LETTER
Construction and Engineering of Positive Feedback Loops Daniel J. Sayut, Yan Niu, and Lianhong Sun* Department of Chemical Engineering, University of Massachusetts Amherst, 686 North Pleasant Street, Amherst, Massachusetts 01003
A B S T R A C T Artificial positive feedback loops (PFLs) have been used as genetic amplifiers for enhancing the responses of weak promoters and in the creation of eukaryotic gene switches. Here we describe the construction and directed evolution of two PFLs based on the LuxR transcriptional activator and its cognate promoter, PluxI. The wild-type PFLs are completely activated by 10 nM of 3-oxo-hexanoyl-homoserine lactone (OHHL). Directed evolution of LuxR increased the sensitivity of the feedback loops, resulting in systems that are completely activated at OHHL concentrations of 5 nM, or ⬃3 molecules per cell. The responses of the PFLs can also be modulated by adjusting inducer concentrations. These highly sensitive yet regulatable PFLs can be used to construct larger artificial genetic networks to gain understanding of the design principles of complex biological systems and are expected to find various applications in industrial fermentation and gene therapy.
*Corresponding author,
[email protected].
Received for review October 11, 2006 and accepted November 3, 2006. Published online December 1, 2006 10.1021/cb6004245 CCC: $33.50 © 2006 by American Chemical Society
692
ACS CHEMICAL BIOLOGY • VOL.1 NO.11
T
here are many examples of positive feedback in biological networks, including the nuclear factor B and p53 signaling pathways, both of which regulate a diverse range of physiological functions (1, 2). Most genetic positive feedback loops (PFLs) exhibit switchlike “all-or-none” bistability, though a binary response is not the only possible output of positive feedback (3, 4). For the response of a PFL to be bistable, the system must contain some degree of ultrasensitivity (5), and its components must be balanced (6). In this regard, the combining of experimental results with stochastic models has proven to be valuable in predicting and characterizing the behavior of artificial positive feedback constructs (7, 8). The bistable responses that result from properly designed PFLs allow for their use as genetic switches to regulate gene expression. In one such example, positive feedback has been used to enhance the transcriptional activity of cell- and tissuespecific promoters, providing a method for the enhancement of gene expression in gene therapy (9, 10). The desired properties of PFLs used for controlling gene expression are that they are simple, tightly regulated, easily inducible, and nontoxic (11, 12). Because of these requirements, it is generally not effective to construct artificial PFLs using natural regulatory elements, as these elements are optimized for survival and reproduction rather than gene expression (13). As a result, an engineering strategy is frequently necessary to optimize the function of artificial genetic circuits (14). Demon-
strating this principle, we have used directed evolution to enhance the properties of PFLs constructed from bacterial quorumsensing components. Bacterial quorum-sensing systems are typically very sensitive to their inducers, activating gene expression at inducer concentrations ⬍1 M (15). Despite these innate sensitivities, we anticipated that more sensitive systems could be obtained by constructing PFLs based on quorum-sensing components. To construct our PFLs, we utilized the LuxR transcriptional activator and PluxI promoter isolated from the quorumsensing system of the marine bacterium Vibrio fischeri (16, 17). The LuxI–LuxR quorum-sensing system has been the focus of many studies due to its natural properties and possible applications in synthetic biology (18–20). In this system, LuxR activates the PluxI promoter in the presence of high concentrations of 3-oxo-hexanoylhomoserine lactone (OHHL) (21). Using these components, we constructed the first PFL, PFL1, by using PluxI to regulate expression of LuxR (Figure 1, panel a). A gfpuv gene upstream of luxR determines the output of the circuit. An alternative PFL, PFL2, which is similar to PFL1 but with an additional constitutively expressed LuxR, was also constructed (Figure 1, panel b). In both of these systems, exogenous addition of OHHL causes activation of the PluxI promoter by LuxR, resulting in expression of luxR and gfpuv. The increased expression of LuxR further enhances the activity of the PluxI promoter, resulting in positive feedback. In www.acschemicalbiology.org
LETTER vation of PFL1 and PFL2 occurred at LuxR 1400 OHHL concentraPFL2 1200 Pluxl PFL1 Pluxl gfpuv luxR tions of 10 nM, gfpuv luxR Isolation of mutants for 1000 FFL corresponding to incorporation into PFLs 800 and use as templates for b ⬃6 molecules per subsequent generations Error-prone 600 Pluxl luxR PCR cell (the volume of 400 X an Escherichia coli LuxR X 200 X cell is ⬃1 ⫻ 0 Mutant library Fluorescent Pluxl 10⫺15 L), a sensigfpuv luxR 0.1 1 100 10 screen OHHL (nM) tivity that is rarely observed for Figure 1. Design and response of PFLs. Schematic diagrams of a) PFL1 and b) PFL2. In PFL1, basal levels of LuxR are inducible genetic activated upon addition of OHHL, resulting in activation of the PluxI promoter and establishment of positive feedback. In PFL2, intracellular concentrations of LuxR are increased by addition of a second LuxR gene constitutively expressed from a systems. The Plac/ara promoter. c) OHHL dose-responses of wild-type PFL1, PFL2, and reference FFL. Data measured using GFPuv responses of the fluorescence. d) Schematic diagram of directed evolution procedure. See main text for discussion. two PFLs are similar, with PFL2 designing PFL1, we assumed that basal 23). The population distributions of cells showing slightly increased activity at OHHL expression of LuxR from the PluxI promoter in the “on” and “off” states leads to an concentrations ⬎1 nM, presumably due to an increased level of LuxR in PFL2. Activation would be high enough to result in system observed graded response at the populaactivation upon addition of OHHL. In addition level (23, 24). Consequently, we charac- of PFL1 in the presence of OHHL indicates tion to the two PFLs, a reference feedforward terized the PFLs and FFLs by measuring their that basal expression from PluxI allows for loop (FFL) composed of a GFPuv protein dose-responses (25) (Figure 1, panel c) and sufficient accumulation of LuxR for system regulated by a PluxI promoter was also con- determining [OHHL]50, which we defined as activation, which is consistent with the native biological functions of the promoter structed to characterize the PFLs. the OHHL concentrations required by each In populations governed by positive feed- circuit for half-maximal activation (Table 1). (16). Despite the basal activity of the PluxI promoter, the background activities of the back, population heterogeneity is deterBoth wild-type PFLs exhibit improved two systems as measured by the GFPuv mined by the level of noise in the regulatory responses to OHHL over the FFL with a loop and the strength of the feedback (8, 22, 10-fold decrease in [OHHL]50. Complete acti- signal are minimal, and OHHL is required for activation. Overall, the high sensitivities of the feedback loops coupled with the ability a TABLE 1. Characteristics of mutant LuxR proteins to stringently control their response make them attractive alternatives to previously Nucleotide Amino acid [OHHL]50,PFL1 (nM) [OHHL]50,PFL2 (nM) constructed genetic circuits for the efficient substitutions substitutions regulation and amplification of gene expresWild type N/A N/A 5.05 ⫾ 0.30 4.48 ⫾ 0.34 sion (12, 26, 27). Mut64 A503T N168I 2.08 ⫾ 0.13 1.65 ⫾ 0.04 To determine if the sensitivity of the PFLs Mut616 T138C I46I 0.54 ⫾ 0.04 0.38 ⫾ 0.02 could be increased, we used directed evoluA193G K65E tion to enhance the activities of the LuxR A311C K104T transcriptional activator (Figure 1, panel d). T606A D202E In a series of papers, directed evolution has Mut620 A343G T115A 1.41 ⫾ 0.06 0.96 ⫾ 0.04 been used to alter LuxR’s specificity for its G484A V162I cognate signal molecule in order to allow for Mut627 T90C S30S 1.85 ⫾ 0.04 1.33 ⫾ 0.06 the creation of unique LuxR proteins that T147A H49Q
a
c
d
Fluorescence (AU)
LuxR
A355T
I119F
The [OHHL]50 value for the FFL is 79.9 ⫾ 5.53.
a
www.acschemicalbiology.org
can be used for the construction of synthetic circuits with minimal cross talk (25, 28). Mutant LuxR proteins that are hypersensiVOL.1 NO.11 • 692–696 • 2006
693
694
VOL.1 NO.11 • 692–696 • 2006
SAYUT ET AL.
cuits allows for the effects of increased intracellular concentrations of the transcription factor to be examined by comparing the responses of the PFL2 circuits to the responses of the PFL1 circuits. The increased responses
Fluorescence (AU)
1400
Mut64
1200 1000 800 600 400
PFL2 PFL1 WT PFL2
200 0 0.1
Fluorescence (AU)
1400
1
10 OHHL (nM)
100
Mut616
1200 1000 800 600 400
PFL2 PFL1 WT PFL2
200 0 0.1
Fluorescence (AU)
1400
1
10 OHHL (nM)
100
Mut620
1200 1000 800 600 400
PFL2 PFL1 WT PFL2
200 0 0.1
1400 Fluorescence (AU)
the OHHL binding pocket (31), and the I119F mutation is adjacent to a residue that is critical for LuxR function (32). One of the mutations, D202E, is positioned in a susFigure 2. Mapping of identified LuxR amino acid substitutions in the crystal structure of pected helix–turn–helix motif (residues TraR. The TraR dimer, target DNA, and OHHL 200–224) and likely alters DNA binding. signaling molecules are shown. The structures Alignment of the mutant sequences with the of the TraR residues that align with the amino crystal structure of the LuxR homologue TraR acid substitutions in LuxR are labeled. Five of supports the hypothesized role of the mutathe identified mutants (H49Q, K65E, K104T, T115A, and I119F) cluster around the signal tions in affecting OHHL binding (Figure 2). binding pocket. Two of the mutants (V162I and Using the mutants identified in our N168I) occur at the interface between the TraR screen, we constructed PFLs as described dimers. Mutation D202E aligns to the DNA above and determined their responses binding domain. While the amino acid (Figure 3). All of the mutant PFL1 and PFL2 sequence homology of LuxR and TraR is
