Threshold effects in gene regulation: When some is not enough

Understanding a complex system requires more than assembling a parts list and wiring the parts together. We also need to understand the functional consequences of the connections, the biochemical properties of the parts and their interactions, the timing of events in a dynamic circuit, and the response of the system to stimuli if we hope to predict systems behavior and the consequences of changes in the system. In a recent issue of PNAS, Kobiler et al. (1) took an important step in this direction for one of the best-understood model gene regulatory circuits, that of bacteriophage λ.

For decades, studies on λ have pioneered this progression toward a systems-level description of a complex regulatory circuit (2, 3). At the same time, since the early days of λ biology, outsiders have been bewildered by the tendency of λ workers to use an apparently endless stream of mutations and multiple-mutant combinations to prove their points. Kobiler et al. (1) offer a relatively mild dose of this treatment, but an introduction to the λ life cycle, the players, and their interactions is helpful.

An Escherichia coli cell infected with λ can follow either of two alternative pathways (Fig. 1A). The choice between these is termed the lysis–lysogeny decision (4, 5). In the lytic pathway (blue in Fig. 1 A), the cell executes a pattern of viral gene expression (Fig. 1B), replicates the viral DNA, makes ≈100 new virions, and finally lyses. In the lysogenic pathway (green in Fig. 1 A), by contrast, the viral genome sets up housekeeping in the host; the lytic genes are repressed by CI repressor, and the viral DNA is physically integrated into the host genome. Once established, the lysogenic state is extremely stable; CI can repress the lytic genes indefinitely. However, this state can switch to the lytic state by a pathway called prophage induction (red in Fig. 1 A), which operates through the host SOS response. DNA damage results in activation of the host RecA protein; activated RecA stimulates the self-cleavage of CI (6), leading to its inactivation and expression of lytic genes.

Fig. 1.

Life cycle of phage λ and transcription patterns. (A) Life cycle. See text for details. (B) Map of part of the λ genome (not to scale). Early lytic transcripts from the pL and pR promoters are indicated; pR′ is constitutive, but in the absence of Q function the transcript terminates at tR′. The dashed segment indicates that a sizable distance lies between cII and Q. Transcription of late genes from pR does not provide sufficient levels to support lytic growth, partly because of repression of pR by Cro. (C) Map of CII-controlled promoters (not to scale). See text for details. The pE (or pRE) promoter directs expression of CI during the lysogenic response (hence, pE for Establishment phase); the pM (or pRM) promoter expresses CI in a lysogenic cell (hence, pM for Maintenance phase), and is activated by CI, not by CII. The location of the 3′ end of the paQ transcript is not known. The int gene encodes integrase. (D) Functional assays. In each case, the indicated fusion is carried on a multicopy plasmid.

Kobiler et al. (1) focus on the lysis–lysogeny decision. This decision depends on the level of another viral protein, the transcriptional activator CII. At high levels of CII, its action at three viral promoters (Fig. 1C) favors the lysogenic pathway. CII activates cI expression from the pE promoter, giving a large burst of CI synthesis and complete shut-off of the lytic pL and pR promoters. Expression of integrase allows integration of the λ genome. Finally, CII activates the paQ promoter, which works as detailed below.

CII is very unstable in vivo. Its level, and therefore the outcome of the regulatory decision, is determined in part by the action of a host protease, FtsH, which degrades CII (7). Cell physiology can influence the outcome of the decision, possibly by regulating the level of FtsH activity. The viral CIII protein, which is also a substrate for FtsH (8), reduces the activity of FtsH toward CII, probably by a competition mechanism. The lysogenic response is favored by infection with several phages, an effect that probably operates, at least in part, by higher rates of synthesis of CII and CIII due to increased gene dosage.

The last player in our story is the Q protein. Q is not involved in the actual lysis–lysogeny decision, but regulating its function is crucial to a proper outcome. Like CII, Q is encoded by the P_R transcript (Fig. 1B). Q is an antitermination factor (9) acting to allow the expression of the late lytic genes, including cell lysis genes, by permitting read-through of the constitutive pR′ transcript through the tR′ terminator (Fig. 1B). Once read-through becomes efficient, the infected cell will eventually lyse. Hence, in a cell following the lysogenic pathway, the circuitry must prevent functional levels of Q from accumulating. One major achievement of Kobiler et al. (1) was to identify mechanisms that likely serve this purpose.

The goals of Kobiler et al. (1) were, first, to obtain a more detailed view of the regulatory events in the lysis–lysogeny decision; second, to ask how the lytic pathway is blocked in a cell following the lysogenic pathway; and third, to evaluate the contributions of various regulatory proteins to the decision process. To address these questions, the authors used functional assays for the activities of CII and Q. In each case, protein function resulted in expression of GFP from a plasmid-borne reporter (Fig. 1D). Cells carrying one or the other fusion were infected with λ or various mutants, fluorescence was followed as a function of time, and rates of expression were inferred from the slope of the time-course (see figure 1 in ref. 1).

This approach revealed two different mechanisms by which the activity of Q is inhibited in cells following the lysogenic pathway. Because the growth conditions favored the lysogenic response, in cells infected with wild-type λ, CII was active but little Q activity was observed. In contrast, Q activity was seen in cells infected with either of two mutants unable to carry out the lysogenic response, cI^– and cII^– mutants. Two features of these infections are crucial: First, in the cII^– infection, there was a substantial delay before Q function appeared. Second, in the cI^– infection, both CII and Q function were observed, as expected because CI plays no role in dictating CII levels at an early stage. The appearance of Q function was delayed by ≈10 min beyond the time seen with the cII^– mutant, suggesting that this additional lag depended on CII function. These and other findings support a model in which two different effects combine to create a long delay in appearance of Q activity, one dependent on CII and the other a threshold effect.

The CII-dependent effect on Q activity required an active paQ promoter; knocking out this promoter prevented the CII-dependent delay, whereas knocking out pE had little or no effect. The mechanism of this effect is unclear; perhaps the paQ transcript acts as an antisense message (10) to inhibit translation of the Q transcript. In any case, the resulting reduction of Q levels would then contribute to the other mechanism for a delay in Q function.

This second mechanism was suggested by the finding that substantial Q protein accumulated, as judged by Western analysis, well before the activity assays showed it to be functional. Kobiler et al. (1) proposed that Q levels must exceed a threshold value before Q can function effectively. A similar finding was made in an uncoupled plasmid-based system (11), but the current findings tie it much more directly to λ biology. The molecular basis of this threshold effect is unclear as yet.

Kobiler et al. (1) also analyzed the effects of two other important regulators on expression of CII and Q. The first is CI repressor. The large difference in Q activity between wild-type and cI^– infections suggests that CI, made from the pE promoter, plays a critical later role in preventing Q function, presumably by repressing pR before Q levels can exceed the activation threshold. In this view, the role of CI in the decision to follow the lysogenic pathway is to confirm this decision once it has been made. Second, λ has another repressor, Cro, which partially represses transcription from pL and pR later in the lytic pathway. λ cro mutants cannot complete the lytic pathway; here, they showed far more CII function than did wild type, but no Q function, presumably because of increased CII function followed by CI repression. These findings suggest that Cro pushes the system toward the lytic pathway in wild type by reducing the level of CII and CIII. Accordingly, the data support a view, previously based on less direct evidence, that CII and Cro act in opposing ways during the decision process.

Although the methodology used is powerful and direct, it is important to note several limitations. First, GFP takes several minutes to mature to the fluorescent form, so its time of synthesis precedes somewhat the time it is detectable. This has little effect on comparisons within the data set, but it limits the ability to compare the kinetics with those seen in other work. Second, the method looks at populations rather than at single cells; if some cells are following each pathway (as in cells infected with two to four phages, see figure 3 in ref. 1), it is difficult to infer the patterns of expression in cells following a particular pathway. The observed patterns were seen in populations strongly biased in one direction by mutation or conditions. If the choice is less biased, the kinetics and patterns of expression may change; for instance, the decision may be delayed somewhat, or its timing may be more heterogeneous. Third, the timing of GFP expression likely varies from cell to cell, either for stochastic reasons (12) or because of variations in the pattern of gene expression in certain cells, as discussed. Focusing on the time at which GFP is first detected may give undue weight to those cells that make it first. Finally, the presence of the CII or Q target sites on multicopy plasmids may perturb the timing of events by acting as sinks for CII or Q, and these effects may differ in the two reporter strains. However, these concerns do not undermine the major findings of the Kobiler et al. (1) article, which are based on large effects.

In sum, Kobiler et al. (1) provide a straightforward explanation for inhibition of lytic functions during the lysogenic response. The threshold behavior of Q function, augmented by the effect of CII-dependent paQ transcription, ensures that a small amount of Q cannot negate the decision to follow the lysogenic pathway. As with other examples of threshold behavior (13), it allows the system to tolerate small excursions onto the path not taken. Clearly, the work goes well beyond the wiring-diagram stage toward providing a systems-level description of this venerable circuit. In this way, it helps us understand how, as in all complex systems, the whole is more than the sum of the parts.