View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by Elsevier - Publisher Connector Developmental Cell Forum

Tuning In to Noise: and Intangible Variation

Nadia C. Whitelaw,1,2 Suyinn Chong,1 and Emma Whitelaw1,* 1Epigenetics Laboratory, Queensland Institute of Medical Research, Brisbane, Queensland 4006, Australia 2School of , University of Queensland, Brisbane, Queensland 4001, Australia *Correspondence: [email protected] DOI 10.1016/j.devcel.2010.11.001

In this special issue of Developmental Cell, we discuss the role of in phenotypic variation as a coun- terpoint to the reviews on chromatin dynamics in development and cancer. We highlight some recent work on the role of chromatin in transcriptional noise in yeast and Caenorhabditis elegans and consider the implica- tions in understanding intangible variation or developmental noise in mammals.

The myriad processes that make us unique involved, but here we are focusing on the signaling components were compared to as individuals lie at the heart of biology. role of chromatin. the when RNAi was used to Molecular biologists and geneticists have Modern molecular techniques, in partic- knockdown each of 1860 genes encoding used classic approaches to tease out the ular the development of single-cell tran- with signaling, chromatin or heritable component of , but scription assays, have given us greater factor functions, i.e., they we have made less progress in unraveling insight into these processes. The first compared 65,000 combinations. The the nonheritable component. The idea that single-cell study of transcriptional noise results showed that knockdown of certain probabilistic or chance events in early in used a two-reporter-gene chromatin genes (e.g., orthologs of the development play a role in the latter is strategy in diploid yeast. Both cyan and NuA4/Tip60 histone acetyltransferase rarely considered, despite long-standing yellow fluorescent proteins (CFP and complex or the nucleosome remodeling evidence from embryologists and natural- YFP) were inserted into the same site on and histone deacetylase [NuRD] complex) ists. In undergraduate textbooks this is homologous chromosomes and each clearly had the greatest effect on en- called ‘‘developmental noise’’ or ‘intan- was driven by the same PHO5 hancing the phenotypes of the mutant gible variation.’’ In essence, highly inbred sequence (Raser and O’Shea, 2004). This strains. They described these chromatin animals, ostensibly isogenic, reared in system allowed the researchers to mea- genes as network ‘‘hubs’’ and showed tightly controlled environments, show a sure extrinsic noise (affecting the expres- that knockdown of these hubs uncovered surprisingly broad range of phenotypes sion of both reporters) and intrinsic noise a broad range of mutant phenotypes. for many measurable traits. (affecting only one of the reporters). The These hub genes interacted with over Interestingly, Waddington’s classic ability to measure intrinsic noise was vital; one-quarter of the signaling mutant genes papers on epigenetics were stimulated in they were able to resolve intrinsic fluctua- from multiple different signal transduction part by his curiosity about developmental tions in expression due to inefficient pathways, suggesting that they can accidents, or ‘‘noise,’’ during the growth promoter activation that was not evident modulate many different processes. The of a complex organism—hence his in- when transcripts were studied from pooled authors propose that chromatin hub terest in asymmetric development or fluc- cells, in which the levels averaged out. genes modulate the phenotypic conse- tuating asymmetry (Waddington, 1957). They found that reduction in the levels of quences of mutation to a large number of Even before the structure of DNA was chromatin remodeling factors (i.e., compo- genes. So, in this system, chromatin modi- elucidated, Waddington foresaw that the nents of SWI/SNF, INO80 or SAGA) that fiers appear to be functioning as general bridge between genotype and phenotype were known to act at the PHO5 promoter buffers of genetic variation. must be complex. Over 50 years later, resulted in increased intrinsic noise, indi- Raj and colleagues, also working with evidence for a role for stochastic events cating that faithful epigenetic gene regula- C. elegans, carefully measured the tran- in development is accumulating and our tion buffers against noise arising from scriptional changes underlying incomplete understanding of the molecular events sluggish promoter transitions. penetrance of mutations affecting intes- underlying developmental noise is be- A role for chromatin in suppressing tinal specification (Raj et al., 2010). The coming clearer (Losick and Desplan, phenotypic variation has also emerged strength of their study lies in the simplicity 2008). Recent studies on transcription in from studies in C. elegans. RNA interfer- of the system; the C. elegans intestine is yeast and Caenorhabditis elegans have ence was used to analyze the phenotypic made up of 20 cells and a well-defined revealed that ‘‘transcriptional noise’’ is consequences of knocking down paired network of only six genes determines regulated by proteins involved in estab- combinations of genes annotated as intestinal cell specification. Mutation of lishing and maintaining the epigenome. having ‘‘signaling,’’ ‘‘chromatin,’’ or ‘‘tran- one of the genes in the pathway, skn-1,is Other systems of buffering cellular noise scription factor’’ function (Lehner et al., embryonic lethal but with incomplete such as or the molecular 2006). The phenotypes of 37 mutant penetrance; in some cases there is no chaperone Hsp90 are also likely to be strains that carry mutations in different intestinal cell specification and in some

Developmental Cell 19, November 16, 2010 ª2010 Elsevier Inc. 649 Developmental Cell Forum

cases there are intestinal cells present in and gives very efficient gene activation. to achieve mutually exclusive expression embryos. By counting transcripts (using When the site was placed within a nucleo- is challenging. One can view this process a version of FISH that makes each mRNA some, expression became bimodal, i.e., as a cell fate decision; each neuron is molecule visible as a single fluorescent an on/off pattern was observed in single- making a different decision. The choice to spot) of various downstream genes in cell assays, indicating stochastic binding express a single olfactory is mutant and wild-type embryos, they of the (Bai et al., stochastic and, as such, can be considered showed that the variation in mutant pheno- 2010). Bimodal expression was also ob- transcriptional noise. Cis-acting sequences types is caused by large variations in the served at an endogenous gene in which and feedback mechanisms ensure that transcript levels of an intermediary gene the transcription factor binding site is natu- only one receptor is expressed (Lomvardas (end-1), which must reach a threshold level rally embedded in a nucleosome. Further- et al., 2006; Serizawa et al., 2003). before it can activate a third gene that acts more, the probability of a cell being ‘‘on’’ At the root of any change in cell fate is as a switch to intestinal cell fate. skn-1 acti- was changed by altering the dosage of a a single event that triggers a cascade of vates end-1 by chromatin remodelling and SWI/SNF remodeler or a histone deacety- subsequent changes. It may well turn the authors hypothesize that inefficient lase. This study suggests that the preserva- out that the capacity for some gene recruitment of the chromatin remodelling tion of nucleosome-depleted regions in promoters to act in a bimodal fashion is machinery underlies the variable expres- promoter elements is a widely conserved a fundamental requirement of multicellu- sion levels. In summary, this study shows epigenetic mechanism that suppresses larity. The study by Raser and O’Shea on that wild-type offspring have epigenetic transcriptional noise. It is interesting that intrinsic transcriptional noise in yeast mechanisms that control large fluctuations some functional promoters lack a nucleo- identified both cis- and trans-acting in and that if these mech- some-depleted region. Could these lie factors that alter the level of noise, sug- anisms are disrupted, stochastic transcrip- adjacent to genes for which bimodal or gesting that noise is an evolvable trait tion in the cell can lead to altered cell fates. variable expression is advantageous? that can be optimized to balance fidelity A somewhat different approach has Intangible variation (developmental and diversity at each gene. It is difficult been taken by Choi and Kim. They have noise) has been studied mainly in in outbred populations, such as humans, used a single-cell proteomic analysis in Drosophila and mammals. In these organ- to measure the contribution of such yeast to study cellular noise and also found isms, in which developmental processes processes to what makes us each unique, that it is largely controlled by chromatin are complex, it will be difficult to design but the idea that transcriptional noise is regulation (Choi and Kim, 2009). They methods of counting transcripts in indi- involved is worthy of consideration. found that genes with the most variable vidual cells at critical stages of a develop- expression (among cells) have a common mental pathway, as has been done in REFERENCES sequence in their promoters that en- worms. Nevertheless, there is every hances the flexibility of the DNA and reason to believe that the same under- Ashe, A., Morgan, D.K., Whitelaw, N.C., Bruxner, attracts nucleosome occupancy. Others lying principles occur. Consistent with T.J., Vickaryous, N.K., Cox, L.L., Butterfield, N.C., Wicking, C., Blewitt, M.E., Wilkins, S.J., et al. had shown previously that nucleosome this, we have observed increased pheno- (2008). Genome Biol. 9, R182. occupancy in these regulatory regions typic noise among inbred littermates, in Bai, L., Charvin, G., Siggia, E.D., and Cross, F.R. prohibits transcription and so nucleosome mice haploinsufficient for chromatin (2010). Dev. Cell 18, 544–555. removal is required for expression. The regulators (N.C.W., unpublished data). authors found that chromatin modifier Similarly, a screen for modifiers of epige- Choi, J.K., and Kim, Y.J. (2009). Nat. Genet. 41, 498–503. levels (e.g., chromatin remodelers, histone netic reprogramming, also carried out in acetyltransferases, deacetyltransferases, an inbred background, identified a num- Lehner, B., Crombie, C., Tischler, J., Fortunato, A., and Fraser, A.G. (2006). Nat. Genet. 38, 896–903. and methyltransferases) had a substantial ber of mutant strains that displayed effect on variability, whereas transcription stochastic death; some mutants died Lomvardas, S., Barnea, G., Pisapia, D.J., Mendel- sohn, M., Kirkland, J., and Axel, R. (2006). Cell 126, factor levels did not (Choi and Kim, and others did not (Ashe et al., 2008). 403–413. 2009). Thus, the variable expression at This can be considered an extreme form these promoters is initially encoded in the of phenotypic noise. Although reports of Losick, R., and Desplan, C. (2008). Science 320, 65–68. promoter sequence, but is ultimately incomplete penetrance in mouse colonies directed by epigenetic processes. are not uncommon, mixed genetic back- Raj, A., Rifkin, S.A., Andersen, E., and van Oude- Numerous genome-wide studies on grounds often make it difficult to rule out naarden, A. (2010). Nature 463, 913–918. nucleosome positioning in eukaryotes underlying genetic explanations. Raser, J.M., and O’Shea, E.K. (2004). Science 304, have found that most promoters have We should not forget that there are 1811–1814. nucleosome depleted regions that lie situations in which transcriptional noise Serizawa, S., Ishii, T., Nakatani, H., Tsuboi, A., upstream of the transcriptional start site. will be advantageous; the development of Nagawa, F., Asano, M., Sudo, K., Sakagami, J., Sakano, H., Ijiri, T., et al. (2000). Nat. Neurosci. 3, A recent study in yeast shows that nucleo- mammalian olfactory neurons, each with 687–693. some occupancy can govern the proba- a specific receptor, is one stunning bility that a gene will be expressed. The example. Each olfactory sensory neuron Serizawa, S., Miyamichi, K., Nakatani, H., Suzuki, M., Saito, M., Yoshihara, Y., and Sakano, H. authors embedded a transcription factor expresses only one allele of one type of (2003). Science 302, 2088–2094. binding site within a nucleosome-bound olfactory receptor (Serizawa et al., 2000). Waddington, C.H., ed. (1957). The Strategy of region of the promoter; normally, the site There are over 1000 olfactory receptor the Genes: A Discussion of Some Aspects of is located in a nucleosome-depleted region genes spread throughout the genome, so Theoretical Biology (London: Allen & Unwin).

650 Developmental Cell 19, November 16, 2010 ª2010 Elsevier Inc.