Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Cell. Author manuscript; available in PMC 2010 September 5.
Published in final edited form as:
PMCID: PMC2848716

A Penetrating Look at Stochasticity in Development


In recent work published in Nature, Raj et al. (2010) use single mRNA molecule quantification to show that variation in gene expression in Caenorhabditis elegans increases in mutants displaying incomplete penetrance. They find that a bimodal response is triggered when noisy expression of an upstream regulator crosses a critical threshold.

Robustness to genetic and environmental variation is an essential feature of all biological systems. In recent years, it has become possible to address the molecular mechanisms that ensure the reproducible outcomes of biological processes. These mechanisms often break down in mutant conditions, leading to variable outcomes. In genetic terms, penetrance refers to the proportion of individuals of a particular mutant genotype displaying a mutant phenotype. In a paper that recently appeared in Nature, Raj and colleagues examine mutant backgrounds of the nematode Caenorhabditis elegans to explore the role of variable gene expression in incomplete penetrance (Raj et al., 2010). Their analysis provides evidence that mutant conditions increase noise in gene expression. They further show that the downstream genes in affected pathways respond to these variable inputs at certain thresholds and that chromatin has a role in modulating variability of gene expression.

Several groups have evaluated well-defined genetic circuits with new and highly quantitative methodologies to derive fundamental principles about the nature of these circuits and of gene regulation in general (for instance, Cağatay et al., 2009; Gregor et al., 2007; Maamar et al., 2007; Mangan et al., 2003; Süel et al., 2006, 2007). Working in a similar vein, Raj and colleagues have now sought to address the mechanisms controlling incomplete penetrance using a technique that allows the counting of single messenger RNA (mRNA) molecules in vivo. They chose to evaluate the genetic network that determines intestinal identity in C. elegans. As illustrated in Figure 1A, this circuit is a meshwork of coherent feedforward loops among transcription factors (for a review, see McGhee, 2007). Coherent feedforward loops delay the response of the most downstream element (Mangan et al., 2003) and thus may contribute to the temporal control of the circuit (Figure 1B). Specifically, skn-1 activates expression of med-1/med-2 (considered together as a single node), end-3, and end-1. med-1/med-2 activate end-3 and end-1, whereas end-3 activates end-1. Additionally, pop-1 mediates a WNT signal to upregulate expression of end-3 and end-1. Together, end-3 and end-1 function in an OR gate (that is, the two genes are redundant) to activate elt-2. elt-2 reinforces this input through a positive feedback loop. As a convergent node of this network motif, elt-2 acts to upregulate hundreds of genes required for intestine fate (Figure 1A).

Figure 1
Gene Expression Variability and Incomplete Penetrance

The authors focus their analysis on the variation in gene expression that occurs in mutants in the most upstream member of this pathway, skn-1, which exhibit intestinal differentiation defects with incomplete penetrance. In wild-type animals, levels of med-1/med-2, end-3, and end-1 mRNA peak with distinct and robust temporal dynamics, culminating in the expression of elt-2 at a sustained maximum (Figure 1B). In skn-1 mutants, expression of med-1/med-2 is completely absent, whereas end-3 levels are minimal. The single mRNA molecule detection technique allows the authors to observe noisiness in end-1 expression that appears to be the only variable input into elt-2 in this mutant condition. This variation is resolved at the level of elt-2 expression, which displays a bimodal ON/OFF distribution of mRNA levels (Figure 1B).

The bimodal response in elt-2 expression suggests that a thresholding mechanism controls this phenomenon. In other words, when end-1 reaches a specific level, the elt-2 positive feedback loop is triggered, resulting in an ON/OFF response. To test this, Raj and coworkers evaluated the relative mRNA levels of end-1 and elt-2. They find that at low end-1 levels, elt-2 expression is never observed. Nevertheless, skn-1 mutant worms with high end-1 mRNA levels similar to wild-type are capable of triggering the elt-2 positive feedback loop. However, some worms exhibiting high levels of end-1 do not trigger elt-2 expression, arguing against a simple thresholding model. The investigators provide an explanation for this apparent contradiction. In null end-3 mutants, end-1 levels remain similar to wild-type levels, but expression of elt-2 is often not induced. Given that end-3 also feeds into elt-2 regulation, this observation suggests that end-3 allows for high end-1 levels to efficiently induce elt-2 expression. In skn-1 mutants, end-3 is nearly absent, which explains why high levels of end-1 are not always capable of elt-2 induction. The frequency of high end-1 levels with low elt-2 levels within embryos is greater than the actual penetrance of intestinal defects. On the basis of this observation, the authors conclude that high levels of end-1 are likely activating elt-2 in most cases, but there is a temporal delay in induction. Given that the single mRNA counting technique only allows for static observations, new methodologies will be needed to further test this hypothesis.

Strangely, for one skn-1 mutant allele, elt-2 expression is activated even at low levels of end-1. This is peculiar, given that skn-1 is the most upstream regulator in this network and is not known to directly regulate elt-2. Therefore, it is surprising that a different type of mutation in this upstream gene would modulate a gene that it only indirectly regulates. This observation suggests that skn-1 probably regulates elt-2 either directly on its promoter or via an unknown mediating factor (Figure 1A). Alternatively, the worms with this particular skn-1 allele may carry a background mutation that affects the threshold response.

Raj and colleagues also examine the role of hda-1, a chromatin regulator that is a repressor of end-1 expression (Calvo et al., 2001) (Figure 1A). The authors conclude that knockdown of hda-1 levels with RNA interference (RNAi) in skn-1 mutants leads to upregulation of end-1 transcripts and a decrease in its variability. Indeed, the coefficient of variation for end-1 levels does decrease with RNAi knockdown of hda-1 in skn-1 mutants as compared to the variation in skn-1 mutants alone. However, this is still markedly higher than the variation observed in the wild-type. It appears that the double-mutant phenotype is complex. Not only is end-1 derepressed, but the variation is also partially limited. Is the partial decrease in variation simply a byproduct of derepression of this gene? Or do multiple mechanisms feed into the control of variation? Or both? These questions are complex, but this paper lays the groundwork for addressing them.

The work of Raj and colleagues begins to address the mechanisms that cause incomplete penetrance. Their highly quantitative single molecule approach is new to developmental biology, which typically makes use of reporter transgenes, antibody staining, and in situ hybridization to assess gene expression. With this methodology, Raj et al. clearly show that variation in expression occurs in specific mutant conditions and that the architecture of a developmental network is able to compensate for noisy expression. However, one of the most challenging problems in the field of gene expression (in wild-type or mutant conditions) is to identify the source of transcriptional stochasticity. Although a role for chromatin state is proposed in this paper, this is hardly surprising given that its regulation is so fundamental to gene expression in general. The next challenge will be to show how these alterations affect variability in gene expression at individual loci.

Robustness compensates for variation caused by the stochastic low level expression of key regulators. However, mechanisms that ensure robustness also provide a buffer in the wild-type, which allows for the evolution of new regulatory interactions (for a review, see Masel and Siegal, 2009). It will be exciting to determine not only how variation occurs due to the break down of wild-type biological programs but also how novel cryptic modes of regulation are revealed when robustness mechanisms are impaired.


  • Cağatay T, Turcotte M, Elowitz MB, Garcia-Ojalvo J, Süel GM. Cell. 2009;139:512–522. [PubMed]
  • Calvo D, Victor M, Gay F, Sui G, Luke MP, Dufourcq P, Wen G, Maduro M, Rothman J, Shi Y. EMBO J. 2001;20:7197–7208. [PubMed]
  • Gregor T, Tank DW, Wieschaus EF, Bialek W. Cell. 2007;130:153–164. [PMC free article] [PubMed]
  • Maamar H, Raj A, Dubnau D. Science. 2007;317:526–529. [PMC free article] [PubMed]
  • Mangan S, Zaslaver A, Alon U. J. Mol. Biol. 2003;334:197–204. [PubMed]
  • Masel J, Siegal ML. Trends Genet. 2009;25:395–403. [PMC free article] [PubMed]
  • McGhee JD. The C. elegans intestine. In: The C. elegans Research Community, editor. WormBook. 2007. 10.1895/wormbook.1.133.1. [PubMed]
  • Raj A, Rifkin SA, Andersen E, van Oudenaarden A. Nature. 2010 Published online February 18, 2010. ■■■. [PMC free article] [PubMed]
  • Süel GM, Garcia-Ojalvo J, Liberman LM, Elowitz MB. Nature. 2006;440:545–550. [PubMed]
  • Süel GM, Kulkarni RP, Dworkin J, Garcia-Ojalvo J, Elowitz MB. Science. 2007;315:1716–1719. [PubMed]