Research

I. Summary

        During my graduate training, I have been interested in exploring the dynamics in stochastic cellular processes at the single cell level, using both quantitative measurements and theoretical models. With a live single cell imaging and tracking method that I developed in the past, I have been able to investigate several dynamic systems in a variation of organisms including the yeast S. cerevisiae, and the bacteria S.elongatus and R. opacus.   

        Recently, I started my journey towards understanding the burst-like nature of transcription kinetics with single molecule resolution in higher eukaryotic cells, using the newly developed single molecule RNA FISH tenichque in our lab.

        In the following sections, I will describe in detail each of my projects.

II. Dynamics in Single Yeast Cells

Epigenetic inheritance of a bi-stable switch in yeast

Stochasticity in gene expression is a major origin of phenotypic variation among isogenic cells. However, another factor that makes single cells distinct is their unique family history. While significant progresses have been made in studying stochastic gene expression, not much work has been done utilizing the single-cell genealogical information. This is mostly due to the technical challenge of reconstructing genealogical relationships at the single cell level.

        We took the advantage of the bistable galactose utilization pathway in the yeast S. cerevisiae and disrupted its major negative feedback loop, so that under neutral media conditions, these cells, with exact same genetic background, could randomly transition between two distinct semi-stable expression states (i.e., ON and OFF), which we called ‘epigenetic states’. By developing a single cell fluorescent microscopy assay and new analytical tools, we were able to obtain both gene expression and genealogical data in growing colonies of budding yeast. We found that, overall, individual cells have exponentially distributed switching times as expected. However, when taking the genealogical information in account, we were surprised to find that cells that were related switched in a synchronized manner.  We constructed a simple stochastic model that captured the essential properties of this process and explained the bulk of our results. In our model, the synchronization observed between closely related cells is a consequence of the large stochastic bursting of a regulatory protein. Such burst-induced correlation will only last for a short period of time, i.e., couple of generations, as measured in our study. This suggested a typical time scale for the epigenetic inherited traits which are generally less stable than genetic inherited traits.

Heritable stochastic switching revealed by single-cell genealogy

B. B. Kaufmann*, Q. Yang*, J. T. Mettetal, and A. van Oudenaarden

(* equal contributions)

PLoS Biology 5, e239 (2007).

Download Reprint (PDF) |  SupInfo | Movies

The impact of network architecture on the evolvability of a eukaryotic signal transduction cascade

        No component in the cell operates in isolation. In fact, it is the complex protein and gene networks that the cell relies on in order to respond to and interact with its surrounding environment. Hence, any statement on the evolution of the components of a cell must take into account their functional interactions. In recent years, many theoretical studies have suggested that network topology is a key determinant of gene evolvability, but experimental evidence has been scarce. We aim to fill the gap by taking an integrated experimental and computational approach to explore the robustness of network dynamics to mutations in an individual protein, in light of its specific topological properties.

        As a model system, we used the well-conserved high osmolarity glycerol (HOG) pathway responsible for mediating the cellular response to osmotic shock in the budding yeast Saccharomyces cerevisiae. In order to study how evolution might impact the HOG pathway, we generated a systematic series of hybrid signalling cascades in which one of the, in total five, S. cerevisiae components has been replaced by its ortholog from several different yeast species. Using the presumably functional orthologs, rather than randomly mutated sequences generated by methods such as error-prone PCR, allowed us to search more efficiently the space of sequences that had a reasonable chance of complementing wildtype behavior and that represented true outcomes of evolution. We found that while large-scale sequence variation is tolerated in the upstream elements of the cascade, the downstream components display a significantly reduced mutational robustness. These experimental findings are consistent with a computational model that points to the role of network context in determining the robustness of the pathway to mutations in its components.

The impact of network architecture on the evolvability of a eukaryotic signal transduction cascade

S. Mukherji, Q. Yang, and A. van Oudenaarden

Science (Under review)

III. Transcriptional Kinetics in Mammalian Cells

Stochastic mRNA synthesis in mammalian cells studied by three-color single-molecule RNA FISH

       Heterogeneity among cells, a universal phenomenon in the noisy biological world, has stimulated much interest in systems biology. One source of this heterogeneity is stochastic effects in the process of gene expression. In particular, recent studies have shown that mRNA synthesis is a key contributor to gene expression variability, with mRNAs being produced in short, random bursts interspersed between longer periods when no mRNAs are produced. In eukaryotes, chromatin remodeling is thought to underlie this burst-like mRNA production.

        With a newly developed RNA FISH technique that allows one to count individual molecules of specific mRNAs in single cells, we are examining variability in mRNA synthesis in mammalian cells to try and understand the basic kinetics of RNA production and the origins of transcriptional bursts. The principle advantage of this method is that it allows one to label endogenous transcripts, thus allowing for very direct measurements of transcription in mammalian cells, which are otherwise very hard to manipulate genetically. We chose three genes that are either up-regulated or down-regulated by the Glucocorticoid Receptor (GR) and labeled the transcripts from these genes with three distinct fluorescent colors. With multi-channel fluorescent microscopy, we simultaneously detect the mRNA numbers of three genes in single A549 cells, enabling us to not only measure fluctuations in the expression of individual genes but also the correlations in these fluctuations between cells. We generated time-course and dose-response data for three genes with dexamethasone induction and extracted the kinetic parameters based on models of burst-like transcription. We found that transcription of these genes is extremely noisy, and that induction sometimes changes burst size and sometimes burst frequency. Interesting correlations in transcription have also emerged from our analysis.

        The following figure is one example of simultaneously imaging three endogenous gene transcripts in a single A549 cell,

with single molecule resolution:

        Fig.A. Merged DIC and DAPI image of a single A549 cell.

        Fig.B. Single mRNA molecules of three endogenous genes from the same cell shown in Fig.A.

                    The individual transcripts are shown as distinct dots in pseudo-colors (COX-2 in red, FLJ1112 in green and FKBP5 in blue).

        Fig.C. Zoom in of the yellow squared region in Fig.B.

IV. Dynamics in Single Cyanobacteria Cells

Timing Systems in Single Cyanobacteria cells

        Circadian oscillators are the basic cellular timing systems found in almost all living creatures on Earth, from algae to humans, to improve their fitness by timing their internal activities to match the patterns of day and night, the day warm and bright and the night cold and dark.

        Cyanobacteria, the simplest organisms known to have circadian behaviour, contain only three core clock proteins: KaiA, KaiB and KaiC, but nevertheless have developed an amazingly precise and self-sustained circadian clock system. Moreover, cyanobacteria divide more than once every 24 hours, meaning that the clock operation is robust to perturbations arising from cell division. Based on these facts, cyanobacteria provide a unique opportunity for us to study the molecular origin of circadian clocks. The results may also provide insight into more complex clock systems in higher organisms.

Two questions have intrigued me the most:

1) How can such a simple prokaryotic clock resist perturbations from frequent cell divisions and high intracellular noise?

2) How does the circadian clock interact with other cellular timing systems such as cell cycle clock?

        However, current studies for the cyanobacteria clock have mainly focused on its biochemical properties in vitro but failed to provide any direct information as to how the clock works in cells. In vivo studies that use luciferase reporter assay have also been limited to bulk measurements due to the low light intensity of the luminescence.

        Hence, it is impossible to answer those two challenging questions with only traditional methods. Instead, I am utilizing the live single-cell fluorescent microscopy to fulfil the goal. In the following two projects, I will describe what I have achieved and am pursuing in order to reveal the anwers:

1) How can such a simple prokaryotic clock resist perturbations from frequent cell divisions and high intracellular noise? Interlocking of the transcription and translation regulatory oscillator and the protein oscillator forms the basis for robust and precise circadian rhythms in cyanobacteria.

        For multi-celled organisms, synchronization between individual cells is easily achieved by cell-to-cell communication, meaning that any one individual cell need not be very accurate in their circadian timing. In cyanobacteria, however, evidence suggests that the cells do not exchange the timing information. Thus, the precise oscillations observed must be due to the inherent fidelity of the circadian mechanism.

        Unlike other organisms, which rely on a delayed transcriptional feedback cycle to generate oscillations, cyanobacteria have a protein oscillator that contains three Kai proteins. Interestingly, adding these proteins together with ATP in a test tube is sufficient to produce sustained oscillations, a finding that has become a major argument that the protein oscillator is the dominant mechanism behind the circadian rhythms observed in cyanobacteria. However, no study to date has demonstrated the necessity of this protein oscillator in vivo. In fact, our data shows that cells still oscillate even with the protein oscillator disrupted by a mutation. More interestingly, these mutants show robust oscillations individually just as the wild type cells but unlike the very regular circadian behaviors in wild type cells, their timing becomes increasingly variable as the cells divide. My single cell approach is essential for this study, since bulk measurement would average away the cell-to-cell variability we observe.

        If each of the oscillators by itself can introduce the oscillation, then what is the advantage to introduce such ‘redundancy’ in cells? We hypothesize that the two oscillators have to couple together to ensure robustness and precision in timing in the absence of the intercellular communication present in higher organisms.. We are building a quantitative model to give us better understanding of the interplay between the transcriptional and protein oscillators.

2) How does the circadian clock interact with other cellular timing systems such as cell cycle clock?

Circadian clock regulates cell cycle by gating the division events at specific phases.

        Cell division is a periodic biological event that is tightly regulated by and coordinated with other cellular events. Only a handful of studies so far have focused on the interaction between the cell cycle and circadian cycle. While cells with a doubling time ranging from 6 hours to indefinite (cell division is shut down) show no defects in circadian rhythm, there are times in the circadian cycle in which cell division is inhibited, called ‘circadian gating’, as first observed by Carl Johnson. The molecular mechanism underlying the circadian gating of cell cycle, however, is completely unknown; in fact, there isn’t even an accurate measurement of the gating itself. By tracking single cell traces and mapping each individual division event onto its circadian phase, we are able to generate a histogram of the point in the circadian cycle when divisions happen and thereby detect the accurate position and range of the gating. By comparing the gating profiles among different mutants, we aim to understand the molecular mechanism underlying the gating. We found that KaiC activity plays an important role for the gating. Its ATPase activity is believed to regulate the cell cycle: the higher ATPase activity, the longer cells grow without dividing and vise versa. Based on our data, the gating position also correlates with the circadian phase position right after the KaiC phosphorylation peak. This is when the KaiC “shuffling” is happening, which is a mechanism thought to play a role in synchronizing KaiC protein states. It may then be that cells have evolved the gating machinery to prevent the interruption of synchronization by cell divisions.