Other than imprinted genes or X-inactivated genes, there are autosomal genes that show random allelic expression pattern. Within a given clonal cell line, the expressed allele can be maternal, paternal or both [1]. Approximately 5-10% of the genes in the human genome behave this way, which can also be tissue-specific [1]. So depending on the location of your gene and the tissue of interest, the gene may fall into this category. There is also a phenomenon called random allelic bias, in which one of the alleles is preferentially expressed, but the extent is not as well explored. More information can be found in the review by Chess [2].

[1] Gimelbrant et al. Science. 2007 Nov 16;318(5853):1136-40. (https://www.researchgate.net/publication/5839816_Widespread_monoallelic_expression_on_human_autosomes?ev=sim_pub)

[2] Chess. Nat Rev Genet. 2012 May 15;13(6):421-8. doi: 10.1038/nrg3239. (https://www.researchgate.net/publication/224957838_Mechanisms_and_consequences_of_widespread_random_monoallelic_expression?ev=sim_pub)

Article Widespread Monoallelic Expression on Human Autosomes

Article Mechanisms and consequences of widespread random monoallelic...

In my opinion, there are a multiple factors that influence the activity of a genes.

Such would be the cell cycle point or differentiation state of a cell/tissue. Resulting differences in the proteome and chromatin state would influence the avalability of the alleles. Maybe a cell cycle synchronization of the in vitro culture used could help.

Besides, there are different classes of genes. Some genes (e.g some housekeeping genes) are required for a specific amount of a gene product and thus leed to haploinsufficiency if an allele get's non-functional.

Maybe I should clarify. In dividing cells in culture, protein X is seen in >90% of the cells. The transcripts for protein X is seen in >80% of the cells. What percentage of cells is actively transcribing? This can be in the case of a housekeeping gene.

For the vast majority of autosomal genes, expression occurs from both alleles simultaneously. In mammals, however, a small proportion  of genes are imprinted, meaning that gene expression occurs from only one allele. The expressed allele is dependent upon its parental origin.   

There is also a phenomenon called Nonsense Mediated Decay, where an inappropriately located stop-codon can lead to the silencing of the allele carrying this alteration, leaving the other allele to be the sole source of coding information for protein synthesis.

For haplotype insufficient genes, they are generally transcibed at both alleles in a normal cell population. However, that does not mean that both alleles should be transcribed in all cells at the same time point. What you observed may exactly reflect the expression status of a haplotype insufficient gene in cell population.

Usually, single allele only expressed in a cell. In some cases, different forms (types) of genes might be present in a cell, so that more than one allele may be expressed in a single cell.

Genes are of course expressed stochastically (see papers from Singer lab), but still 20% is low. Do you see closer to 80 or 90% for a positive control primary transcript FISH, such as actin? If so, then either the FISH probes for the 20% expressor are bad, or you really are seeing very low expression. If that is the case, the mRNA half-life must be very long, if you are seeing >80% processed transcript in the cells. Are you using BACS or oligos for your primary transcript FISH?

I add that yes, the dogma is that both alleles for most active "normal" genes are expressed, but expressed stochastically.

Wow. I think this is very interesting. I wouldn't be surprised if single alleles of non-imprinted genes are expressed sometimes, both at other times, and, why not neither at other times?. I can imagine dynamic changes in the epigenome and interchromosomal interactions that could lead to single allele expression. The only way I can think of to verify this is single cell RNAseq- at sufficient depth to make the quantification valid.

I guess, however, you'd better be sure that the cells you're using are genomically homogeneous and stable...a subpopulation with a deletion..or a nonsense mutation...could give you the same result.

Good point. What cells are you looking at?

Other than imprinted genes or X-inactivated genes, there are autosomal genes that show random allelic expression pattern. Within a given clonal cell line, the expressed allele can be maternal, paternal or both [1]. Approximately 5-10% of the genes in the human genome behave this way, which can also be tissue-specific [1]. So depending on the location of your gene and the tissue of interest, the gene may fall into this category. There is also a phenomenon called random allelic bias, in which one of the alleles is preferentially expressed, but the extent is not as well explored. More information can be found in the review by Chess [2].

[1] Gimelbrant et al. Science. 2007 Nov 16;318(5853):1136-40. (https://www.researchgate.net/publication/5839816_Widespread_monoallelic_expression_on_human_autosomes?ev=sim_pub)

[2] Chess. Nat Rev Genet. 2012 May 15;13(6):421-8. doi: 10.1038/nrg3239. (https://www.researchgate.net/publication/224957838_Mechanisms_and_consequences_of_widespread_random_monoallelic_expression?ev=sim_pub)

Article Widespread Monoallelic Expression on Human Autosomes

Article Mechanisms and consequences of widespread random monoallelic...

Are you looking at an established, immortalized cell line? If so, the number of chromosomes or alleles can vary greatly and is usually represented by a distribution. The ATCC catalog gives an idea of what to expect. For example 3T3 cells range anywhere from 28 to more than 70 chromosomes. Also, gene expression maybe affected by the cell cycle, in addition to stochastic effects.

Sounds like an interesting set of data, but maybe not surprising.

Most regulated transcription probably occurs in pulses, with transcriptional bursts followed by promoter clearance before the next round of transcription. There is a good paper from Gannons lab (Metivier Cell 2003), detailing this for an ER regulated gene.

Both of your alleles may be active, but not oscillating in a synchronised manner. In principle a snapshot of an active gene may show two, one or no active allele, with the frequency of observing two, one or zero dots depending on how that particular gene is regulated.

You can also look at a review by Hager, McNally and Misteli 2009 on transcription dynamics.

To me it sounds very interesting, but more than an enswer I have questions. Have you sequenced the promoter regions of both alelles? is it possible a polimorfism on the promoter region (might be conditioning allele expresion)?

To give you an idea of how a normal , homogeneous cell population may acts in vitro you can take embryonic stem cells as an good example.

Check out this recent paper which describe nicely the allelic expression of key pluripotency genes in ES, as an example for your question:

http://www.cell.com/cell-stem-cell/abstract/S1934-5909(13)00155-0

Overall they claim that nanog and other key genes of plutipotency are biallelically expressed.

I appreciate all the answers but I completely agree with Ryan Yuen's opinion. Apart from the technical issues discussed above such as kind of cell line used for FISH, I believe in a given point of time a cell would not use the entire transcription machinery to express all the non-imprinted genes (both alleles of autosomes). It completely depends on the necessity of the gene's protein function. It might be tissue specific for sure. But what you are observing in your FISH experiments might truly reflect the scenario in the In Vivo situation. Particularly its true in case of developmental genes and CNS specific genes.I feel not more than 25% of the autosomal alleles show biallelic expression in a given point of time. I got one reference to support this idea.

I appreciate any comments on my opinion.

1) http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0013843

Thanks

I may sound like a total idiot in here...but I am so interested that I must ask.

I assume it is RNA FISH otherwise it's not about transcription? and you are talking cells that are heterogenous, so the two different allele has different sequence and are probed with different fluorophore?

Or am I in a totally wrong direction now?

I am just curious on how the probes can differentiate the two allele if they have the same sequence as I am no expert in RNA-FISH.

@ Praveen--really excellent article. A big help.

@ Nick Li, I was referring to FISH using intronic probes that give you one dot for each allele in the nucleus being used. 2 dots=2 alleles active, 1 dot= one allele.

@ everyone else, thanks for your answers. If anyone had 1st hand experience, let me know?

Here is the setting we have been using. We looked at the PW1/Peg3 gene which is expressed in all adult stem cells (Besson et al, PNAS, 2011). We have obtained neurospheres to ask if a subset of cells express both alleles (as the gene is normally expressed uniquely off of the paternally inherited allele). We found a small number that are bi-allelic but only 10% of the PW1+ cells (for protein) positive for either 1 or 2 alleles. The % was higher for more 'standard genes' (i.e. nestin) but still only 20 to 40%. This was surprising but then I realized that this was a pretty wide open question thus any expectations for an outcome may be quickly dashed!

@ David, You are welcome. In fact, its a very interesting question posed by you. Given the advent of NGS technology such as RNA seq etc, I think this opinion becomes more clear to the people who work on autosomal gene expression. Thanks again.

Hi David, given the further information you just provided, my answer earlier seems not be entirely relevant (especially when you are referring to stem cell). I don’t have a 1st hand experience on this matter, so I can’t comment on the expected % of positive cells. On the other hand, I think the following two papers should give you an idea of what to expect for the cell proportion of the allelic expression for non-imprinted genes in stem cells (although it is in ESC):

http://www.nature.com/nature/journal/v483/n7390/full/nature10807.html

http://www.nature.com/nmeth/journal/v10/n9/full/nmeth.2601.html

Basically, there should still be a small proportion of cells showing monoallelic expression even the gene is non-imprinted.

As for the imprinted gene PW1/Peg3, which may be irrelevant, but I was wondering if you have used allele-specific RT-PCR or pyrosequencing for the cDNA to check the allelic ratio on gene expression instead? This may be more informative/quantitative.

@Ryan--indeed we used straightforward sequencing using SNPs for crosses and did NOT see the maternal allele, however, this technique is not too sensitive. We will try pyrosequencing now but if it is only 1% of the total cells harvested, I am not sure it will work which is why we did a cell-by-cell analysis with FISH.

This new article will give some insight about stochastic monoallelic expression.

http://www.sciencemag.org/content/343/6167/193.full

Just to add to the list of (presumably rapidly growing) papers analysing mono-allelic expression: http://elife.elifesciences.org/content/2/e01256

The authors look into the chromatin marks of mono-allelically expressed genes, so perhaps that would be additional information you could use to test your cells.

No, in many cases only one allel is transcribed. In our experience, for example in stress reactions in plants, it is mandatory that only one and the right allel is active.

@ Peter...I did not know this about plants. So what happens in the case of a mutant allele? How does the plant cell 'know' to use the good one?

Hi,

seems to be a question of selection. If you express the wrong one you are probably dead or will not become fertile. Using MACE (NGS-based quantification of 3'-ends of transcripts, see www.genxpro.de) for example, we can see in bulks of "good" versus "bad" offspring the segregation of gene expression of both alleles. Very often you see only one allel expressed in one bulk, the other allel or both alles in the other bulk.

I could send you an example privately.

Control could come via methylation/histone modification/other forms of epigenetic control. The same may be true for human diseases as e.g. in cancer though I very rarely saw such a clearcut yes/no distinction as in plants. May be the tumor cells are too heterogeneous.

And what about HLA (self / nonself recognition) region?

Regardless of imprinting allelic differences in expression are to be expected and are indeed common:

http://www.ncbi.nlm.nih.gov/pubmed/23416756

http://www.ncbi.nlm.nih.gov/pubmed/23913259

If you have need for gene specific probe set sleet me know hi we can help.

Hans

It's complicated. One issue is that most of our detection methods are possibly affected (in terms of detection efficiency) by the allele-specific sequences that permit detection in the first place. Add heterozygous copy number variation into the mix and one needs to do some pretty robust controls to make sure that what appears to be allele-specific expression, really is.

The second consideration is the time resolution of the question. Single-cell transcription studies have shown that expression occurs in stochastic bursts, which are asynchronous across alleles. However, the total expression each allele over a long period might be the same. If your FISH experiments capture a moment in time, they could well be picking up these stochastic bursts of transcription. Is it possible to do a high resolution time course? You might see different alleles flashing on and off. I don't think anyone really knows the answer to your question, so one question then becomes: can your technique itself help to provide the answer?

What about HLA genes?