California Institute of Technology

1200 E. California Blvd

Pasadena, CA 91125

February 18, 2015

To the Editors of ACS Synthetic Biology and Organizers of the SEED2015

Conference:

On behalf of my fellow co-authors, I would like to submit for your consideration a manuscript entitled â??Engineering Transcriptional Regulator E?ector Speci-

ficity using Computational Design and In Vitro Rapid Prototyping: Developing a Vanillin Sensorâ? for a presentation at the SEED conference, as well as publi- cation in ACS Synthetic Biology. This paper is a continutation of the research presented at last yearâ??s SEED conference under the title, â??Engineering Tran-

scriptional Regulator E?ector Specificity Through Rational Design and Rapid

Prototyping�. The work describes the engineering of qacR, a transcriptional regulator to create functional variants responsive to a new e?ector. A copy of

this manuscript and the supplemental information has been uploaded to the bioRxiv preprint server.

Results covered include:

â?¢ Computational protein design to generate a library of qacR mutants that respond to vanillin, a growth-inhibiting small molecule;

â?¢ Rapid screening using an in vitro transcription-translation (TX-TL) sys- tem;

â?¢ Further in vitro charactetrization of hits from the TX-TL screen

â?¢ In vivo characterization of qacR mutants

The further in vitro characterization, and in vivo characterization are new re- sults from further work on the system in the last year. We believe that this work fits well within the scope of the SEED2015 conference and the readers of ACS Synthetic Biology and would be of interest to those in attendance.

Thank you for your consideration.

Emmanuel Lorenzo de los Santos, for the authors

Engineering Transcriptional Regulator Effector

Specificity using Computational Design and In Vitro

Rapid Prototyping: Developing a Vanillin Sensor

Emmanuel L.C. de los Santos,? Joseph T. Meyerowitz, Stephen L. Mayo, and

Richard M. Murray

Division of Biology and Biological Engineering, California Institute of Technology, Pasadena, CA, USA

E-mail: emzodls@caltech.edu
KEYWORDS: synthetic biology, cell-free systems, in vitro biological circuit prototyping, computational protein design, protein engineering

?To whom correspondence should be addressed

1

Abstract

The pursuit of circuits and metabolic pathways of increasing complexity and ro- bustness in synthetic biology will require engineering new regulatory tools. Feedback control based on relevant molecules, including toxic intermediates and environmental signals, would enable genetic circuits to react appropriately to changing conditions. In this work, variants of qacR, a tetR family repressor, were generated by computational protein design and screened in a cell-free transcription-translation (TX-TL) system for responsiveness to a new targeted e?ector. The modified repressors target vanillin, a growth-inhibiting small molecule found in lignocellulosic hydrolysates and other indus- trial processes. Promising candidates from the in vitro screen were further characterized in vitro and in vivo in a gene circuit. The screen yielded two qacR mutants that re- spond to vanillin both in vitro and in vivo. We believe this process, a combination of the generation of variants coupled with in vitro screening, can serve as a framework for designing new sensors for other target compounds.

Introduction

The utility of a synthetic genetic circuit for real world applications is dependent on the ability to e?ectively trigger the circuit. While we can control the expression of target genes with transcriptional regulators, triggers for these transcriptional regulators are limited to a small number of molecules and other inputs (e.g. light) (1 ). As a consequence, most synthetic circuits right now are limited to proof-of-principle demonstrations without being extendable to real world applications. In order to design synthetic control circuits for real world applications such as metabolic engineering or biofuel production, signals from these pathways, such as the level of a toxic metabolic intermediate, need to be transmitted into existing transcriptional control machinery. This work develops a framework to use a combi- nation of sequence generation by computational protein design and rapid prototyping using a cell-free transcription-translation (TX-TL) system to switch e?ector specificity of existing
2

Is there a sensor?

Computationally aided selection of

No sequences for

screen (35,577,057,600 sequences)

Rapid in vitro screening of mutants (27 mutants)

Yes

Optimize to desired specifications

No

Is there a hit?

Yes

in vivo testing and in vitro verification of hits

characterization

(2 mutants)

(2 mutants)

Figure 1: Workflow for generating novel sensors. The in vitro TX-TL platform allows for the rapid screening of sequences selected with the help of computational protein design. Hits from the in vitro screen are then verified by further in vitro testing. In vivo testing and characterization can then be performed to see if they meet the desired specifications. Further refinement of the hits through directed evolution or further computational design can be performed until specifications necessary are achieved. Numbers in parenthesis are the number of sequences considered by the computational algorithm, or the number of mutants assayed at the specified step for vanillin.
transcriptional regulators to respond to targeted small molecules of interest (Figure 1).
The tetR family is a large family of transcriptional regulators found in bacteria. They are named after the tetR repressor, which controls the expression of tetA, an e?ux pump for tetracycline (2 ). They contain two domains, a helical-bundle ligand-binding domain and a helix-turn-helix DNA-binding domain. In the absence of their inducing molecule, tetR repressors bind to DNA, preventing the transcription of downstream genes. Inducer binding to the ligand-binding domain causes a conformational change in the DNA binding domain that causes dissociation from the DNA, allowing transcription of downstream genes. The tetR transcriptional regulation machinery has been used in the design of synthetic circuits, including the repressilator (3 ) and the toggle switch (4 ).
QacR is a tetR-family repressor found in S. aureus that controls the transcription of qacA, an e?ux pump that confers resistance to a large number of quaternary anionic com-
3
pounds. The protein has been studied because it is induced by a broad range of structurally dissimilar compounds (5 ). Structural examination of qacR in complex with di?erent small
molecules has shown that qacR has two di?erent binding regions inside a large binding
pocket. While qacR has multiple binding modes for various inducers, in all cases for which there are structures, binding of the inducer causes a tyrosine explusion which moves one of the helices and alters the conformation of the DNA binding domain, rendering qacR unable to bind DNA (6 â??8 ). Crystal structures of inducer-bound forms of qacR and the qacR-DNA complex coupled with a definitive structural mechanism for qacR induction make it the ideal starting point for computational design of new transcriptional regulators. In this work, we
describe our e?orts to apply our framework to engineer qacR to sense vanillin, a phenolic
growth inhibitor that is a byproduct of lignin degradation performed during the processing of biomass into intermediate feedstock in biofuel production (9 ).

Results and discussion

Computationally Aided Selection of QacR Mutant Sequences

We created a computational model of vanillin to place into a crystal structure of QacR (PDB ID: 1JTO). A computational protein design algorithm was used to find potential vanillin binding sites close to the location of the tyrosine expulsion in the binding pocket of qacR (Figure 2A-B) while being in the proximity of amino acid positions that allowed for favorable pi-stacking and hydrogen bonding interactions. We used targeted ligand placement (11 ) to find potential binding positions for vanillin by defining an idealized binding site for the molecule. The algorithm yielded four potential binding positions for vanillin (Figure 2C). Computational protein sequence design was then used to select amino acid residues at the positions around the potential vanillin binding sites. In order to minimize the possibility of steric clashes in the protein, calculations that considered both the DNA-bound state and the ligand-bound state using a multi-state design algorithm were also performed (14 ). Finally,
4

Figure 2: Computationally Aided Selection of qacR mutants. (A) Overlay of PDB structures of the non- ligand bound (cyan, PDB ID: 1JTO) and ligand bound (yellow, PDB ID: 3BQZ) conformations of qacR. A conformational shift in the binding pocket occurs upon entry of the small molecule causing the protein to dissociate from DNA. (B) A closer look at the binding pocket of qacR, the binding of the ligand causes the displacement of three tyrosine residues shown. (C) Computational model for potential vanillin binding sites.
Vanillin is shown as a di?erent color in each of the four sites. A protein design algorithm was
asked to suggest mutations for amino acids close to the potential binding sites to support
the placement of vanilin in these sites.
we also ran calculations that included an energy bias to favor the wild-type residue. The lowest energy sequences from these four calculations (single-state biased, single-state non- biased, multi-state biased, and multi-state non-biased) were analyzed, and used as a guide to compile a set of ten mutants for in vitro testing.

In Vitro Screening of Generated Sequences

We first decided to validate function of the wild-type protein. This was done by placing green fluorescent protein (GFP) downstream of the qacA promoter sequence (PQacA). While we observed a hundred-fold decrease in fluorescence in cells containing plasmids encoding the wild-type qacR gene in addition to PQacA -GFP, addition of berberine, a native qacR in-
5
ducer, yielded no observable di?erence in fluorescence (Figure S1). We hypothesized that the
inducer was not getting into the cells due to the di?erences in cell wall between gram-positive and gram-negative bacteria. Because of this, we decided to use an in vitro transcription-
translation (TX-TL) system to test the mutants (15 ). We first tested the wild type protein in our TX-TL system. We observed an increase in GFP fluorescence as we increased the con- centration of plasmid encoding pQacA-GFP from 2 nM to 8 nM (Figure 3A). The addition of plasmid encoding the qacR repressor to the system resulted in a decrease in fluorescence. However, we detected high autofluorescence from berberine. We switched to dequalinium, a previously characterized non-fluorescent qacR inducer (6 ). The addition of dequalinium resulted in an increase in fluorescence in the TX-TL system up to 85% of the fluorescence of the positive control, in which there was no DNA encoding the qacR repressor (Figure 3B). These results demonstrated a functional wild-type qacR repressor in TX-TL. After validating the function of wild-type protein in TX-TL, we used the system to look at the functionality of the qacR mutants.
None of the initial ten mutants showed any repression of GFP fluorescence. We analyzed the ligand bound and DNA-bound computational models of one of the qacR mutants that contained only three amino acid substitutions from a qacR mutant that was previously shown to be functional by Peters et al. (8 ). The computational model showed the potential for some mutations to cause steric clashes in the DNA bound state. We created a second library reverting either the 50th and 54th positions (A50F/W54L) or the 119th position (Y119L) back to their wild-type identity (Table S1).
In order to determine if any of the mutants of our library warranted further character- ization, we performed a rapid screen of 17 qacR mutants in TX-TL (Figure S2). Plasmids containing DNA that encoded each of the qacR variants or the wild-type qacR sequence were placed into a TX-TL reaction containing either water, dequalinium or vanillin. QacR activity was monitored by a plasmid encoding GFP downstream of PQacA. Two of the mu-
tants, qacR2 and qacR5, displayed an increase in fluorescence in the presence of vanillin and
6
dequalinium over water (Figure 4). We focused on these two mutants for further in vitro
and in vivo characterization.

(a) (b)

Figure 3: Validation of TX-TL Screening. (A) GFP signal after three hours of a TX-TL reaction. Plasmid encoding GFP downstream of the native qac promoter was added to the TX-TL platform. Higher concentrations of plasmid yielded more GFP signal. (B) Response of wild-type qacR to dequalinium. DNA encoding GFP and wild-type qacR was added to the TX-TL system. Increasing fluorescent signal is observed with increasing concentrations of dequalinium. The highest fluorescent signal is observed when there is no repressor in the system, demonstrating the ability of TX-TL to test for qacR repression and de-repression.

Further in vitro testing of qacR2 and qacR5

In order to verify the response of qacR2, and qacR5 to vanillin, we performed more extensive TX-TL tests on the mutants. TX-TL reactions were set up with a constant amount of reporter (PQacA-deGFP) plasmid and either no repressor (water), or plasmids encoding wild- type qacR, qacR2, or qacR5. Reactions were incubated for 75 minutes at 29 C to produce the repressor protein. This initial reaction was then added to solution containing dequilinium, vanillin, or water. We monitored the rate of GFP production between the first and third hours of the reaction, where the rate of protein production appeared linear.
Figure 5 shows the ratio of GFP fluorescence between the case where there is no repressor, and each of the repressors tested with the di?erent inducers. The wild-type qacR is able to inhibit the production of GFP to around 15% of its maximum value. The mutants are less e?cient at repressing the production of GFP. Three times and four times more repressor
7

4

3.5

vanillin dequalinium

3

2.5

2

1.5

1

0.5

0 wt 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

qacR variants

Figure 4: In Vitro TX-TL screen of qacR mutants found potential candidates for further testing. Fold change in maximum fluorescence between water and inducer for qacR mutants. Seventeen qacR mutants were screened using TX-TL. Plasmids containing DNA encoding each of the qacR variants were placed into the system along with water, dequalinium (native qacR inducer) and vanillin. To monitor qacR response, a plasmid encoding GFP downstream of the native qacA promoter was also added to the system. qacR2 and qacR5 were seclected for further characterization. qacR1 was not selected due to low absolute signal (Figure S2)
DNA was added to the reactions of qacR2 and qacR5 respectively. In spite of the additional DNA, we do not observe the same level of repression that we see with the wild-type protein. Wild-type qacR is well induced by the native inducer, we observed full derepression at the dequalinium concentration used. Induction of qacR2 and qacR5 with dequalinium is also observed, although to a lesser degree than the wild-type protein. Resource limitations due to the increase in repressor DNA added may be contributing to the lower induction level observed. QacR2 and qacR5 display a response to vanillin at the concentration we tested, while no response to vanillin was detected for the wild-type protein. The mutations introduced to the protein decrease the ability of the mutants to repress DNA. This could be due to protein instability, or due to a weaker protein-DNA interaction. However, these mutations also increase the sensitivity of the mutants to vanillin, allowing their response to
8
be detectable in our in vitro platform.
We assumed that the maximum amount of GFP fluorescence that can be achieved for a specific inducer condition was when there is no repressor present. This takes into account potential inhibition of the TX-TL reaction by the inducer. The factors that can a?ect the ability of the particular repressor to reach this the no repressor case are resource limitations due to additional load from the production of the repressor DNA, and response of the repres- sor to the inducer in the reaction. We expect that resource limitations would have a negative e?ect on the ability of the repressor to reach the maximum fluorescence level. Conversely, response to repressor should have a positive e?ect in reaching the maximum fluorescence level.

Figure 5: In vitro testing of qacR2 and qacR5. Ratio of the rate of GFP production between TX-TL reactions with and without repressor DNA. 10 µM of dequalinium and 5 mM of vanillin was used to induce the production of GFP for each of the qacR variants tested.
9

In Vivo Testing of qacR2 and qacR5

In order to further characterize the qacR mutants, and to see if we could detect vanillin in a more complex system, we decided to test the in vivo response of the qacR variants to vanillin. Plasmids containing genes that encode the wild-type qacR sequence, qacR2 or qacR5 downstream of Ptet and GFP downstream of PQacA were cloned into DH5?Z1 cells (Figure 6). For each of the qacR variants, we compared di?erences in fluorescence signal across increasing vanillin concentrations. We tested di?erent repressor concentrations by varying the amount of anhydrous tetracycline (aTc) in the system. Similar to the in vitro experiments, and in order to get an idea for the maximum fluorescence the system could achieve, we grew cells that only contained GFP downstream of PQacA without any repressor. Cells that were grown in higher aTc concentrations had a lower measured optical density (OD), indicating a slower growth rate. We hypothesize that this is due to the toxicity of the qacR repressor to the E. coli strain. Since qacR is not a native protein, it is possible that qacR is binding to locations in the E. coli genome. Interestingly, the di?erences in optical density measurements become less pronounced with increasing vanillin concentration, suggesting that vanillin may provide a mitigating e?ect to this toxicity. In order to account for di?erences in OD, fluorescence measurements were normalized to OD.
The lowest OD measurements were observed for cells encoding the wild-type qacR at 12 ng/mL aTc where very little growth was observed for cells expressing the wild-type protein. At this aTc concentration, all of the cells expressing repressor exhibited lower optical densities when compared to cells that were only expressing fluorescent protein. The di?erences in optical density are less pronounced at lower aTc concentrations. When no aTc is present in the system, cells at the higher vanillin concentrations had lower ODs. At higher aTc concentrations, cells at higher vanillin concentrations had higher ODs. This implies that both the vanillin concentration and the expression of the repressor have an e?ect on cellular growth. The optical densities for the cells at di?erent aTc and vanillin concentrations are shown in Tables S2-S5.
10
Figure 7A shows the e?ect of increasing the aTc concentration on the fluorescence of cells
in the absence of vanillin. Similar to the in vitro tests, fluorescence was normalized to the no repressor case. Increasing the aTc concentration decreased the fluorescence of cells in the absence of vanillin, confirming that the qacR mutants are able to repress the expression of GFP at higher protein concentrations.
The response of wild-type qacR, qacR2, and qacR5 to increasing vanillin concentrations is shown in figure 7B. The response curves for each protein are plotted for the minimum aTc concentraion such that maximum GFP repression is observed. This corresponds to aTc concentrations of 4, 8, and 12 ng/mL for wild-type qacR, qacR2, and qacR5 respectively. This is consistent with the in vitro data that more qacR2 and qacR5 DNA was required to repress the expression of GFP. Similar to the in vitro tests, we expect the ability of the cell to reach the maximum fluorescence level to be dependent on its response to inducer, and toxicity from vanillin and qacR. Indeed, cells expressing the qacR mutants exhibited an increase in fluorescence with increasing vanillin levels demonstrating that they are capable of sensing vanillin. The mutants exhibit a marked increase in sensitivity to vanillin. QacR2 displays a response that goes from approximately 20% of the fluorescence of the cells not expressing any repressor to matching the fluorescence of the non-repressed cells at 1 mM vanillin. QacR5 saturates at around 40% of the fluorescence of the non-repressed cells. This correlates with the in vitro data that show qacR2 achieving close to the non-repressed fluorescence, with qacR5 less sensitive to vanillin (Figure 5).

Framework Enables Engineering of Sensors through Rational Reduc- tion of Design Space

The framework developedâ?? a combination of sequence generation using computationally- aided design, preliminary screening with TX-TL, and in vitro and in vivo validationâ??can be used for other small molecule targets potentially facilitating the design of more actuators in synthetic circuits. While it is possible that the computational model of vanillin binding was
11

Figure 6: Circuit layout for in vivo tests. Genes encoding GFP under the control of the native qac promoter, and our QacR designs under the control of a tet-inducible promoter were

placed in a single plasmid and transformed into DH5?Z1 cells. qacR levels were controlled
using aTc for varying vanillin concentrations. Candidate designs that are responsive to
vanillin should show an increase in fluorescence with increasing vanillin concentrations

(a) (b)

Figure 7: In vivo response of qacR to vanillin. Cells expressing GFP without any repressor were used as a control to normalize for di?erences in fluorescence due to aTc and
vanillin levels. (a) All of the proteins are able to repress the expression of GFP. The wild-type protein is able to inhibit the expression of GFP at lower aTc concentrations, while higher aTc concentrations are necessary for the mutants to achieve a similar level of repression. (b) QacR mutants respond to vanillin in a concentration dependent manner.
inaccurate, the computational design provided value in drastically reducing the number of sequences to test into a figure that was experimentally tractable. Without the computational design to reduce the size of the design space, we would not have had e?ective starting points to attempt the engineering of a vanillin sensor.
12
The use of the in vitro cell-free system in a preliminary screen provides many advantages. It allows the screening of more mutants in a shorter amount of time. The simpler system also reduces the number of variables to consider. Complicating factors such as cell membrane permeability and cell growth do not need to be considered during this part of the screen. Repressors whose native inducers cannot enter the target organism can be used as starting points with the cell-free system. Finally, we can use this framework to target molecules that are known to be toxic to cells and measure engineering results in a cell-free context.
As a result of this process, we now have functional vanillin sensors that can be used in a feedback circuit that dynamically responds to vanillin. QacR2 and qacR5 can be used as a starting point for a synthetic circuit that responds to vanillin concentrations. While we only tested the protein in E. coli recent work has developed a process that facilitates the transfer of prokaryotic transcription factors into eukaryotic cells, increasing the flexibility of the molecules for use in metabolic engineering (16 ). By linking a vanillin sensor to the
expression of a gene that can mitigate the toxic e?ect of vanillin, such as an e?ux pump
or an enzyme which converts vanillin to a less toxic molecule, we can design a dynamic feedback circuit and potentially improve metabolic yield. It remains to be seen whether the sensors developed will have the required dynamic range or sensitivity for a functional feedback circuit; however, if a better sensor is needed these proteins can be used as a starting point for directed evolution in order to obtain a sensor with the desired properties.

Materials and Experimental Methods

Computationally Aided Selection of Mutant Sequences

An in silico model of vanillin was constructed using the Schrödinger software suite. Partial charges for vanillin were computed using Optimization in Jaguar version 7.6 (10 ) using HF/6-311G** as the basis set. Vanillin rotamers were chosen by looking at the ideal angles for the carbon hybrid orbitals. A model of an idealized vanillin binding pocket was designed
13
by looking at the protein data bank for proteins that bound small molecules similar to vanillin. Models of vanillin in the qacR binding pocket were generated using the Phoenix Match algorithm (11 ).
Monte Carlo with simulated annealing (12 ) and FASTER (13 ) were used to sample conformational space. A backbone independent conformer library with a 1.0 Ã? resolution was used for the designed residues (11 ). Designed residues were chosen by compiling a list of amino acid residues within 15 Ã? of vanillin. Potential residues for each site were selected by chemical intuition and context within the structure. Computational models of qacR with vanillin present were scored using the PHOENIX forcefield with the inclusion of an additional geometry bias term that favored pi-stacking and hydrogen bonding interactions (11 ).

Cell Free in vitro Transcription-Translation System and Reactions

The transcription-translation reaction consists of crude cytoplasmic extract from BL21 Rosetta
2 E. coli (15 ). Preliminary tests were done with plasmids and inducers at the specified con- centrations. For the initial screen, the qacR mutants were downstream of a T7 promoter. TX-TL reactions were run with 2 nM of the plasmid encoding the qacR variant, 0.1 nM plasmid encoding T7 RNA polymerase, and 8 nM plasmid encoding PQacA-deGFP. Vanillin was added at a concentration of 2.5 mM and dequalinium was added at 10 µM.
For the in vitro tests to further characterize the hits. Plasmids encoding qacR2 or qacR5 downstream of a tet-responsive promoter were used along with a plasmid encoding deGFP downstream of a qac-responsive promoter. Plasmids were prepared using the Macherey-Nagel NucleoBond Xtra Midi/Maxi Kit. Plasmid DNA was eluted in water and concentrated by vacufuge to the desired concentration. TX-TL reactions were set up as follows: 5 µL of bu?er, 2.5 µL of cell extract and 1.5 µL repressor DNA at a specific concentration was mixed and incubated at 29 C for 75 minutes to facilitate the production of repressor DNA. This mix was then added to a mixture of 1 µL deGFP plasmid and 1 µL of an inducer stock. Measurements were made in a Biotek plate reader at 3 minute intervals using excita-
14
tion/emission wavelengths set at 485/525 nm. Stock repressor plasmid concentrations were
0.24 µM , 0.73 µM , and 0.97 µM for qacR wild-type, qacR2, and qacR5, respectively. The deGFP plasmid concentration was approximately 0.40 µM. Inducer concentrations were 5 mM for vanillin, and 10 µM for dequalinium.
Experimental conditions were done in triplicate and the error bars are the error propa- gated from the standard deviation of the means.

Cell Strain and Media

The circuit was implemented in the E.coli cell strain DH5?Z1, a variant of DH5? which contains a chromosomal integration of the Z1 cassette (17 ). All cell culture was done in optically clear M9ca minimal media (Teknova M8010).

Genes and Plasmids

DNA encoding the qacR genes was constructed using overlap extension PCR. Plasmids used contained chloramphenicol resistance with a p15a origin of replication.

In vivo experiments

Cells were grown in at least two consecutive overnight cultures in M9ca minimal media. On the day of the experiment, overnight cultures were diluted 1:100 and grown for 5 hours to ensure that the cells were in log phase. Cells were then diluted 1:100 into fresh media at the specified experimental condition. Cells were grown in these conditions at 37 C for
12 15 hours in Axygen 96 well plates while shaking at 1100 rpm. Endpoint fluorescence was measured by transferring the cells to clear bottomed 96-well microplates (PerkinElmer, ViewPlate, 6005182) . GFP was read at 488/525 with gain 100.
Analysis of the data was done by taking fluorescence readings for each independent well. Experimental conditions of the qacR proteins were done in triplicate and repeats were
15
averaged. Error bars shown are the error propagated originating from the standard deviation of the mean.

Acknowledgement

The authors thank Jongmin Kim and Jackson Cahn for reading the manuscript. This research was conducted with supprot from the Institute for Collaborative Biotechnologies through grand W911NF-09-0001 from the U.S. Army Research O?ce. Additional support was granted in part by the Benjamin M. Rosen Bioengineering Center, the Gordon and Betty Moore Foundation through Grant GBMF2809 to the Caltech Programmable Molecu- lar Technology Initiative, and DARPA through the Living Foundries Program.

References

[1] Purnick, P. E. M., and Weiss, R. (2009) The second wave of synthetic biology: from modules to systems. Nature reviews. Molecular cell biology 10, 410â??422.
[2] Ramos, J. L., Martinez-Bueno, M., Molina-Henares, A. J., Terán, W., Watanabe, K., Zhang, X. D., Gallegos, M. T., Brennan, R., and Tobes, R. (2005) The TetR family of transcriptional repressors. Microbiology and Molecular Biology Reviews 69, 326â??+.
[3] Elowitz, M. B., and Leibler, S. (2000) A synthetic oscillatory network of transcriptional regulators. Nature 403, 335â??338.
[4] Gardner, T. S., Cantor, C. R., and Collins, J. J. (2000) Construction of a genetic toggle switch in Escherichia coli. Nature 403, 339â??342.
[5] Grkovic, S., Hardie, K. M., Brown, M. H., and Skurray, R. A. (2003) Interactions of the QacR multidrug-binding protein with structurally diverse ligands: implications for the evolution of the binding pocket. Biochemistry 42, 15226â??15236.
16
[6] Schumacher, M. A., Miller, M. C., Grkovic, S., Brown, M. H., Skurray, R. A., and Bren- nan, R. G. (2001) Structural mechanisms of QacR induction and multidrug recognition. Science 294, 2158â??2163.
[7] Schumacher, M. A., and Brennan, R. G. (2003) Deciphering the molecular basis of mul- tidrug recognition: Crystal structures of the Staphylococcus aureus multidrug binding transcription regulator QacR. Research in Microbiology 154, 69â??77.
[8] Peters, K. M., Brooks, B. E., Schumacher, M. A., Skurray, R. A., Brennan, R. G., and Brown, M. H. (2011) A single acidic residue can guide binding site selection but does not govern QacR cationic-drug a?nity. PLoS ONE 6, e15974.
[9] Klinke, H. B., Thomsen, ABâ?? and Ahring, B. K. (2004) Inhibition of ethanol-producing yeast and bacteria by degradation products produced during pre-treatment of biomass. Applied Microbiology and Biotechnology 66, 10â??26.
[10] Bochevarov, A. D., Harder, E., Hughes, T. F., Greenwood, J. R., Braden, D. A., Philipp, D. M., Rinaldo, D., Halls, M. D., Zhang, J., and Friesner, R. A. (2013) Jaguar: A high-performance quantum chemistry software program with strengths in life and materials sciences. International Journal of Quantum Chemistry 113, 2110â??2142.
[11] Lassila, J. K., Privett, H. K., Allen, B. D., and Mayo, S. L. (2006) Combinatorial methods for small-molecule placement in computational enzyme design. Proceedings of the National Academy of Sciences of the United States of America 103, 16710â??16715.
[12] Kuhlman, B., Dantas, G., Ireton, G. C., Varani, G., Stoddard, B. L., and Baker, D. (2003) Design of a novel globular protein fold with atomic-level accuracy. Science 302,
1364â??1368.
[13] Allen, B. D., and Mayo, S. L. (2006) Dramatic performance enhancements for the
FASTER optimization algorithm. Journal of Computational Chemistry 27, 1071â??1075.
17
[14] Allen, B. D., and Mayo, S. L. (2010) An e?cient algorithm for multistate protein design
based on FASTER. Journal of Computational Chemistry 31, 904â??916.
[15] Sun, Z. Z., Hayes, C. A., Shin, J., Caschera, F., Murray, R. M., and Noireaux, V. (2013) Protocols for implementing an Escherichia coli based TX-TL cell-free expression system for synthetic biology. Journal of visualized experiments : JoVE e50762.
[16] Stanton, B. C., Siciliano, V., Ghodasara, A., Wroblewska, L., Clancy, K., Trefzer, A. C., Chesnut, J. D., Weiss, R., and Voigt, C. A. (2014) Systematic Transfer of Prokaryotic Sensors and Circuits to Mammalian Cells. ACS synthetic biology
[17] Lutz, R., and Bujard, H. (1997) Independent and tight regulation of transcriptional units in Escherichia coli via the LacR/O, the TetR/O and AraC/I1-I2 regulatory elements. Nucleic Acids Research 25, 1203â??1210.
18

Supporting information for:

Engineering Transcriptional Regulator Effector Specificity using Computational Design and In Vitro Rapid Prototyping: Developing a Vanillin Sensor

Emmanuel L.C. de los Santos,? Joseph T. Meyerowitz, Stephen L. Mayo, and

Richard M. Murray

Division of Biology and Biological Engineering, California Institute of Technology, Pasadena, CA, USA

E-mail: emzodls@caltech.edu
KEYWORDS: synthetic biology, cell-free systems, in vitro biological circuit prototyping, computational protein design, protein engineering

?To whom correspondence should be addressed

S1
Table S1: QacR Mutants Tested

Protein

Amino Acid Position

50

54

57

58

61

86

89

90

93

96

99

102

116

119

120

126

154

157

161

qacR-wt

F

L

E

E

W

S

T

E

Y

Q

I

F

M

L

E

A

N

N

T

qacR1

F

L

Q

L

Y

S

T

Q

Y

M

Q

S

Q

Y

Q

A

M

Q

M

qacR2

F

L

Q

L

Y

S

T

Q

Y

Q

Q

F

Q

Y

Q

A

M

L

M

qacR3

F

L

Q

L

Y

S

T

Q

Y

Q

I

S

Q

Y

Q

A

M

L

M

qacR4

F

L

Q

L

Y

S

T

Q

Y

M

Q

Q

M

Y

Q

A

M

Q

M

qacR5

F

L

E

E

Y

S

T

Q

Y

M

Q

S

Q

Y

Q

A

N

N

T

qacR6

F

L

E

E

Y

S

T

Q

Y

Q

Q

F

Q

Y

Q

A

N

N

T

qacR7

F

L

E

E

Y

S

T

Q

Y

Q

I

S

Q

Y

Q

A

N

N

T

qacR8

F

L

Q

L

W

S

T

Q

Y

M

Q

S

Q

Y

Q

A

M

Q

M

qacR9

A

W

Q

L

Y

S

T

Q

Y

M

Q

S

Q

L

Q

A

M

Q

M

qacR10

A

W

Q

L

Y

S

T

Q

Y

Q

Q

F

Q

L

Q

A

M

L

M

qacR11

A

W

Q

L

Y

S

T

Q

Y

Q

I

S

Q

L

Q

A

M

L

M

qacR12

A

W

Q

L

Y

S

T

Q

Y

M

Q

Q

M

L

Q

A

M

Q

M

qacR13

A

W

E

E

Y

S

T

Q

Y

M

Q

S

Q

L

Q

A

N

N

T

qacR14

A

W

E

E

Y

S

T

Q

Y

Q

Q

F

Q

L

Q

A

N

N

T

qacR15

A

W

E

E

Y

S

T

Q

Y

Q

I

S

Q

L

Q

A

N

N

T

qacR16

A

W

Q

L

W

S

T

Q

Y

M

Q

S

Q

L

Q

A

M

Q

M

qacR17

A

W

E

E

W

S

T

Q

Y

Q

I

F

M

L

Q

A

N

N

T

S2
Table S2: OD600 of Cells at 4 ng/mL aTc




Table S3: OD600 of Cells at 4 ng/mL aTc Table S4: OD600 of Cells at 8 ng/mL aTc Table S5: OD600 of Cells at 12 ng/mL aTc
S3

Figure S1: Initial qacR induction test. Fluorescence of cells encoding wild-type qacR was compared in the presence and absence of berberine, a native qacR inducer. While we observed repression upon the addition of qacR, we did not observe induction when berberine was added.

qacR mutants screen

4

x 10

6

5

4

3

2

1

0

4

x 10

6

qacR wt

qacR 6

4

x 10

6

5

4

3

2

1

0

4

x 10

6

qacR 1

qacR 7

4

x 10

6

5

4

3

2

1

0

4

x 10

6

qacR 2

qacR 8

4

x 10

6

5

4

3

2

1

0

4

x 10

6

qacR 3

qacR 9

4

x 10

6

5

4

3

2

1

0

4

x 10

6

qacR 4

qacR 10

4

x 10

6

5

4

3

2

1

0

4

x 10

6

qacR 5

qacR 11

5

4

3

2

1

4

x 10

6

5

4

3

2

1

qacR 12

water vanillin dequalinium

5

4

3

2

1

4

x 10

6

5

4

3

2

1

qacR 13

5

4

3

2

1

4

x 10

6

5

4

3

2

1

qacR 14

5

4

3

2

1

4

x 10

6

5

4

3

2

1

qacR 15

5

4

3

2

1

4

x 10

6

5

4

3

2

1

qacR 16

5

4

3

2

1

4

x 10

6

5

4

3

2

1

qacR 17

0 0 200 400 600

0 0 200 400 600

0 0 200 400 600

0 0 200 400 600

Time (min)

0 0 200 400 600

0 0 200 400 600

Figure S2: Timetrace of qacR mutant screen. GFP fluorescence time traces for TX- TL reactions set up to screen for qacR mutants that were sensitive to vanillin. Reactions contained DNA encoding a qacR variant, T7 RNA polymerase, a fluorescent reporter and either water, dequilinium or vanillin.
S4

(a) 0 ng/ml aTc (b) 4 ng/ml aTc

(c) 8 ng/ml aTc (d) 12 ng/ml aTc

Figure S3: Vanillin dosage response curves. GFP normalized to optical density for varying vanillin concentrations. The dotted lines represent the fluoresence level for cultures without aTc in order to observe the decrease in fluorescence due to background expression of the protein
S5