Principles of a switch

Lambda's 'genetic switch' responds in a dramatic all-or-none fashion to an environmental signal, activating transcription of certain genes as it turns off others. The switch illustrates principal features of many biological regulatory processes.

As a higher organism develops, genes turn on and off—that is, they are transcribed or not—in complicated patterns, often in response to signals sent from other cells¹. The bacterial virus (phage) lambda is a model for the study of such phenomena. Lambda's genes are expressed in two distinct patterns, controlled by its switch. The switch also controls the transition between these patterns in response to an extracellular signal. Just two DNA-binding regulatory proteins and the enzyme RNA polymerase are required, and the switch works by simple (and mostly weak) protein-protein and protein-DNA binding reactions. Despite its simplicity, the properties of this switch, which control the action of polymerase, illuminate and are illuminated by other regulatory processes.

Lambda injects its DNA into a bacterium (Escherichia coli) with two possible outcomes. In one scenario, called lysogeny, a single phage gene—the repressor gene cI—is expressed. The repressor protein, despite its name, activates transcription of its own gene as it simultaneously turns off the remaining 50 phage genes. The bacteria then multiply as usual, and the phage DNA, which has become inserted into the bacterial chromosome, replicates passively as the bacteria divide. Lysogeny is perpetuated over many cell generations until a signal (in the form of UV light, for example) destroys repressor, causing a change in gene expression: the repressor gene is turned off, and the previously silent genes are turned on. The phage DNA replicates rapidly, its structural proteins are produced and the bacterium soon bursts, yielding a new crop of phage. Repressor is one of the two specific DNA-binding proteins of the switch. Cro, its antagonist, which is silent in a lysogen, is the other².

In the following sections, I discuss design features and components of the switch and show examples of counterparts in eukaryotes.

Binding of proteins to DNA

A snapshot of the inside of a lambda lysogen would reveal, with high probability, lambda repressors bound to two specific and adjacent sequences—operators—as shown in Figure 1.

Figure 1: The lambda switch.

(a) In a lysogen, lambda repressor (blue dumbbell) preferentially occupies two adjacent operator sites, labeled 1 and 2. This preferential occupancy is determined by two factors: site 1 has the highest affinity for repressor, and a repressor dimer binds there cooperatively with another repressor dimer binding site 2. In this state, repressor activates transcription of its own gene (which proceeds leftward in the figure) as it represses transcription of the cro gene. With lower efficiency, repressor also binds the weak site 3 and thereby turns off transcription of its own gene. Binding of the third site is facilitated by interaction with another repressor dimer bound some 2,500 base pairs away, in an example of cooperative binding accommodated by DNA looping¹³ (not shown). The yellow stars indicate protein-protein contacts of about equal strengths, one mediating cooperative binding of repressor dimers, the other mediating recruitment of RNA polymerase by repressor. (b) UV irradiation results in cleavage of repressor and the onset of transcription of cro and other lytic genes. Cro binds the same three operator sites, but in an order opposite that of repressor: it first binds site 3 and turns off expression of the repressor gene. Later in the lytic cycle, Cro decreases or stops transcription of its own gene by binding sites 1 and 2. Cooperativity has no role in Cro binding, but this is an exceptional case².

Repressor uses two strategies to recognize these sites: each repressor monomer inserts an α-helix in the DNA major groove, where amino acid side chains make specific contacts with the edges of base pairs; and, by virtue of a protein-protein contact, repressor dimers bind cooperatively to these two sites. This contact between adjacently binding repressor dimers is weak (∼2 kcal), but the 10- to 100-fold effect on binding affinity and specificity is crucial. Repressor binds more weakly to the third site (OR3), and Cro also binds these three sites—matters that will be discussed later. The sequences of the three operator sites are similar but not identical. Although their intrinsic affinities for repressor (and Cro) differ by no more than ten-fold, the abilities of the proteins to distinguish between them is important for proper switch function. DNA binding of repressor (or Cro) has no significant effect on DNA structure; rather, it serves simply to adhere the protein to a specific site². Throughout this discussion, terms such as 'binding', 'on' and 'off' are shorthand for concentration-dependent dynamic interactions.

Eukaryotic regulatory proteins recognize specific DNA sequences by similar strategies, with cooperativity being the rule rather than the exception. Not only does cooperativity help ensure specificity of binding, it also is used to integrate information. Thus, two proteins might be present in separate cells with no effect, but when present in the same cell they bind (cooperatively) to DNA. An example of this effect on the differentiation of sensory neurons in the worm Caenorhabditis elegans is shown in ref. 3.

Differential site affinities are exploited to help specify positions in the Drosophila embryo. A concentration gradient of the DNA-binding protein Dorsal is established early during development, increasing from the 'bottom' to the 'top' of the embryo. Strong Dorsal binding sites (associated with a specific set of genes) are occupied throughout the embryo, whereas weaker Dorsal sites are occupied only in regions of the embryo where the concentration of Dorsal is high. The resulting differences in gene expression contribute to proper formation of the fly⁴.

The mechanism of activation of transcription by lambda repressor—my next subject—is similar to that mediating cooperative binding of repressors to DNA.

Activation: recruitment

Again there is an interaction between proteins binding to DNA, but in this case the protein-protein interaction is between repressor and RNA polymerase (Fig. 1). The interaction recruits polymerase to the adjacent promoter (the site of binding of RNA polymerase), where transcription then proceeds spontaneously. The contact between repressor's 'activating region' (a few amino acids on its surface, separate from those directly involved in DNA binding) and polymerase is not very strong—say, 2 kcal—but an increase of 10- to 50-fold in the rate of transcription of a gene is biologically important.

Repressor activates transcription of its own gene because its binding sites on DNA are adjacent to the promoter of the gene (Fig. 1). But such an activator will work on any gene, provided the activator binding site and promoter are suitably positioned. The reaction has limited stereospecific requirements, with different bacterial activators touching different parts of the polymerase surface⁵.

Eukaryotic transcriptional activators also work by recruitment (Box 1). Like bacterial activators, they bear distinguishable DNA-binding and activating regions on their surfaces. It is particularly easy to construct a eukaryotic transcriptional activator: simply attach a DNA-binding domain (even a bacterial one will do) to an activating region. When expressed in a eukaryotic cell, such a fusion protein can activate any gene bearing the appropriate DNA binding site nearby. This works despite the fact that the eukaryotic transcriptional machinery is much more complex than that of bacteria, comprising many proteins in addition to RNA polymerase. The typical eukaryotic activator will recruit whatever factors are required to form the necessary complex, and transcription will ensue. Many peptide modules, when tethered to DNA-binding domains, work as activating regions. The specificity of the activator is determined solely by the location of the DNA site recognized by the DNA-binding domain. For example, the yeast transcriptional activator Gal4, when artificially expressed in another eukaryote, can activate essentially any gene, provided that the gene has had Gal4 binding sites inserted nearby⁵. The mechanism for repression of transcription in eukaryotes can differ from that seen in bacteria, as I describe next.

Box 1: Recruitment

Recruitment has a ubiquitous role in biological regulatory reactions. Most enzymes that work on macromolecules—polymerases, ubiquitylators, histone acetylators, many kinases and phosphatases—have many possible substrates (genes or proteins) upon which they might act. Choosing the right substrate for a heightened effect often depends on apposition of that substrate with the enzyme as mediated by a recruiter or by binding reactions involving scaffolds, attachment to the membrane and so on. In general, recruiting reactions require the continuous action of recruiters for maintenance. For example, a gene does not stay on once activated unless polymerase is continually recruited by the action of an activator. Essentially all recruitment reactions are plagued by the basal-level problem; one of the often unnoted complexities of biological systems is the imposition of inhibitors or other strategies to suppress basal reactions²⁰. Recruitment depends on mass action for specificity²⁵. At abnormally high concentrations of recruiters (or enzymes), specificity can be lost: activators will bind and work where they should not, enzymes will work on sites that they normally leave unmodified. These problems complicate the analyses of diseases, such as cancer, in which regulation has gone awry. RNA molecules can also act as recruiters²⁶.

Repression

To work as a repressor, a protein need only bind a DNA sequence that overlaps a promoter, such that the protein excludes binding of polymerase. It thus is easy to see how a protein (such as lambda repressor) can work as both an activator and a repressor. The dual activity of repressor—turning off Cro as it turns on its own gene—is explained by the position of its binding sites in relation to the flanking promoters. Thus, repressor is positioned so that it contacts polymerase at its own gene's promoter but excludes polymerase from the Cro promoter (Fig. 1). When the switch is flipped, Cro represses transcription of the repressor gene by binding a site (OR3) at which it excludes polymerase binding to the promoter of the repressor gene.

In bacteria, genes that are activated by recruitment are usually also controlled by a specific repressor. Why would one need a repressor to turn off transcription if an activator is required to activate it? Why doesn't eliminating the activator suffice? One general answer lies in the mechanism of activation: recruitment merely increases the rate at which a binding reaction occurs, and that increase is modest—in the absence of an activator, the reaction proceeds spontaneously to some extent, generating a basal level of transcripts from most promoters. Eliminating this basal transcription improves the system. For example, upon induction of a lambda lysogen, Cro blocks basal expression of the repressor gene; absent this effect, the efficiency of induction is compromised⁶ (Fig. 1b). Another advantage of dual control is that either regulatory protein can determine the outcome, providing two inputs that can determine whether or not the gene is transcribed.

Eukaryotic repressors that work at specific genes usually do so not by direct competition, as do their bacterial counterparts, but rather by recruitment. Thus, just as a protein complex is required to transcribe a eukaryotic gene, a repression complex is required to repress it. And, just as an activator uses an activating region to recruit the former, a repressor uses a repressing region (one of another family of short peptides) to recruit the latter⁷.

In eukaryotes, both activators and repressors work when positioned at any of a wide array of positions on DNA, often at considerable distances from their target genes. The interaction between proteins bound to well-separated DNA sites is accommodated by DNA looping. This kind of 'action at a distance' is the rule rather than the exception in eukaryotes.

In eukaryotes, activators are not always coupled with repressors, as they are in bacteria. Many specific repressors can be dispensed with, because, evidently, basal transcription is inhibited at most genes by the wrapping of DNA into structures called nucleosomes⁸. DNA wrapped in nucleosomes is relatively inaccessible, and researchers are now learning how activators nevertheless access their DNA sites and work against this non-specific inhibition^9,10,11.

From these mechanistic considerations I now turn to 'systems' properties of the switch.

Biphasic switch elements

At the heart of lambda's switch lies a double negative: repressor turns off Cro and vice versa. Upon infection, both repressor and Cro are expressed, and, depending upon which factor predominates, lytic or lysogenic growth ensues. If repressor dominates, it turns off the other phage genes, including Cro. Repressor continuously activates its own transcription in a positive feedback loop. The initial transcription of the repressor gene occurs at a promoter (the establishment promoter) separate from the one used in the lysogen (the maintenance promoter), and the initial expression of repressor is directed by a separate activator, cII, that is not expressed in the lysogen².

Positive feedback is another way to make a biphasic switch. In this scenario, once the activator (repressor, in this case) begins to work, its action is reinforced by the positive feedback loop. Positive feedback works best as a switching mechanism if it is coupled to cooperativity, rendering the transition between states hypersensitive, as it is in lambda¹². Next, I describe another feedback loop—a negative one—that helps ensure that a lysogen is poised to switch upon receiving an extracellular signal.

Concentration control

Cooperativity imposes a switch-like response to a change in repressor concentration. As the repressor concentration drops in response to UV light, a threshold is reached at which the repressors cooperatively vacate the operator, thus turning what would otherwise be a linear response into an essentially all-or-none effect. Thus, repressor concentration in a lysogen must be maintained above a specified level, a feat generally ensured by the positive feedback loop.

But it is also important to maintain repressor concentration in a lysogen below some specified level: an increase in repressor concentration of even a factor of 2–3 impedes induction¹³. A negative feedback loop, in which repressor binds site OR3 (Fig. 1), prevents such an increase². Feedback loops and other mechanisms that control concentrations of regulatory proteins are found throughout eukaryotes as well as in lambda.

Perpetuation of epigenetic states

I noted that lysogeny, once established, is self perpetuating in the absence of an inducing signal. As the bacterial cell divides, repressor (present at about 200 monomers per cell) randomly distributes between daughter cells, where it maintains lysogeny by stimulating the production of more repressor. Because lysogeny is self perpetuating once established, and because mutation is involved neither in its establishment nor its dissolution, lambda's “genetic switch” is more properly called an “epigenetic switch”¹⁴.

Much of the story of development of a higher eukaryote from a fertilized egg follows lambda-like patterns. Epigenetic states of gene expression, once established, can be transmitted from cells to their daughters by the passive transfer of recruiters (Fig. 2). This is not surprising, as cases are known in which a cell gives rise to two distinct daughters, but this special effect requires either that only one daughter receive a signal from another cell or that one or more recruiters are inherited by only one daughter. The latter requires a rather elaborate bit of cellular machinery¹⁵. Many epigenetic states in higher-organism development involve double negative and/or positive feedback loops, as well as separate control elements for the establishment and maintenance of those states^16,17,18,19.

Figure 2: Examples of regulatory protein actions in eukaryotes.

(a) Transcription factor Che-1 is a regulatory protein required for the formation and the maintenance of a specific sensory neuron in the worm C. elegans. Transcription of che-1 is initiated by other regulatory proteins that bind the DNA elements labeled 'establishment'. Che-1 binds sequences labeled ASE and activates expression of its own gene in a positive feedback loop. Che-1 also activates transcription of other genes required for identity of this neuron (not shown). The continuous production and activity of Che-1 is required to maintain the differentiated state²⁰. (b) Heterokaryons form when differentiated cells are fused so that their cytoplasms, but not their nuclei, mix. Genes in heterokaryons are expressed as dictated by 'cytoplasmic factors', that is, regulatory proteins. For example, if three fibroblasts are fused with one muscle cell, the muscle nuclei will tend to turn on genes normally expressed in fibroblasts only²¹. (c) Nuclei taken from differentiated cells can be reprogrammed to express genes they do not ordinarily express by transferring them to enucleated cells. Here, a fibroblast nucleus is implanted into an enucleated oocyte to make a pluripotent stem cell²². (d) Artificial introduction of regulatory proteins into cells can change patterns of gene expression and cell states. The introduction of a few specific regulators to cell type A (for example, a fibroblast) can cause it to become cell type B (for example, a pluripotent stem cell)²³.

Closing

Lambda's switch seems to be a highly integrated set of circuits, but it is not hard to see how it might have evolved. It could have begun as a simple but rather crude biphasic switch, comprising just repressor and Cro with their binding sites positioned for mutual repression. The system was probably improved by a series of 'add-ons', each of which made the switch more effective. These add-ons included positive and negative feedback loops involving repressor, a separate circuit for initiating repressor synthesis, repression of basal transcription of the repressor gene by Cro upon induction and cooperativity, all of which require simple, and usually weak, binding interactions.