Microarrays have been used in the study of circadian expression of mammalian genes since 2002 and the consensus was built from those studies that approximately 15% of all the genes expressed in a cell are expressed in a circadian manner. I always felt it was more, much more.
I am no molecular biologist, but I have run a few gels in my life. The biggest problem was to find a control gene - one that does not cycle - to make the comparisons to. Actin, which is often used in such studies as control, cycled in our samples. In the end, we settled on one of the subunits of the ribosome as we could not detect a rhythm in its expression. The operative word is "could not detect". My sampling rate was every 3 hours over a 24-hour period, so it is possible that we could have missed circadian expression of a gene that has multiple peaks, or a single very narrow peak, or a very low amplitude of cycling (it still worked as a control in our case, for different reasons). Thus, my feeling is that everything or almost everything that is expressed in a cell will be expressed in a rhythmic pattern.
If you have heard me talk about clocks (e.g., in the classroom), or have read some of my Clock Tutorials, you know that I tend to say something like "All the genes that code for proteins that are important for the core function of a cell type are expressed in a circadian fashion". So, genes important for liver function will cycle in the liver cells, genes important for muscle function will cycle in muscle cells, etc.
But I omit to note that all such genes that are important for the function of the cell type are all the genes that are expressed in that cell. The genes not used by that cell are not expressed. But I could not go straight out and say "all the genes that are expressed in a cell are expressed in a circadian pattern", because I had no data to support such a notion. Until yesterday.
What happened yesterday?
A paper came out in PLoS-Computational Biology that demonstrates that all genes that are expressed in a cell are expressed with a circadian rhythm (check out press releases here, here, here and here):
Digital Signal Processing Reveals Circadian Baseline Oscillation in Majority of Mammalian Genes by Andrey A. Ptitsyn, Sanjin Zvonic and Jeffrey M. Gimble.
Sampling of tissue for microarray studies (and actually for all molecular techniques looking at circadian patterns of expression) is usually done every 4 hours over a 24-hour period. Noise, errors, low amplitude of cycling of some genes, possibility of missed peaks or multiple peaks - all conspire to lower the estimate of the number of oscillating genes in a tissue. Also, sampling in constant conditions, in which cells of a peripheral oscillator may drift out of phase with each other, can further lower the estimates.
In this paper, authors use several different mathematical methods to analyze multiple previously published datasets, including one of their own made in a light-dark cycle. What is really new is the way they think about it:
The biological pathways and interaction networks are typically presented in a way similar to the graphic depiction of direct current electric circuits. It is assumed that up- or downregulation of a particular component of a biological pathway causes changes in gene expression of downstream elements of the same pathway in a way similar to a direct current electric circuit. We believe that this presentation does not reflect the complex nature of gene expression and omits an important property of biological systems. Biological pathways and interaction networks should be viewed and modeled in a way more similar to the alternating current circuits, in which phase and amplitude are important characteristics of function of each component. Timing is important for understanding the effect of each biological signal as well as the waves spreading through the system perturbed by signal. Like in alternating current circuits, the likely primary source of oscillation is the "energy source" of a cell, the respiratory cycle. The circadian molecular clock based on negative feedback plays the important role of temporary timekeeping, synchronizing oscillation in gene expression levels with the major environmental factors such as daylight. The commonality of oscillation in gene expression reported in this paper suggests that oscillation is a natural and important feature of all or nearly all biological pathways. Genes that display daily variation in their expression level may or may not be linked directly to the circadian molecular clock. Regardless of mechanism, the oscillation of these genes and their encoded protein products will affect all elements of the system as a whole. Hence, the same "alternating current" principle could be applied in modeling all biological processes in mammals and, possibly, in other organisms.
And lo and behold, they found that 99-100% of all genes expressed in a cell oscillate with a circadian period. Yes, some amplitudes are low and some oscillations have multiple peaks, as expected. But the amplitudes (with a few remarkable exceptions) do not differ between cell types. All the genes can be classified into a very small number of groupings by phase and it is the phase that differs markedly between different cell types. While I do not fully understand the math they used in this study, I advise you to go and read their results and discussion, especially concerning the coupling between energetics of the cell and its rhythmic behavior.
So, if this paper is correct, my 'gut feeling' on this issue has been vindicated and I can make my proclamations about it more boldly in the future as I have a reference to back me up.
Just to look at humans, some important things are one shot only:
- sexual attractiveness
- fluid IQ
Looks like it will be a really interesting paper.
I have been frustrated with many biologists over the last few years that have been doing expression array studies and have not taken into account the time of day that the samples are taken. It seems to me that this is an enormous problem. If there is any pattern to the time of day different samples were taken you could have a huge inflation of gene expression differences found on an array. We have been citing the Drosophila array results for a while now as evidence that circadian cycling should be taken into consideration in ALL microarray experiments and I will add this paper to the list (once I have read it of course :) ).
Thanks for the post!
I've been wondering about this ever since I've seen the data from the malaria work (The transcriptome of the intraerythrocytic developmental cycle of Plasmodium falciparum.
Bozdech Z, Llinás M, Pulliam BL, Wong ED, Zhu J, DeRisi JL.- PloS Biol. 2003 Oct; 1(1):E5).
Llinas has presented some awesome pictures of the transcriptional cascade one sees for malaria gene expression... I've always wondered if there's some way to meaningfully quantitate the "cyclical nature" of an orgnaism's transcriptome. Is malaria unusual? Your post would seem to suggest not. Yet I would actually be suprised if some organisms weren't more strictly tied to their rhythms, just because they seem to have such differences in how they live (and how much their eating and sleeping schedules can vary- that has got to affect the cellular metabolism in the long run, right?).
Looking over the paper a little, I can't tell if they even looked beyond mice- certainly many of their data sets are murine. Isn't this a serious limitation? Circadian rhythms are an incredibly blaring example of "mice are not furry people".
There is an extensive analysis of the Arabidopsis transcriptome in many rhythmic and non-rhythmic environmental conditions in this psper: Michael et al. PLoS Genet. 2008 February; 4(2): e14.
The authors conclude that 89% of Arabidopsis transcripts cycle in at least one condition and that most genes have peak expression at a particular time of day, which shifts depending on the environment.