The Strength of Indecisiveness: Oscillatory Behavior for Better Cell Fate Determination

See allHide authors and affiliations

Science's STKE  21 Dec 2004:
Vol. 2004, Issue 264, pp. pe55
DOI: 10.1126/stke.2642004pe55


Oscillatory behavior is very common in many cellular responses. Recently, two pathways involved in response to cell stress, the p53 and nuclear factor kappa B signaling pathways, have been found to show oscillatory behavior. At first sight, there would seem to be no reason for signaling pathways of this type to require oscillations. Recent single-cell studies indicate that oscillatory behavior may be used to allow repeated testing for the continued existence of a signal. I argue that oscillations increase cellular response sensitivity and flexibility by allowing the cell to integrate the results of many periodical evaluations of the signal before making an eventual decision about cell fate, thus reducing the risk of premature commitment.

Driving in New England during the foliage season is inspiring. I am thrilled by the magnificent spectrum of colors—from lemon yellow to burnt orange, bright to dark red, and cinnamon brown (Fig. 1A). It is an extraordinary sight: rows of trees of the same type, grown in the same geographical zone and under similar climatic conditions, with each tree wearing a different leafy costume to this forest party. Even the colors of individual leaves on any given tree may vary. Each leaf, in other words, has its own special phenotype. Recently, several investigators have observed a similar phenomenon of phenotypic variation among individuals in clonal populations of cells. For example, the dynamic response of certain proteins to an extracellular stimulus (in other words, the rate at which changes in the physical states, locations, or amounts of protein take place) can be dramatically different among individuals of the same cell type (Fig. 1B). Differences in cell cycle phase, the basal level of protein expression, or any other stochastic variation can all contribute to differences in the rhythms of the responses. Experiments in which measurements are averaged over a population of cells can often mask the dynamic behavior of the individuals. Therefore, to be able to understand in detail how a dynamic response is controlled, it may be essential to measure the precise timing and amplitude of responses at the level of individual cells.

Fig. 1.

Examples of individual variation. (A) Individual trees during the foliage season show dramatic color differences. (B) The dynamics of proteins involved in biological signaling in stressful conditions show significant individual variation. The image shows a number of genetically identical human cells expressing p53 fused to acyan fluorescent protein (green) and Mdm2 fused to ayellow fluorescent protein (red) after DNA damage. Different colors represent cells in different phases of the dynamic response. [Credit:]

In a recent paper, Nelson et al. (1) took the approach of studying cell responses at the individual level a step further, by using single-cell time-lapse imaging to study not only the dynamics of an initial response but also the consequences of altering this response. Their single-cells studies showed asynchronous oscillations in the localization of nuclear factor kappa B (NF-κB) after cell stimulation, which decreased in frequency with increased IκBα (an inhibitor of NF-κB) transcription. If these oscillations were blocked in individual cells, using a nuclear export inhibitor, transcription from a NF-κB–dependent promoter was also blocked.

NF-κB is a transcription factor that regulates the expression of numerous genes with roles in cellular stress responses, cell growth, cell survival, and apoptosis (24). One of the most prominent NF-κB target genes is IκBα, whose protein product binds to and inhibits NF-κB. The NF-κB–IκBα complex is exported from the nucleus to the cytoplasm, where NF-κB cannot act, creating a negative feedback loop (14). Negative feedback loops composed of one transcription arm and one protein-interaction arm form a very common motif in various species, found far more often than predicted by chance (5). Different feedback parameters can lead to different dynamic behavior over time, such as overdamped, damped, or sustained (undamped) oscillations (Fig. 2). Examples are known of biological negative feedback loops that give rise to either damped or undamped oscillations. For example, negative feedback loops (combined with positive loops) create sustained oscillations in circadian rhythm and cell cycle systems (69). Similarly, the delayed Hes1 feedback loop creates oscillations in Hes1 mRNA and protein; these oscillations underlie the mechanism of the segmentation clock that drives sequential somitogenesis (1012). The NF-κB–IκBα feedback loop shows damped oscillations in the amount of nuclear NF-κB bound to DNA in both population and single-cell experiments, although the single-cell experiments allow a much clearer view of the specific details of the response, such as the precise timing and amplitude of these oscillations (1, 12, 13). There is an even more dramatic difference between population-level and single-cell studies of the dynamics of p53 expression in response to DNA damage; in this case, negative feedback is by way of the Mdm2 protein, which is transcriptionally induced by p53 and causes p53 degradation (1418). Population studies indicated that the p53 response, like the NF-κB response, showed damped oscillations (19). In contrast, single-cell studies showed that in response to DNA damage, p53 is expressed in a series of discrete pulses with fixed mean height and duration, which do not depend on the amount of DNA damage (20). Even more important, genetically identical cells exposed to the same amount of DNA-damaging radiation respond with a variable number of p53 pulses (20).

Fig. 2.

Negative feedback loop dynamics. Oscillatory behavior may lead to flexible control of cell fate decisions. In a nonoscillating system, a decision is made in response to a single test of the signal (overdamped, blue line). In contrast, an oscillating system, whether damped (red; similar to the situation with the nuclear-cytoplasmic shuttling of NF-κB) or undamped (green; similar to the p53 response) can integrate several evaluations of the presence and nature of a signal before making a final commitment to a particular cell fate.

Are these differences between individuals really important? After all, no matter what color the leaves are, they will eventually fall. Do genetically identical cells that show different protein dynamics also end up sharing the same fate? One question that arises in considering this issue is why the oscillations are there in the first place and whether they are relevant to cell fate decisions.

When studying the dynamics of the segmentation clock, cell cycle regulation, or circadian rhythms, it seems smart for the system to choose feedback parameters that give rise to oscillatory behavior. After all, these cellular systems need to change in a cyclical pattern over time, so it makes sense that the transcription factors that drive the cycle should oscillate. In the case of cellular stress, however, the need for oscillations seems much less obvious. In the NF-κB–IκBα feedback loop, once the tumor necrosis factor–α (TNF-α) signal is detected, NF-κB enters the nucleus to activate its target genes, including its own inhibitor, IκBα, which then removes NF-κB from the nucleus, shutting down the transcriptional response (14, 13). If the TNF-α signal is continuous, IκBα is degraded, and NF-κB is free to enter the nucleus again, reactivate transcription, and stimulate the synthesis of more IκBα. Surely, this negative feedback is wasteful compared to leaving NF-κB in the nucleus as long as TNF-α is present. The situation is very similar for p53; again, a wasteful degradation (in this case of p53) is required to maintain the oscillatory behavior.

One explanation for the oscillations in the p53 and NF-κB systems might be that the relevant signal could reside in the frequency of the oscillations. This is known to be the case for intracellular Ca2+ oscillations and for the pulsatile release of cyclic adenosine monophosphate (cAMP) signals in Dictyostelium amoebae (7). In addition, most hormones are secreted in a pulsatile rather than continuous manner, and the temporal pattern in which a hormone appears is often more important than its concentration (7). Another clue for the function of oscillations in the p53 and NF-κB systems lies in the common roles of these proteins. Both NF-κB and p53 are key players in making crucial decisions related to cell fate; p53 activates several alternative pathways, including those involved in DNA repair, growth arrest, senescence, and apoptosis; and NF-κB similarly has a major role in choosing between cell survival and death. Oscillatory behavior might protect the cell from high concentrations of p53 or NF-κB, which might be dangerous to the cell. In addition, in both systems, the decision of which program to activate seems to depend on the specific posttranslational modification status of the NF-κB and p53 transcription factors (2130). For example, several studies show that activation of specific promoters depends on the phosphorylation state of p53 (2326) and that acetylation of specific lysine residues in NF-κB regulates its distinct functions (2730).

One way to make sense of this behavior is to imagine that the oscillatory dynamics of p53 and NF-κB allow the cell to evaluate the stress signal in every pulse. Instead of making one major irreversible decision such as activating fast cell death, cells would make multiple evaluations and decisions; in this scenario, the final commitment (the cell fate) depends on an accumulation of these series of decisions (Fig. 2). The decision time scale is short in comparison to the cell response. Thus, the eventual commitment to one fate or another is a result of the integration of multiple periodical decisions over time, reducing the risk of making a decision prematurely.

In the p53 system, p53 is degraded and re-expressed, so there is no "memory" of previous p53 states to confuse the issue. In the NF-κB system, although the protein is not degraded, it is believed to be stripped of all posttranslational modifications before it exits the nucleus. Consistent with this hypothesis, Nelson et al. showed that NF-κB protein trapped in the nucleus could no longer activate its target genes (1), probably because of dephosphorylation of NF-κB in the nucleus by protein phosphatase A2 (31) or deacetylation by histone deacetylase 3 (32). Thus, it appears that the export of NF-κB induced by IκBα is not in fact essential for turning off the transcriptional response, but instead is required to bring NF-κB back to the cytoplasm, where it can receive new signals.

For this "cumulative alternative-fate" mechanism to operate, I suggest that a system needs to fulfill three conditions: (i) History independence in the state of the transcription factor, generating an "inexperienced" transcription factor in every cycle. This allows the cell to make a fresh decision about which program to activate in every pulse. (ii) History dependence in the response of the cell. The cell must integrate the results of the expression of target genes over time, so that the final commitment is the sum of all the gene expression decisions that have been made in the recent past. (iii) The refresh mechanism must be program-independent; in order to maintain the oscillations, the inhibitor (for example, IκBα or Mdm2) should be expressed and should be effective in inhibiting its target no matter which of the many possible programs are activated.

The reporter gene used by Nelson et al. was driven by a promoter containing a 5 x NF-κB consensus-binding site (1). It will be interesting to see, in future studies, how the expression of specific target genes varies in response to different states of the transcription factor. Extending the type of analysis performed by Nelson et al. should allow us to investigate potential correlations of specific gene expression patterns with the dynamics of the transcription factor, its posttranslational modifications, and the eventual fate of the cell. Then, perhaps, we will understand whether all the leaves are fated to fall or whether some of them might instead turn green.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
View Abstract

Stay Connected to Science Signaling

Navigate This Article