Developmental changes in parental bias and overall gene expression are strongly correlated, suggesting combined roles in regulating gene dosage. Finally, brain-specific deletion of the paternal, but not maternal, allele of the paternally-biased Bcl-x, (Bcl2l1) results in loss of specific neuron types, supporting the functional significance of parental biases. These findings reveal the remarkable complexity of genomic imprinting, with important implications for understanding the normal and diseased brain.
BibTeX
@inproceedings{perez2015quantitative,
  author = {J. Pérez and Nimrod D. Rubinstein and Daniel E. Fernandez and Stephen W. Santoro and Leigh A. Needleman and Olivia Ho-Shing and John J Choi and Mariela Zirlinger and Shau‐Kwaun Chen and Jun S. Liu and Catherine Dulac},
  title = {{Quantitative and functional interrogation of parent-of-origin allelic expression biases in the brain}},
  booktitle = {eLife},
  year = {2015},
  doi = {10.7554/elife.07860},
}
Figure