Gene expression is a stochastic process due to fluctuations in the complex stoichiometry and reaction kinetics of the biochemical reactions, which leads to substantial cell-to-cell variability. The resulting phenotypic fluctuations can only be detected and quantified at the single-cell level within isogenic populations, as shown in our 2002 paper (Ozbudak et al., Nature Genetics). This paper is among the first multidisciplinary studies of gene expression noise (cited more than 1000 times) and has triggered the blossoming of the “stochastic gene expression” field.

One of the most intriguing questions in science is how developmental pattern formation is executed so robustly, despite unavoidable fluctuations in gene expression. This precision necessitates several mechanisms buffering stochastic gene expression. Because of technical difficulties posed by quantitative single-cell measurements, few studies to date have quantified stochastic gene expression in multicellular systems during development, when buffering the process is critical for the precise and reproducible development of an adult organism.

Vertebrae are derived from embryonic segments called somites. Periodic segmentation of somites, somitogenesis, is controlled by a gene expression oscillator called the vertebrate segmentation clock. When somite segmentation goes awry, it results in birth defects such as scoliosis. This system is one of the best examples of robustness of pattern formation during embryonic development.

In this manuscript, we have investigated the magnitude, the underlying causes of and the mechanism limiting cell-to-cell variability in the vertebrate segmentation clock during zebrafish embryonic development. Many studies of stochastic gene expression have been performed with synthetic regulatory networks or promoter reporters for natural networks by using long-lived GFP mRNA and protein. However, many mRNAs coding for critical transcription factors governing embryonic development and adult homeostasis are short-lived. In this study, by examining short-lived segmentation clock RNAs (t½=3 minutes), we investigated gene expression noise for the first time at the fastest-dynamic scale (30 minutes oscillation period) in an intact vertebrate tissue.

We discovered that clock genes have low RNA amplitudes, expression variability is driven by gene-extrinsic sources, and expression noise increases along the axis from posterior progenitor zone to anterior segmentation/differentiation zone. We further showed that cell-to-cell coupling via Notch signaling suppresses expression variability. Our results will shed light on the accuracy of natural clocks in multicellular systems and inspire engineering robust synthetic oscillators.

Oscillations are widespread in biological systems. Notch signaling plays critical roles in controlling the switch from proliferation to differentiation in almost every tissue throughout the metazoan. Their family segmentation clock genes are direct targets of Notch signaling. Hes/Her segmentation clock was the first discovered developmental oscillator. However, Hes/Her protein levels also oscillate in neural progenitor cells, embryonic stem cells, and ovarian cells, thereby controlling the switch from proliferation to differentiation in a number of tissues. Furthermore, Hes/Her proteins are highly expressed in several types of human tumors, and their inhibition has been shown to restore cell differentiation. We anticipate that similar future studies may help to develop new ways of controlling stem cell proliferation and differentiation or new therapies against certain cancer types.