Sunday, January 26, 2014

Notes on: Single-cell RNA-Seq reveals dynamic, random monoallelic gene expression in mammalian cells

Brief background:

We have two copies of each non-sex gene. Each version of the gene is called an allele: one inherited from your genetic mother, one from your genetic father. It is generally thought that each allele is expressed (turned out) at the same intensity. But, there are some examples where this isn't true. The most notable occurs on the X chromosome. Females with two X chromosomes inherited one X chromosome from each parent, but one of these X chromosomes is almost completely inactivated. That means that instead of having biallelic expression (expression from both the maternal and paternal allele), most genes on the X chromosome exhibit monoallelic expression.

In recent work, Deng et al (2014) isolated single cells from two different stains of mice, where they could detect maternal-alleles and paternal-alleles for over 82% of assayed genes (in the other genes, there were not unique variants that allowed deciphering between the two alleles). For each gene, the authors characterized whether they could detect expression from both the maternal and paternal alleles, or from only the maternal or paternal allele. Although the title says, "mammalian," all of the experiments and analysis were conducted in mouse cells and tissues, so far as I can tell.

My notes and thoughts on the paper:
  • The authors state, "…different SNPs within the same gene gave coherent allelic calls (fig. S2)." I am very interested to see what they did in cases where different SNPs did not give the same estimates of allele-specific expression.
  • Mouse paternal X chromosome inactivation is complicated. In single cells, the paternal X chromosome is inactive initially, reactivated starting at the late 2-cell stage, active at the 4-cell stage, then inactivated starting at the 16-cell stage, and completely inactivated again by the early blastocyst stage. Xist appears to be off during early embryogensis, and is only expressed starting at the 16-cell stage - correlating with the re-silencing of the paternal X chromosome in the mouse.
  • X-inactivation near and far from mouse XIC. The spread of X-inactivation is not directly correlated to the distance from the X-inactivation center (XIC).
  • Technology biases estimates of allele-specific expression. Initial observations of allele-specific expression on the autosomes suggested over half of all genes exhibit mono-allelic expression, but as much as 66% of these are false positives due to the loss of RNA molecules with the available technology. After inferring the proportion of losses RNA molecules, the authors propose that 12-24% of genes exhibit monoallelic expression in single cells.
  • Monoallelic expression evens out in tissues. The authors state, "Pooling cells by embryo removed essentially all monoallelic expression, demonstrating a high degree of cell-specific randomness in monoallelic expression." To me this suggests that studies of single cell gene expression may not give the most accurate picture of gene expression within a tissue. 
Additional thoughts:
  • I would very much like to know how estimates of allele-specific expression on the X chromosome varied between the single cell and multicelluar analyses. 
  • The authors claim the patterns of monoallelic expression on the autosomes is likely due to independent allelic expression, but I would like to understand the mechanism more. Is this simply variance in polymerase activity? 
  • If 12-24% of genes are expressed from only one allele, what can we learn from it? Is dosage of these genes less important? Is selection weaker on genes that are more likely to be mono-allelicly expressed? 

 2014 Jan 10;343(6167):193-6. doi: 10.1126/science.1245316.

Single-cell RNA-seq reveals dynamicrandom monoallelic gene expression in mammalian cells.


Here is the Storify of the discussion on twitter about this result, and how it is transcriptional bursting. I wasn't aware of the term, so here's the wikipedia entry for transcriptional bursting. Given this background, it is not so surprising that there is so much variation in expression across the autosomes, but my questions about what kind of classes exhibit measurable levels of this phenomenon in single cells still stands.

Also, I did focus on the X chromosome results because to me they were the most interesting. In marsupials it is always the paternal X that is inactivated - there is no random X-inactivation. My understanding is that in eutherian mammals the paternal X is inherited as inactivated, it is reactivated, then either the maternal or paternal allele is randomly inactivated. That said, there is some evidence of preferential paternal X-inactivation in mice - see Paternally biased X inactivation in mouse neonatal brain.


Kelkar said...

Very interesting, the transcriptional burst phenomenon. Expression noise in bacteria is quite well-known, but was surprised to see how widespread it seems to be in vertebrates. I wonder how well the transcriptional bursts 'translate' to bursts of protein synthesis.

mathbionerd said...

I am a little embarrassed to admit I didn't know either.

Hmm... do you know if transcriptional bursts lead to complete transcripts or partial transcripts? If they're complete transcripts including UTRs, then I don't see why they wouldn't be translated into proteins.

Mark P said...

"After inferring the proportion of losses RNA molecules, the authors propose that 12-24% of genes exhibit monoallelic expression in single cells."

I am not in the field and have a hard time evaluating their degree of certainty but I find this almost impossible to believe, as it seems like this would inevitably uncover heterozygous hits in essential genes. How much of a snapshot in time is this--if protein perdurance were substantially longer than transcript lifetime, parental alleles could be switching on and off really rapidly it might not have the effect I'd expect.

mathbionerd said...

Mark, I think you're right. Each analysis is in a single cell during extremely early development (2-cell, 4-cell, 16-cell stages). Most of this is likely just variance in transcription across single cells, because, as the authors state, the signal disappears when expression across multiple cells is considered.

Further, as Kelkar states, and the authors recognize, the bulk of this signal is likely just transcriptional bursting.

That's why I focused primarily the X chromosome, where this method, when considered over a whole chromosome, might actually give a picture of when whole chromosome silencing is initiated.